i like that very much! may i ask how to learn to create meshes with lite-c and use them for further stuff (e.g. collision detection).

But marching cubes only does the triangle conversion. What did you use to generate your field in the first place?
Can a vertex shader access 3d texel values? In that case it could be done on the GPU.

In GPU Gems 3 there is a chapter on realtime terrain generation on the GPU using a 3d grid like system with the new Geometry shader. But I guess this is a CPU version as the card is a 6600..

This cannot be done easily on 3dgs though can it, we need access to the vertex data?

Edit: With Lite-C we can uses D3D so maybe it is possible ^^

Looks amazing though

Thanks,
Adoado ^^

how do you generate seamless uv coordinates? is this some kind of unwrapping or do you use a procedural approach?

Thanks for all the comments.
Some explanations:
I split the whole scene which is to be generated into smaller tiles (typically 64*64*64), this saves memory (the memory used for creation stays constant, no matter how big your scene will be) and its easier to use several threads to speed things up (but every thread must has its own temporary memory for generation, so this could be a big problem, you can solve that if you use smaller tiles (which are slower though)).
So next is to create the density map for one part, these are double the size of the part (128*128*128) to prevent visible edges between the parts caused by wrong normals and occlusion term. Creating one takes around 50ms. Actually this map are 128 floating points textures which are filled with a shader, and then copied back to an array (faster then doing everything on the cpu). The densitymap specify which parts of the scene is rock (positive values) or air (negative values).
I'm using 4 64*64*64pixel noise 3d textures (filled with rand()s) which gets combined at several different frequencies and amplitudes to form the scene, but you can also encode your own worlds in the shader like planets, or a helix or...

On this map i run the marching cube algorithmus which creates the polygons. Normals and ambient occlusion term gets also calculated. This takes 500ms-2sec (depending on how many polygons needs to be created and if ambient occlusion is calculated)
After this there is also the possibility to generate several LOD meshes.


No, it is saved per vertex at the moment and i don't want to unwrap this meshes.

Its samples the texture 3 times (with XY, YZ, XZ world coordinates) and blend these according to the surface normal.

60fps means ~500k textured polygons on my PC, which i think is enough for a decent terrain .

Some numbers:
Memoryconsumption
1 thread: 350mb
2 threads: 570mb
4 threads: 900mb
Creation time of the babel tower:
1 thread: 16 sec
2 threads: 10,9 sec
4 threads: 10,7 sec

i think i will release a sample demo with the tower level (i like that one), but first i will try i could get some nice clouds around it and i must find a way to get collision with these meshs (any idea?)

http://swollen-eyeballs.org/terrain/babel_00.jpg
http://swollen-eyeballs.org/terrain/babel_01.jpg
http://swollen-eyeballs.org/terrain/babel_02.jpg

thanks a lot, i never thought about splitting the volume. do you use cuda or plain c++ with directx?

it's kind of a late notice, but there's a Procedural Generation Competition going on at TIGSource Forums

Entries are due June 2. Almost everyone makes 2d games, but they DO allow 3d games. It would be nice if some Gamestudio user whips up a 3d procedurally generated game for this contest.

jay. i've also implemented the marching cubes algorithm, my version is running on the cpu only. and i've got clouds... i must ask wether or not i may post an image of it.

another question, though: are you creating a volume texture and passing it to a pixel shader to generate the random values?

and i *think* you're using an optimized version of the standard marching cubes algorithm, or are you outputting vertices on all edges? i think not because you to have gouraud shading working on the mesh. do you know the book gpu gems 3?

joey.

Yeah post your images, maybe we can help each other to create a really fast version.
I'm filling the volume texture on the cpu with rand()s.
I'm saving each created vertex in an extra array, before I create a new vertex I test if there is already a vertex at this position in the array, if so i use the Index of this vertex for the Triangle.
Yeah i know the article in the book, but i only deals with DirectX10 cards.
I've found this article: http://www.mpi-inf.mpg.de/~gziegler/hpmarcher/techreport_histopyramid_isosurface.pdf
According there table there method is 2times faster then my implementation, so i might give it a try.

It`s a cool!