[SciPy-user] SCIPY install with Intel Math Kernel and Intel Fortran

Patrick Marsh patrickmarshwx@gmail....
Sat Jun 20 10:02:43 CDT 2009


Here is the result of the print statement:
<numpy.distutils.fcompiler.intel.IntelItaniumFCompiler instance at
0x2000000002413440> None

Initially, I tried to build using gfortran and the MKL but to no avail
(it failed...error is below).  I figured that MKL required the Intel
compilers, which is how I started down this path.

/usr/bin/gfortran -Wall -Wall -shared
build/temp.linux-ia64-2.5/build/src.linux-ia64-2.5/scipy/lib/lapack/calc_lworkmodule.o
build/temp.linux-ia64-2.5/build/src.linux-ia64-2.5/fortranobject.o
build/temp.linux-ia64-2.5/scipy/lib/lapack/calc_lwork.o
-L/opt/intel/Compiler/11.0/069/mkl/lib/64 -Lbuild/temp.linux-ia64-2.5
-lmkl_lapack32 -lmkl_lapack64 -lmkl -lpthread -lgfortran -o
build/lib.linux-ia64-2.5/scipy/lib/lapack/calc_lwork.so
/usr/lib/gcc/ia64-suse-linux/4.1.2/../../../../ia64-suse-linux/bin/ld:
cannot find -lmkl_lapack32
collect2: ld returned 1 exit status
/usr/lib/gcc/ia64-suse-linux/4.1.2/../../../../ia64-suse-linux/bin/ld:
cannot find -lmkl_lapack32
collect2: ld returned 1 exit status
error: Command "/usr/bin/gfortran -Wall -Wall -shared
build/temp.linux-ia64-2.5/build/src.linux-ia64-2.5/scipy/lib/lapack/calc_lworkmodule.o
build/temp.linux-ia64-2.5/build/src.linux-ia64-2.5/fortranobject.o
build/temp.linux-ia64-2.5/scipy/lib/lapack/calc_lwork.o
-L/opt/intel/Compiler/11.0/069/mkl/lib/64 -Lbuild/temp.linux-ia64-2.5
-lmkl_lapack32 -lmkl_lapack64 -lmkl -lpthread -lgfortran -o
build/lib.linux-ia64-2.5/scipy/lib/lapack/calc_lwork.so" failed with
exit status 1


---
Patrick Marsh
Graduate Research Assistant
School of Meteorology
University of Oklahoma
http://www.patricktmarsh.com




On Sat, Jun 20, 2009 at 12:17 AM, David Cournapeau<cournape@gmail.com> wrote:
> On Sat, Jun 20, 2009 at 1:54 PM, Patrick Marsh<patrickmarshwx@gmail.com> wrote:
>> Greetings,
>>
>> I have run into a problem where I can't build Scipy using the Intel
>> Fortran compilers and the Intel Math Kernel.  I have (at least I
>> believe so) set up my environment as is directed by Intel, but I can't
>> seem to get it to work.  The install routine find the ifort
>> executable, but then fails later.  I've spent many hours banging my
>> head against this so any help is appreciated.
>
> Unfortunately, Itanium is not a well tested platform, since not many
> developers have access to one. From your build log, the problem is not
> MKL, but the fortran compiler not being properly setup. But this is
> hard to debug without being able to run the build by ourselves,
> unfortunately.
>
> From the log, I would first look at the fcompiler customization in the
> build_ext command. In the installed numpy, in
> numpy/distutils/commands/build_ext, there is the following code in the
> run method:
>
>        if need_f77_compiler:
>            ctype = self.fcompiler
>            self._f77_compiler = new_fcompiler(compiler=self.fcompiler,
>                                               verbose=self.verbose,
>                                               dry_run=self.dry_run,
>                                               force=self.force,
>                                               requiref90=False,
>                                               c_compiler=self.compiler)
>            fcompiler = self._f77_compiler
>            if fcompiler:
>                ctype = fcompiler.compiler_type
>                fcompiler.customize(self.distribution)
>            if fcompiler and fcompiler.get_version():
>                fcompiler.customize_cmd(self)
>                fcompiler.show_customization()
>            else:
>                self.warn('f77_compiler=%s is not available.' %
>                          (ctype))
>
> Could you add a print fcompiler, fcompiler.get_version() just before
> the "if fcompiler and fcompiler.get_version()" ? This should confirm
> whether it fails where I think it fails (i.e. the get_version)
>
> cheers,
>
> David
> _______________________________________________
> SciPy-user mailing list
> SciPy-user@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>


More information about the SciPy-user mailing list