copy_fragment#

nvmath.device.copy_fragment(
src: OpaqueTensor,
dst: OpaqueTensor,
alignment=None,
)[source]#

A bidirectional copying method to copy data between register fragments and global memory tensors.

Parameters:

  • src (OpaqueTensor) – The source tensor to copy from.

  • dst (OpaqueTensor) – The destination tensor to copy to.

  • alignment (int, optional) – The alignment (in bytes) for the copy operation.

Refer to the cuBLASDx documentation for more details on how to use this function: https://docs.nvidia.com/cuda/cublasdx/api/other_tensors.html#copying-registers-tensors

On this page