using memcpy instead of packing individual elements is faster

Potential warning in GCC8 due to:
https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess