[petsc-dev] metis config error on XT4

Wed Nov 30 17:21:21 CST 2011

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
"${CMAKE_CURRENT_SOURCE_DIR}" is in 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
>>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111130/30556d3c/attachment.html>