[CP2K:8689] Re: Compilling problems
rdela... at gmail.com
rdela... at gmail.com
Tue Feb 14 13:03:45 UTC 2017
I use ACML because i red that for using MKL ifort compiler is needed, and i
don't have it.
Anyway I'm going to try ATLAS.
I'll tell you. Thank you
Rafa
El martes, 14 de febrero de 2017, 13:50:05 (UTC+1), sassy escribió:
>
> Hi Rafa,
>
> there are two things:
>
> a) why do you use ACML when you got an Intel CPU? ACML is for AMD Opterons
> and
> I am not sure whether it makes sense to use a high performance BLAS
> library
> which was tuned for a different chip family.
> If you want to stay clear of proprietary BLAS libs I would suggest to use
> OpenBLAS or ATLAS.
>
> b) it appears that your library is not in the search path for the
> libraries.
> There are three ways of fixing this.
> - $ export LD_LIBRARY_PATH=/path/to/libacml_mp.so:$LD_LIBRARY_PATH
> - add /path/to/libacml_mp.so in /etc/ld.so.conf and run ldconfig
> - use -Wl,--rpath=/path/to/libacml_mp.so when you compile for the
> linking.
>
> The first will only last for the current shell session and the second is a
> global setting. The third means you hard-code the path into your binary
> file.
>
> Like all things, there are pros and cons for all of that. ;-)
>
> All the best
>
> Jörg
>
> On Tuesday 14 Feb 2017 04:40:09 rdel... at gmail.com <javascript:> wrote:
> > Hello and thank you Jörg
> >
> > I compile the next arch file:
> >
> > CC = gcc
> > CPP =
> > FC = gfortran
> > LD = gfortran
> > AR = ar -r
> > ACML_INC = /home/ralf/ACML/gfortran64_mp/include
> > ACML_LIB = /home/ralf/ACML/gfortran64_mp/lib
> > FFTW_INC = /usr/local/include/
> > FFTW_LIB = /usr/local/lib/
> > LIBINT_INC = /usr/local/libint/1.1.4-stable/include
> > LIBINT_LIB = /usr/local/libint/1.1.4-stable/lib
> > LIBXC_INC = /opt/etsf/include
> > LIBXC_LIB = /opt/etsf/lib
> > DFLAGS = -D__FFTW3 -D__LIBINT -D__LIBXC2\
> > -D__LIBINT_MAX_AM=7 -D__LIBDERIV_MAX_AM1=6 -D__MAX_CONTR=4
> > CPPFLAGS =
> > FCFLAGS = $(DFLAGS) -O2 -ffast-math -ffree-form
> -ffree-line-length-none\
> > -fopenmp -ftree-vectorize -funroll-loops\
> > -mtune=native\
> > -I$(ACML_INC) -I$(FFTW_INC) -I$(LIBINT_INC) -I$(LIBXC_INC)
> > LDFLAGS = $(FCFLAGS) -static-libgfortran
> > LIBS = $(ACML_LIB)/libacml_mp.so\
> > $(FFTW_LIB)/libfftw3.a\
> > $(FFTW_LIB)/libfftw3_threads.a\
> > $(LIBXC_LIB)/libxcf90.a\
> > $(LIBXC_LIB)/libxc.a\
> > $(LIBINT_LIB)/libderiv.a\
> > $(LIBINT_LIB)/libint.a
> >
> > And when I run the tests I obtain in all 2500 cases this kind of error
> >
> >
> /home/ralf/cp2k-4.1/regtesting/make/ssmp/TEST-make-ssmp-2017-02-14_13-35-10/
>
> > optimize_input/regtest-1/driver-stride.inp.out
> > /home/ralf/cp2k-4.1/regtesting/make/ssmp/../../..//exe/make/cp2k.ssmp:
> > error while loading shared libraries: libacml_mp.so: cannot open shared
> > object file: No such file or directory
> > EXIT CODE: 127 MEANING: RUNTIME FAIL
> >
> > I should use libacml_mp.a for the compilling but in my installation this
> > file doesn appear
> >
> > Rafa
> >
> >
> > El martes, 14 de febrero de 2017, 13:11:16 (UTC+1), rdel... at gmail.com
> >
> > escribió:
> > > Hello all.
> > >
> > > I have been trying to compile and execute cp2k in my computer but i'm
> > > having problems.
> > >
> > > I use a modified arch file including my routes to files and libreries.
> > > When I compile it works but wher runing the tests it fails in all the
> > > cases.
> > >
> > > Could anyone help me with this? My computer is an intel i7-920 with 4
> > > cores and 8 threadings. I have a CUDA capable graphic card so i like
> to
> > > enable it (I have installed the last version of the nvidia CUDA kit)
> > > About the software i run ubuntu 16.04.4 with gfortran and gcc, both
> > > versions 5.4.0. Also I have installed the last version of the intel
> MKL
> > > library (aditionally i also have ACML). Finally I'd like to compile
> cp2k
> > > with libint, libxc and fftw. All three are compiled and installed in
> my
> > > computer
> > >
> > > Any help for the creation of the appropiate arch file in wellcome.
> > >
> > > That you all in advance.
> > >
> > > Rafa
>
>
> --
> *************************************************************
> Dr. Jörg Saßmannshausen, MRSC
> University College London
> Department of Chemistry
> 20 Gordon Street
> London
> WC1H 0AJ
>
> email: j.sas... at ucl.ac.uk <javascript:>
> web: http://sassy.formativ.net
>
> Please avoid sending me Word or PowerPoint attachments.
> See http://www.gnu.org/philosophy/no-word-attachments.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170214/45b373e8/attachment.htm>
More information about the CP2K-user
mailing list