webgpu: Remove bufSize from GPUData

[qjia7]

The buffer size and usage are the attributes of GPUBuffer. So bufSize doesn't need to be provided extra in GPUData.

To see the logs from the Cloud Build CI, please join either our discussion or announcement mailing list.

[xhcao]

LGTM

[gyagp]

LGTM

[Linchenn]

Is bufSize still needed in

export interface GPUData {
  tensorRef: Tensor;
  texture?: WebGLTexture;
  buffer?: GPUBuffer;
  texShape?: [number, number];
  bufSize?: number;
}

[Linchenn]

Could you help update this doc, https://github.com/tensorflow/tfjs/blob/master/docs/OPTIMIZATION_PURE_GPU_PIPELINE.md? Or, it could be in a separate pr.

[gyagp]

Is bufSize still needed in

export interface GPUData {
  tensorRef: Tensor;
  texture?: WebGLTexture;
  buffer?: GPUBuffer;
  texShape?: [number, number];
  bufSize?: number;
}

Jiajia's PR only changes the API, which may impact the upcoming release. We will do the related refactoring after holidays.

[Linchenn]

Thanks! LGTM

[qjia7]

Could you help update this doc, https://github.com/tensorflow/tfjs/blob/master/docs/OPTIMIZATION_PURE_GPU_PIPELINE.md? Or, it could be in a separate pr.

PR #7651 fixes it. Thanks.

[Linchenn]

Is bufSize still needed in

export interface GPUData {
  tensorRef: Tensor;
  texture?: WebGLTexture;
  buffer?: GPUBuffer;
  texShape?: [number, number];
  bufSize?: number;
}

@qjia7

[qjia7]

Is bufSize still needed in

export interface GPUData {
  tensorRef: Tensor;
  texture?: WebGLTexture;
  buffer?: GPUBuffer;
  texShape?: [number, number];
  bufSize?: number;
}

@qjia7

bufSize is not needed anymore in GPUData. You can get the size from buffer.size. See here https://www.w3.org/TR/webgpu/#gpubuffer. size is one of the properties of GPUBuffer.

[Linchenn]

Thanks!