SpM Handbook: src/spm.c File Reference

Searching...

No Matches

#include "common.h"
#include <cblas.h>
#include <lapacke.h>

Functions
spm_int_t	spm_create_loc2glob_continuous (const spmatrix_t spm, spm_int_t *l2g_ptr)
	Generate a continuous loc2glob array on each node.
spm_int_t *	spm_get_glob2loc (const spmatrix_t *spm)
	Computes the glob2loc array if needed, and returns it.
spm_int_t *	spm_getandset_glob2loc (spmatrix_t *spm)
	Computes the glob2loc array if needed, and returns it.
int	spm_get_distribution (const spmatrix_t *spm)
	Search the distribution pattern used in the spm structure.
spm_int_t *	spm_get_value_idx_by_col (const spmatrix_t *spm)
	Create an array that represents the shift for each sub-element of the original multidof value array.
spm_int_t *	spm_get_value_idx_by_elt (const spmatrix_t *spm)
	Create an array that represents the shift for each sub-element of the original multidof value array.
int	spmBlasGetNumThreads (void)
	Return the current number of threads used for blas calls.
int	spmBlasSetNumThreads (int nt)
	Set the number of threads for blas calls (BLIS, MKL, OpenBLAS) and return the previous number of blas threads.
int	spmBlasSetNumThreadsOne (void)
	Set the number of threads for blas calls (BLIS, MKL, OpenBLAS) to 1 and return the previous number of blas threads.
SPM basic subroutines
void	spmInitDist (spmatrix_t *spm, SPM_Comm comm)
	Init the spm structure with a specific communicator.
void	spmInit (spmatrix_t *spm)
	Init the spm structure.
void	spmAlloc (spmatrix_t *spm)
	Allocate the arrays of an spm structure.
void	spmExit (spmatrix_t *spm)
	Cleanup the spm structure but do not free the spm pointer.
void	spmBase (spmatrix_t *spm, int baseval)
	Rebase the arrays of the spm to the given value.
spm_int_t	spmFindBase (const spmatrix_t *spm)
	Search the base used in the spm structure.
int	spmConvert (int ofmttype, spmatrix_t *spm)
	Convert the storage format of the spm.
void	spmCopy (const spmatrix_t spm, spmatrix_t newspm)
	Create a copy of the spm.
SPM debug subroutines
void	spm2Dense (const spmatrix_t spm, void A)
	Convert the spm matrix into a dense matrix for test purpose.
void	spmPrintInfo (const spmatrix_t spm, FILE stream)
	Print basic informations about the spm matrix into a given stream.
void	spmPrint (const spmatrix_t spm, FILE stream)
	Print an spm matrix into into a given file.
void	spmExpand (const spmatrix_t spm_in, spmatrix_t spm_out)
	Expand a multi-dof spm matrix into an spm with constant dof set to 1.
SPM BLAS subroutines
double	spmNorm (spm_normtype_t ntype, const spmatrix_t *spm)
	Compute the norm of the spm.
double	spmNormVec (spm_normtype_t ntype, const spmatrix_t spm, const void x, spm_int_t inc)
	Compute the norm of the spm.
double	spmNormMat (spm_normtype_t ntype, const spmatrix_t spm, spm_int_t n, const void A, spm_int_t lda)
	Compute the norm of the spm.
int	spmMatVec (spm_trans_t trans, double alpha, const spmatrix_t spm, const void x, double beta, void *y)
	Compute a matrix-vector product.
int	spmMatMat (spm_trans_t trans, spm_int_t n, double alpha, const spmatrix_t A, const void B, spm_int_t ldb, double beta, void *C, spm_int_t ldc)
	Compute a matrix-matrix product.
void	spmScal (double alpha, spmatrix_t *spm)
	Scale the spm.
void	spmScalMatrix (double alpha, spmatrix_t *spm)
	Scale the spm.
void	spmScalVec (double alpha, const spmatrix_t spm, void x, spm_int_t incx)
	Scale a vector associated to a sparse matrix. Deprecated function replaced by spmScalVec() or spmScalMat().
void	spmScalMat (double alpha, const spmatrix_t spm, spm_int_t n, void A, spm_int_t lda)
	Scale a matrix associated to a sparse matrix.
void	spmScalVector (spm_coeftype_t flt, double alpha, spm_int_t n, void *x, spm_int_t incx)
	Scale a vector according to the spm type.
SPM subroutines to check format
int	spmSort (spmatrix_t *spm)
	Sort the subarray of edges of each vertex.
spm_int_t	spmMergeDuplicate (spmatrix_t *spm)
	Merge multiple entries in a spm by summing their values together.
int	spmCheckAndCorrect (const spmatrix_t spm_in, spmatrix_t spm_out)
	Check the correctness of a spm.
SPM subroutines to check factorization/solve
int	spmCheckAxb (double eps, spm_int_t nrhs, const spmatrix_t spm, void x0, spm_int_t ldx0, void b, spm_int_t ldb, const void x, spm_int_t ldx)
	Check the backward error, and the forward error if x0 is provided.
int	spmGenMat (spm_rhstype_t type, spm_int_t nrhs, const spmatrix_t spm, void alpha, unsigned long long int seed, void *A, spm_int_t lda)
	Generate a set of vectors associated to a given matrix.
int	spmGenVec (spm_rhstype_t type, const spmatrix_t spm, void alpha, unsigned long long int seed, void *x, spm_int_t incx)
	Generate a vector associated to a given matrix.

Detailed Description

SParse Matrix package main routines.

Copyright: 2016-2026 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria, Univ. Bordeaux. All rights reserved.

Version: 1.3.0

Author: Pierre Ramet; Mathieu Faverge; Tony Delarue; Alban Bellot; Matias Hastaran; Matthieu Kuhn; Gregoire Pichon; Alycia Lisito; Gregoire Pichon; Florent Pruvost

Date: 2026-07-06

Definition in file spm.c.

Function Documentation

◆ spm_create_loc2glob_continuous()

spm_int_t spm_create_loc2glob_continuous	(	const spmatrix_t *	spm,
		spm_int_t **	l2g_ptr )

Generate a continuous loc2glob array on each node.

Parameters

[in]	spm	The allocated spm with the correct gN field.
[out]	l2g_ptr	Pointer to the loc2glob array that will be allocated and initialized.

Return values

The	number of unknowns of the local spm.

Definition at line 1852 of file spm.c.

References spmatrix_s::baseval, spmatrix_s::clustnbr, spmatrix_s::clustnum, spmatrix_s::gN, and spm_imin().

Referenced by spm_scatter_init(), and spmGather().

◆ spm_get_glob2loc()

spm_int_t * spm_get_glob2loc ( const spmatrix_t * spm )

Computes the glob2loc array if needed, and returns it.

Parameters

[in,out] spm The sparse matrix for which the glob2loc array must be computed.

Return values

The	pointer to the glob2loc array of the spm.

Definition at line 1896 of file spm.c.

References spmatrix_s::baseval, spmatrix_s::clustnbr, spmatrix_s::clustnum, spmatrix_s::comm, spmatrix_s::glob2loc, spmatrix_s::gN, spmatrix_s::loc2glob, spmatrix_s::n, spmatrix_s::replicated, and SPM_MPI_INT.

Referenced by spm_get_distribution(), and spm_getandset_glob2loc().

◆ spm_getandset_glob2loc()

spm_int_t * spm_getandset_glob2loc ( spmatrix_t * spm )

Computes the glob2loc array if needed, and returns it.

Parameters

[in,out] spm The sparse matrix for which the glob2loc array must be computed.

Return values

The	pointer to the glob2loc array of the spm.

Definition at line 2019 of file spm.c.

References spmatrix_s::clustnbr, spmatrix_s::glob2loc, spmatrix_s::gN, spmatrix_s::replicated, spm_get_glob2loc(), and spmUpdateComputedFields().

Referenced by c_spmSort(), d_spmSort(), p_spmSort(), s_spmSort(), spm_redist_get_newg2l(), spm_scatter_csx_get_locals(), spm_scatter_ijv_get_locals(), spm_symm_check_local_pattern(), and z_spmSort().

◆ spm_get_distribution()

int spm_get_distribution ( const spmatrix_t * spm )

Search the distribution pattern used in the spm structure.

Parameters

[in] spm The sparse matrix structure.

Returns: SpmDistByColumn if the distribution is column based. SpmDistByRow if the distribution is row based. (SpmDistByColumn|SpmDistByRow) if the matrix is not distributed.

Definition at line 2062 of file spm.c.

References spmatrix_s::baseval, spmatrix_s::clustnbr, spmatrix_s::colptr, spmatrix_s::comm, spmatrix_s::fmttype, spmatrix_s::glob2loc, spmatrix_s::gN, spmatrix_s::n, spmatrix_s::nnz, spmatrix_s::replicated, spmatrix_s::rowptr, spm_get_glob2loc(), SpmCSC, SpmCSR, SpmDistByColumn, and SpmDistByRow.