cart-elc

Source code for CART-ELC
git clone git://git.laack.co/cart-elc.git
Log | Files | Refs | README | LICENSE

svd_common.h (19317B)


      1 // This file is part of Eigen, a lightweight C++ template library
      2 // for linear algebra.
      3 //
      4 // Copyright (C) 2008-2014 Gael Guennebaud <gael.guennebaud@inria.fr>
      5 // Copyright (C) 2009 Benoit Jacob <jacob.benoit.1@gmail.com>
      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 #ifndef SVD_DEFAULT
     12 #error a macro SVD_DEFAULT(MatrixType) must be defined prior to including svd_common.h
     13 #endif
     14 
     15 #ifndef SVD_FOR_MIN_NORM
     16 #error a macro SVD_FOR_MIN_NORM(MatrixType) must be defined prior to including svd_common.h
     17 #endif
     18 
     19 #include "svd_fill.h"
     20 #include "solverbase.h"
     21 
     22 // Check that the matrix m is properly reconstructed and that the U and V factors are unitary
     23 // The SVD must have already been computed.
     24 template<typename SvdType, typename MatrixType>
     25 void svd_check_full(const MatrixType& m, const SvdType& svd)
     26 {
     27   Index rows = m.rows();
     28   Index cols = m.cols();
     29 
     30   enum {
     31     RowsAtCompileTime = MatrixType::RowsAtCompileTime,
     32     ColsAtCompileTime = MatrixType::ColsAtCompileTime
     33   };
     34 
     35   typedef typename MatrixType::Scalar Scalar;
     36   typedef typename MatrixType::RealScalar RealScalar;
     37   typedef Matrix<Scalar, RowsAtCompileTime, RowsAtCompileTime> MatrixUType;
     38   typedef Matrix<Scalar, ColsAtCompileTime, ColsAtCompileTime> MatrixVType;
     39 
     40   MatrixType sigma = MatrixType::Zero(rows,cols);
     41   sigma.diagonal() = svd.singularValues().template cast<Scalar>();
     42   MatrixUType u = svd.matrixU();
     43   MatrixVType v = svd.matrixV();
     44   RealScalar scaling = m.cwiseAbs().maxCoeff();
     45   if(scaling<(std::numeric_limits<RealScalar>::min)())
     46   {
     47     VERIFY(sigma.cwiseAbs().maxCoeff() <= (std::numeric_limits<RealScalar>::min)());
     48   }
     49   else
     50   {
     51     VERIFY_IS_APPROX(m/scaling, u * (sigma/scaling) * v.adjoint());
     52   }
     53   VERIFY_IS_UNITARY(u);
     54   VERIFY_IS_UNITARY(v);
     55 }
     56 
     57 // Compare partial SVD defined by computationOptions to a full SVD referenceSvd
     58 template<typename SvdType, typename MatrixType>
     59 void svd_compare_to_full(const MatrixType& m,
     60                          unsigned int computationOptions,
     61                          const SvdType& referenceSvd)
     62 {
     63   typedef typename MatrixType::RealScalar RealScalar;
     64   Index rows = m.rows();
     65   Index cols = m.cols();
     66   Index diagSize = (std::min)(rows, cols);
     67   RealScalar prec = test_precision<RealScalar>();
     68 
     69   SvdType svd(m, computationOptions);
     70 
     71   VERIFY_IS_APPROX(svd.singularValues(), referenceSvd.singularValues());
     72   
     73   if(computationOptions & (ComputeFullV|ComputeThinV))
     74   {
     75     VERIFY( (svd.matrixV().adjoint()*svd.matrixV()).isIdentity(prec) );
     76     VERIFY_IS_APPROX( svd.matrixV().leftCols(diagSize) * svd.singularValues().asDiagonal() * svd.matrixV().leftCols(diagSize).adjoint(),
     77                       referenceSvd.matrixV().leftCols(diagSize) * referenceSvd.singularValues().asDiagonal() * referenceSvd.matrixV().leftCols(diagSize).adjoint());
     78   }
     79   
     80   if(computationOptions & (ComputeFullU|ComputeThinU))
     81   {
     82     VERIFY( (svd.matrixU().adjoint()*svd.matrixU()).isIdentity(prec) );
     83     VERIFY_IS_APPROX( svd.matrixU().leftCols(diagSize) * svd.singularValues().cwiseAbs2().asDiagonal() * svd.matrixU().leftCols(diagSize).adjoint(),
     84                       referenceSvd.matrixU().leftCols(diagSize) * referenceSvd.singularValues().cwiseAbs2().asDiagonal() * referenceSvd.matrixU().leftCols(diagSize).adjoint());
     85   }
     86   
     87   // The following checks are not critical.
     88   // For instance, with Dived&Conquer SVD, if only the factor 'V' is computedt then different matrix-matrix product implementation will be used
     89   // and the resulting 'V' factor might be significantly different when the SVD decomposition is not unique, especially with single precision float.
     90   ++g_test_level;
     91   if(computationOptions & ComputeFullU)  VERIFY_IS_APPROX(svd.matrixU(), referenceSvd.matrixU());
     92   if(computationOptions & ComputeThinU)  VERIFY_IS_APPROX(svd.matrixU(), referenceSvd.matrixU().leftCols(diagSize));
     93   if(computationOptions & ComputeFullV)  VERIFY_IS_APPROX(svd.matrixV().cwiseAbs(), referenceSvd.matrixV().cwiseAbs());
     94   if(computationOptions & ComputeThinV)  VERIFY_IS_APPROX(svd.matrixV(), referenceSvd.matrixV().leftCols(diagSize));
     95   --g_test_level;
     96 }
     97 
     98 //
     99 template<typename SvdType, typename MatrixType>
    100 void svd_least_square(const MatrixType& m, unsigned int computationOptions)
    101 {
    102   typedef typename MatrixType::Scalar Scalar;
    103   typedef typename MatrixType::RealScalar RealScalar;
    104   Index rows = m.rows();
    105   Index cols = m.cols();
    106 
    107   enum {
    108     RowsAtCompileTime = MatrixType::RowsAtCompileTime,
    109     ColsAtCompileTime = MatrixType::ColsAtCompileTime
    110   };
    111 
    112   typedef Matrix<Scalar, RowsAtCompileTime, Dynamic> RhsType;
    113   typedef Matrix<Scalar, ColsAtCompileTime, Dynamic> SolutionType;
    114 
    115   RhsType rhs = RhsType::Random(rows, internal::random<Index>(1, cols));
    116   SvdType svd(m, computationOptions);
    117 
    118        if(internal::is_same<RealScalar,double>::value) svd.setThreshold(1e-8);
    119   else if(internal::is_same<RealScalar,float>::value)  svd.setThreshold(2e-4);
    120 
    121   SolutionType x = svd.solve(rhs);
    122    
    123   RealScalar residual = (m*x-rhs).norm();
    124   RealScalar rhs_norm = rhs.norm();
    125   if(!test_isMuchSmallerThan(residual,rhs.norm()))
    126   {
    127     // ^^^ If the residual is very small, then we have an exact solution, so we are already good.
    128     
    129     // evaluate normal equation which works also for least-squares solutions
    130     if(internal::is_same<RealScalar,double>::value || svd.rank()==m.diagonal().size())
    131     {
    132       using std::sqrt;
    133       // This test is not stable with single precision.
    134       // This is probably because squaring m signicantly affects the precision.      
    135       if(internal::is_same<RealScalar,float>::value) ++g_test_level;
    136       
    137       VERIFY_IS_APPROX(m.adjoint()*(m*x),m.adjoint()*rhs);
    138       
    139       if(internal::is_same<RealScalar,float>::value) --g_test_level;
    140     }
    141     
    142     // Check that there is no significantly better solution in the neighborhood of x
    143     for(Index k=0;k<x.rows();++k)
    144     {
    145       using std::abs;
    146       
    147       SolutionType y(x);
    148       y.row(k) = (RealScalar(1)+2*NumTraits<RealScalar>::epsilon())*x.row(k);
    149       RealScalar residual_y = (m*y-rhs).norm();
    150       VERIFY( test_isMuchSmallerThan(abs(residual_y-residual), rhs_norm) || residual < residual_y );
    151       if(internal::is_same<RealScalar,float>::value) ++g_test_level;
    152       VERIFY( test_isApprox(residual_y,residual) || residual < residual_y );
    153       if(internal::is_same<RealScalar,float>::value) --g_test_level;
    154       
    155       y.row(k) = (RealScalar(1)-2*NumTraits<RealScalar>::epsilon())*x.row(k);
    156       residual_y = (m*y-rhs).norm();
    157       VERIFY( test_isMuchSmallerThan(abs(residual_y-residual), rhs_norm) || residual < residual_y );
    158       if(internal::is_same<RealScalar,float>::value) ++g_test_level;
    159       VERIFY( test_isApprox(residual_y,residual) || residual < residual_y );
    160       if(internal::is_same<RealScalar,float>::value) --g_test_level;
    161     }
    162   }
    163 }
    164 
    165 // check minimal norm solutions, the inoput matrix m is only used to recover problem size
    166 template<typename MatrixType>
    167 void svd_min_norm(const MatrixType& m, unsigned int computationOptions)
    168 {
    169   typedef typename MatrixType::Scalar Scalar;
    170   Index cols = m.cols();
    171 
    172   enum {
    173     ColsAtCompileTime = MatrixType::ColsAtCompileTime
    174   };
    175 
    176   typedef Matrix<Scalar, ColsAtCompileTime, Dynamic> SolutionType;
    177 
    178   // generate a full-rank m x n problem with m<n
    179   enum {
    180     RankAtCompileTime2 = ColsAtCompileTime==Dynamic ? Dynamic : (ColsAtCompileTime)/2+1,
    181     RowsAtCompileTime3 = ColsAtCompileTime==Dynamic ? Dynamic : ColsAtCompileTime+1
    182   };
    183   typedef Matrix<Scalar, RankAtCompileTime2, ColsAtCompileTime> MatrixType2;
    184   typedef Matrix<Scalar, RankAtCompileTime2, 1> RhsType2;
    185   typedef Matrix<Scalar, ColsAtCompileTime, RankAtCompileTime2> MatrixType2T;
    186   Index rank = RankAtCompileTime2==Dynamic ? internal::random<Index>(1,cols) : Index(RankAtCompileTime2);
    187   MatrixType2 m2(rank,cols);
    188   int guard = 0;
    189   do {
    190     m2.setRandom();
    191   } while(SVD_FOR_MIN_NORM(MatrixType2)(m2).setThreshold(test_precision<Scalar>()).rank()!=rank && (++guard)<10);
    192   VERIFY(guard<10);
    193 
    194   RhsType2 rhs2 = RhsType2::Random(rank);
    195   // use QR to find a reference minimal norm solution
    196   HouseholderQR<MatrixType2T> qr(m2.adjoint());
    197   Matrix<Scalar,Dynamic,1> tmp = qr.matrixQR().topLeftCorner(rank,rank).template triangularView<Upper>().adjoint().solve(rhs2);
    198   tmp.conservativeResize(cols);
    199   tmp.tail(cols-rank).setZero();
    200   SolutionType x21 = qr.householderQ() * tmp;
    201   // now check with SVD
    202   SVD_FOR_MIN_NORM(MatrixType2) svd2(m2, computationOptions);
    203   SolutionType x22 = svd2.solve(rhs2);
    204   VERIFY_IS_APPROX(m2*x21, rhs2);
    205   VERIFY_IS_APPROX(m2*x22, rhs2);
    206   VERIFY_IS_APPROX(x21, x22);
    207 
    208   // Now check with a rank deficient matrix
    209   typedef Matrix<Scalar, RowsAtCompileTime3, ColsAtCompileTime> MatrixType3;
    210   typedef Matrix<Scalar, RowsAtCompileTime3, 1> RhsType3;
    211   Index rows3 = RowsAtCompileTime3==Dynamic ? internal::random<Index>(rank+1,2*cols) : Index(RowsAtCompileTime3);
    212   Matrix<Scalar,RowsAtCompileTime3,Dynamic> C = Matrix<Scalar,RowsAtCompileTime3,Dynamic>::Random(rows3,rank);
    213   MatrixType3 m3 = C * m2;
    214   RhsType3 rhs3 = C * rhs2;
    215   SVD_FOR_MIN_NORM(MatrixType3) svd3(m3, computationOptions);
    216   SolutionType x3 = svd3.solve(rhs3);
    217   VERIFY_IS_APPROX(m3*x3, rhs3);
    218   VERIFY_IS_APPROX(m3*x21, rhs3);
    219   VERIFY_IS_APPROX(m2*x3, rhs2);
    220   VERIFY_IS_APPROX(x21, x3);
    221 }
    222 
    223 template<typename MatrixType, typename SolverType>
    224 void svd_test_solvers(const MatrixType& m, const SolverType& solver) {
    225     Index rows, cols, cols2;
    226 
    227     rows = m.rows();
    228     cols = m.cols();
    229 
    230     if(MatrixType::ColsAtCompileTime==Dynamic)
    231     {
    232       cols2 = internal::random<int>(2,EIGEN_TEST_MAX_SIZE);
    233     }
    234     else
    235     {
    236       cols2 = cols;
    237     }
    238     typedef Matrix<typename MatrixType::Scalar, MatrixType::ColsAtCompileTime, MatrixType::ColsAtCompileTime> CMatrixType;
    239     check_solverbase<CMatrixType, MatrixType>(m, solver, rows, cols, cols2);
    240 }
    241 
    242 // Check full, compare_to_full, least_square, and min_norm for all possible compute-options
    243 template<typename SvdType, typename MatrixType>
    244 void svd_test_all_computation_options(const MatrixType& m, bool full_only)
    245 {
    246 //   if (QRPreconditioner == NoQRPreconditioner && m.rows() != m.cols())
    247 //     return;
    248   STATIC_CHECK(( internal::is_same<typename SvdType::StorageIndex,int>::value ));
    249 
    250   SvdType fullSvd(m, ComputeFullU|ComputeFullV);
    251   CALL_SUBTEST(( svd_check_full(m, fullSvd) ));
    252   CALL_SUBTEST(( svd_least_square<SvdType>(m, ComputeFullU | ComputeFullV) ));
    253   CALL_SUBTEST(( svd_min_norm(m, ComputeFullU | ComputeFullV) ));
    254   
    255   #if defined __INTEL_COMPILER
    256   // remark #111: statement is unreachable
    257   #pragma warning disable 111
    258   #endif
    259 
    260   svd_test_solvers(m, fullSvd);
    261 
    262   if(full_only)
    263     return;
    264 
    265   CALL_SUBTEST(( svd_compare_to_full(m, ComputeFullU, fullSvd) ));
    266   CALL_SUBTEST(( svd_compare_to_full(m, ComputeFullV, fullSvd) ));
    267   CALL_SUBTEST(( svd_compare_to_full(m, 0, fullSvd) ));
    268 
    269   if (MatrixType::ColsAtCompileTime == Dynamic) {
    270     // thin U/V are only available with dynamic number of columns
    271     CALL_SUBTEST(( svd_compare_to_full(m, ComputeFullU|ComputeThinV, fullSvd) ));
    272     CALL_SUBTEST(( svd_compare_to_full(m,              ComputeThinV, fullSvd) ));
    273     CALL_SUBTEST(( svd_compare_to_full(m, ComputeThinU|ComputeFullV, fullSvd) ));
    274     CALL_SUBTEST(( svd_compare_to_full(m, ComputeThinU             , fullSvd) ));
    275     CALL_SUBTEST(( svd_compare_to_full(m, ComputeThinU|ComputeThinV, fullSvd) ));
    276     
    277     CALL_SUBTEST(( svd_least_square<SvdType>(m, ComputeFullU | ComputeThinV) ));
    278     CALL_SUBTEST(( svd_least_square<SvdType>(m, ComputeThinU | ComputeFullV) ));
    279     CALL_SUBTEST(( svd_least_square<SvdType>(m, ComputeThinU | ComputeThinV) ));
    280 
    281     CALL_SUBTEST(( svd_min_norm(m, ComputeFullU | ComputeThinV) ));
    282     CALL_SUBTEST(( svd_min_norm(m, ComputeThinU | ComputeFullV) ));
    283     CALL_SUBTEST(( svd_min_norm(m, ComputeThinU | ComputeThinV) ));
    284 
    285     // test reconstruction
    286     Index diagSize = (std::min)(m.rows(), m.cols());
    287     SvdType svd(m, ComputeThinU | ComputeThinV);
    288     VERIFY_IS_APPROX(m, svd.matrixU().leftCols(diagSize) * svd.singularValues().asDiagonal() * svd.matrixV().leftCols(diagSize).adjoint());
    289   }
    290 }
    291 
    292 
    293 // work around stupid msvc error when constructing at compile time an expression that involves
    294 // a division by zero, even if the numeric type has floating point
    295 template<typename Scalar>
    296 EIGEN_DONT_INLINE Scalar zero() { return Scalar(0); }
    297 
    298 // workaround aggressive optimization in ICC
    299 template<typename T> EIGEN_DONT_INLINE  T sub(T a, T b) { return a - b; }
    300 
    301 // This function verifies we don't iterate infinitely on nan/inf values,
    302 // and that info() returns InvalidInput.
    303 template<typename SvdType, typename MatrixType>
    304 void svd_inf_nan()
    305 {
    306   SvdType svd;
    307   typedef typename MatrixType::Scalar Scalar;
    308   Scalar some_inf = Scalar(1) / zero<Scalar>();
    309   VERIFY(sub(some_inf, some_inf) != sub(some_inf, some_inf));
    310   svd.compute(MatrixType::Constant(10,10,some_inf), ComputeFullU | ComputeFullV);
    311   VERIFY(svd.info() == InvalidInput);
    312 
    313   Scalar nan = std::numeric_limits<Scalar>::quiet_NaN();
    314   VERIFY(nan != nan);
    315   svd.compute(MatrixType::Constant(10,10,nan), ComputeFullU | ComputeFullV);
    316   VERIFY(svd.info() == InvalidInput);  
    317 
    318   MatrixType m = MatrixType::Zero(10,10);
    319   m(internal::random<int>(0,9), internal::random<int>(0,9)) = some_inf;
    320   svd.compute(m, ComputeFullU | ComputeFullV);
    321   VERIFY(svd.info() == InvalidInput);
    322 
    323   m = MatrixType::Zero(10,10);
    324   m(internal::random<int>(0,9), internal::random<int>(0,9)) = nan;
    325   svd.compute(m, ComputeFullU | ComputeFullV);
    326   VERIFY(svd.info() == InvalidInput);
    327   
    328   // regression test for bug 791
    329   m.resize(3,3);
    330   m << 0,    2*NumTraits<Scalar>::epsilon(),  0.5,
    331        0,   -0.5,                             0,
    332        nan,  0,                               0;
    333   svd.compute(m, ComputeFullU | ComputeFullV);
    334   VERIFY(svd.info() == InvalidInput);
    335   
    336   m.resize(4,4);
    337   m <<  1, 0, 0, 0,
    338         0, 3, 1, 2e-308,
    339         1, 0, 1, nan,
    340         0, nan, nan, 0;
    341   svd.compute(m, ComputeFullU | ComputeFullV);
    342   VERIFY(svd.info() == InvalidInput);
    343 }
    344 
    345 // Regression test for bug 286: JacobiSVD loops indefinitely with some
    346 // matrices containing denormal numbers.
    347 template<typename>
    348 void svd_underoverflow()
    349 {
    350 #if defined __INTEL_COMPILER
    351 // shut up warning #239: floating point underflow
    352 #pragma warning push
    353 #pragma warning disable 239
    354 #endif
    355   Matrix2d M;
    356   M << -7.90884e-313, -4.94e-324,
    357                  0, 5.60844e-313;
    358   SVD_DEFAULT(Matrix2d) svd;
    359   svd.compute(M,ComputeFullU|ComputeFullV);
    360   CALL_SUBTEST( svd_check_full(M,svd) );
    361   
    362   // Check all 2x2 matrices made with the following coefficients:
    363   VectorXd value_set(9);
    364   value_set << 0, 1, -1, 5.60844e-313, -5.60844e-313, 4.94e-324, -4.94e-324, -4.94e-223, 4.94e-223;
    365   Array4i id(0,0,0,0);
    366   int k = 0;
    367   do
    368   {
    369     M << value_set(id(0)), value_set(id(1)), value_set(id(2)), value_set(id(3));
    370     svd.compute(M,ComputeFullU|ComputeFullV);
    371     CALL_SUBTEST( svd_check_full(M,svd) );
    372 
    373     id(k)++;
    374     if(id(k)>=value_set.size())
    375     {
    376       while(k<3 && id(k)>=value_set.size()) id(++k)++;
    377       id.head(k).setZero();
    378       k=0;
    379     }
    380 
    381   } while((id<int(value_set.size())).all());
    382   
    383 #if defined __INTEL_COMPILER
    384 #pragma warning pop
    385 #endif
    386   
    387   // Check for overflow:
    388   Matrix3d M3;
    389   M3 << 4.4331978442502944e+307, -5.8585363752028680e+307,  6.4527017443412964e+307,
    390         3.7841695601406358e+307,  2.4331702789740617e+306, -3.5235707140272905e+307,
    391        -8.7190887618028355e+307, -7.3453213709232193e+307, -2.4367363684472105e+307;
    392 
    393   SVD_DEFAULT(Matrix3d) svd3;
    394   svd3.compute(M3,ComputeFullU|ComputeFullV); // just check we don't loop indefinitely
    395   CALL_SUBTEST( svd_check_full(M3,svd3) );
    396 }
    397 
    398 // void jacobisvd(const MatrixType& a = MatrixType(), bool pickrandom = true)
    399 
    400 template<typename MatrixType>
    401 void svd_all_trivial_2x2( void (*cb)(const MatrixType&,bool) )
    402 {
    403   MatrixType M;
    404   VectorXd value_set(3);
    405   value_set << 0, 1, -1;
    406   Array4i id(0,0,0,0);
    407   int k = 0;
    408   do
    409   {
    410     M << value_set(id(0)), value_set(id(1)), value_set(id(2)), value_set(id(3));
    411     
    412     cb(M,false);
    413     
    414     id(k)++;
    415     if(id(k)>=value_set.size())
    416     {
    417       while(k<3 && id(k)>=value_set.size()) id(++k)++;
    418       id.head(k).setZero();
    419       k=0;
    420     }
    421     
    422   } while((id<int(value_set.size())).all());
    423 }
    424 
    425 template<typename>
    426 void svd_preallocate()
    427 {
    428   Vector3f v(3.f, 2.f, 1.f);
    429   MatrixXf m = v.asDiagonal();
    430 
    431   internal::set_is_malloc_allowed(false);
    432   VERIFY_RAISES_ASSERT(VectorXf tmp(10);)
    433   SVD_DEFAULT(MatrixXf) svd;
    434   internal::set_is_malloc_allowed(true);
    435   svd.compute(m);
    436   VERIFY_IS_APPROX(svd.singularValues(), v);
    437 
    438   SVD_DEFAULT(MatrixXf) svd2(3,3);
    439   internal::set_is_malloc_allowed(false);
    440   svd2.compute(m);
    441   internal::set_is_malloc_allowed(true);
    442   VERIFY_IS_APPROX(svd2.singularValues(), v);
    443   VERIFY_RAISES_ASSERT(svd2.matrixU());
    444   VERIFY_RAISES_ASSERT(svd2.matrixV());
    445   svd2.compute(m, ComputeFullU | ComputeFullV);
    446   VERIFY_IS_APPROX(svd2.matrixU(), Matrix3f::Identity());
    447   VERIFY_IS_APPROX(svd2.matrixV(), Matrix3f::Identity());
    448   internal::set_is_malloc_allowed(false);
    449   svd2.compute(m);
    450   internal::set_is_malloc_allowed(true);
    451 
    452   SVD_DEFAULT(MatrixXf) svd3(3,3,ComputeFullU|ComputeFullV);
    453   internal::set_is_malloc_allowed(false);
    454   svd2.compute(m);
    455   internal::set_is_malloc_allowed(true);
    456   VERIFY_IS_APPROX(svd2.singularValues(), v);
    457   VERIFY_IS_APPROX(svd2.matrixU(), Matrix3f::Identity());
    458   VERIFY_IS_APPROX(svd2.matrixV(), Matrix3f::Identity());
    459   internal::set_is_malloc_allowed(false);
    460   svd2.compute(m, ComputeFullU|ComputeFullV);
    461   internal::set_is_malloc_allowed(true);
    462 }
    463 
    464 template<typename SvdType,typename MatrixType> 
    465 void svd_verify_assert(const MatrixType& m, bool fullOnly = false)
    466 {
    467   typedef typename MatrixType::Scalar Scalar;
    468   Index rows = m.rows();
    469   Index cols = m.cols();
    470 
    471   enum {
    472     RowsAtCompileTime = MatrixType::RowsAtCompileTime,
    473     ColsAtCompileTime = MatrixType::ColsAtCompileTime
    474   };
    475 
    476   typedef Matrix<Scalar, RowsAtCompileTime, 1> RhsType;
    477   RhsType rhs(rows);
    478   SvdType svd;
    479   VERIFY_RAISES_ASSERT(svd.matrixU())
    480   VERIFY_RAISES_ASSERT(svd.singularValues())
    481   VERIFY_RAISES_ASSERT(svd.matrixV())
    482   VERIFY_RAISES_ASSERT(svd.solve(rhs))
    483   VERIFY_RAISES_ASSERT(svd.transpose().solve(rhs))
    484   VERIFY_RAISES_ASSERT(svd.adjoint().solve(rhs))
    485   MatrixType a = MatrixType::Zero(rows, cols);
    486   a.setZero();
    487   svd.compute(a, 0);
    488   VERIFY_RAISES_ASSERT(svd.matrixU())
    489   VERIFY_RAISES_ASSERT(svd.matrixV())
    490   svd.singularValues();
    491   VERIFY_RAISES_ASSERT(svd.solve(rhs))
    492 
    493   svd.compute(a, ComputeFullU);
    494   svd.matrixU();
    495   VERIFY_RAISES_ASSERT(svd.matrixV())
    496   VERIFY_RAISES_ASSERT(svd.solve(rhs))
    497   svd.compute(a, ComputeFullV);
    498   svd.matrixV();
    499   VERIFY_RAISES_ASSERT(svd.matrixU())
    500   VERIFY_RAISES_ASSERT(svd.solve(rhs))
    501 
    502   if (!fullOnly && ColsAtCompileTime == Dynamic)
    503   {
    504     svd.compute(a, ComputeThinU);
    505     svd.matrixU();
    506     VERIFY_RAISES_ASSERT(svd.matrixV())
    507     VERIFY_RAISES_ASSERT(svd.solve(rhs))
    508     svd.compute(a, ComputeThinV);
    509     svd.matrixV();
    510     VERIFY_RAISES_ASSERT(svd.matrixU())
    511     VERIFY_RAISES_ASSERT(svd.solve(rhs))
    512   }
    513   else
    514   {
    515     VERIFY_RAISES_ASSERT(svd.compute(a, ComputeThinU))
    516     VERIFY_RAISES_ASSERT(svd.compute(a, ComputeThinV))
    517   }
    518 }
    519 
    520 #undef SVD_DEFAULT
    521 #undef SVD_FOR_MIN_NORM