Yes, sorry, I meant SSE intrinsics. As I think I note later in the article, the compiler I used emitted scalar SSE2 for floating point.