The C_ρ criterion^[l] is recently paid much attention to as an important one of variable selection. A parallel algorithm and the number of operations (multiplications and divisions) of selecting the best subset from a11 possible subsets，with regard to a large-scale linear regression model under C_ρ criterion，are presented. The simulation results for the algorithm on the vector super-computers YH-1 and YH-2 are given. The vector speed-up，ratio，s/v, approximates to fifteen and the advantages of the algorithm are shown.