i have discovered a problem running the latest SPSS v16.0.1 software when using python extensions to run their PLS (partial least squares) statistic. SPSS requires numpy and scipy packages. SPSS ships with
python 2.5.1
numpy 1.0.1
scipy 0.5.2
i have tried these and the newer
python 2.5.2
numpy 1.1.0
scipy 0.6.0
but SPSS always crashes when running the PLS command, reporting a fault in the
calc_lwork.pyd
routine. interestingly, this only seems to happen on older PCs, such as PII and AMD-K6-3D; it works fine on a PIV machine. i noticed that the latest numpy install checks for CPU optimisation switches during the install; scipy install does not.
is it possible that the current compilation/install of scipy is not suited to these older processors? if so, is it likely that a similar CPU check as carried out by numpy install could be incorporated into scipy installation?
We have removed CPU-specific optimization in numpy.distutils (upcoming 1.3.0), and building scipy with this version of numpy should not cause any trouble anymore.
Original ticket http://projects.scipy.org/scipy/ticket/710 on 2008-07-31 by trac user pjbenson, assigned to unknown.
i have discovered a problem running the latest SPSS v16.0.1 software when using python extensions to run their PLS (partial least squares) statistic. SPSS requires numpy and scipy packages. SPSS ships with
python 2.5.1 numpy 1.0.1 scipy 0.5.2
i have tried these and the newer
python 2.5.2 numpy 1.1.0 scipy 0.6.0
but SPSS always crashes when running the PLS command, reporting a fault in the
calc_lwork.pyd
routine. interestingly, this only seems to happen on older PCs, such as PII and AMD-K6-3D; it works fine on a PIV machine. i noticed that the latest numpy install checks for CPU optimisation switches during the install; scipy install does not.
is it possible that the current compilation/install of scipy is not suited to these older processors? if so, is it likely that a similar CPU check as carried out by numpy install could be incorporated into scipy installation?