xenia-project/FFmpeg
1
0
Fork 0
mirror of https://github.com/xenia-project/FFmpeg.git synced 2024-11-24 03:59:43 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge commit '58b42345b38b46d11c32e11d9c57517f99d6a601'

Browse Source 
* commit '58b42345b38b46d11c32e11d9c57517f99d6a601':
  dcadec: reorganise context data

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>
This commit is contained in:
Hendrik Leppkes 2015-10-10 09:32:59 +02:00
commit a71fff213d
2 changed files with 197 additions and 169 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

70
libavcodec/dca.h
View File

@ -132,6 +132,47 @@ typedef struct QMF64_table {
float rsin[32];
} QMF64_table;
/* Primary audio coding header */
typedef struct DCAAudioHeader {
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
int subframes; ///< number of subframes
int total_channels; ///< number of channels including extensions
int prim_channels; ///< number of primary audio channels
} DCAAudioHeader;
typedef struct DCAChan {
DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_SUBBANDS][8];
/* Subband samples history (for ADPCM) */
DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_SUBBANDS][4];
int hist_index;
/* Half size is sufficient for core decoding, but for 96 kHz data
* we need QMF with 64 subbands and 1024 samples. */
DECLARE_ALIGNED(32, float, subband_fir_hist)[1024];
DECLARE_ALIGNED(32, float, subband_fir_noidea)[64];
/* Primary audio coding side information */
int prediction_mode[DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
int prediction_vq[DCA_SUBBANDS]; ///< prediction VQ coefs
int bitalloc[DCA_SUBBANDS]; ///< bit allocation index
int transition_mode[DCA_SUBBANDS]; ///< transition mode (transients)
int32_t scale_factor[DCA_SUBBANDS][2];///< scale factors (2 if transient)
int joint_huff; ///< joint subband scale factors codebook
int joint_scale_factor[DCA_SUBBANDS]; ///< joint subband scale factors
int32_t high_freq_vq[DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
} DCAChan;
typedef struct DCAContext {
const AVClass *class; ///< class for AVOptions
AVCodecContext *avctx;
@ -165,28 +206,11 @@ typedef struct DCAContext {
int dialog_norm; ///< dialog normalisation parameter
/* 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
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
DCAAudioHeader audio_header;
/* 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)
int32_t 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
float downmix_coef[DCA_PRIM_CHANNELS_MAX + 1][2]; ///< stereo downmix coefficients
int dynrange_coef; ///< dynamic range coefficient
@ -197,23 +221,17 @@ typedef struct DCAContext {
uint8_t core_downmix_amode; ///< audio channel arrangement of embedded downmix
uint16_t core_downmix_codes[DCA_PRIM_CHANNELS_MAX + 1][4]; ///< embedded downmix coefficients (9-bit codes)
int32_t 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];
/* Half size is sufficient for core decoding, but for 96 kHz data
* we need QMF with 64 subbands and 1024 samples. */
DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][1024];
DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][64];
int hist_index[DCA_PRIM_CHANNELS_MAX];
DECLARE_ALIGNED(32, float, raXin)[32];
DCAChan dca_chan[DCA_PRIM_CHANNELS_MAX];
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;

296
libavcodec/dcadec.c
View File

@ -238,8 +238,8 @@ static int dca_parse_audio_coding_header(DCAContext *s, int base_channel,
return AVERROR_INVALIDDATA;
}
s->total_channels = nchans + base_channel;
s->prim_channels = s->total_channels;
s->audio_header.total_channels = nchans + base_channel;
s->audio_header.prim_channels = s->audio_header.total_channels;
/* obtain speaker layout mask & downmix coefficients for XXCH */
if (xxch) {
@ -266,11 +266,11 @@ static int dca_parse_audio_coding_header(DCAContext *s, int base_channel,
s->xxch_dmix_sf[s->xxch_chset] = scale_factor;
for (i = base_channel; i < s->prim_channels; i++) {
for (i = base_channel; i < s->audio_header.prim_channels; i++) {
mask[i] = get_bits(&s->gb, s->xxch_nbits_spk_mask);
}
for (j = base_channel; j < s->prim_channels; j++) {
for (j = base_channel; j < s->audio_header.prim_channels; j++) {
memset(s->xxch_dmix_coeff[j], 0, sizeof(s->xxch_dmix_coeff[0]));
s->xxch_dmix_embedded |= (embedded_downmix << j);
for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
@ -294,40 +294,44 @@ static int dca_parse_audio_coding_header(DCAContext *s, int base_channel,
}
}
if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
s->prim_channels = DCA_PRIM_CHANNELS_MAX;
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] = 1;
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 (!xxch) {
if (s->crc_present) {
@ -406,7 +410,7 @@ static int dca_parse_frame_header(DCAContext *s)
s->output |= DCA_LFE;
/* 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, 0);
}
@ -445,53 +449,53 @@ 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) {
if (s->dca_chan[j].bitalloc[k] > 26) {
ff_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
j, k, s->bitalloc[j][k]);
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]);
}
}
}
@ -499,14 +503,14 @@ 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) {
if (s->audio_header.scalefactor_huffman[j] == 6) {
scale_table = ff_dca_scale_factor_quant7;
log_size = 7;
} else {
@ -517,45 +521,46 @@ static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
/* 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)) {
@ -580,10 +585,10 @@ 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) {
@ -622,7 +627,7 @@ static void qmf_32_subbands(DCAContext *s, int chans,
{
const float *prCoeff;
int sb_act = s->subband_activity[chans];
int sb_act = s->audio_header.subband_activity[chans];
scale *= sqrt(1 / 8.0);
@ -633,9 +638,9 @@ static void qmf_32_subbands(DCAContext *s, int chans,
prCoeff = ff_dca_fir_32bands_perfect;
s->dcadsp.qmf_32_subbands(samples_in, sb_act, &s->synth, &s->imdct,
s->subband_fir_hist[chans],
&s->hist_index[chans],
s->subband_fir_noidea[chans], prCoeff,
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);
}
@ -670,14 +675,14 @@ static void qmf_64_subbands(DCAContext *s, int chans, float samples_in[64][SAMPL
{
float raXin[64];
float A[32], B[32];
float *raX = s->subband_fir_hist[chans];
float *raZ = s->subband_fir_noidea[chans];
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->subband_activity[chans]; i < 64; i++)
for (i = s->audio_header.subband_activity[chans]; i < 64; i++)
raXin[i] = 0.0;
for (subindex = 0; subindex < SAMPLES_PER_SUBBAND; subindex++) {
for (i = 0; i < s->subband_activity[chans]; i++)
for (i = 0; i < s->audio_header.subband_activity[chans]; i++)
raXin[i] = samples_in[i][subindex];
for (k = 0; k < 32; k++) {
@ -866,8 +871,6 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
const float *quant_step_table;
/* FIXME */
float (*subband_samples)[DCA_SUBBANDS][SAMPLES_PER_SUBBAND] = s->subband_samples[block_index];
LOCAL_ALIGNED_16(int32_t, block, [SAMPLES_PER_SUBBAND * DCA_SUBBANDS]);
/*
@ -880,17 +883,18 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
else
quant_step_table = ff_dca_lossy_quant_d;
for (k = base_channel; k < s->prim_channels; k++) {
for (k = base_channel; k < s->audio_header.prim_channels; k++) {
float (*subband_samples)[8] = s->dca_chan[k].subband_samples[block_index];
float rscale[DCA_SUBBANDS];
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];
int abits = s->dca_chan[k].bitalloc[l];
float quant_step_size = quant_step_table[abits];
@ -899,7 +903,7 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
*/
/* Select quantization index code book */
int sel = s->quant_index_huffman[k][abits];
int sel = s->audio_header.quant_index_huffman[k][abits];
/*
* Extract bits from the bit stream
@ -909,9 +913,10 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
memset(block + SAMPLES_PER_SUBBAND * l, 0, SAMPLES_PER_SUBBAND * sizeof(block[0]));
} else {
/* Deal with transients */
int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
rscale[l] = 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];
rscale[l] = quant_step_size * s->dca_chan[k].scale_factor[l][sfi] *
s->audio_header.scalefactor_adj[k][sel];
if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
if (abits <= 7) {
@ -944,54 +949,61 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
}
}
s->fmt_conv.int32_to_float_fmul_array8(&s->fmt_conv, subband_samples[k][0],
block, rscale, SAMPLES_PER_SUBBAND * s->vq_start_subband[k]);
s->fmt_conv.int32_to_float_fmul_array8(&s->fmt_conv, subband_samples[0],
block, rscale, SAMPLES_PER_SUBBAND * s->audio_header.vq_start_subband[k]);
for (l = 0; l < s->vq_start_subband[k]; l++) {
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;
if (s->predictor_history)
subband_samples[k][l][0] += (ff_dca_adpcm_vb[s->prediction_vq[k][l]][0] *
s->subband_samples_hist[k][l][3] +
ff_dca_adpcm_vb[s->prediction_vq[k][l]][1] *
s->subband_samples_hist[k][l][2] +
ff_dca_adpcm_vb[s->prediction_vq[k][l]][2] *
s->subband_samples_hist[k][l][1] +
ff_dca_adpcm_vb[s->prediction_vq[k][l]][3] *
s->subband_samples_hist[k][l][0]) *
subband_samples[l][0] += (ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][0] *
s->dca_chan[k].subband_samples_hist[l][3] +
ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][1] *
s->dca_chan[k].subband_samples_hist[l][2] +
ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][2] *
s->dca_chan[k].subband_samples_hist[l][1] +
ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][3] *
s->dca_chan[k].subband_samples_hist[l][0]) *
(1.0f / 8192);
for (m = 1; m < SAMPLES_PER_SUBBAND; m++) {
float sum = ff_dca_adpcm_vb[s->prediction_vq[k][l]][0] *
subband_samples[k][l][m - 1];
float sum = ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][0] *
subband_samples[l][m - 1];
for (n = 2; n <= 4; n++)
if (m >= n)
sum += ff_dca_adpcm_vb[s->prediction_vq[k][l]][n - 1] *
subband_samples[k][l][m - n];
sum += ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][n - 1] *
subband_samples[l][m - n];
else if (s->predictor_history)
sum += ff_dca_adpcm_vb[s->prediction_vq[k][l]][n - 1] *
s->subband_samples_hist[k][l][m - n + 4];
subband_samples[k][l][m] += sum * (1.0f / 8192);
sum += ff_dca_adpcm_vb[s->dca_chan[k].prediction_vq[l]][n - 1] *
s->dca_chan[k].subband_samples_hist[l][m - n + 4];
subband_samples[l][m] += sum * (1.0f / 8192);
}
}
}
/* 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
*/
if (s->subband_activity[k] > s->vq_start_subband[k]) {
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;
}
s->dcadsp.decode_hf(subband_samples[k], s->high_freq_vq[k],
s->dcadsp.decode_hf(subband_samples, s->dca_chan[k].high_freq_vq,
ff_dca_high_freq_vq, subsubframe * SAMPLES_PER_SUBBAND,
s->scale_factor[k], s->vq_start_subband[k],
s->subband_activity[k]);
s->dca_chan[k].scale_factor,
s->audio_header.vq_start_subband[k],
s->audio_header.subband_activity[k]);
}
}
@ -1003,17 +1015,11 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
}
}
/* Backup predictor history for adpcm */
for (k = base_channel; k < s->prim_channels; k++)
for (l = 0; l < s->vq_start_subband[k]; l++)
AV_COPY128(s->subband_samples_hist[k][l], &subband_samples[k][l][4]);
return 0;
}
static int dca_filter_channels(DCAContext *s, int block_index, int upsample)
{
float (*subband_samples)[DCA_SUBBANDS][SAMPLES_PER_SUBBAND] = s->subband_samples[block_index];
int k;
if (upsample) {
@ -1024,18 +1030,22 @@ static int dca_filter_channels(DCAContext *s, int block_index, int upsample)
}
/* 64 subbands QMF */
for (k = 0; k < s->prim_channels; k++) {
for (k = 0; k < s->audio_header.prim_channels; k++) {
float (*subband_samples)[SAMPLES_PER_SUBBAND] = s->dca_chan[k].subband_samples[block_index];
if (s->channel_order_tab[k] >= 0)
qmf_64_subbands(s, k, subband_samples[k],
qmf_64_subbands(s, k, subband_samples,
s->samples_chanptr[s->channel_order_tab[k]],
/* Upsampling needs a factor 2 here. */
M_SQRT2 / 32768.0);
}
} else {
/* 32 subbands QMF */
for (k = 0; k < s->prim_channels; k++) {
for (k = 0; k < s->audio_header.prim_channels; k++) {
float (*subband_samples)[SAMPLES_PER_SUBBAND] = s->dca_chan[k].subband_samples[block_index];
if (s->channel_order_tab[k] >= 0)
qmf_32_subbands(s, k, subband_samples[k],
qmf_32_subbands(s, k, subband_samples,
s->samples_chanptr[s->channel_order_tab[k]],
M_SQRT1_2 / 32768.0);
}
@ -1062,7 +1072,7 @@ static int dca_filter_channels(DCAContext *s, int block_index, int upsample)
/* FIXME: This downmixing is probably broken with upsample.
* Probably totally broken also with XLL in general. */
/* Downmixing to Stereo */
if (s->prim_channels + !!s->lfe > 2 &&
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);
@ -1139,7 +1149,7 @@ static int dca_subframe_footer(DCAContext *s, int base_channel)
return AVERROR_INVALIDDATA;
}
for (out = 0; out < ff_dca_channels[s->core_downmix_amode]; out++) {
for (in = 0; in < s->prim_channels + !!s->lfe; in++) {
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,
@ -1185,9 +1195,9 @@ 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;
}
@ -1207,7 +1217,7 @@ 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) {
if (s->current_subframe >= s->audio_header.subframes) {
/* Read subframe footer */
if ((ret = dca_subframe_footer(s, base_channel)))
return ret;
@ -1260,7 +1270,7 @@ int ff_dca_xbr_parse_frame(DCAContext *s)
/* loop over the channel data sets */
/* only decode as many channels as we've decoded base data for */
for(chset = 0, chan_base = 0;
chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->prim_channels;
chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->audio_header.prim_channels;
chan_base += n_xbr_ch[chset++]) {
int start_posn = get_bits_count(&s->gb);
int subsubframe = 0;
@ -1293,7 +1303,7 @@ int ff_dca_xbr_parse_frame(DCAContext *s)
int nbits;
int scale_table_size;
if (s->scalefactor_huffman[chan_base+i] == 6) {
if (s->audio_header.scalefactor_huffman[chan_base+i] == 6) {
scale_table = ff_dca_scale_factor_quant7;
scale_table_size = FF_ARRAY_ELEMS(ff_dca_scale_factor_quant7);
} else {
@ -1312,7 +1322,7 @@ int ff_dca_xbr_parse_frame(DCAContext *s)
}
scale_table_high[i][j][0] = scale_table[index];
if(xbr_tmode && s->transition_mode[i][j]) {
if(xbr_tmode && s->dca_chan[i].transition_mode[j]) {
int index = get_bits(&s->gb, nbits);
if (index >= scale_table_size) {
av_log(s->avctx, AV_LOG_ERROR, "scale table index %d invalid\n", index);
@ -1330,9 +1340,9 @@ int ff_dca_xbr_parse_frame(DCAContext *s)
for(j = 0; j < active_bands[chset][i]; j++) {
const int xbr_abits = abits_high[i][j];
const float quant_step_size = ff_dca_lossless_quant_d[xbr_abits];
const int sfi = xbr_tmode && s->transition_mode[i][j] && subsubframe >= s->transition_mode[i][j];
const int sfi = xbr_tmode && s->dca_chan[i].transition_mode[j] && subsubframe >= s->dca_chan[i].transition_mode[j];
const float rscale = quant_step_size * scale_table_high[i][j][sfi];
float *subband_samples = s->subband_samples[k][chan_base+i][j];
float *subband_samples = s->dca_chan[chan_base+i].subband_samples[k][j];
int block[8];
if(xbr_abits <= 0)
@ -1421,7 +1431,7 @@ int ff_dca_xxch_decode_frame(DCAContext *s)
for (chset = 0; chset < num_chsets; chset++) {
chstart = get_bits_count(&s->gb);
base_channel = s->prim_channels;
base_channel = s->audio_header.prim_channels;
s->xxch_chset = chset;
/* XXCH and Core headers differ, see 6.4.2 "XXCH Channel Set Header" vs.
@ -1479,7 +1489,7 @@ static int scan_for_extensions(AVCodecContext *avctx)
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);
@ -1576,7 +1586,7 @@ static int set_channel_layout(AVCodecContext *avctx, int *channels, int num_core
if (s->amode < 16) {
avctx->channel_layout = ff_dca_core_channel_layout[s->amode];
if (s->prim_channels + !!s->lfe > 2 &&
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
@ -1622,7 +1632,7 @@ static int set_channel_layout(AVCodecContext *avctx, int *channels, int num_core
if (num_core_channels + !!s->lfe > 2 &&
avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) {
*channels = 2;
s->output = s->prim_channels == 2 ? s->amode : DCA_STEREO;
s->output = s->audio_header.prim_channels == 2 ? s->amode : DCA_STEREO;
avctx->channel_layout = AV_CH_LAYOUT_STEREO;
}
else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
@ -1642,7 +1652,7 @@ static int set_channel_layout(AVCodecContext *avctx, int *channels, int num_core
channel_mask = s->xxch_core_spkmask;
{
*channels = s->prim_channels + !!s->lfe;
*channels = s->audio_header.prim_channels + !!s->lfe;
for (i = 0; i < s->xxch_chset; i++) {
channel_mask |= s->xxch_spk_masks[i];
}
@ -1751,9 +1761,9 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data,
}
/* record number of core channels incase less than max channels are requested */
num_core_channels = s->prim_channels;
num_core_channels = s->audio_header.prim_channels;
if (s->prim_channels + !!s->lfe > 2 &&
if (s->audio_header.prim_channels + !!s->lfe > 2 &&
avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) {
/* Stereo downmix coefficients
*
@ -1778,7 +1788,7 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data,
if (num_core_channels + !!s->lfe >
FF_ARRAY_ELEMS(ff_dca_default_coeffs[0])) {
avpriv_request_sample(s->avctx, "Downmixing %d channels",
s->prim_channels + !!s->lfe);
s->audio_header.prim_channels + !!s->lfe);
return AVERROR_PATCHWELCOME;
}
for (i = 0; i < num_core_channels + !!s->lfe; i++) {
@ -1805,7 +1815,7 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data,
avctx->profile = s->profile;
full_channels = channels = s->prim_channels + !!s->lfe;
full_channels = channels = s->audio_header.prim_channels + !!s->lfe;
ret = set_channel_layout(avctx, &channels, num_core_channels);
if (ret < 0)