Libav 0.7.1
|
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 Libav. 00007 * 00008 * Libav 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 * Libav 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 Libav; 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 };