Question

since memcpy should be highly optimized nowadays, does it still make sense to optimize the copy of Ipv6 addresses using explicit loop unrolling ?

#include <netinet/in.h>

struct in6_addr IP_1;
struct in6_addr IP_2;
;
;
IP2.__in6_u.__u6_addr32[0] = IP1.__in6_u.__u6_addr32[0];
IP2.__in6_u.__u6_addr32[1] = IP1.__in6_u.__u6_addr32[1];
IP2.__in6_u.__u6_addr32[2] = IP1.__in6_u.__u6_addr32[2];
IP2.__in6_u.__u6_addr32[3] = IP1.__in6_u.__u6_addr32[3];

Note that the code above is best suited for 32-bit architectures.

Is there a best practice I do not know ?

Answer 1

You should just do IP2 = IP1;, and let the compiler deal with it.