X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fdcadec.c;h=cd4432368c19525fbe207b703b961a75244510d1;hb=2fb6acd9c28907e4f8c0510099a4603ea6caf861;hp=eb12eb2db9dcf98527c7f0beb3b424f797aa5d7b;hpb=61d5313d94c9e6a86d599de781f287b577b7fe11;p=ffmpeg diff --git a/libavcodec/dcadec.c b/libavcodec/dcadec.c index eb12eb2db9d..cd4432368c1 100644 --- a/libavcodec/dcadec.c +++ b/libavcodec/dcadec.c @@ -4,6 +4,8 @@ * Copyright (C) 2004 Benjamin Zores * Copyright (C) 2006 Benjamin Larsson * Copyright (C) 2007 Konstantin Shishkov + * Copyright (C) 2012 Paul B Mahol + * Copyright (C) 2014 Niels Möller * * This file is part of Libav. * @@ -26,40 +28,35 @@ #include #include +#include "libavutil/attributes.h" +#include "libavutil/channel_layout.h" #include "libavutil/common.h" #include "libavutil/float_dsp.h" -#include "libavutil/intmath.h" +#include "libavutil/internal.h" #include "libavutil/intreadwrite.h" #include "libavutil/mathematics.h" -#include "libavutil/audioconvert.h" +#include "libavutil/opt.h" #include "libavutil/samplefmt.h" + #include "avcodec.h" -#include "dsputil.h" +#include "dca.h" +#include "dca_syncwords.h" +#include "dcadata.h" +#include "dcadsp.h" +#include "dcahuff.h" #include "fft.h" +#include "fmtconvert.h" #include "get_bits.h" +#include "internal.h" +#include "mathops.h" +#include "profiles.h" #include "put_bits.h" -#include "dcadata.h" -#include "dcahuff.h" -#include "dca.h" -#include "dca_parser.h" #include "synth_filter.h" -#include "dcadsp.h" -#include "fmtconvert.h" #if ARCH_ARM # include "arm/dca.h" #endif -//#define TRACE - -#define DCA_PRIM_CHANNELS_MAX (7) -#define DCA_SUBBANDS (32) -#define DCA_ABITS_MAX (32) /* Should be 28 */ -#define DCA_SUBSUBFRAMES_MAX (4) -#define DCA_SUBFRAMES_MAX (16) -#define DCA_BLOCKS_MAX (16) -#define DCA_LFE_MAX (3) - enum DCAMode { DCA_MONO = 0, DCA_CHANNEL, @@ -74,39 +71,6 @@ enum DCAMode { DCA_4F2R }; -/* these are unconfirmed but should be mostly correct */ -enum DCAExSSSpeakerMask { - DCA_EXSS_FRONT_CENTER = 0x0001, - DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002, - DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004, - DCA_EXSS_LFE = 0x0008, - DCA_EXSS_REAR_CENTER = 0x0010, - DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020, - DCA_EXSS_REAR_LEFT_RIGHT = 0x0040, - DCA_EXSS_FRONT_HIGH_CENTER = 0x0080, - DCA_EXSS_OVERHEAD = 0x0100, - DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200, - DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400, - DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800, - DCA_EXSS_LFE2 = 0x1000, - DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000, - DCA_EXSS_REAR_HIGH_CENTER = 0x4000, - DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000, -}; - -enum DCAExtensionMask { - DCA_EXT_CORE = 0x001, ///< core in core substream - DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream - DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream - DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream - DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream) - DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS - DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS - DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS - DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS - DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS -}; - /* -1 are reserved or unknown */ static const int dca_ext_audio_descr_mask[] = { DCA_EXT_XCH, @@ -119,9 +83,6 @@ static const int dca_ext_audio_descr_mask[] = { -1, }; -/* extensions that reside in core substream */ -#define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96) - /* 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. @@ -168,86 +129,6 @@ static const uint64_t dca_core_channel_layout[] = { AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_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[][9] = { - { 0, -1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 2, 0, 1, -1, -1, -1, -1, -1, -1}, - { 0, 1, 3, -1, -1, -1, -1, -1, -1}, - { 2, 0, 1, 4, -1, -1, -1, -1, -1}, - { 0, 1, 3, 4, -1, -1, -1, -1, -1}, - { 2, 0, 1, 4, 5, -1, -1, -1, -1}, - { 3, 4, 0, 1, 5, 6, -1, -1, -1}, - { 2, 0, 1, 4, 5, 6, -1, -1, -1}, - { 0, 6, 4, 5, 2, 3, -1, -1, -1}, - { 4, 2, 5, 0, 1, 6, 7, -1, -1}, - { 5, 6, 0, 1, 7, 3, 8, 4, -1}, - { 4, 2, 5, 0, 1, 6, 8, 7, -1}, -}; - -static const int8_t dca_channel_reorder_lfe_xch[][9] = { - { 0, 2, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, 3, -1, -1, -1, -1, -1, -1}, - { 0, 1, 3, -1, -1, -1, -1, -1, -1}, - { 0, 1, 3, -1, -1, -1, -1, -1, -1}, - { 0, 1, 3, -1, -1, -1, -1, -1, -1}, - { 2, 0, 1, 4, -1, -1, -1, -1, -1}, - { 0, 1, 3, 4, -1, -1, -1, -1, -1}, - { 2, 0, 1, 4, 5, -1, -1, -1, -1}, - { 0, 1, 4, 5, 3, -1, -1, -1, -1}, - { 2, 0, 1, 5, 6, 4, -1, -1, -1}, - { 3, 4, 0, 1, 6, 7, 5, -1, -1}, - { 2, 0, 1, 4, 5, 6, 7, -1, -1}, - { 0, 6, 4, 5, 2, 3, 7, -1, -1}, - { 4, 2, 5, 0, 1, 7, 8, 6, -1}, - { 5, 6, 0, 1, 8, 3, 9, 4, 7}, - { 4, 2, 5, 0, 1, 6, 9, 8, 7}, -}; - -static const int8_t dca_channel_reorder_nolfe[][9] = { - { 0, -1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 2, 0, 1, -1, -1, -1, -1, -1, -1}, - { 0, 1, 2, -1, -1, -1, -1, -1, -1}, - { 2, 0, 1, 3, -1, -1, -1, -1, -1}, - { 0, 1, 2, 3, -1, -1, -1, -1, -1}, - { 2, 0, 1, 3, 4, -1, -1, -1, -1}, - { 2, 3, 0, 1, 4, 5, -1, -1, -1}, - { 2, 0, 1, 3, 4, 5, -1, -1, -1}, - { 0, 5, 3, 4, 1, 2, -1, -1, -1}, - { 3, 2, 4, 0, 1, 5, 6, -1, -1}, - { 4, 5, 0, 1, 6, 2, 7, 3, -1}, - { 3, 2, 4, 0, 1, 5, 7, 6, -1}, -}; - -static const int8_t dca_channel_reorder_nolfe_xch[][9] = { - { 0, 1, -1, -1, -1, -1, -1, -1, -1}, - { 0, 1, 2, -1, -1, -1, -1, -1, -1}, - { 0, 1, 2, -1, -1, -1, -1, -1, -1}, - { 0, 1, 2, -1, -1, -1, -1, -1, -1}, - { 0, 1, 2, -1, -1, -1, -1, -1, -1}, - { 2, 0, 1, 3, -1, -1, -1, -1, -1}, - { 0, 1, 2, 3, -1, -1, -1, -1, -1}, - { 2, 0, 1, 3, 4, -1, -1, -1, -1}, - { 0, 1, 3, 4, 2, -1, -1, -1, -1}, - { 2, 0, 1, 4, 5, 3, -1, -1, -1}, - { 2, 3, 0, 1, 5, 6, 4, -1, -1}, - { 2, 0, 1, 3, 4, 5, 6, -1, -1}, - { 0, 5, 3, 4, 1, 2, 6, -1, -1}, - { 3, 2, 4, 0, 1, 6, 7, 5, -1}, - { 4, 5, 0, 1, 7, 2, 8, 3, 6}, - { 3, 2, 4, 0, 1, 5, 8, 7, 6}, -}; - #define DCA_DOLBY 101 /* FIXME */ #define DCA_CHANNEL_BITS 6 @@ -257,13 +138,10 @@ static const int8_t dca_channel_reorder_nolfe_xch[][9] = { #define HEADER_SIZE 14 -#define DCA_MAX_FRAME_SIZE 16384 -#define DCA_MAX_EXSS_HEADER_SIZE 4096 - -#define DCA_BUFFER_PADDING_SIZE 1024 +#define DCA_NSYNCAUX 0x9A1105A0 /** Bit allocation */ -typedef struct { +typedef struct BitAlloc { int offset; ///< code values offset int maxbits[8]; ///< max bits in VLC int wrap; ///< wrap for get_vlc2() @@ -282,126 +160,6 @@ static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba, ba->offset; } -typedef struct { - AVCodecContext *avctx; - AVFrame frame; - /* Frame header */ - int frame_type; ///< type of the current frame - int samples_deficit; ///< deficit sample count - int crc_present; ///< crc is present in the bitstream - int sample_blocks; ///< number of PCM sample blocks - int frame_size; ///< primary frame byte size - 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 - int timestamp; ///< embedded time stamp flag - int aux_data; ///< auxiliary data flag - int hdcd; ///< source material is mastered in HDCD - int ext_descr; ///< extension audio descriptor flag - int ext_coding; ///< extended coding flag - int aspf; ///< audio sync word insertion flag - int lfe; ///< low frequency effects flag - int predictor_history; ///< predictor history flag - int header_crc; ///< header crc check bytes - int multirate_inter; ///< multirate interpolator switch - int version; ///< encoder software revision - int copy_history; ///< copy history - int source_pcm_res; ///< source pcm resolution - int front_sum; ///< front sum/difference flag - int surround_sum; ///< surround sum/difference flag - int dialog_norm; ///< dialog normalisation parameter - - /* Primary audio coding header */ - int subframes; ///< number of subframes - int is_channels_set; ///< check for if the channel number is already set - 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 - int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index - int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book - int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book - int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select - int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select - float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment - - /* Primary audio coding side information */ - int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes - int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count - int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not) - int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs - int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index - int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients) - int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient) - int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook - int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors - int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients - int dynrange_coef; ///< dynamic range coefficient - - int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands - - float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data - int lfe_scale_factor; - - /* Subband samples history (for ADPCM) */ - DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4]; - DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512]; - DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32]; - int hist_index[DCA_PRIM_CHANNELS_MAX]; - DECLARE_ALIGNED(32, float, raXin)[32]; - - int output; ///< type of output - - DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8]; - float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1]; - float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1]; - uint8_t *extra_channels_buffer; - unsigned int extra_channels_buffer_size; - - uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_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; - int current_subsubframe; - - int core_ext_mask; ///< present extensions in the core substream - - /* XCh extension information */ - int xch_present; ///< XCh extension present and valid - int xch_base_channel; ///< index of first (only) channel containing XCH data - - /* ExSS header parser */ - int static_fields; ///< static fields present - int mix_metadata; ///< mixing metadata present - int num_mix_configs; ///< number of mix out configurations - int mix_config_num_ch[4]; ///< number of channels in each mix out configuration - - int profile; - - int debug_flag; ///< used for suppressing repeated error messages output - AVFloatDSPContext fdsp; - FFTContext imdct; - SynthFilterContext synth; - DCADSPContext dcadsp; - FmtConvertContext fmt_conv; -} DCAContext; - -static const uint16_t dca_vlc_offs[] = { - 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364, - 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508, - 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564, - 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240, - 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264, - 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622, -}; - static av_cold void dca_init_vlcs(void) { static int vlcs_initialized = 0; @@ -412,28 +170,28 @@ static av_cold void dca_init_vlcs(void) return; dca_bitalloc_index.offset = 1; - dca_bitalloc_index.wrap = 2; + dca_bitalloc_index.wrap = 2; for (i = 0; i < 5; i++) { - dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]]; - dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i]; + dca_bitalloc_index.vlc[i].table = &dca_table[ff_dca_vlc_offs[i]]; + dca_bitalloc_index.vlc[i].table_allocated = ff_dca_vlc_offs[i + 1] - ff_dca_vlc_offs[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, INIT_VLC_USE_NEW_STATIC); } dca_scalefactor.offset = -64; - dca_scalefactor.wrap = 2; + dca_scalefactor.wrap = 2; for (i = 0; i < 5; i++) { - dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]]; - dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5]; + dca_scalefactor.vlc[i].table = &dca_table[ff_dca_vlc_offs[i + 5]]; + dca_scalefactor.vlc[i].table_allocated = ff_dca_vlc_offs[i + 6] - ff_dca_vlc_offs[i + 5]; init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129, scales_bits[i], 1, 1, scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC); } dca_tmode.offset = 0; - dca_tmode.wrap = 1; + dca_tmode.wrap = 1; for (i = 0; i < 4; i++) { - dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]]; - dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10]; + dca_tmode.vlc[i].table = &dca_table[ff_dca_vlc_offs[i + 10]]; + dca_tmode.vlc[i].table_allocated = ff_dca_vlc_offs[i + 11] - ff_dca_vlc_offs[i + 10]; init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4, tmode_bits[i], 1, 1, tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC); @@ -445,8 +203,8 @@ static av_cold void dca_init_vlcs(void) break; dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i]; dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4); - dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]]; - dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c]; + dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[ff_dca_vlc_offs[c]]; + dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = ff_dca_vlc_offs[c + 1] - ff_dca_vlc_offs[c]; init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j], bitalloc_sizes[i], @@ -466,48 +224,51 @@ static inline void get_array(GetBitContext *gb, int *dst, int len, int bits) static int dca_parse_audio_coding_header(DCAContext *s, int base_channel) { int i, j; - static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 }; + static const uint8_t adj_table[4] = { 16, 18, 20, 23 }; 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->total_channels = get_bits(&s->gb, 3) + 1 + base_channel; - s->prim_channels = s->total_channels; - - if (s->prim_channels > DCA_PRIM_CHANNELS_MAX) - s->prim_channels = DCA_PRIM_CHANNELS_MAX; + s->audio_header.total_channels = get_bits(&s->gb, 3) + 1 + base_channel; + s->audio_header.prim_channels = s->audio_header.total_channels; + if (s->audio_header.prim_channels > DCA_PRIM_CHANNELS_MAX) + s->audio_header.prim_channels = DCA_PRIM_CHANNELS_MAX; - for (i = base_channel; i < s->prim_channels; i++) { - s->subband_activity[i] = get_bits(&s->gb, 5) + 2; - if (s->subband_activity[i] > DCA_SUBBANDS) - s->subband_activity[i] = DCA_SUBBANDS; + for (i = base_channel; i < s->audio_header.prim_channels; i++) { + s->audio_header.subband_activity[i] = get_bits(&s->gb, 5) + 2; + if (s->audio_header.subband_activity[i] > DCA_SUBBANDS) + s->audio_header.subband_activity[i] = DCA_SUBBANDS; } - for (i = base_channel; i < s->prim_channels; i++) { - s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1; - if (s->vq_start_subband[i] > DCA_SUBBANDS) - s->vq_start_subband[i] = DCA_SUBBANDS; + for (i = base_channel; i < s->audio_header.prim_channels; i++) { + s->audio_header.vq_start_subband[i] = get_bits(&s->gb, 5) + 1; + if (s->audio_header.vq_start_subband[i] > DCA_SUBBANDS) + s->audio_header.vq_start_subband[i] = DCA_SUBBANDS; } - get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3); - get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2); - get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3); - get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3); + get_array(&s->gb, s->audio_header.joint_intensity + base_channel, + s->audio_header.prim_channels - base_channel, 3); + get_array(&s->gb, s->audio_header.transient_huffman + base_channel, + s->audio_header.prim_channels - base_channel, 2); + get_array(&s->gb, s->audio_header.scalefactor_huffman + base_channel, + s->audio_header.prim_channels - base_channel, 3); + get_array(&s->gb, s->audio_header.bitalloc_huffman + base_channel, + s->audio_header.prim_channels - base_channel, 3); /* Get codebooks quantization indexes */ if (!base_channel) - memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman)); + memset(s->audio_header.quant_index_huffman, 0, sizeof(s->audio_header.quant_index_huffman)); for (j = 1; j < 11; j++) - for (i = base_channel; i < s->prim_channels; i++) - s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]); + for (i = base_channel; i < s->audio_header.prim_channels; i++) + s->audio_header.quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]); /* Get scale factor adjustment */ for (j = 0; j < 11; j++) - for (i = base_channel; i < s->prim_channels; i++) - s->scalefactor_adj[i][j] = 1; + for (i = base_channel; i < s->audio_header.prim_channels; i++) + s->audio_header.scalefactor_adj[i][j] = 16; for (j = 1; j < 11; j++) - for (i = base_channel; i < s->prim_channels; i++) - if (s->quant_index_huffman[i][j] < thr[j]) - s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)]; + for (i = base_channel; i < s->audio_header.prim_channels; i++) + if (s->audio_header.quant_index_huffman[i][j] < thr[j]) + s->audio_header.scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)]; if (s->crc_present) { /* Audio header CRC check */ @@ -517,33 +278,6 @@ static int dca_parse_audio_coding_header(DCAContext *s, int base_channel) s->current_subframe = 0; s->current_subsubframe = 0; -#ifdef TRACE - av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes); - av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels); - for (i = base_channel; i < s->prim_channels; i++) { - av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", - s->subband_activity[i]); - av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", - s->vq_start_subband[i]); - av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", - s->joint_intensity[i]); - av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", - s->transient_huffman[i]); - av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", - s->scalefactor_huffman[i]); - av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", - s->bitalloc_huffman[i]); - av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:"); - for (j = 0; j < 11; j++) - av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]); - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:"); - for (j = 0; j < 11; j++) - av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]); - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - } -#endif - return 0; } @@ -567,11 +301,11 @@ static int dca_parse_frame_header(DCAContext *s) if (!s->sample_rate) return AVERROR_INVALIDDATA; s->bit_rate_index = get_bits(&s->gb, 5); - s->bit_rate = dca_bit_rates[s->bit_rate_index]; + s->bit_rate = ff_dca_bit_rates[s->bit_rate_index]; if (!s->bit_rate) return AVERROR_INVALIDDATA; - s->downmix = get_bits(&s->gb, 1); + skip_bits1(&s->gb); // always 0 (reserved, cf. ETSI TS 102 114 V1.4.1) s->dynrange = get_bits(&s->gb, 1); s->timestamp = get_bits(&s->gb, 1); s->aux_data = get_bits(&s->gb, 1); @@ -582,6 +316,11 @@ static int dca_parse_frame_header(DCAContext *s) s->lfe = get_bits(&s->gb, 2); s->predictor_history = get_bits(&s->gb, 1); + if (s->lfe > 2) { + av_log(s->avctx, AV_LOG_ERROR, "Invalid LFE value: %d\n", s->lfe); + return AVERROR_INVALIDDATA; + } + /* TODO: check CRC */ if (s->crc_present) s->header_crc = get_bits(&s->gb, 16); @@ -599,57 +338,18 @@ static int dca_parse_frame_header(DCAContext *s) if (s->lfe) s->output |= DCA_LFE; -#ifdef TRACE - av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type); - av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit); - av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present); - av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n", - s->sample_blocks, s->sample_blocks * 32); - 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 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); - av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data); - av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd); - av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr); - av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding); - av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf); - av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe); - av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n", - s->predictor_history); - av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc); - av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n", - s->multirate_inter); - av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version); - av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history); - av_log(s->avctx, AV_LOG_DEBUG, - "source pcm resolution: %i (%i bits/sample)\n", - s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]); - av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum); - av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum); - av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm); - av_log(s->avctx, AV_LOG_DEBUG, "\n"); -#endif - /* Primary audio coding header */ - s->subframes = get_bits(&s->gb, 4) + 1; + s->audio_header.subframes = get_bits(&s->gb, 4) + 1; return dca_parse_audio_coding_header(s, 0); } - static inline int get_scale(GetBitContext *gb, int level, int value, int log2range) { if (level < 5) { /* huffman encoded */ value += get_bitalloc(gb, &dca_scalefactor, level); - value = av_clip(value, 0, (1 << log2range) - 1); + value = av_clip(value, 0, (1 << log2range) - 1); } else if (level < 8) { if (level + 1 > log2range) { skip_bits(gb, level + 1 - log2range); @@ -674,53 +374,53 @@ static int dca_subframe_header(DCAContext *s, int base_channel, int block_index) s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3); } - for (j = base_channel; j < s->prim_channels; j++) { - for (k = 0; k < s->subband_activity[j]; k++) - s->prediction_mode[j][k] = get_bits(&s->gb, 1); + for (j = base_channel; j < s->audio_header.prim_channels; j++) { + for (k = 0; k < s->audio_header.subband_activity[j]; k++) + s->dca_chan[j].prediction_mode[k] = get_bits(&s->gb, 1); } /* Get prediction codebook */ - for (j = base_channel; j < s->prim_channels; j++) { - for (k = 0; k < s->subband_activity[j]; k++) { - if (s->prediction_mode[j][k] > 0) { + for (j = base_channel; j < s->audio_header.prim_channels; j++) { + for (k = 0; k < s->audio_header.subband_activity[j]; k++) { + if (s->dca_chan[j].prediction_mode[k] > 0) { /* (Prediction coefficient VQ address) */ - s->prediction_vq[j][k] = get_bits(&s->gb, 12); + s->dca_chan[j].prediction_vq[k] = get_bits(&s->gb, 12); } } } /* Bit allocation index */ - for (j = base_channel; j < s->prim_channels; j++) { - for (k = 0; k < s->vq_start_subband[j]; k++) { - if (s->bitalloc_huffman[j] == 6) - 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 if (s->bitalloc_huffman[j] == 7) { + for (j = base_channel; j < s->audio_header.prim_channels; j++) { + for (k = 0; k < s->audio_header.vq_start_subband[j]; k++) { + if (s->audio_header.bitalloc_huffman[j] == 6) + s->dca_chan[j].bitalloc[k] = get_bits(&s->gb, 5); + else if (s->audio_header.bitalloc_huffman[j] == 5) + s->dca_chan[j].bitalloc[k] = get_bits(&s->gb, 4); + else if (s->audio_header.bitalloc_huffman[j] == 7) { av_log(s->avctx, AV_LOG_ERROR, "Invalid bit allocation index\n"); return AVERROR_INVALIDDATA; } else { - s->bitalloc[j][k] = - get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]); + s->dca_chan[j].bitalloc[k] = + get_bitalloc(&s->gb, &dca_bitalloc_index, s->audio_header.bitalloc_huffman[j]); } - if (s->bitalloc[j][k] > 26) { - av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n", - j, k, s->bitalloc[j][k]); + if (s->dca_chan[j].bitalloc[k] > 26) { + ff_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n", + j, k, s->dca_chan[j].bitalloc[k]); return AVERROR_INVALIDDATA; } } } /* Transition mode */ - for (j = base_channel; j < s->prim_channels; j++) { - for (k = 0; k < s->subband_activity[j]; k++) { - s->transition_mode[j][k] = 0; + for (j = base_channel; j < s->audio_header.prim_channels; j++) { + for (k = 0; k < s->audio_header.subband_activity[j]; k++) { + s->dca_chan[j].transition_mode[k] = 0; if (s->subsubframes[s->current_subframe] > 1 && - k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) { - s->transition_mode[j][k] = - get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]); + k < s->audio_header.vq_start_subband[j] && s->dca_chan[j].bitalloc[k] > 0) { + s->dca_chan[j].transition_mode[k] = + get_bitalloc(&s->gb, &dca_tmode, s->audio_header.transient_huffman[j]); } } } @@ -728,63 +428,64 @@ static int dca_subframe_header(DCAContext *s, int base_channel, int block_index) if (get_bits_left(&s->gb) < 0) return AVERROR_INVALIDDATA; - for (j = base_channel; j < s->prim_channels; j++) { + for (j = base_channel; j < s->audio_header.prim_channels; j++) { const uint32_t *scale_table; int scale_sum, log_size; - memset(s->scale_factor[j], 0, - s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2); + memset(s->dca_chan[j].scale_factor, 0, + s->audio_header.subband_activity[j] * sizeof(s->dca_chan[j].scale_factor[0][0]) * 2); - if (s->scalefactor_huffman[j] == 6) { - scale_table = scale_factor_quant7; - log_size = 7; + if (s->audio_header.scalefactor_huffman[j] == 6) { + scale_table = ff_dca_scale_factor_quant7; + log_size = 7; } else { - scale_table = scale_factor_quant6; - log_size = 6; + scale_table = ff_dca_scale_factor_quant6; + log_size = 6; } /* When huffman coded, only the difference is encoded */ scale_sum = 0; - for (k = 0; k < s->subband_activity[j]; k++) { - if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) { - scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size); - s->scale_factor[j][k][0] = scale_table[scale_sum]; + for (k = 0; k < s->audio_header.subband_activity[j]; k++) { + if (k >= s->audio_header.vq_start_subband[j] || s->dca_chan[j].bitalloc[k] > 0) { + scale_sum = get_scale(&s->gb, s->audio_header.scalefactor_huffman[j], scale_sum, log_size); + s->dca_chan[j].scale_factor[k][0] = scale_table[scale_sum]; } - if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) { + if (k < s->audio_header.vq_start_subband[j] && s->dca_chan[j].transition_mode[k]) { /* Get second scale factor */ - scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size); - s->scale_factor[j][k][1] = scale_table[scale_sum]; + scale_sum = get_scale(&s->gb, s->audio_header.scalefactor_huffman[j], scale_sum, log_size); + s->dca_chan[j].scale_factor[k][1] = scale_table[scale_sum]; } } } /* Joint subband scale factor codebook select */ - for (j = base_channel; j < s->prim_channels; j++) { + for (j = base_channel; j < s->audio_header.prim_channels; j++) { /* Transmitted only if joint subband coding enabled */ - if (s->joint_intensity[j] > 0) - s->joint_huff[j] = get_bits(&s->gb, 3); + if (s->audio_header.joint_intensity[j] > 0) + s->dca_chan[j].joint_huff = get_bits(&s->gb, 3); } if (get_bits_left(&s->gb) < 0) return AVERROR_INVALIDDATA; /* Scale factors for joint subband coding */ - for (j = base_channel; j < s->prim_channels; j++) { + for (j = base_channel; j < s->audio_header.prim_channels; j++) { int source_channel; /* Transmitted only if joint subband coding enabled */ - if (s->joint_intensity[j] > 0) { + if (s->audio_header.joint_intensity[j] > 0) { int scale = 0; - source_channel = s->joint_intensity[j] - 1; + source_channel = s->audio_header.joint_intensity[j] - 1; /* When huffman coded, only the difference is encoded * (is this valid as well for joint scales ???) */ - for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) { - scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7); - s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */ + for (k = s->audio_header.subband_activity[j]; + k < s->audio_header.subband_activity[source_channel]; k++) { + scale = get_scale(&s->gb, s->dca_chan[j].joint_huff, 64 /* bias */, 7); + s->dca_chan[j].joint_scale_factor[k] = scale; /*joint_scale_table[scale]; */ } if (!(s->debug_flag & 0x02)) { @@ -795,27 +496,6 @@ static int dca_subframe_header(DCAContext *s, int base_channel, int block_index) } } - /* Stereo downmix coefficients */ - if (!base_channel && s->prim_channels > 2) { - if (s->downmix) { - for (j = base_channel; j < s->prim_channels; j++) { - s->downmix_coef[j][0] = get_bits(&s->gb, 7); - s->downmix_coef[j][1] = get_bits(&s->gb, 7); - } - } else { - int am = s->amode & DCA_CHANNEL_MASK; - if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) { - av_log(s->avctx, AV_LOG_ERROR, - "Invalid channel mode %d\n", am); - return AVERROR_INVALIDDATA; - } - for (j = base_channel; j < s->prim_channels; j++) { - s->downmix_coef[j][0] = dca_default_coeffs[am][j][0]; - s->downmix_coef[j][1] = dca_default_coeffs[am][j][1]; - } - } - } - /* Dynamic range coefficient */ if (!base_channel && s->dynrange) s->dynrange_coef = get_bits(&s->gb, 8); @@ -830,15 +510,15 @@ static int dca_subframe_header(DCAContext *s, int base_channel, int block_index) */ /* VQ encoded high frequency subbands */ - for (j = base_channel; j < s->prim_channels; j++) - for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++) + for (j = base_channel; j < s->audio_header.prim_channels; j++) + for (k = s->audio_header.vq_start_subband[j]; k < s->audio_header.subband_activity[j]; k++) /* 1 vector -> 32 samples */ - s->high_freq_vq[j][k] = get_bits(&s->gb, 10); + s->dca_chan[j].high_freq_vq[k] = get_bits(&s->gb, 10); /* Low frequency effect data */ if (!base_channel && s->lfe) { /* LFE samples */ - int lfe_samples = 2 * s->lfe * (4 + block_index); + int lfe_samples = 2 * s->lfe * (4 + block_index); int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]); float lfe_scale; @@ -849,7 +529,7 @@ static int dca_subframe_header(DCAContext *s, int base_channel, int block_index) /* Scale factor index */ skip_bits(&s->gb, 1); - s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 7)]; + s->lfe_scale_factor = ff_dca_scale_factor_quant7[get_bits(&s->gb, 7)]; /* Quantization step size * scale factor */ lfe_scale = 0.035 * s->lfe_scale_factor; @@ -858,127 +538,111 @@ static int dca_subframe_header(DCAContext *s, int base_channel, int block_index) s->lfe_data[j] *= lfe_scale; } -#ifdef TRACE - av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", - s->subsubframes[s->current_subframe]); - av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n", - s->partial_samples[s->current_subframe]); - - for (j = base_channel; j < s->prim_channels; j++) { - av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:"); - for (k = 0; k < s->subband_activity[j]; k++) - av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]); - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - } - for (j = base_channel; j < s->prim_channels; j++) { - for (k = 0; k < s->subband_activity[j]; k++) - av_log(s->avctx, AV_LOG_DEBUG, - "prediction coefs: %f, %f, %f, %f\n", - (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192, - (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192, - (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192, - (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192); - } - for (j = base_channel; j < s->prim_channels; j++) { - av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: "); - for (k = 0; k < s->vq_start_subband[j]; k++) - av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]); - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - } - for (j = base_channel; j < s->prim_channels; j++) { - av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:"); - for (k = 0; k < s->subband_activity[j]; k++) - av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]); - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - } - for (j = base_channel; j < s->prim_channels; j++) { - av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:"); - for (k = 0; k < s->subband_activity[j]; k++) { - if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) - av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]); - if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) - av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]); - } - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - } - for (j = base_channel; 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]); - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - } - } - if (!base_channel && s->prim_channels > 2 && s->downmix) { - av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n"); - for (j = 0; j < s->prim_channels; j++) { - av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j, - dca_downmix_coeffs[s->downmix_coef[j][0]]); - av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j, - dca_downmix_coeffs[s->downmix_coef[j][1]]); - } - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - } - for (j = base_channel; j < s->prim_channels; j++) - 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 (!base_channel && s->lfe) { - int lfe_samples = 2 * s->lfe * (4 + block_index); - int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]); - - av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n"); - for (j = lfe_samples; j < lfe_end_sample; j++) - av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]); - av_log(s->avctx, AV_LOG_DEBUG, "\n"); - } -#endif - return 0; } static void qmf_32_subbands(DCAContext *s, int chans, - float samples_in[32][8], float *samples_out, + float samples_in[DCA_SUBBANDS][SAMPLES_PER_SUBBAND], float *samples_out, float scale) { const float *prCoeff; - int i; - int sb_act = s->subband_activity[chans]; - int subindex; + int sb_act = s->audio_header.subband_activity[chans]; scale *= sqrt(1 / 8.0); /* Select filter */ if (!s->multirate_inter) /* Non-perfect reconstruction */ - prCoeff = fir_32bands_nonperfect; + prCoeff = ff_dca_fir_32bands_nonperfect; else /* Perfect reconstruction */ - prCoeff = fir_32bands_perfect; - - for (i = sb_act; i < 32; i++) - s->raXin[i] = 0.0; - - /* Reconstructed channel sample index */ - for (subindex = 0; subindex < 8; subindex++) { - /* Load in one sample from each subband and clear inactive subbands */ - for (i = 0; i < sb_act; i++) { - unsigned sign = (i - 1) & 2; - uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30; - AV_WN32A(&s->raXin[i], v); + prCoeff = ff_dca_fir_32bands_perfect; + + s->dcadsp.qmf_32_subbands(samples_in, sb_act, &s->synth, &s->imdct, + s->dca_chan[chans].subband_fir_hist, + &s->dca_chan[chans].hist_index, + s->dca_chan[chans].subband_fir_noidea, prCoeff, + samples_out, s->raXin, scale); +} + +static QMF64_table *qmf64_precompute(void) +{ + unsigned i, j; + QMF64_table *table = av_malloc(sizeof(*table)); + if (!table) + return NULL; + + for (i = 0; i < 32; i++) + for (j = 0; j < 32; j++) + table->dct4_coeff[i][j] = cos((2 * i + 1) * (2 * j + 1) * M_PI / 128); + for (i = 0; i < 32; i++) + for (j = 0; j < 32; j++) + table->dct2_coeff[i][j] = cos((2 * i + 1) * j * M_PI / 64); + + /* FIXME: Is the factor 0.125 = 1/8 right? */ + for (i = 0; i < 32; i++) + table->rcos[i] = 0.125 / cos((2 * i + 1) * M_PI / 256); + for (i = 0; i < 32; i++) + table->rsin[i] = -0.125 / sin((2 * i + 1) * M_PI / 256); + + return table; +} + +/* FIXME: Totally unoptimized. Based on the reference code and + * http://multimedia.cx/mirror/dca-transform.pdf, with guessed tweaks + * for doubling the size. */ +static void qmf_64_subbands(DCAContext *s, int chans, + float samples_in[DCA_SUBBANDS_X96K][SAMPLES_PER_SUBBAND], + float *samples_out, float scale) +{ + float raXin[64]; + float A[32], B[32]; + float *raX = s->dca_chan[chans].subband_fir_hist; + float *raZ = s->dca_chan[chans].subband_fir_noidea; + unsigned i, j, k, subindex; + + for (i = s->audio_header.subband_activity[chans]; i < DCA_SUBBANDS_X96K; i++) + raXin[i] = 0.0; + for (subindex = 0; subindex < SAMPLES_PER_SUBBAND; subindex++) { + for (i = 0; i < s->audio_header.subband_activity[chans]; i++) + raXin[i] = samples_in[i][subindex]; + + for (k = 0; k < 32; k++) { + A[k] = 0.0; + for (i = 0; i < 32; i++) + A[k] += (raXin[2 * i] + raXin[2 * i + 1]) * s->qmf64_table->dct4_coeff[k][i]; + } + for (k = 0; k < 32; k++) { + B[k] = raXin[0] * s->qmf64_table->dct2_coeff[k][0]; + for (i = 1; i < 32; i++) + B[k] += (raXin[2 * i] + raXin[2 * i - 1]) * s->qmf64_table->dct2_coeff[k][i]; + } + for (k = 0; k < 32; k++) { + raX[k] = s->qmf64_table->rcos[k] * (A[k] + B[k]); + raX[63 - k] = s->qmf64_table->rsin[k] * (A[k] - B[k]); + } + + for (i = 0; i < DCA_SUBBANDS_X96K; i++) { + float out = raZ[i]; + for (j = 0; j < 1024; j += 128) + out += ff_dca_fir_64bands[j + i] * (raX[j + i] - raX[j + 63 - i]); + *samples_out++ = out * scale; + } + + for (i = 0; i < DCA_SUBBANDS_X96K; i++) { + float hist = 0.0; + for (j = 0; j < 1024; j += 128) + hist += ff_dca_fir_64bands[64 + j + i] * (-raX[i + j] - raX[j + 63 - i]); + + raZ[i] = hist; } - s->synth.synth_filter_float(&s->imdct, - s->subband_fir_hist[chans], - &s->hist_index[chans], - s->subband_fir_noidea[chans], prCoeff, - samples_out, s->raXin, scale); - samples_out += 32; + /* FIXME: Make buffer circular, to avoid this move. */ + memmove(raX + 64, raX, (1024 - 64) * sizeof(*raX)); } } -static void lfe_interpolation_fir(DCAContext *s, int decimation_select, - int num_deci_sample, float *samples_in, - float *samples_out, float scale) +static void lfe_interpolation_fir(DCAContext *s, const float *samples_in, + float *samples_out) { /* samples_in: An array holding decimated samples. * Samples in current subframe starts from samples_in[0], @@ -988,23 +652,26 @@ static void lfe_interpolation_fir(DCAContext *s, int decimation_select, * samples_out: An array holding interpolated samples */ - int decifactor; + int idx; const float *prCoeff; int deciindex; /* Select decimation filter */ - if (decimation_select == 1) { - decifactor = 64; - prCoeff = lfe_fir_128; + if (s->lfe == 1) { + idx = 1; + prCoeff = ff_dca_lfe_fir_128; } else { - decifactor = 32; - prCoeff = lfe_fir_64; + idx = 0; + if (s->exss_ext_mask & DCA_EXT_EXSS_XLL) + prCoeff = ff_dca_lfe_xll_fir_64; + else + prCoeff = ff_dca_lfe_fir_64; } /* Interpolation */ - for (deciindex = 0; deciindex < num_deci_sample; deciindex++) { - s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale); + for (deciindex = 0; deciindex < 2 * s->lfe; deciindex++) { + s->dcadsp.lfe_fir[idx](samples_out, samples_in, prCoeff); samples_in++; - samples_out += 2 * decifactor; + samples_out += 2 * 32 * (1 + idx); } } @@ -1030,29 +697,23 @@ static void lfe_interpolation_fir(DCAContext *s, int decimation_select, op2 \ } -static void dca_downmix(float **samples, int srcfmt, - int downmix_coef[DCA_PRIM_CHANNELS_MAX][2], +static void dca_downmix(float **samples, int srcfmt, int lfe_present, + float coef[DCA_PRIM_CHANNELS_MAX + 1][2], const int8_t *channel_mapping) { int c, l, r, sl, sr, s; int i; float t, u, v; - float coef[DCA_PRIM_CHANNELS_MAX][2]; - - for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) { - coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]]; - coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]]; - } switch (srcfmt) { case DCA_MONO: - case DCA_CHANNEL: - case DCA_STEREO_TOTAL: - case DCA_STEREO_SUMDIFF: case DCA_4F2R: av_log(NULL, 0, "Not implemented!\n"); break; + case DCA_CHANNEL: case DCA_STEREO: + case DCA_STEREO_TOTAL: + case DCA_STEREO_SUMDIFF: break; case DCA_3F: c = channel_mapping[0]; @@ -1087,13 +748,20 @@ static void dca_downmix(float **samples, int srcfmt, MIX_REAR2(samples, sl, sr, 3, coef)); break; } + if (lfe_present) { + int lf_buf = ff_dca_lfe_index[srcfmt]; + int lf_idx = ff_dca_channels[srcfmt]; + for (i = 0; i < 256; i++) { + samples[0][i] += samples[lf_buf][i] * coef[lf_idx][0]; + samples[1][i] += samples[lf_buf][i] * coef[lf_idx][1]; + } + } } - #ifndef decode_blockcodes /* Very compact version of the block code decoder that does not use table * look-up but is slightly slower */ -static int decode_blockcode(int code, int levels, int *values) +static int decode_blockcode(int code, int levels, int32_t *values) { int i; int offset = (levels - 1) >> 1; @@ -1101,13 +769,13 @@ static int decode_blockcode(int code, int levels, int *values) for (i = 0; i < 4; i++) { int div = FASTDIV(code, levels); values[i] = code - offset - div * levels; - code = div; + code = div; } return code; } -static int decode_blockcodes(int code1, int code2, int levels, int *values) +static int decode_blockcodes(int code1, int code2, int levels, int32_t *values) { return decode_blockcode(code1, levels, values) | decode_blockcode(code2, levels, values + 4); @@ -1117,26 +785,11 @@ static int decode_blockcodes(int code1, int code2, int levels, int *values) static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 }; static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 }; -#ifndef int8x8_fmul_int32 -static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale) -{ - float fscale = scale / 16.0; - int i; - for (i = 0; i < 8; i++) - dst[i] = src[i] * fscale; -} -#endif - static int dca_subsubframe(DCAContext *s, int base_channel, int block_index) { int k, l; int subsubframe = s->current_subsubframe; - - const float *quant_step_table; - - /* FIXME */ - float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index]; - LOCAL_ALIGNED_16(int, block, [8]); + const uint32_t *quant_step_table; /* * Audio data @@ -1144,39 +797,41 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index) /* Select quantization step size table */ if (s->bit_rate_index == 0x1f) - quant_step_table = lossless_quant_d; + quant_step_table = ff_dca_lossless_quant; else - quant_step_table = lossy_quant_d; + quant_step_table = ff_dca_lossy_quant; + + for (k = base_channel; k < s->audio_header.prim_channels; k++) { + int32_t (*subband_samples)[8] = s->dca_chan[k].subband_samples[block_index]; - for (k = base_channel; k < s->prim_channels; k++) { if (get_bits_left(&s->gb) < 0) return AVERROR_INVALIDDATA; - for (l = 0; l < s->vq_start_subband[k]; l++) { + for (l = 0; l < s->audio_header.vq_start_subband[k]; l++) { int m; /* Select the mid-tread linear quantizer */ - int abits = s->bitalloc[k][l]; - - float quant_step_size = quant_step_table[abits]; - - /* - * Determine quantization index code book and its type - */ + int abits = s->dca_chan[k].bitalloc[l]; - /* Select quantization index code book */ - int sel = s->quant_index_huffman[k][abits]; + uint32_t quant_step_size = quant_step_table[abits]; /* * Extract bits from the bit stream */ - if (!abits) { - memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0])); - } else { + if (!abits) + memset(subband_samples[l], 0, SAMPLES_PER_SUBBAND * + sizeof(subband_samples[l][0])); + else { + uint32_t rscale; /* Deal with transients */ - int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l]; - float rscale = quant_step_size * s->scale_factor[k][l][sfi] * - s->scalefactor_adj[k][sel]; + int sfi = s->dca_chan[k].transition_mode[l] && + subsubframe >= s->dca_chan[k].transition_mode[l]; + /* Determine quantization index code book and its type. + Select quantization index code book */ + int sel = s->audio_header.quant_index_huffman[k][abits]; + + rscale = (s->dca_chan[k].scale_factor[l][sfi] * + s->audio_header.scalefactor_adj[k][sel] + 8) >> 4; if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) { if (abits <= 7) { @@ -1188,8 +843,8 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index) block_code1 = get_bits(&s->gb, size); block_code2 = get_bits(&s->gb, size); - err = decode_blockcodes(block_code1, block_code2, - levels, block); + err = decode_blockcodes(block_code1, block_code2, + levels, subband_samples[l]); if (err) { av_log(s->avctx, AV_LOG_ERROR, "ERROR: block code look-up failed\n"); @@ -1197,115 +852,165 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index) } } else { /* no coding */ - for (m = 0; m < 8; m++) - block[m] = get_sbits(&s->gb, abits - 3); + for (m = 0; m < SAMPLES_PER_SUBBAND; m++) + subband_samples[l][m] = get_sbits(&s->gb, abits - 3); } } else { /* Huffman coded */ - for (m = 0; m < 8; m++) - block[m] = get_bitalloc(&s->gb, - &dca_smpl_bitalloc[abits], sel); + for (m = 0; m < SAMPLES_PER_SUBBAND; m++) + subband_samples[l][m] = get_bitalloc(&s->gb, + &dca_smpl_bitalloc[abits], sel); } - - s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l], - block, rscale, 8); + s->dcadsp.dequantize(subband_samples[l], quant_step_size, rscale); } + } + for (l = 0; l < s->audio_header.vq_start_subband[k]; l++) { + int m; /* * Inverse ADPCM if in prediction mode */ - if (s->prediction_mode[k][l]) { + if (s->dca_chan[k].prediction_mode[l]) { int n; - for (m = 0; m < 8; m++) { - for (n = 1; n <= 4; n++) + if (s->predictor_history) + subband_samples[l][0] += (ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][0] * + (int64_t)s->dca_chan[k].subband_samples_hist[l][3] + + ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][1] * + (int64_t)s->dca_chan[k].subband_samples_hist[l][2] + + ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][2] * + (int64_t)s->dca_chan[k].subband_samples_hist[l][1] + + ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][3] * + (int64_t)s->dca_chan[k].subband_samples_hist[l][0]) + + (1 << 12) >> 13; + for (m = 1; m < SAMPLES_PER_SUBBAND; m++) { + int64_t sum = ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][0] * + (int64_t)subband_samples[l][m - 1]; + for (n = 2; n <= 4; n++) if (m >= n) - subband_samples[k][l][m] += - (adpcm_vb[s->prediction_vq[k][l]][n - 1] * - subband_samples[k][l][m - n] / 8192); + sum += ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][n - 1] * + (int64_t)subband_samples[l][m - n]; else if (s->predictor_history) - subband_samples[k][l][m] += - (adpcm_vb[s->prediction_vq[k][l]][n - 1] * - s->subband_samples_hist[k][l][m - n + 4] / 8192); + sum += ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][n - 1] * + (int64_t)s->dca_chan[k].subband_samples_hist[l][m - n + 4]; + subband_samples[l][m] += (int32_t)(sum + (1 << 12) >> 13); } } + } + /* Backup predictor history for adpcm */ + for (l = 0; l < DCA_SUBBANDS; l++) + AV_COPY128(s->dca_chan[k].subband_samples_hist[l], &subband_samples[l][4]); + /* * Decode VQ encoded high frequencies */ - for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) { - /* 1 vector -> 32 samples but we only need the 8 samples - * for this subsubframe. */ - int hfvq = s->high_freq_vq[k][l]; - + if (s->audio_header.subband_activity[k] > s->audio_header.vq_start_subband[k]) { if (!s->debug_flag & 0x01) { av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n"); s->debug_flag |= 0x01; } - int8x8_fmul_int32(subband_samples[k][l], - &high_freq_vq[hfvq][subsubframe * 8], - s->scale_factor[k][l][0]); + s->dcadsp.decode_hf(subband_samples, s->dca_chan[k].high_freq_vq, + ff_dca_high_freq_vq, + subsubframe * SAMPLES_PER_SUBBAND, + s->dca_chan[k].scale_factor, + s->audio_header.vq_start_subband[k], + s->audio_header.subband_activity[k]); } } /* Check for DSYNC after subsubframe */ if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) { - if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */ -#ifdef TRACE - av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n"); -#endif - } else { + if (get_bits(&s->gb, 16) != 0xFFFF) { av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n"); + return AVERROR_INVALIDDATA; } } - /* Backup predictor history for adpcm */ - for (k = base_channel; k < s->prim_channels; k++) - for (l = 0; l < s->vq_start_subband[k]; l++) - memcpy(s->subband_samples_hist[k][l], - &subband_samples[k][l][4], - 4 * sizeof(subband_samples[0][0][0])); - return 0; } -static int dca_filter_channels(DCAContext *s, int block_index) +static int dca_filter_channels(DCAContext *s, int block_index, int upsample) { - float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index]; int k; - /* 32 subbands QMF */ - 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_chanptr[s->channel_order_tab[k]], - M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */); - } + if (upsample) { + LOCAL_ALIGNED(32, float, samples, [DCA_SUBBANDS_X96K], [SAMPLES_PER_SUBBAND]); + + if (!s->qmf64_table) { + s->qmf64_table = qmf64_precompute(); + if (!s->qmf64_table) + return AVERROR(ENOMEM); + } + + /* 64 subbands QMF */ + for (k = 0; k < s->audio_header.prim_channels; k++) { + int32_t (*subband_samples)[SAMPLES_PER_SUBBAND] = + s->dca_chan[k].subband_samples[block_index]; + + s->fmt_conv.int32_to_float(samples[0], subband_samples[0], + DCA_SUBBANDS_X96K * SAMPLES_PER_SUBBAND); + + if (s->channel_order_tab[k] >= 0) + qmf_64_subbands(s, k, samples, + s->samples_chanptr[s->channel_order_tab[k]], + /* Upsampling needs a factor 2 here. */ + M_SQRT2 / 32768.0); + } + } else { + /* 32 subbands QMF */ + LOCAL_ALIGNED(32, float, samples, [DCA_SUBBANDS], [SAMPLES_PER_SUBBAND]); + + for (k = 0; k < s->audio_header.prim_channels; k++) { + int32_t (*subband_samples)[SAMPLES_PER_SUBBAND] = + s->dca_chan[k].subband_samples[block_index]; + + s->fmt_conv.int32_to_float(samples[0], subband_samples[0], + DCA_SUBBANDS * SAMPLES_PER_SUBBAND); - /* Down mixing */ - if (s->avctx->request_channels == 2 && s->prim_channels > 2) { - dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab); + if (s->channel_order_tab[k] >= 0) + qmf_32_subbands(s, k, samples, + s->samples_chanptr[s->channel_order_tab[k]], + M_SQRT1_2 / 32768.0); + } } /* Generate LFE samples for this subsubframe FIXME!!! */ - if (s->output & DCA_LFE) { - lfe_interpolation_fir(s, s->lfe, 2 * s->lfe, + if (s->lfe) { + float *samples = s->samples_chanptr[ff_dca_lfe_index[s->amode]]; + lfe_interpolation_fir(s, s->lfe_data + 2 * s->lfe * (block_index + 4), - s->samples_chanptr[dca_lfe_index[s->amode]], - 1.0 / (256.0 * 32768.0)); - /* Outputs 20bits pcm samples */ + samples); + if (upsample) { + unsigned i; + /* Should apply the filter in Table 6-11 when upsampling. For + * now, just duplicate. */ + for (i = 511; i > 0; i--) { + samples[2 * i] = + samples[2 * i + 1] = samples[i]; + } + samples[1] = samples[0]; + } + } + + /* FIXME: This downmixing is probably broken with upsample. + * Probably totally broken also with XLL in general. */ + /* Downmixing to Stereo */ + if (s->audio_header.prim_channels + !!s->lfe > 2 && + s->avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) { + dca_downmix(s->samples_chanptr, s->amode, !!s->lfe, s->downmix_coef, + s->channel_order_tab); } return 0; } - static int dca_subframe_footer(DCAContext *s, int base_channel) { - int aux_data_count = 0, i; + int in, out, aux_data_count, aux_data_end, reserved; + uint32_t nsyncaux; /* * Unpack optional information @@ -1316,13 +1021,89 @@ static int dca_subframe_footer(DCAContext *s, int base_channel) if (s->timestamp) skip_bits_long(&s->gb, 32); - if (s->aux_data) + if (s->aux_data) { aux_data_count = get_bits(&s->gb, 6); - for (i = 0; i < aux_data_count; i++) - get_bits(&s->gb, 8); + // align (32-bit) + skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31); + + aux_data_end = 8 * aux_data_count + get_bits_count(&s->gb); + + if ((nsyncaux = get_bits_long(&s->gb, 32)) != DCA_NSYNCAUX) { + av_log(s->avctx, AV_LOG_ERROR, "nSYNCAUX mismatch %#"PRIx32"\n", + nsyncaux); + return AVERROR_INVALIDDATA; + } + + if (get_bits1(&s->gb)) { // bAUXTimeStampFlag + avpriv_request_sample(s->avctx, + "Auxiliary Decode Time Stamp Flag"); + // align (4-bit) + skip_bits(&s->gb, (-get_bits_count(&s->gb)) & 4); + // 44 bits: nMSByte (8), nMarker (4), nLSByte (28), nMarker (4) + skip_bits_long(&s->gb, 44); + } + + if ((s->core_downmix = get_bits1(&s->gb))) { + int am = get_bits(&s->gb, 3); + switch (am) { + case 0: + s->core_downmix_amode = DCA_MONO; + break; + case 1: + s->core_downmix_amode = DCA_STEREO; + break; + case 2: + s->core_downmix_amode = DCA_STEREO_TOTAL; + break; + case 3: + s->core_downmix_amode = DCA_3F; + break; + case 4: + s->core_downmix_amode = DCA_2F1R; + break; + case 5: + s->core_downmix_amode = DCA_2F2R; + break; + case 6: + s->core_downmix_amode = DCA_3F1R; + break; + default: + av_log(s->avctx, AV_LOG_ERROR, + "Invalid mode %d for embedded downmix coefficients\n", + am); + return AVERROR_INVALIDDATA; + } + for (out = 0; out < ff_dca_channels[s->core_downmix_amode]; out++) { + for (in = 0; in < s->audio_header.prim_channels + !!s->lfe; in++) { + uint16_t tmp = get_bits(&s->gb, 9); + if ((tmp & 0xFF) > 241) { + av_log(s->avctx, AV_LOG_ERROR, + "Invalid downmix coefficient code %"PRIu16"\n", + tmp); + return AVERROR_INVALIDDATA; + } + s->core_downmix_codes[in][out] = tmp; + } + } + } + + align_get_bits(&s->gb); // byte align + skip_bits(&s->gb, 16); // nAUXCRC16 + + /* + * additional data (reserved, cf. ETSI TS 102 114 V1.4.1) + * + * Note: don't check for overreads, aux_data_count can't be trusted. + */ + if ((reserved = (aux_data_end - get_bits_count(&s->gb))) > 0) { + avpriv_request_sample(s->avctx, + "Core auxiliary data reserved content"); + skip_bits_long(&s->gb, reserved); + } + } - if (s->crc_present && (s->downmix || s->dynrange)) + if (s->crc_present && s->dynrange) get_bits(&s->gb, 16); } @@ -1340,25 +1121,19 @@ static int dca_decode_block(DCAContext *s, int base_channel, int block_index) int ret; /* Sanity check */ - if (s->current_subframe >= s->subframes) { + if (s->current_subframe >= s->audio_header.subframes) { av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i", - s->current_subframe, s->subframes); + s->current_subframe, s->audio_header.subframes); return AVERROR_INVALIDDATA; } if (!s->current_subsubframe) { -#ifdef TRACE - av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n"); -#endif /* Read subframe header */ if ((ret = dca_subframe_header(s, base_channel, block_index))) return ret; } /* Read subsubframe */ -#ifdef TRACE - av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n"); -#endif if ((ret = dca_subsubframe(s, base_channel, block_index))) return ret; @@ -1368,10 +1143,7 @@ static int dca_decode_block(DCAContext *s, int base_channel, int block_index) s->current_subsubframe = 0; s->current_subframe++; } - if (s->current_subframe >= s->subframes) { -#ifdef TRACE - av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n"); -#endif + if (s->current_subframe >= s->audio_header.subframes) { /* Read subframe footer */ if ((ret = dca_subframe_footer(s, base_channel))) return ret; @@ -1380,331 +1152,23 @@ static int dca_decode_block(DCAContext *s, int base_channel, int block_index) return 0; } -/** - * Return the number of channels in an ExSS speaker mask (HD) - */ -static int dca_exss_mask2count(int mask) -{ - /* count bits that mean speaker pairs twice */ - return av_popcount(mask) + - av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT | - DCA_EXSS_FRONT_LEFT_RIGHT | - DCA_EXSS_FRONT_HIGH_LEFT_RIGHT | - DCA_EXSS_WIDE_LEFT_RIGHT | - DCA_EXSS_SIDE_LEFT_RIGHT | - DCA_EXSS_SIDE_HIGH_LEFT_RIGHT | - DCA_EXSS_SIDE_REAR_LEFT_RIGHT | - DCA_EXSS_REAR_LEFT_RIGHT | - DCA_EXSS_REAR_HIGH_LEFT_RIGHT)); -} - -/** - * Skip mixing coefficients of a single mix out configuration (HD) - */ -static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch) +static float dca_dmix_code(unsigned code) { - int i; - - for (i = 0; i < channels; i++) { - int mix_map_mask = get_bits(gb, out_ch); - int num_coeffs = av_popcount(mix_map_mask); - skip_bits_long(gb, num_coeffs * 6); - } + int sign = (code >> 8) - 1; + code &= 0xff; + return ((ff_dca_dmixtable[code] ^ sign) - sign) * (1.0 / (1U << 15)); } -/** - * Parse extension substream asset header (HD) - */ -static int dca_exss_parse_asset_header(DCAContext *s) +static int scan_for_extensions(AVCodecContext *avctx) { - int header_pos = get_bits_count(&s->gb); - int header_size; - int channels; - int embedded_stereo = 0; - int embedded_6ch = 0; - int drc_code_present; - int extensions_mask; - int i, j; - - if (get_bits_left(&s->gb) < 16) - return -1; - - /* We will parse just enough to get to the extensions bitmask with which - * we can set the profile value. */ - - header_size = get_bits(&s->gb, 9) + 1; - skip_bits(&s->gb, 3); // asset index - - if (s->static_fields) { - if (get_bits1(&s->gb)) - skip_bits(&s->gb, 4); // asset type descriptor - if (get_bits1(&s->gb)) - skip_bits_long(&s->gb, 24); // language descriptor - - if (get_bits1(&s->gb)) { - /* How can one fit 1024 bytes of text here if the maximum value - * for the asset header size field above was 512 bytes? */ - int text_length = get_bits(&s->gb, 10) + 1; - if (get_bits_left(&s->gb) < text_length * 8) - return -1; - skip_bits_long(&s->gb, text_length * 8); // info text - } - - skip_bits(&s->gb, 5); // bit resolution - 1 - skip_bits(&s->gb, 4); // max sample rate code - channels = get_bits(&s->gb, 8) + 1; - - if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers - int spkr_remap_sets; - int spkr_mask_size = 16; - int num_spkrs[7]; - - if (channels > 2) - embedded_stereo = get_bits1(&s->gb); - if (channels > 6) - embedded_6ch = get_bits1(&s->gb); - - if (get_bits1(&s->gb)) { - spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2; - skip_bits(&s->gb, spkr_mask_size); // spkr activity mask - } - - spkr_remap_sets = get_bits(&s->gb, 3); - - for (i = 0; i < spkr_remap_sets; i++) { - /* std layout mask for each remap set */ - num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size)); - } - - for (i = 0; i < spkr_remap_sets; i++) { - int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1; - if (get_bits_left(&s->gb) < 0) - return -1; - - for (j = 0; j < num_spkrs[i]; j++) { - int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps); - int num_dec_ch = av_popcount(remap_dec_ch_mask); - skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes - } - } - - } else { - skip_bits(&s->gb, 3); // representation type - } - } - - drc_code_present = get_bits1(&s->gb); - if (drc_code_present) - get_bits(&s->gb, 8); // drc code - - if (get_bits1(&s->gb)) - skip_bits(&s->gb, 5); // dialog normalization code - - if (drc_code_present && embedded_stereo) - get_bits(&s->gb, 8); // drc stereo code - - if (s->mix_metadata && get_bits1(&s->gb)) { - skip_bits(&s->gb, 1); // external mix - skip_bits(&s->gb, 6); // post mix gain code - - if (get_bits(&s->gb, 2) != 3) // mixer drc code - skip_bits(&s->gb, 3); // drc limit - else - skip_bits(&s->gb, 8); // custom drc code - - if (get_bits1(&s->gb)) // channel specific scaling - for (i = 0; i < s->num_mix_configs; i++) - skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes - else - skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes - - for (i = 0; i < s->num_mix_configs; i++) { - if (get_bits_left(&s->gb) < 0) - return -1; - dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]); - if (embedded_6ch) - dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]); - if (embedded_stereo) - dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]); - } - } - - switch (get_bits(&s->gb, 2)) { - case 0: extensions_mask = get_bits(&s->gb, 12); break; - case 1: extensions_mask = DCA_EXT_EXSS_XLL; break; - case 2: extensions_mask = DCA_EXT_EXSS_LBR; break; - case 3: extensions_mask = 0; /* aux coding */ break; - } - - /* not parsed further, we were only interested in the extensions mask */ - - if (get_bits_left(&s->gb) < 0) - return -1; - - if (get_bits_count(&s->gb) - header_pos > header_size * 8) { - av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n"); - return -1; - } - skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb)); - - if (extensions_mask & DCA_EXT_EXSS_XLL) - s->profile = FF_PROFILE_DTS_HD_MA; - else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 | - DCA_EXT_EXSS_XXCH)) - s->profile = FF_PROFILE_DTS_HD_HRA; - - if (!(extensions_mask & DCA_EXT_CORE)) - av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n"); - if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask) - av_log(s->avctx, AV_LOG_WARNING, - "DTS extensions detection mismatch (%d, %d)\n", - extensions_mask & DCA_CORE_EXTS, s->core_ext_mask); - - return 0; -} - -/** - * Parse extension substream header (HD) - */ -static void dca_exss_parse_header(DCAContext *s) -{ - int ss_index; - int blownup; - int num_audiop = 1; - int num_assets = 1; - int active_ss_mask[8]; - int i, j; - - if (get_bits_left(&s->gb) < 52) - return; - - skip_bits(&s->gb, 8); // user data - ss_index = get_bits(&s->gb, 2); - - blownup = get_bits1(&s->gb); - skip_bits(&s->gb, 8 + 4 * blownup); // header_size - skip_bits(&s->gb, 16 + 4 * blownup); // hd_size - - s->static_fields = get_bits1(&s->gb); - if (s->static_fields) { - skip_bits(&s->gb, 2); // reference clock code - skip_bits(&s->gb, 3); // frame duration code - - if (get_bits1(&s->gb)) - skip_bits_long(&s->gb, 36); // timestamp - - /* a single stream can contain multiple audio assets that can be - * combined to form multiple audio presentations */ - - num_audiop = get_bits(&s->gb, 3) + 1; - if (num_audiop > 1) { - av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations."); - /* ignore such streams for now */ - return; - } - - num_assets = get_bits(&s->gb, 3) + 1; - if (num_assets > 1) { - av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets."); - /* ignore such streams for now */ - return; - } - - for (i = 0; i < num_audiop; i++) - active_ss_mask[i] = get_bits(&s->gb, ss_index + 1); - - for (i = 0; i < num_audiop; i++) - for (j = 0; j <= ss_index; j++) - if (active_ss_mask[i] & (1 << j)) - skip_bits(&s->gb, 8); // active asset mask - - s->mix_metadata = get_bits1(&s->gb); - if (s->mix_metadata) { - int mix_out_mask_size; - - skip_bits(&s->gb, 2); // adjustment level - mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2; - s->num_mix_configs = get_bits(&s->gb, 2) + 1; - - for (i = 0; i < s->num_mix_configs; i++) { - int mix_out_mask = get_bits(&s->gb, mix_out_mask_size); - s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask); - } - } - } - - for (i = 0; i < num_assets; i++) - skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size - - for (i = 0; i < num_assets; i++) { - if (dca_exss_parse_asset_header(s)) - return; - } - - /* not parsed further, we were only interested in the extensions mask - * from the asset header */ -} - -/** - * Main frame decoding function - * FIXME add arguments - */ -static int dca_decode_frame(AVCodecContext *avctx, void *data, - int *got_frame_ptr, AVPacket *avpkt) -{ - const uint8_t *buf = avpkt->data; - int buf_size = avpkt->size; - - int lfe_samples; - int num_core_channels = 0; - int i, ret; - float **samples_flt; DCAContext *s = avctx->priv_data; - int channels, full_channels; - int core_ss_end; - - - s->xch_present = 0; - - s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer, - DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE); - if (s->dca_buffer_size == AVERROR_INVALIDDATA) { - av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n"); - return AVERROR_INVALIDDATA; - } - - init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8); - if ((ret = dca_parse_frame_header(s)) < 0) { - //seems like the frame is corrupt, try with the next one - return ret; - } - //set AVCodec values with parsed data - avctx->sample_rate = s->sample_rate; - avctx->bit_rate = s->bit_rate; - - s->profile = FF_PROFILE_DTS; - - for (i = 0; i < (s->sample_blocks / 8); i++) { - if ((ret = dca_decode_block(s, 0, i))) { - av_log(avctx, AV_LOG_ERROR, "error decoding block\n"); - return ret; - } - } - - /* record number of core channels incase less than max channels are requested */ - num_core_channels = s->prim_channels; - - if (s->ext_coding) - s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr]; - else - s->core_ext_mask = 0; + int core_ss_end, ret = 0; core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8; /* only scan for extensions if ext_descr was unknown or indicated a * supported XCh extension */ if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) { - /* if ext_descr was unknown, clear s->core_ext_mask so that the * extensions scan can fill it up */ s->core_ext_mask = FFMAX(s->core_ext_mask, 0); @@ -1714,12 +1178,13 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data, while (core_ss_end - get_bits_count(&s->gb) >= 32) { uint32_t bits = get_bits_long(&s->gb, 32); + int i; switch (bits) { - case 0x5a5a5a5a: { + case DCA_SYNCWORD_XCH: { int ext_amode, xch_fsize; - s->xch_base_channel = s->prim_channels; + s->xch_base_channel = s->audio_header.prim_channels; /* validate sync word using XCHFSIZE field */ xch_fsize = show_bits(&s->gb, 10); @@ -1735,8 +1200,9 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data, /* extension amode(number of channels in extension) should be 1 */ /* AFAIK XCh is not used for more channels */ if ((ext_amode = get_bits(&s->gb, 4)) != 1) { - av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not" - " supported!\n", ext_amode); + av_log(avctx, AV_LOG_ERROR, + "XCh extension amode %d not supported!\n", + ext_amode); continue; } @@ -1752,7 +1218,7 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data, s->xch_present = 1; break; } - case 0x47004a03: + case DCA_SYNCWORD_XXCH: /* XXCh: extended channels */ /* usually found either in core or HD part in DTS-HD HRA streams, * but not in DTS-ES which contains XCh extensions instead */ @@ -1789,77 +1255,220 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data, /* check for ExSS (HD part) */ if (s->dca_buffer_size - s->frame_size > 32 && - get_bits_long(&s->gb, 32) == DCA_HD_MARKER) - dca_exss_parse_header(s); + get_bits_long(&s->gb, 32) == DCA_SYNCWORD_SUBSTREAM) + ff_dca_exss_parse_header(s); - avctx->profile = s->profile; + return ret; +} - full_channels = channels = s->prim_channels + !!s->lfe; +static int set_channel_layout(AVCodecContext *avctx, int channels, int num_core_channels) +{ + DCAContext *s = avctx->priv_data; + int i; if (s->amode < 16) { avctx->channel_layout = dca_core_channel_layout[s->amode]; - if (s->xch_present && (!avctx->request_channels || - avctx->request_channels > num_core_channels + !!s->lfe)) { + if (s->audio_header.prim_channels + !!s->lfe > 2 && + avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) { + /* + * Neither the core's auxiliary data nor our default tables contain + * downmix coefficients for the additional channel coded in the XCh + * extension, so when we're doing a Stereo downmix, don't decode it. + */ + s->xch_disable = 1; + } + + if (s->xch_present && !s->xch_disable) { avctx->channel_layout |= AV_CH_BACK_CENTER; if (s->lfe) { avctx->channel_layout |= AV_CH_LOW_FREQUENCY; - s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode]; + s->channel_order_tab = ff_dca_channel_reorder_lfe_xch[s->amode]; } else { - s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode]; + s->channel_order_tab = ff_dca_channel_reorder_nolfe_xch[s->amode]; } } else { - channels = num_core_channels + !!s->lfe; + channels = num_core_channels + !!s->lfe; s->xch_present = 0; /* disable further xch processing */ if (s->lfe) { avctx->channel_layout |= AV_CH_LOW_FREQUENCY; - s->channel_order_tab = dca_channel_reorder_lfe[s->amode]; + s->channel_order_tab = ff_dca_channel_reorder_lfe[s->amode]; } else - s->channel_order_tab = dca_channel_reorder_nolfe[s->amode]; + s->channel_order_tab = ff_dca_channel_reorder_nolfe[s->amode]; } if (channels > !!s->lfe && s->channel_order_tab[channels - 1 - !!s->lfe] < 0) return AVERROR_INVALIDDATA; - if (avctx->request_channels == 2 && s->prim_channels > 2) { - channels = 2; - s->output = DCA_STEREO; + if (num_core_channels + !!s->lfe > 2 && + avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) { + channels = 2; + s->output = s->audio_header.prim_channels == 2 ? s->amode : DCA_STEREO; avctx->channel_layout = AV_CH_LAYOUT_STEREO; + + /* Stereo downmix coefficients + * + * The decoder can only downmix to 2-channel, so we need to ensure + * embedded downmix coefficients are actually targeting 2-channel. + */ + if (s->core_downmix && (s->core_downmix_amode == DCA_STEREO || + s->core_downmix_amode == DCA_STEREO_TOTAL)) { + for (i = 0; i < num_core_channels + !!s->lfe; i++) { + /* Range checked earlier */ + s->downmix_coef[i][0] = dca_dmix_code(s->core_downmix_codes[i][0]); + s->downmix_coef[i][1] = dca_dmix_code(s->core_downmix_codes[i][1]); + } + s->output = s->core_downmix_amode; + } else { + int am = s->amode & DCA_CHANNEL_MASK; + if (am >= FF_ARRAY_ELEMS(ff_dca_default_coeffs)) { + av_log(s->avctx, AV_LOG_ERROR, + "Invalid channel mode %d\n", am); + return AVERROR_INVALIDDATA; + } + if (num_core_channels + !!s->lfe > + FF_ARRAY_ELEMS(ff_dca_default_coeffs[0])) { + avpriv_request_sample(s->avctx, "Downmixing %d channels", + s->audio_header.prim_channels + !!s->lfe); + return AVERROR_PATCHWELCOME; + } + for (i = 0; i < num_core_channels + !!s->lfe; i++) { + s->downmix_coef[i][0] = ff_dca_default_coeffs[am][i][0]; + s->downmix_coef[i][1] = ff_dca_default_coeffs[am][i][1]; + } + } + ff_dlog(s->avctx, "Stereo downmix coeffs:\n"); + for (i = 0; i < num_core_channels + !!s->lfe; i++) { + ff_dlog(s->avctx, "L, input channel %d = %f\n", i, + s->downmix_coef[i][0]); + ff_dlog(s->avctx, "R, input channel %d = %f\n", i, + s->downmix_coef[i][1]); + } + ff_dlog(s->avctx, "\n"); } } else { - av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode); + av_log(avctx, AV_LOG_ERROR, "Nonstandard configuration %d !\n", s->amode); return AVERROR_INVALIDDATA; } + return 0; +} + +/** + * Main frame decoding function + * FIXME add arguments + */ +static int dca_decode_frame(AVCodecContext *avctx, void *data, + int *got_frame_ptr, AVPacket *avpkt) +{ + AVFrame *frame = data; + const uint8_t *buf = avpkt->data; + int buf_size = avpkt->size; + + int lfe_samples; + int num_core_channels = 0; + int i, ret; + float **samples_flt; + DCAContext *s = avctx->priv_data; + int channels, full_channels; + int upsample = 0; + + s->exss_ext_mask = 0; + s->xch_present = 0; + + s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer, + DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE); + if (s->dca_buffer_size == AVERROR_INVALIDDATA) { + av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n"); + return AVERROR_INVALIDDATA; + } - /* There is nothing that prevents a dts frame to change channel configuration - but Libav 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 (s->is_channels_set == 0) { - s->is_channels_set = 1; - avctx->channels = channels; + if ((ret = dca_parse_frame_header(s)) < 0) { + // seems like the frame is corrupt, try with the next one + return ret; } - if (avctx->channels != channels) { - av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of " - "channels changing in stream. Skipping frame.\n"); - return AVERROR_PATCHWELCOME; + // set AVCodec values with parsed data + avctx->sample_rate = s->sample_rate; + avctx->bit_rate = s->bit_rate; + + s->profile = FF_PROFILE_DTS; + + for (i = 0; i < (s->sample_blocks / SAMPLES_PER_SUBBAND); i++) { + if ((ret = dca_decode_block(s, 0, i))) { + av_log(avctx, AV_LOG_ERROR, "error decoding block\n"); + return ret; + } } + /* record number of core channels incase less than max channels are requested */ + num_core_channels = s->audio_header.prim_channels; + + if (s->ext_coding) + s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr]; + else + s->core_ext_mask = 0; + + ret = scan_for_extensions(avctx); + + avctx->profile = s->profile; + + full_channels = channels = s->audio_header.prim_channels + !!s->lfe; + + ret = set_channel_layout(avctx, channels, num_core_channels); + if (ret < 0) + return ret; + avctx->channels = channels; + /* get output buffer */ - s->frame.nb_samples = 256 * (s->sample_blocks / 8); - if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) { + frame->nb_samples = 256 * (s->sample_blocks / SAMPLES_PER_SUBBAND); + if (s->exss_ext_mask & DCA_EXT_EXSS_XLL) { + int xll_nb_samples = s->xll_segments * s->xll_smpl_in_seg; + /* Check for invalid/unsupported conditions first */ + if (s->xll_residual_channels > channels) { + av_log(s->avctx, AV_LOG_WARNING, + "DCA: too many residual channels (%d, core channels %d). Disabling XLL\n", + s->xll_residual_channels, channels); + s->exss_ext_mask &= ~DCA_EXT_EXSS_XLL; + } else if (xll_nb_samples != frame->nb_samples && + 2 * frame->nb_samples != xll_nb_samples) { + av_log(s->avctx, AV_LOG_WARNING, + "DCA: unsupported upsampling (%d XLL samples, %d core samples). Disabling XLL\n", + xll_nb_samples, frame->nb_samples); + s->exss_ext_mask &= ~DCA_EXT_EXSS_XLL; + } else { + if (2 * frame->nb_samples == xll_nb_samples) { + av_log(s->avctx, AV_LOG_INFO, + "XLL: upsampling core channels by a factor of 2\n"); + upsample = 1; + + frame->nb_samples = xll_nb_samples; + // FIXME: Is it good enough to copy from the first channel set? + avctx->sample_rate = s->xll_chsets[0].sampling_frequency; + } + /* If downmixing to stereo, don't decode additional channels. + * FIXME: Using the xch_disable flag for this doesn't seem right. */ + if (!s->xch_disable) + avctx->channels += s->xll_channels - s->xll_residual_channels; + } + } + + /* FIXME: This is an ugly hack, to just revert to the default + * layout if we have additional channels. Need to convert the XLL + * channel masks to libav channel_layout mask. */ + if (av_get_channel_layout_nb_channels(avctx->channel_layout) != avctx->channels) + avctx->channel_layout = 0; + + if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) { av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); return ret; } - samples_flt = (float **) s->frame.extended_data; + samples_flt = (float **) frame->extended_data; /* allocate buffer for extra channels if downmixing */ if (avctx->channels < full_channels) { ret = av_samples_get_buffer_size(NULL, full_channels - channels, - s->frame.nb_samples, + frame->nb_samples, avctx->sample_fmt, 0); if (ret < 0) return ret; @@ -1869,24 +1478,24 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data, if (!s->extra_channels_buffer) return AVERROR(ENOMEM); - ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL, + ret = av_samples_fill_arrays((uint8_t **) s->extra_channels, NULL, s->extra_channels_buffer, full_channels - channels, - s->frame.nb_samples, avctx->sample_fmt, 0); + frame->nb_samples, avctx->sample_fmt, 0); if (ret < 0) return ret; } /* filter to get final output */ - for (i = 0; i < (s->sample_blocks / 8); i++) { + for (i = 0; i < (s->sample_blocks / SAMPLES_PER_SUBBAND); i++) { int ch; - + unsigned block = upsample ? 512 : 256; for (ch = 0; ch < channels; ch++) - s->samples_chanptr[ch] = samples_flt[ch] + i * 256; + s->samples_chanptr[ch] = samples_flt[ch] + i * block; for (; ch < full_channels; ch++) - s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256; + s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * block; - dca_filter_channels(s, i); + dca_filter_channels(s, i, upsample); /* If this was marked as a DTS-ES stream we need to subtract back- */ /* channel from SL & SR to remove matrixed back-channel signal */ @@ -1900,18 +1509,29 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data, } /* update lfe history */ - lfe_samples = 2 * s->lfe * (s->sample_blocks / 8); + lfe_samples = 2 * s->lfe * (s->sample_blocks / SAMPLES_PER_SUBBAND); for (i = 0; i < 2 * s->lfe * 4; i++) s->lfe_data[i] = s->lfe_data[i + lfe_samples]; - *got_frame_ptr = 1; - *(AVFrame *) data = s->frame; + if (s->exss_ext_mask & DCA_EXT_EXSS_XLL) { + ret = ff_dca_xll_decode_audio(s, frame); + if (ret < 0) + return ret; + } + /* AVMatrixEncoding + * + * DCA_STEREO_TOTAL (Lt/Rt) is equivalent to Dolby Surround */ + ret = ff_side_data_update_matrix_encoding(frame, + (s->output & ~DCA_LFE) == DCA_STEREO_TOTAL ? + AV_MATRIX_ENCODING_DOLBY : AV_MATRIX_ENCODING_NONE); + if (ret < 0) + return ret; + + *got_frame_ptr = 1; return buf_size; } - - /** * DCA initialization * @@ -1925,7 +1545,7 @@ static av_cold int dca_decode_init(AVCodecContext *avctx) s->avctx = avctx; dca_init_vlcs(); - avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT); + avpriv_float_dsp_init(&s->fdsp, avctx->flags & AV_CODEC_FLAG_BITEXACT); ff_mdct_init(&s->imdct, 6, 1, 1.0); ff_synth_filter_init(&s->synth); ff_dcadsp_init(&s->dcadsp); @@ -1934,13 +1554,9 @@ static av_cold int dca_decode_init(AVCodecContext *avctx) avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; /* allow downmixing to stereo */ - if (avctx->channels > 0 && avctx->request_channels < avctx->channels && - avctx->request_channels == 2) { - avctx->channels = avctx->request_channels; - } - - avcodec_get_frame_defaults(&s->frame); - avctx->coded_frame = &s->frame; + if (avctx->channels > 2 && + avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) + avctx->channels = 2; return 0; } @@ -1950,29 +1566,36 @@ static av_cold int dca_decode_end(AVCodecContext *avctx) DCAContext *s = avctx->priv_data; ff_mdct_end(&s->imdct); av_freep(&s->extra_channels_buffer); + av_freep(&s->xll_sample_buf); + av_freep(&s->qmf64_table); return 0; } -static const AVProfile profiles[] = { - { FF_PROFILE_DTS, "DTS" }, - { FF_PROFILE_DTS_ES, "DTS-ES" }, - { FF_PROFILE_DTS_96_24, "DTS 96/24" }, - { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" }, - { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" }, - { FF_PROFILE_UNKNOWN }, +static const AVOption options[] = { + { "disable_xch", "disable decoding of the XCh extension", offsetof(DCAContext, xch_disable), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM }, + { "disable_xll", "disable decoding of the XLL extension", offsetof(DCAContext, xll_disable), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM }, + { NULL }, +}; + +static const AVClass dca_decoder_class = { + .class_name = "DCA decoder", + .item_name = av_default_item_name, + .option = options, + .version = LIBAVUTIL_VERSION_INT, }; AVCodec ff_dca_decoder = { .name = "dca", + .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"), .type = AVMEDIA_TYPE_AUDIO, .id = AV_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)"), - .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1, + .capabilities = AV_CODEC_CAP_CHANNEL_CONF | AV_CODEC_CAP_DR1, .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE }, - .profiles = NULL_IF_CONFIG_SMALL(profiles), + .profiles = NULL_IF_CONFIG_SMALL(ff_dca_profiles), + .priv_class = &dca_decoder_class, };