* @author 2005 David Hammerton
* @see http://crazney.net/programs/itunes/alac.html
*
- * Note: This decoder expects a 36- (0x24-)byte QuickTime atom to be
+ * Note: This decoder expects a 36-byte QuickTime atom to be
* passed through the extradata[_size] fields. This atom is tacked onto
* the end of an 'alac' stsd atom and has the following format:
- * bytes 0-3 atom size (0x24), big-endian
- * bytes 4-7 atom type ('alac', not the 'alac' tag from start of stsd)
- * bytes 8-35 data bytes needed by decoder
*
- * Extradata:
- * 32bit size
- * 32bit tag (=alac)
- * 32bit zero?
- * 32bit max sample per frame
- * 8bit ?? (zero?)
+ * 32bit atom size
+ * 32bit tag ("alac")
+ * 32bit tag version (0)
+ * 32bit samples per frame (used when not set explicitly in the frames)
+ * 8bit compatible version (0)
* 8bit sample size
- * 8bit history mult
- * 8bit initial history
- * 8bit kmodifier
- * 8bit channels?
- * 16bit ??
- * 32bit max coded frame size
- * 32bit bitrate?
+ * 8bit history mult (40)
+ * 8bit initial history (14)
+ * 8bit rice param limit (10)
+ * 8bit channels
+ * 16bit maxRun (255)
+ * 32bit max coded frame size (0 means unknown)
+ * 32bit average bitrate (0 means unknown)
* 32bit samplerate
*/
typedef struct {
AVCodecContext *avctx;
+ AVFrame frame;
GetBitContext gb;
- int numchannels;
- int bytespersample;
+ int channels;
/* buffers */
- int32_t *predicterror_buffer[MAX_CHANNELS];
-
- int32_t *outputsamples_buffer[MAX_CHANNELS];
-
+ int32_t *predict_error_buffer[MAX_CHANNELS];
+ int32_t *output_samples_buffer[MAX_CHANNELS];
int32_t *extra_bits_buffer[MAX_CHANNELS];
- /* stuff from setinfo */
- uint32_t setinfo_max_samples_per_frame; /* 0x1000 = 4096 */ /* max samples per frame? */
- uint8_t setinfo_sample_size; /* 0x10 */
- uint8_t setinfo_rice_historymult; /* 0x28 */
- uint8_t setinfo_rice_initialhistory; /* 0x0a */
- uint8_t setinfo_rice_kmodifier; /* 0x0e */
- /* end setinfo stuff */
+ uint32_t max_samples_per_frame;
+ uint8_t sample_size;
+ uint8_t rice_history_mult;
+ uint8_t rice_initial_history;
+ uint8_t rice_limit;
int extra_bits; /**< number of extra bits beyond 16-bit */
} ALACContext;
-static inline int decode_scalar(GetBitContext *gb, int k, int limit, int readsamplesize){
- /* read x - number of 1s before 0 represent the rice */
+static inline int decode_scalar(GetBitContext *gb, int k, int readsamplesize)
+{
int x = get_unary_0_9(gb);
if (x > 8) { /* RICE THRESHOLD */
/* use alternative encoding */
x = get_bits(gb, readsamplesize);
- } else {
- if (k >= limit)
- k = limit;
-
- if (k != 1) {
+ } else if (k != 1) {
int extrabits = show_bits(gb, k);
/* multiply x by 2^k - 1, as part of their strange algorithm */
skip_bits(gb, k);
} else
skip_bits(gb, k - 1);
- }
}
return x;
}
static void bastardized_rice_decompress(ALACContext *alac,
- int32_t *output_buffer,
- int output_size,
- int readsamplesize, /* arg_10 */
- int rice_initialhistory, /* arg424->b */
- int rice_kmodifier, /* arg424->d */
- int rice_historymult, /* arg424->c */
- int rice_kmodifier_mask /* arg424->e */
- )
+ int32_t *output_buffer,
+ int output_size,
+ int readsamplesize,
+ int rice_history_mult)
{
int output_count;
- unsigned int history = rice_initialhistory;
+ unsigned int history = alac->rice_initial_history;
int sign_modifier = 0;
for (output_count = 0; output_count < output_size; output_count++) {
/* read k, that is bits as is */
k = av_log2((history >> 9) + 3);
- x= decode_scalar(&alac->gb, k, rice_kmodifier, readsamplesize);
+ k = FFMIN(k, alac->rice_limit);
+ x = decode_scalar(&alac->gb, k, readsamplesize);
x_modified = sign_modifier + x;
final_val = (x_modified + 1) / 2;
sign_modifier = 0;
/* now update the history */
- history += x_modified * rice_historymult
- - ((history * rice_historymult) >> 9);
+ history += x_modified * rice_history_mult -
+ ((history * rice_history_mult) >> 9);
if (x_modified > 0xffff)
history = 0xffff;
sign_modifier = 1;
k = 7 - av_log2(history) + ((history + 16) >> 6 /* / 64 */);
+ k = FFMIN(k, alac->rice_limit);
- block_size= decode_scalar(&alac->gb, k, rice_kmodifier, 16);
+ block_size = decode_scalar(&alac->gb, k, 16);
if (block_size > 0) {
if(block_size >= output_size - output_count){
return;
}
- if (predictor_coef_num == 0x1f) { /* 11111 - max value of predictor_coef_num */
- /* second-best case scenario for fir decompression,
- * error describes a small difference from the previous sample only
- */
+ if (predictor_coef_num == 31) {
+ /* simple 1st-order prediction */
if (output_size <= 1)
return;
for (i = 0; i < output_size - 1; i++) {
buffer_out[i+1] = val;
}
- /* 4 and 8 are very common cases (the only ones i've seen). these
- * should be unrolled and optimized
- */
+ /* NOTE: 4 and 8 are very common cases that could be optimized. */
/* general case */
if (predictor_coef_num > 0) {
}
}
-static int alac_decode_frame(AVCodecContext *avctx,
- void *outbuffer, int *outputsize,
- AVPacket *avpkt)
+static int alac_decode_frame(AVCodecContext *avctx, void *data,
+ int *got_frame_ptr, AVPacket *avpkt)
{
const uint8_t *inbuffer = avpkt->data;
int input_buffer_size = avpkt->size;
int isnotcompressed;
uint8_t interlacing_shift;
uint8_t interlacing_leftweight;
- int i, ch;
-
- /* short-circuit null buffers */
- if (!inbuffer || !input_buffer_size)
- return -1;
+ int i, ch, ret;
init_get_bits(&alac->gb, inbuffer, input_buffer_size * 8);
return AVERROR_INVALIDDATA;
}
- /* 2^result = something to do with output waiting.
- * perhaps matters if we read > 1 frame in a pass?
- */
- skip_bits(&alac->gb, 4);
-
- skip_bits(&alac->gb, 12); /* unknown, skip 12 bits */
+ skip_bits(&alac->gb, 4); /* element instance tag */
+ skip_bits(&alac->gb, 12); /* unused header bits */
- /* the output sample size is stored soon */
+ /* the number of output samples is stored in the frame */
hassize = get_bits1(&alac->gb);
alac->extra_bits = get_bits(&alac->gb, 2) << 3;
if (hassize) {
/* now read the number of samples as a 32bit integer */
outputsamples = get_bits_long(&alac->gb, 32);
- if(outputsamples > alac->setinfo_max_samples_per_frame){
- av_log(avctx, AV_LOG_ERROR, "outputsamples %d > %d\n", outputsamples, alac->setinfo_max_samples_per_frame);
+ if (outputsamples > alac->max_samples_per_frame) {
+ av_log(avctx, AV_LOG_ERROR, "outputsamples %d > %d\n",
+ outputsamples, alac->max_samples_per_frame);
return -1;
}
} else
- outputsamples = alac->setinfo_max_samples_per_frame;
+ outputsamples = alac->max_samples_per_frame;
- alac->bytespersample = channels * av_get_bytes_per_sample(avctx->sample_fmt);
-
- if(outputsamples > *outputsize / alac->bytespersample){
- av_log(avctx, AV_LOG_ERROR, "sample buffer too small\n");
- return -1;
+ /* get output buffer */
+ if (outputsamples > INT32_MAX) {
+ av_log(avctx, AV_LOG_ERROR, "unsupported block size: %u\n", outputsamples);
+ return AVERROR_INVALIDDATA;
+ }
+ alac->frame.nb_samples = outputsamples;
+ if ((ret = avctx->get_buffer(avctx, &alac->frame)) < 0) {
+ av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
+ return ret;
}
- *outputsize = outputsamples * alac->bytespersample;
- readsamplesize = alac->setinfo_sample_size - alac->extra_bits + channels - 1;
+ readsamplesize = alac->sample_size - alac->extra_bits + channels - 1;
if (readsamplesize > MIN_CACHE_BITS) {
av_log(avctx, AV_LOG_ERROR, "readsamplesize too big (%d)\n", readsamplesize);
return -1;
}
for (ch = 0; ch < channels; ch++) {
bastardized_rice_decompress(alac,
- alac->predicterror_buffer[ch],
+ alac->predict_error_buffer[ch],
outputsamples,
readsamplesize,
- alac->setinfo_rice_initialhistory,
- alac->setinfo_rice_kmodifier,
- ricemodifier[ch] * alac->setinfo_rice_historymult / 4,
- (1 << alac->setinfo_rice_kmodifier) - 1);
-
- if (prediction_type[ch] == 0) {
- /* adaptive fir */
- predictor_decompress_fir_adapt(alac->predicterror_buffer[ch],
- alac->outputsamples_buffer[ch],
- outputsamples,
- readsamplesize,
- predictor_coef_table[ch],
- predictor_coef_num[ch],
- prediction_quantitization[ch]);
- } else {
- av_log(avctx, AV_LOG_ERROR, "FIXME: unhandled prediction type: %i\n", prediction_type[ch]);
- /* I think the only other prediction type (or perhaps this is
- * just a boolean?) runs adaptive fir twice.. like:
- * predictor_decompress_fir_adapt(predictor_error, tempout, ...)
- * predictor_decompress_fir_adapt(predictor_error, outputsamples ...)
- * little strange..
+ ricemodifier[ch] * alac->rice_history_mult / 4);
+
+ /* adaptive FIR filter */
+ if (prediction_type[ch] == 15) {
+ /* Prediction type 15 runs the adaptive FIR twice.
+ * The first pass uses the special-case coef_num = 31, while
+ * the second pass uses the coefs from the bitstream.
+ *
+ * However, this prediction type is not currently used by the
+ * reference encoder.
*/
+ predictor_decompress_fir_adapt(alac->predict_error_buffer[ch],
+ alac->predict_error_buffer[ch],
+ outputsamples, readsamplesize,
+ NULL, 31, 0);
+ } else if (prediction_type[ch] > 0) {
+ av_log(avctx, AV_LOG_WARNING, "unknown prediction type: %i\n",
+ prediction_type[ch]);
}
+ predictor_decompress_fir_adapt(alac->predict_error_buffer[ch],
+ alac->output_samples_buffer[ch],
+ outputsamples, readsamplesize,
+ predictor_coef_table[ch],
+ predictor_coef_num[ch],
+ prediction_quantitization[ch]);
}
} else {
/* not compressed, easy case */
for (i = 0; i < outputsamples; i++) {
for (ch = 0; ch < channels; ch++) {
- alac->outputsamples_buffer[ch][i] = get_sbits_long(&alac->gb,
- alac->setinfo_sample_size);
+ alac->output_samples_buffer[ch][i] = get_sbits_long(&alac->gb,
+ alac->sample_size);
}
}
alac->extra_bits = 0;
av_log(avctx, AV_LOG_ERROR, "Error : Wrong End Of Frame\n");
if (channels == 2 && interlacing_leftweight) {
- decorrelate_stereo(alac->outputsamples_buffer, outputsamples,
+ decorrelate_stereo(alac->output_samples_buffer, outputsamples,
interlacing_shift, interlacing_leftweight);
}
if (alac->extra_bits) {
- append_extra_bits(alac->outputsamples_buffer, alac->extra_bits_buffer,
- alac->extra_bits, alac->numchannels, outputsamples);
+ append_extra_bits(alac->output_samples_buffer, alac->extra_bits_buffer,
+ alac->extra_bits, alac->channels, outputsamples);
}
- switch(alac->setinfo_sample_size) {
+ switch(alac->sample_size) {
case 16:
if (channels == 2) {
- interleave_stereo_16(alac->outputsamples_buffer, outbuffer,
- outputsamples);
+ interleave_stereo_16(alac->output_samples_buffer,
+ (int16_t *)alac->frame.data[0], outputsamples);
} else {
+ int16_t *outbuffer = (int16_t *)alac->frame.data[0];
for (i = 0; i < outputsamples; i++) {
- ((int16_t*)outbuffer)[i] = alac->outputsamples_buffer[0][i];
+ outbuffer[i] = alac->output_samples_buffer[0][i];
}
}
break;
case 24:
if (channels == 2) {
- interleave_stereo_24(alac->outputsamples_buffer, outbuffer,
- outputsamples);
+ interleave_stereo_24(alac->output_samples_buffer,
+ (int32_t *)alac->frame.data[0], outputsamples);
} else {
+ int32_t *outbuffer = (int32_t *)alac->frame.data[0];
for (i = 0; i < outputsamples; i++)
- ((int32_t *)outbuffer)[i] = alac->outputsamples_buffer[0][i] << 8;
+ outbuffer[i] = alac->output_samples_buffer[0][i] << 8;
}
break;
}
if (input_buffer_size * 8 - get_bits_count(&alac->gb) > 8)
av_log(avctx, AV_LOG_ERROR, "Error : %d bits left\n", input_buffer_size * 8 - get_bits_count(&alac->gb));
+ *got_frame_ptr = 1;
+ *(AVFrame *)data = alac->frame;
+
return input_buffer_size;
}
ALACContext *alac = avctx->priv_data;
int ch;
- for (ch = 0; ch < alac->numchannels; ch++) {
- av_freep(&alac->predicterror_buffer[ch]);
- av_freep(&alac->outputsamples_buffer[ch]);
+ for (ch = 0; ch < alac->channels; ch++) {
+ av_freep(&alac->predict_error_buffer[ch]);
+ av_freep(&alac->output_samples_buffer[ch]);
av_freep(&alac->extra_bits_buffer[ch]);
}
static int allocate_buffers(ALACContext *alac)
{
int ch;
- for (ch = 0; ch < alac->numchannels; ch++) {
- int buf_size = alac->setinfo_max_samples_per_frame * sizeof(int32_t);
+ for (ch = 0; ch < alac->channels; ch++) {
+ int buf_size = alac->max_samples_per_frame * sizeof(int32_t);
- FF_ALLOC_OR_GOTO(alac->avctx, alac->predicterror_buffer[ch],
+ FF_ALLOC_OR_GOTO(alac->avctx, alac->predict_error_buffer[ch],
buf_size, buf_alloc_fail);
- FF_ALLOC_OR_GOTO(alac->avctx, alac->outputsamples_buffer[ch],
+ FF_ALLOC_OR_GOTO(alac->avctx, alac->output_samples_buffer[ch],
buf_size, buf_alloc_fail);
FF_ALLOC_OR_GOTO(alac->avctx, alac->extra_bits_buffer[ch],
static int alac_set_info(ALACContext *alac)
{
- const unsigned char *ptr = alac->avctx->extradata;
+ GetByteContext gb;
- ptr += 4; /* size */
- ptr += 4; /* alac */
- ptr += 4; /* 0 ? */
+ bytestream2_init(&gb, alac->avctx->extradata,
+ alac->avctx->extradata_size);
- if(AV_RB32(ptr) >= UINT_MAX/4){
- av_log(alac->avctx, AV_LOG_ERROR, "setinfo_max_samples_per_frame too large\n");
- return -1;
- }
+ bytestream2_skipu(&gb, 12); // size:4, alac:4, version:4
- /* buffer size / 2 ? */
- alac->setinfo_max_samples_per_frame = bytestream_get_be32(&ptr);
- ptr++; /* ??? */
- alac->setinfo_sample_size = *ptr++;
- alac->setinfo_rice_historymult = *ptr++;
- alac->setinfo_rice_initialhistory = *ptr++;
- alac->setinfo_rice_kmodifier = *ptr++;
- alac->numchannels = *ptr++;
- bytestream_get_be16(&ptr); /* ??? */
- bytestream_get_be32(&ptr); /* max coded frame size */
- bytestream_get_be32(&ptr); /* bitrate ? */
- bytestream_get_be32(&ptr); /* samplerate */
+ alac->max_samples_per_frame = bytestream2_get_be32u(&gb);
+ if (alac->max_samples_per_frame >= UINT_MAX/4){
+ av_log(alac->avctx, AV_LOG_ERROR,
+ "max_samples_per_frame too large\n");
+ return AVERROR_INVALIDDATA;
+ }
+ bytestream2_skipu(&gb, 1); // compatible version
+ alac->sample_size = bytestream2_get_byteu(&gb);
+ alac->rice_history_mult = bytestream2_get_byteu(&gb);
+ alac->rice_initial_history = bytestream2_get_byteu(&gb);
+ alac->rice_limit = bytestream2_get_byteu(&gb);
+ alac->channels = bytestream2_get_byteu(&gb);
+ bytestream2_get_be16u(&gb); // maxRun
+ bytestream2_get_be32u(&gb); // max coded frame size
+ bytestream2_get_be32u(&gb); // average bitrate
+ bytestream2_get_be32u(&gb); // samplerate
return 0;
}
return -1;
}
- switch (alac->setinfo_sample_size) {
+ switch (alac->sample_size) {
case 16: avctx->sample_fmt = AV_SAMPLE_FMT_S16;
break;
case 24: avctx->sample_fmt = AV_SAMPLE_FMT_S32;
break;
default: av_log_ask_for_sample(avctx, "Sample depth %d is not supported.\n",
- alac->setinfo_sample_size);
+ alac->sample_size);
return AVERROR_PATCHWELCOME;
}
- if (alac->numchannels < 1) {
+ if (alac->channels < 1) {
av_log(avctx, AV_LOG_WARNING, "Invalid channel count\n");
- alac->numchannels = avctx->channels;
+ alac->channels = avctx->channels;
} else {
- if (alac->numchannels > MAX_CHANNELS)
- alac->numchannels = avctx->channels;
+ if (alac->channels > MAX_CHANNELS)
+ alac->channels = avctx->channels;
else
- avctx->channels = alac->numchannels;
+ avctx->channels = alac->channels;
}
if (avctx->channels > MAX_CHANNELS) {
av_log(avctx, AV_LOG_ERROR, "Unsupported channel count: %d\n",
return ret;
}
+ avcodec_get_frame_defaults(&alac->frame);
+ avctx->coded_frame = &alac->frame;
+
return 0;
}
.init = alac_decode_init,
.close = alac_decode_close,
.decode = alac_decode_frame,
- .long_name = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
+ .capabilities = CODEC_CAP_DR1,
+ .long_name = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
};
projects / ffmpeg / blobdiff