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
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
|
/*
[auto_generated]
boost/numeric/odeint/external/mkl/mkl_operations.hpp
[begin_description]
Wrapper classes for intel math kernel library types.
Get a free, non-commercial download of MKL at
http://software.intel.com/en-us/articles/non-commercial-software-download/
[end_description]
Copyright 2010-2011 Mario Mulansky
Copyright 2011-2013 Karsten Ahnert
Distributed under the Boost Software License, Version 1.0.
(See accompanying file LICENSE_1_0.txt or
copy at http://www.boost.org/LICENSE_1_0.txt)
*/
#ifndef BOOST_NUMERIC_ODEINT_EXTERNAL_MKL_MKL_OPERATIONS_HPP_INCLUDED
#define BOOST_NUMERIC_ODEINT_EXTERNAL_MKL_MKL_OPERATIONS_HPP_INCLUDED
#include <iostream>
#include <mkl_cblas.h>
#include <boost/numeric/odeint/algebra/default_operations.hpp>
/* exemplary example for writing bindings to the Intel MKL library
* see test/mkl for how to use mkl with odeint
* this is a quick and dirty implementation showing the general possibility.
* It works only with containers based on double and sequential memory allocation.
*/
namespace boost {
namespace numeric {
namespace odeint {
/* only defined for doubles */
struct mkl_operations
{
//template< class Fac1 , class Fac2 > struct scale_sum2;
template< class F1 = double , class F2 = F1 >
struct scale_sum2
{
typedef double Fac1;
typedef double Fac2;
const Fac1 m_alpha1;
const Fac2 m_alpha2;
scale_sum2( const Fac1 alpha1 , const Fac2 alpha2 ) : m_alpha1( alpha1 ) , m_alpha2( alpha2 ) { }
template< class T1 , class T2 , class T3 >
void operator()( T1 &t1 , const T2 &t2 , const T3 &t3) const
{ // t1 = m_alpha1 * t2 + m_alpha2 * t3;
// we get Containers that have size() and [i]-access
const int n = t1.size();
//boost::numeric::odeint::copy( t1 , t3 );
if( &(t2[0]) != &(t1[0]) )
{
cblas_dcopy( n , &(t2[0]) , 1 , &(t1[0]) , 1 );
}
cblas_dscal( n , m_alpha1 , &(t1[0]) , 1 );
cblas_daxpy( n , m_alpha2 , &(t3[0]) , 1 , &(t1[0]) , 1 );
//daxpby( &n , &m_alpha2 , &(t3[0]) , &one , &m_alpha1 , &(t1[0]) , &one );
}
};
template< class F1 = double , class F2 = F1 , class F3 = F2 >
struct scale_sum3
{
typedef double Fac1;
typedef double Fac2;
typedef double Fac3;
const Fac1 m_alpha1;
const Fac2 m_alpha2;
const Fac3 m_alpha3;
scale_sum3( const Fac1 alpha1 , const Fac2 alpha2 , const Fac3 alpha3 )
: m_alpha1( alpha1 ) , m_alpha2( alpha2 ) , m_alpha3( alpha3 ) { }
template< class T1 , class T2 , class T3 , class T4 >
void operator()( T1 &t1 , const T2 &t2 , const T3 &t3 , const T4 &t4 ) const
{ // t1 = m_alpha1 * t2 + m_alpha2 * t3 + m_alpha3 * t4;
// we get Containers that have size() and [i]-access
const int n = t1.size();
//boost::numeric::odeint::copy( t1 , t3 );
if( &(t2[0]) != &(t1[0]) )
{
cblas_dcopy( n , &(t2[0]) , 1 , &(t1[0]) , 1 );
}
cblas_dscal( n , m_alpha1 , &(t1[0]) , 1 );
cblas_daxpy( n , m_alpha2 , &(t3[0]) , 1 , &(t1[0]) , 1 );
//daxpby( &n , &m_alpha2 , &(t3[0]) , &one , &m_alpha1 , &(t1[0]) , &one );
cblas_daxpy( n , m_alpha3 , &(t4[0]) , 1 , &(t1[0]) , 1 );
}
};
template< class F1 = double , class F2 = F1 , class F3 = F2 , class F4 = F3 >
struct scale_sum4
{
typedef double Fac1;
typedef double Fac2;
typedef double Fac3;
typedef double Fac4;
const Fac1 m_alpha1;
const Fac2 m_alpha2;
const Fac3 m_alpha3;
const Fac4 m_alpha4;
scale_sum4( const Fac1 alpha1 , const Fac2 alpha2 , const Fac3 alpha3 , const Fac4 alpha4 )
: m_alpha1( alpha1 ) , m_alpha2( alpha2 ) , m_alpha3( alpha3 ) , m_alpha4( alpha4 ) { }
template< class T1 , class T2 , class T3 , class T4 , class T5 >
void operator()( T1 &t1 , const T2 &t2 , const T3 &t3 , const T4 &t4 , const T5 &t5 ) const
{ // t1 = m_alpha1 * t2 + m_alpha2 * t3 + m_alpha3 * t4 + m_alpha4 * t5;
// we get Containers that have size() and [i]-access
const int n = t1.size();
//boost::numeric::odeint::copy( t1 , t3 );
if( &(t2[0]) != &(t1[0]) )
{
cblas_dcopy( n , &(t2[0]) , 1 , &(t1[0]) , 1 );
}
cblas_dscal( n , m_alpha1 , &(t1[0]) , 1 );
cblas_daxpy( n , m_alpha2 , &(t3[0]) , 1 , &(t1[0]) , 1 );
//daxpby( &n , &m_alpha2 , &(t3[0]) , &one , &m_alpha1 , &(t1[0]) , &one );
cblas_daxpy( n , m_alpha3 , &(t4[0]) , 1 , &(t1[0]) , 1 );
cblas_daxpy( n , m_alpha4 , &(t5[0]) , 1 , &(t1[0]) , 1 );
}
};
template< class F1 = double , class F2 = F1 , class F3 = F2 , class F4 = F3 , class F5 = F4 >
struct scale_sum5
{
typedef double Fac1;
typedef double Fac2;
typedef double Fac3;
typedef double Fac4;
typedef double Fac5;
const Fac1 m_alpha1;
const Fac2 m_alpha2;
const Fac3 m_alpha3;
const Fac4 m_alpha4;
const Fac5 m_alpha5;
scale_sum5( const Fac1 alpha1 , const Fac2 alpha2 , const Fac3 alpha3 , const Fac4 alpha4 , const Fac5 alpha5 )
: m_alpha1( alpha1 ) , m_alpha2( alpha2 ) , m_alpha3( alpha3 ) , m_alpha4( alpha4 ) , m_alpha5( alpha5 ) { }
template< class T1 , class T2 , class T3 , class T4 , class T5 , class T6 >
void operator()( T1 &t1 , const T2 &t2 , const T3 &t3 , const T4 &t4 , const T5 &t5 , const T6 &t6 ) const
{ // t1 = m_alpha1 * t2 + m_alpha2 * t3 + m_alpha3 * t4 + m_alpha4 * t5 + m_alpha5 * t6;
// we get Containers that have size() and [i]-access
const int n = t1.size();
//boost::numeric::odeint::copy( t1 , t3 );
if( &(t2[0]) != &(t1[0]) )
{
cblas_dcopy( n , &(t2[0]) , 1 , &(t1[0]) , 1 );
}
cblas_dscal( n , m_alpha1 , &(t1[0]) , 1 );
cblas_daxpy( n , m_alpha2 , &(t3[0]) , 1 , &(t1[0]) , 1 );
//daxpby( &n , &m_alpha2 , &(t3[0]) , &one , &m_alpha1 , &(t1[0]) , &one );
cblas_daxpy( n , m_alpha3 , &(t4[0]) , 1 , &(t1[0]) , 1 );
cblas_daxpy( n , m_alpha4 , &(t5[0]) , 1 , &(t1[0]) , 1 );
cblas_daxpy( n , m_alpha5 , &(t6[0]) , 1 , &(t1[0]) , 1 );
}
};
};
} // odeint
} // numeric
} // boost
#endif // BOOST_NUMERIC_ODEINT_EXTERNAL_MKL_MKL_OPERATIONS_HPP_INCLUDED
|
