The computation of the CS decomposition is the key to the stable computation of the Generalized Singular Value Decomposition, and is also important in other applications. This paper describes our implementation of a technique to compute the CS decomposition, and investigates the transfer of this technique to systolic computer architectures with parallel computation.