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 * Note: Rounding-to-nearest used unless otherwise stated 23 * 24 */ 25 26#define USE_FIXED 1 27 28#include "aac.h" 29#include "config.h" 30#include "libavutil/attributes.h" 31#include "libavutil/intfloat.h" 32#include "sbrdsp.h" 33 34static SoftFloat sbr_sum_square_c(int (*x)[2], int n) 35{ 36 SoftFloat ret; 37 uint64_t accu = 0, round; 38 uint64_t accu0 = 0, accu1 = 0, accu2 = 0, accu3 = 0; 39 int i, nz, nz0; 40 unsigned u; 41 42 nz = 0; 43 for (i = 0; i < n; i += 2) { 44 accu0 += (int64_t)x[i + 0][0] * x[i + 0][0]; 45 accu1 += (int64_t)x[i + 0][1] * x[i + 0][1]; 46 accu2 += (int64_t)x[i + 1][0] * x[i + 1][0]; 47 accu3 += (int64_t)x[i + 1][1] * x[i + 1][1]; 48 if ((accu0|accu1|accu2|accu3) > UINT64_MAX - INT32_MIN*(int64_t)INT32_MIN || i+2>=n) { 49 accu0 >>= nz; 50 accu1 >>= nz; 51 accu2 >>= nz; 52 accu3 >>= nz; 53 while ((accu0|accu1|accu2|accu3) > (UINT64_MAX - accu) >> 2) { 54 accu0 >>= 1; 55 accu1 >>= 1; 56 accu2 >>= 1; 57 accu3 >>= 1; 58 accu >>= 1; 59 nz ++; 60 } 61 accu += accu0 + accu1 + accu2 + accu3; 62 accu0 = accu1 = accu2 = accu3 = 0; 63 } 64 } 65 66 nz0 = 15 - nz; 67 68 u = accu >> 32; 69 if (u) { 70 nz = 33; 71 while (u < 0x80000000U) { 72 u <<= 1; 73 nz--; 74 } 75 } else 76 nz = 1; 77 78 round = 1ULL << (nz-1); 79 u = ((accu + round) >> nz); 80 u >>= 1; 81 ret = av_int2sf(u, nz0 - nz); 82 83 return ret; 84} 85 86static void sbr_neg_odd_64_c(int *x) 87{ 88 int i; 89 for (i = 1; i < 64; i += 2) 90 x[i] = -(unsigned)x[i]; 91} 92 93static void sbr_qmf_pre_shuffle_c(int *z) 94{ 95 int k; 96 z[64] = z[0]; 97 z[65] = z[1]; 98 for (k = 1; k < 32; k++) { 99 z[64+2*k ] = -z[64 - k]; 100 z[64+2*k+1] = z[ k + 1]; 101 } 102} 103 104static void sbr_qmf_post_shuffle_c(int W[32][2], const int *z) 105{ 106 int k; 107 for (k = 0; k < 32; k++) { 108 W[k][0] = -z[63-k]; 109 W[k][1] = z[k]; 110 } 111} 112 113static void sbr_qmf_deint_neg_c(int *v, const int *src) 114{ 115 int i; 116 for (i = 0; i < 32; i++) { 117 v[ i] = (int)(0x10U + src[63 - 2*i ]) >> 5; 118 v[63 - i] = (int)(0x10U - src[63 - 2*i - 1]) >> 5; 119 } 120} 121 122static av_always_inline SoftFloat autocorr_calc(int64_t accu) 123{ 124 int nz, mant, expo; 125 unsigned round; 126 int i = (int)(accu >> 32); 127 if (i == 0) { 128 nz = 1; 129 } else { 130 nz = 0; 131 while (FFABS(i) < 0x40000000) { 132 i *= 2; 133 nz++; 134 } 135 nz = 32-nz; 136 } 137 138 round = 1U << (nz-1); 139 mant = (int)((accu + round) >> nz); 140 mant = (mant + 0x40LL)>>7; 141 mant *= 64; 142 expo = nz + 15; 143 return av_int2sf(mant, 30 - expo); 144} 145 146static av_always_inline void autocorrelate(const int x[40][2], SoftFloat phi[3][2][2], int lag) 147{ 148 int i; 149 int64_t real_sum, imag_sum; 150 int64_t accu_re = 0, accu_im = 0; 151 152 if (lag) { 153 for (i = 1; i < 38; i++) { 154 accu_re += (uint64_t)x[i][0] * x[i+lag][0]; 155 accu_re += (uint64_t)x[i][1] * x[i+lag][1]; 156 accu_im += (uint64_t)x[i][0] * x[i+lag][1]; 157 accu_im -= (uint64_t)x[i][1] * x[i+lag][0]; 158 } 159 160 real_sum = accu_re; 161 imag_sum = accu_im; 162 163 accu_re += (uint64_t)x[ 0][0] * x[lag][0]; 164 accu_re += (uint64_t)x[ 0][1] * x[lag][1]; 165 accu_im += (uint64_t)x[ 0][0] * x[lag][1]; 166 accu_im -= (uint64_t)x[ 0][1] * x[lag][0]; 167 168 phi[2-lag][1][0] = autocorr_calc(accu_re); 169 phi[2-lag][1][1] = autocorr_calc(accu_im); 170 171 if (lag == 1) { 172 accu_re = real_sum; 173 accu_im = imag_sum; 174 accu_re += (uint64_t)x[38][0] * x[39][0]; 175 accu_re += (uint64_t)x[38][1] * x[39][1]; 176 accu_im += (uint64_t)x[38][0] * x[39][1]; 177 accu_im -= (uint64_t)x[38][1] * x[39][0]; 178 179 phi[0][0][0] = autocorr_calc(accu_re); 180 phi[0][0][1] = autocorr_calc(accu_im); 181 } 182 } else { 183 for (i = 1; i < 38; i++) { 184 accu_re += (uint64_t)x[i][0] * x[i][0]; 185 accu_re += (uint64_t)x[i][1] * x[i][1]; 186 } 187 real_sum = accu_re; 188 accu_re += (uint64_t)x[ 0][0] * x[ 0][0]; 189 accu_re += (uint64_t)x[ 0][1] * x[ 0][1]; 190 191 phi[2][1][0] = autocorr_calc(accu_re); 192 193 accu_re = real_sum; 194 accu_re += (uint64_t)x[38][0] * x[38][0]; 195 accu_re += (uint64_t)x[38][1] * x[38][1]; 196 197 phi[1][0][0] = autocorr_calc(accu_re); 198 } 199} 200 201static void sbr_autocorrelate_c(const int x[40][2], SoftFloat phi[3][2][2]) 202{ 203 autocorrelate(x, phi, 0); 204 autocorrelate(x, phi, 1); 205 autocorrelate(x, phi, 2); 206} 207 208static void sbr_hf_gen_c(int (*X_high)[2], const int (*X_low)[2], 209 const int alpha0[2], const int alpha1[2], 210 int bw, int start, int end) 211{ 212 int alpha[4]; 213 int i; 214 int64_t accu; 215 216 accu = (int64_t)alpha0[0] * bw; 217 alpha[2] = (int)((accu + 0x40000000) >> 31); 218 accu = (int64_t)alpha0[1] * bw; 219 alpha[3] = (int)((accu + 0x40000000) >> 31); 220 accu = (int64_t)bw * bw; 221 bw = (int)((accu + 0x40000000) >> 31); 222 accu = (int64_t)alpha1[0] * bw; 223 alpha[0] = (int)((accu + 0x40000000) >> 31); 224 accu = (int64_t)alpha1[1] * bw; 225 alpha[1] = (int)((accu + 0x40000000) >> 31); 226 227 for (i = start; i < end; i++) { 228 accu = (int64_t)X_low[i][0] * 0x20000000; 229 accu += (int64_t)X_low[i - 2][0] * alpha[0]; 230 accu -= (int64_t)X_low[i - 2][1] * alpha[1]; 231 accu += (int64_t)X_low[i - 1][0] * alpha[2]; 232 accu -= (int64_t)X_low[i - 1][1] * alpha[3]; 233 X_high[i][0] = (int)((accu + 0x10000000) >> 29); 234 235 accu = (int64_t)X_low[i][1] * 0x20000000; 236 accu += (int64_t)X_low[i - 2][1] * alpha[0]; 237 accu += (int64_t)X_low[i - 2][0] * alpha[1]; 238 accu += (int64_t)X_low[i - 1][1] * alpha[2]; 239 accu += (int64_t)X_low[i - 1][0] * alpha[3]; 240 X_high[i][1] = (int)((accu + 0x10000000) >> 29); 241 } 242} 243 244static void sbr_hf_g_filt_c(int (*Y)[2], const int (*X_high)[40][2], 245 const SoftFloat *g_filt, int m_max, intptr_t ixh) 246{ 247 int m; 248 int64_t accu; 249 250 for (m = 0; m < m_max; m++) { 251 if (22 - g_filt[m].exp < 61) { 252 int64_t r = 1LL << (22-g_filt[m].exp); 253 accu = (int64_t)X_high[m][ixh][0] * ((g_filt[m].mant + 0x40)>>7); 254 Y[m][0] = (int)((accu + r) >> (23-g_filt[m].exp)); 255 256 accu = (int64_t)X_high[m][ixh][1] * ((g_filt[m].mant + 0x40)>>7); 257 Y[m][1] = (int)((accu + r) >> (23-g_filt[m].exp)); 258 } 259 } 260} 261 262static av_always_inline int sbr_hf_apply_noise(int (*Y)[2], 263 const SoftFloat *s_m, 264 const SoftFloat *q_filt, 265 int noise, 266 int phi_sign0, 267 int phi_sign1, 268 int m_max) 269{ 270 int m; 271 272 for (m = 0; m < m_max; m++) { 273 unsigned y0 = Y[m][0]; 274 unsigned y1 = Y[m][1]; 275 noise = (noise + 1) & 0x1ff; 276 if (s_m[m].mant) { 277 int shift, round; 278 279 shift = 22 - s_m[m].exp; 280 if (shift < 1) { 281 av_log(NULL, AV_LOG_ERROR, "Overflow in sbr_hf_apply_noise, shift=%d\n", shift); 282 return AVERROR(ERANGE); 283 } else if (shift < 30) { 284 round = 1 << (shift-1); 285 y0 += (s_m[m].mant * phi_sign0 + round) >> shift; 286 y1 += (s_m[m].mant * phi_sign1 + round) >> shift; 287 } 288 } else { 289 int shift, round, tmp; 290 int64_t accu; 291 292 shift = 22 - q_filt[m].exp; 293 if (shift < 1) { 294 av_log(NULL, AV_LOG_ERROR, "Overflow in sbr_hf_apply_noise, shift=%d\n", shift); 295 return AVERROR(ERANGE); 296 } else if (shift < 30) { 297 round = 1 << (shift-1); 298 299 accu = (int64_t)q_filt[m].mant * ff_sbr_noise_table_fixed[noise][0]; 300 tmp = (int)((accu + 0x40000000) >> 31); 301 y0 += (tmp + round) >> shift; 302 303 accu = (int64_t)q_filt[m].mant * ff_sbr_noise_table_fixed[noise][1]; 304 tmp = (int)((accu + 0x40000000) >> 31); 305 y1 += (tmp + round) >> shift; 306 } 307 } 308 Y[m][0] = y0; 309 Y[m][1] = y1; 310 phi_sign1 = -phi_sign1; 311 } 312 return 0; 313} 314 315#include "sbrdsp_template.c" 316