diff options
author | Reinhard Tartler <siretart@tauware.de> | 2010-07-28 20:14:28 +0000 |
---|---|---|
committer | Reinhard Tartler <siretart@tauware.de> | 2010-07-28 20:14:28 +0000 |
commit | ba90f93852a3a0f9f9edc014338d3ac381be14f3 (patch) | |
tree | f0d59aaf268531bf652a2c184149f1d96fd3ac56 | |
parent | 93883b0c390c87f85e6f7bb22428d789bc842b12 (diff) | |
download | ffmpeg-ba90f93852a3a0f9f9edc014338d3ac381be14f3.tar.gz |
Backport AAC-HE v2 from trunk
This patch has seen testing for a couple of weeks in ubuntu maverick and debian/experimental w/o negative feedback so far.
Originally committed as revision 24576 to svn://svn.ffmpeg.org/ffmpeg/branches/0.6
-rw-r--r-- | libavcodec/Makefile | 2 | ||||
-rw-r--r-- | libavcodec/aac.h | 13 | ||||
-rw-r--r-- | libavcodec/aac_tablegen.c | 39 | ||||
-rw-r--r-- | libavcodec/aac_tablegen.h | 42 | ||||
-rw-r--r-- | libavcodec/aac_tablegen_decl.h | 34 | ||||
-rw-r--r-- | libavcodec/aacdec.c (renamed from libavcodec/aac.c) | 174 | ||||
-rw-r--r-- | libavcodec/aacenc.c | 25 | ||||
-rw-r--r-- | libavcodec/aacenc.h | 2 | ||||
-rw-r--r-- | libavcodec/aacps.c | 1037 | ||||
-rw-r--r-- | libavcodec/aacps.h | 82 | ||||
-rw-r--r-- | libavcodec/aacps_tablegen.c | 93 | ||||
-rw-r--r-- | libavcodec/aacps_tablegen.h | 212 | ||||
-rw-r--r-- | libavcodec/aacpsdata.c | 163 | ||||
-rw-r--r-- | libavcodec/aacsbr.c | 157 | ||||
-rw-r--r-- | libavcodec/aactab.c | 127 | ||||
-rw-r--r-- | libavcodec/aactab.h | 7 | ||||
-rw-r--r-- | libavcodec/sbr.h | 6 |
17 files changed, 1908 insertions, 307 deletions
diff --git a/libavcodec/Makefile b/libavcodec/Makefile index 00e6b57fd3..4547ef60ca 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -43,7 +43,7 @@ OBJS-$(CONFIG_VAAPI) += vaapi.o OBJS-$(CONFIG_VDPAU) += vdpau.o # decoders/encoders/hardware accelerators -OBJS-$(CONFIG_AAC_DECODER) += aac.o aactab.o aacsbr.o +OBJS-$(CONFIG_AAC_DECODER) += aacdec.o aactab.o aacsbr.o aacps.o OBJS-$(CONFIG_AAC_ENCODER) += aacenc.o aaccoder.o \ aacpsy.o aactab.o \ psymodel.o iirfilter.o \ diff --git a/libavcodec/aac.h b/libavcodec/aac.h index 1a8aa6c145..94f578ff38 100644 --- a/libavcodec/aac.h +++ b/libavcodec/aac.h @@ -38,12 +38,6 @@ #include <stdint.h> -#define AAC_INIT_VLC_STATIC(num, size) \ - INIT_VLC_STATIC(&vlc_spectral[num], 8, ff_aac_spectral_sizes[num], \ - ff_aac_spectral_bits[num], sizeof( ff_aac_spectral_bits[num][0]), sizeof( ff_aac_spectral_bits[num][0]), \ - ff_aac_spectral_codes[num], sizeof(ff_aac_spectral_codes[num][0]), sizeof(ff_aac_spectral_codes[num][0]), \ - size); - #define MAX_CHANNELS 64 #define MAX_ELEM_ID 16 @@ -241,7 +235,7 @@ typedef struct { * main AAC context */ typedef struct { - AVCodecContext * avccontext; + AVCodecContext *avctx; MPEG4AudioConfig m4ac; @@ -255,8 +249,9 @@ typedef struct { enum ChannelPosition che_pos[4][MAX_ELEM_ID]; /**< channel element channel mapping with the * first index as the first 4 raw data block types */ - ChannelElement * che[4][MAX_ELEM_ID]; - ChannelElement * tag_che_map[4][MAX_ELEM_ID]; + ChannelElement *che[4][MAX_ELEM_ID]; + ChannelElement *tag_che_map[4][MAX_ELEM_ID]; + uint8_t tags_seen_this_frame[4][MAX_ELEM_ID]; int tags_mapped; /** @} */ diff --git a/libavcodec/aac_tablegen.c b/libavcodec/aac_tablegen.c new file mode 100644 index 0000000000..de1c9578a7 --- /dev/null +++ b/libavcodec/aac_tablegen.c @@ -0,0 +1,39 @@ +/* + * Generate a header file for hardcoded AAC tables + * + * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com> + * + * This file is part of FFmpeg. + * + * 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. + * + * 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 FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <stdlib.h> +#define CONFIG_HARDCODED_TABLES 0 +#include "aac_tablegen.h" +#include "tableprint.h" + +int main(void) +{ + ff_aac_tableinit(); + + write_fileheader(); + + printf("const float ff_aac_pow2sf_tab[428] = {\n"); + write_float_array(ff_aac_pow2sf_tab, 428); + printf("};\n"); + + return 0; +} diff --git a/libavcodec/aac_tablegen.h b/libavcodec/aac_tablegen.h new file mode 100644 index 0000000000..630071dd1d --- /dev/null +++ b/libavcodec/aac_tablegen.h @@ -0,0 +1,42 @@ +/* + * Header file for hardcoded AAC tables + * + * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com> + * + * This file is part of FFmpeg. + * + * 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. + * + * 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 FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AAC_TABLEGEN_H +#define AAC_TABLEGEN_H + +#include "aac_tablegen_decl.h" + +#if CONFIG_HARDCODED_TABLES +#include "libavcodec/aac_tables.h" +#else +#include "../libavutil/mathematics.h" +float ff_aac_pow2sf_tab[428]; + +void ff_aac_tableinit(void) +{ + int i; + for (i = 0; i < 428; i++) + ff_aac_pow2sf_tab[i] = pow(2, (i - 200) / 4.); +} +#endif /* CONFIG_HARDCODED_TABLES */ + +#endif /* AAC_TABLEGEN_H */ diff --git a/libavcodec/aac_tablegen_decl.h b/libavcodec/aac_tablegen_decl.h new file mode 100644 index 0000000000..7157c80461 --- /dev/null +++ b/libavcodec/aac_tablegen_decl.h @@ -0,0 +1,34 @@ +/* + * Header file for hardcoded AAC tables + * + * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com> + * + * This file is part of FFmpeg. + * + * 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. + * + * 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 FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AAC_TABLEGEN_INIT_H +#define AAC_TABLEGEN_INIT_H + +#if CONFIG_HARDCODED_TABLES +#define ff_aac_tableinit() +extern const float ff_aac_pow2sf_tab[428]; +#else +void ff_aac_tableinit(void); +extern float ff_aac_pow2sf_tab[428]; +#endif /* CONFIG_HARDCODED_TABLES */ + +#endif /* AAC_TABLEGEN_INIT_H */ diff --git a/libavcodec/aac.c b/libavcodec/aacdec.c index 3330e1cefe..e375d1f98b 100644 --- a/libavcodec/aac.c +++ b/libavcodec/aacdec.c @@ -67,7 +67,7 @@ * Y (not in this code) Layer-2 * Y (not in this code) Layer-3 * N SinuSoidal Coding (Transient, Sinusoid, Noise) - * N (planned) Parametric Stereo + * Y Parametric Stereo * N Direct Stream Transfer * * Note: - HE AAC v1 comprises LC AAC with Spectral Band Replication. @@ -113,6 +113,22 @@ static const char overread_err[] = "Input buffer exhausted before END element fo static ChannelElement *get_che(AACContext *ac, int type, int elem_id) { + /* Some buggy encoders appear to set all elem_ids to zero and rely on + channels always occurring in the same order. This is expressly forbidden + by the spec but we will try to work around it. + */ + int err_printed = 0; + while (ac->tags_seen_this_frame[type][elem_id] && elem_id < MAX_ELEM_ID) { + if (ac->output_configured < OC_LOCKED && !err_printed) { + av_log(ac->avctx, AV_LOG_WARNING, "Duplicate channel tag found, attempting to remap.\n"); + err_printed = 1; + } + elem_id++; + } + if (elem_id == MAX_ELEM_ID) + return NULL; + ac->tags_seen_this_frame[type][elem_id] = 1; + if (ac->tag_che_map[type][elem_id]) { return ac->tag_che_map[type][elem_id]; } @@ -127,8 +143,8 @@ static ChannelElement *get_che(AACContext *ac, int type, int elem_id) } case 6: /* Some streams incorrectly code 5.1 audio as SCE[0] CPE[0] CPE[1] SCE[1] - instead of SCE[0] CPE[0] CPE[0] LFE[0]. If we seem to have - encountered such a stream, transfer the LFE[0] element to SCE[1] */ + instead of SCE[0] CPE[0] CPE[1] LFE[0]. If we seem to have + encountered such a stream, transfer the LFE[0] element to the SCE[1]'s mapping */ if (ac->tags_mapped == tags_per_config[ac->m4ac.chan_config] - 1 && (type == TYPE_LFE || type == TYPE_SCE)) { ac->tags_mapped++; return ac->tag_che_map[type][elem_id] = ac->che[TYPE_LFE][0]; @@ -184,7 +200,8 @@ static av_cold int che_configure(AACContext *ac, ff_aac_sbr_ctx_init(&ac->che[type][id]->sbr); if (type != TYPE_CCE) { ac->output_data[(*channels)++] = ac->che[type][id]->ch[0].ret; - if (type == TYPE_CPE) { + if (type == TYPE_CPE || + (type == TYPE_SCE && ac->m4ac.ps == 1)) { ac->output_data[(*channels)++] = ac->che[type][id]->ch[1].ret; } } @@ -209,9 +226,10 @@ static av_cold int output_configure(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID], int channel_config, enum OCStatus oc_type) { - AVCodecContext *avctx = ac->avccontext; + AVCodecContext *avctx = ac->avctx; int i, type, channels = 0, ret; + if (new_che_pos != che_pos) memcpy(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0])); if (channel_config) { @@ -292,7 +310,7 @@ static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_EL sampling_index = get_bits(gb, 4); if (ac->m4ac.sampling_index != sampling_index) - av_log(ac->avccontext, AV_LOG_WARNING, "Sample rate index in program config element does not match the sample rate index configured by the container.\n"); + av_log(ac->avctx, AV_LOG_WARNING, "Sample rate index in program config element does not match the sample rate index configured by the container.\n"); num_front = get_bits(gb, 4); num_side = get_bits(gb, 4); @@ -323,7 +341,7 @@ static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_EL /* comment field, first byte is length */ comment_len = get_bits(gb, 8) * 8; if (get_bits_left(gb) < comment_len) { - av_log(ac->avccontext, AV_LOG_ERROR, overread_err); + av_log(ac->avctx, AV_LOG_ERROR, overread_err); return -1; } skip_bits_long(gb, comment_len); @@ -343,7 +361,7 @@ static av_cold int set_default_channel_config(AACContext *ac, int channel_config) { if (channel_config < 1 || channel_config > 7) { - av_log(ac->avccontext, AV_LOG_ERROR, "invalid default channel configuration (%d)\n", + av_log(ac->avctx, AV_LOG_ERROR, "invalid default channel configuration (%d)\n", channel_config); return -1; } @@ -388,7 +406,7 @@ static int decode_ga_specific_config(AACContext *ac, GetBitContext *gb, int extension_flag, ret; if (get_bits1(gb)) { // frameLengthFlag - av_log_missing_feature(ac->avccontext, "960/120 MDCT window is", 1); + av_log_missing_feature(ac->avctx, "960/120 MDCT window is", 1); return -1; } @@ -452,9 +470,11 @@ static int decode_audio_specific_config(AACContext *ac, void *data, if ((i = ff_mpeg4audio_get_config(&ac->m4ac, data, data_size)) < 0) return -1; if (ac->m4ac.sampling_index > 12) { - av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index); + av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index); return -1; } + if (ac->m4ac.sbr == 1 && ac->m4ac.ps == -1) + ac->m4ac.ps = 1; skip_bits_long(&gb, i); @@ -465,7 +485,7 @@ static int decode_audio_specific_config(AACContext *ac, void *data, return -1; break; default: - av_log(ac->avccontext, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n", + av_log(ac->avctx, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n", ac->m4ac.sbr == 1? "SBR+" : "", ac->m4ac.object_type); return -1; } @@ -508,20 +528,25 @@ static void reset_predictor_group(PredictorState *ps, int group_num) reset_predict_state(&ps[i]); } -static av_cold int aac_decode_init(AVCodecContext *avccontext) +#define AAC_INIT_VLC_STATIC(num, size) \ + INIT_VLC_STATIC(&vlc_spectral[num], 8, ff_aac_spectral_sizes[num], \ + ff_aac_spectral_bits[num], sizeof( ff_aac_spectral_bits[num][0]), sizeof( ff_aac_spectral_bits[num][0]), \ + ff_aac_spectral_codes[num], sizeof(ff_aac_spectral_codes[num][0]), sizeof(ff_aac_spectral_codes[num][0]), \ + size); + +static av_cold int aac_decode_init(AVCodecContext *avctx) { - AACContext *ac = avccontext->priv_data; - int i; + AACContext *ac = avctx->priv_data; - ac->avccontext = avccontext; - ac->m4ac.sample_rate = avccontext->sample_rate; + ac->avctx = avctx; + ac->m4ac.sample_rate = avctx->sample_rate; - if (avccontext->extradata_size > 0) { - if (decode_audio_specific_config(ac, avccontext->extradata, avccontext->extradata_size)) + if (avctx->extradata_size > 0) { + if (decode_audio_specific_config(ac, avctx->extradata, avctx->extradata_size)) return -1; } - avccontext->sample_fmt = SAMPLE_FMT_S16; + avctx->sample_fmt = SAMPLE_FMT_S16; AAC_INIT_VLC_STATIC( 0, 304); AAC_INIT_VLC_STATIC( 1, 270); @@ -537,7 +562,7 @@ static av_cold int aac_decode_init(AVCodecContext *avccontext) ff_aac_sbr_init(); - dsputil_init(&ac->dsp, avccontext); + dsputil_init(&ac->dsp, avctx); ac->random_state = 0x1f2e3d4c; @@ -555,10 +580,7 @@ static av_cold int aac_decode_init(AVCodecContext *avccontext) ac->sf_offset = 60; } -#if !CONFIG_HARDCODED_TABLES - for (i = 0; i < 428; i++) - ff_aac_pow2sf_tab[i] = pow(2, (i - 200) / 4.); -#endif /* CONFIG_HARDCODED_TABLES */ + ff_aac_tableinit(); 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]), @@ -591,7 +613,7 @@ static int skip_data_stream_element(AACContext *ac, GetBitContext *gb) align_get_bits(gb); if (get_bits_left(gb) < 8 * count) { - av_log(ac->avccontext, AV_LOG_ERROR, overread_err); + av_log(ac->avctx, AV_LOG_ERROR, overread_err); return -1; } skip_bits_long(gb, 8 * count); @@ -605,7 +627,7 @@ static int decode_prediction(AACContext *ac, IndividualChannelStream *ics, 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"); + av_log(ac->avctx, AV_LOG_ERROR, "Invalid Predictor Reset Group.\n"); return -1; } } @@ -624,7 +646,7 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, GetBitContext *gb, int common_window) { if (get_bits1(gb)) { - av_log(ac->avccontext, AV_LOG_ERROR, "Reserved bit set.\n"); + av_log(ac->avctx, AV_LOG_ERROR, "Reserved bit set.\n"); memset(ics, 0, sizeof(IndividualChannelStream)); return -1; } @@ -665,11 +687,11 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, 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"); + av_log(ac->avctx, AV_LOG_ERROR, "Prediction is not allowed in AAC-LC.\n"); memset(ics, 0, sizeof(IndividualChannelStream)); return -1; } else { - av_log_missing_feature(ac->avccontext, "Predictor bit set but LTP is", 1); + av_log_missing_feature(ac->avctx, "Predictor bit set but LTP is", 1); memset(ics, 0, sizeof(IndividualChannelStream)); return -1; } @@ -677,7 +699,7 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, } if (ics->max_sfb > ics->num_swb) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Number of scalefactor bands in group (%d) exceeds limit (%d).\n", ics->max_sfb, ics->num_swb); memset(ics, 0, sizeof(IndividualChannelStream)); @@ -708,18 +730,18 @@ static int decode_band_types(AACContext *ac, enum BandType band_type[120], int sect_len_incr; int sect_band_type = get_bits(gb, 4); if (sect_band_type == 12) { - av_log(ac->avccontext, AV_LOG_ERROR, "invalid band type\n"); + av_log(ac->avctx, AV_LOG_ERROR, "invalid band type\n"); return -1; } while ((sect_len_incr = get_bits(gb, bits)) == (1 << bits) - 1) sect_end += sect_len_incr; sect_end += sect_len_incr; if (get_bits_left(gb) < 0) { - av_log(ac->avccontext, AV_LOG_ERROR, overread_err); + av_log(ac->avctx, AV_LOG_ERROR, overread_err); return -1; } if (sect_end > ics->max_sfb) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Number of bands (%d) exceeds limit (%d).\n", sect_end, ics->max_sfb); return -1; @@ -764,7 +786,7 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, for (; i < run_end; i++, idx++) { offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60; if (offset[2] > 255U) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "%s (%d) out of range.\n", sf_str[2], offset[2]); return -1; } @@ -777,7 +799,7 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, else offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60; if (offset[1] > 255U) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "%s (%d) out of range.\n", sf_str[1], offset[1]); return -1; } @@ -787,7 +809,7 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, for (; i < run_end; i++, idx++) { offset[0] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60; if (offset[0] > 255U) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "%s (%d) out of range.\n", sf_str[0], offset[0]); return -1; } @@ -844,7 +866,7 @@ static int decode_tns(AACContext *ac, TemporalNoiseShaping *tns, 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.\n", + av_log(ac->avctx, AV_LOG_ERROR, "TNS filter order %d is greater than maximum %d.\n", tns->order[w][filt], tns_max_order); tns->order[w][filt] = 0; return -1; @@ -1163,7 +1185,7 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], b = 31 - av_log2(~b); if (b > 8) { - av_log(ac->avccontext, AV_LOG_ERROR, "error in spectral data, ESC overflow\n"); + av_log(ac->avctx, AV_LOG_ERROR, "error in spectral data, ESC overflow\n"); return -1; } @@ -1216,7 +1238,7 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], return 0; err_cb_overflow: - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Read beyond end of ff_aac_codebook_vectors[%d][]. index %d >= %d\n", band_type[idx], err_idx, ff_aac_spectral_sizes[band_type[idx]]); return -1; @@ -1337,18 +1359,18 @@ static int decode_ics(AACContext *ac, SingleChannelElement *sce, if (!scale_flag) { if ((pulse_present = get_bits1(gb))) { if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) { - av_log(ac->avccontext, AV_LOG_ERROR, "Pulse tool not allowed in eight short sequence.\n"); + av_log(ac->avctx, AV_LOG_ERROR, "Pulse tool not allowed in eight short sequence.\n"); return -1; } if (decode_pulses(&pulse, gb, ics->swb_offset, ics->num_swb)) { - av_log(ac->avccontext, AV_LOG_ERROR, "Pulse data corrupt or invalid.\n"); + av_log(ac->avctx, 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); + av_log_missing_feature(ac->avctx, "SSR", 1); return -1; } } @@ -1448,7 +1470,7 @@ static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe) cpe->ch[1].ics.use_kb_window[1] = i; ms_present = get_bits(gb, 2); if (ms_present == 3) { - av_log(ac->avccontext, AV_LOG_ERROR, "ms_present = 3 is reserved.\n"); + av_log(ac->avctx, AV_LOG_ERROR, "ms_present = 3 is reserved.\n"); return -1; } else if (ms_present) decode_mid_side_stereo(cpe, gb, ms_present); @@ -1635,16 +1657,20 @@ static int decode_extension_payload(AACContext *ac, GetBitContext *gb, int cnt, crc_flag++; case EXT_SBR_DATA: if (!che) { - av_log(ac->avccontext, AV_LOG_ERROR, "SBR was found before the first channel element.\n"); + av_log(ac->avctx, AV_LOG_ERROR, "SBR was found before the first channel element.\n"); return res; } else if (!ac->m4ac.sbr) { - av_log(ac->avccontext, AV_LOG_ERROR, "SBR signaled to be not-present but was found in the bitstream.\n"); + av_log(ac->avctx, AV_LOG_ERROR, "SBR signaled to be not-present but was found in the bitstream.\n"); skip_bits_long(gb, 8 * cnt - 4); return res; } else if (ac->m4ac.sbr == -1 && ac->output_configured == OC_LOCKED) { - av_log(ac->avccontext, AV_LOG_ERROR, "Implicit SBR was found with a first occurrence after the first frame.\n"); + av_log(ac->avctx, AV_LOG_ERROR, "Implicit SBR was found with a first occurrence after the first frame.\n"); skip_bits_long(gb, 8 * cnt - 4); return res; + } else if (ac->m4ac.ps == -1 && ac->output_configured < OC_LOCKED && ac->avctx->channels == 1) { + ac->m4ac.sbr = 1; + ac->m4ac.ps = 1; + output_configure(ac, ac->che_pos, ac->che_pos, ac->m4ac.chan_config, ac->output_configured); } else { ac->m4ac.sbr = 1; } @@ -1728,7 +1754,7 @@ static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce, float // 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, + av_log(ac->avctx, 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 < 1024; i += 128) @@ -1794,7 +1820,7 @@ static void apply_dependent_coupling(AACContext *ac, 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, + av_log(ac->avctx, AV_LOG_ERROR, "Dependent coupling is not supported together with LTP\n"); return; } @@ -1924,58 +1950,65 @@ static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb) } else if (ac->output_configured != OC_LOCKED) { ac->output_configured = OC_NONE; } - if (ac->output_configured != OC_LOCKED) + if (ac->output_configured != OC_LOCKED) { ac->m4ac.sbr = -1; + ac->m4ac.ps = -1; + } ac->m4ac.sample_rate = hdr_info.sample_rate; ac->m4ac.sampling_index = hdr_info.sampling_index; ac->m4ac.object_type = hdr_info.object_type; - if (!ac->avccontext->sample_rate) - ac->avccontext->sample_rate = hdr_info.sample_rate; + if (!ac->avctx->sample_rate) + ac->avctx->sample_rate = hdr_info.sample_rate; if (hdr_info.num_aac_frames == 1) { if (!hdr_info.crc_absent) skip_bits(gb, 16); } else { - av_log_missing_feature(ac->avccontext, "More than one AAC RDB per ADTS frame is", 0); + av_log_missing_feature(ac->avctx, "More than one AAC RDB per ADTS frame is", 0); return -1; } } return size; } -static int aac_decode_frame(AVCodecContext *avccontext, void *data, +static int aac_decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPacket *avpkt) { const uint8_t *buf = avpkt->data; int buf_size = avpkt->size; - AACContext *ac = avccontext->priv_data; + AACContext *ac = avctx->priv_data; ChannelElement *che = NULL, *che_prev = NULL; GetBitContext gb; enum RawDataBlockType elem_type, elem_type_prev = TYPE_END; int err, elem_id, data_size_tmp; int buf_consumed; - int samples = 1024, multiplier; + int samples = 0, multiplier; int buf_offset; init_get_bits(&gb, buf, buf_size * 8); if (show_bits(&gb, 12) == 0xfff) { if (parse_adts_frame_header(ac, &gb) < 0) { - av_log(avccontext, AV_LOG_ERROR, "Error decoding AAC frame header.\n"); + av_log(avctx, AV_LOG_ERROR, "Error decoding AAC frame header.\n"); return -1; } if (ac->m4ac.sampling_index > 12) { - av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index); + av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index); return -1; } } + memset(ac->tags_seen_this_frame, 0, sizeof(ac->tags_seen_this_frame)); // parse while ((elem_type = get_bits(&gb, 3)) != TYPE_END) { elem_id = get_bits(&gb, 4); - if (elem_type < TYPE_DSE && !(che=get_che(ac, elem_type, elem_id))) { - av_log(ac->avccontext, AV_LOG_ERROR, "channel element %d.%d is not allocated\n", elem_type, elem_id); - return -1; + if (elem_type < TYPE_DSE) { + if (!(che=get_che(ac, elem_type, elem_id))) { + av_log(ac->avctx, AV_LOG_ERROR, "channel element %d.%d is not allocated\n", + elem_type, elem_id); + return -1; + } + samples = 1024; } switch (elem_type) { @@ -2006,7 +2039,7 @@ static int aac_decode_frame(AVCodecContext *avccontext, void *data, if ((err = decode_pce(ac, new_che_pos, &gb))) break; if (ac->output_configured > OC_TRIAL_PCE) - av_log(avccontext, AV_LOG_ERROR, + av_log(avctx, AV_LOG_ERROR, "Not evaluating a further program_config_element as this construct is dubious at best.\n"); else err = output_configure(ac, ac->che_pos, new_che_pos, 0, OC_TRIAL_PCE); @@ -2017,7 +2050,7 @@ static int aac_decode_frame(AVCodecContext *avccontext, void *data, if (elem_id == 15) elem_id += get_bits(&gb, 8) - 1; if (get_bits_left(&gb) < 8 * elem_id) { - av_log(avccontext, AV_LOG_ERROR, overread_err); + av_log(avctx, AV_LOG_ERROR, overread_err); return -1; } while (elem_id > 0) @@ -2037,7 +2070,7 @@ static int aac_decode_frame(AVCodecContext *avccontext, void *data, return err; if (get_bits_left(&gb) < 3) { - av_log(avccontext, AV_LOG_ERROR, overread_err); + av_log(avctx, AV_LOG_ERROR, overread_err); return -1; } } @@ -2047,20 +2080,21 @@ static int aac_decode_frame(AVCodecContext *avccontext, void *data, multiplier = (ac->m4ac.sbr == 1) ? ac->m4ac.ext_sample_rate > ac->m4ac.sample_rate : 0; samples <<= multiplier; if (ac->output_configured < OC_LOCKED) { - avccontext->sample_rate = ac->m4ac.sample_rate << multiplier; - avccontext->frame_size = samples; + avctx->sample_rate = ac->m4ac.sample_rate << multiplier; + avctx->frame_size = samples; } - data_size_tmp = samples * avccontext->channels * sizeof(int16_t); + data_size_tmp = samples * avctx->channels * sizeof(int16_t); if (*data_size < data_size_tmp) { - av_log(avccontext, AV_LOG_ERROR, + av_log(avctx, AV_LOG_ERROR, "Output buffer too small (%d) or trying to output too many samples (%d) for this frame.\n", *data_size, data_size_tmp); return -1; } *data_size = data_size_tmp; - ac->dsp.float_to_int16_interleave(data, (const float **)ac->output_data, samples, avccontext->channels); + if (samples) + ac->dsp.float_to_int16_interleave(data, (const float **)ac->output_data, samples, avctx->channels); if (ac->output_configured) ac->output_configured = OC_LOCKED; @@ -2073,9 +2107,9 @@ static int aac_decode_frame(AVCodecContext *avccontext, void *data, return buf_size > buf_offset ? buf_consumed : buf_size; } -static av_cold int aac_decode_close(AVCodecContext *avccontext) +static av_cold int aac_decode_close(AVCodecContext *avctx) { - AACContext *ac = avccontext->priv_data; + AACContext *ac = avctx->priv_data; int i, type; for (i = 0; i < MAX_ELEM_ID; i++) { diff --git a/libavcodec/aacenc.c b/libavcodec/aacenc.c index 90dff15dd5..8c913e0dbc 100644 --- a/libavcodec/aacenc.c +++ b/libavcodec/aacenc.c @@ -201,13 +201,11 @@ static av_cold int aac_encode_init(AVCodecContext *avctx) lengths[1] = ff_aac_num_swb_128[i]; ff_psy_init(&s->psy, avctx, 2, sizes, lengths); s->psypp = ff_psy_preprocess_init(avctx); - s->coder = &ff_aac_coders[0]; + s->coder = &ff_aac_coders[2]; s->lambda = avctx->global_quality ? avctx->global_quality : 120; -#if !CONFIG_HARDCODED_TABLES - for (i = 0; i < 428; i++) - ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.); -#endif /* CONFIG_HARDCODED_TABLES */ + + ff_aac_tableinit(); if (avctx->channels > 5) av_log(avctx, AV_LOG_ERROR, "This encoder does not yet enforce the restrictions on LFEs. " @@ -234,25 +232,21 @@ static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s, s->output[i] = sce->saved[i]; } if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) { - j = channel; - for (i = 0; i < 1024; i++, j += avctx->channels) { + for (i = 0, j = channel; i < 1024; i++, j += avctx->channels) { s->output[i+1024] = audio[j] * lwindow[1024 - i - 1]; sce->saved[i] = audio[j] * lwindow[i]; } } else { - j = channel; - for (i = 0; i < 448; i++, j += avctx->channels) + for (i = 0, j = channel; i < 448; i++, j += avctx->channels) s->output[i+1024] = audio[j]; - for (i = 448; i < 576; i++, j += avctx->channels) + for (; i < 576; i++, j += avctx->channels) s->output[i+1024] = audio[j] * swindow[576 - i - 1]; memset(s->output+1024+576, 0, sizeof(s->output[0]) * 448); - j = channel; - for (i = 0; i < 1024; i++, j += avctx->channels) + for (i = 0, j = channel; i < 1024; i++, j += avctx->channels) sce->saved[i] = audio[j]; } ff_mdct_calc(&s->mdct1024, sce->coeffs, s->output); } else { - j = channel; for (k = 0; k < 1024; k += 128) { for (i = 448 + k; i < 448 + k + 256; i++) s->output[i - 448 - k] = (i < 1024) @@ -262,8 +256,7 @@ static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s, s->dsp.vector_fmul_reverse(s->output+128, s->output+128, swindow, 128); ff_mdct_calc(&s->mdct128, sce->coeffs + k, s->output); } - j = channel; - for (i = 0; i < 1024; i++, j += avctx->channels) + for (i = 0, j = channel; i < 1024; i++, j += avctx->channels) sce->saved[i] = audio[j]; } } @@ -562,6 +555,7 @@ static int aac_encode_frame(AVCodecContext *avctx, cpe = &s->cpe[i]; for (j = 0; j < chans; j++) { s->cur_channel = start_ch + j; + ff_psy_set_band_info(&s->psy, s->cur_channel, cpe->ch[j].coeffs, &wi[j]); s->coder->search_for_quantizers(avctx, s, &cpe->ch[j], s->lambda); } cpe->common_window = 0; @@ -592,7 +586,6 @@ static int aac_encode_frame(AVCodecContext *avctx, } for (j = 0; j < chans; j++) { s->cur_channel = start_ch + j; - ff_psy_set_band_info(&s->psy, s->cur_channel, cpe->ch[j].coeffs, &wi[j]); encode_individual_channel(avctx, s, &cpe->ch[j], cpe->common_window); } start_ch += chans; diff --git a/libavcodec/aacenc.h b/libavcodec/aacenc.h index e99be98048..86c68d3e16 100644 --- a/libavcodec/aacenc.h +++ b/libavcodec/aacenc.h @@ -64,7 +64,7 @@ typedef struct AACEncContext { int cur_channel; int last_frame; float lambda; - DECLARE_ALIGNED(16, int, qcoefs)[96][2]; ///< quantized coefficients + DECLARE_ALIGNED(16, int, qcoefs)[96]; ///< quantized coefficients DECLARE_ALIGNED(16, float, scoefs)[1024]; ///< scaled coefficients } AACEncContext; diff --git a/libavcodec/aacps.c b/libavcodec/aacps.c new file mode 100644 index 0000000000..fc124d1972 --- /dev/null +++ b/libavcodec/aacps.c @@ -0,0 +1,1037 @@ +/* + * MPEG-4 Parametric Stereo decoding functions + * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com> + * + * This file is part of FFmpeg. + * + * 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. + * + * 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 FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <stdint.h> +#include "libavutil/common.h" +#include "libavutil/mathematics.h" +#include "avcodec.h" +#include "get_bits.h" +#include "aacps.h" +#include "aacps_tablegen.h" +#include "aacpsdata.c" + +#define PS_BASELINE 0 //< Operate in Baseline PS mode + //< Baseline implies 10 or 20 stereo bands, + //< mixing mode A, and no ipd/opd + +#define numQMFSlots 32 //numTimeSlots * RATE + +static const int8_t num_env_tab[2][4] = { + { 0, 1, 2, 4, }, + { 1, 2, 3, 4, }, +}; + +static const int8_t nr_iidicc_par_tab[] = { + 10, 20, 34, 10, 20, 34, +}; + +static const int8_t nr_iidopd_par_tab[] = { + 5, 11, 17, 5, 11, 17, +}; + +enum { + huff_iid_df1, + huff_iid_dt1, + huff_iid_df0, + huff_iid_dt0, + huff_icc_df, + huff_icc_dt, + huff_ipd_df, + huff_ipd_dt, + huff_opd_df, + huff_opd_dt, +}; + +static const int huff_iid[] = { + huff_iid_df0, + huff_iid_df1, + huff_iid_dt0, + huff_iid_dt1, +}; + +static VLC vlc_ps[10]; + +/** + * Read Inter-channel Intensity Difference/Inter-Channel Coherence/ + * Inter-channel Phase Difference/Overall Phase Difference parameters from the + * bitstream. + * + * @param avctx contains the current codec context + * @param gb pointer to the input bitstream + * @param ps pointer to the Parametric Stereo context + * @param par pointer to the parameter to be read + * @param e envelope to decode + * @param dt 1: time delta-coded, 0: frequency delta-coded + */ +#define READ_PAR_DATA(PAR, OFFSET, MASK, ERR_CONDITION) \ +static int read_ ## PAR ## _data(AVCodecContext *avctx, GetBitContext *gb, PSContext *ps, \ + int8_t (*PAR)[PS_MAX_NR_IIDICC], int table_idx, int e, int dt) \ +{ \ + int b, num = ps->nr_ ## PAR ## _par; \ + VLC_TYPE (*vlc_table)[2] = vlc_ps[table_idx].table; \ + if (dt) { \ + int e_prev = e ? e - 1 : ps->num_env_old - 1; \ + e_prev = FFMAX(e_prev, 0); \ + for (b = 0; b < num; b++) { \ + int val = PAR[e_prev][b] + get_vlc2(gb, vlc_table, 9, 3) - OFFSET; \ + if (MASK) val &= MASK; \ + PAR[e][b] = val; \ + if (ERR_CONDITION) \ + goto err; \ + } \ + } else { \ + int val = 0; \ + for (b = 0; b < num; b++) { \ + val += get_vlc2(gb, vlc_table, 9, 3) - OFFSET; \ + if (MASK) val &= MASK; \ + PAR[e][b] = val; \ + if (ERR_CONDITION) \ + goto err; \ + } \ + } \ + return 0; \ +err: \ + av_log(avctx, AV_LOG_ERROR, "illegal "#PAR"\n"); \ + return -1; \ +} + +READ_PAR_DATA(iid, huff_offset[table_idx], 0, FFABS(ps->iid_par[e][b]) > 7 + 8 * ps->iid_quant) +READ_PAR_DATA(icc, huff_offset[table_idx], 0, ps->icc_par[e][b] > 7U) +READ_PAR_DATA(ipdopd, 0, 0x07, 0) + +static int ps_read_extension_data(GetBitContext *gb, PSContext *ps, int ps_extension_id) +{ + int e; + int count = get_bits_count(gb); + + if (ps_extension_id) + return 0; + + ps->enable_ipdopd = get_bits1(gb); + if (ps->enable_ipdopd) { + for (e = 0; e < ps->num_env; e++) { + int dt = get_bits1(gb); + read_ipdopd_data(NULL, gb, ps, ps->ipd_par, dt ? huff_ipd_dt : huff_ipd_df, e, dt); + dt = get_bits1(gb); + read_ipdopd_data(NULL, gb, ps, ps->opd_par, dt ? huff_opd_dt : huff_opd_df, e, dt); + } + } + skip_bits1(gb); //reserved_ps + return get_bits_count(gb) - count; +} + +static void ipdopd_reset(int8_t *opd_hist, int8_t *ipd_hist) +{ + int i; + for (i = 0; i < PS_MAX_NR_IPDOPD; i++) { + opd_hist[i] = 0; + ipd_hist[i] = 0; + } +} + +int ff_ps_read_data(AVCodecContext *avctx, GetBitContext *gb_host, PSContext *ps, int bits_left) +{ + int e; + int bit_count_start = get_bits_count(gb_host); + int header; + int bits_consumed; + GetBitContext gbc = *gb_host, *gb = &gbc; + + header = get_bits1(gb); + if (header) { //enable_ps_header + ps->enable_iid = get_bits1(gb); + if (ps->enable_iid) { + int iid_mode = get_bits(gb, 3); + if (iid_mode > 5) { + av_log(avctx, AV_LOG_ERROR, "iid_mode %d is reserved.\n", + iid_mode); + goto err; + } + ps->nr_iid_par = nr_iidicc_par_tab[iid_mode]; + ps->iid_quant = iid_mode > 2; + ps->nr_ipdopd_par = nr_iidopd_par_tab[iid_mode]; + } + ps->enable_icc = get_bits1(gb); + if (ps->enable_icc) { + ps->icc_mode = get_bits(gb, 3); + if (ps->icc_mode > 5) { + av_log(avctx, AV_LOG_ERROR, "icc_mode %d is reserved.\n", + ps->icc_mode); + goto err; + } + ps->nr_icc_par = nr_iidicc_par_tab[ps->icc_mode]; + } + ps->enable_ext = get_bits1(gb); + } + + ps->frame_class = get_bits1(gb); + ps->num_env_old = ps->num_env; + ps->num_env = num_env_tab[ps->frame_class][get_bits(gb, 2)]; + + ps->border_position[0] = -1; + if (ps->frame_class) { + for (e = 1; e <= ps->num_env; e++) + ps->border_position[e] = get_bits(gb, 5); + } else + for (e = 1; e <= ps->num_env; e++) + ps->border_position[e] = (e * numQMFSlots >> ff_log2_tab[ps->num_env]) - 1; + + if (ps->enable_iid) { + for (e = 0; e < ps->num_env; e++) { + int dt = get_bits1(gb); + if (read_iid_data(avctx, gb, ps, ps->iid_par, huff_iid[2*dt+ps->iid_quant], e, dt)) + goto err; + } + } else + memset(ps->iid_par, 0, sizeof(ps->iid_par)); + + if (ps->enable_icc) + for (e = 0; e < ps->num_env; e++) { + int dt = get_bits1(gb); + if (read_icc_data(avctx, gb, ps, ps->icc_par, dt ? huff_icc_dt : huff_icc_df, e, dt)) + goto err; + } + else + memset(ps->icc_par, 0, sizeof(ps->icc_par)); + + if (ps->enable_ext) { + int cnt = get_bits(gb, 4); + if (cnt == 15) { + cnt += get_bits(gb, 8); + } + cnt *= 8; + while (cnt > 7) { + int ps_extension_id = get_bits(gb, 2); + cnt -= 2 + ps_read_extension_data(gb, ps, ps_extension_id); + } + if (cnt < 0) { + av_log(avctx, AV_LOG_ERROR, "ps extension overflow %d", cnt); + goto err; + } + skip_bits(gb, cnt); + } + + ps->enable_ipdopd &= !PS_BASELINE; + + //Fix up envelopes + if (!ps->num_env || ps->border_position[ps->num_env] < numQMFSlots - 1) { + //Create a fake envelope + int source = ps->num_env ? ps->num_env - 1 : ps->num_env_old - 1; + if (source >= 0 && source != ps->num_env) { + if (ps->enable_iid) { + memcpy(ps->iid_par+ps->num_env, ps->iid_par+source, sizeof(ps->iid_par[0])); + } + if (ps->enable_icc) { + memcpy(ps->icc_par+ps->num_env, ps->icc_par+source, sizeof(ps->icc_par[0])); + } + if (ps->enable_ipdopd) { + memcpy(ps->ipd_par+ps->num_env, ps->ipd_par+source, sizeof(ps->ipd_par[0])); + memcpy(ps->opd_par+ps->num_env, ps->opd_par+source, sizeof(ps->opd_par[0])); + } + } + ps->num_env++; + ps->border_position[ps->num_env] = numQMFSlots - 1; + } + + + ps->is34bands_old = ps->is34bands; + if (!PS_BASELINE && (ps->enable_iid || ps->enable_icc)) + ps->is34bands = (ps->enable_iid && ps->nr_iid_par == 34) || + (ps->enable_icc && ps->nr_icc_par == 34); + + //Baseline + if (!ps->enable_ipdopd) { + memset(ps->ipd_par, 0, sizeof(ps->ipd_par)); + memset(ps->opd_par, 0, sizeof(ps->opd_par)); + } + + if (header) + ps->start = 1; + + bits_consumed = get_bits_count(gb) - bit_count_start; + if (bits_consumed <= bits_left) { + skip_bits_long(gb_host, bits_consumed); + return bits_consumed; + } + av_log(avctx, AV_LOG_ERROR, "Expected to read %d PS bits actually read %d.\n", bits_left, bits_consumed); +err: + ps->start = 0; + skip_bits_long(gb_host, bits_left); + return bits_left; +} + +/** Split one subband into 2 subsubbands with a symmetric real filter. + * The filter must have its non-center even coefficients equal to zero. */ +static void hybrid2_re(float (*in)[2], float (*out)[32][2], const float filter[7], int len, int reverse) +{ + int i, j; + for (i = 0; i < len; i++, in++) { + float re_in = filter[6] * in[6][0]; //real inphase + float re_op = 0.0f; //real out of phase + float im_in = filter[6] * in[6][1]; //imag inphase + float im_op = 0.0f; //imag out of phase + for (j = 0; j < 6; j += 2) { + re_op += filter[j+1] * (in[j+1][0] + in[12-j-1][0]); + im_op += filter[j+1] * (in[j+1][1] + in[12-j-1][1]); + } + out[ reverse][i][0] = re_in + re_op; + out[ reverse][i][1] = im_in + im_op; + out[!reverse][i][0] = re_in - re_op; + out[!reverse][i][1] = im_in - im_op; + } +} + +/** Split one subband into 6 subsubbands with a complex filter */ +static void hybrid6_cx(float (*in)[2], float (*out)[32][2], const float (*filter)[7][2], int len) +{ + int i, j, ssb; + int N = 8; + float temp[8][2]; + + for (i = 0; i < len; i++, in++) { + for (ssb = 0; ssb < N; ssb++) { + float sum_re = filter[ssb][6][0] * in[6][0], sum_im = filter[ssb][6][0] * in[6][1]; + for (j = 0; j < 6; j++) { + float in0_re = in[j][0]; + float in0_im = in[j][1]; + float in1_re = in[12-j][0]; + float in1_im = in[12-j][1]; + sum_re += filter[ssb][j][0] * (in0_re + in1_re) - filter[ssb][j][1] * (in0_im - in1_im); + sum_im += filter[ssb][j][0] * (in0_im + in1_im) + filter[ssb][j][1] * (in0_re - in1_re); + } + temp[ssb][0] = sum_re; + temp[ssb][1] = sum_im; + } + out[0][i][0] = temp[6][0]; + out[0][i][1] = temp[6][1]; + out[1][i][0] = temp[7][0]; + out[1][i][1] = temp[7][1]; + out[2][i][0] = temp[0][0]; + out[2][i][1] = temp[0][1]; + out[3][i][0] = temp[1][0]; + out[3][i][1] = temp[1][1]; + out[4][i][0] = temp[2][0] + temp[5][0]; + out[4][i][1] = temp[2][1] + temp[5][1]; + out[5][i][0] = temp[3][0] + temp[4][0]; + out[5][i][1] = temp[3][1] + temp[4][1]; + } +} + +static void hybrid4_8_12_cx(float (*in)[2], float (*out)[32][2], const float (*filter)[7][2], int N, int len) +{ + int i, j, ssb; + + for (i = 0; i < len; i++, in++) { + for (ssb = 0; ssb < N; ssb++) { + float sum_re = filter[ssb][6][0] * in[6][0], sum_im = filter[ssb][6][0] * in[6][1]; + for (j = 0; j < 6; j++) { + float in0_re = in[j][0]; + float in0_im = in[j][1]; + float in1_re = in[12-j][0]; + float in1_im = in[12-j][1]; + sum_re += filter[ssb][j][0] * (in0_re + in1_re) - filter[ssb][j][1] * (in0_im - in1_im); + sum_im += filter[ssb][j][0] * (in0_im + in1_im) + filter[ssb][j][1] * (in0_re - in1_re); + } + out[ssb][i][0] = sum_re; + out[ssb][i][1] = sum_im; + } + } +} + +static void hybrid_analysis(float out[91][32][2], float in[5][44][2], float L[2][38][64], int is34, int len) +{ + int i, j; + for (i = 0; i < 5; i++) { + for (j = 0; j < 38; j++) { + in[i][j+6][0] = L[0][j][i]; + in[i][j+6][1] = L[1][j][i]; + } + } + if (is34) { + hybrid4_8_12_cx(in[0], out, f34_0_12, 12, len); + hybrid4_8_12_cx(in[1], out+12, f34_1_8, 8, len); + hybrid4_8_12_cx(in[2], out+20, f34_2_4, 4, len); + hybrid4_8_12_cx(in[3], out+24, f34_2_4, 4, len); + hybrid4_8_12_cx(in[4], out+28, f34_2_4, 4, len); + for (i = 0; i < 59; i++) { + for (j = 0; j < len; j++) { + out[i+32][j][0] = L[0][j][i+5]; + out[i+32][j][1] = L[1][j][i+5]; + } + } + } else { + hybrid6_cx(in[0], out, f20_0_8, len); + hybrid2_re(in[1], out+6, g1_Q2, len, 1); + hybrid2_re(in[2], out+8, g1_Q2, len, 0); + for (i = 0; i < 61; i++) { + for (j = 0; j < len; j++) { + out[i+10][j][0] = L[0][j][i+3]; + out[i+10][j][1] = L[1][j][i+3]; + } + } + } + //update in_buf + for (i = 0; i < 5; i++) { + memcpy(in[i], in[i]+32, 6 * sizeof(in[i][0])); + } +} + +static void hybrid_synthesis(float out[2][38][64], float in[91][32][2], int is34, int len) +{ + int i, n; + if (is34) { + for (n = 0; n < len; n++) { + memset(out[0][n], 0, 5*sizeof(out[0][n][0])); + memset(out[1][n], 0, 5*sizeof(out[1][n][0])); + for (i = 0; i < 12; i++) { + out[0][n][0] += in[ i][n][0]; + out[1][n][0] += in[ i][n][1]; + } + for (i = 0; i < 8; i++) { + out[0][n][1] += in[12+i][n][0]; + out[1][n][1] += in[12+i][n][1]; + } + for (i = 0; i < 4; i++) { + out[0][n][2] += in[20+i][n][0]; + out[1][n][2] += in[20+i][n][1]; + out[0][n][3] += in[24+i][n][0]; + out[1][n][3] += in[24+i][n][1]; + out[0][n][4] += in[28+i][n][0]; + out[1][n][4] += in[28+i][n][1]; + } + } + for (i = 0; i < 59; i++) { + for (n = 0; n < len; n++) { + out[0][n][i+5] = in[i+32][n][0]; + out[1][n][i+5] = in[i+32][n][1]; + } + } + } else { + for (n = 0; n < len; n++) { + out[0][n][0] = in[0][n][0] + in[1][n][0] + in[2][n][0] + + in[3][n][0] + in[4][n][0] + in[5][n][0]; + out[1][n][0] = in[0][n][1] + in[1][n][1] + in[2][n][1] + + in[3][n][1] + in[4][n][1] + in[5][n][1]; + out[0][n][1] = in[6][n][0] + in[7][n][0]; + out[1][n][1] = in[6][n][1] + in[7][n][1]; + out[0][n][2] = in[8][n][0] + in[9][n][0]; + out[1][n][2] = in[8][n][1] + in[9][n][1]; + } + for (i = 0; i < 61; i++) { + for (n = 0; n < len; n++) { + out[0][n][i+3] = in[i+10][n][0]; + out[1][n][i+3] = in[i+10][n][1]; + } + } + } +} + +/// All-pass filter decay slope +#define DECAY_SLOPE 0.05f +/// Number of frequency bands that can be addressed by the parameter index, b(k) +static const int NR_PAR_BANDS[] = { 20, 34 }; +/// Number of frequency bands that can be addressed by the sub subband index, k +static const int NR_BANDS[] = { 71, 91 }; +/// Start frequency band for the all-pass filter decay slope +static const int DECAY_CUTOFF[] = { 10, 32 }; +/// Number of all-pass filer bands +static const int NR_ALLPASS_BANDS[] = { 30, 50 }; +/// First stereo band using the short one sample delay +static const int SHORT_DELAY_BAND[] = { 42, 62 }; + +/** Table 8.46 */ +static void map_idx_10_to_20(int8_t *par_mapped, const int8_t *par, int full) +{ + int b; + if (full) + b = 9; + else { + b = 4; + par_mapped[10] = 0; + } + for (; b >= 0; b--) { + par_mapped[2*b+1] = par_mapped[2*b] = par[b]; + } +} + +static void map_idx_34_to_20(int8_t *par_mapped, const int8_t *par, int full) +{ + par_mapped[ 0] = (2*par[ 0] + par[ 1]) / 3; + par_mapped[ 1] = ( par[ 1] + 2*par[ 2]) / 3; + par_mapped[ 2] = (2*par[ 3] + par[ 4]) / 3; + par_mapped[ 3] = ( par[ 4] + 2*par[ 5]) / 3; + par_mapped[ 4] = ( par[ 6] + par[ 7]) / 2; + par_mapped[ 5] = ( par[ 8] + par[ 9]) / 2; + par_mapped[ 6] = par[10]; + par_mapped[ 7] = par[11]; + par_mapped[ 8] = ( par[12] + par[13]) / 2; + par_mapped[ 9] = ( par[14] + par[15]) / 2; + par_mapped[10] = par[16]; + if (full) { + par_mapped[11] = par[17]; + par_mapped[12] = par[18]; + par_mapped[13] = par[19]; + par_mapped[14] = ( par[20] + par[21]) / 2; + par_mapped[15] = ( par[22] + par[23]) / 2; + par_mapped[16] = ( par[24] + par[25]) / 2; + par_mapped[17] = ( par[26] + par[27]) / 2; + par_mapped[18] = ( par[28] + par[29] + par[30] + par[31]) / 4; + par_mapped[19] = ( par[32] + par[33]) / 2; + } +} + +static void map_val_34_to_20(float par[PS_MAX_NR_IIDICC]) +{ + par[ 0] = (2*par[ 0] + par[ 1]) * 0.33333333f; + par[ 1] = ( par[ 1] + 2*par[ 2]) * 0.33333333f; + par[ 2] = (2*par[ 3] + par[ 4]) * 0.33333333f; + par[ 3] = ( par[ 4] + 2*par[ 5]) * 0.33333333f; + par[ 4] = ( par[ 6] + par[ 7]) * 0.5f; + par[ 5] = ( par[ 8] + par[ 9]) * 0.5f; + par[ 6] = par[10]; + par[ 7] = par[11]; + par[ 8] = ( par[12] + par[13]) * 0.5f; + par[ 9] = ( par[14] + par[15]) * 0.5f; + par[10] = par[16]; + par[11] = par[17]; + par[12] = par[18]; + par[13] = par[19]; + par[14] = ( par[20] + par[21]) * 0.5f; + par[15] = ( par[22] + par[23]) * 0.5f; + par[16] = ( par[24] + par[25]) * 0.5f; + par[17] = ( par[26] + par[27]) * 0.5f; + par[18] = ( par[28] + par[29] + par[30] + par[31]) * 0.25f; + par[19] = ( par[32] + par[33]) * 0.5f; +} + +static void map_idx_10_to_34(int8_t *par_mapped, const int8_t *par, int full) +{ + if (full) { + par_mapped[33] = par[9]; + par_mapped[32] = par[9]; + par_mapped[31] = par[9]; + par_mapped[30] = par[9]; + par_mapped[29] = par[9]; + par_mapped[28] = par[9]; + par_mapped[27] = par[8]; + par_mapped[26] = par[8]; + par_mapped[25] = par[8]; + par_mapped[24] = par[8]; + par_mapped[23] = par[7]; + par_mapped[22] = par[7]; + par_mapped[21] = par[7]; + par_mapped[20] = par[7]; + par_mapped[19] = par[6]; + par_mapped[18] = par[6]; + par_mapped[17] = par[5]; + par_mapped[16] = par[5]; + } else { + par_mapped[16] = 0; + } + par_mapped[15] = par[4]; + par_mapped[14] = par[4]; + par_mapped[13] = par[4]; + par_mapped[12] = par[4]; + par_mapped[11] = par[3]; + par_mapped[10] = par[3]; + par_mapped[ 9] = par[2]; + par_mapped[ 8] = par[2]; + par_mapped[ 7] = par[2]; + par_mapped[ 6] = par[2]; + par_mapped[ 5] = par[1]; + par_mapped[ 4] = par[1]; + par_mapped[ 3] = par[1]; + par_mapped[ 2] = par[0]; + par_mapped[ 1] = par[0]; + par_mapped[ 0] = par[0]; +} + +static void map_idx_20_to_34(int8_t *par_mapped, const int8_t *par, int full) +{ + if (full) { + par_mapped[33] = par[19]; + par_mapped[32] = par[19]; + par_mapped[31] = par[18]; + par_mapped[30] = par[18]; + par_mapped[29] = par[18]; + par_mapped[28] = par[18]; + par_mapped[27] = par[17]; + par_mapped[26] = par[17]; + par_mapped[25] = par[16]; + par_mapped[24] = par[16]; + par_mapped[23] = par[15]; + par_mapped[22] = par[15]; + par_mapped[21] = par[14]; + par_mapped[20] = par[14]; + par_mapped[19] = par[13]; + par_mapped[18] = par[12]; + par_mapped[17] = par[11]; + } + par_mapped[16] = par[10]; + par_mapped[15] = par[ 9]; + par_mapped[14] = par[ 9]; + par_mapped[13] = par[ 8]; + par_mapped[12] = par[ 8]; + par_mapped[11] = par[ 7]; + par_mapped[10] = par[ 6]; + par_mapped[ 9] = par[ 5]; + par_mapped[ 8] = par[ 5]; + par_mapped[ 7] = par[ 4]; + par_mapped[ 6] = par[ 4]; + par_mapped[ 5] = par[ 3]; + par_mapped[ 4] = (par[ 2] + par[ 3]) / 2; + par_mapped[ 3] = par[ 2]; + par_mapped[ 2] = par[ 1]; + par_mapped[ 1] = (par[ 0] + par[ 1]) / 2; + par_mapped[ 0] = par[ 0]; +} + +static void map_val_20_to_34(float par[PS_MAX_NR_IIDICC]) +{ + par[33] = par[19]; + par[32] = par[19]; + par[31] = par[18]; + par[30] = par[18]; + par[29] = par[18]; + par[28] = par[18]; + par[27] = par[17]; + par[26] = par[17]; + par[25] = par[16]; + par[24] = par[16]; + par[23] = par[15]; + par[22] = par[15]; + par[21] = par[14]; + par[20] = par[14]; + par[19] = par[13]; + par[18] = par[12]; + par[17] = par[11]; + par[16] = par[10]; + par[15] = par[ 9]; + par[14] = par[ 9]; + par[13] = par[ 8]; + par[12] = par[ 8]; + par[11] = par[ 7]; + par[10] = par[ 6]; + par[ 9] = par[ 5]; + par[ 8] = par[ 5]; + par[ 7] = par[ 4]; + par[ 6] = par[ 4]; + par[ 5] = par[ 3]; + par[ 4] = (par[ 2] + par[ 3]) * 0.5f; + par[ 3] = par[ 2]; + par[ 2] = par[ 1]; + par[ 1] = (par[ 0] + par[ 1]) * 0.5f; + par[ 0] = par[ 0]; +} + +static void decorrelation(PSContext *ps, float (*out)[32][2], const float (*s)[32][2], int is34) +{ + float power[34][PS_QMF_TIME_SLOTS] = {{0}}; + float transient_gain[34][PS_QMF_TIME_SLOTS]; + float *peak_decay_nrg = ps->peak_decay_nrg; + float *power_smooth = ps->power_smooth; + float *peak_decay_diff_smooth = ps->peak_decay_diff_smooth; + float (*delay)[PS_QMF_TIME_SLOTS + PS_MAX_DELAY][2] = ps->delay; + float (*ap_delay)[PS_AP_LINKS][PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2] = ps->ap_delay; + const int8_t *k_to_i = is34 ? k_to_i_34 : k_to_i_20; + const float peak_decay_factor = 0.76592833836465f; + const float transient_impact = 1.5f; + const float a_smooth = 0.25f; //< Smoothing coefficient + int i, k, m, n; + int n0 = 0, nL = 32; + static const int link_delay[] = { 3, 4, 5 }; + static const float a[] = { 0.65143905753106f, + 0.56471812200776f, + 0.48954165955695f }; + + if (is34 != ps->is34bands_old) { + memset(ps->peak_decay_nrg, 0, sizeof(ps->peak_decay_nrg)); + memset(ps->power_smooth, 0, sizeof(ps->power_smooth)); + memset(ps->peak_decay_diff_smooth, 0, sizeof(ps->peak_decay_diff_smooth)); + memset(ps->delay, 0, sizeof(ps->delay)); + memset(ps->ap_delay, 0, sizeof(ps->ap_delay)); + } + + for (n = n0; n < nL; n++) { + for (k = 0; k < NR_BANDS[is34]; k++) { + int i = k_to_i[k]; + power[i][n] += s[k][n][0] * s[k][n][0] + s[k][n][1] * s[k][n][1]; + } + } + + //Transient detection + for (i = 0; i < NR_PAR_BANDS[is34]; i++) { + for (n = n0; n < nL; n++) { + float decayed_peak = peak_decay_factor * peak_decay_nrg[i]; + float denom; + peak_decay_nrg[i] = FFMAX(decayed_peak, power[i][n]); + power_smooth[i] += a_smooth * (power[i][n] - power_smooth[i]); + peak_decay_diff_smooth[i] += a_smooth * (peak_decay_nrg[i] - power[i][n] - peak_decay_diff_smooth[i]); + denom = transient_impact * peak_decay_diff_smooth[i]; + transient_gain[i][n] = (denom > power_smooth[i]) ? + power_smooth[i] / denom : 1.0f; + } + } + + //Decorrelation and transient reduction + // PS_AP_LINKS - 1 + // ----- + // | | Q_fract_allpass[k][m]*z^-link_delay[m] - a[m]*g_decay_slope[k] + //H[k][z] = z^-2 * phi_fract[k] * | | ---------------------------------------------------------------- + // | | 1 - a[m]*g_decay_slope[k]*Q_fract_allpass[k][m]*z^-link_delay[m] + // m = 0 + //d[k][z] (out) = transient_gain_mapped[k][z] * H[k][z] * s[k][z] + for (k = 0; k < NR_ALLPASS_BANDS[is34]; k++) { + int b = k_to_i[k]; + float g_decay_slope = 1.f - DECAY_SLOPE * (k - DECAY_CUTOFF[is34]); + float ag[PS_AP_LINKS]; + g_decay_slope = av_clipf(g_decay_slope, 0.f, 1.f); + memcpy(delay[k], delay[k]+nL, PS_MAX_DELAY*sizeof(delay[k][0])); + memcpy(delay[k]+PS_MAX_DELAY, s[k], numQMFSlots*sizeof(delay[k][0])); + for (m = 0; m < PS_AP_LINKS; m++) { + memcpy(ap_delay[k][m], ap_delay[k][m]+numQMFSlots, 5*sizeof(ap_delay[k][m][0])); + ag[m] = a[m] * g_decay_slope; + } + for (n = n0; n < nL; n++) { + float in_re = delay[k][n+PS_MAX_DELAY-2][0] * phi_fract[is34][k][0] - + delay[k][n+PS_MAX_DELAY-2][1] * phi_fract[is34][k][1]; + float in_im = delay[k][n+PS_MAX_DELAY-2][0] * phi_fract[is34][k][1] + + delay[k][n+PS_MAX_DELAY-2][1] * phi_fract[is34][k][0]; + for (m = 0; m < PS_AP_LINKS; m++) { + float a_re = ag[m] * in_re; + float a_im = ag[m] * in_im; + float link_delay_re = ap_delay[k][m][n+5-link_delay[m]][0]; + float link_delay_im = ap_delay[k][m][n+5-link_delay[m]][1]; + float fractional_delay_re = Q_fract_allpass[is34][k][m][0]; + float fractional_delay_im = Q_fract_allpass[is34][k][m][1]; + ap_delay[k][m][n+5][0] = in_re; + ap_delay[k][m][n+5][1] = in_im; + in_re = link_delay_re * fractional_delay_re - link_delay_im * fractional_delay_im - a_re; + in_im = link_delay_re * fractional_delay_im + link_delay_im * fractional_delay_re - a_im; + ap_delay[k][m][n+5][0] += ag[m] * in_re; + ap_delay[k][m][n+5][1] += ag[m] * in_im; + } + out[k][n][0] = transient_gain[b][n] * in_re; + out[k][n][1] = transient_gain[b][n] * in_im; + } + } + for (; k < SHORT_DELAY_BAND[is34]; k++) { + memcpy(delay[k], delay[k]+nL, PS_MAX_DELAY*sizeof(delay[k][0])); + memcpy(delay[k]+PS_MAX_DELAY, s[k], numQMFSlots*sizeof(delay[k][0])); + for (n = n0; n < nL; n++) { + //H = delay 14 + out[k][n][0] = transient_gain[k_to_i[k]][n] * delay[k][n+PS_MAX_DELAY-14][0]; + out[k][n][1] = transient_gain[k_to_i[k]][n] * delay[k][n+PS_MAX_DELAY-14][1]; + } + } + for (; k < NR_BANDS[is34]; k++) { + memcpy(delay[k], delay[k]+nL, PS_MAX_DELAY*sizeof(delay[k][0])); + memcpy(delay[k]+PS_MAX_DELAY, s[k], numQMFSlots*sizeof(delay[k][0])); + for (n = n0; n < nL; n++) { + //H = delay 1 + out[k][n][0] = transient_gain[k_to_i[k]][n] * delay[k][n+PS_MAX_DELAY-1][0]; + out[k][n][1] = transient_gain[k_to_i[k]][n] * delay[k][n+PS_MAX_DELAY-1][1]; + } + } +} + +static void remap34(int8_t (**p_par_mapped)[PS_MAX_NR_IIDICC], + int8_t (*par)[PS_MAX_NR_IIDICC], + int num_par, int num_env, int full) +{ + int8_t (*par_mapped)[PS_MAX_NR_IIDICC] = *p_par_mapped; + int e; + if (num_par == 20 || num_par == 11) { + for (e = 0; e < num_env; e++) { + map_idx_20_to_34(par_mapped[e], par[e], full); + } + } else if (num_par == 10 || num_par == 5) { + for (e = 0; e < num_env; e++) { + map_idx_10_to_34(par_mapped[e], par[e], full); + } + } else { + *p_par_mapped = par; + } +} + +static void remap20(int8_t (**p_par_mapped)[PS_MAX_NR_IIDICC], + int8_t (*par)[PS_MAX_NR_IIDICC], + int num_par, int num_env, int full) +{ + int8_t (*par_mapped)[PS_MAX_NR_IIDICC] = *p_par_mapped; + int e; + if (num_par == 34 || num_par == 17) { + for (e = 0; e < num_env; e++) { + map_idx_34_to_20(par_mapped[e], par[e], full); + } + } else if (num_par == 10 || num_par == 5) { + for (e = 0; e < num_env; e++) { + map_idx_10_to_20(par_mapped[e], par[e], full); + } + } else { + *p_par_mapped = par; + } +} + +static void stereo_processing(PSContext *ps, float (*l)[32][2], float (*r)[32][2], int is34) +{ + int e, b, k, n; + + float (*H11)[PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC] = ps->H11; + float (*H12)[PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC] = ps->H12; + float (*H21)[PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC] = ps->H21; + float (*H22)[PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC] = ps->H22; + int8_t *opd_hist = ps->opd_hist; + int8_t *ipd_hist = ps->ipd_hist; + int8_t iid_mapped_buf[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; + int8_t icc_mapped_buf[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; + int8_t ipd_mapped_buf[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; + int8_t opd_mapped_buf[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; + int8_t (*iid_mapped)[PS_MAX_NR_IIDICC] = iid_mapped_buf; + int8_t (*icc_mapped)[PS_MAX_NR_IIDICC] = icc_mapped_buf; + int8_t (*ipd_mapped)[PS_MAX_NR_IIDICC] = ipd_mapped_buf; + int8_t (*opd_mapped)[PS_MAX_NR_IIDICC] = opd_mapped_buf; + const int8_t *k_to_i = is34 ? k_to_i_34 : k_to_i_20; + const float (*H_LUT)[8][4] = (PS_BASELINE || ps->icc_mode < 3) ? HA : HB; + + //Remapping + memcpy(H11[0][0], H11[0][ps->num_env_old], PS_MAX_NR_IIDICC*sizeof(H11[0][0][0])); + memcpy(H11[1][0], H11[1][ps->num_env_old], PS_MAX_NR_IIDICC*sizeof(H11[1][0][0])); + memcpy(H12[0][0], H12[0][ps->num_env_old], PS_MAX_NR_IIDICC*sizeof(H12[0][0][0])); + memcpy(H12[1][0], H12[1][ps->num_env_old], PS_MAX_NR_IIDICC*sizeof(H12[1][0][0])); + memcpy(H21[0][0], H21[0][ps->num_env_old], PS_MAX_NR_IIDICC*sizeof(H21[0][0][0])); + memcpy(H21[1][0], H21[1][ps->num_env_old], PS_MAX_NR_IIDICC*sizeof(H21[1][0][0])); + memcpy(H22[0][0], H22[0][ps->num_env_old], PS_MAX_NR_IIDICC*sizeof(H22[0][0][0])); + memcpy(H22[1][0], H22[1][ps->num_env_old], PS_MAX_NR_IIDICC*sizeof(H22[1][0][0])); + if (is34) { + remap34(&iid_mapped, ps->iid_par, ps->nr_iid_par, ps->num_env, 1); + remap34(&icc_mapped, ps->icc_par, ps->nr_icc_par, ps->num_env, 1); + if (ps->enable_ipdopd) { + remap34(&ipd_mapped, ps->ipd_par, ps->nr_ipdopd_par, ps->num_env, 0); + remap34(&opd_mapped, ps->opd_par, ps->nr_ipdopd_par, ps->num_env, 0); + } + if (!ps->is34bands_old) { + map_val_20_to_34(H11[0][0]); + map_val_20_to_34(H11[1][0]); + map_val_20_to_34(H12[0][0]); + map_val_20_to_34(H12[1][0]); + map_val_20_to_34(H21[0][0]); + map_val_20_to_34(H21[1][0]); + map_val_20_to_34(H22[0][0]); + map_val_20_to_34(H22[1][0]); + ipdopd_reset(ipd_hist, opd_hist); + } + } else { + remap20(&iid_mapped, ps->iid_par, ps->nr_iid_par, ps->num_env, 1); + remap20(&icc_mapped, ps->icc_par, ps->nr_icc_par, ps->num_env, 1); + if (ps->enable_ipdopd) { + remap20(&ipd_mapped, ps->ipd_par, ps->nr_ipdopd_par, ps->num_env, 0); + remap20(&opd_mapped, ps->opd_par, ps->nr_ipdopd_par, ps->num_env, 0); + } + if (ps->is34bands_old) { + map_val_34_to_20(H11[0][0]); + map_val_34_to_20(H11[1][0]); + map_val_34_to_20(H12[0][0]); + map_val_34_to_20(H12[1][0]); + map_val_34_to_20(H21[0][0]); + map_val_34_to_20(H21[1][0]); + map_val_34_to_20(H22[0][0]); + map_val_34_to_20(H22[1][0]); + ipdopd_reset(ipd_hist, opd_hist); + } + } + + //Mixing + for (e = 0; e < ps->num_env; e++) { + for (b = 0; b < NR_PAR_BANDS[is34]; b++) { + float h11, h12, h21, h22; + h11 = H_LUT[iid_mapped[e][b] + 7 + 23 * ps->iid_quant][icc_mapped[e][b]][0]; + h12 = H_LUT[iid_mapped[e][b] + 7 + 23 * ps->iid_quant][icc_mapped[e][b]][1]; + h21 = H_LUT[iid_mapped[e][b] + 7 + 23 * ps->iid_quant][icc_mapped[e][b]][2]; + h22 = H_LUT[iid_mapped[e][b] + 7 + 23 * ps->iid_quant][icc_mapped[e][b]][3]; + if (!PS_BASELINE && ps->enable_ipdopd && b < ps->nr_ipdopd_par) { + //The spec say says to only run this smoother when enable_ipdopd + //is set but the reference decoder appears to run it constantly + float h11i, h12i, h21i, h22i; + float ipd_adj_re, ipd_adj_im; + int opd_idx = opd_hist[b] * 8 + opd_mapped[e][b]; + int ipd_idx = ipd_hist[b] * 8 + ipd_mapped[e][b]; + float opd_re = pd_re_smooth[opd_idx]; + float opd_im = pd_im_smooth[opd_idx]; + float ipd_re = pd_re_smooth[ipd_idx]; + float ipd_im = pd_im_smooth[ipd_idx]; + opd_hist[b] = opd_idx & 0x3F; + ipd_hist[b] = ipd_idx & 0x3F; + + ipd_adj_re = opd_re*ipd_re + opd_im*ipd_im; + ipd_adj_im = opd_im*ipd_re - opd_re*ipd_im; + h11i = h11 * opd_im; + h11 = h11 * opd_re; + h12i = h12 * ipd_adj_im; + h12 = h12 * ipd_adj_re; + h21i = h21 * opd_im; + h21 = h21 * opd_re; + h22i = h22 * ipd_adj_im; + h22 = h22 * ipd_adj_re; + H11[1][e+1][b] = h11i; + H12[1][e+1][b] = h12i; + H21[1][e+1][b] = h21i; + H22[1][e+1][b] = h22i; + } + H11[0][e+1][b] = h11; + H12[0][e+1][b] = h12; + H21[0][e+1][b] = h21; + H22[0][e+1][b] = h22; + } + for (k = 0; k < NR_BANDS[is34]; k++) { + float h11r, h12r, h21r, h22r; + float h11i, h12i, h21i, h22i; + float h11r_step, h12r_step, h21r_step, h22r_step; + float h11i_step, h12i_step, h21i_step, h22i_step; + int start = ps->border_position[e]; + int stop = ps->border_position[e+1]; + float width = 1.f / (stop - start); + b = k_to_i[k]; + h11r = H11[0][e][b]; + h12r = H12[0][e][b]; + h21r = H21[0][e][b]; + h22r = H22[0][e][b]; + if (!PS_BASELINE && ps->enable_ipdopd) { + //Is this necessary? ps_04_new seems unchanged + if ((is34 && k <= 13 && k >= 9) || (!is34 && k <= 1)) { + h11i = -H11[1][e][b]; + h12i = -H12[1][e][b]; + h21i = -H21[1][e][b]; + h22i = -H22[1][e][b]; + } else { + h11i = H11[1][e][b]; + h12i = H12[1][e][b]; + h21i = H21[1][e][b]; + h22i = H22[1][e][b]; + } + } + //Interpolation + h11r_step = (H11[0][e+1][b] - h11r) * width; + h12r_step = (H12[0][e+1][b] - h12r) * width; + h21r_step = (H21[0][e+1][b] - h21r) * width; + h22r_step = (H22[0][e+1][b] - h22r) * width; + if (!PS_BASELINE && ps->enable_ipdopd) { + h11i_step = (H11[1][e+1][b] - h11i) * width; + h12i_step = (H12[1][e+1][b] - h12i) * width; + h21i_step = (H21[1][e+1][b] - h21i) * width; + h22i_step = (H22[1][e+1][b] - h22i) * width; + } + for (n = start + 1; n <= stop; n++) { + //l is s, r is d + float l_re = l[k][n][0]; + float l_im = l[k][n][1]; + float r_re = r[k][n][0]; + float r_im = r[k][n][1]; + h11r += h11r_step; + h12r += h12r_step; + h21r += h21r_step; + h22r += h22r_step; + if (!PS_BASELINE && ps->enable_ipdopd) { + h11i += h11i_step; + h12i += h12i_step; + h21i += h21i_step; + h22i += h22i_step; + + l[k][n][0] = h11r*l_re + h21r*r_re - h11i*l_im - h21i*r_im; + l[k][n][1] = h11r*l_im + h21r*r_im + h11i*l_re + h21i*r_re; + r[k][n][0] = h12r*l_re + h22r*r_re - h12i*l_im - h22i*r_im; + r[k][n][1] = h12r*l_im + h22r*r_im + h12i*l_re + h22i*r_re; + } else { + l[k][n][0] = h11r*l_re + h21r*r_re; + l[k][n][1] = h11r*l_im + h21r*r_im; + r[k][n][0] = h12r*l_re + h22r*r_re; + r[k][n][1] = h12r*l_im + h22r*r_im; + } + } + } + } +} + +int ff_ps_apply(AVCodecContext *avctx, PSContext *ps, float L[2][38][64], float R[2][38][64], int top) +{ + float Lbuf[91][32][2]; + float Rbuf[91][32][2]; + const int len = 32; + int is34 = ps->is34bands; + + top += NR_BANDS[is34] - 64; + memset(ps->delay+top, 0, (NR_BANDS[is34] - top)*sizeof(ps->delay[0])); + if (top < NR_ALLPASS_BANDS[is34]) + memset(ps->ap_delay + top, 0, (NR_ALLPASS_BANDS[is34] - top)*sizeof(ps->ap_delay[0])); + + hybrid_analysis(Lbuf, ps->in_buf, L, is34, len); + decorrelation(ps, Rbuf, Lbuf, is34); + stereo_processing(ps, Lbuf, Rbuf, is34); + hybrid_synthesis(L, Lbuf, is34, len); + hybrid_synthesis(R, Rbuf, is34, len); + + return 0; +} + +#define PS_INIT_VLC_STATIC(num, size) \ + INIT_VLC_STATIC(&vlc_ps[num], 9, ps_tmp[num].table_size / ps_tmp[num].elem_size, \ + ps_tmp[num].ps_bits, 1, 1, \ + ps_tmp[num].ps_codes, ps_tmp[num].elem_size, ps_tmp[num].elem_size, \ + size); + +#define PS_VLC_ROW(name) \ + { name ## _codes, name ## _bits, sizeof(name ## _codes), sizeof(name ## _codes[0]) } + +av_cold void ff_ps_init(void) { + // Syntax initialization + static const struct { + const void *ps_codes, *ps_bits; + const unsigned int table_size, elem_size; + } ps_tmp[] = { + PS_VLC_ROW(huff_iid_df1), + PS_VLC_ROW(huff_iid_dt1), + PS_VLC_ROW(huff_iid_df0), + PS_VLC_ROW(huff_iid_dt0), + PS_VLC_ROW(huff_icc_df), + PS_VLC_ROW(huff_icc_dt), + PS_VLC_ROW(huff_ipd_df), + PS_VLC_ROW(huff_ipd_dt), + PS_VLC_ROW(huff_opd_df), + PS_VLC_ROW(huff_opd_dt), + }; + + PS_INIT_VLC_STATIC(0, 1544); + PS_INIT_VLC_STATIC(1, 832); + PS_INIT_VLC_STATIC(2, 1024); + PS_INIT_VLC_STATIC(3, 1036); + PS_INIT_VLC_STATIC(4, 544); + PS_INIT_VLC_STATIC(5, 544); + PS_INIT_VLC_STATIC(6, 512); + PS_INIT_VLC_STATIC(7, 512); + PS_INIT_VLC_STATIC(8, 512); + PS_INIT_VLC_STATIC(9, 512); + + ps_tableinit(); +} + +av_cold void ff_ps_ctx_init(PSContext *ps) +{ +} diff --git a/libavcodec/aacps.h b/libavcodec/aacps.h new file mode 100644 index 0000000000..5fc211a986 --- /dev/null +++ b/libavcodec/aacps.h @@ -0,0 +1,82 @@ +/* + * MPEG-4 Parametric Stereo definitions and declarations + * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com> + * + * This file is part of FFmpeg. + * + * 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. + * + * 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 FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVCODEC_PS_H +#define AVCODEC_PS_H + +#include <stdint.h> + +#include "avcodec.h" +#include "get_bits.h" + +#define PS_MAX_NUM_ENV 5 +#define PS_MAX_NR_IIDICC 34 +#define PS_MAX_NR_IPDOPD 17 +#define PS_MAX_SSB 91 +#define PS_MAX_AP_BANDS 50 +#define PS_QMF_TIME_SLOTS 32 +#define PS_MAX_DELAY 14 +#define PS_AP_LINKS 3 +#define PS_MAX_AP_DELAY 5 + +typedef struct { + int start; + int enable_iid; + int iid_quant; + int nr_iid_par; + int nr_ipdopd_par; + int enable_icc; + int icc_mode; + int nr_icc_par; + int enable_ext; + int frame_class; + int num_env_old; + int num_env; + int enable_ipdopd; + int border_position[PS_MAX_NUM_ENV+1]; + int8_t iid_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; //<Inter-channel Intensity Difference Parameters + int8_t icc_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; //<Inter-Channel Coherence Parameters + /* ipd/opd is iid/icc sized so that the same functions can handle both */ + int8_t ipd_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; //<Inter-channel Phase Difference Parameters + int8_t opd_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; //<Overall Phase Difference Parameters + int is34bands; + int is34bands_old; + + float in_buf[5][44][2]; + float delay[PS_MAX_SSB][PS_QMF_TIME_SLOTS + PS_MAX_DELAY][2]; + float ap_delay[PS_MAX_AP_BANDS][PS_AP_LINKS][PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2]; + float peak_decay_nrg[34]; + float power_smooth[34]; + float peak_decay_diff_smooth[34]; + float H11[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC]; + float H12[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC]; + float H21[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC]; + float H22[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC]; + int8_t opd_hist[PS_MAX_NR_IIDICC]; + int8_t ipd_hist[PS_MAX_NR_IIDICC]; +} PSContext; + +void ff_ps_init(void); +void ff_ps_ctx_init(PSContext *ps); +int ff_ps_read_data(AVCodecContext *avctx, GetBitContext *gb, PSContext *ps, int bits_left); +int ff_ps_apply(AVCodecContext *avctx, PSContext *ps, float L[2][38][64], float R[2][38][64], int top); + +#endif /* AVCODEC_PS_H */ diff --git a/libavcodec/aacps_tablegen.c b/libavcodec/aacps_tablegen.c new file mode 100644 index 0000000000..dc7797f6b8 --- /dev/null +++ b/libavcodec/aacps_tablegen.c @@ -0,0 +1,93 @@ +/* + * Generate a header file for hardcoded Parametric Stereo tables + * + * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com> + * + * This file is part of FFmpeg. + * + * 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. + * + * 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 FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <stdlib.h> +#define CONFIG_HARDCODED_TABLES 0 +#include "aacps_tablegen.h" +#include "tableprint.h" + +void write_float_3d_array (const void *p, int b, int c, int d) +{ + int i; + const float *f = p; + for (i = 0; i < b; i++) { + printf("{\n"); + write_float_2d_array(f, c, d); + printf("},\n"); + f += c * d; + } +} + +void write_float_4d_array (const void *p, int a, int b, int c, int d) +{ + int i; + const float *f = p; + for (i = 0; i < a; i++) { + printf("{\n"); + write_float_3d_array(f, b, c, d); + printf("},\n"); + f += b * c * d; + } +} + +int main(void) +{ + ps_tableinit(); + + write_fileheader(); + + printf("static const float pd_re_smooth[8*8*8] = {\n"); + write_float_array(pd_re_smooth, 8*8*8); + printf("};\n"); + printf("static const float pd_im_smooth[8*8*8] = {\n"); + write_float_array(pd_im_smooth, 8*8*8); + printf("};\n"); + + printf("static const float HA[46][8][4] = {\n"); + write_float_3d_array(HA, 46, 8, 4); + printf("};\n"); + printf("static const float HB[46][8][4] = {\n"); + write_float_3d_array(HB, 46, 8, 4); + printf("};\n"); + + printf("static const float f20_0_8[8][7][2] = {\n"); + write_float_3d_array(f20_0_8, 8, 7, 2); + printf("};\n"); + printf("static const float f34_0_12[12][7][2] = {\n"); + write_float_3d_array(f34_0_12, 12, 7, 2); + printf("};\n"); + printf("static const float f34_1_8[8][7][2] = {\n"); + write_float_3d_array(f34_1_8, 8, 7, 2); + printf("};\n"); + printf("static const float f34_2_4[4][7][2] = {\n"); + write_float_3d_array(f34_2_4, 4, 7, 2); + printf("};\n"); + + printf("static const float Q_fract_allpass[2][50][3][2] = {\n"); + write_float_4d_array(Q_fract_allpass, 2, 50, 3, 2); + printf("};\n"); + printf("static const float phi_fract[2][50][2] = {\n"); + write_float_3d_array(phi_fract, 2, 50, 2); + printf("};\n"); + + return 0; +} diff --git a/libavcodec/aacps_tablegen.h b/libavcodec/aacps_tablegen.h new file mode 100644 index 0000000000..3c1b27decd --- /dev/null +++ b/libavcodec/aacps_tablegen.h @@ -0,0 +1,212 @@ +/* + * Header file for hardcoded Parametric Stereo tables + * + * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com> + * + * This file is part of FFmpeg. + * + * 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. + * + * 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 FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AACPS_TABLEGEN_H +#define AACPS_TABLEGEN_H + +#include <stdint.h> + +#if CONFIG_HARDCODED_TABLES +#define ps_tableinit() +#include "libavcodec/aacps_tables.h" +#else +#include "../libavutil/common.h" +#include "../libavutil/mathematics.h" +#define NR_ALLPASS_BANDS20 30 +#define NR_ALLPASS_BANDS34 50 +#define PS_AP_LINKS 3 +static float pd_re_smooth[8*8*8]; +static float pd_im_smooth[8*8*8]; +static float HA[46][8][4]; +static float HB[46][8][4]; +static float f20_0_8 [ 8][7][2]; +static float f34_0_12[12][7][2]; +static float f34_1_8 [ 8][7][2]; +static float f34_2_4 [ 4][7][2]; +static float Q_fract_allpass[2][50][3][2]; +static float phi_fract[2][50][2]; + +static const float g0_Q8[] = { + 0.00746082949812f, 0.02270420949825f, 0.04546865930473f, 0.07266113929591f, + 0.09885108575264f, 0.11793710567217f, 0.125f +}; + +static const float g0_Q12[] = { + 0.04081179924692f, 0.03812810994926f, 0.05144908135699f, 0.06399831151592f, + 0.07428313801106f, 0.08100347892914f, 0.08333333333333f +}; + +static const float g1_Q8[] = { + 0.01565675600122f, 0.03752716391991f, 0.05417891378782f, 0.08417044116767f, + 0.10307344158036f, 0.12222452249753f, 0.125f +}; + +static const float g2_Q4[] = { + -0.05908211155639f, -0.04871498374946f, 0.0f, 0.07778723915851f, + 0.16486303567403f, 0.23279856662996f, 0.25f +}; + +static void make_filters_from_proto(float (*filter)[7][2], const float *proto, int bands) +{ + int q, n; + for (q = 0; q < bands; q++) { + for (n = 0; n < 7; n++) { + double theta = 2 * M_PI * (q + 0.5) * (n - 6) / bands; + filter[q][n][0] = proto[n] * cos(theta); + filter[q][n][1] = proto[n] * -sin(theta); + } + } +} + +static void ps_tableinit(void) +{ + static const float ipdopd_sin[] = { 0, M_SQRT1_2, 1, M_SQRT1_2, 0, -M_SQRT1_2, -1, -M_SQRT1_2 }; + static const float ipdopd_cos[] = { 1, M_SQRT1_2, 0, -M_SQRT1_2, -1, -M_SQRT1_2, 0, M_SQRT1_2 }; + int pd0, pd1, pd2; + + static const float iid_par_dequant[] = { + //iid_par_dequant_default + 0.05623413251903, 0.12589254117942, 0.19952623149689, 0.31622776601684, + 0.44668359215096, 0.63095734448019, 0.79432823472428, 1, + 1.25892541179417, 1.58489319246111, 2.23872113856834, 3.16227766016838, + 5.01187233627272, 7.94328234724282, 17.7827941003892, + //iid_par_dequant_fine + 0.00316227766017, 0.00562341325190, 0.01, 0.01778279410039, + 0.03162277660168, 0.05623413251903, 0.07943282347243, 0.11220184543020, + 0.15848931924611, 0.22387211385683, 0.31622776601684, 0.39810717055350, + 0.50118723362727, 0.63095734448019, 0.79432823472428, 1, + 1.25892541179417, 1.58489319246111, 1.99526231496888, 2.51188643150958, + 3.16227766016838, 4.46683592150963, 6.30957344480193, 8.91250938133745, + 12.5892541179417, 17.7827941003892, 31.6227766016838, 56.2341325190349, + 100, 177.827941003892, 316.227766016837, + }; + static const float icc_invq[] = { + 1, 0.937, 0.84118, 0.60092, 0.36764, 0, -0.589, -1 + }; + static const float acos_icc_invq[] = { + 0, 0.35685527, 0.57133466, 0.92614472, 1.1943263, M_PI/2, 2.2006171, M_PI + }; + int iid, icc; + + int k, m; + static const int8_t f_center_20[] = { + -3, -1, 1, 3, 5, 7, 10, 14, 18, 22, + }; + static const int8_t f_center_34[] = { + 2, 6, 10, 14, 18, 22, 26, 30, + 34,-10, -6, -2, 51, 57, 15, 21, + 27, 33, 39, 45, 54, 66, 78, 42, + 102, 66, 78, 90,102,114,126, 90, + }; + static const float fractional_delay_links[] = { 0.43f, 0.75f, 0.347f }; + const float fractional_delay_gain = 0.39f; + + for (pd0 = 0; pd0 < 8; pd0++) { + float pd0_re = ipdopd_cos[pd0]; + float pd0_im = ipdopd_sin[pd0]; + for (pd1 = 0; pd1 < 8; pd1++) { + float pd1_re = ipdopd_cos[pd1]; + float pd1_im = ipdopd_sin[pd1]; + for (pd2 = 0; pd2 < 8; pd2++) { + float pd2_re = ipdopd_cos[pd2]; + float pd2_im = ipdopd_sin[pd2]; + float re_smooth = 0.25f * pd0_re + 0.5f * pd1_re + pd2_re; + float im_smooth = 0.25f * pd0_im + 0.5f * pd1_im + pd2_im; + float pd_mag = 1 / sqrt(im_smooth * im_smooth + re_smooth * re_smooth); + pd_re_smooth[pd0*64+pd1*8+pd2] = re_smooth * pd_mag; + pd_im_smooth[pd0*64+pd1*8+pd2] = im_smooth * pd_mag; + } + } + } + + for (iid = 0; iid < 46; iid++) { + float c = iid_par_dequant[iid]; //<Linear Inter-channel Intensity Difference + float c1 = (float)M_SQRT2 / sqrtf(1.0f + c*c); + float c2 = c * c1; + for (icc = 0; icc < 8; icc++) { + /*if (PS_BASELINE || ps->icc_mode < 3)*/ { + float alpha = 0.5f * acos_icc_invq[icc]; + float beta = alpha * (c1 - c2) * (float)M_SQRT1_2; + HA[iid][icc][0] = c2 * cosf(beta + alpha); + HA[iid][icc][1] = c1 * cosf(beta - alpha); + HA[iid][icc][2] = c2 * sinf(beta + alpha); + HA[iid][icc][3] = c1 * sinf(beta - alpha); + } /* else */ { + float alpha, gamma, mu, rho; + float alpha_c, alpha_s, gamma_c, gamma_s; + rho = FFMAX(icc_invq[icc], 0.05f); + alpha = 0.5f * atan2f(2.0f * c * rho, c*c - 1.0f); + mu = c + 1.0f / c; + mu = sqrtf(1 + (4 * rho * rho - 4)/(mu * mu)); + gamma = atanf(sqrtf((1.0f - mu)/(1.0f + mu))); + if (alpha < 0) alpha += M_PI/2; + alpha_c = cosf(alpha); + alpha_s = sinf(alpha); + gamma_c = cosf(gamma); + gamma_s = sinf(gamma); + HB[iid][icc][0] = M_SQRT2 * alpha_c * gamma_c; + HB[iid][icc][1] = M_SQRT2 * alpha_s * gamma_c; + HB[iid][icc][2] = -M_SQRT2 * alpha_s * gamma_s; + HB[iid][icc][3] = M_SQRT2 * alpha_c * gamma_s; + } + } + } + + for (k = 0; k < NR_ALLPASS_BANDS20; k++) { + double f_center, theta; + if (k < FF_ARRAY_ELEMS(f_center_20)) + f_center = f_center_20[k] * 0.125; + else + f_center = k - 6.5f; + for (m = 0; m < PS_AP_LINKS; m++) { + theta = -M_PI * fractional_delay_links[m] * f_center; + Q_fract_allpass[0][k][m][0] = cos(theta); + Q_fract_allpass[0][k][m][1] = sin(theta); + } + theta = -M_PI*fractional_delay_gain*f_center; + phi_fract[0][k][0] = cos(theta); + phi_fract[0][k][1] = sin(theta); + } + for (k = 0; k < NR_ALLPASS_BANDS34; k++) { + double f_center, theta; + if (k < FF_ARRAY_ELEMS(f_center_34)) + f_center = f_center_34[k] / 24.; + else + f_center = k - 26.5f; + for (m = 0; m < PS_AP_LINKS; m++) { + theta = -M_PI * fractional_delay_links[m] * f_center; + Q_fract_allpass[1][k][m][0] = cos(theta); + Q_fract_allpass[1][k][m][1] = sin(theta); + } + theta = -M_PI*fractional_delay_gain*f_center; + phi_fract[1][k][0] = cos(theta); + phi_fract[1][k][1] = sin(theta); + } + + make_filters_from_proto(f20_0_8, g0_Q8, 8); + make_filters_from_proto(f34_0_12, g0_Q12, 12); + make_filters_from_proto(f34_1_8, g1_Q8, 8); + make_filters_from_proto(f34_2_4, g2_Q4, 4); +} +#endif /* CONFIG_HARDCODED_TABLES */ + +#endif /* AACPS_TABLEGEN_H */ diff --git a/libavcodec/aacpsdata.c b/libavcodec/aacpsdata.c new file mode 100644 index 0000000000..7431caebc6 --- /dev/null +++ b/libavcodec/aacpsdata.c @@ -0,0 +1,163 @@ +/* + * MPEG-4 Parametric Stereo data tables + * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com> + * + * This file is part of FFmpeg. + * + * 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. + * + * 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 FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +static const uint8_t huff_iid_df1_bits[] = { + 18, 18, 18, 18, 18, 18, 18, 18, 18, 17, 18, 17, 17, 16, 16, 15, 14, 14, + 13, 12, 12, 11, 10, 10, 8, 7, 6, 5, 4, 3, 1, 3, 4, 5, 6, 7, + 8, 9, 10, 11, 11, 12, 13, 14, 14, 15, 16, 16, 17, 17, 18, 17, 18, 18, + 18, 18, 18, 18, 18, 18, 18, +}; + +static const uint32_t huff_iid_df1_codes[] = { + 0x01FEB4, 0x01FEB5, 0x01FD76, 0x01FD77, 0x01FD74, 0x01FD75, 0x01FE8A, + 0x01FE8B, 0x01FE88, 0x00FE80, 0x01FEB6, 0x00FE82, 0x00FEB8, 0x007F42, + 0x007FAE, 0x003FAF, 0x001FD1, 0x001FE9, 0x000FE9, 0x0007EA, 0x0007FB, + 0x0003FB, 0x0001FB, 0x0001FF, 0x00007C, 0x00003C, 0x00001C, 0x00000C, + 0x000000, 0x000001, 0x000001, 0x000002, 0x000001, 0x00000D, 0x00001D, + 0x00003D, 0x00007D, 0x0000FC, 0x0001FC, 0x0003FC, 0x0003F4, 0x0007EB, + 0x000FEA, 0x001FEA, 0x001FD6, 0x003FD0, 0x007FAF, 0x007F43, 0x00FEB9, + 0x00FE83, 0x01FEB7, 0x00FE81, 0x01FE89, 0x01FE8E, 0x01FE8F, 0x01FE8C, + 0x01FE8D, 0x01FEB2, 0x01FEB3, 0x01FEB0, 0x01FEB1, +}; + +static const uint8_t huff_iid_dt1_bits[] = { + 16, 16, 16, 16, 16, 16, 16, 16, 16, 15, 15, 15, 15, 15, 15, 14, 14, 13, + 13, 13, 12, 12, 11, 10, 9, 9, 7, 6, 5, 3, 1, 2, 5, 6, 7, 8, + 9, 10, 11, 11, 12, 12, 13, 13, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16, + 16, 16, 16, 16, 16, 16, 16, +}; + +static const uint16_t huff_iid_dt1_codes[] = { + 0x004ED4, 0x004ED5, 0x004ECE, 0x004ECF, 0x004ECC, 0x004ED6, 0x004ED8, + 0x004F46, 0x004F60, 0x002718, 0x002719, 0x002764, 0x002765, 0x00276D, + 0x0027B1, 0x0013B7, 0x0013D6, 0x0009C7, 0x0009E9, 0x0009ED, 0x0004EE, + 0x0004F7, 0x000278, 0x000139, 0x00009A, 0x00009F, 0x000020, 0x000011, + 0x00000A, 0x000003, 0x000001, 0x000000, 0x00000B, 0x000012, 0x000021, + 0x00004C, 0x00009B, 0x00013A, 0x000279, 0x000270, 0x0004EF, 0x0004E2, + 0x0009EA, 0x0009D8, 0x0013D7, 0x0013D0, 0x0027B2, 0x0027A2, 0x00271A, + 0x00271B, 0x004F66, 0x004F67, 0x004F61, 0x004F47, 0x004ED9, 0x004ED7, + 0x004ECD, 0x004ED2, 0x004ED3, 0x004ED0, 0x004ED1, +}; + +static const uint8_t huff_iid_df0_bits[] = { + 17, 17, 17, 17, 16, 15, 13, 10, 9, 7, 6, 5, 4, 3, 1, 3, 4, 5, + 6, 6, 8, 11, 13, 14, 14, 15, 17, 18, 18, +}; + +static const uint32_t huff_iid_df0_codes[] = { + 0x01FFFB, 0x01FFFC, 0x01FFFD, 0x01FFFA, 0x00FFFC, 0x007FFC, 0x001FFD, + 0x0003FE, 0x0001FE, 0x00007E, 0x00003C, 0x00001D, 0x00000D, 0x000005, + 0x000000, 0x000004, 0x00000C, 0x00001C, 0x00003D, 0x00003E, 0x0000FE, + 0x0007FE, 0x001FFC, 0x003FFC, 0x003FFD, 0x007FFD, 0x01FFFE, 0x03FFFE, + 0x03FFFF, +}; + +static const uint8_t huff_iid_dt0_bits[] = { + 19, 19, 19, 20, 20, 20, 17, 15, 12, 10, 8, 6, 4, 2, 1, 3, 5, 7, + 9, 11, 13, 14, 17, 19, 20, 20, 20, 20, 20, +}; + +static const uint32_t huff_iid_dt0_codes[] = { + 0x07FFF9, 0x07FFFA, 0x07FFFB, 0x0FFFF8, 0x0FFFF9, 0x0FFFFA, 0x01FFFD, + 0x007FFE, 0x000FFE, 0x0003FE, 0x0000FE, 0x00003E, 0x00000E, 0x000002, + 0x000000, 0x000006, 0x00001E, 0x00007E, 0x0001FE, 0x0007FE, 0x001FFE, + 0x003FFE, 0x01FFFC, 0x07FFF8, 0x0FFFFB, 0x0FFFFC, 0x0FFFFD, 0x0FFFFE, + 0x0FFFFF, +}; + +static const uint8_t huff_icc_df_bits[] = { + 14, 14, 12, 10, 7, 5, 3, 1, 2, 4, 6, 8, 9, 11, 13, +}; + +static const uint16_t huff_icc_df_codes[] = { + 0x3FFF, 0x3FFE, 0x0FFE, 0x03FE, 0x007E, 0x001E, 0x0006, 0x0000, + 0x0002, 0x000E, 0x003E, 0x00FE, 0x01FE, 0x07FE, 0x1FFE, +}; + +static const uint8_t huff_icc_dt_bits[] = { + 14, 13, 11, 9, 7, 5, 3, 1, 2, 4, 6, 8, 10, 12, 14, +}; + +static const uint16_t huff_icc_dt_codes[] = { + 0x3FFE, 0x1FFE, 0x07FE, 0x01FE, 0x007E, 0x001E, 0x0006, 0x0000, + 0x0002, 0x000E, 0x003E, 0x00FE, 0x03FE, 0x0FFE, 0x3FFF, +}; + +static const uint8_t huff_ipd_df_bits[] = { + 1, 3, 4, 4, 4, 4, 4, 4, +}; + +static const uint8_t huff_ipd_df_codes[] = { + 0x01, 0x00, 0x06, 0x04, 0x02, 0x03, 0x05, 0x07, +}; + +static const uint8_t huff_ipd_dt_bits[] = { + 1, 3, 4, 5, 5, 4, 4, 3, +}; + +static const uint8_t huff_ipd_dt_codes[] = { + 0x01, 0x02, 0x02, 0x03, 0x02, 0x00, 0x03, 0x03, +}; + +static const uint8_t huff_opd_df_bits[] = { + 1, 3, 4, 4, 5, 5, 4, 3, +}; + +static const uint8_t huff_opd_df_codes[] = { + 0x01, 0x01, 0x06, 0x04, 0x0F, 0x0E, 0x05, 0x00, +}; + +static const uint8_t huff_opd_dt_bits[] = { + 1, 3, 4, 5, 5, 4, 4, 3, +}; + +static const uint8_t huff_opd_dt_codes[] = { + 0x01, 0x02, 0x01, 0x07, 0x06, 0x00, 0x02, 0x03, +}; + +static const int8_t huff_offset[] = { + 30, 30, + 14, 14, + 7, 7, + 0, 0, + 0, 0, +}; + +///Table 8.48 +static const int8_t k_to_i_20[] = { + 1, 0, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 14, 15, + 15, 15, 16, 16, 16, 16, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, + 18, 18, 18, 18, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, + 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19 +}; +///Table 8.49 +static const int8_t k_to_i_34[] = { + 0, 1, 2, 3, 4, 5, 6, 6, 7, 2, 1, 0, 10, 10, 4, 5, 6, 7, 8, + 9, 10, 11, 12, 9, 14, 11, 12, 13, 14, 15, 16, 13, 16, 17, 18, 19, 20, 21, + 22, 22, 23, 23, 24, 24, 25, 25, 26, 26, 27, 27, 27, 28, 28, 28, 29, 29, 29, + 30, 30, 30, 31, 31, 31, 31, 32, 32, 32, 32, 33, 33, 33, 33, 33, 33, 33, 33, + 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33 +}; + +static const float g1_Q2[] = { + 0.0f, 0.01899487526049f, 0.0f, -0.07293139167538f, + 0.0f, 0.30596630545168f, 0.5f +}; diff --git a/libavcodec/aacsbr.c b/libavcodec/aacsbr.c index 0de89c2315..ca5c52cce3 100644 --- a/libavcodec/aacsbr.c +++ b/libavcodec/aacsbr.c @@ -31,6 +31,7 @@ #include "aacsbr.h" #include "aacsbrdata.h" #include "fft.h" +#include "aacps.h" #include <stdint.h> #include <float.h> @@ -71,9 +72,6 @@ enum { static VLC vlc_sbr[10]; static const int8_t vlc_sbr_lav[10] = { 60, 60, 24, 24, 31, 31, 12, 12, 31, 12 }; -static DECLARE_ALIGNED(16, float, analysis_cos_pre)[64]; -static DECLARE_ALIGNED(16, float, analysis_sin_pre)[64]; -static DECLARE_ALIGNED(16, float, analysis_cossin_post)[32][2]; static const DECLARE_ALIGNED(16, float, zero64)[64]; #define SBR_INIT_VLC_STATIC(num, size) \ @@ -87,7 +85,7 @@ static const DECLARE_ALIGNED(16, float, zero64)[64]; av_cold void ff_aac_sbr_init(void) { - int n, k; + int n; static const struct { const void *sbr_codes, *sbr_bits; const unsigned int table_size, elem_size; @@ -116,16 +114,6 @@ av_cold void ff_aac_sbr_init(void) SBR_INIT_VLC_STATIC(8, 592); SBR_INIT_VLC_STATIC(9, 512); - for (n = 0; n < 64; n++) { - float pre = M_PI * n / 64; - analysis_cos_pre[n] = cosf(pre); - analysis_sin_pre[n] = sinf(pre); - } - for (k = 0; k < 32; k++) { - float post = M_PI * (k + 0.5) / 128; - analysis_cossin_post[k][0] = 4.0 * cosf(post); - analysis_cossin_post[k][1] = -4.0 * sinf(post); - } for (n = 1; n < 320; n++) sbr_qmf_window_us[320 + n] = sbr_qmf_window_us[320 - n]; sbr_qmf_window_us[384] = -sbr_qmf_window_us[384]; @@ -133,6 +121,8 @@ av_cold void ff_aac_sbr_init(void) for (n = 0; n < 320; n++) sbr_qmf_window_ds[n] = sbr_qmf_window_us[2*n]; + + ff_ps_init(); } av_cold void ff_aac_sbr_ctx_init(SpectralBandReplication *sbr) @@ -142,13 +132,14 @@ av_cold void ff_aac_sbr_ctx_init(SpectralBandReplication *sbr) sbr->data[0].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128); sbr->data[1].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128); ff_mdct_init(&sbr->mdct, 7, 1, 1.0/64); - ff_rdft_init(&sbr->rdft, 6, IDFT_R2C); + ff_mdct_init(&sbr->mdct_ana, 7, 1, -2.0); + ff_ps_ctx_init(&sbr->ps); } av_cold void ff_aac_sbr_ctx_close(SpectralBandReplication *sbr) { ff_mdct_end(&sbr->mdct); - ff_rdft_end(&sbr->rdft); + ff_mdct_end(&sbr->mdct_ana); } static int qsort_comparison_function_int16(const void *a, const void *b) @@ -293,15 +284,15 @@ static void make_bands(int16_t* bands, int start, int stop, int num_bands) bands[num_bands-1] = stop - previous; } -static int check_n_master(AVCodecContext *avccontext, int n_master, int bs_xover_band) +static int check_n_master(AVCodecContext *avctx, int n_master, int bs_xover_band) { // Requirements (14496-3 sp04 p205) if (n_master <= 0) { - av_log(avccontext, AV_LOG_ERROR, "Invalid n_master: %d\n", n_master); + av_log(avctx, AV_LOG_ERROR, "Invalid n_master: %d\n", n_master); return -1; } if (bs_xover_band >= n_master) { - av_log(avccontext, AV_LOG_ERROR, + av_log(avctx, AV_LOG_ERROR, "Invalid bitstream, crossover band index beyond array bounds: %d\n", bs_xover_band); return -1; @@ -349,7 +340,7 @@ static int sbr_make_f_master(AACContext *ac, SpectralBandReplication *sbr, sbr_offset_ptr = sbr_offset[5]; break; default: - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Unsupported sample rate for SBR: %d\n", sbr->sample_rate); return -1; } @@ -367,7 +358,7 @@ static int sbr_make_f_master(AACContext *ac, SpectralBandReplication *sbr, } else if (spectrum->bs_stop_freq == 15) { sbr->k[2] = 3*sbr->k[0]; } else { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Invalid bs_stop_freq: %d\n", spectrum->bs_stop_freq); return -1; } @@ -382,18 +373,17 @@ static int sbr_make_f_master(AACContext *ac, SpectralBandReplication *sbr, max_qmf_subbands = 32; if (sbr->k[2] - sbr->k[0] > max_qmf_subbands) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Invalid bitstream, too many QMF subbands: %d\n", sbr->k[2] - sbr->k[0]); return -1; } if (!spectrum->bs_freq_scale) { - unsigned int dk; - int k2diff; + int dk, k2diff; dk = spectrum->bs_alter_scale + 1; sbr->n_master = ((sbr->k[2] - sbr->k[0] + (dk&2)) >> dk) << 1; - if (check_n_master(ac->avccontext, sbr->n_master, sbr->spectrum_params.bs_xover_band)) + if (check_n_master(ac->avctx, sbr->n_master, sbr->spectrum_params.bs_xover_band)) return -1; for (k = 1; k <= sbr->n_master; k++) @@ -428,7 +418,7 @@ static int sbr_make_f_master(AACContext *ac, SpectralBandReplication *sbr, num_bands_0 = lrintf(half_bands * log2f(sbr->k[1] / (float)sbr->k[0])) * 2; if (num_bands_0 <= 0) { // Requirements (14496-3 sp04 p205) - av_log(ac->avccontext, AV_LOG_ERROR, "Invalid num_bands_0: %d\n", num_bands_0); + av_log(ac->avctx, AV_LOG_ERROR, "Invalid num_bands_0: %d\n", num_bands_0); return -1; } @@ -442,7 +432,7 @@ static int sbr_make_f_master(AACContext *ac, SpectralBandReplication *sbr, vk0[0] = sbr->k[0]; for (k = 1; k <= num_bands_0; k++) { if (vk0[k] <= 0) { // Requirements (14496-3 sp04 p205) - av_log(ac->avccontext, AV_LOG_ERROR, "Invalid vDk0[%d]: %d\n", k, vk0[k]); + av_log(ac->avctx, AV_LOG_ERROR, "Invalid vDk0[%d]: %d\n", k, vk0[k]); return -1; } vk0[k] += vk0[k-1]; @@ -472,14 +462,14 @@ static int sbr_make_f_master(AACContext *ac, SpectralBandReplication *sbr, vk1[0] = sbr->k[1]; for (k = 1; k <= num_bands_1; k++) { if (vk1[k] <= 0) { // Requirements (14496-3 sp04 p205) - av_log(ac->avccontext, AV_LOG_ERROR, "Invalid vDk1[%d]: %d\n", k, vk1[k]); + av_log(ac->avctx, AV_LOG_ERROR, "Invalid vDk1[%d]: %d\n", k, vk1[k]); return -1; } vk1[k] += vk1[k-1]; } sbr->n_master = num_bands_0 + num_bands_1; - if (check_n_master(ac->avccontext, sbr->n_master, sbr->spectrum_params.bs_xover_band)) + if (check_n_master(ac->avctx, sbr->n_master, sbr->spectrum_params.bs_xover_band)) return -1; memcpy(&sbr->f_master[0], vk0, (num_bands_0 + 1) * sizeof(sbr->f_master[0])); @@ -488,7 +478,7 @@ static int sbr_make_f_master(AACContext *ac, SpectralBandReplication *sbr, } else { sbr->n_master = num_bands_0; - if (check_n_master(ac->avccontext, sbr->n_master, sbr->spectrum_params.bs_xover_band)) + if (check_n_master(ac->avctx, sbr->n_master, sbr->spectrum_params.bs_xover_band)) return -1; memcpy(sbr->f_master, vk0, (num_bands_0 + 1) * sizeof(sbr->f_master[0])); } @@ -524,7 +514,7 @@ static int sbr_hf_calc_npatches(AACContext *ac, SpectralBandReplication *sbr) // illegal however the Coding Technologies decoder check stream has a final // count of 6 patches if (sbr->num_patches > 5) { - av_log(ac->avccontext, AV_LOG_ERROR, "Too many patches: %d\n", sbr->num_patches); + av_log(ac->avctx, AV_LOG_ERROR, "Too many patches: %d\n", sbr->num_patches); return -1; } @@ -563,12 +553,12 @@ static int sbr_make_f_derived(AACContext *ac, SpectralBandReplication *sbr) // Requirements (14496-3 sp04 p205) if (sbr->kx[1] + sbr->m[1] > 64) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Stop frequency border too high: %d\n", sbr->kx[1] + sbr->m[1]); return -1; } if (sbr->kx[1] > 32) { - av_log(ac->avccontext, AV_LOG_ERROR, "Start frequency border too high: %d\n", sbr->kx[1]); + av_log(ac->avctx, AV_LOG_ERROR, "Start frequency border too high: %d\n", sbr->kx[1]); return -1; } @@ -580,7 +570,7 @@ static int sbr_make_f_derived(AACContext *ac, SpectralBandReplication *sbr) sbr->n_q = FFMAX(1, lrintf(sbr->spectrum_params.bs_noise_bands * log2f(sbr->k[2] / (float)sbr->kx[1]))); // 0 <= bs_noise_bands <= 3 if (sbr->n_q > 5) { - av_log(ac->avccontext, AV_LOG_ERROR, "Too many noise floor scale factors: %d\n", sbr->n_q); + av_log(ac->avctx, AV_LOG_ERROR, "Too many noise floor scale factors: %d\n", sbr->n_q); return -1; } @@ -638,7 +628,7 @@ static int read_sbr_grid(AACContext *ac, SpectralBandReplication *sbr, ch_data->bs_amp_res = 0; if (ch_data->bs_num_env > 4) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Invalid bitstream, too many SBR envelopes in FIXFIX type SBR frame: %d\n", ch_data->bs_num_env); return -1; @@ -693,7 +683,7 @@ static int read_sbr_grid(AACContext *ac, SpectralBandReplication *sbr, ch_data->bs_num_env = num_rel_lead + num_rel_trail + 1; if (ch_data->bs_num_env > 5) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Invalid bitstream, too many SBR envelopes in VARVAR type SBR frame: %d\n", ch_data->bs_num_env); return -1; @@ -714,7 +704,7 @@ static int read_sbr_grid(AACContext *ac, SpectralBandReplication *sbr, } if (bs_pointer > ch_data->bs_num_env + 1) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Invalid bitstream, bs_pointer points to a middle noise border outside the time borders table: %d\n", bs_pointer); return -1; @@ -722,7 +712,7 @@ static int read_sbr_grid(AACContext *ac, SpectralBandReplication *sbr, for (i = 1; i <= ch_data->bs_num_env; i++) { if (ch_data->t_env[i-1] > ch_data->t_env[i]) { - av_log(ac->avccontext, AV_LOG_ERROR, "Non monotone time borders\n"); + av_log(ac->avctx, AV_LOG_ERROR, "Non monotone time borders\n"); return -1; } } @@ -903,25 +893,24 @@ static void read_sbr_extension(AACContext *ac, SpectralBandReplication *sbr, GetBitContext *gb, int bs_extension_id, int *num_bits_left) { -//TODO - implement ps_data for parametric stereo parsing switch (bs_extension_id) { case EXTENSION_ID_PS: if (!ac->m4ac.ps) { - av_log(ac->avccontext, AV_LOG_ERROR, "Parametric Stereo signaled to be not-present but was found in the bitstream.\n"); + av_log(ac->avctx, AV_LOG_ERROR, "Parametric Stereo signaled to be not-present but was found in the bitstream.\n"); skip_bits_long(gb, *num_bits_left); // bs_fill_bits *num_bits_left = 0; } else { -#if 0 - *num_bits_left -= ff_ps_data(gb, ps); +#if 1 + *num_bits_left -= ff_ps_read_data(ac->avctx, gb, &sbr->ps, *num_bits_left); #else - av_log_missing_feature(ac->avccontext, "Parametric Stereo is", 0); + av_log_missing_feature(ac->avctx, "Parametric Stereo is", 0); skip_bits_long(gb, *num_bits_left); // bs_fill_bits *num_bits_left = 0; #endif } break; default: - av_log_missing_feature(ac->avccontext, "Reserved SBR extensions are", 1); + av_log_missing_feature(ac->avctx, "Reserved SBR extensions are", 1); skip_bits_long(gb, *num_bits_left); // bs_fill_bits *num_bits_left = 0; break; @@ -1006,7 +995,7 @@ static unsigned int read_sbr_data(AACContext *ac, SpectralBandReplication *sbr, return get_bits_count(gb) - cnt; } } else { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Invalid bitstream - cannot apply SBR to element type %d\n", id_aac); sbr->start = 0; return get_bits_count(gb) - cnt; @@ -1021,6 +1010,11 @@ static unsigned int read_sbr_data(AACContext *ac, SpectralBandReplication *sbr, num_bits_left -= 2; read_sbr_extension(ac, sbr, gb, get_bits(gb, 2), &num_bits_left); // bs_extension_id } + if (num_bits_left < 0) { + av_log(ac->avctx, AV_LOG_ERROR, "SBR Extension over read.\n"); + } + if (num_bits_left > 0) + skip_bits(gb, num_bits_left); } return get_bits_count(gb) - cnt; @@ -1033,7 +1027,7 @@ static void sbr_reset(AACContext *ac, SpectralBandReplication *sbr) if (err >= 0) err = sbr_make_f_derived(ac, sbr); if (err < 0) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "SBR reset failed. Switching SBR to pure upsampling mode.\n"); sbr->start = 0; } @@ -1085,7 +1079,7 @@ int ff_decode_sbr_extension(AACContext *ac, SpectralBandReplication *sbr, bytes_read = ((num_sbr_bits + num_align_bits + 4) >> 3); if (bytes_read > cnt) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "Expected to read %d SBR bytes actually read %d.\n", cnt, bytes_read); } return cnt; @@ -1139,7 +1133,7 @@ static void sbr_dequant(SpectralBandReplication *sbr, int id_aac) * @param x pointer to the beginning of the first sample window * @param W array of complex-valued samples split into subbands */ -static void sbr_qmf_analysis(DSPContext *dsp, RDFTContext *rdft, const float *in, float *x, +static void sbr_qmf_analysis(DSPContext *dsp, FFTContext *mdct, const float *in, float *x, float z[320], float W[2][32][32][2], float scale) { @@ -1152,23 +1146,23 @@ static void sbr_qmf_analysis(DSPContext *dsp, RDFTContext *rdft, const float *in memcpy(x+288, in, 1024*sizeof(*x)); for (i = 0; i < 32; i++) { // numTimeSlots*RATE = 16*2 as 960 sample frames // are not supported - float re, im; dsp->vector_fmul_reverse(z, sbr_qmf_window_ds, x, 320); for (k = 0; k < 64; k++) { float f = z[k] + z[k + 64] + z[k + 128] + z[k + 192] + z[k + 256]; - z[k] = f * analysis_cos_pre[k]; - z[k+64] = f; + z[k] = f; } - ff_rdft_calc(rdft, z); - re = z[0] * 0.5f; - im = 0.5f * dsp->scalarproduct_float(z+64, analysis_sin_pre, 64); - W[1][i][0][0] = re * analysis_cossin_post[0][0] - im * analysis_cossin_post[0][1]; - W[1][i][0][1] = re * analysis_cossin_post[0][1] + im * analysis_cossin_post[0][0]; + //Shuffle to IMDCT + z[64] = z[0]; for (k = 1; k < 32; k++) { - re = z[2*k ] - re; - im = z[2*k+1] - im; - W[1][i][k][0] = re * analysis_cossin_post[k][0] - im * analysis_cossin_post[k][1]; - W[1][i][k][1] = re * analysis_cossin_post[k][1] + im * analysis_cossin_post[k][0]; + z[64+2*k-1] = z[ k]; + z[64+2*k ] = -z[64-k]; + } + z[64+63] = z[32]; + + ff_imdct_half(mdct, z, z+64); + for (k = 0; k < 32; k++) { + W[1][i][k][0] = -z[63-k]; + W[1][i][k][1] = z[k]; } x += 32; } @@ -1179,7 +1173,7 @@ static void sbr_qmf_analysis(DSPContext *dsp, RDFTContext *rdft, const float *in * (14496-3 sp04 p206) */ static void sbr_qmf_synthesis(DSPContext *dsp, FFTContext *mdct, - float *out, float X[2][32][64], + float *out, float X[2][38][64], float mdct_buf[2][64], float *v0, int *v_off, const unsigned int div, float bias, float scale) @@ -1197,21 +1191,22 @@ static void sbr_qmf_synthesis(DSPContext *dsp, FFTContext *mdct, *v_off -= 128 >> div; } v = v0 + *v_off; - for (n = 1; n < 64 >> div; n+=2) { - X[1][i][n] = -X[1][i][n]; - } - if (div) { - memset(X[0][i]+32, 0, 32*sizeof(float)); - memset(X[1][i]+32, 0, 32*sizeof(float)); - } - ff_imdct_half(mdct, mdct_buf[0], X[0][i]); - ff_imdct_half(mdct, mdct_buf[1], X[1][i]); if (div) { for (n = 0; n < 32; n++) { - v[ n] = -mdct_buf[0][63 - 2*n] + mdct_buf[1][2*n ]; - v[ 63 - n] = mdct_buf[0][62 - 2*n] + mdct_buf[1][2*n + 1]; + X[0][i][ n] = -X[0][i][n]; + X[0][i][32+n] = X[1][i][31-n]; + } + ff_imdct_half(mdct, mdct_buf[0], X[0][i]); + for (n = 0; n < 32; n++) { + v[ n] = mdct_buf[0][63 - 2*n]; + v[63 - n] = -mdct_buf[0][62 - 2*n]; } } else { + for (n = 1; n < 64; n+=2) { + X[1][i][n] = -X[1][i][n]; + } + ff_imdct_half(mdct, mdct_buf[0], X[0][i]); + ff_imdct_half(mdct, mdct_buf[1], X[1][i]); for (n = 0; n < 64; n++) { v[ n] = -mdct_buf[0][63 - n] + mdct_buf[1][ n ]; v[127 - n] = mdct_buf[0][63 - n] + mdct_buf[1][ n ]; @@ -1380,7 +1375,7 @@ static int sbr_hf_gen(AACContext *ac, SpectralBandReplication *sbr, g--; if (g < 0) { - av_log(ac->avccontext, AV_LOG_ERROR, + av_log(ac->avctx, AV_LOG_ERROR, "ERROR : no subband found for frequency %d\n", k); return -1; } @@ -1414,7 +1409,7 @@ static int sbr_hf_gen(AACContext *ac, SpectralBandReplication *sbr, } /// Generate the subband filtered lowband -static int sbr_x_gen(SpectralBandReplication *sbr, float X[2][32][64], +static int sbr_x_gen(SpectralBandReplication *sbr, float X[2][38][64], const float X_low[32][40][2], const float Y[2][38][64][2], int ch) { @@ -1436,7 +1431,7 @@ static int sbr_x_gen(SpectralBandReplication *sbr, float X[2][32][64], } for (k = 0; k < sbr->kx[1]; k++) { - for (i = i_Temp; i < i_f; i++) { + for (i = i_Temp; i < 38; i++) { X[0][i][k] = X_low[k][i + ENVELOPE_ADJUSTMENT_OFFSET][0]; X[1][i][k] = X_low[k][i + ENVELOPE_ADJUSTMENT_OFFSET][1]; } @@ -1730,7 +1725,7 @@ void ff_sbr_apply(AACContext *ac, SpectralBandReplication *sbr, int id_aac, } for (ch = 0; ch < nch; ch++) { /* decode channel */ - sbr_qmf_analysis(&ac->dsp, &sbr->rdft, ch ? R : L, sbr->data[ch].analysis_filterbank_samples, + sbr_qmf_analysis(&ac->dsp, &sbr->mdct_ana, ch ? R : L, sbr->data[ch].analysis_filterbank_samples, (float*)sbr->qmf_filter_scratch, sbr->data[ch].W, 1/(-1024 * ac->sf_scale)); sbr_lf_gen(ac, sbr, sbr->X_low, sbr->data[ch].W); @@ -1752,6 +1747,16 @@ void ff_sbr_apply(AACContext *ac, SpectralBandReplication *sbr, int id_aac, /* synthesis */ sbr_x_gen(sbr, sbr->X[ch], sbr->X_low, sbr->data[ch].Y, ch); } + + if (ac->m4ac.ps == 1) { + if (sbr->ps.start) { + ff_ps_apply(ac->avctx, &sbr->ps, sbr->X[0], sbr->X[1], sbr->kx[1] + sbr->m[1]); + } else { + memcpy(sbr->X[1], sbr->X[0], sizeof(sbr->X[0])); + } + nch = 2; + } + sbr_qmf_synthesis(&ac->dsp, &sbr->mdct, L, sbr->X[0], sbr->qmf_filter_scratch, sbr->data[0].synthesis_filterbank_samples, &sbr->data[0].synthesis_filterbank_samples_offset, diff --git a/libavcodec/aactab.c b/libavcodec/aactab.c index eaa5486822..17102a66f5 100644 --- a/libavcodec/aactab.c +++ b/libavcodec/aactab.c @@ -29,6 +29,7 @@ #include "libavutil/mem.h" #include "aac.h" +#include "aac_tablegen.h" #include <stdint.h> @@ -1204,129 +1205,3 @@ const uint8_t ff_tns_max_bands_128[] = { 9, 9, 10, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14 }; // @} - - -#if CONFIG_HARDCODED_TABLES - -/** - * Table of pow(2, (i - 200)/4.) used for different purposes depending on the - * range of indices to the table: - * [ 0, 255] scale factor decoding when using C dsp.float_to_int16 - * [60, 315] scale factor decoding when using SIMD dsp.float_to_int16 - * [45, 300] intensity stereo position decoding mapped in reverse order i.e. 0->300, 1->299, ..., 254->46, 255->45 - */ -const float ff_aac_pow2sf_tab[428] = { - 8.88178420e-16, 1.05622810e-15, 1.25607397e-15, 1.49373210e-15, - 1.77635684e-15, 2.11245619e-15, 2.51214793e-15, 2.98746420e-15, - 3.55271368e-15, 4.22491238e-15, 5.02429587e-15, 5.97492839e-15, - 7.10542736e-15, 8.44982477e-15, 1.00485917e-14, 1.19498568e-14, - 1.42108547e-14, 1.68996495e-14, 2.00971835e-14, 2.38997136e-14, - 2.84217094e-14, 3.37992991e-14, 4.01943669e-14, 4.77994272e-14, - 5.68434189e-14, 6.75985982e-14, 8.03887339e-14, 9.55988543e-14, - 1.13686838e-13, 1.35197196e-13, 1.60777468e-13, 1.91197709e-13, - 2.27373675e-13, 2.70394393e-13, 3.21554936e-13, 3.82395417e-13, - 4.54747351e-13, 5.40788785e-13, 6.43109871e-13, 7.64790834e-13, - 9.09494702e-13, 1.08157757e-12, 1.28621974e-12, 1.52958167e-12, - 1.81898940e-12, 2.16315514e-12, 2.57243948e-12, 3.05916334e-12, - 3.63797881e-12, 4.32631028e-12, 5.14487897e-12, 6.11832668e-12, - 7.27595761e-12, 8.65262056e-12, 1.02897579e-11, 1.22366534e-11, - 1.45519152e-11, 1.73052411e-11, 2.05795159e-11, 2.44733067e-11, - 2.91038305e-11, 3.46104823e-11, 4.11590317e-11, 4.89466134e-11, - 5.82076609e-11, 6.92209645e-11, 8.23180635e-11, 9.78932268e-11, - 1.16415322e-10, 1.38441929e-10, 1.64636127e-10, 1.95786454e-10, - 2.32830644e-10, 2.76883858e-10, 3.29272254e-10, 3.91572907e-10, - 4.65661287e-10, 5.53767716e-10, 6.58544508e-10, 7.83145814e-10, - 9.31322575e-10, 1.10753543e-09, 1.31708902e-09, 1.56629163e-09, - 1.86264515e-09, 2.21507086e-09, 2.63417803e-09, 3.13258326e-09, - 3.72529030e-09, 4.43014173e-09, 5.26835606e-09, 6.26516652e-09, - 7.45058060e-09, 8.86028346e-09, 1.05367121e-08, 1.25303330e-08, - 1.49011612e-08, 1.77205669e-08, 2.10734243e-08, 2.50606661e-08, - 2.98023224e-08, 3.54411338e-08, 4.21468485e-08, 5.01213321e-08, - 5.96046448e-08, 7.08822677e-08, 8.42936970e-08, 1.00242664e-07, - 1.19209290e-07, 1.41764535e-07, 1.68587394e-07, 2.00485328e-07, - 2.38418579e-07, 2.83529071e-07, 3.37174788e-07, 4.00970657e-07, - 4.76837158e-07, 5.67058141e-07, 6.74349576e-07, 8.01941314e-07, - 9.53674316e-07, 1.13411628e-06, 1.34869915e-06, 1.60388263e-06, - 1.90734863e-06, 2.26823256e-06, 2.69739830e-06, 3.20776526e-06, - 3.81469727e-06, 4.53646513e-06, 5.39479661e-06, 6.41553051e-06, - 7.62939453e-06, 9.07293026e-06, 1.07895932e-05, 1.28310610e-05, - 1.52587891e-05, 1.81458605e-05, 2.15791864e-05, 2.56621220e-05, - 3.05175781e-05, 3.62917210e-05, 4.31583729e-05, 5.13242441e-05, - 6.10351562e-05, 7.25834421e-05, 8.63167458e-05, 1.02648488e-04, - 1.22070312e-04, 1.45166884e-04, 1.72633492e-04, 2.05296976e-04, - 2.44140625e-04, 2.90333768e-04, 3.45266983e-04, 4.10593953e-04, - 4.88281250e-04, 5.80667537e-04, 6.90533966e-04, 8.21187906e-04, - 9.76562500e-04, 1.16133507e-03, 1.38106793e-03, 1.64237581e-03, - 1.95312500e-03, 2.32267015e-03, 2.76213586e-03, 3.28475162e-03, - 3.90625000e-03, 4.64534029e-03, 5.52427173e-03, 6.56950324e-03, - 7.81250000e-03, 9.29068059e-03, 1.10485435e-02, 1.31390065e-02, - 1.56250000e-02, 1.85813612e-02, 2.20970869e-02, 2.62780130e-02, - 3.12500000e-02, 3.71627223e-02, 4.41941738e-02, 5.25560260e-02, - 6.25000000e-02, 7.43254447e-02, 8.83883476e-02, 1.05112052e-01, - 1.25000000e-01, 1.48650889e-01, 1.76776695e-01, 2.10224104e-01, - 2.50000000e-01, 2.97301779e-01, 3.53553391e-01, 4.20448208e-01, - 5.00000000e-01, 5.94603558e-01, 7.07106781e-01, 8.40896415e-01, - 1.00000000e+00, 1.18920712e+00, 1.41421356e+00, 1.68179283e+00, - 2.00000000e+00, 2.37841423e+00, 2.82842712e+00, 3.36358566e+00, - 4.00000000e+00, 4.75682846e+00, 5.65685425e+00, 6.72717132e+00, - 8.00000000e+00, 9.51365692e+00, 1.13137085e+01, 1.34543426e+01, - 1.60000000e+01, 1.90273138e+01, 2.26274170e+01, 2.69086853e+01, - 3.20000000e+01, 3.80546277e+01, 4.52548340e+01, 5.38173706e+01, - 6.40000000e+01, 7.61092554e+01, 9.05096680e+01, 1.07634741e+02, - 1.28000000e+02, 1.52218511e+02, 1.81019336e+02, 2.15269482e+02, - 2.56000000e+02, 3.04437021e+02, 3.62038672e+02, 4.30538965e+02, - 5.12000000e+02, 6.08874043e+02, 7.24077344e+02, 8.61077929e+02, - 1.02400000e+03, 1.21774809e+03, 1.44815469e+03, 1.72215586e+03, - 2.04800000e+03, 2.43549617e+03, 2.89630938e+03, 3.44431172e+03, - 4.09600000e+03, 4.87099234e+03, 5.79261875e+03, 6.88862343e+03, - 8.19200000e+03, 9.74198469e+03, 1.15852375e+04, 1.37772469e+04, - 1.63840000e+04, 1.94839694e+04, 2.31704750e+04, 2.75544937e+04, - 3.27680000e+04, 3.89679387e+04, 4.63409500e+04, 5.51089875e+04, - 6.55360000e+04, 7.79358775e+04, 9.26819000e+04, 1.10217975e+05, - 1.31072000e+05, 1.55871755e+05, 1.85363800e+05, 2.20435950e+05, - 2.62144000e+05, 3.11743510e+05, 3.70727600e+05, 4.40871900e+05, - 5.24288000e+05, 6.23487020e+05, 7.41455200e+05, 8.81743800e+05, - 1.04857600e+06, 1.24697404e+06, 1.48291040e+06, 1.76348760e+06, - 2.09715200e+06, 2.49394808e+06, 2.96582080e+06, 3.52697520e+06, - 4.19430400e+06, 4.98789616e+06, 5.93164160e+06, 7.05395040e+06, - 8.38860800e+06, 9.97579232e+06, 1.18632832e+07, 1.41079008e+07, - 1.67772160e+07, 1.99515846e+07, 2.37265664e+07, 2.82158016e+07, - 3.35544320e+07, 3.99031693e+07, 4.74531328e+07, 5.64316032e+07, - 6.71088640e+07, 7.98063385e+07, 9.49062656e+07, 1.12863206e+08, - 1.34217728e+08, 1.59612677e+08, 1.89812531e+08, 2.25726413e+08, - 2.68435456e+08, 3.19225354e+08, 3.79625062e+08, 4.51452825e+08, - 5.36870912e+08, 6.38450708e+08, 7.59250125e+08, 9.02905651e+08, - 1.07374182e+09, 1.27690142e+09, 1.51850025e+09, 1.80581130e+09, - 2.14748365e+09, 2.55380283e+09, 3.03700050e+09, 3.61162260e+09, - 4.29496730e+09, 5.10760567e+09, 6.07400100e+09, 7.22324521e+09, - 8.58993459e+09, 1.02152113e+10, 1.21480020e+10, 1.44464904e+10, - 1.71798692e+10, 2.04304227e+10, 2.42960040e+10, 2.88929808e+10, - 3.43597384e+10, 4.08608453e+10, 4.85920080e+10, 5.77859616e+10, - 6.87194767e+10, 8.17216907e+10, 9.71840160e+10, 1.15571923e+11, - 1.37438953e+11, 1.63443381e+11, 1.94368032e+11, 2.31143847e+11, - 2.74877907e+11, 3.26886763e+11, 3.88736064e+11, 4.62287693e+11, - 5.49755814e+11, 6.53773525e+11, 7.77472128e+11, 9.24575386e+11, - 1.09951163e+12, 1.30754705e+12, 1.55494426e+12, 1.84915077e+12, - 2.19902326e+12, 2.61509410e+12, 3.10988851e+12, 3.69830155e+12, - 4.39804651e+12, 5.23018820e+12, 6.21977702e+12, 7.39660309e+12, - 8.79609302e+12, 1.04603764e+13, 1.24395540e+13, 1.47932062e+13, - 1.75921860e+13, 2.09207528e+13, 2.48791081e+13, 2.95864124e+13, - 3.51843721e+13, 4.18415056e+13, 4.97582162e+13, 5.91728247e+13, - 7.03687442e+13, 8.36830112e+13, 9.95164324e+13, 1.18345649e+14, - 1.40737488e+14, 1.67366022e+14, 1.99032865e+14, 2.36691299e+14, - 2.81474977e+14, 3.34732045e+14, 3.98065730e+14, 4.73382598e+14, - 5.62949953e+14, 6.69464090e+14, 7.96131459e+14, 9.46765196e+14, - 1.12589991e+15, 1.33892818e+15, 1.59226292e+15, 1.89353039e+15, - 2.25179981e+15, 2.67785636e+15, 3.18452584e+15, 3.78706078e+15, - 4.50359963e+15, 5.35571272e+15, 6.36905167e+15, 7.57412156e+15, - 9.00719925e+15, 1.07114254e+16, 1.27381033e+16, 1.51482431e+16, - 1.80143985e+16, 2.14228509e+16, 2.54762067e+16, 3.02964863e+16, - 3.60287970e+16, 4.28457018e+16, 5.09524134e+16, 6.05929725e+16, - 7.20575940e+16, 8.56914035e+16, 1.01904827e+17, 1.21185945e+17, -}; - -#else - -float ff_aac_pow2sf_tab[428]; - -#endif /* CONFIG_HARDCODED_TABLES */ diff --git a/libavcodec/aactab.h b/libavcodec/aactab.h index 283d6c3a2d..c6213999b5 100644 --- a/libavcodec/aactab.h +++ b/libavcodec/aactab.h @@ -32,6 +32,7 @@ #include "libavutil/mem.h" #include "aac.h" +#include "aac_tablegen_decl.h" #include <stdint.h> @@ -73,10 +74,4 @@ extern const uint16_t * const ff_swb_offset_128 [13]; extern const uint8_t ff_tns_max_bands_1024[13]; extern const uint8_t ff_tns_max_bands_128 [13]; -#if CONFIG_HARDCODED_TABLES -extern const float ff_aac_pow2sf_tab[428]; -#else -extern float ff_aac_pow2sf_tab[428]; -#endif /* CONFIG_HARDCODED_TABLES */ - #endif /* AVCODEC_AACTAB_H */ diff --git a/libavcodec/sbr.h b/libavcodec/sbr.h index 2d9f4a519d..1c87a2e564 100644 --- a/libavcodec/sbr.h +++ b/libavcodec/sbr.h @@ -31,6 +31,7 @@ #include <stdint.h> #include "fft.h" +#include "aacps.h" /** * Spectral Band Replication header - spectrum parameters that invoke a reset if they differ from the previous header. @@ -133,6 +134,7 @@ typedef struct { ///The number of frequency bands in f_master unsigned n_master; SBRData data[2]; + PSContext ps; ///N_Low and N_High respectively, the number of frequency bands for low and high resolution unsigned n[2]; ///Number of noise floor bands @@ -157,7 +159,7 @@ typedef struct { ///QMF output of the HF generator float X_high[64][40][2]; ///QMF values of the reconstructed signal - DECLARE_ALIGNED(16, float, X)[2][2][32][64]; + DECLARE_ALIGNED(16, float, X)[2][2][38][64]; ///Zeroth coefficient used to filter the subband signals float alpha0[64][2]; ///First coefficient used to filter the subband signals @@ -176,7 +178,7 @@ typedef struct { float s_m[7][48]; float gain[7][48]; DECLARE_ALIGNED(16, float, qmf_filter_scratch)[5][64]; - RDFTContext rdft; + FFTContext mdct_ana; FFTContext mdct; } SpectralBandReplication; |