Improve performance of Python integrator (NVE_Opt version)

class NVE_Opt(LAMMPSIntegrator):

    """ Tuned Python implementation of fix/nve """

    """ Performance-optimized Python implementation of fix/nve """

    def __init__(self, ptr, group_name="all"):

        super(NVE_Opt, self).__init__(ptr)

        assert(self.group_name == "all")

        mass = self.mass

        dtfm = dtf / np.take(mass, atype)

        dtfm.reshape((nlocal, 1))

        for i in range(x.shape[0]):

            vi = v[i,:] 

            vi += dtfm[i] * f[i,:]

            x[i,:] += dtv * vi

        for d in range(x.shape[1]):

            v[:,d] += dtfm[:,0] * f[:,d]

            x[:,d] += dtv * v[:,d]

    def final_integrate(self):

        nlocal = self.lmp.extract_global("nlocal", 0)

        mass = self.mass

        dtfm = dtf / np.take(mass, atype)

        dtfm.reshape((nlocal, 1))

        for i in range(v.shape[0]):

            v[i,:] += dtfm[i] * f[i,:]

        for d in range(v.shape[1]):

            v[:,d] += dtfm[:,0] * f[:,d]