win_eigen: TriangularMatrixVector

Go to the documentation of this file.
00001 /*
00002  Copyright (c) 2011, Intel Corporation. All rights reserved.
00003 
00004  Redistribution and use in source and binary forms, with or without modification,
00005  are permitted provided that the following conditions are met:
00006 
00007  * Redistributions of source code must retain the above copyright notice, this
00008    list of conditions and the following disclaimer.
00009  * Redistributions in binary form must reproduce the above copyright notice,
00010    this list of conditions and the following disclaimer in the documentation
00011    and/or other materials provided with the distribution.
00012  * Neither the name of Intel Corporation nor the names of its contributors may
00013    be used to endorse or promote products derived from this software without
00014    specific prior written permission.
00015 
00016  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
00017  ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
00018  WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
00019  DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
00020  ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
00021  (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
00022  LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
00023  ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
00024  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
00025  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
00026 
00027  ********************************************************************************
00028  *   Content : Eigen bindings to Intel(R) MKL
00029  *   Triangular matrix-vector product functionality based on ?TRMV.
00030  ********************************************************************************
00031 */
00032 
00033 #ifndef EIGEN_TRIANGULAR_MATRIX_VECTOR_MKL_H
00034 #define EIGEN_TRIANGULAR_MATRIX_VECTOR_MKL_H
00035 
00036 namespace Eigen { 
00037 
00038 namespace internal {
00039 
00040 /**********************************************************************
00041 * This file implements triangular matrix-vector multiplication using BLAS
00042 **********************************************************************/
00043 
00044 // trmv/hemv specialization
00045 
00046 template<typename Index, int Mode, typename LhsScalar, bool ConjLhs, typename RhsScalar, bool ConjRhs, int StorageOrder>
00047 struct triangular_matrix_vector_product_trmv :
00048   triangular_matrix_vector_product<Index,Mode,LhsScalar,ConjLhs,RhsScalar,ConjRhs,StorageOrder,BuiltIn> {};
00049 
00050 #define EIGEN_MKL_TRMV_SPECIALIZE(Scalar) \
00051 template<typename Index, int Mode, bool ConjLhs, bool ConjRhs> \
00052 struct triangular_matrix_vector_product<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs,ColMajor,Specialized> { \
00053  static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \
00054                                      const Scalar* _rhs, Index rhsIncr, Scalar* _res, Index resIncr, Scalar alpha) { \
00055       triangular_matrix_vector_product_trmv<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs,ColMajor>::run( \
00056         _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \
00057   } \
00058 }; \
00059 template<typename Index, int Mode, bool ConjLhs, bool ConjRhs> \
00060 struct triangular_matrix_vector_product<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs,RowMajor,Specialized> { \
00061  static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \
00062                                      const Scalar* _rhs, Index rhsIncr, Scalar* _res, Index resIncr, Scalar alpha) { \
00063       triangular_matrix_vector_product_trmv<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs,RowMajor>::run( \
00064         _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \
00065   } \
00066 };
00067 
00068 EIGEN_MKL_TRMV_SPECIALIZE(double)
00069 EIGEN_MKL_TRMV_SPECIALIZE(float)
00070 EIGEN_MKL_TRMV_SPECIALIZE(dcomplex)
00071 EIGEN_MKL_TRMV_SPECIALIZE(scomplex)
00072 
00073 // implements col-major: res += alpha * op(triangular) * vector
00074 #define EIGEN_MKL_TRMV_CM(EIGTYPE, MKLTYPE, EIGPREFIX, MKLPREFIX) \
00075 template<typename Index, int Mode, bool ConjLhs, bool ConjRhs> \
00076 struct triangular_matrix_vector_product_trmv<Index,Mode,EIGTYPE,ConjLhs,EIGTYPE,ConjRhs,ColMajor> { \
00077   enum { \
00078     IsLower = (Mode&Lower) == Lower, \
00079     SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \
00080     IsUnitDiag  = (Mode&UnitDiag) ? 1 : 0, \
00081     IsZeroDiag  = (Mode&ZeroDiag) ? 1 : 0, \
00082     LowUp = IsLower ? Lower : Upper \
00083   }; \
00084  static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \
00085                              const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \
00086  { \
00087    if (ConjLhs || IsZeroDiag) { \
00088      triangular_matrix_vector_product<Index,Mode,EIGTYPE,ConjLhs,EIGTYPE,ConjRhs,ColMajor,BuiltIn>::run( \
00089        _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \
00090      return; \
00091    }\
00092    Index size = (std::min)(_rows,_cols); \
00093    Index rows = IsLower ? _rows : size; \
00094    Index cols = IsLower ? size : _cols; \
00095 \
00096    typedef VectorX##EIGPREFIX VectorRhs; \
00097    EIGTYPE *x, *y;\
00098 \
00099 /* Set x*/ \
00100    Map<const VectorRhs, 0, InnerStride<> > rhs(_rhs,cols,InnerStride<>(rhsIncr)); \
00101    VectorRhs x_tmp; \
00102    if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \
00103    x = x_tmp.data(); \
00104 \
00105 /* Square part handling */\
00106 \
00107    char trans, uplo, diag; \
00108    MKL_INT m, n, lda, incx, incy; \
00109    EIGTYPE const *a; \
00110    MKLTYPE alpha_, beta_; \
00111    assign_scalar_eig2mkl<MKLTYPE, EIGTYPE>(alpha_, alpha); \
00112    assign_scalar_eig2mkl<MKLTYPE, EIGTYPE>(beta_, EIGTYPE(1)); \
00113 \
00114 /* Set m, n */ \
00115    n = (MKL_INT)size; \
00116    lda = lhsStride; \
00117    incx = 1; \
00118    incy = resIncr; \
00119 \
00120 /* Set uplo, trans and diag*/ \
00121    trans = 'N'; \
00122    uplo = IsLower ? 'L' : 'U'; \
00123    diag = IsUnitDiag ? 'U' : 'N'; \
00124 \
00125 /* call ?TRMV*/ \
00126    MKLPREFIX##trmv(&uplo, &trans, &diag, &n, (const MKLTYPE*)_lhs, &lda, (MKLTYPE*)x, &incx); \
00127 \
00128 /* Add op(a_tr)rhs into res*/ \
00129    MKLPREFIX##axpy(&n, &alpha_,(const MKLTYPE*)x, &incx, (MKLTYPE*)_res, &incy); \
00130 /* Non-square case - doesn't fit to MKL ?TRMV. Fall to default triangular product*/ \
00131    if (size<(std::max)(rows,cols)) { \
00132      typedef Matrix<EIGTYPE, Dynamic, Dynamic> MatrixLhs; \
00133      if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \
00134      x = x_tmp.data(); \
00135      if (size<rows) { \
00136        y = _res + size*resIncr; \
00137        a = _lhs + size; \
00138        m = rows-size; \
00139        n = size; \
00140      } \
00141      else { \
00142        x += size; \
00143        y = _res; \
00144        a = _lhs + size*lda; \
00145        m = size; \
00146        n = cols-size; \
00147      } \
00148      MKLPREFIX##gemv(&trans, &m, &n, &alpha_, (const MKLTYPE*)a, &lda, (const MKLTYPE*)x, &incx, &beta_, (MKLTYPE*)y, &incy); \
00149    } \
00150   } \
00151 };
00152 
00153 EIGEN_MKL_TRMV_CM(double, double, d, d)
00154 EIGEN_MKL_TRMV_CM(dcomplex, MKL_Complex16, cd, z)
00155 EIGEN_MKL_TRMV_CM(float, float, f, s)
00156 EIGEN_MKL_TRMV_CM(scomplex, MKL_Complex8, cf, c)
00157 
00158 // implements row-major: res += alpha * op(triangular) * vector
00159 #define EIGEN_MKL_TRMV_RM(EIGTYPE, MKLTYPE, EIGPREFIX, MKLPREFIX) \
00160 template<typename Index, int Mode, bool ConjLhs, bool ConjRhs> \
00161 struct triangular_matrix_vector_product_trmv<Index,Mode,EIGTYPE,ConjLhs,EIGTYPE,ConjRhs,RowMajor> { \
00162   enum { \
00163     IsLower = (Mode&Lower) == Lower, \
00164     SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \
00165     IsUnitDiag  = (Mode&UnitDiag) ? 1 : 0, \
00166     IsZeroDiag  = (Mode&ZeroDiag) ? 1 : 0, \
00167     LowUp = IsLower ? Lower : Upper \
00168   }; \
00169  static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \
00170                              const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \
00171  { \
00172    if (IsZeroDiag) { \
00173      triangular_matrix_vector_product<Index,Mode,EIGTYPE,ConjLhs,EIGTYPE,ConjRhs,RowMajor,BuiltIn>::run( \
00174        _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \
00175      return; \
00176    }\
00177    Index size = (std::min)(_rows,_cols); \
00178    Index rows = IsLower ? _rows : size; \
00179    Index cols = IsLower ? size : _cols; \
00180 \
00181    typedef VectorX##EIGPREFIX VectorRhs; \
00182    EIGTYPE *x, *y;\
00183 \
00184 /* Set x*/ \
00185    Map<const VectorRhs, 0, InnerStride<> > rhs(_rhs,cols,InnerStride<>(rhsIncr)); \
00186    VectorRhs x_tmp; \
00187    if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \
00188    x = x_tmp.data(); \
00189 \
00190 /* Square part handling */\
00191 \
00192    char trans, uplo, diag; \
00193    MKL_INT m, n, lda, incx, incy; \
00194    EIGTYPE const *a; \
00195    MKLTYPE alpha_, beta_; \
00196    assign_scalar_eig2mkl<MKLTYPE, EIGTYPE>(alpha_, alpha); \
00197    assign_scalar_eig2mkl<MKLTYPE, EIGTYPE>(beta_, EIGTYPE(1)); \
00198 \
00199 /* Set m, n */ \
00200    n = (MKL_INT)size; \
00201    lda = lhsStride; \
00202    incx = 1; \
00203    incy = resIncr; \
00204 \
00205 /* Set uplo, trans and diag*/ \
00206    trans = ConjLhs ? 'C' : 'T'; \
00207    uplo = IsLower ? 'U' : 'L'; \
00208    diag = IsUnitDiag ? 'U' : 'N'; \
00209 \
00210 /* call ?TRMV*/ \
00211    MKLPREFIX##trmv(&uplo, &trans, &diag, &n, (const MKLTYPE*)_lhs, &lda, (MKLTYPE*)x, &incx); \
00212 \
00213 /* Add op(a_tr)rhs into res*/ \
00214    MKLPREFIX##axpy(&n, &alpha_,(const MKLTYPE*)x, &incx, (MKLTYPE*)_res, &incy); \
00215 /* Non-square case - doesn't fit to MKL ?TRMV. Fall to default triangular product*/ \
00216    if (size<(std::max)(rows,cols)) { \
00217      typedef Matrix<EIGTYPE, Dynamic, Dynamic> MatrixLhs; \
00218      if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \
00219      x = x_tmp.data(); \
00220      if (size<rows) { \
00221        y = _res + size*resIncr; \
00222        a = _lhs + size*lda; \
00223        m = rows-size; \
00224        n = size; \
00225      } \
00226      else { \
00227        x += size; \
00228        y = _res; \
00229        a = _lhs + size; \
00230        m = size; \
00231        n = cols-size; \
00232      } \
00233      MKLPREFIX##gemv(&trans, &n, &m, &alpha_, (const MKLTYPE*)a, &lda, (const MKLTYPE*)x, &incx, &beta_, (MKLTYPE*)y, &incy); \
00234    } \
00235   } \
00236 };
00237 
00238 EIGEN_MKL_TRMV_RM(double, double, d, d)
00239 EIGEN_MKL_TRMV_RM(dcomplex, MKL_Complex16, cd, z)
00240 EIGEN_MKL_TRMV_RM(float, float, f, s)
00241 EIGEN_MKL_TRMV_RM(scomplex, MKL_Complex8, cf, c)
00242 
00243 } // end namespase internal
00244 
00245 } // end namespace Eigen
00246 
00247 #endif // EIGEN_TRIANGULAR_MATRIX_VECTOR_MKL_H