1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
|
/*
* Tiny Vector Matrix Library
* Dense Vector Matrix Libary of Tiny size using Expression Templates
*
* Copyright (C) 2001 - 2003 Olaf Petzold <opetzold@users.sourceforge.net>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* $Id: Gemm.h,v 1.8 2004/06/16 09:30:07 opetzold Exp $
*/
#ifndef TVMET_LOOP_GEMM_H
#define TVMET_LOOP_GEMM_H
namespace tvmet {
namespace loop {
/**
* \class gemm Gemm.h "tvmet/loop/Gemm.h"
* \brief class for matrix-matrix product using loop unrolling.
* using formula
* \f[
* M_1\,M_2
* \f]
* \par Example:
* \code
* template<class T, int Rows1, int Cols1, int Cols2>
* inline
* void
* prod(const Matrix<T, Rows1, Cols1>& lhs, const Matrix<T, Cols1, Cols2>& rhs,
* Matrix<T, Rows1, Cols2>& dest)
* {
* for (int i = 0; i != Rows1; ++i) {
* for (int j = 0; j != Cols2; ++j) {
* dest(i, j) = tvmet::loop::gemm<Rows1, Cols1, Cols2>().prod(lhs, rhs, i, j);
* }
* }
* }
* \endcode
* \note The number of rows of rhs matrix have to be equal to cols of lhs matrix.
* The result is a (Rows1 x Cols2) matrix.
*/
template<int Rows1, int Cols1,
int Cols2>
class gemm
{
gemm(const gemm&);
gemm& operator=(const gemm&);
private:
enum {
count = Cols1,
N = (count+7)/8
};
public:
gemm() { }
public:
template<class E1, class E2>
static inline
typename PromoteTraits<
typename E1::value_type,
typename E2::value_type
>::value_type
prod(const E1& lhs, const E2& rhs, int i, int j) {
typename PromoteTraits<
typename E1::value_type,
typename E2::value_type
>::value_type sum(0);
int k(0);
int n(N);
// Duff's device
switch(count % 8) {
case 0: do { sum += lhs(i, k) * rhs(k, j); ++k;
case 7: sum += lhs(i, k) * rhs(k, j); ++k;
case 6: sum += lhs(i, k) * rhs(k, j); ++k;
case 5: sum += lhs(i, k) * rhs(k, j); ++k;
case 4: sum += lhs(i, k) * rhs(k, j); ++k;
case 3: sum += lhs(i, k) * rhs(k, j); ++k;
case 2: sum += lhs(i, k) * rhs(k, j); ++k;
case 1: sum += lhs(i, k) * rhs(k, j); ++k;
} while(--n != 0);
}
return sum;
}
};
} // namespace loop
} // namespace tvmet
#endif /* TVMET_LOOP_GEMM_H */
// Local Variables:
// mode:C++
// End:
|
