GPUContext API

WebGPU device and resource management.

---

Constructor

constructor()

Creates a new GPUContext instance. Does not initialize the GPU device.

---

Methods

initialize

async initialize(config?: GPUContextConfig): Promise<void>

Initialize WebGPU adapter and device.

Parameters:

interface GPUContextConfig {
  powerPreference?: 'low-power' | 'high-performance';
  forceFallbackAdapter?: boolean;
}

Throws:

• WebGPUNotSupportedError: If WebGPU is not available
• DeviceInitializationError: If GPU device initialization fails

Example:

const context = new GPUContext();
await context.initialize({ powerPreference: 'high-performance' });

---

getDevice

getDevice(): GPUDevice

Get the WebGPU device. Must call initialize() first.

Throws:

• Error: If context is not initialized

---

createBuffer

createBuffer(descriptor: GPUBufferDescriptor): GPUBuffer

Create a GPU buffer.

Parameters:

• descriptor: WebGPU buffer descriptor

Example:

const buffer = context.createBuffer({
  size: 1024,
  usage: GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_DST
});

---

createShaderModule

createShaderModule(code: string, label?: string): GPUShaderModule

Create a shader module from WGSL code.

Parameters:

• code: WGSL shader code
• label: Optional debug label

---

createCommandEncoder

createCommandEncoder(descriptor?: GPUCommandEncoderDescriptor): GPUCommandEncoder

Create a command encoder for recording GPU commands.

---

submit

submit(commandBuffers: GPUCommandBuffer[]): void

Submit command buffers to the GPU queue.

Example:

const encoder = context.createCommandEncoder();
// ... encode commands
const commandBuffer = encoder.finish();
context.submit([commandBuffer]);

---

writeBuffer

writeBuffer(buffer: GPUBuffer, data: ArrayBuffer | ArrayBufferView, bufferOffset?: number): void

Write data to a GPU buffer.

Parameters:

• buffer: Target GPU buffer
• data: Data to write
• bufferOffset: Offset in buffer (default: 0)

---

deferDestroy

deferDestroy(buffer: GPUBuffer | null | undefined): void

Queue a buffer for destruction after GPU work completes. Safe async cleanup pattern.

Example:

const tempBuffer = context.createBuffer({...});
context.submit(commandsUsingBuffer);
context.deferDestroy(tempBuffer);
// Buffer destroyed after GPU work completes

---

sync

async sync(): Promise<void>

Wait for all submitted GPU work to complete and flush deferred destructions.

---

destroy

destroy(): void

Release all GPU resources. After calling this, the context cannot be reused.

---

Properties

isInitialized

readonly isInitialized: boolean

Whether the context has been initialized.

---

Error Handling

import { 
  GPUContext, 
  WebGPUNotSupportedError, 
  DeviceInitializationError 
} from 'tiny-dl-inference';

try {
  const context = new GPUContext();
  await context.initialize();
} catch (error) {
  if (error instanceof WebGPUNotSupportedError) {
    console.log('Please use a WebGPU-compatible browser');
  } else if (error instanceof DeviceInitializationError) {
    console.log('GPU initialization failed:', error.message);
  }
}