require_vector_vs; require_rv64; P_LOOP(32, { int32_t aop = P_SH(ps1, 0); int32_t bop = P_SH(ps2, 1); if ((INT16_MIN != aop) | (INT16_MIN != bop)) { pd = aop * bop; pd >>= 15; } else { pd = INT16_MAX; P_SET_OV(1); } pd = (int16_t)pd; })