[SciPy-user] Building Scipy on Dell 64bit Xeon

Nicolas Pernetty nicopernetty at yahoo.fr
Tue Jul 12 17:46:09 CDT 2005


On Tue, 12 Jul 2005 14:51:08 -0700 (PDT), David Noriega
<neruocomp at yahoo.com> wrote :

> Well I did some more looking around and found that my
> problem was that I didnt have the 64bit package for
> libg2c.  Its a bit weird with this system since it
> installs both i386 and x86_64 for some packages.  Well
> that fixed this problem,

Good !


> I have a new one. It seems I
> need to build the ATLAS libraries with -fPIC
> 
> /usr/bin/g77 -shared
> build/temp.linux-x86_64-2.4/build/src/Lib/optimize/lbfgsb-0.9/_lbfgsb
> module.o build/temp.linux-x86_64-2.4/build/src/fortranobject.o
> build/temp.linux-x86_64-2.4/Lib/optimize/lbfgsb-0.9/routines.o
> -L/usr/local/lib/atlas -Lbuild/temp.linux-x86_64-2.4
> -llapack -lf77blas -lcblas -latlas -lg2c -o
> build/lib.linux-x86_64-2.4/scipy/optimize/_lbfgsb.so
> /usr/bin/ld:
> /usr/local/lib/atlas/liblapack.a(dpotrf.o): relocation
> R_X86_64_32 against `a local symbol' can not be used
> when making a shared object; recompile with -fPIC
> /usr/local/lib/atlas/liblapack.a: could not read
> symbols: Bad value

-fPIC again...

> 
> I'll keep you posted on my progress.  I also wanted to
> point out something in scipy_distutils.  The way
> cpuinfo.py is currently setup, it will not find that
> this system uses a Xeon processor.  also it will fail
> to recongize it as having sse3.  Looking at
> /proc/cpuinfo, Xeon is not in all caps and sse3, while
> it should be there, only comes up as ss.  But that
> might just be a kernel thing as I fixed it in
> cpuinfo.py.

I've just managed to compile SciPy on a RedHat Advanced Server 3 on an
IA-64, so I guess you're near to succeed too !
I've also tried to use the default atlas and lapack given by RedHat but
it somehow didn't fit (couldn't remember the errors though).
After having much trouble compiling both for Solaris and HP-UX, doing it
on a Linux box was much more simpler (didn't say easy), so I recompile
atlas and lapack and managed to compile SciPy.

Numeric is really simple to compile, could you try to compile it with
your atlas/lapack to see if you still got the error ? (you have to
manually edit setup.py to indicate paths to atlas/lapack) That would
prove that the problem is inside atlas/lapack.

If that's the case I could help you compiling them.

Keep us informed !

Regards,

> --- David Noriega <neruocomp at yahoo.com> wrote:
> 
> > Well thats easier said then done on a fedora/redhat
> > system.  I know enought to get rpmbuild to rebuild
> > gcc
> > for me, but how would I get it to recompile gcc with
> > -fPIC.  Any ideas as to which file I should edit and
> > where to put "-fPIC"?
> > 
> > --- Nicolas Pernetty <nicopernetty at yahoo.fr> wrote:
> > 
> > > On Tue, 12 Jul 2005 08:07:10 -0700 (PDT), David
> > > Noriega
> > > <neruocomp at yahoo.com> wrote :
> > > 
> > > > Hello, I've been put in charge of setting up a
> > > Dell
> > > > 64bit Xeon workstation and on the side I've been
> > > tring
> > > > to get scipy to install, but I haven't had any
> > > luck. 
> > > > I compiled LAPACK, BLAS, and ATLAS sucsesfuly
> > and
> > > > scipy finds them.  Its when I try to build scipy
> > > that
> > > > I get problems.  I try just plain "./setup.py
> > > build"
> > > > and I get these errors:
> > > 
> > > Hello,
> > > 
> > > > /usr/bin/ld:
> > > >
> > >
> >
> /usr/lib/gcc-lib/x86_64-redhat-linux/3.2.3/libg2c.a(s_stop.o):
> > > > relocation R_X86_64_32 against `a local symbol'
> > > can
> > > > not be used when making a shared object;
> > recompile
> > > > with -fPIC
> > > 
> > > It seems that your gcc installation is somewhat
> > > messed up (libg2c hasn't
> > > been compiled with -fPIC)
> > > IMHO, best you can do to see if it's the problem
> > is
> > > to compile your own
> > > gcc (try 3.4.3 instead of 4.0.0 which seems a
> > little
> > > buggy). Don't be
> > > afraid : it's not that hard.
> > > 
> > > Regards,
> > > 
> > > _______________________________________________
> > > SciPy-user mailing list
> > > SciPy-user at scipy.net
> > > http://www.scipy.net/mailman/listinfo/scipy-user
> > > 
> > 
> > 
> > Don't fear that philosophy's an impious way
> > --superstition's more likely to lead folk astray.
> > 
> > ~Lucretius, De rerum natura, Book One
> > 
> > http://mindbender.deviantart.com
> > 
> > __________________________________________________
> > Do You Yahoo!?
> > Tired of spam?  Yahoo! Mail has the best spam
> > protection around 
> > http://mail.yahoo.com 
> > 
> > _______________________________________________
> > SciPy-user mailing list
> > SciPy-user at scipy.net
> > http://www.scipy.net/mailman/listinfo/scipy-user
> > 
> 
> 
> Don't fear that philosophy's an impious way
> --superstition's more likely to lead folk astray.
> 
> ~Lucretius, De rerum natura, Book One
> 
> http://mindbender.deviantart.com
> 
> 
> 		
> __________________________________ 
> Yahoo! Mail 
> Stay connected, organized, and protected. Take the tour: 
> http://tour.mail.yahoo.com/mailtour.html 
> 
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.net
> http://www.scipy.net/mailman/listinfo/scipy-user



More information about the SciPy-user mailing list