Even better / potentially more surprising:

    unsigned mult(unsigned x, unsigned y) {
        unsigned y0 = y;
        while (x--) y = add_v1(y, y0);
        return y;
    }

optimizes to:

    mult(unsigned int, unsigned int):
      madd w0, w1, w0, w1
      ret

(and this produces the same result when substituting any of the `add_vN`s from TFA)