cudaBindTexture2D vs cudaBindTextureToArray

In the topic posted in 2009, 2D locality caching works on cudaBindTextureToArray only.

(cudaBindTexture2D vs cudaBindTextureToArray)

  1. Is the 2D locality caching works on cudaBindTexture2D today?
  2. The buffer pointer parameter of cudaBindTexture2D, could be a unified memory or not?
  3. In jetson nano, How to speed up global memory multi-reading in 2D pattern?

Thanks for help!