WACS.WASI.GFX.Webgpu 0.1.0-preview

WACS.WASI.GFX.Webgpu

WASI-GFX wasi:webgpu@0.0.1 host bindings for WACS. The fourth wasi-gfx WIT package, mirroring the browser WebGPU spec verbatim.

This is the contract assembly — [WitSource]-tagged interfaces (generated from wit/webgpu.wit) plus the WitBindings dispatcher that ties them into the WACS runtime. The actual GPU backend lives in WACS.WASI.GFX.Silk and wraps Silk.NET.WebGPU / wgpu-native.

Surface covered

• gpu.request-adapter / adapter.request-device / device.queue.
• Buffers: device.create-buffer, buffer.{map-async, get-mapped-range-get-with-copy, get-mapped-range-set, unmap, destroy}.
• Textures + views: device.create-texture, texture.{create-view, destroy}.
• Shaders + pipelines: device.create-shader-module, device.{create-bind-group-layout, create-pipeline-layout, create-bind-group}, device.{create-compute-pipeline, create-render-pipeline}.
• Command encoding: device.create-command-encoder, command-encoder.{begin-compute-pass, begin-render-pass, copy-buffer-to-buffer, copy-texture-to-buffer, finish}, compute-pass-encoder + render-pass-encoder operations, queue.{submit, write-buffer, on-submitted-work-done}.
• Swap-chain bridge: device.connect-graphics-context(ctx) — the wgpu-side hook that fuses an OS surface (held by wasi:surface) to a wgpu device for get-current-buffer + present (driven by wasi:graphics-context.context).
• Async work: gpu-future (request-device, map-async, on-submitted-work-done) signals through ManualResetPollable so the guest's wasi:io/poll.poll(...) wakes on completion.

Status

Feature-complete for the parity fixtures shipped in this repo. Verified end-to-end through both the interpreter and transpiler paths against Spec.Test/components/fixtures/:

• wasi-webgpu-hello-compute — compute pipeline + map-async readback.
• wasi-webgpu-hello-render — render pipeline + copy-texture-to-buffer.
• wasi-webgpu-game-of-life — bind-group ping-pong + multi-pass compute.
• wasi-webgpu-game-of-life-windowed — full swap-chain through surface + graphics-context + webgpu.

See for the usage guide and the for backend internals (wgpu-native binding, swap-chain bridge, threading model).