1/*
2 Copyright (c) 2011, Intel Corporation. All rights reserved.
3
4 Redistribution and use in source and binary forms, with or without modification,
5 are permitted provided that the following conditions are met:
6
7 * Redistributions of source code must retain the above copyright notice, this
8   list of conditions and the following disclaimer.
9 * Redistributions in binary form must reproduce the above copyright notice,
10   this list of conditions and the following disclaimer in the documentation
11   and/or other materials provided with the distribution.
12 * Neither the name of Intel Corporation nor the names of its contributors may
13   be used to endorse or promote products derived from this software without
14   specific prior written permission.
15
16 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
17 ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
18 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
19 DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
20 ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
21 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
22 LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
23 ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
25 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26
27 ********************************************************************************
28 *   Content : Eigen bindings to Intel(R) MKL
29 *   General matrix-vector product functionality based on ?GEMV.
30 ********************************************************************************
31*/
32
33#ifndef EIGEN_GENERAL_MATRIX_VECTOR_MKL_H
34#define EIGEN_GENERAL_MATRIX_VECTOR_MKL_H
35
36namespace Eigen {
37
38namespace internal {
39
40/**********************************************************************
41* This file implements general matrix-vector multiplication using BLAS
42* gemv function via partial specialization of
43* general_matrix_vector_product::run(..) method for float, double,
44* std::complex<float> and std::complex<double> types
45**********************************************************************/
46
47// gemv specialization
48
49template<typename Index, typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, typename RhsScalar, bool ConjugateRhs>
50struct general_matrix_vector_product_gemv :
51  general_matrix_vector_product<Index,LhsScalar,LhsStorageOrder,ConjugateLhs,RhsScalar,ConjugateRhs,BuiltIn> {};
52
53#define EIGEN_MKL_GEMV_SPECIALIZE(Scalar) \
54template<typename Index, bool ConjugateLhs, bool ConjugateRhs> \
55struct general_matrix_vector_product<Index,Scalar,ColMajor,ConjugateLhs,Scalar,ConjugateRhs,Specialized> { \
56static void run( \
57  Index rows, Index cols, \
58  const Scalar* lhs, Index lhsStride, \
59  const Scalar* rhs, Index rhsIncr, \
60  Scalar* res, Index resIncr, Scalar alpha) \
61{ \
62  if (ConjugateLhs) { \
63    general_matrix_vector_product<Index,Scalar,ColMajor,ConjugateLhs,Scalar,ConjugateRhs,BuiltIn>::run( \
64      rows, cols, lhs, lhsStride, rhs, rhsIncr, res, resIncr, alpha); \
65  } else { \
66    general_matrix_vector_product_gemv<Index,Scalar,ColMajor,ConjugateLhs,Scalar,ConjugateRhs>::run( \
67      rows, cols, lhs, lhsStride, rhs, rhsIncr, res, resIncr, alpha); \
68  } \
69} \
70}; \
71template<typename Index, bool ConjugateLhs, bool ConjugateRhs> \
72struct general_matrix_vector_product<Index,Scalar,RowMajor,ConjugateLhs,Scalar,ConjugateRhs,Specialized> { \
73static void run( \
74  Index rows, Index cols, \
75  const Scalar* lhs, Index lhsStride, \
76  const Scalar* rhs, Index rhsIncr, \
77  Scalar* res, Index resIncr, Scalar alpha) \
78{ \
79    general_matrix_vector_product_gemv<Index,Scalar,RowMajor,ConjugateLhs,Scalar,ConjugateRhs>::run( \
80      rows, cols, lhs, lhsStride, rhs, rhsIncr, res, resIncr, alpha); \
81} \
82}; \
83
84EIGEN_MKL_GEMV_SPECIALIZE(double)
85EIGEN_MKL_GEMV_SPECIALIZE(float)
86EIGEN_MKL_GEMV_SPECIALIZE(dcomplex)
87EIGEN_MKL_GEMV_SPECIALIZE(scomplex)
88
89#define EIGEN_MKL_GEMV_SPECIALIZATION(EIGTYPE,MKLTYPE,MKLPREFIX) \
90template<typename Index, int LhsStorageOrder, bool ConjugateLhs, bool ConjugateRhs> \
91struct general_matrix_vector_product_gemv<Index,EIGTYPE,LhsStorageOrder,ConjugateLhs,EIGTYPE,ConjugateRhs> \
92{ \
93typedef Matrix<EIGTYPE,Dynamic,1,ColMajor> GEMVVector;\
94\
95static void run( \
96  Index rows, Index cols, \
97  const EIGTYPE* lhs, Index lhsStride, \
98  const EIGTYPE* rhs, Index rhsIncr, \
99  EIGTYPE* res, Index resIncr, EIGTYPE alpha) \
100{ \
101  MKL_INT m=rows, n=cols, lda=lhsStride, incx=rhsIncr, incy=resIncr; \
102  MKLTYPE alpha_, beta_; \
103  const EIGTYPE *x_ptr, myone(1); \
104  char trans=(LhsStorageOrder==ColMajor) ? 'N' : (ConjugateLhs) ? 'C' : 'T'; \
105  if (LhsStorageOrder==RowMajor) { \
106    m=cols; \
107    n=rows; \
108  }\
109  assign_scalar_eig2mkl(alpha_, alpha); \
110  assign_scalar_eig2mkl(beta_, myone); \
111  GEMVVector x_tmp; \
112  if (ConjugateRhs) { \
113    Map<const GEMVVector, 0, InnerStride<> > map_x(rhs,cols,1,InnerStride<>(incx)); \
114    x_tmp=map_x.conjugate(); \
115    x_ptr=x_tmp.data(); \
116    incx=1; \
117  } else x_ptr=rhs; \
118  MKLPREFIX##gemv(&trans, &m, &n, &alpha_, (const MKLTYPE*)lhs, &lda, (const MKLTYPE*)x_ptr, &incx, &beta_, (MKLTYPE*)res, &incy); \
119}\
120};
121
122EIGEN_MKL_GEMV_SPECIALIZATION(double,   double,        d)
123EIGEN_MKL_GEMV_SPECIALIZATION(float,    float,         s)
124EIGEN_MKL_GEMV_SPECIALIZATION(dcomplex, MKL_Complex16, z)
125EIGEN_MKL_GEMV_SPECIALIZATION(scomplex, MKL_Complex8,  c)
126
127} // end namespase internal
128
129} // end namespace Eigen
130
131#endif // EIGEN_GENERAL_MATRIX_VECTOR_MKL_H
132