ARM-optimized memcpy is real engineering: Android's bionic and ARM's optimized-routines repo carry multiple variants tuned per core, using the load/store-pair instructions you practice here. Apple's M-series memcpy performance is part of why the machines feel fast — bulk copies underlie everything.