[SciPy-dev] installation problems with SciPy

Giovanni Samaey Giovanni.Samaey at cs.kuleuven.ac.be
Fri Jun 25 12:10:49 CDT 2004


Pearu Peterson wrote:

>On Fri, 25 Jun 2004, Giovanni Samaey wrote:
>
>  
>
>>just changing LD_LIBRARY_PATH did not make a difference.
>>
>>I tried reinstalling scipy with the LD_LIBRARY_PATH set, and somehow
>>this installation crashed with the following final lines:
>>
>>building extension "scipy.io.numpyio" sources
>>building extension "scipy.linalg.fblas" sources
>>generating fblas interface
>>20
>>error: build/src/atlas321/fblas.pyf: No such file or directory
>>
>>Now I am confused again... How did I break this?
>>    
>>
>
>Did you remove build directory before reinstalling scipy?
>  
>
I did.  Apparently the problem was indeed that the atlas installation on 
our systems
is outdated.  (Like almost all other software that is managed by the 
system group ;-) )

Installing the atlas binary in some directory dir/ and pointing both 
LD_LIBRARY_PATH
and ATLAS environment variables to dir/ worked.  Now everything runs 
smoothly!

(I tried installing the binary myself before but this didn't work.  This 
will probably also
have been due to not setting LD_LIBRARY_PATH.)

Now the output of python -c 'import atlas_version' is

ATLAS version 3.6.0 built by pearu on Sun Apr 11 01:13:26 EEST 2004:
   UNAME    : Linux p4.home.net 2.6.0-p4.10 #1 SMP Fri Jan 2 13:37:10 
EET 2004 i686 GNU/Linux
   INSTFLG  :
   MMDEF    : /home/pearu/src/ATLAS/CONFIG/ARCHS/P4SSE2/gcc/gemm
   ARCHDEF  : /home/pearu/src/ATLAS/CONFIG/ARCHS/P4SSE2/gcc/misc
   F2CDEFS  : -DAdd_ -DStringSunStyle
   CACHEEDGE: 1048576
   F77      : /home/pearu/bin/g77, version GNU Fortran (GCC) 3.3.3 
(Debian 20040314)
   F77FLAGS : -fomit-frame-pointer -O -fno-second-underscore
   CC       : /home/pearu/bin/gcc, version gcc-3.3 (GCC) 3.3.3 (Debian 
20040314)
   CC FLAGS : -fomit-frame-pointer -O3 -funroll-all-loops
   MCC      : /home/pearu/bin/gcc, version gcc-3.3 (GCC) 3.3.3 (Debian 
20040314)
   MCCFLAGS : -fomit-frame-pointer -O

which looks like it is what it was supposed to be...

Thanks for the patient help!  I would never have done this on my own!

Bye,

Giovanni




More information about the Scipy-dev mailing list