This can be more efficient than UNALIGNED_LOAD64 + UNALIGNED_STORE64 on some platforms, in particular ARM.
source not stored for this graph (policy: none)
no test coverage detected