Samuel Bouchet
cd9814e494
CPU smooth mesher optimizations (560ms → 17ms): - VoxelData grid cache eliminates redundant readVoxel calls - Pre-cached 27 neighbor chunk pointers (readVoxelFast) - smoothNear dilation (8 lookups/cell instead of 56) - Early exit via containsSmooth flag on chunks - Thread-local scratch buffers (SmoothScratch ~600KB) - wi::jobsystem parallelization across all cores - Persistent staging vectors for upload TopingSystem optimizations (58ms → 6ms): - collectInstancesParallel() with per-chunk local vectors - Neighbor chunk pointer caching GPU compute Surface Nets (Phase 5.3): - Two-pass compute shader: centroid grid + emit with smooth normals - Pass 1 (voxelSmoothCentroidCS): computes centroids + solid flags for cells [-1..32], cross-chunk neighbor voxel reading - Pass 2 (voxelSmoothCS): reads ONLY from centroid grid, computes area-weighted smooth normals from 12 incident edges per vertex - Batched dispatch: all centroid passes then all emit passes with single UAV→SRV barrier (instead of 2 barriers per chunk) - Smooth chunk filtering: only dispatches chunks with containsSmooth - Centroid grid buffer dynamically sized per smooth chunk count - 1-frame readback delay with auto-redispatch on first frame
118 lines
4.6 KiB
Markdown
118 lines
4.6 KiB
Markdown
# BVLE Voxels
|
||
|
||
Prototype de moteur voxel hybride basé sur [Wicked Engine](https://github.com/turanszkij/WickedEngine) (MIT, C++17, DX12/Vulkan).
|
||
|
||
Cible : **60+ fps en 1440p**, monde de 512×512×256 voxels visibles.
|
||
|
||
## Prérequis
|
||
|
||
| Outil | Version | Installation |
|
||
|-------|---------|-------------|
|
||
| **Windows** | 10/11 (x64) | — |
|
||
| **CMake** | 3.19+ | `winget install Kitware.CMake` |
|
||
| **Visual Studio 2022 Build Tools** | 17.x | `winget install Microsoft.VisualStudio.2022.BuildTools` |
|
||
| **Windows SDK** | 10.0.26100+ | `winget install Microsoft.WindowsSDK.10.0.26100` |
|
||
| **GPU** | DX12 feature level 12.0+ | AMD RDNA 2+ / Nvidia RTX 3060+ recommandé |
|
||
|
||
> Le SDK **10.0.26100** est requis car les headers DX12 fournis par Wicked Engine ne sont pas compatibles avec le SDK 22621.
|
||
|
||
## Installation
|
||
|
||
```bash
|
||
# 1. Cloner le dépôt
|
||
git clone <url> bvle-voxels
|
||
cd bvle-voxels
|
||
|
||
# 2. Cloner Wicked Engine dans engine/
|
||
git clone --depth 1 https://github.com/turanszkij/WickedEngine.git engine
|
||
|
||
# 3. Configurer CMake
|
||
cmake -B build -G "Visual Studio 17 2022" -A x64 -DCMAKE_SYSTEM_VERSION=10.0.26100.0
|
||
|
||
# 4. Compiler
|
||
cmake --build build --config Release --target BVLEVoxels --parallel
|
||
|
||
# 5. Lancer
|
||
./build/Release/BVLEVoxels.exe
|
||
```
|
||
|
||
## Commandes de lancement
|
||
|
||
| Commande | Description |
|
||
|----------|-------------|
|
||
| `BVLEVoxels.exe` | Mode normal (monde procédural, rayon 4 chunks) |
|
||
| `BVLEVoxels.exe debug` | Mode debug face-color (+X=Rouge, -X=Rouge sombre, etc.) |
|
||
| `BVLEVoxels.exe debugdevice` | Active la couche de debug D3D12 |
|
||
| `BVLEVoxels.exe vulkan` | Force le backend Vulkan |
|
||
|
||
## Contrôles
|
||
|
||
| Touche | Action |
|
||
|--------|--------|
|
||
| **WASD** | Déplacement caméra |
|
||
| **Espace / Ctrl** | Monter / Descendre |
|
||
| **Shift** | Vitesse ×3 |
|
||
| **Clic droit** | Capturer/libérer la souris |
|
||
| **~** (tilde) | Console Wicked Engine |
|
||
|
||
## Architecture
|
||
|
||
```
|
||
bvle-voxels/
|
||
├── CMakeLists.txt # Build CMake
|
||
├── engine/ # Wicked Engine (git clone --depth 1)
|
||
├── src/
|
||
│ ├── voxel/ # Bibliothèque VoxelEngine (static lib)
|
||
│ │ ├── VoxelTypes.h # Types (VoxelData, PackedQuad, ChunkPos)
|
||
│ │ ├── VoxelWorld.h/.cpp # Monde voxel (hashmap, génération Perlin)
|
||
│ │ ├── VoxelMesher.h/.cpp # Binary Greedy Mesher CPU
|
||
│ │ └── VoxelRenderer.h/.cpp# Renderer GPU-driven + VoxelRenderPath
|
||
│ └── app/
|
||
│ └── main.cpp # Point d'entrée Win32
|
||
├── shaders/ # Sources HLSL
|
||
│ ├── voxelCommon.hlsli # Root signature, CB, structs partagés
|
||
│ ├── voxelVS.hlsl # Vertex shader (vertex pulling)
|
||
│ ├── voxelPS.hlsl # Pixel shader (triplanar + lighting)
|
||
│ ├── voxelCullCS.hlsl # Compute: frustum + backface culling
|
||
│ └── voxelMeshCS.hlsl # Compute: GPU mesher (binary, baseline)
|
||
└── voxel_engine_spec.md # Document de spécification complet (Markdown)
|
||
```
|
||
|
||
## Pipeline de rendu (Phase 2 — GPU-driven)
|
||
|
||
```
|
||
CPU: mesh dirty chunks (greedy merge) → pack quads + chunkInfo dans mega-buffers → upload GPU
|
||
GPU: frustum cull compute → indirect args → DrawInstancedIndirectCount (1 appel = N draws)
|
||
```
|
||
|
||
**Buffers GPU :**
|
||
|
||
| Buffer | Type | Slot | Rôle |
|
||
|--------|------|------|------|
|
||
| `megaQuadBuffer_` | StructuredBuffer\<PackedQuad\> | SRV t0 | 2M quads max (16 MB) |
|
||
| `chunkInfoBuffer_` | StructuredBuffer\<GPUChunkInfo\> | SRV t2 | 2048 chunks max |
|
||
| `indirectArgsBuffer_` | RWStructuredBuffer | UAV u0 | Indirect draw args |
|
||
| `drawCountBuffer_` | RWByteAddressBuffer | UAV u1 | Compteur atomique de draws |
|
||
|
||
**Caractéristiques Phase 2 :**
|
||
- Mega-buffer unique pour tous les quads de tous les chunks
|
||
- Vertex pulling via `SV_VertexID` + push constants (`b999`)
|
||
- Frustum culling CPU (wi::primitive::Frustum)
|
||
- Backface culling par face group (6 directions × chunk)
|
||
- GPU frustum cull compute shader (prêt, activation via flag)
|
||
- GPU mesher compute shader baseline (binary, sans greedy merge)
|
||
- Tri des quads par direction de face dans le mesher CPU
|
||
- GPU timestamp queries pour benchmark
|
||
|
||
## Phases de développement
|
||
|
||
- [x] **Phase 1** — Setup, meshing CPU, rendu basique
|
||
- [x] **Phase 2** — GPU-driven pipeline, mega-buffer, culling, compute shaders
|
||
- [x] **Phase 3** — Texture blending (triplanar, height-based)
|
||
- [x] **Phase 4** — Toping (rebords, bordures procédurales)
|
||
- [x] **Phase 5** — Rendu smooth (Surface Nets / Marching Cubes)
|
||
- [ ] **Phase 6** — Ray tracing hybride (RT shadows + AO)
|
||
|
||
## Licence
|
||
|
||
Wicked Engine est sous licence MIT. Le code spécifique BVLE est propriétaire.
|