Level coding - voxel based(?)

Started by
4 comments, last by AdeptStrain 11 years, 4 months ago
Hi all,


This is my first technical post so please bear this in mind.

My second game (although it's gonna be a long term thing for me) is a clone of the old Gravity Force 2. Although with my own flavour of course.

I'm doing this is basic DirectX11 (not even bothering with textures yet) but I'm not sure how to implement a destructible landscape.

PLEASE NOTE that I am teaching myself and have no cash - so can only hire out books from the library...

My current thought is have a minecraft-esque visual theme to the landscape - blocks that can be destroyed. But I'm unsure how to implement this properly. I've so far got a multiD boolean array 250*250 (totalling 62,500 pieces). For every frame I've been going through the array and seeing if a block is there or not and then simply updated the constant buffer and display respectively.

I originally tried this with a 100*1000 array (100,000 pieces) but this caused drastic slowdown of the frame rate. And I'm only drawing a cube for each block and the basic triangluar player ship.

My question is this;

Is this kind of thinking valid? Is there a better way to achieve a similar result?

Many thanks for the assistance all.
Advertisement
I think minecraft does it as follows:
-Divide the world cells into chunks (lets say 16*16)
-For each chunk, generate a mesh containing only the cells that are next to a transparent cell (and thus visible. if its 2D and you want to see all of them then dont do this of course...)
*regenerate the mesh each time the chunk is modified
-Each frame render each of the meshes


so you only tell the GPU to render the meshes each frame, instead of the individual blocks

o3o


so you only tell the GPU to render the meshes each frame, instead of the individual blocks

QFE.

The key here is be able to render chunks of nearby cubes all at once. Graphics cards are good at drawing lots of triangles quickly - feed it only a few triangles at a time, and you are starving the GPU.

Tristam MacDonald. Ex-BigTech Software Engineer. Future farmer. [https://trist.am]

Notch discusses a bit how he does it here.

At a very basic level, you can do a 3D perlin noise on a section (using Minecraft's example, let's say 16 x 16 x 128). This should give you a value between -1 and -1. Personally I would convert that value to a 0 - 1 value ( noiseValue * 2 - 1 ) and save it to texture. Once you've done that, you can then just make the assumption that "Any value less than 0.5 is solid, anything great is empty." And read from that texture any time you load the game. Someone destroys a block just update the texture with the appropriate value (anything above 0.5 in this example).
Okay,

What about the array? Is this not a good way to keep track of the grid?

Okay,

What about the array? Is this not a good way to keep track of the grid?


You can use an array, the key is loading/unloading sections. How you store the data (array, texture, some binary format, XML, whatever) doesn't really matter (at least in this case, there are pros and cons to all stored data approaches). Loading 9 sections (let's say the spot you're standing on and it's nearest neighbors) is do-able and fine. Loading an entire world (which is impossible in games like Minecraft as the world is created as you walk between sections) is going to be memory prohibitive (among other things).

This topic is closed to new replies.

Advertisement