CPU implementation of the coupled preconditioned conjugate gradient method. More...

Inheritance diagram for cg_cpld::cg_cpld_t:

Collaboration diagram for cg_cpld::cg_cpld_t:

Public Member Functions
procedure, pass(this)	init (this, n, max_iter, m, rel_tol, abs_tol, monitor)
	Initialise a coupled CPU PCG solver.

procedure, pass(this)	free (this)
	Free a coupled CPU PCG solver.

procedure, pass(this)	solve (this, ax, x, f, n, coef, bc_projector, gs_h, niter)
	Reject scalar solves, which are not supported by this type.

procedure, pass(this)	solve_coupled (this, ax, x, y, z, fx, fy, fz, n, coef, bc_projector, gs_h, niter)
	Solve a three-component coupled system with the CPU PCG method.

procedure(ksp_init_intrf), deferred, pass(this)	init (this, n, max_iter, m, rel_tol, abs_tol, monitor)
	Constructor.

procedure, pass(this)	ksp_init (this, max_iter, rel_tol, abs_tol, m, monitor)
	Base type constructor.

procedure, pass(this)	ksp_free (this)
	Base type destructor.

procedure, pass(this)	set_pc (this, m)
	Set preconditioner.

procedure(ksp_method), deferred, pass	solve (this, ax, x, f, n, coef, bc_projector, gs_h, niter)
	Solve the system.

procedure(ksp_method_coupled), deferred, pass	solve_coupled (this, ax, x, y, z, fx, fy, fz, n, coef, bc_projector, gs_h, niter)
	Solve the system (coupled version).

procedure, pass(this)	monitor_start (this, name)
	Monitor start.

procedure, pass(this)	monitor_stop (this)
	Monitor stop.

procedure, pass(this)	monitor_iter (this, iter, rnorm)
	Monitor iteration.

procedure, pass(this)	is_converged (this, iter, residual)
	Check for convergence.

procedure(ksp_t_free), deferred, pass	free (this)
	Destructor.

Public Attributes
real(kind=rp), dimension(:), pointer	w1 => null()
	First component of the operator action \(w = A p\).

real(kind=rp), dimension(:), pointer	w2 => null()
	Second component of the operator action \(w = A p\).

real(kind=rp), dimension(:), pointer	w3 => null()
	Third component of the operator action \(w = A p\).

real(kind=rp), dimension(:), pointer	r1 => null()
	First component of the residual \(r = f - A x\).

real(kind=rp), dimension(:), pointer	r2 => null()
	Second component of the residual \(r = f - A x\).

real(kind=rp), dimension(:), pointer	r3 => null()
	Third component of the residual \(r = f - A x\).

real(kind=rp), dimension(:), pointer	p1 => null()
	First component of the search direction \(p\).

real(kind=rp), dimension(:), pointer	p2 => null()
	Second component of the search direction \(p\).

real(kind=rp), dimension(:), pointer	p3 => null()
	Third component of the search direction \(p\).

real(kind=rp), dimension(:), pointer	z1 => null()
	First component of the preconditioned residual \(z = M^{-1}r\).

real(kind=rp), dimension(:), pointer	z2 => null()
	Second component of the preconditioned residual \(z = M^{-1}r\).

real(kind=rp), dimension(:), pointer	z3 => null()
	Third component of the preconditioned residual \(z = M^{-1}r\).

class(pc_t), pointer	m => null()
	Preconditioner.

real(kind=rp)	rel_tol
	Relative tolerance.

real(kind=rp)	abs_tol
	Absolute tolerance.

integer	max_iter
	Maximum number of iterations.

class(pc_t), allocatable	m_ident
	Internal preconditioner (Identity)

logical	monitor
	Turn on/off monitoring.

Detailed Description

Workspace pointers are associated with host arrays from the scratch registry only for the duration of a solve. The method uses a single Krylov recurrence and combined inner products over all three components.

Definition at line 59 of file cg_coupled.f90.

Member Function/Subroutine Documentation

◆ free() [1/2]

procedure, pass(this) cg_cpld::cg_cpld_t::free ( class(cg_cpld_t), intent(inout) this )

Definition at line 88 of file cg_coupled.f90.

◆ free() [2/2]

procedure(ksp_t_free), deferred, pass krylov::ksp_t::free ( class(ksp_t), intent(inout) this )

pure virtualinherited

Definition at line 103 of file krylov.f90.

◆ init() [1/2]

procedure, pass(this) cg_cpld::cg_cpld_t::init	(	class(cg_cpld_t), intent(inout), target	this,
		integer, intent(in)	n,
		integer, intent(in)	max_iter,
		class(pc_t), intent(in), optional, target	m,
		real(kind=rp), intent(in), optional	rel_tol,
		real(kind=rp), intent(in), optional	abs_tol,
		logical, intent(in), optional	monitor
	)

Parameters

n	Number of degrees of freedom per component.
max_iter	Maximum number of iterations.
M	Optional preconditioner. An identity preconditioner is used if absent.
rel_tol	Optional relative convergence tolerance.
abs_tol	Optional absolute convergence tolerance.
monitor	Optional switch for logging the residual at each iteration.

Definition at line 86 of file cg_coupled.f90.

◆ init() [2/2]

procedure(ksp_init_intrf), deferred, pass(this) krylov::ksp_t::init	(	class(ksp_t), intent(inout), target	this,
		integer, intent(in)	n,
		integer, intent(in)	max_iter,
		class(pc_t), intent(in), optional, target	m,
		real(kind=rp), intent(in), optional	rel_tol,
		real(kind=rp), intent(in), optional	abs_tol,
		logical, intent(in), optional	monitor
	)

pure virtualinherited

Definition at line 83 of file krylov.f90.

◆ is_converged()

procedure, pass(this) krylov::ksp_t::is_converged	(	class(ksp_t), intent(in)	this,
		integer, intent(in)	iter,
		real(kind=rp), intent(in)	residual
	)

inherited

This function checks if the Krylov solver has converged. The solver is considered converged if the residual is less than the absolute tolerance.

Parameters

residual	Residual
iter	Iteration number

Definition at line 101 of file krylov.f90.

◆ ksp_free()

procedure, pass(this) krylov::ksp_t::ksp_free ( class(ksp_t), intent(inout) this )

inherited

Definition at line 87 of file krylov.f90.

◆ ksp_init()

procedure, pass(this) krylov::ksp_t::ksp_init	(	class(ksp_t), intent(inout), target	this,
		integer, intent(in)	max_iter,
		real(kind=rp), intent(in), optional	rel_tol,
		real(kind=rp), intent(in), optional	abs_tol,
		class(pc_t), intent(in), optional, target	m,
		logical, intent(in), optional	monitor
	)

inherited

Parameters

max_iter	Maximum number of iterations.
rel_tol	Relative tolarance for converence.
rel_tol	Absolute tolarance for converence.
M	The preconditioner.

Definition at line 85 of file krylov.f90.

◆ monitor_iter()

procedure, pass(this) krylov::ksp_t::monitor_iter	(	class(ksp_t), intent(in)	this,
		integer, intent(in)	iter,
		real(kind=rp), intent(in)	rnorm
	)

inherited

Definition at line 99 of file krylov.f90.

◆ monitor_start()

procedure, pass(this) krylov::ksp_t::monitor_start	(	class(ksp_t), intent(in)	this,
		character(len=*)	name
	)

inherited

Definition at line 95 of file krylov.f90.

◆ monitor_stop()

procedure, pass(this) krylov::ksp_t::monitor_stop ( class(ksp_t), intent(in) this )

inherited

Definition at line 97 of file krylov.f90.

◆ set_pc()

procedure, pass(this) krylov::ksp_t::set_pc	(	class(ksp_t), intent(inout)	this,
		class(pc_t), intent(in), target	m
	)

inherited

Parameters

M	The preconditioner.

Definition at line 89 of file krylov.f90.

◆ solve() [1/2]

procedure, pass(this) cg_cpld::cg_cpld_t::solve	(	class(cg_cpld_t), intent(inout)	this,
		class(ax_t), intent(in)	ax,
		type(field_t), intent(inout)	x,
		real(kind=rp), dimension(n), intent(in)	f,
		integer, intent(in)	n,
		type(coef_t), intent(inout)	coef,
		class(scalar_bc_projector_t), intent(inout)	bc_projector,
		type(gs_t), intent(inout)	gs_h,
		integer, intent(in), optional	niter
	)

Parameters

Ax	Linear operator.
x	Solution field.
f	Right-hand side.
n	Number of degrees of freedom.
coef	Spectral element coefficients and multiplicity weights.
bc_projector	Projector for Dirichlet boundary nodes.
gs_h	Gather-scatter handle used to assemble the operator result.
niter	Optional maximum number of iterations.

Returns: Unused convergence information.

Definition at line 90 of file cg_coupled.f90.

◆ solve() [2/2]

procedure(ksp_method), deferred, pass krylov::ksp_t::solve	(	class(ksp_t), intent(inout)	this,
		class(ax_t), intent(in)	ax,
		type(field_t), intent(inout)	x,
		real(kind=rp), dimension(n), intent(in)	f,
		integer, intent(in)	n,
		type(coef_t), intent(inout)	coef,
		class(scalar_bc_projector_t), intent(inout)	bc_projector,
		type(gs_t), intent(inout)	gs_h,
		integer, intent(in), optional	niter
	)

pure virtualinherited

Definition at line 91 of file krylov.f90.

◆ solve_coupled() [1/2]

procedure, pass(this) cg_cpld::cg_cpld_t::solve_coupled	(	class(cg_cpld_t), intent(inout)	this,
		class(ax_t), intent(in)	ax,
		type(field_t), intent(inout)	x,
		type(field_t), intent(inout)	y,
		type(field_t), intent(inout)	z,
		real(kind=rp), dimension(n), intent(in)	fx,
		real(kind=rp), dimension(n), intent(in)	fy,
		real(kind=rp), dimension(n), intent(in)	fz,
		integer, intent(in)	n,
		type(coef_t), intent(inout)	coef,
		class(vector_bc_projector_t), intent(inout)	bc_projector,
		type(gs_t), intent(inout)	gs_h,
		integer, intent(in), optional	niter
	)

The initial guesses are discarded. All inner products combine the three components, so the method advances one Krylov recurrence for the complete coupled system.

Parameters

Ax	Coupled linear operator.
x	Solution field for the first component.
y	Solution field for the second component.
z	Solution field for the third component.
fx	Right-hand side for the first component.
fy	Right-hand side for the second component.
fz	Right-hand side for the third component.
n	Number of degrees of freedom per component.
coef	Spectral element coefficients and multiplicity weights.
bc_projector	Projector for vector boundary nodes.
gs_h	Gather-scatter handle used to assemble operator results.
niter	Optional maximum number of iterations, overriding the configured value.

Returns: Identical combined convergence information for all components.

Definition at line 92 of file cg_coupled.f90.

◆ solve_coupled() [2/2]

procedure(ksp_method_coupled), deferred, pass krylov::ksp_t::solve_coupled	(	class(ksp_t), intent(inout)	this,
		class(ax_t), intent(in)	ax,
		type(field_t), intent(inout)	x,
		type(field_t), intent(inout)	y,
		type(field_t), intent(inout)	z,
		real(kind=rp), dimension(n), intent(in)	fx,
		real(kind=rp), dimension(n), intent(in)	fy,
		real(kind=rp), dimension(n), intent(in)	fz,
		integer, intent(in)	n,
		type(coef_t), intent(inout)	coef,
		class(vector_bc_projector_t), intent(inout)	bc_projector,
		type(gs_t), intent(inout)	gs_h,
		integer, intent(in), optional	niter
	)

pure virtualinherited

Definition at line 93 of file krylov.f90.

Member Data Documentation

◆ abs_tol

real(kind=rp) krylov::ksp_t::abs_tol

inherited

Definition at line 77 of file krylov.f90.

◆ m

class(pc_t), pointer krylov::ksp_t::m => null()

inherited

Definition at line 75 of file krylov.f90.

◆ m_ident

class(pc_t), allocatable krylov::ksp_t::m_ident

inherited

Definition at line 79 of file krylov.f90.

◆ max_iter

integer krylov::ksp_t::max_iter

inherited

Definition at line 78 of file krylov.f90.

◆ monitor

logical krylov::ksp_t::monitor

inherited

Definition at line 80 of file krylov.f90.

◆ p1

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::p1 => null()

Definition at line 73 of file cg_coupled.f90.

◆ p2

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::p2 => null()

Definition at line 75 of file cg_coupled.f90.

◆ p3

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::p3 => null()

Definition at line 77 of file cg_coupled.f90.

◆ r1

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::r1 => null()

Definition at line 67 of file cg_coupled.f90.

◆ r2

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::r2 => null()

Definition at line 69 of file cg_coupled.f90.

◆ r3

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::r3 => null()

Definition at line 71 of file cg_coupled.f90.

◆ rel_tol

real(kind=rp) krylov::ksp_t::rel_tol

inherited

Definition at line 76 of file krylov.f90.

◆ w1

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::w1 => null()

Definition at line 61 of file cg_coupled.f90.

◆ w2

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::w2 => null()

Definition at line 63 of file cg_coupled.f90.

◆ w3

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::w3 => null()

Definition at line 65 of file cg_coupled.f90.

◆ z1

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::z1 => null()

Definition at line 79 of file cg_coupled.f90.

◆ z2

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::z2 => null()

Definition at line 81 of file cg_coupled.f90.

◆ z3

real(kind=rp), dimension(:), pointer cg_cpld::cg_cpld_t::z3 => null()

Definition at line 83 of file cg_coupled.f90.

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

/home/runner/work/neko/neko/src/krylov/bcknd/cpu/cg_coupled.f90

Public Member Functions

Public Attributes

Detailed Description

Member Function/Subroutine Documentation

◆ free() [1/2]

◆ free() [2/2]

◆ init() [1/2]

◆ init() [2/2]

◆ is_converged()

◆ ksp_free()

◆ ksp_init()

◆ monitor_iter()

◆ monitor_start()

◆ monitor_stop()

◆ set_pc()

◆ solve() [1/2]

◆ solve() [2/2]

◆ solve_coupled() [1/2]

◆ solve_coupled() [2/2]

Member Data Documentation

◆ abs_tol

◆ m

◆ m_ident

◆ max_iter

◆ monitor

◆ p1

◆ p2

◆ p3

◆ r1

◆ r2

◆ r3

◆ rel_tol

◆ w1

◆ w2

◆ w3

◆ z1

◆ z2

◆ z3