cusp::csr_matrix< IndexType, ValueType, MemorySpace > Class Template Reference

#include <csr_matrix.h>

Inheritance diagram for cusp::csr_matrix< IndexType, ValueType, MemorySpace >:

Public Methods
	csr_matrix (void)
	csr_matrix (size_t num_rows, size_t num_cols, size_t num_entries)
template<typename MatrixType >
	csr_matrix (const MatrixType &matrix)
void	resize (const size_t num_rows, const size_t num_cols, const size_t num_entries)
void	swap (csr_matrix &matrix)
template<typename MatrixType >
csr_matrix &	operator= (const MatrixType &matrix)

Public Members
row_offsets_array_type	row_offsets
column_indices_array_type	column_indices
values_array_type	values

Detailed description

template<typename IndexType, typename ValueType, class MemorySpace>
class cusp::csr_matrix< IndexType, ValueType, MemorySpace >

Compressed sparse row (CSR) representation a sparse matrix.

Template Parameters

IndexType	Type used for matrix indices (e.g. int).
ValueType	Type used for matrix values (e.g. float).
MemorySpace	A memory space (e.g. cusp::host_memory or cusp::device_memory)

Overview

A csr_matrix is a sparse matrix container that stores an offset to the first entry of each row in matrix and one column entry per nonzero. The matrix may reside in either "host" or "device" memory depending on the MemorySpace. All entries in the csr_matrix are sorted according to row and internally sorted within each row by column index.

Note

The matrix entries within the same row must be sorted by column index.

The matrix should not contain duplicate entries.

Example

The following code snippet demonstrates how to create a 4-by-3 csr_matrix on the host with 6 nonzeros and then copies the matrix to the device.

// include the csr_matrix header file
#include <cusp/csr_matrix.h>
#include <cusp/print.h>
 
int main()
{
  // allocate storage for (4,3) matrix with 4 nonzeros
  cusp::csr_matrix<int,float,cusp::host_memory> A(4,3,6);
 
  // initialize matrix entries on host
  A.row_offsets[0] = 0;  // first offset is always zero
  A.row_offsets[1] = 2;
  A.row_offsets[2] = 2;
  A.row_offsets[3] = 3;
  A.row_offsets[4] = 6; // last offset is always num_entries
 
  A.column_indices[0] = 0; A.values[0] = 10;
  A.column_indices[1] = 2; A.values[1] = 20;
  A.column_indices[2] = 2; A.values[2] = 30;
  A.column_indices[3] = 0; A.values[3] = 40;
  A.column_indices[4] = 1; A.values[4] = 50;
  A.column_indices[5] = 2; A.values[5] = 60;
 
  // A now represents the following matrix
  //    [10  0 20]
  //    [ 0  0  0]
  //    [ 0  0 30]
  //    [40 50 60]
 
  // copy to the device
  cusp::csr_matrix<int,float,cusp::device_memory> B(A);
 
  cusp::print(B);
}

Definition at line 107 of file csr_matrix.h.

Constructor & Destructor Documentation

csr_matrix() [1/3]

template<typename IndexType , typename ValueType , class MemorySpace >

cusp::csr_matrix< IndexType, ValueType, MemorySpace >::csr_matrix ( void  )

inline

Construct an empty csr_matrix.

Definition at line 162 of file csr_matrix.h.

csr_matrix() [2/3]

template<typename IndexType , typename ValueType , class MemorySpace >

cusp::csr_matrix< IndexType, ValueType, MemorySpace >::csr_matrix ( size_t  num_rows, size_t  num_cols, size_t  num_entries  )

inline

Construct a csr_matrix with a specific shape and number of nonzero entries.

Parameters

num_rows	Number of rows.
num_cols	Number of columns.
num_entries	Number of nonzero matrix entries.

Definition at line 170 of file csr_matrix.h.

csr_matrix() [3/3]

template<typename IndexType , typename ValueType , class MemorySpace >

template<typename MatrixType >

cusp::csr_matrix< IndexType, ValueType, MemorySpace >::csr_matrix

(

const MatrixType &

matrix

)

Construct a csr_matrix from another matrix.

Template Parameters

MatrixType

Type of input matrix used to create this csr_matrix.

Parameters

matrix

Another sparse or dense matrix.

Member Function Documentation

operator=()

template<typename IndexType , typename ValueType , class MemorySpace >

template<typename MatrixType >

csr_matrix & cusp::csr_matrix< IndexType, ValueType, MemorySpace >::operator=

(

const MatrixType &

matrix

)

Assignment from another matrix.

Template Parameters

MatrixType

Type of input matrix to copy into this csr_matrix.

Parameters

matrix

Another sparse or dense matrix.

resize()

template<typename IndexType , typename ValueType , class MemorySpace >

void cusp::csr_matrix< IndexType, ValueType, MemorySpace >::resize	(	const size_t	num_rows,
		const size_t	num_cols,
		const size_t	num_entries
	)

Resize matrix dimensions and underlying storage

Parameters

num_rows	Number of rows.
num_cols	Number of columns.
num_entries	Number of nonzero matrix entries.

swap()

template<typename IndexType , typename ValueType , class MemorySpace >

void cusp::csr_matrix< IndexType, ValueType, MemorySpace >::swap

(

csr_matrix< IndexType, ValueType, MemorySpace > &

matrix

)

Swap the contents of two csr_matrix objects.

Parameters

matrix

Another csr_matrix with the same IndexType and ValueType.

Member Data Documentation

column_indices

template<typename IndexType , typename ValueType , class MemorySpace >

column_indices_array_type cusp::csr_matrix< IndexType, ValueType, MemorySpace >::column_indices

Storage for the column indices of the CSR data structure.

Definition at line 154 of file csr_matrix.h.

row_offsets

template<typename IndexType , typename ValueType , class MemorySpace >

row_offsets_array_type cusp::csr_matrix< IndexType, ValueType, MemorySpace >::row_offsets

Storage for the row offsets of the CSR data structure. Also called the "row pointer" array.

Definition at line 150 of file csr_matrix.h.

values

template<typename IndexType , typename ValueType , class MemorySpace >

values_array_type cusp::csr_matrix< IndexType, ValueType, MemorySpace >::values

Storage for the nonzero entries of the CSR data structure.

Definition at line 158 of file csr_matrix.h.