Is it though? The memcpy implementations I remember reading about do all sorts of exotic things depending on the size, which processor, etc. As you mention, so

Is it though? The memcpy implementations I remember reading about do all sorts of exotic things depending on the size, which processor, etc. As you mention, sometimes it's still slower and depends on microcode. A dedicated memcpy instruction shouldn't have that issue right?