00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00048 #include "libavutil/audioconvert.h"
00049 #include "avcodec.h"
00050 #include "get_bits.h"
00051 #include "bytestream.h"
00052 #include "unary.h"
00053 #include "mathops.h"
00054
00055 #define ALAC_EXTRADATA_SIZE 36
00056 #define MAX_CHANNELS 8
00057
00058 typedef struct {
00059 AVCodecContext *avctx;
00060 AVFrame frame;
00061 GetBitContext gb;
00062 int channels;
00063
00064 int32_t *predict_error_buffer[2];
00065 int32_t *output_samples_buffer[2];
00066 int32_t *extra_bits_buffer[2];
00067
00068 uint32_t max_samples_per_frame;
00069 uint8_t sample_size;
00070 uint8_t rice_history_mult;
00071 uint8_t rice_initial_history;
00072 uint8_t rice_limit;
00073
00074 int extra_bits;
00075 int nb_samples;
00077 int direct_output;
00078 } ALACContext;
00079
00080 enum RawDataBlockType {
00081
00082 TYPE_SCE,
00083 TYPE_CPE,
00084 TYPE_CCE,
00085 TYPE_LFE,
00086 TYPE_DSE,
00087 TYPE_PCE,
00088 TYPE_FIL,
00089 TYPE_END
00090 };
00091
00092 static const uint8_t alac_channel_layout_offsets[8][8] = {
00093 { 0 },
00094 { 0, 1 },
00095 { 2, 0, 1 },
00096 { 2, 0, 1, 3 },
00097 { 2, 0, 1, 3, 4 },
00098 { 2, 0, 1, 4, 5, 3 },
00099 { 2, 0, 1, 4, 5, 6, 3 },
00100 { 2, 6, 7, 0, 1, 4, 5, 3 }
00101 };
00102
00103 static const uint16_t alac_channel_layouts[8] = {
00104 AV_CH_LAYOUT_MONO,
00105 AV_CH_LAYOUT_STEREO,
00106 AV_CH_LAYOUT_SURROUND,
00107 AV_CH_LAYOUT_4POINT0,
00108 AV_CH_LAYOUT_5POINT0_BACK,
00109 AV_CH_LAYOUT_5POINT1_BACK,
00110 AV_CH_LAYOUT_6POINT1_BACK,
00111 AV_CH_LAYOUT_7POINT1_WIDE_BACK
00112 };
00113
00114 static inline unsigned int decode_scalar(GetBitContext *gb, int k, int bps)
00115 {
00116 unsigned int x = get_unary_0_9(gb);
00117
00118 if (x > 8) {
00119
00120 x = get_bits_long(gb, bps);
00121 } else if (k != 1) {
00122 int extrabits = show_bits(gb, k);
00123
00124
00125 x = (x << k) - x;
00126
00127 if (extrabits > 1) {
00128 x += extrabits - 1;
00129 skip_bits(gb, k);
00130 } else
00131 skip_bits(gb, k - 1);
00132 }
00133 return x;
00134 }
00135
00136 static int rice_decompress(ALACContext *alac, int32_t *output_buffer,
00137 int nb_samples, int bps, int rice_history_mult)
00138 {
00139 int i;
00140 unsigned int history = alac->rice_initial_history;
00141 int sign_modifier = 0;
00142
00143 for (i = 0; i < nb_samples; i++) {
00144 int k;
00145 unsigned int x;
00146
00147 if(get_bits_left(&alac->gb) <= 0)
00148 return -1;
00149
00150
00151 k = av_log2((history >> 9) + 3);
00152 k = FFMIN(k, alac->rice_limit);
00153 x = decode_scalar(&alac->gb, k, bps);
00154 x += sign_modifier;
00155 sign_modifier = 0;
00156 output_buffer[i] = (x >> 1) ^ -(x & 1);
00157
00158
00159 if (x > 0xffff)
00160 history = 0xffff;
00161 else
00162 history += x * rice_history_mult -
00163 ((history * rice_history_mult) >> 9);
00164
00165
00166 if ((history < 128) && (i + 1 < nb_samples)) {
00167 int block_size;
00168
00169
00170 k = 7 - av_log2(history) + ((history + 16) >> 6);
00171 k = FFMIN(k, alac->rice_limit);
00172 block_size = decode_scalar(&alac->gb, k, 16);
00173
00174 if (block_size > 0) {
00175 if (block_size >= nb_samples - i) {
00176 av_log(alac->avctx, AV_LOG_ERROR,
00177 "invalid zero block size of %d %d %d\n", block_size,
00178 nb_samples, i);
00179 block_size = nb_samples - i - 1;
00180 }
00181 memset(&output_buffer[i + 1], 0,
00182 block_size * sizeof(*output_buffer));
00183 i += block_size;
00184 }
00185 if (block_size <= 0xffff)
00186 sign_modifier = 1;
00187 history = 0;
00188 }
00189 }
00190 return 0;
00191 }
00192
00193 static inline int sign_only(int v)
00194 {
00195 return v ? FFSIGN(v) : 0;
00196 }
00197
00198 static void lpc_prediction(int32_t *error_buffer, int32_t *buffer_out,
00199 int nb_samples, int bps, int16_t *lpc_coefs,
00200 int lpc_order, int lpc_quant)
00201 {
00202 int i;
00203 int32_t *pred = buffer_out;
00204
00205
00206 *buffer_out = *error_buffer;
00207
00208 if (nb_samples <= 1)
00209 return;
00210
00211 if (!lpc_order) {
00212 memcpy(&buffer_out[1], &error_buffer[1],
00213 (nb_samples - 1) * sizeof(*buffer_out));
00214 return;
00215 }
00216
00217 if (lpc_order == 31) {
00218
00219 for (i = 1; i < nb_samples; i++) {
00220 buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i],
00221 bps);
00222 }
00223 return;
00224 }
00225
00226
00227 for (i = 1; i <= lpc_order; i++)
00228 buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i], bps);
00229
00230
00231
00232 for (; i < nb_samples; i++) {
00233 int j;
00234 int val = 0;
00235 int error_val = error_buffer[i];
00236 int error_sign;
00237 int d = *pred++;
00238
00239
00240 for (j = 0; j < lpc_order; j++)
00241 val += (pred[j] - d) * lpc_coefs[j];
00242 val = (val + (1 << (lpc_quant - 1))) >> lpc_quant;
00243 val += d + error_val;
00244 buffer_out[i] = sign_extend(val, bps);
00245
00246
00247 error_sign = sign_only(error_val);
00248 if (error_sign) {
00249 for (j = 0; j < lpc_order && error_val * error_sign > 0; j++) {
00250 int sign;
00251 val = d - pred[j];
00252 sign = sign_only(val) * error_sign;
00253 lpc_coefs[j] -= sign;
00254 val *= sign;
00255 error_val -= (val >> lpc_quant) * (j + 1);
00256 }
00257 }
00258 }
00259 }
00260
00261 static void decorrelate_stereo(int32_t *buffer[2], int nb_samples,
00262 int decorr_shift, int decorr_left_weight)
00263 {
00264 int i;
00265
00266 for (i = 0; i < nb_samples; i++) {
00267 int32_t a, b;
00268
00269 a = buffer[0][i];
00270 b = buffer[1][i];
00271
00272 a -= (b * decorr_left_weight) >> decorr_shift;
00273 b += a;
00274
00275 buffer[0][i] = b;
00276 buffer[1][i] = a;
00277 }
00278 }
00279
00280 static void append_extra_bits(int32_t *buffer[2], int32_t *extra_bits_buffer[2],
00281 int extra_bits, int channels, int nb_samples)
00282 {
00283 int i, ch;
00284
00285 for (ch = 0; ch < channels; ch++)
00286 for (i = 0; i < nb_samples; i++)
00287 buffer[ch][i] = (buffer[ch][i] << extra_bits) | extra_bits_buffer[ch][i];
00288 }
00289
00290 static int decode_element(AVCodecContext *avctx, void *data, int ch_index,
00291 int channels)
00292 {
00293 ALACContext *alac = avctx->priv_data;
00294 int has_size, bps, is_compressed, decorr_shift, decorr_left_weight, ret;
00295 uint32_t output_samples;
00296 int i, ch;
00297
00298 skip_bits(&alac->gb, 4);
00299 skip_bits(&alac->gb, 12);
00300
00301
00302 has_size = get_bits1(&alac->gb);
00303
00304 alac->extra_bits = get_bits(&alac->gb, 2) << 3;
00305 bps = alac->sample_size - alac->extra_bits + channels - 1;
00306 if (bps > 32) {
00307 av_log(avctx, AV_LOG_ERROR, "bps is unsupported: %d\n", bps);
00308 return AVERROR_PATCHWELCOME;
00309 }
00310
00311
00312 is_compressed = !get_bits1(&alac->gb);
00313
00314 if (has_size)
00315 output_samples = get_bits_long(&alac->gb, 32);
00316 else
00317 output_samples = alac->max_samples_per_frame;
00318 if (!output_samples || output_samples > alac->max_samples_per_frame) {
00319 av_log(avctx, AV_LOG_ERROR, "invalid samples per frame: %d\n",
00320 output_samples);
00321 return AVERROR_INVALIDDATA;
00322 }
00323 if (!alac->nb_samples) {
00324
00325 alac->frame.nb_samples = output_samples;
00326 if ((ret = avctx->get_buffer(avctx, &alac->frame)) < 0) {
00327 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
00328 return ret;
00329 }
00330 } else if (output_samples != alac->nb_samples) {
00331 av_log(avctx, AV_LOG_ERROR, "sample count mismatch: %u != %d\n",
00332 output_samples, alac->nb_samples);
00333 return AVERROR_INVALIDDATA;
00334 }
00335 alac->nb_samples = output_samples;
00336 if (alac->direct_output) {
00337 for (ch = 0; ch < channels; ch++)
00338 alac->output_samples_buffer[ch] = (int32_t *)alac->frame.extended_data[ch_index + ch];
00339 }
00340
00341 if (is_compressed) {
00342 int16_t lpc_coefs[2][32];
00343 int lpc_order[2];
00344 int prediction_type[2];
00345 int lpc_quant[2];
00346 int rice_history_mult[2];
00347
00348 decorr_shift = get_bits(&alac->gb, 8);
00349 decorr_left_weight = get_bits(&alac->gb, 8);
00350
00351 for (ch = 0; ch < channels; ch++) {
00352 prediction_type[ch] = get_bits(&alac->gb, 4);
00353 lpc_quant[ch] = get_bits(&alac->gb, 4);
00354 rice_history_mult[ch] = get_bits(&alac->gb, 3);
00355 lpc_order[ch] = get_bits(&alac->gb, 5);
00356
00357
00358 for (i = lpc_order[ch] - 1; i >= 0; i--)
00359 lpc_coefs[ch][i] = get_sbits(&alac->gb, 16);
00360 }
00361
00362 if (alac->extra_bits) {
00363 for (i = 0; i < alac->nb_samples; i++) {
00364 if(get_bits_left(&alac->gb) <= 0)
00365 return -1;
00366 for (ch = 0; ch < channels; ch++)
00367 alac->extra_bits_buffer[ch][i] = get_bits(&alac->gb, alac->extra_bits);
00368 }
00369 }
00370 for (ch = 0; ch < channels; ch++) {
00371 int ret=rice_decompress(alac, alac->predict_error_buffer[ch],
00372 alac->nb_samples, bps,
00373 rice_history_mult[ch] * alac->rice_history_mult / 4);
00374 if(ret<0)
00375 return ret;
00376
00377
00378 if (prediction_type[ch] == 15) {
00379
00380
00381
00382
00383
00384
00385
00386 lpc_prediction(alac->predict_error_buffer[ch],
00387 alac->predict_error_buffer[ch],
00388 alac->nb_samples, bps, NULL, 31, 0);
00389 } else if (prediction_type[ch] > 0) {
00390 av_log(avctx, AV_LOG_WARNING, "unknown prediction type: %i\n",
00391 prediction_type[ch]);
00392 }
00393 lpc_prediction(alac->predict_error_buffer[ch],
00394 alac->output_samples_buffer[ch], alac->nb_samples,
00395 bps, lpc_coefs[ch], lpc_order[ch], lpc_quant[ch]);
00396 }
00397 } else {
00398
00399 for (i = 0; i < alac->nb_samples; i++) {
00400 if(get_bits_left(&alac->gb) <= 0)
00401 return -1;
00402 for (ch = 0; ch < channels; ch++) {
00403 alac->output_samples_buffer[ch][i] =
00404 get_sbits_long(&alac->gb, alac->sample_size);
00405 }
00406 }
00407 alac->extra_bits = 0;
00408 decorr_shift = 0;
00409 decorr_left_weight = 0;
00410 }
00411
00412 if (channels == 2 && decorr_left_weight) {
00413 decorrelate_stereo(alac->output_samples_buffer, alac->nb_samples,
00414 decorr_shift, decorr_left_weight);
00415 }
00416
00417 if (alac->extra_bits) {
00418 append_extra_bits(alac->output_samples_buffer, alac->extra_bits_buffer,
00419 alac->extra_bits, channels, alac->nb_samples);
00420 }
00421
00422 if(av_sample_fmt_is_planar(avctx->sample_fmt)) {
00423 switch(alac->sample_size) {
00424 case 16: {
00425 for (ch = 0; ch < channels; ch++) {
00426 int16_t *outbuffer = (int16_t *)alac->frame.extended_data[ch_index + ch];
00427 for (i = 0; i < alac->nb_samples; i++)
00428 *outbuffer++ = alac->output_samples_buffer[ch][i];
00429 }}
00430 break;
00431 case 24: {
00432 for (ch = 0; ch < channels; ch++) {
00433 for (i = 0; i < alac->nb_samples; i++)
00434 alac->output_samples_buffer[ch][i] <<= 8;
00435 }}
00436 break;
00437 }
00438 }else{
00439 switch(alac->sample_size) {
00440 case 16: {
00441 int16_t *outbuffer = ((int16_t *)alac->frame.extended_data[0]) + ch_index;
00442 for (i = 0; i < alac->nb_samples; i++) {
00443 for (ch = 0; ch < channels; ch++)
00444 *outbuffer++ = alac->output_samples_buffer[ch][i];
00445 outbuffer += alac->channels - channels;
00446 }
00447 }
00448 break;
00449 case 24: {
00450 int32_t *outbuffer = ((int32_t *)alac->frame.extended_data[0]) + ch_index;
00451 for (i = 0; i < alac->nb_samples; i++) {
00452 for (ch = 0; ch < channels; ch++)
00453 *outbuffer++ = alac->output_samples_buffer[ch][i] << 8;
00454 outbuffer += alac->channels - channels;
00455 }
00456 }
00457 break;
00458 case 32: {
00459 int32_t *outbuffer = ((int32_t *)alac->frame.extended_data[0]) + ch_index;
00460 for (i = 0; i < alac->nb_samples; i++) {
00461 for (ch = 0; ch < channels; ch++)
00462 *outbuffer++ = alac->output_samples_buffer[ch][i];
00463 outbuffer += alac->channels - channels;
00464 }
00465 }
00466 break;
00467 }
00468 }
00469
00470 return 0;
00471 }
00472
00473 static int alac_decode_frame(AVCodecContext *avctx, void *data,
00474 int *got_frame_ptr, AVPacket *avpkt)
00475 {
00476 ALACContext *alac = avctx->priv_data;
00477 enum RawDataBlockType element;
00478 int channels;
00479 int ch, ret, got_end;
00480
00481 init_get_bits(&alac->gb, avpkt->data, avpkt->size * 8);
00482
00483 got_end = 0;
00484 alac->nb_samples = 0;
00485 ch = 0;
00486 while (get_bits_left(&alac->gb) >= 3) {
00487 element = get_bits(&alac->gb, 3);
00488 if (element == TYPE_END) {
00489 got_end = 1;
00490 break;
00491 }
00492 if (element > TYPE_CPE && element != TYPE_LFE) {
00493 av_log(avctx, AV_LOG_ERROR, "syntax element unsupported: %d\n", element);
00494 return AVERROR_PATCHWELCOME;
00495 }
00496
00497 channels = (element == TYPE_CPE) ? 2 : 1;
00498 if (ch + channels > alac->channels) {
00499 av_log(avctx, AV_LOG_ERROR, "invalid element channel count\n");
00500 return AVERROR_INVALIDDATA;
00501 }
00502
00503 ret = decode_element(avctx, data,
00504 alac_channel_layout_offsets[alac->channels - 1][ch],
00505 channels);
00506 if (ret < 0 && get_bits_left(&alac->gb))
00507 return ret;
00508
00509 ch += channels;
00510 }
00511 if (!got_end) {
00512 av_log(avctx, AV_LOG_ERROR, "no end tag found. incomplete packet.\n");
00513 return AVERROR_INVALIDDATA;
00514 }
00515
00516 if (avpkt->size * 8 - get_bits_count(&alac->gb) > 8) {
00517 av_log(avctx, AV_LOG_ERROR, "Error : %d bits left\n",
00518 avpkt->size * 8 - get_bits_count(&alac->gb));
00519 }
00520
00521 *got_frame_ptr = 1;
00522 *(AVFrame *)data = alac->frame;
00523
00524 return avpkt->size;
00525 }
00526
00527 static av_cold int alac_decode_close(AVCodecContext *avctx)
00528 {
00529 ALACContext *alac = avctx->priv_data;
00530
00531 int ch;
00532 for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
00533 av_freep(&alac->predict_error_buffer[ch]);
00534 if (!alac->direct_output)
00535 av_freep(&alac->output_samples_buffer[ch]);
00536 av_freep(&alac->extra_bits_buffer[ch]);
00537 }
00538
00539 return 0;
00540 }
00541
00542 static int allocate_buffers(ALACContext *alac)
00543 {
00544 int ch;
00545 int buf_size = alac->max_samples_per_frame * sizeof(int32_t);
00546
00547 for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
00548 FF_ALLOC_OR_GOTO(alac->avctx, alac->predict_error_buffer[ch],
00549 buf_size, buf_alloc_fail);
00550
00551 alac->direct_output = alac->sample_size > 16 && av_sample_fmt_is_planar(alac->avctx->sample_fmt);
00552 if (!alac->direct_output) {
00553 FF_ALLOC_OR_GOTO(alac->avctx, alac->output_samples_buffer[ch],
00554 buf_size, buf_alloc_fail);
00555 }
00556
00557 FF_ALLOC_OR_GOTO(alac->avctx, alac->extra_bits_buffer[ch],
00558 buf_size, buf_alloc_fail);
00559 }
00560 return 0;
00561 buf_alloc_fail:
00562 alac_decode_close(alac->avctx);
00563 return AVERROR(ENOMEM);
00564 }
00565
00566 static int alac_set_info(ALACContext *alac)
00567 {
00568 GetByteContext gb;
00569
00570 bytestream2_init(&gb, alac->avctx->extradata,
00571 alac->avctx->extradata_size);
00572
00573 bytestream2_skipu(&gb, 12);
00574
00575 alac->max_samples_per_frame = bytestream2_get_be32u(&gb);
00576 if (!alac->max_samples_per_frame || alac->max_samples_per_frame > INT_MAX) {
00577 av_log(alac->avctx, AV_LOG_ERROR, "max samples per frame invalid: %u\n",
00578 alac->max_samples_per_frame);
00579 return AVERROR_INVALIDDATA;
00580 }
00581 bytestream2_skipu(&gb, 1);
00582 alac->sample_size = bytestream2_get_byteu(&gb);
00583 alac->rice_history_mult = bytestream2_get_byteu(&gb);
00584 alac->rice_initial_history = bytestream2_get_byteu(&gb);
00585 alac->rice_limit = bytestream2_get_byteu(&gb);
00586 alac->channels = bytestream2_get_byteu(&gb);
00587 bytestream2_get_be16u(&gb);
00588 bytestream2_get_be32u(&gb);
00589 bytestream2_get_be32u(&gb);
00590 bytestream2_get_be32u(&gb);
00591
00592 return 0;
00593 }
00594
00595 static av_cold int alac_decode_init(AVCodecContext * avctx)
00596 {
00597 int ret;
00598 int req_packed;
00599 ALACContext *alac = avctx->priv_data;
00600 alac->avctx = avctx;
00601
00602
00603 if (alac->avctx->extradata_size != ALAC_EXTRADATA_SIZE) {
00604 av_log(avctx, AV_LOG_ERROR, "expected %d extradata bytes\n",
00605 ALAC_EXTRADATA_SIZE);
00606 return -1;
00607 }
00608 if (alac_set_info(alac)) {
00609 av_log(avctx, AV_LOG_ERROR, "set_info failed\n");
00610 return -1;
00611 }
00612
00613 req_packed = LIBAVCODEC_VERSION_MAJOR < 55 && !av_sample_fmt_is_planar(avctx->request_sample_fmt);
00614 switch (alac->sample_size) {
00615 case 16: avctx->sample_fmt = req_packed ? AV_SAMPLE_FMT_S16 : AV_SAMPLE_FMT_S16P;
00616 break;
00617 case 24:
00618 case 32: avctx->sample_fmt = req_packed ? AV_SAMPLE_FMT_S32 : AV_SAMPLE_FMT_S32P;
00619 break;
00620 default: av_log_ask_for_sample(avctx, "Sample depth %d is not supported.\n",
00621 alac->sample_size);
00622 return AVERROR_PATCHWELCOME;
00623 }
00624
00625 if (alac->channels < 1) {
00626 av_log(avctx, AV_LOG_WARNING, "Invalid channel count\n");
00627 alac->channels = avctx->channels;
00628 } else {
00629 if (alac->channels > MAX_CHANNELS)
00630 alac->channels = avctx->channels;
00631 else
00632 avctx->channels = alac->channels;
00633 }
00634 if (avctx->channels > MAX_CHANNELS) {
00635 av_log(avctx, AV_LOG_ERROR, "Unsupported channel count: %d\n",
00636 avctx->channels);
00637 return AVERROR_PATCHWELCOME;
00638 }
00639 avctx->channel_layout = alac_channel_layouts[alac->channels - 1];
00640
00641 if ((ret = allocate_buffers(alac)) < 0) {
00642 av_log(avctx, AV_LOG_ERROR, "Error allocating buffers\n");
00643 return ret;
00644 }
00645
00646 avcodec_get_frame_defaults(&alac->frame);
00647 avctx->coded_frame = &alac->frame;
00648
00649 return 0;
00650 }
00651
00652 AVCodec ff_alac_decoder = {
00653 .name = "alac",
00654 .type = AVMEDIA_TYPE_AUDIO,
00655 .id = AV_CODEC_ID_ALAC,
00656 .priv_data_size = sizeof(ALACContext),
00657 .init = alac_decode_init,
00658 .close = alac_decode_close,
00659 .decode = alac_decode_frame,
00660 .capabilities = CODEC_CAP_DR1,
00661 .long_name = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
00662 };