Routines for n-dimensional arrays. More...

Data Structures
struct	array
	N-dimensional array. More...

Typedefs
typedef struct array	Array

Functions
int	array_from_buffer (Array *A, int ndim, int shape, char dtype, char order, void *data)
	Init array from given buffer. More...

int	array_new (Array *A, int ndim, int shape, char dtype, char order)
	Init array object and allocate memory for its buffer. More...

int	array_new_like (Array *A, Array B)
	Init new array object with the shape, dtype and order of other. More...

int	array_new_copy (Array *A, Array B, char order)
	Init new array object as copy of other with given data layout. More...

void	array_free (Array *A)

void	array_info (Array *A)

void	array_print (Array *A)

int	array_to_matrix (Array *A, char kind)

int	array_trans_inplace (Array *A)
	Transpose `A` by changing shape, stride and order. More...

int	array_dot (Array A, Array B, Array **C)
	GEMM for `A` and `B` into `C`. More...

int	array_SVD (Array A, Array U, Array S, Array *V)
	Compute SVD of a given 2-dimensional `A`. More...

int	SVD_get_rank (Array S, double tol, char type, int rank)
	Returns rank by given singular values `S`, tolerance and type of norm. More...

int	array_scale (Array A, char kind, Array S)
	Apply row or column scaling to A. More...

int	array_diff (Array A, Array B, double *result)
	Measure Frobenius error of approximation of `A` by `B`. More...

int	array_norm (Array A, double result)
	Measure Frobenius norm of `A`. More...

int	array_convert (Array *A, Array B, char dtype)
	Create `A` as a copy of `B` with different data type (precision). More...

int	array_cholesky (Array *A, char uplo)
	Cholesky factorization for `A`. More...

Detailed Description

Routines for n-dimensional arrays.

Typedef Documentation

◆ Array

typedef struct array Array

Typedef for array

Function Documentation

◆ array_cholesky()

int array_cholesky	(	Array *	A,
		char	uplo
	)

Cholesky factorization for A.

Parameters

[in,out]	A	Pointer to Array object.
[in]	uplo	Store result in upper or in lower part of `A`.

Returns: Error code STARSH_ERRNO.

◆ array_convert()

int array_convert	(	Array **	A,
		Array *	B,
		char	dtype
	)

Create A as a copy of B with different data type (precision).

Parameters

[out]	A	Address of pointer to Array object.
[in]	B	Pointer to Array object.
[in]	dtype	New data type (precision).

Returns: Error code STARSH_ERRNO.

◆ array_diff()

int array_diff	(	Array *	A,
		Array *	B,
		double *	result
	)

Measure Frobenius error of approximation of A by B.

Parameters

[in]	A	Pointer to Array object.
[in]	B	Pointer to Array object.
[out]	result	Address of result.

Returns: Error code STARSH_ERRNO.

◆ array_dot()

int array_dot	(	Array *	A,
		Array *	B,
		Array **	C
	)

GEMM for A and B into C.

Multiplication is performed by last dimension of A and first dimension of B. These dimensions, data types and ordering of A and B should be equal.

Parameters

[in]	A	Pointer to Array object.
[in]	B	Pointer to Array object.
[out]	C	Address of pointer to Array object.

Returns: Error code STARSH_ERRNO.

◆ array_free()

void array_free ( Array * A )

Free array object.

◆ array_from_buffer()

int array_from_buffer	(	Array **	A,
		int	ndim,
		int *	shape,
		char	dtype,
		char	order,
		void *	data
	)

Init array from given buffer.

Memory is automatically allocated, do not forget to free it by array_free().

Parameters

[out]	A	Output array.
[in]	ndim	Number of dimensions of array, `ndim > 1`.
[in]	shape	Size of array in each dimension.
[in]	dtype	Precision of array element.
[in]	order	Fortran (column-major) or C (row-major) order.
[in]	data	Data buffer.

Returns: Error code STARSH_ERRNO.

◆ array_info()

void array_info ( Array * A )

Print all the data from array object A.

◆ array_new()

int array_new	(	Array **	A,
		int	ndim,
		int *	shape,
		char	dtype,
		char	order
	)

Init array object and allocate memory for its buffer.

Parameters

[out]	A	Address of pointer to array object.
[in]	ndim	Number of dimensions of array, `ndim > 1`.
[in]	shape	Size of array in each dimension.
[in]	dtype	Precision of array element.
[in]	order	Fortran (column-major) or C (row-major) order.

Returns: Error code STARSH_ERRNO.

◆ array_new_copy()

int array_new_copy	(	Array **	A,
		Array *	B,
		char	order
	)

Init new array object as copy of other with given data layout.

Parameters

[out]	A	Address of pointer to array object.
[in]	B	Other array object.
[in]	order	Fortran (column-major) or C (row-major) order.

Returns: Error code STARSH_ERRNO.

◆ array_new_like()

int array_new_like	(	Array **	A,
		Array *	B
	)

Init new array object with the shape, dtype and order of other.

Parameters

[out]	A	Address of pointer to array object.
[in]	B	Other array object.

Returns: Error code STARSH_ERRNO.

◆ array_norm()

int array_norm	(	Array *	A,
		double *	result
	)

Measure Frobenius norm of A.

Parameters

[in]	A	Pointer to Array object.
[out]	result	Address of result.

Returns: Error code STARSH_ERRNO.

◆ array_print()

void array_print ( Array * A )

Print array. Different rows of A are printed on different output rows.

◆ array_scale()

int array_scale	(	Array *	A,
		char	kind,
		Array *	S
	)

Apply row or column scaling to A.

Parameters

[in,out]	A	Pointer to Array object.
[in]	kind	Row or column scaling.
[in]	S	Scaling factors in Array object.

Returns: Error code STARSH_ERRNO.

◆ array_SVD()

int array_SVD	(	Array *	A,
		Array **	U,
		Array **	S,
		Array **	V
	)

Compute SVD of a given 2-dimensional A.

Parameters

[in]	A	Pointer to Array object.
[out]	U	Address of pointer to Array object.
[out]	S	Address of pointer to Array object.
[out]	V	Address of pointer to Array object.

Returns: Error code STARSH_ERRNO.

◆ array_to_matrix()

int array_to_matrix	(	Array *	A,
		char	kind
	)

Convert array to 2-dimensional matrix by glueing different dimensions.

Convert N-dimensional A to 2-dimensional A (matrix) by collapsing dimensions. This collapse can be assumed as attempt to look at A as at a matrix with long rows (kind == 'R') or long columns (kind == 'C'). If kind is 'R', dimensions from 1 to the last are collapsed into columns. If kind is 'C', dimensions from 0 to the last minus one are collapsed into rows. Example: A of shape (2,3,4,5) will be collapsed to A of shape (2,60) if kind is 'R' or to A of shape (24,5) if kind is 'C'.

Parameters

[in,out]	A	Pointer to Array object.
[in]	kind	kind of collapse (into long rows or into long columns).

Returns: Error code STARSH_ERRNO.

◆ array_trans_inplace()

int array_trans_inplace ( Array * A )

Transpose A by changing shape, stride and order.

Parameters

[in,out] A Pointer to Array object.

Returns: Error code STARSH_ERRNO.

◆ SVD_get_rank()

int SVD_get_rank	(	Array *	S,
		double	tol,
		char	type,
		int *	rank
	)

Returns rank by given singular values S, tolerance and type of norm.

Parameters

[in]	S	Pointer to Array object.
[in]	tol	Relative error threshold.
[in]	type	Frobenius or spectral norm.
[out]	rank	Address of rank.

Returns: Error code STARSH_ERRNO.

Data Structures

Typedefs

Functions

Detailed Description

Typedef Documentation

◆ Array

Function Documentation

◆ array_cholesky()

◆ array_convert()

◆ array_diff()

◆ array_dot()

◆ array_free()

◆ array_from_buffer()

◆ array_info()

◆ array_new()

◆ array_new_copy()

◆ array_new_like()

◆ array_norm()

◆ array_print()

◆ array_scale()

◆ array_SVD()

◆ array_to_matrix()

◆ array_trans_inplace()

◆ SVD_get_rank()