A systolic array for computing B A-1

This paper is devoted to the design of a new systolic array of n (n + 1) elementary processors which computes B A
^-1
within 4n + p - 2 time steps, where A is a dense (nonsingular) n by n matrix, B is a dense p by n matrix. The array can be directly extended to compute the vector Y = B A
^-1
R, where R is a vector with n components. Such computations arise frequently in signal processing applications.