[SciPy-user] BLAS, LAPACK, ATLAS libraries.
David Cournapeau
david at ar.media.kyoto-u.ac.jp
Mon Jul 2 22:10:12 EDT 2007
fie.pye at atlas.cz wrote:
> Hello.
>
> I would like to install NumPy and SciPy modules. I compiled and install BLAS, LAPACK and ATLAS libraries. After installation and importing NumPy I obtained message:
>
>>>> import numpy
> Traceback (most recent call last):
> File "", line 1, in
> File "/usr/local/python/python2.5.1/lib/python2.5/site-packages/numpy/__init__.py", line 43, in
> import linalg
> File "/usr/local/python/python2.5.1/lib/python2.5/site-packages/numpy/linalg/__init__.py", line 4, in
> from linalg import *
> File "/usr/local/python/python2.5.1/lib/python2.5/site-packages/numpy/linalg/linalg.py", line 25, in
> from numpy.linalg import lapack_lite
> ImportError: liblapack.so: cannot open shared object file: No such file or directory
> It seams to me that BLAS and LAPACK installation is OK.
Well, it may not be :) If you do (once you exported LD_LIBRARY_PATH):
ldd
/usr/local/python/python2.5.1/lib/python2.5/site-packages/numpy/linalg/lapack_lite.so
What do you get ? The most useful information, actually, is to give us
the log of python setup.py config &> log when building numpy and scipy.
>
> # LAPACK installation Make.inc
> PLAT = _LINUX
> FORTRAN = gfortran
> OPTS = -funroll-all-loops -O3 -fpic
> DRVOPTS = $(OPTS)
> NOOPT =
Here you forgot NOOPT = -fpic. Then, how did you compile atlas ? You
should try first without atlas at all, I think: compile blas, lapack,
and check that python setup.py config picks up the libraries.
As a note, there are source rpms available for blas/lapack/atlas/scipy:
the binary rpms do not support CENTOS, but I think building the rpm on
CENTOS should not be too difficult. Everything is taken care of in the
rpms: options, compatible fortran ABI, full pic version of liblapack to
build complete LAPACK based on ATLAS, etc...
http://software.opensuse.org/download/home:/ashigabou/ (pick up source
rpm for FC 6 or 7, this should be the more similar to centos).
If this does not work, I would like to hear about it,
David
More information about the SciPy-User
mailing list