• Home
  • History
  • Annotate
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1  // This file is part of Eigen, a lightweight C++ template library
2  // for linear algebra.
3  //
4  // Copyright (C) 2008 Benoit Jacob <jacob.benoit.1@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 "main.h"
11  
matrixRedux(const MatrixType & m)12  template<typename MatrixType> void matrixRedux(const MatrixType& m)
13  {
14    typedef typename MatrixType::Index Index;
15    typedef typename MatrixType::Scalar Scalar;
16    typedef typename MatrixType::RealScalar RealScalar;
17  
18    Index rows = m.rows();
19    Index cols = m.cols();
20  
21    MatrixType m1 = MatrixType::Random(rows, cols);
22  
23    // The entries of m1 are uniformly distributed in [0,1], so m1.prod() is very small. This may lead to test
24    // failures if we underflow into denormals. Thus, we scale so that entires are close to 1.
25    MatrixType m1_for_prod = MatrixType::Ones(rows, cols) + RealScalar(0.2) * m1;
26  
27    VERIFY_IS_MUCH_SMALLER_THAN(MatrixType::Zero(rows, cols).sum(), Scalar(1));
28    VERIFY_IS_APPROX(MatrixType::Ones(rows, cols).sum(), Scalar(float(rows*cols))); // the float() here to shut up excessive MSVC warning about int->complex conversion being lossy
29    Scalar s(0), p(1), minc(numext::real(m1.coeff(0))), maxc(numext::real(m1.coeff(0)));
30    for(int j = 0; j < cols; j++)
31    for(int i = 0; i < rows; i++)
32    {
33      s += m1(i,j);
34      p *= m1_for_prod(i,j);
35      minc = (std::min)(numext::real(minc), numext::real(m1(i,j)));
36      maxc = (std::max)(numext::real(maxc), numext::real(m1(i,j)));
37    }
38    const Scalar mean = s/Scalar(RealScalar(rows*cols));
39  
40    VERIFY_IS_APPROX(m1.sum(), s);
41    VERIFY_IS_APPROX(m1.mean(), mean);
42    VERIFY_IS_APPROX(m1_for_prod.prod(), p);
43    VERIFY_IS_APPROX(m1.real().minCoeff(), numext::real(minc));
44    VERIFY_IS_APPROX(m1.real().maxCoeff(), numext::real(maxc));
45  
46    // test slice vectorization assuming assign is ok
47    Index r0 = internal::random<Index>(0,rows-1);
48    Index c0 = internal::random<Index>(0,cols-1);
49    Index r1 = internal::random<Index>(r0+1,rows)-r0;
50    Index c1 = internal::random<Index>(c0+1,cols)-c0;
51    VERIFY_IS_APPROX(m1.block(r0,c0,r1,c1).sum(), m1.block(r0,c0,r1,c1).eval().sum());
52    VERIFY_IS_APPROX(m1.block(r0,c0,r1,c1).mean(), m1.block(r0,c0,r1,c1).eval().mean());
53    VERIFY_IS_APPROX(m1_for_prod.block(r0,c0,r1,c1).prod(), m1_for_prod.block(r0,c0,r1,c1).eval().prod());
54    VERIFY_IS_APPROX(m1.block(r0,c0,r1,c1).real().minCoeff(), m1.block(r0,c0,r1,c1).real().eval().minCoeff());
55    VERIFY_IS_APPROX(m1.block(r0,c0,r1,c1).real().maxCoeff(), m1.block(r0,c0,r1,c1).real().eval().maxCoeff());
56  
57    // test empty objects
58    VERIFY_IS_APPROX(m1.block(r0,c0,0,0).sum(),   Scalar(0));
59    VERIFY_IS_APPROX(m1.block(r0,c0,0,0).prod(),  Scalar(1));
60  }
61  
vectorRedux(const VectorType & w)62  template<typename VectorType> void vectorRedux(const VectorType& w)
63  {
64    using std::abs;
65    typedef typename VectorType::Index Index;
66    typedef typename VectorType::Scalar Scalar;
67    typedef typename NumTraits<Scalar>::Real RealScalar;
68    Index size = w.size();
69  
70    VectorType v = VectorType::Random(size);
71    VectorType v_for_prod = VectorType::Ones(size) + Scalar(0.2) * v; // see comment above declaration of m1_for_prod
72  
73    for(int i = 1; i < size; i++)
74    {
75      Scalar s(0), p(1);
76      RealScalar minc(numext::real(v.coeff(0))), maxc(numext::real(v.coeff(0)));
77      for(int j = 0; j < i; j++)
78      {
79        s += v[j];
80        p *= v_for_prod[j];
81        minc = (std::min)(minc, numext::real(v[j]));
82        maxc = (std::max)(maxc, numext::real(v[j]));
83      }
84      VERIFY_IS_MUCH_SMALLER_THAN(abs(s - v.head(i).sum()), Scalar(1));
85      VERIFY_IS_APPROX(p, v_for_prod.head(i).prod());
86      VERIFY_IS_APPROX(minc, v.real().head(i).minCoeff());
87      VERIFY_IS_APPROX(maxc, v.real().head(i).maxCoeff());
88    }
89  
90    for(int i = 0; i < size-1; i++)
91    {
92      Scalar s(0), p(1);
93      RealScalar minc(numext::real(v.coeff(i))), maxc(numext::real(v.coeff(i)));
94      for(int j = i; j < size; j++)
95      {
96        s += v[j];
97        p *= v_for_prod[j];
98        minc = (std::min)(minc, numext::real(v[j]));
99        maxc = (std::max)(maxc, numext::real(v[j]));
100      }
101      VERIFY_IS_MUCH_SMALLER_THAN(abs(s - v.tail(size-i).sum()), Scalar(1));
102      VERIFY_IS_APPROX(p, v_for_prod.tail(size-i).prod());
103      VERIFY_IS_APPROX(minc, v.real().tail(size-i).minCoeff());
104      VERIFY_IS_APPROX(maxc, v.real().tail(size-i).maxCoeff());
105    }
106  
107    for(int i = 0; i < size/2; i++)
108    {
109      Scalar s(0), p(1);
110      RealScalar minc(numext::real(v.coeff(i))), maxc(numext::real(v.coeff(i)));
111      for(int j = i; j < size-i; j++)
112      {
113        s += v[j];
114        p *= v_for_prod[j];
115        minc = (std::min)(minc, numext::real(v[j]));
116        maxc = (std::max)(maxc, numext::real(v[j]));
117      }
118      VERIFY_IS_MUCH_SMALLER_THAN(abs(s - v.segment(i, size-2*i).sum()), Scalar(1));
119      VERIFY_IS_APPROX(p, v_for_prod.segment(i, size-2*i).prod());
120      VERIFY_IS_APPROX(minc, v.real().segment(i, size-2*i).minCoeff());
121      VERIFY_IS_APPROX(maxc, v.real().segment(i, size-2*i).maxCoeff());
122    }
123  
124    // test empty objects
125    VERIFY_IS_APPROX(v.head(0).sum(),   Scalar(0));
126    VERIFY_IS_APPROX(v.tail(0).prod(),  Scalar(1));
127    VERIFY_RAISES_ASSERT(v.head(0).mean());
128    VERIFY_RAISES_ASSERT(v.head(0).minCoeff());
129    VERIFY_RAISES_ASSERT(v.head(0).maxCoeff());
130  }
131  
test_redux()132  void test_redux()
133  {
134    // the max size cannot be too large, otherwise reduxion operations obviously generate large errors.
135    int maxsize = (std::min)(100,EIGEN_TEST_MAX_SIZE);
136    TEST_SET_BUT_UNUSED_VARIABLE(maxsize);
137    for(int i = 0; i < g_repeat; i++) {
138      CALL_SUBTEST_1( matrixRedux(Matrix<float, 1, 1>()) );
139      CALL_SUBTEST_1( matrixRedux(Array<float, 1, 1>()) );
140      CALL_SUBTEST_2( matrixRedux(Matrix2f()) );
141      CALL_SUBTEST_2( matrixRedux(Array2f()) );
142      CALL_SUBTEST_3( matrixRedux(Matrix4d()) );
143      CALL_SUBTEST_3( matrixRedux(Array4d()) );
144      CALL_SUBTEST_4( matrixRedux(MatrixXcf(internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
145      CALL_SUBTEST_4( matrixRedux(ArrayXXcf(internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
146      CALL_SUBTEST_5( matrixRedux(MatrixXd (internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
147      CALL_SUBTEST_5( matrixRedux(ArrayXXd (internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
148      CALL_SUBTEST_6( matrixRedux(MatrixXi (internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
149      CALL_SUBTEST_6( matrixRedux(ArrayXXi (internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
150    }
151    for(int i = 0; i < g_repeat; i++) {
152      CALL_SUBTEST_7( vectorRedux(Vector4f()) );
153      CALL_SUBTEST_7( vectorRedux(Array4f()) );
154      CALL_SUBTEST_5( vectorRedux(VectorXd(internal::random<int>(1,maxsize))) );
155      CALL_SUBTEST_5( vectorRedux(ArrayXd(internal::random<int>(1,maxsize))) );
156      CALL_SUBTEST_8( vectorRedux(VectorXf(internal::random<int>(1,maxsize))) );
157      CALL_SUBTEST_8( vectorRedux(ArrayXf(internal::random<int>(1,maxsize))) );
158    }
159  }
160