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
|
#include "private/cpu.h"
#ifndef FLAC__INTEGER_ONLY_LIBRARY
#ifndef FLAC__NO_ASM
#if defined FLAC__CPU_AARCH64 && FLAC__HAS_NEONINTRIN
#include "private/lpc.h"
#include "FLAC/assert.h"
#include "FLAC/format.h"
#include <arm_neon.h>
inline float32x4_t shufffleVector(float32x4_t vec)
{
float32x2_t hi = vget_high_f32(vec);
float32x2_t lo = vget_low_f32(vec);
float32x2x2_t qr0_z = vzip_f32(hi, lo);
return vcombine_f32(qr0_z.val[0], qr0_z.val[1]);
}
inline float32x4_t shufffleVector_2103(float32x4_t vec)
{
float32_t *tempPtr;
float32_t temp = vgetq_lane_f32(vec, 0);
vec = vcopyq_laneq_f32(vec, 0, vec, 2);
tempPtr = &temp;
vst1q_lane_f32(tempPtr, vec, 2);
return vec;
}
void FLAC__lpc_compute_autocorrelation_intrin_neon_lag_4(const FLAC__real data[], uint32_t data_len, uint32_t lag, FLAC__real autoc[])
{
int i;
int limit = data_len - 4;
float32x4_t sum0 = vdupq_n_f32(0.0f);
(void)lag;
FLAC__ASSERT(lag <= 4);
FLAC__ASSERT(lag <= data_len);
for (i = 0; i <= limit; i++)
{
float32x4_t d, d0;
d0 = vld1q_f32(data + i);
d = shufffleVector(d0);
sum0 = vaddq_f32(sum0, vmulq_f32(d0, d));
}
{
float32x4_t d0 = vdupq_n_f32(0.0f);
limit++;
if (limit < 0)
limit = 0;
for (i = data_len - 1; i >= limit; i--)
{
float32x4_t d = vld1q_lane_f32(data + i, vdupq_n_f32(0.0f), 0);
d = shufffleVector(d);
d0 = shufffleVector_2103(d0);
d0 = vcopyq_laneq_f32(d0, 0, d, 0);
sum0 = vaddq_f32(sum0, vmulq_f32(d, d0));
}
}
vst1q_f32(autoc, sum0);
}
void FLAC__lpc_compute_autocorrelation_intrin_neon_lag_8(const FLAC__real data[], uint32_t data_len, uint32_t lag, FLAC__real autoc[])
{
int i;
int limit = data_len - 8;
float32x4_t sum0 = vdupq_n_f32(0.0f);
float32x4_t sum1 = vdupq_n_f32(0.0f);
(void)lag;
FLAC__ASSERT(lag <= 8);
FLAC__ASSERT(lag <= data_len);
for (i = 0; i <= limit; i++)
{
float32x4_t d, d0, d1;
d0 = vld1q_f32(data + i);
d1 = vld1q_f32(data + i + 4);
d = shufffleVector(d0);
sum0 = vaddq_f32(sum0, vmulq_f32(d0, d));
sum1 = vaddq_f32(sum1, vmulq_f32(d1, d));
}
{
float32x4_t d0 = vdupq_n_f32(0.0f);
float32x4_t d1 = vdupq_n_f32(0.0f);
limit++;
if (limit < 0)
limit = 0;
for (i = data_len - 1; i >= limit; i--)
{
float32x4_t d = vld1q_lane_f32(data + i, vdupq_n_f32(0.0f), 0);
d = shufffleVector(d);
d1 = shufffleVector_2103(d1);
d0 = shufffleVector_2103(d0);
d1 = vcopyq_laneq_f32(d1, 0, d0, 0);
d0 = vcopyq_laneq_f32(d0, 0, d, 0);
sum1 = vaddq_f32(sum1, vmulq_f32(d, d1));
sum0 = vaddq_f32(sum0, vmulq_f32(d, d0));
}
}
vst1q_f32(autoc, sum0);
vst1q_f32(autoc + 4, sum1);
}
void FLAC__lpc_compute_autocorrelation_intrin_neon_lag_12(const FLAC__real data[], uint32_t data_len, uint32_t lag, FLAC__real autoc[])
{
int i;
int limit = data_len - 12;
float32x4_t sum0 = vdupq_n_f32(0.0f);
float32x4_t sum1 = vdupq_n_f32(0.0f);
float32x4_t sum2 = vdupq_n_f32(0.0f);
(void)lag;
FLAC__ASSERT(lag <= 12);
FLAC__ASSERT(lag <= data_len);
for (i = 0; i <= limit; i++)
{
float32x4_t d, d0, d1, d2;
d0 = vld1q_f32(data + i);
d1 = vld1q_f32(data + i + 4);
d1 = vld1q_f32(data + i + 8);
d = shufffleVector(d0);
sum0 = vaddq_f32(sum0, vmulq_f32(d0, d));
sum1 = vaddq_f32(sum1, vmulq_f32(d1, d));
sum2 = vaddq_f32(sum2, vmulq_f32(d2, d));
}
{
float32x4_t d0 = vdupq_n_f32(0.0f);
float32x4_t d1 = vdupq_n_f32(0.0f);
float32x4_t d2 = vdupq_n_f32(0.0f);
limit++;
if (limit < 0)
limit = 0;
for (i = data_len - 1; i >= limit; i--)
{
float32x4_t d = vld1q_lane_f32(data + i, vdupq_n_f32(0.0f), 0);
d = shufffleVector(d);
d2 = shufffleVector_2103(d2);
d1 = shufffleVector_2103(d1);
d0 = shufffleVector_2103(d0);
d2 = vcopyq_laneq_f32(d2, 0, d1, 0);
d1 = vcopyq_laneq_f32(d1, 0, d0, 0);
d0 = vcopyq_laneq_f32(d0, 0, d, 0);
sum2 = vaddq_f32(sum2, vmulq_f32(d, d2));
sum1 = vaddq_f32(sum1, vmulq_f32(d, d1));
sum0 = vaddq_f32(sum0, vmulq_f32(d, d0));
}
}
vst1q_f32(autoc, sum0);
vst1q_f32(autoc + 4, sum1);
vst1q_f32(autoc + 8, sum2);
}
void FLAC__lpc_compute_autocorrelation_intrin_neon_lag_16(const FLAC__real data[], uint32_t data_len, uint32_t lag, FLAC__real autoc[])
{
int i;
int limit = data_len - 16;
float32x4_t sum0 = vdupq_n_f32(0.0f);
float32x4_t sum1 = vdupq_n_f32(0.0f);
float32x4_t sum2 = vdupq_n_f32(0.0f);
float32x4_t sum3 = vdupq_n_f32(0.0f);
(void)lag;
FLAC__ASSERT(lag <= 16);
FLAC__ASSERT(lag <= data_len);
for (i = 0; i <= limit; i++)
{
float32x4_t d, d0, d1, d2, d3;
d0 = vld1q_f32(data + i);
d1 = vld1q_f32(data + i + 4);
d1 = vld1q_f32(data + i + 8);
d1 = vld1q_f32(data + i + 12);
d = shufffleVector(d0);
sum0 = vaddq_f32(sum0, vmulq_f32(d0, d));
sum1 = vaddq_f32(sum1, vmulq_f32(d1, d));
sum2 = vaddq_f32(sum2, vmulq_f32(d2, d));
sum3 = vaddq_f32(sum3, vmulq_f32(d3, d));
}
{
float32x4_t d0 = vdupq_n_f32(0.0f);
float32x4_t d1 = vdupq_n_f32(0.0f);
float32x4_t d2 = vdupq_n_f32(0.0f);
float32x4_t d3 = vdupq_n_f32(0.0f);
limit++;
if (limit < 0)
limit = 0;
for (i = data_len - 1; i >= limit; i--)
{
float32x4_t d = vld1q_lane_f32(data + i, vdupq_n_f32(0.0f), 0);
d = shufffleVector(d);
d3 = shufffleVector_2103(d3);
d2 = shufffleVector_2103(d2);
d1 = shufffleVector_2103(d1);
d0 = shufffleVector_2103(d0);
d3 = vcopyq_laneq_f32(d3, 0, d2, 0);
d2 = vcopyq_laneq_f32(d2, 0, d1, 0);
d1 = vcopyq_laneq_f32(d1, 0, d0, 0);
d0 = vcopyq_laneq_f32(d0, 0, d, 0);
sum3 = vaddq_f32(sum3, vmulq_f32(d, d3));
sum2 = vaddq_f32(sum2, vmulq_f32(d, d2));
sum1 = vaddq_f32(sum1, vmulq_f32(d, d1));
sum0 = vaddq_f32(sum0, vmulq_f32(d, d0));
}
}
vst1q_f32(autoc, sum0);
vst1q_f32(autoc + 4, sum1);
vst1q_f32(autoc + 8, sum2);
vst1q_f32(autoc + 12, sum3);
}
#endif /* FLAC__CPU_AARCH64 && FLAC__HAS_ARCH64INTRIN */
#endif /* FLAC__NO_ASM */
#endif /* FLAC__INTEGER_ONLY_LIBRARY */
|
