How does atomicity work with the memory accesses?
Can you give an example or perhaps assembly code.
I'm thinking I may have misunderstood the question as asking why the atomic instructions in ARMv8.1 are wanted when their work can be done using the acquire/release and exclusive load/store instructions on the base architecture. If you are asking instead why it is extremely desirable to be able to support atomic operations irrespective of the architecture. then here's an introduction and WIkipedia has lots of more detailed entries if one searches on the various terms
Atomic Operations in Hardware