src/SVD/JacobiSVD_LAPACKE.h

0001 /*
0002  Copyright (c) 2011, Intel Corporation. All rights reserved.
0003
0004  Redistribution and use in source and binary forms, with or without modification,
0005  are permitted provided that the following conditions are met:
0006
0007  * Redistributions of source code must retain the above copyright notice, this
0008    list of conditions and the following disclaimer.
0009  * Redistributions in binary form must reproduce the above copyright notice,
0010    this list of conditions and the following disclaimer in the documentation
0011    and/or other materials provided with the distribution.
0012  * Neither the name of Intel Corporation nor the names of its contributors may
0013    be used to endorse or promote products derived from this software without
0014    specific prior written permission.
0015
0016  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
0017  ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
0018  WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
0019  DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
0020  ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
0021  (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
0022  LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
0023  ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
0024  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
0025  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
0026
0027  ********************************************************************************
0028  *   Content : Eigen bindings to LAPACKe
0029  *    Singular Value Decomposition - SVD.
0030  ********************************************************************************
0031 */
0032
0033 #ifndef EIGEN_JACOBISVD_LAPACKE_H
0034 #define EIGEN_JACOBISVD_LAPACKE_H
0035
0036 namespace Eigen {
0037
0038 /** \internal Specialization for the data types supported by LAPACKe */
0039
0040 #define EIGEN_LAPACKE_SVD(EIGTYPE, LAPACKE_TYPE, LAPACKE_RTYPE, LAPACKE_PREFIX, EIGCOLROW, LAPACKE_COLROW) \
0041 template<> inline \
0042 JacobiSVD<Matrix<EIGTYPE, Dynamic, Dynamic, EIGCOLROW, Dynamic, Dynamic>, ColPivHouseholderQRPreconditioner>& \
0043 JacobiSVD<Matrix<EIGTYPE, Dynamic, Dynamic, EIGCOLROW, Dynamic, Dynamic>, ColPivHouseholderQRPreconditioner>::compute(const Matrix<EIGTYPE, Dynamic, Dynamic, EIGCOLROW, Dynamic, Dynamic>& matrix, unsigned int computationOptions) \
0044 { \
0045   typedef Matrix<EIGTYPE, Dynamic, Dynamic, EIGCOLROW, Dynamic, Dynamic> MatrixType; \
0046   /*typedef MatrixType::Scalar Scalar;*/ \
0047   /*typedef MatrixType::RealScalar RealScalar;*/ \
0048   allocate(matrix.rows(), matrix.cols(), computationOptions); \
0049 \
0050   /*const RealScalar precision = RealScalar(2) * NumTraits<Scalar>::epsilon();*/ \
0051   m_nonzeroSingularValues = m_diagSize; \
0052 \
0053   lapack_int lda = internal::convert_index<lapack_int>(matrix.outerStride()), ldu, ldvt; \
0054   lapack_int matrix_order = LAPACKE_COLROW; \
0055   char jobu, jobvt; \
0056   LAPACKE_TYPE *u, *vt, dummy; \
0057   jobu  = (m_computeFullU) ? 'A' : (m_computeThinU) ? 'S' : 'N'; \
0058   jobvt = (m_computeFullV) ? 'A' : (m_computeThinV) ? 'S' : 'N'; \
0059   if (computeU()) { \
0060     ldu  = internal::convert_index<lapack_int>(m_matrixU.outerStride()); \
0061     u    = (LAPACKE_TYPE*)m_matrixU.data(); \
0062   } else { ldu=1; u=&dummy; }\
0063   MatrixType localV; \
0064   lapack_int vt_rows = (m_computeFullV) ? internal::convert_index<lapack_int>(m_cols) : (m_computeThinV) ? internal::convert_index<lapack_int>(m_diagSize) : 1; \
0065   if (computeV()) { \
0066     localV.resize(vt_rows, m_cols); \
0067     ldvt  = internal::convert_index<lapack_int>(localV.outerStride()); \
0068     vt   = (LAPACKE_TYPE*)localV.data(); \
0069   } else { ldvt=1; vt=&dummy; }\
0070   Matrix<LAPACKE_RTYPE, Dynamic, Dynamic> superb; superb.resize(m_diagSize, 1); \
0071   MatrixType m_temp; m_temp = matrix; \
0072   LAPACKE_##LAPACKE_PREFIX##gesvd( matrix_order, jobu, jobvt, internal::convert_index<lapack_int>(m_rows), internal::convert_index<lapack_int>(m_cols), (LAPACKE_TYPE*)m_temp.data(), lda, (LAPACKE_RTYPE*)m_singularValues.data(), u, ldu, vt, ldvt, superb.data()); \
0073   if (computeV()) m_matrixV = localV.adjoint(); \
0074  /* for(int i=0;i<m_diagSize;i++) if (m_singularValues.coeffRef(i) < precision) { m_nonzeroSingularValues--; m_singularValues.coeffRef(i)=RealScalar(0);}*/ \
0075   m_isInitialized = true; \
0076   return *this; \
0077 }
0078
0079 EIGEN_LAPACKE_SVD(double,   double,                double, d, ColMajor, LAPACK_COL_MAJOR)
0080 EIGEN_LAPACKE_SVD(float,    float,                 float , s, ColMajor, LAPACK_COL_MAJOR)
0081 EIGEN_LAPACKE_SVD(dcomplex, lapack_complex_double, double, z, ColMajor, LAPACK_COL_MAJOR)
0082 EIGEN_LAPACKE_SVD(scomplex, lapack_complex_float,  float , c, ColMajor, LAPACK_COL_MAJOR)
0083
0084 EIGEN_LAPACKE_SVD(double,   double,                double, d, RowMajor, LAPACK_ROW_MAJOR)
0085 EIGEN_LAPACKE_SVD(float,    float,                 float , s, RowMajor, LAPACK_ROW_MAJOR)
0086 EIGEN_LAPACKE_SVD(dcomplex, lapack_complex_double, double, z, RowMajor, LAPACK_ROW_MAJOR)
0087 EIGEN_LAPACKE_SVD(scomplex, lapack_complex_float,  float , c, RowMajor, LAPACK_ROW_MAJOR)
0088
0089 } // end namespace Eigen
0090
0091 #endif // EIGEN_JACOBISVD_LAPACKE_H