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
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
|
/*
==============================================================================
This file is part of the JUCE library.
Copyright (c) 2017 - ROLI Ltd.
JUCE is an open source library subject to commercial or open-source
licensing.
By using JUCE, you agree to the terms of both the JUCE 5 End-User License
Agreement and JUCE 5 Privacy Policy (both updated and effective as of the
27th April 2017).
End User License Agreement: www.juce.com/juce-5-licence
Privacy Policy: www.juce.com/juce-5-privacy-policy
Or: You may also use this code under the terms of the GPL v3 (see
www.gnu.org/licenses).
JUCE IS PROVIDED "AS IS" WITHOUT ANY WARRANTY, AND ALL WARRANTIES, WHETHER
EXPRESSED OR IMPLIED, INCLUDING MERCHANTABILITY AND FITNESS FOR PURPOSE, ARE
DISCLAIMED.
==============================================================================
*/
namespace juce
{
namespace dsp
{
/** A template specialisation to find corresponding mask type for primitives. */
namespace SIMDInternal
{
template <typename Primitive> struct MaskTypeFor { using type = Primitive; };
template <> struct MaskTypeFor <float> { using type = uint32_t; };
template <> struct MaskTypeFor <double> { using type = uint64_t; };
template <> struct MaskTypeFor <char> { using type = uint8_t; };
template <> struct MaskTypeFor <int8_t> { using type = uint8_t; };
template <> struct MaskTypeFor <int16_t> { using type = uint16_t; };
template <> struct MaskTypeFor <int32_t> { using type = uint32_t; };
template <> struct MaskTypeFor <int64_t> { using type = uint64_t; };
template <> struct MaskTypeFor <std::complex<float>> { using type = uint32_t; };
template <> struct MaskTypeFor <std::complex<double>> { using type = uint64_t; };
template <typename Primitive> struct PrimitiveType { using type = typename std::remove_cv<Primitive>::type; };
template <typename Primitive> struct PrimitiveType<std::complex<Primitive>> { using type = typename std::remove_cv<Primitive>::type; };
template <int n> struct Log2Helper { enum { value = Log2Helper<n/2>::value + 1 }; };
template <> struct Log2Helper<1> { enum { value = 0 }; };
}
/**
Useful fallback routines to use if the native SIMD op is not supported. You
should never need to use this directly. Use juce_SIMDRegister instead.
@tags{DSP}
*/
template <typename ScalarType, typename vSIMDType>
struct SIMDFallbackOps
{
static constexpr size_t n = sizeof (vSIMDType) / sizeof (ScalarType);
static constexpr size_t mask = (sizeof (vSIMDType) / sizeof (ScalarType)) - 1;
static constexpr size_t bits = SIMDInternal::Log2Helper<(int) n>::value;
// helper types
using MaskType = typename SIMDInternal::MaskTypeFor<ScalarType>::type;
union UnionType { vSIMDType v; ScalarType s[n]; };
union UnionMaskType { vSIMDType v; MaskType m[n]; };
// fallback methods
static forcedinline vSIMDType add (vSIMDType a, vSIMDType b) noexcept { return apply<ScalarAdd> (a, b); }
static forcedinline vSIMDType sub (vSIMDType a, vSIMDType b) noexcept { return apply<ScalarSub> (a, b); }
static forcedinline vSIMDType mul (vSIMDType a, vSIMDType b) noexcept { return apply<ScalarMul> (a, b); }
static forcedinline vSIMDType bit_and (vSIMDType a, vSIMDType b) noexcept { return bitapply<ScalarAnd> (a, b); }
static forcedinline vSIMDType bit_or (vSIMDType a, vSIMDType b) noexcept { return bitapply<ScalarOr > (a, b); }
static forcedinline vSIMDType bit_xor (vSIMDType a, vSIMDType b) noexcept { return bitapply<ScalarXor> (a, b); }
static forcedinline vSIMDType bit_notand (vSIMDType a, vSIMDType b) noexcept { return bitapply<ScalarNot> (a, b); }
static forcedinline vSIMDType min (vSIMDType a, vSIMDType b) noexcept { return apply<ScalarMin> (a, b); }
static forcedinline vSIMDType max (vSIMDType a, vSIMDType b) noexcept { return apply<ScalarMax> (a, b); }
static forcedinline vSIMDType equal (vSIMDType a, vSIMDType b) noexcept { return cmp<ScalarEq > (a, b); }
static forcedinline vSIMDType notEqual (vSIMDType a, vSIMDType b) noexcept { return cmp<ScalarNeq> (a, b); }
static forcedinline vSIMDType greaterThan (vSIMDType a, vSIMDType b) noexcept { return cmp<ScalarGt > (a, b); }
static forcedinline vSIMDType greaterThanOrEqual (vSIMDType a, vSIMDType b) noexcept { return cmp<ScalarGeq> (a, b); }
static forcedinline ScalarType get (vSIMDType v, size_t i) noexcept
{
UnionType u {v};
return u.s[i];
}
static forcedinline vSIMDType set (vSIMDType v, size_t i, ScalarType s) noexcept
{
UnionType u {v};
u.s[i] = s;
return u.v;
}
static forcedinline vSIMDType bit_not (vSIMDType av) noexcept
{
UnionMaskType a {av};
for (size_t i = 0; i < n; ++i)
a.m[i] = ~a.m[i];
return a.v;
}
static forcedinline ScalarType sum (vSIMDType av) noexcept
{
UnionType a {av};
auto retval = static_cast<ScalarType> (0);
for (size_t i = 0; i < n; ++i)
retval += a.s[i];
return retval;
}
static forcedinline vSIMDType truncate (vSIMDType av) noexcept
{
UnionType a {av};
for (size_t i = 0; i < n; ++i)
{
jassert (a.s[i] >= ScalarType (0));
a.s[i] = static_cast <ScalarType> (static_cast<int> (a.s[i]));
}
return a.v;
}
static forcedinline vSIMDType multiplyAdd (vSIMDType av, vSIMDType bv, vSIMDType cv) noexcept
{
UnionType a {av}, b {bv}, c {cv};
for (size_t i = 0; i < n; ++i)
a.s[i] += b.s[i] * c.s[i];
return a.v;
}
//==============================================================================
static forcedinline bool allEqual (vSIMDType av, vSIMDType bv) noexcept
{
UnionType a {av}, b {bv};
for (size_t i = 0; i < n; ++i)
if (a.s[i] != b.s[i])
return false;
return true;
}
//==============================================================================
static forcedinline vSIMDType cmplxmul (vSIMDType av, vSIMDType bv) noexcept
{
UnionType a {av}, b {bv}, r;
const int m = n >> 1;
for (int i = 0; i < m; ++i)
{
std::complex<ScalarType> result
= std::complex<ScalarType> (a.s[i<<1], a.s[(i<<1)|1])
* std::complex<ScalarType> (b.s[i<<1], b.s[(i<<1)|1]);
r.s[i<<1] = result.real();
r.s[(i<<1)|1] = result.imag();
}
return r.v;
}
struct ScalarAdd { static forcedinline ScalarType op (ScalarType a, ScalarType b) noexcept { return a + b; } };
struct ScalarSub { static forcedinline ScalarType op (ScalarType a, ScalarType b) noexcept { return a - b; } };
struct ScalarMul { static forcedinline ScalarType op (ScalarType a, ScalarType b) noexcept { return a * b; } };
struct ScalarMin { static forcedinline ScalarType op (ScalarType a, ScalarType b) noexcept { return jmin (a, b); } };
struct ScalarMax { static forcedinline ScalarType op (ScalarType a, ScalarType b) noexcept { return jmax (a, b); } };
struct ScalarAnd { static forcedinline MaskType op (MaskType a, MaskType b) noexcept { return a & b; } };
struct ScalarOr { static forcedinline MaskType op (MaskType a, MaskType b) noexcept { return a | b; } };
struct ScalarXor { static forcedinline MaskType op (MaskType a, MaskType b) noexcept { return a ^ b; } };
struct ScalarNot { static forcedinline MaskType op (MaskType a, MaskType b) noexcept { return (~a) & b; } };
struct ScalarEq { static forcedinline bool op (ScalarType a, ScalarType b) noexcept { return (a == b); } };
struct ScalarNeq { static forcedinline bool op (ScalarType a, ScalarType b) noexcept { return (a != b); } };
struct ScalarGt { static forcedinline bool op (ScalarType a, ScalarType b) noexcept { return (a > b); } };
struct ScalarGeq { static forcedinline bool op (ScalarType a, ScalarType b) noexcept { return (a >= b); } };
// generic apply routines for operations above
template <typename Op>
static forcedinline vSIMDType apply (vSIMDType av, vSIMDType bv) noexcept
{
UnionType a {av}, b {bv};
for (size_t i = 0; i < n; ++i)
a.s[i] = Op::op (a.s[i], b.s[i]);
return a.v;
}
template <typename Op>
static forcedinline vSIMDType cmp (vSIMDType av, vSIMDType bv) noexcept
{
UnionType a {av}, b {bv};
UnionMaskType r;
for (size_t i = 0; i < n; ++i)
r.m[i] = Op::op (a.s[i], b.s[i]) ? static_cast<MaskType> (-1) : static_cast<MaskType> (0);
return r.v;
}
template <typename Op>
static forcedinline vSIMDType bitapply (vSIMDType av, vSIMDType bv) noexcept
{
UnionMaskType a {av}, b {bv};
for (size_t i = 0; i < n; ++i)
a.m[i] = Op::op (a.m[i], b.m[i]);
return a.v;
}
static forcedinline vSIMDType expand (ScalarType s) noexcept
{
UnionType r;
for (size_t i = 0; i < n; ++i)
r.s[i] = s;
return r.v;
}
static forcedinline vSIMDType load (const ScalarType* a) noexcept
{
UnionType r;
for (size_t i = 0; i < n; ++i)
r.s[i] = a[i];
return r.v;
}
static forcedinline void store (vSIMDType av, ScalarType* dest) noexcept
{
UnionType a {av};
for (size_t i = 0; i < n; ++i)
dest[i] = a.s[i];
}
template <unsigned int shuffle_idx>
static forcedinline vSIMDType shuffle (vSIMDType av) noexcept
{
UnionType a {av}, r;
// the compiler will unroll this loop and the index can
// be computed at compile-time, so this will be super fast
for (size_t i = 0; i < n; ++i)
r.s[i] = a.s[(shuffle_idx >> (bits * i)) & mask];
return r.v;
}
};
} // namespace dsp
} // namespace juce
|
