Introduction Type Safety Tensor Expressions Tensor Indexing Einsum Einops Autograd Training a Model Profiling & Memory PyTorch Migration Best Practices Runtimes Performance PyTorch Compatibility Benchmarks DType Coverage

dawn.WGPUQueue

export interface WGPUQueue {
  /**
   * Opaque native handle to this queue.
   */
  readonly handle: QueueHandle;

  /**
   * Submit one or more command buffers for GPU execution.
   *
   * Commands are executed asynchronously. Use {@link onSubmittedWorkDone} to
   * wait for completion.
   *
   * @param commandBuffers - Array of command buffers to submit
   */
  submit(commandBuffers: WGPUCommandBuffer[]): void;

  /**
   * Write data to a buffer from the CPU.
   *
   * The data is copied from host memory to the GPU buffer.
   * This is a synchronous operation from the CPU perspective.
   *
   * @param buffer - Destination buffer (must have COPY_DST usage)
   * @param bufferOffset - Byte offset in buffer
   * @param data - Data to write (ArrayBuffer or typed array)
   */
  writeBuffer(buffer: WGPUBuffer, bufferOffset: number, data: ArrayBuffer | ArrayBufferView): void;

  /**
   * Wait for all submitted work to complete.
   *
   * @returns Promise that resolves when GPU is idle
   */
  onSubmittedWorkDone(): Promise<void>;
}

readonlyhandle(QueueHandle): – Opaque native handle to this queue.
submit((commandBuffers: WGPUCommandBuffer[]) => void): – Submit one or more command buffers for GPU execution. Commands are executed asynchronously. Use onSubmittedWorkDone to wait for completion.
writeBuffer((buffer: WGPUBuffer, bufferOffset: number, data: ArrayBuffer | ArrayBufferView) => void): – Write data to a buffer from the CPU. The data is copied from host memory to the GPU buffer. This is a synchronous operation from the CPU perspective.
onSubmittedWorkDone(() => Promise<void>): – Wait for all submitted work to complete.

GPU command queue.

Submits command buffers and data writes to the GPU for execution.