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
270
271
272
273
274
|
/*
* Copyright (C) 2020 Linux Studio Plugins Project <https://lsp-plug.in/>
* (C) 2020 Vladimir Sadovnikov <sadko4u@gmail.com>
*
* This file is part of lsp-dsp-lib
* Created on: 31 мар. 2020 г.
*
* lsp-dsp-lib is free software: you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* any later version.
*
* lsp-dsp-lib 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 Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with lsp-dsp-lib. If not, see <https://www.gnu.org/licenses/>.
*/
#ifndef PRIVATE_DSP_ARCH_X86_AVX_IMPL
#error "This header should not be included directly"
#endif /* PRIVATE_DSP_ARCH_X86_AVX_IMPL */
#include <private/dsp/arch/x86/avx/fft/const.h>
#include <private/dsp/arch/x86/avx/fft/butterfly.h>
#include <private/dsp/arch/x86/avx/fft/normalize.h>
#define FFT_SCRAMBLE_SELF_DIRECT_NAME scramble_self_direct8
#define FFT_SCRAMBLE_SELF_REVERSE_NAME scramble_self_reverse8
#define FFT_SCRAMBLE_COPY_DIRECT_NAME scramble_copy_direct8
#define FFT_SCRAMBLE_COPY_REVERSE_NAME scramble_copy_reverse8
#define FFT_TYPE uint8_t
#define FFT_FMA(a, b) a
#include <private/dsp/arch/x86/avx/fft/scramble.h>
#define FFT_SCRAMBLE_SELF_DIRECT_NAME scramble_self_direct16
#define FFT_SCRAMBLE_SELF_REVERSE_NAME scramble_self_reverse16
#define FFT_SCRAMBLE_COPY_DIRECT_NAME scramble_copy_direct16
#define FFT_SCRAMBLE_COPY_REVERSE_NAME scramble_copy_reverse16
#define FFT_TYPE uint16_t
#define FFT_FMA(a, b) a
#include <private/dsp/arch/x86/avx/fft/scramble.h>
#define FFT_SCRAMBLE_SELF_DIRECT_NAME scramble_self_direct8_fma3
#define FFT_SCRAMBLE_SELF_REVERSE_NAME scramble_self_reverse8_fma3
#define FFT_SCRAMBLE_COPY_DIRECT_NAME scramble_copy_direct8_fma3
#define FFT_SCRAMBLE_COPY_REVERSE_NAME scramble_copy_reverse8_fma3
#define FFT_TYPE uint8_t
#define FFT_FMA(a, b) b
#include <private/dsp/arch/x86/avx/fft/scramble.h>
#define FFT_SCRAMBLE_SELF_DIRECT_NAME scramble_self_direct16_fma3
#define FFT_SCRAMBLE_SELF_REVERSE_NAME scramble_self_reverse16_fma3
#define FFT_SCRAMBLE_COPY_DIRECT_NAME scramble_copy_direct16_fma3
#define FFT_SCRAMBLE_COPY_REVERSE_NAME scramble_copy_reverse16_fma3
#define FFT_TYPE uint16_t
#define FFT_FMA(a, b) b
#include <private/dsp/arch/x86/avx/fft/scramble.h>
namespace lsp
{
namespace avx
{
static inline void small_direct_fft(float *dst_re, float *dst_im, const float *src_re, const float *src_im, size_t rank)
{
if (rank == 2)
{
float s0_re = src_re[0] + src_re[1];
float s1_re = src_re[0] - src_re[1];
float s2_re = src_re[2] + src_re[3];
float s3_re = src_re[2] - src_re[3];
float s0_im = src_im[0] + src_im[1];
float s1_im = src_im[0] - src_im[1];
float s2_im = src_im[2] + src_im[3];
float s3_im = src_im[2] - src_im[3];
dst_re[0] = s0_re + s2_re;
dst_re[1] = s1_re + s3_im;
dst_re[2] = s0_re - s2_re;
dst_re[3] = s1_re - s3_im;
dst_im[0] = s0_im + s2_im;
dst_im[1] = s1_im - s3_re;
dst_im[2] = s0_im - s2_im;
dst_im[3] = s1_im + s3_re;
}
else if (rank == 1)
{
// s0' = s0 + s1
// s1' = s0 - s1
float s1_re = src_re[1];
float s1_im = src_im[1];
dst_re[1] = src_re[0] - s1_re;
dst_im[1] = src_im[0] - s1_im;
dst_re[0] = src_re[0] + s1_re;
dst_im[0] = src_im[0] + s1_im;
}
else
{
dst_re[0] = src_re[0];
dst_im[0] = src_im[0];
}
}
static inline void small_reverse_fft(float *dst_re, float *dst_im, const float *src_re, const float *src_im, size_t rank)
{
if (rank == 2)
{
float s0_re = src_re[0] + src_re[1];
float s1_re = src_re[0] - src_re[1];
float s2_re = src_re[2] + src_re[3];
float s3_re = src_re[2] - src_re[3];
float s0_im = src_im[0] + src_im[1];
float s1_im = src_im[0] - src_im[1];
float s2_im = src_im[2] + src_im[3];
float s3_im = src_im[2] - src_im[3];
dst_re[0] = s0_re + s2_re;
dst_re[1] = s1_re + s3_im;
dst_re[2] = s0_re - s2_re;
dst_re[3] = s1_re - s3_im;
dst_im[0] = s0_im + s2_im;
dst_im[1] = s1_im - s3_re;
dst_im[2] = s0_im - s2_im;
dst_im[3] = s1_im + s3_re;
}
else if (rank == 1)
{
// s0' = s0 + s1
// s1' = s0 - s1
float s1_re = src_re[1];
float s1_im = src_im[1];
dst_re[1] = src_re[0] - s1_re;
dst_im[1] = src_im[0] - s1_im;
dst_re[0] = src_re[0] + s1_re;
dst_im[0] = src_im[0] + s1_im;
}
else
{
dst_re[0] = src_re[0];
dst_im[0] = src_im[0];
}
}
void direct_fft(float *dst_re, float *dst_im, const float *src_re, const float *src_im, size_t rank)
{
// Check bounds
if (rank <= 2)
{
small_direct_fft(dst_re, dst_im, src_re, src_im, rank);
return;
}
if ((dst_re == src_re) || (dst_im == src_im) || (rank < 4))
{
dsp::move(dst_re, src_re, 1 << rank);
dsp::move(dst_im, src_im, 1 << rank);
if (rank <= 8)
scramble_self_direct8(dst_re, dst_im, rank);
else
scramble_self_direct16(dst_re, dst_im, rank);
}
else
{
if (rank <= 12)
scramble_copy_direct8(dst_re, dst_im, src_re, src_im, rank-4);
else
scramble_copy_direct16(dst_re, dst_im, src_re, src_im, rank-4);
}
for (size_t i=3; i < rank; ++i)
butterfly_direct8p(dst_re, dst_im, i, 1 << (rank - i - 1));
}
void direct_fft_fma3(float *dst_re, float *dst_im, const float *src_re, const float *src_im, size_t rank)
{
// Check bounds
if (rank <= 2)
{
small_direct_fft(dst_re, dst_im, src_re, src_im, rank);
return;
}
if ((dst_re == src_re) || (dst_im == src_im) || (rank < 4))
{
dsp::move(dst_re, src_re, 1 << rank);
dsp::move(dst_im, src_im, 1 << rank);
if (rank <= 8)
scramble_self_direct8_fma3(dst_re, dst_im, rank);
else
scramble_self_direct16_fma3(dst_re, dst_im, rank);
}
else
{
if (rank <= 12)
scramble_copy_direct8_fma3(dst_re, dst_im, src_re, src_im, rank-4);
else
scramble_copy_direct16_fma3(dst_re, dst_im, src_re, src_im, rank-4);
}
for (size_t i=3; i < rank; ++i)
butterfly_direct8p_fma3(dst_re, dst_im, i, 1 << (rank - i - 1));
}
void reverse_fft(float *dst_re, float *dst_im, const float *src_re, const float *src_im, size_t rank)
{
// Check bounds
if (rank <= 2)
{
small_reverse_fft(dst_re, dst_im, src_re, src_im, rank);
return;
}
if ((dst_re == src_re) || (dst_im == src_im) || (rank < 4))
{
dsp::move(dst_re, src_re, 1 << rank);
dsp::move(dst_im, src_im, 1 << rank);
if (rank <= 8)
scramble_self_reverse8(dst_re, dst_im, rank);
else
scramble_self_reverse16(dst_re, dst_im, rank);
}
else
{
if (rank <= 12)
scramble_copy_reverse8(dst_re, dst_im, src_re, src_im, rank-4);
else
scramble_copy_reverse16(dst_re, dst_im, src_re, src_im, rank-4);
}
for (size_t i=3; i < rank; ++i)
butterfly_reverse8p(dst_re, dst_im, i, 1 << (rank - i - 1));
dsp::normalize_fft2(dst_re, dst_im, rank);
}
void reverse_fft_fma3(float *dst_re, float *dst_im, const float *src_re, const float *src_im, size_t rank)
{
// Check bounds
if (rank <= 2)
{
small_reverse_fft(dst_re, dst_im, src_re, src_im, rank);
return;
}
if ((dst_re == src_re) || (dst_im == src_im) || (rank < 4))
{
dsp::move(dst_re, src_re, 1 << rank);
dsp::move(dst_im, src_im, 1 << rank);
if (rank <= 8)
scramble_self_reverse8_fma3(dst_re, dst_im, rank);
else
scramble_self_reverse16_fma3(dst_re, dst_im, rank);
}
else
{
if (rank <= 12)
scramble_copy_reverse8_fma3(dst_re, dst_im, src_re, src_im, rank-4);
else
scramble_copy_reverse16_fma3(dst_re, dst_im, src_re, src_im, rank-4);
}
for (size_t i=3; i < rank; ++i)
butterfly_reverse8p_fma3(dst_re, dst_im, i, 1 << (rank - i - 1));
dsp::normalize_fft2(dst_re, dst_im, rank);
}
}
}
|
