eigen2_sparse_product.cpp revision c981c48f5bc9aefeffc0bcb0cc3934c2fae179dd 
1// This file is part of Eigen, a lightweight C++ template library
2// for linear algebra. Eigen itself is part of the KDE project.
3//
4// Copyright (C) 2008 Daniel Gomez Ferro <dgomezferro@gmail.com>
5//
6// This Source Code Form is subject to the terms of the Mozilla
7// Public License v. 2.0. If a copy of the MPL was not distributed
8// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
9
10#include "sparse.h"
11
12template<typename SparseMatrixType> void sparse_product(const SparseMatrixType& ref)
13{
14  const int rows = ref.rows();
15  const int cols = ref.cols();
16  typedef typename SparseMatrixType::Scalar Scalar;
17  enum { Flags = SparseMatrixType::Flags };
18
19  double density = std::max(8./(rows*cols), 0.01);
20  typedef Matrix<Scalar,Dynamic,Dynamic> DenseMatrix;
21  typedef Matrix<Scalar,Dynamic,1> DenseVector;
22
23  // test matrix-matrix product
24  {
25    DenseMatrix refMat2 = DenseMatrix::Zero(rows, rows);
26    DenseMatrix refMat3 = DenseMatrix::Zero(rows, rows);
27    DenseMatrix refMat4 = DenseMatrix::Zero(rows, rows);
28    DenseMatrix dm4 = DenseMatrix::Zero(rows, rows);
29    SparseMatrixType m2(rows, rows);
30    SparseMatrixType m3(rows, rows);
31    SparseMatrixType m4(rows, rows);
32    initSparse<Scalar>(density, refMat2, m2);
33    initSparse<Scalar>(density, refMat3, m3);
34    initSparse<Scalar>(density, refMat4, m4);
35    VERIFY_IS_APPROX(m4=m2*m3, refMat4=refMat2*refMat3);
36    VERIFY_IS_APPROX(m4=m2.transpose()*m3, refMat4=refMat2.transpose()*refMat3);
37    VERIFY_IS_APPROX(m4=m2.transpose()*m3.transpose(), refMat4=refMat2.transpose()*refMat3.transpose());
38    VERIFY_IS_APPROX(m4=m2*m3.transpose(), refMat4=refMat2*refMat3.transpose());
39
40    // sparse * dense
41    VERIFY_IS_APPROX(dm4=m2*refMat3, refMat4=refMat2*refMat3);
42    VERIFY_IS_APPROX(dm4=m2*refMat3.transpose(), refMat4=refMat2*refMat3.transpose());
43    VERIFY_IS_APPROX(dm4=m2.transpose()*refMat3, refMat4=refMat2.transpose()*refMat3);
44    VERIFY_IS_APPROX(dm4=m2.transpose()*refMat3.transpose(), refMat4=refMat2.transpose()*refMat3.transpose());
45
46    // dense * sparse
47    VERIFY_IS_APPROX(dm4=refMat2*m3, refMat4=refMat2*refMat3);
48    VERIFY_IS_APPROX(dm4=refMat2*m3.transpose(), refMat4=refMat2*refMat3.transpose());
49    VERIFY_IS_APPROX(dm4=refMat2.transpose()*m3, refMat4=refMat2.transpose()*refMat3);
50    VERIFY_IS_APPROX(dm4=refMat2.transpose()*m3.transpose(), refMat4=refMat2.transpose()*refMat3.transpose());
51
52    VERIFY_IS_APPROX(m3=m3*m3, refMat3=refMat3*refMat3);
53  }
54
55  // test matrix - diagonal product
56  if(false) // it compiles, but the precision is terrible. probably doesn't matter in this branch....
57  {
58    DenseMatrix refM2 = DenseMatrix::Zero(rows, rows);
59    DenseMatrix refM3 = DenseMatrix::Zero(rows, rows);
60    DiagonalMatrix<DenseVector> d1(DenseVector::Random(rows));
61    SparseMatrixType m2(rows, rows);
62    SparseMatrixType m3(rows, rows);
63    initSparse<Scalar>(density, refM2, m2);
64    initSparse<Scalar>(density, refM3, m3);
65    VERIFY_IS_APPROX(m3=m2*d1, refM3=refM2*d1);
66    VERIFY_IS_APPROX(m3=m2.transpose()*d1, refM3=refM2.transpose()*d1);
67    VERIFY_IS_APPROX(m3=d1*m2, refM3=d1*refM2);
68    VERIFY_IS_APPROX(m3=d1*m2.transpose(), refM3=d1 * refM2.transpose());
69  }
70
71  // test self adjoint products
72  {
73    DenseMatrix b = DenseMatrix::Random(rows, rows);
74    DenseMatrix x = DenseMatrix::Random(rows, rows);
75    DenseMatrix refX = DenseMatrix::Random(rows, rows);
76    DenseMatrix refUp = DenseMatrix::Zero(rows, rows);
77    DenseMatrix refLo = DenseMatrix::Zero(rows, rows);
78    DenseMatrix refS = DenseMatrix::Zero(rows, rows);
79    SparseMatrixType mUp(rows, rows);
80    SparseMatrixType mLo(rows, rows);
81    SparseMatrixType mS(rows, rows);
82    do {
83      initSparse<Scalar>(density, refUp, mUp, ForceRealDiag|/*ForceNonZeroDiag|*/MakeUpperTriangular);
84    } while (refUp.isZero());
85    refLo = refUp.transpose().conjugate();
86    mLo = mUp.transpose().conjugate();
87    refS = refUp + refLo;
88    refS.diagonal() *= 0.5;
89    mS = mUp + mLo;
90    for (int k=0; k<mS.outerSize(); ++k)
91      for (typename SparseMatrixType::InnerIterator it(mS,k); it; ++it)
92        if (it.index() == k)
93          it.valueRef() *= 0.5;
94
95    VERIFY_IS_APPROX(refS.adjoint(), refS);
96    VERIFY_IS_APPROX(mS.transpose().conjugate(), mS);
97    VERIFY_IS_APPROX(mS, refS);
98    VERIFY_IS_APPROX(x=mS*b, refX=refS*b);
99    VERIFY_IS_APPROX(x=mUp.template marked<UpperTriangular|SelfAdjoint>()*b, refX=refS*b);
100    VERIFY_IS_APPROX(x=mLo.template marked<LowerTriangular|SelfAdjoint>()*b, refX=refS*b);
101    VERIFY_IS_APPROX(x=mS.template marked<SelfAdjoint>()*b, refX=refS*b);
102  }
103
104}
105
106void test_eigen2_sparse_product()
107{
108  for(int i = 0; i < g_repeat; i++) {
109    CALL_SUBTEST_1( sparse_product(SparseMatrix<double>(8, 8)) );
110    CALL_SUBTEST_2( sparse_product(SparseMatrix<std::complex<double> >(16, 16)) );
111    CALL_SUBTEST_1( sparse_product(SparseMatrix<double>(33, 33)) );
112
113    CALL_SUBTEST_3( sparse_product(DynamicSparseMatrix<double>(8, 8)) );
114  }
115}
116