1// Ceres Solver - A fast non-linear least squares minimizer
2// Copyright 2013 Google Inc. All rights reserved.
3// http://code.google.com/p/ceres-solver/
4//
5// Redistribution and use in source and binary forms, with or without
6// modification, are permitted provided that the following conditions are met:
7//
8// * Redistributions of source code must retain the above copyright notice,
9//   this list of conditions and the following disclaimer.
10// * Redistributions in binary form must reproduce the above copyright notice,
11//   this list of conditions and the following disclaimer in the documentation
12//   and/or other materials provided with the distribution.
13// * Neither the name of Google Inc. nor the names of its contributors may be
14//   used to endorse or promote products derived from this software without
15//   specific prior written permission.
16//
17// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27// POSSIBILITY OF SUCH DAMAGE.
28//
29// Author: sameeragarwal@google.com (Sameer Agarwal)
30
31#include <algorithm>
32#include "ceres/compressed_col_sparse_matrix_utils.h"
33#include "ceres/internal/port.h"
34#include "ceres/suitesparse.h"
35#include "ceres/triplet_sparse_matrix.h"
36#include "glog/logging.h"
37#include "gtest/gtest.h"
38
39namespace ceres {
40namespace internal {
41
42TEST(_, BlockPermutationToScalarPermutation) {
43  vector<int> blocks;
44  //  Block structure
45  //  0  --1-  ---2---  ---3---  4
46  // [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
47  blocks.push_back(1);
48  blocks.push_back(2);
49  blocks.push_back(3);
50  blocks.push_back(3);
51  blocks.push_back(1);
52
53  // Block ordering
54  // [1, 0, 2, 4, 5]
55  vector<int> block_ordering;
56  block_ordering.push_back(1);
57  block_ordering.push_back(0);
58  block_ordering.push_back(2);
59  block_ordering.push_back(4);
60  block_ordering.push_back(3);
61
62  // Expected ordering
63  // [1, 2, 0, 3, 4, 5, 9, 6, 7, 8]
64  vector<int> expected_scalar_ordering;
65  expected_scalar_ordering.push_back(1);
66  expected_scalar_ordering.push_back(2);
67  expected_scalar_ordering.push_back(0);
68  expected_scalar_ordering.push_back(3);
69  expected_scalar_ordering.push_back(4);
70  expected_scalar_ordering.push_back(5);
71  expected_scalar_ordering.push_back(9);
72  expected_scalar_ordering.push_back(6);
73  expected_scalar_ordering.push_back(7);
74  expected_scalar_ordering.push_back(8);
75
76  vector<int> scalar_ordering;
77  BlockOrderingToScalarOrdering(blocks,
78                                block_ordering,
79                                &scalar_ordering);
80  EXPECT_EQ(scalar_ordering.size(), expected_scalar_ordering.size());
81  for (int i = 0; i < expected_scalar_ordering.size(); ++i) {
82    EXPECT_EQ(scalar_ordering[i], expected_scalar_ordering[i]);
83  }
84}
85
86// Helper function to fill the sparsity pattern of a TripletSparseMatrix.
87int FillBlock(const vector<int>& row_blocks,
88              const vector<int>& col_blocks,
89              const int row_block_id,
90              const int col_block_id,
91              int* rows,
92              int* cols) {
93  int row_pos = 0;
94  for (int i = 0; i < row_block_id; ++i) {
95    row_pos += row_blocks[i];
96  }
97
98  int col_pos = 0;
99  for (int i = 0; i < col_block_id; ++i) {
100    col_pos += col_blocks[i];
101  }
102
103  int offset = 0;
104  for (int r = 0; r < row_blocks[row_block_id]; ++r) {
105    for (int c = 0; c < col_blocks[col_block_id]; ++c, ++offset) {
106      rows[offset] = row_pos + r;
107      cols[offset] = col_pos + c;
108    }
109  }
110  return offset;
111}
112
113TEST(_, ScalarMatrixToBlockMatrix) {
114  // Block sparsity.
115  //
116  //     [1 2 3 2]
117  // [1]  x   x
118  // [2]    x   x
119  // [2]  x x
120  // num_nonzeros = 1 + 3 + 4 + 4 + 1 + 2 = 15
121
122  vector<int> col_blocks;
123  col_blocks.push_back(1);
124  col_blocks.push_back(2);
125  col_blocks.push_back(3);
126  col_blocks.push_back(2);
127
128  vector<int> row_blocks;
129  row_blocks.push_back(1);
130  row_blocks.push_back(2);
131  row_blocks.push_back(2);
132
133  TripletSparseMatrix tsm(5, 8, 18);
134  int* rows = tsm.mutable_rows();
135  int* cols = tsm.mutable_cols();
136  fill(tsm.mutable_values(), tsm.mutable_values() + 18, 1.0);
137  int offset = 0;
138
139#define CERES_TEST_FILL_BLOCK(row_block_id, col_block_id) \
140  offset += FillBlock(row_blocks, col_blocks, \
141                      row_block_id, col_block_id, \
142                      rows + offset, cols + offset);
143
144  CERES_TEST_FILL_BLOCK(0, 0);
145  CERES_TEST_FILL_BLOCK(2, 0);
146  CERES_TEST_FILL_BLOCK(1, 1);
147  CERES_TEST_FILL_BLOCK(2, 1);
148  CERES_TEST_FILL_BLOCK(0, 2);
149  CERES_TEST_FILL_BLOCK(1, 3);
150#undef CERES_TEST_FILL_BLOCK
151
152  tsm.set_num_nonzeros(offset);
153
154  SuiteSparse ss;
155  scoped_ptr<cholmod_sparse> ccsm(ss.CreateSparseMatrix(&tsm));
156
157  vector<int> expected_block_rows;
158  expected_block_rows.push_back(0);
159  expected_block_rows.push_back(2);
160  expected_block_rows.push_back(1);
161  expected_block_rows.push_back(2);
162  expected_block_rows.push_back(0);
163  expected_block_rows.push_back(1);
164
165  vector<int> expected_block_cols;
166  expected_block_cols.push_back(0);
167  expected_block_cols.push_back(2);
168  expected_block_cols.push_back(4);
169  expected_block_cols.push_back(5);
170  expected_block_cols.push_back(6);
171
172  vector<int> block_rows;
173  vector<int> block_cols;
174  CompressedColumnScalarMatrixToBlockMatrix(
175      reinterpret_cast<const int*>(ccsm->i),
176      reinterpret_cast<const int*>(ccsm->p),
177      row_blocks,
178      col_blocks,
179      &block_rows,
180      &block_cols);
181
182  EXPECT_EQ(block_cols.size(), expected_block_cols.size());
183  EXPECT_EQ(block_rows.size(), expected_block_rows.size());
184
185  for (int i = 0; i < expected_block_cols.size(); ++i) {
186    EXPECT_EQ(block_cols[i], expected_block_cols[i]);
187  }
188
189  for (int i = 0; i < expected_block_rows.size(); ++i) {
190    EXPECT_EQ(block_rows[i], expected_block_rows[i]);
191  }
192
193  ss.Free(ccsm.release());
194}
195
196class SolveUpperTriangularTest : public ::testing::Test {
197 protected:
198  void SetUp() {
199    cols.resize(5);
200    rows.resize(7);
201    values.resize(7);
202
203    cols[0] = 0;
204    rows[0] = 0;
205    values[0] = 0.50754;
206
207    cols[1] = 1;
208    rows[1] = 1;
209    values[1] = 0.80483;
210
211    cols[2] = 2;
212    rows[2] = 1;
213    values[2] = 0.14120;
214    rows[3] = 2;
215    values[3] = 0.3;
216
217    cols[3] = 4;
218    rows[4] = 0;
219    values[4] = 0.77696;
220    rows[5] = 1;
221    values[5] = 0.41860;
222    rows[6] = 3;
223    values[6] = 0.88979;
224
225    cols[4] = 7;
226  }
227
228  vector<int> cols;
229  vector<int> rows;
230  vector<double> values;
231};
232
233TEST_F(SolveUpperTriangularTest, SolveInPlace) {
234  double rhs_and_solution[] = {1.0, 1.0, 2.0, 2.0};
235  const double expected[] = { -1.4706, -1.0962, 6.6667, 2.2477};
236
237  SolveUpperTriangularInPlace<int>(cols.size() - 1,
238                                   &rows[0],
239                                   &cols[0],
240                                   &values[0],
241                                   rhs_and_solution);
242
243  for (int i = 0; i < 4; ++i) {
244    EXPECT_NEAR(rhs_and_solution[i], expected[i], 1e-4) << i;
245  }
246}
247
248TEST_F(SolveUpperTriangularTest, TransposeSolveInPlace) {
249  double rhs_and_solution[] = {1.0, 1.0, 2.0, 2.0};
250  double expected[] = {1.970288,  1.242498,  6.081864, -0.057255};
251
252  SolveUpperTriangularTransposeInPlace<int>(cols.size() - 1,
253                                            &rows[0],
254                                            &cols[0],
255                                            &values[0],
256                                            rhs_and_solution);
257
258  for (int i = 0; i < 4; ++i) {
259    EXPECT_NEAR(rhs_and_solution[i], expected[i], 1e-4) << i;
260  }
261}
262
263TEST_F(SolveUpperTriangularTest, RTRSolveWithSparseRHS) {
264  double solution[4];
265  double expected[] = { 6.8420e+00,   1.0057e+00,  -1.4907e-16,  -1.9335e+00,
266                        1.0057e+00,   2.2275e+00,  -1.9493e+00,  -6.5693e-01,
267                       -1.4907e-16,  -1.9493e+00,   1.1111e+01,   9.7381e-17,
268                       -1.9335e+00,  -6.5693e-01,   9.7381e-17,   1.2631e+00 };
269
270  for (int i = 0; i < 4; ++i) {
271    SolveRTRWithSparseRHS<int>(cols.size() - 1,
272                               &rows[0],
273                               &cols[0],
274                               &values[0],
275                               i,
276                               solution);
277    for (int j = 0; j < 4; ++j) {
278      EXPECT_NEAR(solution[j], expected[4 * i + j], 1e-3) << i;
279    }
280  }
281}
282
283}  // namespace internal
284}  // namespace ceres
285