oft_petsc_solvers Module Reference

Detailed Description

PETSc solver implementations.

Wrappers for PETSc solvers implementing the common oft_solver interface.

• Conjugate-Gradient
• Generalized Minimum Residual
• LU Factorization (SUPERLU, MUMPS, etc.)

Wrappers for PETSc preconditioners implementing the common oft_solver interface.

• Diagonal scaling
• Additive Schwarz
• Block Jacobi
• Multi-Grid
• LU Factorization (SUPERLU, MUMPS, etc.)

Authors

Chris Hansen

Date

January 2013

Data Types
type	oft_petsc_asprecond
	PETSc Additive-Schwarz preconditioner. More...
type	oft_petsc_cg_solver
	PETSc Conjugate-Gradient solver. More...
type	oft_petsc_diagprecond
	PETSc diagonal preconditioning. More...
type	oft_petsc_direct_solver
	PETSc direct solver (LU factorization) More...
type	oft_petsc_factordef
	PETSc factorization package information. More...
type	oft_petsc_gmres_solver
	PETSc GMRES solver. More...
type	oft_petsc_jacobi_solver
	PETSc Point-Jacobi solver. More...
type	oft_petsc_luprecond
	PETSc LU factorization/solve. More...
type	oft_petsc_mgprecond
	PETSc multi-grid preconditioner. More...
type	oft_petsc_precond
	PETSc abstract preconditioner class. More...
type	oft_petsc_sjacobi_solver
	PETSc symmetric Point-Jacobi solver. More...
type	oft_petsc_solver
	PETSc abstract solver class. More...

Functions/Subroutines
subroutine	asprecond_setup (self, pc, dist)
	Setup additiv Schwarz preconditioner.
subroutine	asprecond_setup_xml (self, solver_node, level)
	Setup additive Schwarz preconditioner from XML definition.
subroutine	cg_setup_ksp (self, ksp)
	Setup PETSc ksp object for CG.
subroutine	cg_setup_xml (self, solver_node, level)
	Setup CG solver from XML definition.
subroutine	cg_solver_apply (self, u, g)
	Solve a linear system using PETSc's implementation of the Conjugate-Gradient method.
subroutine	cg_view (self)
	Print solver configuration.
subroutine	create_lu_pc (self, pc)
	Create PETSc LU PC object.
subroutine	diagprecond_setup (self, pc, dist)
	Setup diagonal preconditioner.
subroutine	diagprecond_view (self)
	Print solver configuration.
subroutine	direct_setup_ksp (self, ksp)
	Setup PETSc ksp object for direct solver.
subroutine	direct_setup_xml (self, solver_node, level)
	Setup direct solver from XML definition.
subroutine	direct_solver_apply (self, u, g)
	Solve a linear system using a direct solver through PETSc (LU factorization)
subroutine	direct_view (self)
	Print solver configuration.
subroutine	gmres_setup_ksp (self, ksp)
	Setup PETSc ksp object for FGMRES.
subroutine	gmres_setup_xml (self, solver_node, level)
	Setup FGMRES solver from XML definition.
subroutine	gmres_solver_apply (self, u, g)
	Solve a linear system using PETSc's implementation of the FGMRES method.
subroutine	gmres_view (self)
	Print solver configuration.
subroutine	jacobi_solver_apply (self, u, g)
	Solve a linear system using PETSc's implementation of the Point-Jacobi method.
subroutine	lu_pc_load_xml (self, solver_node, level)
	Setup LU preconditioner object from XML definition.
subroutine	luprecond_setup (self, pc, dist)
	Setup LU factorization preconditioner.
subroutine	mgprecond_setup (self, pc, dist)
	Setup multi-grid preconditioner.
logical function	petsc_cg_solver_cast (self, source)
	Cast a solver object to a oft_petsc_cg_solver.
logical function	petsc_diagprecond_cast (self, source)
	Cast a solver object to a oft_petsc_diagprecond.
logical function	petsc_direct_solver_cast (self, source)
	Cast a solver object to a oft_petsc_direct_solver.
logical function	petsc_gmres_solver_cast (self, source)
	Cast a solver object to a oft_petsc_gmres_solver.
logical function	petsc_jacobi_solver_cast (self, source)
	Cast a solver object to a oft_petsc_jacobi_solver.
logical function	petsc_precond_cast (self, source)
	Cast a solver object to a oft_petsc_precond.
logical function	petsc_sjacobi_solver_cast (self, source)
	Cast a solver object to a oft_petsc_sjacobi_solver.
logical function	petsc_solver_cast (self, source)
	Cast a vector object to a oft_petsc_vector.
subroutine	petsc_solver_delete (self)
	Delete PETSc solver.
subroutine	petsc_solver_setup (self)
	Update solver after changing settings/operators.
subroutine	petsc_solver_setup_ksp (self, ksp)
	Setup PETSc ksp object.
subroutine	petsc_solver_setup_pm (ksp)
	Setup performance monitor for PETSc solver.
recursive subroutine	petsc_solver_update (self, new_pattern)
	Update solver after updating settings/matrix.
subroutine	precond_apply (self, u, g)
	Precondition a linear system using one iteration of a PETSc preconditioner.
subroutine	precond_delete (self)
	Delete PETSc solver.
subroutine	precond_dummy_setup (self, pc, dist)
	Setup preconditioner.
recursive subroutine	precond_update (self, new_pattern)
	Update preconditioner after changing settings/operators.
subroutine	set_solver_package (pc, sol_type, ierr)
	Set factorization package for PETSc PC object.
subroutine	sjacobi_setup_xml (self, solver_node, level)
	Setup symmetric Jacobi solver from XML definition.
subroutine	sjacobi_solver_apply (self, u, g)
	Apply 1-step of a symmetric Jacobi smoother with PETSc matrices.

Function/Subroutine Documentation

asprecond_setup()

subroutine asprecond_setup	(	class(oft_petsc_asprecond), intent(inout)	self,
		type(tpc), intent(inout)	pc,
		logical, intent(in)	dist
	)

Setup additiv Schwarz preconditioner.

Parameters

[in,out]	self	Solver object
[in,out]	pc	PETSc PC object
[in]	dist	Flag for local vs distributed solve?

asprecond_setup_xml()

subroutine asprecond_setup_xml	(	class(oft_petsc_asprecond), intent(inout)	self,
		type(xml_node), intent(in), pointer	solver_node,
		integer(i4), intent(in), optional	level
	)

Setup additive Schwarz preconditioner from XML definition.

Parameters

[in,out]	self	Solver object
[in]	solver_node	XML element containing solver definition
[in]	level	Level in MG hierarchy (optional)

cg_setup_ksp()

subroutine cg_setup_ksp	(	class(oft_petsc_cg_solver), intent(inout)	self,
		type(tksp), intent(inout)	ksp
	)

Setup PETSc ksp object for CG.

Parameters

[in,out]	self	Solver object
[in,out]	ksp	PETSc KSP object

cg_setup_xml()

subroutine cg_setup_xml	(	class(oft_petsc_cg_solver), intent(inout)	self,
		type(xml_node), intent(in), pointer	solver_node,
		integer(i4), intent(in), optional	level
	)

Setup CG solver from XML definition.

Parameters

[in,out]	self	Solver object
[in]	solver_node	XML element containing solver definition
[in]	level	Level in MG hierarchy (optional)

cg_solver_apply()

subroutine cg_solver_apply	(	class(oft_petsc_cg_solver), intent(inout)	self,
		class(oft_vector), intent(inout)	u,
		class(oft_vector), intent(inout)	g
	)

Solve a linear system using PETSc's implementation of the Conjugate-Gradient method.

Parameters

[in,out]	self	Solver object
[in,out]	u	Guess (input), Solution (output)
[in,out]	g	RHS (input), Residual (output)

cg_view()

subroutine cg_view

(

class(oft_petsc_cg_solver), intent(inout)

self

)

Print solver configuration.

Parameters

[in,out]

self

Solver object

create_lu_pc()

subroutine create_lu_pc	(	class(oft_petsc_factordef), intent(inout)	self,
		type(tpc), intent(inout)	pc
	)

Create PETSc LU PC object.

Parameters

[in,out]	self	Solver object
[in,out]	pc	PETSc PC object

diagprecond_setup()

subroutine diagprecond_setup	(	class(oft_petsc_diagprecond), intent(inout)	self,
		type(tpc), intent(inout)	pc,
		logical, intent(in)	dist
	)

Setup diagonal preconditioner.

Parameters

[in,out]	self	Solver object
[in,out]	pc	PETSc PC object
[in]	dist	Flag for local vs distributed solve?

diagprecond_view()

subroutine diagprecond_view

(

class(oft_petsc_diagprecond), intent(inout)

self

)

Print solver configuration.

Parameters

[in,out]

self

Solver object

direct_setup_ksp()

subroutine direct_setup_ksp	(	class(oft_petsc_direct_solver), intent(inout)	self,
		type(tksp), intent(inout)	ksp
	)

Setup PETSc ksp object for direct solver.

Parameters

[in,out]	self	Solver object
[in,out]	ksp	PETSc KSP solver object

direct_setup_xml()

subroutine direct_setup_xml	(	class(oft_petsc_direct_solver), intent(inout)	self,
		type(xml_node), intent(in), pointer	solver_node,
		integer(i4), intent(in), optional	level
	)

Setup direct solver from XML definition.

Parameters

[in,out]	self	Solver object
[in]	solver_node	XML element containing solver definition
[in]	level	Level in MG hierarchy (optional)

direct_solver_apply()

subroutine direct_solver_apply	(	class(oft_petsc_direct_solver), intent(inout)	self,
		class(oft_vector), intent(inout)	u,
		class(oft_vector), intent(inout)	g
	)

Solve a linear system using a direct solver through PETSc (LU factorization)

Parameters

[in,out]	self	Solver object
[in,out]	u	Guess (input), Solution (output)
[in,out]	g	RHS (input), Residual (output)

direct_view()

subroutine direct_view

(

class(oft_petsc_direct_solver), intent(inout)

self

)

Print solver configuration.

Parameters

[in,out]

self

Solver object

gmres_setup_ksp()

subroutine gmres_setup_ksp	(	class(oft_petsc_gmres_solver), intent(inout)	self,
		type(tksp), intent(inout)	ksp
	)

Setup PETSc ksp object for FGMRES.

Parameters

[in,out]	self	Solver object
[in,out]	ksp	PETSc KSP object

gmres_setup_xml()

subroutine gmres_setup_xml	(	class(oft_petsc_gmres_solver), intent(inout)	self,
		type(xml_node), intent(in), pointer	solver_node,
		integer(i4), intent(in), optional	level
	)

Setup FGMRES solver from XML definition.

Parameters

[in,out]	self	Solver object
[in]	solver_node	XML element containing solver definition
[in]	level	Level in MG hierarchy (optional)

gmres_solver_apply()

subroutine gmres_solver_apply	(	class(oft_petsc_gmres_solver), intent(inout)	self,
		class(oft_vector), intent(inout)	u,
		class(oft_vector), intent(inout)	g
	)

Solve a linear system using PETSc's implementation of the FGMRES method.

Parameters

[in,out]	self	Solver object
[in,out]	u	Guess (input), Solution (output)
[in,out]	g	RHS (input), Residual (output)

gmres_view()

subroutine gmres_view

(

class(oft_petsc_gmres_solver), intent(inout)

self

)

Print solver configuration.

Parameters

[in,out]

self

Solver object

jacobi_solver_apply()

subroutine jacobi_solver_apply	(	class(oft_petsc_jacobi_solver), intent(inout)	self,
		class(oft_vector), intent(inout)	u,
		class(oft_vector), intent(inout)	g
	)

Solve a linear system using PETSc's implementation of the Point-Jacobi method.

Parameters

[in,out]	self	Solver object
[in,out]	u	Guess (input), Solution (output)
[in,out]	g	RHS (input), Residual (output)

lu_pc_load_xml()

subroutine lu_pc_load_xml	(	class(oft_petsc_factordef), intent(inout)	self,
		type(xml_node), intent(in), pointer	solver_node,
		integer(i4), intent(in), optional	level
	)

Setup LU preconditioner object from XML definition.

Parameters

[in,out]	self	Solver object
[in]	solver_node	XML element containing solver definition
[in]	level	Level in MG hierarchy (optional)

luprecond_setup()

subroutine luprecond_setup	(	class(oft_petsc_luprecond), intent(inout)	self,
		type(tpc), intent(inout)	pc,
		logical, intent(in)	dist
	)

Setup LU factorization preconditioner.

Parameters

[in,out]	self	Solver object
[in,out]	pc	PETSc PC object
[in]	dist	Flag for local vs distributed solve?

mgprecond_setup()

subroutine mgprecond_setup	(	class(oft_petsc_mgprecond), intent(inout)	self,
		type(tpc), intent(inout)	pc,
		logical, intent(in)	dist
	)

Setup multi-grid preconditioner.

Parameters

[in,out]	self	Solver object
[in,out]	pc	PETSc PC object
[in]	dist	Flag for local vs distributed solve?

petsc_cg_solver_cast()

logical function petsc_cg_solver_cast	(	type(oft_petsc_cg_solver), intent(out), pointer	self,
		class(oft_solver), intent(in), target	source
	)

Cast a solver object to a oft_petsc_cg_solver.

The source solver must be oft_petsc_cg_solver or a child class, otherwise pointer will be returned as null and success == .FALSE.

Parameters

[out]	self	Reference to source object with desired class
[in]	source	Source solver to cast

Returns

Cast success flag

petsc_diagprecond_cast()

logical function petsc_diagprecond_cast	(	type(oft_petsc_diagprecond), intent(out), pointer	self,
		class(oft_solver), intent(in), target	source
	)

Cast a solver object to a oft_petsc_diagprecond.

The source solver must be oft_petsc_diagprecond or a child class, otherwise pointer will be returned as null and success == .FALSE.

Parameters

[out]	self	Reference to source object with desired class
[in]	source	Source solver to cast

Returns

Cast success flag

petsc_direct_solver_cast()

logical function petsc_direct_solver_cast	(	type(oft_petsc_direct_solver), intent(out), pointer	self,
		class(oft_solver), intent(in), target	source
	)

Cast a solver object to a oft_petsc_direct_solver.

The source solver must be oft_petsc_direct_solver or a child class, otherwise pointer will be returned as null and success == .FALSE.

Parameters

[out]	self	Reference to source object with desired class
[in]	source	Source solver to cast

Returns

Cast success flag

petsc_gmres_solver_cast()

logical function petsc_gmres_solver_cast	(	type(oft_petsc_gmres_solver), intent(out), pointer	self,
		class(oft_solver), intent(in), target	source
	)

Cast a solver object to a oft_petsc_gmres_solver.

The source solver must be oft_petsc_gmres_solver or a child class, otherwise pointer will be returned as null and success == .FALSE.

Parameters

[out]	self	Reference to source object with desired class
[in]	source	Source solver to cast

Returns

Cast success flag

petsc_jacobi_solver_cast()

logical function petsc_jacobi_solver_cast	(	type(oft_petsc_jacobi_solver), intent(out), pointer	self,
		class(oft_solver), intent(in), target	source
	)

Cast a solver object to a oft_petsc_jacobi_solver.

The source solver must be oft_petsc_jacobi_solver or a child class, otherwise pointer will be returned as null and success == .FALSE.

Parameters

[out]	self	Reference to source object with desired class
[in]	source	Source solver to cast

Returns

Cast success flag

petsc_precond_cast()

logical function petsc_precond_cast	(	class(oft_petsc_precond), intent(out), pointer	self,
		class(oft_solver), intent(in), target	source
	)

Cast a solver object to a oft_petsc_precond.

The source solver must be oft_petsc_precond or a child class, otherwise pointer will be returned as null and success == .FALSE.

Parameters

[out]	self	Reference to source object with desired class
[in]	source	Source solver to cast

Returns

Cast success flag

petsc_sjacobi_solver_cast()

logical function petsc_sjacobi_solver_cast	(	type(oft_petsc_sjacobi_solver), intent(out), pointer	self,
		class(oft_solver), intent(in), target	source
	)

Cast a solver object to a oft_petsc_sjacobi_solver.

The source solver must be oft_petsc_sjacobi_solver or a child class, otherwise pointer will be returned as null and success == .FALSE.

Parameters

[out]	self	Reference to source object with desired class
[in]	source	Source solver to cast

Returns

Cast success flag

petsc_solver_cast()

logical function petsc_solver_cast	(	class(oft_petsc_solver), intent(out), pointer	self,
		class(oft_solver), intent(in), target	source
	)

Cast a vector object to a oft_petsc_vector.

The source vector must be oft_petsc_vector or a child class, otherwise pointer will be returned as null and success == .FALSE.

Parameters

[out]	self	Reference to source object with desired class
[in]	source	Source solver to cast

Returns

Cast success flag

petsc_solver_delete()

subroutine petsc_solver_delete

(

class(oft_petsc_solver), intent(inout)

self

)

Delete PETSc solver.

Parameters

[in,out]

self

Solver object

petsc_solver_setup()

subroutine petsc_solver_setup

(

class(oft_petsc_solver), intent(inout)

self

)

Update solver after changing settings/operators.

Parameters

[in,out]

self

Solver object

petsc_solver_setup_ksp()

subroutine petsc_solver_setup_ksp	(	class(oft_petsc_solver), intent(inout)	self,
		type(tksp), intent(inout)	ksp
	)

Setup PETSc ksp object.

Parameters

[in,out]

self

Solver object

petsc_solver_setup_pm()

subroutine petsc_solver_setup_pm

(

type(tksp), intent(inout)

ksp

)

Setup performance monitor for PETSc solver.

petsc_solver_update()

recursive subroutine petsc_solver_update	(	class(oft_petsc_solver), intent(inout)	self,
		logical, intent(in), optional	new_pattern
	)

Update solver after updating settings/matrix.

Parameters

[in,out]	self	Solver object
[in]	new_pattern	Update matrix pattern (optional)

precond_apply()

subroutine precond_apply	(	class(oft_petsc_precond), intent(inout)	self,
		class(oft_vector), intent(inout)	u,
		class(oft_vector), intent(inout)	g
	)

Precondition a linear system using one iteration of a PETSc preconditioner.

Parameters

[in,out]	self	Solver object
[in,out]	u	Guess (input), Solution (output)
[in,out]	g	RHS (input), Residual (output)

precond_delete()

subroutine precond_delete

(

class(oft_petsc_precond), intent(inout)

self

)

Delete PETSc solver.

Parameters

[in,out]

self

Solver object

precond_dummy_setup()

subroutine precond_dummy_setup	(	class(oft_petsc_precond), intent(inout)	self,
		type(tpc), intent(inout)	pc,
		logical, intent(in)	dist
	)

Setup preconditioner.

Parameters

[in,out]	self	Solver object
[in,out]	pc	PETSc PC object
[in]	dist	Flag for local vs distributed solve?

precond_update()

recursive subroutine precond_update	(	class(oft_petsc_precond), intent(inout)	self,
		logical, intent(in), optional	new_pattern
	)

Update preconditioner after changing settings/operators.

Parameters

[in,out]	self	Solver object
[in]	new_pattern	Update matrix pattern (optional)

set_solver_package()

subroutine set_solver_package	(	type(tpc), intent(inout)	pc,
		character(len=*), intent(in)	sol_type,
		integer(i4), intent(out)	ierr
	)

Set factorization package for PETSc PC object.

Parameters

[in,out]	pc	PETSc PC object
[in]	sol_type	Solver type
[out]	ierr	Error flag

sjacobi_setup_xml()

subroutine sjacobi_setup_xml	(	class(oft_petsc_sjacobi_solver), intent(inout)	self,
		type(xml_node), intent(in), pointer	solver_node,
		integer(i4), intent(in), optional	level
	)

Setup symmetric Jacobi solver from XML definition.

Parameters

[in,out]	self	Solver object
[in]	solver_node	XML element containing solver definition
[in]	level	Level in MG hierarchy (optional)

sjacobi_solver_apply()

subroutine sjacobi_solver_apply	(	class(oft_petsc_sjacobi_solver), intent(inout)	self,
		class(oft_vector), intent(inout)	u,
		class(oft_vector), intent(inout)	g
	)

Apply 1-step of a symmetric Jacobi smoother with PETSc matrices.

Parameters

[in,out]	self	Solver object
[in,out]	u	Guess (input), Solution (output)
[in,out]	g	RHS (input), Residual (output)