Multi-dimensional indexing

Detailed Description

Map multi-dimensional indices to a linear/flat index and vice versa and calculate the resulting memory layout when taking slices of already existing arrays/views.

Classes
class	nda::idx_map< Rank, StaticExtents, StrideOrder, LayoutProp >
	Layout that specifies how to map multi-dimensional indices to a linear/flat index. More...

Functions
template<int R, uint64_t SE, uint64_t SO, layout_prop_e LP, typename... Args>
__inline__ decltype(auto)	nda::slice_static::slice_idx_map (idx_map< R, SE, SO, LP > const &idxm, Args const &...args)
	Determine the resulting nda::idx_map when taking a slice of a given nda::idx_map.

Variables
template<int Rank>
constexpr uint64_t	nda::C_stride_order = nda::encode(nda::permutations::identity<Rank>())
	C/Row-major stride order.
template<int Rank>
constexpr uint64_t	nda::Fortran_stride_order = nda::encode(nda::permutations::reverse_identity<Rank>())
	Fortran/Column-major stride order.

Function Documentation

slice_idx_map()

template<int R, uint64_t SE, uint64_t SO, layout_prop_e LP, typename... Args>

__inline__ decltype(auto) nda::slice_static::slice_idx_map	(	idx_map< R, SE, SO, LP > const &	idxm,
		Args const &...	args )

#include <nda/layout/slice_static.hpp>

Determine the resulting nda::idx_map when taking a slice of a given nda::idx_map.

Let n_args be the number of given long, nda::range, nda::range::all_t or nda::ellipsis arguments.

The rank R' of the resulting slice is determined by the rank R of the original nda::idx_map and the number n_long of long arguments, i.e. R' = R - n_long.

The number of allowed nda::ellipsis objects is restricted to at most one. If an nda::ellipsis object is present and n_args <= R, the ellipsis is expanded in terms of nda::range::all_t objects to cover the remaining R - n_args + 1 dimensions. Otherwise, the ellipsis is ignored.

After ellipsis expansion, the only arguments contributing to the slice are nda::range and nda::range::all_t objects. Together with the original nda::idx_map, they determine the resulting nda::idx_map.

Template Parameters

R	Rank of the original nda::idx_map.
SE	Static extents of the original nda::idx_map.
SO	Stride order of the original nda::idx_map.
LP	Layout properties of the original nda::idx_map.
Args	Given argument types.

Parameters

idxm	Original nda::idx_map.
args	Arguments consisting of long, nda::range, nda::range::all_t or nda::ellipsis objects.

Returns

Resulting nda::idx_map of the slice.

Definition at line 325 of file slice_static.hpp.

Variable Documentation

C_stride_order

template<int Rank>

uint64_t nda::C_stride_order = nda::encode(nda::permutations::identity<Rank>())

constexpr

#include <nda/layout/idx_map.hpp>

C/Row-major stride order.

The last dimension varies the fastest, the first dimension the slowest.

Template Parameters

Rank	Number of dimensions.

Definition at line 65 of file idx_map.hpp.

Fortran_stride_order

template<int Rank>

uint64_t nda::Fortran_stride_order = nda::encode(nda::permutations::reverse_identity<Rank>())

constexpr

#include <nda/layout/idx_map.hpp>

Fortran/Column-major stride order.

The first dimension varies the fastest, the last dimension the slowest.

Template Parameters

Rank	Number of dimensions.

Definition at line 57 of file idx_map.hpp.