I see that you have installed MPICH2 in your home directory. This is fine except you need to make sure that you aren't mixing and matching that library with other MPI libs sitting in the system search area. If all your paths are not consistent throughout your builds (PETSC, and libMesh) you can end up with bizarre behavior at runtime because you may be compiling against one library but running against another with compatible symbol names. This might also explain why you might be seeing that spurious MPICH1 error that John eluded to earlier. The other thing that I do is set my compiler environment variables (CC, CXX, F90, FC, etc) to the mpi wrappers directly instead of using the 'with-mpi' options in our configure command. This is again to ensure consistency through the builds of libMesh and the libraries underneath it.
Here are a few other troubleshooting steps you may take: - remove the '--enable-shared' option from the configuration line - this may reveal problems with your libraries earlier than if you dynamically link at runtime - use the '-n' option to 'make' so that you can visually inspect the compile lines when compiling libMesh (you can check that the correct compiler wrapper is being called (mpicxx instead of g++, also you can test your search paths after seeing those lines) Let me know if any of those suggestions help - we regularly run on all versions of Ubuntu, multiple versions of OpenSuSE 11 and 12, and SLES with this exact set of libraries. Cody On Sat, Jul 10, 2010 at 8:03 PM, Yujie <recrusa...@gmail.com> wrote: > Thank you very much, Cody. > > My system is AMD X86_64 CPU. OS is Red Hat Enterprise 5.0. The > configure command is > ./configure --enable-shared --enable-complex --enable-mpi > --enable-petsc --enable-tecplot --enable-metis --enable-tetgen > --enable-gmv --with-mpi=/home/yujie/codes/mpich2/ --enable-amr > --enable-perflog --enable-vtk|tee configresult.txt" > " > > Best Regards, > Yujie > > On Sat, Jul 10, 2010 at 5:58 PM, Cody Permann <codyperm...@gmail.com> > wrote: > > Please send details on your system, OS and your configure options and we > may > > be able to see where the problem lies. > > > > Cody > > > > On Sat, Jul 10, 2010 at 11:39 AM, Derek Gaston <fried...@gmail.com> > wrote: > >> > >> I don't know exactly what the problem is here.... But I thought I > >> would let you know that we use MPICH2 with Petsc and libmesh > >> everyday... On both 64bit macs and Linux... And everything works fine. > >> > >> So it is possible. Sorry I've never seen this error before... Maybe > >> Cody can help.... > >> > >> Derek > >> > >> Sent from my iPad > >> > >> On Jul 10, 2010, at 7:10 AM, John Peterson > >> <peter...@cfdlab.ae.utexas.edu> wrote: > >> > >> > On Fri, Jul 9, 2010 at 9:47 PM, Yujie <recrusa...@gmail.com> wrote: > >> >> Dear Libmesh Developers > >> >> > >> >> I am updating MPICH from MPICH1 to MPICH2. I can compile PETSc and > >> >> Libmesh successfully. I also use "make test" to successfully test > >> >> PETSc. However, when I run my codes with mpiexec. I got the following > >> >> errors > >> >> " > >> >> 0 - <NO ERROR MESSAGE> : Could not convert index 1140850688 into a > >> >> pointer > >> >> The index may be an incorrect argument. > >> >> Possible sources of this problem are a missing "include 'mpif.h'", > >> >> a misspelled MPI object (e.g., MPI_COM_WORLD instead of > MPI_COMM_WORLD) > >> >> or a misspelled user variable for an MPI object (e.g., > >> >> com instead of comm). > >> >> [0] Aborting program ! > >> >> [0] Aborting program! > >> >> p0_17843: p4_error: : 9039 > >> >> p0_17842: p4_error: : 9039 > >> >> 0 - <NO ERROR MESSAGE> : Could not convert index 1140850688 into a > >> >> pointer > >> >> The index may be an incorrect argument. > >> >> Possible sources of this problem are a missing "include 'mpif.h'", > >> >> a misspelled MPI object (e.g., MPI_COM_WORLD instead of > MPI_COMM_WORLD) > >> >> or a misspelled user variable for an MPI object (e.g., > >> >> com instead of comm). > >> >> [0] Aborting program ! > >> >> [0] Aborting program! > >> >> p0_23019: p4_error: : 9039" > >> >> > >> >> Could you give me some comments? thanks a lot. > >> > > >> > Hi, > >> > > >> > If I recall correctly, this is a problem with the Fortran side of the > >> > mpich1 implementation storing the address of a pointer as a 4 byte > >> > integer. This does not work on 64-bit systems where pointers are 8 > >> > bytes...if you google the error message you'll get more information. > >> > > >> > You say you are switching to mpich2 though... so I'm not even sure why > >> > you would be seeing this error. My advice is to let petsc download > >> > MPI for you during it's installation. That has worked for me in the > >> > past. > >> > > >> > -- > >> > John > >> > > >> > > >> > > ------------------------------------------------------------------------------ > >> > This SF.net email is sponsored by Sprint > >> > What will you do first with EVO, the first 4G phone? > >> > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > >> > _______________________________________________ > >> > Libmesh-users mailing list > >> > Libmesh-users@lists.sourceforge.net > >> > https://lists.sourceforge.net/lists/listinfo/libmesh-users > >> > >> > >> > ------------------------------------------------------------------------------ > >> This SF.net email is sponsored by Sprint > >> What will you do first with EVO, the first 4G phone? > >> Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > >> _______________________________________________ > >> Libmesh-users mailing list > >> Libmesh-users@lists.sourceforge.net > >> https://lists.sourceforge.net/lists/listinfo/libmesh-users > > > > > ------------------------------------------------------------------------------ This SF.net email is sponsored by Sprint What will you do first with EVO, the first 4G phone? Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
