On May 31, 2023, at 4:16 PM, Kenneth C Hall <kenneth.c.hall@duke.edu> wrote:

Matt,

 

Thanks for your quick reply.  I think what you say makes sense.

 

You asked what my code does. The MySolver program performs one iteration of a CFD iteration. The CFD scheme is an explicit scheme that uses multigrid, Mach number preconditioning, and residual smoothing.  Typically, I have to call MySolver on the order of 40 to 100 times to get acceptable convergence. 

 

And in fact, I have another version of this Petsc code that uses SNESNGMRES to solve the problem with MySolver providing the residuals as R =  N(x) - x. 

But I would like a version where I am using just MySolver, without any other operations applied to it.  So I am trying to plug MySolver into the PETSc system to provide monitoring and other features, and for consistency and comparison to these other (more appropriate!) uses of PETSc.

 

Thanks.

Kenneth

 

 

From: Matthew Knepley <knepley@gmail.com>
Date: Wednesday, May 31, 2023 at 3:48 PM
To: Kenneth C Hall <kenneth.c.hall@duke.edu>
Cc: petsc-users@mcs.anl.gov <petsc-users@mcs.anl.gov>
Subject: Re: [petsc-users] Using SNESSHELL as a wrapper for a CFD solver.

On Wed, May 31, 2023 at 3:21 PM Kenneth C Hall <kenneth.c.hall@duke.edu> wrote:

Hi,

 

I am doing a number of problems using PETSc/SLEPc, but I also work on some non-PETSc/SLEPc flow solvers.   I would like to use PETSc as a wrapper for this non-PETSc flow solver for compatibility,  so I can use the tolerance monitoring, options, viewers, and for direct comparison to PETSc methods I am using.

 

Here is what I am trying to do… I have a CFD solver that iterates with a nonlinear iterator of the form x := N(x).  This can be expressed in a fortran routine of the form,

 

SUBROUTINE MySolver(x) 

or 

SUBROUTINE MySolver(x,y)

 

In the first case, x is over written. In the second, y = N(x).  In any event, I want to do something like what is shown in the subroutine at the bottom of this email. 

 

The code below “works” in the sense that MySolver is called, but it is called exactly *once*. But MyMonitor and MyConverged are *not* called. Again, I want to iterate so MySolver should be called many times, as should MyMonitor and MyConverged.

 

The SNESolve() method is called once per nonlinear solve, just as KSPSolve() is called once per linear solve. There may be iteration inside the method, but that is handled inside the particular implementation. For example, both Newton's method and Nonlinear Conjugate Gradient iterate, but the iteration is internal to both, and they both call the monitor and convergence test at each internal iterate.

 

So, if your nonlinear solver should iterate, it should happen inside the SNESSolve call for the SNESSHELL object. Does this make sense? What does your solver do?

 

  Thanks,

 

     Matt

 

The SNESView before and after SNESSolve looks like this:

 

SNES Object: 1 MPI process

  type: shell

  SNES has not been set up so information may be incomplete

  maximum iterations=50, maximum function evaluations=10000

  tolerances: relative=1e-50, absolute=1e-10, solution=1e+06

  total number of function evaluations=0

  norm schedule ALWAYS

 

SNES Object: 1 MPI process

  type: shell

  maximum iterations=50, maximum function evaluations=10000

  tolerances: relative=1e-50, absolute=1e-10, solution=1e+06

  total number of function evaluations=0

  norm schedule ALWAYS

 

Any suggestions on how to do what I am trying to accomplish?

 

Thanks.

Kenneth Hall

 

 

 

#include <petsc/finclude/petsc.h>

#include "macros.h"

 

   MODULE SolveWithSNESShell_module

      USE MyPetscModule

   CONTAINS

!

!====================================================================================================

   SUBROUTINE MySolver(snes, x, ierr)

!====================================================================================================

!!

!!

!====================================================================================================

!

      USE MyPetscModule

      IMPLICIT NONE

!

!.... declared passed variables

      SNES     :: snes

      Vec      :: x

      PetscErrorCode :: ierr

!

!.... code to find residual  x := N(x)

!.... (or alternatively      y := N(x))

           

   END SUBROUTINE MySolver

!

!====================================================================================================

   SUBROUTINE MyMonitor(snes, its, rnorm, ierr)

!====================================================================================================

!!

!!

!====================================================================================================

!

      USE MyPetscModule

      IMPLICIT NONE

!

!.... Declare passed variables

      SNES           :: snes

      PetscInt       :: its

      PetscReal      :: rnorm

      PetscErrorCode :: ierr

!

!.... Code to print out convergence history

!.... Code to print out convergence history

 

   END SUBROUTINE MyMonitor

 

!====================================================================================================

   SUBROUTINE MyConverged(snes, it, xnorm, ynorm, znorm, reason, ierr)

      USE MyPetscModule

      IMPLICIT NONE

 

      SNES     :: snes

      PetscInt :: it,ctx

      PetscReal :: xnorm, ynorm, znorm

      KSPConvergedReason :: reason

      PetscErrorCode :: ierr

 

!      ... add convergence test here ...

!      set reason to a positive value if convergence has been achieved

 

 

   END SUBROUTINE MyConverged

   END MODULE SolveWithSNESShell_module

!

!====================================================================================================

   SUBROUTINE SolveWithSNESShell

!====================================================================================================

!!

!!

!====================================================================================================

!

      USE SolveWithSNESShell_module

      IMPLICIT NONE

!

!.... Declare passed variables

      INTEGER            :: level_tmp

!

!.... Declare local variables

      INTEGER            :: iz

      INTEGER            :: imax

      INTEGER            :: jmax

      INTEGER            :: kmax

      SNES               :: snes

      KSP                :: ksp

      Vec                :: x

      Vec                :: y

      PetscViewer        :: viewer

      PetscErrorCode     :: ierr

      PetscReal          :: rtol   = 1.0D-10   !! relative tolerance

      PetscReal          :: atol   = 1.0D-50   !! absolute tolerance

      PetscReal          :: dtol   = 1.0D+06   !! divergence tolerance

      PetscInt           :: maxits = 50

      PetscInt           :: maxf   = 10000

      character(len=1000):: args

!

!.... count the number of degrees of freedom.

      level = level_tmp

      n = 0

      DO iz = 1, hb(level)%nzone

         imax = hb(level)%zone(iz)%imax - 1

         jmax = hb(level)%zone(iz)%jmax - 1

         kmax = hb(level)%zone(iz)%kmax - 1

         n = n + imax * jmax * kmax

      END DO

      n = n * neqn

!

!.... Initialize PETSc

      PetscCall(PetscInitialize(PETSC_NULL_CHARACTER, ierr))

!

!.... Log

      PetscCall(PetscLogDefaultBegin(ierr))

!

!.... Hard-wired options.

!     PetscCall(PetscOptionsInsertString(PETSC_NULL_OPTIONS, "command line style option here" , ierr))

!

!.... Command line options.

      call GET_COMMAND(args)

      PetscCall(PetscOptionsInsertString(PETSC_NULL_OPTIONS, args, ierr))

!

!.... view command line table

      PetscCall(PetscViewerASCIIOpen(PETSC_COMM_SELF, PETSC_VIEWER_STDOUT_SELF, viewer, ierr))

      PetscCall(PetscOptionsView(PETSC_NULL_OPTIONS, viewer, ierr))

      PetscCall(PetscViewerDestroy(viewer, ierr))

!

!.... Create PETSc vectors

      PetscCall(VecCreateSeq(PETSC_COMM_SELF, n, x, ierr))

      PetscCall(VecCreateSeq(PETSC_COMM_SELF, n, y, ierr))

      PetscCall(VecSet(x, 0.0d0, ierr))

      PetscCall(VecSet(y, 0.0d0, ierr))

 

!.... SNES context

      PetscCall(SNESCreate(PETSC_COMM_SELF, snes, ierr))

      PetscCall(SNESSetType(snes, SNESSHELL, ierr))

      PetscCall(SNESShellSetSolve(snes, MySolver, ierr))

 

!!!!  PetscCall(SNESSetFunction(snes, x, MySolver, PETSC_NULL_INTEGER, ierr))

!!!!  this line causes a segmentation error if uncommented.

 

      PetscCall(SNESSetConvergenceTest(snes, MyConverged, 0, PETSC_NULL_FUNCTION, ierr))

!

!.... Set SNES options

      PetscCall(SNESSetFromOptions(snes, ierr))

 

!.... Set tolerances

      PetscCall(SNESSetTolerances(snes, rtol, atol, dtol, maxits, maxf, ierr))

!

!.... SNES montior

      PetscCall(SNESMonitorSet(snes, MyMonitor, PETSC_NULL_INTEGER, PETSC_NULL_FUNCTION,ierr))

!

!.... Set the initial solution

      CALL HBToVecX(x)

!

!.... View snes context

      PetscCall(SNESView(snes, viewer, ierr))

!

!.... Solve SNES problem

      PetscCall(SNESSolve(snes, PETSC_NULL_VEC, x, ierr))

!

!.... View snes context

      PetscCall(SNESView(snes, viewer, ierr))

!

!.... dump the logs

!     call PetscLogDump(ierr)  ! Why does this cause error

!

!.... Destroy PETSc objects

      PetscCall(SNESDestroy(snes, ierr))

      PetscCall(VecDestroy(x, ierr))

      PetscCall(VecDestroy(y, ierr))

      PetscCall(PetscViewerDestroy(viewer, ierr))

!

!.... Finish

      PetscCall(PetscFinalize(ierr))

 

   END SUBROUTINE SolveWithSNESShell

 

 

--

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

 

https://www.cse.buffalo.edu/~knepley/