require_vector_vs; P_LOOP(32, { int64_t addop = 0; int64_t mres = 0; bool sat = false; if ((INT32_MIN != ps1) | (INT16_MIN != P_SH(ps2, 1))) { mres = ((int64_t) ps1 * P_SH(ps2, 1)) << 1; addop = ((mres >> 15) + 1) >> 1; } else { addop = INT32_MAX; P_SET_OV(1); } pd = (sat_add(pd, addop, sat)); P_SET_OV(sat); })