FFmpeg: libavcodec/atrac3.c Source File

00001 /*
00002  * Atrac 3 compatible decoder
00003  * Copyright (c) 2006-2008 Maxim Poliakovski
00004  * Copyright (c) 2006-2008 Benjamin Larsson
00005  *
00006  * This file is part of FFmpeg.
00007  *
00008  * FFmpeg is free software; you can redistribute it and/or
00009  * modify it under the terms of the GNU Lesser General Public
00010  * License as published by the Free Software Foundation; either
00011  * version 2.1 of the License, or (at your option) any later version.
00012  *
00013  * FFmpeg is distributed in the hope that it will be useful,
00014  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00015  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00016  * Lesser General Public License for more details.
00017  *
00018  * You should have received a copy of the GNU Lesser General Public
00019  * License along with FFmpeg; if not, write to the Free Software
00020  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00021  */
00022 
00035 #include <math.h>
00036 #include <stddef.h>
00037 #include <stdio.h>
00038 
00039 #include "avcodec.h"
00040 #include "get_bits.h"
00041 #include "dsputil.h"
00042 #include "bytestream.h"
00043 #include "fft.h"
00044 
00045 #include "atrac.h"
00046 #include "atrac3data.h"
00047 
00048 #define JOINT_STEREO    0x12
00049 #define STEREO          0x2
00050 
00051 
00052 /* These structures are needed to store the parsed gain control data. */
00053 typedef struct {
00054     int   num_gain_data;
00055     int   levcode[8];
00056     int   loccode[8];
00057 } gain_info;
00058 
00059 typedef struct {
00060     gain_info   gBlock[4];
00061 } gain_block;
00062 
00063 typedef struct {
00064     int     pos;
00065     int     numCoefs;
00066     float   coef[8];
00067 } tonal_component;
00068 
00069 typedef struct {
00070     int               bandsCoded;
00071     int               numComponents;
00072     tonal_component   components[64];
00073     float             prevFrame[1024];
00074     int               gcBlkSwitch;
00075     gain_block        gainBlock[2];
00076 
00077     DECLARE_ALIGNED(32, float, spectrum)[1024];
00078     DECLARE_ALIGNED(32, float, IMDCT_buf)[1024];
00079 
00080     float             delayBuf1[46]; 
00081     float             delayBuf2[46];
00082     float             delayBuf3[46];
00083 } channel_unit;
00084 
00085 typedef struct {
00086     GetBitContext       gb;
00088 
00089     int                 channels;
00090     int                 codingMode;
00091     int                 bit_rate;
00092     int                 sample_rate;
00093     int                 samples_per_channel;
00094     int                 samples_per_frame;
00095 
00096     int                 bits_per_frame;
00097     int                 bytes_per_frame;
00098     int                 pBs;
00099     channel_unit*       pUnits;
00101 
00102 
00103     int                 matrix_coeff_index_prev[4];
00104     int                 matrix_coeff_index_now[4];
00105     int                 matrix_coeff_index_next[4];
00106     int                 weighting_delay[6];
00108 
00109 
00110     float               outSamples[2048];
00111     uint8_t*            decoded_bytes_buffer;
00112     float               tempBuf[1070];
00114 
00115 
00116     int                 atrac3version;
00117     int                 delay;
00118     int                 scrambled_stream;
00119     int                 frame_factor;
00121 
00122     FFTContext          mdct_ctx;
00123 } ATRAC3Context;
00124 
00125 static DECLARE_ALIGNED(32, float, mdct_window)[512];
00126 static VLC              spectral_coeff_tab[7];
00127 static float            gain_tab1[16];
00128 static float            gain_tab2[31];
00129 static DSPContext       dsp;
00130 
00131 
00141 static void IMLT(ATRAC3Context *q, float *pInput, float *pOutput, int odd_band)
00142 {
00143     int     i;
00144 
00145     if (odd_band) {
00155         for (i=0; i<128; i++)
00156             FFSWAP(float, pInput[i], pInput[255-i]);
00157     }
00158 
00159     q->mdct_ctx.imdct_calc(&q->mdct_ctx,pOutput,pInput);
00160 
00161     /* Perform windowing on the output. */
00162     dsp.vector_fmul(pOutput, pOutput, mdct_window, 512);
00163 
00164 }
00165 
00166 
00175 static int decode_bytes(const uint8_t* inbuffer, uint8_t* out, int bytes){
00176     int i, off;
00177     uint32_t c;
00178     const uint32_t* buf;
00179     uint32_t* obuf = (uint32_t*) out;
00180 
00181     off = (intptr_t)inbuffer & 3;
00182     buf = (const uint32_t*) (inbuffer - off);
00183     c = av_be2ne32((0x537F6103 >> (off*8)) | (0x537F6103 << (32-(off*8))));
00184     bytes += 3 + off;
00185     for (i = 0; i < bytes/4; i++)
00186         obuf[i] = c ^ buf[i];
00187 
00188     if (off)
00189         av_log_ask_for_sample(NULL, "Offset of %d not handled.\n", off);
00190 
00191     return off;
00192 }
00193 
00194 
00195 static av_cold void init_atrac3_transforms(ATRAC3Context *q) {
00196     float enc_window[256];
00197     int i;
00198 
00199     /* Generate the mdct window, for details see
00200      * http://wiki.multimedia.cx/index.php?title=RealAudio_atrc#Windows */
00201     for (i=0 ; i<256; i++)
00202         enc_window[i] = (sin(((i + 0.5) / 256.0 - 0.5) * M_PI) + 1.0) * 0.5;
00203 
00204     if (!mdct_window[0])
00205         for (i=0 ; i<256; i++) {
00206             mdct_window[i] = enc_window[i]/(enc_window[i]*enc_window[i] + enc_window[255-i]*enc_window[255-i]);
00207             mdct_window[511-i] = mdct_window[i];
00208         }
00209 
00210     /* Initialize the MDCT transform. */
00211     ff_mdct_init(&q->mdct_ctx, 9, 1, 1.0);
00212 }
00213 
00218 static av_cold int atrac3_decode_close(AVCodecContext *avctx)
00219 {
00220     ATRAC3Context *q = avctx->priv_data;
00221 
00222     av_free(q->pUnits);
00223     av_free(q->decoded_bytes_buffer);
00224     ff_mdct_end(&q->mdct_ctx);
00225 
00226     return 0;
00227 }
00228 
00239 static void readQuantSpectralCoeffs (GetBitContext *gb, int selector, int codingFlag, int* mantissas, int numCodes)
00240 {
00241     int   numBits, cnt, code, huffSymb;
00242 
00243     if (selector == 1)
00244         numCodes /= 2;
00245 
00246     if (codingFlag != 0) {
00247         /* constant length coding (CLC) */
00248         numBits = CLCLengthTab[selector];
00249 
00250         if (selector > 1) {
00251             for (cnt = 0; cnt < numCodes; cnt++) {
00252                 if (numBits)
00253                     code = get_sbits(gb, numBits);
00254                 else
00255                     code = 0;
00256                 mantissas[cnt] = code;
00257             }
00258         } else {
00259             for (cnt = 0; cnt < numCodes; cnt++) {
00260                 if (numBits)
00261                     code = get_bits(gb, numBits); //numBits is always 4 in this case
00262                 else
00263                     code = 0;
00264                 mantissas[cnt*2] = seTab_0[code >> 2];
00265                 mantissas[cnt*2+1] = seTab_0[code & 3];
00266             }
00267         }
00268     } else {
00269         /* variable length coding (VLC) */
00270         if (selector != 1) {
00271             for (cnt = 0; cnt < numCodes; cnt++) {
00272                 huffSymb = get_vlc2(gb, spectral_coeff_tab[selector-1].table, spectral_coeff_tab[selector-1].bits, 3);
00273                 huffSymb += 1;
00274                 code = huffSymb >> 1;
00275                 if (huffSymb & 1)
00276                     code = -code;
00277                 mantissas[cnt] = code;
00278             }
00279         } else {
00280             for (cnt = 0; cnt < numCodes; cnt++) {
00281                 huffSymb = get_vlc2(gb, spectral_coeff_tab[selector-1].table, spectral_coeff_tab[selector-1].bits, 3);
00282                 mantissas[cnt*2] = decTable1[huffSymb*2];
00283                 mantissas[cnt*2+1] = decTable1[huffSymb*2+1];
00284             }
00285         }
00286     }
00287 }
00288 
00297 static int decodeSpectrum (GetBitContext *gb, float *pOut)
00298 {
00299     int   numSubbands, codingMode, cnt, first, last, subbWidth, *pIn;
00300     int   subband_vlc_index[32], SF_idxs[32];
00301     int   mantissas[128];
00302     float SF;
00303 
00304     numSubbands = get_bits(gb, 5); // number of coded subbands
00305     codingMode = get_bits1(gb); // coding Mode: 0 - VLC/ 1-CLC
00306 
00307     /* Get the VLC selector table for the subbands, 0 means not coded. */
00308     for (cnt = 0; cnt <= numSubbands; cnt++)
00309         subband_vlc_index[cnt] = get_bits(gb, 3);
00310 
00311     /* Read the scale factor indexes from the stream. */
00312     for (cnt = 0; cnt <= numSubbands; cnt++) {
00313         if (subband_vlc_index[cnt] != 0)
00314             SF_idxs[cnt] = get_bits(gb, 6);
00315     }
00316 
00317     for (cnt = 0; cnt <= numSubbands; cnt++) {
00318         first = subbandTab[cnt];
00319         last = subbandTab[cnt+1];
00320 
00321         subbWidth = last - first;
00322 
00323         if (subband_vlc_index[cnt] != 0) {
00324             /* Decode spectral coefficients for this subband. */
00325             /* TODO: This can be done faster is several blocks share the
00326              * same VLC selector (subband_vlc_index) */
00327             readQuantSpectralCoeffs (gb, subband_vlc_index[cnt], codingMode, mantissas, subbWidth);
00328 
00329             /* Decode the scale factor for this subband. */
00330             SF = ff_atrac_sf_table[SF_idxs[cnt]] * iMaxQuant[subband_vlc_index[cnt]];
00331 
00332             /* Inverse quantize the coefficients. */
00333             for (pIn=mantissas ; first<last; first++, pIn++)
00334                 pOut[first] = *pIn * SF;
00335         } else {
00336             /* This subband was not coded, so zero the entire subband. */
00337             memset(pOut+first, 0, subbWidth*sizeof(float));
00338         }
00339     }
00340 
00341     /* Clear the subbands that were not coded. */
00342     first = subbandTab[cnt];
00343     memset(pOut+first, 0, (1024 - first) * sizeof(float));
00344     return numSubbands;
00345 }
00346 
00355 static int decodeTonalComponents (GetBitContext *gb, tonal_component *pComponent, int numBands)
00356 {
00357     int i,j,k,cnt;
00358     int   components, coding_mode_selector, coding_mode, coded_values_per_component;
00359     int   sfIndx, coded_values, max_coded_values, quant_step_index, coded_components;
00360     int   band_flags[4], mantissa[8];
00361     float  *pCoef;
00362     float  scalefactor;
00363     int   component_count = 0;
00364 
00365     components = get_bits(gb,5);
00366 
00367     /* no tonal components */
00368     if (components == 0)
00369         return 0;
00370 
00371     coding_mode_selector = get_bits(gb,2);
00372     if (coding_mode_selector == 2)
00373         return -1;
00374 
00375     coding_mode = coding_mode_selector & 1;
00376 
00377     for (i = 0; i < components; i++) {
00378         for (cnt = 0; cnt <= numBands; cnt++)
00379             band_flags[cnt] = get_bits1(gb);
00380 
00381         coded_values_per_component = get_bits(gb,3);
00382 
00383         quant_step_index = get_bits(gb,3);
00384         if (quant_step_index <= 1)
00385             return -1;
00386 
00387         if (coding_mode_selector == 3)
00388             coding_mode = get_bits1(gb);
00389 
00390         for (j = 0; j < (numBands + 1) * 4; j++) {
00391             if (band_flags[j >> 2] == 0)
00392                 continue;
00393 
00394             coded_components = get_bits(gb,3);
00395 
00396             for (k=0; k<coded_components; k++) {
00397                 sfIndx = get_bits(gb,6);
00398                 if (component_count >= 64)
00399                     return AVERROR_INVALIDDATA;
00400                 pComponent[component_count].pos = j * 64 + (get_bits(gb,6));
00401                 max_coded_values = 1024 - pComponent[component_count].pos;
00402                 coded_values = coded_values_per_component + 1;
00403                 coded_values = FFMIN(max_coded_values,coded_values);
00404 
00405                 scalefactor = ff_atrac_sf_table[sfIndx] * iMaxQuant[quant_step_index];
00406 
00407                 readQuantSpectralCoeffs(gb, quant_step_index, coding_mode, mantissa, coded_values);
00408 
00409                 pComponent[component_count].numCoefs = coded_values;
00410 
00411                 /* inverse quant */
00412                 pCoef = pComponent[component_count].coef;
00413                 for (cnt = 0; cnt < coded_values; cnt++)
00414                     pCoef[cnt] = mantissa[cnt] * scalefactor;
00415 
00416                 component_count++;
00417             }
00418         }
00419     }
00420 
00421     return component_count;
00422 }
00423 
00432 static int decodeGainControl (GetBitContext *gb, gain_block *pGb, int numBands)
00433 {
00434     int   i, cf, numData;
00435     int   *pLevel, *pLoc;
00436 
00437     gain_info   *pGain = pGb->gBlock;
00438 
00439     for (i=0 ; i<=numBands; i++)
00440     {
00441         numData = get_bits(gb,3);
00442         pGain[i].num_gain_data = numData;
00443         pLevel = pGain[i].levcode;
00444         pLoc = pGain[i].loccode;
00445 
00446         for (cf = 0; cf < numData; cf++){
00447             pLevel[cf]= get_bits(gb,4);
00448             pLoc  [cf]= get_bits(gb,5);
00449             if(cf && pLoc[cf] <= pLoc[cf-1])
00450                 return -1;
00451         }
00452     }
00453 
00454     /* Clear the unused blocks. */
00455     for (; i<4 ; i++)
00456         pGain[i].num_gain_data = 0;
00457 
00458     return 0;
00459 }
00460 
00471 static void gainCompensateAndOverlap (float *pIn, float *pPrev, float *pOut, gain_info *pGain1, gain_info *pGain2)
00472 {
00473     /* gain compensation function */
00474     float  gain1, gain2, gain_inc;
00475     int   cnt, numdata, nsample, startLoc, endLoc;
00476 
00477 
00478     if (pGain2->num_gain_data == 0)
00479         gain1 = 1.0;
00480     else
00481         gain1 = gain_tab1[pGain2->levcode[0]];
00482 
00483     if (pGain1->num_gain_data == 0) {
00484         for (cnt = 0; cnt < 256; cnt++)
00485             pOut[cnt] = pIn[cnt] * gain1 + pPrev[cnt];
00486     } else {
00487         numdata = pGain1->num_gain_data;
00488         pGain1->loccode[numdata] = 32;
00489         pGain1->levcode[numdata] = 4;
00490 
00491         nsample = 0; // current sample = 0
00492 
00493         for (cnt = 0; cnt < numdata; cnt++) {
00494             startLoc = pGain1->loccode[cnt] * 8;
00495             endLoc = startLoc + 8;
00496 
00497             gain2 = gain_tab1[pGain1->levcode[cnt]];
00498             gain_inc = gain_tab2[(pGain1->levcode[cnt+1] - pGain1->levcode[cnt])+15];
00499 
00500             /* interpolate */
00501             for (; nsample < startLoc; nsample++)
00502                 pOut[nsample] = (pIn[nsample] * gain1 + pPrev[nsample]) * gain2;
00503 
00504             /* interpolation is done over eight samples */
00505             for (; nsample < endLoc; nsample++) {
00506                 pOut[nsample] = (pIn[nsample] * gain1 + pPrev[nsample]) * gain2;
00507                 gain2 *= gain_inc;
00508             }
00509         }
00510 
00511         for (; nsample < 256; nsample++)
00512             pOut[nsample] = (pIn[nsample] * gain1) + pPrev[nsample];
00513     }
00514 
00515     /* Delay for the overlapping part. */
00516     memcpy(pPrev, &pIn[256], 256*sizeof(float));
00517 }
00518 
00528 static int addTonalComponents (float *pSpectrum, int numComponents, tonal_component *pComponent)
00529 {
00530     int   cnt, i, lastPos = -1;
00531     float   *pIn, *pOut;
00532 
00533     for (cnt = 0; cnt < numComponents; cnt++){
00534         lastPos = FFMAX(pComponent[cnt].pos + pComponent[cnt].numCoefs, lastPos);
00535         pIn = pComponent[cnt].coef;
00536         pOut = &(pSpectrum[pComponent[cnt].pos]);
00537 
00538         for (i=0 ; i<pComponent[cnt].numCoefs ; i++)
00539             pOut[i] += pIn[i];
00540     }
00541 
00542     return lastPos;
00543 }
00544 
00545 
00546 #define INTERPOLATE(old,new,nsample) ((old) + (nsample)*0.125*((new)-(old)))
00547 
00548 static void reverseMatrixing(float *su1, float *su2, int *pPrevCode, int *pCurrCode)
00549 {
00550     int    i, band, nsample, s1, s2;
00551     float    c1, c2;
00552     float    mc1_l, mc1_r, mc2_l, mc2_r;
00553 
00554     for (i=0,band = 0; band < 4*256; band+=256,i++) {
00555         s1 = pPrevCode[i];
00556         s2 = pCurrCode[i];
00557         nsample = 0;
00558 
00559         if (s1 != s2) {
00560             /* Selector value changed, interpolation needed. */
00561             mc1_l = matrixCoeffs[s1*2];
00562             mc1_r = matrixCoeffs[s1*2+1];
00563             mc2_l = matrixCoeffs[s2*2];
00564             mc2_r = matrixCoeffs[s2*2+1];
00565 
00566             /* Interpolation is done over the first eight samples. */
00567             for(; nsample < 8; nsample++) {
00568                 c1 = su1[band+nsample];
00569                 c2 = su2[band+nsample];
00570                 c2 = c1 * INTERPOLATE(mc1_l,mc2_l,nsample) + c2 * INTERPOLATE(mc1_r,mc2_r,nsample);
00571                 su1[band+nsample] = c2;
00572                 su2[band+nsample] = c1 * 2.0 - c2;
00573             }
00574         }
00575 
00576         /* Apply the matrix without interpolation. */
00577         switch (s2) {
00578             case 0:     /* M/S decoding */
00579                 for (; nsample < 256; nsample++) {
00580                     c1 = su1[band+nsample];
00581                     c2 = su2[band+nsample];
00582                     su1[band+nsample] = c2 * 2.0;
00583                     su2[band+nsample] = (c1 - c2) * 2.0;
00584                 }
00585                 break;
00586 
00587             case 1:
00588                 for (; nsample < 256; nsample++) {
00589                     c1 = su1[band+nsample];
00590                     c2 = su2[band+nsample];
00591                     su1[band+nsample] = (c1 + c2) * 2.0;
00592                     su2[band+nsample] = c2 * -2.0;
00593                 }
00594                 break;
00595             case 2:
00596             case 3:
00597                 for (; nsample < 256; nsample++) {
00598                     c1 = su1[band+nsample];
00599                     c2 = su2[band+nsample];
00600                     su1[band+nsample] = c1 + c2;
00601                     su2[band+nsample] = c1 - c2;
00602                 }
00603                 break;
00604             default:
00605                 assert(0);
00606         }
00607     }
00608 }
00609 
00610 static void getChannelWeights (int indx, int flag, float ch[2]){
00611 
00612     if (indx == 7) {
00613         ch[0] = 1.0;
00614         ch[1] = 1.0;
00615     } else {
00616         ch[0] = (float)(indx & 7) / 7.0;
00617         ch[1] = sqrt(2 - ch[0]*ch[0]);
00618         if(flag)
00619             FFSWAP(float, ch[0], ch[1]);
00620     }
00621 }
00622 
00623 static void channelWeighting (float *su1, float *su2, int *p3)
00624 {
00625     int   band, nsample;
00626     /* w[x][y] y=0 is left y=1 is right */
00627     float w[2][2];
00628 
00629     if (p3[1] != 7 || p3[3] != 7){
00630         getChannelWeights(p3[1], p3[0], w[0]);
00631         getChannelWeights(p3[3], p3[2], w[1]);
00632 
00633         for(band = 1; band < 4; band++) {
00634             /* scale the channels by the weights */
00635             for(nsample = 0; nsample < 8; nsample++) {
00636                 su1[band*256+nsample] *= INTERPOLATE(w[0][0], w[0][1], nsample);
00637                 su2[band*256+nsample] *= INTERPOLATE(w[1][0], w[1][1], nsample);
00638             }
00639 
00640             for(; nsample < 256; nsample++) {
00641                 su1[band*256+nsample] *= w[1][0];
00642                 su2[band*256+nsample] *= w[1][1];
00643             }
00644         }
00645     }
00646 }
00647 
00648 
00660 static int decodeChannelSoundUnit (ATRAC3Context *q, GetBitContext *gb, channel_unit *pSnd, float *pOut, int channelNum, int codingMode)
00661 {
00662     int   band, result=0, numSubbands, lastTonal, numBands;
00663 
00664     if (codingMode == JOINT_STEREO && channelNum == 1) {
00665         if (get_bits(gb,2) != 3) {
00666             av_log(NULL,AV_LOG_ERROR,"JS mono Sound Unit id != 3.\n");
00667             return -1;
00668         }
00669     } else {
00670         if (get_bits(gb,6) != 0x28) {
00671             av_log(NULL,AV_LOG_ERROR,"Sound Unit id != 0x28.\n");
00672             return -1;
00673         }
00674     }
00675 
00676     /* number of coded QMF bands */
00677     pSnd->bandsCoded = get_bits(gb,2);
00678 
00679     result = decodeGainControl (gb, &(pSnd->gainBlock[pSnd->gcBlkSwitch]), pSnd->bandsCoded);
00680     if (result) return result;
00681 
00682     pSnd->numComponents = decodeTonalComponents (gb, pSnd->components, pSnd->bandsCoded);
00683     if (pSnd->numComponents == -1) return -1;
00684 
00685     numSubbands = decodeSpectrum (gb, pSnd->spectrum);
00686 
00687     /* Merge the decoded spectrum and tonal components. */
00688     lastTonal = addTonalComponents (pSnd->spectrum, pSnd->numComponents, pSnd->components);
00689 
00690 
00691     /* calculate number of used MLT/QMF bands according to the amount of coded spectral lines */
00692     numBands = (subbandTab[numSubbands] - 1) >> 8;
00693     if (lastTonal >= 0)
00694         numBands = FFMAX((lastTonal + 256) >> 8, numBands);
00695 
00696 
00697     /* Reconstruct time domain samples. */
00698     for (band=0; band<4; band++) {
00699         /* Perform the IMDCT step without overlapping. */
00700         if (band <= numBands) {
00701             IMLT(q, &(pSnd->spectrum[band*256]), pSnd->IMDCT_buf, band&1);
00702         } else
00703             memset(pSnd->IMDCT_buf, 0, 512 * sizeof(float));
00704 
00705         /* gain compensation and overlapping */
00706         gainCompensateAndOverlap (pSnd->IMDCT_buf, &(pSnd->prevFrame[band*256]), &(pOut[band*256]),
00707                                     &((pSnd->gainBlock[1 - (pSnd->gcBlkSwitch)]).gBlock[band]),
00708                                     &((pSnd->gainBlock[pSnd->gcBlkSwitch]).gBlock[band]));
00709     }
00710 
00711     /* Swap the gain control buffers for the next frame. */
00712     pSnd->gcBlkSwitch ^= 1;
00713 
00714     return 0;
00715 }
00716 
00724 static int decodeFrame(ATRAC3Context *q, const uint8_t* databuf)
00725 {
00726     int   result, i;
00727     float   *p1, *p2, *p3, *p4;
00728     uint8_t *ptr1;
00729 
00730     if (q->codingMode == JOINT_STEREO) {
00731 
00732         /* channel coupling mode */
00733         /* decode Sound Unit 1 */
00734         init_get_bits(&q->gb,databuf,q->bits_per_frame);
00735 
00736         result = decodeChannelSoundUnit(q,&q->gb, q->pUnits, q->outSamples, 0, JOINT_STEREO);
00737         if (result != 0)
00738             return (result);
00739 
00740         /* Framedata of the su2 in the joint-stereo mode is encoded in
00741          * reverse byte order so we need to swap it first. */
00742         if (databuf == q->decoded_bytes_buffer) {
00743             uint8_t *ptr2 = q->decoded_bytes_buffer+q->bytes_per_frame-1;
00744             ptr1 = q->decoded_bytes_buffer;
00745             for (i = 0; i < (q->bytes_per_frame/2); i++, ptr1++, ptr2--) {
00746                 FFSWAP(uint8_t,*ptr1,*ptr2);
00747             }
00748         } else {
00749             const uint8_t *ptr2 = databuf+q->bytes_per_frame-1;
00750             for (i = 0; i < q->bytes_per_frame; i++)
00751                 q->decoded_bytes_buffer[i] = *ptr2--;
00752         }
00753 
00754         /* Skip the sync codes (0xF8). */
00755         ptr1 = q->decoded_bytes_buffer;
00756         for (i = 4; *ptr1 == 0xF8; i++, ptr1++) {
00757             if (i >= q->bytes_per_frame)
00758                 return -1;
00759         }
00760 
00761 
00762         /* set the bitstream reader at the start of the second Sound Unit*/
00763         init_get_bits(&q->gb,ptr1,q->bits_per_frame);
00764 
00765         /* Fill the Weighting coeffs delay buffer */
00766         memmove(q->weighting_delay,&(q->weighting_delay[2]),4*sizeof(int));
00767         q->weighting_delay[4] = get_bits1(&q->gb);
00768         q->weighting_delay[5] = get_bits(&q->gb,3);
00769 
00770         for (i = 0; i < 4; i++) {
00771             q->matrix_coeff_index_prev[i] = q->matrix_coeff_index_now[i];
00772             q->matrix_coeff_index_now[i] = q->matrix_coeff_index_next[i];
00773             q->matrix_coeff_index_next[i] = get_bits(&q->gb,2);
00774         }
00775 
00776         /* Decode Sound Unit 2. */
00777         result = decodeChannelSoundUnit(q,&q->gb, &q->pUnits[1], &q->outSamples[1024], 1, JOINT_STEREO);
00778         if (result != 0)
00779             return (result);
00780 
00781         /* Reconstruct the channel coefficients. */
00782         reverseMatrixing(q->outSamples, &q->outSamples[1024], q->matrix_coeff_index_prev, q->matrix_coeff_index_now);
00783 
00784         channelWeighting(q->outSamples, &q->outSamples[1024], q->weighting_delay);
00785 
00786     } else {
00787         /* normal stereo mode or mono */
00788         /* Decode the channel sound units. */
00789         for (i=0 ; i<q->channels ; i++) {
00790 
00791             /* Set the bitstream reader at the start of a channel sound unit. */
00792             init_get_bits(&q->gb, databuf+((i*q->bytes_per_frame)/q->channels), (q->bits_per_frame)/q->channels);
00793 
00794             result = decodeChannelSoundUnit(q,&q->gb, &q->pUnits[i], &q->outSamples[i*1024], i, q->codingMode);
00795             if (result != 0)
00796                 return (result);
00797         }
00798     }
00799 
00800     /* Apply the iQMF synthesis filter. */
00801     p1= q->outSamples;
00802     for (i=0 ; i<q->channels ; i++) {
00803         p2= p1+256;
00804         p3= p2+256;
00805         p4= p3+256;
00806         atrac_iqmf (p1, p2, 256, p1, q->pUnits[i].delayBuf1, q->tempBuf);
00807         atrac_iqmf (p4, p3, 256, p3, q->pUnits[i].delayBuf2, q->tempBuf);
00808         atrac_iqmf (p1, p3, 512, p1, q->pUnits[i].delayBuf3, q->tempBuf);
00809         p1 +=1024;
00810     }
00811 
00812     return 0;
00813 }
00814 
00815 
00822 static int atrac3_decode_frame(AVCodecContext *avctx,
00823             void *data, int *data_size,
00824             AVPacket *avpkt) {
00825     const uint8_t *buf = avpkt->data;
00826     int buf_size = avpkt->size;
00827     ATRAC3Context *q = avctx->priv_data;
00828     int result = 0, i;
00829     const uint8_t* databuf;
00830     int16_t* samples = data;
00831 
00832     if (buf_size < avctx->block_align) {
00833         av_log(avctx, AV_LOG_ERROR,
00834                "Frame too small (%d bytes). Truncated file?\n", buf_size);
00835         *data_size = 0;
00836         return buf_size;
00837     }
00838 
00839     /* Check if we need to descramble and what buffer to pass on. */
00840     if (q->scrambled_stream) {
00841         decode_bytes(buf, q->decoded_bytes_buffer, avctx->block_align);
00842         databuf = q->decoded_bytes_buffer;
00843     } else {
00844         databuf = buf;
00845     }
00846 
00847     result = decodeFrame(q, databuf);
00848 
00849     if (result != 0) {
00850         av_log(NULL,AV_LOG_ERROR,"Frame decoding error!\n");
00851         return -1;
00852     }
00853 
00854     if (q->channels == 1) {
00855         /* mono */
00856         for (i = 0; i<1024; i++)
00857             samples[i] = av_clip_int16(round(q->outSamples[i]));
00858         *data_size = 1024 * sizeof(int16_t);
00859     } else {
00860         /* stereo */
00861         for (i = 0; i < 1024; i++) {
00862             samples[i*2] = av_clip_int16(round(q->outSamples[i]));
00863             samples[i*2+1] = av_clip_int16(round(q->outSamples[1024+i]));
00864         }
00865         *data_size = 2048 * sizeof(int16_t);
00866     }
00867 
00868     return avctx->block_align;
00869 }
00870 
00871 
00878 static av_cold int atrac3_decode_init(AVCodecContext *avctx)
00879 {
00880     int i;
00881     const uint8_t *edata_ptr = avctx->extradata;
00882     ATRAC3Context *q = avctx->priv_data;
00883     static VLC_TYPE atrac3_vlc_table[4096][2];
00884     static int vlcs_initialized = 0;
00885 
00886     /* Take data from the AVCodecContext (RM container). */
00887     q->sample_rate = avctx->sample_rate;
00888     q->channels = avctx->channels;
00889     q->bit_rate = avctx->bit_rate;
00890     q->bits_per_frame = avctx->block_align * 8;
00891     q->bytes_per_frame = avctx->block_align;
00892 
00893     /* Take care of the codec-specific extradata. */
00894     if (avctx->extradata_size == 14) {
00895         /* Parse the extradata, WAV format */
00896         av_log(avctx,AV_LOG_DEBUG,"[0-1] %d\n",bytestream_get_le16(&edata_ptr));  //Unknown value always 1
00897         q->samples_per_channel = bytestream_get_le32(&edata_ptr);
00898         q->codingMode = bytestream_get_le16(&edata_ptr);
00899         av_log(avctx,AV_LOG_DEBUG,"[8-9] %d\n",bytestream_get_le16(&edata_ptr));  //Dupe of coding mode
00900         q->frame_factor = bytestream_get_le16(&edata_ptr);  //Unknown always 1
00901         av_log(avctx,AV_LOG_DEBUG,"[12-13] %d\n",bytestream_get_le16(&edata_ptr));  //Unknown always 0
00902 
00903         /* setup */
00904         q->samples_per_frame = 1024 * q->channels;
00905         q->atrac3version = 4;
00906         q->delay = 0x88E;
00907         if (q->codingMode)
00908             q->codingMode = JOINT_STEREO;
00909         else
00910             q->codingMode = STEREO;
00911 
00912         q->scrambled_stream = 0;
00913 
00914         if ((q->bytes_per_frame == 96*q->channels*q->frame_factor) || (q->bytes_per_frame == 152*q->channels*q->frame_factor) || (q->bytes_per_frame == 192*q->channels*q->frame_factor)) {
00915         } else {
00916             av_log(avctx,AV_LOG_ERROR,"Unknown frame/channel/frame_factor configuration %d/%d/%d\n", q->bytes_per_frame, q->channels, q->frame_factor);
00917             return -1;
00918         }
00919 
00920     } else if (avctx->extradata_size == 10) {
00921         /* Parse the extradata, RM format. */
00922         q->atrac3version = bytestream_get_be32(&edata_ptr);
00923         q->samples_per_frame = bytestream_get_be16(&edata_ptr);
00924         q->delay = bytestream_get_be16(&edata_ptr);
00925         q->codingMode = bytestream_get_be16(&edata_ptr);
00926 
00927         q->samples_per_channel = q->samples_per_frame / q->channels;
00928         q->scrambled_stream = 1;
00929 
00930     } else {
00931         av_log(NULL,AV_LOG_ERROR,"Unknown extradata size %d.\n",avctx->extradata_size);
00932     }
00933     /* Check the extradata. */
00934 
00935     if (q->atrac3version != 4) {
00936         av_log(avctx,AV_LOG_ERROR,"Version %d != 4.\n",q->atrac3version);
00937         return -1;
00938     }
00939 
00940     if (q->samples_per_frame != 1024 && q->samples_per_frame != 2048) {
00941         av_log(avctx,AV_LOG_ERROR,"Unknown amount of samples per frame %d.\n",q->samples_per_frame);
00942         return -1;
00943     }
00944 
00945     if (q->delay != 0x88E) {
00946         av_log(avctx,AV_LOG_ERROR,"Unknown amount of delay %x != 0x88E.\n",q->delay);
00947         return -1;
00948     }
00949 
00950     if (q->codingMode == STEREO) {
00951         av_log(avctx,AV_LOG_DEBUG,"Normal stereo detected.\n");
00952     } else if (q->codingMode == JOINT_STEREO) {
00953         av_log(avctx,AV_LOG_DEBUG,"Joint stereo detected.\n");
00954     } else {
00955         av_log(avctx,AV_LOG_ERROR,"Unknown channel coding mode %x!\n",q->codingMode);
00956         return -1;
00957     }
00958 
00959     if (avctx->channels <= 0 || avctx->channels > 2 /*|| ((avctx->channels * 1024) != q->samples_per_frame)*/) {
00960         av_log(avctx,AV_LOG_ERROR,"Channel configuration error!\n");
00961         return -1;
00962     }
00963 
00964 
00965     if(avctx->block_align >= UINT_MAX/2)
00966         return -1;
00967 
00968     /* Pad the data buffer with FF_INPUT_BUFFER_PADDING_SIZE,
00969      * this is for the bitstream reader. */
00970     if ((q->decoded_bytes_buffer = av_mallocz((avctx->block_align+(4-avctx->block_align%4) + FF_INPUT_BUFFER_PADDING_SIZE)))  == NULL)
00971         return AVERROR(ENOMEM);
00972 
00973 
00974     /* Initialize the VLC tables. */
00975     if (!vlcs_initialized) {
00976         for (i=0 ; i<7 ; i++) {
00977             spectral_coeff_tab[i].table = &atrac3_vlc_table[atrac3_vlc_offs[i]];
00978             spectral_coeff_tab[i].table_allocated = atrac3_vlc_offs[i + 1] - atrac3_vlc_offs[i];
00979             init_vlc (&spectral_coeff_tab[i], 9, huff_tab_sizes[i],
00980                 huff_bits[i], 1, 1,
00981                 huff_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
00982         }
00983         vlcs_initialized = 1;
00984     }
00985 
00986     init_atrac3_transforms(q);
00987 
00988     atrac_generate_tables();
00989 
00990     /* Generate gain tables. */
00991     for (i=0 ; i<16 ; i++)
00992         gain_tab1[i] = powf (2.0, (4 - i));
00993 
00994     for (i=-15 ; i<16 ; i++)
00995         gain_tab2[i+15] = powf (2.0, i * -0.125);
00996 
00997     /* init the joint-stereo decoding data */
00998     q->weighting_delay[0] = 0;
00999     q->weighting_delay[1] = 7;
01000     q->weighting_delay[2] = 0;
01001     q->weighting_delay[3] = 7;
01002     q->weighting_delay[4] = 0;
01003     q->weighting_delay[5] = 7;
01004 
01005     for (i=0; i<4; i++) {
01006         q->matrix_coeff_index_prev[i] = 3;
01007         q->matrix_coeff_index_now[i] = 3;
01008         q->matrix_coeff_index_next[i] = 3;
01009     }
01010 
01011     dsputil_init(&dsp, avctx);
01012 
01013     q->pUnits = av_mallocz(sizeof(channel_unit)*q->channels);
01014     if (!q->pUnits) {
01015         av_free(q->decoded_bytes_buffer);
01016         return AVERROR(ENOMEM);
01017     }
01018 
01019     avctx->sample_fmt = AV_SAMPLE_FMT_S16;
01020     return 0;
01021 }
01022 
01023 
01024 AVCodec ff_atrac3_decoder =
01025 {
01026     .name = "atrac3",
01027     .type = AVMEDIA_TYPE_AUDIO,
01028     .id = CODEC_ID_ATRAC3,
01029     .priv_data_size = sizeof(ATRAC3Context),
01030     .init = atrac3_decode_init,
01031     .close = atrac3_decode_close,
01032     .decode = atrac3_decode_frame,
01033     .long_name = NULL_IF_CONFIG_SMALL("Atrac 3 (Adaptive TRansform Acoustic Coding 3)"),
01034 };