vectorwiseop.cpp (11489B)
1 // This file is part of Eigen, a lightweight C++ template library 2 // for linear algebra. 3 // 4 // Copyright (C) 2011 Benoit Jacob <jacob.benoit.1@gmail.com> 5 // Copyright (C) 2015 Gael Guennebaud <gael.guennebaud@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 #define TEST_ENABLE_TEMPORARY_TRACKING 12 #define EIGEN_NO_STATIC_ASSERT 13 14 #include "main.h" 15 16 template<typename ArrayType> void vectorwiseop_array(const ArrayType& m) 17 { 18 typedef typename ArrayType::Scalar Scalar; 19 typedef Array<Scalar, ArrayType::RowsAtCompileTime, 1> ColVectorType; 20 typedef Array<Scalar, 1, ArrayType::ColsAtCompileTime> RowVectorType; 21 22 Index rows = m.rows(); 23 Index cols = m.cols(); 24 Index r = internal::random<Index>(0, rows-1), 25 c = internal::random<Index>(0, cols-1); 26 27 ArrayType m1 = ArrayType::Random(rows, cols), 28 m2(rows, cols), 29 m3(rows, cols); 30 31 ColVectorType colvec = ColVectorType::Random(rows); 32 RowVectorType rowvec = RowVectorType::Random(cols); 33 34 // test addition 35 36 m2 = m1; 37 m2.colwise() += colvec; 38 VERIFY_IS_APPROX(m2, m1.colwise() + colvec); 39 VERIFY_IS_APPROX(m2.col(c), m1.col(c) + colvec); 40 41 VERIFY_RAISES_ASSERT(m2.colwise() += colvec.transpose()); 42 VERIFY_RAISES_ASSERT(m1.colwise() + colvec.transpose()); 43 44 m2 = m1; 45 m2.rowwise() += rowvec; 46 VERIFY_IS_APPROX(m2, m1.rowwise() + rowvec); 47 VERIFY_IS_APPROX(m2.row(r), m1.row(r) + rowvec); 48 49 VERIFY_RAISES_ASSERT(m2.rowwise() += rowvec.transpose()); 50 VERIFY_RAISES_ASSERT(m1.rowwise() + rowvec.transpose()); 51 52 // test substraction 53 54 m2 = m1; 55 m2.colwise() -= colvec; 56 VERIFY_IS_APPROX(m2, m1.colwise() - colvec); 57 VERIFY_IS_APPROX(m2.col(c), m1.col(c) - colvec); 58 59 VERIFY_RAISES_ASSERT(m2.colwise() -= colvec.transpose()); 60 VERIFY_RAISES_ASSERT(m1.colwise() - colvec.transpose()); 61 62 m2 = m1; 63 m2.rowwise() -= rowvec; 64 VERIFY_IS_APPROX(m2, m1.rowwise() - rowvec); 65 VERIFY_IS_APPROX(m2.row(r), m1.row(r) - rowvec); 66 67 VERIFY_RAISES_ASSERT(m2.rowwise() -= rowvec.transpose()); 68 VERIFY_RAISES_ASSERT(m1.rowwise() - rowvec.transpose()); 69 70 // test multiplication 71 72 m2 = m1; 73 m2.colwise() *= colvec; 74 VERIFY_IS_APPROX(m2, m1.colwise() * colvec); 75 VERIFY_IS_APPROX(m2.col(c), m1.col(c) * colvec); 76 77 VERIFY_RAISES_ASSERT(m2.colwise() *= colvec.transpose()); 78 VERIFY_RAISES_ASSERT(m1.colwise() * colvec.transpose()); 79 80 m2 = m1; 81 m2.rowwise() *= rowvec; 82 VERIFY_IS_APPROX(m2, m1.rowwise() * rowvec); 83 VERIFY_IS_APPROX(m2.row(r), m1.row(r) * rowvec); 84 85 VERIFY_RAISES_ASSERT(m2.rowwise() *= rowvec.transpose()); 86 VERIFY_RAISES_ASSERT(m1.rowwise() * rowvec.transpose()); 87 88 // test quotient 89 90 m2 = m1; 91 m2.colwise() /= colvec; 92 VERIFY_IS_APPROX(m2, m1.colwise() / colvec); 93 VERIFY_IS_APPROX(m2.col(c), m1.col(c) / colvec); 94 95 VERIFY_RAISES_ASSERT(m2.colwise() /= colvec.transpose()); 96 VERIFY_RAISES_ASSERT(m1.colwise() / colvec.transpose()); 97 98 m2 = m1; 99 m2.rowwise() /= rowvec; 100 VERIFY_IS_APPROX(m2, m1.rowwise() / rowvec); 101 VERIFY_IS_APPROX(m2.row(r), m1.row(r) / rowvec); 102 103 VERIFY_RAISES_ASSERT(m2.rowwise() /= rowvec.transpose()); 104 VERIFY_RAISES_ASSERT(m1.rowwise() / rowvec.transpose()); 105 106 m2 = m1; 107 // yes, there might be an aliasing issue there but ".rowwise() /=" 108 // is supposed to evaluate " m2.colwise().sum()" into a temporary to avoid 109 // evaluating the reduction multiple times 110 if(ArrayType::RowsAtCompileTime>2 || ArrayType::RowsAtCompileTime==Dynamic) 111 { 112 m2.rowwise() /= m2.colwise().sum(); 113 VERIFY_IS_APPROX(m2, m1.rowwise() / m1.colwise().sum()); 114 } 115 116 // all/any 117 Array<bool,Dynamic,Dynamic> mb(rows,cols); 118 mb = (m1.real()<=0.7).colwise().all(); 119 VERIFY( (mb.col(c) == (m1.real().col(c)<=0.7).all()).all() ); 120 mb = (m1.real()<=0.7).rowwise().all(); 121 VERIFY( (mb.row(r) == (m1.real().row(r)<=0.7).all()).all() ); 122 123 mb = (m1.real()>=0.7).colwise().any(); 124 VERIFY( (mb.col(c) == (m1.real().col(c)>=0.7).any()).all() ); 125 mb = (m1.real()>=0.7).rowwise().any(); 126 VERIFY( (mb.row(r) == (m1.real().row(r)>=0.7).any()).all() ); 127 } 128 129 template<typename MatrixType> void vectorwiseop_matrix(const MatrixType& m) 130 { 131 typedef typename MatrixType::Scalar Scalar; 132 typedef typename NumTraits<Scalar>::Real RealScalar; 133 typedef Matrix<Scalar, MatrixType::RowsAtCompileTime, 1> ColVectorType; 134 typedef Matrix<Scalar, 1, MatrixType::ColsAtCompileTime> RowVectorType; 135 typedef Matrix<RealScalar, MatrixType::RowsAtCompileTime, 1> RealColVectorType; 136 typedef Matrix<RealScalar, 1, MatrixType::ColsAtCompileTime> RealRowVectorType; 137 typedef Matrix<Scalar,Dynamic,Dynamic> MatrixX; 138 139 Index rows = m.rows(); 140 Index cols = m.cols(); 141 Index r = internal::random<Index>(0, rows-1), 142 c = internal::random<Index>(0, cols-1); 143 144 MatrixType m1 = MatrixType::Random(rows, cols), 145 m2(rows, cols), 146 m3(rows, cols); 147 148 ColVectorType colvec = ColVectorType::Random(rows); 149 RowVectorType rowvec = RowVectorType::Random(cols); 150 RealColVectorType rcres; 151 RealRowVectorType rrres; 152 153 // test broadcast assignment 154 m2 = m1; 155 m2.colwise() = colvec; 156 for(Index j=0; j<cols; ++j) 157 VERIFY_IS_APPROX(m2.col(j), colvec); 158 m2.rowwise() = rowvec; 159 for(Index i=0; i<rows; ++i) 160 VERIFY_IS_APPROX(m2.row(i), rowvec); 161 if(rows>1) 162 VERIFY_RAISES_ASSERT(m2.colwise() = colvec.transpose()); 163 if(cols>1) 164 VERIFY_RAISES_ASSERT(m2.rowwise() = rowvec.transpose()); 165 166 // test addition 167 168 m2 = m1; 169 m2.colwise() += colvec; 170 VERIFY_IS_APPROX(m2, m1.colwise() + colvec); 171 VERIFY_IS_APPROX(m2.col(c), m1.col(c) + colvec); 172 173 if(rows>1) 174 { 175 VERIFY_RAISES_ASSERT(m2.colwise() += colvec.transpose()); 176 VERIFY_RAISES_ASSERT(m1.colwise() + colvec.transpose()); 177 } 178 179 m2 = m1; 180 m2.rowwise() += rowvec; 181 VERIFY_IS_APPROX(m2, m1.rowwise() + rowvec); 182 VERIFY_IS_APPROX(m2.row(r), m1.row(r) + rowvec); 183 184 if(cols>1) 185 { 186 VERIFY_RAISES_ASSERT(m2.rowwise() += rowvec.transpose()); 187 VERIFY_RAISES_ASSERT(m1.rowwise() + rowvec.transpose()); 188 } 189 190 // test substraction 191 192 m2 = m1; 193 m2.colwise() -= colvec; 194 VERIFY_IS_APPROX(m2, m1.colwise() - colvec); 195 VERIFY_IS_APPROX(m2.col(c), m1.col(c) - colvec); 196 197 if(rows>1) 198 { 199 VERIFY_RAISES_ASSERT(m2.colwise() -= colvec.transpose()); 200 VERIFY_RAISES_ASSERT(m1.colwise() - colvec.transpose()); 201 } 202 203 m2 = m1; 204 m2.rowwise() -= rowvec; 205 VERIFY_IS_APPROX(m2, m1.rowwise() - rowvec); 206 VERIFY_IS_APPROX(m2.row(r), m1.row(r) - rowvec); 207 208 if(cols>1) 209 { 210 VERIFY_RAISES_ASSERT(m2.rowwise() -= rowvec.transpose()); 211 VERIFY_RAISES_ASSERT(m1.rowwise() - rowvec.transpose()); 212 } 213 214 // ------ partial reductions ------ 215 216 #define TEST_PARTIAL_REDUX_BASIC(FUNC,ROW,COL,PREPROCESS) { \ 217 ROW = m1 PREPROCESS .colwise().FUNC ; \ 218 for(Index k=0; k<cols; ++k) VERIFY_IS_APPROX(ROW(k), m1.col(k) PREPROCESS .FUNC ); \ 219 COL = m1 PREPROCESS .rowwise().FUNC ; \ 220 for(Index k=0; k<rows; ++k) VERIFY_IS_APPROX(COL(k), m1.row(k) PREPROCESS .FUNC ); \ 221 } 222 223 TEST_PARTIAL_REDUX_BASIC(sum(), rowvec,colvec,EIGEN_EMPTY); 224 TEST_PARTIAL_REDUX_BASIC(prod(), rowvec,colvec,EIGEN_EMPTY); 225 TEST_PARTIAL_REDUX_BASIC(mean(), rowvec,colvec,EIGEN_EMPTY); 226 TEST_PARTIAL_REDUX_BASIC(minCoeff(), rrres, rcres, .real()); 227 TEST_PARTIAL_REDUX_BASIC(maxCoeff(), rrres, rcres, .real()); 228 TEST_PARTIAL_REDUX_BASIC(norm(), rrres, rcres, EIGEN_EMPTY); 229 TEST_PARTIAL_REDUX_BASIC(squaredNorm(),rrres, rcres, EIGEN_EMPTY); 230 TEST_PARTIAL_REDUX_BASIC(redux(internal::scalar_sum_op<Scalar,Scalar>()),rowvec,colvec,EIGEN_EMPTY); 231 232 VERIFY_IS_APPROX(m1.cwiseAbs().colwise().sum(), m1.colwise().template lpNorm<1>()); 233 VERIFY_IS_APPROX(m1.cwiseAbs().rowwise().sum(), m1.rowwise().template lpNorm<1>()); 234 VERIFY_IS_APPROX(m1.cwiseAbs().colwise().maxCoeff(), m1.colwise().template lpNorm<Infinity>()); 235 VERIFY_IS_APPROX(m1.cwiseAbs().rowwise().maxCoeff(), m1.rowwise().template lpNorm<Infinity>()); 236 237 // regression for bug 1158 238 VERIFY_IS_APPROX(m1.cwiseAbs().colwise().sum().x(), m1.col(0).cwiseAbs().sum()); 239 240 // test normalized 241 m2 = m1.colwise().normalized(); 242 VERIFY_IS_APPROX(m2.col(c), m1.col(c).normalized()); 243 m2 = m1.rowwise().normalized(); 244 VERIFY_IS_APPROX(m2.row(r), m1.row(r).normalized()); 245 246 // test normalize 247 m2 = m1; 248 m2.colwise().normalize(); 249 VERIFY_IS_APPROX(m2.col(c), m1.col(c).normalized()); 250 m2 = m1; 251 m2.rowwise().normalize(); 252 VERIFY_IS_APPROX(m2.row(r), m1.row(r).normalized()); 253 254 // test with partial reduction of products 255 Matrix<Scalar,MatrixType::RowsAtCompileTime,MatrixType::RowsAtCompileTime> m1m1 = m1 * m1.transpose(); 256 VERIFY_IS_APPROX( (m1 * m1.transpose()).colwise().sum(), m1m1.colwise().sum()); 257 Matrix<Scalar,1,MatrixType::RowsAtCompileTime> tmp(rows); 258 VERIFY_EVALUATION_COUNT( tmp = (m1 * m1.transpose()).colwise().sum(), 1); 259 260 m2 = m1.rowwise() - (m1.colwise().sum()/RealScalar(m1.rows())).eval(); 261 m1 = m1.rowwise() - (m1.colwise().sum()/RealScalar(m1.rows())); 262 VERIFY_IS_APPROX( m1, m2 ); 263 VERIFY_EVALUATION_COUNT( m2 = (m1.rowwise() - m1.colwise().sum()/RealScalar(m1.rows())), (MatrixType::RowsAtCompileTime!=1 ? 1 : 0) ); 264 265 // test empty expressions 266 VERIFY_IS_APPROX(m1.matrix().middleCols(0,0).rowwise().sum().eval(), MatrixX::Zero(rows,1)); 267 VERIFY_IS_APPROX(m1.matrix().middleRows(0,0).colwise().sum().eval(), MatrixX::Zero(1,cols)); 268 VERIFY_IS_APPROX(m1.matrix().middleCols(0,fix<0>).rowwise().sum().eval(), MatrixX::Zero(rows,1)); 269 VERIFY_IS_APPROX(m1.matrix().middleRows(0,fix<0>).colwise().sum().eval(), MatrixX::Zero(1,cols)); 270 271 VERIFY_IS_APPROX(m1.matrix().middleCols(0,0).rowwise().prod().eval(), MatrixX::Ones(rows,1)); 272 VERIFY_IS_APPROX(m1.matrix().middleRows(0,0).colwise().prod().eval(), MatrixX::Ones(1,cols)); 273 VERIFY_IS_APPROX(m1.matrix().middleCols(0,fix<0>).rowwise().prod().eval(), MatrixX::Ones(rows,1)); 274 VERIFY_IS_APPROX(m1.matrix().middleRows(0,fix<0>).colwise().prod().eval(), MatrixX::Ones(1,cols)); 275 276 VERIFY_IS_APPROX(m1.matrix().middleCols(0,0).rowwise().squaredNorm().eval(), MatrixX::Zero(rows,1)); 277 278 VERIFY_RAISES_ASSERT(m1.real().middleCols(0,0).rowwise().minCoeff().eval()); 279 VERIFY_RAISES_ASSERT(m1.real().middleRows(0,0).colwise().maxCoeff().eval()); 280 VERIFY_IS_EQUAL(m1.real().middleRows(0,0).rowwise().maxCoeff().eval().rows(),0); 281 VERIFY_IS_EQUAL(m1.real().middleCols(0,0).colwise().maxCoeff().eval().cols(),0); 282 VERIFY_IS_EQUAL(m1.real().middleRows(0,fix<0>).rowwise().maxCoeff().eval().rows(),0); 283 VERIFY_IS_EQUAL(m1.real().middleCols(0,fix<0>).colwise().maxCoeff().eval().cols(),0); 284 } 285 286 EIGEN_DECLARE_TEST(vectorwiseop) 287 { 288 CALL_SUBTEST_1( vectorwiseop_array(Array22cd()) ); 289 CALL_SUBTEST_2( vectorwiseop_array(Array<double, 3, 2>()) ); 290 CALL_SUBTEST_3( vectorwiseop_array(ArrayXXf(3, 4)) ); 291 CALL_SUBTEST_4( vectorwiseop_matrix(Matrix4cf()) ); 292 CALL_SUBTEST_5( vectorwiseop_matrix(Matrix4f()) ); 293 CALL_SUBTEST_5( vectorwiseop_matrix(Vector4f()) ); 294 CALL_SUBTEST_5( vectorwiseop_matrix(Matrix<float,4,5>()) ); 295 CALL_SUBTEST_6( vectorwiseop_matrix(MatrixXd(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) ); 296 CALL_SUBTEST_7( vectorwiseop_matrix(VectorXd(internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) ); 297 CALL_SUBTEST_7( vectorwiseop_matrix(RowVectorXd(internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) ); 298 }