bun-webgpu

WebGPU (Dawn) FFI bindings for Bun.

Usage

import { setupGlobals } from 'bun-webgpu';

setupGlobals();

const adapter = navigator.gpu.requestAdapter();
const device = await adapter?.requestDevice();

// ... use WebGPU ...

Find out more about WebGPU types here.

Building

Note: There are prebuilt binaries in the pipeline artifacts.

Prerequisites

• Bun: Ensure you have BunJS installed. (https://bun.sh)
• Zig: The native components of this library are written in Zig. Ensure Zig is installed and available in your PATH. (https://ziglang.org/learn/getting-started/)
• Pre-built Dawn Libraries: This project relies on pre-built Dawn libraries.
  • See packages/bun-webgpu/dawn/README.md for details on how to download the required Dawn shared libraries.
  • Basically just run bun run ./dawn/download_artifacts.ts

Building the FFI Liberary

The package.json includes scripts to build the native library components.

• build:dev
• build:prod

Conformance Test Suite (CTS)

To run the CTS, build the library first, then run the tests:

./run-cts.sh 'webgpu:api,operation,adapter,requestDevice:always_returns_device:*'

Run all webgpu tests for example with ./run-cts.sh 'webgpu:*'

Current results for webgpu:api,* are:

** Summary **
Passed  w/o warnings = 39994 / 51160 =  78.17%
Passed with warnings =     0 / 51160 =   0.00%
Skipped              =  8290 / 51160 =  16.20%
Failed               =  2876 / 51160 =   5.62%

Testing

To run tests, build the library first, then run the tests:

bun test