[SYCL][CUDA] memory access across peer devices

[zjin-lcf]

I tried to run DPCT on a CUDA sample and it shows the following message for the function cudaDeviceCanAccessPeer(&can_access_peer, i, j);

SYCL currently does not support memory access across peer devices.

[JackAKirk]

There is not yet a corresponding function to cudaDeviceCanAccessPeer in DPC++.

See my answer here:

#6749 (comment)

and discussion here:

#6104

for a summary for the P2P status.