[petsc-dev] (PETSc-3.1-p6 & SUPERLU_dist) + ("-ksp_view" CLI option & KSPSetOptionsPrefix()/PCSetOptionsPrefix() ) = bug?

[Filippo Spiga]

Dear Satish,
   thanks for your fast reply! I am compiling PETSc with the patch you
provided.

I also have already found a workaround: I've skip the automatic
download of SuperLU_dist and
I've specified manually the version 2.3 instead of the 2.4 (I had a backup
of that package). Well, using that version of SuperLU_dist the problem does
not appear.

I will keep you updated about the efficacy of the patch.

Cheers

--
Filippo SPIGA, MSc Computer Science
~ homepage: http://tinyurl.com/fspiga ~

«Nobody will drive us out of Cantor's paradise.»
    -- David Hilbert

*****
Disclaimer: "Please note this message and any attachment are CONFIDENTIAL an
may be privileged or otherwise protected from disclosure. The contents are
not to be disclosed to anyone other than the addressee. Unauthorized
recipients are requested to preserve this confidentiality and to advise the
sender immediately of any error in transmission."



On Wed, Dec 15, 2010 at 2:46 PM, Satish Balay <balay at mcs.anl.gov> wrote:

> sorry missed the attachment.. its attached now.
>
> satish
>
> On Wed, 15 Dec 2010, Satish Balay wrote:
>
> > Can you try the attached patch - and see if this goes away?
> >
> > patch -Np1 < superlu_dist.patch
> >
> > Satish
> >
> > On Wed, 15 Dec 2010, Filippo Spiga wrote:
> >
> > > Dear all,
> > >     Sorry for the subject of this email, I had no better ideas to
> introduce
> > > the topic (-: I spent a couple of hours yesterday night trying to
> figure out
> > > about a strange behavior of PETSc-3.1-p6 using SUPERLU_dist. This
> problem
> > > arises if I combine the usage of
> KSPSetOptionsPrefix()/PCSetOptionsPrefix()
> > > and "-ksp_view" option. This is the error that I obtain:
> > >
> > > KSP Object:(lop_)
> > >   type: gmres
> > >     GMRES: restart=30, using Classical (unmodified) Gram-Schmidt
> > > Orthogonalization with no iterative refinement
> > >     GMRES: happy breakdown tolerance 1e-30
> > >   maximum iterations=1000, initial guess is zero
> > >   tolerances:  relative=1e-14, absolute=1e-14, divergence=10000
> > >   left preconditioning
> > >   using PRECONDITIONED norm type for convergence test
> > > PC Object:(lop_)
> > >   type: lu
> > >     LU: out-of-place factorization
> > >     tolerance for zero pivot 1e-12
> > >     matrix ordering: nd
> > >     factor fill ratio given 0, needed 0
> > >       Factored matrix follows:
> > >         Matrix Object:
> > >           type=seqaij, rows=242, cols=242
> > >           package used to perform factorization: superlu_dist
> > >           total: nonzeros=0, allocated nonzeros=242
> > >             SuperLU_DIST run parameters:
> > >               Process grid nprow 1 x npcol 1
> > >               Equilibrate matrix TRUE
> > >               Matrix input mode 0
> > >               Replace tiny pivots TRUE
> > >               Use iterative refinement FALSE
> > >               Processors in row 1 col partition 1
> > >               Row permutation LargeDiag
> > > [0|09:06:54]: unknown: MatFactorInfo_SuperLU_DIST() line 705 in
> > > src/mat/impls/aij/mpi/superlu_dist/superlu_dist.c: Unknown column
> > > permutation
> > >
> > >
> > > Inside the code, I have these lines:
> > >
> > > ierr = KSPCreate(PETSC_COMM_WORLD,&ksp); CHKERRQ(ierr);
> > > ierr = KSPSetOptionsPrefix(ksp,"lop_"); CHKERRQ(ierr);
> > > ierr = KSPGetPC(ksp,&precond); CHKERRQ(ierr);
> > > ierr = PCSetOptionsPrefix(precond,"lop_"); CHKERRQ(ierr);
> > > ierr = KSPSetFromOptions(ksp); CHKERRQ(ierr);
> > > ierr = KSPSolve(ksp,b,*u); CHKERRQ(ierr);
> > >
> > > so all the options that I pass to PETSc (using "PetscOptionsSetValue()"
> or
> > > the command line) must have the "lop_" prefix. These are my options:
> > >
> > > -lop_ksp_rtol 1.e-10 -lop_ksp_atol  1.e-10 -lop_ksp_monitor
> > > -lop_ksp_converged_reason -lop_ksp_type gmres
> -lop_ksp_monitor_true_residual
> > > -lop_pc_type lu -lop_pc_factor_mat_solver_package superlu_dist
> -lop_ksp_view
> > >
> > >
> > > If I remove "-lop_ksp_view", the error that I reported disappears (and
> also
> > > disappear the final summary
> > > If I remove -lop_pc_type lu -lop_pc_factor_mat_solver_package
> superlu_dist"
> > > and I use HYPRE ("-lop_pc_type hypre") and I keep the option
> "-lop_ksp_view"
> > > as valid option, the problem does not appear.
> > > If I remove "-lop_ksp_view" from command line but after "KSPSolve()" I
> add
> > > "KSPView()", the problem persists.
> > > If I suppress the prefix ("-ksp_rtol 1.e-10 -ksp_atol ... -ksp_view"),
> the
> > > problem does not appear.
> > >
> > > PETSc-3.1-p3, PETSc-3.1-p4 and PETSc-3.1-p5 are not affected of this
> problem
> > > (other machines that I use have these versions installed, the code runs
> > > without problems). In the meanwhile I am recompiling PETSc-3.1-p6 on
> another
> > > machine.
> > >
> > > Do you have any suggestion to skip/solve this problem? (if it is a real
> > > problem, maybe I did something wrong...)
> > >
> > > Many thanks in advance,
> > > Cheers
> > >
> > > --
> > > Filippo SPIGA, MSc Computer Science
> > > ~ homepage: http://tinyurl.com/fspiga ~
> > >
> > > «Nobody will drive us out of Cantor's paradise.»
> > >     -- David Hilbert
> > >
> > > *****
> > > Disclaimer: "Please note this message and any attachment are
> CONFIDENTIAL an
> > > may be privileged or otherwise protected from disclosure. The contents
> are
> > > not to be disclosed to anyone other than the addressee. Unauthorized
> > > recipients are requested to preserve this confidentiality and to advise
> the
> > > sender immediately of any error in transmission."
> > >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20101215/2074206c/attachment.html>