| 1 | /* |
| 2 | * WMA compatible encoder |
| 3 | * Copyright (c) 2007 Michael Niedermayer |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 7 | * FFmpeg is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU Lesser General Public |
| 9 | * License as published by the Free Software Foundation; either |
| 10 | * version 2.1 of the License, or (at your option) any later version. |
| 11 | * |
| 12 | * FFmpeg is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * Lesser General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU Lesser General Public |
| 18 | * License along with FFmpeg; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | */ |
| 21 | |
| 22 | #include "libavutil/attributes.h" |
| 23 | |
| 24 | #include "avcodec.h" |
| 25 | #include "internal.h" |
| 26 | #include "wma.h" |
| 27 | #include "libavutil/avassert.h" |
| 28 | |
| 29 | |
| 30 | static av_cold int encode_init(AVCodecContext *avctx) |
| 31 | { |
| 32 | WMACodecContext *s = avctx->priv_data; |
| 33 | int i, flags1, flags2, block_align; |
| 34 | uint8_t *extradata; |
| 35 | |
| 36 | s->avctx = avctx; |
| 37 | |
| 38 | if (avctx->channels > MAX_CHANNELS) { |
| 39 | av_log(avctx, AV_LOG_ERROR, |
| 40 | "too many channels: got %i, need %i or fewer\n", |
| 41 | avctx->channels, MAX_CHANNELS); |
| 42 | return AVERROR(EINVAL); |
| 43 | } |
| 44 | |
| 45 | if (avctx->sample_rate > 48000) { |
| 46 | av_log(avctx, AV_LOG_ERROR, "sample rate is too high: %d > 48kHz\n", |
| 47 | avctx->sample_rate); |
| 48 | return AVERROR(EINVAL); |
| 49 | } |
| 50 | |
| 51 | if (avctx->bit_rate < 24 * 1000) { |
| 52 | av_log(avctx, AV_LOG_ERROR, |
| 53 | "bitrate too low: got %i, need 24000 or higher\n", |
| 54 | avctx->bit_rate); |
| 55 | return AVERROR(EINVAL); |
| 56 | } |
| 57 | |
| 58 | /* extract flag infos */ |
| 59 | flags1 = 0; |
| 60 | flags2 = 1; |
| 61 | if (avctx->codec->id == AV_CODEC_ID_WMAV1) { |
| 62 | extradata = av_malloc(4); |
| 63 | avctx->extradata_size = 4; |
| 64 | AV_WL16(extradata, flags1); |
| 65 | AV_WL16(extradata + 2, flags2); |
| 66 | } else if (avctx->codec->id == AV_CODEC_ID_WMAV2) { |
| 67 | extradata = av_mallocz(10); |
| 68 | avctx->extradata_size = 10; |
| 69 | AV_WL32(extradata, flags1); |
| 70 | AV_WL16(extradata + 4, flags2); |
| 71 | } else { |
| 72 | av_assert0(0); |
| 73 | } |
| 74 | avctx->extradata = extradata; |
| 75 | s->use_exp_vlc = flags2 & 0x0001; |
| 76 | s->use_bit_reservoir = flags2 & 0x0002; |
| 77 | s->use_variable_block_len = flags2 & 0x0004; |
| 78 | if (avctx->channels == 2) |
| 79 | s->ms_stereo = 1; |
| 80 | |
| 81 | ff_wma_init(avctx, flags2); |
| 82 | |
| 83 | /* init MDCT */ |
| 84 | for (i = 0; i < s->nb_block_sizes; i++) |
| 85 | ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 0, 1.0); |
| 86 | |
| 87 | block_align = avctx->bit_rate * (int64_t) s->frame_len / |
| 88 | (avctx->sample_rate * 8); |
| 89 | block_align = FFMIN(block_align, MAX_CODED_SUPERFRAME_SIZE); |
| 90 | avctx->block_align = block_align; |
| 91 | |
| 92 | avctx->frame_size = |
| 93 | avctx->delay = s->frame_len; |
| 94 | |
| 95 | return 0; |
| 96 | } |
| 97 | |
| 98 | static void apply_window_and_mdct(AVCodecContext *avctx, const AVFrame *frame) |
| 99 | { |
| 100 | WMACodecContext *s = avctx->priv_data; |
| 101 | float **audio = (float **) frame->extended_data; |
| 102 | int len = frame->nb_samples; |
| 103 | int window_index = s->frame_len_bits - s->block_len_bits; |
| 104 | FFTContext *mdct = &s->mdct_ctx[window_index]; |
| 105 | int ch; |
| 106 | const float *win = s->windows[window_index]; |
| 107 | int window_len = 1 << s->block_len_bits; |
| 108 | float n = 2.0 * 32768.0 / window_len; |
| 109 | |
| 110 | for (ch = 0; ch < avctx->channels; ch++) { |
| 111 | memcpy(s->output, s->frame_out[ch], window_len * sizeof(*s->output)); |
| 112 | s->fdsp.vector_fmul_scalar(s->frame_out[ch], audio[ch], n, len); |
| 113 | s->fdsp.vector_fmul_reverse(&s->output[window_len], s->frame_out[ch], |
| 114 | win, len); |
| 115 | s->fdsp.vector_fmul(s->frame_out[ch], s->frame_out[ch], win, len); |
| 116 | mdct->mdct_calc(mdct, s->coefs[ch], s->output); |
| 117 | } |
| 118 | } |
| 119 | |
| 120 | // FIXME use for decoding too |
| 121 | static void init_exp(WMACodecContext *s, int ch, const int *exp_param) |
| 122 | { |
| 123 | int n; |
| 124 | const uint16_t *ptr; |
| 125 | float v, *q, max_scale, *q_end; |
| 126 | |
| 127 | ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits]; |
| 128 | q = s->exponents[ch]; |
| 129 | q_end = q + s->block_len; |
| 130 | max_scale = 0; |
| 131 | while (q < q_end) { |
| 132 | /* XXX: use a table */ |
| 133 | v = pow(10, *exp_param++ *(1.0 / 16.0)); |
| 134 | max_scale = FFMAX(max_scale, v); |
| 135 | n = *ptr++; |
| 136 | do { |
| 137 | *q++ = v; |
| 138 | } while (--n); |
| 139 | } |
| 140 | s->max_exponent[ch] = max_scale; |
| 141 | } |
| 142 | |
| 143 | static void encode_exp_vlc(WMACodecContext *s, int ch, const int *exp_param) |
| 144 | { |
| 145 | int last_exp; |
| 146 | const uint16_t *ptr; |
| 147 | float *q, *q_end; |
| 148 | |
| 149 | ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits]; |
| 150 | q = s->exponents[ch]; |
| 151 | q_end = q + s->block_len; |
| 152 | if (s->version == 1) { |
| 153 | last_exp = *exp_param++; |
| 154 | av_assert0(last_exp - 10 >= 0 && last_exp - 10 < 32); |
| 155 | put_bits(&s->pb, 5, last_exp - 10); |
| 156 | q += *ptr++; |
| 157 | } else |
| 158 | last_exp = 36; |
| 159 | while (q < q_end) { |
| 160 | int exp = *exp_param++; |
| 161 | int code = exp - last_exp + 60; |
| 162 | av_assert1(code >= 0 && code < 120); |
| 163 | put_bits(&s->pb, ff_aac_scalefactor_bits[code], |
| 164 | ff_aac_scalefactor_code[code]); |
| 165 | /* XXX: use a table */ |
| 166 | q += *ptr++; |
| 167 | last_exp = exp; |
| 168 | } |
| 169 | } |
| 170 | |
| 171 | static int encode_block(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], |
| 172 | int total_gain) |
| 173 | { |
| 174 | int v, bsize, ch, coef_nb_bits, parse_exponents; |
| 175 | float mdct_norm; |
| 176 | int nb_coefs[MAX_CHANNELS]; |
| 177 | static const int fixed_exp[25] = { |
| 178 | 20, 20, 20, 20, 20, |
| 179 | 20, 20, 20, 20, 20, |
| 180 | 20, 20, 20, 20, 20, |
| 181 | 20, 20, 20, 20, 20, |
| 182 | 20, 20, 20, 20, 20 |
| 183 | }; |
| 184 | |
| 185 | // FIXME remove duplication relative to decoder |
| 186 | if (s->use_variable_block_len) { |
| 187 | av_assert0(0); // FIXME not implemented |
| 188 | } else { |
| 189 | /* fixed block len */ |
| 190 | s->next_block_len_bits = s->frame_len_bits; |
| 191 | s->prev_block_len_bits = s->frame_len_bits; |
| 192 | s->block_len_bits = s->frame_len_bits; |
| 193 | } |
| 194 | |
| 195 | s->block_len = 1 << s->block_len_bits; |
| 196 | // av_assert0((s->block_pos + s->block_len) <= s->frame_len); |
| 197 | bsize = s->frame_len_bits - s->block_len_bits; |
| 198 | |
| 199 | // FIXME factor |
| 200 | v = s->coefs_end[bsize] - s->coefs_start; |
| 201 | for (ch = 0; ch < s->avctx->channels; ch++) |
| 202 | nb_coefs[ch] = v; |
| 203 | { |
| 204 | int n4 = s->block_len / 2; |
| 205 | mdct_norm = 1.0 / (float) n4; |
| 206 | if (s->version == 1) |
| 207 | mdct_norm *= sqrt(n4); |
| 208 | } |
| 209 | |
| 210 | if (s->avctx->channels == 2) |
| 211 | put_bits(&s->pb, 1, !!s->ms_stereo); |
| 212 | |
| 213 | for (ch = 0; ch < s->avctx->channels; ch++) { |
| 214 | // FIXME only set channel_coded when needed, instead of always |
| 215 | s->channel_coded[ch] = 1; |
| 216 | if (s->channel_coded[ch]) |
| 217 | init_exp(s, ch, fixed_exp); |
| 218 | } |
| 219 | |
| 220 | for (ch = 0; ch < s->avctx->channels; ch++) { |
| 221 | if (s->channel_coded[ch]) { |
| 222 | WMACoef *coefs1; |
| 223 | float *coefs, *exponents, mult; |
| 224 | int i, n; |
| 225 | |
| 226 | coefs1 = s->coefs1[ch]; |
| 227 | exponents = s->exponents[ch]; |
| 228 | mult = pow(10, total_gain * 0.05) / s->max_exponent[ch]; |
| 229 | mult *= mdct_norm; |
| 230 | coefs = src_coefs[ch]; |
| 231 | if (s->use_noise_coding && 0) { |
| 232 | av_assert0(0); // FIXME not implemented |
| 233 | } else { |
| 234 | coefs += s->coefs_start; |
| 235 | n = nb_coefs[ch]; |
| 236 | for (i = 0; i < n; i++) { |
| 237 | double t = *coefs++ / (exponents[i] * mult); |
| 238 | if (t < -32768 || t > 32767) |
| 239 | return -1; |
| 240 | |
| 241 | coefs1[i] = lrint(t); |
| 242 | } |
| 243 | } |
| 244 | } |
| 245 | } |
| 246 | |
| 247 | v = 0; |
| 248 | for (ch = 0; ch < s->avctx->channels; ch++) { |
| 249 | int a = s->channel_coded[ch]; |
| 250 | put_bits(&s->pb, 1, a); |
| 251 | v |= a; |
| 252 | } |
| 253 | |
| 254 | if (!v) |
| 255 | return 1; |
| 256 | |
| 257 | for (v = total_gain - 1; v >= 127; v -= 127) |
| 258 | put_bits(&s->pb, 7, 127); |
| 259 | put_bits(&s->pb, 7, v); |
| 260 | |
| 261 | coef_nb_bits = ff_wma_total_gain_to_bits(total_gain); |
| 262 | |
| 263 | if (s->use_noise_coding) { |
| 264 | for (ch = 0; ch < s->avctx->channels; ch++) { |
| 265 | if (s->channel_coded[ch]) { |
| 266 | int i, n; |
| 267 | n = s->exponent_high_sizes[bsize]; |
| 268 | for (i = 0; i < n; i++) { |
| 269 | put_bits(&s->pb, 1, s->high_band_coded[ch][i] = 0); |
| 270 | if (0) |
| 271 | nb_coefs[ch] -= s->exponent_high_bands[bsize][i]; |
| 272 | } |
| 273 | } |
| 274 | } |
| 275 | } |
| 276 | |
| 277 | parse_exponents = 1; |
| 278 | if (s->block_len_bits != s->frame_len_bits) |
| 279 | put_bits(&s->pb, 1, parse_exponents); |
| 280 | |
| 281 | if (parse_exponents) { |
| 282 | for (ch = 0; ch < s->avctx->channels; ch++) { |
| 283 | if (s->channel_coded[ch]) { |
| 284 | if (s->use_exp_vlc) { |
| 285 | encode_exp_vlc(s, ch, fixed_exp); |
| 286 | } else { |
| 287 | av_assert0(0); // FIXME not implemented |
| 288 | // encode_exp_lsp(s, ch); |
| 289 | } |
| 290 | } |
| 291 | } |
| 292 | } else |
| 293 | av_assert0(0); // FIXME not implemented |
| 294 | |
| 295 | for (ch = 0; ch < s->avctx->channels; ch++) { |
| 296 | if (s->channel_coded[ch]) { |
| 297 | int run, tindex; |
| 298 | WMACoef *ptr, *eptr; |
| 299 | tindex = (ch == 1 && s->ms_stereo); |
| 300 | ptr = &s->coefs1[ch][0]; |
| 301 | eptr = ptr + nb_coefs[ch]; |
| 302 | |
| 303 | run = 0; |
| 304 | for (; ptr < eptr; ptr++) { |
| 305 | if (*ptr) { |
| 306 | int level = *ptr; |
| 307 | int abs_level = FFABS(level); |
| 308 | int code = 0; |
| 309 | if (abs_level <= s->coef_vlcs[tindex]->max_level) |
| 310 | if (run < s->coef_vlcs[tindex]->levels[abs_level - 1]) |
| 311 | code = run + s->int_table[tindex][abs_level - 1]; |
| 312 | |
| 313 | av_assert2(code < s->coef_vlcs[tindex]->n); |
| 314 | put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code], |
| 315 | s->coef_vlcs[tindex]->huffcodes[code]); |
| 316 | |
| 317 | if (code == 0) { |
| 318 | if (1 << coef_nb_bits <= abs_level) |
| 319 | return -1; |
| 320 | |
| 321 | put_bits(&s->pb, coef_nb_bits, abs_level); |
| 322 | put_bits(&s->pb, s->frame_len_bits, run); |
| 323 | } |
| 324 | // FIXME the sign is flipped somewhere |
| 325 | put_bits(&s->pb, 1, level < 0); |
| 326 | run = 0; |
| 327 | } else |
| 328 | run++; |
| 329 | } |
| 330 | if (run) |
| 331 | put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1], |
| 332 | s->coef_vlcs[tindex]->huffcodes[1]); |
| 333 | } |
| 334 | if (s->version == 1 && s->avctx->channels >= 2) |
| 335 | avpriv_align_put_bits(&s->pb); |
| 336 | } |
| 337 | return 0; |
| 338 | } |
| 339 | |
| 340 | static int encode_frame(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], |
| 341 | uint8_t *buf, int buf_size, int total_gain) |
| 342 | { |
| 343 | init_put_bits(&s->pb, buf, buf_size); |
| 344 | |
| 345 | if (s->use_bit_reservoir) |
| 346 | av_assert0(0); // FIXME not implemented |
| 347 | else if (encode_block(s, src_coefs, total_gain) < 0) |
| 348 | return INT_MAX; |
| 349 | |
| 350 | avpriv_align_put_bits(&s->pb); |
| 351 | |
| 352 | return put_bits_count(&s->pb) / 8 - s->avctx->block_align; |
| 353 | } |
| 354 | |
| 355 | static int encode_superframe(AVCodecContext *avctx, AVPacket *avpkt, |
| 356 | const AVFrame *frame, int *got_packet_ptr) |
| 357 | { |
| 358 | WMACodecContext *s = avctx->priv_data; |
| 359 | int i, total_gain, ret, error; |
| 360 | |
| 361 | s->block_len_bits = s->frame_len_bits; // required by non variable block len |
| 362 | s->block_len = 1 << s->block_len_bits; |
| 363 | |
| 364 | apply_window_and_mdct(avctx, frame); |
| 365 | |
| 366 | if (s->ms_stereo) { |
| 367 | float a, b; |
| 368 | int i; |
| 369 | |
| 370 | for (i = 0; i < s->block_len; i++) { |
| 371 | a = s->coefs[0][i] * 0.5; |
| 372 | b = s->coefs[1][i] * 0.5; |
| 373 | s->coefs[0][i] = a + b; |
| 374 | s->coefs[1][i] = a - b; |
| 375 | } |
| 376 | } |
| 377 | |
| 378 | if ((ret = ff_alloc_packet2(avctx, avpkt, 2 * MAX_CODED_SUPERFRAME_SIZE)) < 0) |
| 379 | return ret; |
| 380 | |
| 381 | total_gain = 128; |
| 382 | for (i = 64; i; i >>= 1) { |
| 383 | error = encode_frame(s, s->coefs, avpkt->data, avpkt->size, |
| 384 | total_gain - i); |
| 385 | if (error <= 0) |
| 386 | total_gain -= i; |
| 387 | } |
| 388 | |
| 389 | while(total_gain <= 128 && error > 0) |
| 390 | error = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain++); |
| 391 | if (error > 0) { |
| 392 | av_log(avctx, AV_LOG_ERROR, "Invalid input data or requested bitrate too low, cannot encode\n"); |
| 393 | avpkt->size = 0; |
| 394 | return AVERROR(EINVAL); |
| 395 | } |
| 396 | av_assert0((put_bits_count(&s->pb) & 7) == 0); |
| 397 | i= avctx->block_align - (put_bits_count(&s->pb)+7)/8; |
| 398 | av_assert0(i>=0); |
| 399 | while(i--) |
| 400 | put_bits(&s->pb, 8, 'N'); |
| 401 | |
| 402 | flush_put_bits(&s->pb); |
| 403 | av_assert0(put_bits_ptr(&s->pb) - s->pb.buf == avctx->block_align); |
| 404 | |
| 405 | if (frame->pts != AV_NOPTS_VALUE) |
| 406 | avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay); |
| 407 | |
| 408 | avpkt->size = avctx->block_align; |
| 409 | *got_packet_ptr = 1; |
| 410 | return 0; |
| 411 | } |
| 412 | |
| 413 | #if CONFIG_WMAV1_ENCODER |
| 414 | AVCodec ff_wmav1_encoder = { |
| 415 | .name = "wmav1", |
| 416 | .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 1"), |
| 417 | .type = AVMEDIA_TYPE_AUDIO, |
| 418 | .id = AV_CODEC_ID_WMAV1, |
| 419 | .priv_data_size = sizeof(WMACodecContext), |
| 420 | .init = encode_init, |
| 421 | .encode2 = encode_superframe, |
| 422 | .close = ff_wma_end, |
| 423 | .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP, |
| 424 | AV_SAMPLE_FMT_NONE }, |
| 425 | }; |
| 426 | #endif |
| 427 | #if CONFIG_WMAV2_ENCODER |
| 428 | AVCodec ff_wmav2_encoder = { |
| 429 | .name = "wmav2", |
| 430 | .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 2"), |
| 431 | .type = AVMEDIA_TYPE_AUDIO, |
| 432 | .id = AV_CODEC_ID_WMAV2, |
| 433 | .priv_data_size = sizeof(WMACodecContext), |
| 434 | .init = encode_init, |
| 435 | .encode2 = encode_superframe, |
| 436 | .close = ff_wma_end, |
| 437 | .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP, |
| 438 | AV_SAMPLE_FMT_NONE }, |
| 439 | }; |
| 440 | #endif |