X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fdca.c;h=64fe38e2ae434ff990e3f682c300aae06caeae08;hb=120b4557f3deb855d3438a336a1e610d03c1d8ce;hp=5081eaae93920e647c67b2a9472352d5d0637a36;hpb=6369e6ebc3ee644e990324b8612c85edbd8141a0;p=ffmpeg diff --git a/libavcodec/dca.c b/libavcodec/dca.c index 5081eaae939..64fe38e2ae4 100644 --- a/libavcodec/dca.c +++ b/libavcodec/dca.c @@ -23,7 +23,7 @@ */ /** - * @file dca.c + * @file libavcodec/dca.c */ #include @@ -32,18 +32,11 @@ #include "avcodec.h" #include "dsputil.h" -#include "bitstream.h" +#include "get_bits.h" +#include "put_bits.h" #include "dcadata.h" #include "dcahuff.h" -#include "parser.h" - -/** DCA syncwords, also used for bitstream type detection */ -//@{ -#define DCA_MARKER_RAW_BE 0x7FFE8001 -#define DCA_MARKER_RAW_LE 0xFE7F0180 -#define DCA_MARKER_14B_BE 0x1FFFE800 -#define DCA_MARKER_14B_LE 0xFF1F00E8 -//@} +#include "dca.h" //#define TRACE @@ -67,6 +60,78 @@ enum DCAMode { DCA_4F2R }; +/* Tables for mapping dts channel configurations to libavcodec multichannel api. + * Some compromises have been made for special configurations. Most configurations + * are never used so complete accuracy is not needed. + * + * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead. + * S -> side, when both rear and back are configured move one of them to the side channel + * OV -> center back + * All 2 channel configurations -> CH_LAYOUT_STEREO + */ + +static const int64_t dca_core_channel_layout[] = { + CH_FRONT_CENTER, ///< 1, A + CH_LAYOUT_STEREO, ///< 2, A + B (dual mono) + CH_LAYOUT_STEREO, ///< 2, L + R (stereo) + CH_LAYOUT_STEREO, ///< 2, (L+R) + (L-R) (sum-difference) + CH_LAYOUT_STEREO, ///< 2, LT +RT (left and right total) + CH_LAYOUT_STEREO|CH_FRONT_CENTER, ///< 3, C+L+R + CH_LAYOUT_STEREO|CH_BACK_CENTER, ///< 3, L+R+S + CH_LAYOUT_STEREO|CH_FRONT_CENTER|CH_BACK_CENTER, ///< 4, C + L + R+ S + CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_SIDE_RIGHT, ///< 4, L + R +SL+ SR + CH_LAYOUT_STEREO|CH_FRONT_CENTER|CH_SIDE_LEFT|CH_SIDE_RIGHT, ///< 5, C + L + R+ SL+SR + CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_SIDE_RIGHT|CH_FRONT_LEFT_OF_CENTER|CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR + CH_LAYOUT_STEREO|CH_BACK_LEFT|CH_BACK_RIGHT|CH_FRONT_CENTER|CH_BACK_CENTER, ///< 6, C + L + R+ LR + RR + OV + CH_FRONT_CENTER|CH_FRONT_RIGHT_OF_CENTER|CH_FRONT_LEFT_OF_CENTER|CH_BACK_CENTER|CH_BACK_LEFT|CH_BACK_RIGHT, ///< 6, CF+ CR+LF+ RF+LR + RR + CH_FRONT_LEFT_OF_CENTER|CH_FRONT_CENTER|CH_FRONT_RIGHT_OF_CENTER|CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR + CH_FRONT_LEFT_OF_CENTER|CH_FRONT_RIGHT_OF_CENTER|CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_SIDE_RIGHT|CH_BACK_LEFT|CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2+ SR1 + SR2 + CH_FRONT_LEFT_OF_CENTER|CH_FRONT_CENTER|CH_FRONT_RIGHT_OF_CENTER|CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_BACK_CENTER|CH_SIDE_RIGHT, ///< 8, CL + C+ CR + L + R + SL + S+ SR +}; + +static const int8_t dca_lfe_index[] = { + 1,2,2,2,2,3,2,3,2,3,2,3,1,3,2,3 +}; + +static const int8_t dca_channel_reorder_lfe[][8] = { + { 0, -1, -1, -1, -1, -1, -1, -1}, + { 0, 1, -1, -1, -1, -1, -1, -1}, + { 0, 1, -1, -1, -1, -1, -1, -1}, + { 0, 1, -1, -1, -1, -1, -1, -1}, + { 0, 1, -1, -1, -1, -1, -1, -1}, + { 2, 0, 1, -1, -1, -1, -1, -1}, + { 0, 1, 3, -1, -1, -1, -1, -1}, + { 2, 0, 1, 4, -1, -1, -1, -1}, + { 0, 1, 3, 4, -1, -1, -1, -1}, + { 2, 0, 1, 4, 5, -1, -1, -1}, + { 3, 4, 0, 1, 5, 6, -1, -1}, + { 2, 0, 1, 4, 5, 6, -1, -1}, + { 0, 6, 4, 5, 2, 3, -1, -1}, + { 4, 2, 5, 0, 1, 6, 7, -1}, + { 5, 6, 0, 1, 7, 3, 8, 4}, + { 4, 2, 5, 0, 1, 6, 8, 7}, +}; + +static const int8_t dca_channel_reorder_nolfe[][8] = { + { 0, -1, -1, -1, -1, -1, -1, -1}, + { 0, 1, -1, -1, -1, -1, -1, -1}, + { 0, 1, -1, -1, -1, -1, -1, -1}, + { 0, 1, -1, -1, -1, -1, -1, -1}, + { 0, 1, -1, -1, -1, -1, -1, -1}, + { 2, 0, 1, -1, -1, -1, -1, -1}, + { 0, 1, 2, -1, -1, -1, -1, -1}, + { 2, 0, 1, 3, -1, -1, -1, -1}, + { 0, 1, 2, 3, -1, -1, -1, -1}, + { 2, 0, 1, 3, 4, -1, -1, -1}, + { 2, 3, 0, 1, 4, 5, -1, -1}, + { 2, 0, 1, 3, 4, 5, -1, -1}, + { 0, 5, 3, 4, 1, 2, -1, -1}, + { 3, 2, 4, 0, 1, 5, 6, -1}, + { 4, 5, 0, 1, 6, 2, 7, 3}, + { 3, 2, 4, 0, 1, 5, 7, 6}, +}; + + #define DCA_DOLBY 101 /* FIXME */ #define DCA_CHANNEL_BITS 6 @@ -75,9 +140,8 @@ enum DCAMode { #define DCA_LFE 0x80 #define HEADER_SIZE 14 -#define CONVERT_BIAS 384 -#define DCA_MAX_FRAME_SIZE 16383 +#define DCA_MAX_FRAME_SIZE 16384 /** Bit allocation */ typedef struct { @@ -92,10 +156,7 @@ static BitAlloc dca_tmode; ///< transition mode VLCs static BitAlloc dca_scalefactor; ///< scalefactor VLCs static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs -/** Pre-calculated cosine modulation coefs for the QMF */ -static float cos_mod[544]; - -static int av_always_inline get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx) +static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx) { return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) + ba->offset; } @@ -111,6 +172,7 @@ typedef struct { int amode; ///< audio channels arrangement int sample_rate; ///< audio sampling rate int bit_rate; ///< transmission bit rate + int bit_rate_index; ///< transmission bit rate index int downmix; ///< embedded downmix enabled int dynrange; ///< embedded dynamic range flag @@ -133,6 +195,7 @@ typedef struct { /* Primary audio coding header */ int subframes; ///< number of subframes + int total_channels; ///< number of channels including extensions int prim_channels; ///< number of primary audio channels int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband @@ -164,18 +227,21 @@ typedef struct { /* Subband samples history (for ADPCM) */ float subband_samples_hist[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4]; - float subband_fir_hist[DCA_PRIM_CHANNELS_MAX][512]; - float subband_fir_noidea[DCA_PRIM_CHANNELS_MAX][64]; + DECLARE_ALIGNED_16(float, subband_fir_hist[DCA_PRIM_CHANNELS_MAX][512]); + float subband_fir_noidea[DCA_PRIM_CHANNELS_MAX][32]; + int hist_index[DCA_PRIM_CHANNELS_MAX]; int output; ///< type of output - int bias; ///< output bias + float add_bias; ///< output bias + float scale_bias; ///< output scale DECLARE_ALIGNED_16(float, samples[1536]); /* 6 * 256 = 1536, might only need 5 */ - DECLARE_ALIGNED_16(int16_t, tsamples[1536]); + const float *samples_chanptr[6]; uint8_t dca_buffer[DCA_MAX_FRAME_SIZE]; int dca_buffer_size; ///< how much data is in the dca_buffer + const int8_t* channel_order_tab; ///< channel reordering table, lfe and non lfe GetBitContext gb; /* Current position in DCA frame */ int current_subframe; @@ -183,34 +249,35 @@ typedef struct { int debug_flag; ///< used for suppressing repeated error messages output DSPContext dsp; + MDCTContext imdct; } DCAContext; -static void dca_init_vlcs(void) +static av_cold void dca_init_vlcs(void) { - static int vlcs_inited = 0; + static int vlcs_initialized = 0; int i, j; - if (vlcs_inited) + if (vlcs_initialized) return; dca_bitalloc_index.offset = 1; - dca_bitalloc_index.wrap = 1; + dca_bitalloc_index.wrap = 2; for (i = 0; i < 5; i++) init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12, bitalloc_12_bits[i], 1, 1, - bitalloc_12_codes[i], 2, 2, 1); + bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_STATIC); dca_scalefactor.offset = -64; dca_scalefactor.wrap = 2; for (i = 0; i < 5; i++) init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129, scales_bits[i], 1, 1, - scales_codes[i], 2, 2, 1); + scales_codes[i], 2, 2, INIT_VLC_USE_STATIC); dca_tmode.offset = 0; dca_tmode.wrap = 1; for (i = 0; i < 4; i++) init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4, tmode_bits[i], 1, 1, - tmode_codes[i], 2, 2, 1); + tmode_codes[i], 2, 2, INIT_VLC_USE_STATIC); for(i = 0; i < 10; i++) for(j = 0; j < 7; j++){ @@ -220,9 +287,9 @@ static void dca_init_vlcs(void) init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j], bitalloc_sizes[i], bitalloc_bits[i][j], 1, 1, - bitalloc_codes[i][j], 2, 2, 1); + bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_STATIC); } - vlcs_inited = 1; + vlcs_initialized = 1; } static inline void get_array(GetBitContext *gb, int *dst, int len, int bits) @@ -238,8 +305,6 @@ static int dca_parse_frame_header(DCAContext * s) static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 }; static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 }; - s->bias = CONVERT_BIAS; - init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8); /* Sync code */ @@ -257,7 +322,8 @@ static int dca_parse_frame_header(DCAContext * s) s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)]; if (!s->sample_rate) return -1; - s->bit_rate = dca_bit_rates[get_bits(&s->gb, 5)]; + s->bit_rate_index = get_bits(&s->gb, 5); + s->bit_rate = dca_bit_rates[s->bit_rate_index]; if (!s->bit_rate) return -1; @@ -297,10 +363,10 @@ static int dca_parse_frame_header(DCAContext * s) av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size); av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n", s->amode, dca_channels[s->amode]); - av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i (%i Hz)\n", - s->sample_rate, dca_sample_rates[s->sample_rate]); - av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i (%i bits/s)\n", - s->bit_rate, dca_bit_rates[s->bit_rate]); + av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n", + s->sample_rate); + av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n", + s->bit_rate); av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix); av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange); av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp); @@ -328,7 +394,10 @@ static int dca_parse_frame_header(DCAContext * s) /* Primary audio coding header */ s->subframes = get_bits(&s->gb, 4) + 1; - s->prim_channels = get_bits(&s->gb, 3) + 1; + s->total_channels = get_bits(&s->gb, 3) + 1; + s->prim_channels = s->total_channels; + if (s->prim_channels > DCA_PRIM_CHANNELS_MAX) + s->prim_channels = DCA_PRIM_CHANNELS_MAX; /* We only support DTS core */ for (i = 0; i < s->prim_channels; i++) { @@ -435,7 +504,11 @@ static int dca_subframe_header(DCAContext * s) s->bitalloc[j][k] = get_bits(&s->gb, 5); else if (s->bitalloc_huffman[j] == 5) s->bitalloc[j][k] = get_bits(&s->gb, 4); - else { + else if (s->bitalloc_huffman[j] == 7) { + av_log(s->avctx, AV_LOG_ERROR, + "Invalid bit allocation index\n"); + return -1; + } else { s->bitalloc[j][k] = get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]); } @@ -461,15 +534,15 @@ static int dca_subframe_header(DCAContext * s) } for (j = 0; j < s->prim_channels; j++) { - uint32_t *scale_table; + const uint32_t *scale_table; int scale_sum; memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2); if (s->scalefactor_huffman[j] == 6) - scale_table = (uint32_t *) scale_factor_quant7; + scale_table = scale_factor_quant7; else - scale_table = (uint32_t *) scale_factor_quant6; + scale_table = scale_factor_quant6; /* When huffman coded, only the difference is encoded */ scale_sum = 0; @@ -620,6 +693,7 @@ static int dca_subframe_header(DCAContext * s) } for (j = 0; j < s->prim_channels; j++) { if (s->joint_intensity[j] > 0) { + int source_channel = s->joint_intensity[j] - 1; av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n"); for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]); @@ -638,6 +712,7 @@ static int dca_subframe_header(DCAContext * s) for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++) av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]); if(s->lfe){ + int lfe_samples = 2 * s->lfe * s->subsubframes; av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n"); for (j = lfe_samples; j < lfe_samples * 2; j++) av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]); @@ -652,69 +727,64 @@ static void qmf_32_subbands(DCAContext * s, int chans, float samples_in[32][8], float *samples_out, float scale, float bias) { - float *prCoeff; - int i, j, k; - float praXin[33], *raXin = &praXin[1]; + const float *prCoeff; + int i, j; + DECLARE_ALIGNED_16(float, raXin[32]); - float *subband_fir_hist = s->subband_fir_hist[chans]; + int hist_index= s->hist_index[chans]; float *subband_fir_hist2 = s->subband_fir_noidea[chans]; - int chindex = 0, subindex; + int subindex; - praXin[0] = 0.0; + scale *= sqrt(1/8.0); /* Select filter */ if (!s->multirate_inter) /* Non-perfect reconstruction */ - prCoeff = (float *) fir_32bands_nonperfect; + prCoeff = fir_32bands_nonperfect; else /* Perfect reconstruction */ - prCoeff = (float *) fir_32bands_perfect; + prCoeff = fir_32bands_perfect; /* Reconstructed channel sample index */ for (subindex = 0; subindex < 8; subindex++) { - float t1, t2, sum[16], diff[16]; - + float *subband_fir_hist = s->subband_fir_hist[chans] + hist_index; /* Load in one sample from each subband and clear inactive subbands */ - for (i = 0; i < s->subband_activity[chans]; i++) - raXin[i] = samples_in[i][subindex]; + for (i = 0; i < s->subband_activity[chans]; i++){ + if((i-1)&2) raXin[i] = -samples_in[i][subindex]; + else raXin[i] = samples_in[i][subindex]; + } for (; i < 32; i++) raXin[i] = 0.0; - /* Multiply by cosine modulation coefficients and - * create temporary arrays SUM and DIFF */ - for (j = 0, k = 0; k < 16; k++) { - t1 = 0.0; - t2 = 0.0; - for (i = 0; i < 16; i++, j++){ - t1 += (raXin[2 * i] + raXin[2 * i + 1]) * cos_mod[j]; - t2 += (raXin[2 * i] + raXin[2 * i - 1]) * cos_mod[j + 256]; - } - sum[k] = t1 + t2; - diff[k] = t1 - t2; - } - - j = 512; - /* Store history */ - for (k = 0; k < 16; k++) - subband_fir_hist[k] = cos_mod[j++] * sum[k]; - for (k = 0; k < 16; k++) - subband_fir_hist[32-k-1] = cos_mod[j++] * diff[k]; + ff_imdct_half(&s->imdct, subband_fir_hist, raXin); /* Multiply by filter coefficients */ - for (k = 31, i = 0; i < 32; i++, k--) - for (j = 0; j < 512; j += 64){ - subband_fir_hist2[i] += prCoeff[i+j] * ( subband_fir_hist[i+j] - subband_fir_hist[j+k]); - subband_fir_hist2[i+32] += prCoeff[i+j+32]*(-subband_fir_hist[i+j] - subband_fir_hist[j+k]); + for (i = 0; i < 16; i++){ + float a= subband_fir_hist2[i ]; + float b= subband_fir_hist2[i+16]; + float c= 0; + float d= 0; + for (j = 0; j < 512-hist_index; j += 64){ + a += prCoeff[i+j ]*(-subband_fir_hist[15-i+j]); + b += prCoeff[i+j+16]*( subband_fir_hist[ i+j]); + c += prCoeff[i+j+32]*( subband_fir_hist[16+i+j]); + d += prCoeff[i+j+48]*( subband_fir_hist[31-i+j]); } + for ( ; j < 512; j += 64){ + a += prCoeff[i+j ]*(-subband_fir_hist[15-i+j-512]); + b += prCoeff[i+j+16]*( subband_fir_hist[ i+j-512]); + c += prCoeff[i+j+32]*( subband_fir_hist[16+i+j-512]); + d += prCoeff[i+j+48]*( subband_fir_hist[31-i+j-512]); + } + samples_out[i ] = a * scale + bias; + samples_out[i+16] = b * scale + bias; + subband_fir_hist2[i ] = c; + subband_fir_hist2[i+16] = d; + } + samples_out+= 32; - /* Create 32 PCM output samples */ - for (i = 0; i < 32; i++) - samples_out[chindex++] = subband_fir_hist2[i] * scale + bias; - - /* Update working arrays */ - memmove(&subband_fir_hist[32], &subband_fir_hist[0], (512 - 32) * sizeof(float)); - memmove(&subband_fir_hist2[0], &subband_fir_hist2[32], 32 * sizeof(float)); - memset(&subband_fir_hist2[32], 0, 32 * sizeof(float)); + hist_index = (hist_index-32)&511; } + s->hist_index[chans]= hist_index; } static void lfe_interpolation_fir(int decimation_select, @@ -752,7 +822,7 @@ static void lfe_interpolation_fir(int decimation_select, //FIXME the coeffs are symetric, fix that for (j = 0; j < 512 / decifactor; j++) rTmp += samples_in[deciindex - j] * prCoeff[k + j * decifactor]; - samples_out[interp_index++] = rTmp / scale + bias; + samples_out[interp_index++] = (rTmp * scale) + bias; } } } @@ -848,7 +918,7 @@ static int dca_subsubframe(DCAContext * s) int k, l; int subsubframe = s->current_subsubframe; - float *quant_step_table; + const float *quant_step_table; /* FIXME */ float subband_samples[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8]; @@ -858,10 +928,10 @@ static int dca_subsubframe(DCAContext * s) */ /* Select quantization step size table */ - if (s->bit_rate == 0x1f) - quant_step_table = (float *) lossless_quant_d; + if (s->bit_rate_index == 0x1f) + quant_step_table = lossless_quant_d; else - quant_step_table = (float *) lossy_quant_d; + quant_step_table = lossy_quant_d; for (k = 0; k < s->prim_channels; k++) { for (l = 0; l < s->vq_start_subband[k]; l++) { @@ -987,9 +1057,9 @@ static int dca_subsubframe(DCAContext * s) for (k = 0; k < s->prim_channels; k++) { /* static float pcm_to_double[8] = {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/ - qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * k], - 2.0 / 3 /*pcm_to_double[s->source_pcm_res] */ , - 0 /*s->bias */ ); + qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]], + M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ , + s->add_bias ); } /* Down mixing */ @@ -1001,13 +1071,12 @@ static int dca_subsubframe(DCAContext * s) /* Generate LFE samples for this subsubframe FIXME!!! */ if (s->output & DCA_LFE) { int lfe_samples = 2 * s->lfe * s->subsubframes; - int i_channels = dca_channels[s->output & DCA_CHANNEL_MASK]; lfe_interpolation_fir(s->lfe, 2 * s->lfe, s->lfe_data + lfe_samples + 2 * s->lfe * subsubframe, - &s->samples[256 * i_channels], - 8388608.0, s->bias); + &s->samples[256 * dca_lfe_index[s->amode]], + (1.0/256.0)*s->scale_bias, s->add_bias); /* Outputs 20bits pcm samples */ } @@ -1097,26 +1166,30 @@ static int dca_decode_block(DCAContext * s) /** * Convert bitstream to one representation based on sync marker */ -static int dca_convert_bitstream(uint8_t * src, int src_size, uint8_t * dst, +static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst, int max_size) { uint32_t mrk; int i, tmp; - uint16_t *ssrc = (uint16_t *) src, *sdst = (uint16_t *) dst; + const uint16_t *ssrc = (const uint16_t *) src; + uint16_t *sdst = (uint16_t *) dst; PutBitContext pb; - if((unsigned)src_size > (unsigned)max_size) - return -1; + if((unsigned)src_size > (unsigned)max_size) { +// av_log(NULL, AV_LOG_ERROR, "Input frame size larger then DCA_MAX_FRAME_SIZE!\n"); +// return -1; + src_size = max_size; + } mrk = AV_RB32(src); switch (mrk) { case DCA_MARKER_RAW_BE: - memcpy(dst, src, FFMIN(src_size, max_size)); - return FFMIN(src_size, max_size); + memcpy(dst, src, src_size); + return src_size; case DCA_MARKER_RAW_LE: - for (i = 0; i < (FFMIN(src_size, max_size) + 1) >> 1; i++) + for (i = 0; i < (src_size + 1) >> 1; i++) *sdst++ = bswap_16(*ssrc++); - return FFMIN(src_size, max_size); + return src_size; case DCA_MARKER_14B_BE: case DCA_MARKER_14B_LE: init_put_bits(&pb, dst, max_size); @@ -1137,10 +1210,12 @@ static int dca_convert_bitstream(uint8_t * src, int src_size, uint8_t * dst, */ static int dca_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; - int i, j, k; + int i; int16_t *samples = data; DCAContext *s = avctx->priv_data; int channels; @@ -1148,13 +1223,14 @@ static int dca_decode_frame(AVCodecContext * avctx, s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer, DCA_MAX_FRAME_SIZE); if (s->dca_buffer_size == -1) { - av_log(avctx, AV_LOG_ERROR, "Not a DCA frame\n"); + av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n"); return -1; } init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8); if (dca_parse_frame_header(s) < 0) { //seems like the frame is corrupt, try with the next one + *data_size=0; return buf_size; } //set AVCodec values with parsed data @@ -1162,35 +1238,42 @@ static int dca_decode_frame(AVCodecContext * avctx, avctx->bit_rate = s->bit_rate; channels = s->prim_channels + !!s->lfe; - if(avctx->channels == 0) { - avctx->channels = channels; - } else if(channels < avctx->channels) { - av_log(avctx, AV_LOG_WARNING, "DTS source channels are less than " - "specified: output to %d channels.\n", channels); - avctx->channels = channels; - } - if(avctx->channels == 2) { - s->output = DCA_STEREO; - } else if(avctx->channels != channels) { - av_log(avctx, AV_LOG_ERROR, "Cannot downmix DTS to %d channels.\n", - avctx->channels); + + if (s->amode<16) { + avctx->channel_layout = dca_core_channel_layout[s->amode]; + + if (s->lfe) { + avctx->channel_layout |= CH_LOW_FREQUENCY; + s->channel_order_tab = dca_channel_reorder_lfe[s->amode]; + } else + s->channel_order_tab = dca_channel_reorder_nolfe[s->amode]; + + if(avctx->request_channels == 2 && s->prim_channels > 2) { + channels = 2; + s->output = DCA_STEREO; + avctx->channel_layout = CH_LAYOUT_STEREO; + } + } else { + av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode); return -1; } - channels = avctx->channels; + + /* There is nothing that prevents a dts frame to change channel configuration + but FFmpeg doesn't support that so only set the channels if it is previously + unset. Ideally during the first probe for channels the crc should be checked + and only set avctx->channels when the crc is ok. Right now the decoder could + set the channels based on a broken first frame.*/ + if (!avctx->channels) + avctx->channels = channels; + if(*data_size < (s->sample_blocks / 8) * 256 * sizeof(int16_t) * channels) return -1; - *data_size = 0; + *data_size = 256 / 8 * s->sample_blocks * sizeof(int16_t) * channels; for (i = 0; i < (s->sample_blocks / 8); i++) { dca_decode_block(s); - s->dsp.float_to_int16(s->tsamples, s->samples, 256 * channels); - /* interleave samples */ - for (j = 0; j < 256; j++) { - for (k = 0; k < channels; k++) - samples[k] = s->tsamples[j + k * 256]; - samples += channels; - } - *data_size += 256 * sizeof(int16_t) * channels; + s->dsp.float_to_int16_interleave(samples, s->samples_chanptr, 256, channels); + samples += 256 * channels; } return buf_size; @@ -1198,159 +1281,59 @@ static int dca_decode_frame(AVCodecContext * avctx, -/** - * Build the cosine modulation tables for the QMF - * - * @param s pointer to the DCAContext - */ - -static void pre_calc_cosmod(DCAContext * s) -{ - int i, j, k; - static int cosmod_inited = 0; - - if(cosmod_inited) return; - for (j = 0, k = 0; k < 16; k++) - for (i = 0; i < 16; i++) - cos_mod[j++] = cos((2 * i + 1) * (2 * k + 1) * M_PI / 64); - - for (k = 0; k < 16; k++) - for (i = 0; i < 16; i++) - cos_mod[j++] = cos((i) * (2 * k + 1) * M_PI / 32); - - for (k = 0; k < 16; k++) - cos_mod[j++] = 0.25 / (2 * cos((2 * k + 1) * M_PI / 128)); - - for (k = 0; k < 16; k++) - cos_mod[j++] = -0.25 / (2.0 * sin((2 * k + 1) * M_PI / 128)); - - cosmod_inited = 1; -} - - /** * DCA initialization * * @param avctx pointer to the AVCodecContext */ -static int dca_decode_init(AVCodecContext * avctx) +static av_cold int dca_decode_init(AVCodecContext * avctx) { DCAContext *s = avctx->priv_data; + int i; s->avctx = avctx; dca_init_vlcs(); - pre_calc_cosmod(s); dsputil_init(&s->dsp, avctx); - return 0; -} - + ff_mdct_init(&s->imdct, 6, 1); -AVCodec dca_decoder = { - .name = "dca", - .type = CODEC_TYPE_AUDIO, - .id = CODEC_ID_DTS, - .priv_data_size = sizeof(DCAContext), - .init = dca_decode_init, - .decode = dca_decode_frame, -}; - -#ifdef CONFIG_DCA_PARSER + for(i = 0; i < 6; i++) + s->samples_chanptr[i] = s->samples + i * 256; + avctx->sample_fmt = SAMPLE_FMT_S16; -typedef struct DCAParseContext { - ParseContext pc; - uint32_t lastmarker; -} DCAParseContext; - -#define IS_MARKER(state, i, buf, buf_size) \ - ((state == DCA_MARKER_14B_LE && (i < buf_size-2) && (buf[i+1] & 0xF0) == 0xF0 && buf[i+2] == 0x07) \ - || (state == DCA_MARKER_14B_BE && (i < buf_size-2) && buf[i+1] == 0x07 && (buf[i+2] & 0xF0) == 0xF0) \ - || state == DCA_MARKER_RAW_LE || state == DCA_MARKER_RAW_BE) + if(s->dsp.float_to_int16 == ff_float_to_int16_c) { + s->add_bias = 385.0f; + s->scale_bias = 1.0 / 32768.0; + } else { + s->add_bias = 0.0f; + s->scale_bias = 1.0; -/** - * finds the end of the current frame in the bitstream. - * @return the position of the first byte of the next frame, or -1 - */ -static int dca_find_frame_end(DCAParseContext * pc1, const uint8_t * buf, - int buf_size) -{ - int start_found, i; - uint32_t state; - ParseContext *pc = &pc1->pc; - - start_found = pc->frame_start_found; - state = pc->state; - - i = 0; - if (!start_found) { - for (i = 0; i < buf_size; i++) { - state = (state << 8) | buf[i]; - if (IS_MARKER(state, i, buf, buf_size)) { - if (pc1->lastmarker && state == pc1->lastmarker) { - start_found = 1; - break; - } else if (!pc1->lastmarker) { - start_found = 1; - pc1->lastmarker = state; - break; - } - } - } - } - if (start_found) { - for (; i < buf_size; i++) { - state = (state << 8) | buf[i]; - if (state == pc1->lastmarker && IS_MARKER(state, i, buf, buf_size)) { - pc->frame_start_found = 0; - pc->state = -1; - return i - 3; - } + /* allow downmixing to stereo */ + if (avctx->channels > 0 && avctx->request_channels < avctx->channels && + avctx->request_channels == 2) { + avctx->channels = avctx->request_channels; } } - pc->frame_start_found = start_found; - pc->state = state; - return END_NOT_FOUND; -} -static int dca_parse_init(AVCodecParserContext * s) -{ - DCAParseContext *pc1 = s->priv_data; - pc1->lastmarker = 0; return 0; } -static int dca_parse(AVCodecParserContext * s, - AVCodecContext * avctx, - uint8_t ** poutbuf, int *poutbuf_size, - const uint8_t * buf, int buf_size) +static av_cold int dca_decode_end(AVCodecContext * avctx) { - DCAParseContext *pc1 = s->priv_data; - ParseContext *pc = &pc1->pc; - int next; - - if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) { - next = buf_size; - } else { - next = dca_find_frame_end(pc1, buf, buf_size); - - if (ff_combine_frame(pc, next, (uint8_t **) & buf, &buf_size) < 0) { - *poutbuf = NULL; - *poutbuf_size = 0; - return buf_size; - } - } - *poutbuf = (uint8_t *) buf; - *poutbuf_size = buf_size; - return next; + DCAContext *s = avctx->priv_data; + ff_mdct_end(&s->imdct); + return 0; } -AVCodecParser dca_parser = { - {CODEC_ID_DTS}, - sizeof(DCAParseContext), - dca_parse_init, - dca_parse, - ff_parse_close, +AVCodec dca_decoder = { + .name = "dca", + .type = CODEC_TYPE_AUDIO, + .id = CODEC_ID_DTS, + .priv_data_size = sizeof(DCAContext), + .init = dca_decode_init, + .decode = dca_decode_frame, + .close = dca_decode_end, + .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"), }; -#endif /* CONFIG_DCA_PARSER */