sbrdsp.c 9.61 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
/*
 * AAC Spectral Band Replication decoding functions
 * Copyright (c) 2008-2009 Robert Swain ( rob opendot cl )
 * Copyright (c) 2009-2010 Alex Converse <alex.converse@gmail.com>
 *
 * This file is part of Libav.
 *
 * Libav 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 2.1 of the License, or (at your option) any later version.
 *
 * Libav 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 Libav; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

23
#include "config.h"
24
#include "libavutil/attributes.h"
25
#include "libavutil/intfloat.h"
26 27 28 29 30 31 32 33 34 35 36 37 38
#include "sbrdsp.h"

static void sbr_sum64x5_c(float *z)
{
    int k;
    for (k = 0; k < 64; k++) {
        float f = z[k] + z[k + 64] + z[k + 128] + z[k + 192] + z[k + 256];
        z[k] = f;
    }
}

static float sbr_sum_square_c(float (*x)[2], int n)
{
39
    float sum0 = 0.0f, sum1 = 0.0f;
40 41
    int i;

42 43 44 45 46 47 48
    for (i = 0; i < n; i += 2)
    {
        sum0 += x[i + 0][0] * x[i + 0][0];
        sum1 += x[i + 0][1] * x[i + 0][1];
        sum0 += x[i + 1][0] * x[i + 1][0];
        sum1 += x[i + 1][1] * x[i + 1][1];
    }
49

50
    return sum0 + sum1;
51 52 53 54
}

static void sbr_neg_odd_64_c(float *x)
{
55
    union av_intfloat32 *xi = (union av_intfloat32*) x;
56
    int i;
57 58 59 60
    for (i = 1; i < 64; i += 4) {
        xi[i + 0].i ^= 1U << 31;
        xi[i + 2].i ^= 1U << 31;
    }
61 62 63 64
}

static void sbr_qmf_pre_shuffle_c(float *z)
{
65
    union av_intfloat32 *zi = (union av_intfloat32*) z;
66
    int k;
67 68 69 70 71 72 73
    zi[64].i = zi[0].i;
    zi[65].i = zi[1].i;
    for (k = 1; k < 31; k += 2) {
        zi[64 + 2 * k + 0].i = zi[64 - k].i ^ (1U << 31);
        zi[64 + 2 * k + 1].i = zi[ k + 1].i;
        zi[64 + 2 * k + 2].i = zi[63 - k].i ^ (1U << 31);
        zi[64 + 2 * k + 3].i = zi[ k + 2].i;
74
    }
75 76
    zi[64 + 2 * 31 + 0].i = zi[64 - 31].i ^ (1U << 31);
    zi[64 + 2 * 31 + 1].i = zi[31 +  1].i;
77 78 79 80
}

static void sbr_qmf_post_shuffle_c(float W[32][2], const float *z)
{
81 82
    const union av_intfloat32 *zi = (const union av_intfloat32*) z;
    union av_intfloat32 *Wi       = (union av_intfloat32*) W;
83
    int k;
84 85 86 87 88
    for (k = 0; k < 32; k += 2) {
        Wi[2 * k + 0].i = zi[63 - k].i ^ (1U << 31);
        Wi[2 * k + 1].i = zi[ k + 0].i;
        Wi[2 * k + 2].i = zi[62 - k].i ^ (1U << 31);
        Wi[2 * k + 3].i = zi[ k + 1].i;
89 90 91 92 93
    }
}

static void sbr_qmf_deint_neg_c(float *v, const float *src)
{
94 95
    const union av_intfloat32 *si = (const union av_intfloat32*)src;
    union av_intfloat32 *vi = (union av_intfloat32*)v;
96 97
    int i;
    for (i = 0; i < 32; i++) {
98 99
        vi[     i].i = si[63 - 2 * i    ].i;
        vi[63 - i].i = si[63 - 2 * i - 1].i ^ (1U << 31);
100 101 102 103 104 105 106 107 108 109 110 111
    }
}

static void sbr_qmf_deint_bfly_c(float *v, const float *src0, const float *src1)
{
    int i;
    for (i = 0; i < 64; i++) {
        v[      i] = src0[i] - src1[63 - i];
        v[127 - i] = src0[i] + src1[63 - i];
    }
}

112 113 114 115 116

#if 0
    /* This code is slower because it multiplies memory accesses.
     * It is left for educational purposes and because it may offer
     * a better reference for writing arch-specific DSP functions. */
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
static av_always_inline void autocorrelate(const float x[40][2],
                                           float phi[3][2][2], int lag)
{
    int i;
    float real_sum = 0.0f;
    float imag_sum = 0.0f;
    if (lag) {
        for (i = 1; i < 38; i++) {
            real_sum += x[i][0] * x[i+lag][0] + x[i][1] * x[i+lag][1];
            imag_sum += x[i][0] * x[i+lag][1] - x[i][1] * x[i+lag][0];
        }
        phi[2-lag][1][0] = real_sum + x[ 0][0] * x[lag][0] + x[ 0][1] * x[lag][1];
        phi[2-lag][1][1] = imag_sum + x[ 0][0] * x[lag][1] - x[ 0][1] * x[lag][0];
        if (lag == 1) {
            phi[0][0][0] = real_sum + x[38][0] * x[39][0] + x[38][1] * x[39][1];
            phi[0][0][1] = imag_sum + x[38][0] * x[39][1] - x[38][1] * x[39][0];
        }
    } else {
        for (i = 1; i < 38; i++) {
            real_sum += x[i][0] * x[i][0] + x[i][1] * x[i][1];
        }
        phi[2][1][0] = real_sum + x[ 0][0] * x[ 0][0] + x[ 0][1] * x[ 0][1];
        phi[1][0][0] = real_sum + x[38][0] * x[38][0] + x[38][1] * x[38][1];
    }
}

static void sbr_autocorrelate_c(const float x[40][2], float phi[3][2][2])
{
    autocorrelate(x, phi, 0);
    autocorrelate(x, phi, 1);
    autocorrelate(x, phi, 2);
148
}
149
#else
150 151
static void sbr_autocorrelate_c(const float x[40][2], float phi[3][2][2])
{
152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171
    float real_sum2 = x[0][0] * x[2][0] + x[0][1] * x[2][1];
    float imag_sum2 = x[0][0] * x[2][1] - x[0][1] * x[2][0];
    float real_sum1 = 0.0f, imag_sum1 = 0.0f, real_sum0 = 0.0f;
    int   i;
    for (i = 1; i < 38; i++) {
        real_sum0 += x[i][0] * x[i    ][0] + x[i][1] * x[i    ][1];
        real_sum1 += x[i][0] * x[i + 1][0] + x[i][1] * x[i + 1][1];
        imag_sum1 += x[i][0] * x[i + 1][1] - x[i][1] * x[i + 1][0];
        real_sum2 += x[i][0] * x[i + 2][0] + x[i][1] * x[i + 2][1];
        imag_sum2 += x[i][0] * x[i + 2][1] - x[i][1] * x[i + 2][0];
    }
    phi[2 - 2][1][0] = real_sum2;
    phi[2 - 2][1][1] = imag_sum2;
    phi[2    ][1][0] = real_sum0 + x[ 0][0] * x[ 0][0] + x[ 0][1] * x[ 0][1];
    phi[1    ][0][0] = real_sum0 + x[38][0] * x[38][0] + x[38][1] * x[38][1];
    phi[2 - 1][1][0] = real_sum1 + x[ 0][0] * x[ 1][0] + x[ 0][1] * x[ 1][1];
    phi[2 - 1][1][1] = imag_sum1 + x[ 0][0] * x[ 1][1] - x[ 0][1] * x[ 1][0];
    phi[0    ][0][0] = real_sum1 + x[38][0] * x[39][0] + x[38][1] * x[39][1];
    phi[0    ][0][1] = imag_sum1 + x[38][0] * x[39][1] - x[38][1] * x[39][0];
#endif
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
}

static void sbr_hf_gen_c(float (*X_high)[2], const float (*X_low)[2],
                         const float alpha0[2], const float alpha1[2],
                         float bw, int start, int end)
{
    float alpha[4];
    int i;

    alpha[0] = alpha1[0] * bw * bw;
    alpha[1] = alpha1[1] * bw * bw;
    alpha[2] = alpha0[0] * bw;
    alpha[3] = alpha0[1] * bw;

    for (i = start; i < end; i++) {
        X_high[i][0] =
            X_low[i - 2][0] * alpha[0] -
            X_low[i - 2][1] * alpha[1] +
            X_low[i - 1][0] * alpha[2] -
            X_low[i - 1][1] * alpha[3] +
            X_low[i][0];
        X_high[i][1] =
            X_low[i - 2][1] * alpha[0] +
            X_low[i - 2][0] * alpha[1] +
            X_low[i - 1][1] * alpha[2] +
            X_low[i - 1][0] * alpha[3] +
            X_low[i][1];
    }
}

static void sbr_hf_g_filt_c(float (*Y)[2], const float (*X_high)[40][2],
203
                            const float *g_filt, int m_max, intptr_t ixh)
204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
{
    int m;

    for (m = 0; m < m_max; m++) {
        Y[m][0] = X_high[m][ixh][0] * g_filt[m];
        Y[m][1] = X_high[m][ixh][1] * g_filt[m];
    }
}

static av_always_inline void sbr_hf_apply_noise(float (*Y)[2],
                                                const float *s_m,
                                                const float *q_filt,
                                                int noise,
                                                float phi_sign0,
                                                float phi_sign1,
                                                int m_max)
{
    int m;

    for (m = 0; m < m_max; m++) {
        float y0 = Y[m][0];
        float y1 = Y[m][1];
        noise = (noise + 1) & 0x1ff;
        if (s_m[m]) {
            y0 += s_m[m] * phi_sign0;
            y1 += s_m[m] * phi_sign1;
        } else {
            y0 += q_filt[m] * ff_sbr_noise_table[noise][0];
            y1 += q_filt[m] * ff_sbr_noise_table[noise][1];
        }
        Y[m][0] = y0;
        Y[m][1] = y1;
        phi_sign1 = -phi_sign1;
    }
}

static void sbr_hf_apply_noise_0(float (*Y)[2], const float *s_m,
                                 const float *q_filt, int noise,
                                 int kx, int m_max)
{
    sbr_hf_apply_noise(Y, s_m, q_filt, noise, 1.0, 0.0, m_max);
}

static void sbr_hf_apply_noise_1(float (*Y)[2], const float *s_m,
                                 const float *q_filt, int noise,
                                 int kx, int m_max)
{
    float phi_sign = 1 - 2 * (kx & 1);
    sbr_hf_apply_noise(Y, s_m, q_filt, noise, 0.0, phi_sign, m_max);
}

static void sbr_hf_apply_noise_2(float (*Y)[2], const float *s_m,
                                 const float *q_filt, int noise,
                                 int kx, int m_max)
{
    sbr_hf_apply_noise(Y, s_m, q_filt, noise, -1.0, 0.0, m_max);
}

static void sbr_hf_apply_noise_3(float (*Y)[2], const float *s_m,
                                 const float *q_filt, int noise,
                                 int kx, int m_max)
{
    float phi_sign = 1 - 2 * (kx & 1);
    sbr_hf_apply_noise(Y, s_m, q_filt, noise, 0.0, -phi_sign, m_max);
}

av_cold void ff_sbrdsp_init(SBRDSPContext *s)
{
    s->sum64x5 = sbr_sum64x5_c;
    s->sum_square = sbr_sum_square_c;
    s->neg_odd_64 = sbr_neg_odd_64_c;
    s->qmf_pre_shuffle = sbr_qmf_pre_shuffle_c;
    s->qmf_post_shuffle = sbr_qmf_post_shuffle_c;
    s->qmf_deint_neg = sbr_qmf_deint_neg_c;
    s->qmf_deint_bfly = sbr_qmf_deint_bfly_c;
    s->autocorrelate = sbr_autocorrelate_c;
    s->hf_gen = sbr_hf_gen_c;
    s->hf_g_filt = sbr_hf_g_filt_c;

    s->hf_apply_noise[0] = sbr_hf_apply_noise_0;
    s->hf_apply_noise[1] = sbr_hf_apply_noise_1;
    s->hf_apply_noise[2] = sbr_hf_apply_noise_2;
    s->hf_apply_noise[3] = sbr_hf_apply_noise_3;
287 288 289

    if (ARCH_ARM)
        ff_sbrdsp_init_arm(s);
290
    if (ARCH_X86)
291
        ff_sbrdsp_init_x86(s);
292
}