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
|
// softclip
// Copyright (C) 2008, 2009 Tim Blechmann
//
// 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 2 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; see the file COPYING. If not, write to
// the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
// Boston, MA 02111-1307, USA.
#ifndef SIMD_SOFTCLIP_HPP
#define SIMD_SOFTCLIP_HPP
#include <cassert>
#include <cmath>
#include "vec.hpp"
#include "detail/unroll_helpers.hpp"
#if defined(__GNUC__) && defined(NDEBUG)
#define always_inline inline __attribute__((always_inline))
#else
#define always_inline inline
#endif
namespace nova
{
namespace detail
{
template <typename FloatType>
struct softclip
{
always_inline FloatType operator()(FloatType arg) const
{
FloatType abs = fabs(arg);
if (abs < FloatType(0.5))
return arg;
else
return (abs - FloatType(0.25)) / arg;
}
};
#ifdef SSE
/* this computes both parts of the branch
*
* benchmarks (core2) showed:
* 6.5 seconds for simdfied code
* 5.3 seconds for non-simd code for samples with abs(sample) < 0.5
* 17 seconds when 50% of the samples have abs(sample) < 0.5
* 26 seconds for samples with abs(sample) > 0.5
*
* on intel nethalem cpus, the simdfied code is faster than all non-simd code
* */
template <>
struct softclip<vec<float> >
{
typedef vec<float> vec_type;
always_inline vec_type operator()(vec_type arg) const
{
const vec_type const05 = vec_type::gen_05();
const vec_type const025 = vec_type::gen_025();
vec_type abs = abs(arg);
vec_type selecter = mask_lt(abs, const05);
vec_type alt_ret = (abs - const025) / arg;
return select(alt_ret, arg, selecter);
}
};
#endif
}
template <typename float_type>
inline void softclip_vec(float_type * out, const float_type * in, unsigned int n)
{
detail::apply_on_vector(out, in, n, detail::softclip<float>());
}
template <typename float_type>
inline void softclip_vec_simd(float_type * out, const float_type * in, unsigned int n)
{
softclip_vec(out, in, n);
}
template <unsigned int n, typename float_type>
inline void softclip_vec_simd(float_type * out, const float_type * in)
{
softclip_vec_simd(out, in, n);
}
#ifdef SSE
template <>
inline void softclip_vec_simd<float>(float_type * out, const float_type * in, unsigned int n)
{
detail::generate_simd_loop(out, in, n, detail::softclip<vec<float> >());
}
#endif
} /* namespace nova */
#undef always_inline
#endif /* SIMD_SOFTCLIP_HPP */
|
