AGX - Volta based - concurrent memory access


I have a AGX - Pascal Based, currently it is impossible to make concurrent READ access on memory by CPU/GPU simultaneously (with unify memory). I mean.
CPU does: C = A + B and GPU does D = A * B, I get segfault.

Does the AGX - Volta Based can do such think ?

All the best,


UVM is supported only on dGPU (Discrete GPU from Nvidia). Not on Integrated GPU.
This device attribute cudaDevAttrConcurrentManagedAccess can be checked to see if concurrent accesses can be made