Linear algebra (dolfinx::la)

namespace la

Linear algebra interface.

Interface to linear algebra data structures and solvers

Enums

enum class Norm

Norm types.

Values:

enumerator l1

enumerator l2

enumerator linf

enumerator frobenius

Functions

template<typename T, class Allocator>
T inner_product(const Vector<T, Allocator> &a, const Vector<T, Allocator> &b)

Compute the inner product of two vectors. The two vectors must have the same parallel layout.

Note

Collective MPI operation

Parameters

• a – A vector

• b – A vector

Returns

Returns a^{H} b (a^{T} b if a and b are real)

template<typename T, class Allocator>
auto squared_norm(const Vector<T, Allocator> &a)

Compute the squared L2 norm of vector.

Note

Collective MPI operation

template<typename T, class Allocator>
auto norm(const Vector<T, Allocator> &a, Norm type = Norm::l2)

Compute the norm of the vector.

Note

Collective MPI operation

Parameters

• a – A vector

• type – Norm type (supported types are \(L^2\) and \(L^\infty\))

template<typename T, typename U>
void orthonormalize(const std::span<Vector<T, U>> &basis, double tol = 1.0e-10)

Orthonormalize a set of vectors.

Parameters

• basis – [inout] The set of vectors to orthonormalise. The vectors must have identical parallel layouts. The vectors are modified in-place.

• tol – [in] The tolerance used to detect a linear dependency

template<typename T, typename U>
bool is_orthonormal(const std::span<const Vector<T, U>> &basis, double tol = 1.0e-10)

Test if basis is orthonormal.

Parameters

• basis – [in] The set of vectors to check

• tol – [in] The tolerance used to test for orthonormality

Returns

True is basis is orthonormal, otherwise false

template<typename T, class Allocator = std::allocator<T>>
class MatrixCSR

#include <MatrixCSR.h>

Distributed sparse matrix.

The matrix storage format is compressed sparse row. The matrix is partitioned row-wise across MPI rank.

Todo:

Handle block sizes

Note

Highly “experimental” storage of a matrix in CSR format which can be assembled into using the usual dolfinx assembly routines Matrix internal data can be accessed for interfacing with other code.

Template Parameters

• T – The data type for the matrix

• Allocator – The memory allocator type for the data storage

Public Types

using value_type = T

The value type.

using allocator_type = Allocator

The allocator type.

Public Functions

inline auto mat_set_values()

Insertion functor for setting values in matrix. It is typically used in finite element assembly functions.

Todo:

clarify setting on non-owned enrties

Parameters

A – Matrix to insert into

Returns

Function for inserting values into A

inline auto mat_add_values()

Insertion functor for accumulating values in matrix. It is typically used in finite element assembly functions.

Parameters

A – Matrix to insert into

Returns

Function for inserting values into A

inline MatrixCSR(const SparsityPattern &p, const Allocator &alloc = Allocator())

Create a distributed matrix.

Parameters

• p – [in] The sparsty pattern the describes the parallel distribution and the non-zero structure

• alloc – [in] The memory allocator for the data storafe

MatrixCSR(MatrixCSR &&A) = default

Move constructor.

Todo:

Check handling of MPI_Request

inline void set(T x)

Set all non-zero local entries to a value including entries in ghost rows.

Parameters

x – [in] The value to set non-zero matrix entries to

inline void set(const std::span<const T> &x, const std::span<const std::int32_t> &rows, const std::span<const std::int32_t> &cols)

Set values in the matrix.

Note

Only entries included in the sparsity pattern used to initialize the matrix can be set

Note

All indices are local to the calling MPI rank and entries cannot be set in ghost rows.

Note

This should be called after finalize. Using before finalize will set the values correctly, but incoming values may get added to them during a subsequent finalize operation.

Parameters

• x – [in] The m by n dense block of values (row-major) to set in the matrix

• rows – [in] The row indices of x

• cols – [in] The column indices of x

inline void add(const std::span<const T> &x, const std::span<const std::int32_t> &rows, const std::span<const std::int32_t> &cols)

Accumulate values in the matrix.

Note

Only entries included in the sparsity pattern used to initialize the matrix can be accumulated in to

Note

All indices are local to the calling MPI rank and entries may go into ghost rows.

Note

Use finalize after all entries have been added to send ghost rows to owners. Adding more entries after finalize is allowed, but another call to finalize will then be required.

Parameters

• x – [in] The m by n dense block of values (row-major) to add to the matrix

• rows – [in] The row indices of x

• cols – [in] The column indices of x

inline std::int32_t num_owned_rows() const

Number of local rows excluding ghost rows.

inline std::int32_t num_all_rows() const

Number of local rows including ghost rows.

inline std::vector<T> to_dense() const

Copy to a dense matrix.

Note

This function is typically used for debugging and not used in production

Note

Ghost rows are also returned, and these can be truncated manually by using num_owned_rows() if required.

Returns

Dense copy of the part of the matrix on the calling rank. Storage is row-major.

inline void finalize()

Transfer ghost row data to the owning ranks accumulating received values on the owned rows, and zeroing any existing data in ghost rows.

inline void finalize_begin()

Begin transfer of ghost row data to owning ranks, where it will be accumulated into existing owned rows.

Note

Calls to this function must be followed by MatrixCSR::finalize_end(). Between the two calls matrix values must not be changed.

Note

This function does not change the matrix data. Data update only occurs with finalize_end().

inline void finalize_end()

End transfer of ghost row data to owning ranks.

Note

Must be preceded by MatrixCSR::finalize_begin()

Note

Matrix data received from other processes will be accumulated into locally owned rows, and ghost rows will be zeroed.

inline double norm_squared() const

Compute the Frobenius norm squared.

inline const std::array<std::shared_ptr<const common::IndexMap>, 2> &index_maps() const

Index maps for the row and column space. The row IndexMap contains ghost entries for rows which may be inserted into and the column IndexMap contains all local and ghost columns that may exist in the owned rows.

Returns

Row (0) and column (1) index maps

inline std::vector<T> &values()

Get local data values.

Note

Includes ghost values

inline const std::vector<T> &values() const

Get local values (const version)

Note

Includes ghost values

inline const std::vector<std::int32_t> &row_ptr() const

Get local row pointers.

Note

Includes pointers to ghost rows

inline const std::vector<std::int32_t> &cols() const

Get local column indices.

Note

Includes columns in ghost rows

inline const std::vector<std::int32_t> &off_diag_offset() const

Get the start of off-diagonal (unowned columns) on each row, allowing the matrix to be split (virtually) into two parts. Operations (such as matrix-vector multiply) between the owned parts of the matrix and vector can then be performed separately from operations on the unowned parts.

Note

Includes ghost rows, which should be truncated manually if not required.

class SLEPcEigenSolver

#include <slepc.h>

This class provides an eigenvalue solver for PETSc matrices. It is a wrapper for the SLEPc eigenvalue solver.

Public Functions

explicit SLEPcEigenSolver(MPI_Comm comm)

Create eigenvalue solver.

SLEPcEigenSolver(EPS eps, bool inc_ref_count)

Create eigenvalue solver from EPS object.

SLEPcEigenSolver(SLEPcEigenSolver &&solver)

Move constructor.

~SLEPcEigenSolver()

Destructor.

SLEPcEigenSolver &operator=(SLEPcEigenSolver &&solver)

Move assignment.

void set_operators(const Mat A, const Mat B)

Set opeartors (B may be nullptr for regular eigenvalues problems)

void solve()

Compute all eigenpairs of the matrix A (solve \(A x = \lambda x\))

void solve(std::int64_t n)

Compute the n first eigenpairs of the matrix A (solve \(A x = \lambda x\))

std::complex<PetscReal> get_eigenvalue(int i) const

Get ith eigenvalue.

void get_eigenpair(PetscScalar &lr, PetscScalar &lc, Vec r, Vec c, int i) const

Get ith eigenpair.

int get_iteration_number() const

Get the number of iterations used by the solver.

std::int64_t get_number_converged() const

Get the number of converged eigenvalues.

void set_options_prefix(std::string options_prefix)

Sets the prefix used by PETSc when searching the PETSc options database.

std::string get_options_prefix() const

Returns the prefix used by PETSc when searching the PETSc options database.

void set_from_options() const

Set options from PETSc options database.

EPS eps() const

Return SLEPc EPS pointer.

MPI_Comm comm() const

Return MPI communicator.

class SparsityPattern

#include <SparsityPattern.h>

This class provides a sparsity pattern data structure that can be used to initialize sparse matrices. After assembly, column indices are always sorted in increasing order. Ghost entries are kept after assembly.

Public Functions

SparsityPattern(MPI_Comm comm, const std::array<std::shared_ptr<const common::IndexMap>, 2> &maps, const std::array<int, 2> &bs)

Create an empty sparsity pattern with specified dimensions.

Parameters

• comm – [in] The communicator that the pattenr is defined on

• maps – [in] The index maps describing the [0] row and [1] column index ranges (up to a block size)

• bs – [in] The block sizes for the [0] row and [1] column maps

SparsityPattern(MPI_Comm comm, const std::vector<std::vector<const SparsityPattern*>> &patterns, const std::array<std::vector<std::pair<std::reference_wrapper<const common::IndexMap>, int>>, 2> &maps, const std::array<std::vector<int>, 2> &bs)

Create a new sparsity pattern by concatenating sub-patterns, e.g. pattern =[ pattern00 ][ pattern 01] [ pattern10 ][ pattern 11].

Parameters

• comm – [in] The MPI communicator

• patterns – [in] Rectangular array of sparsity pattern. The patterns must not be finalised. Null block are permited

• maps – [in] Pairs of (index map, block size) for each row block (maps[0]) and column blocks (maps[1])

• bs – [in] Block sizes for the sparsity pattern entries

SparsityPattern(SparsityPattern &&pattern) = default

Move constructor.

~SparsityPattern() = default

Destructor.

SparsityPattern &operator=(SparsityPattern &&pattern) = default

Move assignment.

void insert(const std::span<const std::int32_t> &rows, const std::span<const std::int32_t> &cols)

Insert non-zero locations using local (process-wise) indices.

void insert_diagonal(const std::span<const std::int32_t> &rows)

Insert non-zero locations on the diagonal.

Parameters

rows – [in] The rows in local (process-wise) indices. The indices must exist in the row IndexMap.

void assemble()

Finalize sparsity pattern and communicate off-process entries.

std::shared_ptr<const common::IndexMap> index_map(int dim) const

Index map for given dimension dimension. Returns the index map for rows and columns that will be set by the current MPI rank.

Parameters

dim – [in] The requested map, row (0) or column (1)

Returns

The index map

std::vector<std::int64_t> column_indices() const

Global indices of non-zero columns on owned rows.

Note

The ghosts are computed only once SparsityPattern::assemble has been called

Returns

The global index non-zero columns on this process, including ghosts

common::IndexMap column_index_map() const

Builds the index map for columns after assembly of the sparsity pattern.

Todo:

Should this be compted and stored when finalising the SparsityPattern?

Returns

Map for all non-zero columns on this process, including ghosts

int block_size(int dim) const

Return index map block size for dimension dim.

std::int64_t num_nonzeros() const

Number of nonzeros on this rank after assembly, including ghost rows.

std::int32_t nnz_diag(std::int32_t row) const

Number of non-zeros in owned columns (diagonal block) on a given row.

Note

Can also be used on ghost rows

std::int32_t nnz_off_diag(std::int32_t row) const

Number of non-zeros in unowned columns (off-diagonal block) on a given row.

Note

Can also be used on ghost rows

const graph::AdjacencyList<std::int32_t> &graph() const

Sparsity pattern graph after assembly. Uses local indices for the columns.

Note

Column global indices can be obtained from SparsityPattern::column_index_map()

Note

Includes ghost rows

std::span<const int> off_diagonal_offset() const

Row-wise start of off-diagonal (unowned columns) on each row.

Note

Includes ghost rows

MPI_Comm comm() const

Return MPI communicator.

template<typename T, class Allocator = std::allocator<T>>
class Vector

#include <Vector.h>

Distributed vector.

Public Types

using value_type = T

The value type.

using allocator_type = Allocator

The allocator type.

Public Functions

inline Vector(const std::shared_ptr<const common::IndexMap> &map, int bs, const Allocator &alloc = Allocator())

Create a distributed vector.

inline Vector(const Vector &x)

Copy constructor.

inline Vector(Vector &&x)

Move constructor.

Vector &operator=(Vector &&x) = default

Move Assignment operator.

inline void set(T v)

Set all entries (including ghosts)

Parameters

v – [in] The value to set all entries to (on calling rank)

inline void scatter_fwd_begin()

Begin scatter of local data from owner to ghosts on other ranks.

Note

Collective MPI operation

inline void scatter_fwd_end()

End scatter of local data from owner to ghosts on other ranks.

Note

Collective MPI operation

inline void scatter_fwd()

Scatter local data to ghost positions on other ranks.

Note

Collective MPI operation

inline void scatter_rev_begin()

Start scatter of ghost data to owner.

Note

Collective MPI operation

template<class BinaryOperation>
inline void scatter_rev_end(BinaryOperation op)

End scatter of ghost data to owner. This process may receive data from more than one process, and the received data can be summed or inserted into the local portion of the vector.

Note

Collective MPI operation

Parameters

op – The operation to perform when adding/setting received values (add or insert)

template<class BinaryOperation>
inline void scatter_rev(BinaryOperation op)

Scatter ghost data to owner. This process may receive data from more than one process, and the received data can be summed or inserted into the local portion of the vector.

Note

Collective MPI operation

Parameters

op – IndexMap operation (add or insert)

inline std::shared_ptr<const common::IndexMap> map() const

Get IndexMap.

inline constexpr int bs() const

Get block size.

inline std::span<const T> array() const

Get local part of the vector (const version)

inline std::span<T> mutable_array()

Get local part of the vector.

inline constexpr allocator_type allocator() const

Get the allocator associated with the container.

namespace impl

Functions

template<typename U, typename V, typename W, typename X>
void set_csr(U &&data, const V &cols, const V &row_ptr, const W &x, const X &xrows, const X &xcols, [[maybe_unused]] typename X::value_type local_size)

Set data in a CSR matrix.

Parameters

• data – [out] The CSR matrix data

• cols – [in] The CSR column indices

• row_ptr – [in] The pointer to the ith row in the CSR data

• x – [in] The m by n dense block of values (row-major) to set in the matrix

• xrows – [in] The row indices of x

• xcols – [in] The column indices of x

• local_size – [in] The maximum row index that can be set. Used when debugging is own to check that rows beyond a permitted range are not being set.

template<typename U, typename V, typename W, typename X>
void add_csr(U &&data, const V &cols, const V &row_ptr, const W &x, const X &xrows, const X &xcols)

Add data to a CSR matrix.

Parameters

• data – [out] The CSR matrix data

• cols – [in] The CSR column indices

• row_ptr – [in] The pointer to the ith row in the CSR data

• x – [in] The m by n dense block of values (row-major) to add to the matrix

• xrows – [in] The row indices of x

• xcols – [in] The column indices of x

namespace petsc

Functions

void error(int error_code, std::string filename, std::string petsc_function)

Print error message for PETSc calls that return an error.

std::vector<Vec> create_vectors(MPI_Comm comm, const std::vector<std::span<const PetscScalar>> &x)

Create PETsc vectors from the local data. The data is copied into the PETSc vectors and is not shared.

Note

Caller is responsible for destroying the returned object

Parameters

• comm – [in] The MPI communicator

• x – [in] The vector data owned by the calling rank. All components must have the same length.

Returns

Array of PETSc vectors

Vec create_vector(const common::IndexMap &map, int bs)

Create a ghosted PETSc Vec.

Note

Caller is responsible for destroying the returned object

Parameters

• map – [in] The index map describing the parallel layout (by block)

• bs – [in] The block size

Returns

A PETSc Vec

Vec create_vector(MPI_Comm comm, std::array<std::int64_t, 2> range, const std::span<const std::int64_t> &ghosts, int bs)

Create a ghosted PETSc Vec from a local range and ghost indices.

Note

Caller is responsible for freeing the returned object

Parameters

• comm – [in] The MPI communicator

• range – [in] The local ownership range (by blocks)

• ghosts – [in] Ghost blocks

• bs – [in] The block size. The total number of local entries is bs * (range[1] - range[0]).

Returns

A PETSc Vec

Vec create_vector_wrap(const common::IndexMap &map, int bs, const std::span<const PetscScalar> &x)

Create a PETSc Vec that wraps the data in an array.

Note

The array x must be kept alive to use the PETSc Vec object

Note

The caller should call VecDestroy to free the return PETSc vector

Parameters

• map – [in] The index map that describes the parallel layout of the distributed vector (by block)

• bs – [in] Block size

• x – [in] The local part of the vector, including ghost entries

Returns

A PETSc Vec object that shares the data in x

template<typename Allocator>
Vec create_vector_wrap(const la::Vector<PetscScalar, Allocator> &x)

Create a PETSc Vec that wraps the data in an array.

Parameters

x – [in] The vector to be wrapped

Returns

A PETSc Vec object that shares the data in x

std::vector<IS> create_index_sets(const std::vector<std::pair<std::reference_wrapper<const common::IndexMap>, int>> &maps)

Todo:

This function could take just the local sizes

Compute PETSc IndexSets (IS) for a stack of index maps. E.g., if map[0] = {0, 1, 2, 3, 4, 5, 6} and map[1] = {0, 1, 2, 4} (in local indices) then IS[0] = {0, 1, 2, 3, 4, 5, 6} and IS[1] = {7, 8, 9, 10}.

Note

The caller is responsible for destruction of each IS.

Parameters

maps – [in] Vector of IndexMaps and corresponding block sizes

Returns

Vector of PETSc Index Sets, created onPETSC_COMM_SELF

std::vector<std::vector<PetscScalar>> get_local_vectors(const Vec x, const std::vector<std::pair<std::reference_wrapper<const common::IndexMap>, int>> &maps)

Copy blocks from Vec into local vectors.

void scatter_local_vectors(Vec x, const std::vector<std::span<const PetscScalar>> &x_b, const std::vector<std::pair<std::reference_wrapper<const common::IndexMap>, int>> &maps)

Scatter local vectors to Vec.

Mat create_matrix(MPI_Comm comm, const SparsityPattern &sp, const std::string &type = std::string())

Create a PETSc Mat. Caller is responsible for destroying the returned object.

MatNullSpace create_nullspace(MPI_Comm comm, const std::span<const Vec> &basis)

Create PETSc MatNullSpace. Caller is responsible for destruction returned object.

Parameters

• comm – [in] The MPI communicator

• basis – [in] The nullspace basis vectors

Returns

A PETSc nullspace object

class Vector

#include <petsc.h>

A simple wrapper for a PETSc vector pointer (Vec). Its main purpose is to assist with memory/lifetime management of PETSc Vec objects.

Access the underlying PETSc Vec pointer using the function Vector::vec() and use the full PETSc interface.

Public Functions

Vector(const common::IndexMap &map, int bs)

Create a vector.

Note

Collective

Parameters

• map – [in] Index map describing the parallel layout

• bs – [in] the block size

Vector(Vector &&x)

Move constructor.

Vector(Vec x, bool inc_ref_count)

Create holder of a PETSc Vec object/pointer. The Vec x object should already be created. If inc_ref_count is true, the reference counter of the Vec object will be increased. The Vec reference count will always be decreased upon destruction of the PETScVector.

Note

Collective

Parameters

• x – [in] The PETSc Vec

• inc_ref_count – [in] True if the reference count of x should be incremented

virtual ~Vector()

Destructor.

Vector &operator=(Vector &&x)

Move Assignment operator.

Vector copy() const

Create a copy of the vector.

Note

Collective

std::int64_t size() const

Return global size of the vector.

std::int32_t local_size() const

Return local size of vector (belonging to the call rank)

std::array<std::int64_t, 2> local_range() const

Return ownership range for calling rank.

MPI_Comm comm() const

Return MPI communicator.

void set_options_prefix(std::string options_prefix)

Sets the prefix used by PETSc when searching the options database.

std::string get_options_prefix() const

Returns the prefix used by PETSc when searching the options database.

void set_from_options()

Call PETSc function VecSetFromOptions on the underlying Vec object.

Vec vec() const

Return pointer to PETSc Vec object.

class Operator

#include <petsc.h>

This class is a base class for matrices that can be used in petsc::KrylovSolver.

Subclassed by Matrix

Public Functions

Operator(Mat A, bool inc_ref_count)

Constructor.

Operator(Operator &&A)

Move constructor.

virtual ~Operator()

Destructor.

Operator &operator=(const Operator &A) = delete

Assignment operator (deleted)

Operator &operator=(Operator &&A)

Move assignment operator.

std::array<std::int64_t, 2> size() const

Return number of rows and columns (num_rows, num_cols). PETSc returns -1 if size has not been set.

Vec create_vector(std::size_t dim) const

Initialize vector to be compatible with the matrix-vector product y = Ax. In the parallel case, size and layout are both important.

Parameters

dim – [in] The dimension (axis): dim = 0 –> z = y, dim = 1 –> z = x

Mat mat() const

Return PETSc Mat pointer.

class Matrix : public Operator

#include <petsc.h>

It is a simple wrapper for a PETSc matrix pointer (Mat). Its main purpose is to assist memory management of PETSc Mat objects.

For advanced usage, access the PETSc Mat pointer using the function mat() and use the standard PETSc interface.

Public Types

enum class AssemblyType : std::int32_t

Assembly type FINAL - corresponds to PETSc MAT_FINAL_ASSEMBLY FLUSH - corresponds to PETSc MAT_FLUSH_ASSEMBLY.

Values:

enumerator FINAL

enumerator FLUSH

Public Functions

Matrix(MPI_Comm comm, const SparsityPattern &sp, const std::string &type = std::string())

Create holder for a PETSc Mat object from a sparsity pattern.

Matrix(Mat A, bool inc_ref_count)

Create holder of a PETSc Mat object/pointer. The Mat A object should already be created. If inc_ref_count is true, the reference counter of the Mat will be increased. The Mat reference count will always be decreased upon destruction of the petsc::Matrix.

Matrix(Matrix &&A) = default

Move constructor (falls through to base class move constructor)

~Matrix() = default

Destructor.

Matrix &operator=(const Matrix &A) = delete

Assignment operator (deleted)

Matrix &operator=(Matrix &&A) = default

Move assignment operator.

void apply(AssemblyType type)

Finalize assembly of tensor. The following values are recognized for the mode parameter:

Parameters

type – FINAL - corresponds to PETSc MatAssemblyBegin+End(MAT_FINAL_ASSEMBLY) FLUSH - corresponds to PETSc MatAssemblyBegin+End(MAT_FLUSH_ASSEMBLY)

double norm(Norm norm_type) const

Return norm of matrix.

void set_options_prefix(std::string options_prefix)

Sets the prefix used by PETSc when searching the options database.

std::string get_options_prefix() const

Returns the prefix used by PETSc when searching the options database.

void set_from_options()

Call PETSc function MatSetFromOptions on the PETSc Mat object.

Public Static Functions

static inline auto set_fn(Mat A, InsertMode mode)

Return a function with an interface for adding or inserting values into the matrix A (calls MatSetValuesLocal)

Parameters

• A – [in] The matrix to set values in

• mode – [in] The PETSc insert mode (ADD_VALUES, INSERT_VALUES, …)

static inline auto set_block_fn(Mat A, InsertMode mode)

Return a function with an interface for adding or inserting values into the matrix A using blocked indices (calls MatSetValuesBlockedLocal)

Parameters

• A – [in] The matrix to set values in

• mode – [in] The PETSc insert mode (ADD_VALUES, INSERT_VALUES, …)

static inline auto set_block_expand_fn(Mat A, int bs0, int bs1, InsertMode mode)

Return a function with an interface for adding or inserting blocked values to the matrix A using non-blocked insertion (calls MatSetValuesLocal). Internally it expands the blocked indices into non-blocked arrays.

Parameters

• A – [in] The matrix to set values in

• bs0 – [in] Block size for the matrix rows

• bs1 – [in] Block size for the matrix columns

• mode – [in] The PETSc insert mode (ADD_VALUES, INSERT_VALUES, …)

class KrylovSolver

#include <petsc.h>

This class implements Krylov methods for linear systems of the form Ax = b. It is a wrapper for the Krylov solvers of PETSc.

Public Functions

explicit KrylovSolver(MPI_Comm comm)

Create Krylov solver for a particular method and named preconditioner.

KrylovSolver(KSP ksp, bool inc_ref_count)

Create solver wrapper of a PETSc KSP object.

Parameters

• ksp – [in] The PETSc KSP object. It should already have been created

• inc_ref_count – [in] Increment the reference count on ksp if true

KrylovSolver(KrylovSolver &&solver)

Move constructor.

~KrylovSolver()

Destructor.

KrylovSolver &operator=(KrylovSolver &&solver)

Move assignment.

void set_operator(const Mat A)

Set operator (Mat)

void set_operators(const Mat A, const Mat P)

Set operator and preconditioner matrix (Mat)

int solve(Vec x, const Vec b, bool transpose = false) const

Solve linear system Ax = b and return number of iterations (A^t x = b if transpose is true)

void set_options_prefix(std::string options_prefix)

Sets the prefix used by PETSc when searching the PETSc options database.

std::string get_options_prefix() const

Returns the prefix used by PETSc when searching the PETSc options database.

void set_from_options() const

Set options from PETSc options database.

KSP ksp() const

Return PETSc KSP pointer.

namespace options

These class provides static functions that permit users to set and retrieve PETSc options via the PETSc option/parameter system. The option must not be prefixed by ‘-’, e.g.

la::petsc::options::set("mat_mumps_icntl_14", 40);

Functions

void set(std::string option)

Set PETSc option that takes no value.

template<typename T>
void set(std::string option, const T value)

Generic function for setting PETSc option.

void clear(std::string option)

Clear a PETSc option.

void clear()

Clear PETSc global options database.