Real Valued Single Precision

Admissibility condition
hpro_ admcond_t hpro_s_	admcond_alg (const hpro_ adm_t crit, const double eta, const hpro_s_ matrix_t S, const hpro_ permutation_t row_perm, const hpro_ permutation_t col_perm, int *info)

Cluster Trees
Root of a cluster tree with stored index permutations.
hpro_ clustertree_t hpro_s_	clt_build_bsp_nd (const hpro_ coord_t coord, const hpro_s_ matrix_t S, const hpro_ bsp_t bsptype, const unsigned int nmin, int *info)
hpro_ clustertree_t hpro_s_	clt_build_alg (const hpro_s_ matrix_t S, const hpro_ alg_t algtype, const unsigned int nmin, int *info)
hpro_ clustertree_t hpro_s_	clt_build_alg_nd (const hpro_s_ matrix_t S, const hpro_ alg_t algtype, const unsigned int nmin, int *info)
hpro_ clustertree_t hpro_s_	clt_build_alg_part (const hpro_s_ matrix_t S, const unsigned int *partition, const hpro_ alg_t algtype, const unsigned int nmin, int *info)

Vectors
size_t hpro_s_	vector_size (const hpro_s_ vector_t v, int *info)
hpro_s_ vector_t hpro_s_	vector_copy (const hpro_s_ vector_t v, int *info)
size_t hpro_s_	vector_bytesize (const hpro_s_ vector_t v, int *info)
void hpro_s_	vector_free (hpro_s_ vector_t v, int *info)
float hpro_s_	vector_entry_get (const hpro_s_ vector_t x, const size_t i, int *info)
void hpro_s_	vector_entry_set (const hpro_s_ vector_t x, const size_t i, const float f, int *info)
hpro_s_ vector_t hpro_s_	vector_build (const size_t size, int *info)
void hpro_s_	vector_import (hpro_s_ vector_t v, const float *arr, int *info)
void hpro_s_	vector_export (const hpro_s_ vector_t v, float *arr, int *info)
void hpro_s_	vector_fill (hpro_s_ vector_t x, const float f, int *info)
void hpro_s_	vector_fill_rand (hpro_s_ vector_t x, int *info)
void hpro_s_	vector_assign (hpro_s_ vector_t y, const hpro_s_ vector_t x, int *info)
void hpro_s_	vector_scale (hpro_s_ vector_t x, const float f, int *info)
void hpro_s_	vector_axpy (const float alpha, const hpro_s_ vector_t x, hpro_s_ vector_t y, int *info)
float hpro_s_	vector_dot (const hpro_s_ vector_t x, const hpro_s_ vector_t y, int *info)
float hpro_s_	vector_norm2 (const hpro_s_ vector_t x, int *info)
float hpro_s_	vector_norm_inf (const hpro_s_ vector_t x, int *info)
void hpro_s_	vector_permute (hpro_s_ vector_t v, const hpro_ permutation_t perm, int *info)

Matrices
void hpro_s_	linearoperator_free (hpro_s_ linearoperator_t A, int *info)
hpro_s_ vector_t hpro_s_	linearoperator_range_vector (const hpro_s_ linearoperator_t A, int *info)
hpro_s_ vector_t hpro_s_	linearoperator_domain_vector (const hpro_s_ linearoperator_t A, int *info)
hpro_s_ linearoperator_t hpro_s_	perm_linearoperator (const hpro_ permutation_t P, const hpro_s_ linearoperator_t A, const hpro_ permutation_t R, int *info)
void hpro_s_	linearoperator_apply (const hpro_s_ linearoperator_t A, const hpro_s_ vector_t x, hpro_s_ vector_t y, const hpro_ matop_t matop, int *info)
void hpro_s_	linearoperator_apply_add (const float alpha, const hpro_s_ linearoperator_t A, const hpro_s_ vector_t x, hpro_s_ vector_t y, const hpro_ matop_t matop, int *info)
size_t hpro_s_	matrix_rows (const hpro_s_ matrix_t A, int *info)
size_t hpro_s_	matrix_cols (const hpro_s_ matrix_t A, int *info)
hpro_s_ matrix_t hpro_s_	matrix_copy (const hpro_s_ matrix_t A, int *info)
hpro_s_ matrix_t hpro_s_	matrix_copy_acc (const hpro_s_ matrix_t A, const hpro_ acc_t acc, const int coarsen, int *info)
void hpro_s_	matrix_copyto (const hpro_s_ matrix_t A, hpro_s_ matrix_t B, int *info)
void hpro_s_	matrix_copyto_acc (const hpro_s_ matrix_t A, hpro_s_ matrix_t B, const hpro_ acc_t acc, const int coarsen, int *info)
hpro_s_ matrix_t hpro_s_	matrix_copy_blockdiag (const hpro_s_ matrix_t A, const unsigned int lvl, const size_t blocksize, int *info)
hpro_s_ matrix_t hpro_s_	matrix_copy_blockdiag_acc (const hpro_s_ matrix_t A, const hpro_ acc_t acc, const unsigned int lvl, const size_t blocksize, int *info)
hpro_s_ matrix_t hpro_s_	matrix_copy_nearfield (const hpro_s_ matrix_t A, const int without_farfield, int *info)
void hpro_s_	matrix_restrict_nearfield (hpro_s_ matrix_t A, const int delete_farfield, int *info)
void hpro_s_	matrix_nearfield_to_crs (const hpro_s_ matrix_t A, size_t *nnz, int **rowptr, int **colind, float **coeffs, int *info)
size_t hpro_s_	matrix_bytesize (const hpro_s_ matrix_t A, int *info)
void hpro_s_	matrix_free (hpro_s_ matrix_t A, int *info)
int hpro_s_	matrix_is_sym (const hpro_s_ matrix_t A, int *info)
int hpro_s_	matrix_is_herm (const hpro_s_ matrix_t A, int *info)
int hpro_s_	matrix_is_complex (const hpro_s_ matrix_t A, int *info)
float hpro_s_	matrix_entry_get (const hpro_s_ matrix_t A, const size_t i, const size_t j, int *info)
hpro_s_ vector_t hpro_s_	matrix_row_vector (const hpro_s_ matrix_t A, int *info)
hpro_s_ vector_t hpro_s_	matrix_col_vector (const hpro_s_ matrix_t A, int *info)
hpro_s_ matrix_t hpro_s_	matrix_import_crs (const size_t rows, const size_t cols, const size_t nnz, const int *rowptr, const int *colind, const float *coeffs, const int sym, int *info)
hpro_s_ matrix_t hpro_s_	matrix_import_ccs (const size_t rows, const size_t cols, const size_t nnz, const int *colptr, const int *rowind, const float *coeffs, const int sym, int *info)
hpro_s_ matrix_t hpro_s_	matrix_import_dense (const size_t rows, const size_t cols, const float *D, const int sym, int *info)
hpro_s_ matrix_t hpro_s_	matrix_build_sparse (const hpro_ blockclustertree_t bct, const hpro_s_ matrix_t S, const hpro_ acc_t acc, int *info)
hpro_s_ matrix_t hpro_s_	matrix_build_dense (const hpro_ blockclustertree_t bct, const hpro_s_ matrix_t D, const hpro_ lrapx_t lrapx, const hpro_ acc_t acc, int *info)
hpro_s_ matrix_t hpro_s_	matrix_build_coeff (const hpro_ blockclustertree_t bct, const hpro_s_ coeff_t f, void *arg, const hpro_ lrapx_t lrapx, const hpro_ acc_t acc, const int sym, int *info)
hpro_s_ matrix_t hpro_s_	matrix_build_identity (const hpro_ blockclustertree_t bct, int *info)
hpro_s_ matrix_t hpro_s_	matrix_assemble_block (const size_t brows, const size_t bcols, hpro_s_ matrix_t *submat, const int copy, int *info)
hpro_ permutation_t hpro_s_	matrix_row_perm_i2e (const hpro_s_ matrix_t A, int *info) HPRO_DEPRECATED
hpro_ permutation_t hpro_s_	matrix_row_perm_e2i (const hpro_s_ matrix_t A, int *info) HPRO_DEPRECATED
void hpro_s_	matrix_permute (hpro_s_ matrix_t A, const hpro_ permutation_t row_perm, const hpro_ permutation_t col_perm, int *info)
hpro_s_ linearoperator_t hpro_s_	perm_matrix (const hpro_ permutation_t P, const hpro_s_ matrix_t A, const hpro_ permutation_t R, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_product2 (const hpro_ matop_t matop1, const hpro_s_ linearoperator_t A1, const hpro_ matop_t matop2, const hpro_s_ linearoperator_t A2, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_product3 (const hpro_ matop_t matop1, const hpro_s_ linearoperator_t A1, const hpro_ matop_t matop2, const hpro_s_ linearoperator_t A2, const hpro_ matop_t matop3, const hpro_s_ linearoperator_t A3, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_sum2 (const float alpha1, const hpro_ matop_t matop1, const hpro_s_ linearoperator_t A1, const float alpha2, const hpro_ matop_t matop2, const hpro_s_ linearoperator_t A2, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_sum3 (const float alpha1, const hpro_ matop_t matop1, const hpro_s_ linearoperator_t A1, const float alpha2, const hpro_ matop_t matop2, const hpro_s_ linearoperator_t A2, const float alpha3, const hpro_ matop_t matop3, const hpro_s_ linearoperator_t A3, int *info)
hpro_s_ matrix_t hpro_s_	matrix_to_dense (const hpro_s_ matrix_t A, int *info)
hpro_s_ matrix_t hpro_s_	matrix_to_rank (const hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ matrix_t hpro_s_	matrix_approx_rank (const hpro_s_ matrix_t A, const hpro_ lrapx_t lrapx, const hpro_ acc_t acc, int *info)
void hpro_s_	matrix_make_symmetric (hpro_s_ matrix_t A, int *info)
void hpro_s_	matrix_make_hermitian (hpro_s_ matrix_t A, int *info)
void hpro_s_	matrix_print_ps (const hpro_s_ matrix_t A, const char *filename, const int options, int *info)

Algebra
void hpro_s_	matrix_mulvec (const float alpha, const hpro_s_ matrix_t A, const hpro_s_ vector_t x, const float beta, hpro_s_ vector_t y, const hpro_ matop_t matop, int *info)
void hpro_s_	matrix_transpose (const hpro_s_ matrix_t A, int *info)
void hpro_s_	matrix_scale (const float f, const hpro_s_ matrix_t A, int *info)
void hpro_s_	matrix_add (const float alpha, const hpro_s_ matrix_t A, const float beta, hpro_s_ matrix_t B, const hpro_ acc_t acc, int *info)
void hpro_s_	matrix_mul (const float alpha, const hpro_ matop_t matop_A, const hpro_s_ matrix_t A, const hpro_ matop_t matop_B, const hpro_s_ matrix_t B, const float beta, hpro_s_ matrix_t C, const hpro_ acc_t acc, int *info)
hpro_s_ matrix_t hpro_s_	matrix_inv (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ vector_t hpro_s_	matrix_inv_diag (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_factorise (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_factorise_inv (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_lu (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_ldl (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_ll (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_waz (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_lu_inv (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_ldl_inv (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
hpro_s_ linearoperator_t hpro_s_	matrix_ll_inv (hpro_s_ matrix_t A, const hpro_ acc_t acc, int *info)
void hpro_s_	matrix_solve_lower_left (const hpro_s_ matrix_t L, const hpro_ matop_t op_L, hpro_s_ matrix_t A, const hpro_ acc_t acc, const int pointwise, const int unit_diag, int *info)
void hpro_s_	matrix_solve_lower_right (hpro_s_ matrix_t A, const hpro_s_ matrix_t L, const hpro_ matop_t op_L, const hpro_ acc_t acc, const int pointwise, const int unit_diag, int *info)
void hpro_s_	matrix_solve_upper_right (hpro_s_ matrix_t A, const hpro_s_ matrix_t U, const hpro_ acc_t acc, const int pointwise, const int unit_diag, int *info)
void hpro_s_	matrix_solve_diag_left (const hpro_s_ matrix_t D, const hpro_ matop_t op_D, hpro_s_ matrix_t A, const hpro_ acc_t acc, const int pointwise, int *info)
void hpro_s_	matrix_solve_diag_right (hpro_s_ matrix_t A, const hpro_s_ matrix_t D, const hpro_ matop_t op_D, const hpro_ acc_t acc, const int pointwise, int *info)
void hpro_s_	matrix_add_identity (hpro_s_ matrix_t A, const float lambda, int *info)
float hpro_s_	matrix_norm_frobenius (const hpro_s_ matrix_t A, int *info)
float hpro_s_	matrix_norm_frobenius_diff (const hpro_s_ matrix_t A, const hpro_s_ matrix_t B, int *info)
float hpro_s_	matrix_norm_spectral (const hpro_s_ matrix_t A, int *info)
float hpro_s_	matrix_norm_spectral_inv (const hpro_s_ matrix_t A, int *info)
float hpro_s_	matrix_norm_spectral_diff (const hpro_s_ matrix_t A, const hpro_s_ matrix_t B, int *info)
float hpro_s_	matrix_norm_inv_approx (const hpro_s_ matrix_t A, const hpro_s_ matrix_t B, int *info)
float hpro_s_	linearoperator_norm_spectral (const hpro_s_ linearoperator_t A, int *info)
float hpro_s_	linearoperator_norm_spectral_inv (const hpro_s_ linearoperator_t A, int *info)
float hpro_s_	linearoperator_norm_spectral_diff (const hpro_s_ linearoperator_t A, const hpro_s_ linearoperator_t B, int *info)
float hpro_s_	linearoperator_norm_inv_approx (const hpro_s_ linearoperator_t A, const hpro_s_ linearoperator_t B, int *info)

Solver
void hpro_s_	solver_solve (const hpro_ solver_t solver, const hpro_s_ linearoperator_t A, hpro_s_ vector_t x, const hpro_s_ vector_t b, const hpro_s_ linearoperator_t W, hpro_ solve_info_t *solve_info, int *info)

Input/Output
hpro_s_ matrix_t hpro_s_	hformat_load_matrix (const char *filename, int *info)
hpro_s_ linearoperator_t hpro_s_	hformat_load_linearoperator (const char *filename, int *info)
hpro_s_ vector_t hpro_s_	hformat_load_vector (const char *filename, int *info)
void hpro_s_	hformat_save_linearoperator (const hpro_s_ linearoperator_t A, const char *filename, int *info)
void hpro_s_	hformat_save_matrix (const hpro_s_ matrix_t A, const char *filename, int *info)
void hpro_s_	hformat_save_vector (const hpro_s_ vector_t v, const char *filename, int *info)
hpro_s_ matrix_t hpro_s_	samg_load_matrix (const char *filename, int *info)
void hpro_s_	samg_save_matrix (const hpro_s_ matrix_t A, const char *filename, int *info)
hpro_s_ vector_t hpro_s_	samg_load_vector (const char *filename, int *info)
void hpro_s_	samg_save_vector (const hpro_s_ vector_t x, const char *filename, int *info)
hpro_s_ matrix_t hpro_s_	matlab_load_matrix (const char *filename, const char *matname, int *info)
void hpro_s_	matlab_save_matrix (const hpro_s_ matrix_t M, const char *filename, const char *matname, int *info)
hpro_s_ vector_t hpro_s_	matlab_load_vector (const char *filename, const char *vecname, int *info)
void hpro_s_	matlab_save_vector (const hpro_s_ vector_t v, const char *filename, const char *vecname, int *info)
hpro_s_ matrix_t hpro_s_	pltmg_load_matrix (const char *filename, int *info)
hpro_s_ matrix_t hpro_s_	hb_load_matrix (const char *filename, int *info)
void hpro_s_	hb_save_matrix (const hpro_s_ matrix_t M, const char *filename, int *info)
hpro_s_ matrix_t hpro_s_	mm_load_matrix (const char *filename, int *info)
hpro_s_ matrix_t hpro_s_	load_matrix (const char *filename, int *info)
hpro_s_ vector_t hpro_s_	load_vector (const char *filename, int *info)

Detailed Description

This modules defines single precision real valued functions for the usage of 𝖧𝖫𝖨𝖡𝗉𝗋𝗈 in C programs.

Function Documentation

admcond_alg()

hpro_ admcond_t hpro_s_ admcond_alg	(	const hpro_ adm_t	crit,
		const double	eta,
		const hpro_s_ matrix_t	S,
		const hpro_ permutation_t	row_perm,
		const hpro_ permutation_t	col_perm,
		int *	info
	)

create admissibility condition based on algebraic connectivity in S

• row_perm and col_perm define the (optional) mapping between internal to external ordering for row and column index sets, respectively (see hlib_ct_perm_i2e); set to NULL if not present

clt_build_alg()

hpro_ clustertree_t hpro_s_ clt_build_alg	(	const hpro_s_ matrix_t	S,
		const hpro_ alg_t	algtype,
		const unsigned int	nmin,
		int *	info
	)

build clustertree via algebraic clustering based on given sparse matrix S and partitioning type algtype.

clt_build_alg_nd()

hpro_ clustertree_t hpro_s_ clt_build_alg_nd	(	const hpro_s_ matrix_t	S,
		const hpro_ alg_t	algtype,
		const unsigned int	nmin,
		int *	info
	)

build clustertree via algebraic clustering with nested dissection based on given sparse matrix S and partitioning type algtype.

clt_build_alg_part()

hpro_ clustertree_t hpro_s_ clt_build_alg_part	(	const hpro_s_ matrix_t	S,
		const unsigned int *	partition,
		const hpro_ alg_t	algtype,
		const unsigned int	nmin,
		int *	info
	)

build clustertree via algebraic clustering based on matrix S and use a predefined partition for first step

• partition must have size "rows(S)" and contain ids ∈ [0,k-1], where k is the total number of partitions
• index "i" will go into group partition[i]

clt_build_bsp_nd()

hpro_ clustertree_t hpro_s_ clt_build_bsp_nd	(	const hpro_ coord_t	coord,
		const hpro_s_ matrix_t	S,
		const hpro_ bsp_t	bsptype,
		const unsigned int	nmin,
		int *	info
	)

build cluster tree using binary space partitioning and nested dissection based on given sparse matrix S

• for period see HPRO_S_PF(clt_build_bsp)

Parameters

coord	coordinates of indices
S	sparse mat. for connectivity
bsptype	partitioning type
nmin	minimal cluster size

hb_load_matrix()

hpro_s_ matrix_t hpro_s_ hb_load_matrix	(	const char *	filename,
		int *	info
	)

read matrix from file filename

hb_save_matrix()

void hpro_s_ hb_save_matrix	(	const hpro_s_ matrix_t	M,
		const char *	filename,
		int *	info
	)

save matrix M to file filename

hformat_load_linearoperator()

hpro_s_ linearoperator_t hpro_s_ hformat_load_linearoperator	(	const char *	filename,
		int *	info
	)

read matrix from file filename

hformat_load_matrix()

hpro_s_ matrix_t hpro_s_ hformat_load_matrix	(	const char *	filename,
		int *	info
	)

read matrix from file filename

hformat_load_vector()

hpro_s_ vector_t hpro_s_ hformat_load_vector	(	const char *	filename,
		int *	info
	)

read vector from file filename

hformat_save_linearoperator()

void hpro_s_ hformat_save_linearoperator	(	const hpro_s_ linearoperator_t	A,
		const char *	filename,
		int *	info
	)

save linear operator A to file filename

hformat_save_matrix()

void hpro_s_ hformat_save_matrix	(	const hpro_s_ matrix_t	A,
		const char *	filename,
		int *	info
	)

save matrix A to file filename

hformat_save_vector()

void hpro_s_ hformat_save_vector	(	const hpro_s_ vector_t	v,
		const char *	filename,
		int *	info
	)

save vector v to file filename

linearoperator_apply()

void hpro_s_ linearoperator_apply	(	const hpro_s_ linearoperator_t	A,
		const hpro_s_ vector_t	x,
		hpro_s_ vector_t	y,
		const hpro_ matop_t	matop,
		int *	info
	)

computes \( y := op(A) \cdot x \)

linearoperator_apply_add()

void hpro_s_ linearoperator_apply_add	(	const float	alpha,
		const hpro_s_ linearoperator_t	A,
		const hpro_s_ vector_t	x,
		hpro_s_ vector_t	y,
		const hpro_ matop_t	matop,
		int *	info
	)

computes \( y := y + \alpha op(A) \cdot x \)

linearoperator_domain_vector()

hpro_s_ vector_t hpro_s_ linearoperator_domain_vector	(	const hpro_s_ linearoperator_t	A,
		int *	info
	)

create vectors from the domain of the linear operator A

linearoperator_free()

void hpro_s_ linearoperator_free	(	hpro_s_ linearoperator_t	A,
		int *	info
	)

free resources coupled with linear operator A

linearoperator_norm_inv_approx()

float hpro_s_ linearoperator_norm_inv_approx	(	const hpro_s_ linearoperator_t	A,
		const hpro_s_ linearoperator_t	B,
		int *	info
	)

compute inversion error |I-BA|_2, where B is supposed to be A^-1

linearoperator_norm_spectral()

float hpro_s_ linearoperator_norm_spectral	(	const hpro_s_ linearoperator_t	A,
		int *	info
	)

compute spectral norm of matrix

linearoperator_norm_spectral_diff()

float hpro_s_ linearoperator_norm_spectral_diff	(	const hpro_s_ linearoperator_t	A,
		const hpro_s_ linearoperator_t	B,
		int *	info
	)

compute relative spectral norm of A-B, e.g. |A-B|_2/|A|_2

linearoperator_norm_spectral_inv()

float hpro_s_ linearoperator_norm_spectral_inv	(	const hpro_s_ linearoperator_t	A,
		int *	info
	)

compute spectral norm of inverse matrix

linearoperator_range_vector()

hpro_s_ vector_t hpro_s_ linearoperator_range_vector	(	const hpro_s_ linearoperator_t	A,
		int *	info
	)

create vectors from the range of the linear operator A

load_matrix()

hpro_s_ matrix_t hpro_s_ load_matrix	(	const char *	filename,
		int *	info
	)

read matrix from file filename

load_vector()

hpro_s_ vector_t hpro_s_ load_vector	(	const char *	filename,
		int *	info
	)

read vector from file filename

matlab_load_matrix()

hpro_s_ matrix_t hpro_s_ matlab_load_matrix	(	const char *	filename,
		const char *	matname,
		int *	info
	)

read matrix named matname in Matlab V7 format from file filename; if matname == NULL or matname == "", the first matrix in filename will be read

matlab_load_vector()

hpro_s_ vector_t hpro_s_ matlab_load_vector	(	const char *	filename,
		const char *	vecname,
		int *	info
	)

read vector named vname from file filename in Matlab V7 format

matlab_save_matrix()

void hpro_s_ matlab_save_matrix	(	const hpro_s_ matrix_t	M,
		const char *	filename,
		const char *	matname,
		int *	info
	)

save matrix M named matname in Matlab V7 format to file filename

matlab_save_vector()

void hpro_s_ matlab_save_vector	(	const hpro_s_ vector_t	v,
		const char *	filename,
		const char *	vecname,
		int *	info
	)

save vector named vname to file filename in Matlab V7 format

matrix_add()

void hpro_s_ matrix_add	(	const float	alpha,
		const hpro_s_ matrix_t	A,
		const float	beta,
		hpro_s_ matrix_t	B,
		const hpro_ acc_t	acc,
		int *	info
	)

compute B := alpha A + beta B with block-wise precision acc

matrix_add_identity()

void hpro_s_ matrix_add_identity	(	hpro_s_ matrix_t	A,
		const float	lambda,
		int *	info
	)

compute A := A + lambda I

matrix_approx_rank()

hpro_s_ matrix_t hpro_s_ matrix_approx_rank	(	const hpro_s_ matrix_t	A,
		const hpro_ lrapx_t	lrapx,
		const hpro_ acc_t	acc,
		int *	info
	)

convert matrix to lowrank format using given approximation method with accuracy defined by acc

• currently only dense matrices supported

matrix_assemble_block()

hpro_s_ matrix_t hpro_s_ matrix_assemble_block	(	const size_t	brows,
		const size_t	bcols,
		hpro_s_ matrix_t *	submat,
		const int	copy,
		int *	info
	)

assemble brows × bcols block matrix out of sub blocks submat

• sub matrices have to be stored column wise in submat, e.g.,

  \begin{eqnarray*} A_{00} & A_{01} \\ A_{10} & A_{11} \end{eqnarray*}

  is stored { A_00, A_10, A_01, A_11 },
• if copy is non-zero, sub matrices are copied, otherwise used directly in new block matrix,
• index sets of sub matrices are adjusted to be consistent in new block matrix,
• if sub matrices are H-matrices, permutations from internal to external ordering (and vice versa) are also adjusted.

Parameters

brows	number of block rows
bcols	number of block columns
submat	array of sub matrices
copy	copy matrices if true

matrix_build_coeff()

hpro_s_ matrix_t hpro_s_ matrix_build_coeff	(	const hpro_ blockclustertree_t	bct,
		const hpro_s_ coeff_t	f,
		void *	arg,
		const hpro_ lrapx_t	lrapx,
		const hpro_ acc_t	acc,
		const int	sym,
		int *	info
	)

build H-matrix over block cluster tree bct of precision acc out of dense matrix given by a matrix coefficient function f by using a low-rank approximation for admissible blocks

Parameters

bct	block ct. defining indexsets
f	coeff. function defining mat.
arg	add. argument to coeff. fn.
lrapx	low-rank approximation method
acc	accuracy of the approximation
sym	if non-zero, matrix is sym.

matrix_build_dense()

hpro_s_ matrix_t hpro_s_ matrix_build_dense	(	const hpro_ blockclustertree_t	bct,
		const hpro_s_ matrix_t	D,
		const hpro_ lrapx_t	lrapx,
		const hpro_ acc_t	acc,
		int *	info
	)

build H-matrix over block cluster tree bct with contents defined by given dense matrix D with low-rank approximation method lrapx and block-wise accuracy acc

Parameters

bct	block ct. defining indexsets
D	dense mat. to be converted
lrapx	low-rank approximation method
acc	accuracy of the approximation

matrix_build_identity()

hpro_s_ matrix_t hpro_s_ matrix_build_identity	(	const hpro_ blockclustertree_t	bct,
		int *	info
	)

build H-matrix over block cluster tree bct representing identity

matrix_build_sparse()

hpro_s_ matrix_t hpro_s_ matrix_build_sparse	(	const hpro_ blockclustertree_t	bct,
		const hpro_s_ matrix_t	S,
		const hpro_ acc_t	acc,
		int *	info
	)

build H-matrix over block cluster tree bct with contents defined by given sparse matrix S

Parameters

bct	block ct. defining indexsets
S	sparse mat. to be converted
acc	accuracy of the approximation

matrix_bytesize()

size_t hpro_s_ matrix_bytesize	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return size of matrix in bytes

matrix_col_vector()

hpro_s_ vector_t hpro_s_ matrix_col_vector	(	const hpro_s_ matrix_t	A,
		int *	info
	)

create vectors matching the column indexsets of matrix A

matrix_cols()

size_t hpro_s_ matrix_cols	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return number of columns of matrix

matrix_copy()

hpro_s_ matrix_t hpro_s_ matrix_copy	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return copy of matrix A

matrix_copy_acc()

hpro_s_ matrix_t hpro_s_ matrix_copy_acc	(	const hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		const int	coarsen,
		int *	info
	)

return copy of matrix A with accuracy acc; if coarsen != 0, copy is coarsend

matrix_copy_blockdiag()

hpro_s_ matrix_t hpro_s_ matrix_copy_blockdiag	(	const hpro_s_ matrix_t	A,
		const unsigned int	lvl,
		const size_t	blocksize,
		int *	info
	)

copy blockdiagonal part of first lvl levels or blocks of size blocksize of matrix

matrix_copy_blockdiag_acc()

hpro_s_ matrix_t hpro_s_ matrix_copy_blockdiag_acc	(	const hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		const unsigned int	lvl,
		const size_t	blocksize,
		int *	info
	)

copy blockdiagonal part of first lvl levels or blocks of size blocksize of matrix with given accuracy acc

matrix_copy_nearfield()

hpro_s_ matrix_t hpro_s_ matrix_copy_nearfield	(	const hpro_s_ matrix_t	A,
		const int	without_farfield,
		int *	info
	)

copy nearfield part of matrix

• if without_farfield is non-zero, farfield matrices (e.g., low-rank) will be skipped, otherwise set to rank 0

matrix_copyto()

void hpro_s_ matrix_copyto	(	const hpro_s_ matrix_t	A,
		hpro_s_ matrix_t	B,
		int *	info
	)

copy matrix A to B

matrix_copyto_acc()

void hpro_s_ matrix_copyto_acc	(	const hpro_s_ matrix_t	A,
		hpro_s_ matrix_t	B,
		const hpro_ acc_t	acc,
		const int	coarsen,
		int *	info
	)

copy matrix A to B with accuracy acc; if coarsen != 0, B is coarsend

matrix_entry_get()

float hpro_s_ matrix_entry_get	(	const hpro_s_ matrix_t	A,
		const size_t	i,
		const size_t	j,
		int *	info
	)

get single entry at position (i,j) in matrix A

matrix_factorise()

hpro_s_ linearoperator_t hpro_s_ matrix_factorise	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

factorise matrix A up to block-wise precision acc

• depending on form of A, e.g. if unsymmetric, symmetric or hermitian, an appropriate factorisation method is chosen
• A will be overwritten by the factors
• the return value is a matrix which can be used to evaluate the factorisation, e.g. for matrix-vector mult. (this is not possible with A after factorisation as it holds just the data, albeit, it is still needed)

matrix_factorise_inv()

hpro_s_ linearoperator_t hpro_s_ matrix_factorise_inv	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

same as hlib_matrix_factorise but return matrix object for evaluation of inverse of A.

matrix_free()

void hpro_s_ matrix_free	(	hpro_s_ matrix_t	A,
		int *	info
	)

free resources coupled with matrix A

matrix_import_ccs()

hpro_s_ matrix_t hpro_s_ matrix_import_ccs	(	const size_t	rows,
		const size_t	cols,
		const size_t	nnz,
		const int *	colptr,
		const int *	rowind,
		const float *	coeffs,
		const int	sym,
		int *	info
	)

import given real sparse matrix in CCS format into internal sparse matrix format

• the data is copied into internal representation

Parameters

rows	number of rows and
cols	columns of the matrix
nnz	number of non-zero coeff.
colptr	array of column pointers
rowind	array of row indices
coeffs	array of matrix coefficients
sym	if non-zero, matrix is sym.

matrix_import_crs()

hpro_s_ matrix_t hpro_s_ matrix_import_crs	(	const size_t	rows,
		const size_t	cols,
		const size_t	nnz,
		const int *	rowptr,
		const int *	colind,
		const float *	coeffs,
		const int	sym,
		int *	info
	)

import given real sparse matrix in CRS format into internal sparse matrix format

• the data is copied into internal representation

Parameters

rows	number of rows and
cols	columns of the matrix
nnz	number of non-zero coeff.
rowptr	array of row pointers
colind	array of column indices
coeffs	array of matrix coefficients
sym	if non-zero, matrix is sym.

matrix_import_dense()

hpro_s_ matrix_t hpro_s_ matrix_import_dense	(	const size_t	rows,
		const size_t	cols,
		const float *	D,
		const int	sym,
		int *	info
	)

import given real dense matrix in column major format into internal dense matrix format

• the data is copied into internal representation

Parameters

rows	number of rows and
cols	columns of the matrix
D	accuracy of the approximation
sym	if non-zero, matrix is sym.

matrix_inv()

hpro_s_ matrix_t hpro_s_ matrix_inv	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

compute inverse of A with block-wise precision acc; A will be overwritten with A^-1 upon exit

• the return value is either A, or a new representation for the inverse of A (A is still needed then)

matrix_inv_diag()

hpro_s_ vector_t hpro_s_ matrix_inv_diag	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

compute diagonal of inverse of A with local accuracy acc

• A will be overwritten during computation

matrix_is_complex()

int hpro_s_ matrix_is_complex	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return 1 if matrix A is complex valued and 0 otherwise

matrix_is_herm()

int hpro_s_ matrix_is_herm	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return 1 if matrix A is hermitian, e.g. A = A^H

matrix_is_sym()

int hpro_s_ matrix_is_sym	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return 1 if matrix A is symmetric, e.g. A = A^T

matrix_ldl()

hpro_s_ linearoperator_t hpro_s_ matrix_ldl	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

LDL factorisation of matrix A up to block-wise precision acc

• A will be overwritten by L*D*L^T, which only allows matrix vector mult.
• in case of hermitian matrices, LDL^H is computed
• for return value see "hlib_matrix_lu"

matrix_ldl_inv()

hpro_s_ linearoperator_t hpro_s_ matrix_ldl_inv	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

LDL factorisation of matrix A up to block-wise precision acc

• A will be overwritten by L*D*L^T, which only allows matrix vector mult.
• in case of hermitian matrices, LDL^H is computed
• for return value see "hlib_matrix_lu_inv"

matrix_ll()

hpro_s_ linearoperator_t hpro_s_ matrix_ll	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

Cholesky factorisation of matrix A up to block-wise precision acc

• A will be overwritten by L*L^T, which only allows matrix vector mult.
• in case of hermitian matrices, L*L^H is computed
• for return value see "hlib_matrix_lu"

matrix_ll_inv()

hpro_s_ linearoperator_t hpro_s_ matrix_ll_inv	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

Cholesky factorisation of matrix A up to block-wise precision acc

• A will be overwritten by L*L^T, which only allows matrix vector mult.
• in case of hermitian matrices, L*L^H is computed
• for return value see "hlib_matrix_lu_inv"

matrix_lu()

hpro_s_ linearoperator_t hpro_s_ matrix_lu	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

LU factorise matrix A up to block-wise precision acc

• A will be overwritten by L*U, which only allows matrix vector mult.
• the return value is a matrix which can be used to evaluate LU, e.g. for matrix-vector multiplication (this is not possible with A after factorisation as it holds just the data, albeit, it is still needed)

matrix_lu_inv()

hpro_s_ linearoperator_t hpro_s_ matrix_lu_inv	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

LU factorise matrix A up to block-wise precision acc

• A will be overwritten by L*U, which only allows matrix vector mult.
• the return value is a matrix which can be used to evaluate (LU)^-1, e.g. for matrix-vector multiplication (this is not possible with A after factorisation as it holds just the data, albeit, it is still needed)

matrix_make_hermitian()

void hpro_s_ matrix_make_hermitian	(	hpro_s_ matrix_t	A,
		int *	info
	)

make given matrix hermitian by computing (A+A^H)/2, afterwards the upper right half is removed and the matrix has format "hermitian"

matrix_make_symmetric()

void hpro_s_ matrix_make_symmetric	(	hpro_s_ matrix_t	A,
		int *	info
	)

make given matrix symmetric by computing (A+A^T)/2, afterwards the upper right half is removed and the matrix has format "symmetric"

matrix_mul()

void hpro_s_ matrix_mul	(	const float	alpha,
		const hpro_ matop_t	matop_A,
		const hpro_s_ matrix_t	A,
		const hpro_ matop_t	matop_B,
		const hpro_s_ matrix_t	B,
		const float	beta,
		hpro_s_ matrix_t	C,
		const hpro_ acc_t	acc,
		int *	info
	)

compute C := alpha op(A) op(B) + beta C with block-wise precision acc

matrix_mulvec()

void hpro_s_ matrix_mulvec	(	const float	alpha,
		const hpro_s_ matrix_t	A,
		const hpro_s_ vector_t	x,
		const float	beta,
		hpro_s_ vector_t	y,
		const hpro_ matop_t	matop,
		int *	info
	)

compute y := alpha op(A) x + beta y, where op(A) is either A or A^T

Parameters

x	source vector of dimension columns(op(A))
y	dest. vector of dimension rows(op(A))
matop	matrix operation, e.g. transpose

matrix_nearfield_to_crs()

void hpro_s_ matrix_nearfield_to_crs	(	const hpro_s_ matrix_t	A,
		size_t *	nnz,
		int **	rowptr,
		int **	colind,
		float **	coeffs,
		int *	info
	)

returns nearfield part of given matrix in CRS format

• CRS is stored in nnz, rowptr, colind and coeffs

Parameters

nnz	number of non-zero coeff.
rowptr	array of row pointers
colind	array of column indices
coeffs	array of matrix coefficients

matrix_norm_frobenius()

float hpro_s_ matrix_norm_frobenius	(	const hpro_s_ matrix_t	A,
		int *	info
	)

compute Frobenius norm of matrix

matrix_norm_frobenius_diff()

float hpro_s_ matrix_norm_frobenius_diff	(	const hpro_s_ matrix_t	A,
		const hpro_s_ matrix_t	B,
		int *	info
	)

compute Frobenius norm of A-B

matrix_norm_inv_approx()

float hpro_s_ matrix_norm_inv_approx	(	const hpro_s_ matrix_t	A,
		const hpro_s_ matrix_t	B,
		int *	info
	)

compute inversion error |I-BA|_2, where B is supposed to be A^-1

matrix_norm_spectral()

float hpro_s_ matrix_norm_spectral	(	const hpro_s_ matrix_t	A,
		int *	info
	)

compute spectral norm of matrix

matrix_norm_spectral_diff()

float hpro_s_ matrix_norm_spectral_diff	(	const hpro_s_ matrix_t	A,
		const hpro_s_ matrix_t	B,
		int *	info
	)

compute relative spectral norm of A-B, e.g. |A-B|_2/|A|_2

matrix_norm_spectral_inv()

float hpro_s_ matrix_norm_spectral_inv	(	const hpro_s_ matrix_t	A,
		int *	info
	)

compute spectral norm of inverse matrix

matrix_permute()

void hpro_s_ matrix_permute	(	hpro_s_ matrix_t	A,
		const hpro_ permutation_t	row_perm,
		const hpro_ permutation_t	col_perm,
		int *	info
	)

reorder matrix A entries with given row and column permutations

• only supported for dense matrices
• NULL permutation will be handled as identity

matrix_print_ps()

void hpro_s_ matrix_print_ps	(	const hpro_s_ matrix_t	A,
		const char *	filename,
		const int	options,
		int *	info
	)

print matrix A in postscript format to file filename options (may be ORed) :

• HPRO_MATIO_SVD : print singular value decomposition in each block
• HPRO_MATIO_ENTRY : print each entry of matrix
  
• HPRO_MATIO_PATTERN : print sparsity pattern (non-zero entries)

Parameters

options

options defining output

matrix_product2()

hpro_s_ linearoperator_t hpro_s_ matrix_product2	(	const hpro_ matop_t	matop1,
		const hpro_s_ linearoperator_t	A1,
		const hpro_ matop_t	matop2,
		const hpro_s_ linearoperator_t	A2,
		int *	info
	)

represents product of matrices \(A_1 \cdot A_2\)

• all matrix objects are only referenced, not copied

matrix_product3()

hpro_s_ linearoperator_t hpro_s_ matrix_product3	(	const hpro_ matop_t	matop1,
		const hpro_s_ linearoperator_t	A1,
		const hpro_ matop_t	matop2,
		const hpro_s_ linearoperator_t	A2,
		const hpro_ matop_t	matop3,
		const hpro_s_ linearoperator_t	A3,
		int *	info
	)

represents product of matrices \(A_1 \cdot A_2 \cdot A3\)

• all matrix objects are only referenced, not copied

matrix_restrict_nearfield()

void hpro_s_ matrix_restrict_nearfield	(	hpro_s_ matrix_t	A,
		const int	delete_farfield,
		int *	info
	)

restrict given matrix to nearfield part

• if delete_farfield is non-zero, farfield matrices (e.g., low-rank) will be removed, otherwise set to rank 0

matrix_row_perm_e2i()

hpro_ permutation_t hpro_s_ matrix_row_perm_e2i	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return mapping from external (user) to internal (in H-matrix) ordering for rows and columns.

• returned object is reference to internal permutation: do NOT free
• if matrix has no permutation, the return value is NULL

matrix_row_perm_i2e()

hpro_ permutation_t hpro_s_ matrix_row_perm_i2e	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return mapping from internal (in H-matrix) to external (user) ordering for rows and columns.

• returned object is reference to internal permutation: do NOT free
• if matrix has no permutation, the return value is NULL

matrix_row_vector()

hpro_s_ vector_t hpro_s_ matrix_row_vector	(	const hpro_s_ matrix_t	A,
		int *	info
	)

create vectors matching the row indexsets of matrix A

matrix_rows()

size_t hpro_s_ matrix_rows	(	const hpro_s_ matrix_t	A,
		int *	info
	)

return number of rows of matrix

matrix_scale()

void hpro_s_ matrix_scale	(	const float	f,
		const hpro_s_ matrix_t	A,
		int *	info
	)

scale matrix A by factor f

matrix_solve_diag_left()

void hpro_s_ matrix_solve_diag_left	(	const hpro_s_ matrix_t	D,
		const hpro_ matop_t	op_D,
		hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		const int	pointwise,
		int *	info
	)

solve op(D) B = A with diagonal D and known A

• B will overwrite A
• if pointwise is true, evaluation is performed pointwise, otherwise blockwise, e.g. diagonal leaf matrices are considered full and not diagonal

matrix_solve_diag_right()

void hpro_s_ matrix_solve_diag_right	(	hpro_s_ matrix_t	A,
		const hpro_s_ matrix_t	D,
		const hpro_ matop_t	op_D,
		const hpro_ acc_t	acc,
		const int	pointwise,
		int *	info
	)

solve B op(D) = A with diagonal D and known A

• B will overwrite A
• if pointwise is true, evaluation is performed pointwise, otherwise blockwise, e.g. diagonal leaf matrices are considered full and not diagonal

matrix_solve_lower_left()

void hpro_s_ matrix_solve_lower_left	(	const hpro_s_ matrix_t	L,
		const hpro_ matop_t	op_L,
		hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		const int	pointwise,
		const int	unit_diag,
		int *	info
	)

solve L B = A with lower triangular L and known A

• B will overwrite A
• if pointwise is true, evaluation is performed pointwise, otherwise blockwise, e.g. diagonal leaf matrices are considered full and not triangular
• if unit_diag is true, the pointwise/blockwise diagonal is considered to be the identity

matrix_solve_lower_right()

void hpro_s_ matrix_solve_lower_right	(	hpro_s_ matrix_t	A,
		const hpro_s_ matrix_t	L,
		const hpro_ matop_t	op_L,
		const hpro_ acc_t	acc,
		const int	pointwise,
		const int	unit_diag,
		int *	info
	)

solve B op(L) = A with lower triangular L and known A

• B will overwrite A
• if pointwise is true, evaluation is performed pointwise, otherwise blockwise, e.g. diagonal leaf matrices are considered full and not triangular
• if unit_diag is true, the pointwise/blockwise diagonal is considered to be the identity

matrix_solve_upper_right()

void hpro_s_ matrix_solve_upper_right	(	hpro_s_ matrix_t	A,
		const hpro_s_ matrix_t	U,
		const hpro_ acc_t	acc,
		const int	pointwise,
		const int	unit_diag,
		int *	info
	)

solve B U = A with upper triangular U and known A

• B will overwrite A
• if pointwise is true, evaluation is performed pointwise, otherwise blockwise, e.g. diagonal leaf matrices are considered full and not triangular
• if unit_diag is true, the pointwise/blockwise diagonal is considered to be the identity

matrix_sum2()

hpro_s_ linearoperator_t hpro_s_ matrix_sum2	(	const float	alpha1,
		const hpro_ matop_t	matop1,
		const hpro_s_ linearoperator_t	A1,
		const float	alpha2,
		const hpro_ matop_t	matop2,
		const hpro_s_ linearoperator_t	A2,
		int *	info
	)

represents product of matrices \(A_1 + A_2\)

• all matrix objects are only referenced, not copied

matrix_sum3()

hpro_s_ linearoperator_t hpro_s_ matrix_sum3	(	const float	alpha1,
		const hpro_ matop_t	matop1,
		const hpro_s_ linearoperator_t	A1,
		const float	alpha2,
		const hpro_ matop_t	matop2,
		const hpro_s_ linearoperator_t	A2,
		const float	alpha3,
		const hpro_ matop_t	matop3,
		const hpro_s_ linearoperator_t	A3,
		int *	info
	)

represents product of matrices \(A_1 + A_2 + A3\)

• all matrix objects are only referenced, not copied

matrix_to_dense()

hpro_s_ matrix_t hpro_s_ matrix_to_dense	(	const hpro_s_ matrix_t	A,
		int *	info
	)

convert matrix to dense format

matrix_to_rank()

hpro_s_ matrix_t hpro_s_ matrix_to_rank	(	const hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

convert matrix to lowrank format using best approximation (SVD) with truncation defined by acc

matrix_transpose()

void hpro_s_ matrix_transpose	(	const hpro_s_ matrix_t	A,
		int *	info
	)

transpose matrix A

matrix_waz()

hpro_s_ linearoperator_t hpro_s_ matrix_waz	(	hpro_s_ matrix_t	A,
		const hpro_ acc_t	acc,
		int *	info
	)

Factorise matrix A into \(WAZ=I\), e.g., factorization of inverse of A

• A will be overwritten by \(W^{-1} Z^{-1}\).
• the return value is a linear operator which can be used to evaluate \(A^{-1}\), e.g. for matrix-vector multiplication (this is not possible with A after factorisation as it holds just the data, albeit, it is still needed)

mm_load_matrix()

hpro_s_ matrix_t hpro_s_ mm_load_matrix	(	const char *	filename,
		int *	info
	)

read matrix from file filename

perm_linearoperator()

hpro_s_ linearoperator_t hpro_s_ perm_linearoperator	(	const hpro_ permutation_t	P,
		const hpro_s_ linearoperator_t	A,
		const hpro_ permutation_t	R,
		int *	info
	)

represents permuted operator \(P \cdot A \cdot R\)

perm_matrix()

hpro_s_ linearoperator_t hpro_s_ perm_matrix	(	const hpro_ permutation_t	P,
		const hpro_s_ matrix_t	A,
		const hpro_ permutation_t	R,
		int *	info
	)

represents permuted matrix \(P \cdot A \cdot R\)

• all objects (P, A and R) are only referenced, not copied

pltmg_load_matrix()

hpro_s_ matrix_t hpro_s_ pltmg_load_matrix	(	const char *	filename,
		int *	info
	)

read matrix from file filename

samg_load_matrix()

hpro_s_ matrix_t hpro_s_ samg_load_matrix	(	const char *	filename,
		int *	info
	)

read matrix A from file "\a filename" in SAMG format

• expecting format definition in basename(filename)>.frm

samg_load_vector()

hpro_s_ vector_t hpro_s_ samg_load_vector	(	const char *	filename,
		int *	info
	)

read vector from file filename in SAMG format

• expecting format definition in basename(filename)>.frm

samg_save_matrix()

void hpro_s_ samg_save_matrix	(	const hpro_s_ matrix_t	A,
		const char *	filename,
		int *	info
	)

save matrix A to file "\a filename" in SAMG format

• also writing format definition to basename(filename)>.frm

samg_save_vector()

void hpro_s_ samg_save_vector	(	const hpro_s_ vector_t	x,
		const char *	filename,
		int *	info
	)

save vector to file filename in SAMG format

solver_solve()

void hpro_s_ solver_solve	(	const hpro_ solver_t	solver,
		const hpro_s_ linearoperator_t	A,
		hpro_s_ vector_t	x,
		const hpro_s_ vector_t	b,
		const hpro_s_ linearoperator_t	W,
		hpro_ solve_info_t *	solve_info,
		int *	info
	)

solve A x = b using solver with optional preconditioning, e.g. W A x = W b if W != NULL; information about the solving process is stored in solve_info

vector_assign()

void hpro_s_ vector_assign	(	hpro_s_ vector_t	y,
		const hpro_s_ vector_t	x,
		int *	info
	)

copy content of vector x to vector y x and y have to be of equal type

vector_axpy()

void hpro_s_ vector_axpy	(	const float	alpha,
		const hpro_s_ vector_t	x,
		hpro_s_ vector_t	y,
		int *	info
	)

compute y := y + a x

vector_build()

hpro_s_ vector_t hpro_s_ vector_build	(	const size_t	size,
		int *	info
	)

create scalar vectors of size size initialised with 0

vector_bytesize()

size_t hpro_s_ vector_bytesize	(	const hpro_s_ vector_t	v,
		int *	info
	)

return size of matrix in bytes

vector_copy()

hpro_s_ vector_t hpro_s_ vector_copy	(	const hpro_s_ vector_t	v,
		int *	info
	)

copy vectormatrix

vector_dot()

float hpro_s_ vector_dot	(	const hpro_s_ vector_t	x,
		const hpro_s_ vector_t	y,
		int *	info
	)

compute dot product <x,y> = x^H * y

vector_entry_get()

float hpro_s_ vector_entry_get	(	const hpro_s_ vector_t	x,
		const size_t	i,
		int *	info
	)

get single entry at position i in vector x

vector_entry_set()

void hpro_s_ vector_entry_set	(	const hpro_s_ vector_t	x,
		const size_t	i,
		const float	f,
		int *	info
	)

set single entry at position i in vector x

vector_export()

void hpro_s_ vector_export	(	const hpro_s_ vector_t	v,
		float *	arr,
		int *	info
	)

copy data from vector v into given C arrays

• arr has to be of size hlib_vector_size( v )

vector_fill()

void hpro_s_ vector_fill	(	hpro_s_ vector_t	x,
		const float	f,
		int *	info
	)

fill vector x with constant value f

vector_fill_rand()

void hpro_s_ vector_fill_rand	(	hpro_s_ vector_t	x,
		int *	info
	)

fill vector x with random values

vector_free()

void hpro_s_ vector_free	(	hpro_s_ vector_t	v,
		int *	info
	)

free resources coupled with vector v

vector_import()

void hpro_s_ vector_import	(	hpro_s_ vector_t	v,
		const float *	arr,
		int *	info
	)

copy data from C array arr into vector v

• arr has to be of size hlib_vector_size( v )

vector_norm2()

float hpro_s_ vector_norm2	(	const hpro_s_ vector_t	x,
		int *	info
	)

compute euklidean norm of x

vector_norm_inf()

float hpro_s_ vector_norm_inf	(	const hpro_s_ vector_t	x,
		int *	info
	)

compute infinity norm of x

vector_permute()

void hpro_s_ vector_permute	(	hpro_s_ vector_t	v,
		const hpro_ permutation_t	perm,
		int *	info
	)

reorder vector entries with given permutation

vector_scale()

void hpro_s_ vector_scale	(	hpro_s_ vector_t	x,
		const float	f,
		int *	info
	)

scale vector x by f

vector_size()

size_t hpro_s_ vector_size	(	const hpro_s_ vector_t	v,
		int *	info
	)

return size of vectors