1// This file is part of Eigen, a lightweight C++ template library
2// for linear algebra.
3//
4// Copyright (C) 2008-2010 Gael Guennebaud <gael.guennebaud@inria.fr>
5// Copyright (C) 2010 Jitse Niesen <jitse@maths.leeds.ac.uk>
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 EIGEN_SELFADJOINTEIGENSOLVER_H
12#define EIGEN_SELFADJOINTEIGENSOLVER_H
13
14#include "./Tridiagonalization.h"
15
16namespace Eigen {
17
18template<typename _MatrixType>
19class GeneralizedSelfAdjointEigenSolver;
20
21namespace internal {
22template<typename SolverType,int Size,bool IsComplex> struct direct_selfadjoint_eigenvalues;
23}
24
25/** \eigenvalues_module \ingroup Eigenvalues_Module
26  *
27  *
28  * \class SelfAdjointEigenSolver
29  *
30  * \brief Computes eigenvalues and eigenvectors of selfadjoint matrices
31  *
32  * \tparam _MatrixType the type of the matrix of which we are computing the
33  * eigendecomposition; this is expected to be an instantiation of the Matrix
34  * class template.
35  *
36  * A matrix \f$ A \f$ is selfadjoint if it equals its adjoint. For real
37  * matrices, this means that the matrix is symmetric: it equals its
38  * transpose. This class computes the eigenvalues and eigenvectors of a
39  * selfadjoint matrix. These are the scalars \f$ \lambda \f$ and vectors
40  * \f$ v \f$ such that \f$ Av = \lambda v \f$.  The eigenvalues of a
41  * selfadjoint matrix are always real. If \f$ D \f$ is a diagonal matrix with
42  * the eigenvalues on the diagonal, and \f$ V \f$ is a matrix with the
43  * eigenvectors as its columns, then \f$ A = V D V^{-1} \f$ (for selfadjoint
44  * matrices, the matrix \f$ V \f$ is always invertible). This is called the
45  * eigendecomposition.
46  *
47  * The algorithm exploits the fact that the matrix is selfadjoint, making it
48  * faster and more accurate than the general purpose eigenvalue algorithms
49  * implemented in EigenSolver and ComplexEigenSolver.
50  *
51  * Only the \b lower \b triangular \b part of the input matrix is referenced.
52  *
53  * Call the function compute() to compute the eigenvalues and eigenvectors of
54  * a given matrix. Alternatively, you can use the
55  * SelfAdjointEigenSolver(const MatrixType&, int) constructor which computes
56  * the eigenvalues and eigenvectors at construction time. Once the eigenvalue
57  * and eigenvectors are computed, they can be retrieved with the eigenvalues()
58  * and eigenvectors() functions.
59  *
60  * The documentation for SelfAdjointEigenSolver(const MatrixType&, int)
61  * contains an example of the typical use of this class.
62  *
63  * To solve the \em generalized eigenvalue problem \f$ Av = \lambda Bv \f$ and
64  * the likes, see the class GeneralizedSelfAdjointEigenSolver.
65  *
66  * \sa MatrixBase::eigenvalues(), class EigenSolver, class ComplexEigenSolver
67  */
68template<typename _MatrixType> class SelfAdjointEigenSolver
69{
70  public:
71
72    typedef _MatrixType MatrixType;
73    enum {
74      Size = MatrixType::RowsAtCompileTime,
75      ColsAtCompileTime = MatrixType::ColsAtCompileTime,
76      Options = MatrixType::Options,
77      MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime
78    };
79
80    /** \brief Scalar type for matrices of type \p _MatrixType. */
81    typedef typename MatrixType::Scalar Scalar;
82    typedef typename MatrixType::Index Index;
83
84    /** \brief Real scalar type for \p _MatrixType.
85      *
86      * This is just \c Scalar if #Scalar is real (e.g., \c float or
87      * \c double), and the type of the real part of \c Scalar if #Scalar is
88      * complex.
89      */
90    typedef typename NumTraits<Scalar>::Real RealScalar;
91
92    friend struct internal::direct_selfadjoint_eigenvalues<SelfAdjointEigenSolver,Size,NumTraits<Scalar>::IsComplex>;
93
94    /** \brief Type for vector of eigenvalues as returned by eigenvalues().
95      *
96      * This is a column vector with entries of type #RealScalar.
97      * The length of the vector is the size of \p _MatrixType.
98      */
99    typedef typename internal::plain_col_type<MatrixType, RealScalar>::type RealVectorType;
100    typedef Tridiagonalization<MatrixType> TridiagonalizationType;
101
102    /** \brief Default constructor for fixed-size matrices.
103      *
104      * The default constructor is useful in cases in which the user intends to
105      * perform decompositions via compute(). This constructor
106      * can only be used if \p _MatrixType is a fixed-size matrix; use
107      * SelfAdjointEigenSolver(Index) for dynamic-size matrices.
108      *
109      * Example: \include SelfAdjointEigenSolver_SelfAdjointEigenSolver.cpp
110      * Output: \verbinclude SelfAdjointEigenSolver_SelfAdjointEigenSolver.out
111      */
112    SelfAdjointEigenSolver()
113        : m_eivec(),
114          m_eivalues(),
115          m_subdiag(),
116          m_isInitialized(false)
117    { }
118
119    /** \brief Constructor, pre-allocates memory for dynamic-size matrices.
120      *
121      * \param [in]  size  Positive integer, size of the matrix whose
122      * eigenvalues and eigenvectors will be computed.
123      *
124      * This constructor is useful for dynamic-size matrices, when the user
125      * intends to perform decompositions via compute(). The \p size
126      * parameter is only used as a hint. It is not an error to give a wrong
127      * \p size, but it may impair performance.
128      *
129      * \sa compute() for an example
130      */
131    SelfAdjointEigenSolver(Index size)
132        : m_eivec(size, size),
133          m_eivalues(size),
134          m_subdiag(size > 1 ? size - 1 : 1),
135          m_isInitialized(false)
136    {}
137
138    /** \brief Constructor; computes eigendecomposition of given matrix.
139      *
140      * \param[in]  matrix  Selfadjoint matrix whose eigendecomposition is to
141      *    be computed. Only the lower triangular part of the matrix is referenced.
142      * \param[in]  options Can be #ComputeEigenvectors (default) or #EigenvaluesOnly.
143      *
144      * This constructor calls compute(const MatrixType&, int) to compute the
145      * eigenvalues of the matrix \p matrix. The eigenvectors are computed if
146      * \p options equals #ComputeEigenvectors.
147      *
148      * Example: \include SelfAdjointEigenSolver_SelfAdjointEigenSolver_MatrixType.cpp
149      * Output: \verbinclude SelfAdjointEigenSolver_SelfAdjointEigenSolver_MatrixType.out
150      *
151      * \sa compute(const MatrixType&, int)
152      */
153    SelfAdjointEigenSolver(const MatrixType& matrix, int options = ComputeEigenvectors)
154      : m_eivec(matrix.rows(), matrix.cols()),
155        m_eivalues(matrix.cols()),
156        m_subdiag(matrix.rows() > 1 ? matrix.rows() - 1 : 1),
157        m_isInitialized(false)
158    {
159      compute(matrix, options);
160    }
161
162    /** \brief Computes eigendecomposition of given matrix.
163      *
164      * \param[in]  matrix  Selfadjoint matrix whose eigendecomposition is to
165      *    be computed. Only the lower triangular part of the matrix is referenced.
166      * \param[in]  options Can be #ComputeEigenvectors (default) or #EigenvaluesOnly.
167      * \returns    Reference to \c *this
168      *
169      * This function computes the eigenvalues of \p matrix.  The eigenvalues()
170      * function can be used to retrieve them.  If \p options equals #ComputeEigenvectors,
171      * then the eigenvectors are also computed and can be retrieved by
172      * calling eigenvectors().
173      *
174      * This implementation uses a symmetric QR algorithm. The matrix is first
175      * reduced to tridiagonal form using the Tridiagonalization class. The
176      * tridiagonal matrix is then brought to diagonal form with implicit
177      * symmetric QR steps with Wilkinson shift. Details can be found in
178      * Section 8.3 of Golub \& Van Loan, <i>%Matrix Computations</i>.
179      *
180      * The cost of the computation is about \f$ 9n^3 \f$ if the eigenvectors
181      * are required and \f$ 4n^3/3 \f$ if they are not required.
182      *
183      * This method reuses the memory in the SelfAdjointEigenSolver object that
184      * was allocated when the object was constructed, if the size of the
185      * matrix does not change.
186      *
187      * Example: \include SelfAdjointEigenSolver_compute_MatrixType.cpp
188      * Output: \verbinclude SelfAdjointEigenSolver_compute_MatrixType.out
189      *
190      * \sa SelfAdjointEigenSolver(const MatrixType&, int)
191      */
192    SelfAdjointEigenSolver& compute(const MatrixType& matrix, int options = ComputeEigenvectors);
193
194    /** \brief Computes eigendecomposition of given matrix using a direct algorithm
195      *
196      * This is a variant of compute(const MatrixType&, int options) which
197      * directly solves the underlying polynomial equation.
198      *
199      * Currently only 3x3 matrices for which the sizes are known at compile time are supported (e.g., Matrix3d).
200      *
201      * This method is usually significantly faster than the QR algorithm
202      * but it might also be less accurate. It is also worth noting that
203      * for 3x3 matrices it involves trigonometric operations which are
204      * not necessarily available for all scalar types.
205      *
206      * \sa compute(const MatrixType&, int options)
207      */
208    SelfAdjointEigenSolver& computeDirect(const MatrixType& matrix, int options = ComputeEigenvectors);
209
210    /** \brief Returns the eigenvectors of given matrix.
211      *
212      * \returns  A const reference to the matrix whose columns are the eigenvectors.
213      *
214      * \pre The eigenvectors have been computed before.
215      *
216      * Column \f$ k \f$ of the returned matrix is an eigenvector corresponding
217      * to eigenvalue number \f$ k \f$ as returned by eigenvalues().  The
218      * eigenvectors are normalized to have (Euclidean) norm equal to one. If
219      * this object was used to solve the eigenproblem for the selfadjoint
220      * matrix \f$ A \f$, then the matrix returned by this function is the
221      * matrix \f$ V \f$ in the eigendecomposition \f$ A = V D V^{-1} \f$.
222      *
223      * Example: \include SelfAdjointEigenSolver_eigenvectors.cpp
224      * Output: \verbinclude SelfAdjointEigenSolver_eigenvectors.out
225      *
226      * \sa eigenvalues()
227      */
228    const MatrixType& eigenvectors() const
229    {
230      eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized.");
231      eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues.");
232      return m_eivec;
233    }
234
235    /** \brief Returns the eigenvalues of given matrix.
236      *
237      * \returns A const reference to the column vector containing the eigenvalues.
238      *
239      * \pre The eigenvalues have been computed before.
240      *
241      * The eigenvalues are repeated according to their algebraic multiplicity,
242      * so there are as many eigenvalues as rows in the matrix. The eigenvalues
243      * are sorted in increasing order.
244      *
245      * Example: \include SelfAdjointEigenSolver_eigenvalues.cpp
246      * Output: \verbinclude SelfAdjointEigenSolver_eigenvalues.out
247      *
248      * \sa eigenvectors(), MatrixBase::eigenvalues()
249      */
250    const RealVectorType& eigenvalues() const
251    {
252      eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized.");
253      return m_eivalues;
254    }
255
256    /** \brief Computes the positive-definite square root of the matrix.
257      *
258      * \returns the positive-definite square root of the matrix
259      *
260      * \pre The eigenvalues and eigenvectors of a positive-definite matrix
261      * have been computed before.
262      *
263      * The square root of a positive-definite matrix \f$ A \f$ is the
264      * positive-definite matrix whose square equals \f$ A \f$. This function
265      * uses the eigendecomposition \f$ A = V D V^{-1} \f$ to compute the
266      * square root as \f$ A^{1/2} = V D^{1/2} V^{-1} \f$.
267      *
268      * Example: \include SelfAdjointEigenSolver_operatorSqrt.cpp
269      * Output: \verbinclude SelfAdjointEigenSolver_operatorSqrt.out
270      *
271      * \sa operatorInverseSqrt(),
272      *     \ref MatrixFunctions_Module "MatrixFunctions Module"
273      */
274    MatrixType operatorSqrt() const
275    {
276      eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized.");
277      eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues.");
278      return m_eivec * m_eivalues.cwiseSqrt().asDiagonal() * m_eivec.adjoint();
279    }
280
281    /** \brief Computes the inverse square root of the matrix.
282      *
283      * \returns the inverse positive-definite square root of the matrix
284      *
285      * \pre The eigenvalues and eigenvectors of a positive-definite matrix
286      * have been computed before.
287      *
288      * This function uses the eigendecomposition \f$ A = V D V^{-1} \f$ to
289      * compute the inverse square root as \f$ V D^{-1/2} V^{-1} \f$. This is
290      * cheaper than first computing the square root with operatorSqrt() and
291      * then its inverse with MatrixBase::inverse().
292      *
293      * Example: \include SelfAdjointEigenSolver_operatorInverseSqrt.cpp
294      * Output: \verbinclude SelfAdjointEigenSolver_operatorInverseSqrt.out
295      *
296      * \sa operatorSqrt(), MatrixBase::inverse(),
297      *     \ref MatrixFunctions_Module "MatrixFunctions Module"
298      */
299    MatrixType operatorInverseSqrt() const
300    {
301      eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized.");
302      eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues.");
303      return m_eivec * m_eivalues.cwiseInverse().cwiseSqrt().asDiagonal() * m_eivec.adjoint();
304    }
305
306    /** \brief Reports whether previous computation was successful.
307      *
308      * \returns \c Success if computation was succesful, \c NoConvergence otherwise.
309      */
310    ComputationInfo info() const
311    {
312      eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized.");
313      return m_info;
314    }
315
316    /** \brief Maximum number of iterations.
317      *
318      * The algorithm terminates if it does not converge within m_maxIterations * n iterations, where n
319      * denotes the size of the matrix. This value is currently set to 30 (copied from LAPACK).
320      */
321    static const int m_maxIterations = 30;
322
323    #ifdef EIGEN2_SUPPORT
324    SelfAdjointEigenSolver(const MatrixType& matrix, bool computeEigenvectors)
325      : m_eivec(matrix.rows(), matrix.cols()),
326        m_eivalues(matrix.cols()),
327        m_subdiag(matrix.rows() > 1 ? matrix.rows() - 1 : 1),
328        m_isInitialized(false)
329    {
330      compute(matrix, computeEigenvectors);
331    }
332
333    SelfAdjointEigenSolver(const MatrixType& matA, const MatrixType& matB, bool computeEigenvectors = true)
334        : m_eivec(matA.cols(), matA.cols()),
335          m_eivalues(matA.cols()),
336          m_subdiag(matA.cols() > 1 ? matA.cols() - 1 : 1),
337          m_isInitialized(false)
338    {
339      static_cast<GeneralizedSelfAdjointEigenSolver<MatrixType>*>(this)->compute(matA, matB, computeEigenvectors ? ComputeEigenvectors : EigenvaluesOnly);
340    }
341
342    void compute(const MatrixType& matrix, bool computeEigenvectors)
343    {
344      compute(matrix, computeEigenvectors ? ComputeEigenvectors : EigenvaluesOnly);
345    }
346
347    void compute(const MatrixType& matA, const MatrixType& matB, bool computeEigenvectors = true)
348    {
349      compute(matA, matB, computeEigenvectors ? ComputeEigenvectors : EigenvaluesOnly);
350    }
351    #endif // EIGEN2_SUPPORT
352
353  protected:
354    MatrixType m_eivec;
355    RealVectorType m_eivalues;
356    typename TridiagonalizationType::SubDiagonalType m_subdiag;
357    ComputationInfo m_info;
358    bool m_isInitialized;
359    bool m_eigenvectorsOk;
360};
361
362/** \internal
363  *
364  * \eigenvalues_module \ingroup Eigenvalues_Module
365  *
366  * Performs a QR step on a tridiagonal symmetric matrix represented as a
367  * pair of two vectors \a diag and \a subdiag.
368  *
369  * \param matA the input selfadjoint matrix
370  * \param hCoeffs returned Householder coefficients
371  *
372  * For compilation efficiency reasons, this procedure does not use eigen expression
373  * for its arguments.
374  *
375  * Implemented from Golub's "Matrix Computations", algorithm 8.3.2:
376  * "implicit symmetric QR step with Wilkinson shift"
377  */
378namespace internal {
379template<int StorageOrder,typename RealScalar, typename Scalar, typename Index>
380static void tridiagonal_qr_step(RealScalar* diag, RealScalar* subdiag, Index start, Index end, Scalar* matrixQ, Index n);
381}
382
383template<typename MatrixType>
384SelfAdjointEigenSolver<MatrixType>& SelfAdjointEigenSolver<MatrixType>
385::compute(const MatrixType& matrix, int options)
386{
387  using std::abs;
388  eigen_assert(matrix.cols() == matrix.rows());
389  eigen_assert((options&~(EigVecMask|GenEigMask))==0
390          && (options&EigVecMask)!=EigVecMask
391          && "invalid option parameter");
392  bool computeEigenvectors = (options&ComputeEigenvectors)==ComputeEigenvectors;
393  Index n = matrix.cols();
394  m_eivalues.resize(n,1);
395
396  if(n==1)
397  {
398    m_eivalues.coeffRef(0,0) = numext::real(matrix.coeff(0,0));
399    if(computeEigenvectors)
400      m_eivec.setOnes(n,n);
401    m_info = Success;
402    m_isInitialized = true;
403    m_eigenvectorsOk = computeEigenvectors;
404    return *this;
405  }
406
407  // declare some aliases
408  RealVectorType& diag = m_eivalues;
409  MatrixType& mat = m_eivec;
410
411  // map the matrix coefficients to [-1:1] to avoid over- and underflow.
412  mat = matrix.template triangularView<Lower>();
413  RealScalar scale = mat.cwiseAbs().maxCoeff();
414  if(scale==RealScalar(0)) scale = RealScalar(1);
415  mat.template triangularView<Lower>() /= scale;
416  m_subdiag.resize(n-1);
417  internal::tridiagonalization_inplace(mat, diag, m_subdiag, computeEigenvectors);
418
419  Index end = n-1;
420  Index start = 0;
421  Index iter = 0; // total number of iterations
422
423  while (end>0)
424  {
425    for (Index i = start; i<end; ++i)
426      if (internal::isMuchSmallerThan(abs(m_subdiag[i]),(abs(diag[i])+abs(diag[i+1]))))
427        m_subdiag[i] = 0;
428
429    // find the largest unreduced block
430    while (end>0 && m_subdiag[end-1]==0)
431    {
432      end--;
433    }
434    if (end<=0)
435      break;
436
437    // if we spent too many iterations, we give up
438    iter++;
439    if(iter > m_maxIterations * n) break;
440
441    start = end - 1;
442    while (start>0 && m_subdiag[start-1]!=0)
443      start--;
444
445    internal::tridiagonal_qr_step<MatrixType::Flags&RowMajorBit ? RowMajor : ColMajor>(diag.data(), m_subdiag.data(), start, end, computeEigenvectors ? m_eivec.data() : (Scalar*)0, n);
446  }
447
448  if (iter <= m_maxIterations * n)
449    m_info = Success;
450  else
451    m_info = NoConvergence;
452
453  // Sort eigenvalues and corresponding vectors.
454  // TODO make the sort optional ?
455  // TODO use a better sort algorithm !!
456  if (m_info == Success)
457  {
458    for (Index i = 0; i < n-1; ++i)
459    {
460      Index k;
461      m_eivalues.segment(i,n-i).minCoeff(&k);
462      if (k > 0)
463      {
464        std::swap(m_eivalues[i], m_eivalues[k+i]);
465        if(computeEigenvectors)
466          m_eivec.col(i).swap(m_eivec.col(k+i));
467      }
468    }
469  }
470
471  // scale back the eigen values
472  m_eivalues *= scale;
473
474  m_isInitialized = true;
475  m_eigenvectorsOk = computeEigenvectors;
476  return *this;
477}
478
479
480namespace internal {
481
482template<typename SolverType,int Size,bool IsComplex> struct direct_selfadjoint_eigenvalues
483{
484  static inline void run(SolverType& eig, const typename SolverType::MatrixType& A, int options)
485  { eig.compute(A,options); }
486};
487
488template<typename SolverType> struct direct_selfadjoint_eigenvalues<SolverType,3,false>
489{
490  typedef typename SolverType::MatrixType MatrixType;
491  typedef typename SolverType::RealVectorType VectorType;
492  typedef typename SolverType::Scalar Scalar;
493
494  static inline void computeRoots(const MatrixType& m, VectorType& roots)
495  {
496    using std::sqrt;
497    using std::atan2;
498    using std::cos;
499    using std::sin;
500    const Scalar s_inv3 = Scalar(1.0)/Scalar(3.0);
501    const Scalar s_sqrt3 = sqrt(Scalar(3.0));
502
503    // The characteristic equation is x^3 - c2*x^2 + c1*x - c0 = 0.  The
504    // eigenvalues are the roots to this equation, all guaranteed to be
505    // real-valued, because the matrix is symmetric.
506    Scalar c0 = m(0,0)*m(1,1)*m(2,2) + Scalar(2)*m(1,0)*m(2,0)*m(2,1) - m(0,0)*m(2,1)*m(2,1) - m(1,1)*m(2,0)*m(2,0) - m(2,2)*m(1,0)*m(1,0);
507    Scalar c1 = m(0,0)*m(1,1) - m(1,0)*m(1,0) + m(0,0)*m(2,2) - m(2,0)*m(2,0) + m(1,1)*m(2,2) - m(2,1)*m(2,1);
508    Scalar c2 = m(0,0) + m(1,1) + m(2,2);
509
510    // Construct the parameters used in classifying the roots of the equation
511    // and in solving the equation for the roots in closed form.
512    Scalar c2_over_3 = c2*s_inv3;
513    Scalar a_over_3 = (c1 - c2*c2_over_3)*s_inv3;
514    if (a_over_3 > Scalar(0))
515      a_over_3 = Scalar(0);
516
517    Scalar half_b = Scalar(0.5)*(c0 + c2_over_3*(Scalar(2)*c2_over_3*c2_over_3 - c1));
518
519    Scalar q = half_b*half_b + a_over_3*a_over_3*a_over_3;
520    if (q > Scalar(0))
521      q = Scalar(0);
522
523    // Compute the eigenvalues by solving for the roots of the polynomial.
524    Scalar rho = sqrt(-a_over_3);
525    Scalar theta = atan2(sqrt(-q),half_b)*s_inv3;
526    Scalar cos_theta = cos(theta);
527    Scalar sin_theta = sin(theta);
528    roots(0) = c2_over_3 + Scalar(2)*rho*cos_theta;
529    roots(1) = c2_over_3 - rho*(cos_theta + s_sqrt3*sin_theta);
530    roots(2) = c2_over_3 - rho*(cos_theta - s_sqrt3*sin_theta);
531
532    // Sort in increasing order.
533    if (roots(0) >= roots(1))
534      std::swap(roots(0),roots(1));
535    if (roots(1) >= roots(2))
536    {
537      std::swap(roots(1),roots(2));
538      if (roots(0) >= roots(1))
539        std::swap(roots(0),roots(1));
540    }
541  }
542
543  static inline void run(SolverType& solver, const MatrixType& mat, int options)
544  {
545    using std::sqrt;
546    eigen_assert(mat.cols() == 3 && mat.cols() == mat.rows());
547    eigen_assert((options&~(EigVecMask|GenEigMask))==0
548            && (options&EigVecMask)!=EigVecMask
549            && "invalid option parameter");
550    bool computeEigenvectors = (options&ComputeEigenvectors)==ComputeEigenvectors;
551
552    MatrixType& eivecs = solver.m_eivec;
553    VectorType& eivals = solver.m_eivalues;
554
555    // map the matrix coefficients to [-1:1] to avoid over- and underflow.
556    Scalar scale = mat.cwiseAbs().maxCoeff();
557    MatrixType scaledMat = mat / scale;
558
559    // compute the eigenvalues
560    computeRoots(scaledMat,eivals);
561
562    // compute the eigen vectors
563    if(computeEigenvectors)
564    {
565      Scalar safeNorm2 = Eigen::NumTraits<Scalar>::epsilon();
566      safeNorm2 *= safeNorm2;
567      if((eivals(2)-eivals(0))<=Eigen::NumTraits<Scalar>::epsilon())
568      {
569        eivecs.setIdentity();
570      }
571      else
572      {
573        scaledMat = scaledMat.template selfadjointView<Lower>();
574        MatrixType tmp;
575        tmp = scaledMat;
576
577        Scalar d0 = eivals(2) - eivals(1);
578        Scalar d1 = eivals(1) - eivals(0);
579        int k =  d0 > d1 ? 2 : 0;
580        d0 = d0 > d1 ? d1 : d0;
581
582        tmp.diagonal().array () -= eivals(k);
583        VectorType cross;
584        Scalar n;
585        n = (cross = tmp.row(0).cross(tmp.row(1))).squaredNorm();
586
587        if(n>safeNorm2)
588          eivecs.col(k) = cross / sqrt(n);
589        else
590        {
591          n = (cross = tmp.row(0).cross(tmp.row(2))).squaredNorm();
592
593          if(n>safeNorm2)
594            eivecs.col(k) = cross / sqrt(n);
595          else
596          {
597            n = (cross = tmp.row(1).cross(tmp.row(2))).squaredNorm();
598
599            if(n>safeNorm2)
600              eivecs.col(k) = cross / sqrt(n);
601            else
602            {
603              // the input matrix and/or the eigenvaues probably contains some inf/NaN,
604              // => exit
605              // scale back to the original size.
606              eivals *= scale;
607
608              solver.m_info = NumericalIssue;
609              solver.m_isInitialized = true;
610              solver.m_eigenvectorsOk = computeEigenvectors;
611              return;
612            }
613          }
614        }
615
616        tmp = scaledMat;
617        tmp.diagonal().array() -= eivals(1);
618
619        if(d0<=Eigen::NumTraits<Scalar>::epsilon())
620          eivecs.col(1) = eivecs.col(k).unitOrthogonal();
621        else
622        {
623          n = (cross = eivecs.col(k).cross(tmp.row(0).normalized())).squaredNorm();
624          if(n>safeNorm2)
625            eivecs.col(1) = cross / sqrt(n);
626          else
627          {
628            n = (cross = eivecs.col(k).cross(tmp.row(1))).squaredNorm();
629            if(n>safeNorm2)
630              eivecs.col(1) = cross / sqrt(n);
631            else
632            {
633              n = (cross = eivecs.col(k).cross(tmp.row(2))).squaredNorm();
634              if(n>safeNorm2)
635                eivecs.col(1) = cross / sqrt(n);
636              else
637              {
638                // we should never reach this point,
639                // if so the last two eigenvalues are likely to ve very closed to each other
640                eivecs.col(1) = eivecs.col(k).unitOrthogonal();
641              }
642            }
643          }
644
645          // make sure that eivecs[1] is orthogonal to eivecs[2]
646          Scalar d = eivecs.col(1).dot(eivecs.col(k));
647          eivecs.col(1) = (eivecs.col(1) - d * eivecs.col(k)).normalized();
648        }
649
650        eivecs.col(k==2 ? 0 : 2) = eivecs.col(k).cross(eivecs.col(1)).normalized();
651      }
652    }
653    // Rescale back to the original size.
654    eivals *= scale;
655
656    solver.m_info = Success;
657    solver.m_isInitialized = true;
658    solver.m_eigenvectorsOk = computeEigenvectors;
659  }
660};
661
662// 2x2 direct eigenvalues decomposition, code from Hauke Heibel
663template<typename SolverType> struct direct_selfadjoint_eigenvalues<SolverType,2,false>
664{
665  typedef typename SolverType::MatrixType MatrixType;
666  typedef typename SolverType::RealVectorType VectorType;
667  typedef typename SolverType::Scalar Scalar;
668
669  static inline void computeRoots(const MatrixType& m, VectorType& roots)
670  {
671    using std::sqrt;
672    const Scalar t0 = Scalar(0.5) * sqrt( numext::abs2(m(0,0)-m(1,1)) + Scalar(4)*m(1,0)*m(1,0));
673    const Scalar t1 = Scalar(0.5) * (m(0,0) + m(1,1));
674    roots(0) = t1 - t0;
675    roots(1) = t1 + t0;
676  }
677
678  static inline void run(SolverType& solver, const MatrixType& mat, int options)
679  {
680    using std::sqrt;
681    eigen_assert(mat.cols() == 2 && mat.cols() == mat.rows());
682    eigen_assert((options&~(EigVecMask|GenEigMask))==0
683            && (options&EigVecMask)!=EigVecMask
684            && "invalid option parameter");
685    bool computeEigenvectors = (options&ComputeEigenvectors)==ComputeEigenvectors;
686
687    MatrixType& eivecs = solver.m_eivec;
688    VectorType& eivals = solver.m_eivalues;
689
690    // map the matrix coefficients to [-1:1] to avoid over- and underflow.
691    Scalar scale = mat.cwiseAbs().maxCoeff();
692    scale = (std::max)(scale,Scalar(1));
693    MatrixType scaledMat = mat / scale;
694
695    // Compute the eigenvalues
696    computeRoots(scaledMat,eivals);
697
698    // compute the eigen vectors
699    if(computeEigenvectors)
700    {
701      scaledMat.diagonal().array () -= eivals(1);
702      Scalar a2 = numext::abs2(scaledMat(0,0));
703      Scalar c2 = numext::abs2(scaledMat(1,1));
704      Scalar b2 = numext::abs2(scaledMat(1,0));
705      if(a2>c2)
706      {
707        eivecs.col(1) << -scaledMat(1,0), scaledMat(0,0);
708        eivecs.col(1) /= sqrt(a2+b2);
709      }
710      else
711      {
712        eivecs.col(1) << -scaledMat(1,1), scaledMat(1,0);
713        eivecs.col(1) /= sqrt(c2+b2);
714      }
715
716      eivecs.col(0) << eivecs.col(1).unitOrthogonal();
717    }
718
719    // Rescale back to the original size.
720    eivals *= scale;
721
722    solver.m_info = Success;
723    solver.m_isInitialized = true;
724    solver.m_eigenvectorsOk = computeEigenvectors;
725  }
726};
727
728}
729
730template<typename MatrixType>
731SelfAdjointEigenSolver<MatrixType>& SelfAdjointEigenSolver<MatrixType>
732::computeDirect(const MatrixType& matrix, int options)
733{
734  internal::direct_selfadjoint_eigenvalues<SelfAdjointEigenSolver,Size,NumTraits<Scalar>::IsComplex>::run(*this,matrix,options);
735  return *this;
736}
737
738namespace internal {
739template<int StorageOrder,typename RealScalar, typename Scalar, typename Index>
740static void tridiagonal_qr_step(RealScalar* diag, RealScalar* subdiag, Index start, Index end, Scalar* matrixQ, Index n)
741{
742  using std::abs;
743  RealScalar td = (diag[end-1] - diag[end])*RealScalar(0.5);
744  RealScalar e = subdiag[end-1];
745  // Note that thanks to scaling, e^2 or td^2 cannot overflow, however they can still
746  // underflow thus leading to inf/NaN values when using the following commented code:
747//   RealScalar e2 = numext::abs2(subdiag[end-1]);
748//   RealScalar mu = diag[end] - e2 / (td + (td>0 ? 1 : -1) * sqrt(td*td + e2));
749  // This explain the following, somewhat more complicated, version:
750  RealScalar mu = diag[end];
751  if(td==0)
752    mu -= abs(e);
753  else
754  {
755    RealScalar e2 = numext::abs2(subdiag[end-1]);
756    RealScalar h = numext::hypot(td,e);
757    if(e2==0)  mu -= (e / (td + (td>0 ? 1 : -1))) * (e / h);
758    else       mu -= e2 / (td + (td>0 ? h : -h));
759  }
760
761  RealScalar x = diag[start] - mu;
762  RealScalar z = subdiag[start];
763  for (Index k = start; k < end; ++k)
764  {
765    JacobiRotation<RealScalar> rot;
766    rot.makeGivens(x, z);
767
768    // do T = G' T G
769    RealScalar sdk = rot.s() * diag[k] + rot.c() * subdiag[k];
770    RealScalar dkp1 = rot.s() * subdiag[k] + rot.c() * diag[k+1];
771
772    diag[k] = rot.c() * (rot.c() * diag[k] - rot.s() * subdiag[k]) - rot.s() * (rot.c() * subdiag[k] - rot.s() * diag[k+1]);
773    diag[k+1] = rot.s() * sdk + rot.c() * dkp1;
774    subdiag[k] = rot.c() * sdk - rot.s() * dkp1;
775
776
777    if (k > start)
778      subdiag[k - 1] = rot.c() * subdiag[k-1] - rot.s() * z;
779
780    x = subdiag[k];
781
782    if (k < end - 1)
783    {
784      z = -rot.s() * subdiag[k+1];
785      subdiag[k + 1] = rot.c() * subdiag[k+1];
786    }
787
788    // apply the givens rotation to the unit matrix Q = Q * G
789    if (matrixQ)
790    {
791      // FIXME if StorageOrder == RowMajor this operation is not very efficient
792      Map<Matrix<Scalar,Dynamic,Dynamic,StorageOrder> > q(matrixQ,n,n);
793      q.applyOnTheRight(k,k+1,rot);
794    }
795  }
796}
797
798} // end namespace internal
799
800} // end namespace Eigen
801
802#endif // EIGEN_SELFADJOINTEIGENSOLVER_H
803