[petsc-users] VecValidValues() reports NaN found

Mon May 5 08:12:05 CDT 2014

On Mon, May 5, 2014 at 7:56 AM, linjing bo <francium87 at hotmail.com> wrote:

> I use JACOBI. The message showed is with JACOBI.
>
>
> Wired situation is that the backtrack information shows the location is
> before actually apply PC, so I guess the rhs vec is not changed at this
> point.
>
> Another wired thing is : Because the original code is to complex. I write
> out the A matrix in Ax=b, and write a small test code to read in this
> matrix and solve it, no error showed. The KSP, PC are all set to be the
> same.
>
> When I try to using ILU, more wired error happens, the backtrack info
> shows it died in a Flops logging function:
>

1) Run in serial until it works

2) It looks like you have memory overwriting problems. Run with valgrind

   Matt

> [2]PETSC ERROR: --------------------- Error Message
> ------------------------------------
> [2]PETSC ERROR: Argument out of
> range!
> [2]PETSC ERROR: Cannot log negative
> flops!
> [2]PETSC ERROR:
> ------------------------------------------------------------------------
> [2]PETSC ERROR: Petsc Release Version 3.4.4, Mar, 13,
> 2014
> [2]PETSC ERROR: See docs/changes/index.html for recent
> updates.
> [2]PETSC ERROR: See docs/faq.html for hints about trouble
> shooting.
> [2]PETSC ERROR: See docs/index.html for manual
> pages.
> [2]PETSC ERROR:
> ------------------------------------------------------------------------
> [2]PETSC ERROR: ./gtc on a arch-linux2-c-debug named node2.indac.info by
> jlin Mon May  5 20:51:27
> 2014
>
> [2]PETSC ERROR: Libraries linked from
> /opt/sfw/petsc/3.4.4/intel/openmpi/lib
> [2]PETSC ERROR: Configure run at Sat Apr 26 20:19:41
> 2014
> [2]PETSC ERROR: Configure options
> --prefix=/opt/sfw/petsc/3.4.4/intel/openmpi
> --with-mpi-dir=/opt/sfw/openmpi/1.6.3/intel
> --with-blas-lapack-dir=/opt/sfw/intel/composer_xe_2011_sp1.7.256/mkl/lib/intel64
> --with-mpiexec=mpiexec
>
> [2]PETSC ERROR:
> ------------------------------------------------------------------------
>
> [2]PETSC ERROR: PetscLogFlops() line 204 in
> /tmp/petsc-3.4.4/include/petsclog.h
> [2]PETSC ERROR: MatLUFactorNumeric_SeqAIJ() line 552 in
> /tmp/petsc-3.4.4/src/mat/impls/aij/seq/aijfact.c
>
> [2]PETSC ERROR: MatLUFactorNumeric() line 2889 in
> /tmp/petsc-3.4.4/src/mat/interface/matrix.c
> [2]PETSC ERROR: PCSetUp_ILU() line 232 in
> /tmp/petsc-3.4.4/src/ksp/pc/impls/factor/ilu/ilu.c
> [2]PETSC ERROR: PCSetUp() line 890 in
> /tmp/petsc-3.4.4/src/ksp/pc/interface/precon.c
> [2]PETSC ERROR: KSPSetUp() line 278 in
> /tmp/petsc-3.4.4/src/ksp/ksp/interface/itfunc.c
> [2]PETSC ERROR: KSPSolve() line 399 in
> /tmp/petsc-3.4.4/src/ksp/ksp/interface/itfunc.c
>
>
>
>
> ------------------------------
> Date: Mon, 5 May 2014 07:27:52 -0500
> Subject: Re: [petsc-users] VecValidValues() reports NaN found
> From: knepley at gmail.com
> To: francium87 at hotmail.com
> CC: petsc-users at mcs.anl.gov
>
> On Mon, May 5, 2014 at 7:25 AM, linjing bo <francium87 at hotmail.com> wrote:
>
>  Hi, I'm trying to use PETSc's ksp method to solve a linear system. When
> running, Error is reported by VecValidValues() that NaN or Inf is found
> with error message listed below
>
>
> [3]PETSC ERROR: --------------------- Error Message
> ------------------------------------
> [3]PETSC ERROR: Floating point
> exception!
> [3]PETSC ERROR: Vec entry at local location 0 is not-a-number or infinite
> at beginning of function: Parameter number
> 2!
>
> [3]PETSC ERROR:
> ------------------------------------------------------------------------
>
> [3]PETSC ERROR: Petsc Release Version 3.4.4, Mar, 13,
> 2014
> [3]PETSC ERROR: See docs/changes/index.html for recent
> updates.
> [3]PETSC ERROR: See docs/faq.html for hints about trouble
> shooting.
> [3]PETSC ERROR: See docs/index.html for manual
> pages.
> [3]PETSC ERROR:
> ------------------------------------------------------------------------
>
> [3]PETSC ERROR: ./gtc on a arch-linux2-c-debug named node2.indac.info by
> jlin Mon May  5 20:03:20
> 2014
>
> [3]PETSC ERROR: Libraries linked from
> /opt/sfw/petsc/3.4.4/intel/openmpi/lib
> [3]PETSC ERROR: Configure run at Sat Apr 26 20:19:41
> 2014
> [3]PETSC ERROR: Configure options
> --prefix=/opt/sfw/petsc/3.4.4/intel/openmpi
> --with-mpi-dir=/opt/sfw/openmpi/1.6.3/intel
> --with-blas-lapack-dir=/opt/sfw/intel/composer_xe_2011_sp1.7.256/mkl/lib/intel64
> --with-mpiexec=mpiexec
>
> [3]PETSC ERROR:
> ------------------------------------------------------------------------
>
> [3]PETSC ERROR: VecValidValues() line 28 in
> /tmp/petsc-3.4.4/src/vec/vec/interface/rvector.c
>
>
> It looks like the vector after preconditioner application is bad. What is
> the preconditioner?
>
>   Matt
>
>
>
> [3]PETSC ERROR: PCApply() line 436 in
> /tmp/petsc-3.4.4/src/ksp/pc/interface/precon.c
> [3]PETSC ERROR: KSP_PCApply() line 227 in
> /tmp/petsc-3.4.4/include/petsc-private/kspimpl.h
> [3]PETSC ERROR: KSPInitialResidual() line 64 in
> /tmp/petsc-3.4.4/src/ksp/ksp/interface/itres.c
> [3]PETSC ERROR: KSPSolve_GMRES() line 239 in
> /tmp/petsc-3.4.4/src/ksp/ksp/impls/gmres/gmres.c
> [3]PETSC ERROR: KSPSolve() line 441 in
> /tmp/petsc-3.4.4/src/ksp/ksp/interface/itfunc.c
>
>
> After read the source code shown by backtrack informations, I realize the
> problem is in the right hand side vector. So I make a trial of set right
> hand side vector to ONE by VecSet, But the program still shows error
> message above, and using VecView or VecGetValue to investigate the first
> value of rhs vec shows the value is 1.0 as I set it to. Hope I clearly
> describe the problem. The code related is listed below
>
> ---------------------------Solver section--------------------------
>
>   call VecSet( pet_bp_b, one, ierr)
>
>   vecidx=[0,1]
>   call VecGetValues( pet_bp_b, 2, vecidx, first, ierr)
>   write(*,*) ' first two values ', first(1), first(2)
>
>   call KSPSetInitialGuessNonzero(solver_bp,Petsc_True,ierr)
>   call KSPSolve(solver_bp,pet_bp_b,pet_bp_x,ierr)
>   call KSPView(solver_bp, PETSC_VIEWER_STDOUT_SELF,ierr)
>   CHKERRQ(ierr)
>
>
>
>
>
> --
> What most experimenters take for granted before they begin their
> experiments is infinitely more interesting than any results to which their
> experiments lead.
> -- Norbert Wiener
>

-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20140505/0afe5b2d/attachment.html>