00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00022 #include "avcodec.h"
00023 #include "sinewin.h"
00024 #include "wma.h"
00025 #include "wmadata.h"
00026
00027 #undef NDEBUG
00028 #include <assert.h>
00029
00030
00031
00032 static void init_coef_vlc(VLC *vlc, uint16_t **prun_table,
00033 float **plevel_table, uint16_t **pint_table,
00034 const CoefVLCTable *vlc_table)
00035 {
00036 int n = vlc_table->n;
00037 const uint8_t *table_bits = vlc_table->huffbits;
00038 const uint32_t *table_codes = vlc_table->huffcodes;
00039 const uint16_t *levels_table = vlc_table->levels;
00040 uint16_t *run_table, *level_table, *int_table;
00041 float *flevel_table;
00042 int i, l, j, k, level;
00043
00044 init_vlc(vlc, VLCBITS, n, table_bits, 1, 1, table_codes, 4, 4, 0);
00045
00046 run_table = av_malloc(n * sizeof(uint16_t));
00047 level_table = av_malloc(n * sizeof(uint16_t));
00048 flevel_table= av_malloc(n * sizeof(*flevel_table));
00049 int_table = av_malloc(n * sizeof(uint16_t));
00050 i = 2;
00051 level = 1;
00052 k = 0;
00053 while (i < n) {
00054 int_table[k] = i;
00055 l = levels_table[k++];
00056 for (j = 0; j < l; j++) {
00057 run_table[i] = j;
00058 level_table[i] = level;
00059 flevel_table[i]= level;
00060 i++;
00061 }
00062 level++;
00063 }
00064 *prun_table = run_table;
00065 *plevel_table = flevel_table;
00066 *pint_table = int_table;
00067 av_free(level_table);
00068 }
00069
00077 int av_cold ff_wma_get_frame_len_bits(int sample_rate, int version,
00078 unsigned int decode_flags)
00079 {
00080
00081 int frame_len_bits;
00082
00083 if (sample_rate <= 16000) {
00084 frame_len_bits = 9;
00085 } else if (sample_rate <= 22050 ||
00086 (sample_rate <= 32000 && version == 1)) {
00087 frame_len_bits = 10;
00088 } else if (sample_rate <= 48000) {
00089 frame_len_bits = 11;
00090 } else if (sample_rate <= 96000) {
00091 frame_len_bits = 12;
00092 } else {
00093 frame_len_bits = 13;
00094 }
00095
00096 if (version == 3) {
00097 int tmp = decode_flags & 0x6;
00098 if (tmp == 0x2) {
00099 ++frame_len_bits;
00100 } else if (tmp == 0x4) {
00101 --frame_len_bits;
00102 } else if (tmp == 0x6) {
00103 frame_len_bits -= 2;
00104 }
00105 }
00106
00107 return frame_len_bits;
00108 }
00109
00110 int ff_wma_init(AVCodecContext *avctx, int flags2)
00111 {
00112 WMACodecContext *s = avctx->priv_data;
00113 int i;
00114 float bps1, high_freq;
00115 volatile float bps;
00116 int sample_rate1;
00117 int coef_vlc_table;
00118
00119 if ( avctx->sample_rate <= 0 || avctx->sample_rate > 50000
00120 || avctx->channels <= 0 || avctx->channels > 8
00121 || avctx->bit_rate <= 0)
00122 return -1;
00123
00124 s->sample_rate = avctx->sample_rate;
00125 s->nb_channels = avctx->channels;
00126 s->bit_rate = avctx->bit_rate;
00127 s->block_align = avctx->block_align;
00128
00129 dsputil_init(&s->dsp, avctx);
00130 ff_fmt_convert_init(&s->fmt_conv, avctx);
00131
00132 if (avctx->codec->id == CODEC_ID_WMAV1) {
00133 s->version = 1;
00134 } else {
00135 s->version = 2;
00136 }
00137
00138
00139 s->frame_len_bits = ff_wma_get_frame_len_bits(s->sample_rate, s->version, 0);
00140
00141 s->frame_len = 1 << s->frame_len_bits;
00142 if (s->use_variable_block_len) {
00143 int nb_max, nb;
00144 nb = ((flags2 >> 3) & 3) + 1;
00145 if ((s->bit_rate / s->nb_channels) >= 32000)
00146 nb += 2;
00147 nb_max = s->frame_len_bits - BLOCK_MIN_BITS;
00148 if (nb > nb_max)
00149 nb = nb_max;
00150 s->nb_block_sizes = nb + 1;
00151 } else {
00152 s->nb_block_sizes = 1;
00153 }
00154
00155
00156 s->use_noise_coding = 1;
00157 high_freq = s->sample_rate * 0.5;
00158
00159
00160 sample_rate1 = s->sample_rate;
00161 if (s->version == 2) {
00162 if (sample_rate1 >= 44100) {
00163 sample_rate1 = 44100;
00164 } else if (sample_rate1 >= 22050) {
00165 sample_rate1 = 22050;
00166 } else if (sample_rate1 >= 16000) {
00167 sample_rate1 = 16000;
00168 } else if (sample_rate1 >= 11025) {
00169 sample_rate1 = 11025;
00170 } else if (sample_rate1 >= 8000) {
00171 sample_rate1 = 8000;
00172 }
00173 }
00174
00175 bps = (float)s->bit_rate / (float)(s->nb_channels * s->sample_rate);
00176 s->byte_offset_bits = av_log2((int)(bps * s->frame_len / 8.0 + 0.5)) + 2;
00177
00178
00179
00180 bps1 = bps;
00181 if (s->nb_channels == 2)
00182 bps1 = bps * 1.6;
00183 if (sample_rate1 == 44100) {
00184 if (bps1 >= 0.61) {
00185 s->use_noise_coding = 0;
00186 } else {
00187 high_freq = high_freq * 0.4;
00188 }
00189 } else if (sample_rate1 == 22050) {
00190 if (bps1 >= 1.16) {
00191 s->use_noise_coding = 0;
00192 } else if (bps1 >= 0.72) {
00193 high_freq = high_freq * 0.7;
00194 } else {
00195 high_freq = high_freq * 0.6;
00196 }
00197 } else if (sample_rate1 == 16000) {
00198 if (bps > 0.5) {
00199 high_freq = high_freq * 0.5;
00200 } else {
00201 high_freq = high_freq * 0.3;
00202 }
00203 } else if (sample_rate1 == 11025) {
00204 high_freq = high_freq * 0.7;
00205 } else if (sample_rate1 == 8000) {
00206 if (bps <= 0.625) {
00207 high_freq = high_freq * 0.5;
00208 } else if (bps > 0.75) {
00209 s->use_noise_coding = 0;
00210 } else {
00211 high_freq = high_freq * 0.65;
00212 }
00213 } else {
00214 if (bps >= 0.8) {
00215 high_freq = high_freq * 0.75;
00216 } else if (bps >= 0.6) {
00217 high_freq = high_freq * 0.6;
00218 } else {
00219 high_freq = high_freq * 0.5;
00220 }
00221 }
00222 av_dlog(s->avctx, "flags2=0x%x\n", flags2);
00223 av_dlog(s->avctx, "version=%d channels=%d sample_rate=%d bitrate=%d block_align=%d\n",
00224 s->version, s->nb_channels, s->sample_rate, s->bit_rate,
00225 s->block_align);
00226 av_dlog(s->avctx, "bps=%f bps1=%f high_freq=%f bitoffset=%d\n",
00227 bps, bps1, high_freq, s->byte_offset_bits);
00228 av_dlog(s->avctx, "use_noise_coding=%d use_exp_vlc=%d nb_block_sizes=%d\n",
00229 s->use_noise_coding, s->use_exp_vlc, s->nb_block_sizes);
00230
00231
00232 {
00233 int a, b, pos, lpos, k, block_len, i, j, n;
00234 const uint8_t *table;
00235
00236 if (s->version == 1) {
00237 s->coefs_start = 3;
00238 } else {
00239 s->coefs_start = 0;
00240 }
00241 for (k = 0; k < s->nb_block_sizes; k++) {
00242 block_len = s->frame_len >> k;
00243
00244 if (s->version == 1) {
00245 lpos = 0;
00246 for (i = 0; i < 25; i++) {
00247 a = ff_wma_critical_freqs[i];
00248 b = s->sample_rate;
00249 pos = ((block_len * 2 * a) + (b >> 1)) / b;
00250 if (pos > block_len)
00251 pos = block_len;
00252 s->exponent_bands[0][i] = pos - lpos;
00253 if (pos >= block_len) {
00254 i++;
00255 break;
00256 }
00257 lpos = pos;
00258 }
00259 s->exponent_sizes[0] = i;
00260 } else {
00261
00262 table = NULL;
00263 a = s->frame_len_bits - BLOCK_MIN_BITS - k;
00264 if (a < 3) {
00265 if (s->sample_rate >= 44100) {
00266 table = exponent_band_44100[a];
00267 } else if (s->sample_rate >= 32000) {
00268 table = exponent_band_32000[a];
00269 } else if (s->sample_rate >= 22050) {
00270 table = exponent_band_22050[a];
00271 }
00272 }
00273 if (table) {
00274 n = *table++;
00275 for (i = 0; i < n; i++)
00276 s->exponent_bands[k][i] = table[i];
00277 s->exponent_sizes[k] = n;
00278 } else {
00279 j = 0;
00280 lpos = 0;
00281 for (i = 0; i < 25; i++) {
00282 a = ff_wma_critical_freqs[i];
00283 b = s->sample_rate;
00284 pos = ((block_len * 2 * a) + (b << 1)) / (4 * b);
00285 pos <<= 2;
00286 if (pos > block_len)
00287 pos = block_len;
00288 if (pos > lpos)
00289 s->exponent_bands[k][j++] = pos - lpos;
00290 if (pos >= block_len)
00291 break;
00292 lpos = pos;
00293 }
00294 s->exponent_sizes[k] = j;
00295 }
00296 }
00297
00298
00299 s->coefs_end[k] = (s->frame_len - ((s->frame_len * 9) / 100)) >> k;
00300
00301 s->high_band_start[k] = (int)((block_len * 2 * high_freq) /
00302 s->sample_rate + 0.5);
00303 n = s->exponent_sizes[k];
00304 j = 0;
00305 pos = 0;
00306 for (i = 0; i < n; i++) {
00307 int start, end;
00308 start = pos;
00309 pos += s->exponent_bands[k][i];
00310 end = pos;
00311 if (start < s->high_band_start[k])
00312 start = s->high_band_start[k];
00313 if (end > s->coefs_end[k])
00314 end = s->coefs_end[k];
00315 if (end > start)
00316 s->exponent_high_bands[k][j++] = end - start;
00317 }
00318 s->exponent_high_sizes[k] = j;
00319 #if 0
00320 tprintf(s->avctx, "%5d: coefs_end=%d high_band_start=%d nb_high_bands=%d: ",
00321 s->frame_len >> k,
00322 s->coefs_end[k],
00323 s->high_band_start[k],
00324 s->exponent_high_sizes[k]);
00325 for (j = 0; j < s->exponent_high_sizes[k]; j++)
00326 tprintf(s->avctx, " %d", s->exponent_high_bands[k][j]);
00327 tprintf(s->avctx, "\n");
00328 #endif
00329 }
00330 }
00331
00332 #ifdef TRACE
00333 {
00334 int i, j;
00335 for (i = 0; i < s->nb_block_sizes; i++) {
00336 tprintf(s->avctx, "%5d: n=%2d:",
00337 s->frame_len >> i,
00338 s->exponent_sizes[i]);
00339 for (j = 0; j < s->exponent_sizes[i]; j++)
00340 tprintf(s->avctx, " %d", s->exponent_bands[i][j]);
00341 tprintf(s->avctx, "\n");
00342 }
00343 }
00344 #endif
00345
00346
00347 for (i = 0; i < s->nb_block_sizes; i++) {
00348 ff_init_ff_sine_windows(s->frame_len_bits - i);
00349 s->windows[i] = ff_sine_windows[s->frame_len_bits - i];
00350 }
00351
00352 s->reset_block_lengths = 1;
00353
00354 if (s->use_noise_coding) {
00355
00356
00357 if (s->use_exp_vlc) {
00358 s->noise_mult = 0.02;
00359 } else {
00360 s->noise_mult = 0.04;
00361 }
00362
00363 #ifdef TRACE
00364 for (i = 0; i < NOISE_TAB_SIZE; i++)
00365 s->noise_table[i] = 1.0 * s->noise_mult;
00366 #else
00367 {
00368 unsigned int seed;
00369 float norm;
00370 seed = 1;
00371 norm = (1.0 / (float)(1LL << 31)) * sqrt(3) * s->noise_mult;
00372 for (i = 0; i < NOISE_TAB_SIZE; i++) {
00373 seed = seed * 314159 + 1;
00374 s->noise_table[i] = (float)((int)seed) * norm;
00375 }
00376 }
00377 #endif
00378 }
00379
00380
00381 coef_vlc_table = 2;
00382 if (s->sample_rate >= 32000) {
00383 if (bps1 < 0.72) {
00384 coef_vlc_table = 0;
00385 } else if (bps1 < 1.16) {
00386 coef_vlc_table = 1;
00387 }
00388 }
00389 s->coef_vlcs[0]= &coef_vlcs[coef_vlc_table * 2 ];
00390 s->coef_vlcs[1]= &coef_vlcs[coef_vlc_table * 2 + 1];
00391 init_coef_vlc(&s->coef_vlc[0], &s->run_table[0], &s->level_table[0], &s->int_table[0],
00392 s->coef_vlcs[0]);
00393 init_coef_vlc(&s->coef_vlc[1], &s->run_table[1], &s->level_table[1], &s->int_table[1],
00394 s->coef_vlcs[1]);
00395
00396 return 0;
00397 }
00398
00399 int ff_wma_total_gain_to_bits(int total_gain)
00400 {
00401 if (total_gain < 15) return 13;
00402 else if (total_gain < 32) return 12;
00403 else if (total_gain < 40) return 11;
00404 else if (total_gain < 45) return 10;
00405 else return 9;
00406 }
00407
00408 int ff_wma_end(AVCodecContext *avctx)
00409 {
00410 WMACodecContext *s = avctx->priv_data;
00411 int i;
00412
00413 for (i = 0; i < s->nb_block_sizes; i++)
00414 ff_mdct_end(&s->mdct_ctx[i]);
00415
00416 if (s->use_exp_vlc) {
00417 free_vlc(&s->exp_vlc);
00418 }
00419 if (s->use_noise_coding) {
00420 free_vlc(&s->hgain_vlc);
00421 }
00422 for (i = 0; i < 2; i++) {
00423 free_vlc(&s->coef_vlc[i]);
00424 av_free(s->run_table[i]);
00425 av_free(s->level_table[i]);
00426 av_free(s->int_table[i]);
00427 }
00428
00429 return 0;
00430 }
00431
00437 unsigned int ff_wma_get_large_val(GetBitContext* gb)
00438 {
00440 int n_bits = 8;
00442 if (get_bits1(gb)) {
00443 n_bits += 8;
00444 if (get_bits1(gb)) {
00445 n_bits += 8;
00446 if (get_bits1(gb)) {
00447 n_bits += 7;
00448 }
00449 }
00450 }
00451 return get_bits_long(gb, n_bits);
00452 }
00453
00470 int ff_wma_run_level_decode(AVCodecContext* avctx, GetBitContext* gb,
00471 VLC *vlc,
00472 const float *level_table, const uint16_t *run_table,
00473 int version, WMACoef *ptr, int offset,
00474 int num_coefs, int block_len, int frame_len_bits,
00475 int coef_nb_bits)
00476 {
00477 int code, level, sign;
00478 const uint32_t *ilvl = (const uint32_t*)level_table;
00479 uint32_t *iptr = (uint32_t*)ptr;
00480 const unsigned int coef_mask = block_len - 1;
00481 for (; offset < num_coefs; offset++) {
00482 code = get_vlc2(gb, vlc->table, VLCBITS, VLCMAX);
00483 if (code > 1) {
00485 offset += run_table[code];
00486 sign = get_bits1(gb) - 1;
00487 iptr[offset & coef_mask] = ilvl[code] ^ sign<<31;
00488 } else if (code == 1) {
00490 break;
00491 } else {
00493 if (!version) {
00494 level = get_bits(gb, coef_nb_bits);
00497 offset += get_bits(gb, frame_len_bits);
00498 } else {
00499 level = ff_wma_get_large_val(gb);
00501 if (get_bits1(gb)) {
00502 if (get_bits1(gb)) {
00503 if (get_bits1(gb)) {
00504 av_log(avctx,AV_LOG_ERROR,
00505 "broken escape sequence\n");
00506 return -1;
00507 } else
00508 offset += get_bits(gb, frame_len_bits) + 4;
00509 } else
00510 offset += get_bits(gb, 2) + 1;
00511 }
00512 }
00513 sign = get_bits1(gb) - 1;
00514 ptr[offset & coef_mask] = (level^sign) - sign;
00515 }
00516 }
00518 if (offset > num_coefs) {
00519 av_log(avctx, AV_LOG_ERROR, "overflow in spectral RLE, ignoring\n");
00520 return -1;
00521 }
00522
00523 return 0;
00524 }
00525