/*
- * 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 "bitstream.h"
#include "bytestream.h"
#include "fft.h"
-#include "fmtconvert.h"
-#include "get_bits.h"
#include "internal.h"
+#include "vlc.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 {
- GetBitContext gb;
+ BitstreamContext bc;
//@{
/** stream data */
int coding_mode;
int scrambled_stream;
//@}
+ 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];
return off;
}
-static av_cold void init_atrac3_window(void)
+static av_cold void init_imdct_window(void)
{
int i, j;
return 0;
}
-/*
+/**
* Mantissa decoding
*
* @param selector which table the output values are coded with
* @param mantissas mantissa output table
* @param num_codes number of values to get
*/
-static void read_quant_spectral_coeffs(GetBitContext *gb, int selector,
+static void read_quant_spectral_coeffs(BitstreamContext *bc, int selector,
int coding_flag, int *mantissas,
int num_codes)
{
if (selector > 1) {
for (i = 0; i < num_codes; i++) {
if (num_bits)
- code = get_sbits(gb, num_bits);
+ code = bitstream_read_signed(bc, num_bits);
else
code = 0;
mantissas[i] = code;
} else {
for (i = 0; i < num_codes; i++) {
if (num_bits)
- code = get_bits(gb, num_bits); // num_bits is always 4 in this case
+ code = bitstream_read(bc, num_bits); // num_bits is always 4 in this case
else
code = 0;
mantissas[i * 2 ] = mantissa_clc_tab[code >> 2];
/* variable length coding (VLC) */
if (selector != 1) {
for (i = 0; i < num_codes; i++) {
- huff_symb = get_vlc2(gb, spectral_coeff_tab[selector-1].table,
- spectral_coeff_tab[selector-1].bits, 3);
+ huff_symb = bitstream_read_vlc(bc, spectral_coeff_tab[selector-1].table,
+ spectral_coeff_tab[selector-1].bits, 3);
huff_symb += 1;
code = huff_symb >> 1;
if (huff_symb & 1)
}
} else {
for (i = 0; i < num_codes; i++) {
- huff_symb = get_vlc2(gb, spectral_coeff_tab[selector - 1].table,
- spectral_coeff_tab[selector - 1].bits, 3);
+ huff_symb = bitstream_read_vlc(bc, spectral_coeff_tab[selector - 1].table,
+ spectral_coeff_tab[selector - 1].bits, 3);
mantissas[i * 2 ] = mantissa_vlc_tab[huff_symb * 2 ];
mantissas[i * 2 + 1] = mantissa_vlc_tab[huff_symb * 2 + 1];
}
}
}
-/*
+/**
* Restore the quantized band spectrum coefficients
*
* @return subband count, fix for broken specification/files
*/
-static int decode_spectrum(GetBitContext *gb, float *output)
+static int decode_spectrum(BitstreamContext *bc, float *output)
{
int num_subbands, coding_mode, i, j, first, last, subband_size;
int subband_vlc_index[32], sf_index[32];
int mantissas[128];
float scale_factor;
- num_subbands = get_bits(gb, 5); // number of coded subbands
- coding_mode = get_bits1(gb); // coding Mode: 0 - VLC/ 1-CLC
+ num_subbands = bitstream_read(bc, 5); // number of coded subbands
+ coding_mode = bitstream_read_bit(bc); // coding Mode: 0 - VLC/ 1 - CLC
/* get the VLC selector table for the subbands, 0 means not coded */
for (i = 0; i <= num_subbands; i++)
- subband_vlc_index[i] = get_bits(gb, 3);
+ subband_vlc_index[i] = bitstream_read(bc, 3);
/* read the scale factor indexes from the stream */
for (i = 0; i <= num_subbands; i++) {
if (subband_vlc_index[i] != 0)
- sf_index[i] = get_bits(gb, 6);
+ sf_index[i] = bitstream_read(bc, 6);
}
for (i = 0; i <= num_subbands; i++) {
/* decode spectral coefficients for this subband */
/* TODO: This can be done faster is several blocks share the
* same VLC selector (subband_vlc_index) */
- read_quant_spectral_coeffs(gb, subband_vlc_index[i], coding_mode,
+ read_quant_spectral_coeffs(bc, subband_vlc_index[i], coding_mode,
mantissas, subband_size);
/* decode the scale factor for this subband */
return num_subbands;
}
-/*
+/**
* Restore the quantized tonal components
*
* @param components tonal components
* @param num_bands number of coded bands
*/
-static int decode_tonal_components(GetBitContext *gb,
+static int decode_tonal_components(BitstreamContext *bc,
TonalComponent *components, int num_bands)
{
int i, b, c, m;
int band_flags[4], mantissa[8];
int component_count = 0;
- nb_components = get_bits(gb, 5);
+ nb_components = bitstream_read(bc, 5);
/* no tonal components */
if (nb_components == 0)
return 0;
- coding_mode_selector = get_bits(gb, 2);
+ coding_mode_selector = bitstream_read(bc, 2);
if (coding_mode_selector == 2)
return AVERROR_INVALIDDATA;
int coded_values_per_component, quant_step_index;
for (b = 0; b <= num_bands; b++)
- band_flags[b] = get_bits1(gb);
+ band_flags[b] = bitstream_read_bit(bc);
- coded_values_per_component = get_bits(gb, 3);
+ coded_values_per_component = bitstream_read(bc, 3);
- quant_step_index = get_bits(gb, 3);
+ quant_step_index = bitstream_read(bc, 3);
if (quant_step_index <= 1)
return AVERROR_INVALIDDATA;
if (coding_mode_selector == 3)
- coding_mode = get_bits1(gb);
+ coding_mode = bitstream_read_bit(bc);
for (b = 0; b < (num_bands + 1) * 4; b++) {
int coded_components;
if (band_flags[b >> 2] == 0)
continue;
- coded_components = get_bits(gb, 3);
+ coded_components = bitstream_read(bc, 3);
for (c = 0; c < coded_components; c++) {
TonalComponent *cmp = &components[component_count];
int sf_index, coded_values, max_coded_values;
float scale_factor;
- sf_index = get_bits(gb, 6);
+ sf_index = bitstream_read(bc, 6);
if (component_count >= 64)
return AVERROR_INVALIDDATA;
- cmp->pos = b * 64 + get_bits(gb, 6);
+ cmp->pos = b * 64 + bitstream_read(bc, 6);
max_coded_values = SAMPLES_PER_FRAME - cmp->pos;
coded_values = coded_values_per_component + 1;
scale_factor = ff_atrac_sf_table[sf_index] *
inv_max_quant[quant_step_index];
- read_quant_spectral_coeffs(gb, quant_step_index, coding_mode,
+ read_quant_spectral_coeffs(bc, quant_step_index, coding_mode,
mantissa, coded_values);
cmp->num_coefs = coded_values;
return component_count;
}
-/*
+/**
* Decode gain parameters for the coded bands
*
* @param block the gainblock for the current band
* @param num_bands amount of coded bands
*/
-static int decode_gain_control(GetBitContext *gb, GainBlock *block,
+static int decode_gain_control(BitstreamContext *bc, 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 = bitstream_read(bc, 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] = bitstream_read(bc, 4);
+ loc[j] = bitstream_read(bc, 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(*prev));
-}
-
-/*
+/**
* Combine the tonal band spectrum and regular band spectrum
*
* @param spectrum output spectrum buffer
}
}
-/*
+/**
* Decode a Sound Unit
*
* @param snd the channel unit to be used
* @param channel_num channel number
* @param coding_mode the coding mode (JOINT_STEREO or regular stereo/mono)
*/
-static int decode_channel_sound_unit(ATRAC3Context *q, GetBitContext *gb,
+static int decode_channel_sound_unit(ATRAC3Context *q, BitstreamContext *bc,
ChannelUnit *snd, float *output,
int channel_num, int coding_mode)
{
GainBlock *gain2 = &snd->gain_block[1 - snd->gc_blk_switch];
if (coding_mode == JOINT_STEREO && channel_num == 1) {
- if (get_bits(gb, 2) != 3) {
+ if (bitstream_read(bc, 2) != 3) {
av_log(NULL,AV_LOG_ERROR,"JS mono Sound Unit id != 3.\n");
return AVERROR_INVALIDDATA;
}
} else {
- if (get_bits(gb, 6) != 0x28) {
+ if (bitstream_read(bc, 6) != 0x28) {
av_log(NULL,AV_LOG_ERROR,"Sound Unit id != 0x28.\n");
return AVERROR_INVALIDDATA;
}
}
/* number of coded QMF bands */
- snd->bands_coded = get_bits(gb, 2);
+ snd->bands_coded = bitstream_read(bc, 2);
- ret = decode_gain_control(gb, gain2, snd->bands_coded);
+ ret = decode_gain_control(bc, gain2, snd->bands_coded);
if (ret)
return ret;
- snd->num_components = decode_tonal_components(gb, snd->components,
+ snd->num_components = decode_tonal_components(bc, 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);
+ num_subbands = decode_spectrum(bc, snd->spectrum);
/* Merge the decoded spectrum and tonal components. */
last_tonal = add_tonal_components(snd->spectrum, snd->num_components,
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. */
if (q->coding_mode == JOINT_STEREO) {
/* channel coupling mode */
/* decode Sound Unit 1 */
- init_get_bits(&q->gb, databuf, avctx->block_align * 8);
+ bitstream_init8(&q->bc, databuf, avctx->block_align);
- ret = decode_channel_sound_unit(q, &q->gb, q->units, out_samples[0], 0,
+ ret = decode_channel_sound_unit(q, &q->bc, q->units, out_samples[0], 0,
JOINT_STEREO);
if (ret != 0)
return ret;
/* set the bitstream reader at the start of the second Sound Unit*/
- init_get_bits(&q->gb, ptr1, avctx->block_align * 8);
+ bitstream_init8(&q->bc, ptr1, avctx->block_align - i);
/* Fill the Weighting coeffs delay buffer */
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);
+ q->weighting_delay[4] = bitstream_read_bit(&q->bc);
+ q->weighting_delay[5] = bitstream_read(&q->bc, 3);
for (i = 0; i < 4; i++) {
q->matrix_coeff_index_prev[i] = q->matrix_coeff_index_now[i];
q->matrix_coeff_index_now[i] = q->matrix_coeff_index_next[i];
- q->matrix_coeff_index_next[i] = get_bits(&q->gb, 2);
+ q->matrix_coeff_index_next[i] = bitstream_read(&q->bc, 2);
}
/* Decode Sound Unit 2. */
- ret = decode_channel_sound_unit(q, &q->gb, &q->units[1],
+ ret = decode_channel_sound_unit(q, &q->bc, &q->units[1],
out_samples[1], 1, JOINT_STEREO);
if (ret != 0)
return ret;
/* Decode the channel sound units. */
for (i = 0; i < avctx->channels; i++) {
/* Set the bitstream reader at the start of a channel sound unit. */
- init_get_bits(&q->gb,
- databuf + i * avctx->block_align / avctx->channels,
- avctx->block_align * 8 / avctx->channels);
+ bitstream_init8(&q->bc,
+ databuf + i * avctx->block_align / avctx->channels,
+ avctx->block_align / avctx->channels);
- ret = decode_channel_sound_unit(q, &q->gb, &q->units[i],
+ ret = decode_channel_sound_unit(q, &q->bc, &q->units[i],
out_samples[i], i, q->coding_mode);
if (ret != 0)
return ret;
return avctx->block_align;
}
-static void atrac3_init_static_data(AVCodec *codec)
+static av_cold void atrac3_init_static_data(AVCodec *codec)
{
int i;
- init_atrac3_window();
+ init_imdct_window();
ff_atrac_generate_tables();
/* Initialize the VLC tables. */
huff_bits[i], 1, 1,
huff_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
}
-
- /* 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);
}
static av_cold int atrac3_decode_init(AVCodecContext *avctx)
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;
return AVERROR(EINVAL);
q->decoded_bytes_buffer = av_mallocz(FFALIGN(avctx->block_align, 4) +
- FF_INPUT_BUFFER_PADDING_SIZE);
- if (q->decoded_bytes_buffer == NULL)
+ AV_INPUT_BUFFER_PADDING_SIZE);
+ if (!q->decoded_bytes_buffer)
return AVERROR(ENOMEM);
avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
q->matrix_coeff_index_next[i] = 3;
}
- avpriv_float_dsp_init(&q->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
- ff_fmt_convert_init(&q->fmt_conv, avctx);
+ ff_atrac_init_gain_compensation(&q->gainc_ctx, 4, 3);
+ avpriv_float_dsp_init(&q->fdsp, avctx->flags & AV_CODEC_FLAG_BITEXACT);
q->units = av_mallocz(sizeof(*q->units) * avctx->channels);
if (!q->units) {
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_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)"),
+ .capabilities = AV_CODEC_CAP_SUBFRAMES | AV_CODEC_CAP_DR1,
.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
AV_SAMPLE_FMT_NONE },
};