1/*
2 * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21#include "config.h"
22#include "libavutil/attributes.h"
23#include "aacpsdsp.h"
24
25static void ps_add_squares_c(float *dst, const float (*src)[2], int n)
26{
27    int i;
28    for (i = 0; i < n; i++)
29        dst[i] += src[i][0] * src[i][0] + src[i][1] * src[i][1];
30}
31
32static void ps_mul_pair_single_c(float (*dst)[2], float (*src0)[2], float *src1,
33                                 int n)
34{
35    int i;
36    for (i = 0; i < n; i++) {
37        dst[i][0] = src0[i][0] * src1[i];
38        dst[i][1] = src0[i][1] * src1[i];
39    }
40}
41
42static void ps_hybrid_analysis_c(float (*out)[2], float (*in)[2],
43                                 const float (*filter)[8][2],
44                                 int stride, int n)
45{
46    int i, j;
47
48    for (i = 0; i < n; i++) {
49        float sum_re = filter[i][6][0] * in[6][0];
50        float sum_im = filter[i][6][0] * in[6][1];
51
52        for (j = 0; j < 6; j++) {
53            float in0_re = in[j][0];
54            float in0_im = in[j][1];
55            float in1_re = in[12-j][0];
56            float in1_im = in[12-j][1];
57            sum_re += filter[i][j][0] * (in0_re + in1_re) -
58                      filter[i][j][1] * (in0_im - in1_im);
59            sum_im += filter[i][j][0] * (in0_im + in1_im) +
60                      filter[i][j][1] * (in0_re - in1_re);
61        }
62        out[i * stride][0] = sum_re;
63        out[i * stride][1] = sum_im;
64    }
65}
66
67static void ps_hybrid_analysis_ileave_c(float (*out)[32][2], float L[2][38][64],
68                                        int i, int len)
69{
70    int j;
71
72    for (; i < 64; i++) {
73        for (j = 0; j < len; j++) {
74            out[i][j][0] = L[0][j][i];
75            out[i][j][1] = L[1][j][i];
76        }
77    }
78}
79
80static void ps_hybrid_synthesis_deint_c(float out[2][38][64],
81                                        float (*in)[32][2],
82                                        int i, int len)
83{
84    int n;
85
86    for (; i < 64; i++) {
87        for (n = 0; n < len; n++) {
88            out[0][n][i] = in[i][n][0];
89            out[1][n][i] = in[i][n][1];
90        }
91    }
92}
93
94static void ps_decorrelate_c(float (*out)[2], float (*delay)[2],
95                             float (*ap_delay)[PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2],
96                             const float phi_fract[2], const float (*Q_fract)[2],
97                             const float *transient_gain,
98                             float g_decay_slope,
99                             int len)
100{
101    static const float a[] = { 0.65143905753106f,
102                               0.56471812200776f,
103                               0.48954165955695f };
104    float ag[PS_AP_LINKS];
105    int m, n;
106
107    for (m = 0; m < PS_AP_LINKS; m++)
108        ag[m] = a[m] * g_decay_slope;
109
110    for (n = 0; n < len; n++) {
111        float in_re = delay[n][0] * phi_fract[0] - delay[n][1] * phi_fract[1];
112        float in_im = delay[n][0] * phi_fract[1] + delay[n][1] * phi_fract[0];
113        for (m = 0; m < PS_AP_LINKS; m++) {
114            float a_re                = ag[m] * in_re;
115            float a_im                = ag[m] * in_im;
116            float link_delay_re       = ap_delay[m][n+2-m][0];
117            float link_delay_im       = ap_delay[m][n+2-m][1];
118            float fractional_delay_re = Q_fract[m][0];
119            float fractional_delay_im = Q_fract[m][1];
120            float apd_re = in_re;
121            float apd_im = in_im;
122            in_re = link_delay_re * fractional_delay_re -
123                    link_delay_im * fractional_delay_im - a_re;
124            in_im = link_delay_re * fractional_delay_im +
125                    link_delay_im * fractional_delay_re - a_im;
126            ap_delay[m][n+5][0] = apd_re + ag[m] * in_re;
127            ap_delay[m][n+5][1] = apd_im + ag[m] * in_im;
128        }
129        out[n][0] = transient_gain[n] * in_re;
130        out[n][1] = transient_gain[n] * in_im;
131    }
132}
133
134static void ps_stereo_interpolate_c(float (*l)[2], float (*r)[2],
135                                    float h[2][4], float h_step[2][4],
136                                    int len)
137{
138    float h0 = h[0][0];
139    float h1 = h[0][1];
140    float h2 = h[0][2];
141    float h3 = h[0][3];
142    float hs0 = h_step[0][0];
143    float hs1 = h_step[0][1];
144    float hs2 = h_step[0][2];
145    float hs3 = h_step[0][3];
146    int n;
147
148    for (n = 0; n < len; n++) {
149        //l is s, r is d
150        float l_re = l[n][0];
151        float l_im = l[n][1];
152        float r_re = r[n][0];
153        float r_im = r[n][1];
154        h0 += hs0;
155        h1 += hs1;
156        h2 += hs2;
157        h3 += hs3;
158        l[n][0] = h0 * l_re + h2 * r_re;
159        l[n][1] = h0 * l_im + h2 * r_im;
160        r[n][0] = h1 * l_re + h3 * r_re;
161        r[n][1] = h1 * l_im + h3 * r_im;
162    }
163}
164
165static void ps_stereo_interpolate_ipdopd_c(float (*l)[2], float (*r)[2],
166                                           float h[2][4], float h_step[2][4],
167                                           int len)
168{
169    float h00  = h[0][0],      h10  = h[1][0];
170    float h01  = h[0][1],      h11  = h[1][1];
171    float h02  = h[0][2],      h12  = h[1][2];
172    float h03  = h[0][3],      h13  = h[1][3];
173    float hs00 = h_step[0][0], hs10 = h_step[1][0];
174    float hs01 = h_step[0][1], hs11 = h_step[1][1];
175    float hs02 = h_step[0][2], hs12 = h_step[1][2];
176    float hs03 = h_step[0][3], hs13 = h_step[1][3];
177    int n;
178
179    for (n = 0; n < len; n++) {
180        //l is s, r is d
181        float l_re = l[n][0];
182        float l_im = l[n][1];
183        float r_re = r[n][0];
184        float r_im = r[n][1];
185        h00 += hs00;
186        h01 += hs01;
187        h02 += hs02;
188        h03 += hs03;
189        h10 += hs10;
190        h11 += hs11;
191        h12 += hs12;
192        h13 += hs13;
193
194        l[n][0] = h00 * l_re + h02 * r_re - h10 * l_im - h12 * r_im;
195        l[n][1] = h00 * l_im + h02 * r_im + h10 * l_re + h12 * r_re;
196        r[n][0] = h01 * l_re + h03 * r_re - h11 * l_im - h13 * r_im;
197        r[n][1] = h01 * l_im + h03 * r_im + h11 * l_re + h13 * r_re;
198    }
199}
200
201av_cold void ff_psdsp_init(PSDSPContext *s)
202{
203    s->add_squares            = ps_add_squares_c;
204    s->mul_pair_single        = ps_mul_pair_single_c;
205    s->hybrid_analysis        = ps_hybrid_analysis_c;
206    s->hybrid_analysis_ileave = ps_hybrid_analysis_ileave_c;
207    s->hybrid_synthesis_deint = ps_hybrid_synthesis_deint_c;
208    s->decorrelate            = ps_decorrelate_c;
209    s->stereo_interpolate[0]  = ps_stereo_interpolate_c;
210    s->stereo_interpolate[1]  = ps_stereo_interpolate_ipdopd_c;
211
212    if (ARCH_ARM)
213        ff_psdsp_init_arm(s);
214    if (ARCH_MIPS)
215        ff_psdsp_init_mips(s);
216}
217