This version has been used to compute the first large model using the B65536 (alphas=5) and B32768 (alphas=0) matrices.
