I found that an explicitly decreased performance happened when I used the parallel setting mode in SMACv1 and SMACv2.