[SciPy-User] scipy.stats

Bruce Southey bsouthey@gmail....
Tue Aug 23 13:04:30 CDT 2011


On Tue, Aug 23, 2011 at 7:14 AM, Rita <rmorgan466@gmail.com> wrote:
> Any ideas?
>
> On Sat, Aug 20, 2011 at 7:40 AM, Rita <rmorgan466@gmail.com> wrote:
>>
>> It should be
>>
>>
>> 'icc -L /opt/intel/lib/intel64 -L /opt/intel/ipp/em64t/lib -L
>> /opt/intel/mkl/lib/em64t -L /usr/lib64 -L /usr/lib -I/opt/intel/mkl/include
>> -I /opt/intel/include -fPIC -O3 -openmp -limf -lmkl_core -lmkl_intel_lp64
>> -lmkl_intel_thread     -lstdc++ -DMKL_ILP64'
>>
>> Here is how I am doing the compilation
>>
>>
>>
>>
>>
>> On Sat, Aug 20, 2011 at 6:38 AM, Rita <rmorgan466@gmail.com> wrote:
>>>
>>> Thanks Bruce. I have already seen this
>>> Here are more details of my build.
>>>
>>> My Intel compiler exists here, /opt/intel/
>>>
>>> self.cc_exe = 'icc -L /opt/intel/lib/intel64 -L /opt/intel/ipp/em64t/lib
>>> -L /opt/intel/mkl/lib/em64t -L /usr/lib64 -L /usr/lib -I
>>> /opt/intel/ipp/em64t/in    clude -I /etg/source/Linux/include -I
>>> /opt/intel/mkl/include -I /opt/intel/include -fPIC -O3 -openmp -limf
>>> -lmkl_core -lmkl_intel_lp64 -lmkl_intel_thread     -lstdc++ -DMKL_ILP64'
>>> Here is how I am doing the compilation
>>>
>>> CC=icc CXX=icpc AR=xiar /opt/python-2.7.2/bin/python setup.py config
>>> --compiler=intel  --fcompiler=intelem build_clib --compiler=intel
>>> --fcompiler=intelem build_ext --compiler=intel install
>>>
>>> /opt/intel/ipp is what I was using for the math library. This compiles
>>> but I keep getting that problem
>>>
>>> I use the same compile statement to compile scipy
>>>
>>>
>>> On Fri, Aug 19, 2011 at 8:00 PM, Bruce Southey <bsouthey@gmail.com>
>>> wrote:
>>>>
>>>> On Fri, Aug 19, 2011 at 5:53 PM, Rita <rmorgan466@gmail.com> wrote:
>>>> > I apologize for the vague question.
>>>> > OS: Linux
>>>> > Compiler: Intel compiler suite. Version 11 (this also includes fortran
>>>> > compiler)
>>>> > MKL: 10.3
>>>> > Numpy version: 1.6.1
>>>> > When I do numpy.config() I see it properly compiled against Intel's
>>>> > BLAS and
>>>> > LAPACK
>>>> >
>>>> > Where are the build logs located? Do you need to build log for Numpy
>>>> > also?
>>>> >
>>>> > On Fri, Aug 19, 2011 at 11:00 AM, Ralf Gommers
>>>> > <ralf.gommers@googlemail.com>
>>>> > wrote:
>>>> >>
>>>> >>
>>>> >> On Thu, Aug 18, 2011 at 12:20 PM, Rita <rmorgan466@gmail.com> wrote:
>>>> >>>
>>>> >>> I am trying to import scipy.stats but I keep getting an import
>>>> >>> Error,
>>>> >>> ...scipy/special/_cephes.so: undefined symbol: __libm_sse2_sincos
>>>> >>> I compiled Numpy with Intel C compiler and Scipy compiled ok but
>>>> >>> just
>>>> >>> cant get this working.
>>>> >>> Any advise?
>>>> >>
>>>> >> The symbol is defined in an Intel math library. You'll need to give
>>>> >> us
>>>> >> more details in order to say more than that. What exact compilers and
>>>> >> MKL
>>>> >> did you use, what OS? Build command and build log?
>>>> >>
>>>> >> Cheers,
>>>> >> Ralf
>>>> >>
>>>> >>
>>>> >>
>>>>
>>>> A quick google indicates that you need to ensure that you link to the
>>>> appropriate Intel Math library:
>>>>
>>>> http://software.intel.com/en-us/articles/unresolved-external-symbol-libm-sse2/
>>>>
>>>> Also what is the cpu type?
>>>>
>>>> Bruce
>>>> _______________________________________________
>>>> SciPy-User mailing list
>>>> SciPy-User@scipy.org
>>>> http://mail.scipy.org/mailman/listinfo/scipy-user
>>>
>>>
>>>
>>> --
>>> --- Get your facts first, then you can distort them as you please.--
>>
>>
>>
>> --
>> --- Get your facts first, then you can distort them as you please.--
>
>
>
> --
> --- Get your facts first, then you can distort them as you please.--
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
>

Just use Enthought's version:
http://www.enthought.com/products/epd.php

I do not use Intel's compiler so without more details it just appears
that you have not given icc the correct paths to the libraries it
needs when linking.

Bruce


More information about the SciPy-User mailing list