{
PutBitContext pb;
AACEncContext *s = avctx->priv_data;
+ int channels = s->channels - (s->channels == 8 ? 1 : 0);
init_put_bits(&pb, avctx->extradata, avctx->extradata_size);
put_bits(&pb, 5, s->profile+1); //profile
put_bits(&pb, 4, s->samplerate_index); //sample rate index
- put_bits(&pb, 4, s->channels);
+ put_bits(&pb, 4, channels);
//GASpecificConfig
put_bits(&pb, 1, 0); //frame length - 1024 samples
put_bits(&pb, 1, 0); //does not depend on core coder
flush_put_bits(&pb);
}
+void ff_quantize_band_cost_cache_init(struct AACEncContext *s)
+{
+ int sf, g;
+ for (sf = 0; sf < 256; sf++) {
+ for (g = 0; g < 128; g++) {
+ s->quantize_band_cost_cache[sf][g].bits = -1;
+ }
+ }
+}
+
#define WINDOW_FUNC(type) \
static void apply_ ##type ##_window(AVFloatDSPContext *fdsp, \
SingleChannelElement *sce, \
start += ics->swb_sizes[g];
continue;
}
+ if (cpe->ms_mask[w*16 + g])
+ p *= -1;
for (i = 0; i < ics->swb_sizes[g]; i++) {
float sum = (cpe->ch[0].coeffs[start+i] + p*cpe->ch[1].coeffs[start+i])*scale;
cpe->ch[0].coeffs[start+i] = sum;
for (w2 = 0; w2 < ics->group_len[w]; w2++) {
int start = (w+w2) * 128;
for (g = 0; g < ics->num_swb; g++) {
- if (!cpe->ms_mask[w*16 + g]) {
+ if (!cpe->ms_mask[w*16 + g] && !cpe->is_mask[w*16 + g]) {
start += ics->swb_sizes[g];
continue;
}
put_ics_info(s, &sce->ics);
if (s->coder->encode_main_pred)
s->coder->encode_main_pred(s, sce);
+ if (s->coder->encode_ltp_info)
+ s->coder->encode_ltp_info(s, sce, 0);
}
encode_band_info(s, sce);
encode_scale_factors(avctx, s, sce);
float **samples = s->planar_samples, *samples2, *la, *overlap;
ChannelElement *cpe;
SingleChannelElement *sce;
+ IndividualChannelStream *ics;
int i, its, ch, w, chans, tag, start_ch, ret, frame_bits;
+ int target_bits, rate_bits, too_many_bits, too_few_bits;
int ms_mode = 0, is_mode = 0, tns_mode = 0, pred_mode = 0;
int chan_el_counter[4];
FFPsyWindowInfo windows[AAC_MAX_CHANNELS];
chans = tag == TYPE_CPE ? 2 : 1;
cpe = &s->cpe[i];
for (ch = 0; ch < chans; ch++) {
- IndividualChannelStream *ics = &cpe->ch[ch].ics;
- int cur_channel = start_ch + ch;
+ sce = &cpe->ch[ch];
+ ics = &sce->ics;
+ s->cur_channel = start_ch + ch;
float clip_avoidance_factor;
- overlap = &samples[cur_channel][0];
+ overlap = &samples[s->cur_channel][0];
samples2 = overlap + 1024;
la = samples2 + (448+64);
if (!frame)
*/
ics->num_swb = s->samplerate_index >= 8 ? 1 : 3;
} else {
- wi[ch] = s->psy.model->window(&s->psy, samples2, la, cur_channel,
+ wi[ch] = s->psy.model->window(&s->psy, samples2, la, s->cur_channel,
ics->window_sequence[0]);
}
ics->window_sequence[1] = ics->window_sequence[0];
ics->clip_avoidance_factor = 1.0f;
}
- apply_window_and_mdct(s, &cpe->ch[ch], overlap);
+ apply_window_and_mdct(s, sce, overlap);
+
+ if (s->options.ltp && s->coder->update_ltp) {
+ s->coder->update_ltp(s, sce);
+ apply_window[sce->ics.window_sequence[0]](s->fdsp, sce, &sce->ltp_state[0]);
+ s->mdct1024.mdct_calc(&s->mdct1024, sce->lcoeffs, sce->ret_buf);
+ }
+
if (isnan(cpe->ch->coeffs[0])) {
av_log(avctx, AV_LOG_ERROR, "Input contains NaN\n");
return AVERROR(EINVAL);
}
- avoid_clipping(s, &cpe->ch[ch]);
+ avoid_clipping(s, sce);
}
start_ch += chans;
}
return ret;
frame_bits = its = 0;
do {
- int target_bits, too_many_bits, too_few_bits;
-
init_put_bits(&s->pb, avpkt->data, avpkt->size);
if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & AV_CODEC_FLAG_BITEXACT))
sce = &cpe->ch[ch];
coeffs[ch] = sce->coeffs;
sce->ics.predictor_present = 0;
- memset(&sce->ics.prediction_used, 0, sizeof(sce->ics.prediction_used));
+ sce->ics.ltp.present = 0;
+ memset(sce->ics.ltp.used, 0, sizeof(sce->ics.ltp.used));
+ memset(sce->ics.prediction_used, 0, sizeof(sce->ics.prediction_used));
memset(&sce->tns, 0, sizeof(TemporalNoiseShaping));
for (w = 0; w < 128; w++)
if (sce->band_type[w] > RESERVED_BT)
s->psy.model->analyze(&s->psy, start_ch, coeffs, wi);
if (s->psy.bitres.alloc > 0) {
/* Lambda unused here on purpose, we need to take psy's unscaled allocation */
- target_bits += s->psy.bitres.alloc;
+ target_bits += s->psy.bitres.alloc
+ * (s->lambda / (avctx->global_quality ? avctx->global_quality : 120));
s->psy.bitres.alloc /= chans;
}
s->cur_type = tag;
for (ch = 0; ch < chans; ch++) {
s->cur_channel = start_ch + ch;
+ if (s->options.pns && s->coder->mark_pns)
+ s->coder->mark_pns(s, avctx, &cpe->ch[ch]);
s->coder->search_for_quantizers(avctx, s, &cpe->ch[ch], s->lambda);
}
if (chans > 1
s->coder->search_for_pred(s, sce);
if (cpe->ch[ch].ics.predictor_present) pred_mode = 1;
}
- if (s->coder->adjust_common_prediction)
- s->coder->adjust_common_prediction(s, cpe);
+ if (s->coder->adjust_common_pred)
+ s->coder->adjust_common_pred(s, cpe);
for (ch = 0; ch < chans; ch++) {
sce = &cpe->ch[ch];
s->cur_channel = start_ch + ch;
}
s->cur_channel = start_ch;
}
- if (s->options.stereo_mode) { /* Mid/Side stereo */
- if (s->options.stereo_mode == -1 && s->coder->search_for_ms)
+ if (s->options.mid_side) { /* Mid/Side stereo */
+ if (s->options.mid_side == -1 && s->coder->search_for_ms)
s->coder->search_for_ms(s, cpe);
else if (cpe->common_window)
memset(cpe->ms_mask, 1, sizeof(cpe->ms_mask));
- for (w = 0; w < 128; w++)
- cpe->ms_mask[w] = cpe->is_mask[w] ? 0 : cpe->ms_mask[w];
apply_mid_side_stereo(cpe);
}
adjust_frame_information(cpe, chans);
+ if (s->options.ltp) { /* LTP */
+ for (ch = 0; ch < chans; ch++) {
+ sce = &cpe->ch[ch];
+ s->cur_channel = start_ch + ch;
+ if (s->coder->search_for_ltp)
+ s->coder->search_for_ltp(s, sce, cpe->common_window);
+ if (sce->ics.ltp.present) pred_mode = 1;
+ }
+ s->cur_channel = start_ch;
+ if (s->coder->adjust_common_ltp)
+ s->coder->adjust_common_ltp(s, cpe);
+ }
if (chans == 2) {
put_bits(&s->pb, 1, cpe->common_window);
if (cpe->common_window) {
put_ics_info(s, &cpe->ch[0].ics);
if (s->coder->encode_main_pred)
s->coder->encode_main_pred(s, &cpe->ch[0]);
+ if (s->coder->encode_ltp_info)
+ s->coder->encode_ltp_info(s, &cpe->ch[0], 1);
encode_ms_info(&s->pb, cpe);
if (cpe->ms_mode) ms_mode = 1;
}
}
/* rate control stuff
- * target either the nominal bitrate, or what psy's bit reservoir says to target
- * whichever is greatest
+ * allow between the nominal bitrate, and what psy's bit reservoir says to target
+ * but drift towards the nominal bitrate always
*/
-
frame_bits = put_bits_count(&s->pb);
- target_bits = FFMAX(target_bits, avctx->bit_rate * 1024 / avctx->sample_rate);
- target_bits = FFMIN(target_bits, 6144 * s->channels - 3);
+ rate_bits = avctx->bit_rate * 1024 / avctx->sample_rate;
+ rate_bits = FFMIN(rate_bits, 6144 * s->channels - 3);
+ too_many_bits = FFMAX(target_bits, rate_bits);
+ too_many_bits = FFMIN(too_many_bits, 6144 * s->channels - 3);
+ too_few_bits = FFMIN(FFMAX(rate_bits - rate_bits/4, target_bits), too_many_bits);
/* When using ABR, be strict (but only for increasing) */
- too_many_bits = target_bits + target_bits/2;
- too_few_bits = target_bits - target_bits/8;
+ too_few_bits = too_few_bits - too_few_bits/8;
+ too_many_bits = too_many_bits + too_many_bits/2;
if ( its == 0 /* for steady-state Q-scale tracking */
|| (its < 5 && (frame_bits < too_few_bits || frame_bits > too_many_bits))
|| frame_bits >= 6144 * s->channels - 3 )
{
- float ratio = ((float)target_bits) / frame_bits;
+ float ratio = ((float)rate_bits) / frame_bits;
if (frame_bits >= too_few_bits && frame_bits <= too_many_bits) {
/*
s->lambda = FFMIN(s->lambda * ratio, 65536.f);
/* Keep iterating if we must reduce and lambda is in the sky */
- if (s->lambda < 300.f || ratio > 0.9f) {
+ if ((s->lambda < 300.f || ratio > 0.9f) && (s->lambda > 10.f || ratio < 1.1f)) {
break;
} else {
if (is_mode || ms_mode || tns_mode || pred_mode) {
}
} while (1);
+ if (s->options.ltp && s->coder->ltp_insert_new_frame)
+ s->coder->ltp_insert_new_frame(s);
+
put_bits(&s->pb, 3, TYPE_END);
flush_put_bits(&s->pb);
avctx->frame_bits = put_bits_count(&s->pb);
+ s->lambda_sum += s->lambda;
+ s->lambda_count++;
if (!frame)
s->last_frame++;
{
AACEncContext *s = avctx->priv_data;
+ av_log(avctx, AV_LOG_INFO, "Qavg: %.3f\n", s->lambda_sum / s->lambda_count);
+
ff_mdct_end(&s->mdct1024);
ff_mdct_end(&s->mdct128);
ff_psy_end(&s->psy);
uint8_t grouping[AAC_MAX_CHANNELS];
int lengths[2];
+ s->channels = avctx->channels;
+ s->chan_map = aac_chan_configs[s->channels-1];
+ s->random_state = 0x1f2e3d4c;
+ s->lambda = avctx->global_quality > 0 ? avctx->global_quality : 120;
+ avctx->extradata_size = 5;
avctx->frame_size = 1024;
+ avctx->initial_padding = 1024;
+ avctx->bit_rate = (int)FFMIN(
+ 6144 * s->channels / 1024.0 * avctx->sample_rate,
+ avctx->bit_rate);
+ avctx->profile = avctx->profile == FF_PROFILE_UNKNOWN ? FF_PROFILE_AAC_LOW :
+ avctx->profile;
for (i = 0; i < 16; i++)
if (avctx->sample_rate == avpriv_mpeg4audio_sample_rates[i])
break;
+ s->samplerate_index = i;
- s->channels = avctx->channels;
-
- ERROR_IF(i == 16 || i >= ff_aac_swb_size_1024_len || i >= ff_aac_swb_size_128_len,
+ ERROR_IF(s->samplerate_index == 16 ||
+ s->samplerate_index >= ff_aac_swb_size_1024_len ||
+ s->samplerate_index >= ff_aac_swb_size_128_len,
"Unsupported sample rate %d\n", avctx->sample_rate);
- ERROR_IF(s->channels > AAC_MAX_CHANNELS,
+ ERROR_IF(s->channels > AAC_MAX_CHANNELS || s->channels == 7,
"Unsupported number of channels: %d\n", s->channels);
WARN_IF(1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * s->channels,
"Too many bits per frame requested, clamping to max\n");
- if (avctx->profile == FF_PROFILE_AAC_MAIN) {
+
+ for (i = 0; i < FF_ARRAY_ELEMS(aacenc_profiles); i++)
+ if (avctx->profile == aacenc_profiles[i])
+ break;
+ ERROR_IF(i == FF_ARRAY_ELEMS(aacenc_profiles),
+ "Unsupported encoding profile: %d\n", avctx->profile);
+ if (avctx->profile == FF_PROFILE_MPEG2_AAC_LOW) {
+ avctx->profile = FF_PROFILE_AAC_LOW;
+ ERROR_IF(s->options.pred,
+ "Main prediction unavailable in the \"mpeg2_aac_low\" profile\n");
+ ERROR_IF(s->options.ltp,
+ "LTP prediction unavailable in the \"mpeg2_aac_low\" profile\n");
+ WARN_IF(s->options.pns,
+ "PNS unavailable in the \"mpeg2_aac_low\" profile, turning off\n");
+ s->options.pns = 0;
+ } else if (avctx->profile == FF_PROFILE_AAC_LTP) {
+ s->options.ltp = 1;
+ ERROR_IF(s->options.pred,
+ "Main prediction unavailable in the \"aac_ltp\" profile\n");
+ } else if (avctx->profile == FF_PROFILE_AAC_MAIN) {
s->options.pred = 1;
- } else if ((avctx->profile == FF_PROFILE_AAC_LOW ||
- avctx->profile == FF_PROFILE_UNKNOWN) && s->options.pred) {
- s->profile = 0; /* Main */
- WARN_IF(1, "Prediction requested, changing profile to AAC-Main\n");
- } else if (avctx->profile == FF_PROFILE_AAC_LOW ||
- avctx->profile == FF_PROFILE_UNKNOWN) {
- s->profile = 1; /* Low */
- } else {
- ERROR_IF(1, "Unsupported profile %d\n", avctx->profile);
+ ERROR_IF(s->options.ltp,
+ "LTP prediction unavailable in the \"aac_main\" profile\n");
+ } else if (s->options.ltp) {
+ avctx->profile = FF_PROFILE_AAC_LTP;
+ WARN_IF(1,
+ "Chainging profile to \"aac_ltp\"\n");
+ ERROR_IF(s->options.pred,
+ "Main prediction unavailable in the \"aac_ltp\" profile\n");
+ } else if (s->options.pred) {
+ avctx->profile = FF_PROFILE_AAC_MAIN;
+ WARN_IF(1,
+ "Chainging profile to \"aac_main\"\n");
+ ERROR_IF(s->options.pred,
+ "LTP prediction unavailable in the \"aac_main\" profile\n");
}
+ s->profile = avctx->profile;
+ s->coder = &ff_aac_coders[s->options.coder];
- if (s->options.aac_coder != AAC_CODER_TWOLOOP) {
+ if (s->options.coder != AAC_CODER_TWOLOOP) {
s->options.intensity_stereo = 0;
s->options.pns = 0;
}
- avctx->bit_rate = (int)FFMIN(
- 6144 * s->channels / 1024.0 * avctx->sample_rate,
- avctx->bit_rate);
-
- s->samplerate_index = i;
-
- s->chan_map = aac_chan_configs[s->channels-1];
-
if ((ret = dsp_init(avctx, s)) < 0)
goto fail;
if ((ret = alloc_buffers(avctx, s)) < 0)
goto fail;
- avctx->extradata_size = 5;
put_audio_specific_config(avctx);
- sizes[0] = ff_aac_swb_size_1024[i];
- sizes[1] = ff_aac_swb_size_128[i];
- lengths[0] = ff_aac_num_swb_1024[i];
- lengths[1] = ff_aac_num_swb_128[i];
+ sizes[0] = ff_aac_swb_size_1024[s->samplerate_index];
+ sizes[1] = ff_aac_swb_size_128[s->samplerate_index];
+ lengths[0] = ff_aac_num_swb_1024[s->samplerate_index];
+ lengths[1] = ff_aac_num_swb_128[s->samplerate_index];
for (i = 0; i < s->chan_map[0]; i++)
grouping[i] = s->chan_map[i + 1] == TYPE_CPE;
if ((ret = ff_psy_init(&s->psy, avctx, 2, sizes, lengths,
s->chan_map[0], grouping)) < 0)
goto fail;
s->psypp = ff_psy_preprocess_init(avctx);
- s->coder = &ff_aac_coders[s->options.aac_coder];
ff_lpc_init(&s->lpc, 2*avctx->frame_size, TNS_MAX_ORDER, FF_LPC_TYPE_LEVINSON);
if (HAVE_MIPSDSPR1)
ff_aac_coder_init_mips(s);
- s->lambda = avctx->global_quality > 0 ? avctx->global_quality : 120;
- s->random_state = 0x1f2e3d4c;
-
ff_aac_tableinit();
- avctx->initial_padding = 1024;
ff_af_queue_init(avctx, &s->afq);
return 0;
#define AACENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
static const AVOption aacenc_options[] = {
- {"stereo_mode", "Stereo coding method", offsetof(AACEncContext, options.stereo_mode), AV_OPT_TYPE_INT, {.i64 = 0}, -1, 1, AACENC_FLAGS, "stereo_mode"},
- {"auto", "Selected by the Encoder", 0, AV_OPT_TYPE_CONST, {.i64 = -1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
- {"ms_off", "Disable Mid/Side coding", 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
- {"ms_force", "Force Mid/Side for the whole frame if possible", 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
- {"aac_coder", "Coding algorithm", offsetof(AACEncContext, options.aac_coder), AV_OPT_TYPE_INT, {.i64 = AAC_CODER_TWOLOOP}, 0, AAC_CODER_NB-1, AACENC_FLAGS, "aac_coder"},
- {"faac", "FAAC-inspired method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_FAAC}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
- {"anmr", "ANMR method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_ANMR}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
- {"twoloop", "Two loop searching method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_TWOLOOP}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
- {"fast", "Constant quantizer", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_FAST}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
- {"aac_pns", "Perceptual Noise Substitution", offsetof(AACEncContext, options.pns), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, AACENC_FLAGS},
- {"aac_is", "Intensity stereo coding", offsetof(AACEncContext, options.intensity_stereo), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, AACENC_FLAGS},
- {"aac_tns", "Temporal noise shaping", offsetof(AACEncContext, options.tns), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AACENC_FLAGS},
- {"aac_pred", "AAC-Main prediction", offsetof(AACEncContext, options.pred), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AACENC_FLAGS},
+ {"aac_coder", "Coding algorithm", offsetof(AACEncContext, options.coder), AV_OPT_TYPE_INT, {.i64 = AAC_CODER_TWOLOOP}, -1, AAC_CODER_NB-1, AACENC_FLAGS, "coder"},
+ {"faac", "FAAC-inspired method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_FAAC}, INT_MIN, INT_MAX, AACENC_FLAGS, "coder"},
+ {"anmr", "ANMR method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_ANMR}, INT_MIN, INT_MAX, AACENC_FLAGS, "coder"},
+ {"twoloop", "Two loop searching method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_TWOLOOP}, INT_MIN, INT_MAX, AACENC_FLAGS, "coder"},
+ {"fast", "Constant quantizer", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_FAST}, INT_MIN, INT_MAX, AACENC_FLAGS, "coder"},
+ {"aac_ms", "Force M/S stereo coding", offsetof(AACEncContext, options.mid_side), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, AACENC_FLAGS},
+ {"aac_is", "Intensity stereo coding", offsetof(AACEncContext, options.intensity_stereo), AV_OPT_TYPE_BOOL, {.i64 = 1}, -1, 1, AACENC_FLAGS},
+ {"aac_pns", "Perceptual noise substitution", offsetof(AACEncContext, options.pns), AV_OPT_TYPE_BOOL, {.i64 = 1}, -1, 1, AACENC_FLAGS},
+ {"aac_tns", "Temporal noise shaping", offsetof(AACEncContext, options.tns), AV_OPT_TYPE_BOOL, {.i64 = 0}, -1, 1, AACENC_FLAGS},
+ {"aac_ltp", "Long term prediction", offsetof(AACEncContext, options.ltp), AV_OPT_TYPE_BOOL, {.i64 = 0}, -1, 1, AACENC_FLAGS},
+ {"aac_pred", "AAC-Main prediction", offsetof(AACEncContext, options.pred), AV_OPT_TYPE_BOOL, {.i64 = 0}, -1, 1, AACENC_FLAGS},
{NULL}
};
projects / ffmpeg / blobdiff