Lots of memcpy() implementations are still more efficient than a dumb byte-by-byte copy. They'll copy the (unaligned) head and the tail in bytes, but the bulk of the data using whatever data type and method is fastest.