All the subroutines associated to solving non-linear systems, Schur complement, etc. More...

Functions/Subroutines
subroutine, public	boundedsolutioncorrections (state, packed_state, Mdims, CV_funs, small_findrm, small_colm, Field_name, for_sat, min_max_limits)
	The sparcity of the local CV connectivity is in: small_findrm, small_colm. ngl_its=max no of global iterations e.g. 100. error_tol = tolerance on the iterations. More...

subroutine, public	fpi_backtracking (nphase, Mdims, ndgln, state, packed_state, sat_bak, backtrack_sat, backtrack_par_from_schema, Previous_convergence, satisfactory_convergence, new_backtrack_par, Max_sat_its, its, nonlinear_iteration, useful_sats, res, res_ratio, first_res)
	:In this subroutine we applied some corrections and backtrack_par on the saturations obtained from the saturation equation this subroutine is detailed in Salinas et al. 2017 doi:10.1002/fld.4357 The method ensures convergence "independent" of the time step. More...

subroutine, public	set_saturation_to_sum_one (mdims, packed_state, state, do_not_update_halos)
	:This subroutines eliminates the oscillations in the saturation that are bigger than a certain tolerance and also sets the saturation to be between bounds More...

subroutine, public	non_porous_ensure_sum_to_one (Mdims, packed_state, do_not_update_halos)
	: This subroutines eliminates the oscillations in the saturation that are bigger than a certain tolerance and also sets the saturation to be between bounds More...

subroutine, public	initialise_saturation_sums_one (mdims, packed_state, find_scapegoat_phase)
	:Ensure that the saturations at the beginning sum to one, if they do not all the error is compensated in the scapegoat_phase. Normally the last More...

subroutine, public	auto_backtracking (Mdims, backtrack_par_factor, courant_number_in, first_time_step, nonlinear_iteration, I_am_temperature)
	: The maximum backtracking factor is calculated based on the Courant number and physical effects ocurring in the domain More...

subroutine, public	get_anderson_acceleration_new_guess (N, M, NewField, History_field, stored_residuals, max_its, prev_small_matrix, restart_now)
	This subroutine provides a new guess based on previous updates and residuals. Use this subroutine to speed up any system being solved by looping. A new guess is computed and returned Method explained in DOI.10.1137/10078356X. More...

subroutine, public	scale_petsc_matrix (Mat_petsc)
	In this subroutine the matrix is re-scaled based on the formula D^-0.5 * A * D^-0.5 X'= D^-0.5 b; and next X = D^-0.5 * X'; IMPORTANT: the step X = D^-0.5 * X' needs to be done elsewhere store the diagonal before calling this This should allow to deal with high ranges of viscosity ratio for example A is-written. More...

subroutine, public	duplicate_petsc_matrix (MAT_A, MAT_B)

subroutine, public	petsc_stokes_solver (packed_state, Mdims, Mmat, ndgln, Mspars, final_phase, pmat, P_all, deltaP, rhs_p, solver_option_path, Dmat)
	In this subroutine the Schur complement is generated and solved using PETSc to update the pressure field Matrices need to be in petsc format and pmat is the preconditioned matrix, i.e. pmat = A11- A10(AproxA00^-1)A01. More...

subroutine	convert_c_and_ct_mat_to_petsc_format (packed_state, Mdims, Mmat, ndgln, Mspars, NPHASE)
	: This subroutine converts the C (gradient) and CT (Divergence) matrices into PETSc format Mainly devoted to be used by the PETSc stokes schur solver This can be used when moving from the ICFERST matrix format to PETSc as reference More...

Detailed Description

All the subroutines associated to solving non-linear systems, Schur complement, etc.

Function/Subroutine Documentation

◆ auto_backtracking()

subroutine, public solvers_module::auto_backtracking	(	type(multi_dimensions), intent(in)	Mdims,
		real, intent(inout)	backtrack_par_factor,
		real, dimension(:), intent(inout)	courant_number_in,
		logical, intent(in)	first_time_step,
		integer, intent(in)	nonlinear_iteration,
		logical, intent(in), optional	I_am_temperature
	)

: The maximum backtracking factor is calculated based on the Courant number and physical effects ocurring in the domain

Parameters

Mdims	Data type storing all the dimensions describing the mesh, fields, nodes, etc
backtrack_par_factor	INOUT backtracking value to use
courant_number_in	Courant number and shock front courant number
first_time_step	true if the first one
nonlinear_iteration	current non-linear iteration
I_am_temperature	If doing temperature true (DEPRECATED PROBABLY)

Here is the caller graph for this function:

◆ boundedsolutioncorrections()

subroutine, public solvers_module::boundedsolutioncorrections	(	type( state_type ), dimension( : ), intent(inout)	state,
		type( state_type ), intent(inout)	packed_state,
		type(multi_dimensions), intent(in)	Mdims,
		type(multi_shape_funs), intent(in)	CV_funs,
		integer, dimension( : ), intent(in)	small_findrm,
		integer, dimension( : ), intent(in)	small_colm,
		character( len=* ), intent(in)	Field_name,
		logical, intent(in), optional	for_sat,
		real, dimension(2), optional	min_max_limits
	)

The sparcity of the local CV connectivity is in: small_findrm, small_colm. ngl_its=max no of global iterations e.g. 100. error_tol = tolerance on the iterations.

nloc_its: This iteration is very good at avoiding spreading the modifications too far - however it can stagnate. nloc_its2: This iteration is very good at avoiding stagnating but does spread the modifcations far. us a single iteration because of this as default... nits_nod: iterations at a nod - this iteration is very good at avoiding spreading the modifications too far - however it can stagnate.

Parameters

state	Linked list containing all the fields defined in diamond and considered by Fluidity
packed_state	Linked list containing all the fields used by IC-FERST, memory partially shared with state
Mdims	Data type storing all the dimensions describing the mesh, fields, nodes, etc
CV_funs	Shape functions for the CV mesh
small_findrm	sparcity of the local CV connectivity is in: small_findrm, small_colm.
small_colm	sparcity of the local CV connectivity is in: small_findrm, small_colm.
Field_name	Name of the field
for_sat	True if doing it for Phase Volume Fraction
min_max_limits	Maximum and minimum values

Here is the call graph for this function:

Here is the caller graph for this function:

◆ convert_c_and_ct_mat_to_petsc_format()

subroutine solvers_module::convert_c_and_ct_mat_to_petsc_format	(	type( state_type ), intent(inout)	packed_state,
		type(multi_dimensions), intent(in)	Mdims,
		type(multi_matrices), intent(inout)	Mmat,
		type(multi_ndgln), intent(in)	ndgln,
		type(multi_sparsities), intent(in)	Mspars,
		integer, intent(in)	NPHASE
	)

: This subroutine converts the C (gradient) and CT (Divergence) matrices into PETSc format Mainly devoted to be used by the PETSc stokes schur solver This can be used when moving from the ICFERST matrix format to PETSc as reference

Parameters

packed_state	Linked list containing all the fields used by IC-FERST, memory partially shared with state
Mdims	Data type storing all the dimensions describing the mesh, fields, nodes, etc
Mmat	Matrices for ICFERST
ndgln	Global to local variables
Mspars	Sparsity of the matrices
nphase	number of phases

Here is the caller graph for this function:

◆ duplicate_petsc_matrix()

subroutine, public solvers_module::duplicate_petsc_matrix	(	type(petsc_csr_matrix), intent(in)	MAT_A,
		type(petsc_csr_matrix), intent(inout)	MAT_B
	)

◆ fpi_backtracking()

subroutine, public solvers_module::fpi_backtracking	(	integer, intent(in)	nphase,
		type( multi_dimensions ), intent(in)	Mdims,
		type(multi_ndgln), intent(in)	ndgln,
		type( state_type ), dimension( : ), intent(in)	state,
		type( state_type ), intent(inout)	packed_state,
		real, dimension(:, :), intent(in)	sat_bak,
		real, dimension(:, :), intent(in)	backtrack_sat,
		real, intent(in)	backtrack_par_from_schema,
		real, intent(inout)	Previous_convergence,
		logical, intent(inout)	satisfactory_convergence,
		real, intent(inout)	new_backtrack_par,
		integer, intent(in)	Max_sat_its,
		integer, intent(in)	its,
		integer, intent(in)	nonlinear_iteration,
		integer, intent(inout)	useful_sats,
		real, intent(in)	res,
		real, intent(in)	res_ratio,
		real, intent(in)	first_res
	)

:In this subroutine we applied some corrections and backtrack_par on the saturations obtained from the saturation equation this subroutine is detailed in Salinas et al. 2017 doi:10.1002/fld.4357 The method ensures convergence "independent" of the time step.

Parameters

nphase	Number of phases
Mdims	Data type storing all the dimensions describing the mesh, fields, nodes, etc
ndgln	Global to local variables
state	Linked list containing all the fields defined in diamond and considered by Fluidity
packed_state	Linked list containing all the fields used by IC-FERST, memory partially shared with state
sat_bak	Backup of the saturation field
backtrack_sat	Backtrack parameter used
backtrack_par_from_schema	obtained from diamond/input file
Previous_convergence	Convergence obtained in the previous attempt
satisfactory_convergence	If converged true
new_backtrack_par	New obtained backtracking parameter
Max_sat_its	Maximum allowed number of saturation iterations
its	Iterations taken?
nonlinear_iteration	current non_linear iteration
useful_sats	Number of useful iterations of the saturation iteration method
res	residual obtained using the new sigmas obtained with the new saturation in the transport equation
res_ratio	ratio of residuals between to iterations
first_res	residual obtained at the first attemp, used as reference

Here is the call graph for this function:

Here is the caller graph for this function:

◆ get_anderson_acceleration_new_guess()

subroutine, public solvers_module::get_anderson_acceleration_new_guess	(	integer, intent(in)	N,
		integer, intent(in)	M,
		real, dimension(n), intent(out)	NewField,
		real, dimension(n, m+2), intent(inout)	History_field,
		real, dimension(n, m+1), intent(inout)	stored_residuals,
		integer, optional	max_its,
		real, dimension(:,:), intent(inout), optional, allocatable	prev_small_matrix,
		logical, intent(inout)	restart_now
	)

This subroutine provides a new guess based on previous updates and residuals. Use this subroutine to speed up any system being solved by looping. A new guess is computed and returned Method explained in DOI.10.1137/10078356X.

Author: Pablo Salinas

Parameters

N	Size if the field of interest. Used also to turn vector/tensor fields into scalar fields internally here
M	Size of the field of iterations available - 2
NewField	New guess obtained by the Anderson Acceleration method
History_field	Past results obtained by the outer solver
stored_residuals	Past residuals obtained as the (- guessed value + G(guess value) )
max_its	Maximum number of iterations
prev_small_matrix	Contains the previous A'*A, speeds up the method. On exit is updated (M,M) Prepared to be passed unallocated from M = 1 and allocated internally

Here is the caller graph for this function:

◆ initialise_saturation_sums_one()

subroutine, public solvers_module::initialise_saturation_sums_one	(	type( multi_dimensions ), intent(in)	mdims,
		type( state_type ), intent(inout)	packed_state,
		logical, intent(in), optional	find_scapegoat_phase
	)

:Ensure that the saturations at the beginning sum to one, if they do not all the error is compensated in the scapegoat_phase. Normally the last

Parameters

Mdims	Data type storing all the dimensions describing the mesh, fields, nodes, etc
find_scapegoat_phase	Against which phase we correct the error

◆ non_porous_ensure_sum_to_one()

subroutine, public solvers_module::non_porous_ensure_sum_to_one	(	type( multi_dimensions ), intent(in)	Mdims,
		type( state_type ), intent(inout)	packed_state,
		logical, intent(in), optional	do_not_update_halos
	)

: This subroutines eliminates the oscillations in the saturation that are bigger than a certain tolerance and also sets the saturation to be between bounds

Parameters

Mdims	Data type storing all the dimensions describing the mesh, fields, nodes, etc
packed_state	Linked list containing all the fields used by IC-FERST, memory partially shared with state
do_not_update_halos	If true do not update halos, to save time in parallel

Here is the caller graph for this function:

◆ petsc_stokes_solver()

subroutine, public solvers_module::petsc_stokes_solver	(	type( state_type ), intent(inout)	packed_state,
		type(multi_dimensions), intent(in)	Mdims,
		type(multi_matrices), intent(inout)	Mmat,
		type(multi_ndgln), intent(in)	ndgln,
		type(multi_sparsities), intent(in)	Mspars,
		integer, intent(in)	final_phase,
		type(petsc_csr_matrix), intent(inout)	pmat,
		type( tensor_field ), intent(inout)	P_all,
		type( vector_field ), intent(inout)	deltaP,
		type( vector_field ), intent(in)	rhs_p,
		character(len=option_path_len), intent(in)	solver_option_path,
		type(petsc_csr_matrix), intent(inout), optional	Dmat
	)

In this subroutine the Schur complement is generated and solved using PETSc to update the pressure field Matrices need to be in petsc format and pmat is the preconditioned matrix, i.e. pmat = A11- A10(AproxA00^-1)A01.

Author: Pablo Salinas

Parameters

packed_state	Linked list containing all the fields used by IC-FERST, memory partially shared with state
Mdims	Data type storing all the dimensions describing the mesh, fields, nodes, etc
Mspars	Sparsity of the matrices
ndgln	Global to local variables
final_phase	This is the final phase to be assembled, in this way we can assemble from phase 1 to final_phase not necessarily being for all the phases
pmat	INOUT preconditioned matrix, i.e. pmat = A11- A10(AproxA00^-1)A01
P_all	Pressure field
deltaP	Update of pressure from before solving
rhs_p	RHS of the Pressure system
solver_option_path	Path of the solver to be used to solve here the system
Dmat	(optional) is the Matrix multipliying pressure in the continuity equation

Here is the call graph for this function:

Here is the caller graph for this function:

◆ scale_petsc_matrix()

subroutine, public solvers_module::scale_petsc_matrix ( type(petsc_csr_matrix), intent(inout) Mat_petsc )

In this subroutine the matrix is re-scaled based on the formula D^-0.5 * A * D^-0.5 X'= D^-0.5 b; and next X = D^-0.5 * X'; IMPORTANT: the step X = D^-0.5 * X' needs to be done elsewhere store the diagonal before calling this This should allow to deal with high ranges of viscosity ratio for example A is-written.

Author: Pablo Salinas

Here is the caller graph for this function:

◆ set_saturation_to_sum_one()

subroutine, public solvers_module::set_saturation_to_sum_one	(	type( multi_dimensions ), intent(in)	mdims,
		type( state_type ), intent(inout)	packed_state,
		type( state_type ), dimension(:), intent(in)	state,
		logical, intent(in), optional	do_not_update_halos
	)

:This subroutines eliminates the oscillations in the saturation that are bigger than a certain tolerance and also sets the saturation to be between bounds

Parameters

Mdims	Data type storing all the dimensions describing the mesh, fields, nodes, etc
packed_state	Linked list containing all the fields used by IC-FERST, memory partially shared with state
state	Linked list containing all the fields defined in diamond and considered by Fluidity
do_not_update_halos	If true do not update halos, to save time in parallel

Here is the caller graph for this function:

Functions/Subroutines

Detailed Description

Function/Subroutine Documentation

◆ auto_backtracking()

◆ boundedsolutioncorrections()

◆ convert_c_and_ct_mat_to_petsc_format()

◆ duplicate_petsc_matrix()

◆ fpi_backtracking()

◆ get_anderson_acceleration_new_guess()

◆ initialise_saturation_sums_one()

◆ non_porous_ensure_sum_to_one()

◆ petsc_stokes_solver()

◆ scale_petsc_matrix()

◆ set_saturation_to_sum_one()