Re: [Libmesh-users] parallel run

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

This is a problem due to a MPI C++ binding conflict with stdlib.

to fix this easily try to add by any way the following flag to compilation :

=2DDMPICH_IGNORE_CXX_SEEK

this will fix the SEEK_SET error stuff

Le Jeudi 1 Juin 2006 00:11, Hae-Won Choi a =E9crit=A0:
> Hi, I have tried several different option of PetSc compiling.
> I tried gnu based mpi compiler for BGL.
> But again when I enable-petsc for libmesh,
> I got following error messages:
>
> haewon/libmesh> make
> Compiling C++ (in optimized mode) src/base/dof_map.C...
> In file included from
> /bgl/BlueLight/V1R2M1_020_2006-060110/ppc/bglsys/include/mpi.h:1107,
>                   from
> /ptmp/haewon/libmesh/include/base/libmesh_common.h:39,
>                   from /ptmp/haewon/libmesh/include/base/dof_map.h:33,
>                   from src/base/dof_map.C:27:
> /bgl/BlueLight/V1R2M1_020_2006-060110/ppc/bglsys/include/mpicxx.h:26:2:
> #error "SEEK_SET is #defined but must not be for the C++ binding of
> MPI"
> /bgl/BlueLight/V1R2M1_020_2006-060110/ppc/bglsys/include/mpicxx.h:30:2:
> #error "SEEK_CUR is #defined but must not be for the C++ binding of
> MPI"
> /bgl/BlueLight/V1R2M1_020_2006-060110/ppc/bglsys/include/mpicxx.h:35:2:
> #error "SEEK_END is #defined but must not be for the C++ binding of
> MPI"
> make: *** [src/base/dof_map.powerpc64-unknown-linux-gnu.opt.o] Error 1
> haewon/libmesh>
>
> I'll appreciate your advice.
>
> Sincerely,
>
> Hae-Won Choi
> --------------------------------------------------------------------
> Ph.D., Post Doctoral Research Associate
> Scientific Computing Division (SCD)
> National Center for Atmospheric Research (NCAR)
> 1850 Table Mesa Dr., Boulder, CO 80305, USA
> Phone: (303) 497-1271
> Email:   ha...@uc...
> --------------------------------------------------------------------
>
> On May 31, 2006, at 10:26 AM, John Peterson wrote:
> > We're happy to help...having libmesh running on BGL is pretty cool!
> > If you can post the compiler error messages, we may be able to tell
> > you more.  AIX compatibility has not been rigorously tested in a
> > few months.  If you haven't already, take a look at the libmesh
> > installation wiki
> >
> > http://libmesh.sourceforge.net/wiki/index.php/Installation
> >
> > -J
> >
> > Hae-Won Choi writes:
> >> Many thanks for your email.
> >> I had tried several computer accounts such as BlueGene/L, AIX and
> >> Alpha-HP
> >> that I have access in.
> >> Main problem was some conflicts with PetSc.
> >> I could not make properly install libmesh on BlueGene/L and AIX but
> >> I could only install this on Alpha-HP system.
> >> I first installed PetSc on BlueGene/L based on specific BleueGene/L
> >> configuration
> >> but after installed PetSc then I could not install libmesh since some
> >> compiling error was involved.
> >> That was main difficulty that I had.
> >> I could not figure it out.
> >> That's why I move on to Alpha-HP.
> >>
> >> Nevertheless, I really want to install libmesh on BGL since I will
> >> eventually access to
> >> 2000 of processors on BGL.
> >> I'll further appreciate give some specific tips for installing libmesh
> >> on BGL system.
> >> Thank you.
> >>
> >> Sincerely,
> >>
> >> Hae-Won Choi
> >> --------------------------------------------------------------------
> >> Ph.D., Post Doctoral Research Associate
> >> Scientific Computing Division (SCD)
> >> National Center for Atmospheric Research (NCAR)
> >> 1850 Table Mesa Dr., Boulder, CO 80305, USA
> >> Phone: (303) 497-1271
> >> Email:   ha...@uc...
> >> --------------------------------------------------------------------
> >>
> >> On May 31, 2006, at 8:43 AM, Kirk, Benjamin (JSC-EG) wrote:
> >>> Running in parallel is somewhat system dependent, but for an
> >>> MPICH-based
> >>> MPI install you can simply do
> >>>
> >>> $ mpirun -np NP ./ex4 -d 3 -n 20 -log_summary
> >>>
> >>> Where NP is the number of processors you want to run on.
> >>> -log_summary
> >>> is not necessary but can be useful.
> >>> Of course, if you need to submit to a queue to run parallel jobs the
> >>> procedure will be more involved and dependent on your particular
> >>> system.
> >>>
> >>>
> >>> So, I am sorry to hear the install was difficult.  What exactly were
> >>> the
> >>> complications?
> >>>
> >>> -Ben
> >>>
> >>>
> >>>
> >>> -----Original Message-----
> >>> From: lib...@li...
> >>> [mailto:lib...@li...] On Behalf Of
> >>> Hae-Won
> >>> Choi
> >>> Sent: Wednesday, May 31, 2006 9:04 AM
> >>> To: lib...@li...
> >>> Subject: [Libmesh-users] parallel run
> >>>
> >>> Hi, I had very hard time installing libmesh on my account.
> >>> It seems all serial run of examples are working all right.
> >>> But I'm interested running examples 4 and 9 in parallel.
> >>> Can you let me know how I can run above mentioned examples in
> >>> parallel?
> >>> What is the output looks like?
> >>> Thank  you.
> >>>
> >>> Sincerely,
> >>>
> >>> Hae-Won Choi
> >
> > -------------------------------------------------------
> > All the advantages of Linux Managed Hosting--Without the Cost and Risk!
> > Fully trained technicians. The highest number of Red Hat
> > certifications in
> > the hosting industry. Fanatical Support. Click to learn more
> > http://sel.as-us.falkag.net/sel?
> > cmd=3Dlnk&kid=3D107521&bid=3D248729&dat=3D121642
> > _______________________________________________
> > Libmesh-users mailing list
> > Lib...@li...
> > https://lists.sourceforge.net/lists/listinfo/libmesh-users
>
> -------------------------------------------------------
> All the advantages of Linux Managed Hosting--Without the Cost and Risk!
> Fully trained technicians. The highest number of Red Hat certifications in
> the hosting industry. Fanatical Support. Click to learn more
> http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D107521&bid=3D248729&dat=
=3D121642
> _______________________________________________
> Libmesh-users mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-users