Skip to main content
gamedev.net
🔒 Locked

Terrain rendering.

Started by mrmrcoleman Feb 19, 2005 at 5:29 PM 2 replies 900+ views
Original Post
mrmrcoleman
mrmrcoleman
Hello. I have had experience with terrain rendering a few years ago whilst at university. Back then I played with quad-trees for occulsion and got a ROAM implementation going but I haven't been able to keep up with advances as I have been working in 2d for the last couple of years (albeit in Direct3d). What type of algorithms are being used to render the terrains that I see in Halo, Halo 2, Return to castle wolfenstein etc? I realise that there is probably some sort of page swapping to enable the huge areas that are used, but what is being used for LOD, for culling, etc? Are these maps made form heightfields, TIN's, modelled in polygons? Also, does anybody know of any good references, tutorials, books which can help me get back up to speed? All I seem to be able to find is technical documents which don't give away very much in terms of implementations, and also don't give me any idea of there usefulness in a game environment. Thanks for any help. Mark Coleman
snk_kid
snk_kid
From the very little i've read about the subject it will envolve hybrid spatial data structures often quadtree like structures for the massive areas where inside each leaf different data structures can be used depending on its contents, and your wright areas about to come into view are paged in the quadtree is used for that aswell.

You might be intereted in read this.
noisecrime
noisecrime
Hi,

With regard RTCW I don't believe there is anything special going on with Terrain, its just brute forced. Afterall RTCW and RTCW:ET is heavily based on the Quake 3 engine, which had no fancy terrain handling, and from what I remember of modding/mapping there is nothing special about the terrains. However if you look at some sample maps you'll proberbly find that the terrain areas are not that big as they are often seperated by portal geometry - e.g. long tunnels beneath a mountain. So the overall map might look quite big, but terrain is kept to a reasonable level within each section.

The general consenus these days is to simply brute force the terrain. Why spend valuable cpu cycles on ROAM when the gpu is sitting idle? Obviously though, you can't just keep chucking more and more terrain into your engine, there are limits. So some form of spatial scheme is needed and LOD, quadtrees are still favourable other buzzwords are chunked LOD and Geomipmappng.

There have been some very good threads on gamedev about terrain, sadly without the search feature its a bit hard to track them down. But nothing to stop you from searching backwards page by page (the graphic prog and theory forum is best) ;)

Anyway try some of these links, from the forum faq
vterrain.org: Very extensive resource covering all aspects of terrain visualization.
Methods for Realistic Landscape Imaging: Fractal landscape generation and rendering (by F. Kenton Musgrave).
Terrain LOD techniques: ROAM, CLOD, quadtrees.
Chunked LODs: Hardware friendly terrain LOD technique, a mixture of ROAM and geo-mipmapping (by Thatcher Ulrich).

and I guess its always worth pointing you to
Yann's Greatest hits, not so much on actual terrain implementations, but plenty on spatial partitioning schemes, occlusion and making your terrain look pretty.
mrmrcoleman
mrmrcoleman
Thanks for the input guys. I guess the thing which is keeping me from understanding a lot of this stuff is a lack of understanding of what a GPU can do for me, last time I cared graphics cards were good for blitting and vertex buffers.

Guess I need to look into that a bit more also, does anybody know of any books which would get me up to date on all the features of Direct3d and also what the new graphics hardware does for me? I may be missing the point on this one but I'm sure you'll put me straight!

Mark Coleman

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.

Sign in to reply to this topic.