Wasi-Webgpu: Build Beautiful Graphics and Safely Run AI with Any GPU
summary
Project renames to wasi-gfx and mechanically converts WebGPU Web IDL into WIT via a WebIDL→WIT tool, exposing the full WebGPU API through the WASI Component Model with host pieces wasi-surface, wasi-graphics-context and wasi-framebuffer. A toy host runtime (graph-time) runs compiled Wasm apps: wgpu-based Rust code and a full Bevy game both link and run as Wasm, demos load procedural-modeling plugins into a VM map viewer, and early benchmarks show about 85% native performance.
AI-written summary · about
chapters (11)
- 0:00Wasi-WebGPU introduction
- 2:50Why WebGPU needs WASI
- 5:16GPU access use cases
- 7:44Wasi-GFX architecture
- 10:09Bevy complex examples
- 11:42AI inference with Wasi-GFX
- 12:27Plugin architecture and GPU sharing
- 14:31Procedural modeling demo
- 18:23Wasi-GFX future plans
- 21:10Community and next steps
- 21:32Q&A
related talks
31mGPUs Unleashed! Make Your Games More Powerful With wasi-gfx
38mWebGPU: Next-generation 3D Graphics on the Web
25mPathfinder to GPU Offload in WASM
24mContent Authenticity Initiative Trustmark With WASI WebGPU
32mBringing WebAssembly outside the web with WASI
56mKEYNOTE: WebAssembly Update: The Proposals, Wasi, And The Component Model
50mWeb Assembly
37mGetting Started with AI and WebAssembly