When working in 3D, there are many many things you have to worry about - occlusion, lighting, depth of field, skies, BSP, audio, physics and all sorts of other types of crap. For this challenge I'll be covering one particular thing: landscapes, particularly outdoor landscapes.

So imagine a MASSIVE outdoor landscape, WoW type size. This is filled with rolling hills, sheer cliffs, deserts, trees, and buildings and caves which you can enter and explore. Thing is, you can't keep that amount of stuff in memory at once - you'd run out of memory, or even if you didn't, the rendering would go at like less than a frame per second.

So what I'm looking for is a theoretical approach to managing this type of outdoor environment. How will it be stored in memory? How will it be loaded from disk? How will the renderer know what to show?

Note that this challenge, like the last, is theoretical: I'm looking for an efficient but elegant approach, but no code needed (although if you want to use it to demonstrate a point, that's fine).

The best answer will win a free copy of Onlink!

This is mostly done by dividing the big map into little pieces. Which are small enough to load on the fly. And when you hit the edge of the map (like say 5m before the end of the map, considering your map is 1km big(radius?)) it will start loading the second map.
Mostly this map is stored in 1 big file. And cut into sequence in memory. (just pointers I guess).

For more info, try reading : Visual Basic Game Programming For Teens

Friendly greetings
Skyblownet (Owner of skyblownet.com also known as sbn)

-

What also is being used is. You make a very basic (totally no detail) map and let this load fully into memory (would take like less then 1mb of ram?). And on the fly you render the detailed map. Like says radius of about 10meter (just viewing distance of the player).
And when you're on the move again
the 'machine' will forget the fully detail render.

-
There are some other techniques which I don't know about (or not fully) I just explained two common used systems.

So say I'm standing on top of a cliff, surveying my mighty kingdom. The problem returns of way too much information, except this time the system is trying to break it down into chunks too.

I'm looking for a continuous environment here.

The LOD point is not a bad point, do you want to elaborate on that a little?

From my experiences from gaming, many games have a draw distance. I guess you could have your program only draw in a certain radius defined in an options panel. The rest of the map would be just hills or whatever is there with very low quality textures, or none at all. And for buildings, they would also appear as models with low textures. So basically, the user can choose how far the game will draw based on their computer capabilities. (And if you've ever played Guild Wars, there is a AutoConfigure feature)

So basically, it would be a continuous environment, where the game would load only what it needs for the current draw area.

IDEA 1

I you are to have a circle of space lets sat 1 meter in diameter and have the loading point .25 meters from the circle edge. Then whenever you hit that point you start loading 1 meter out. You have the draw distance then .20 meters.

IDEA 2

Use idea 1 but store much more data, by storing data in main memory and then moving the data to video memory when needed.

You could have a LOD thing like Oblivion, or you could have a variable draw distance, set by the client, and just fog out anything beyond the draw distance (chalk it up to morning mists :P ), that way a person with a seriously kickass computer can see a lot further than someone with a not-so-kickass computer.

I believe Morrowind used the second idea (along with the "cut the map into sections so that when the player gets near the edge of a section, begin to load the next") rather effectively (though you don't get grand and amazing vistas like in Halo 3 or anything).

-- Griffinhart

Let us assume we're looking at a city. Just or kicks. Generally speaking with buildings, there can be tons and tons of lines showing floors and windows etc.

However, there comes a point where you are so far from the building that you probably can't tell the lines for floors and such. At that point, you could have a diminished map of the structure. A simple box struct rather than multiple floors etc.

Also, Mister Lynn, please catch me online sometime.

This'd be easier for everyone if you were actually online some of the time.

I was thinking more of rolling outdoor natural environments actually. A city would obviously be a simple quadtree with occlusion culling and LOD.

...

Seriously, though. Even with landspaces you can use lower quality maps at far distances.

I think fog is good, because kickass computers will automatically show more.

And AF is for seeing those far away lines :)

What did you think they were talking about when they talked about LOD points?

Your idea was already mentioned by cpu5594 and FinalWarrior. Why'd you repost the same idea?

Seriously Miah. Stop spamming. :P

@prophile: So what's the next part of the challenge? Or have we already won?

-- Griffinhart

You've given me an overview of the first part. Now actually do the challenge, kthx.

I wasn't paying attention. My bad. :3