• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 
2 // This file is part of Eigen, a lightweight C++ template library
3 // for linear algebra.
4 //
5 // Copyright (C) 2012  Désiré Nuentsa-Wakam <desire.nuentsa_wakam@inria.fr>
6 //
7 // This Source Code Form is subject to the terms of the Mozilla
8 // Public License v. 2.0. If a copy of the MPL was not distributed
9 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
10 
11 #ifndef EIGEN_ORDERING_H
12 #define EIGEN_ORDERING_H
13 
14 namespace Eigen {
15 
16 #include "Eigen_Colamd.h"
17 
18 namespace internal {
19 
20 /** \internal
21   * \ingroup OrderingMethods_Module
22   * \returns the symmetric pattern A^T+A from the input matrix A.
23   * FIXME: The values should not be considered here
24   */
25 template<typename MatrixType>
ordering_helper_at_plus_a(const MatrixType & mat,MatrixType & symmat)26 void ordering_helper_at_plus_a(const MatrixType& mat, MatrixType& symmat)
27 {
28   MatrixType C;
29   C = mat.transpose(); // NOTE: Could be  costly
30   for (int i = 0; i < C.rows(); i++)
31   {
32       for (typename MatrixType::InnerIterator it(C, i); it; ++it)
33         it.valueRef() = 0.0;
34   }
35   symmat = C + mat;
36 }
37 
38 }
39 
40 #ifndef EIGEN_MPL2_ONLY
41 
42 /** \ingroup OrderingMethods_Module
43   * \class AMDOrdering
44   *
45   * Functor computing the \em approximate \em minimum \em degree ordering
46   * If the matrix is not structurally symmetric, an ordering of A^T+A is computed
47   * \tparam  Index The type of indices of the matrix
48   * \sa COLAMDOrdering
49   */
50 template <typename Index>
51 class AMDOrdering
52 {
53   public:
54     typedef PermutationMatrix<Dynamic, Dynamic, Index> PermutationType;
55 
56     /** Compute the permutation vector from a sparse matrix
57      * This routine is much faster if the input matrix is column-major
58      */
59     template <typename MatrixType>
operator()60     void operator()(const MatrixType& mat, PermutationType& perm)
61     {
62       // Compute the symmetric pattern
63       SparseMatrix<typename MatrixType::Scalar, ColMajor, Index> symm;
64       internal::ordering_helper_at_plus_a(mat,symm);
65 
66       // Call the AMD routine
67       //m_mat.prune(keep_diag());
68       internal::minimum_degree_ordering(symm, perm);
69     }
70 
71     /** Compute the permutation with a selfadjoint matrix */
72     template <typename SrcType, unsigned int SrcUpLo>
operator()73     void operator()(const SparseSelfAdjointView<SrcType, SrcUpLo>& mat, PermutationType& perm)
74     {
75       SparseMatrix<typename SrcType::Scalar, ColMajor, Index> C; C = mat;
76 
77       // Call the AMD routine
78       // m_mat.prune(keep_diag()); //Remove the diagonal elements
79       internal::minimum_degree_ordering(C, perm);
80     }
81 };
82 
83 #endif // EIGEN_MPL2_ONLY
84 
85 /** \ingroup OrderingMethods_Module
86   * \class NaturalOrdering
87   *
88   * Functor computing the natural ordering (identity)
89   *
90   * \note Returns an empty permutation matrix
91   * \tparam  Index The type of indices of the matrix
92   */
93 template <typename Index>
94 class NaturalOrdering
95 {
96   public:
97     typedef PermutationMatrix<Dynamic, Dynamic, Index> PermutationType;
98 
99     /** Compute the permutation vector from a column-major sparse matrix */
100     template <typename MatrixType>
operator()101     void operator()(const MatrixType& /*mat*/, PermutationType& perm)
102     {
103       perm.resize(0);
104     }
105 
106 };
107 
108 /** \ingroup OrderingMethods_Module
109   * \class COLAMDOrdering
110   *
111   * Functor computing the \em column \em approximate \em minimum \em degree ordering
112   * The matrix should be in column-major and \b compressed format (see SparseMatrix::makeCompressed()).
113   */
114 template<typename Index>
115 class COLAMDOrdering
116 {
117   public:
118     typedef PermutationMatrix<Dynamic, Dynamic, Index> PermutationType;
119     typedef Matrix<Index, Dynamic, 1> IndexVector;
120 
121     /** Compute the permutation vector \a perm form the sparse matrix \a mat
122       * \warning The input sparse matrix \a mat must be in compressed mode (see SparseMatrix::makeCompressed()).
123       */
124     template <typename MatrixType>
operator()125     void operator() (const MatrixType& mat, PermutationType& perm)
126     {
127       eigen_assert(mat.isCompressed() && "COLAMDOrdering requires a sparse matrix in compressed mode. Call .makeCompressed() before passing it to COLAMDOrdering");
128 
129       Index m = mat.rows();
130       Index n = mat.cols();
131       Index nnz = mat.nonZeros();
132       // Get the recommended value of Alen to be used by colamd
133       Index Alen = internal::colamd_recommended(nnz, m, n);
134       // Set the default parameters
135       double knobs [COLAMD_KNOBS];
136       Index stats [COLAMD_STATS];
137       internal::colamd_set_defaults(knobs);
138 
139       IndexVector p(n+1), A(Alen);
140       for(Index i=0; i <= n; i++)   p(i) = mat.outerIndexPtr()[i];
141       for(Index i=0; i < nnz; i++)  A(i) = mat.innerIndexPtr()[i];
142       // Call Colamd routine to compute the ordering
143       Index info = internal::colamd(m, n, Alen, A.data(), p.data(), knobs, stats);
144       EIGEN_UNUSED_VARIABLE(info);
145       eigen_assert( info && "COLAMD failed " );
146 
147       perm.resize(n);
148       for (Index i = 0; i < n; i++) perm.indices()(p(i)) = i;
149     }
150 };
151 
152 } // end namespace Eigen
153 
154 #endif
155