product_extra.cpp (15711B)
1 // This file is part of Eigen, a lightweight C++ template library 2 // for linear algebra. 3 // 4 // Copyright (C) 2006-2008 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 12 template<typename MatrixType> void product_extra(const MatrixType& m) 13 { 14 typedef typename MatrixType::Scalar Scalar; 15 typedef Matrix<Scalar, 1, Dynamic> RowVectorType; 16 typedef Matrix<Scalar, Dynamic, 1> ColVectorType; 17 typedef Matrix<Scalar, Dynamic, Dynamic, 18 MatrixType::Flags&RowMajorBit> OtherMajorMatrixType; 19 20 Index rows = m.rows(); 21 Index cols = m.cols(); 22 23 MatrixType m1 = MatrixType::Random(rows, cols), 24 m2 = MatrixType::Random(rows, cols), 25 m3(rows, cols), 26 mzero = MatrixType::Zero(rows, cols), 27 identity = MatrixType::Identity(rows, rows), 28 square = MatrixType::Random(rows, rows), 29 res = MatrixType::Random(rows, rows), 30 square2 = MatrixType::Random(cols, cols), 31 res2 = MatrixType::Random(cols, cols); 32 RowVectorType v1 = RowVectorType::Random(rows), vrres(rows); 33 ColVectorType vc2 = ColVectorType::Random(cols), vcres(cols); 34 OtherMajorMatrixType tm1 = m1; 35 36 Scalar s1 = internal::random<Scalar>(), 37 s2 = internal::random<Scalar>(), 38 s3 = internal::random<Scalar>(); 39 40 VERIFY_IS_APPROX(m3.noalias() = m1 * m2.adjoint(), m1 * m2.adjoint().eval()); 41 VERIFY_IS_APPROX(m3.noalias() = m1.adjoint() * square.adjoint(), m1.adjoint().eval() * square.adjoint().eval()); 42 VERIFY_IS_APPROX(m3.noalias() = m1.adjoint() * m2, m1.adjoint().eval() * m2); 43 VERIFY_IS_APPROX(m3.noalias() = (s1 * m1.adjoint()) * m2, (s1 * m1.adjoint()).eval() * m2); 44 VERIFY_IS_APPROX(m3.noalias() = ((s1 * m1).adjoint()) * m2, (numext::conj(s1) * m1.adjoint()).eval() * m2); 45 VERIFY_IS_APPROX(m3.noalias() = (- m1.adjoint() * s1) * (s3 * m2), (- m1.adjoint() * s1).eval() * (s3 * m2).eval()); 46 VERIFY_IS_APPROX(m3.noalias() = (s2 * m1.adjoint() * s1) * m2, (s2 * m1.adjoint() * s1).eval() * m2); 47 VERIFY_IS_APPROX(m3.noalias() = (-m1*s2) * s1*m2.adjoint(), (-m1*s2).eval() * (s1*m2.adjoint()).eval()); 48 49 // a very tricky case where a scale factor has to be automatically conjugated: 50 VERIFY_IS_APPROX( m1.adjoint() * (s1*m2).conjugate(), (m1.adjoint()).eval() * ((s1*m2).conjugate()).eval()); 51 52 53 // test all possible conjugate combinations for the four matrix-vector product cases: 54 55 VERIFY_IS_APPROX((-m1.conjugate() * s2) * (s1 * vc2), 56 (-m1.conjugate()*s2).eval() * (s1 * vc2).eval()); 57 VERIFY_IS_APPROX((-m1 * s2) * (s1 * vc2.conjugate()), 58 (-m1*s2).eval() * (s1 * vc2.conjugate()).eval()); 59 VERIFY_IS_APPROX((-m1.conjugate() * s2) * (s1 * vc2.conjugate()), 60 (-m1.conjugate()*s2).eval() * (s1 * vc2.conjugate()).eval()); 61 62 VERIFY_IS_APPROX((s1 * vc2.transpose()) * (-m1.adjoint() * s2), 63 (s1 * vc2.transpose()).eval() * (-m1.adjoint()*s2).eval()); 64 VERIFY_IS_APPROX((s1 * vc2.adjoint()) * (-m1.transpose() * s2), 65 (s1 * vc2.adjoint()).eval() * (-m1.transpose()*s2).eval()); 66 VERIFY_IS_APPROX((s1 * vc2.adjoint()) * (-m1.adjoint() * s2), 67 (s1 * vc2.adjoint()).eval() * (-m1.adjoint()*s2).eval()); 68 69 VERIFY_IS_APPROX((-m1.adjoint() * s2) * (s1 * v1.transpose()), 70 (-m1.adjoint()*s2).eval() * (s1 * v1.transpose()).eval()); 71 VERIFY_IS_APPROX((-m1.transpose() * s2) * (s1 * v1.adjoint()), 72 (-m1.transpose()*s2).eval() * (s1 * v1.adjoint()).eval()); 73 VERIFY_IS_APPROX((-m1.adjoint() * s2) * (s1 * v1.adjoint()), 74 (-m1.adjoint()*s2).eval() * (s1 * v1.adjoint()).eval()); 75 76 VERIFY_IS_APPROX((s1 * v1) * (-m1.conjugate() * s2), 77 (s1 * v1).eval() * (-m1.conjugate()*s2).eval()); 78 VERIFY_IS_APPROX((s1 * v1.conjugate()) * (-m1 * s2), 79 (s1 * v1.conjugate()).eval() * (-m1*s2).eval()); 80 VERIFY_IS_APPROX((s1 * v1.conjugate()) * (-m1.conjugate() * s2), 81 (s1 * v1.conjugate()).eval() * (-m1.conjugate()*s2).eval()); 82 83 VERIFY_IS_APPROX((-m1.adjoint() * s2) * (s1 * v1.adjoint()), 84 (-m1.adjoint()*s2).eval() * (s1 * v1.adjoint()).eval()); 85 86 // test the vector-matrix product with non aligned starts 87 Index i = internal::random<Index>(0,m1.rows()-2); 88 Index j = internal::random<Index>(0,m1.cols()-2); 89 Index r = internal::random<Index>(1,m1.rows()-i); 90 Index c = internal::random<Index>(1,m1.cols()-j); 91 Index i2 = internal::random<Index>(0,m1.rows()-1); 92 Index j2 = internal::random<Index>(0,m1.cols()-1); 93 94 VERIFY_IS_APPROX(m1.col(j2).adjoint() * m1.block(0,j,m1.rows(),c), m1.col(j2).adjoint().eval() * m1.block(0,j,m1.rows(),c).eval()); 95 VERIFY_IS_APPROX(m1.block(i,0,r,m1.cols()) * m1.row(i2).adjoint(), m1.block(i,0,r,m1.cols()).eval() * m1.row(i2).adjoint().eval()); 96 97 // test negative strides 98 { 99 Map<MatrixType,Unaligned,Stride<Dynamic,Dynamic> > map1(&m1(rows-1,cols-1), rows, cols, Stride<Dynamic,Dynamic>(-m1.outerStride(),-1)); 100 Map<MatrixType,Unaligned,Stride<Dynamic,Dynamic> > map2(&m2(rows-1,cols-1), rows, cols, Stride<Dynamic,Dynamic>(-m2.outerStride(),-1)); 101 Map<RowVectorType,Unaligned,InnerStride<-1> > mapv1(&v1(v1.size()-1), v1.size(), InnerStride<-1>(-1)); 102 Map<ColVectorType,Unaligned,InnerStride<-1> > mapvc2(&vc2(vc2.size()-1), vc2.size(), InnerStride<-1>(-1)); 103 VERIFY_IS_APPROX(MatrixType(map1), m1.reverse()); 104 VERIFY_IS_APPROX(MatrixType(map2), m2.reverse()); 105 VERIFY_IS_APPROX(m3.noalias() = MatrixType(map1) * MatrixType(map2).adjoint(), m1.reverse() * m2.reverse().adjoint()); 106 VERIFY_IS_APPROX(m3.noalias() = map1 * map2.adjoint(), m1.reverse() * m2.reverse().adjoint()); 107 VERIFY_IS_APPROX(map1 * vc2, m1.reverse() * vc2); 108 VERIFY_IS_APPROX(m1 * mapvc2, m1 * mapvc2); 109 VERIFY_IS_APPROX(map1.adjoint() * v1.transpose(), m1.adjoint().reverse() * v1.transpose()); 110 VERIFY_IS_APPROX(m1.adjoint() * mapv1.transpose(), m1.adjoint() * v1.reverse().transpose()); 111 } 112 113 // regression test 114 MatrixType tmp = m1 * m1.adjoint() * s1; 115 VERIFY_IS_APPROX(tmp, m1 * m1.adjoint() * s1); 116 117 // regression test for bug 1343, assignment to arrays 118 Array<Scalar,Dynamic,1> a1 = m1 * vc2; 119 VERIFY_IS_APPROX(a1.matrix(),m1*vc2); 120 Array<Scalar,Dynamic,1> a2 = s1 * (m1 * vc2); 121 VERIFY_IS_APPROX(a2.matrix(),s1*m1*vc2); 122 Array<Scalar,1,Dynamic> a3 = v1 * m1; 123 VERIFY_IS_APPROX(a3.matrix(),v1*m1); 124 Array<Scalar,Dynamic,Dynamic> a4 = m1 * m2.adjoint(); 125 VERIFY_IS_APPROX(a4.matrix(),m1*m2.adjoint()); 126 } 127 128 // Regression test for bug reported at http://forum.kde.org/viewtopic.php?f=74&t=96947 129 void mat_mat_scalar_scalar_product() 130 { 131 Eigen::Matrix2Xd dNdxy(2, 3); 132 dNdxy << -0.5, 0.5, 0, 133 -0.3, 0, 0.3; 134 double det = 6.0, wt = 0.5; 135 VERIFY_IS_APPROX(dNdxy.transpose()*dNdxy*det*wt, det*wt*dNdxy.transpose()*dNdxy); 136 } 137 138 template <typename MatrixType> 139 void zero_sized_objects(const MatrixType& m) 140 { 141 typedef typename MatrixType::Scalar Scalar; 142 const int PacketSize = internal::packet_traits<Scalar>::size; 143 const int PacketSize1 = PacketSize>1 ? PacketSize-1 : 1; 144 Index rows = m.rows(); 145 Index cols = m.cols(); 146 147 { 148 MatrixType res, a(rows,0), b(0,cols); 149 VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(rows,cols) ); 150 VERIFY_IS_APPROX( (res=a*a.transpose()), MatrixType::Zero(rows,rows) ); 151 VERIFY_IS_APPROX( (res=b.transpose()*b), MatrixType::Zero(cols,cols) ); 152 VERIFY_IS_APPROX( (res=b.transpose()*a.transpose()), MatrixType::Zero(cols,rows) ); 153 } 154 155 { 156 MatrixType res, a(rows,cols), b(cols,0); 157 res = a*b; 158 VERIFY(res.rows()==rows && res.cols()==0); 159 b.resize(0,rows); 160 res = b*a; 161 VERIFY(res.rows()==0 && res.cols()==cols); 162 } 163 164 { 165 Matrix<Scalar,PacketSize,0> a; 166 Matrix<Scalar,0,1> b; 167 Matrix<Scalar,PacketSize,1> res; 168 VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(PacketSize,1) ); 169 VERIFY_IS_APPROX( (res=a.lazyProduct(b)), MatrixType::Zero(PacketSize,1) ); 170 } 171 172 { 173 Matrix<Scalar,PacketSize1,0> a; 174 Matrix<Scalar,0,1> b; 175 Matrix<Scalar,PacketSize1,1> res; 176 VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(PacketSize1,1) ); 177 VERIFY_IS_APPROX( (res=a.lazyProduct(b)), MatrixType::Zero(PacketSize1,1) ); 178 } 179 180 { 181 Matrix<Scalar,PacketSize,Dynamic> a(PacketSize,0); 182 Matrix<Scalar,Dynamic,1> b(0,1); 183 Matrix<Scalar,PacketSize,1> res; 184 VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(PacketSize,1) ); 185 VERIFY_IS_APPROX( (res=a.lazyProduct(b)), MatrixType::Zero(PacketSize,1) ); 186 } 187 188 { 189 Matrix<Scalar,PacketSize1,Dynamic> a(PacketSize1,0); 190 Matrix<Scalar,Dynamic,1> b(0,1); 191 Matrix<Scalar,PacketSize1,1> res; 192 VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(PacketSize1,1) ); 193 VERIFY_IS_APPROX( (res=a.lazyProduct(b)), MatrixType::Zero(PacketSize1,1) ); 194 } 195 } 196 197 template<int> 198 void bug_127() 199 { 200 // Bug 127 201 // 202 // a product of the form lhs*rhs with 203 // 204 // lhs: 205 // rows = 1, cols = 4 206 // RowsAtCompileTime = 1, ColsAtCompileTime = -1 207 // MaxRowsAtCompileTime = 1, MaxColsAtCompileTime = 5 208 // 209 // rhs: 210 // rows = 4, cols = 0 211 // RowsAtCompileTime = -1, ColsAtCompileTime = -1 212 // MaxRowsAtCompileTime = 5, MaxColsAtCompileTime = 1 213 // 214 // was failing on a runtime assertion, because it had been mis-compiled as a dot product because Product.h was using the 215 // max-sizes to detect size 1 indicating vectors, and that didn't account for 0-sized object with max-size 1. 216 217 Matrix<float,1,Dynamic,RowMajor,1,5> a(1,4); 218 Matrix<float,Dynamic,Dynamic,ColMajor,5,1> b(4,0); 219 a*b; 220 } 221 222 template<int> void bug_817() 223 { 224 ArrayXXf B = ArrayXXf::Random(10,10), C; 225 VectorXf x = VectorXf::Random(10); 226 C = (x.transpose()*B.matrix()); 227 B = (x.transpose()*B.matrix()); 228 VERIFY_IS_APPROX(B,C); 229 } 230 231 template<int> 232 void unaligned_objects() 233 { 234 // Regression test for the bug reported here: 235 // http://forum.kde.org/viewtopic.php?f=74&t=107541 236 // Recall the matrix*vector kernel avoid unaligned loads by loading two packets and then reassemble then. 237 // There was a mistake in the computation of the valid range for fully unaligned objects: in some rare cases, 238 // memory was read outside the allocated matrix memory. Though the values were not used, this might raise segfault. 239 for(int m=450;m<460;++m) 240 { 241 for(int n=8;n<12;++n) 242 { 243 MatrixXf M(m, n); 244 VectorXf v1(n), r1(500); 245 RowVectorXf v2(m), r2(16); 246 247 M.setRandom(); 248 v1.setRandom(); 249 v2.setRandom(); 250 for(int o=0; o<4; ++o) 251 { 252 r1.segment(o,m).noalias() = M * v1; 253 VERIFY_IS_APPROX(r1.segment(o,m), M * MatrixXf(v1)); 254 r2.segment(o,n).noalias() = v2 * M; 255 VERIFY_IS_APPROX(r2.segment(o,n), MatrixXf(v2) * M); 256 } 257 } 258 } 259 } 260 261 template<typename T> 262 EIGEN_DONT_INLINE 263 Index test_compute_block_size(Index m, Index n, Index k) 264 { 265 Index mc(m), nc(n), kc(k); 266 internal::computeProductBlockingSizes<T,T>(kc, mc, nc); 267 return kc+mc+nc; 268 } 269 270 template<typename T> 271 Index compute_block_size() 272 { 273 Index ret = 0; 274 ret += test_compute_block_size<T>(0,1,1); 275 ret += test_compute_block_size<T>(1,0,1); 276 ret += test_compute_block_size<T>(1,1,0); 277 ret += test_compute_block_size<T>(0,0,1); 278 ret += test_compute_block_size<T>(0,1,0); 279 ret += test_compute_block_size<T>(1,0,0); 280 ret += test_compute_block_size<T>(0,0,0); 281 return ret; 282 } 283 284 template<typename> 285 void aliasing_with_resize() 286 { 287 Index m = internal::random<Index>(10,50); 288 Index n = internal::random<Index>(10,50); 289 MatrixXd A, B, C(m,n), D(m,m); 290 VectorXd a, b, c(n); 291 C.setRandom(); 292 D.setRandom(); 293 c.setRandom(); 294 double s = internal::random<double>(1,10); 295 296 A = C; 297 B = A * A.transpose(); 298 A = A * A.transpose(); 299 VERIFY_IS_APPROX(A,B); 300 301 A = C; 302 B = (A * A.transpose())/s; 303 A = (A * A.transpose())/s; 304 VERIFY_IS_APPROX(A,B); 305 306 A = C; 307 B = (A * A.transpose()) + D; 308 A = (A * A.transpose()) + D; 309 VERIFY_IS_APPROX(A,B); 310 311 A = C; 312 B = D + (A * A.transpose()); 313 A = D + (A * A.transpose()); 314 VERIFY_IS_APPROX(A,B); 315 316 A = C; 317 B = s * (A * A.transpose()); 318 A = s * (A * A.transpose()); 319 VERIFY_IS_APPROX(A,B); 320 321 A = C; 322 a = c; 323 b = (A * a)/s; 324 a = (A * a)/s; 325 VERIFY_IS_APPROX(a,b); 326 } 327 328 template<int> 329 void bug_1308() 330 { 331 int n = 10; 332 MatrixXd r(n,n); 333 VectorXd v = VectorXd::Random(n); 334 r = v * RowVectorXd::Ones(n); 335 VERIFY_IS_APPROX(r, v.rowwise().replicate(n)); 336 r = VectorXd::Ones(n) * v.transpose(); 337 VERIFY_IS_APPROX(r, v.rowwise().replicate(n).transpose()); 338 339 Matrix4d ones44 = Matrix4d::Ones(); 340 Matrix4d m44 = Matrix4d::Ones() * Matrix4d::Ones(); 341 VERIFY_IS_APPROX(m44,Matrix4d::Constant(4)); 342 VERIFY_IS_APPROX(m44.noalias()=ones44*Matrix4d::Ones(), Matrix4d::Constant(4)); 343 VERIFY_IS_APPROX(m44.noalias()=ones44.transpose()*Matrix4d::Ones(), Matrix4d::Constant(4)); 344 VERIFY_IS_APPROX(m44.noalias()=Matrix4d::Ones()*ones44, Matrix4d::Constant(4)); 345 VERIFY_IS_APPROX(m44.noalias()=Matrix4d::Ones()*ones44.transpose(), Matrix4d::Constant(4)); 346 347 typedef Matrix<double,4,4,RowMajor> RMatrix4d; 348 RMatrix4d r44 = Matrix4d::Ones() * Matrix4d::Ones(); 349 VERIFY_IS_APPROX(r44,Matrix4d::Constant(4)); 350 VERIFY_IS_APPROX(r44.noalias()=ones44*Matrix4d::Ones(), Matrix4d::Constant(4)); 351 VERIFY_IS_APPROX(r44.noalias()=ones44.transpose()*Matrix4d::Ones(), Matrix4d::Constant(4)); 352 VERIFY_IS_APPROX(r44.noalias()=Matrix4d::Ones()*ones44, Matrix4d::Constant(4)); 353 VERIFY_IS_APPROX(r44.noalias()=Matrix4d::Ones()*ones44.transpose(), Matrix4d::Constant(4)); 354 VERIFY_IS_APPROX(r44.noalias()=ones44*RMatrix4d::Ones(), Matrix4d::Constant(4)); 355 VERIFY_IS_APPROX(r44.noalias()=ones44.transpose()*RMatrix4d::Ones(), Matrix4d::Constant(4)); 356 VERIFY_IS_APPROX(r44.noalias()=RMatrix4d::Ones()*ones44, Matrix4d::Constant(4)); 357 VERIFY_IS_APPROX(r44.noalias()=RMatrix4d::Ones()*ones44.transpose(), Matrix4d::Constant(4)); 358 359 // RowVector4d r4; 360 m44.setOnes(); 361 r44.setZero(); 362 VERIFY_IS_APPROX(r44.noalias() += m44.row(0).transpose() * RowVector4d::Ones(), ones44); 363 r44.setZero(); 364 VERIFY_IS_APPROX(r44.noalias() += m44.col(0) * RowVector4d::Ones(), ones44); 365 r44.setZero(); 366 VERIFY_IS_APPROX(r44.noalias() += Vector4d::Ones() * m44.row(0), ones44); 367 r44.setZero(); 368 VERIFY_IS_APPROX(r44.noalias() += Vector4d::Ones() * m44.col(0).transpose(), ones44); 369 } 370 371 EIGEN_DECLARE_TEST(product_extra) 372 { 373 for(int i = 0; i < g_repeat; i++) { 374 CALL_SUBTEST_1( product_extra(MatrixXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) ); 375 CALL_SUBTEST_2( product_extra(MatrixXd(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) ); 376 CALL_SUBTEST_2( mat_mat_scalar_scalar_product() ); 377 CALL_SUBTEST_3( product_extra(MatrixXcf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE/2), internal::random<int>(1,EIGEN_TEST_MAX_SIZE/2))) ); 378 CALL_SUBTEST_4( product_extra(MatrixXcd(internal::random<int>(1,EIGEN_TEST_MAX_SIZE/2), internal::random<int>(1,EIGEN_TEST_MAX_SIZE/2))) ); 379 CALL_SUBTEST_1( zero_sized_objects(MatrixXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) ); 380 } 381 CALL_SUBTEST_5( bug_127<0>() ); 382 CALL_SUBTEST_5( bug_817<0>() ); 383 CALL_SUBTEST_5( bug_1308<0>() ); 384 CALL_SUBTEST_6( unaligned_objects<0>() ); 385 CALL_SUBTEST_7( compute_block_size<float>() ); 386 CALL_SUBTEST_7( compute_block_size<double>() ); 387 CALL_SUBTEST_7( compute_block_size<std::complex<double> >() ); 388 CALL_SUBTEST_8( aliasing_with_resize<void>() ); 389 390 }