[petsc-users] pctype hmpi

Fri Aug 31 15:18:54 CDT 2012

On Aug 31, 2012, at 3:09 PM, George Pau <gpau at lbl.gov> wrote:

> Hi Barry,
> 
> For the hmpi_spawn_size,  the options in my .petscrc are
> -info
> -pc_view
> pc_type hmpi

  How come there is no - in front of this one?

> -ksp_type preonly
> -ksp_view
> -hmpi_pc_monitor
> -hmpi_ksp_monitor
> -hmpi_ksp_type cg
> -hmpi_pc_type hypre
> -hmpi_pc_hypre_type boomeramg
> -hmpi_spawn_size 3
> 
> mpiexec.mpich2 -n 1 myprogram
> 
> [0] petscinitialize_(): (Fortran):PETSc successfully started: procs 1
> [0] PetscGetHostName(): Rejecting domainname, likely is NIS gilbert.(none)
> [0] petscinitialize_(): Running on machine: gilbert 
> 
> [0] PetscCommDuplicate(): Duplicating a communicator 1140850688 -2080374784 max tags = 2147483647
> [0] MatSetUp(): Warning not preallocating matrix storage
> [0] MatAssemblyEnd_SeqAIJ(): Matrix size: 360 X 360; storage space: 3978 unneeded,3222 used
> [0] MatAssemblyEnd_SeqAIJ(): Number of mallocs during MatSetValues() is 360
> [0] MatAssemblyEnd_SeqAIJ(): Maximum nonzeros in any row is 9
> [0] Mat_CheckInode(): Found 120 nodes of 360. Limit used: 5. Using Inode routines
> [0] PetscCommDuplicate(): Using internal PETSc communicator 1140850688 -2080374784
> [0] PetscCommDuplicate(): Using internal PETSc communicator 1140850688 -2080374784
> 
> Fatal error in PMPI_Bcast: Invalid communicator, error stack:
> PMPI_Bcast(1478): MPI_Bcast(buf=0x7fff30dacecc, count=1, MPI_INT, root=0, comm=0x0) failed
> PMPI_Bcast(1418): Invalid communicator
> 
> I inserted some print statement between the ksp calls and found that the error occurs in 
> 
> call KSPSetFromOptions(ksp, pierr)
> 
> 2. If I change hmpi_spawn_size 3 to hmpi_merge_size 2 and launch my job by

   How come there is no - in front of hmpi_merge_size 2?

   Can you try putting all the arguments as command line arguments instead of in a file? It shouldn't matter but it seems like some of the arguments are being ignored.

   Barry

> 
> mpiexec.mpich2 -n 2 myprogram
> 
> [0] petscinitialize_(): (Fortran):PETSc successfully started: procs 2
> [0] PetscGetHostName(): Rejecting domainname, likely is NIS gilbert.(none)
> [0] petscinitialize_(): Running on machine: gilbert
> [1] petscinitialize_(): (Fortran):PETSc successfully started: procs 2
> [1] PetscGetHostName(): Rejecting domainname, likely is NIS gilbert.(none)
> [1] petscinitialize_(): Running on machine: gilbert
> 
> [0] PetscCommDuplicate(): Duplicating a communicator 1140850688 -2080374780 max tags = 2147483647
> [0] MatSetUp(): Warning not preallocating matrix storage
> [1] PetscCommDuplicate(): Duplicating a communicator 1140850688 -2080374782 max tags = 2147483647
> [0] PetscCommDuplicate(): Duplicating a communicator 1140850689 -2080374777 max tags = 2147483647
> [0] PetscCommDuplicate(): Using internal PETSc communicator 1140850689 -2080374777
> [1] PetscCommDuplicate(): Duplicating a communicator 1140850689 -2080374780 max tags = 2147483647
> [1] PetscCommDuplicate(): Using internal PETSc communicator 1140850689 -2080374780
> [0] MatStashScatterBegin_Private(): No of messages: 1 
> [0] MatStashScatterBegin_Private(): Mesg_to: 1: size: 12896 
> [0] MatAssemblyBegin_MPIAIJ(): Stash has 1611 entries, uses 0 mallocs.
> [1] MatAssemblyBegin_MPIAIJ(): Stash has 0 entries, uses 0 mallocs.
> [0] MatAssemblyEnd_SeqAIJ(): Matrix size: 180 X 180; storage space: 1998 unneeded,1602 used
> [0] MatAssemblyEnd_SeqAIJ(): Number of mallocs during MatSetValues() is 180
> [0] MatAssemblyEnd_SeqAIJ(): Maximum nonzeros in any row is 9
> [0] Mat_CheckInode(): Found 60 nodes of 180. Limit used: 5. Using Inode routines
> [1] MatAssemblyEnd_SeqAIJ(): Matrix size: 180 X 180; storage space: 1998 unneeded,1602 used
> [1] MatAssemblyEnd_SeqAIJ(): Number of mallocs during MatSetValues() is 180
> [1] MatAssemblyEnd_SeqAIJ(): Maximum nonzeros in any row is 9
> [1] Mat_CheckInode(): Found 60 nodes of 180. Limit used: 5. Using Inode routines
> [0] PetscCommDuplicate(): Using internal PETSc communicator 1140850689 -2080374777
> [1] PetscCommDuplicate(): Using internal PETSc communicator 1140850689 -2080374780
> [0] PetscCommDuplicate(): Using internal PETSc communicator 1140850689 -2080374777
> [1] PetscCommDuplicate(): Using internal PETSc communicator 1140850689 -2080374780
> [0] VecScatterCreateCommon_PtoS(): Using blocksize 1 scatter
> [0] VecScatterCreate(): General case: MPI to Seq
> [1] MatAssemblyEnd_SeqAIJ(): Matrix size: 180 X 3; storage space: 396 unneeded,9 used
> [1] MatAssemblyEnd_SeqAIJ(): Number of mallocs during MatSetValues() is 3
> [1] MatAssemblyEnd_SeqAIJ(): Maximum nonzeros in any row is 3
> [1] PetscCommDuplicate(): Using internal PETSc communicator 1140850688 -2080374782
> [0] MatAssemblyEnd_SeqAIJ(): Matrix size: 180 X 3; storage space: 396 unneeded,9 used
> [0] MatAssemblyEnd_SeqAIJ(): Number of mallocs during MatSetValues() is 3
> [0] MatAssemblyEnd_SeqAIJ(): Maximum nonzeros in any row is 3
> [0] PetscCommDuplicate(): Using internal PETSc communicator 1140850688 -2080374780
> [0] VecAssemblyBegin_MPI(): Stash has 180 entries, uses 1 mallocs.
> [0] VecAssemblyBegin_MPI(): Block-Stash has 0 entries, uses 0 mallocs.
> [1] PetscCommDuplicate(): Using internal PETSc communicator 1140850688 -2080374782
> [0] PetscCommDuplicate(): Using internal PETSc communicator 1140850688 -2080374780
> [0] PCSetUp(): Setting up new PC
> [0] PetscCommDuplicate(): Using internal PETSc communicator 1140850688 -2080374780
> 
> [0]PETSC ERROR: --------------------- Error Message ------------------------------------
> [0]PETSC ERROR: Nonconforming object sizes!
> [0]PETSC ERROR: HMPI preconditioner only works for sequential solves!
> [0]PETSC ERROR: ------------------------------------------------------------------------
> [0]PETSC ERROR: Petsc Release Version 3.3.0, Patch 3, Wed Aug 29 11:26:24 CDT 2012 
> [0]PETSC ERROR: See docs/changes/index.html for recent updates.
> [0]PETSC ERROR: See docs/faq.html for hints about trouble shooting.
> [0]PETSC ERROR: See docs/index.html for manual pages.
> [0]PETSC ERROR: ------------------------------------------------------------------------
> [0]PETSC ERROR: ../../esd-tough2/xt2_eos4 on a arch-linu named gilbert by gpau Fri Aug 31 13:00:31 2012
> [0]PETSC ERROR: Libraries linked from /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/lib
> [0]PETSC ERROR: Configure run at Thu Aug 30 15:27:17 2012
> [0]PETSC ERROR: Configure options --with-debugging=0 --with-mpi-dir=/usr/lib/mpich2 --download-hypre=1 --prefix=/home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib
> [0]PETSC ERROR: ------------------------------------------------------------------------
> [0]PETSC ERROR: PCCreate_HMPI() line 283 in /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/tpls/petsc/petsc-3.3-p3-source/src/ksp/pc/impls/openmp/hpc.c
> [0]PETSC ERROR: PCSetType() line 83 in /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/tpls/petsc/petsc-3.3-p3-source/src/ksp/pc/interface/pcset.c
> [0]PETSC ERROR: PCSetFromOptions() line 188 in /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/tpls/petsc/petsc-3.3-p3-source/src/ksp/pc/interface/pcset.c
> [0]PETSC ERROR: KSPSetFromOptions() line 287 in /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/tpls/petsc/petsc-3.3-p3-source/src/ksp/ksp/interface/itcl.c
> [0]PETSC ERROR: --------------------- Error Message ------------------------------------
> [0]PETSC ERROR: No support for this operation for this object type!
> [0]PETSC ERROR: PC does not have apply!
> [0]PETSC ERROR: ------------------------------------------------------------------------
> [0]PETSC ERROR: Petsc Release Version 3.3.0, Patch 3, Wed Aug 29 11:26:24 CDT 2012 
> [0]PETSC ERROR: See docs/changes/index.html for recent updates.
> [0]PETSC ERROR: See docs/faq.html for hints about trouble shooting.
> [0]PETSC ERROR: See docs/index.html for manual pages.
> [0]PETSC ERROR: ------------------------------------------------------------------------
> [0]PETSC ERROR: ../../esd-tough2/xt2_eos4 on a arch-linu named gilbert by gpau Fri Aug 31 13:00:31 2012
> [0]PETSC ERROR: Libraries linked from /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/lib
> [0]PETSC ERROR: Configure run at Thu Aug 30 15:27:17 2012
> [0]PETSC ERROR: Configure options --with-debugging=0 --with-mpi-dir=/usr/lib/mpich2 --download-hypre=1 --prefix=/home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib
> [0]PETSC ERROR: ------------------------------------------------------------------------
> [0]PETSC ERROR: PCApply() line 382 in /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/tpls/petsc/petsc-3.3-p3-source/src/ksp/pc/interface/precon.c
> [0]PETSC ERROR: KSPInitialResidual() line 64 in /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/tpls/petsc/petsc-3.3-p3-source/src/ksp/ksp/interface/itres.c
> [0]PETSC ERROR: KSPSolve_GMRES() line 230 in /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/tpls/petsc/petsc-3.3-p3-source/src/ksp/ksp/impls/gmres/gmres.c
> [0]PETSC ERROR: KSPSolve() line 446 in /home/gpau/tough_codes/esd-tough2/build/Linux-x86_64-Debug-MPI-EOS4/toughlib/tpls/petsc/petsc-3.3-p3-source/src/ksp/ksp/interface/itfunc.c
> 
> I note that the error appears to occur at the same point.
> 
> George
> 
> 
> On Fri, Aug 31, 2012 at 11:31 AM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> 
> On Aug 31, 2012, at 1:27 PM, George Pau <gpau at lbl.gov> wrote:
> 
> > Hi Barry,
> >
> > 1. It is the exact same error related to MPI_ERR_COMM and MPI_Bcast.
> 
>    That should not happen. Please run and send all the output including the exact command line used
> 
> 
> >  I am currently using the MPICH2 distribution provided by ubuntu but if MPICH version that Petsc download with -download-mpich works, I can use that.
> > 2. If I use hmpi_merge_size, I will need to launch mpiexec with more than 1 cpus.  But, petsc will complain that the pctype hmpi can only be used in serial.
> 
>    That should not happen. Run with 2 MPI processes and -hmpi_merge_size 2  and send the complete error message.
> 
> 
>   Barry
> 
> >
> > George
> >
> >
> > On Aug 31, 2012, at 11:17 AM, Barry Smith wrote:
> >
> >>
> >> On Aug 30, 2012, at 10:02 PM, George Pau <gpau at lbl.gov> wrote:
> >>
> >>> Hi Barry,
> >>>
> >>> I tried with the addition of
> >>>
> >>> -hmpi_spawn_size 3
> >>>
> >>> but I am still getting the same error though.
> >>
> >>   The EXACT same error? Or some other error?
> >>
> >>    What happens if you run with the -hmpi_merge_size <size> option instead?
> >>
> >>  Barry
> >>
> >> 1) I am getting a crash with the spawn version I suspect is due to bugs in the MPICH version I am using related to spawn.
> >>
> >> 2) I am getting errors with the merge version due to Apple's ASLR which they make hard to turn off.
> >>
> >>
> >>> I am using mpich2.  Any other options to try?
> >>>
> >>> George
> >>>
> >>>
> >>> On Aug 30, 2012, at 7:28 PM, Barry Smith wrote:
> >>>
> >>>>
> >>>> On Aug 30, 2012, at 7:24 PM, George Pau <gpau at lbl.gov> wrote:
> >>>>
> >>>>> Hi,
> >>>>>
> >>>>> I have some issues using the -pctype hmpi.  I used the same setting found at
> >>>>>
> >>>>> http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/PC/PCHMPI.html
> >>>>>
> >>>>> i.e.
> >>>>> -pc_type hmpi
> >>>>> -ksp_type preonly
> >>>>> -hmpi_ksp_type cg
> >>>>> -hmpi_pc_type hypre
> >>>>> -hmpi_pc_hypre_type boomeramg
> >>>>>
> >>>>> My command  is
> >>>>>
> >>>>> mpiexec -n 1 myprogram
> >>>>
> >>>> Sorry the documentation doesn't make this clearer. You need to start PETSc with special options to get the "worker" processes initialized. From the manual page for PCHMPI it has
> >>>>
> >>>> See PetscHMPIMerge() and PetscHMPISpawn() for two ways to start up MPI for use with this preconditioner
> >>>>
> >>>> This will tell you want option to start PETSc up with.
> >>>>
> >>>> I will fix the PC so that it prints a far more useful error message.
> >>>>
> >>>>
> >>>>
> >>>> Barry
> >>>>
> >>>>
> >>>>>
> >>>>> But, I get
> >>>>>
> >>>>> [gilbert:4041] *** An error occurred in MPI_Bcast
> >>>>> [gilbert:4041] *** on communicator MPI_COMM_WORLD
> >>>>> [gilbert:4041] *** MPI_ERR_COMM: invalid communicator
> >>>>> [gilbert:4041] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
> >>>>>
> >>>>> with openmpi.  I get similar error with mpich2
> >>>>>
> >>>>> Fatal error in PMPI_Bcast: Invalid communicator, error stack:
> >>>>> PMPI_Bcast(1478): MPI_Bcast(buf=0x7fffb683479c, count=1, MPI_INT, root=0, comm=0x0) failed
> >>>>> PMPI_Bcast(1418): Invalid communicator
> >>>>>
> >>>>> I couldn't figure out what is wrong.    My petsc is  version 3.3.3 and the configuration is -with-debugging=0 --with-mpi-dir=/usr/lib/openmpi --download-hypre=1 and I am on a Ubuntu machine.
> >>>>>
> >>>>> Note that with the default pc_type and ksp_type, everything is fine.  It was also tested with multiple processors.  I wondering whether there are some options that I am not specifying correctly?
> >>>>>
> >>>>> --
> >>>>> George Pau
> >>>>> Earth Sciences Division
> >>>>> Lawrence Berkeley National Laboratory
> >>>>> One Cyclotron, MS 74-120
> >>>>> Berkeley, CA 94720
> >>>>>
> >>>>> (510) 486-7196
> >>>>> gpau at lbl.gov
> >>>>> http://esd.lbl.gov/about/staff/georgepau/
> >>>>>
> >>>>
> >>>
> >>
> >
> 
> 
> 
> 
> -- 
> George Pau
> Earth Sciences Division
> Lawrence Berkeley National Laboratory
> One Cyclotron, MS 74-120
> Berkeley, CA 94720
> 
> (510) 486-7196
> gpau at lbl.gov
> http://esd.lbl.gov/about/staff/georgepau/
> 