* N (code in SoC repo) Long Term Prediction
* Y intensity stereo
* Y channel coupling
- * N frequency domain prediction
+ * Y frequency domain prediction
* Y Perceptual Noise Substitution
* Y Mid/Side stereo
* N Scalable Inverse AAC Quantization
#include "avcodec.h"
+#include "internal.h"
#include "bitstream.h"
#include "dsputil.h"
+#include "lpc.h"
#include "aac.h"
#include "aactab.h"
#include <math.h>
#include <string.h>
-#ifndef CONFIG_HARDCODED_TABLES
- static float ff_aac_ivquant_tab[IVQUANT_SIZE];
- static float ff_aac_pow2sf_tab[316];
-#endif /* CONFIG_HARDCODED_TABLES */
-
static VLC vlc_scalefactors;
static VLC vlc_spectral[11];
* Set up default 1:1 output mapping.
*
* For a 5.1 stream the output order will be:
- * [ Front Left ] [ Front Right ] [ Center ] [ LFE ] [ Surround Left ] [ Surround Right ]
+ * [ Center ] [ Front Left ] [ Front Right ] [ LFE ] [ Surround Left ] [ Surround Right ]
*/
for(i = 0; i < MAX_ELEM_ID; i++) {
*/
static int decode_pce(AACContext * ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
GetBitContext * gb) {
- int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc;
+ int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc, sampling_index;
skip_bits(gb, 2); // object_type
- ac->m4ac.sampling_index = get_bits(gb, 4);
- if(ac->m4ac.sampling_index > 11) {
+ sampling_index = get_bits(gb, 4);
+ if(sampling_index > 11) {
av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
return -1;
}
+ ac->m4ac.sampling_index = sampling_index;
ac->m4ac.sample_rate = ff_mpeg4audio_sample_rates[ac->m4ac.sampling_index];
num_front = get_bits(gb, 4);
num_side = get_bits(gb, 4);
int extension_flag, ret;
if(get_bits1(gb)) { // frameLengthFlag
- av_log_missing_feature(ac->avccontext, "960/120 MDCT window is", 1);
+ ff_log_missing_feature(ac->avccontext, "960/120 MDCT window is", 1);
return -1;
}
skip_bits_long(&gb, i);
switch (ac->m4ac.object_type) {
+ case AOT_AAC_MAIN:
case AOT_AAC_LC:
if (decode_ga_specific_config(ac, &gb, ac->m4ac.chan_config))
return -1;
return previous_val * 1664525 + 1013904223;
}
+static void reset_predict_state(PredictorState * ps) {
+ ps->r0 = 0.0f;
+ ps->r1 = 0.0f;
+ ps->cor0 = 0.0f;
+ ps->cor1 = 0.0f;
+ ps->var0 = 1.0f;
+ ps->var1 = 1.0f;
+}
+
+static void reset_all_predictors(PredictorState * ps) {
+ int i;
+ for (i = 0; i < MAX_PREDICTORS; i++)
+ reset_predict_state(&ps[i]);
+}
+
+static void reset_predictor_group(PredictorState * ps, int group_num) {
+ int i;
+ for (i = group_num-1; i < MAX_PREDICTORS; i+=30)
+ reset_predict_state(&ps[i]);
+}
+
static av_cold int aac_decode_init(AVCodecContext * avccontext) {
AACContext * ac = avccontext->priv_data;
int i;
}
#ifndef CONFIG_HARDCODED_TABLES
- for (i = 1 - IVQUANT_SIZE/2; i < IVQUANT_SIZE/2; i++)
- ff_aac_ivquant_tab[i + IVQUANT_SIZE/2 - 1] = cbrt(fabs(i)) * i;
- for (i = 0; i < 316; i++)
+ for (i = 0; i < 428; i++)
ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.);
#endif /* CONFIG_HARDCODED_TABLES */
- INIT_VLC_STATIC(&vlc_scalefactors, 7, sizeof(ff_aac_scalefactor_code)/sizeof(ff_aac_scalefactor_code[0]),
+ INIT_VLC_STATIC(&vlc_scalefactors,7,FF_ARRAY_ELEMS(ff_aac_scalefactor_code),
ff_aac_scalefactor_bits, sizeof(ff_aac_scalefactor_bits[0]), sizeof(ff_aac_scalefactor_bits[0]),
ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]),
352);
ff_mdct_init(&ac->mdct, 11, 1);
ff_mdct_init(&ac->mdct_small, 8, 1);
+ // window initialization
+ ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
+ ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
+ ff_sine_window_init(ff_sine_1024, 1024);
+ ff_sine_window_init(ff_sine_128, 128);
+
return 0;
}
skip_bits_long(gb, 8 * count);
}
+static int decode_prediction(AACContext * ac, IndividualChannelStream * ics, GetBitContext * gb) {
+ int sfb;
+ if (get_bits1(gb)) {
+ ics->predictor_reset_group = get_bits(gb, 5);
+ if (ics->predictor_reset_group == 0 || ics->predictor_reset_group > 30) {
+ av_log(ac->avccontext, AV_LOG_ERROR, "Invalid Predictor Reset Group.\n");
+ return -1;
+ }
+ }
+ for (sfb = 0; sfb < FFMIN(ics->max_sfb, ff_aac_pred_sfb_max[ac->m4ac.sampling_index]); sfb++) {
+ ics->prediction_used[sfb] = get_bits1(gb);
+ }
+ return 0;
+}
+
/**
* Decode Individual Channel Stream info; reference: table 4.6.
*
ics->use_kb_window[0] = get_bits1(gb);
ics->num_window_groups = 1;
ics->group_len[0] = 1;
-
- if (get_bits1(gb)) {
- av_log_missing_feature(ac->avccontext, "Predictor bit set but LTP is", 1);
- memset(ics, 0, sizeof(IndividualChannelStream));
- return -1;
+ if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
+ int i;
+ ics->max_sfb = get_bits(gb, 4);
+ for (i = 0; i < 7; i++) {
+ if (get_bits1(gb)) {
+ ics->group_len[ics->num_window_groups-1]++;
+ } else {
+ ics->num_window_groups++;
+ ics->group_len[ics->num_window_groups-1] = 1;
+ }
+ }
+ ics->num_windows = 8;
+ ics->swb_offset = swb_offset_128[ac->m4ac.sampling_index];
+ ics->num_swb = ff_aac_num_swb_128[ac->m4ac.sampling_index];
+ ics->tns_max_bands = tns_max_bands_128[ac->m4ac.sampling_index];
+ ics->predictor_present = 0;
+ } else {
+ ics->max_sfb = get_bits(gb, 6);
+ ics->num_windows = 1;
+ ics->swb_offset = swb_offset_1024[ac->m4ac.sampling_index];
+ ics->num_swb = ff_aac_num_swb_1024[ac->m4ac.sampling_index];
+ ics->tns_max_bands = tns_max_bands_1024[ac->m4ac.sampling_index];
+ ics->predictor_present = get_bits1(gb);
+ ics->predictor_reset_group = 0;
+ if (ics->predictor_present) {
+ if (ac->m4ac.object_type == AOT_AAC_MAIN) {
+ if (decode_prediction(ac, ics, gb)) {
+ memset(ics, 0, sizeof(IndividualChannelStream));
+ return -1;
+ }
+ } else if (ac->m4ac.object_type == AOT_AAC_LC) {
+ av_log(ac->avccontext, AV_LOG_ERROR, "Prediction is not allowed in AAC-LC.\n");
+ memset(ics, 0, sizeof(IndividualChannelStream));
+ return -1;
+ } else {
+ ff_log_missing_feature(ac->avccontext, "Predictor bit set but LTP is", 1);
+ memset(ics, 0, sizeof(IndividualChannelStream));
+ return -1;
+ }
}
}
return 0;
}
-/**
- * inverse quantization
- *
- * @param a quantized value to be dequantized
- * @return Returns dequantized value.
- */
-static inline float ivquant(int a) {
- if (a + (unsigned int)IVQUANT_SIZE/2 - 1 < (unsigned int)IVQUANT_SIZE - 1)
- return ff_aac_ivquant_tab[a + IVQUANT_SIZE/2 - 1];
- else
- return cbrtf(fabsf(a)) * a;
-}
-
/**
* Decode band types (section_data payload); reference: table 4.46.
*
sect_len, ics->max_sfb);
return -1;
}
+ for (; k < sect_len; k++) {
+ band_type [idx] = sect_band_type;
+ band_type_run_end[idx++] = sect_len;
+ }
}
}
return 0;
int offset[3] = { global_gain, global_gain - 90, 100 };
int noise_flag = 1;
static const char *sf_str[3] = { "Global gain", "Noise gain", "Intensity stereo position" };
- ics->intensity_present = 0;
for (g = 0; g < ics->num_window_groups; g++) {
for (i = 0; i < ics->max_sfb;) {
int run_end = band_type_run_end[idx];
for(; i < run_end; i++, idx++)
sf[idx] = 0.;
}else if((band_type[idx] == INTENSITY_BT) || (band_type[idx] == INTENSITY_BT2)) {
- ics->intensity_present = 1;
for(; i < run_end; i++, idx++) {
offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
if(offset[2] > 255U) {
"%s (%d) out of range.\n", sf_str[1], offset[1]);
return -1;
}
- sf[idx] = -ff_aac_pow2sf_tab[ offset[1] + sf_offset];
+ sf[idx] = -ff_aac_pow2sf_tab[ offset[1] + sf_offset + 100];
}
}else {
for(; i < run_end; i++, idx++) {
/**
* Decode pulse data; reference: table 4.7.
*/
-static void decode_pulses(Pulse * pulse, GetBitContext * gb) {
- int i;
+static int decode_pulses(Pulse * pulse, GetBitContext * gb, const uint16_t * swb_offset, int num_swb) {
+ int i, pulse_swb;
pulse->num_pulse = get_bits(gb, 2) + 1;
- pulse->start = get_bits(gb, 6);
- for (i = 0; i < pulse->num_pulse; i++) {
- pulse->offset[i] = get_bits(gb, 5);
- pulse->amp [i] = get_bits(gb, 4);
+ pulse_swb = get_bits(gb, 6);
+ if (pulse_swb >= num_swb)
+ return -1;
+ pulse->pos[0] = swb_offset[pulse_swb];
+ pulse->pos[0] += get_bits(gb, 5);
+ if (pulse->pos[0] > 1023)
+ return -1;
+ pulse->amp[0] = get_bits(gb, 4);
+ for (i = 1; i < pulse->num_pulse; i++) {
+ pulse->pos[i] = get_bits(gb, 5) + pulse->pos[i-1];
+ if (pulse->pos[i] > 1023)
+ return -1;
+ pulse->amp[i] = get_bits(gb, 4);
}
+ return 0;
+}
+
+/**
+ * Decode Temporal Noise Shaping data; reference: table 4.48.
+ *
+ * @return Returns error status. 0 - OK, !0 - error
+ */
+static int decode_tns(AACContext * ac, TemporalNoiseShaping * tns,
+ GetBitContext * gb, const IndividualChannelStream * ics) {
+ int w, filt, i, coef_len, coef_res, coef_compress;
+ const int is8 = ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE;
+ const int tns_max_order = is8 ? 7 : ac->m4ac.object_type == AOT_AAC_MAIN ? 20 : 12;
+ for (w = 0; w < ics->num_windows; w++) {
+ if ((tns->n_filt[w] = get_bits(gb, 2 - is8))) {
+ coef_res = get_bits1(gb);
+
+ for (filt = 0; filt < tns->n_filt[w]; filt++) {
+ int tmp2_idx;
+ tns->length[w][filt] = get_bits(gb, 6 - 2*is8);
+
+ if ((tns->order[w][filt] = get_bits(gb, 5 - 2*is8)) > tns_max_order) {
+ av_log(ac->avccontext, AV_LOG_ERROR, "TNS filter order %d is greater than maximum %d.",
+ tns->order[w][filt], tns_max_order);
+ tns->order[w][filt] = 0;
+ return -1;
+ }
+ if (tns->order[w][filt]) {
+ tns->direction[w][filt] = get_bits1(gb);
+ coef_compress = get_bits1(gb);
+ coef_len = coef_res + 3 - coef_compress;
+ tmp2_idx = 2*coef_compress + coef_res;
+
+ for (i = 0; i < tns->order[w][filt]; i++)
+ tns->coef[w][filt][i] = tns_tmp2_map[tmp2_idx][get_bits(gb, coef_len)];
+ }
+ }
+ }
+ }
+ return 0;
}
/**
}
/**
- * Add pulses with particular amplitudes to the quantized spectral data; reference: 4.6.3.3.
+ * Decode spectral data; reference: table 4.50.
+ * Dequantize and scale spectral data; reference: 4.6.3.3.
+ *
+ * @param coef array of dequantized, scaled spectral data
+ * @param sf array of scalefactors or intensity stereo positions
+ * @param pulse_present set if pulses are present
+ * @param pulse pointer to pulse data struct
+ * @param band_type array of the used band type
*
- * @param pulse pointer to pulse data struct
- * @param icoef array of quantized spectral data
+ * @return Returns error status. 0 - OK, !0 - error
*/
-static void add_pulses(int icoef[1024], const Pulse * pulse, const IndividualChannelStream * ics) {
- int i, off = ics->swb_offset[pulse->start];
- for (i = 0; i < pulse->num_pulse; i++) {
- int ic;
- off += pulse->offset[i];
- ic = (icoef[off] - 1)>>31;
- icoef[off] += (pulse->amp[i]^ic) - ic;
+static int decode_spectrum_and_dequant(AACContext * ac, float coef[1024], GetBitContext * gb, float sf[120],
+ int pulse_present, const Pulse * pulse, const IndividualChannelStream * ics, enum BandType band_type[120]) {
+ int i, k, g, idx = 0;
+ const int c = 1024/ics->num_windows;
+ const uint16_t * offsets = ics->swb_offset;
+ float *coef_base = coef;
+ static const float sign_lookup[] = { 1.0f, -1.0f };
+
+ for (g = 0; g < ics->num_windows; g++)
+ memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float)*(c - offsets[ics->max_sfb]));
+
+ for (g = 0; g < ics->num_window_groups; g++) {
+ for (i = 0; i < ics->max_sfb; i++, idx++) {
+ const int cur_band_type = band_type[idx];
+ const int dim = cur_band_type >= FIRST_PAIR_BT ? 2 : 4;
+ const int is_cb_unsigned = IS_CODEBOOK_UNSIGNED(cur_band_type);
+ int group;
+ if (cur_band_type == ZERO_BT || cur_band_type == INTENSITY_BT2 || cur_band_type == INTENSITY_BT) {
+ for (group = 0; group < ics->group_len[g]; group++) {
+ memset(coef + group * 128 + offsets[i], 0, (offsets[i+1] - offsets[i])*sizeof(float));
+ }
+ }else if (cur_band_type == NOISE_BT) {
+ for (group = 0; group < ics->group_len[g]; group++) {
+ float scale;
+ float band_energy = 0;
+ for (k = offsets[i]; k < offsets[i+1]; k++) {
+ ac->random_state = lcg_random(ac->random_state);
+ coef[group*128+k] = ac->random_state;
+ band_energy += coef[group*128+k]*coef[group*128+k];
+ }
+ scale = sf[idx] / sqrtf(band_energy);
+ for (k = offsets[i]; k < offsets[i+1]; k++) {
+ coef[group*128+k] *= scale;
+ }
+ }
+ }else {
+ for (group = 0; group < ics->group_len[g]; group++) {
+ for (k = offsets[i]; k < offsets[i+1]; k += dim) {
+ const int index = get_vlc2(gb, vlc_spectral[cur_band_type - 1].table, 6, 3);
+ const int coef_tmp_idx = (group << 7) + k;
+ const float *vq_ptr;
+ int j;
+ if(index >= ff_aac_spectral_sizes[cur_band_type - 1]) {
+ av_log(ac->avccontext, AV_LOG_ERROR,
+ "Read beyond end of ff_aac_codebook_vectors[%d][]. index %d >= %d\n",
+ cur_band_type - 1, index, ff_aac_spectral_sizes[cur_band_type - 1]);
+ return -1;
+ }
+ vq_ptr = &ff_aac_codebook_vectors[cur_band_type - 1][index * dim];
+ if (is_cb_unsigned) {
+ if (vq_ptr[0]) coef[coef_tmp_idx ] = sign_lookup[get_bits1(gb)];
+ if (vq_ptr[1]) coef[coef_tmp_idx + 1] = sign_lookup[get_bits1(gb)];
+ if (dim == 4) {
+ if (vq_ptr[2]) coef[coef_tmp_idx + 2] = sign_lookup[get_bits1(gb)];
+ if (vq_ptr[3]) coef[coef_tmp_idx + 3] = sign_lookup[get_bits1(gb)];
+ }
+ }else {
+ coef[coef_tmp_idx ] = 1.0f;
+ coef[coef_tmp_idx + 1] = 1.0f;
+ if (dim == 4) {
+ coef[coef_tmp_idx + 2] = 1.0f;
+ coef[coef_tmp_idx + 3] = 1.0f;
+ }
+ }
+ if (cur_band_type == ESC_BT) {
+ for (j = 0; j < 2; j++) {
+ if (vq_ptr[j] == 64.0f) {
+ int n = 4;
+ /* The total length of escape_sequence must be < 22 bits according
+ to the specification (i.e. max is 11111111110xxxxxxxxxx). */
+ while (get_bits1(gb) && n < 15) n++;
+ if(n == 15) {
+ av_log(ac->avccontext, AV_LOG_ERROR, "error in spectral data, ESC overflow\n");
+ return -1;
+ }
+ n = (1<<n) + get_bits(gb, n);
+ coef[coef_tmp_idx + j] *= cbrtf(n) * n;
+ }else
+ coef[coef_tmp_idx + j] *= vq_ptr[j];
+ }
+ }else
+ {
+ coef[coef_tmp_idx ] *= vq_ptr[0];
+ coef[coef_tmp_idx + 1] *= vq_ptr[1];
+ if (dim == 4) {
+ coef[coef_tmp_idx + 2] *= vq_ptr[2];
+ coef[coef_tmp_idx + 3] *= vq_ptr[3];
+ }
+ }
+ coef[coef_tmp_idx ] *= sf[idx];
+ coef[coef_tmp_idx + 1] *= sf[idx];
+ if (dim == 4) {
+ coef[coef_tmp_idx + 2] *= sf[idx];
+ coef[coef_tmp_idx + 3] *= sf[idx];
+ }
+ }
+ }
+ }
+ }
+ coef += ics->group_len[g]<<7;
}
+
+ if (pulse_present) {
+ idx = 0;
+ for(i = 0; i < pulse->num_pulse; i++){
+ float co = coef_base[ pulse->pos[i] ];
+ while(offsets[idx + 1] <= pulse->pos[i])
+ idx++;
+ if (band_type[idx] != NOISE_BT && sf[idx]) {
+ float ico = -pulse->amp[i];
+ if (co) {
+ co /= sf[idx];
+ ico = co / sqrtf(sqrtf(fabsf(co))) + (co > 0 ? -ico : ico);
+ }
+ coef_base[ pulse->pos[i] ] = cbrtf(fabsf(ico)) * ico * sf[idx];
+ }
+ }
+ }
+ return 0;
+}
+
+static av_always_inline float flt16_round(float pf) {
+ int exp;
+ pf = frexpf(pf, &exp);
+ pf = ldexpf(roundf(ldexpf(pf, 8)), exp-8);
+ return pf;
+}
+
+static av_always_inline float flt16_even(float pf) {
+ int exp;
+ pf = frexpf(pf, &exp);
+ pf = ldexpf(rintf(ldexpf(pf, 8)), exp-8);
+ return pf;
+}
+
+static av_always_inline float flt16_trunc(float pf) {
+ int exp;
+ pf = frexpf(pf, &exp);
+ pf = ldexpf(truncf(ldexpf(pf, 8)), exp-8);
+ return pf;
+}
+
+static void predict(AACContext * ac, PredictorState * ps, float* coef, int output_enable) {
+ const float a = 0.953125; // 61.0/64
+ const float alpha = 0.90625; // 29.0/32
+ float e0, e1;
+ float pv;
+ float k1, k2;
+
+ k1 = ps->var0 > 1 ? ps->cor0 * flt16_even(a / ps->var0) : 0;
+ k2 = ps->var1 > 1 ? ps->cor1 * flt16_even(a / ps->var1) : 0;
+
+ pv = flt16_round(k1 * ps->r0 + k2 * ps->r1);
+ if (output_enable)
+ *coef += pv * ac->sf_scale;
+
+ e0 = *coef / ac->sf_scale;
+ e1 = e0 - k1 * ps->r0;
+
+ ps->cor1 = flt16_trunc(alpha * ps->cor1 + ps->r1 * e1);
+ ps->var1 = flt16_trunc(alpha * ps->var1 + 0.5 * (ps->r1 * ps->r1 + e1 * e1));
+ ps->cor0 = flt16_trunc(alpha * ps->cor0 + ps->r0 * e0);
+ ps->var0 = flt16_trunc(alpha * ps->var0 + 0.5 * (ps->r0 * ps->r0 + e0 * e0));
+
+ ps->r1 = flt16_trunc(a * (ps->r0 - k1 * e0));
+ ps->r0 = flt16_trunc(a * e0);
+}
+
+/**
+ * Apply AAC-Main style frequency domain prediction.
+ */
+static void apply_prediction(AACContext * ac, SingleChannelElement * sce) {
+ int sfb, k;
+
+ if (!sce->ics.predictor_initialized) {
+ reset_all_predictors(sce->predictor_state);
+ sce->ics.predictor_initialized = 1;
+ }
+
+ if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
+ for (sfb = 0; sfb < ff_aac_pred_sfb_max[ac->m4ac.sampling_index]; sfb++) {
+ for (k = sce->ics.swb_offset[sfb]; k < sce->ics.swb_offset[sfb + 1]; k++) {
+ predict(ac, &sce->predictor_state[k], &sce->coeffs[k],
+ sce->ics.predictor_present && sce->ics.prediction_used[sfb]);
+ }
+ }
+ if (sce->ics.predictor_reset_group)
+ reset_predictor_group(sce->predictor_state, sce->ics.predictor_reset_group);
+ } else
+ reset_all_predictors(sce->predictor_state);
}
/**
* @return Returns error status. 0 - OK, !0 - error
*/
static int decode_ics(AACContext * ac, SingleChannelElement * sce, GetBitContext * gb, int common_window, int scale_flag) {
- int icoeffs[1024];
Pulse pulse;
TemporalNoiseShaping * tns = &sce->tns;
IndividualChannelStream * ics = &sce->ics;
float * out = sce->coeffs;
int global_gain, pulse_present = 0;
- /* These two assignments are to silence some GCC warnings about the
- * variables being used uninitialised when in fact they always are.
+ /* This assignment is to silence a GCC warning about the variable being used
+ * uninitialized when in fact it always is.
*/
pulse.num_pulse = 0;
- pulse.start = 0;
global_gain = get_bits(gb, 8);
av_log(ac->avccontext, AV_LOG_ERROR, "Pulse tool not allowed in eight short sequence.\n");
return -1;
}
- decode_pulses(&pulse, gb);
+ if (decode_pulses(&pulse, gb, ics->swb_offset, ics->num_swb)) {
+ av_log(ac->avccontext, AV_LOG_ERROR, "Pulse data corrupt or invalid.\n");
+ return -1;
+ }
}
if ((tns->present = get_bits1(gb)) && decode_tns(ac, tns, gb, ics))
return -1;
if (get_bits1(gb)) {
- av_log_missing_feature(ac->avccontext, "SSR", 1);
+ ff_log_missing_feature(ac->avccontext, "SSR", 1);
return -1;
}
}
- if (decode_spectrum(ac, icoeffs, gb, ics, sce->band_type) < 0)
+ if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0)
return -1;
- if (pulse_present)
- add_pulses(icoeffs, &pulse, ics);
- dequant(ac, out, icoeffs, sce->sf, ics, sce->band_type);
+
+ if(ac->m4ac.object_type == AOT_AAC_MAIN && !common_window)
+ apply_prediction(ac, sce);
+
return 0;
}
+/**
+ * Mid/Side stereo decoding; reference: 4.6.8.1.3.
+ */
+static void apply_mid_side_stereo(ChannelElement * cpe) {
+ const IndividualChannelStream * ics = &cpe->ch[0].ics;
+ float *ch0 = cpe->ch[0].coeffs;
+ float *ch1 = cpe->ch[1].coeffs;
+ int g, i, k, group, idx = 0;
+ const uint16_t * offsets = ics->swb_offset;
+ for (g = 0; g < ics->num_window_groups; g++) {
+ for (i = 0; i < ics->max_sfb; i++, idx++) {
+ if (cpe->ms_mask[idx] &&
+ cpe->ch[0].band_type[idx] < NOISE_BT && cpe->ch[1].band_type[idx] < NOISE_BT) {
+ for (group = 0; group < ics->group_len[g]; group++) {
+ for (k = offsets[i]; k < offsets[i+1]; k++) {
+ float tmp = ch0[group*128 + k] - ch1[group*128 + k];
+ ch0[group*128 + k] += ch1[group*128 + k];
+ ch1[group*128 + k] = tmp;
+ }
+ }
+ }
+ }
+ ch0 += ics->group_len[g]*128;
+ ch1 += ics->group_len[g]*128;
+ }
+}
+
+/**
+ * intensity stereo decoding; reference: 4.6.8.2.3
+ *
+ * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
+ * [1] mask is decoded from bitstream; [2] mask is all 1s;
+ * [3] reserved for scalable AAC
+ */
+static void apply_intensity_stereo(ChannelElement * cpe, int ms_present) {
+ const IndividualChannelStream * ics = &cpe->ch[1].ics;
+ SingleChannelElement * sce1 = &cpe->ch[1];
+ float *coef0 = cpe->ch[0].coeffs, *coef1 = cpe->ch[1].coeffs;
+ const uint16_t * offsets = ics->swb_offset;
+ int g, group, i, k, idx = 0;
+ int c;
+ float scale;
+ for (g = 0; g < ics->num_window_groups; g++) {
+ for (i = 0; i < ics->max_sfb;) {
+ if (sce1->band_type[idx] == INTENSITY_BT || sce1->band_type[idx] == INTENSITY_BT2) {
+ const int bt_run_end = sce1->band_type_run_end[idx];
+ for (; i < bt_run_end; i++, idx++) {
+ c = -1 + 2 * (sce1->band_type[idx] - 14);
+ if (ms_present)
+ c *= 1 - 2 * cpe->ms_mask[idx];
+ scale = c * sce1->sf[idx];
+ for (group = 0; group < ics->group_len[g]; group++)
+ for (k = offsets[i]; k < offsets[i+1]; k++)
+ coef1[group*128 + k] = scale * coef0[group*128 + k];
+ }
+ } else {
+ int bt_run_end = sce1->band_type_run_end[idx];
+ idx += bt_run_end - i;
+ i = bt_run_end;
+ }
+ }
+ coef0 += ics->group_len[g]*128;
+ coef1 += ics->group_len[g]*128;
+ }
+}
+
/**
* Decode a channel_pair_element; reference: table 4.4.
*
if ((ret = decode_ics(ac, &cpe->ch[1], gb, common_window, 0)))
return ret;
- if (common_window && ms_present)
- apply_mid_side_stereo(cpe);
+ if (common_window) {
+ if (ms_present)
+ apply_mid_side_stereo(cpe);
+ if (ac->m4ac.object_type == AOT_AAC_MAIN) {
+ apply_prediction(ac, &cpe->ch[0]);
+ apply_prediction(ac, &cpe->ch[1]);
+ }
+ }
- if (cpe->ch[1].ics.intensity_present)
- apply_intensity_stereo(cpe, ms_present);
+ apply_intensity_stereo(cpe, ms_present);
return 0;
}
+/**
+ * Decode coupling_channel_element; reference: table 4.8.
+ *
+ * @param elem_id Identifies the instance of a syntax element.
+ *
+ * @return Returns error status. 0 - OK, !0 - error
+ */
+static int decode_cce(AACContext * ac, GetBitContext * gb, ChannelElement * che) {
+ int num_gain = 0;
+ int c, g, sfb, ret;
+ int sign;
+ float scale;
+ SingleChannelElement * sce = &che->ch[0];
+ ChannelCoupling * coup = &che->coup;
+
coup->coupling_point = 2*get_bits1(gb);
coup->num_coupled = get_bits(gb, 3);
for (c = 0; c <= coup->num_coupled; c++) {
if (coup->ch_select[c] == 3)
num_gain++;
} else
- coup->ch_select[c] = 1;
+ coup->ch_select[c] = 2;
}
coup->coupling_point += get_bits1(gb);
}
sign = get_bits(gb, 1);
- scale = pow(2., pow(2., get_bits(gb, 2) - 3));
+ scale = pow(2., pow(2., (int)get_bits(gb, 2) - 3));
if ((ret = decode_ics(ac, sce, gb, 0, 0)))
return ret;
for (c = 0; c < num_gain; c++) {
+ int idx = 0;
int cge = 1;
int gain = 0;
float gain_cache = 1.;
if (c) {
cge = coup->coupling_point == AFTER_IMDCT ? 1 : get_bits1(gb);
gain = cge ? get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60: 0;
- gain_cache = pow(scale, gain);
+ gain_cache = pow(scale, -gain);
}
- for (g = 0; g < sce->ics.num_window_groups; g++)
- for (sfb = 0; sfb < sce->ics.max_sfb; sfb++, idx++)
+ for (g = 0; g < sce->ics.num_window_groups; g++) {
+ for (sfb = 0; sfb < sce->ics.max_sfb; sfb++, idx++) {
if (sce->band_type[idx] != ZERO_BT) {
if (!cge) {
int t = get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
if (t) {
int s = 1;
+ t = gain += t;
if (sign) {
s -= 2 * (t & 0x1);
t >>= 1;
}
- gain += t;
- gain_cache = pow(scale, gain) * s;
+ gain_cache = pow(scale, -t) * s;
}
}
coup->gain[c][idx] = gain_cache;
}
+ }
+ }
}
return 0;
}
*/
static int decode_sbr_extension(AACContext * ac, GetBitContext * gb, int crc, int cnt) {
// TODO : sbr_extension implementation
- av_log_missing_feature(ac->avccontext, "SBR", 0);
+ ff_log_missing_feature(ac->avccontext, "SBR", 0);
skip_bits_long(gb, 8*cnt - 4); // -4 due to reading extension type
return cnt;
}
return res;
}
+/**
+ * Decode Temporal Noise Shaping filter coefficients and apply all-pole filters; reference: 4.6.9.3.
+ *
+ * @param decode 1 if tool is used normally, 0 if tool is used in LTP.
+ * @param coef spectral coefficients
+ */
+static void apply_tns(float coef[1024], TemporalNoiseShaping * tns, IndividualChannelStream * ics, int decode) {
+ const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
+ int w, filt, m, i;
+ int bottom, top, order, start, end, size, inc;
+ float lpc[TNS_MAX_ORDER];
+
+ for (w = 0; w < ics->num_windows; w++) {
+ bottom = ics->num_swb;
+ for (filt = 0; filt < tns->n_filt[w]; filt++) {
+ top = bottom;
+ bottom = FFMAX(0, top - tns->length[w][filt]);
+ order = tns->order[w][filt];
+ if (order == 0)
+ continue;
+
+ // tns_decode_coef
+ compute_lpc_coefs(tns->coef[w][filt], order, lpc, 0, 0, 0);
+
+ start = ics->swb_offset[FFMIN(bottom, mmm)];
+ end = ics->swb_offset[FFMIN( top, mmm)];
+ if ((size = end - start) <= 0)
+ continue;
+ if (tns->direction[w][filt]) {
+ inc = -1; start = end - 1;
+ } else {
+ inc = 1;
+ }
+ start += w * 128;
+
+ // ar filter
+ for (m = 0; m < size; m++, start += inc)
+ for (i = 1; i <= FFMIN(m, order); i++)
+ coef[start] -= coef[start - i*inc] * lpc[i-1];
+ }
+ }
+}
+
/**
* Conduct IMDCT and windowing.
*/
float * in = sce->coeffs;
float * out = sce->ret;
float * saved = sce->saved;
- const float * lwindow = ics->use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_aac_sine_long_1024;
- const float * swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_aac_sine_short_128;
- const float * lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_aac_sine_long_1024;
- const float * swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_aac_sine_short_128;
+ const float * swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
+ const float * lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
+ const float * swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
float * buf = ac->buf_mdct;
+ float * temp = ac->temp;
int i;
+ // imdct
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
if (ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE)
av_log(ac->avccontext, AV_LOG_WARNING,
"Transition from an ONLY_LONG or LONG_STOP to an EIGHT_SHORT sequence detected. "
"If you heard an audible artifact, please submit the sample to the FFmpeg developers.\n");
- for (i = 0; i < 2048; i += 256) {
- ff_imdct_calc(&ac->mdct_small, buf + i, in + i/2);
- ac->dsp.vector_fmul_reverse(ac->revers + i/2, buf + i + 128, swindow, 128);
- }
- for (i = 0; i < 448; i++) out[i] = saved[i] + ac->add_bias;
-
- ac->dsp.vector_fmul_add_add(out + 448 + 0*128, buf + 0*128, swindow_prev, saved + 448 , ac->add_bias, 128, 1);
- ac->dsp.vector_fmul_add_add(out + 448 + 1*128, buf + 2*128, swindow, ac->revers + 0*128, ac->add_bias, 128, 1);
- ac->dsp.vector_fmul_add_add(out + 448 + 2*128, buf + 4*128, swindow, ac->revers + 1*128, ac->add_bias, 128, 1);
- ac->dsp.vector_fmul_add_add(out + 448 + 3*128, buf + 6*128, swindow, ac->revers + 2*128, ac->add_bias, 128, 1);
- ac->dsp.vector_fmul_add_add(out + 448 + 4*128, buf + 8*128, swindow, ac->revers + 3*128, ac->add_bias, 64, 1);
-
-#if 0
- vector_fmul_add_add_add(&ac->dsp, out + 448 + 1*128, buf + 2*128, swindow, saved + 448 + 1*128, ac->revers + 0*128, ac->add_bias, 128);
- vector_fmul_add_add_add(&ac->dsp, out + 448 + 2*128, buf + 4*128, swindow, saved + 448 + 2*128, ac->revers + 1*128, ac->add_bias, 128);
- vector_fmul_add_add_add(&ac->dsp, out + 448 + 3*128, buf + 6*128, swindow, saved + 448 + 3*128, ac->revers + 2*128, ac->add_bias, 128);
- vector_fmul_add_add_add(&ac->dsp, out + 448 + 4*128, buf + 8*128, swindow, saved + 448 + 4*128, ac->revers + 3*128, ac->add_bias, 64);
-#endif
-
- ac->dsp.vector_fmul_add_add(saved, buf + 1024 + 64, swindow + 64, ac->revers + 3*128+64, 0, 64, 1);
- ac->dsp.vector_fmul_add_add(saved + 64, buf + 1024 + 2*128, swindow, ac->revers + 4*128, 0, 128, 1);
- ac->dsp.vector_fmul_add_add(saved + 192, buf + 1024 + 4*128, swindow, ac->revers + 5*128, 0, 128, 1);
- ac->dsp.vector_fmul_add_add(saved + 320, buf + 1024 + 6*128, swindow, ac->revers + 6*128, 0, 128, 1);
- memcpy( saved + 448, ac->revers + 7*128, 128 * sizeof(float));
- memset( saved + 576, 0, 448 * sizeof(float));
+ for (i = 0; i < 1024; i += 128)
+ ff_imdct_half(&ac->mdct_small, buf + i, in + i);
+ } else
+ ff_imdct_half(&ac->mdct, buf, in);
+
+ /* window overlapping
+ * NOTE: To simplify the overlapping code, all 'meaningless' short to long
+ * and long to short transitions are considered to be short to short
+ * transitions. This leaves just two cases (long to long and short to short)
+ * with a little special sauce for EIGHT_SHORT_SEQUENCE.
+ */
+ if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
+ (ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
+ ac->dsp.vector_fmul_window( out, saved, buf, lwindow_prev, ac->add_bias, 512);
} else {
- ff_imdct_calc(&ac->mdct, buf, in);
- if (ics->window_sequence[0] == LONG_STOP_SEQUENCE) {
- for (i = 0; i < 448; i++) out[i] = saved[i] + ac->add_bias;
- ac->dsp.vector_fmul_add_add(out + 448, buf + 448, swindow_prev, saved + 448, ac->add_bias, 128, 1);
- for (i = 576; i < 1024; i++) out[i] = buf[i] + saved[i] + ac->add_bias;
- } else {
- ac->dsp.vector_fmul_add_add(out, buf, lwindow_prev, saved, ac->add_bias, 1024, 1);
- }
- if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
- memcpy(saved, buf + 1024, 448 * sizeof(float));
- ac->dsp.vector_fmul_reverse(saved + 448, buf + 1024 + 448, swindow, 128);
- memset(saved + 576, 0, 448 * sizeof(float));
+ for (i = 0; i < 448; i++)
+ out[i] = saved[i] + ac->add_bias;
+
+ if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
+ ac->dsp.vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, ac->add_bias, 64);
+ ac->dsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, ac->add_bias, 64);
+ ac->dsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, ac->add_bias, 64);
+ ac->dsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, ac->add_bias, 64);
+ ac->dsp.vector_fmul_window(temp, buf + 3*128 + 64, buf + 4*128, swindow, ac->add_bias, 64);
+ memcpy( out + 448 + 4*128, temp, 64 * sizeof(float));
} else {
- ac->dsp.vector_fmul_reverse(saved, buf + 1024, lwindow, 1024);
+ ac->dsp.vector_fmul_window(out + 448, saved + 448, buf, swindow_prev, ac->add_bias, 64);
+ for (i = 576; i < 1024; i++)
+ out[i] = buf[i-512] + ac->add_bias;
}
}
+
+ // buffer update
+ if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
+ for (i = 0; i < 64; i++)
+ saved[i] = temp[64 + i] - ac->add_bias;
+ ac->dsp.vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 0, 64);
+ ac->dsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 0, 64);
+ ac->dsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 0, 64);
+ memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(float));
+ } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
+ memcpy( saved, buf + 512, 448 * sizeof(float));
+ memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(float));
+ } else { // LONG_STOP or ONLY_LONG
+ memcpy( saved, buf + 512, 512 * sizeof(float));
+ }
}
/**
*
* @param index index into coupling gain array
*/
-static void apply_dependent_coupling(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index) {
- IndividualChannelStream * ics = &cc->ch[0].ics;
+static void apply_dependent_coupling(AACContext * ac, SingleChannelElement * target, ChannelElement * cce, int index) {
+ IndividualChannelStream * ics = &cce->ch[0].ics;
const uint16_t * offsets = ics->swb_offset;
- float * dest = sce->coeffs;
- const float * src = cc->ch[0].coeffs;
+ float * dest = target->coeffs;
+ const float * src = cce->ch[0].coeffs;
int g, i, group, k, idx = 0;
if(ac->m4ac.object_type == AOT_AAC_LTP) {
av_log(ac->avccontext, AV_LOG_ERROR,
}
for (g = 0; g < ics->num_window_groups; g++) {
for (i = 0; i < ics->max_sfb; i++, idx++) {
- if (cc->ch[0].band_type[idx] != ZERO_BT) {
+ if (cce->ch[0].band_type[idx] != ZERO_BT) {
for (group = 0; group < ics->group_len[g]; group++) {
for (k = offsets[i]; k < offsets[i+1]; k++) {
// XXX dsputil-ize
- dest[group*128+k] += cc->coup.gain[index][idx] * src[group*128+k];
+ dest[group*128+k] += cce->coup.gain[index][idx] * src[group*128+k];
}
}
}
*
* @param index index into coupling gain array
*/
-static void apply_independent_coupling(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index) {
+static void apply_independent_coupling(AACContext * ac, SingleChannelElement * target, ChannelElement * cce, int index) {
int i;
for (i = 0; i < 1024; i++)
- sce->ret[i] += cc->coup.gain[index][0] * (cc->ch[0].ret[i] - ac->add_bias);
+ target->ret[i] += cce->coup.gain[index][0] * (cce->ch[0].ret[i] - ac->add_bias);
}
+/**
+ * channel coupling transformation interface
+ *
+ * @param index index into coupling gain array
+ * @param apply_coupling_method pointer to (in)dependent coupling function
+ */
+static void apply_channel_coupling(AACContext * ac, ChannelElement * cc,
+ enum RawDataBlockType type, int elem_id, enum CouplingPoint coupling_point,
+ void (*apply_coupling_method)(AACContext * ac, SingleChannelElement * target, ChannelElement * cce, int index))
+{
+ int i, c;
+
+ for (i = 0; i < MAX_ELEM_ID; i++) {
+ ChannelElement *cce = ac->che[TYPE_CCE][i];
+ int index = 0;
+
+ if (cce && cce->coup.coupling_point == coupling_point) {
+ ChannelCoupling * coup = &cce->coup;
+
+ for (c = 0; c <= coup->num_coupled; c++) {
+ if (coup->type[c] == type && coup->id_select[c] == elem_id) {
+ if (coup->ch_select[c] != 1) {
+ apply_coupling_method(ac, &cc->ch[0], cce, index);
+ if (coup->ch_select[c] != 0)
+ index++;
+ }
+ if (coup->ch_select[c] != 2)
+ apply_coupling_method(ac, &cc->ch[1], cce, index++);
+ } else
+ index += 1 + (coup->ch_select[c] == 3);
+ }
+ }
+ }
+}
+
+/**
+ * Convert spectral data to float samples, applying all supported tools as appropriate.
+ */
+static void spectral_to_sample(AACContext * ac) {
+ int i, type;
+ for(type = 3; type >= 0; type--) {
+ for (i = 0; i < MAX_ELEM_ID; i++) {
+ ChannelElement *che = ac->che[type][i];
+ if(che) {
+ if(type <= TYPE_CPE)
+ apply_channel_coupling(ac, che, type, i, BEFORE_TNS, apply_dependent_coupling);
+ if(che->ch[0].tns.present)
+ apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1);
+ if(che->ch[1].tns.present)
+ apply_tns(che->ch[1].coeffs, &che->ch[1].tns, &che->ch[1].ics, 1);
+ if(type <= TYPE_CPE)
+ apply_channel_coupling(ac, che, type, i, BETWEEN_TNS_AND_IMDCT, apply_dependent_coupling);
+ if(type != TYPE_CCE || che->coup.coupling_point == AFTER_IMDCT)
+ imdct_and_windowing(ac, &che->ch[0]);
+ if(type == TYPE_CPE)
+ imdct_and_windowing(ac, &che->ch[1]);
+ if(type <= TYPE_CCE)
+ apply_channel_coupling(ac, che, type, i, AFTER_IMDCT, apply_independent_coupling);
}
}
}
ac->che[TYPE_SCE][elem_id] = ac->che[TYPE_LFE][0];
ac->che[TYPE_LFE][0] = NULL;
}
- if(elem_type && elem_type < TYPE_DSE) {
+ if(elem_type < TYPE_DSE) {
if(!ac->che[elem_type][elem_id])
return -1;
if(elem_type != TYPE_CCE)
break;
case TYPE_CCE:
- err = decode_cce(ac, &gb, ac->che[TYPE_SCE][elem_id]);
+ err = decode_cce(ac, &gb, ac->che[TYPE_CCE][elem_id]);
break;
case TYPE_LFE:
if (!ac->is_saved) {
ac->is_saved = 1;
*data_size = 0;
- return 0;
+ return buf_size;
}
data_size_tmp = 1024 * avccontext->channels * sizeof(int16_t);