how to perform memcpy in __device__ function

Accelerated Computing CUDA CUDA Programming and Performance
1

is there any way to perform memcpy-like in device function?

of just “=”?