Custom-Virtualized-Geometry-Renderer

KO | EN

gitlite — search

#python #java #python3 #arduino #golang #machine-learning #rust #html #flask #javascript #seismology #nodejs

Custom-Virtualized-Geometry-Renderer

★ 34 Open GitHub ↗

A custom Virtualized Geometry LOD Rendering Engine : cluster-based real-time mesh simplification and LOD streaming system inspired by UE5's Nanite. This engine dynamically switches between multiple Levels of Detail using a virtualized geometry hierarchy, supporting view-dependent streaming and cluster-based simplification.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Custom-Virtualized-Geometry-Renderer 💻
A custom Virtualized Geometry LOD Rendering Engine : cluster-based real-time mesh simplification and LOD streaming system inspired by UE5's Nanite. This engine dynamically switches between multiple Levels of Detail using a virtualized geometry hierarchy, supporting view-dependent streaming and cluster-based simplification.

## ✨ Features
- ♻️ **Dynamic LOD transitions** via a **graph** to reduce cluster overdraw  
- 🧩 **Cluster-based mesh partitioning** for efficient culling and updates  
- 🚀 **View-dependent streaming** with bounding volume–based LOD switching  
- 🍭 **Debug views:** render **clusters** or **triangles (primitives)**
- ⚡️ **Ray Traced** light, shadows, ambient occlusion
---
## 🔧 Technical Highlights
- 🛠️ Built from the ground up
- 📦 Minimal external dependencies: [GLFW](https://www.glfw.org/) and [Dear ImGui](https://github.com/ocornut/imgui)
- ⏭️ Custom SIMD (SSE/AVX/FMA) math for vector ops
- 🌲 **BVH** (Bounding Volume Hierarchy) for cluster spatial organization
- ✂️ **QEM** (Quadric Error Metrics) for geometry simplification
- 🎯 Per-cluster frustum culling
- 📥 View-frustum based streaming of cluster geometry
- 🧵 **Multi-threaded** mesh processing; rendering on its own thread
- 🌋 **Vulkan RHI** backend with dynamic rendering and indirect drawing

---
## 🎮 Ideal For
- Researchers exploring real-time mesh simplification  
- Engine developers experimenting with virtualized geometry  
- Anyone inspired by **Nanite** and interested in similar techniques (here using **Vulkan**)
---
## 🎥 Demo
A packaged demo app is available under **Releases**. Example meshes are included.<br>  
**Note:** not every mesh will work—see the project note below for details.<br>
❗️🛡️ On some system **Windows Defender**  will be alerted. So to run click on **More Info** -> **Run anyway**.<br>
**Video** to [this](https://youtu.be/D8U2XRY2Bt8) : https://youtu.be/D8U2XRY2Bt8

### Triangle View
![primitive_view](https://github.com/user-attachments/assets/a9ed94d2-7118-47bd-86d3-0f5ba690bb0b)

### Cluster View
![cluster_view](https://github.com/user-attachments/assets/84c4f374-0396-4df2-bed2-5ee1261dde27)

## Ray Tracing 
![RT](https://github.com/user-attachments/assets/86614106-443e-4036-9c90-bf44b65dcc72)

---
## ⚠️🚧⚠️ Important Note to Project
This project was developed as part of my degree and is **experimental**. It is **not** a production-ready, bullet-proof engine.
- Core data structures trade memory for simplicity and will need re-architecture for real-world scale and stability.
- Some meshes—especially **non-manifold** or highly irregular topology—may fail to process or render correctly.
- APIs, file formats, and behavior may change without notice.
**Provided “as is”, without warranty of any kind.** Use at your own risk.

## ✅ Requirements
**OS**
- Windows 10 or 11, 64-bit
 
**GPU & Drivers**
- A Vulkan-capable discrete or integrated GPU
- Vulkan **1.4** driver recommended (no SDK required to run)

**CPU**
- x86_64 CPU
- **AVX2 + FMA recommended** (this build uses SIMD paths for speed)
- SSE2-only CPUs may work with a non-SIMD build (slower). If you need that, build with SIMD disabled.

Custom-Virtualized-Geometry-Renderer

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README