[SciPy-User] site.cfg problems : scipy can't find BLAS

josef.pktd@gmai... josef.pktd@gmai...
Thu Apr 8 15:04:10 CDT 2010


On Thu, Apr 8, 2010 at 3:18 PM, John Reid <j.reid@mail.cryst.bbk.ac.uk> wrote:
> josef.pktd@gmail.com wrote:
>> On Thu, Apr 8, 2010 at 8:38 AM, John Reid <j.reid@mail.cryst.bbk.ac.uk> wrote:
>>> Hi,
>>>
>>> I'm trying to build numpy and scipy from the latest svn source. I have
>>> followed the instructions in site.cfg.example to configure my own
>>> site.cfg and I'm using the same site.cfg for both numpy and scipy. numpy
>>> builds and installs fine and the tests pass. scipy won't build and
>>> throws out lots of warnings about the BLAS libraries and source. I have
>>> GotoBLAS2 installed in /usr/local/lib but it doesn't seem to find it.
>>>
>>> Here is my site.cfg:
>>> # Defaults
>>> # ========
>>> # The settings given here will apply to all other sections if not
>>> overridden.
>>> # This is a good place to add general library and include directories like
>>> # /usr/local/{lib,include}
>>> #
>>> [DEFAULT]
>>> library_dirs =
>>> /usr/local/lib:/home/john/local/src/SuiteSparse-3.4.0/UMFPACK/Lib:/home/john/local/src/SuiteSparse-3.4.0/AMD/Lib
>>> include_dirs =
>>> /usr/local/include:/home/john/local/src/SuiteSparse-3.4.0/AMD/Include:/home/john/local/src/SuiteSparse-3.4.0/UMFPACK/Include
>>
>> I don't know Linux nor gotblas, but I think that here you are telling
>> numpy distutils that all libraries and include are under AMD and
>> UMFPACK
>
> I do have /usr/local/lib at the front of these library lists. That is
> where the libgoto.a and libgoto.so are stored. How else should I tell it
> where these libraries are?

My mistake, I didn't know : works as separator
the only other thing, I can think of, is whether it looks for the
correct filenames, is goto2_penrynp-r1.13
the actual filename?

my impression is that
#[blas_opt]
#libraries = ptf77blas, ptcblas, atlas
#
#[lapack_opt]
#libraries = lapack, ptf77blas, ptcblas, atlas

refer to actual filenames of the libraries (maybe minus prefix and extension).

But I'm usually debugging by trial and error (and I'm not too bad on
it on Windows), so the guess in your case might be as bad as my first
guess.

Josef

>
> Thanks,
> John.
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>


More information about the SciPy-User mailing list