cfltest_mod Module Reference

This contains the CFL test. It will perform the local advective CFL and Galilean transfromation calculations, compute the global values of these, then check the cfl criterion and determine an absolute new dtm. Depending upon the maximum and increment values this is then smoothed into a new dtm value. The value of dtm is physically set to this new value at the start of the next timestep. More...

Functions/Subroutines
type(component_descriptor_type) function, public	cfltest_get_descriptor ()
	Provides the descriptor back to the caller and is used in component registration. More...
subroutine	initialisation_callback (current_state)
	Called at initialisation, will read in configuration and use either configured or default values. More...
subroutine	timestep_callback (current_state)
	Called at each timestep, this will only do the CFL computation every nncfl timesteps (or every timestep up to nncfl) but will ratchet up to the absolute (target) dtm as needed. More...
subroutine	update_dtm_based_on_absolute (current_state, cfl_number)
	Updates the (new) dtm value, which is actioned after time step completion, based upon the absolute value. This is incremented towards the absolute if that is too large a step, and even if the absolute value has not been updated in this timestep, this ratcheting will still occur if needed. More...
subroutine	perform_cfl_and_galilean_transformation_calculation (current_state)
	Performs the CFL and Galilean transformation calculations. First locally and then will determine the global value of each calculation. If U, V or W are not active then these are set to 0 and the calculation use these values. More...
subroutine	get_global_values (local_zumin, local_zumax, local_zvmin, local_zvmax, local_cvel_z, local_cvis, global_zumin, global_zumax, global_zvmin, global_zvmax, global_cvel_z, global_cvis, parallel_state)
	Gets the global reduction values based upon the local contributions of CFL and Galilean transformations provided. More...

Variables
real(kind=default_precision)	tollerance
real(kind=default_precision)	cvismax
real(kind=default_precision)	cvelmax
real(kind=default_precision)	dtmmax
real(kind=default_precision)	dtmmin
real(kind=default_precision)	rincmax
logical	l_monitor_cfl

Detailed Description

This contains the CFL test. It will perform the local advective CFL and Galilean transfromation calculations, compute the global values of these, then check the cfl criterion and determine an absolute new dtm. Depending upon the maximum and increment values this is then smoothed into a new dtm value. The value of dtm is physically set to this new value at the start of the next timestep.

Function/Subroutine Documentation

cfltest_get_descriptor()

type(component_descriptor_type) function, public cfltest_mod::cfltest_get_descriptor

Provides the descriptor back to the caller and is used in component registration.

Returns

The termination check component descriptor

Definition at line 31 of file cfltest.F90.

    cfltest_get_descriptor%name="cfltest"
    cfltest_get_descriptor%version=0.1
    cfltest_get_descriptor%initialisation=>initialisation_callback
    cfltest_get_descriptor%timestep=>timestep_callback

Here is the call graph for this function:

get_global_values()

subroutine cfltest_mod::get_global_values ( real(kind=default_precision), intent(in)  local_zumin, real(kind=default_precision), intent(in)  local_zumax, real(kind=default_precision), intent(in)  local_zvmin, real(kind=default_precision), intent(in)  local_zvmax, real(kind=default_precision), intent(in)  local_cvel_z, real(kind=default_precision), intent(in)  local_cvis, real(kind=default_precision), intent(out)  global_zumin, real(kind=default_precision), intent(out)  global_zumax, real(kind=default_precision), intent(out)  global_zvmin, real(kind=default_precision), intent(out)  global_zvmax, real(kind=default_precision), intent(out)  global_cvel_z, real(kind=default_precision), intent(out)  global_cvis, type(parallel_state_type), intent(inout)  parallel_state  )

private

Gets the global reduction values based upon the local contributions of CFL and Galilean transformations provided.

Parameters

local_zumin	Local contribution to ZU minimum
local_zumax	Local contribution to ZU maximum
local_zvmin	Local contribution to ZV minimum
local_zvmax	Local contribution to ZV maximum
local_cvel_z	Local contribution to component of Courant number in z
local_cvis	Local contribution to viscous Courant number
global_zumin	Globally reduced value of ZU minimum
global_zumax	Globally reduced value of ZU maximum
global_zvmin	Globally reduced value of ZV minimum
global_zvmax	Globally reduced value of ZV maximum
global_cvel_z	Globally reduced value of contribution to component of Courant number in z
global_cvis	Globally reduced value of contribution to viscous Courant number
parallel_state	The parallel state

Definition at line 200 of file cfltest.F90.

    type(parallel_state_type), intent(inout) :: parallel_state
    real(kind=default_precision), intent(in) :: local_zumin, local_zumax, local_zvmin, local_zvmax, local_cvel_z, local_cvis
    real(kind=default_precision), intent(out) :: global_zumin, global_zumax, global_zvmin, global_zvmax, global_cvel_z, global_cvis
 
    integer :: ierr
 
    call mpi_allreduce(local_zumax, global_zumax, 1, precision_type, mpi_max, parallel_state%monc_communicator, ierr)
    call mpi_allreduce(local_zvmax, global_zvmax, 1, precision_type, mpi_max, parallel_state%monc_communicator, ierr)
    call mpi_allreduce(local_cvel_z, global_cvel_z, 1, precision_type, mpi_max, parallel_state%monc_communicator, ierr)
    call mpi_allreduce(local_cvis, global_cvis, 1, precision_type, mpi_max, parallel_state%monc_communicator, ierr)
    call mpi_allreduce(local_zumin, global_zumin, 1, precision_type, mpi_min, parallel_state%monc_communicator, ierr)
    call mpi_allreduce(local_zvmin, global_zvmin, 1, precision_type, mpi_min, parallel_state%monc_communicator, ierr)

Here is the caller graph for this function:

initialisation_callback()

subroutine cfltest_mod::initialisation_callback ( type(model_state_type), intent(inout), target  current_state )

private

Called at initialisation, will read in configuration and use either configured or default values.

Parameters

current_state

The current model state

Definition at line 40 of file cfltest.F90.

    type(model_state_type), intent(inout), target :: current_state
 
    current_state%cfl_frequency=options_get_integer(current_state%options_database, "cfl_frequency")
    tollerance=options_get_real(current_state%options_database, "cfl_tollerance")
    cvismax=options_get_real(current_state%options_database, "cfl_cvismax")
    cvelmax=options_get_real(current_state%options_database, "cfl_cvelmax")
    dtmmax=options_get_real(current_state%options_database, "cfl_dtmmax")
    dtmmin=options_get_real(current_state%options_database, "cfl_dtmmin")
    rincmax=options_get_real(current_state%options_database, "cfl_rincmax")
 
    l_monitor_cfl = options_get_logical(current_state%options_database,"cfl_monitor")
 
    allocate(current_state%abswmax(current_state%local_grid%local_domain_end_index(z_index)))

Here is the caller graph for this function:

perform_cfl_and_galilean_transformation_calculation()

subroutine cfltest_mod::perform_cfl_and_galilean_transformation_calculation ( type(model_state_type), intent(inout), target  current_state )

private

Performs the CFL and Galilean transformation calculations. First locally and then will determine the global value of each calculation. If U, V or W are not active then these are set to 0 and the calculation use these values.

Parameters

current_state

The current model state

Definition at line 138 of file cfltest.F90.

    type(model_state_type), intent(inout), target :: current_state
 
    integer :: k
    real(kind=default_precision) :: global_zumin, global_zumax, global_zvmin, &
         global_zvmax, global_cvel_z, global_cvis
 
#ifdef U_ACTIVE
    current_state%local_zumin=current_state%local_zumin+current_state%ugal               ! _undo Gal-trfm                        
    current_state%local_zumax=current_state%local_zumax+current_state%ugal
#else
    current_state%local_zumin=0.0_default_precision
    current_state%local_zumax=0.0_default_precision
#endif
#ifdef V_ACTIVE
    current_state%local_zvmin=current_state%local_zvmin+current_state%vgal               ! _undo Gal-trfm                        
    current_state%local_zvmax=current_state%local_zvmax+current_state%vgal
#else
    current_state%local_zvmin=0.0_default_precision
    current_state%local_zvmax=0.0_default_precision
#endif
#ifdef W_ACTIVE
    current_state%local_cvel_z=current_state%cvel_z
    do k=2,current_state%local_grid%local_domain_end_index(z_index)-1
      !  CVELZ will be multiplied by DTM in TESTCFL
      current_state%local_cvel_z=max(current_state%local_cvel_z, &
           current_state%abswmax(k)*current_state%global_grid%configuration%vertical%rdzn(k+1))
    end do
#else
    current_state%local_cvel_z=0.0_default_precision
#endif
    call get_global_values(current_state%local_zumin, current_state%local_zumax, current_state%local_zvmin, &
         current_state%local_zvmax, current_state%local_cvel_z, current_state%cvis, &
         global_zumin, global_zumax, global_zvmin, global_zvmax, global_cvel_z, global_cvis, current_state%parallel)
 
    if (current_state%galilean_transformation) then
      if (.not.current_state%fix_ugal)current_state%ugal=0.5_default_precision*(global_zumin+global_zumax)
      if (.not.current_state%fix_vgal)current_state%vgal=0.5_default_precision*(global_zvmin+global_zvmax)
    else
      current_state%ugal=0.0_default_precision
      current_state%vgal=0.0_default_precision
    end if
    current_state%cvel_z=global_cvel_z
    current_state%cvel_x=max(abs(global_zumax-current_state%ugal), abs(global_zumin-current_state%ugal))
    current_state%cvel_y=max(abs(global_zvmax-current_state%vgal), abs(global_zvmin-current_state%vgal))
    current_state%cvis=global_cvis

Here is the call graph for this function:

Here is the caller graph for this function:

timestep_callback()

subroutine cfltest_mod::timestep_callback ( type(model_state_type), intent(inout), target  current_state )

private

Called at each timestep, this will only do the CFL computation every nncfl timesteps (or every timestep up to nncfl) but will ratchet up to the absolute (target) dtm as needed.

Parameters

current_state

The current model state

Definition at line 59 of file cfltest.F90.

    type(model_state_type), intent(inout), target :: current_state
 
    real(kind=default_precision) :: cfl_number
 
    if (mod(current_state%timestep, current_state%cfl_frequency) == 1 .or. &
         current_state%timestep-current_state%start_timestep .le. current_state%cfl_frequency) then
      current_state%cvel=0.0_default_precision
      current_state%cvel_x=0.0_default_precision
      current_state%cvel_y=0.0_default_precision
      current_state%cvel_z=0.0_default_precision
 
      call perform_cfl_and_galilean_transformation_calculation(current_state)
 
      current_state%cvel=(current_state%cvel_x*current_state%global_grid%configuration%horizontal%cx+current_state%cvel_y*&
           current_state%global_grid%configuration%horizontal%cy+current_state%cvel_z)*current_state%dtm
      current_state%cvis=current_state%cvis*(current_state%dtm * 4)
 
      cfl_number=current_state%cvis/cvismax+current_state%cvel/cvelmax
 
      current_state%absolute_new_dtm=current_state%dtm
      current_state%update_dtm=.false.
      if (cfl_number .gt. 0.0_default_precision) then
        if (cfl_number .lt. (1.0_default_precision-tollerance) .or. cfl_number .gt. (1.0_default_precision+tollerance)) then
          current_state%absolute_new_dtm=current_state%dtm/cfl_number
        end if
      end if
    end if
    call update_dtm_based_on_absolute(current_state, cfl_number)
    current_state%cvis=0.0_default_precision

Here is the call graph for this function:

Here is the caller graph for this function:

update_dtm_based_on_absolute()

subroutine cfltest_mod::update_dtm_based_on_absolute ( type(model_state_type), intent(inout), target  current_state, real(kind=default_precision), intent(in)  cfl_number  )

private

Updates the (new) dtm value, which is actioned after time step completion, based upon the absolute value. This is incremented towards the absolute if that is too large a step, and even if the absolute value has not been updated in this timestep, this ratcheting will still occur if needed.

Parameters

current_state

The current model state

Definition at line 95 of file cfltest.F90.

    type(model_state_type), intent(inout), target :: current_state
    real(kind=default_precision), intent(in)      :: cfl_number
 
    if (current_state%dtm .ne. current_state%absolute_new_dtm .and. &
         (current_state%dtm .ne. dtmmax .or. current_state%absolute_new_dtm .lt. dtmmax)) then
 
      current_state%update_dtm=.true.
 
      current_state%dtm_new=min(current_state%dtm*(1.0_default_precision+rincmax), current_state%absolute_new_dtm, dtmmax)
 
      !! --- Diagnostic Writing -----------------
      if (current_state%parallel%my_rank==0) then
        if (log_get_logging_level() .eq. log_debug) then
          call log_log(log_debug, "dtm changed from "//trim(conv_to_string(current_state%dtm, 5))//" to "//&
               trim(conv_to_string(current_state%dtm_new, 5)))
        end if
        if (current_state%dtm_new .lt. dtmmin) then
          call log_log(log_error, "Timestep too small, dtmnew="//trim(conv_to_string(current_state%dtm_new, 5))//&
               " dtmmin="//trim(conv_to_string(dtmmin, 5)))
        end if
        if (l_monitor_cfl) then
          call log_log(log_info, " --- CFL Monitoring Information --- ")
          call log_log(log_info, "dtm changed from "//trim(conv_to_string(current_state%dtm, 5))//" to "//&
                                 trim(conv_to_string(current_state%dtm_new, 5)))
          if (cfl_number .gt. 0.0) then 
            call log_log(log_info, "cfl_number :  "//trim(conv_to_string(cfl_number))//"  (change divisor)")
            call log_log(log_info, "cvis       :  "//trim(conv_to_string(current_state%cvis)) )
            call log_log(log_info, "cvel       :  "//trim(conv_to_string(current_state%cvel)) )
          else
            call log_log(log_info, "dtm change due to ratcheting only. Target dtm unchanged.")
          end if
          call log_log(log_info, "target dtm :  "//trim(conv_to_string(current_state%absolute_new_dtm)) )
          call log_newline()
 
        end if ! l_monitor_cfl
      end if ! Diagnostic Writing
    end if

Here is the caller graph for this function:

Variable Documentation

cvelmax

real(kind=default_precision) cfltest_mod::cvelmax

private

Definition at line 23 of file cfltest.F90.

cvismax

real(kind=default_precision) cfltest_mod::cvismax

private

Definition at line 23 of file cfltest.F90.

dtmmax

real(kind=default_precision) cfltest_mod::dtmmax

private

Definition at line 23 of file cfltest.F90.

dtmmin

real(kind=default_precision) cfltest_mod::dtmmin

private

Definition at line 23 of file cfltest.F90.

l_monitor_cfl

logical cfltest_mod::l_monitor_cfl

private

Definition at line 24 of file cfltest.F90.

  logical l_monitor_cfl

rincmax

real(kind=default_precision) cfltest_mod::rincmax

private

Definition at line 23 of file cfltest.F90.

tollerance

real(kind=default_precision) cfltest_mod::tollerance

private

Definition at line 23 of file cfltest.F90.

  real(kind=default_precision) :: tollerance, cvismax, cvelmax, dtmmax, dtmmin, rincmax