float* coeffs; ///< pointer to the subframe decode buffer
uint16_t num_vec_coeffs; ///< number of vector coded coefficients
DECLARE_ALIGNED(16, float, out)[WMALL_BLOCK_MAX_SIZE + WMALL_BLOCK_MAX_SIZE / 2]; ///< output buffer
+ int transient_counter; ///< number of transient samples from the beginning of transient zone
} WmallChannelCtx;
/**
uint32_t frame_num; ///< current frame number (not used for decoding)
GetBitContext gb; ///< bitstream reader context
int buf_bit_size; ///< buffer size in bits
- float* samples; ///< current samplebuffer pointer
- float* samples_end; ///< maximum samplebuffer pointer
+ int16_t* samples_16; ///< current samplebuffer pointer (16-bit)
+ int16_t* samples_16_end; ///< maximum samplebuffer pointer
+ int16_t* samples_32; ///< current samplebuffer pointer (24-bit)
+ int16_t* samples_32_end; ///< maximum samplebuffer pointer
uint8_t drc_gain; ///< gain for the DRC tool
int8_t skip_frame; ///< skip output step
int8_t parsed_all_subframes; ///< all subframes decoded?
WmallChannelCtx channel[WMALL_MAX_CHANNELS]; ///< per channel data
// WMA lossless
-
+
uint8_t do_arith_coding;
uint8_t do_ac_filter;
uint8_t do_inter_ch_decorr;
int8_t acfilter_order;
int8_t acfilter_scaling;
- int acfilter_coeffs[16];
+ int64_t acfilter_coeffs[16];
+ int acfilter_prevvalues[2][16];
int8_t mclms_order;
int8_t mclms_scaling;
int16_t mclms_coeffs[128];
int16_t mclms_coeffs_cur[4];
- int mclms_prevvalues[64]; // FIXME: should be 32-bit / 16-bit depending on bit-depth
+ int16_t mclms_prevvalues[64]; // FIXME: should be 32-bit / 16-bit depending on bit-depth
int16_t mclms_updates[64];
int mclms_recent;
int quant_stepsize;
struct {
- int order;
- int scaling;
- int coefsend;
- int bitsend;
- int16_t coefs[256];
- int lms_prevvalues[512]; // FIXME: see above
+ int order;
+ int scaling;
+ int coefsend;
+ int bitsend;
+ int16_t coefs[256];
+ int16_t lms_prevvalues[512]; // FIXME: see above
int16_t lms_updates[512]; // and here too
int recent;
} cdlms[2][9]; /* XXX: Here, 2 is the max. no. of channels allowed,
int lpc_scaling;
int lpc_intbits;
- int channel_coeffs[2][2048];
+ int channel_coeffs[2][2048]; // FIXME: should be 32-bit / 16-bit depending on bit-depth
} WmallDecodeCtx;
#define dprintf(pctx, ...) av_log(pctx, AV_LOG_DEBUG, __VA_ARGS__)
-//static int num_logged_tiles;
+static int num_logged_tiles = 0;
+static int num_logged_subframes = 0;
+static int num_lms_update_call = 0;
/**
*@brief helper function to print the most important members of the context
PRINT("num channels", s->num_channels);
}
+static void dump_int_buffer(uint8_t *buffer, int size, int length, int delimiter)
+{
+ int i;
+
+ for (i=0 ; i<length ; i++) {
+ if (!(i%delimiter))
+ av_log(0, 0, "\n[%d] ", i);
+ av_log(0, 0, "%d, ", *(int16_t *)(buffer + i * size));
+ }
+ av_log(0, 0, "\n");
+}
+
/**
*@brief Uninitialize the decoder and free all resources.
*@param avctx codec context
int log2_max_num_subframes;
int num_possible_block_sizes;
- //num_logged_tiles = 0;
s->avctx = avctx;
dsputil_init(&s->dsp, avctx);
init_put_bits(&s->pb, s->frame_data, MAX_FRAMESIZE);
- avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
-
if (avctx->extradata_size >= 18) {
s->decode_flags = AV_RL16(edata_ptr+14);
channel_mask = AV_RL32(edata_ptr+2);
s->bits_per_sample = AV_RL16(edata_ptr);
+ if (s->bits_per_sample == 16)
+ avctx->sample_fmt = AV_SAMPLE_FMT_S16;
+ else if (s->bits_per_sample == 24)
+ avctx->sample_fmt = AV_SAMPLE_FMT_S32;
+ else {
+ av_log(avctx, AV_LOG_ERROR, "Unknown bit-depth: %d\n",
+ s->bits_per_sample);
+ return AVERROR_INVALIDDATA;
+ }
/** dump the extradata */
for (i = 0; i < avctx->extradata_size; i++)
dprintf(avctx, "[%x] ", avctx->extradata[i]);
for (c = 0; c < s->num_channels; c++) {
if (num_samples[c] == min_channel_len) {
if (fixed_channel_layout || channels_for_cur_subframe == 1 ||
- (min_channel_len == s->samples_per_frame - s->min_samples_per_subframe)) {
+ (min_channel_len == s->samples_per_frame - s->min_samples_per_subframe)) {
contains_subframe[c] = 1;
- }
+ }
else {
contains_subframe[c] = get_bits1(&s->gb);
- }
+ }
} else
contains_subframe[c] = 0;
}
if (num_samples[c] > s->samples_per_frame) {
av_log(s->avctx, AV_LOG_ERROR, "broken frame: "
"channel len(%d) > samples_per_frame(%d)\n",
- num_samples[c], s->samples_per_frame);
+ num_samples[c], s->samples_per_frame);
return AVERROR_INVALIDDATA;
}
} else if (num_samples[c] <= min_channel_len) {
{
unsigned int iLog2 = 0;
while ((i >> iLog2) > 1)
- iLog2++;
+ iLog2++;
return iLog2;
}
s->acfilter_scaling = get_bits(&s->gb, 4);
for(i = 0; i < s->acfilter_order; i++) {
- s->acfilter_coeffs[i] = get_bits(&s->gb, s->acfilter_scaling) + 1;
+ s->acfilter_coeffs[i] = get_bits(&s->gb, s->acfilter_scaling) + 1;
}
}
s->mclms_order = (get_bits(&s->gb, 4) + 1) * 2;
s->mclms_scaling = get_bits(&s->gb, 4);
if(get_bits1(&s->gb)) {
- // mclms_send_coef
- int i;
- int send_coef_bits;
- int cbits = av_log2(s->mclms_scaling + 1);
- assert(cbits == my_log2(s->mclms_scaling + 1));
- if(1 << cbits < s->mclms_scaling + 1)
- cbits++;
-
- send_coef_bits = (cbits ? get_bits(&s->gb, cbits) : 0) + 2;
-
- for(i = 0; i < s->mclms_order * s->num_channels * s->num_channels; i++) {
- s->mclms_coeffs[i] = get_bits(&s->gb, send_coef_bits);
- }
-
- for(i = 0; i < s->num_channels; i++) {
- int c;
- for(c = 0; c < i; c++) {
- s->mclms_coeffs_cur[i * s->num_channels + c] = get_bits(&s->gb, send_coef_bits);
- }
- }
+ // mclms_send_coef
+ int i;
+ int send_coef_bits;
+ int cbits = av_log2(s->mclms_scaling + 1);
+ assert(cbits == my_log2(s->mclms_scaling + 1));
+ if(1 << cbits < s->mclms_scaling + 1)
+ cbits++;
+
+ send_coef_bits = (cbits ? get_bits(&s->gb, cbits) : 0) + 2;
+
+ for(i = 0; i < s->mclms_order * s->num_channels * s->num_channels; i++) {
+ s->mclms_coeffs[i] = get_bits(&s->gb, send_coef_bits);
+ }
+
+ for(i = 0; i < s->num_channels; i++) {
+ int c;
+ for(c = 0; c < i; c++) {
+ s->mclms_coeffs_cur[i * s->num_channels + c] = get_bits(&s->gb, send_coef_bits);
+ }
+ }
}
}
int cdlms_send_coef = get_bits1(&s->gb);
for(c = 0; c < s->num_channels; c++) {
- s->cdlms_ttl[c] = get_bits(&s->gb, 3) + 1;
- for(i = 0; i < s->cdlms_ttl[c]; i++) {
- s->cdlms[c][i].order = (get_bits(&s->gb, 7) + 1) * 8;
- }
-
- for(i = 0; i < s->cdlms_ttl[c]; i++) {
- s->cdlms[c][i].scaling = get_bits(&s->gb, 4);
- }
-
- if(cdlms_send_coef) {
- for(i = 0; i < s->cdlms_ttl[c]; i++) {
- int cbits, shift_l, shift_r, j;
- cbits = av_log2(s->cdlms[c][i].order);
- if(1 << cbits < s->cdlms[c][i].order)
- cbits++;
- s->cdlms[c][i].coefsend = get_bits(&s->gb, cbits) + 1;
-
- cbits = av_log2(s->cdlms[c][i].scaling + 1);
- if(1 << cbits < s->cdlms[c][i].scaling + 1)
- cbits++;
-
- s->cdlms[c][i].bitsend = get_bits(&s->gb, cbits) + 2;
- shift_l = 32 - s->cdlms[c][i].bitsend;
- shift_r = 32 - 2 - s->cdlms[c][i].scaling;
- for(j = 0; j < s->cdlms[c][i].coefsend; j++) {
- s->cdlms[c][i].coefs[j] =
- (get_bits(&s->gb, s->cdlms[c][i].bitsend) << shift_l) >> shift_r;
- }
- }
- }
+ s->cdlms_ttl[c] = get_bits(&s->gb, 3) + 1;
+ for(i = 0; i < s->cdlms_ttl[c]; i++) {
+ s->cdlms[c][i].order = (get_bits(&s->gb, 7) + 1) * 8;
+ }
+
+ for(i = 0; i < s->cdlms_ttl[c]; i++) {
+ s->cdlms[c][i].scaling = get_bits(&s->gb, 4);
+ }
+
+ if(cdlms_send_coef) {
+ for(i = 0; i < s->cdlms_ttl[c]; i++) {
+ int cbits, shift_l, shift_r, j;
+ cbits = av_log2(s->cdlms[c][i].order);
+ if(1 << cbits < s->cdlms[c][i].order)
+ cbits++;
+ s->cdlms[c][i].coefsend = get_bits(&s->gb, cbits) + 1;
+
+ cbits = av_log2(s->cdlms[c][i].scaling + 1);
+ if(1 << cbits < s->cdlms[c][i].scaling + 1)
+ cbits++;
+
+ s->cdlms[c][i].bitsend = get_bits(&s->gb, cbits) + 2;
+ shift_l = 32 - s->cdlms[c][i].bitsend;
+ shift_r = 32 - 2 - s->cdlms[c][i].scaling;
+ for(j = 0; j < s->cdlms[c][i].coefsend; j++) {
+ s->cdlms[c][i].coefs[j] =
+ (get_bits(&s->gb, s->cdlms[c][i].bitsend) << shift_l) >> shift_r;
+ }
+ }
+ }
}
}
int i = 0;
unsigned int ave_mean;
s->transient[ch] = get_bits1(&s->gb);
- if(s->transient[ch])
- s->transient_pos[ch] = get_bits(&s->gb, av_log2(tile_size));
-
+ if(s->transient[ch]) {
+ s->transient_pos[ch] = get_bits(&s->gb, av_log2(tile_size));
+ if (s->transient_pos[ch])
+ s->transient[ch] = 0;
+ s->channel[ch].transient_counter =
+ FFMAX(s->channel[ch].transient_counter, s->samples_per_frame / 2);
+ } else if (s->channel[ch].transient_counter)
+ s->transient[ch] = 1;
+
if(s->seekable_tile) {
- ave_mean = get_bits(&s->gb, s->bits_per_sample);
- s->ave_sum[ch] = ave_mean << (s->movave_scaling + 1);
-// s->ave_sum[ch] *= 2;
+ ave_mean = get_bits(&s->gb, s->bits_per_sample);
+ s->ave_sum[ch] = ave_mean << (s->movave_scaling + 1);
+// s->ave_sum[ch] *= 2;
}
if(s->seekable_tile) {
- if(s->do_inter_ch_decorr)
- s->channel_residues[ch][0] = get_sbits(&s->gb, s->bits_per_sample + 1);
- else
- s->channel_residues[ch][0] = get_sbits(&s->gb, s->bits_per_sample);
- i++;
+ if(s->do_inter_ch_decorr)
+ s->channel_residues[ch][0] = get_sbits(&s->gb, s->bits_per_sample + 1);
+ else
+ s->channel_residues[ch][0] = get_sbits(&s->gb, s->bits_per_sample);
+ i++;
}
+ //av_log(0, 0, "%8d: ", num_logged_tiles++);
for(; i < tile_size; i++) {
- int quo = 0, rem, rem_bits, residue;
- while(get_bits1(&s->gb))
- quo++;
- if(quo >= 32)
- quo += get_bits_long(&s->gb, get_bits(&s->gb, 5) + 1);
-
- ave_mean = (s->ave_sum[ch] + (1 << s->movave_scaling)) >> (s->movave_scaling + 1);
- rem_bits = av_ceil_log2(ave_mean);
- rem = rem_bits ? get_bits(&s->gb, rem_bits) : 0;
- residue = (quo << rem_bits) + rem;
-
- s->ave_sum[ch] = residue + s->ave_sum[ch] - (s->ave_sum[ch] >> s->movave_scaling);
-
- if(residue & 1)
- residue = -(residue >> 1) - 1;
- else
- residue = residue >> 1;
- s->channel_residues[ch][i] = residue;
-
- //if (num_logged_tiles < 1)
- av_log(0, 0, "%4d ", residue);
-// dprintf(s->avctx, "%5d: %5d %10d %12d %12d %5d %-16d %04x\n",i, quo, ave_mean, s->ave_sum[ch], rem, rem_bits, s->channel_residues[ch][i], show_bits(&s->gb, 16));
+ int quo = 0, rem, rem_bits, residue;
+ while(get_bits1(&s->gb))
+ quo++;
+ if(quo >= 32)
+ quo += get_bits_long(&s->gb, get_bits(&s->gb, 5) + 1);
+
+ ave_mean = (s->ave_sum[ch] + (1 << s->movave_scaling)) >> (s->movave_scaling + 1);
+ rem_bits = av_ceil_log2(ave_mean);
+ rem = rem_bits ? get_bits(&s->gb, rem_bits) : 0;
+ residue = (quo << rem_bits) + rem;
+
+ s->ave_sum[ch] = residue + s->ave_sum[ch] - (s->ave_sum[ch] >> s->movave_scaling);
+
+ if(residue & 1)
+ residue = -(residue >> 1) - 1;
+ else
+ residue = residue >> 1;
+ s->channel_residues[ch][i] = residue;
}
- av_log(0, 0, "\n");
- //num_logged_tiles++;
+ //dump_int_buffer(s->channel_residues[ch], 4, tile_size, 16);
return 0;
s->lpc_intbits = get_bits(&s->gb, 3) + 1;
cbits = s->lpc_scaling + s->lpc_intbits;
for(ch = 0; ch < s->num_channels; ch++) {
- for(i = 0; i < s->lpc_order; i++) {
- s->lpc_coefs[ch][i] = get_sbits(&s->gb, cbits);
- }
+ for(i = 0; i < s->lpc_order; i++) {
+ s->lpc_coefs[ch][i] = get_sbits(&s->gb, cbits);
+ }
}
}
{
int ich, ilms;
- memset(s->acfilter_coeffs, 0, 16 * sizeof(int));
- memset(s->lpc_coefs , 0, 40 * 2 * sizeof(int));
+ memset(s->acfilter_coeffs , 0, 16 * sizeof(int));
+ memset(s->acfilter_prevvalues, 0, 16 * 2 * sizeof(int)); // may be wrong
+ memset(s->lpc_coefs , 0, 40 * 2 * sizeof(int));
memset(s->mclms_coeffs , 0, 128 * sizeof(int16_t));
memset(s->mclms_coeffs_cur, 0, 4 * sizeof(int16_t));
for (ich = 0; ich < s->num_channels; ich++) {
for (ilms = 0; ilms < s->cdlms_ttl[ich]; ilms++) {
memset(s->cdlms[ich][ilms].coefs , 0, 256 * sizeof(int16_t));
- memset(s->cdlms[ich][ilms].lms_prevvalues, 0, 512 * sizeof(int));
+ memset(s->cdlms[ich][ilms].lms_prevvalues, 0, 512 * sizeof(int16_t));
memset(s->cdlms[ich][ilms].lms_updates , 0, 512 * sizeof(int16_t));
}
s->ave_sum[ich] = 0;
}
}
+/**
+ *@brief Resets filter parameters and transient area at new seekable tile
+ */
static void reset_codec(WmallDecodeCtx *s)
{
int ich, ilms;
s->mclms_recent = s->mclms_order * s->num_channels;
- for (ich = 0; ich < s->num_channels; ich++)
+ for (ich = 0; ich < s->num_channels; ich++) {
for (ilms = 0; ilms < s->cdlms_ttl[ich]; ilms++)
s->cdlms[ich][ilms].recent = s->cdlms[ich][ilms].order;
+ /* first sample of a seekable subframe is considered as the starting of
+ a transient area which is samples_per_frame samples long */
+ s->channel[ich].transient_counter = s->samples_per_frame;
+ s->transient[ich] = 1;
+ s->transient_pos[ich] = 0;
+ }
}
+static void mclms_update(WmallDecodeCtx *s, int icoef, int *pred)
+{
+ int i, j, ich;
+ int pred_error;
+ int order = s->mclms_order;
+ int num_channels = s->num_channels;
+ int range = 1 << (s->bits_per_sample - 1);
+ int bps = s->bits_per_sample > 16 ? 4 : 2; // bytes per sample
+
+ for (ich = 0; ich < num_channels; ich++) {
+ pred_error = s->channel_residues[ich][icoef] - pred[ich];
+ if (pred_error > 0) {
+ for (i = 0; i < order * num_channels; i++)
+ s->mclms_coeffs[i + ich * order * num_channels] +=
+ s->mclms_updates[s->mclms_recent + i];
+ for (j = 0; j < ich; j++) {
+ if (s->channel_residues[j][icoef] > 0)
+ s->mclms_coeffs_cur[ich * num_channels + j] += 1;
+ else if (s->channel_residues[j][icoef] < 0)
+ s->mclms_coeffs_cur[ich * num_channels + j] -= 1;
+ }
+ } else if (pred_error < 0) {
+ for (i = 0; i < order * num_channels; i++)
+ s->mclms_coeffs[i + ich * order * num_channels] -=
+ s->mclms_updates[s->mclms_recent + i];
+ for (j = 0; j < ich; j++) {
+ if (s->channel_residues[j][icoef] > 0)
+ s->mclms_coeffs_cur[ich * num_channels + j] -= 1;
+ else if (s->channel_residues[j][icoef] < 0)
+ s->mclms_coeffs_cur[ich * num_channels + j] += 1;
+ }
+ }
+ }
+
+ for (ich = num_channels - 1; ich >= 0; ich--) {
+ s->mclms_recent--;
+ s->mclms_prevvalues[s->mclms_recent] = s->channel_residues[ich][icoef];
+ if (s->channel_residues[ich][icoef] > range - 1)
+ s->mclms_prevvalues[s->mclms_recent] = range - 1;
+ else if (s->channel_residues[ich][icoef] < -range)
+ s->mclms_prevvalues[s->mclms_recent] = -range;
+
+ s->mclms_updates[s->mclms_recent] = 0;
+ if (s->channel_residues[ich][icoef] > 0)
+ s->mclms_updates[s->mclms_recent] = 1;
+ else if (s->channel_residues[ich][icoef] < 0)
+ s->mclms_updates[s->mclms_recent] = -1;
+ }
+
+ if (s->mclms_recent == 0) {
+ memcpy(&s->mclms_prevvalues[order * num_channels],
+ s->mclms_prevvalues,
+ bps * order * num_channels);
+ memcpy(&s->mclms_updates[order * num_channels],
+ s->mclms_updates,
+ bps * order * num_channels);
+ s->mclms_recent = num_channels * order;
+ }
+}
+
+static void mclms_predict(WmallDecodeCtx *s, int icoef, int *pred)
+{
+ int ich, i;
+ int order = s->mclms_order;
+ int num_channels = s->num_channels;
+
+ for (ich = 0; ich < num_channels; ich++) {
+ if (!s->is_channel_coded[ich])
+ continue;
+ pred[ich] = 0;
+ for (i = 0; i < order * num_channels; i++)
+ pred[ich] += s->mclms_prevvalues[i + s->mclms_recent] *
+ s->mclms_coeffs[i + order * num_channels * ich];
+ for (i = 0; i < ich; i++)
+ pred[ich] += s->channel_residues[i][icoef] *
+ s->mclms_coeffs_cur[i + num_channels * ich];
+ pred[ich] += 1 << s->mclms_scaling - 1;
+ pred[ich] >>= s->mclms_scaling;
+ s->channel_residues[ich][icoef] += pred[ich];
+ }
+}
+
+static void revert_mclms(WmallDecodeCtx *s, int tile_size)
+{
+ int icoef, pred[s->num_channels];
+ for (icoef = 0; icoef < tile_size; icoef++) {
+ mclms_predict(s, icoef, pred);
+ mclms_update(s, icoef, pred);
+ }
+}
+
static int lms_predict(WmallDecodeCtx *s, int ich, int ilms)
{
- int32_t pred, icoef;
+ int pred = 0;
+ int icoef;
int recent = s->cdlms[ich][ilms].recent;
for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
pred += s->cdlms[ich][ilms].coefs[icoef] *
s->cdlms[ich][ilms].lms_prevvalues[icoef + recent];
- pred += (1 << (s->cdlms[ich][ilms].scaling - 1));
+ //pred += (1 << (s->cdlms[ich][ilms].scaling - 1));
/* XXX: Table 29 has:
iPred >= cdlms[iCh][ilms].scaling;
seems to me like a missing > */
- pred >>= s->cdlms[ich][ilms].scaling;
+ //pred >>= s->cdlms[ich][ilms].scaling;
return pred;
}
-static void lms_update(WmallDecodeCtx *s, int ich, int ilms, int32_t input, int32_t pred)
+static void lms_update(WmallDecodeCtx *s, int ich, int ilms, int input, int residue)
{
int icoef;
int recent = s->cdlms[ich][ilms].recent;
- int range = 1 << (s->bits_per_sample - 1);
+ int range = 1 << s->bits_per_sample - 1;
int bps = s->bits_per_sample > 16 ? 4 : 2; // bytes per sample
- if (input > pred) {
+ if (residue < 0) {
for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
- s->cdlms[ich][ilms].coefs[icoef] +=
+ s->cdlms[ich][ilms].coefs[icoef] -=
s->cdlms[ich][ilms].lms_updates[icoef + recent];
- } else {
+ } else if (residue > 0) {
for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
- s->cdlms[ich][ilms].coefs[icoef] -=
- s->cdlms[ich][ilms].lms_updates[icoef]; // XXX: [icoef + recent] ?
+ s->cdlms[ich][ilms].coefs[icoef] +=
+ s->cdlms[ich][ilms].lms_updates[icoef + recent]; /* spec mistakenly
+ dropped the recent */
}
- s->cdlms[ich][ilms].recent--;
- s->cdlms[ich][ilms].lms_prevvalues[recent] = av_clip(input, -range, range - 1);
- if (input > pred)
- s->cdlms[ich][ilms].lms_updates[recent] = s->update_speed[ich];
- else if (input < pred)
+ if (recent)
+ recent--;
+ else {
+ /* XXX: This memcpy()s will probably fail if a fixed 32-bit buffer is used.
+ follow kshishkov's suggestion of using a union. */
+ memcpy(&s->cdlms[ich][ilms].lms_prevvalues[s->cdlms[ich][ilms].order],
+ s->cdlms[ich][ilms].lms_prevvalues,
+ bps * s->cdlms[ich][ilms].order);
+ memcpy(&s->cdlms[ich][ilms].lms_updates[s->cdlms[ich][ilms].order],
+ s->cdlms[ich][ilms].lms_updates,
+ bps * s->cdlms[ich][ilms].order);
+ recent = s->cdlms[ich][ilms].order - 1;
+ }
+
+ s->cdlms[ich][ilms].lms_prevvalues[recent] = av_clip(input, -range, range - 1);
+ if (!input)
+ s->cdlms[ich][ilms].lms_updates[recent] = 0;
+ else if (input < 0)
s->cdlms[ich][ilms].lms_updates[recent] = -s->update_speed[ich];
+ else
+ s->cdlms[ich][ilms].lms_updates[recent] = s->update_speed[ich];
/* XXX: spec says:
cdlms[iCh][ilms].updates[iRecent + cdlms[iCh][ilms].order >> 4] >>= 2;
seperate buffers? Here I've assumed that the two are same which makes
more sense to me.
*/
- s->cdlms[ich][ilms].lms_updates[recent + s->cdlms[ich][ilms].order >> 4] >>= 2;
- s->cdlms[ich][ilms].lms_updates[recent + s->cdlms[ich][ilms].order >> 3] >>= 1;
- /* XXX: recent + (s->cdlms[ich][ilms].order >> 4) ? */
+ s->cdlms[ich][ilms].lms_updates[recent + (s->cdlms[ich][ilms].order >> 4)] >>= 2;
+ s->cdlms[ich][ilms].lms_updates[recent + (s->cdlms[ich][ilms].order >> 3)] >>= 1;
+ s->cdlms[ich][ilms].recent = recent;
+}
- if (s->cdlms[ich][ilms].recent == 0) {
- /* XXX: This memcpy()s will probably fail if a fixed 32-bit buffer is used.
- follow kshishkov's suggestion of using a union. */
- memcpy(s->cdlms[ich][ilms].lms_prevvalues + s->cdlms[ich][ilms].order,
- s->cdlms[ich][ilms].lms_prevvalues,
- bps * s->cdlms[ich][ilms].order);
- memcpy(s->cdlms[ich][ilms].lms_updates + s->cdlms[ich][ilms].order,
- s->cdlms[ich][ilms].lms_updates,
- bps * s->cdlms[ich][ilms].order);
- s->cdlms[ich][ilms].recent = s->cdlms[ich][ilms].order;
+static void use_high_update_speed(WmallDecodeCtx *s, int ich)
+{
+ int ilms, recent, icoef;
+ for (ilms = s->cdlms_ttl[ich] - 1; ilms >= 0; ilms--) {
+ recent = s->cdlms[ich][ilms].recent;
+ if (s->update_speed[ich] == 16)
+ continue;
+ if (s->bV3RTM) {
+ for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+ s->cdlms[ich][ilms].lms_updates[icoef + recent] *= 2;
+ } else {
+ for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+ s->cdlms[ich][ilms].lms_updates[icoef] *= 2;
+ }
+ }
+ s->update_speed[ich] = 16;
+}
+
+static void use_normal_update_speed(WmallDecodeCtx *s, int ich)
+{
+ int ilms, recent, icoef;
+ for (ilms = s->cdlms_ttl[ich] - 1; ilms >= 0; ilms--) {
+ recent = s->cdlms[ich][ilms].recent;
+ if (s->update_speed[ich] == 8)
+ continue;
+ if (s->bV3RTM) {
+ for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+ s->cdlms[ich][ilms].lms_updates[icoef + recent] /= 2;
+ } else {
+ for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+ s->cdlms[ich][ilms].lms_updates[icoef] /= 2;
+ }
+ }
+ s->update_speed[ich] = 8;
+}
+
+static void revert_cdlms(WmallDecodeCtx *s, int ch, int coef_begin, int coef_end)
+{
+ int icoef;
+ int pred;
+ int ilms, num_lms;
+ int residue, input;
+
+ num_lms = s->cdlms_ttl[ch];
+ for (ilms = num_lms - 1; ilms >= 0; ilms--) {
+ //s->cdlms[ch][ilms].recent = s->cdlms[ch][ilms].order;
+ for (icoef = coef_begin; icoef < coef_end; icoef++) {
+ pred = 1 << (s->cdlms[ch][ilms].scaling - 1);
+ residue = s->channel_residues[ch][icoef];
+ pred += lms_predict(s, ch, ilms);
+ input = residue + (pred >> s->cdlms[ch][ilms].scaling);
+ lms_update(s, ch, ilms, input, residue);
+ s->channel_residues[ch][icoef] = input;
+ }
+ }
+}
+
+static void revert_inter_ch_decorr(WmallDecodeCtx *s, int tile_size)
+{
+ int icoef;
+ if (s->num_channels != 2)
+ return;
+ else {
+ for (icoef = 0; icoef < tile_size; icoef++) {
+ s->channel_residues[0][icoef] -= s->channel_residues[1][icoef] >> 1;
+ s->channel_residues[1][icoef] += s->channel_residues[0][icoef];
+ }
+ }
+}
+
+static void revert_acfilter(WmallDecodeCtx *s, int tile_size)
+{
+ int ich, icoef;
+ int pred;
+ int i, j;
+ int64_t *filter_coeffs = s->acfilter_coeffs;
+ int scaling = s->acfilter_scaling;
+ int order = s->acfilter_order;
+
+ for (ich = 0; ich < s->num_channels; ich++) {
+ int *prevvalues = s->acfilter_prevvalues[ich];
+ for (i = 0; i < order; i++) {
+ pred = 0;
+ for (j = 0; j < order; j++) {
+ if (i <= j)
+ pred += filter_coeffs[j] * prevvalues[j - i];
+ else
+ pred += s->channel_residues[ich][i - j - 1] * filter_coeffs[j];
+ }
+ pred >>= scaling;
+ s->channel_residues[ich][i] += pred;
+ }
+ for (i = order; i < tile_size; i++) {
+ pred = 0;
+ for (j = 0; j < order; j++)
+ pred += s->channel_residues[ich][i - j - 1] * filter_coeffs[j];
+ pred >>= scaling;
+ s->channel_residues[ich][i] += pred;
+ }
+ for (j = 0; j < order; j++)
+ prevvalues[j] = s->channel_residues[ich][tile_size - j - 1];
}
}
{
int offset = s->samples_per_frame;
int subframe_len = s->samples_per_frame;
- int i;
+ int i, j;
int total_samples = s->samples_per_frame * s->num_channels;
int rawpcm_tile;
int padding_zeroes;
if(s->seekable_tile) {
clear_codec_buffers(s);
- s->do_arith_coding = get_bits1(&s->gb);
- if(s->do_arith_coding) {
- dprintf(s->avctx, "do_arith_coding == 1");
- abort();
- }
- s->do_ac_filter = get_bits1(&s->gb);
- s->do_inter_ch_decorr = get_bits1(&s->gb);
- s->do_mclms = get_bits1(&s->gb);
-
- if(s->do_ac_filter)
- decode_ac_filter(s);
-
- if(s->do_mclms)
- decode_mclms(s);
-
- decode_cdlms(s);
- s->movave_scaling = get_bits(&s->gb, 3);
- s->quant_stepsize = get_bits(&s->gb, 8) + 1;
-
- reset_codec(s);
+ s->do_arith_coding = get_bits1(&s->gb);
+ if(s->do_arith_coding) {
+ dprintf(s->avctx, "do_arith_coding == 1");
+ abort();
+ }
+ s->do_ac_filter = get_bits1(&s->gb);
+ s->do_inter_ch_decorr = get_bits1(&s->gb);
+ s->do_mclms = get_bits1(&s->gb);
+
+ if(s->do_ac_filter)
+ decode_ac_filter(s);
+
+ if(s->do_mclms)
+ decode_mclms(s);
+
+ decode_cdlms(s);
+ s->movave_scaling = get_bits(&s->gb, 3);
+ s->quant_stepsize = get_bits(&s->gb, 8) + 1;
+
+ reset_codec(s);
}
rawpcm_tile = get_bits1(&s->gb);
for(i = 0; i < s->num_channels; i++) {
- s->is_channel_coded[i] = 1;
+ s->is_channel_coded[i] = 1;
}
if(!rawpcm_tile) {
- for(i = 0; i < s->num_channels; i++) {
- s->is_channel_coded[i] = get_bits1(&s->gb);
- }
-
- if(s->bV3RTM) {
- // LPC
- s->do_lpc = get_bits1(&s->gb);
- if(s->do_lpc) {
- decode_lpc(s);
- }
- } else {
- s->do_lpc = 0;
- }
+ for(i = 0; i < s->num_channels; i++) {
+ s->is_channel_coded[i] = get_bits1(&s->gb);
+ }
+
+ if(s->bV3RTM) {
+ // LPC
+ s->do_lpc = get_bits1(&s->gb);
+ if(s->do_lpc) {
+ decode_lpc(s);
+ }
+ } else {
+ s->do_lpc = 0;
+ }
}
if(get_bits1(&s->gb)) {
- padding_zeroes = get_bits(&s->gb, 5);
+ padding_zeroes = get_bits(&s->gb, 5);
} else {
- padding_zeroes = 0;
+ padding_zeroes = 0;
}
if(rawpcm_tile) {
-
- int bits = s->bits_per_sample - padding_zeroes;
- int j;
- dprintf(s->avctx, "RAWPCM %d bits per sample. total %d bits, remain=%d\n", bits,
- bits * s->num_channels * subframe_len, get_bits_count(&s->gb));
- for(i = 0; i < s->num_channels; i++) {
- for(j = 0; j < subframe_len; j++) {
- s->channel_coeffs[i][j] = get_sbits(&s->gb, bits);
-// dprintf(s->avctx, "PCM[%d][%d] = 0x%04x\n", i, j, s->channel_coeffs[i][j]);
- }
- }
+
+ int bits = s->bits_per_sample - padding_zeroes;
+ dprintf(s->avctx, "RAWPCM %d bits per sample. total %d bits, remain=%d\n", bits,
+ bits * s->num_channels * subframe_len, get_bits_count(&s->gb));
+ for(i = 0; i < s->num_channels; i++) {
+ for(j = 0; j < subframe_len; j++) {
+ s->channel_coeffs[i][j] = get_sbits(&s->gb, bits);
+// dprintf(s->avctx, "PCM[%d][%d] = 0x%04x\n", i, j, s->channel_coeffs[i][j]);
+ }
+ }
} else {
- for(i = 0; i < s->num_channels; i++)
- if(s->is_channel_coded[i])
- decode_channel_residues(s, i, subframe_len);
+ for(i = 0; i < s->num_channels; i++)
+ if(s->is_channel_coded[i]) {
+ decode_channel_residues(s, i, subframe_len);
+ if (s->seekable_tile)
+ use_high_update_speed(s, i);
+ else
+ use_normal_update_speed(s, i);
+ revert_cdlms(s, i, 0, subframe_len);
+ }
}
+ if (s->do_mclms)
+ revert_mclms(s, subframe_len);
+ if (s->do_inter_ch_decorr)
+ revert_inter_ch_decorr(s, subframe_len);
+ if(s->do_ac_filter)
+ revert_acfilter(s, subframe_len);
+
+ /* Dequantize */
+ if (s->quant_stepsize != 1)
+ for (i = 0; i < s->num_channels; i++)
+ for (j = 0; j < subframe_len; j++)
+ s->channel_residues[i][j] *= s->quant_stepsize;
+
+ // Write to proper output buffer depending on bit-depth
+ for (i = 0; i < subframe_len; i++)
+ for (j = 0; j < s->num_channels; j++) {
+ if (s->bits_per_sample == 16)
+ *s->samples_16++ = (int16_t) s->channel_residues[j][i];
+ else
+ *s->samples_32++ = s->channel_residues[j][i];
+ }
/** handled one subframe */
}
++s->channel[c].cur_subframe;
}
+ num_logged_subframes++;
return 0;
}
int more_frames = 0;
int len = 0;
int i;
+ int buffer_len;
/** check for potential output buffer overflow */
- if (s->num_channels * s->samples_per_frame > s->samples_end - s->samples) {
+ if (s->bits_per_sample == 16)
+ buffer_len = s->samples_16_end - s->samples_16;
+ else
+ buffer_len = s->samples_32_end - s->samples_32;
+ if (s->num_channels * s->samples_per_frame > buffer_len) {
/** return an error if no frame could be decoded at all */
av_log(s->avctx, AV_LOG_ERROR,
"not enough space for the output samples\n");
if (s->skip_frame) {
s->skip_frame = 0;
- } else
- s->samples += s->num_channels * s->samples_per_frame;
+ }
if (s->len_prefix) {
if (len != (get_bits_count(gb) - s->frame_offset) + 2) {
} else {
/*
while (get_bits_count(gb) < s->num_saved_bits && get_bits1(gb) == 0) {
- dprintf(s->avctx, "skip1\n");
+ dprintf(s->avctx, "skip1\n");
}
*/
}
int num_bits_prev_frame;
int packet_sequence_number;
- s->samples = data;
- s->samples_end = (float*)((int8_t*)data + *data_size);
+ if (s->bits_per_sample == 16) {
+ s->samples_16 = (int16_t *) data;
+ s->samples_16_end = (int16_t *) ((int8_t*)data + *data_size);
+ } else {
+ s->samples_32 = (int *) data;
+ s->samples_32_end = (int *) ((int8_t*)data + *data_size);
+ }
*data_size = 0;
if (s->packet_done || s->packet_loss) {
+ int seekable_frame_in_packet, spliced_packet;
s->packet_done = 0;
/** sanity check for the buffer length */
/** parse packet header */
init_get_bits(gb, buf, s->buf_bit_size);
packet_sequence_number = get_bits(gb, 4);
- int seekable_frame_in_packet = get_bits1(gb);
- int spliced_packet = get_bits1(gb);
+ seekable_frame_in_packet = get_bits1(gb);
+ spliced_packet = get_bits1(gb);
/** get number of bits that need to be added to the previous frame */
num_bits_prev_frame = get_bits(gb, s->log2_frame_size);
/** decode the cross packet frame if it is valid */
if (!s->packet_loss)
- decode_frame(s);
+ decode_frame(s);
} else if (s->num_saved_bits - s->frame_offset) {
dprintf(avctx, "ignoring %x previously saved bits\n",
s->num_saved_bits - s->frame_offset);
s->buf_bit_size = (avpkt->size - s->next_packet_start) << 3;
init_get_bits(gb, avpkt->data, s->buf_bit_size);
skip_bits(gb, s->packet_offset);
-
+
if (s->len_prefix && remaining_bits(s, gb) > s->log2_frame_size &&
(frame_size = show_bits(gb, s->log2_frame_size)) &&
frame_size <= remaining_bits(s, gb)) {
s->packet_done = !decode_frame(s);
} else {
s->packet_done = 1;
- }
+ }
}
if (s->packet_done && !s->packet_loss &&
save_bits(s, gb, remaining_bits(s, gb), 0);
}
- *data_size = 0; // (int8_t *)s->samples - (int8_t *)data;
+ if (s->bits_per_sample == 16)
+ *data_size = (int8_t *)s->samples_16 - (int8_t *)data;
+ else
+ *data_size = (int8_t *)s->samples_32 - (int8_t *)data;
s->packet_offset = get_bits_count(gb) & 7;
return (s->packet_loss) ? AVERROR_INVALIDDATA : get_bits_count(gb) >> 3;
NULL,
decode_end,
decode_packet,
- .capabilities = CODEC_CAP_SUBFRAMES,
+ .capabilities = CODEC_CAP_SUBFRAMES | CODEC_CAP_EXPERIMENTAL,
.flush= flush,
.long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 9 Lossless"),
};
projects / ffmpeg / blobdiff