amrex::HypreNodeLap Class Reference

IJ-based nodal Laplacian that mirrors MLNodeLinOp's discretization. More...

#include <AMReX_HypreNodeLap.H>

Public Types
using	Int = HYPRE_Int
using	AtomicInt = std::conditional_t< sizeof(Int)==4, unsigned int, unsigned long long >

Public Member Functions
	HypreNodeLap (const BoxArray &grids_, const DistributionMapping &dmap_, const Geometry &geom_, const FabFactory< FArrayBox > &factory_, const iMultiFab &owner_mask_, const iMultiFab &dirichlet_mask_, MPI_Comm comm_, MLNodeLinOp const *linop_, int verbose_, std::string options_namespace_)
	Construct an IJ-based nodal Laplacian mirroring MLNodeLinOp.
	~HypreNodeLap ()
	HypreNodeLap (HypreNodeLap const &)=delete
	HypreNodeLap (HypreNodeLap &&)=delete
HypreNodeLap &	operator= (HypreNodeLap const &)=delete
HypreNodeLap &	operator= (HypreNodeLap &&)=delete
void	solve (MultiFab &soln, const MultiFab &rhs, Real rel_tol, Real abs_tol, int max_iter)
	Solve the nodal Laplacian using HYPRE's IJ interface.
void	setHypreOptionsNamespace (const std::string &ns)
	Override the ParmParse namespace used when configuring HYPRE (ns).
Int	fill_local_node_id ()
	Assign local ids to each owned node.
Int	fill_local_node_id_gpu ()
	GPU-specific local-id fill routine; returns the owned-node count.
Int	fill_local_node_id_cpu ()
	CPU fallback local-id fill routine; returns the owned-node count.
void	fill_global_node_id ()
	Convert local ids to globally unique ids recognized by HYPRE.
void	loadVectors (MultiFab &soln, const MultiFab &rhs)
	Copy AMReX solution guess soln and RHS rhs into HYPRE's IJ vectors prior to solving.
void	getSolution (MultiFab &soln)
	Copy the solved data soln out of HYPRE and back into AMReX storage.

Static Public Member Functions
static void	adjust_singular_matrix (Int const *ncols, Int const *cols, Int const *rows, Real *mat)
	Apply the fix-up HYPRE expects for singular matrices (Dirichlet pinning).

Detailed Description

IJ-based nodal Laplacian that mirrors MLNodeLinOp's discretization.

Member Typedef Documentation

AtomicInt

using amrex::HypreNodeLap::AtomicInt = std::conditional_t<sizeof(Int) == 4, unsigned int, unsigned long long>

Int

using amrex::HypreNodeLap::Int = HYPRE_Int

Constructor & Destructor Documentation

HypreNodeLap() [1/3]

amrex::HypreNodeLap::HypreNodeLap	(	const BoxArray &	grids_,
		const DistributionMapping &	dmap_,
		const Geometry &	geom_,
		const FabFactory< FArrayBox > &	factory_,
		const iMultiFab &	owner_mask_,
		const iMultiFab &	dirichlet_mask_,
		MPI_Comm	comm_,
		MLNodeLinOp const *	linop_,
		int	verbose_,
		std::string	options_namespace_
	)

Construct an IJ-based nodal Laplacian mirroring MLNodeLinOp.

Parameters

grids_	Cell-centered grids that define the nodal mesh.
dmap_	Distribution mapping for grids_.
geom_	Geometry describing the level.
factory_	Fab factory referenced when creating workspace.
owner_mask_	Ownership mask used to deduplicate shared nodes.
dirichlet_mask_	Mask describing Dirichlet nodes.
comm_	MPI communicator supplied to HYPRE.
linop_	MLNodeLinOp used as the reference discretization.
verbose_	Verbosity forwarded to HYPRE.
options_namespace_	ParmParse namespace for runtime options.

~HypreNodeLap()

amrex::HypreNodeLap::~HypreNodeLap ( )

default

HypreNodeLap() [2/3]

amrex::HypreNodeLap::HypreNodeLap ( HypreNodeLap const &  )

delete

HypreNodeLap() [3/3]

amrex::HypreNodeLap::HypreNodeLap ( HypreNodeLap &&  )

delete

Member Function Documentation

adjust_singular_matrix()

void amrex::HypreNodeLap::adjust_singular_matrix ( Int const *  ncols, Int const *  cols, Int const *  rows, Real *  mat  )

static

Apply the fix-up HYPRE expects for singular matrices (Dirichlet pinning).

Parameters

ncols	Number of columns in each CSR row.
cols	Column indices.
rows	Row indices.
mat	Matrix entries updated in place.

fill_global_node_id()

void amrex::HypreNodeLap::fill_global_node_id

(

)

Convert local ids to globally unique ids recognized by HYPRE.

fill_local_node_id()

HypreNodeLap::Int amrex::HypreNodeLap::fill_local_node_id

(

)

Assign local ids to each owned node.

Returns

Total owned-node count on this rank.

fill_local_node_id_cpu()

HypreNodeLap::Int amrex::HypreNodeLap::fill_local_node_id_cpu

(

)

CPU fallback local-id fill routine; returns the owned-node count.

fill_local_node_id_gpu()

HypreNodeLap::Int amrex::HypreNodeLap::fill_local_node_id_gpu

(

)

GPU-specific local-id fill routine; returns the owned-node count.

getSolution()

void amrex::HypreNodeLap::getSolution

(

MultiFab &

soln

)

Copy the solved data soln out of HYPRE and back into AMReX storage.

loadVectors()

void amrex::HypreNodeLap::loadVectors	(	MultiFab &	soln,
		const MultiFab &	rhs
	)

Copy AMReX solution guess soln and RHS rhs into HYPRE's IJ vectors prior to solving.

operator=() [1/2]

HypreNodeLap & amrex::HypreNodeLap::operator= ( HypreNodeLap &&  )

delete

operator=() [2/2]

HypreNodeLap & amrex::HypreNodeLap::operator= ( HypreNodeLap const &  )

delete

setHypreOptionsNamespace()

void amrex::HypreNodeLap::setHypreOptionsNamespace ( const std::string &  ns )

inline

Override the ParmParse namespace used when configuring HYPRE (ns).

solve()

void amrex::HypreNodeLap::solve	(	MultiFab &	soln,
		const MultiFab &	rhs,
		Real	rel_tol,
		Real	abs_tol,
		int	max_iter
	)

Solve the nodal Laplacian using HYPRE's IJ interface.

Parameters

soln	Solution MultiFab (nodal) updated on return.
rhs	RHS MultiFab (nodal).
rel_tol	Relative tolerance.
abs_tol	Absolute tolerance.
max_iter	Maximum allowed iterations.

---

The documentation for this class was generated from the following files:

• /home/runner/work/amrex/amrex/Src/Extern/HYPRE/AMReX_HypreNodeLap.H
• /home/runner/work/amrex/amrex/Src/Extern/HYPRE/AMReX_HypreNodeLap.cpp