[SciPy-user] lpmn - python crash with AMD processor

ennesnospam1@aeroakust... ennesnospam1@aeroakust...
Mon Jan 7 09:22:40 CST 2008


> >>> I tried on eight different machines with different scipy versions
<=
> >>> 0.6.0, all under WinXP SP 2 (32 bit). Two had Intel processors ->
> >>> everything fine. Six had AMDs (Athlon, Duron) -> segmentation
fault.
> >>> From your post it seems that it works fine under Linux.
Unfortunately
> >>> Linux is no option for me (at least in this case).
> >>>
> >
> >> What are the kind of processor you have and which binary did you
take ?
> >> For Athlon XP processors, you cannot take the usual binary bacause
it
> > uses
> >> SSE2 instructions which are not supported by Athlon XP processors.
You
> >> should take the binary for Pentium II (which does not support the
SSE2
> >> instruction set).
> >
> > On my present machine it is an AMD Sempron, but I had the same
problem
> > with other AMDs like AlthonXP and Duron. It is not very clear for me
> > _what_ binary you mean. The specfun.pyd? I have no means of
determining
> > what version it has (I found version "0.0" in the crash report, but
I
> > guess that would be of no help) . Python itself and any function in
> > scipy/numpy/ETS etc. works fine on all processors.
> > My scipy install is from the scipy-0.6.0.0004_s-py2.5-win32.egg
(which
> > is not marked as exclusively built for Intel), but I found the same
> > behaviour in earlier versions.
> >
> > Where do I find this PII version you mentioned?
> >
> I think he was referring to the binary version of the ATLAS library
that
> you have installed on your system. Google for ATLAS pre-built binaries
> to find the place to go.
I doubt this will do the trick - all other, esp. numpy works fine. And,
from the source
http://svn.scipy.org/svn/scipy/trunk/scipy/special/specfun/specfun.f I
learn that there is no call to any foreign function from SUBROUTINE
LPMN (this has not much to do with linear algebra).

Thanks anyway,
Ennes






More information about the SciPy-user mailing list