A method of generating a low-discrepancy sequence at high speed including the steps of: (a) expanding an i-th coordinate value u.sup.(1) (1.ltoreq.i.ltoreq.k) of an n-th element (U.sub.n (1), . . . , U.sub.n (k)) of the low-discrepancy sequence in radix b, and storing m components generated by the expansion of each said coordinate value in sequence into a first storage, wherein the n is an integer equal to or more than 1 and the b is a prime number equal to or more than k; (b) expanding the number n in radix b, and storing the generated expansion (n.sub.m,.sup.(1), n.sub.m-1, . . . , n.sub.1) in sequence into a second storage; (c) scanning the second storage in sequence, and detecting a minimum order j, wherein n is not equal to b-1; (d) in response to the step (c), reading out j-th row components of a generator matrix T.sup.(1) for the i-th coordinate (1.ltoreq.i.ltoreq.k) of the low-discrepancy sequence, stored in advance, and the components generated by the expansion of said i-th coordinate value of the n-th element of the low-discrepancy sequence in sequence, and adding said j-th row components and the components generated by the expansion of the i-th coordinate value of th n-th element of the low-discrepancy sequence in modulo b, both read out in the same order, and storing the result of the addition in sequence into third storage.