juj/wasm_webgpu

Overview

wasm_webgpu is a small open‑source library that lets C/C++ programs talk to the WebGPU graphics API when they are compiled to WebAssembly with Emscripten. It provides a thin, 1‑to‑1 mapping of the browser‑side WebGPU JavaScript API so developers can write GPU‑accelerated rendering code in native C/C++ and run it in the browser (or, experimentally, via Dawn outside the browser).

What it does

Feature	What it means for you
Header + implementation files (lib_webgpu.*)	Drop these into any Emscripten project, include lib_webgpu.h, and you get C‑style wrappers such as wgpu_adapter_request_device_sync() that call the underlying WebGPU JS functions.
Exact API mapping	Types like GPUAdapter become WGpuAdapter; functions keep the same shape (wgpu_canvas_context_configure). This makes the spec a reliable reference when writing C code.
Performance‑focused	The binding layer is deliberately low‑level and avoids extra JavaScript objects. It works with Closure Compiler for aggressive minification and tries to keep runtime overhead and generated code size tiny.
Custom helpers	Functions for reading/writing mapped buffers and loading images from URLs simplify common WebGPU tasks that are awkward to marshal between JS and Wasm.
JSPI (synchronous API) support	Optional flags (-sJSPI) expose synchronous versions of normally async WebGPU calls, handy for quick prototypes or test harnesses.
OffscreenCanvas & workers	APIs to create and control an OffscreenCanvas from a WebWorker or pthread, enabling rendering on background threads.
Memory model flexibility	Works with Emscripten’s 2 GB, 4 GB, and Wasm64 memory configurations, so you can target large data sets if needed.
Dawn backend (experimental)	A separate source file (lib_webgpu_dawn.cpp) lets you compile against the Dawn implementation of WebGPU when you’re not in a browser, though it’s not the primary focus.

Who might use it

• Game or graphics engine developers who want to keep their core rendering code in C/C++ and ship it to the web without rewriting everything in JavaScript or WebGL.
• Research codebases that already use GPU compute in native C++ and need a quick path to run in a browser for demos or reproducibility.
• Tooling authors building WebGPU‑based visualizers, editors, or test suites that benefit from the synchronous API and minimal code‑size footprint.

Getting started (quick recap)

1. Install Emscripten.
2. Clone the repo, copy the lib/ files into your project, and add the library to the Emscripten link step (--js-library …).
3. Use the provided CMakeLists in samples/ as a template.
4. Build with emcmake cmake … && ninja, then serve the generated HTML with emrun ..

Limitations / caveats

• The library mirrors the current WebGPU spec exactly; breaking changes can appear when the spec evolves, and there is no ABI stability guarantee.
• The Dawn backend is outdated and not actively maintained.
• It is a binding layer only—you still need to write the actual rendering logic; the supplied samples are intentionally simple and not flashy.
• Using JSPI can introduce unexpected event‑loop interactions, so profiling is recommended for production‑grade code.

Bottom line

wasm_webgpu is a genuine, low‑level bridge that makes WebGPU accessible from C/C++ compiled to WebAssembly. It is aimed at developers who care about performance, code‑size, and staying close to the official WebGPU specification, rather than those looking for a high‑level graphics engine.