November 2011 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] metis config error on XT4
by Mark F. Adams 30 Nov '11

30 Nov '11

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(a)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(a)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(a)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(a)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(a)kaust.edu.sa> >>> To: "For users of the development version of PETSc" <petsc-dev(a)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(a)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

1 0

Re: [petsc-dev] metis config error on XT4
by Jack Poulson 30 Nov '11

30 Nov '11

I am using very recent versions of CMake (2.8.5 or 2.8.6, depending on the machine). On Wed, Nov 30, 2011 at 5:30 PM, Stefano Zampini <stefano.zampini(a)gmail.com>wrote: > What version of cmake are you using? > > 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(a)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(a)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(a)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(a)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(a)kaust.edu.sa> >>>>> *To: *"For users of the development version of PETSc" < >>>>> petsc-dev(a)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(a)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 >

1 0

Re: [petsc-dev] metis config error on XT4
by Stefano Zampini 30 Nov '11

30 Nov '11

What version of cmake are you using? 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(a)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(a)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(a)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(a)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(a)kaust.edu.sa> >>>> *To: *"For users of the development version of PETSc" < >>>> petsc-dev(a)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(a)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

1 0

Re: [petsc-dev] metis config error on XT4
by Jack Poulson 30 Nov '11

30 Nov '11

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(a)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(a)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(a)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(a)kaust.edu.sa> >>> *To: *"For users of the development version of PETSc" < >>> petsc-dev(a)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(a)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 >>>> >>> >>> >>> >>> >>> >>> >> > >

1 0

Re: [petsc-dev] metis config error on XT4
by Mark F. Adams 30 Nov '11

30 Nov '11

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(a)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(a)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(a)kaust.edu.sa> >> To: "For users of the development version of PETSc" <petsc-dev(a)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(a)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 >> >> >> > > > >

1 0

Re: [petsc-dev] metis config error on XT4
by Sean Farley 30 Nov '11

30 Nov '11

> > No, but perhaps I should. My understanding is that those lines only need > to be changed when ParMetis is included as a subproject, as they seem to > have configured the build system to use relative paths. > I would still have to be the one to apply the patch since PETSc is using a heavily patched version of ParMETIS. Mark, I'll try this out on Franklin later tonight or tomorrow.

1 0

Re: [petsc-dev] metis config error on XT4
by Barry Smith 30 Nov '11

30 Nov '11

On Nov 30, 2011, at 4:44 PM, Jed Brown wrote: > On Wed, Nov 30, 2011 at 16:40, Jack Poulson <jack.poulson(a)gmail.com> wrote: > No, but perhaps I should. My understanding is that those lines only need to be changed when ParMetis is included as a subproject, as they seem to have configured the build system to use relative paths. > > That's a misuse of CMake. Normally you build from a separate build directory (so that you can clean reliably and independently between configurations). That ParMetis misuses CMake by default isn't an excuse for not allowing it to be used properly. A well designed buildsystem would not allow itself to be misused.

1 0

Re: [petsc-dev] metis config error on XT4
by Jed Brown 30 Nov '11

30 Nov '11

On Wed, Nov 30, 2011 at 16:40, Jack Poulson <jack.poulson(a)gmail.com> wrote: > No, but perhaps I should. My understanding is that those lines only need > to be changed when ParMetis is included as a subproject, as they seem to > have configured the build system to use relative paths. That's a misuse of CMake. Normally you build from a separate build directory (so that you can clean reliably and independently between configurations). That ParMetis misuses CMake by default isn't an excuse for not allowing it to be used properly.

1 0

Re: [petsc-dev] metis config error on XT4
by Jack Poulson 30 Nov '11

30 Nov '11

On Wed, Nov 30, 2011 at 4:36 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > On Wed, Nov 30, 2011 at 16:35, Jack Poulson <jack.poulson(a)gmail.com>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") >> > > Did you send George this patch? > No, but perhaps I should. My understanding is that those lines only need to be changed when ParMetis is included as a subproject, as they seem to have configured the build system to use relative paths. Jack

1 0

Re: [petsc-dev] metis config error on XT4
by Jed Brown 30 Nov '11

30 Nov '11

On Wed, Nov 30, 2011 at 16:35, Jack Poulson <jack.poulson(a)gmail.com> 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") > Did you send George this patch?

1 0