sparseMat.h File Reference

Sparse matrices. More...

#include <set>
#include <map>
#include <vector>
#include <iostream>
#include <cassert>
#include <algorithm>
#include <execution>
#include <mutex>
#include "config.h"
#include "algebraCore.h"

Go to the source code of this file.

Classes
class	algebra::w_sparseMat
	Write-mode sparse matrix. More...
class	algebra::r_sparseMat
	Read-mode square sparse matrix. More...
struct	algebra::r_sparseMat::SparseVector

Namespaces
	algebra

Typedefs
using	algebra::MatrixShape = std::vector< std::set< int > >

Detailed Description

Sparse matrices.

The main class provided here is r_sparseMat, a "read-mode" square sparse matrix. It is meant to efficiently compute matrix-vector products, where the vectors are full (not sparse), represented as std::vector<double>. Such a matrix can be constructed in two ways:

• If the set of valid index pairs (locations that can have non-zero values) is not known in advance, one can construct a w_sparseMat (write-mode sparse matrix), populate it with its insert() method, then use it to construct the read-mode sparse matrix.
• If the set of valid index pairs is known, one can store it in a MatrixShape, then use this shape to construct the read-mode sparse matrix with all values initialized to zero.

Once constructed, the set of valid index pairs is immutable. However, the associated values can be modified with clear() and add(). Rewriting the values this way is way more efficient than reconstructing the matrix from scratch.