Various quantized/gemm backends scopes

There a bunch of backends / dependencies supported by PyTorch for different cases:

  • fbgemm
  • fbgemm_gpu
  • nnpack
  • qnnpack
  • xnnpack
  • mkldnn / onednn
  • onnxruntime
  • tensorrt
  • cudnn
  • etc?

Is there anywhere a single docs summary page listing them all compactly and explaining for which usecases are these needed?

Thanks!