Functions
AMREX_GPU_HOST_DEVICE AMREX_FORCE_INLINE void	avgdown_masked (int i, int j, int k, int n, amrex::Array4< amrex::Real > const &crse, amrex::Array4< amrex::Real const > const &fine, amrex::Array4< amrex::Real const > const &fmsk, amrex::Array4< amrex::Real const > const &cmsk, int ccomp, int fcomp, amrex::IntVect const &ratio) noexcept

AMREX_GPU_HOST_DEVICE AMREX_FORCE_INLINE void	avgdown_faces_masked (int i, int j, int k, int n, amrex::Array4< amrex::Real > const &crse, amrex::Array4< amrex::Real const > const &fine, amrex::Array4< amrex::Real const > const &fmsk, amrex::Array4< amrex::Real const > const &cmsk, int ccomp, int fcomp, amrex::IntVect const &ratio, int idir) noexcept

Detailed Description

Two-way nesting's fine-to-coarse update, weighted by the land/sea mask.

Ported from ROMS fine2coarse2d/fine2coarse3d in Nonlinear/nesting.F, the AreaAvg = .FALSE. branch that every call site there uses:

num  = sum of (fine value * fine mask) over the fine points under the coarse one
den  = sum of min(1, fine mask) over the same
crse = num/den * coarse mask

Dividing by the count of wet fine points rather than by how many there are means a coarse point only partly covered by water takes the mean of that water, not a value diluted toward zero by land. That is not conservative; ROMS makes the same trade deliberately. The coarse mask has the final say on wet or dry.

Where every fine point is wet these have to reproduce amrex_avgdown and amrex_avgdown_faces bit for bit, or every existing multilevel answer moves in its last bit.

Function Documentation

◆ avgdown_faces_masked()

AMREX_GPU_HOST_DEVICE AMREX_FORCE_INLINE void REMORAMaskedAvgDown::avgdown_faces_masked	(	int	i,
		int	j,
		int	k,
		int	n,
		amrex::Array4< amrex::Real > const &	crse,
		amrex::Array4< amrex::Real const > const &	fine,
		amrex::Array4< amrex::Real const > const &	fmsk,
		amrex::Array4< amrex::Real const > const &	cmsk,
		int	ccomp,
		int	fcomp,
		amrex::IntVect const &	ratio,
		int	idir
	)

noexcept

Face-centered fields: only the fine faces in the coarse face's own plane, the stencil amrex_avgdown_faces uses, so "wet" means the flow really does cross here.

Definition at line 69 of file REMORA_MaskedAverageDown.H.

Referenced by REMORA::average_down_masked(), and REMORA::average_down_with_grow_cells().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ avgdown_masked()

AMREX_GPU_HOST_DEVICE AMREX_FORCE_INLINE void REMORAMaskedAvgDown::avgdown_masked	(	int	i,
		int	j,
		int	k,
		int	n,
		amrex::Array4< amrex::Real > const &	crse,
		amrex::Array4< amrex::Real const > const &	fine,
		amrex::Array4< amrex::Real const > const &	fmsk,
		amrex::Array4< amrex::Real const > const &	cmsk,
		int	ccomp,
		int	fcomp,
		amrex::IntVect const &	ratio
	)