That's not accessing memory, so there's no reason it couldn't be (reasonably) efficiently pipelined without decoding to µops.