[petsc-dev] metis config error on XT4
Mark F. Adams
mark.adams at columbia.edu
Wed Nov 30 17:44:04 CST 2011
On Nov 30, 2011, at 6:30 PM, Stefano Zampini wrote:
> What version of cmake are you using?
>
I am using an older version. I need to install it myself.
Sean: don't worry about Franklin, I'll get to it next week.
> I had a similar problem for metis 5.0.1: I got metis compiled by using version of cmake 2.6 which is older than the requested one by dropping the request in Cmakelists
>
> 2011/12/1 Jack Poulson <jack.poulson at gmail.com>
> I'm not too surprised; I haven't looked at how ParMETIS is being used within PETSc. The problem is definitely that ParMETIS isn't properly locating its GKlib subproject, and I think that the issue is in the lines I pointed out, but I'm not sure as to what the appropriate replacement for i"${CMAKE_CURRENT_SOURCE_DIR}" isin your situation.
>
> Jack
>
>
> On Wed, Nov 30, 2011 at 5:18 PM, Mark F. Adams <mark.adams at columbia.edu> wrote:
> This did not seem to help.
> Mark
>
> On Nov 30, 2011, at 5:35 PM, Jack Poulson wrote:
>
>> I have actually run into a similar problem with ParMetis. I fixed it on my end by changing the ParMetis CMakeLists.txt lines
>>
>> set(GKLIB_PATH METIS/GKlib CACHE PATH "path to GKlib")
>> set(METIS_PATH METIS CACHE PATH "path to METIS")
>>
>> to
>>
>> set(GKLIB_PATH ${CMAKE_CURRENT_SOURCE_DIR}/metis/GKlib CACHE PATH "path to GKlib
>> ")
>> set(METIS_PATH ${CMAKE_CURRENT_SOURCE_DIR}/metis CACHE PATH "path to METIS")
>>
>> Your mileage may vary.
>>
>> Jack
>>
>> On Wed, Nov 30, 2011 at 4:32 PM, Aron Ahmadia <aron.ahmadia at kaust.edu.sa> wrote:
>> Mark,
>>
>> Sorry, no help here, except that cmake is pretty easy to install from source, so that might be my recommendation if you don't want to wait for help from their systems team.
>>
>> A
>>
>> On Thu, Dec 1, 2011 at 1:30 AM, Mark F. Adams <mark.adams at columbia.edu> wrote:
>> Thanks Tim, I do use gcc. And this is on Franklin at NERSC.
>>
>> Cmake is just kicking my ass, now I get (appended).
>>
>> I do have an old version of cmake here:
>>
>> madams/petsc-dev> which cmake
>> /usr/common/usg/cmake/2.8.3/bin/cmake
>>
>> any ideas?
>> Mark
>>
>> =============================================================================== Configuring METIS; this may take several minutes =============================================================================== *******************************************************************************
>> UNABLE to CONFIGURE with GIVEN OPTIONS (see configure.log for details):
>> -------------------------------------------------------------------------------
>> Error running configure on METIS: Could not execute "cd /global/u2/m/madams/petsc-dev/externalpackages/metis-5.0.2 && make distclean && make config prefix=/global/homes/m/madams/petsc-dev/arch-xt5-opt cc="cc" doubleprecision=1":
>> rm -rf build/Linux-x86_64
>> rm -rf build/Linux-x86_64
>> mkdir -p build/Linux-x86_64
>> cd build/Linux-x86_64 && cmake /global/u2/m/madams/petsc-dev/externalpackages/metis-5.0.2 -DCMAKE_VERBOSE_MAKEFILE=1 -DGKLIB_PATH= -DCMAKE_INSTALL_PREFIX=/global/homes/m/madams/petsc-dev/arch-xt5-opt -DCMAKE_C_COMPILER=cc -DMETIS_USE_DOUBLEPRECISION=1
>> -- The C compiler identification is GNU
>> -- The CXX compiler identification is GNU
>> -- Check for working C compiler: /opt/cray/xt-asyncpe/5.01/bin/cc
>> -- Check for working C compiler: /opt/cray/xt-asyncpe/5.01/bin/cc -- works
>> -- Detecting C compiler ABI info
>> -- Detecting C compiler ABI info - done
>> -- Check for working CXX compiler: /opt/gcc/4.5.3/bin/c++
>> -- Check for working CXX compiler: /opt/gcc/4.5.3/bin/c++ -- works
>> -- Detecting CXX compiler ABI info
>> -- Detecting CXX compiler ABI info - done
>> -- Configuring incomplete, errors occurred!
>> CMake Error at CMakeLists.txt:21 (include):
>> include could not find load file:
>>
>> /GKlibSystem.cmake
>>
>>
>> make: *** [config] Error 1
>>
>>
>>
>> On Nov 30, 2011, at 4:00 PM, Tim Gallagher wrote:
>>
>>> Without knowing how your system is configured, this may not help. But on the Cray systems we use (they have the modules package installed), if we are using a compiler other than GNU (Intel, PGI, etc) but we want to use cmake, we still have to load the gcc module (not the programming environment, just the module).
>>>
>>> On our machines, it means we may load the PrgEnv-Intel module (so cc points to icc, ftn points to ifort, etc) but we also have to load gcc/<version> to get cmake to work.
>>>
>>> Tim
>>>
>>> From: "Aron Ahmadia" <aron.ahmadia at kaust.edu.sa>
>>> To: "For users of the development version of PETSc" <petsc-dev at mcs.anl.gov>
>>> Sent: Wednesday, November 30, 2011 3:49:25 PM
>>> Subject: Re: [petsc-dev] metis config error on XT4
>>>
>>> looks like a broken cmake...
>>>
>>> What happens when you just run: cmake by itself?
>>>
>>> A
>>>
>>> On Wed, Nov 30, 2011 at 11:47 PM, Mark F. Adams <mark.adams at columbia.edu> wrote:
>>> I'm getting this error on Franklin at NERSC:
>>> *******************************************************************************
>>> UNABLE to CONFIGURE with GIVEN OPTIONS (see configure.log for details):
>>> -------------------------------------------------------------------------------
>>> Error running configure on METIS: Could not execute "cd /global/u2/m/madams/petsc-dev/externalpackages/metis-5.0.2 && make distclean && make config prefix=/global/homes/m/madams/petsc-dev/arch-xt5-opt cc="cc" doubleprecision=1":
>>> rm -rf build/Linux-x86_64
>>> rm -rf build/Linux-x86_64
>>> mkdir -p build/Linux-x86_64
>>> cd build/Linux-x86_64 && cmake /global/u2/m/madams/petsc-dev/externalpackages/metis-5.0.2 -DCMAKE_VERBOSE_MAKEFILE=1 -DGKLIB_PATH= -DCMAKE_INSTALL_PREFIX=/global/homes/m/madams/petsc-dev/arch-xt5-opt -DCMAKE_C_COMPILER=cc -DMETIS_USE_DOUBLEPRECISION=1
>>> cmake: /opt/gcc/4.5.3/snos/lib64/libstdc++.so.6: version `GLIBCXX_3.4.15' not found (required by cmake)
>>> make: *** [config] Error 1
>>> *******************************************************************************
>>>
>>> any ideas?
>>>
>>> Mark
>>>
>>>
>>>
>>
>>
>>
>>
>
>
>
>
>
> --
> Stefano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111130/35280eab/attachment.html>
More information about the petsc-dev
mailing list