LocalInterpCtx Class Reference

The "local interpolation context" describes the processor's part of the source NetCDF file (for regridding). More...

#include <LocalInterpCtx.hh>

Public Member Functions
	LocalInterpCtx (grid_info g, const double zlevsIN[], const double zblevsIN[], IceGrid &grid)
	Construct a local interpolation context from arrays of parameters.
	~LocalInterpCtx ()
	Deallocate memory.
int	kBelowHeight (const double height)
	Get the index for the source grid below a given vertical elevation within the ice.
int	kbBelowHeight (const double elevation)
	Get the index for the source grid below a given vertical elevation within the bedrock.
PetscErrorCode	printGrid ()
	Print out the grid information stored in the local interpolation context.
PetscErrorCode	printArray ()
	Print out the actual array information stored in the local interpolation context.
Public Attributes
double	fstart [3]
double	delta [3]
int	start [5]
int	count [5]
double *	a
	temporary buffer
int	a_len
	the size of the buffer
int	nz
	number of z-levels
int	nzb
	number of zb-levels
double *	zlevs
double *	zblevs
bool	no_regrid_ice
bool	no_regrid_bedrock
bool	report_range
MPI_Comm	com
	MPI Communicator (for printing, mostly).
PetscMPIInt	rank
	MPI rank, to allocate a_raw on proc 0 only.

Detailed Description

The "local interpolation context" describes the processor's part of the source NetCDF file (for regridding).

The local interpolation context contains the details of how the processor's block of the new computational domain fits into the domain of the netCDF file. Note that each vertical column of the grid is owned by exactly one processor.

For any particular dimension, we have a new computational domain $[a,b]$ with spacing $h$ so there are $n = (b - a) / h$ interior cells, indexed by $\{i_0, \dots, i_n\}$ . The local processor owns a range $\{i_m, \dots, i_{m'}\}$ . Suppose the netCDF file has domain $[A,B]$ , spacing $H$ , and cells. In order to interpolate onto these points, we need the indices $\{I_m, \dots, I_{m'}\}$ of the netCDF file so that

$[x(i_m), x(i_{m'})] \quad \text{is a subset of} \quad [x(I_m), x(I_{m'})]$

In the code, xbdy. We have obtained the netCDF bounds $x(I_0)$ and $x(I_N)$ in the bdy array and the number of elements $N+1$ in the dim array.

This is a summary of the fields of a LocalInterpCtx, and a translation between scalars in the previous paragraphs and the fields:

$I_m = \operatorname{floor}((x(i_m) - A) / H)$ is called start
$I_{m'} - I_m = \operatorname{ceil}((x(i_{m'}) - X(I_m)) / H$ is count - 1
is called fstart
is called delta

delta and fstart are not used in the vertical dimension because the spacing is not generally equal. Rather, zlevs and zblevs contain the needed information.

The arrays start and count have 5 integer entries, corresponding to the dimensions $t, x, y, z, zb$ .

Definition at line 74 of file LocalInterpCtx.hh.

Constructor & Destructor Documentation

LocalInterpCtx	(	grid_info	g,
		const double	zlevsIN[],
		const double	zblevsIN[],
		IceGrid &	grid
	)

Construct a local interpolation context from arrays of parameters.

This method constructs a class from existing information already read from a NetCDF file and stored in arrays.

The essential quantities to compute are where each processor should start within the NetCDF file grid (start[]) and how many grid points, from the starting place, the processor has. The resulting portion of the grid is stored in array a (a field of the LocalInterCtx).

We make conservative choices about start[] and count[]. In particular, the portions owned by processors must overlap at one point in the NetCDF file grid, but they may overlap more than that (as computed here).

Note this constructor doesn't extract new information from the NetCDF file or do communication. The information from the NetCDF file must already be extracted, validly stored in a grid_info structure g and preallocated arrays zlevsIN[], zblevsIN[], and each processor must have the same values for all of them.

The IceGrid is used to determine what ranges of the target arrays (i.e. Vecs into which NetCDF information will be interpolated) are owned by each processor.

Definition at line 49 of file LocalInterpCtx.cc.

References a, a_len, IceGrid.com, com, count, delta, IceGrid.dx, IceGrid.dy, e, fstart, IceGrid.Lbz, grid_info.Lx, IceGrid.Lx, grid_info.Ly, IceGrid.Ly, Ly, IceGrid.Lz, no_regrid_bedrock, no_regrid_ice, nz, nzb, grid_info.print(), IceGrid.rank, rank, report_range, start, grid_info.t_len, grid_info.time, verbPrintf(), check_stationarity.X, grid_info.x0, IceGrid.x0, grid_info.x_len, grid_info.x_max, grid_info.x_min, IceGrid.xm, IceGrid.xs, grid_info.y0, IceGrid.y0, grid_info.y_len, grid_info.y_max, grid_info.y_min, IceGrid.ym, IceGrid.ys, grid_info.z_len, grid_info.z_max, grid_info.zb_len, grid_info.zb_min, zblevs, and zlevs.