[Pw_forum] Compilation issue (phenom II and ACML)

Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu
Thu May 7 12:38:31 CEST 2009


On Thu, 2009-05-07 at 17:29 +0700, Vit wrote:
> I found the solution...
>  ./configure LDFLAGS="-L/opt/acml4.2.0/gfortran64/lib/ -lacml" works

> (for MP LDFLAGS="-L/opt/acml4.2.0/gfortran64_mp/lib/ -lacml_mp" , but
> as the result I have

victor,
have you checked the ACML docs? most likely, you have to add -lpthread
to link with -lacml_mp. last time i checked, the threading support in
-lacml_mp was quite poor, so i would not bother trying to make it work.

> The following libraries have been found:
>   BLAS_LIBS=
>   LAPACK_LIBS= -llapack
>   FFT_LIBS=
> Please check if this is what you expect.
> 
> As I see QE tryes to use lapack from the system, but not ACML and FFT
> is from internal (DFLAGS =  -D__GFORTRAN -D__FFTW
> -D__USE_INTERNAL_FFTW).
> ACML can provede blas, lapack, fft libraryes. Why did it recognize only blas?

the configure script only uses what works with the embedded test scripts
and what is supported by Q-E. ACML has _a_ FFT, but the API is different
from what is supported. you'd have to program an interface. but don't
worry about that, the FFTW library (even the internal one) is quite
efficient and beats "vendor optimized" versions on a regular basis. 

cheers,
   axel.

> Best wishes,
> Koroteev Victor,
> Nikolaev Institute of Inorganic Chemistry SB RAS.
> 
> On Mon, May 4, 2009 at 6:19 AM, Axel Kohlmeyer
> <akohlmey at cmm.chem.upenn.edu> wrote:
> > On Sun, May 3, 2009 at 8:36 AM, Vit <vitruss at gmail.com> wrote:
> >> Good day, everyone!
> >> I'm trying to use PWSCF on my own computer with Phenom II x4 with Arch
> >> Linux onboard. The User guide says that I can get huge benefit from
> >> using ACML from AMD.
> >> I've downloaded and installed it. Now I'm trying to compile QE with
> >> this library. If I use just "configure" it uses internal BLAS, LAPACK
> >> and FFTW libraries. If I try
> >> LDFLAGS="-L/opt/acml4.2.0/gfortran64_mp/lib/ -lacml" it says:
> >
> > hmmm... if you want to run with MPI, you should not use
> > the multi-threaded ACML but the serial one. see the comments
> > on using OMP_NUM_THREADS=1 with intel's MKL v10.
> >
> >> checking build system type... x86_64-unknown-linux-gnu
> >> checking architecture... amd64
> >> checking for Fortran 77 compiler default output file name...
> >> configure: error: Fortran 77 compiler cannot create executables
> >> See `config.log' for more details.
> >>
> >> What's wrong?
> > [...]
> >
> >> configure:1695: checking for Fortran 77 compiler default output file
> >> name
> >> configure:1698: mpif77  -L/opt/acml4.2.0/gfortran64_mp/lib/ -lacml
> >> conftest.f  >&5
> >> /usr/bin/ld: cannot find -lacml
> >> collect2: ld returned 1 exit status
> >
> > looks like your ACML installation does not work. did you test
> > it on a simpler program before?
> >
> > cheers,
> >   axel
> >
> >> configure:1701: $? = 1
> >> configure: failed program was:
> >> |       program main
> >> |
> >> |       end
> >> configure:1740: error: Fortran 77 compiler cannot create executables
> >> See `config.log' for more details.
> >
> >
> >
> > --
> > =======================================================================
> > Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
> >  Center for Molecular Modeling   --   University of Pennsylvania
> > Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
> > tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
> > =======================================================================
> > If you make something idiot-proof, the universe creates a better idiot.
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://www.democritos.it/mailman/listinfo/pw_forum
> >
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.




More information about the users mailing list