Hey David! Awesome work =)



Don't have any shaders sadly, it's the first Asus EEEPC model ever released. Intel GMA 900 with OpenGL 1.1 and thanks for all the fish.




I saw that thread, didn't understand one bit of it (I haven't learned shaders yet, professionally I come from 2D console game development and, as a hobbie, I simply don't have the hardware to work with).
Regardless, fragment programming won't help me =(




I have dynamic chunk load/saving with RLE to disk, going from 128kb to 5kb in like 2 picoseconds.
I meant merging coplanar facets, using what beyzend had done in this thread.
Seeing as beyzend code is for PositionMaterial voxels and I use PositionMaterialNormal (And couldn't get that to work), I'm mashing some code of my own to merge those faces. Something more general and separate from mesh extraction.



Yup, that's me =(
At least, I got Polyvox to work with Boost instead of VS2010. That sucker got my lil' netbook on overdrive =(

I'll keep you posted

The I'll be interested how it works out, and in particular where you find that the bottlenecks are. I tried running Thermite on my girlfriends laptop and the performance wasn't great, but I didn't have the resources to investigate the problem (she might not even have had proper drivers installed). Your framerate is already looking much better.

Then we'll have to think about texturing. I imagine that for each vertex you will want to provide several sets of texture coordinates. Something like this (pseudocode, as my OpenGL is rusty!):


Is this using some existing library or your own code? I was going to roll my own simple RLE compresser (I alrady have some code actually and it's just 100 lines or so) but I'm open to other ideas. I'm trying not to introduce external depenancies, though.

I already have something like this, but it's a mess and only works with the Marching Cubes surface so far (not the Minecraft-style cubic surface). 'SurfaceMesh<VertexType>::decimate' is the place to look if you are curious, not to be confused with 'Mesh::decimate...' which is to be deleted.

Anyway, I'll look forward to seeing how you get on

That's a screenshot from my work's PC. I get 200-500 FPS in my netbook, using GMABoost to overclock the beast for 10-25% more FPS.



That looks yummy, I'll have to eat it (And then show you something functional!).



My own code, about 100 lines too. If I ever introduce the 3byte runlenght (Making stuff like "aaabbcddde" compress to "aa1bb0cdd1e" instead of "a3b2c1d3e1"), I'll post the code.
Still, this second method works better with larger entropy data. Maybe compressing volumes works better with simple RLE.
On the other hand, I have been trying to find a good way to compress mesh data, as extracting the mesh takes about 80% more time than loading from an uncompressed file.
I got to merge triangles into quads today, reducing the vertex index file weight by 25%.
Next step is merging adjacent coplanar quads



Yeah, I looked into that function while trying to simplify the mesh.
Didn't work, of course =p

I know what it's like to have a library on the works, documentation changing revision by revision and stuff. Still, I almost gave up after seeing the online docs all empty and placeholderish.

If you don't have docs yet, an empty wiki looks better than an empty section in your web... and we get to give something back to the project too

Well, I'm taking a break 'til tomorrow.

Sayounara!

Hey, I got Bullet to load my chunks dynamically =)
I'm standing in my future home ^^

I really REALLY need to clean the code, I spent the last three nights sleeprogramming.
Turns out REM sleep is really good when you need to optimize code

Also, I think I'll have to find an alternative to loading display lists on the main thread. It freezes my main loop for half a second with each loaded chunk =(

Very nice, so is Bullet working with your quad mesh or the original PolyVox one? I ask becase I wasn't aware that Bullet supported quads.

When I had Bullet support in Thermite3D I was going through the OgreBullet wrapper, but when I put it back in I think I'm more inclined to do it natively like you are doing here.

I'm making a btCompoundShape filled with one btBoxShape per voxel.
Performance wise, I had to dynamically page the chunks or the system fell to pieces.
I can have 5-10 chunks of 16x16x128 btBoxShapes with acceptable performance.

EDIT: And this way, if a voxel changes it's material, I only need to add/remove a btBoxShape.

Ok, that's an interesting approach. And presumably you're only creating boxes for voxels which are visible, not the ones which are inside the volume?

Nope
All the voxels are there xD
That would be faster, but I only have a performance impact when doing debug rendering of the btCompoundShapes...

Hi, thanks so much for releasing this lib - i'm a 2nd year PhD Computer Graphics student at Durham Uni (UK) & am excited to integrate this with my tool. Also writing a game as a side project and hope to re-use the tech

I found this tool when researching improvements to my current voxelizer (subdivision method/slice GPU renderer), and marching cubes implementations. Quite excited, but sadly I need to finish a paper due in May before I can really integrate this, and move onto more exciting developments.

Chris

Welcome aboard! So what is your research topic? This engine actually started as a result of my own PhD research - I was doing medical visualisation but soon thought it was more fun to apply it to games