EBHelmholtzOp Class Reference

Helmholtz operator for equations like alpha*a(x)*phi(x) + beta*div(b(x)*grad(phi(x))) = rho. More...

#include <CD_EBHelmholtzOp.H>

Inheritance diagram for EBHelmholtzOp:

Collaboration diagram for EBHelmholtzOp:

Classes
struct	DomainBndryBoxes
	Cached domain-boundary boxes for a single grid patch. More...

Public Types
enum class	Smoother {   NoRelax , PointJacobi , GauSaiRedBlack , GauSaiMultiColor ,   Chebyshev , RestrictedAdditiveSchwarz }
	Relaxation method for the operators.

Public Member Functions
	EBHelmholtzOp ()=delete
	Disallowed default constructor.
	EBHelmholtzOp (const EBHelmholtzOp &a_other)=delete
	Disallowed copy constructor.
	EBHelmholtzOp (const EBHelmholtzOp &&a_other)=delete
	Disallowed move constructor.
	EBHelmholtzOp (const Location::Cell a_dataLocation, const EBLevelGrid &a_eblgFine, const EBLevelGrid &a_eblg, const EBLevelGrid &a_eblgCoFi, const EBLevelGrid &a_eblgCoar, const EBLevelGrid &a_eblgCoarMG, const RefCountedPtr< LevelData< BaseFab< bool > > > &a_validCells, const RefCountedPtr< EBMultigridInterpolator > &a_interpolator, const RefCountedPtr< EBReflux > &a_fluxReg, const RefCountedPtr< EBCoarAve > &a_coarAve, const RefCountedPtr< EBHelmholtzDomainBC > &a_domainBC, const RefCountedPtr< EBHelmholtzEBBC > &a_ebBC, const RealVect &a_probLo, const Real &a_dx, const int &a_refToFine, const int &a_refToCoar, const bool &a_hasFine, const bool &a_hasCoar, const bool &a_hasMGObjects, const Real &a_alpha, const Real &a_beta, const RefCountedPtr< LevelData< EBCellFAB > > &a_Acoef, const RefCountedPtr< LevelData< EBFluxFAB > > &a_Bcoef, const RefCountedPtr< LevelData< BaseIVFAB< Real > > > &a_BcoefIrreg, const IntVect &a_ghostPhi, const IntVect &a_ghostRhs, const Smoother &a_smoother, const Real &a_relaxFactor, const int &a_chebyOrder, const Real &a_chebyEigRatio, const int &a_rasInnerSweeps, const bool a_refluxFree=false)
	Full constructor.
virtual	~EBHelmholtzOp ()
	Dtor.
EBHelmholtzOp &	operator= (const EBHelmholtzOp &a_oper)=delete
	No copy assignment allowed.
EBHelmholtzOp &	operator= (const EBHelmholtzOp &&a_oper)=delete
	No move assignment allowed.
void	turnOffCFInterp ()
	Turn off BCs.
void	turnOnCFInterp ()
	Turn on BCs.
void	turnOffExchange ()
	Turn off exchange operation.
void	turnOnExchange ()
	Turn on exchange operation.
void	turnOffCoarsening ()
	Turn off coarsening operation.
void	turnOnCoarsening ()
	Turn on coarsening operation.
void	setAcoAndBco (const RefCountedPtr< LevelData< EBCellFAB > > &a_Acoef, const RefCountedPtr< LevelData< EBFluxFAB > > &a_Bcoef, const RefCountedPtr< LevelData< BaseIVFAB< Real > > > &a_BcoefIrreg)
	Update with new A and B coefficients.
const RefCountedPtr< LevelData< EBCellFAB > > &	getAcoef ()
	Get the Helmholtz A-coefficient on cell centers.
const RefCountedPtr< LevelData< EBFluxFAB > > &	getBcoef ()
	Get the Helmholtz B-coefficient on faces.
const RefCountedPtr< LevelData< BaseIVFAB< Real > > > &	getBcoefIrreg ()
	Get the Helmholtz B-coefficient on the EB.
void	coarsenCell (LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_phiFine)
	Coarsen data from fine to coar level.
void	coarsenFlux (LevelData< EBFluxFAB > &a_flux, const LevelData< EBFluxFAB > &a_fineFlux)
	Coarsen fluxes on the fine level onto this level.
void	pointJacobiKernel (EBCellFAB &a_Lcorr, EBCellFAB &a_corr, const EBCellFAB &a_resid, const EBCellFAB &a_Acoef, const EBFluxFAB &a_Bcoef, const BaseIVFAB< Real > &a_BcoefIrreg, const Box &a_cellBox, const DataIndex &a_dit) const noexcept
	Point Jacobi kernel.
void	chebyshevKernel (EBCellFAB &a_Lcorr, EBCellFAB &a_corr, const EBCellFAB &a_resid, const EBCellFAB &a_Acoef, const EBFluxFAB &a_Bcoef, const BaseIVFAB< Real > &a_BcoefIrreg, const Box &a_cellBox, const DataIndex &a_dit, const Real a_omega) const noexcept
	Chebyshev smoother kernel.
void	gauSaiRedBlackKernel (EBCellFAB &a_Lcorr, EBCellFAB &a_corr, const EBCellFAB &a_resid, const EBCellFAB &a_Acoef, const EBFluxFAB &a_Bcoef, const BaseIVFAB< Real > &a_BcoefIrreg, const Box &a_cellBox, const DataIndex &a_dit, const int &a_redBlack) const noexcept
	Red-black Gauss-Seidel kernel.
void	gauSaiMultiColorKernel (EBCellFAB &a_Lcorr, EBCellFAB &a_corr, const EBCellFAB &a_resid, const EBCellFAB &a_Acoef, const EBFluxFAB &a_Bcoef, const BaseIVFAB< Real > &a_BcoefIrreg, const Box &a_cellBox, const DataIndex &a_dit, const IntVect &a_color) const noexcept
	Multi-color Gauss-Seidel kernel.
void	residual (LevelData< EBCellFAB > &a_residual, const LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_rhs, const bool a_homogeneousPhysBc) override
	Compute residual on this level.
void	preCond (LevelData< EBCellFAB > &a_corr, const LevelData< EBCellFAB > &a_residual) override final
	Precondition system before bottom solve.
void	interpolateCF (LevelData< EBCellFAB > &a_phiFine, const LevelData< EBCellFAB > *a_phiCoar, const bool a_homogeneousCFBC)
	Apply coarse-fine boundary conditions.
void	homogeneousCFInterp (LevelData< EBCellFAB > &a_phi)
	Do homogeneous coarse-fine interpolation.
void	inhomogeneousCFInterp (LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_phiCoar)
	Inhomogeneous coarse-fine interpolation.
void	gauSaiMultiColor (LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_rhs, const IntVect &a_color) const
	Multi-colored Gauss-Seidel kernel. Public because MFHelmholtzOp may want to use use.
void	applyOp (LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > *const a_phiCoar, const bool a_homogeneousPhysBC, const bool a_homogeneousCFBC)
	Apply operator on this level. This is a more general version which can turn on/off homogeneous and CF bcs.
void	applyOp (LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_phi, bool a_homogeneousPhysBc) override final
	Apply operator.
void	applyOp (EBCellFAB &a_Lphi, EBCellFAB &a_phi, const EBCellFAB &a_Acoef, const EBFluxFAB &a_Bcoef, const BaseIVFAB< Real > &a_BcoefIrreg, const Box &a_cellBox, const DataIndex &a_dit, const bool a_homogeneousPhysBC) const noexcept
	Apply operator in a grid box.
void	applyOpRegular (EBCellFAB &a_Lphi, EBCellFAB &a_phi, const EBCellFAB &a_Acoef, const EBFluxFAB &a_Bcoef, const BaseIVFAB< Real > &a_BcoefIrreg, const Box &a_cellBox, const DataIndex &a_dit, const bool a_homogeneousPhysBC) const noexcept
	Apply operator in regular cells.
void	applyDomainFlux (EBCellFAB &a_phi, const EBFluxFAB &a_Bcoef, const Box &a_cellBox, const DataIndex &a_dit, const bool a_homogeneousPhysBc) const noexcept
	Apply domain flux.
void	fillDomainFlux (EBFluxFAB &a_flux, const EBCellFAB &a_phi, const Box &a_cellBox, const DataIndex &a_dit, const bool a_homogeneousPhysBC)
	Fill domain flux. This fills the flux on the domain face using centered differencing ala applyDomainFlux.
void	applyOpIrregular (EBCellFAB &a_Lphi, const EBCellFAB &a_phi, const EBCellFAB &a_Acoef, const EBFluxFAB &a_Bcoef, const BaseIVFAB< Real > &a_BcoefIrreg, const BaseIVFAB< Real > &a_alphaDiagWeight, const Box &a_cellBox, const DataIndex &a_dit, const bool a_homogeneousPhysBC) const noexcept
	Apply operator in irregular cells.
void	create (LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs) override final
	Create data which clones the layout of the other.
void	assign (LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs) override final
	Assign data.
void	assignCopier (LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs, const Copier &a_copier) override final
	Assign lhs.
void	assignLocal (LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs) override final
	Local assignment function.
void	buildCopier (Copier &a_copier, const LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs) override
	Build copier.
Real	dotProduct (const LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs) override final
	Compute the dot product??
void	dotProductMaskedLocal (Real &a_sumKappaXY, Real &a_sumVolume, const LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs, const LevelData< BaseFab< bool > > &a_mask, const bool a_needVolume) const noexcept
	Rank-local (un-reduced) partials for the masked AMR-composite Krylov inner product.
void	incr (LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs, const Real a_scale) override final
	Increment operator.
void	axby (LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_x, const LevelData< EBCellFAB > &a_y, const Real a_a, const Real a_b) override final
	Set a_lhs = a*x + b*y.
void	scale (LevelData< EBCellFAB > &a_lhs, const Real &a_scale) override final
	Scale data. Returns a_lhs = a_lhs*a_scale.
Real	norm (const LevelData< EBCellFAB > &a_rhs, const int a_order) override final
	Compute norm of data.
void	setToZero (LevelData< EBCellFAB > &a_lhs) override final
	Set data to zero.
void	createCoarser (LevelData< EBCellFAB > &a_coarse, const LevelData< EBCellFAB > &a_fine, bool a_ghosted) override final
	Create coarsened data.
void	relax (LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_residual, int a_iterations) override final
	Relaxation method. This does smoothing for the system L(correction) = residual.
void	restrictResidual (LevelData< EBCellFAB > &a_resCoar, LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_rhs) override final
	Restrict residual onto coarse level.
void	prolongIncrement (LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_correctCoarse) override final
	Prolongation method.
int	refToCoarser () override final
	Return coarsening factor to coarser level (1 if there is no coarser level);.
void	AMROperator (LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_phiFine, const LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_phiCoar, const bool a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *a_finerOp) override final
	Apply the AMR operator, i.e. compute L(phi) in an AMR context.
void	refluxFreeAMROperator (LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_phiFine, const LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_phiCoar, const bool a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *a_finerOp)
	Apply the AMR operator, i.e. compute L(phi) in an AMR context.
void	AMRResidual (LevelData< EBCellFAB > &a_residual, const LevelData< EBCellFAB > &a_phiFine, const LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_phiCoar, const LevelData< EBCellFAB > &a_rhs, bool a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *a_finerOp) override final
	Compute residual on this level. AMR version.
void	AMRResidualNF (LevelData< EBCellFAB > &a_residual, const LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_phiCoar, const LevelData< EBCellFAB > &a_rhs, bool a_homogeneousPhysBC) override final
	Compute AMR residual on finest AMR level.
void	AMRResidualNC (LevelData< EBCellFAB > &a_residual, const LevelData< EBCellFAB > &a_phiFine, const LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_rhs, bool a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *a_finerOp) override final
	Compute AMR residual on coarsest.
void	AMROperatorNF (LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_phi, const LevelData< EBCellFAB > &a_phiCoar, bool a_homogeneousPhysBC) override final
	Apply the AMR operator, i.e. compute L(phi) in an AMR context, assuming no finer levels.
void	AMROperatorNC (LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_phiFine, const LevelData< EBCellFAB > &a_phi, bool a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *a_finerOp) override final
	Apply the AMR operator, i.e. compute L(phi) in an AMR context, assuming no coarser AMR levels.
void	AMRRestrict (LevelData< EBCellFAB > &a_residualCoarse, const LevelData< EBCellFAB > &a_residual, const LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_coarseCorrection, bool a_skip_res) override final
	Restrict residual.
void	AMRProlong (LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_coarseCorrection) override final
	Prolongation onto AMR level.
void	AMRUpdateResidual (LevelData< EBCellFAB > &a_residual, const LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_coarseCorrection) override final
	Update AMR residual.
void	setAlphaAndBeta (const Real &a_alpha, const Real &a_beta) override final
	Set alpha coefficient and beta coefficient (can change as diffusion solvers progress)
void	createCoarsened (LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs, const int &a_refRat) override final
	Create coarsening of data holder.
void	diagonalScale (LevelData< EBCellFAB > &a_rhs, bool a_kappaWeighted) override final
	Do diagonal scaling.
void	divideByIdentityCoef (LevelData< EBCellFAB > &a_rhs) override final
	Divide by the a-coefficient.
void	applyOpNoBoundary (LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_phi) override final
	Apply operator but turn off all BCs.
void	fillGrad (const LevelData< EBCellFAB > &a_phi) override final
	Not called, I think.
void	getFlux (EBFluxFAB &a_flux, const LevelData< EBCellFAB > &a_data, const Box &a_grid, const DataIndex &a_dit, Real a_scale) override final
	Fill flux.
void	allocateFlux () const noexcept
	Allocate m_flux.
void	deallocateFlux () const noexcept
	Deallocate m_flux.
LevelData< EBFluxFAB > &	getFlux () const
	Returns m_flux. This is used in refluxing routines.
const LevelData< EBCellFAB > &	getRelaxationCoeff () const noexcept
	Returns m_relCoef. Used by MFHelmholtzOp.
int	getChebyOrder () const noexcept
	Returns m_chebyOrder. Used by MFHelmholtzOp to replicate Chebyshev parameters.
Real	getChebyEigRatio () const noexcept
	Returns m_chebyEigRatio. Used by MFHelmholtzOp to replicate Chebyshev parameters.
Real	getSpectralRadius () const noexcept
	Returns m_spectralRadius. Used by MFHelmholtzOp to replicate Chebyshev parameters.
int	getRasInnerSweeps () const noexcept
	Returns m_rasInnerSweeps. Used by MFHelmholtzOp to replicate the Schwarz block solve.
LayoutData< VoFIterator > &	getVofIterIrreg () const noexcept
	Return all-cut VoFIterators (one per patch). Used by MFHelmholtzOp.

Static Public Member Functions
static void	scaleLocal (LevelData< EBCellFAB > &a_lhs, const LevelData< EBCellFAB > &a_rhs) noexcept
	Local scaling function. Multiplies the left-hand side by the right-hand side.

Protected Member Functions
void	relaxPointJacobi (LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_residual, const int a_iterations)
	Jacobi relaxation.
void	relaxGSRedBlack (LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_residual, const int a_iterations)
	Jacobi relaxation.
void	relaxGSMultiColor (LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_residual, const int a_iterations)
	Multi-colored gauss-seidel relaxation.
void	relaxChebyshev (LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_residual, const int a_iterations)
	Chebyshev polynomial smoother.
void	relaxRestrictedAdditiveSchwarz (LevelData< EBCellFAB > &a_correction, const LevelData< EBCellFAB > &a_residual, const int a_iterations)
	Restricted additive Schwarz smoother (block smoother).
void	computeDiagWeight ()
	Calculate the weight of the diagonal term.
void	computeRelaxationCoefficient ()
	Calculate relaxation coefficient.
void	computeSpectralRadius ()
	Estimate the spectral radius of D^{-1}A (Jacobi-preconditioned operator) via the Gershgorin bound.
void	makeAggStencil ()
	Compute aggregated stencils.
void	defineStencils ()
	Define stencils.
VoFStencil	getFaceCenterFluxStencil (const FaceIndex &a_face, const DataIndex &a_dit) const
	Get the face-centered flux stencil.
VoFStencil	getFaceCentroidFluxStencil (const FaceIndex &a_face, const DataIndex &a_dit) const
	Get the face-centroid flux stencil.
void	computeFlux (EBFaceFAB &a_fluxCentroid, const EBCellFAB &a_phi, const Box &a_cellBox, const DataIndex &a_dit, const int a_dir)
	Compute centroid fluxes in a direction.
void	computeFaceCenteredFlux (EBFaceFAB &a_fluxCenter, const EBCellFAB &a_phi, const Box &a_cellBox, const DataIndex &a_dit, const int a_dir)
	Compute face-centered fluxes.
void	computeFaceCentroidFlux (EBFaceFAB &a_flux, const EBCellFAB &a_phi, const Box &a_cellBox, const DataIndex &a_dit, const int a_dir)
	Compute face-centroid fluxes.
void	computeFlux (const LevelData< EBCellFAB > &a_phi)
	Fill face centroid-centered fluxes on this level.
void	reflux (LevelData< EBCellFAB > &a_Lphi, const LevelData< EBCellFAB > &a_phiFine, const LevelData< EBCellFAB > &a_phi, AMRLevelOp< LevelData< EBCellFAB > > &a_finerOp)
	Reflux algorithm.

Protected Attributes
Timer	m_timer
	Timer so user can profile.
Interval	m_interval
	Interval.
Smoother	m_smoother
	Relaxation method.
Location::Cell	m_dataLocation
	Data centering.
bool	m_refluxFree
	Use reflux-free formulation or not.
bool	m_profile
	Profile the operator.
bool	m_hasMGObjects
	True if there is a multigrid level below this operator.
bool	m_hasFine
	True if there's a finer level.
bool	m_hasCoar
	True if there's a coarser level.
int	m_refToCoar
	Refinement factor to coarse level.
int	m_refToFine
	Refinement factor to fine level.
int	m_numSmoothPreCond
	Number of smoothings in the preconditioner.
bool	m_doInterpCF
	Do coarse-fine interpolation or not.
bool	m_doExchange
	Turn on/off exchange operation.
bool	m_doCoarsen
	Turn on/off exchange operation.
IntVect	m_ghostPhi
	Ghost cells for phi.
IntVect	m_ghostRhs
	Ghost cells for rhs (note, the operator rhs)
Real	m_alpha
	Alpha-coefficient.
Real	m_beta
	Beta-coefficient.
Real	m_dx
	Grid resolution;.
Real	m_relaxFactor
	Successive over-relaxation factor (Must be >= 1).
int	m_chebyOrder
	Number of Chebyshev polynomial steps per smoother invocation.
Real	m_chebyEigRatio
	Ratio lambda_max/lambda_min for the Chebyshev eigenvalue window.
Real	m_spectralRadius
	Spectral radius estimate for the Jacobi-preconditioned operator D^{-1}A.
int	m_rasInnerSweeps
	Number of inner (frozen-ghost) red-black sweeps per outer restricted-additive-Schwarz iteration.
RealVect	m_vecDx
	Vector resolution.
RealVect	m_probLo
	Lower-left corner of domain.
Copier	m_exchangeCopier
	Pre-built exchange copier.
Copier	m_exchangeCopierFine
	Pre-built exchange copier.
EBLevelGrid	m_eblgFine
	Fine level grid (if the operator has a fine level)
EBLevelGrid	m_eblg
	Grid.
EBLevelGrid	m_eblgCoFi
	Coarsened of m_eblg.
EBLevelGrid	m_eblgCoar
	Coarse level grid (if the operator has a coarse level)
EBLevelGrid	m_eblgCoarMG
	Coarser grids (multigrid level)
EBMGRestrict	m_restrictOp
	Restriction operator for AMR levels.
EBMGRestrict	m_restrictOpMG
	Restriction operator if this is an MG level.
EBMGProlong	m_prolongOp
	Prolongation operator for AMR levels.
EBMGProlong	m_prolongOpMG
	Prolongation operator for MG levels.
RefCountedPtr< LevelData< BaseFab< bool > > >	m_validCells
	Valid grid cells (might be nullptr on MG levels)
RefCountedPtr< EBHelmholtzDomainBC >	m_domainBc
	Domain bc object.
RefCountedPtr< EBHelmholtzEBBC >	m_ebBc
	Domain bc object.
RefCountedPtr< EBMultigridInterpolator >	m_interpolator
	Interpolator object.
RefCountedPtr< EBReflux >	m_fluxReg
	Flux register.
RefCountedPtr< EBCoarAve >	m_coarAve
	Conservative coarsener.
RefCountedPtr< LevelData< EBCellFAB > >	m_Acoef
	A-coefficient in Helmholtz equation.
RefCountedPtr< LevelData< EBFluxFAB > >	m_Bcoef
	B-coefficient in Helmholtz equation.
RefCountedPtr< LevelData< BaseIVFAB< Real > > >	m_BcoefIrreg
	B-coefficient in Helmholtz equation, but on EB faces.
LevelData< EBCellFAB >	m_relCoef
	Relaxation coefficient.
LevelData< EBFluxFAB > *	m_flux
	For holding fluxes.
LayoutData< BaseIFFAB< Real > >	m_interpolant [SpaceDim]
	Interpolant for when we want centroid fluxes.
LayoutData< BaseIFFAB< FaceStencil > >	m_interpStencil [SpaceDim]
	Face centroid interpolation stencil.
LayoutData< BaseIFFAB< VoFStencil > >	m_centroidFluxStencil [SpaceDim]
	Face centroid flux stencil. Defined on all faces connecting one or more irregular vofs.
LayoutData< BaseIVFAB< VoFStencil > >	m_relaxStencils
	Operator stencils in irregular cells (and ones that border irregular cells if using a centroid discretization).
LayoutData< RefCountedPtr< VCAggStencil > >	m_aggRelaxStencil
	For making irregular stencil applications go faster.
LayoutData< BaseIVFAB< Real > >	m_alphaDiagWeight
	Weights of diagonal alpha terms.
LayoutData< BaseIVFAB< Real > >	m_betaDiagWeight
	Weights of diagonal beta terms.
LayoutData< VoFIterator >	m_vofIterIrreg
	VoFIterator for irregular cells.
LayoutData< VoFIterator >	m_vofIterMulti
	VoFIterator for "multi-cells".
LayoutData< VoFIterator >	m_vofIterStenc
	VoFIterator which iterates over all cells that are 1) a cut-cell or 2) borders a cut-cell.
LayoutData< VoFIterator >	m_vofIterDomLo [SpaceDim]
	VoF iterators for lo domain side.
LayoutData< VoFIterator >	m_vofIterDomHi [SpaceDim]
	VoF iterators for hi domain side.
std::map< std::pair< int, Side::LoHiSide >, Box >	m_sideBox
	Domain boxes on each side.
LayoutData< DomainBndryBoxes >	m_domainBndryBoxes
	Precomputed domain-boundary boxes for each grid patch.
Vector< IntVect >	m_colors
	"Colors" for the relaxation methods

Static Protected Attributes
static constexpr int	m_nComp = 1
	Number of components that we solve for (always one..)
static constexpr int	m_comp = 0
	Component that we solve for.

Detailed Description

Helmholtz operator for equations like alpha*a(x)*phi(x) + beta*div(b(x)*grad(phi(x))) = rho.

This can be used with TGA time stepping.

Constructor & Destructor Documentation

EBHelmholtzOp() [1/3]

EBHelmholtzOp::EBHelmholtzOp ( const EBHelmholtzOp &  a_other )

delete

Disallowed copy constructor.

Parameters

[in]

a_other

Other operator

EBHelmholtzOp() [2/3]

EBHelmholtzOp::EBHelmholtzOp ( const EBHelmholtzOp &&  a_other )

delete

Disallowed move constructor.

Parameters

[in]

a_other

Other operator

EBHelmholtzOp() [3/3]

EBHelmholtzOp::EBHelmholtzOp	(	const Location::Cell	a_dataLocation,
		const EBLevelGrid &	a_eblgFine,
		const EBLevelGrid &	a_eblg,
		const EBLevelGrid &	a_eblgCoFi,
		const EBLevelGrid &	a_eblgCoar,
		const EBLevelGrid &	a_eblgCoarMG,
		const RefCountedPtr< LevelData< BaseFab< bool > > > &	a_validCells,
		const RefCountedPtr< EBMultigridInterpolator > &	a_interpolator,
		const RefCountedPtr< EBReflux > &	a_fluxReg,
		const RefCountedPtr< EBCoarAve > &	a_coarAve,
		const RefCountedPtr< EBHelmholtzDomainBC > &	a_domainBC,
		const RefCountedPtr< EBHelmholtzEBBC > &	a_ebBC,
		const RealVect &	a_probLo,
		const Real &	a_dx,
		const int &	a_refToFine,
		const int &	a_refToCoar,
		const bool &	a_hasFine,
		const bool &	a_hasCoar,
		const bool &	a_hasMGObjects,
		const Real &	a_alpha,
		const Real &	a_beta,
		const RefCountedPtr< LevelData< EBCellFAB > > &	a_Acoef,
		const RefCountedPtr< LevelData< EBFluxFAB > > &	a_Bcoef,
		const RefCountedPtr< LevelData< BaseIVFAB< Real > > > &	a_BcoefIrreg,
		const IntVect &	a_ghostPhi,
		const IntVect &	a_ghostRhs,
		const Smoother &	a_smoother,
		const Real &	a_relaxFactor,
		const int &	a_chebyOrder,
		const Real &	a_chebyEigRatio,
		const int &	a_rasInnerSweeps,
		const bool	a_refluxFree = false
	)

Full constructor.

Parameters

[in]	a_dataLocation	Data location, either cell center or cell centroid
[in]	a_eblgFine	Fine grids
[in]	a_eblg	Grids on this level
[in]	a_eblgCoFi	Coarsening of fine level grids
[in]	a_eblgCoar	Coarse grids
[in]	a_eblgCoarMG	Multigrid-grids
[in]	a_validCells	Valid grid cells
[in]	a_interpolator	Interpolator
[in]	a_fluxReg	Flux register
[in]	a_coarAve	Coarsener
[in]	a_domainBC	Domain BC
[in]	a_ebBC	Boundary conditions on EBs
[in]	a_probLo	Lower-left corner of computational domain
[in]	a_dx	Grid resolution
[in]	a_refToFine	Refinement ratio to fine level
[in]	a_refToCoar	Refinement ratio to coarse level
[in]	a_hasFine	Has fine level or not
[in]	a_hasCoar	Has coarse level or not
[in]	a_hasMGObjects	Has multigrid-objects (special objects between AMR levels, or below the AMR levels)
[in]	a_alpha	Operator alpha
[in]	a_beta	Operator beta
[in]	a_Acoef	Operator A-coefficient
[in]	a_Bcoef	Operator B-coefficient
[in]	a_ghostPhi	Ghost cells in solution vector
[in]	a_ghostRhs	Ghost cells in right-hand side
[in]	a_smoother	Which smoother to use
[in]	a_relaxFactor	Relaxation factor
[in]	a_chebyOrder	Chebyshev polynomial degree (only used when a_smoother is Smoother::Chebyshev)
[in]	a_chebyEigRatio	Chebyshev eigenvalue ratio lambda_max/lambda_min (only used when a_smoother is Smoother::Chebyshev)
[in]	a_rasInnerSweeps	Number of inner frozen-ghost sweeps per restricted-additive-Schwarz iteration (only used when a_smoother is Smoother::RestrictedAdditiveSchwarz)
[in]	a_refluxFree	If true, use the reflux-free AMR operator (skips coarse-fine flux refluxing)
[in]	a_BcoefIrreg	B-coefficient on EB faces

Member Function Documentation

AMROperator()

void EBHelmholtzOp::AMROperator ( LevelData< EBCellFAB > &  a_Lphi, const LevelData< EBCellFAB > &  a_phiFine, const LevelData< EBCellFAB > &  a_phi, const LevelData< EBCellFAB > &  a_phiCoar, const bool  a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *  a_finerOp  )

finaloverride

Apply the AMR operator, i.e. compute L(phi) in an AMR context.

Parameters

[in]	a_phiFine	Phi on fine level
[in]	a_phi	Phi on this level
[in]	a_phiCoar	Phi on coar level
[in]	a_homogeneousPhysBC	Use homogeneous physical BCs or not
[in]	a_finerOp	Finer operator

This involves ghost cell interpolation if there's a coarse level, and refluxing if there's a fine level.

Parameters

[in]

a_Lphi

Lphi

AMROperatorNC()

void EBHelmholtzOp::AMROperatorNC ( LevelData< EBCellFAB > &  a_Lphi, const LevelData< EBCellFAB > &  a_phiFine, const LevelData< EBCellFAB > &  a_phi, bool  a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *  a_finerOp  )

finaloverride

Apply the AMR operator, i.e. compute L(phi) in an AMR context, assuming no coarser AMR levels.

Parameters

[out]	a_Lphi	L(phi)
[in]	a_phiFine	Phi on fine level
[in]	a_phi	Phi on this level
[in]	a_homogeneousPhysBC	Use homogeneous physical BCs or not
[in]	a_finerOp	Finer operator

This involves ghost cell interpolation if there's a coarse level, and refluxing if there's a fine level.

AMROperatorNF()

void EBHelmholtzOp::AMROperatorNF ( LevelData< EBCellFAB > &  a_Lphi, const LevelData< EBCellFAB > &  a_phi, const LevelData< EBCellFAB > &  a_phiCoar, bool  a_homogeneousPhysBC  )

finaloverride

Apply the AMR operator, i.e. compute L(phi) in an AMR context, assuming no finer levels.

Parameters

[out]	a_Lphi	L(phi)
[in]	a_phi	Phi on this level
[in]	a_phiCoar	Phi on coar level
[in]	a_homogeneousPhysBC	Use homogeneous physical BCs or not

This involves ghost cell interpolation if there's a coarse level, and refluxing if there's a fine level.

AMRProlong()

void EBHelmholtzOp::AMRProlong ( LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_coarseCorrection  )

finaloverride

Prolongation onto AMR level.

Parameters

[out]	a_correction	Interpolated correction
[in]	a_coarseCorrection	Correction on coarse level

AMRResidual()

void EBHelmholtzOp::AMRResidual ( LevelData< EBCellFAB > &  a_residual, const LevelData< EBCellFAB > &  a_phiFine, const LevelData< EBCellFAB > &  a_phi, const LevelData< EBCellFAB > &  a_phiCoar, const LevelData< EBCellFAB > &  a_rhs, bool  a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *  a_finerOp  )

finaloverride

Compute residual on this level. AMR version.

Parameters

[out]	a_residual	Residual on this level
[in]	a_phiFine	Phi on fine level
[in]	a_phi	Phi on this level
[in]	a_phiCoar	Phi on coar level
[in]	a_rhs	Right-hand side on this level
[in]	a_homogeneousPhysBC	Use homogeneous physical BCs or not
[in]	a_finerOp	Finer operator

AMRResidualNC()

void EBHelmholtzOp::AMRResidualNC ( LevelData< EBCellFAB > &  a_residual, const LevelData< EBCellFAB > &  a_phiFine, const LevelData< EBCellFAB > &  a_phi, const LevelData< EBCellFAB > &  a_rhs, bool  a_homogeneousPhysBC, AMRLevelOp< LevelData< EBCellFAB > > *  a_finerOp  )

finaloverride

Compute AMR residual on coarsest.

Parameters

[out]	a_residual	Residual on this level
[in]	a_phiFine	Phi on fine level
[in]	a_phi	Phi on this level
[in]	a_rhs	Right-hand side on this level
[in]	a_homogeneousPhysBC	Use homogeneous physical BCs or not
[in]	a_finerOp	Finer operator

AMRResidualNF()

void EBHelmholtzOp::AMRResidualNF ( LevelData< EBCellFAB > &  a_residual, const LevelData< EBCellFAB > &  a_phi, const LevelData< EBCellFAB > &  a_phiCoar, const LevelData< EBCellFAB > &  a_rhs, bool  a_homogeneousPhysBC  )

finaloverride

Compute AMR residual on finest AMR level.

Parameters

[out]	a_residual	Residual on this level
[in]	a_phi	Phi on this level
[in]	a_phiCoar	Phi on coar level
[in]	a_rhs	Right-hand side on this level
[in]	a_homogeneousPhysBC	Use homogeneous physical BCs or not

AMRRestrict()

void EBHelmholtzOp::AMRRestrict ( LevelData< EBCellFAB > &  a_residualCoarse, const LevelData< EBCellFAB > &  a_residual, const LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_coarseCorrection, bool  a_skip_res  )

finaloverride

Restrict residual.

Parameters

[out]	a_residualCoarse	Coarse residual
[out]	a_residual	Residual
[out]	a_correction	Correction on this level
[out]	a_coarseCorrection	Coarse level correction
[in]	a_skip_res	I have no idea what this one is supposed to do.

AMRUpdateResidual()

void EBHelmholtzOp::AMRUpdateResidual ( LevelData< EBCellFAB > &  a_residual, const LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_coarseCorrection  )

finaloverride

Update AMR residual.

Parameters

[in]	a_residual	Residual
[in]	a_correction	Correction
[in]	a_coarseCorrection	Coarse-level correction

applyDomainFlux()

void EBHelmholtzOp::applyDomainFlux ( EBCellFAB &  a_phi, const EBFluxFAB &  a_Bcoef, const Box &  a_cellBox, const DataIndex &  a_dit, const bool  a_homogeneousPhysBc  ) const

noexcept

Apply domain flux.

Parameters

[in,out]	a_phi	Cell data
[in]	a_Bcoef	Helmholtz B-coefficient
[in]	a_cellBox	Computation box
[in]	a_dit	Data index

Note

This sets the ghost cells of phi to phi(i-1) = phi(i) + F*dx/bco such that when we take the centered difference we get -bco*(phi(i) - phi(i-1))/dx = F.

Parameters

[in]

a_homogeneousPhysBc

Homogeneous phys bc

applyOp() [1/3]

void EBHelmholtzOp::applyOp ( EBCellFAB &  a_Lphi, EBCellFAB &  a_phi, const EBCellFAB &  a_Acoef, const EBFluxFAB &  a_Bcoef, const BaseIVFAB< Real > &  a_BcoefIrreg, const Box &  a_cellBox, const DataIndex &  a_dit, const bool  a_homogeneousPhysBC  ) const

noexcept

Apply operator in a grid box.

Parameters

[out]	a_Lphi	L(phi)
[in]	a_phi	Phi
[in]	a_cellBox	Grid box
[in]	a_dit	Data indxe
[in]	a_homogeneousPhysBC	Homogeneous physical BCs or not
[in]	a_Acoef	Description
[in]	a_Bcoef	Description
[in]	a_BcoefIrreg	Description

applyOp() [2/3]

void EBHelmholtzOp::applyOp ( LevelData< EBCellFAB > &  a_Lphi, const LevelData< EBCellFAB > &  a_phi, bool  a_homogeneousPhysBc  )

finaloverride

Apply operator.

Parameters

[out]	a_Lphi	L(phi)
[in]	a_phi	Phi
[in]	a_homogeneousPhysBc	Homogeneous physical BCs or not

This computes a_Lphi = L(a_phi) using homogeneous physical BCs or not

applyOp() [3/3]

void EBHelmholtzOp::applyOp	(	LevelData< EBCellFAB > &	a_Lphi,
		const LevelData< EBCellFAB > &	a_phi,
		const LevelData< EBCellFAB > *const	a_phiCoar,
		const bool	a_homogeneousPhysBC,
		const bool	a_homogeneousCFBC
	)

Apply operator on this level. This is a more general version which can turn on/off homogeneous and CF bcs.

Parameters

[out]	a_Lphi	L(phi)
[out]	a_phi	Phi on this level
[out]	a_phiCoar	Coarse-level phi. If you have a coar this
[in]	a_homogeneousPhysBC	Use homogeneous physical BCs or not
[in]	a_homogeneousCFBC	Use homogeneous coarse-fine bcs or not

Note

a_phi is not really const and we do a nasty cast. Leaving it as const because we only modify the ghost cells!

applyOpIrregular()

void EBHelmholtzOp::applyOpIrregular ( EBCellFAB &  a_Lphi, const EBCellFAB &  a_phi, const EBCellFAB &  a_Acoef, const EBFluxFAB &  a_Bcoef, const BaseIVFAB< Real > &  a_BcoefIrreg, const BaseIVFAB< Real > &  a_alphaDiagWeight, const Box &  a_cellBox, const DataIndex &  a_dit, const bool  a_homogeneousPhysBC  ) const

noexcept

Apply operator in irregular cells.

Parameters

[out]	a_Lphi	L(phi)
[in]	a_phi	Cell-centered data
[in]	a_cellBox	Cell box
[in]	a_dit	Grid index
[in]	a_homogeneousPhysBC	Homogeneous physical BCs or not
[in]	a_Acoef	Description
[in]	a_Bcoef	Description
[in]	a_BcoefIrreg	Description
[in]	a_alphaDiagWeight	Description

applyOpNoBoundary()

void EBHelmholtzOp::applyOpNoBoundary ( LevelData< EBCellFAB > &  a_Lphi, const LevelData< EBCellFAB > &  a_phi  )

finaloverride

Apply operator but turn off all BCs.

Parameters

[in,out]	a_Lphi	L(phi) on output
[in]	a_phi	Phi

applyOpRegular()

void EBHelmholtzOp::applyOpRegular ( EBCellFAB &  a_Lphi, EBCellFAB &  a_phi, const EBCellFAB &  a_Acoef, const EBFluxFAB &  a_Bcoef, const BaseIVFAB< Real > &  a_BcoefIrreg, const Box &  a_cellBox, const DataIndex &  a_dit, const bool  a_homogeneousPhysBC  ) const

noexcept

Apply operator in regular cells.

Parameters

[out]	a_Lphi	L(phi)
[in]	a_phi	Phi
[in]	a_cellBox	Grid box
[in]	a_dit	Data indxe
[in]	a_homogeneousPhysBC	Homogeneous physical BCs or not
[in]	a_Acoef	Description
[in]	a_Bcoef	Description
[in]	a_BcoefIrreg	Description

assign()

void EBHelmholtzOp::assign ( LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs  )

finaloverride

Assign data.

This does a copy from rhs to lhs

Parameters

[in,out]	a_lhs	Description
[in,out]	a_rhs	Description

assignCopier()

void EBHelmholtzOp::assignCopier ( LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs, const Copier &  a_copier  )

finaloverride

Assign lhs.

This is the version that is called by AMRMultiGrid::VCycle. Note that the other version might be called by other operators (e.g., EBBackwardEuler)

Parameters

[out]	a_lhs	Outgoing data
[in]	a_rhs	Incoming data
[in]	a_copier	Copier

assignLocal()

void EBHelmholtzOp::assignLocal ( LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs  )

finaloverride

Local assignment function.

Parameters

[out]	a_lhs	Equal to a_rhs on output
[in]	a_rhs	Data

axby()

void EBHelmholtzOp::axby ( LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_x, const LevelData< EBCellFAB > &  a_y, const Real  a_a, const Real  a_b  )

finaloverride

Set a_lhs = a*x + b*y.

Parameters

[out]	a_lhs	Result data
[in]	a_x	x-data
[in]	a_y	y-data
[in]	a_a	Scaling factor
[in]	a_b	Scaling factor

buildCopier()

void EBHelmholtzOp::buildCopier ( Copier &  a_copier, const LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs  )

override

Build copier.

Parameters

[out]	a_copier	Copier for copying between a_lhs and a_rhs
[in]	a_lhs	Copying from
[in]	a_rhs	Copying to

chebyshevKernel()

void EBHelmholtzOp::chebyshevKernel ( EBCellFAB &  a_Lcorr, EBCellFAB &  a_corr, const EBCellFAB &  a_resid, const EBCellFAB &  a_Acoef, const EBFluxFAB &  a_Bcoef, const BaseIVFAB< Real > &  a_BcoefIrreg, const Box &  a_cellBox, const DataIndex &  a_dit, const Real  a_omega  ) const

noexcept

Chebyshev smoother kernel.

Same as pointJacobiKernel but scales the update by a_omega (the Chebyshev step size).

Parameters

[in,out]	a_Lcorr	Storage for computing L(a_corr)
[in,out]	a_corr	Correction
[in]	a_resid	Residual
[in]	a_Acoef	A-coefficient
[in]	a_Bcoef	B-coefficient
[in]	a_BcoefIrreg	B-coefficient on EB faces
[in]	a_cellBox	Grid box
[in]	a_dit	Data index
[in]	a_omega	Chebyshev step-size scaling factor

coarsenCell()

void EBHelmholtzOp::coarsenCell	(	LevelData< EBCellFAB > &	a_phi,
		const LevelData< EBCellFAB > &	a_phiFine
	)

Coarsen data from fine to coar level.

Parameters

[in]	a_phi	Data on this level
[in]	a_phiFine	Data on finer level

Note

This routine is used in AMROperator to ensure that EB fluxes which are described by stencils get consistent data under the fine level.

coarsenFlux()

void EBHelmholtzOp::coarsenFlux	(	LevelData< EBFluxFAB > &	a_flux,
		const LevelData< EBFluxFAB > &	a_fineFlux
	)

Coarsen fluxes on the fine level onto this level.

User must update fluxes before calling this routine.

Parameters

[in]	a_flux	Flux on this level.
[in]	a_fineFlux	Flux on finer level

Note

This routine is used in the reflux-free AMROperator

computeFaceCenteredFlux()

void EBHelmholtzOp::computeFaceCenteredFlux ( EBFaceFAB &  a_fluxCenter, const EBCellFAB &  a_phi, const Box &  a_cellBox, const DataIndex &  a_dit, const int  a_dir  )

protected

Compute face-centered fluxes.

Parameters

[out]	a_fluxCenter	Flux on center
[in]	a_phi	Cell-centered data
[in]	a_cellBox	Cell-centered compute box.
[in]	a_dit	Data index
[in]	a_dir	Coordinate direction

Note

This computes the flux for all dir-oriented faces in a_cellBox except for boundary faces.

computeFaceCentroidFlux()

void EBHelmholtzOp::computeFaceCentroidFlux ( EBFaceFAB &  a_flux, const EBCellFAB &  a_phi, const Box &  a_cellBox, const DataIndex &  a_dit, const int  a_dir  )

protected

Compute face-centroid fluxes.

Parameters

[out]	a_flux	Flux on face centroid
[in]	a_phi	Cell-centered data
[in]	a_cellBox	Cell-centered compute box.
[in]	a_dit	Data index
[in]	a_dir	Coordinate direction

Note

This computes the flux for all dir-oriented faces in a_cellBox except for boundary faces.

computeFlux() [1/2]

void EBHelmholtzOp::computeFlux ( const LevelData< EBCellFAB > &  a_phi )

protected

Fill face centroid-centered fluxes on this level.

Parameters

[in]

a_phi

Cell-centered data. Must have updated ghost cells for this to make any sense.

Note

This computes the flux onto m_flux everywhere except on boundary faces.

computeFlux() [2/2]

void EBHelmholtzOp::computeFlux ( EBFaceFAB &  a_fluxCentroid, const EBCellFAB &  a_phi, const Box &  a_cellBox, const DataIndex &  a_dit, const int  a_dir  )

protected

Compute centroid fluxes in a direction.

Parameters

[out]	a_fluxCentroid	Flux on face centroid
[in]	a_phi	Cell-centered data
[in]	a_cellBox	Cell-centered compute box.
[in]	a_dit	Data index
[in]	a_dir	Coordinate direction

Note

This computes the flux for all dir-oriented faces in a_cellBox except for boundary faces.

computeSpectralRadius()

void EBHelmholtzOp::computeSpectralRadius ( )

protected

Estimate the spectral radius of D^{-1}A (Jacobi-preconditioned operator) via the Gershgorin bound.

rho_i = 2 - alpha * A_i * relCoef_i. Equals 2 for pure Laplacian and decreases with alpha > 0. Must be called after computeRelaxationCoefficient() and computeDiagWeight().

create()

void EBHelmholtzOp::create ( LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs  )

finaloverride

Create data which clones the layout of the other.

Parameters

[out]	a_lhs	Data clone (returned data is not initialized)
[out]	a_rhs	Data layout to be cloned

createCoarsened()

void EBHelmholtzOp::createCoarsened ( LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs, const int &  a_refRat  )

finaloverride

Create coarsening of data holder.

Parameters

[out]	a_lhs	Coarsened data
[in]	a_rhs	Fine data
[in]	a_refRat	Coarsening factor

createCoarser()

void EBHelmholtzOp::createCoarser ( LevelData< EBCellFAB > &  a_coarse, const LevelData< EBCellFAB > &  a_fine, bool  a_ghosted  )

finaloverride

Create coarsened data.

Parameters

[out]	a_coarse	Coarse data
[in]	a_fine	Fine data
[in]	a_ghosted	Include ghost cells or nto

diagonalScale()

void EBHelmholtzOp::diagonalScale ( LevelData< EBCellFAB > &  a_rhs, bool  a_kappaWeighted  )

finaloverride

Do diagonal scaling.

Parameters

[in,out]	a_rhs	Data to be scaled
[in]	a_kappaWeighted	Use kappa-weighted scaling or not

divideByIdentityCoef()

void EBHelmholtzOp::divideByIdentityCoef ( LevelData< EBCellFAB > &  a_rhs )

finaloverride

Divide by the a-coefficient.

Parameters

[in,out]

a_rhs

Divided data

dotProduct()

Real EBHelmholtzOp::dotProduct ( const LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs  )

finaloverride

Compute the dot product??

Returns

Return value

Parameters

[in,out]	a_lhs	Description
[in,out]	a_rhs	Description

dotProductMaskedLocal()

void EBHelmholtzOp::dotProductMaskedLocal ( Real &  a_sumKappaXY, Real &  a_sumVolume, const LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs, const LevelData< BaseFab< bool > > &  a_mask, const bool  a_needVolume  ) const

noexcept

Rank-local (un-reduced) partials for the masked AMR-composite Krylov inner product.

Same kappa-weighted inner product as dotProduct, except cells where a_mask is false (e.g. cells covered by a finer AMR level) are excluded from the numerator; the volume normalisation counts all valid cells, matching Chombo's MultilevelLinearOp AMR dot product. Unlike dotProduct this performs NO MPI reduction: it returns the rank-local partial sums so the caller (AMRMultigridKrylovOp) can batch the per-level partials into a single MPI_Allreduce. The volume is geometry-only and is reduced/cached once by the caller, so a_needVolume can be set false on later calls to skip recomputing it.

Parameters

[out]	a_sumKappaXY	Rank-local masked numerator (sum of kappa*X*Y over unmasked cells)
[out]	a_sumVolume	Rank-local volume (sum over all valid cells); set to zero when a_needVolume is false
[in]	a_lhs	First operand
[in]	a_rhs	Second operand
[in]	a_mask	Per-cell mask; false marks cells excluded from the numerator
[in]	a_needVolume	If false, skip accumulating a_sumVolume (caller already cached it)

fillDomainFlux()

void EBHelmholtzOp::fillDomainFlux	(	EBFluxFAB &	a_flux,
		const EBCellFAB &	a_phi,
		const Box &	a_cellBox,
		const DataIndex &	a_dit,
		const bool	a_homogeneousPhysBC
	)

Fill domain flux. This fills the flux on the domain face using centered differencing ala applyDomainFlux.

a_flux is replaced by the user-specified flux on the domain faces.

Parameters

[in,out]	a_flux	Flux data holder.
[in,out]	a_phi	Cell-centered data
[in]	a_cellBox	Computation box
[in]	a_dit	Data index
[in]	a_homogeneousPhysBC	If true, use homogeneous physical BCs

fillGrad()

void EBHelmholtzOp::fillGrad ( const LevelData< EBCellFAB > &  a_phi )

finaloverride

Not called, I think.

Parameters

[in]

a_phi

Phi

gauSaiMultiColor()

void EBHelmholtzOp::gauSaiMultiColor	(	LevelData< EBCellFAB > &	a_phi,
		const LevelData< EBCellFAB > &	a_Lphi,
		const LevelData< EBCellFAB > &	a_rhs,
		const IntVect &	a_color
	)		const

Multi-colored Gauss-Seidel kernel. Public because MFHelmholtzOp may want to use use.

Parameters

[in]	a_phi	Correction
[in]	a_rhs	Residual
[in]	a_color	"Color":
[in]	a_Lphi	Description

gauSaiMultiColorKernel()

void EBHelmholtzOp::gauSaiMultiColorKernel ( EBCellFAB &  a_Lcorr, EBCellFAB &  a_corr, const EBCellFAB &  a_resid, const EBCellFAB &  a_Acoef, const EBFluxFAB &  a_Bcoef, const BaseIVFAB< Real > &  a_BcoefIrreg, const Box &  a_cellBox, const DataIndex &  a_dit, const IntVect &  a_color  ) const

noexcept

Multi-color Gauss-Seidel kernel.

Parameters

[in,out]	a_Lcorr	Storage for computing L(a_corr)
[in,out]	a_corr	Correction
[in]	a_resid	Residual
[in]	a_Acoef	A-coefficient
[in]	a_Bcoef	B-coefficient
[in]	a_BcoefIrreg	B-coefficient on EB faces
[in]	a_cellBox	Grid box
[in]	a_dit	Data index
[in]	a_color	Color

gauSaiRedBlackKernel()

void EBHelmholtzOp::gauSaiRedBlackKernel ( EBCellFAB &  a_Lcorr, EBCellFAB &  a_corr, const EBCellFAB &  a_resid, const EBCellFAB &  a_Acoef, const EBFluxFAB &  a_Bcoef, const BaseIVFAB< Real > &  a_BcoefIrreg, const Box &  a_cellBox, const DataIndex &  a_dit, const int &  a_redBlack  ) const

noexcept

Red-black Gauss-Seidel kernel.

Parameters

[in,out]	a_Lcorr	Storage for computing L(a_corr)
[in,out]	a_corr	Correction
[in]	a_resid	Residual
[in]	a_Acoef	A-coefficient
[in]	a_Bcoef	B-coefficient
[in]	a_BcoefIrreg	B-coefficient on EB faces
[in]	a_redBlack	Red or black
[in]	a_cellBox	Grid box
[in]	a_dit	Data index

getAcoef()

const RefCountedPtr< LevelData< EBCellFAB > > & EBHelmholtzOp::getAcoef

(

)

Get the Helmholtz A-coefficient on cell centers.

Returns

m_Acoef

getBcoef()

const RefCountedPtr< LevelData< EBFluxFAB > > & EBHelmholtzOp::getBcoef

(

)

Get the Helmholtz B-coefficient on faces.

Returns

m_Bcoef

getBcoefIrreg()

const RefCountedPtr< LevelData< BaseIVFAB< Real > > > & EBHelmholtzOp::getBcoefIrreg

(

)

Get the Helmholtz B-coefficient on the EB.

Returns

m_Bcoef

getChebyEigRatio()

Real EBHelmholtzOp::getChebyEigRatio ( ) const

noexcept

Returns m_chebyEigRatio. Used by MFHelmholtzOp to replicate Chebyshev parameters.

Returns

Chebyshev eigenvalue ratio (lambda_max / lambda_min)

getChebyOrder()

int EBHelmholtzOp::getChebyOrder ( ) const

noexcept

Returns m_chebyOrder. Used by MFHelmholtzOp to replicate Chebyshev parameters.

Returns

Chebyshev polynomial degree

getFaceCenterFluxStencil()

VoFStencil EBHelmholtzOp::getFaceCenterFluxStencil ( const FaceIndex &  a_face, const DataIndex &  a_dit  ) const

protected

Get the face-centered flux stencil.

Parameters

[in]	a_face	Face
[in]	a_dit	Data index. Need because we multiply by B-coefficient

Returns

Returns a stencil for the face-centered flux for the given face

Note

If the face is a boundary face the returned stencil is empty.

getFaceCentroidFluxStencil()

VoFStencil EBHelmholtzOp::getFaceCentroidFluxStencil ( const FaceIndex &  a_face, const DataIndex &  a_dit  ) const

protected

Get the face-centroid flux stencil.

Parameters

[in]	a_face	Face
[in]	a_dit	Data index. Need because we multiply by B-coefficient

Returns

Returns a stencil for the face-centroid flux for the given face

Note

The current version just interpolates the centered fluxes.

getFlux() [1/2]

LevelData< EBFluxFAB > & EBHelmholtzOp::getFlux

(

)

const

Returns m_flux. This is used in refluxing routines.

Returns

Flux

getFlux() [2/2]

void EBHelmholtzOp::getFlux ( EBFluxFAB &  a_flux, const LevelData< EBCellFAB > &  a_data, const Box &  a_grid, const DataIndex &  a_dit, Real  a_scale  )

finaloverride

Fill flux.

Parameters

[out]	a_flux	Flux
[in]	a_data	Data for which we will compute the flux
[in]	a_grid	Grid
[in]	a_dit	Corresponding data index.
[in]	a_scale	Scaling factor

getRasInnerSweeps()

int EBHelmholtzOp::getRasInnerSweeps ( ) const

noexcept

Returns m_rasInnerSweeps. Used by MFHelmholtzOp to replicate the Schwarz block solve.

Returns

Number of inner (frozen-ghost) red-black sweeps per outer Schwarz iteration

getRelaxationCoeff()

const LevelData< EBCellFAB > & EBHelmholtzOp::getRelaxationCoeff ( ) const

noexcept

Returns m_relCoef. Used by MFHelmholtzOp.

Returns

Relaxation coeff

getSpectralRadius()

Real EBHelmholtzOp::getSpectralRadius ( ) const

noexcept

Returns m_spectralRadius. Used by MFHelmholtzOp to replicate Chebyshev parameters.

Returns

Gershgorin bound on rho(D^{-1}A)

getVofIterIrreg()

LayoutData< VoFIterator > & EBHelmholtzOp::getVofIterIrreg ( ) const

noexcept

Return all-cut VoFIterators (one per patch). Used by MFHelmholtzOp.

Returns

All-cut VoF iterators over irregular cells on this level.

homogeneousCFInterp()

void EBHelmholtzOp::homogeneousCFInterp

(

LevelData< EBCellFAB > &

a_phi

)

Do homogeneous coarse-fine interpolation.

Parameters

[in]

a_phi

Data

incr()

void EBHelmholtzOp::incr ( LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs, const Real  a_scale  )

finaloverride

Increment operator.

Parameters

[in,out]	a_lhs	Data to be incremented
[in]	a_rhs	Incrementation data
[in]	a_scale	Scaling factor

inhomogeneousCFInterp()

void EBHelmholtzOp::inhomogeneousCFInterp	(	LevelData< EBCellFAB > &	a_phi,
		const LevelData< EBCellFAB > &	a_phiCoar
	)

Inhomogeneous coarse-fine interpolation.

Parameters

[in,out]	a_phiCoar	Coarse data.
[in]	a_phi	Phi

interpolateCF()

void EBHelmholtzOp::interpolateCF	(	LevelData< EBCellFAB > &	a_phiFine,
		const LevelData< EBCellFAB > *	a_phiCoar,
		const bool	a_homogeneousCFBC
	)

Apply coarse-fine boundary conditions.

Parameters

[in,out]	a_phiFine	Fine data
[in,out]	a_phiCoar	Coarse data
[in]	a_homogeneousCFBC	Homogeneous CF or not (i.e. coar data is zero);

norm()

Real EBHelmholtzOp::norm ( const LevelData< EBCellFAB > &  a_rhs, const int  a_order  )

finaloverride

Compute norm of data.

Parameters

[in]	a_rhs	Data to compute norm of
[in]	a_order	Norm order (currently ignored)

a_order is currently ignored as we call a static routine in EBAMRPoissonOp for now.

Returns

Norm value

operator=() [1/2]

EBHelmholtzOp & EBHelmholtzOp::operator= ( const EBHelmholtzOp &&  a_oper )

delete

No move assignment allowed.

Parameters

[in]

a_oper

Oper

operator=() [2/2]

EBHelmholtzOp & EBHelmholtzOp::operator= ( const EBHelmholtzOp &  a_oper )

delete

No copy assignment allowed.

Parameters

[in]

a_oper

Oper

pointJacobiKernel()

void EBHelmholtzOp::pointJacobiKernel ( EBCellFAB &  a_Lcorr, EBCellFAB &  a_corr, const EBCellFAB &  a_resid, const EBCellFAB &  a_Acoef, const EBFluxFAB &  a_Bcoef, const BaseIVFAB< Real > &  a_BcoefIrreg, const Box &  a_cellBox, const DataIndex &  a_dit  ) const

noexcept

Point Jacobi kernel.

Parameters

[in,out]	a_Lcorr	Storage for computing L(a_corr)
[in,out]	a_corr	Correction
[in]	a_resid	Residual
[in]	a_Acoef	A-coefficient
[in]	a_Bcoef	B-coefficient
[in]	a_BcoefIrreg	B-coefficient on EB faces
[in]	a_cellBox	Grid box
[in]	a_dit	Data index

preCond()

void EBHelmholtzOp::preCond ( LevelData< EBCellFAB > &  a_corr, const LevelData< EBCellFAB > &  a_residual  )

finaloverride

Precondition system before bottom solve.

Parameters

[in]	a_corr	Correction
[in]	a_residual	Residual

This just runs a few relaxations.

prolongIncrement()

void EBHelmholtzOp::prolongIncrement ( LevelData< EBCellFAB > &  a_phi, const LevelData< EBCellFAB > &  a_correctCoarse  )

finaloverride

Prolongation method.

Parameters

[out]	a_phi	Correction on this level
[out]	a_correctCoarse	Correction on coarse level

reflux()

void EBHelmholtzOp::reflux ( LevelData< EBCellFAB > &  a_Lphi, const LevelData< EBCellFAB > &  a_phiFine, const LevelData< EBCellFAB > &  a_phi, AMRLevelOp< LevelData< EBCellFAB > > &  a_finerOp  )

protected

Reflux algorithm.

Parameters

[in,out]	a_Lphi	L(phiFine, phiCoar). This is modified along the CF interface to consistently account for fine-level fluxes.
[in]	a_phiFine	Data on finer level
[in]	a_phi	Data on this operator's level
[in]	a_finerOp	Operator on fine level

refluxFreeAMROperator()

void EBHelmholtzOp::refluxFreeAMROperator	(	LevelData< EBCellFAB > &	a_Lphi,
		const LevelData< EBCellFAB > &	a_phiFine,
		const LevelData< EBCellFAB > &	a_phi,
		const LevelData< EBCellFAB > &	a_phiCoar,
		const bool	a_homogeneousPhysBC,
		AMRLevelOp< LevelData< EBCellFAB > > *	a_finerOp
	)

Apply the AMR operator, i.e. compute L(phi) in an AMR context.

This is the reflux-free version of AMROperator.

Parameters

[in]	a_phiFine	Phi on fine level
[in]	a_phi	Phi on this level
[in]	a_phiCoar	Phi on coar level
[in]	a_homogeneousPhysBC	Use homogeneous physical BCs or not
[in]	a_finerOp	Finer operator

This involves ghost cell interpolation if there's a coarse level, and refluxing if there's a fine level.

Parameters

[in]

a_Lphi

Lphi

refToCoarser()

int EBHelmholtzOp::refToCoarser ( )

finaloverride

Return coarsening factor to coarser level (1 if there is no coarser level);.

Returns

Return value

relax()

void EBHelmholtzOp::relax ( LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_residual, int  a_iterations  )

finaloverride

Relaxation method. This does smoothing for the system L(correction) = residual.

Parameters

[in,out]	a_correction	Correction
[in]	a_residual	Residual
[in]	a_iterations	Number of iterations

relaxChebyshev()

void EBHelmholtzOp::relaxChebyshev ( LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_residual, const int  a_iterations  )

protected

Chebyshev polynomial smoother.

Applies m_chebyOrder Chebyshev-Richardson steps of the form correction += omega_i * relCoef * (residual - L(correction)). The step sizes omega_i are derived from Chebyshev nodes on [lambdaMax/m_chebyEigRatio, lambdaMax] where lambdaMax = 2 (safe bound for the Jacobi-preconditioned Laplacian-type operator).

Parameters

[in,out]	a_correction	Correction
[in]	a_residual	Residual
[in]	a_iterations	Number of outer smoother invocations (each applies m_chebyOrder steps)

relaxGSMultiColor()

void EBHelmholtzOp::relaxGSMultiColor ( LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_residual, const int  a_iterations  )

protected

Multi-colored gauss-seidel relaxation.

Parameters

[in,out]	a_correction	Correction
[in]	a_residual	Residual
[in]	a_iterations	Number of iterations

relaxGSRedBlack()

void EBHelmholtzOp::relaxGSRedBlack ( LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_residual, const int  a_iterations  )

protected

Jacobi relaxation.

Parameters

[in,out]	a_correction	Correction
[in]	a_residual	Residual
[in]	a_iterations	Number of iterations

relaxPointJacobi()

void EBHelmholtzOp::relaxPointJacobi ( LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_residual, const int  a_iterations  )

protected

Jacobi relaxation.

Parameters

[in,out]	a_correction	Correction
[in]	a_residual	Residual
[in]	a_iterations	Number of iterations

relaxRestrictedAdditiveSchwarz()

void EBHelmholtzOp::relaxRestrictedAdditiveSchwarz ( LevelData< EBCellFAB > &  a_correction, const LevelData< EBCellFAB > &  a_residual, const int  a_iterations  )

protected

Restricted additive Schwarz smoother (block smoother).

Each outer iteration exchanges ghost cells once and then performs m_rasInnerSweeps red-black Gauss-Seidel sweeps per patch with the ghost cells held frozen (Dirichlet). The blocks are the disjoint patches, so the update is automatically restricted to the valid region and no additive damping is required (overlap = 0). Because the ghost exchange is amortised over the inner sweeps, this smoother performs fewer halo exchanges per multigrid cycle than point relaxation.

Parameters

[in,out]	a_correction	Correction
[in]	a_residual	Residual
[in]	a_iterations	Number of outer Schwarz iterations

residual()

void EBHelmholtzOp::residual ( LevelData< EBCellFAB > &  a_residual, const LevelData< EBCellFAB > &  a_phi, const LevelData< EBCellFAB > &  a_rhs, const bool  a_homogeneousPhysBc  )

override

Compute residual on this level.

Parameters

[out]	a_residual	Residual rhs - L(phi)
[in]	a_phi	phi
[in]	a_rhs	Right-hand side of system
[in]	a_homogeneousPhysBc	Homogeneous phys bc

restrictResidual()

void EBHelmholtzOp::restrictResidual ( LevelData< EBCellFAB > &  a_resCoar, LevelData< EBCellFAB > &  a_phi, const LevelData< EBCellFAB > &  a_rhs  )

finaloverride

Restrict residual onto coarse level.

Parameters

[in,out]	a_resCoar	Coarse residual
[in,out]	a_phi	Phi on this level
[in]	a_rhs	Rhs on this level

scale()

void EBHelmholtzOp::scale ( LevelData< EBCellFAB > &  a_lhs, const Real &  a_scale  )

finaloverride

Scale data. Returns a_lhs = a_lhs*a_scale.

Parameters

[in,out]	a_lhs	Data to be scaled
[in]	a_scale	Scaling factor

scaleLocal()

void EBHelmholtzOp::scaleLocal ( LevelData< EBCellFAB > &  a_lhs, const LevelData< EBCellFAB > &  a_rhs  )

staticnoexcept

Local scaling function. Multiplies the left-hand side by the right-hand side.

Parameters

[out]	a_lhs	Equal to a_lhs*a_rhs on output
[in]	a_rhs	Scaling data

setAcoAndBco()

void EBHelmholtzOp::setAcoAndBco	(	const RefCountedPtr< LevelData< EBCellFAB > > &	a_Acoef,
		const RefCountedPtr< LevelData< EBFluxFAB > > &	a_Bcoef,
		const RefCountedPtr< LevelData< BaseIVFAB< Real > > > &	a_BcoefIrreg
	)

Update with new A and B coefficients.

Note

This will call defineStencils but will not recompute the stencils from the BC objects.

Parameters

[in]	a_Acoef	Operator A-coefficient
[in]	a_Bcoef	Operator B-coefficient
[in]	a_BcoefIrreg	Operator B-coefficient (on EB faces)

setAlphaAndBeta()

void EBHelmholtzOp::setAlphaAndBeta ( const Real &  a_alpha, const Real &  a_beta  )

finaloverride

Set alpha coefficient and beta coefficient (can change as diffusion solvers progress)

Parameters

[in]	a_alpha	Alpha-coefficient
[in]	a_beta	Beta-coefficient

setToZero()

void EBHelmholtzOp::setToZero ( LevelData< EBCellFAB > &  a_lhs )

finaloverride

Set data to zero.

Parameters

[in,out]

a_lhs

Data to be set to zero.

turnOffCoarsening()

void EBHelmholtzOp::turnOffCoarsening

(

)

Turn off coarsening operation.

Note

Done when MFHelmholtzOp does the coarsening and we don't want EBHelmholtzOp to redo it.

turnOffExchange()

void EBHelmholtzOp::turnOffExchange

(

)

Turn off exchange operation.

Note

Done when MFHelmholtzOp does the exchange and we don't want EBHelmholtzOp to redo it.

turnOnCoarsening()

void EBHelmholtzOp::turnOnCoarsening

(

)

Turn on coarsening operation.

Note

Done when MFHelmholtzOp does the coarsening and we don't want EBHelmholtzOp to redo it.

turnOnExchange()

void EBHelmholtzOp::turnOnExchange

(

)

Turn on exchange operation.

Note

Done when MFHelmholtzOp does the exchange and we don't want EBHelmholtzOp to redo it.

Member Data Documentation

m_aggRelaxStencil

LayoutData<RefCountedPtr<VCAggStencil> > EBHelmholtzOp::m_aggRelaxStencil

protected

For making irregular stencil applications go faster.

This wraps m_relaxStencils in VCAggStencil (which computes explicit stencil offsets)

m_chebyEigRatio

Real EBHelmholtzOp::m_chebyEigRatio

protected

Ratio lambda_max/lambda_min for the Chebyshev eigenvalue window.

Targets the upper (1 - 1/m_chebyEigRatio) fraction of the spectrum.

m_doCoarsen

bool EBHelmholtzOp::m_doCoarsen

protected

Turn on/off exchange operation.

Note

Used by MFHelmholtzOp for optimizing how calls in EBHelmholtzOp are made

m_doExchange

bool EBHelmholtzOp::m_doExchange

protected

Turn on/off exchange operation.

Note

Used by MFHelmholtzOp for optimizing how calls in EBHelmholtzOp are made

m_doInterpCF

bool EBHelmholtzOp::m_doInterpCF

protected

Do coarse-fine interpolation or not.

Note

Used by MFHelmholtzOp for optimizing how calls in EBHelmholtzOp are made

m_domainBndryBoxes

LayoutData<DomainBndryBoxes> EBHelmholtzOp::m_domainBndryBoxes

protected

Precomputed domain-boundary boxes for each grid patch.

Filled in EBHelmholtzOp::defineStencils and consumed by applyDomainFlux/fillDomainFlux.

m_relaxStencils

LayoutData<BaseIVFAB<VoFStencil> > EBHelmholtzOp::m_relaxStencils

protected

Operator stencils in irregular cells (and ones that border irregular cells if using a centroid discretization).

This stencil is => sum(fluxes)/dx, not including boundary faces or EB faces. I.e. this is the same as kappa*div(F) with the exclusion of faces where we have boundary conditions.

m_spectralRadius

Real EBHelmholtzOp::m_spectralRadius

protected

Spectral radius estimate for the Jacobi-preconditioned operator D^{-1}A.

Computed via the Gershgorin bound after defineStencils(). Used as lambda_max in the Chebyshev smoother.

---

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

• Source/Elliptic/CD_EBHelmholtzOp.H
• Source/Elliptic/CD_EBHelmholtzOp.cpp