[petsc-users] Error with MatSchurComplement

Barry Smith bsmith at mcs.anl.gov
Tue Feb 16 23:57:21 CST 2016

  Hmm, it seems this should never happen. It is like inconsistent (and in correct) IS are being generated. 

   1 is an odd number of processes, does it happen then?

   Are you providing the IS yourself to the PCFIELDSPLIT or using the default? If you are providing them is it possible that you have a bug in generating them for certain configurations? Of course it could also be a bug in PETSc. 

   I recommend trying to produce the crash on a small a problem as possible then run in the debugger to check the IS at each stage to see why one of them becomes bonkers.

   It could be we should have more asserts in PETSc to insure that input IS are compatible.


> On Feb 16, 2016, at 10:37 PM, Bhalla, Amneet Pal S <amneetb at live.unc.edu> wrote:
> Hi Folks,
> I am getting error with odd number of processors with PCFIELDSPLIT. The error log is attached. Any hints where to start digging?
> +++++++++++++++++++++++++++++++++++++++++++++++++++
> At beginning of timestep # 0
> Simulation time is 0
>     Residual norms for ib_ solve.
>     0 KSP unpreconditioned resid norm 5.093934451651e+05 true resid norm 5.093934451651e+05 ||r(i)||/||b|| 1.000000000000e+00
> [2]PETSC ERROR: --------------------- Error Message --------------------------------------------------------------
> [2]PETSC ERROR: Nonconforming object sizes
> [2]PETSC ERROR: Local columns of A10 1072 do not equal local rows of A00 1104
> [2]PETSC ERROR: See http://www.mcs.anl.gov/petsc/documentation/faq.html for trouble shooting.
> [2]PETSC ERROR: Petsc Release Version 3.6.1, unknown 
> [2]PETSC ERROR: ./main2d on a darwin-dbg named Amneets-MBP.attlocal.net by Taylor Tue Feb 16 20:30:45 2016
> [2]PETSC ERROR: Configure options --CC=mpicc --CXX=mpicxx --FC=mpif90 --PETSC_ARCH=darwin-dbg --with-debugging=1 --with-c++-support=1 --with-hypre=1 --download-hypre=1 --with-hdf5=yes --with-hdf5-dir=/Users/Taylor/Documents/SOFTWARES/HDF5/
> [2]PETSC ERROR: #1 MatSchurComplementSetSubMatrices() line 261 in /Users/Taylor/Documents/SOFTWARES/PETSc-BitBucket/PETSc/src/ksp/ksp/utils/schurm.c
> [2]PETSC ERROR: #2 PCSetUp_FieldSplit() line 714 in /Users/Taylor/Documents/SOFTWARES/PETSc-BitBucket/PETSc/src/ksp/pc/impls/fieldsplit/fieldsplit.c
> [2]PETSC ERROR: #3 PCSetUp() line 983 in /Users/Taylor/Documents/SOFTWARES/PETSc-BitBucket/PETSc/src/ksp/pc/interface/precon.c
> [2]PETSC ERROR: #4 KSPSetUp() line 332 in /Users/Taylor/Documents/SOFTWARES/PETSc-BitBucket/PETSc/src/ksp/ksp/interface/itfunc.c
> [2]PETSC ERROR: #5 KSPSolve() line 546 in /Users/Taylor/Documents/SOFTWARES/PETSc-BitBucket/PETSc/src/ksp/ksp/interface/itfunc.c
> [2]PETSC ERROR: #6 solveSystem() line 229 in ../../../IBAMR/ibtk/lib/../src/solvers/impls/PETScLevelSolver.cpp
> P=00002:Program abort called in file ``../../../IBAMR/ibtk/lib/../src/solvers/impls/PETScLevelSolver.cpp'' at line 229
> P=00002:
> application called MPI_Abort(MPI_COMM_WORLD, -1) - process 2
> *** Warning, AutoPtr is deprecated and will be removed in a future library version! ./include/libmesh/auto_ptr.h, line 271, compiled Feb  3 2016 at 10:29:17 ***
> — Amneet 

More information about the petsc-users mailing list