summaryrefslogtreecommitdiff
path: root/libavcodec/aacenc.c
diff options
context:
space:
mode:
Diffstat (limited to 'libavcodec/aacenc.c')
-rw-r--r--libavcodec/aacenc.c799
1 files changed, 569 insertions, 230 deletions
diff --git a/libavcodec/aacenc.c b/libavcodec/aacenc.c
index b7f60fb872..4d0abb107f 100644
--- a/libavcodec/aacenc.c
+++ b/libavcodec/aacenc.c
@@ -2,20 +2,20 @@
* AAC encoder
* Copyright (C) 2008 Konstantin Shishkov
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
@@ -27,9 +27,10 @@
/***********************************
* TODOs:
* add sane pulse detection
- * add temporal noise shaping
***********************************/
+#include "libavutil/libm.h"
+#include "libavutil/thread.h"
#include "libavutil/float_dsp.h"
#include "libavutil/opt.h"
#include "avcodec.h"
@@ -42,143 +43,93 @@
#include "aac.h"
#include "aactab.h"
#include "aacenc.h"
+#include "aacenctab.h"
+#include "aacenc_utils.h"
#include "psymodel.h"
-#define AAC_MAX_CHANNELS 6
+static AVOnce aac_table_init = AV_ONCE_INIT;
-#define ERROR_IF(cond, ...) \
- if (cond) { \
- av_log(avctx, AV_LOG_ERROR, __VA_ARGS__); \
- return AVERROR(EINVAL); \
+static void put_pce(PutBitContext *pb, AVCodecContext *avctx)
+{
+ int i, j;
+ AACEncContext *s = avctx->priv_data;
+ AACPCEInfo *pce = &s->pce;
+ const int bitexact = avctx->flags & AV_CODEC_FLAG_BITEXACT;
+ const char *aux_data = bitexact ? "Lavc" : LIBAVCODEC_IDENT;
+
+ put_bits(pb, 4, 0);
+
+ put_bits(pb, 2, avctx->profile);
+ put_bits(pb, 4, s->samplerate_index);
+
+ put_bits(pb, 4, pce->num_ele[0]); /* Front */
+ put_bits(pb, 4, pce->num_ele[1]); /* Side */
+ put_bits(pb, 4, pce->num_ele[2]); /* Back */
+ put_bits(pb, 2, pce->num_ele[3]); /* LFE */
+ put_bits(pb, 3, 0); /* Assoc data */
+ put_bits(pb, 4, 0); /* CCs */
+
+ put_bits(pb, 1, 0); /* Stereo mixdown */
+ put_bits(pb, 1, 0); /* Mono mixdown */
+ put_bits(pb, 1, 0); /* Something else */
+
+ for (i = 0; i < 4; i++) {
+ for (j = 0; j < pce->num_ele[i]; j++) {
+ if (i < 3)
+ put_bits(pb, 1, pce->pairing[i][j]);
+ put_bits(pb, 4, pce->index[i][j]);
+ }
}
-float ff_aac_pow34sf_tab[428];
-
-static const uint8_t swb_size_1024_96[] = {
- 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8,
- 12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44,
- 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
-};
-
-static const uint8_t swb_size_1024_64[] = {
- 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8,
- 12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36,
- 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40
-};
-
-static const uint8_t swb_size_1024_48[] = {
- 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
- 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
- 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32,
- 96
-};
-
-static const uint8_t swb_size_1024_32[] = {
- 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
- 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
- 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
-};
-
-static const uint8_t swb_size_1024_24[] = {
- 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28,
- 32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64
-};
-
-static const uint8_t swb_size_1024_16[] = {
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28,
- 32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64
-};
-
-static const uint8_t swb_size_1024_8[] = {
- 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
- 16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28,
- 32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80
-};
-
-static const uint8_t * const swb_size_1024[] = {
- swb_size_1024_96, swb_size_1024_96, swb_size_1024_64,
- swb_size_1024_48, swb_size_1024_48, swb_size_1024_32,
- swb_size_1024_24, swb_size_1024_24, swb_size_1024_16,
- swb_size_1024_16, swb_size_1024_16, swb_size_1024_8
-};
-
-static const uint8_t swb_size_128_96[] = {
- 4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36
-};
-
-static const uint8_t swb_size_128_48[] = {
- 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16
-};
-
-static const uint8_t swb_size_128_24[] = {
- 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20
-};
-
-static const uint8_t swb_size_128_16[] = {
- 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20
-};
-
-static const uint8_t swb_size_128_8[] = {
- 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20
-};
-
-static const uint8_t * const swb_size_128[] = {
- /* the last entry on the following row is swb_size_128_64 but is a
- duplicate of swb_size_128_96 */
- swb_size_128_96, swb_size_128_96, swb_size_128_96,
- swb_size_128_48, swb_size_128_48, swb_size_128_48,
- swb_size_128_24, swb_size_128_24, swb_size_128_16,
- swb_size_128_16, swb_size_128_16, swb_size_128_8
-};
-
-/** default channel configurations */
-static const uint8_t aac_chan_configs[6][5] = {
- {1, TYPE_SCE}, // 1 channel - single channel element
- {1, TYPE_CPE}, // 2 channels - channel pair
- {2, TYPE_SCE, TYPE_CPE}, // 3 channels - center + stereo
- {3, TYPE_SCE, TYPE_CPE, TYPE_SCE}, // 4 channels - front center + stereo + back center
- {3, TYPE_SCE, TYPE_CPE, TYPE_CPE}, // 5 channels - front center + stereo + back stereo
- {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE
-};
-
-/**
- * Table to remap channels from Libav's default order to AAC order.
- */
-static const uint8_t aac_chan_maps[AAC_MAX_CHANNELS][AAC_MAX_CHANNELS] = {
- { 0 },
- { 0, 1 },
- { 2, 0, 1 },
- { 2, 0, 1, 3 },
- { 2, 0, 1, 3, 4 },
- { 2, 0, 1, 4, 5, 3 },
-};
+ avpriv_align_put_bits(pb);
+ put_bits(pb, 8, strlen(aux_data));
+ avpriv_put_string(pb, aux_data, 0);
+}
/**
* Make AAC audio config object.
* @see 1.6.2.1 "Syntax - AudioSpecificConfig"
*/
-static void put_audio_specific_config(AVCodecContext *avctx)
+static int put_audio_specific_config(AVCodecContext *avctx)
{
PutBitContext pb;
AACEncContext *s = avctx->priv_data;
+ int channels = (!s->needs_pce)*(s->channels - (s->channels == 8 ? 1 : 0));
+ const int max_size = 32;
+
+ avctx->extradata = av_mallocz(max_size);
+ if (!avctx->extradata)
+ return AVERROR(ENOMEM);
- init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8);
- put_bits(&pb, 5, 2); //object type - AAC-LC
+ init_put_bits(&pb, avctx->extradata, max_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
put_bits(&pb, 1, 0); //is not extension
+ if (s->needs_pce)
+ put_pce(&pb, avctx);
//Explicitly Mark SBR absent
put_bits(&pb, 11, 0x2b7); //sync extension
put_bits(&pb, 5, AOT_SBR);
put_bits(&pb, 1, 0);
flush_put_bits(&pb);
+ avctx->extradata_size = put_bits_count(&pb) >> 3;
+
+ return 0;
+}
+
+void ff_quantize_band_cost_cache_init(struct AACEncContext *s)
+{
+ ++s->quantize_band_cost_cache_generation;
+ if (s->quantize_band_cost_cache_generation == 0) {
+ memset(s->quantize_band_cost_cache, 0, sizeof(s->quantize_band_cost_cache));
+ s->quantize_band_cost_cache_generation = 1;
+ }
}
#define WINDOW_FUNC(type) \
@@ -250,16 +201,17 @@ static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
float *audio)
{
int i;
- float *output = sce->ret_buf;
+ const float *output = sce->ret_buf;
- apply_window[sce->ics.window_sequence[0]](&s->fdsp, sce, audio);
+ apply_window[sce->ics.window_sequence[0]](s->fdsp, sce, audio);
if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE)
s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
else
for (i = 0; i < 1024; i += 128)
- s->mdct128.mdct_calc(&s->mdct128, sce->coeffs + i, output + i*2);
+ s->mdct128.mdct_calc(&s->mdct128, &sce->coeffs[i], output + i*2);
memcpy(audio, audio + 1024, sizeof(audio[0]) * 1024);
+ memcpy(sce->pcoeffs, sce->coeffs, sizeof(sce->pcoeffs));
}
/**
@@ -275,7 +227,7 @@ static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
put_bits(&s->pb, 1, info->use_kb_window[0]);
if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
put_bits(&s->pb, 6, info->max_sfb);
- put_bits(&s->pb, 1, 0); // no prediction
+ put_bits(&s->pb, 1, !!info->predictor_present);
} else {
put_bits(&s->pb, 4, info->max_sfb);
for (w = 1; w < 8; w++)
@@ -304,27 +256,18 @@ static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
static void adjust_frame_information(ChannelElement *cpe, int chans)
{
int i, w, w2, g, ch;
- int start, maxsfb, cmaxsfb;
+ int maxsfb, cmaxsfb;
for (ch = 0; ch < chans; ch++) {
IndividualChannelStream *ics = &cpe->ch[ch].ics;
- start = 0;
maxsfb = 0;
cpe->ch[ch].pulse.num_pulse = 0;
- for (w = 0; w < ics->num_windows*16; w += 16) {
- for (g = 0; g < ics->num_swb; g++) {
- //apply M/S
- if (cpe->common_window && !ch && cpe->ms_mask[w + g]) {
- for (i = 0; i < ics->swb_sizes[g]; i++) {
- cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
- cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
- }
- }
- start += ics->swb_sizes[g];
+ for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
+ for (w2 = 0; w2 < ics->group_len[w]; w2++) {
+ for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w*16+cmaxsfb-1]; cmaxsfb--)
+ ;
+ maxsfb = FFMAX(maxsfb, cmaxsfb);
}
- for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--)
- ;
- maxsfb = FFMAX(maxsfb, cmaxsfb);
}
ics->max_sfb = maxsfb;
@@ -360,6 +303,67 @@ static void adjust_frame_information(ChannelElement *cpe, int chans)
}
}
+static void apply_intensity_stereo(ChannelElement *cpe)
+{
+ int w, w2, g, i;
+ IndividualChannelStream *ics = &cpe->ch[0].ics;
+ if (!cpe->common_window)
+ return;
+ for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
+ for (w2 = 0; w2 < ics->group_len[w]; w2++) {
+ int start = (w+w2) * 128;
+ for (g = 0; g < ics->num_swb; g++) {
+ int p = -1 + 2 * (cpe->ch[1].band_type[w*16+g] - 14);
+ float scale = cpe->ch[0].is_ener[w*16+g];
+ if (!cpe->is_mask[w*16 + g]) {
+ 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;
+ cpe->ch[1].coeffs[start+i] = 0.0f;
+ }
+ start += ics->swb_sizes[g];
+ }
+ }
+ }
+}
+
+static void apply_mid_side_stereo(ChannelElement *cpe)
+{
+ int w, w2, g, i;
+ IndividualChannelStream *ics = &cpe->ch[0].ics;
+ if (!cpe->common_window)
+ return;
+ for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
+ for (w2 = 0; w2 < ics->group_len[w]; w2++) {
+ int start = (w+w2) * 128;
+ for (g = 0; g < ics->num_swb; g++) {
+ /* ms_mask can be used for other purposes in PNS and I/S,
+ * so must not apply M/S if any band uses either, even if
+ * ms_mask is set.
+ */
+ if (!cpe->ms_mask[w*16 + g] || cpe->is_mask[w*16 + g]
+ || cpe->ch[0].band_type[w*16 + g] >= NOISE_BT
+ || cpe->ch[1].band_type[w*16 + g] >= NOISE_BT) {
+ start += ics->swb_sizes[g];
+ continue;
+ }
+ for (i = 0; i < ics->swb_sizes[g]; i++) {
+ float L = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) * 0.5f;
+ float R = L - cpe->ch[1].coeffs[start+i];
+ cpe->ch[0].coeffs[start+i] = L;
+ cpe->ch[1].coeffs[start+i] = R;
+ }
+ start += ics->swb_sizes[g];
+ }
+ }
+ }
+}
+
/**
* Encode scalefactor band coding type.
*/
@@ -367,6 +371,9 @@ static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
{
int w;
+ if (s->coder->set_special_band_scalefactors)
+ s->coder->set_special_band_scalefactors(s, sce);
+
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
}
@@ -377,16 +384,30 @@ static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s,
SingleChannelElement *sce)
{
- int off = sce->sf_idx[0], diff;
+ int diff, off_sf = sce->sf_idx[0], off_pns = sce->sf_idx[0] - NOISE_OFFSET;
+ int off_is = 0, noise_flag = 1;
int i, w;
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
for (i = 0; i < sce->ics.max_sfb; i++) {
if (!sce->zeroes[w*16 + i]) {
- diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO;
- if (diff < 0 || diff > 120)
- av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n");
- off = sce->sf_idx[w*16 + i];
+ if (sce->band_type[w*16 + i] == NOISE_BT) {
+ diff = sce->sf_idx[w*16 + i] - off_pns;
+ off_pns = sce->sf_idx[w*16 + i];
+ if (noise_flag-- > 0) {
+ put_bits(&s->pb, NOISE_PRE_BITS, diff + NOISE_PRE);
+ continue;
+ }
+ } else if (sce->band_type[w*16 + i] == INTENSITY_BT ||
+ sce->band_type[w*16 + i] == INTENSITY_BT2) {
+ diff = sce->sf_idx[w*16 + i] - off_is;
+ off_is = sce->sf_idx[w*16 + i];
+ } else {
+ diff = sce->sf_idx[w*16 + i] - off_sf;
+ off_sf = sce->sf_idx[w*16 + i];
+ }
+ diff += SCALE_DIFF_ZERO;
+ av_assert0(diff >= 0 && diff <= 120);
put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
}
}
@@ -426,18 +447,41 @@ static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
start += sce->ics.swb_sizes[i];
continue;
}
- for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++)
- s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128,
- sce->ics.swb_sizes[i],
+ for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++) {
+ s->coder->quantize_and_encode_band(s, &s->pb,
+ &sce->coeffs[start + w2*128],
+ NULL, sce->ics.swb_sizes[i],
sce->sf_idx[w*16 + i],
sce->band_type[w*16 + i],
- s->lambda);
+ s->lambda,
+ sce->ics.window_clipping[w]);
+ }
start += sce->ics.swb_sizes[i];
}
}
}
/**
+ * Downscale spectral coefficients for near-clipping windows to avoid artifacts
+ */
+static void avoid_clipping(AACEncContext *s, SingleChannelElement *sce)
+{
+ int start, i, j, w;
+
+ if (sce->ics.clip_avoidance_factor < 1.0f) {
+ for (w = 0; w < sce->ics.num_windows; w++) {
+ start = 0;
+ for (i = 0; i < sce->ics.max_sfb; i++) {
+ float *swb_coeffs = &sce->coeffs[start + w*128];
+ for (j = 0; j < sce->ics.swb_sizes[i]; j++)
+ swb_coeffs[j] *= sce->ics.clip_avoidance_factor;
+ start += sce->ics.swb_sizes[i];
+ }
+ }
+ }
+}
+
+/**
* Encode one channel of audio data.
*/
static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
@@ -445,12 +489,19 @@ static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
int common_window)
{
put_bits(&s->pb, 8, sce->sf_idx[0]);
- if (!common_window)
+ if (!common_window) {
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);
encode_pulses(s, &sce->pulse);
- put_bits(&s->pb, 1, 0); //tns
+ put_bits(&s->pb, 1, !!sce->tns.present);
+ if (s->coder->encode_tns_info)
+ s->coder->encode_tns_info(s, sce);
put_bits(&s->pb, 1, 0); //ssr
encode_spectral_coeffs(s, sce);
return 0;
@@ -478,13 +529,13 @@ static void put_bitstream_info(AACEncContext *s, const char *name)
/*
* Copy input samples.
- * Channels are reordered from Libav's default order to AAC order.
+ * Channels are reordered from libavcodec's default order to AAC order.
*/
static void copy_input_samples(AACEncContext *s, const AVFrame *frame)
{
int ch;
int end = 2048 + (frame ? frame->nb_samples : 0);
- const uint8_t *channel_map = aac_chan_maps[s->channels - 1];
+ const uint8_t *channel_map = s->reorder_map;
/* copy and remap input samples */
for (ch = 0; ch < s->channels; ch++) {
@@ -508,18 +559,21 @@ static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
AACEncContext *s = avctx->priv_data;
float **samples = s->planar_samples, *samples2, *la, *overlap;
ChannelElement *cpe;
- int i, ch, w, g, chans, tag, start_ch, ret;
+ 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];
- int frame_bits;
FFPsyWindowInfo windows[AAC_MAX_CHANNELS];
- if (s->last_frame == 2)
- return 0;
-
/* add current frame to queue */
if (frame) {
if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
return ret;
+ } else {
+ if (!s->afq.remaining_samples || (!s->afq.frame_alloc && !s->afq.frame_count))
+ return 0;
}
copy_input_samples(s, frame);
@@ -536,18 +590,22 @@ static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
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;
- overlap = &samples[cur_channel][0];
+ int k;
+ float clip_avoidance_factor;
+ sce = &cpe->ch[ch];
+ ics = &sce->ics;
+ s->cur_channel = start_ch + ch;
+ overlap = &samples[s->cur_channel][0];
samples2 = overlap + 1024;
la = samples2 + (448+64);
if (!frame)
la = NULL;
if (tag == TYPE_LFE) {
- wi[ch].window_type[0] = ONLY_LONG_SEQUENCE;
+ wi[ch].window_type[0] = wi[ch].window_type[1] = ONLY_LONG_SEQUENCE;
wi[ch].window_shape = 0;
wi[ch].num_windows = 1;
wi[ch].grouping[0] = 1;
+ wi[ch].clipping[0] = 0;
/* Only the lowest 12 coefficients are used in a LFE channel.
* The expression below results in only the bottom 8 coefficients
@@ -555,7 +613,7 @@ static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
*/
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];
@@ -565,24 +623,71 @@ static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
ics->num_windows = wi[ch].num_windows;
ics->swb_sizes = s->psy.bands [ics->num_windows == 8];
ics->num_swb = tag == TYPE_LFE ? ics->num_swb : s->psy.num_bands[ics->num_windows == 8];
+ ics->max_sfb = FFMIN(ics->max_sfb, ics->num_swb);
+ ics->swb_offset = wi[ch].window_type[0] == EIGHT_SHORT_SEQUENCE ?
+ ff_swb_offset_128 [s->samplerate_index]:
+ ff_swb_offset_1024[s->samplerate_index];
+ ics->tns_max_bands = wi[ch].window_type[0] == EIGHT_SHORT_SEQUENCE ?
+ ff_tns_max_bands_128 [s->samplerate_index]:
+ ff_tns_max_bands_1024[s->samplerate_index];
+
for (w = 0; w < ics->num_windows; w++)
ics->group_len[w] = wi[ch].grouping[w];
- apply_window_and_mdct(s, &cpe->ch[ch], overlap);
+ /* Calculate input sample maximums and evaluate clipping risk */
+ clip_avoidance_factor = 0.0f;
+ for (w = 0; w < ics->num_windows; w++) {
+ const float *wbuf = overlap + w * 128;
+ const int wlen = 2048 / ics->num_windows;
+ float max = 0;
+ int j;
+ /* mdct input is 2 * output */
+ for (j = 0; j < wlen; j++)
+ max = FFMAX(max, fabsf(wbuf[j]));
+ wi[ch].clipping[w] = max;
+ }
+ for (w = 0; w < ics->num_windows; w++) {
+ if (wi[ch].clipping[w] > CLIP_AVOIDANCE_FACTOR) {
+ ics->window_clipping[w] = 1;
+ clip_avoidance_factor = FFMAX(clip_avoidance_factor, wi[ch].clipping[w]);
+ } else {
+ ics->window_clipping[w] = 0;
+ }
+ }
+ if (clip_avoidance_factor > CLIP_AVOIDANCE_FACTOR) {
+ ics->clip_avoidance_factor = CLIP_AVOIDANCE_FACTOR / clip_avoidance_factor;
+ } else {
+ ics->clip_avoidance_factor = 1.0f;
+ }
+
+ 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);
+ }
+
+ for (k = 0; k < 1024; k++) {
+ if (!(fabs(cpe->ch[ch].coeffs[k]) < 1E16)) { // Ensure headroom for energy calculation
+ av_log(avctx, AV_LOG_ERROR, "Input contains (near) NaN/+-Inf\n");
+ return AVERROR(EINVAL);
+ }
+ }
+ avoid_clipping(s, sce);
}
start_ch += chans;
}
- if ((ret = ff_alloc_packet(avpkt, 768 * s->channels))) {
- av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
+ if ((ret = ff_alloc_packet2(avctx, avpkt, 8192 * s->channels, 0)) < 0)
return ret;
- }
-
+ frame_bits = its = 0;
do {
init_put_bits(&s->pb, avpkt->data, avpkt->size);
if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & AV_CODEC_FLAG_BITEXACT))
put_bitstream_info(s, LIBAVCODEC_IDENT);
start_ch = 0;
+ target_bits = 0;
memset(chan_el_counter, 0, sizeof(chan_el_counter));
for (i = 0; i < s->chan_map[0]; i++) {
FFPsyWindowInfo* wi = windows + start_ch;
@@ -590,16 +695,39 @@ static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
tag = s->chan_map[i+1];
chans = tag == TYPE_CPE ? 2 : 1;
cpe = &s->cpe[i];
+ cpe->common_window = 0;
+ memset(cpe->is_mask, 0, sizeof(cpe->is_mask));
+ memset(cpe->ms_mask, 0, sizeof(cpe->ms_mask));
put_bits(&s->pb, 3, tag);
put_bits(&s->pb, 4, chan_el_counter[tag]++);
- for (ch = 0; ch < chans; ch++)
- coeffs[ch] = cpe->ch[ch].coeffs;
+ for (ch = 0; ch < chans; ch++) {
+ sce = &cpe->ch[ch];
+ coeffs[ch] = sce->coeffs;
+ sce->ics.predictor_present = 0;
+ 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)
+ sce->band_type[w] = 0;
+ }
+ s->psy.bitres.alloc = -1;
+ s->psy.bitres.bits = s->last_frame_pb_count / s->channels;
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
+ * (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);
}
- cpe->common_window = 0;
if (chans > 1
&& wi[0].window_type[0] == wi[1].window_type[0]
&& wi[0].window_shape == wi[1].window_shape) {
@@ -612,23 +740,73 @@ static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
}
}
}
+ for (ch = 0; ch < chans; ch++) { /* TNS and PNS */
+ sce = &cpe->ch[ch];
+ s->cur_channel = start_ch + ch;
+ if (s->options.tns && s->coder->search_for_tns)
+ s->coder->search_for_tns(s, sce);
+ if (s->options.tns && s->coder->apply_tns_filt)
+ s->coder->apply_tns_filt(s, sce);
+ if (sce->tns.present)
+ tns_mode = 1;
+ if (s->options.pns && s->coder->search_for_pns)
+ s->coder->search_for_pns(s, avctx, sce);
+ }
s->cur_channel = start_ch;
- if (s->options.stereo_mode && cpe->common_window) {
- if (s->options.stereo_mode > 0) {
- IndividualChannelStream *ics = &cpe->ch[0].ics;
- for (w = 0; w < ics->num_windows; w += ics->group_len[w])
- for (g = 0; g < ics->num_swb; g++)
- cpe->ms_mask[w*16+g] = 1;
- } else if (s->coder->search_for_ms) {
- s->coder->search_for_ms(s, cpe, s->lambda);
+ if (s->options.intensity_stereo) { /* Intensity Stereo */
+ if (s->coder->search_for_is)
+ s->coder->search_for_is(s, avctx, cpe);
+ if (cpe->is_mode) is_mode = 1;
+ apply_intensity_stereo(cpe);
+ }
+ if (s->options.pred) { /* Prediction */
+ for (ch = 0; ch < chans; ch++) {
+ sce = &cpe->ch[ch];
+ s->cur_channel = start_ch + ch;
+ if (s->options.pred && s->coder->search_for_pred)
+ s->coder->search_for_pred(s, sce);
+ if (cpe->ch[ch].ics.predictor_present) pred_mode = 1;
+ }
+ 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;
+ if (s->options.pred && s->coder->apply_main_pred)
+ s->coder->apply_main_pred(s, sce);
}
+ s->cur_channel = start_ch;
+ }
+ 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));
+ 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;
}
}
for (ch = 0; ch < chans; ch++) {
@@ -638,34 +816,77 @@ static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
start_ch += chans;
}
- frame_bits = put_bits_count(&s->pb);
- if (frame_bits <= 6144 * s->channels - 3) {
- s->psy.bitres.bits = frame_bits / s->channels;
+ if (avctx->flags & AV_CODEC_FLAG_QSCALE) {
+ /* When using a constant Q-scale, don't mess with lambda */
break;
}
- s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
+ /* rate control stuff
+ * 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);
+ 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_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)rate_bits) / frame_bits;
+
+ if (frame_bits >= too_few_bits && frame_bits <= too_many_bits) {
+ /*
+ * This path is for steady-state Q-scale tracking
+ * When frame bits fall within the stable range, we still need to adjust
+ * lambda to maintain it like so in a stable fashion (large jumps in lambda
+ * create artifacts and should be avoided), but slowly
+ */
+ ratio = sqrtf(sqrtf(ratio));
+ ratio = av_clipf(ratio, 0.9f, 1.1f);
+ } else {
+ /* Not so fast though */
+ ratio = sqrtf(ratio);
+ }
+ s->lambda = FFMIN(s->lambda * ratio, 65536.f);
+ /* Keep iterating if we must reduce and lambda is in the sky */
+ if (ratio > 0.9f && ratio < 1.1f) {
+ break;
+ } else {
+ if (is_mode || ms_mode || tns_mode || pred_mode) {
+ for (i = 0; i < s->chan_map[0]; i++) {
+ // Must restore coeffs
+ chans = tag == TYPE_CPE ? 2 : 1;
+ cpe = &s->cpe[i];
+ for (ch = 0; ch < chans; ch++)
+ memcpy(cpe->ch[ch].coeffs, cpe->ch[ch].pcoeffs, sizeof(cpe->ch[ch].coeffs));
+ }
+ }
+ its++;
+ }
+ } else {
+ break;
+ }
} 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);
- frame_bits = put_bits_count(&s->pb);
-#if FF_API_STAT_BITS
-FF_DISABLE_DEPRECATION_WARNINGS
- avctx->frame_bits = frame_bits;
-FF_ENABLE_DEPRECATION_WARNINGS
-#endif
-
- // rate control stuff
- if (!(avctx->flags & AV_CODEC_FLAG_QSCALE)) {
- float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
- s->lambda *= ratio;
- s->lambda = FFMIN(s->lambda, 65536.f);
- }
- if (!frame)
- s->last_frame++;
+ s->last_frame_pb_count = put_bits_count(&s->pb);
+
+ s->lambda_sum += s->lambda;
+ s->lambda_count++;
ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
&avpkt->duration);
@@ -679,13 +900,17 @@ static av_cold int aac_encode_end(AVCodecContext *avctx)
{
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);
+ ff_lpc_end(&s->lpc);
if (s->psypp)
ff_psy_preprocess_end(s->psypp);
av_freep(&s->buffer.samples);
av_freep(&s->cpe);
+ av_freep(&s->fdsp);
ff_af_queue_close(&s->afq);
return 0;
}
@@ -694,7 +919,9 @@ static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
{
int ret = 0;
- avpriv_float_dsp_init(&s->fdsp, avctx->flags & AV_CODEC_FLAG_BITEXACT);
+ s->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
+ if (!s->fdsp)
+ return AVERROR(ENOMEM);
// window init
ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
@@ -702,9 +929,9 @@ static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
ff_init_ff_sine_windows(10);
ff_init_ff_sine_windows(7);
- if (ret = ff_mdct_init(&s->mdct1024, 11, 0, 32768.0))
+ if ((ret = ff_mdct_init(&s->mdct1024, 11, 0, 32768.0)) < 0)
return ret;
- if (ret = ff_mdct_init(&s->mdct128, 8, 0, 32768.0))
+ if ((ret = ff_mdct_init(&s->mdct128, 8, 0, 32768.0)) < 0)
return ret;
return 0;
@@ -713,9 +940,8 @@ static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
static av_cold int alloc_buffers(AVCodecContext *avctx, AACEncContext *s)
{
int ch;
- FF_ALLOCZ_OR_GOTO(avctx, s->buffer.samples, 3 * 1024 * s->channels * sizeof(s->buffer.samples[0]), alloc_fail);
- FF_ALLOCZ_OR_GOTO(avctx, s->cpe, sizeof(ChannelElement) * s->chan_map[0], alloc_fail);
- FF_ALLOCZ_OR_GOTO(avctx, avctx->extradata, 5 + AV_INPUT_BUFFER_PADDING_SIZE, alloc_fail);
+ FF_ALLOCZ_ARRAY_OR_GOTO(avctx, s->buffer.samples, s->channels, 3 * 1024 * sizeof(s->buffer.samples[0]), alloc_fail);
+ FF_ALLOCZ_ARRAY_OR_GOTO(avctx, s->cpe, s->chan_map[0], sizeof(ChannelElement), alloc_fail);
for(ch = 0; ch < s->channels; ch++)
s->planar_samples[ch] = s->buffer.samples + 3 * 1024 * ch;
@@ -725,6 +951,11 @@ alloc_fail:
return AVERROR(ENOMEM);
}
+static av_cold void aac_encode_init_tables(void)
+{
+ ff_aac_tableinit();
+}
+
static av_cold int aac_encode_init(AVCodecContext *avctx)
{
AACEncContext *s = avctx->priv_data;
@@ -733,28 +964,117 @@ static av_cold int aac_encode_init(AVCodecContext *avctx)
uint8_t grouping[AAC_MAX_CHANNELS];
int lengths[2];
+ /* Constants */
+ s->last_frame_pb_count = 0;
avctx->frame_size = 1024;
+ avctx->initial_padding = 1024;
+ s->lambda = avctx->global_quality > 0 ? avctx->global_quality : 120;
- for (i = 0; i < 16; i++)
- if (avctx->sample_rate == avpriv_mpeg4audio_sample_rates[i])
+ /* Channel map and unspecified bitrate guessing */
+ s->channels = avctx->channels;
+
+ s->needs_pce = 1;
+ for (i = 0; i < FF_ARRAY_ELEMS(aac_normal_chan_layouts); i++) {
+ if (avctx->channel_layout == aac_normal_chan_layouts[i]) {
+ s->needs_pce = s->options.pce;
break;
+ }
+ }
- s->channels = avctx->channels;
+ if (s->needs_pce) {
+ char buf[64];
+ for (i = 0; i < FF_ARRAY_ELEMS(aac_pce_configs); i++)
+ if (avctx->channel_layout == aac_pce_configs[i].layout)
+ break;
+ av_get_channel_layout_string(buf, sizeof(buf), -1, avctx->channel_layout);
+ ERROR_IF(i == FF_ARRAY_ELEMS(aac_pce_configs), "Unsupported channel layout \"%s\"\n", buf);
+ av_log(avctx, AV_LOG_INFO, "Using a PCE to encode channel layout \"%s\"\n", buf);
+ s->pce = aac_pce_configs[i];
+ s->reorder_map = s->pce.reorder_map;
+ s->chan_map = s->pce.config_map;
+ } else {
+ s->reorder_map = aac_chan_maps[s->channels - 1];
+ s->chan_map = aac_chan_configs[s->channels - 1];
+ }
- ERROR_IF(i == 16,
+ if (!avctx->bit_rate) {
+ for (i = 1; i <= s->chan_map[0]; i++) {
+ avctx->bit_rate += s->chan_map[i] == TYPE_CPE ? 128000 : /* Pair */
+ s->chan_map[i] == TYPE_LFE ? 16000 : /* LFE */
+ 69000 ; /* SCE */
+ }
+ }
+
+ /* Samplerate */
+ for (i = 0; i < 16; i++)
+ if (avctx->sample_rate == avpriv_mpeg4audio_sample_rates[i])
+ break;
+ s->samplerate_index = i;
+ 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,
- "Unsupported number of channels: %d\n", s->channels);
- ERROR_IF(avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW,
- "Unsupported profile %d\n", avctx->profile);
- ERROR_IF(1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * s->channels,
- "Too many bits %f > %d per frame requested\n",
+
+ /* Bitrate limiting */
+ WARN_IF(1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * s->channels,
+ "Too many bits %f > %d per frame requested, clamping to max\n",
1024.0 * avctx->bit_rate / avctx->sample_rate,
6144 * s->channels);
+ avctx->bit_rate = (int64_t)FFMIN(6144 * s->channels / 1024.0 * avctx->sample_rate,
+ avctx->bit_rate);
+
+ /* Profile and option setting */
+ avctx->profile = avctx->profile == FF_PROFILE_UNKNOWN ? FF_PROFILE_AAC_LOW :
+ avctx->profile;
+ for (i = 0; i < FF_ARRAY_ELEMS(aacenc_profiles); i++)
+ if (avctx->profile == aacenc_profiles[i])
+ break;
+ 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;
+ 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.ltp,
+ "LTP prediction unavailable in the \"aac_main\" profile\n");
+ }
+ s->profile = avctx->profile;
+
+ /* Coder limitations */
+ s->coder = &ff_aac_coders[s->options.coder];
+ if (s->options.coder == AAC_CODER_ANMR) {
+ ERROR_IF(avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL,
+ "The ANMR coder is considered experimental, add -strict -2 to enable!\n");
+ s->options.intensity_stereo = 0;
+ s->options.pns = 0;
+ }
+ ERROR_IF(s->options.ltp && avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL,
+ "The LPT profile requires experimental compliance, add -strict -2 to enable!\n");
- s->samplerate_index = i;
-
- s->chan_map = aac_chan_configs[s->channels-1];
+ /* M/S introduces horrible artifacts with multichannel files, this is temporary */
+ if (s->channels > 3)
+ s->options.mid_side = 0;
if ((ret = dsp_init(avctx, s)) < 0)
goto fail;
@@ -762,29 +1082,34 @@ static av_cold int aac_encode_init(AVCodecContext *avctx)
if ((ret = alloc_buffers(avctx, s)) < 0)
goto fail;
- avctx->extradata_size = 5;
- put_audio_specific_config(avctx);
+ if ((ret = put_audio_specific_config(avctx)))
+ goto fail;
- sizes[0] = swb_size_1024[i];
- sizes[1] = 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[2];
+ ff_lpc_init(&s->lpc, 2*avctx->frame_size, TNS_MAX_ORDER, FF_LPC_TYPE_LEVINSON);
+ s->random_state = 0x1f2e3d4c;
- s->lambda = avctx->global_quality ? avctx->global_quality : 120;
+ s->abs_pow34 = abs_pow34_v;
+ s->quant_bands = quantize_bands;
- ff_aac_tableinit();
+ if (ARCH_X86)
+ ff_aac_dsp_init_x86(s);
- for (i = 0; i < 428; i++)
- ff_aac_pow34sf_tab[i] = sqrt(ff_aac_pow2sf_tab[i] * sqrt(ff_aac_pow2sf_tab[i]));
+ if (HAVE_MIPSDSP)
+ ff_aac_coder_init_mips(s);
+
+ if ((ret = ff_thread_once(&aac_table_init, &aac_encode_init_tables)) != 0)
+ return AVERROR_UNKNOWN;
- avctx->initial_padding = 1024;
ff_af_queue_init(avctx, &s->afq);
return 0;
@@ -795,10 +1120,17 @@ fail:
#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.coder), AV_OPT_TYPE_INT, {.i64 = AAC_CODER_FAST}, 0, AAC_CODER_NB-1, 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", "Default fast search", 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 = 1}, -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},
+ {"aac_pce", "Forces the use of PCEs", offsetof(AACEncContext, options.pce), AV_OPT_TYPE_BOOL, {.i64 = 0}, -1, 1, AACENC_FLAGS},
{NULL}
};
@@ -809,6 +1141,11 @@ static const AVClass aacenc_class = {
.version = LIBAVUTIL_VERSION_INT,
};
+static const AVCodecDefault aac_encode_defaults[] = {
+ { "b", "0" },
+ { NULL }
+};
+
AVCodec ff_aac_encoder = {
.name = "aac",
.long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
@@ -818,8 +1155,10 @@ AVCodec ff_aac_encoder = {
.init = aac_encode_init,
.encode2 = aac_encode_frame,
.close = aac_encode_end,
- .capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY |
- AV_CODEC_CAP_EXPERIMENTAL,
+ .defaults = aac_encode_defaults,
+ .supported_samplerates = mpeg4audio_sample_rates,
+ .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE,
+ .capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY,
.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
AV_SAMPLE_FMT_NONE },
.priv_class = &aacenc_class,
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-14 18:03:46 +0000