Fast voxel meshing algorithm - creates 'greedy' meshes with support for voxel types, baked light & Ambient Occlusion. UVs can easily be added but the vertex structure would have to be changed from a single integer.
> git clone https://github.com/cgerikj/binary-greedy-meshing --recursive
> cd binary-greedy-meshing
> mkdir build && cd build
> cmake .. -G "Visual Studio 16 2019"
> start binaryMesher.sln
- Noclip: WASD
- Toggle wireframe: X
- Regenerate: Spacebar
- Cycle mesh type: Tab
Meshing duration is printed to the console.
The mesher lives in src/mesher.h
Input data:
- std::vector<uint8_t> voxels (values 0-31 usable)
- std::vector<uint8_t> light (values 0-15 usable)
The input data includes duplicate edge data from neighboring chunks which is used for visibility culling and AO. Input data is ordered in YXZ and is 64^3 which results in a 62^3 mesh.
Output data:
- std::vector<uint32_t> of vertices in chunk-space.
Vertex data is packed into one unsigned integer:
- x, y, z: 6 bit each (0-63)
- Type: 5 bit (0-31)
- Light: 4 bit (0-15)
- Normal: 3 bit (0-5)
- AO: 2 bit
Meshes can be offset to world space using a per-draw uniform or by packing xyz in gl_BaseInstance if rendering with glMultiDrawArraysIndirect.
| Mesh | Wireframe |
|---|---|
 |
 |
 |
 |
 |
 |
 |
 |
Average execution time running on Ryzen 3800x.
| Scene | Milliseconds | Vertices |
|---|---|---|
| 3d hills | 1.139 | 46798 |
| Red sphere | 1.108 | 71532 |
| White noise | 33.315 | 2142608 |
| 3d checkerboard | 36.832 | 4289904 |
| Empty | 0.174 | 0 |
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent