What exactly does the managed memory flag do and what changes?

[cbuchner1]

have you tried a cudaDeviceSynchronize() following the kernel call?

Quoting Robert Crovella, from another thread

@Robert_Crovella

Jetson managed memory does not allow concurrent access. Therefore, you should remember to issue cudaDeviceSynchronize(), sometime after launching kernels, and before you intend to access managed data in host code.