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 * Note: Rounding-to-nearest used unless otherwise stated
21 *
22 */
23#include <stdint.h>
24
25#include "config.h"
26#include "libavutil/attributes.h"
27#include "aacpsdsp.h"
28
29static void ps_add_squares_c(INTFLOAT *dst, const INTFLOAT (*src)[2], int n)
30{
31    int i;
32    for (i = 0; i < n; i++)
33        dst[i] += (UINTFLOAT)AAC_MADD28(src[i][0], src[i][0], src[i][1], src[i][1]);
34}
35
36static void ps_mul_pair_single_c(INTFLOAT (*dst)[2], INTFLOAT (*src0)[2], INTFLOAT *src1,
37                                 int n)
38{
39    int i;
40    for (i = 0; i < n; i++) {
41        dst[i][0] = AAC_MUL16(src0[i][0], src1[i]);
42        dst[i][1] = AAC_MUL16(src0[i][1], src1[i]);
43    }
44}
45
46static void ps_hybrid_analysis_c(INTFLOAT (*out)[2], INTFLOAT (*in)[2],
47                                 const INTFLOAT (*filter)[8][2],
48                                 ptrdiff_t stride, int n)
49{
50    int i, j;
51
52    for (i = 0; i < n; i++) {
53        INT64FLOAT sum_re = (INT64FLOAT)filter[i][6][0] * in[6][0];
54        INT64FLOAT sum_im = (INT64FLOAT)filter[i][6][0] * in[6][1];
55
56        for (j = 0; j < 6; j++) {
57            INT64FLOAT in0_re = in[j][0];
58            INT64FLOAT in0_im = in[j][1];
59            INT64FLOAT in1_re = in[12-j][0];
60            INT64FLOAT in1_im = in[12-j][1];
61            sum_re += (INT64FLOAT)filter[i][j][0] * (in0_re + in1_re) -
62                      (INT64FLOAT)filter[i][j][1] * (in0_im - in1_im);
63            sum_im += (INT64FLOAT)filter[i][j][0] * (in0_im + in1_im) +
64                      (INT64FLOAT)filter[i][j][1] * (in0_re - in1_re);
65        }
66#if USE_FIXED
67        out[i * stride][0] = (int)((sum_re + 0x40000000) >> 31);
68        out[i * stride][1] = (int)((sum_im + 0x40000000) >> 31);
69#else
70        out[i * stride][0] = sum_re;
71        out[i * stride][1] = sum_im;
72#endif /* USE_FIXED */
73    }
74}
75
76static void ps_hybrid_analysis_ileave_c(INTFLOAT (*out)[32][2], INTFLOAT L[2][38][64],
77                                      int i, int len)
78{
79    int j;
80
81    for (; i < 64; i++) {
82        for (j = 0; j < len; j++) {
83            out[i][j][0] = L[0][j][i];
84            out[i][j][1] = L[1][j][i];
85        }
86    }
87}
88
89static void ps_hybrid_synthesis_deint_c(INTFLOAT out[2][38][64],
90                                      INTFLOAT (*in)[32][2],
91                                      int i, int len)
92{
93    int n;
94
95    for (; i < 64; i++) {
96        for (n = 0; n < len; n++) {
97            out[0][n][i] = in[i][n][0];
98            out[1][n][i] = in[i][n][1];
99        }
100    }
101}
102
103static void ps_decorrelate_c(INTFLOAT (*out)[2], INTFLOAT (*delay)[2],
104                             INTFLOAT (*ap_delay)[PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2],
105                             const INTFLOAT phi_fract[2], const INTFLOAT (*Q_fract)[2],
106                             const INTFLOAT *transient_gain,
107                             INTFLOAT g_decay_slope,
108                             int len)
109{
110    static const INTFLOAT a[] = { Q31(0.65143905753106f),
111                               Q31(0.56471812200776f),
112                               Q31(0.48954165955695f) };
113    INTFLOAT ag[PS_AP_LINKS];
114    int m, n;
115
116    for (m = 0; m < PS_AP_LINKS; m++)
117        ag[m] = AAC_MUL30(a[m], g_decay_slope);
118
119    for (n = 0; n < len; n++) {
120        INTFLOAT in_re = AAC_MSUB30(delay[n][0], phi_fract[0], delay[n][1], phi_fract[1]);
121        INTFLOAT in_im = AAC_MADD30(delay[n][0], phi_fract[1], delay[n][1], phi_fract[0]);
122        for (m = 0; m < PS_AP_LINKS; m++) {
123            INTFLOAT a_re                = AAC_MUL31(ag[m], in_re);
124            INTFLOAT a_im                = AAC_MUL31(ag[m], in_im);
125            INTFLOAT link_delay_re       = ap_delay[m][n+2-m][0];
126            INTFLOAT link_delay_im       = ap_delay[m][n+2-m][1];
127            INTFLOAT fractional_delay_re = Q_fract[m][0];
128            INTFLOAT fractional_delay_im = Q_fract[m][1];
129            INTFLOAT apd_re = in_re;
130            INTFLOAT apd_im = in_im;
131            in_re = AAC_MSUB30(link_delay_re, fractional_delay_re,
132                    link_delay_im, fractional_delay_im);
133            in_re -= (UINTFLOAT)a_re;
134            in_im = AAC_MADD30(link_delay_re, fractional_delay_im,
135                    link_delay_im, fractional_delay_re);
136            in_im -= (UINTFLOAT)a_im;
137            ap_delay[m][n+5][0] = apd_re + (UINTFLOAT)AAC_MUL31(ag[m], in_re);
138            ap_delay[m][n+5][1] = apd_im + (UINTFLOAT)AAC_MUL31(ag[m], in_im);
139        }
140        out[n][0] = AAC_MUL16(transient_gain[n], in_re);
141        out[n][1] = AAC_MUL16(transient_gain[n], in_im);
142    }
143}
144
145static void ps_stereo_interpolate_c(INTFLOAT (*l)[2], INTFLOAT (*r)[2],
146                                    INTFLOAT h[2][4], INTFLOAT h_step[2][4],
147                                    int len)
148{
149    INTFLOAT h0 = h[0][0];
150    INTFLOAT h1 = h[0][1];
151    INTFLOAT h2 = h[0][2];
152    INTFLOAT h3 = h[0][3];
153    UINTFLOAT hs0 = h_step[0][0];
154    UINTFLOAT hs1 = h_step[0][1];
155    UINTFLOAT hs2 = h_step[0][2];
156    UINTFLOAT hs3 = h_step[0][3];
157    int n;
158
159    for (n = 0; n < len; n++) {
160        //l is s, r is d
161        INTFLOAT l_re = l[n][0];
162        INTFLOAT l_im = l[n][1];
163        INTFLOAT r_re = r[n][0];
164        INTFLOAT r_im = r[n][1];
165        h0 += hs0;
166        h1 += hs1;
167        h2 += hs2;
168        h3 += hs3;
169        l[n][0] = AAC_MADD30(h0, l_re, h2, r_re);
170        l[n][1] = AAC_MADD30(h0, l_im, h2, r_im);
171        r[n][0] = AAC_MADD30(h1, l_re, h3, r_re);
172        r[n][1] = AAC_MADD30(h1, l_im, h3, r_im);
173    }
174}
175
176static void ps_stereo_interpolate_ipdopd_c(INTFLOAT (*l)[2], INTFLOAT (*r)[2],
177                                           INTFLOAT h[2][4], INTFLOAT h_step[2][4],
178                                           int len)
179{
180    INTFLOAT h00  = h[0][0],      h10  = h[1][0];
181    INTFLOAT h01  = h[0][1],      h11  = h[1][1];
182    INTFLOAT h02  = h[0][2],      h12  = h[1][2];
183    INTFLOAT h03  = h[0][3],      h13  = h[1][3];
184    UINTFLOAT hs00 = h_step[0][0], hs10 = h_step[1][0];
185    UINTFLOAT hs01 = h_step[0][1], hs11 = h_step[1][1];
186    UINTFLOAT hs02 = h_step[0][2], hs12 = h_step[1][2];
187    UINTFLOAT hs03 = h_step[0][3], hs13 = h_step[1][3];
188    int n;
189
190    for (n = 0; n < len; n++) {
191        //l is s, r is d
192        INTFLOAT l_re = l[n][0];
193        INTFLOAT l_im = l[n][1];
194        INTFLOAT r_re = r[n][0];
195        INTFLOAT r_im = r[n][1];
196        h00 += hs00;
197        h01 += hs01;
198        h02 += hs02;
199        h03 += hs03;
200        h10 += hs10;
201        h11 += hs11;
202        h12 += hs12;
203        h13 += hs13;
204
205        l[n][0] = AAC_MSUB30_V8(h00, l_re, h02, r_re, h10, l_im, h12, r_im);
206        l[n][1] = AAC_MADD30_V8(h00, l_im, h02, r_im, h10, l_re, h12, r_re);
207        r[n][0] = AAC_MSUB30_V8(h01, l_re, h03, r_re, h11, l_im, h13, r_im);
208        r[n][1] = AAC_MADD30_V8(h01, l_im, h03, r_im, h11, l_re, h13, r_re);
209    }
210}
211
212av_cold void AAC_RENAME(ff_psdsp_init)(PSDSPContext *s)
213{
214    s->add_squares            = ps_add_squares_c;
215    s->mul_pair_single        = ps_mul_pair_single_c;
216    s->hybrid_analysis        = ps_hybrid_analysis_c;
217    s->hybrid_analysis_ileave = ps_hybrid_analysis_ileave_c;
218    s->hybrid_synthesis_deint = ps_hybrid_synthesis_deint_c;
219    s->decorrelate            = ps_decorrelate_c;
220    s->stereo_interpolate[0]  = ps_stereo_interpolate_c;
221    s->stereo_interpolate[1]  = ps_stereo_interpolate_ipdopd_c;
222
223#if !USE_FIXED
224#if ARCH_ARM
225    ff_psdsp_init_arm(s);
226#elif ARCH_AARCH64
227    ff_psdsp_init_aarch64(s);
228#elif ARCH_MIPS
229    ff_psdsp_init_mips(s);
230#elif ARCH_X86
231    ff_psdsp_init_x86(s);
232#endif
233#endif /* !USE_FIXED */
234}
235