Memcpy in ARM Assembly

Write an optimised memcpy in ARM assembly that copies word-at-a-time when both pointers are aligned and falls back to bytes otherwise.

Constraints

• Signature: void *memcpy(void *dst, const void *src, size_t n) with AAPCS registers.
• Copy words when both pointers are 4-byte aligned and at least 4 bytes remain.
• Handle the leading and trailing unaligned bytes.
• Must return the original destination pointer.
• Do not use LDM/STM with more than 4 registers.
• Preserve r4 and above if used.