oft_mf_matrix Type Reference

Detailed Description

Wrapper matrix for matrix-free Jacobians.

Computes a Finite Difference approximation to the Jacobian of a function \( F'(v) = [F(u0 + h * v) - F(u0)]/h \)

Inheritance diagram for oft_mf_matrix:

Public Member Functions
procedure	add_values (self, i_inds, j_inds, b, n, m, iblock, jblock, loc_cache)
	Add values to a matrix.
procedure(mat_add_values), deferred	add_values (self, i_inds, j_inds, b, n, m, iblock, jblock, loc_cache)
	Add values to the matrix.
generic	apply (self, a, b)
	Apply the matrix.
generic	apply (self, a, b)
	Apply the matrix.
procedure	apply_complex (self, a, b)
	Apply the matrix to a field and optionally add it to an existing field.
procedure	apply_real (self, a, b)
	Apply the matrix to a field.
procedure	apply_real (self, a, b)
	Compute Jacobian approximation.
procedure(mat_apply_vec), deferred	apply_real (self, a, b)
generic	applyt (self, a, b)
	Apply the matrix transpose.
generic	applyt (self, a, b)
	Apply the matrix transpose.
procedure	applyt_complex (self, a, b)
	Apply the matrix to a field.
procedure	applyt_complex (self, a, b)
	Apply the matrix to a field and optionally add it to an existing field.
procedure	applyt_real (self, a, b)
	Apply the matrix to a field.
procedure(mat_apply_vec), deferred	applyt_real (self, a, b)
procedure	assemble (self, diag)
	Finish assembly of matrix and optionally extract diagonals.
procedure(mat_assemble), deferred	assemble (self, diag)
	Complete matrix assembly.
procedure	atomic_add_values (self, i_inds, j_inds, b, n, m, iblock, jblock, loc_cache)
	Add values to a matrix.
procedure(mat_add_values), deferred	atomic_add_values (self, i_inds, j_inds, b, n, m, iblock, jblock, loc_cache)
	Add values atomically to the matrix.
procedure	delete (self)
	Delete matrix.
procedure	delete (self)
	Cleanup internal storage.
procedure	delete (self)
	Delete matrix.
procedure	set_values (self, i_inds, j_inds, b, n, m, iblock, jblock)
	Set values of a matrix.
procedure(mat_set_values), deferred	set_values (self, i_inds, j_inds, b, n, m, iblock, jblock)
	Set values of the matrix.
procedure	setup (self, a, f, utyp)
	Setup MF jacobian operator.
procedure	update (self, a)
	Update linearization point.
procedure	zero (self)
	Zero all entries in matrix.
procedure(mat_zero), deferred	zero (self)
	Zero all elements.
procedure	zero_rows (self, nrows, irows, iblock, keep_diag)
	Zero all entries in the specified rows.
procedure(mat_zero_rows), deferred	zero_rows (self, nrows, irows, iblock, keep_diag)
	Zero all elements in a given row.

Public Attributes
real(r8)	b0 = 1.d-6
	Approximate relative error in function eval.
class(oft_vector), pointer	d => NULL()
	Diagonal entries for scaling.
class(oft_matrix), pointer	f => NULL()
	Non-linear function.
class(oft_vector), pointer	f0 => NULL()
	Linearization value "F(u0)".
logical	force_local = .FALSE.
	Do not stitch resulting vector (Native ONLY)
type(oft_map), dimension(:), pointer	i_map => NULL()
	Row block mapping.
type(oft_map), dimension(:), pointer	j_map => NULL()
	Column block mapping.
integer(i4)	nc
	Local number of columns.
integer(i8)	ncg
	Gobal number of columns.
integer(i4)	ncslice = 0
	Number of owned columns.
integer(i4)	ni = 0
	Number of row blocks.
integer(i4)	nj = 0
	Number of column blocks.
integer(i4)	nr
	Local number of rows.
integer(i8)	nrg
	Gobal number of rows.
integer(i4)	nrslice = 0
	Number of owned rows.
class(oft_vector), pointer	tmp => NULL()
	Internal work vector.
class(oft_vector), pointer	u0 => NULL()
	Linearization point.
class(oft_vector), pointer	utyp => NULL()
	Typical values vector.

Member Function/Subroutine Documentation

add_values() [1/2]

procedure add_values ( class(oft_noop_matrix), intent(inout)  self, integer(i4), dimension(n), intent(in)  i_inds, integer(i4), dimension(m), intent(in)  j_inds, real(r8), dimension(n,m), intent(in)  b, integer(i4), intent(in)  n, integer(i4), intent(in)  m, integer(i4), intent(in), optional  iblock, integer(i4), intent(in), optional  jblock, integer(i4), dimension(n,m), intent(inout), optional  loc_cache  )

inherited

Add values to a matrix.

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	i_inds	Row indices of entries to set [n]
[in]	j_inds	Column indices of entries to set [m]
[in]	b	Values to add [n,m]
[in]	n	Number of rows in local matrix
[in]	m	Number of columns in local matrix
[in]	iblock	Row block (optional)
[in]	jblock	Column block (optional)

add_values() [2/2]

procedure(mat_add_values), deferred add_values ( class(oft_matrix), intent(inout)  self, integer(i4), dimension(n), intent(in)  i_inds, integer(i4), dimension(m), intent(in)  j_inds, real(r8), dimension(n,m), intent(in)  b, integer(i4), intent(in)  n, integer(i4), intent(in)  m, integer(i4), intent(in), optional  iblock, integer(i4), intent(in), optional  jblock, integer(i4), dimension(n,m), intent(inout), optional  loc_cache  )

pure virtualinherited

Add values to the matrix.

apply() [1/2]

generic apply ( class(oft_matrix), intent(inout)  self, class(oft_cvector), intent(inout), target  a, class(oft_cvector), intent(inout)  b  )

inherited

Apply the matrix.

b = self^T * a

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	a	Source field
[in,out]	b	Result of matrix product
[in]	beta	Factor for matrix addition

apply() [2/2]

generic apply ( class(oft_matrix), intent(inout)  self, class(oft_vector), intent(inout), target  a, class(oft_vector), intent(inout)  b  )

inherited

Apply the matrix.

apply_complex()

procedure apply_complex ( class(oft_matrix), intent(inout)  self, class(oft_cvector), intent(inout), target  a, class(oft_cvector), intent(inout)  b  )

inherited

Apply the matrix to a field and optionally add it to an existing field.

b = self^T * a

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	a	Source field
[in,out]	b	Result of matrix product
[in]	beta	Factor for matrix addition

apply_real() [1/3]

procedure apply_real ( class(oft_noop_matrix), intent(inout)  self, class(oft_vector), intent(inout), target  a, class(oft_vector), intent(inout)  b  )

inherited

Apply the matrix to a field.

b = self * a

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	a	Source field
[out]	b	Result of matrix product

apply_real() [2/3]

procedure apply_real	(	class(oft_mf_matrix), intent(inout)	self,
		class(oft_vector), intent(inout), target	a,
		class(oft_vector), intent(inout)	b
	)

Compute Jacobian approximation.

Parameters

[in]	a	Source field
[in,out]	b	F(a)

apply_real() [3/3]

procedure(mat_apply_vec), deferred apply_real ( class(oft_matrix), intent(inout)  self, class(oft_vector), intent(inout), target  a, class(oft_vector), intent(inout)  b  )

pure virtualinherited

applyt() [1/2]

generic applyt ( class(oft_matrix), intent(inout)  self, class(oft_cvector), intent(inout), target  a, class(oft_cvector), intent(inout)  b  )

inherited

Apply the matrix transpose.

b = self^T * a

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	a	Source field
[in,out]	b	Result of matrix product
[in]	beta	Factor for matrix addition

applyt() [2/2]

generic applyt ( class(oft_matrix), intent(inout)  self, class(oft_vector), intent(inout), target  a, class(oft_vector), intent(inout)  b  )

inherited

Apply the matrix transpose.

applyt_complex() [1/2]

procedure applyt_complex ( class(oft_noop_matrix), intent(inout)  self, class(oft_cvector), intent(inout), target  a, class(oft_cvector), intent(inout)  b  )

inherited

Apply the matrix to a field.

b = self * a

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	a	Source field
[out]	b	Result of matrix product

applyt_complex() [2/2]

procedure applyt_complex ( class(oft_matrix), intent(inout)  self, class(oft_cvector), intent(inout), target  a, class(oft_cvector), intent(inout)  b  )

inherited

Apply the matrix to a field and optionally add it to an existing field.

b = self^T * a

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	a	Source field
[in,out]	b	Result of matrix product
[in]	beta	Factor for matrix addition

applyt_real() [1/2]

procedure applyt_real ( class(oft_noop_matrix), intent(inout)  self, class(oft_vector), intent(inout), target  a, class(oft_vector), intent(inout)  b  )

inherited

Apply the matrix to a field.

b = self * a

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	a	Source field
[out]	b	Result of matrix product

applyt_real() [2/2]

procedure(mat_apply_vec), deferred applyt_real ( class(oft_matrix), intent(inout)  self, class(oft_vector), intent(inout), target  a, class(oft_vector), intent(inout)  b  )

pure virtualinherited

assemble() [1/2]

procedure assemble ( class(oft_noop_matrix), intent(inout)  self, class(oft_vector), intent(inout), optional, target  diag  )

inherited

Finish assembly of matrix and optionally extract diagonals.

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in,out]

diag

Diagonal entries of matrix [nr] (optional)

assemble() [2/2]

procedure(mat_assemble), deferred assemble ( class(oft_matrix), intent(inout)  self, class(oft_vector), intent(inout), optional, target  diag  )

pure virtualinherited

Complete matrix assembly.

atomic_add_values() [1/2]

procedure atomic_add_values ( class(oft_noop_matrix), intent(inout)  self, integer(i4), dimension(n), intent(in)  i_inds, integer(i4), dimension(m), intent(in)  j_inds, real(r8), dimension(n,m), intent(in)  b, integer(i4), intent(in)  n, integer(i4), intent(in)  m, integer(i4), intent(in), optional  iblock, integer(i4), intent(in), optional  jblock, integer(i4), dimension(n,m), intent(inout), optional  loc_cache  )

inherited

Add values to a matrix.

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	i_inds	Row indices of entries to set [n]
[in]	j_inds	Column indices of entries to set [m]
[in]	b	Values to add [n,m]
[in]	n	Number of rows in local matrix
[in]	m	Number of columns in local matrix
[in]	iblock	Row block (optional)
[in]	jblock	Column block (optional)

atomic_add_values() [2/2]

procedure(mat_add_values), deferred atomic_add_values ( class(oft_matrix), intent(inout)  self, integer(i4), dimension(n), intent(in)  i_inds, integer(i4), dimension(m), intent(in)  j_inds, real(r8), dimension(n,m), intent(in)  b, integer(i4), intent(in)  n, integer(i4), intent(in)  m, integer(i4), intent(in), optional  iblock, integer(i4), intent(in), optional  jblock, integer(i4), dimension(n,m), intent(inout), optional  loc_cache  )

pure virtualinherited

Add values atomically to the matrix.

delete() [1/3]

procedure delete ( class(oft_noop_matrix), intent(inout)  self )

inherited

Delete matrix.

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

delete() [2/3]

procedure delete

(

class(oft_mf_matrix), intent(inout)

self

)

Cleanup internal storage.

delete() [3/3]

procedure delete ( class(oft_matrix), intent(inout)  self )

inherited

Delete matrix.

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

set_values() [1/2]

procedure set_values ( class(oft_noop_matrix), intent(inout)  self, integer(i4), dimension(n), intent(in)  i_inds, integer(i4), dimension(m), intent(in)  j_inds, real(r8), dimension(n,m), intent(in)  b, integer(i4), intent(in)  n, integer(i4), intent(in)  m, integer(i4), intent(in), optional  iblock, integer(i4), intent(in), optional  jblock  )

inherited

Set values of a matrix.

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	i_inds	Row indices of entries to set [n]
[in]	j_inds	Column indices of entries to set [m]
[in]	b	Values to set [n,m]
[in]	n	Number of rows in local matrix
[in]	m	Number of columns in local matrix
[in]	iblock	Row block (optional)
[in]	jblock	Column block (optional)

set_values() [2/2]

procedure(mat_set_values), deferred set_values ( class(oft_matrix), intent(inout)  self, integer(i4), dimension(n), intent(in)  i_inds, integer(i4), dimension(m), intent(in)  j_inds, real(r8), dimension(n,m), intent(in)  b, integer(i4), intent(in)  n, integer(i4), intent(in)  m, integer(i4), intent(in), optional  iblock, integer(i4), intent(in), optional  jblock  )

pure virtualinherited

Set values of the matrix.

setup()

procedure setup	(	class(oft_mf_matrix), intent(inout)	self,
		class(oft_vector), intent(inout)	a,
		class(oft_matrix), intent(in), target	f,
		class(oft_vector), intent(inout), optional	utyp
	)

Setup MF jacobian operator.

Parameters

[in]	a	Field defining domain and range spaces
[in]	f	Non-linear function defining the Jacobian

update()

procedure update	(	class(oft_mf_matrix), intent(inout)	self,
		class(oft_vector), intent(inout)	a
	)

Update linearization point.

Parameters

[in]

a

New linearization point

zero() [1/2]

procedure zero ( class(oft_noop_matrix), intent(inout)  self )

inherited

Zero all entries in matrix.

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

zero() [2/2]

procedure(mat_zero), deferred zero ( class(oft_matrix), intent(inout)  self )

pure virtualinherited

Zero all elements.

zero_rows() [1/2]

procedure zero_rows ( class(oft_noop_matrix), intent(inout)  self, integer(i4), intent(in)  nrows, integer(i4), dimension(nrows), intent(in)  irows, integer(i4), intent(in), optional  iblock, logical, intent(in), optional  keep_diag  )

inherited

Zero all entries in the specified rows.

Note

This subroutine is a dummy routine used to specify the interface of the member function and catch errors in uninitialized matrices.

Parameters

[in]	nrows	Number of rows to zero
[in]	irows	Indices of rows to zero [nrows]
[in]	iblock	Row block (optional)
[in]	keep_diag	Keep diagonal entries

zero_rows() [2/2]

procedure(mat_zero_rows), deferred zero_rows ( class(oft_matrix), intent(inout)  self, integer(i4), intent(in)  nrows, integer(i4), dimension(nrows), intent(in)  irows, integer(i4), intent(in), optional  iblock, logical, intent(in), optional  keep_diag  )

pure virtualinherited

Zero all elements in a given row.

Member Data Documentation

b0

real(r8) b0 = 1.d-6

Approximate relative error in function eval.

d

class(oft_vector), pointer d => NULL()

inherited

Diagonal entries for scaling.

f

class(oft_matrix), pointer f => NULL()

Non-linear function.

f0

class(oft_vector), pointer f0 => NULL()

Linearization value "F(u0)".

force_local

logical force_local = .FALSE.

inherited

Do not stitch resulting vector (Native ONLY)

i_map

type(oft_map), dimension(:), pointer i_map => NULL()

inherited

Row block mapping.

j_map

type(oft_map), dimension(:), pointer j_map => NULL()

inherited

Column block mapping.

nc

integer(i4) nc

inherited

Local number of columns.

ncg

integer(i8) ncg

inherited

Gobal number of columns.

ncslice

integer(i4) ncslice = 0

inherited

Number of owned columns.

ni

integer(i4) ni = 0

inherited

Number of row blocks.

nj

integer(i4) nj = 0

inherited

Number of column blocks.

nr

integer(i4) nr

inherited

Local number of rows.

nrg

integer(i8) nrg

inherited

Gobal number of rows.

nrslice

integer(i4) nrslice = 0

inherited

Number of owned rows.

tmp

class(oft_vector), pointer tmp => NULL()

Internal work vector.

u0

class(oft_vector), pointer u0 => NULL()

Linearization point.

utyp

class(oft_vector), pointer utyp => NULL()

Typical values vector.

---

The documentation for this type was generated from the following file:

• /home/runner/work/OpenFUSIONToolkit/OpenFUSIONToolkit/src/lin_alg/deriv_matrices.F90