s_bmul_2x2
void s_bmul_2x2(mp_digit *r, const mp_digit a1, const mp_digit a0, const mp_digit b1,
s_bmul_2x2(zm, a1, a2^a0, b1, b2^b0); /* fill middle 4 words */
s_bmul_2x2(r, a1, a0, b1, b0); /* fill bottom 4 words */
s_bmul_2x2(r+4, a3, a2, b3, b2); /* fill top 4 words */
s_bmul_2x2(zm, a3^a1, a2^a0, b3^b1, b2^b0); /* fill middle 4 words */
s_bmul_2x2(r, a1, a0, b1, b0); /* fill bottom 4 words */