A WebGPU Implementation of Augmented Vertex Block Descent: Running Physics Simulations in the Browser

WebGPU Physics: Running Augmented Vertex Block Descent Simulations in the Browser

A developer has implemented Augmented Vertex Block Descent (AVBD), a physics simulation algorithm, using WebGPU, enabling real-time physics simulations to run entirely in the browser. The project has gained 115 points on Hacker News with 14 comments.

What Is WebGPU

WebGPU is the successor to WebGL, providing:

• Modern GPU access: Closer to metal GPU programming from JavaScript
• Compute shaders: General-purpose GPU computing (GPGPU) in the browser
• Better performance: Lower overhead and more predictable performance than WebGL
• Wider adoption: Chrome, Firefox, and Safari all support or are implementing WebGPU

What Is Augmented Vertex Block Descent

AVBD is a constraint-based physics simulation method:

• Particle systems: Simulates objects as collections of particles
• Distance constraints: Maintains relationships between particles (rigid body, soft body, cloth)
• GPU acceleration: Massively parallel constraint solving on the GPU
• Augmented approach: Improved convergence and stability over standard Position-Based Dynamics

The Implementation

This project brings together:

1. WebGPU compute shaders: Physics calculations run on the GPU
2. WebGPU render pipeline: Visualization renders directly from simulation data
3. JavaScript control: Interactive parameters and user input handling
4. Real-time performance: Smooth 60fps simulations of complex scenes

Why This Matters

1. No installation: Physics simulations that run in any modern browser
2. GPU compute in the browser: WebGPU makes GPGPU accessible to web developers
3. Educational: Interactive physics demonstrations for teaching and learning
4. Creative coding: New possibilities for browser-based art and games
5. Zero setup: No build tools, no compilation, just open a URL

Broader Trend

WebGPU is enabling a new class of browser applications:

• AI inference: Running ML models in the browser
• Video processing: Real-time video effects
• CAD applications: Browser-based design tools
• Game engines: Full game engines running in the browser
• Scientific visualization: Interactive data visualization

Technical Achievement

Running physics simulations efficiently in the browser requires careful optimization:

• Minimizing GPU-CPU data transfer
• Efficient memory layout for GPU access patterns
• Proper synchronization between compute and render passes
• Handling WebGPU compatibility across browsers

Source: GitHub (jure) / HN — 115 points, 14 comments