[petsc-dev] Fwd: [petsc-users] questions about the SNES Function Norm

[Peter Brune]

Yes, you've got to be careful about this in the case of preconditioning.
 I'm still pretty sure that SNESGetFunctionNorm() (which is ancient
http://www.mcs.anl.gov/petsc/petsc-2.2.0/docs/manualpages/SNES/SNESGetFunctionNorm.html)
can be replaced transparently with VecNorm();  SNESSetFunctionNorm()
is
used in one place and can be eliminated, unless we're going to put wonky
logic in there about when things are set.


On Thu, May 1, 2014 at 2:02 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:

>
> On May 1, 2014, at 1:32 PM, Peter Brune <prbrune at gmail.com> wrote:
>
> > Yeah that bit of interface can most likely go away entirely.
>  SNESSet/GetFunctionNorm is barely used in the code and can be mostly
> replaced with calls to VecNorm().  It's odd that snes->norm isn't the same
> as the norm of the function at the end of an iteration, but is of course
> totally possible given the multitude of potential exit conditions.  Actual
> output would, you know, be helpful in diagnosing what exactly's going on
> instead of forcing us to stab blindly into the dark.
>
>    Hmm, what about the weird case with nonlinear preconditioners where the
> “function norm” may not be the function norm. Like in
>
>   /* check if the function is valid based upon how the inner solver is
> preconditioned */
>     if (normschedule != SNES_NORM_NONE && normschedule !=
> SNES_NORM_INITIAL_ONLY && (npcside == PC_RIGHT || functype ==
> SNES_FUNCTION_UNPRECONDITIONED)) {
>       ierr = SNESGetFunction(snes->pc,&FPC,NULL,NULL);CHKERRQ(ierr);
>       if (FPC) {
>         if (fnorm) {ierr =
> SNESGetFunctionNorm(snes->pc,fnorm);CHKERRQ(ierr);}
>         ierr = VecCopy(FPC,F);CHKERRQ(ierr);
>       } else {
>         SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Preconditioner
> has no function");
>       }
>     } else {
>       ierr = SNESGetSolution(snes->pc,&XPC);CHKERRQ(ierr);
>       if (XPC) {
>         ierr = SNESComputeFunction(snes->pc,XPC,F);CHKERRQ(ierr);
>         if (fnorm) {ierr = VecNorm(F,NORM_2,fnorm);CHKERRQ(ierr);}
>
> or
>
>     ierr = PetscInfo4(snes,"fnorm=%18.16e, gnorm=%18.16e, ynorm=%18.16e,
> lssucceed=%d\n",(double)fnorm,(double)gnorm,(double)ynorm,(int)lssucceed);CHKERRQ(ierr);
>
>     if (snes->pc && snes->pcside == PC_RIGHT) {
>       ierr =
> PetscLogEventBegin(SNES_NPCSolve,snes->pc,X,0,0);CHKERRQ(ierr);
>       ierr = SNESSolve(snes->pc,snes->vec_rhs,X);CHKERRQ(ierr);
>       ierr = PetscLogEventEnd(SNES_NPCSolve,snes->pc,X,0,0);CHKERRQ(ierr);
>       ierr = SNESGetConvergedReason(snes->pc,&reason);CHKERRQ(ierr);
>       if (reason < 0 && reason != SNES_DIVERGED_MAX_IT) {
>         snes->reason = SNES_DIVERGED_INNER;
>         PetscFunctionReturn(0);
>       }
>       ierr = SNESGetNPCFunction(snes,F,&fnorm);CHKERRQ(ierr);
>     }
>
>     ierr = SNESSetIterationNumber(snes, i+1);CHKERRQ(ierr);
>     ierr = SNESSetFunctionNorm(snes, fnorm);CHKERRQ(ierr);
>
>   and a few other cases?
>
>   Barry
>
> >
> > - Peter
> >
> >
> > On Thu, May 1, 2014 at 1:00 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> >
> >   Peter,
> >
> >      This SNESSetFunctionNorm() seems pretty likely to be fragile where
> it doesn’t always get set. Is there any way we can test that it always gets
> updated properly when the function gets re-evaluated?   For example what if
> one place in the code the function is evaluated but the norm does not get
> computed and set with SNESSetFunctionNorm()?
> >
> >    Thanks
> >
> >     Barry
> >
> >
> > Begin forwarded message:
> >
> >> From: Xiangdong <epscodes at gmail.com>
> >> Subject: Re: [petsc-users] questions about the SNES Function Norm
> >> Date: May 1, 2014 at 10:32:44 AM CDT
> >> To: Matthew Knepley <knepley at gmail.com>
> >> Cc: Barry Smith <bsmith at mcs.anl.gov>, PETSc users list <
> petsc-users at mcs.anl.gov>
> >>
> >> Under what condition, SNESGetFunctionNorm() will output different
> results from SENEGetFunction + VecNorm (with NORM_2)?
> >>
> >> For most of my test cases, it is the same. However, when I have some
> special (trivial) initial guess to the SNES problem, I see different norms.
> >>
> >> Another phenomenon I noticed with this is that KSP in SNES squeeze my
> matrix by eliminating rows. I have a Jacobian supposed to be 50-by-50. When
> I use KSPGetOperators/rhs/solutions, I found that the operator is 25-by-25,
> and the rhs and solution is with length 25. Do you have any clue on what
> triggered this? To my surprise, when I output the Jacobian inside the
> FormJacobianLocal, it outputs the correct matrix 50-by-50 with correct
> numerical entries. Why does the operator obtained from KSP is different and
> got rows eliminated? These rows got eliminated have only one entries per
> row, but the rhs in that row is not zero. Eliminating these rows would give
> wrong solutions.
> >>
> >> Thank you.
> >>
> >> Xiangdong
> >>
> >>
> >>
> >>
> >>
> >>
> >> On Tue, Apr 29, 2014 at 3:12 PM, Matthew Knepley <knepley at gmail.com>
> wrote:
> >> On Tue, Apr 29, 2014 at 2:09 PM, Xiangdong <epscodes at gmail.com> wrote:
> >> It turns out to a be a bug  in my FormFunctionLocal(DMDALocalInfo
> *info,PetscScalar **x,PetscScalar **f,AppCtx *user). I forgot to initialize
> the array f. Zero the array f solved the problem and gave consistent result.
> >>
> >> Just curious, why does not petsc initialize the array f to zero by
> default inside petsc when passing the f array to FormFunctionLocal?
> >>
> >> If you directly set entires, you might not want us to spend the time
> writing those zeros.
> >>
> >> I have another quick question about the array x passed to
> FormFunctionLocal. If I want to know the which x is evaluated, how can I
> output x in a vector format? Currently, I created a global vector vecx and
> a local vector vecx_local, get the array of vecx_local_array, copy the x to
> vecx_local_array,  scatter to global vecx and output vecx. Is there a quick
> way to restore the array x to a vector and output?
> >>
> >> I cannot think of a better way than that.
> >>
> >>    Matt
> >>
> >> Thank you.
> >>
> >> Best,
> >> Xiangdong
> >>
> >>
> >>
> >> On Mon, Apr 28, 2014 at 10:28 PM, Barry Smith <bsmith at mcs.anl.gov>
> wrote:
> >>
> >> On Apr 28, 2014, at 3:23 PM, Xiangdong <epscodes at gmail.com> wrote:
> >>
> >> > Hello everyone,
> >> >
> >> > When I run snes program,
> >>
> >>                ^^^^ what SNES program”?
> >>
> >> > it outputs "SNES Function norm 1.23456789e+10". It seems that this
> norm is different from residue norm (even if solving F(x)=0)
> >>
> >>    Please send the full output where you see this.
> >>
> >> > and also differ from norm of the Jacobian. What is the definition of
> this "SNES Function Norm”?
> >>
> >>    The SNES Function Norm as printed by PETSc is suppose to the 2-norm
> of F(x) - b (where b is usually zero) and this is also the same thing as
> the “residue norm”
> >>
> >>    Barry
> >>
> >> >
> >> > Thank you.
> >> >
> >> > Best,
> >> > Xiangdong
> >>
> >>
> >>
> >>
> >>
> >> --
> >> 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-dev/attachments/20140501/b9c2bdc4/attachment.html>