/*
- * Atrac 3 compatible decoder
+ * ATRAC3 compatible decoder
* Copyright (c) 2006-2008 Maxim Poliakovski
* Copyright (c) 2006-2008 Benjamin Larsson
*
/**
* @file
- * Atrac 3 compatible decoder.
+ * ATRAC3 compatible decoder.
* This decoder handles Sony's ATRAC3 data.
*
- * Container formats used to store atrac 3 data:
+ * Container formats used to store ATRAC3 data:
* RealMedia (.rm), RIFF WAV (.wav, .at3), Sony OpenMG (.oma, .aa3).
*
* To use this decoder, a calling application must supply the extradata
#include <stddef.h>
#include <stdio.h>
+#include "libavutil/attributes.h"
#include "libavutil/float_dsp.h"
#include "avcodec.h"
#include "bytestream.h"
#include "fft.h"
#include "fmtconvert.h"
#include "get_bits.h"
+#include "internal.h"
#include "atrac.h"
#include "atrac3data.h"
#define SAMPLES_PER_FRAME 1024
#define MDCT_SIZE 512
-typedef struct GainInfo {
- int num_gain_data;
- int lev_code[8];
- int loc_code[8];
-} GainInfo;
-
typedef struct GainBlock {
- GainInfo g_block[4];
+ AtracGainInfo g_block[4];
} GainBlock;
typedef struct TonalComponent {
} ChannelUnit;
typedef struct ATRAC3Context {
- AVFrame frame;
GetBitContext gb;
//@{
/** stream data */
int coding_mode;
- int bit_rate;
- int sample_rate;
- int samples_per_channel;
- int samples_per_frame;
ChannelUnit *units;
//@}
//@{
/** extradata */
int scrambled_stream;
- int frame_factor;
//@}
+ AtracGCContext gainc_ctx;
FFTContext mdct_ctx;
FmtConvertContext fmt_conv;
AVFloatDSPContext fdsp;
} ATRAC3Context;
static DECLARE_ALIGNED(32, float, mdct_window)[MDCT_SIZE];
+static VLC_TYPE atrac3_vlc_table[4096][2];
static VLC spectral_coeff_tab[7];
-static float gain_tab1[16];
-static float gain_tab2[31];
-
-/*
+/**
* Regular 512 points IMDCT without overlapping, with the exception of the
* swapping of odd bands caused by the reverse spectra of the QMF.
*
off = (intptr_t)input & 3;
buf = (const uint32_t *)(input - off);
- c = av_be2ne32((0x537F6103 >> (off * 8)) | (0x537F6103 << (32 - (off * 8))));
+ if (off)
+ c = av_be2ne32((0x537F6103U >> (off * 8)) | (0x537F6103U << (32 - (off * 8))));
+ else
+ c = av_be2ne32(0x537F6103U);
bytes += 3 + off;
for (i = 0; i < bytes / 4; i++)
output[i] = c ^ buf[i];
if (off)
- av_log_ask_for_sample(NULL, "Offset of %d not handled.\n", off);
+ avpriv_request_sample(NULL, "Offset of %d", off);
return off;
}
-static av_cold int init_atrac3_transforms(ATRAC3Context *q)
+static av_cold void init_imdct_window(void)
{
- float enc_window[256];
- int i;
+ int i, j;
/* generate the mdct window, for details see
* http://wiki.multimedia.cx/index.php?title=RealAudio_atrc#Windows */
- for (i = 0; i < 256; i++)
- enc_window[i] = (sin(((i + 0.5) / 256.0 - 0.5) * M_PI) + 1.0) * 0.5;
-
- if (!mdct_window[0]) {
- for (i = 0; i < 256; i++) {
- mdct_window[i] = enc_window[i] /
- (enc_window[ i] * enc_window[ i] +
- enc_window[255 - i] * enc_window[255 - i]);
- mdct_window[511 - i] = mdct_window[i];
- }
+ for (i = 0, j = 255; i < 128; i++, j--) {
+ float wi = sin(((i + 0.5) / 256.0 - 0.5) * M_PI) + 1.0;
+ float wj = sin(((j + 0.5) / 256.0 - 0.5) * M_PI) + 1.0;
+ float w = 0.5 * (wi * wi + wj * wj);
+ mdct_window[i] = mdct_window[511 - i] = wi / w;
+ mdct_window[j] = mdct_window[511 - j] = wj / w;
}
-
- /* initialize the MDCT transform */
- return ff_mdct_init(&q->mdct_ctx, 9, 1, 1.0 / 32768);
}
static av_cold int atrac3_decode_close(AVCodecContext *avctx)
return 0;
}
-/*
+/**
* Mantissa decoding
*
* @param selector which table the output values are coded with
}
}
-/*
+/**
* Restore the quantized band spectrum coefficients
*
* @return subband count, fix for broken specification/files
output[first] = mantissas[j] * scale_factor;
} else {
/* this subband was not coded, so zero the entire subband */
- memset(output + first, 0, subband_size * sizeof(float));
+ memset(output + first, 0, subband_size * sizeof(*output));
}
}
/* clear the subbands that were not coded */
first = subband_tab[i];
- memset(output + first, 0, (SAMPLES_PER_FRAME - first) * sizeof(float));
+ memset(output + first, 0, (SAMPLES_PER_FRAME - first) * sizeof(*output));
return num_subbands;
}
-/*
+/**
* Restore the quantized tonal components
*
* @param components tonal components
return component_count;
}
-/*
+/**
* Decode gain parameters for the coded bands
*
* @param block the gainblock for the current band
static int decode_gain_control(GetBitContext *gb, GainBlock *block,
int num_bands)
{
- int i, cf, num_data;
+ int i, j;
int *level, *loc;
- GainInfo *gain = block->g_block;
+ AtracGainInfo *gain = block->g_block;
for (i = 0; i <= num_bands; i++) {
- num_data = get_bits(gb, 3);
- gain[i].num_gain_data = num_data;
+ gain[i].num_points = get_bits(gb, 3);
level = gain[i].lev_code;
loc = gain[i].loc_code;
- for (cf = 0; cf < gain[i].num_gain_data; cf++) {
- level[cf] = get_bits(gb, 4);
- loc [cf] = get_bits(gb, 5);
- if (cf && loc[cf] <= loc[cf - 1])
+ for (j = 0; j < gain[i].num_points; j++) {
+ level[j] = get_bits(gb, 4);
+ loc[j] = get_bits(gb, 5);
+ if (j && loc[j] <= loc[j - 1])
return AVERROR_INVALIDDATA;
}
}
/* Clear the unused blocks. */
for (; i < 4 ; i++)
- gain[i].num_gain_data = 0;
+ gain[i].num_points = 0;
return 0;
}
-/*
- * Apply gain parameters and perform the MDCT overlapping part
- *
- * @param input input buffer
- * @param prev previous buffer to perform overlap against
- * @param output output buffer
- * @param gain1 current band gain info
- * @param gain2 next band gain info
- */
-static void gain_compensate_and_overlap(float *input, float *prev,
- float *output, GainInfo *gain1,
- GainInfo *gain2)
-{
- float g1, g2, gain_inc;
- int i, j, num_data, start_loc, end_loc;
-
-
- if (gain2->num_gain_data == 0)
- g1 = 1.0;
- else
- g1 = gain_tab1[gain2->lev_code[0]];
-
- if (gain1->num_gain_data == 0) {
- for (i = 0; i < 256; i++)
- output[i] = input[i] * g1 + prev[i];
- } else {
- num_data = gain1->num_gain_data;
- gain1->loc_code[num_data] = 32;
- gain1->lev_code[num_data] = 4;
-
- for (i = 0, j = 0; i < num_data; i++) {
- start_loc = gain1->loc_code[i] * 8;
- end_loc = start_loc + 8;
-
- g2 = gain_tab1[gain1->lev_code[i]];
- gain_inc = gain_tab2[gain1->lev_code[i + 1] -
- gain1->lev_code[i ] + 15];
-
- /* interpolate */
- for (; j < start_loc; j++)
- output[j] = (input[j] * g1 + prev[j]) * g2;
-
- /* interpolation is done over eight samples */
- for (; j < end_loc; j++) {
- output[j] = (input[j] * g1 + prev[j]) * g2;
- g2 *= gain_inc;
- }
- }
-
- for (; j < 256; j++)
- output[j] = input[j] * g1 + prev[j];
- }
-
- /* Delay for the overlapping part. */
- memcpy(prev, &input[256], 256 * sizeof(float));
-}
-
-/*
+/**
* Combine the tonal band spectrum and regular band spectrum
*
* @param spectrum output spectrum buffer
output = &spectrum[components[i].pos];
for (j = 0; j < components[i].num_coefs; j++)
- output[i] += input[i];
+ output[j] += input[j];
}
return last_pos;
}
}
-/*
+/**
* Decode a Sound Unit
*
* @param snd the channel unit to be used
snd->num_components = decode_tonal_components(gb, snd->components,
snd->bands_coded);
- if (snd->num_components == -1)
- return -1;
+ if (snd->num_components < 0)
+ return snd->num_components;
num_subbands = decode_spectrum(gb, snd->spectrum);
if (band <= num_bands)
imlt(q, &snd->spectrum[band * 256], snd->imdct_buf, band & 1);
else
- memset(snd->imdct_buf, 0, 512 * sizeof(float));
+ memset(snd->imdct_buf, 0, 512 * sizeof(*snd->imdct_buf));
/* gain compensation and overlapping */
- gain_compensate_and_overlap(snd->imdct_buf,
- &snd->prev_frame[band * 256],
- &output[band * 256],
- &gain1->g_block[band],
- &gain2->g_block[band]);
+ ff_atrac_gain_compensation(&q->gainc_ctx, snd->imdct_buf,
+ &snd->prev_frame[band * 256],
+ &gain1->g_block[band], &gain2->g_block[band],
+ 256, &output[band * 256]);
}
/* Swap the gain control buffers for the next frame. */
/* set the bitstream reader at the start of the second Sound Unit*/
- init_get_bits(&q->gb, ptr1, avctx->block_align * 8);
+ init_get_bits(&q->gb, ptr1, (avctx->block_align - i) * 8);
/* Fill the Weighting coeffs delay buffer */
- memmove(q->weighting_delay, &q->weighting_delay[2], 4 * sizeof(int));
+ memmove(q->weighting_delay, &q->weighting_delay[2],
+ 4 * sizeof(*q->weighting_delay));
q->weighting_delay[4] = get_bits1(&q->gb);
q->weighting_delay[5] = get_bits(&q->gb, 3);
static int atrac3_decode_frame(AVCodecContext *avctx, void *data,
int *got_frame_ptr, AVPacket *avpkt)
{
+ AVFrame *frame = data;
const uint8_t *buf = avpkt->data;
int buf_size = avpkt->size;
ATRAC3Context *q = avctx->priv_data;
}
/* get output buffer */
- q->frame.nb_samples = SAMPLES_PER_FRAME;
- if ((ret = avctx->get_buffer(avctx, &q->frame)) < 0) {
+ frame->nb_samples = SAMPLES_PER_FRAME;
+ if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
return ret;
}
databuf = buf;
}
- ret = decode_frame(avctx, databuf, (float **)q->frame.extended_data);
+ ret = decode_frame(avctx, databuf, (float **)frame->extended_data);
if (ret) {
av_log(NULL, AV_LOG_ERROR, "Frame decoding error!\n");
return ret;
}
- *got_frame_ptr = 1;
- *(AVFrame *)data = q->frame;
+ *got_frame_ptr = 1;
return avctx->block_align;
}
+static av_cold void atrac3_init_static_data(AVCodec *codec)
+{
+ int i;
+
+ init_imdct_window();
+ ff_atrac_generate_tables();
+
+ /* Initialize the VLC tables. */
+ for (i = 0; i < 7; i++) {
+ spectral_coeff_tab[i].table = &atrac3_vlc_table[atrac3_vlc_offs[i]];
+ spectral_coeff_tab[i].table_allocated = atrac3_vlc_offs[i + 1] -
+ atrac3_vlc_offs[i ];
+ init_vlc(&spectral_coeff_tab[i], 9, huff_tab_sizes[i],
+ huff_bits[i], 1, 1,
+ huff_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
+ }
+}
+
static av_cold int atrac3_decode_init(AVCodecContext *avctx)
{
int i, ret;
- int version, delay;
+ int version, delay, samples_per_frame, frame_factor;
const uint8_t *edata_ptr = avctx->extradata;
ATRAC3Context *q = avctx->priv_data;
- static VLC_TYPE atrac3_vlc_table[4096][2];
- static int vlcs_initialized = 0;
-
- /* Take data from the AVCodecContext (RM container). */
- q->sample_rate = avctx->sample_rate;
- q->bit_rate = avctx->bit_rate;
if (avctx->channels <= 0 || avctx->channels > 2) {
av_log(avctx, AV_LOG_ERROR, "Channel configuration error!\n");
/* Parse the extradata, WAV format */
av_log(avctx, AV_LOG_DEBUG, "[0-1] %d\n",
bytestream_get_le16(&edata_ptr)); // Unknown value always 1
- q->samples_per_channel = bytestream_get_le32(&edata_ptr);
+ edata_ptr += 4; // samples per channel
q->coding_mode = bytestream_get_le16(&edata_ptr);
av_log(avctx, AV_LOG_DEBUG,"[8-9] %d\n",
bytestream_get_le16(&edata_ptr)); //Dupe of coding mode
- q->frame_factor = bytestream_get_le16(&edata_ptr); // Unknown always 1
+ frame_factor = bytestream_get_le16(&edata_ptr); // Unknown always 1
av_log(avctx, AV_LOG_DEBUG,"[12-13] %d\n",
bytestream_get_le16(&edata_ptr)); // Unknown always 0
/* setup */
- q->samples_per_frame = SAMPLES_PER_FRAME * avctx->channels;
+ samples_per_frame = SAMPLES_PER_FRAME * avctx->channels;
version = 4;
delay = 0x88E;
q->coding_mode = q->coding_mode ? JOINT_STEREO : STEREO;
q->scrambled_stream = 0;
- if (avctx->block_align != 96 * avctx->channels * q->frame_factor &&
- avctx->block_align != 152 * avctx->channels * q->frame_factor &&
- avctx->block_align != 192 * avctx->channels * q->frame_factor) {
+ if (avctx->block_align != 96 * avctx->channels * frame_factor &&
+ avctx->block_align != 152 * avctx->channels * frame_factor &&
+ avctx->block_align != 192 * avctx->channels * frame_factor) {
av_log(avctx, AV_LOG_ERROR, "Unknown frame/channel/frame_factor "
"configuration %d/%d/%d\n", avctx->block_align,
- avctx->channels, q->frame_factor);
+ avctx->channels, frame_factor);
return AVERROR_INVALIDDATA;
}
} else if (avctx->extradata_size == 10) {
/* Parse the extradata, RM format. */
version = bytestream_get_be32(&edata_ptr);
- q->samples_per_frame = bytestream_get_be16(&edata_ptr);
+ samples_per_frame = bytestream_get_be16(&edata_ptr);
delay = bytestream_get_be16(&edata_ptr);
q->coding_mode = bytestream_get_be16(&edata_ptr);
- q->samples_per_channel = q->samples_per_frame / avctx->channels;
q->scrambled_stream = 1;
} else {
av_log(NULL, AV_LOG_ERROR, "Unknown extradata size %d.\n",
avctx->extradata_size);
+ return AVERROR(EINVAL);
}
/* Check the extradata */
return AVERROR_INVALIDDATA;
}
- if (q->samples_per_frame != SAMPLES_PER_FRAME &&
- q->samples_per_frame != SAMPLES_PER_FRAME * 2) {
+ if (samples_per_frame != SAMPLES_PER_FRAME &&
+ samples_per_frame != SAMPLES_PER_FRAME * 2) {
av_log(avctx, AV_LOG_ERROR, "Unknown amount of samples per frame %d.\n",
- q->samples_per_frame);
+ samples_per_frame);
return AVERROR_INVALIDDATA;
}
if (q->coding_mode == STEREO)
av_log(avctx, AV_LOG_DEBUG, "Normal stereo detected.\n");
- else if (q->coding_mode == JOINT_STEREO)
+ else if (q->coding_mode == JOINT_STEREO) {
+ if (avctx->channels != 2)
+ return AVERROR_INVALIDDATA;
av_log(avctx, AV_LOG_DEBUG, "Joint stereo detected.\n");
- else {
+ } else {
av_log(avctx, AV_LOG_ERROR, "Unknown channel coding mode %x!\n",
q->coding_mode);
return AVERROR_INVALIDDATA;
if (avctx->block_align >= UINT_MAX / 2)
return AVERROR(EINVAL);
- q->decoded_bytes_buffer = av_mallocz(avctx->block_align +
- (4 - avctx->block_align % 4) +
+ q->decoded_bytes_buffer = av_mallocz(FFALIGN(avctx->block_align, 4) +
FF_INPUT_BUFFER_PADDING_SIZE);
if (q->decoded_bytes_buffer == NULL)
return AVERROR(ENOMEM);
-
- /* Initialize the VLC tables. */
- if (!vlcs_initialized) {
- for (i = 0; i < 7; i++) {
- spectral_coeff_tab[i].table = &atrac3_vlc_table[atrac3_vlc_offs[i]];
- spectral_coeff_tab[i].table_allocated = atrac3_vlc_offs[i + 1] -
- atrac3_vlc_offs[i ];
- init_vlc(&spectral_coeff_tab[i], 9, huff_tab_sizes[i],
- huff_bits[i], 1, 1,
- huff_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
- }
- vlcs_initialized = 1;
- }
-
avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
- if ((ret = init_atrac3_transforms(q))) {
+ /* initialize the MDCT transform */
+ if ((ret = ff_mdct_init(&q->mdct_ctx, 9, 1, 1.0 / 32768)) < 0) {
av_log(avctx, AV_LOG_ERROR, "Error initializing MDCT\n");
av_freep(&q->decoded_bytes_buffer);
return ret;
}
- ff_atrac_generate_tables();
-
- /* Generate gain tables */
- for (i = 0; i < 16; i++)
- gain_tab1[i] = powf(2.0, (4 - i));
-
- for (i = -15; i < 16; i++)
- gain_tab2[i + 15] = powf(2.0, i * -0.125);
-
/* init the joint-stereo decoding data */
q->weighting_delay[0] = 0;
q->weighting_delay[1] = 7;
q->matrix_coeff_index_next[i] = 3;
}
+ ff_atrac_init_gain_compensation(&q->gainc_ctx, 4, 3);
avpriv_float_dsp_init(&q->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
ff_fmt_convert_init(&q->fmt_conv, avctx);
- q->units = av_mallocz(sizeof(ChannelUnit) * avctx->channels);
+ q->units = av_mallocz(sizeof(*q->units) * avctx->channels);
if (!q->units) {
atrac3_decode_close(avctx);
return AVERROR(ENOMEM);
}
- avcodec_get_frame_defaults(&q->frame);
- avctx->coded_frame = &q->frame;
-
return 0;
}
AVCodec ff_atrac3_decoder = {
.name = "atrac3",
+ .long_name = NULL_IF_CONFIG_SMALL("ATRAC3 (Adaptive TRansform Acoustic Coding 3)"),
.type = AVMEDIA_TYPE_AUDIO,
.id = AV_CODEC_ID_ATRAC3,
.priv_data_size = sizeof(ATRAC3Context),
.init = atrac3_decode_init,
+ .init_static_data = atrac3_init_static_data,
.close = atrac3_decode_close,
.decode = atrac3_decode_frame,
.capabilities = CODEC_CAP_SUBFRAMES | CODEC_CAP_DR1,
- .long_name = NULL_IF_CONFIG_SMALL("Atrac 3 (Adaptive TRansform Acoustic Coding 3)"),
.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
AV_SAMPLE_FMT_NONE },
};