/*
* Atrac 3 compatible decoder
- * Copyright (c) 2006-2007 Maxim Poliakovski
- * Copyright (c) 2006-2007 Benjamin Larsson
+ * Copyright (c) 2006-2008 Maxim Poliakovski
+ * Copyright (c) 2006-2008 Benjamin Larsson
*
* This file is part of FFmpeg.
*
*/
/**
- * @file atrac3.c
+ * @file libavcodec/atrac3.c
* Atrac 3 compatible decoder.
- * This decoder handles RealNetworks, RealAudio atrc data.
- * Atrac 3 is identified by the codec name atrc in RealMedia files.
+ * This decoder handles Sony's ATRAC3 data.
+ *
+ * Container formats used to store atrac 3 data:
+ * RealMedia (.rm), RIFF WAV (.wav, .at3), Sony OpenMG (.oma, .aa3).
*
* To use this decoder, a calling application must supply the extradata
- * bytes provided from the RealMedia container: 10 bytes or 14 bytes
- * from the WAV container.
+ * bytes provided in the containers above.
*/
#include <math.h>
#include <stdio.h>
#include "avcodec.h"
-#include "bitstream.h"
+#include "get_bits.h"
#include "dsputil.h"
#include "bytestream.h"
+#include "atrac.h"
#include "atrac3data.h"
#define JOINT_STEREO 0x12
float outSamples[2048];
uint8_t* decoded_bytes_buffer;
float tempBuf[1070];
- DECLARE_ALIGNED_16(float,mdct_tmp[512]);
//@}
//@{
/** extradata */
} ATRAC3Context;
static DECLARE_ALIGNED_16(float,mdct_window[512]);
-static float qmf_window[48];
static VLC spectral_coeff_tab[7];
-static float SFTable[64];
static float gain_tab1[16];
static float gain_tab2[31];
-static MDCTContext mdct_ctx;
+static FFTContext mdct_ctx;
static DSPContext dsp;
-/* quadrature mirror synthesis filter */
-
-/**
- * Quadrature mirror synthesis filter.
- *
- * @param inlo lower part of spectrum
- * @param inhi higher part of spectrum
- * @param nIn size of spectrum buffer
- * @param pOut out buffer
- * @param delayBuf delayBuf buffer
- * @param temp temp buffer
- */
-
-
-static void iqmf (float *inlo, float *inhi, unsigned int nIn, float *pOut, float *delayBuf, float *temp)
-{
- int i, j;
- float *p1, *p3;
-
- memcpy(temp, delayBuf, 46*sizeof(float));
-
- p3 = temp + 46;
-
- /* loop1 */
- for(i=0; i<nIn; i+=2){
- p3[2*i+0] = inlo[i ] + inhi[i ];
- p3[2*i+1] = inlo[i ] - inhi[i ];
- p3[2*i+2] = inlo[i+1] + inhi[i+1];
- p3[2*i+3] = inlo[i+1] - inhi[i+1];
- }
-
- /* loop2 */
- p1 = temp;
- for (j = nIn; j != 0; j--) {
- float s1 = 0.0;
- float s2 = 0.0;
-
- for (i = 0; i < 48; i += 2) {
- s1 += p1[i] * qmf_window[i];
- s2 += p1[i+1] * qmf_window[i+1];
- }
-
- pOut[0] = s2;
- pOut[1] = s1;
-
- p1 += 2;
- pOut += 2;
- }
-
- /* Update the delay buffer. */
- memcpy(delayBuf, temp + nIn*2, 46*sizeof(float));
-}
-
/**
* Regular 512 points IMDCT without overlapping, with the exception of the swapping of odd bands
* caused by the reverse spectra of the QMF.
* @param pInput float input
* @param pOutput float output
* @param odd_band 1 if the band is an odd band
- * @param mdct_tmp aligned temporary buffer for the mdct
*/
-static void IMLT(float *pInput, float *pOutput, int odd_band, float* mdct_tmp)
+static void IMLT(float *pInput, float *pOutput, int odd_band)
{
int i;
FFSWAP(float, pInput[i], pInput[255-i]);
}
- mdct_ctx.fft.imdct_calc(&mdct_ctx,pOutput,pInput,mdct_tmp);
+ ff_imdct_calc(&mdct_ctx,pOutput,pInput);
/* Perform windowing on the output. */
dsp.vector_fmul(pOutput,mdct_window,512);
* @param out pointer to 8 bit array of outdata
*/
-static int decode_bytes(uint8_t* inbuffer, uint8_t* out, int bytes){
+static int decode_bytes(const uint8_t* inbuffer, uint8_t* out, int bytes){
int i, off;
uint32_t c;
- uint32_t* buf;
+ const uint32_t* buf;
uint32_t* obuf = (uint32_t*) out;
- off = (int)((long)inbuffer & 3);
- buf = (uint32_t*) (inbuffer - off);
+ off = (intptr_t)inbuffer & 3;
+ buf = (const uint32_t*) (inbuffer - off);
c = be2me_32((0x537F6103 >> (off*8)) | (0x537F6103 << (32-(off*8))));
bytes += 3 + off;
for (i = 0; i < bytes/4; i++)
}
-static void init_atrac3_transforms(ATRAC3Context *q) {
+static av_cold void init_atrac3_transforms(ATRAC3Context *q) {
float enc_window[256];
- float s;
int i;
/* Generate the mdct window, for details see
mdct_window[511-i] = mdct_window[i];
}
- /* Generate the QMF window. */
- for (i=0 ; i<24; i++) {
- s = qmf_48tap_half[i] * 2.0;
- qmf_window[i] = s;
- qmf_window[47 - i] = s;
- }
-
/* Initialize the MDCT transform. */
- ff_mdct_init(&mdct_ctx, 9, 1);
+ ff_mdct_init(&mdct_ctx, 9, 1, 1.0);
}
/**
* Atrac3 uninit, free all allocated memory
*/
-static int atrac3_decode_close(AVCodecContext *avctx)
+static av_cold int atrac3_decode_close(AVCodecContext *avctx)
{
ATRAC3Context *q = avctx->priv_data;
if (codingFlag != 0) {
/* constant length coding (CLC) */
- //FIXME we don't have any samples coded in CLC mode
numBits = CLCLengthTab[selector];
if (selector > 1) {
float SF;
numSubbands = get_bits(gb, 5); // number of coded subbands
- codingMode = get_bits(gb, 1); // coding Mode: 0 - VLC/ 1-CLC
+ codingMode = get_bits1(gb); // coding Mode: 0 - VLC/ 1-CLC
/* Get the VLC selector table for the subbands, 0 means not coded. */
for (cnt = 0; cnt <= numSubbands; cnt++)
readQuantSpectralCoeffs (gb, subband_vlc_index[cnt], codingMode, mantissas, subbWidth);
/* Decode the scale factor for this subband. */
- SF = SFTable[SF_idxs[cnt]] * iMaxQuant[subband_vlc_index[cnt]];
+ SF = sf_table[SF_idxs[cnt]] * iMaxQuant[subband_vlc_index[cnt]];
/* Inverse quantize the coefficients. */
for (pIn=mantissas ; first<last; first++, pIn++)
* Restore the quantized tonal components
*
* @param gb the GetBit context
- * @param numComponents tonal components to report back
* @param pComponent tone component
* @param numBands amount of coded bands
*/
-static int decodeTonalComponents (GetBitContext *gb, int *numComponents, tonal_component *pComponent, int numBands)
+static int decodeTonalComponents (GetBitContext *gb, tonal_component *pComponent, int numBands)
{
int i,j,k,cnt;
- int component_count, components, coding_mode_selector, coding_mode, coded_values_per_component;
+ int components, coding_mode_selector, coding_mode, coded_values_per_component;
int sfIndx, coded_values, max_coded_values, quant_step_index, coded_components;
int band_flags[4], mantissa[8];
float *pCoef;
float scalefactor;
-
- component_count = 0;
- *numComponents = 0;
+ int component_count = 0;
components = get_bits(gb,5);
coded_values = coded_values_per_component + 1;
coded_values = FFMIN(max_coded_values,coded_values);
- scalefactor = SFTable[sfIndx] * iMaxQuant[quant_step_index];
+ scalefactor = sf_table[sfIndx] * iMaxQuant[quant_step_index];
readQuantSpectralCoeffs(gb, quant_step_index, coding_mode, mantissa, coded_values);
pComponent[component_count].numCoefs = coded_values;
/* inverse quant */
- pCoef = pComponent[k].coef;
+ pCoef = pComponent[component_count].coef;
for (cnt = 0; cnt < coded_values; cnt++)
pCoef[cnt] = mantissa[cnt] * scalefactor;
}
}
- *numComponents = component_count;
-
- return 0;
+ return component_count;
}
/**
/**
* Combine the tonal band spectrum and regular band spectrum
+ * Return position of the last tonal coefficient
*
* @param pSpectrum output spectrum buffer
* @param numComponents amount of tonal components
* @param pComponent tonal components for this band
*/
-static void addTonalComponents (float *pSpectrum, int numComponents, tonal_component *pComponent)
+static int addTonalComponents (float *pSpectrum, int numComponents, tonal_component *pComponent)
{
- int cnt, i;
+ int cnt, i, lastPos = -1;
float *pIn, *pOut;
for (cnt = 0; cnt < numComponents; cnt++){
+ lastPos = FFMAX(pComponent[cnt].pos + pComponent[cnt].numCoefs, lastPos);
pIn = pComponent[cnt].coef;
pOut = &(pSpectrum[pComponent[cnt].pos]);
for (i=0 ; i<pComponent[cnt].numCoefs ; i++)
pOut[i] += pIn[i];
}
+
+ return lastPos;
}
static int decodeChannelSoundUnit (ATRAC3Context *q, GetBitContext *gb, channel_unit *pSnd, float *pOut, int channelNum, int codingMode)
{
- int band, result=0, numSubbands, numBands;
+ int band, result=0, numSubbands, lastTonal, numBands;
if (codingMode == JOINT_STEREO && channelNum == 1) {
if (get_bits(gb,2) != 3) {
result = decodeGainControl (gb, &(pSnd->gainBlock[pSnd->gcBlkSwitch]), pSnd->bandsCoded);
if (result) return result;
- result = decodeTonalComponents (gb, &pSnd->numComponents, pSnd->components, pSnd->bandsCoded);
- if (result) return result;
+ pSnd->numComponents = decodeTonalComponents (gb, pSnd->components, pSnd->bandsCoded);
+ if (pSnd->numComponents == -1) return -1;
numSubbands = decodeSpectrum (gb, pSnd->spectrum);
/* Merge the decoded spectrum and tonal components. */
- addTonalComponents (pSnd->spectrum, pSnd->numComponents, pSnd->components);
+ lastTonal = addTonalComponents (pSnd->spectrum, pSnd->numComponents, pSnd->components);
- /* Convert number of subbands into number of MLT/QMF bands */
+ /* calculate number of used MLT/QMF bands according to the amount of coded spectral lines */
numBands = (subbandTab[numSubbands] - 1) >> 8;
+ if (lastTonal >= 0)
+ numBands = FFMAX((lastTonal + 256) >> 8, numBands);
/* Reconstruct time domain samples. */
for (band=0; band<4; band++) {
/* Perform the IMDCT step without overlapping. */
if (band <= numBands) {
- IMLT(&(pSnd->spectrum[band*256]), pSnd->IMDCT_buf, band&1,q->mdct_tmp);
+ IMLT(&(pSnd->spectrum[band*256]), pSnd->IMDCT_buf, band&1);
} else
memset(pSnd->IMDCT_buf, 0, 512 * sizeof(float));
* @param databuf the input data
*/
-static int decodeFrame(ATRAC3Context *q, uint8_t* databuf)
+static int decodeFrame(ATRAC3Context *q, const uint8_t* databuf)
{
int result, i;
float *p1, *p2, *p3, *p4;
- uint8_t *ptr1, *ptr2;
+ uint8_t *ptr1;
if (q->codingMode == JOINT_STEREO) {
/* Framedata of the su2 in the joint-stereo mode is encoded in
* reverse byte order so we need to swap it first. */
- ptr1 = databuf;
- ptr2 = databuf+q->bytes_per_frame-1;
- for (i = 0; i < (q->bytes_per_frame/2); i++, ptr1++, ptr2--) {
- FFSWAP(uint8_t,*ptr1,*ptr2);
+ if (databuf == q->decoded_bytes_buffer) {
+ uint8_t *ptr2 = q->decoded_bytes_buffer+q->bytes_per_frame-1;
+ ptr1 = q->decoded_bytes_buffer;
+ for (i = 0; i < (q->bytes_per_frame/2); i++, ptr1++, ptr2--) {
+ FFSWAP(uint8_t,*ptr1,*ptr2);
+ }
+ } else {
+ const uint8_t *ptr2 = databuf+q->bytes_per_frame-1;
+ for (i = 0; i < q->bytes_per_frame; i++)
+ q->decoded_bytes_buffer[i] = *ptr2--;
}
/* Skip the sync codes (0xF8). */
- ptr1 = databuf;
+ ptr1 = q->decoded_bytes_buffer;
for (i = 4; *ptr1 == 0xF8; i++, ptr1++) {
if (i >= q->bytes_per_frame)
return -1;
/* Fill the Weighting coeffs delay buffer */
memmove(q->weighting_delay,&(q->weighting_delay[2]),4*sizeof(int));
- q->weighting_delay[4] = get_bits(&q->gb,1);
+ q->weighting_delay[4] = get_bits1(&q->gb);
q->weighting_delay[5] = get_bits(&q->gb,3);
for (i = 0; i < 4; i++) {
p2= p1+256;
p3= p2+256;
p4= p3+256;
- iqmf (p1, p2, 256, p1, q->pUnits[i].delayBuf1, q->tempBuf);
- iqmf (p4, p3, 256, p3, q->pUnits[i].delayBuf2, q->tempBuf);
- iqmf (p1, p3, 512, p1, q->pUnits[i].delayBuf3, q->tempBuf);
+ atrac_iqmf (p1, p2, 256, p1, q->pUnits[i].delayBuf1, q->tempBuf);
+ atrac_iqmf (p4, p3, 256, p3, q->pUnits[i].delayBuf2, q->tempBuf);
+ atrac_iqmf (p1, p3, 512, p1, q->pUnits[i].delayBuf3, q->tempBuf);
p1 +=1024;
}
static int atrac3_decode_frame(AVCodecContext *avctx,
void *data, int *data_size,
- uint8_t *buf, int buf_size) {
+ AVPacket *avpkt) {
+ const uint8_t *buf = avpkt->data;
+ int buf_size = avpkt->size;
ATRAC3Context *q = avctx->priv_data;
int result = 0, i;
- uint8_t* databuf;
+ const uint8_t* databuf;
int16_t* samples = data;
if (buf_size < avctx->block_align)
if (q->channels == 1) {
/* mono */
for (i = 0; i<1024; i++)
- samples[i] = av_clip(round(q->outSamples[i]), -32768, 32767);
+ samples[i] = av_clip_int16(round(q->outSamples[i]));
*data_size = 1024 * sizeof(int16_t);
} else {
/* stereo */
for (i = 0; i < 1024; i++) {
- samples[i*2] = av_clip(round(q->outSamples[i]), -32768, 32767);
- samples[i*2+1] = av_clip(round(q->outSamples[1024+i]), -32768, 32767);
+ samples[i*2] = av_clip_int16(round(q->outSamples[i]));
+ samples[i*2+1] = av_clip_int16(round(q->outSamples[1024+i]));
}
*data_size = 2048 * sizeof(int16_t);
}
* @param avctx pointer to the AVCodecContext
*/
-static int atrac3_decode_init(AVCodecContext *avctx)
+static av_cold int atrac3_decode_init(AVCodecContext *avctx)
{
int i;
- uint8_t *edata_ptr = avctx->extradata;
+ 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;
/* Pad the data buffer with FF_INPUT_BUFFER_PADDING_SIZE,
* this is for the bitstream reader. */
if ((q->decoded_bytes_buffer = av_mallocz((avctx->block_align+(4-avctx->block_align%4) + FF_INPUT_BUFFER_PADDING_SIZE))) == NULL)
- return -1;
+ return AVERROR(ENOMEM);
/* Initialize the VLC tables. */
- for (i=0 ; i<7 ; 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_STATIC);
+ 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;
}
init_atrac3_transforms(q);
- /* Generate the scale factors. */
- for (i=0 ; i<64 ; i++)
- SFTable[i] = pow(2.0, (i - 15) / 3.0);
+ atrac_generate_tables();
/* Generate gain tables. */
for (i=0 ; i<16 ; i++)
dsputil_init(&dsp, avctx);
q->pUnits = av_mallocz(sizeof(channel_unit)*q->channels);
+ if (!q->pUnits) {
+ av_free(q->decoded_bytes_buffer);
+ return AVERROR(ENOMEM);
+ }
+ avctx->sample_fmt = SAMPLE_FMT_S16;
return 0;
}
AVCodec atrac3_decoder =
{
- .name = "atrac 3",
+ .name = "atrac3",
.type = CODEC_TYPE_AUDIO,
.id = CODEC_ID_ATRAC3,
.priv_data_size = sizeof(ATRAC3Context),
.init = atrac3_decode_init,
.close = atrac3_decode_close,
.decode = atrac3_decode_frame,
+ .long_name = NULL_IF_CONFIG_SMALL("Atrac 3 (Adaptive TRansform Acoustic Coding 3)"),
};
projects / ffmpeg / blobdiff