[petsc-users] Installation failure
Satish Balay
balay at mcs.anl.gov
Tue Jul 28 10:23:58 CDT 2020
>>>>>>>
CC linux-intel/obj/sys/error/fp.o
In file included from /usr/include/complex.h(152),
from /opt/intel/compilers_and_libraries_2016.3.210/linux/compiler/include/complex.h(22),
from /home/rmondaini/libraries/petsc-3.13.3/include/petscsystypes.h(234),
from /home/rmondaini/libraries/petsc-3.13.3/include/petscsys.h(36),
from /home/rmondaini/libraries/petsc-3.13.3/src/sys/error/fp.c(19):
/usr/include/x86_64-linux-gnu/bits/cmathcalls.h(55): error: identifier "_Float32" is undefined
__MATHCALL (cacos, (_Mdouble_complex_ __z));
^
CC linux-intel/obj/sys/error/errstop.o
In file included from /home/rmondaini/libraries/petsc-3.13.3/include/petscmath.h(13),
from /home/rmondaini/libraries/petsc-3.13.3/include/petscsys.h(322),
from /home/rmondaini/libraries/petsc-3.13.3/src/sys/error/errstop.c(2):
/opt/intel/compilers_and_libraries_2016.3.210/linux/compiler/include/math.h(1214): error: identifier "_LIB_VERSION_TYPE" is undefined
_LIBIMF_EXT _LIB_VERSIONIMF_TYPE _LIBIMF_PUBVAR _LIB_VERSIONIMF;
^
<<<<<
Looks like this version of intel compiler has issues with system (gcc) include files.
Can you use gcc/g++ instead for this build?
Satish
On Tue, 28 Jul 2020, Rubem Mondaini wrote:
> Thanks a bunch, Satish!
>
> I can configure PETSc now (see configure.log attached), but the compilation breaks down (see make.log). I am wondering if my problem is related to this: https://community.intel.com/t5/Intel-C-Compiler/LIB-VERSION-TYPE-is-undefined-icc-and-icps-errors-under-Debian/td-p/1160017
>
>
> > -----Original Messages-----
> > From: "Satish Balay" <balay at mcs.anl.gov>
> > Sent Time: 2020-07-28 13:38:09 (Tuesday)
> > To: "Rubem Mondaini" <rmondaini at csrc.ac.cn>
> > Cc: "Barry Smith" <bsmith at petsc.dev>, petsc-users at mcs.anl.gov
> > Subject: Re: [petsc-users] [SPAM] Re: Installation failure
> >
> > something went wrong - but petsc configure didn't capture this error.
> >
> > You can retry with:
> >
> > ./configure PETSC_ARCH=linux-intel --with-cc=mpiicc --with-cxx=0 --with-fc=mpiifort --with-blaslapack-dir=/opt/intel/compilers_and_libraries_2016/linux/mkl --with-scalar-type=complex --with-64-bit-indices
> >
> > And perhaps add one more option: LIBS="-Bstatic -lifcore -Bdynamic"
> >
> > Note: --with-64-bit-blas-indices is not appropriate here.
> >
> > Satish
> >
> > On Tue, 28 Jul 2020, Rubem Mondaini wrote:
> >
> > > Dear Barry,
> > >
> > > thank you very much for the reply. I removed the options "--with-mpi-include"
> > > and "--with-mpi-lib", but I still get the same error. I attach the
> > > configure.log file.
> > >
> > > I can confirm that I can compile and run other codes with my compilers,
> > > however.
> > >
> > > Many thanks!
> > >
> > > Rubem
> > >
> > > On 7/28/20 10:32 AM, Barry Smith wrote:
> > > > Hmm, configure is a mess.
> > > >
> > > > Since you explicitly use MPI compilers
> > > >
> > > > --with-cc=mpiicc --with-cxx=mpiicpc --with-fc=mpiifort
> > > >
> > > > the
> > > >
> > > > --with-mpi-include=/opt/intel/compilers_and_libraries_2016/linux/mpi/intel64/include
> > > > --with-mpi-lib=/opt/intel/compilers_and_libraries_2016.3.210/linux/mpi/intel64/lib/libmpicxx.a
> > > >
> > > > are not needed and should not be provided (Satish I thought this would
> > > > generate an error? Why doesn't it?)
> > > >
> > > > Please try without those arguments and send the configure.log if it
> > > > fails.
> > > >
> > > > The error message isn't completely right.
> > > >
> > > > Cannot run executables created with C.
> > > >
> > > > This is because the call the call to
> > > > self.setCompilers.checkCompiler('C') doesn't have enough context to
> > > > print a decent error message.
> > > >
> > > > Barry
> > > >
> > > >
> > > >> On Jul 27, 2020, at 2:06 AM, Rubem Mondaini <rmondaini at csrc.ac.cn> wrote:
> > > >>
> > > >> Dears,
> > > >>
> > > >> I have tried to install the PETSc library on my machine and I am failing. I
> > > >> have managed to install it on another one, which possess similar
> > > >> configurations for the compilers, and there things do seem to be very much
> > > >> fine.
> > > >>
> > > >> I am attaching the corresponding configure.log. Is there something simple
> > > >> that I am missing?
> > > >>
> > > >> Best,
> > > >>
> > > >>
> > > >> <configure.log>
> > >
>
More information about the petsc-users
mailing list