1 |
|
|
/* |
2 |
|
|
* AAC Spectral Band Replication decoding functions |
3 |
|
|
* Copyright (c) 2008-2009 Robert Swain ( rob opendot cl ) |
4 |
|
|
* Copyright (c) 2009-2010 Alex Converse <alex.converse@gmail.com> |
5 |
|
|
* |
6 |
|
|
* This file is part of FFmpeg. |
7 |
|
|
* |
8 |
|
|
* FFmpeg is free software; you can redistribute it and/or |
9 |
|
|
* modify it under the terms of the GNU Lesser General Public |
10 |
|
|
* License as published by the Free Software Foundation; either |
11 |
|
|
* version 2.1 of the License, or (at your option) any later version. |
12 |
|
|
* |
13 |
|
|
* FFmpeg is distributed in the hope that it will be useful, |
14 |
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
15 |
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
16 |
|
|
* Lesser General Public License for more details. |
17 |
|
|
* |
18 |
|
|
* You should have received a copy of the GNU Lesser General Public |
19 |
|
|
* License along with FFmpeg; if not, write to the Free Software |
20 |
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
21 |
|
|
*/ |
22 |
|
|
|
23 |
|
|
#define USE_FIXED 0 |
24 |
|
|
|
25 |
|
|
#include "aac.h" |
26 |
|
|
#include "config.h" |
27 |
|
|
#include "libavutil/attributes.h" |
28 |
|
|
#include "libavutil/intfloat.h" |
29 |
|
|
#include "sbrdsp.h" |
30 |
|
|
|
31 |
|
251188 |
static float sbr_sum_square_c(float (*x)[2], int n) |
32 |
|
|
{ |
33 |
|
251188 |
float sum0 = 0.0f, sum1 = 0.0f; |
34 |
|
|
int i; |
35 |
|
|
|
36 |
✓✓ |
3003652 |
for (i = 0; i < n; i += 2) |
37 |
|
|
{ |
38 |
|
2752464 |
sum0 += x[i + 0][0] * x[i + 0][0]; |
39 |
|
2752464 |
sum1 += x[i + 0][1] * x[i + 0][1]; |
40 |
|
2752464 |
sum0 += x[i + 1][0] * x[i + 1][0]; |
41 |
|
2752464 |
sum1 += x[i + 1][1] * x[i + 1][1]; |
42 |
|
|
} |
43 |
|
|
|
44 |
|
251188 |
return sum0 + sum1; |
45 |
|
|
} |
46 |
|
|
|
47 |
|
355459 |
static void sbr_neg_odd_64_c(float *x) |
48 |
|
|
{ |
49 |
|
355459 |
union av_intfloat32 *xi = (union av_intfloat32*) x; |
50 |
|
|
int i; |
51 |
✓✓ |
6042803 |
for (i = 1; i < 64; i += 4) { |
52 |
|
5687344 |
xi[i + 0].i ^= 1U << 31; |
53 |
|
5687344 |
xi[i + 2].i ^= 1U << 31; |
54 |
|
|
} |
55 |
|
355459 |
} |
56 |
|
|
|
57 |
|
365091 |
static void sbr_qmf_pre_shuffle_c(float *z) |
58 |
|
|
{ |
59 |
|
365091 |
union av_intfloat32 *zi = (union av_intfloat32*) z; |
60 |
|
|
int k; |
61 |
|
365091 |
zi[64].i = zi[0].i; |
62 |
|
365091 |
zi[65].i = zi[1].i; |
63 |
✓✓ |
5841456 |
for (k = 1; k < 31; k += 2) { |
64 |
|
5476365 |
zi[64 + 2 * k + 0].i = zi[64 - k].i ^ (1U << 31); |
65 |
|
5476365 |
zi[64 + 2 * k + 1].i = zi[ k + 1].i; |
66 |
|
5476365 |
zi[64 + 2 * k + 2].i = zi[63 - k].i ^ (1U << 31); |
67 |
|
5476365 |
zi[64 + 2 * k + 3].i = zi[ k + 2].i; |
68 |
|
|
} |
69 |
|
|
|
70 |
|
365091 |
zi[64 + 2 * 31 + 0].i = zi[64 - 31].i ^ (1U << 31); |
71 |
|
365091 |
zi[64 + 2 * 31 + 1].i = zi[31 + 1].i; |
72 |
|
365091 |
} |
73 |
|
|
|
74 |
|
365091 |
static void sbr_qmf_post_shuffle_c(float W[32][2], const float *z) |
75 |
|
|
{ |
76 |
|
365091 |
const union av_intfloat32 *zi = (const union av_intfloat32*) z; |
77 |
|
365091 |
union av_intfloat32 *Wi = (union av_intfloat32*) W; |
78 |
|
|
int k; |
79 |
✓✓ |
6206547 |
for (k = 0; k < 32; k += 2) { |
80 |
|
5841456 |
Wi[2 * k + 0].i = zi[63 - k].i ^ (1U << 31); |
81 |
|
5841456 |
Wi[2 * k + 1].i = zi[ k + 0].i; |
82 |
|
5841456 |
Wi[2 * k + 2].i = zi[62 - k].i ^ (1U << 31); |
83 |
|
5841456 |
Wi[2 * k + 3].i = zi[ k + 1].i; |
84 |
|
|
} |
85 |
|
365091 |
} |
86 |
|
|
|
87 |
|
73603 |
static void sbr_qmf_deint_neg_c(float *v, const float *src) |
88 |
|
|
{ |
89 |
|
73603 |
const union av_intfloat32 *si = (const union av_intfloat32*)src; |
90 |
|
73603 |
union av_intfloat32 *vi = (union av_intfloat32*)v; |
91 |
|
|
int i; |
92 |
✓✓ |
2428899 |
for (i = 0; i < 32; i++) { |
93 |
|
2355296 |
vi[ i].i = si[63 - 2 * i ].i; |
94 |
|
2355296 |
vi[63 - i].i = si[63 - 2 * i - 1].i ^ (1U << 31); |
95 |
|
|
} |
96 |
|
73603 |
} |
97 |
|
|
|
98 |
|
|
#if 0 |
99 |
|
|
/* This code is slower because it multiplies memory accesses. |
100 |
|
|
* It is left for educational purposes and because it may offer |
101 |
|
|
* a better reference for writing arch-specific DSP functions. */ |
102 |
|
|
static av_always_inline void autocorrelate(const float x[40][2], |
103 |
|
|
float phi[3][2][2], int lag) |
104 |
|
|
{ |
105 |
|
|
int i; |
106 |
|
|
float real_sum = 0.0f; |
107 |
|
|
float imag_sum = 0.0f; |
108 |
|
|
if (lag) { |
109 |
|
|
for (i = 1; i < 38; i++) { |
110 |
|
|
real_sum += x[i][0] * x[i+lag][0] + x[i][1] * x[i+lag][1]; |
111 |
|
|
imag_sum += x[i][0] * x[i+lag][1] - x[i][1] * x[i+lag][0]; |
112 |
|
|
} |
113 |
|
|
phi[2-lag][1][0] = real_sum + x[ 0][0] * x[lag][0] + x[ 0][1] * x[lag][1]; |
114 |
|
|
phi[2-lag][1][1] = imag_sum + x[ 0][0] * x[lag][1] - x[ 0][1] * x[lag][0]; |
115 |
|
|
if (lag == 1) { |
116 |
|
|
phi[0][0][0] = real_sum + x[38][0] * x[39][0] + x[38][1] * x[39][1]; |
117 |
|
|
phi[0][0][1] = imag_sum + x[38][0] * x[39][1] - x[38][1] * x[39][0]; |
118 |
|
|
} |
119 |
|
|
} else { |
120 |
|
|
for (i = 1; i < 38; i++) { |
121 |
|
|
real_sum += x[i][0] * x[i][0] + x[i][1] * x[i][1]; |
122 |
|
|
} |
123 |
|
|
phi[2][1][0] = real_sum + x[ 0][0] * x[ 0][0] + x[ 0][1] * x[ 0][1]; |
124 |
|
|
phi[1][0][0] = real_sum + x[38][0] * x[38][0] + x[38][1] * x[38][1]; |
125 |
|
|
} |
126 |
|
|
} |
127 |
|
|
|
128 |
|
|
static void sbr_autocorrelate_c(const float x[40][2], float phi[3][2][2]) |
129 |
|
|
{ |
130 |
|
|
autocorrelate(x, phi, 0); |
131 |
|
|
autocorrelate(x, phi, 1); |
132 |
|
|
autocorrelate(x, phi, 2); |
133 |
|
|
} |
134 |
|
|
#else |
135 |
|
124327 |
static void sbr_autocorrelate_c(const float x[40][2], float phi[3][2][2]) |
136 |
|
|
{ |
137 |
|
124327 |
float real_sum2 = x[0][0] * x[2][0] + x[0][1] * x[2][1]; |
138 |
|
124327 |
float imag_sum2 = x[0][0] * x[2][1] - x[0][1] * x[2][0]; |
139 |
|
124327 |
float real_sum1 = 0.0f, imag_sum1 = 0.0f, real_sum0 = 0.0f; |
140 |
|
|
int i; |
141 |
✓✓ |
4724426 |
for (i = 1; i < 38; i++) { |
142 |
|
4600099 |
real_sum0 += x[i][0] * x[i ][0] + x[i][1] * x[i ][1]; |
143 |
|
4600099 |
real_sum1 += x[i][0] * x[i + 1][0] + x[i][1] * x[i + 1][1]; |
144 |
|
4600099 |
imag_sum1 += x[i][0] * x[i + 1][1] - x[i][1] * x[i + 1][0]; |
145 |
|
4600099 |
real_sum2 += x[i][0] * x[i + 2][0] + x[i][1] * x[i + 2][1]; |
146 |
|
4600099 |
imag_sum2 += x[i][0] * x[i + 2][1] - x[i][1] * x[i + 2][0]; |
147 |
|
|
} |
148 |
|
124327 |
phi[2 - 2][1][0] = real_sum2; |
149 |
|
124327 |
phi[2 - 2][1][1] = imag_sum2; |
150 |
|
124327 |
phi[2 ][1][0] = real_sum0 + x[ 0][0] * x[ 0][0] + x[ 0][1] * x[ 0][1]; |
151 |
|
124327 |
phi[1 ][0][0] = real_sum0 + x[38][0] * x[38][0] + x[38][1] * x[38][1]; |
152 |
|
124327 |
phi[2 - 1][1][0] = real_sum1 + x[ 0][0] * x[ 1][0] + x[ 0][1] * x[ 1][1]; |
153 |
|
124327 |
phi[2 - 1][1][1] = imag_sum1 + x[ 0][0] * x[ 1][1] - x[ 0][1] * x[ 1][0]; |
154 |
|
124327 |
phi[0 ][0][0] = real_sum1 + x[38][0] * x[39][0] + x[38][1] * x[39][1]; |
155 |
|
124327 |
phi[0 ][0][1] = imag_sum1 + x[38][0] * x[39][1] - x[38][1] * x[39][0]; |
156 |
|
124327 |
} |
157 |
|
|
#endif |
158 |
|
|
|
159 |
|
172098 |
static void sbr_hf_gen_c(float (*X_high)[2], const float (*X_low)[2], |
160 |
|
|
const float alpha0[2], const float alpha1[2], |
161 |
|
|
float bw, int start, int end) |
162 |
|
|
{ |
163 |
|
|
float alpha[4]; |
164 |
|
|
int i; |
165 |
|
|
|
166 |
|
172098 |
alpha[0] = alpha1[0] * bw * bw; |
167 |
|
172098 |
alpha[1] = alpha1[1] * bw * bw; |
168 |
|
172098 |
alpha[2] = alpha0[0] * bw; |
169 |
|
172098 |
alpha[3] = alpha0[1] * bw; |
170 |
|
|
|
171 |
✓✓ |
5685186 |
for (i = start; i < end; i++) { |
172 |
|
5513088 |
X_high[i][0] = |
173 |
|
5513088 |
X_low[i - 2][0] * alpha[0] - |
174 |
|
5513088 |
X_low[i - 2][1] * alpha[1] + |
175 |
|
5513088 |
X_low[i - 1][0] * alpha[2] - |
176 |
|
5513088 |
X_low[i - 1][1] * alpha[3] + |
177 |
|
5513088 |
X_low[i][0]; |
178 |
|
5513088 |
X_high[i][1] = |
179 |
|
5513088 |
X_low[i - 2][1] * alpha[0] + |
180 |
|
5513088 |
X_low[i - 2][0] * alpha[1] + |
181 |
|
5513088 |
X_low[i - 1][1] * alpha[2] + |
182 |
|
5513088 |
X_low[i - 1][0] * alpha[3] + |
183 |
|
5513088 |
X_low[i][1]; |
184 |
|
|
} |
185 |
|
172098 |
} |
186 |
|
|
|
187 |
|
212195 |
static void sbr_hf_g_filt_c(float (*Y)[2], const float (*X_high)[40][2], |
188 |
|
|
const float *g_filt, int m_max, intptr_t ixh) |
189 |
|
|
{ |
190 |
|
|
int m; |
191 |
|
|
|
192 |
✓✓ |
5716739 |
for (m = 0; m < m_max; m++) { |
193 |
|
5504544 |
Y[m][0] = X_high[m][ixh][0] * g_filt[m]; |
194 |
|
5504544 |
Y[m][1] = X_high[m][ixh][1] * g_filt[m]; |
195 |
|
|
} |
196 |
|
212195 |
} |
197 |
|
|
|
198 |
|
209994 |
static av_always_inline void sbr_hf_apply_noise(float (*Y)[2], |
199 |
|
|
const float *s_m, |
200 |
|
|
const float *q_filt, |
201 |
|
|
int noise, |
202 |
|
|
float phi_sign0, |
203 |
|
|
float phi_sign1, |
204 |
|
|
int m_max) |
205 |
|
|
{ |
206 |
|
|
int m; |
207 |
|
|
|
208 |
✓✓ |
5659882 |
for (m = 0; m < m_max; m++) { |
209 |
|
5449888 |
float y0 = Y[m][0]; |
210 |
|
5449888 |
float y1 = Y[m][1]; |
211 |
|
5449888 |
noise = (noise + 1) & 0x1ff; |
212 |
✓✓ |
5449888 |
if (s_m[m]) { |
213 |
|
39374 |
y0 += s_m[m] * phi_sign0; |
214 |
|
39374 |
y1 += s_m[m] * phi_sign1; |
215 |
|
|
} else { |
216 |
|
5410514 |
y0 += q_filt[m] * ff_sbr_noise_table[noise][0]; |
217 |
|
5410514 |
y1 += q_filt[m] * ff_sbr_noise_table[noise][1]; |
218 |
|
|
} |
219 |
|
5449888 |
Y[m][0] = y0; |
220 |
|
5449888 |
Y[m][1] = y1; |
221 |
|
5449888 |
phi_sign1 = -phi_sign1; |
222 |
|
|
} |
223 |
|
209994 |
} |
224 |
|
|
|
225 |
|
|
#include "sbrdsp_template.c" |