This note develops an array algorithm for optimal filtering of discrete-time Markovian jump linear systems (DTMJLSs). The known advantages of this kind of algorithm, which was originally developed for normal state-space systems, remain valid when it is applied to linear systems subject to Markovian jumps. It is numerically more stable in the sense that it presents better conditioning and reduced dynamical range. A numerical example, based on fixed-point implementations, is presented in order to demonstrate the advantage of this algorithm.