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
|
/***************************************************************************
* 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_SSE3_HPP
#define XSIMD_SSE3_HPP
#include "../types/xsimd_sse3_register.hpp"
#include <type_traits>
namespace xsimd
{
namespace kernel
{
using namespace types;
// haddp
template <class A>
inline batch<float, A> haddp(batch<float, A> const* row, requires_arch<sse3>) noexcept
{
return _mm_hadd_ps(_mm_hadd_ps(row[0], row[1]),
_mm_hadd_ps(row[2], row[3]));
}
template <class A>
inline batch<double, A> haddp(batch<double, A> const* row, requires_arch<sse3>) noexcept
{
return _mm_hadd_pd(row[0], row[1]);
}
// load_unaligned
template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
inline batch<T, A> load_unaligned(T const* mem, convert<T>, requires_arch<sse3>) noexcept
{
return _mm_lddqu_si128((__m128i const*)mem);
}
// reduce_add
template <class A>
inline float reduce_add(batch<float, A> const& self, requires_arch<sse3>) noexcept
{
__m128 tmp0 = _mm_hadd_ps(self, self);
__m128 tmp1 = _mm_hadd_ps(tmp0, tmp0);
return _mm_cvtss_f32(tmp1);
}
template <class A>
inline double reduce_add(batch<double, A> const& self, requires_arch<sse3>) noexcept
{
__m128d tmp0 = _mm_hadd_pd(self, self);
return _mm_cvtsd_f64(tmp0);
}
}
}
#endif
|
