!
!   Solves the time dependent Bratu problem using pseudo-timestepping
!
!   This code demonstrates how one may solve a nonlinear problem
!   with pseudo-timestepping. In this simple example, the pseudo-timestep
!   is the same for all grid points, i.e., this is equivalent to using
!   the backward Euler method with a variable timestep.
!
!   Note: This example does not require pseudo-timestepping since it
!   is an easy nonlinear problem, but it is included to demonstrate how
!   the pseudo-timestepping may be done.
!
!   See snes/tutorials/ex4.c[ex4f.F] and
!   snes/tutorials/ex5.c[ex5f.F] where the problem is described
!   and solved using the method of Newton alone.
!
!
!23456789012345678901234567890123456789012345678901234567890123456789012
#include <petsc/finclude/petscts.h>
module ex1fmodule
  use petscts
  implicit none
contains
!
!  --------------------  Evaluate Function F(x) ---------------------
!
  subroutine FormFunction(ts, t, X, F, user, ierr)

    TS ts
    PetscReal t
    Vec X, F
    PetscReal user(3)
    PetscErrorCode ierr
    PetscInt i, j, row, mx, my
    PetscReal two, lambda
    PetscReal hx, hy, hxdhy, hydhx
    PetscScalar ut, ub, ul, ur, u
    PetscScalar uxx, uyy, sc
    PetscScalar, pointer :: xx(:), ff(:)
    PetscInt, parameter :: param = 1, lmx = 2, lmy = 3

    two = 2.0

    mx = int(user(lmx))
    my = int(user(lmy))
    lambda = user(param)

    hx = 1.0/real(mx - 1)
    hy = 1.0/real(my - 1)
    sc = hx*hy
    hxdhy = hx/hy
    hydhx = hy/hx

    PetscCall(VecGetArrayRead(X, xx, ierr))
    PetscCall(VecGetArray(F, ff, ierr))
    do j = 1, my
      do i = 1, mx
        row = i + (j - 1)*mx
        if (i == 1 .or. j == 1 .or. i == mx .or. j == my) then
          ff(row) = xx(row)
        else
          u = xx(row)
          ub = xx(row - mx)
          ul = xx(row - 1)
          ut = xx(row + mx)
          ur = xx(row + 1)
          uxx = (-ur + two*u - ul)*hydhx
          uyy = (-ut + two*u - ub)*hxdhy
          ff(row) = -uxx - uyy + sc*lambda*exp(u)
        end if
      end do
    end do

    PetscCall(VecRestoreArrayRead(X, xx, ierr))
    PetscCall(VecRestoreArray(F, ff, ierr))
  end
!
!  --------------------  Evaluate Jacobian of F(x) --------------------
!
  subroutine FormJacobian(ts, ctime, X, JJ, B, user, ierr)

    TS ts
    Vec X
    Mat JJ, B
    PetscReal user(3), ctime
    PetscErrorCode ierr
    Mat jac
    PetscInt i, j, row(1), mx, my
    PetscInt col(5), i1, i5
    PetscScalar two, one, lambda
    PetscScalar v(5), sc
    PetscScalar, pointer :: xx(:)
    PetscReal hx, hy, hxdhy, hydhx

    PetscInt, parameter :: param = 1, lmx = 2, lmy = 3

    i1 = 1
    i5 = 5
    jac = B
    two = 2.0
    one = 1.0

    mx = int(user(lmx))
    my = int(user(lmy))
    lambda = user(param)

    hx = 1.0/real(mx - 1)
    hy = 1.0/real(my - 1)
    sc = hx*hy
    hxdhy = hx/hy
    hydhx = hy/hx

    PetscCall(VecGetArrayRead(X, xx, ierr))
    do j = 1, my
      do i = 1, mx
!
!      When inserting into PETSc matrices, indices start at 0
!
        row(1) = i - 1 + (j - 1)*mx
        if (i == 1 .or. j == 1 .or. i == mx .or. j == my) then
          PetscCall(MatSetValues(jac, i1, [row], i1, [row], [one], INSERT_VALUES, ierr))
        else
          v(1) = hxdhy
          col(1) = row(1) - mx
          v(2) = hydhx
          col(2) = row(1) - 1
          v(3) = -two*(hydhx + hxdhy) + sc*lambda*exp(xx(row(1)))
          col(3) = row(1)
          v(4) = hydhx
          col(4) = row(1) + 1
          v(5) = hxdhy
          col(5) = row(1) + mx
          PetscCall(MatSetValues(jac, i1, [row], i5, col, v, INSERT_VALUES, ierr))
        end if
      end do
    end do
    PetscCall(MatAssemblyBegin(jac, MAT_FINAL_ASSEMBLY, ierr))
    PetscCall(MatAssemblyEnd(jac, MAT_FINAL_ASSEMBLY, ierr))
    PetscCall(VecRestoreArray(X, xx, ierr))
  end
!
!  --------------------  Form initial approximation -----------------
!
  subroutine FormInitialGuess(X, user, ierr)

    Vec X
    PetscReal user(3)
    PetscInt i, j, row, mx, my
    PetscErrorCode ierr
    PetscReal one, lambda
    PetscReal temp1, temp, hx, hy
    PetscScalar, pointer :: xx(:)
    PetscInt, parameter :: param = 1, lmx = 2, lmy = 3

    one = 1.0

    mx = int(user(lmx))
    my = int(user(lmy))
    lambda = user(param)

    hy = one/(my - 1)
    hx = one/(mx - 1)

    PetscCall(VecGetArray(X, xx, ierr))
    temp1 = lambda/(lambda + one)
    do j = 1, my
      temp = min(j - 1, my - j)*hy
      do i = 1, mx
        row = i + (j - 1)*mx
        if (i == 1 .or. j == 1 .or. i == mx .or. j == my) then
          xx(row) = 0.0
        else
          xx(row) = temp1*sqrt(min(min(i - 1, mx - i)*hx, temp))
        end if
      end do
    end do
    PetscCall(VecRestoreArray(X, xx, ierr))
  end
end module ex1fmodule
program main
  use petscts
  use ex1fmodule
  implicit none
!
!  Create an application context to contain data needed by the
!  application-provided call-back routines, FormJacobian() and
!  FormFunction(). We use a double precision array with three
!  entries indexed by param, lmx, lmy.
!
  PetscReal user(3)
  PetscInt i5
  PetscInt, parameter :: param = 1, lmx = 2, lmy = 3
!
!   Data for problem
!
  TS ts
  Vec x, r
  Mat J
  PetscInt its, N, i1000, itmp
  PetscBool flg
  PetscErrorCode ierr
  PetscReal param_max, param_min, dt
  PetscReal tmax
  PetscReal ftime
  TSConvergedReason reason

  i5 = 5
  param_max = 6.81
  param_min = 0

  PetscCallA(PetscInitialize(ierr))
  user(lmx) = 4
  user(lmy) = 4
  user(param) = 6.0

!
!     Allow user to set the grid dimensions and nonlinearity parameter at run-time
!
  PetscCallA(PetscOptionsGetReal(PETSC_NULL_OPTIONS, PETSC_NULL_CHARACTER, '-mx', user(lmx), flg, ierr))
  itmp = 4
  PetscCallA(PetscOptionsGetInt(PETSC_NULL_OPTIONS, PETSC_NULL_CHARACTER, '-my', itmp, flg, ierr))
  user(lmy) = itmp
  PetscCallA(PetscOptionsGetReal(PETSC_NULL_OPTIONS, PETSC_NULL_CHARACTER, '-param', user(param), flg, ierr))
  if (user(param) >= param_max .or. user(param) <= param_min) then
    print *, 'Parameter is out of range'
  end if
  if (user(lmx) > user(lmy)) then
    dt = .5/user(lmx)
  else
    dt = .5/user(lmy)
  end if
  PetscCallA(PetscOptionsGetReal(PETSC_NULL_OPTIONS, PETSC_NULL_CHARACTER, '-dt', dt, flg, ierr))
  N = int(user(lmx)*user(lmy))

!
!      Create vectors to hold the solution and function value
!
  PetscCallA(VecCreateSeq(PETSC_COMM_SELF, N, x, ierr))
  PetscCallA(VecDuplicate(x, r, ierr))

!
!    Create matrix to hold Jacobian. Preallocate 5 nonzeros per row
!    in the sparse matrix. Note that this is not the optimal strategy see
!    the Performance chapter of the users manual for information on
!    preallocating memory in sparse matrices.
!
  i5 = 5
  PetscCallA(MatCreateSeqAIJ(PETSC_COMM_SELF, N, N, i5, PETSC_NULL_INTEGER_ARRAY, J, ierr))

!
!     Create timestepper context
!

  PetscCallA(TSCreate(PETSC_COMM_WORLD, ts, ierr))
  PetscCallA(TSSetProblemType(ts, TS_NONLINEAR, ierr))

!
!     Tell the timestepper context where to compute solutions
!

  PetscCallA(TSSetSolution(ts, x, ierr))

!
!    Provide the call-back for the nonlinear function we are
!    evaluating. Thus whenever the timestepping routines need the
!    function they will call this routine. Note the final argument
!    is the application context used by the call-back functions.
!

  PetscCallA(TSSetRHSFunction(ts, PETSC_NULL_VEC, FormFunction, user, ierr))

!
!     Set the Jacobian matrix and the function used to compute
!     Jacobians.
!

  PetscCallA(TSSetRHSJacobian(ts, J, J, FormJacobian, user, ierr))

!
!       For the initial guess for the problem
!

  PetscCallA(FormInitialGuess(x, user, ierr))

!
!       This indicates that we are using pseudo timestepping to
!     find a steady state solution to the nonlinear problem.
!

  PetscCallA(TSSetType(ts, TSPSEUDO, ierr))

!
!       Set the initial time to start at (this is arbitrary for
!     steady state problems and the initial timestep given above
!

  PetscCallA(TSSetTimeStep(ts, dt, ierr))

!
!      Set a large number of timesteps and final duration time
!     to insure convergence to steady state.
!
  i1000 = 1000
  tmax = 1.e12
  PetscCallA(TSSetMaxSteps(ts, i1000, ierr))
  PetscCallA(TSSetMaxTime(ts, tmax, ierr))
  PetscCallA(TSSetExactFinalTime(ts, TS_EXACTFINALTIME_STEPOVER, ierr))

!
!      Set any additional options from the options database. This
!     includes all options for the nonlinear and linear solvers used
!     internally the timestepping routines.
!

  PetscCallA(TSSetFromOptions(ts, ierr))

  PetscCallA(TSSetUp(ts, ierr))

!
!      Perform the solve. This is where the timestepping takes place.
!
  PetscCallA(TSSolve(ts, x, ierr))
  PetscCallA(TSGetSolveTime(ts, ftime, ierr))
  PetscCallA(TSGetStepNumber(ts, its, ierr))
  PetscCallA(TSGetConvergedReason(ts, reason, ierr))

  write (6, 100) its, ftime, reason
100 format('Number of pseudo time-steps ', i5, ' final time ', 1pe9.2, ' reason ', i3)

!
!     Free the data structures constructed above
!

  PetscCallA(VecDestroy(x, ierr))
  PetscCallA(VecDestroy(r, ierr))
  PetscCallA(MatDestroy(J, ierr))
  PetscCallA(TSDestroy(ts, ierr))
  PetscCallA(PetscFinalize(ierr))
end
!/*TEST
!
!    test:
!      TODO: broken
!      args: -ksp_gmres_cgs_refinement_type refine_always -snes_type newtonls -ts_monitor_pseudo -ts_max_snes_failures 3 -ts_pseudo_frtol 1.e-5 -snes_stol 1e-5
!
!TEST*/
