loop_spe_multiprocessing not parallel any more

Since we updated to MPDAF v1.2 on some machines here, I find that a piece of code that I used to compute the S/N for Voronoi binning is not using more than 100% CPU any more, i.e. does not run in parallel.

Was something done to loop_spe_multiprocessing() to break it between 1.1.16 and 1.2 or do I have to use it differently?

def f_SN(spec, lmin, lmax):
    scut = spec.get_lambda(lmin, lmax = lmax)
    sn = scut.data / np.sqrt(scut.var)
    if scut.data.count() > 0 and sn.count() > 0:
        return sn.mean()
    return 0.

def SN_par(filename, outfilename):
    print 'Opening %s.' % filename
    cube = mpdaf.obj.Cube(filename)
    print 'Opened %s.' % filename
    # create image of mean S/N of the continuum
    sn = cube.loop_spe_multiprocessing(f = f_SN, lmin = 5600., lmax = 6500.)
    print 'Will write to %s.' % outfilename
    sn.write(outfilename)