$darkmode
Eigen  5.0.1-dev
BDCSVD_LAPACKE.h
1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2022 Melven Roehrig-Zoellner <Melven.Roehrig-Zoellner@DLR.de>
5 // Copyright (c) 2011, Intel Corporation. All rights reserved.
6 //
7 // This file is based on the JacobiSVD_LAPACKE.h originally from Intel -
8 // see license notice below:
9 /*
10  Redistribution and use in source and binary forms, with or without modification,
11  are permitted provided that the following conditions are met:
12 
13  * Redistributions of source code must retain the above copyright notice, this
14  list of conditions and the following disclaimer.
15  * Redistributions in binary form must reproduce the above copyright notice,
16  this list of conditions and the following disclaimer in the documentation
17  and/or other materials provided with the distribution.
18  * Neither the name of Intel Corporation nor the names of its contributors may
19  be used to endorse or promote products derived from this software without
20  specific prior written permission.
21 
22  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
23  ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
24  WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
25  DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
26  ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
27  (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
29  ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 
33  ********************************************************************************
34  * Content : Eigen bindings to LAPACKe
35  * Singular Value Decomposition - SVD (divide and conquer variant)
36  ********************************************************************************
37 */
38 #ifndef EIGEN_BDCSVD_LAPACKE_H
39 #define EIGEN_BDCSVD_LAPACKE_H
40 
41 namespace Eigen {
42 
43 namespace internal {
44 
45 namespace lapacke_helpers {
46 
49 // defining a derived class to allow access to protected members
50 template <typename MatrixType_, int Options>
51 class BDCSVD_LAPACKE : public BDCSVD<MatrixType_, Options> {
52  typedef BDCSVD<MatrixType_, Options> SVD;
53  typedef typename SVD::MatrixType MatrixType;
54  typedef typename SVD::Scalar Scalar;
55  typedef typename SVD::RealScalar RealScalar;
56 
57  public:
58  // construct this by moving from a parent object
59  BDCSVD_LAPACKE(SVD&& svd) : SVD(std::move(svd)) {}
60 
61  template <typename Derived>
62  void compute_impl_lapacke(const MatrixBase<Derived>& matrix, unsigned int computationOptions) {
63  SVD::allocate(matrix.rows(), matrix.cols(), computationOptions);
64 
65  SVD::m_nonzeroSingularValues = SVD::m_diagSize;
66 
67  // prepare arguments to ?gesdd
68  const lapack_int matrix_order = lapack_storage_of(matrix);
69  const char jobz = (SVD::m_computeFullU || SVD::m_computeFullV) ? 'A'
70  : (SVD::m_computeThinU || SVD::m_computeThinV) ? 'S'
71  : 'N';
72  const lapack_int u_cols = (jobz == 'A') ? to_lapack(SVD::rows()) : (jobz == 'S') ? to_lapack(SVD::diagSize()) : 1;
73  const lapack_int vt_rows = (jobz == 'A') ? to_lapack(SVD::cols()) : (jobz == 'S') ? to_lapack(SVD::diagSize()) : 1;
74  lapack_int ldu, ldvt;
75  Scalar *u, *vt, dummy;
76  MatrixType localU;
77  if (SVD::computeU() && !(SVD::m_computeThinU && SVD::m_computeFullV)) {
78  ldu = to_lapack(SVD::m_matrixU.outerStride());
79  u = SVD::m_matrixU.data();
80  } else if (SVD::computeV()) {
81  localU.resize(SVD::rows(), u_cols);
82  ldu = to_lapack(localU.outerStride());
83  u = localU.data();
84  } else {
85  ldu = 1;
86  u = &dummy;
87  }
88  MatrixType localV;
89  if (SVD::computeU() || SVD::computeV()) {
90  localV.resize(vt_rows, SVD::cols());
91  ldvt = to_lapack(localV.outerStride());
92  vt = localV.data();
93  } else {
94  ldvt = 1;
95  vt = &dummy;
96  }
97  MatrixType temp;
98  temp = matrix;
99 
100  // actual call to ?gesdd
101  lapack_int info = gesdd(matrix_order, jobz, to_lapack(SVD::rows()), to_lapack(SVD::cols()), to_lapack(temp.data()),
102  to_lapack(temp.outerStride()), (RealScalar*)SVD::m_singularValues.data(), to_lapack(u), ldu,
103  to_lapack(vt), ldvt);
104 
105  // Check the result of the LAPACK call
106  if (info < 0 || !SVD::m_singularValues.allFinite()) {
107  // this includes info == -4 => NaN entry in A
108  SVD::m_info = InvalidInput;
109  } else if (info > 0) {
110  SVD::m_info = NoConvergence;
111  } else {
112  SVD::m_info = Success;
113  if (SVD::m_computeThinU && SVD::m_computeFullV) {
114  SVD::m_matrixU = localU.leftCols(SVD::m_matrixU.cols());
115  }
116  if (SVD::computeV()) {
117  SVD::m_matrixV = localV.adjoint().leftCols(SVD::m_matrixV.cols());
118  }
119  }
120  SVD::m_isInitialized = true;
121  }
122 };
123 
124 template <typename MatrixType_, int Options, typename Derived>
125 BDCSVD<MatrixType_, Options>& BDCSVD_wrapper(BDCSVD<MatrixType_, Options>& svd, const MatrixBase<Derived>& matrix,
126  int computationOptions) {
127  // we need to move to the wrapper type and back
128  BDCSVD_LAPACKE<MatrixType_, Options> tmpSvd(std::move(svd));
129  tmpSvd.compute_impl_lapacke(matrix, computationOptions);
130  svd = std::move(tmpSvd);
131  return svd;
132 }
133 
134 } // end namespace lapacke_helpers
135 
136 } // end namespace internal
137 
138 #define EIGEN_LAPACKE_SDD(EIGTYPE, EIGCOLROW, OPTIONS) \
139  template <> \
140  template <typename Derived> \
141  inline BDCSVD<Matrix<EIGTYPE, Dynamic, Dynamic, EIGCOLROW, Dynamic, Dynamic>, OPTIONS>& \
142  BDCSVD<Matrix<EIGTYPE, Dynamic, Dynamic, EIGCOLROW, Dynamic, Dynamic>, OPTIONS>::compute_impl( \
143  const MatrixBase<Derived>& matrix, unsigned int computationOptions) { \
144  return internal::lapacke_helpers::BDCSVD_wrapper(*this, matrix, computationOptions); \
145  }
146 
147 #define EIGEN_LAPACK_SDD_OPTIONS(OPTIONS) \
148  EIGEN_LAPACKE_SDD(double, ColMajor, OPTIONS) \
149  EIGEN_LAPACKE_SDD(float, ColMajor, OPTIONS) \
150  EIGEN_LAPACKE_SDD(dcomplex, ColMajor, OPTIONS) \
151  EIGEN_LAPACKE_SDD(scomplex, ColMajor, OPTIONS) \
152  \
153  EIGEN_LAPACKE_SDD(double, RowMajor, OPTIONS) \
154  EIGEN_LAPACKE_SDD(float, RowMajor, OPTIONS) \
155  EIGEN_LAPACKE_SDD(dcomplex, RowMajor, OPTIONS) \
156  EIGEN_LAPACKE_SDD(scomplex, RowMajor, OPTIONS)
157 
158 EIGEN_LAPACK_SDD_OPTIONS(0)
159 EIGEN_LAPACK_SDD_OPTIONS(ComputeThinU)
160 EIGEN_LAPACK_SDD_OPTIONS(ComputeThinV)
161 EIGEN_LAPACK_SDD_OPTIONS(ComputeFullU)
162 EIGEN_LAPACK_SDD_OPTIONS(ComputeFullV)
163 EIGEN_LAPACK_SDD_OPTIONS(ComputeThinU | ComputeThinV)
164 EIGEN_LAPACK_SDD_OPTIONS(ComputeFullU | ComputeFullV)
165 EIGEN_LAPACK_SDD_OPTIONS(ComputeThinU | ComputeFullV)
166 EIGEN_LAPACK_SDD_OPTIONS(ComputeFullU | ComputeThinV)
167 
168 #undef EIGEN_LAPACK_SDD_OPTIONS
169 
170 #undef EIGEN_LAPACKE_SDD
171 
172 } // end namespace Eigen
173 
174 #endif // EIGEN_BDCSVD_LAPACKE_H
Definition: Constants.h:389
ComputationInfo info() const
Reports whether previous computation was successful.
Definition: SVDBase.h:300
Definition: Constants.h:395
Namespace containing all symbols from the Eigen library.
Definition: B01_Experimental.dox:1
Definition: BFloat16.h:231
bool computeV() const
Definition: SVDBase.h:275
Definition: Constants.h:447
Definition: Constants.h:440
bool computeU() const
Definition: SVDBase.h:273
Definition: Constants.h:393
Definition: Constants.h:391
Definition: Constants.h:444