Memory layour and CUDA question

I am giving a talk on the Nano to the London Julia Meetup on Friday.
Can asomeoen point me to the memory setup on the Nano - I know thet the system memory is shared with the GPU so there is no need for memory copies across to a GPU.
I am working with the CUDAdrv.jl module in Julia.
The test fail when this routine is called cuMemHostRegister_v2
I think this may be due to the memory layout on the Nano.
I know this question should be better answered on a CUDA forum - please point me in the right direction.


In general, cudaHostRegister() is not supported on the ARM platforms.
This is because the caching attribute of an existing allocation can’t be changed on the fly.

The only special case is Jetson Xavier.
cudaHostRegister() is supported on the Xavier since the new I/O Coherence feature.


Thankyou for that. That is very helpful.