sparse_block.cpp
Go to the documentation of this file.
1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2008-2015 Gael Guennebaud <gael.guennebaud@inria.fr>
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 
12 template<typename T>
15 {
16  return A.row(i);
17 }
18 
19 template<typename T>
22 {
23  return A.col(i);
24 }
25 
26 template<typename SparseMatrixType> void sparse_block(const SparseMatrixType& ref)
27 {
28  const Index rows = ref.rows();
29  const Index cols = ref.cols();
30  const Index inner = ref.innerSize();
31  const Index outer = ref.outerSize();
32 
33  typedef typename SparseMatrixType::Scalar Scalar;
34  typedef typename SparseMatrixType::StorageIndex StorageIndex;
35 
36  double density = (std::max)(8./(rows*cols), 0.01);
39  typedef Matrix<Scalar,1,Dynamic> RowDenseVector;
40  typedef SparseVector<Scalar> SparseVectorType;
41 
42  Scalar s1 = internal::random<Scalar>();
43  {
44  SparseMatrixType m(rows, cols);
45  DenseMatrix refMat = DenseMatrix::Zero(rows, cols);
46  initSparse<Scalar>(density, refMat, m);
47 
48  VERIFY_IS_APPROX(m, refMat);
49 
50  // test InnerIterators and Block expressions
51  for (int t=0; t<10; ++t)
52  {
53  Index j = internal::random<Index>(0,cols-2);
54  Index i = internal::random<Index>(0,rows-2);
55  Index w = internal::random<Index>(1,cols-j);
56  Index h = internal::random<Index>(1,rows-i);
57 
58  VERIFY_IS_APPROX(m.block(i,j,h,w), refMat.block(i,j,h,w));
59  for(Index c=0; c<w; c++)
60  {
61  VERIFY_IS_APPROX(m.block(i,j,h,w).col(c), refMat.block(i,j,h,w).col(c));
62  for(Index r=0; r<h; r++)
63  {
64  VERIFY_IS_APPROX(m.block(i,j,h,w).col(c).coeff(r), refMat.block(i,j,h,w).col(c).coeff(r));
65  VERIFY_IS_APPROX(m.block(i,j,h,w).coeff(r,c), refMat.block(i,j,h,w).coeff(r,c));
66  }
67  }
68  for(Index r=0; r<h; r++)
69  {
70  VERIFY_IS_APPROX(m.block(i,j,h,w).row(r), refMat.block(i,j,h,w).row(r));
71  for(Index c=0; c<w; c++)
72  {
73  VERIFY_IS_APPROX(m.block(i,j,h,w).row(r).coeff(c), refMat.block(i,j,h,w).row(r).coeff(c));
74  VERIFY_IS_APPROX(m.block(i,j,h,w).coeff(r,c), refMat.block(i,j,h,w).coeff(r,c));
75  }
76  }
77 
78  VERIFY_IS_APPROX(m.middleCols(j,w), refMat.middleCols(j,w));
79  VERIFY_IS_APPROX(m.middleRows(i,h), refMat.middleRows(i,h));
80  for(Index r=0; r<h; r++)
81  {
82  VERIFY_IS_APPROX(m.middleCols(j,w).row(r), refMat.middleCols(j,w).row(r));
83  VERIFY_IS_APPROX(m.middleRows(i,h).row(r), refMat.middleRows(i,h).row(r));
84  for(Index c=0; c<w; c++)
85  {
86  VERIFY_IS_APPROX(m.col(c).coeff(r), refMat.col(c).coeff(r));
87  VERIFY_IS_APPROX(m.row(r).coeff(c), refMat.row(r).coeff(c));
88 
89  VERIFY_IS_APPROX(m.middleCols(j,w).coeff(r,c), refMat.middleCols(j,w).coeff(r,c));
90  VERIFY_IS_APPROX(m.middleRows(i,h).coeff(r,c), refMat.middleRows(i,h).coeff(r,c));
91  if(m.middleCols(j,w).coeff(r,c) != Scalar(0))
92  {
93  VERIFY_IS_APPROX(m.middleCols(j,w).coeffRef(r,c), refMat.middleCols(j,w).coeff(r,c));
94  }
95  if(m.middleRows(i,h).coeff(r,c) != Scalar(0))
96  {
97  VERIFY_IS_APPROX(m.middleRows(i,h).coeff(r,c), refMat.middleRows(i,h).coeff(r,c));
98  }
99  }
100  }
101  for(Index c=0; c<w; c++)
102  {
103  VERIFY_IS_APPROX(m.middleCols(j,w).col(c), refMat.middleCols(j,w).col(c));
104  VERIFY_IS_APPROX(m.middleRows(i,h).col(c), refMat.middleRows(i,h).col(c));
105  }
106  }
107 
108  for(Index c=0; c<cols; c++)
109  {
110  VERIFY_IS_APPROX(m.col(c) + m.col(c), (m + m).col(c));
111  VERIFY_IS_APPROX(m.col(c) + m.col(c), refMat.col(c) + refMat.col(c));
112  }
113 
114  for(Index r=0; r<rows; r++)
115  {
116  VERIFY_IS_APPROX(m.row(r) + m.row(r), (m + m).row(r));
117  VERIFY_IS_APPROX(m.row(r) + m.row(r), refMat.row(r) + refMat.row(r));
118  }
119  }
120 
121  // test innerVector()
122  {
123  DenseMatrix refMat2 = DenseMatrix::Zero(rows, cols);
124  SparseMatrixType m2(rows, cols);
125  initSparse<Scalar>(density, refMat2, m2);
126  Index j0 = internal::random<Index>(0,outer-1);
127  Index j1 = internal::random<Index>(0,outer-1);
128  Index r0 = internal::random<Index>(0,rows-1);
129  Index c0 = internal::random<Index>(0,cols-1);
130 
131  VERIFY_IS_APPROX(m2.innerVector(j0), innervec(refMat2,j0));
132  VERIFY_IS_APPROX(m2.innerVector(j0)+m2.innerVector(j1), innervec(refMat2,j0)+innervec(refMat2,j1));
133 
134  m2.innerVector(j0) *= Scalar(2);
135  innervec(refMat2,j0) *= Scalar(2);
136  VERIFY_IS_APPROX(m2, refMat2);
137 
138  m2.row(r0) *= Scalar(3);
139  refMat2.row(r0) *= Scalar(3);
140  VERIFY_IS_APPROX(m2, refMat2);
141 
142  m2.col(c0) *= Scalar(4);
143  refMat2.col(c0) *= Scalar(4);
144  VERIFY_IS_APPROX(m2, refMat2);
145 
146  m2.row(r0) /= Scalar(3);
147  refMat2.row(r0) /= Scalar(3);
148  VERIFY_IS_APPROX(m2, refMat2);
149 
150  m2.col(c0) /= Scalar(4);
151  refMat2.col(c0) /= Scalar(4);
152  VERIFY_IS_APPROX(m2, refMat2);
153 
154  SparseVectorType v1;
155  VERIFY_IS_APPROX(v1 = m2.col(c0) * 4, refMat2.col(c0)*4);
156  VERIFY_IS_APPROX(v1 = m2.row(r0) * 4, refMat2.row(r0).transpose()*4);
157 
158  SparseMatrixType m3(rows,cols);
159  m3.reserve(VectorXi::Constant(outer,int(inner/2)));
160  for(Index j=0; j<outer; ++j)
161  for(Index k=0; k<(std::min)(j,inner); ++k)
162  m3.insertByOuterInner(j,k) = internal::convert_index<StorageIndex>(k+1);
163  for(Index j=0; j<(std::min)(outer, inner); ++j)
164  {
165  VERIFY(j==numext::real(m3.innerVector(j).nonZeros()));
166  if(j>0)
167  VERIFY(j==numext::real(m3.innerVector(j).lastCoeff()));
168  }
169  m3.makeCompressed();
170  for(Index j=0; j<(std::min)(outer, inner); ++j)
171  {
172  VERIFY(j==numext::real(m3.innerVector(j).nonZeros()));
173  if(j>0)
174  VERIFY(j==numext::real(m3.innerVector(j).lastCoeff()));
175  }
176 
177  VERIFY(m3.innerVector(j0).nonZeros() == m3.transpose().innerVector(j0).nonZeros());
178 
179 // m2.innerVector(j0) = 2*m2.innerVector(j1);
180 // refMat2.col(j0) = 2*refMat2.col(j1);
181 // VERIFY_IS_APPROX(m2, refMat2);
182  }
183 
184  // test innerVectors()
185  {
186  DenseMatrix refMat2 = DenseMatrix::Zero(rows, cols);
187  SparseMatrixType m2(rows, cols);
188  initSparse<Scalar>(density, refMat2, m2);
189  if(internal::random<float>(0,1)>0.5f) m2.makeCompressed();
190  Index j0 = internal::random<Index>(0,outer-2);
191  Index j1 = internal::random<Index>(0,outer-2);
192  Index n0 = internal::random<Index>(1,outer-(std::max)(j0,j1));
193  if(SparseMatrixType::IsRowMajor)
194  VERIFY_IS_APPROX(m2.innerVectors(j0,n0), refMat2.block(j0,0,n0,cols));
195  else
196  VERIFY_IS_APPROX(m2.innerVectors(j0,n0), refMat2.block(0,j0,rows,n0));
197  if(SparseMatrixType::IsRowMajor)
198  VERIFY_IS_APPROX(m2.innerVectors(j0,n0)+m2.innerVectors(j1,n0),
199  refMat2.middleRows(j0,n0)+refMat2.middleRows(j1,n0));
200  else
201  VERIFY_IS_APPROX(m2.innerVectors(j0,n0)+m2.innerVectors(j1,n0),
202  refMat2.block(0,j0,rows,n0)+refMat2.block(0,j1,rows,n0));
203 
204  VERIFY_IS_APPROX(m2, refMat2);
205 
206  VERIFY(m2.innerVectors(j0,n0).nonZeros() == m2.transpose().innerVectors(j0,n0).nonZeros());
207 
208  m2.innerVectors(j0,n0) = m2.innerVectors(j0,n0) + m2.innerVectors(j1,n0);
209  if(SparseMatrixType::IsRowMajor)
210  refMat2.middleRows(j0,n0) = (refMat2.middleRows(j0,n0) + refMat2.middleRows(j1,n0)).eval();
211  else
212  refMat2.middleCols(j0,n0) = (refMat2.middleCols(j0,n0) + refMat2.middleCols(j1,n0)).eval();
213 
214  VERIFY_IS_APPROX(m2, refMat2);
215  }
216 
217  // test generic blocks
218  {
219  DenseMatrix refMat2 = DenseMatrix::Zero(rows, cols);
220  SparseMatrixType m2(rows, cols);
221  initSparse<Scalar>(density, refMat2, m2);
222  Index j0 = internal::random<Index>(0,outer-2);
223  Index j1 = internal::random<Index>(0,outer-2);
224  Index n0 = internal::random<Index>(1,outer-(std::max)(j0,j1));
225  if(SparseMatrixType::IsRowMajor)
226  VERIFY_IS_APPROX(m2.block(j0,0,n0,cols), refMat2.block(j0,0,n0,cols));
227  else
228  VERIFY_IS_APPROX(m2.block(0,j0,rows,n0), refMat2.block(0,j0,rows,n0));
229 
230  if(SparseMatrixType::IsRowMajor)
231  VERIFY_IS_APPROX(m2.block(j0,0,n0,cols)+m2.block(j1,0,n0,cols),
232  refMat2.block(j0,0,n0,cols)+refMat2.block(j1,0,n0,cols));
233  else
234  VERIFY_IS_APPROX(m2.block(0,j0,rows,n0)+m2.block(0,j1,rows,n0),
235  refMat2.block(0,j0,rows,n0)+refMat2.block(0,j1,rows,n0));
236 
237  Index i = internal::random<Index>(0,m2.outerSize()-1);
238  if(SparseMatrixType::IsRowMajor) {
239  m2.innerVector(i) = m2.innerVector(i) * s1;
240  refMat2.row(i) = refMat2.row(i) * s1;
241  VERIFY_IS_APPROX(m2,refMat2);
242  } else {
243  m2.innerVector(i) = m2.innerVector(i) * s1;
244  refMat2.col(i) = refMat2.col(i) * s1;
245  VERIFY_IS_APPROX(m2,refMat2);
246  }
247 
248  Index r0 = internal::random<Index>(0,rows-2);
249  Index c0 = internal::random<Index>(0,cols-2);
250  Index r1 = internal::random<Index>(1,rows-r0);
251  Index c1 = internal::random<Index>(1,cols-c0);
252 
253  VERIFY_IS_APPROX(DenseVector(m2.col(c0)), refMat2.col(c0));
254  VERIFY_IS_APPROX(m2.col(c0), refMat2.col(c0));
255 
256  VERIFY_IS_APPROX(RowDenseVector(m2.row(r0)), refMat2.row(r0));
257  VERIFY_IS_APPROX(m2.row(r0), refMat2.row(r0));
258 
259  VERIFY_IS_APPROX(m2.block(r0,c0,r1,c1), refMat2.block(r0,c0,r1,c1));
260  VERIFY_IS_APPROX((2*m2).block(r0,c0,r1,c1), (2*refMat2).block(r0,c0,r1,c1));
261 
262  if(m2.nonZeros()>0)
263  {
264  VERIFY_IS_APPROX(m2, refMat2);
265  SparseMatrixType m3(rows, cols);
266  DenseMatrix refMat3(rows, cols); refMat3.setZero();
267  Index n = internal::random<Index>(1,10);
268  for(Index k=0; k<n; ++k)
269  {
270  Index o1 = internal::random<Index>(0,outer-1);
271  Index o2 = internal::random<Index>(0,outer-1);
272  if(SparseMatrixType::IsRowMajor)
273  {
274  m3.innerVector(o1) = m2.row(o2);
275  refMat3.row(o1) = refMat2.row(o2);
276  }
277  else
278  {
279  m3.innerVector(o1) = m2.col(o2);
280  refMat3.col(o1) = refMat2.col(o2);
281  }
282  if(internal::random<bool>())
283  m3.makeCompressed();
284  }
285  if(m3.nonZeros()>0)
286  VERIFY_IS_APPROX(m3, refMat3);
287  }
288  }
289 }
290 
292 {
293  for(int i = 0; i < g_repeat; i++) {
294  int r = Eigen::internal::random<int>(1,200), c = Eigen::internal::random<int>(1,200);
295  if(Eigen::internal::random<int>(0,4) == 0) {
296  r = c; // check square matrices in 25% of tries
297  }
299  CALL_SUBTEST_1(( sparse_block(SparseMatrix<double>(1, 1)) ));
300  CALL_SUBTEST_1(( sparse_block(SparseMatrix<double>(8, 8)) ));
301  CALL_SUBTEST_1(( sparse_block(SparseMatrix<double>(r, c)) ));
302  CALL_SUBTEST_2(( sparse_block(SparseMatrix<std::complex<double>, ColMajor>(r, c)) ));
303  CALL_SUBTEST_2(( sparse_block(SparseMatrix<std::complex<double>, RowMajor>(r, c)) ));
304 
305  CALL_SUBTEST_3(( sparse_block(SparseMatrix<double,ColMajor,long int>(r, c)) ));
306  CALL_SUBTEST_3(( sparse_block(SparseMatrix<double,RowMajor,long int>(r, c)) ));
307 
308  r = Eigen::internal::random<int>(1,100);
309  c = Eigen::internal::random<int>(1,100);
310  if(Eigen::internal::random<int>(0,4) == 0) {
311  r = c; // check square matrices in 25% of tries
312  }
313 
314  CALL_SUBTEST_4(( sparse_block(SparseMatrix<double,ColMajor,short int>(short(r), short(c))) ));
315  CALL_SUBTEST_4(( sparse_block(SparseMatrix<double,RowMajor,short int>(short(r), short(c))) ));
316  }
317 }
Matrix3f m
Matrix< Scalar, Dynamic, Dynamic > DenseMatrix
SCALAR Scalar
Definition: bench_gemm.cpp:33
#define max(a, b)
Definition: datatypes.h:20
float real
Definition: datatypes.h:10
m m block(1, 0, 2, 2)<< 4
Vector v1
const StorageIndex & row() const
Definition: SparseUtil.h:164
#define min(a, b)
Definition: datatypes.h:19
MatrixType m2(n_dims)
int n
Scalar Scalar * c
Definition: benchVecAdd.cpp:17
Matrix< Scalar, Dynamic, 1 > DenseVector
const StorageIndex & col() const
Definition: SparseUtil.h:167
#define VERIFY_IS_APPROX(a, b)
m row(1)
a sparse vector class
Definition: SparseUtil.h:54
static int g_repeat
Definition: main.h:144
EIGEN_DEFAULT_DENSE_INDEX_TYPE Index
The Index type as used for the API.
Definition: Meta.h:33
void test_sparse_block()
RowVector3d w
const double h
static const double r1
#define VERIFY(a)
Definition: main.h:325
Eigen::internal::enable_if<(T::Flags &RowMajorBit)==RowMajorBit, typename T::RowXpr >::type innervec(T &A, Index i)
m col(1)
internal::nested_eval< T, 1 >::type eval(const T &xpr)
The matrix class, also used for vectors and row-vectors.
std::ptrdiff_t j
void sparse_block(const SparseMatrixType &ref)
static const Key c1
#define EIGEN_UNUSED_VARIABLE(var)
Definition: Macros.h:618
Point2 t(10, 10)


gtsam
Author(s):
autogenerated on Sat May 8 2021 02:44:17