Skip to content

WebGPU API Reference

These pages list the WebGPU interfaces used by the tutorials. Each page stays with members that exist in the specification. The complete definitions are in the WebGPU specification.

From the adapter to the queue

interface GPU {
  requestAdapter(options?: GPURequestAdapterOptions): Promise<GPUAdapter | null>;
  getPreferredCanvasFormat(): GPUTextureFormat;
}

interface GPUAdapter {
  readonly features: GPUSupportedFeatures;
  readonly limits: GPUSupportedLimits;
  readonly info: GPUAdapterInfo;
  requestDevice(descriptor?: GPUDeviceDescriptor): Promise<GPUDevice>;
}

GPUDevice creates the resources. Common methods:

Method	Object
`createBuffer`	GPUBuffer
`createTexture`	GPUTexture
`createSampler`	GPUSampler
`createBindGroupLayout` / `createBindGroup`	GPUBindGroup
`createPipelineLayout`	Combines bind group layouts into a pipeline layout
`createShaderModule`	Compiles WGSL
`createRenderPipeline`	GPURenderPipeline
`createComputePipeline`	GPUComputePipeline
`createCommandEncoder`	GPUCommandEncoder
`pushErrorScope` / `popErrorScope`	Capture validation errors
`destroy`	Destroy the device and resources it created

The device also exposes read-only features, limits, and queue. lost is a promise that resolves when the device is lost. Validation errors that are not captured by an error scope are delivered as uncapturederror events.

getPreferredCanvasFormat() returns only "rgba8unorm" or "bgra8unorm". Use that value for the canvas color attachment and the fragment target format.

How resources reach a shader

Vertex and index data are buffers, bound with setVertexBuffer() and setIndexBuffer().
Uniforms, storage buffers, textures, and samplers go in a bind group and are bound with setBindGroup().
Commands are recorded into a GPUCommandEncoder. finish() returns a GPUCommandBuffer, and GPUQueue.submit() runs it.

Tutorial order:

Further reading