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
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
|
/***************************************************************************
* Copyright (c) Johan Mabille, Sylvain Corlay, Wolf Vollprecht and *
* Martin Renou *
* Copyright (c) QuantStack *
* Copyright (c) Serge Guelton *
* *
* Distributed under the terms of the BSD 3-Clause License. *
* *
* The full license is in the file LICENSE, distributed with this software. *
****************************************************************************/
#ifndef XSIMD_GENERIC_ARITHMETIC_HPP
#define XSIMD_GENERIC_ARITHMETIC_HPP
#include <complex>
#include <limits>
#include <type_traits>
#include "./xsimd_generic_details.hpp"
namespace xsimd
{
namespace kernel
{
using namespace types;
// bitwise_lshift
template <class A, class T, class /*=typename std::enable_if<std::is_integral<T>::value, void>::type*/>
inline batch<T, A> bitwise_lshift(batch<T, A> const& self, batch<T, A> const& other, requires_arch<generic>) noexcept
{
return detail::apply([](T x, T y) noexcept
{ return x << y; },
self, other);
}
// bitwise_rshift
template <class A, class T, class /*=typename std::enable_if<std::is_integral<T>::value, void>::type*/>
inline batch<T, A> bitwise_rshift(batch<T, A> const& self, batch<T, A> const& other, requires_arch<generic>) noexcept
{
return detail::apply([](T x, T y) noexcept
{ return x >> y; },
self, other);
}
// decr
template <class A, class T>
inline batch<T, A> decr(batch<T, A> const& self, requires_arch<generic>) noexcept
{
return self - T(1);
}
// decr_if
template <class A, class T, class Mask>
inline batch<T, A> decr_if(batch<T, A> const& self, Mask const& mask, requires_arch<generic>) noexcept
{
return select(mask, decr(self), self);
}
// div
template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
inline batch<T, A> div(batch<T, A> const& self, batch<T, A> const& other, requires_arch<generic>) noexcept
{
return detail::apply([](T x, T y) noexcept -> T
{ return x / y; },
self, other);
}
// fma
template <class A, class T>
inline batch<T, A> fma(batch<T, A> const& x, batch<T, A> const& y, batch<T, A> const& z, requires_arch<generic>) noexcept
{
return x * y + z;
}
template <class A, class T>
inline batch<std::complex<T>, A> fma(batch<std::complex<T>, A> const& x, batch<std::complex<T>, A> const& y, batch<std::complex<T>, A> const& z, requires_arch<generic>) noexcept
{
auto res_r = fms(x.real(), y.real(), fms(x.imag(), y.imag(), z.real()));
auto res_i = fma(x.real(), y.imag(), fma(x.imag(), y.real(), z.imag()));
return { res_r, res_i };
}
// fms
template <class A, class T>
inline batch<T, A> fms(batch<T, A> const& x, batch<T, A> const& y, batch<T, A> const& z, requires_arch<generic>) noexcept
{
return x * y - z;
}
template <class A, class T>
inline batch<std::complex<T>, A> fms(batch<std::complex<T>, A> const& x, batch<std::complex<T>, A> const& y, batch<std::complex<T>, A> const& z, requires_arch<generic>) noexcept
{
auto res_r = fms(x.real(), y.real(), fma(x.imag(), y.imag(), z.real()));
auto res_i = fma(x.real(), y.imag(), fms(x.imag(), y.real(), z.imag()));
return { res_r, res_i };
}
// fnma
template <class A, class T>
inline batch<T, A> fnma(batch<T, A> const& x, batch<T, A> const& y, batch<T, A> const& z, requires_arch<generic>) noexcept
{
return -x * y + z;
}
template <class A, class T>
inline batch<std::complex<T>, A> fnma(batch<std::complex<T>, A> const& x, batch<std::complex<T>, A> const& y, batch<std::complex<T>, A> const& z, requires_arch<generic>) noexcept
{
auto res_r = -fms(x.real(), y.real(), fma(x.imag(), y.imag(), z.real()));
auto res_i = -fma(x.real(), y.imag(), fms(x.imag(), y.real(), z.imag()));
return { res_r, res_i };
}
// fnms
template <class A, class T>
inline batch<T, A> fnms(batch<T, A> const& x, batch<T, A> const& y, batch<T, A> const& z, requires_arch<generic>) noexcept
{
return -x * y - z;
}
template <class A, class T>
inline batch<std::complex<T>, A> fnms(batch<std::complex<T>, A> const& x, batch<std::complex<T>, A> const& y, batch<std::complex<T>, A> const& z, requires_arch<generic>) noexcept
{
auto res_r = -fms(x.real(), y.real(), fms(x.imag(), y.imag(), z.real()));
auto res_i = -fma(x.real(), y.imag(), fma(x.imag(), y.real(), z.imag()));
return { res_r, res_i };
}
// hadd
template <class A, class T, class /*=typename std::enable_if<std::is_integral<T>::value, void>::type*/>
inline T hadd(batch<T, A> const& self, requires_arch<generic>) noexcept
{
alignas(A::alignment()) T buffer[batch<T, A>::size];
self.store_aligned(buffer);
T res = 0;
for (T val : buffer)
{
res += val;
}
return res;
}
// incr
template <class A, class T>
inline batch<T, A> incr(batch<T, A> const& self, requires_arch<generic>) noexcept
{
return self + T(1);
}
// incr_if
template <class A, class T, class Mask>
inline batch<T, A> incr_if(batch<T, A> const& self, Mask const& mask, requires_arch<generic>) noexcept
{
return select(mask, incr(self), self);
}
// mul
template <class A, class T, class /*=typename std::enable_if<std::is_integral<T>::value, void>::type*/>
inline batch<T, A> mul(batch<T, A> const& self, batch<T, A> const& other, requires_arch<generic>) noexcept
{
return detail::apply([](T x, T y) noexcept -> T
{ return x * y; },
self, other);
}
// rotl
template <class A, class T, class STy>
inline batch<T, A> rotl(batch<T, A> const& self, STy other, requires_arch<generic>) noexcept
{
constexpr auto N = std::numeric_limits<T>::digits;
return (self << other) | (self >> (N - other));
}
// rotr
template <class A, class T, class STy>
inline batch<T, A> rotr(batch<T, A> const& self, STy other, requires_arch<generic>) noexcept
{
constexpr auto N = std::numeric_limits<T>::digits;
return (self >> other) | (self << (N - other));
}
// sadd
template <class A>
inline batch<float, A> sadd(batch<float, A> const& self, batch<float, A> const& other, requires_arch<generic>) noexcept
{
return add(self, other); // no saturated arithmetic on floating point numbers
}
template <class A, class T, class /*=typename std::enable_if<std::is_integral<T>::value, void>::type*/>
inline batch<T, A> sadd(batch<T, A> const& self, batch<T, A> const& other, requires_arch<generic>) noexcept
{
if (std::is_signed<T>::value)
{
auto mask = (other >> (8 * sizeof(T) - 1));
auto self_pos_branch = min(std::numeric_limits<T>::max() - other, self);
auto self_neg_branch = max(std::numeric_limits<T>::min() - other, self);
return other + select(batch_bool<T, A>(mask.data), self_neg_branch, self_pos_branch);
}
else
{
const auto diffmax = std::numeric_limits<T>::max() - self;
const auto mindiff = min(diffmax, other);
return self + mindiff;
}
}
template <class A>
inline batch<double, A> sadd(batch<double, A> const& self, batch<double, A> const& other, requires_arch<generic>) noexcept
{
return add(self, other); // no saturated arithmetic on floating point numbers
}
// ssub
template <class A>
inline batch<float, A> ssub(batch<float, A> const& self, batch<float, A> const& other, requires_arch<generic>) noexcept
{
return sub(self, other); // no saturated arithmetic on floating point numbers
}
template <class A, class T, class /*=typename std::enable_if<std::is_integral<T>::value, void>::type*/>
inline batch<T, A> ssub(batch<T, A> const& self, batch<T, A> const& other, requires_arch<generic>) noexcept
{
if (std::is_signed<T>::value)
{
return sadd(self, -other);
}
else
{
const auto diff = min(self, other);
return self - diff;
}
}
template <class A>
inline batch<double, A> ssub(batch<double, A> const& self, batch<double, A> const& other, requires_arch<generic>) noexcept
{
return sub(self, other); // no saturated arithmetic on floating point numbers
}
}
}
#endif
|
