> > Moreover, np.linalg.norm is slow compare to blas. > > Is there a way/plan to fix it? > > Xavier > Have a look at: http://fseoane.net/blog/2011/computing-the-vector-norm/ greetings Till