Searching for New Convolutional Codes using the Cell Broadband Engine Architecture

The Bidirectional Efficient Algorithm for Searching code Trees (BEAST), which is an algorithm to efficiently determine the free distance and spectral components of convolutional encoders, is implemented for the Cell Broadband Engine Architecture, efficiently utilizing the underlying hardware. Exhaustive and random searches are carried out, presenting new rate R = 1/2 convolutional encoding matrices with memory m = 26-29 and larger free distances and/or fewer spectral components than previously known encoding matrices of same rate and complexity. The main result of this paper consists in determining the previously unknown optimum free distance convolutional code with memory m = 26.