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
182
183
|
/* Ergo, version 3.8, a program for linear scaling electronic structure
* calculations.
* Copyright (C) 2019 Elias Rudberg, Emanuel H. Rubensson, Pawel Salek,
* and Anastasia Kruchinina.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program 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 General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
* Primary academic reference:
* Ergo: An open-source program for linear-scaling electronic structure
* calculations,
* Elias Rudberg, Emanuel H. Rubensson, Pawel Salek, and Anastasia
* Kruchinina,
* SoftwareX 7, 107 (2018),
* <http://dx.doi.org/10.1016/j.softx.2018.03.005>
*
* For further information about Ergo, see <http://www.ergoscf.org>.
*/
/** @file vector_intrin.h
*
* @brief Vector template for convenient access to SIMD operations.
*
* @author Emanuel H. Rubensson
* @date 2009
*
*/
/*
* Templates for efficient gemm kernels.
* For architectures with SSE2 or higher.
*
* Copyright Emanuel Rubensson, 2009
*
*
*
*
*/
#ifndef VECTOR_INTRIN
#define VECTOR_INTRIN
#include "common.h"
#include "g_intrin.h"
/** Vector class template for access to SIMD operations.
*
* Currently supports a limited set of double and single precision SSE operations.
*
*/
template<typename Treal, typename Treg>
class Vector_intrin {
public:
inline void ALWAYS_INLINE load_p(Treal const * ptr) {
values = _mm_load_p(ptr);
}
inline void ALWAYS_INLINE load1_p(Treal const * ptr) {
values = _mm_load1_p(ptr);
}
inline void ALWAYS_INLINE store_p(Treal * ptr) const {
_mm_store_p ( ptr, values );
}
inline Vector_intrin<Treal, Treg>& ALWAYS_INLINE operator*= ( Vector_intrin<Treal, Treg> const & other ) {
values = _mm_mul_p ( other.values, values );
return *this;
}
inline Vector_intrin<Treal, Treg>& ALWAYS_INLINE operator+= ( Vector_intrin<Treal, Treg> const & other ) {
values = _mm_add_p ( other.values, values );
return *this;
}
inline Vector_intrin<Treal, Treg>& ALWAYS_INLINE operator+= ( Treal const * ptr ) {
Treg tmp;
tmp = _mm_load_p(ptr);
values = _mm_add_p ( tmp, values );
return *this;
}
#if 0
inline void ALWAYS_INLINE xor_p( Vector_intrin<Treal, Treg> const & other ) {
values = _mm_xor_p ( other.values, values );
}
#endif
inline void ALWAYS_INLINE set_to_zero() {
values = _mm_xor_p ( values, values );
}
protected:
Treg values;
private:
};
template<typename Treal>
class Vector_intrin<Treal, Treal> {
public:
inline void ALWAYS_INLINE load_p(Treal const * ptr) {
values = *ptr;
}
inline void ALWAYS_INLINE load1_p(Treal const * ptr) {
values = *ptr;
}
inline void ALWAYS_INLINE store_p(Treal * ptr) const {
*ptr = values;
}
inline Vector_intrin<Treal, Treal>& ALWAYS_INLINE operator*= ( Vector_intrin<Treal, Treal> const & other ) {
values *= other.values;
return *this;
}
inline Vector_intrin<Treal, Treal>& ALWAYS_INLINE operator+= ( Vector_intrin<Treal, Treal> const & other ) {
values += other.values;
return *this;
}
inline Vector_intrin<Treal, Treal>& ALWAYS_INLINE operator+= ( Treal const * ptr ) {
values += *ptr;
return *this;
}
#if 0
inline void ALWAYS_INLINE xor_p( Vector_intrin<Treal, Treg> const & other ) {
values = _mm_xor_p ( other.values, values );
}
#endif
inline void ALWAYS_INLINE set_to_zero() {
values = 0;
}
protected:
Treal values;
private:
};
#if 0
template<>
class Vector_intrin<double, double> {
public:
inline void ALWAYS_INLINE load_p(double const * ptr) {
values[0] = *ptr;
values[1] = ptr[1];
}
inline void ALWAYS_INLINE load1_p(double const * ptr) {
values[0] = *ptr;
values[1] = *ptr;
}
inline void ALWAYS_INLINE store_p(double * ptr) const {
ptr[0] = values[0];
ptr[1] = values[1];
}
inline Vector_intrin<double, double>& ALWAYS_INLINE operator*= ( Vector_intrin<double, double> const & other ) {
values[0] *= other.values[0];
values[1] *= other.values[1];
}
inline Vector_intrin<double, double>& ALWAYS_INLINE operator+= ( Vector_intrin<double, double> const & other ) {
values[0] += other.values[0];
values[1] += other.values[1];
}
inline Vector_intrin<double, double>& ALWAYS_INLINE operator+= ( double const * ptr ) {
values[0] += *ptr;
values[1] += ptr[1];
}
#if 0
inline void ALWAYS_INLINE xor_p( Vector_intrin<double, Treg> const & other ) {
values = _mm_xor_p ( other.values, values );
}
#endif
inline void ALWAYS_INLINE set_to_zero() {
values[0] = 0;
values[1] = 0;
}
protected:
double values[2];
private:
};
#endif
#endif // VECTOR_INTRIN
|
