X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Faacdec.c;h=bd5ba0037390a373e37890df8d66ac59a233b922;hb=2eba6898c96fe3c920da0a0e30b829d05bf5cce1;hp=945ed395fea2bc114b22e44c17e13d472d714504;hpb=19ed4b8647982cbe45ab1f8e6e9ab4e052e3f74b;p=ffmpeg diff --git a/libavcodec/aacdec.c b/libavcodec/aacdec.c index 945ed395fea..bd5ba003739 100644 --- a/libavcodec/aacdec.c +++ b/libavcodec/aacdec.c @@ -3,20 +3,24 @@ * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org ) * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com ) * - * This file is part of FFmpeg. + * AAC LATM decoder + * Copyright (c) 2008-2010 Paul Kendall + * Copyright (c) 2010 Janne Grunau * - * FFmpeg is free software; you can redistribute it and/or + * This file is part of Libav. + * + * Libav 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, + * Libav 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 + * License along with Libav; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -38,7 +42,7 @@ * Y filterbank - standard * N (code in SoC repo) filterbank - Scalable Sample Rate * Y Temporal Noise Shaping - * N (code in SoC repo) Long Term Prediction + * Y Long Term Prediction * Y intensity stereo * Y channel coupling * Y frequency domain prediction @@ -81,7 +85,10 @@ #include "get_bits.h" #include "dsputil.h" #include "fft.h" +#include "fmtconvert.h" #include "lpc.h" +#include "kbdwin.h" +#include "sinewin.h" #include "aac.h" #include "aactab.h" @@ -90,7 +97,8 @@ #include "sbr.h" #include "aacsbr.h" #include "mpeg4audio.h" -#include "aac_parser.h" +#include "aacadtsdec.h" +#include "libavutil/intfloat.h" #include #include @@ -101,86 +109,27 @@ # include "arm/aac.h" #endif -union float754 { - float f; - uint32_t i; -}; - static VLC vlc_scalefactors; static VLC vlc_spectral[11]; static const char overread_err[] = "Input buffer exhausted before END element found\n"; -static ChannelElement *get_che(AACContext *ac, int type, int elem_id) +static int count_channels(uint8_t (*layout)[3], int tags) { - /* 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]; - } - if (ac->tags_mapped >= tags_per_config[ac->m4ac.chan_config]) { - return NULL; - } - switch (ac->m4ac.chan_config) { - case 7: - if (ac->tags_mapped == 3 && type == TYPE_CPE) { - ac->tags_mapped++; - return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][2]; - } - 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[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]; - } - case 5: - if (ac->tags_mapped == 2 && type == TYPE_CPE) { - ac->tags_mapped++; - return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][1]; - } - case 4: - if (ac->tags_mapped == 2 && ac->m4ac.chan_config == 4 && type == TYPE_SCE) { - ac->tags_mapped++; - return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][1]; - } - case 3: - case 2: - if (ac->tags_mapped == (ac->m4ac.chan_config != 2) && type == TYPE_CPE) { - ac->tags_mapped++; - return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][0]; - } else if (ac->m4ac.chan_config == 2) { - return NULL; - } - case 1: - if (!ac->tags_mapped && type == TYPE_SCE) { - ac->tags_mapped++; - return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][0]; - } - default: - return NULL; + int i, sum = 0; + for (i = 0; i < tags; i++) { + int syn_ele = layout[i][0]; + int pos = layout[i][2]; + sum += (1 + (syn_ele == TYPE_CPE)) * + (pos != AAC_CHANNEL_OFF && pos != AAC_CHANNEL_CC); } + return sum; } /** * Check for the channel element in the current channel position configuration. * If it exists, make sure the appropriate element is allocated and map the - * channel order to match the internal FFmpeg channel layout. + * channel order to match the internal Libav channel layout. * * @param che_pos current channel position configuration * @param type channel element type @@ -190,18 +139,19 @@ static ChannelElement *get_che(AACContext *ac, int type, int elem_id) * @return Returns error status. 0 - OK, !0 - error */ static av_cold int che_configure(AACContext *ac, - enum ChannelPosition che_pos[4][MAX_ELEM_ID], - int type, int id, - int *channels) + enum ChannelPosition che_pos, + int type, int id, int *channels) { - if (che_pos[type][id]) { - if (!ac->che[type][id] && !(ac->che[type][id] = av_mallocz(sizeof(ChannelElement)))) - return AVERROR(ENOMEM); - ff_aac_sbr_ctx_init(&ac->che[type][id]->sbr); + if (che_pos) { + if (!ac->che[type][id]) { + if (!(ac->che[type][id] = av_mallocz(sizeof(ChannelElement)))) + return AVERROR(ENOMEM); + ff_aac_sbr_ctx_init(ac, &ac->che[type][id]->sbr); + } if (type != TYPE_CCE) { ac->output_data[(*channels)++] = ac->che[type][id]->ch[0].ret; if (type == TYPE_CPE || - (type == TYPE_SCE && ac->m4ac.ps == 1)) { + (type == TYPE_SCE && ac->oc[1].m4ac.ps == 1)) { ac->output_data[(*channels)++] = ac->che[type][id]->ch[1].ret; } } @@ -213,104 +163,413 @@ static av_cold int che_configure(AACContext *ac, return 0; } +struct elem_to_channel { + uint64_t av_position; + uint8_t syn_ele; + uint8_t elem_id; + uint8_t aac_position; +}; + +static int assign_pair(struct elem_to_channel e2c_vec[MAX_ELEM_ID], + uint8_t (*layout_map)[3], int offset, int tags, uint64_t left, + uint64_t right, int pos) +{ + if (layout_map[offset][0] == TYPE_CPE) { + e2c_vec[offset] = (struct elem_to_channel) { + .av_position = left | right, .syn_ele = TYPE_CPE, + .elem_id = layout_map[offset ][1], .aac_position = pos }; + return 1; + } else { + e2c_vec[offset] = (struct elem_to_channel) { + .av_position = left, .syn_ele = TYPE_SCE, + .elem_id = layout_map[offset ][1], .aac_position = pos }; + e2c_vec[offset + 1] = (struct elem_to_channel) { + .av_position = right, .syn_ele = TYPE_SCE, + .elem_id = layout_map[offset + 1][1], .aac_position = pos }; + return 2; + } +} + +static int count_paired_channels(uint8_t (*layout_map)[3], int tags, int pos, int *current) { + int num_pos_channels = 0; + int first_cpe = 0; + int sce_parity = 0; + int i; + for (i = *current; i < tags; i++) { + if (layout_map[i][2] != pos) + break; + if (layout_map[i][0] == TYPE_CPE) { + if (sce_parity) { + if (pos == AAC_CHANNEL_FRONT && !first_cpe) { + sce_parity = 0; + } else { + return -1; + } + } + num_pos_channels += 2; + first_cpe = 1; + } else { + num_pos_channels++; + sce_parity ^= 1; + } + } + if (sce_parity && + ((pos == AAC_CHANNEL_FRONT && first_cpe) || pos == AAC_CHANNEL_SIDE)) + return -1; + *current = i; + return num_pos_channels; +} + +static uint64_t sniff_channel_order(uint8_t (*layout_map)[3], int tags) +{ + int i, n, total_non_cc_elements; + struct elem_to_channel e2c_vec[4*MAX_ELEM_ID] = {{ 0 }}; + int num_front_channels, num_side_channels, num_back_channels; + uint64_t layout; + + if (FF_ARRAY_ELEMS(e2c_vec) < tags) + return 0; + + i = 0; + num_front_channels = + count_paired_channels(layout_map, tags, AAC_CHANNEL_FRONT, &i); + if (num_front_channels < 0) + return 0; + num_side_channels = + count_paired_channels(layout_map, tags, AAC_CHANNEL_SIDE, &i); + if (num_side_channels < 0) + return 0; + num_back_channels = + count_paired_channels(layout_map, tags, AAC_CHANNEL_BACK, &i); + if (num_back_channels < 0) + return 0; + + i = 0; + if (num_front_channels & 1) { + e2c_vec[i] = (struct elem_to_channel) { + .av_position = AV_CH_FRONT_CENTER, .syn_ele = TYPE_SCE, + .elem_id = layout_map[i][1], .aac_position = AAC_CHANNEL_FRONT }; + i++; + num_front_channels--; + } + if (num_front_channels >= 4) { + i += assign_pair(e2c_vec, layout_map, i, tags, + AV_CH_FRONT_LEFT_OF_CENTER, + AV_CH_FRONT_RIGHT_OF_CENTER, + AAC_CHANNEL_FRONT); + num_front_channels -= 2; + } + if (num_front_channels >= 2) { + i += assign_pair(e2c_vec, layout_map, i, tags, + AV_CH_FRONT_LEFT, + AV_CH_FRONT_RIGHT, + AAC_CHANNEL_FRONT); + num_front_channels -= 2; + } + while (num_front_channels >= 2) { + i += assign_pair(e2c_vec, layout_map, i, tags, + UINT64_MAX, + UINT64_MAX, + AAC_CHANNEL_FRONT); + num_front_channels -= 2; + } + + if (num_side_channels >= 2) { + i += assign_pair(e2c_vec, layout_map, i, tags, + AV_CH_SIDE_LEFT, + AV_CH_SIDE_RIGHT, + AAC_CHANNEL_FRONT); + num_side_channels -= 2; + } + while (num_side_channels >= 2) { + i += assign_pair(e2c_vec, layout_map, i, tags, + UINT64_MAX, + UINT64_MAX, + AAC_CHANNEL_SIDE); + num_side_channels -= 2; + } + + while (num_back_channels >= 4) { + i += assign_pair(e2c_vec, layout_map, i, tags, + UINT64_MAX, + UINT64_MAX, + AAC_CHANNEL_BACK); + num_back_channels -= 2; + } + if (num_back_channels >= 2) { + i += assign_pair(e2c_vec, layout_map, i, tags, + AV_CH_BACK_LEFT, + AV_CH_BACK_RIGHT, + AAC_CHANNEL_BACK); + num_back_channels -= 2; + } + if (num_back_channels) { + e2c_vec[i] = (struct elem_to_channel) { + .av_position = AV_CH_BACK_CENTER, .syn_ele = TYPE_SCE, + .elem_id = layout_map[i][1], .aac_position = AAC_CHANNEL_BACK }; + i++; + num_back_channels--; + } + + if (i < tags && layout_map[i][2] == AAC_CHANNEL_LFE) { + e2c_vec[i] = (struct elem_to_channel) { + .av_position = AV_CH_LOW_FREQUENCY, .syn_ele = TYPE_LFE, + .elem_id = layout_map[i][1], .aac_position = AAC_CHANNEL_LFE }; + i++; + } + while (i < tags && layout_map[i][2] == AAC_CHANNEL_LFE) { + e2c_vec[i] = (struct elem_to_channel) { + .av_position = UINT64_MAX, .syn_ele = TYPE_LFE, + .elem_id = layout_map[i][1], .aac_position = AAC_CHANNEL_LFE }; + i++; + } + + // Must choose a stable sort + total_non_cc_elements = n = i; + do { + int next_n = 0; + for (i = 1; i < n; i++) { + if (e2c_vec[i-1].av_position > e2c_vec[i].av_position) { + FFSWAP(struct elem_to_channel, e2c_vec[i-1], e2c_vec[i]); + next_n = i; + } + } + n = next_n; + } while (n > 0); + + layout = 0; + for (i = 0; i < total_non_cc_elements; i++) { + layout_map[i][0] = e2c_vec[i].syn_ele; + layout_map[i][1] = e2c_vec[i].elem_id; + layout_map[i][2] = e2c_vec[i].aac_position; + if (e2c_vec[i].av_position != UINT64_MAX) { + layout |= e2c_vec[i].av_position; + } + } + + return layout; +} + +/** + * Save current output configuration if and only if it has been locked. + */ +static void push_output_configuration(AACContext *ac) { + if (ac->oc[1].status == OC_LOCKED) { + ac->oc[0] = ac->oc[1]; + } + ac->oc[1].status = OC_NONE; +} + +/** + * Restore the previous output configuration if and only if the current + * configuration is unlocked. + */ +static void pop_output_configuration(AACContext *ac) { + if (ac->oc[1].status != OC_LOCKED) { + ac->oc[1] = ac->oc[0]; + ac->avctx->channels = ac->oc[1].channels; + ac->avctx->channel_layout = ac->oc[1].channels; + } +} + /** * Configure output channel order based on the current program configuration element. * - * @param che_pos current channel position configuration - * @param new_che_pos New channel position configuration - we only do something if it differs from the current one. - * * @return Returns error status. 0 - OK, !0 - error */ -static av_cold int output_configure(AACContext *ac, - enum ChannelPosition che_pos[4][MAX_ELEM_ID], - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID], - int channel_config, enum OCStatus oc_type) +static int output_configure(AACContext *ac, + uint8_t layout_map[MAX_ELEM_ID*4][3], int tags, + int channel_config, enum OCStatus oc_type) { AVCodecContext *avctx = ac->avctx; - int i, type, channels = 0, ret; + int i, channels = 0, ret; + uint64_t layout = 0; - if (new_che_pos != che_pos) - memcpy(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0])); + if (ac->oc[1].layout_map != layout_map) { + memcpy(ac->oc[1].layout_map, layout_map, tags * sizeof(layout_map[0])); + ac->oc[1].layout_map_tags = tags; + } - if (channel_config) { - for (i = 0; i < tags_per_config[channel_config]; i++) { - if ((ret = che_configure(ac, che_pos, - aac_channel_layout_map[channel_config - 1][i][0], - aac_channel_layout_map[channel_config - 1][i][1], - &channels))) - return ret; + // Try to sniff a reasonable channel order, otherwise output the + // channels in the order the PCE declared them. + if (avctx->request_channel_layout != AV_CH_LAYOUT_NATIVE) + layout = sniff_channel_order(layout_map, tags); + for (i = 0; i < tags; i++) { + int type = layout_map[i][0]; + int id = layout_map[i][1]; + int position = layout_map[i][2]; + // Allocate or free elements depending on if they are in the + // current program configuration. + ret = che_configure(ac, position, type, id, &channels); + if (ret < 0) + return ret; + } + if (ac->oc[1].m4ac.ps == 1 && channels == 2) { + if (layout == AV_CH_FRONT_CENTER) { + layout = AV_CH_FRONT_LEFT|AV_CH_FRONT_RIGHT; + } else { + layout = 0; } + } - memset(ac->tag_che_map, 0, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0])); - ac->tags_mapped = 0; - - avctx->channel_layout = aac_channel_layout[channel_config - 1]; - } else { - /* Allocate or free elements depending on if they are in the - * current program configuration. - * - * Set up default 1:1 output mapping. - * - * For a 5.1 stream the output order will be: - * [ Center ] [ Front Left ] [ Front Right ] [ LFE ] [ Surround Left ] [ Surround Right ] - */ - - for (i = 0; i < MAX_ELEM_ID; i++) { - for (type = 0; type < 4; type++) { - if ((ret = che_configure(ac, che_pos, type, i, &channels))) - return ret; - } - } + memcpy(ac->tag_che_map, ac->che, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0])); + avctx->channel_layout = ac->oc[1].channel_layout = layout; + avctx->channels = ac->oc[1].channels = channels; + ac->oc[1].status = oc_type; - memcpy(ac->tag_che_map, ac->che, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0])); - ac->tags_mapped = 4 * MAX_ELEM_ID; + return 0; +} - avctx->channel_layout = 0; +/** + * Set up channel positions based on a default channel configuration + * as specified in table 1.17. + * + * @return Returns error status. 0 - OK, !0 - error + */ +static int set_default_channel_config(AVCodecContext *avctx, + uint8_t (*layout_map)[3], + int *tags, + int channel_config) +{ + if (channel_config < 1 || channel_config > 7) { + av_log(avctx, AV_LOG_ERROR, "invalid default channel configuration (%d)\n", + channel_config); + return -1; } + *tags = tags_per_config[channel_config]; + memcpy(layout_map, aac_channel_layout_map[channel_config-1], *tags * sizeof(*layout_map)); + return 0; +} - avctx->channels = channels; +static ChannelElement *get_che(AACContext *ac, int type, int elem_id) +{ + // For PCE based channel configurations map the channels solely based on tags. + if (!ac->oc[1].m4ac.chan_config) { + return ac->tag_che_map[type][elem_id]; + } + // Allow single CPE stereo files to be signalled with mono configuration. + if (!ac->tags_mapped && type == TYPE_CPE && ac->oc[1].m4ac.chan_config == 1) { + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int layout_map_tags; + push_output_configuration(ac); + + if (set_default_channel_config(ac->avctx, layout_map, &layout_map_tags, + 2) < 0) + return NULL; + if (output_configure(ac, layout_map, layout_map_tags, + 2, OC_TRIAL_FRAME) < 0) + return NULL; - ac->output_configured = oc_type; + ac->oc[1].m4ac.chan_config = 2; + } + // And vice-versa + if (!ac->tags_mapped && type == TYPE_SCE && ac->oc[1].m4ac.chan_config == 2) { + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int layout_map_tags; + push_output_configuration(ac); + + if (set_default_channel_config(ac->avctx, layout_map, &layout_map_tags, + 1) < 0) + return NULL; + if (output_configure(ac, layout_map, layout_map_tags, + 1, OC_TRIAL_FRAME) < 0) + return NULL; - return 0; + ac->oc[1].m4ac.chan_config = 1; + } + // For indexed channel configurations map the channels solely based on position. + switch (ac->oc[1].m4ac.chan_config) { + case 7: + if (ac->tags_mapped == 3 && type == TYPE_CPE) { + ac->tags_mapped++; + return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][2]; + } + 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[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->oc[1].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]; + } + case 5: + if (ac->tags_mapped == 2 && type == TYPE_CPE) { + ac->tags_mapped++; + return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][1]; + } + case 4: + if (ac->tags_mapped == 2 && ac->oc[1].m4ac.chan_config == 4 && type == TYPE_SCE) { + ac->tags_mapped++; + return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][1]; + } + case 3: + case 2: + if (ac->tags_mapped == (ac->oc[1].m4ac.chan_config != 2) && type == TYPE_CPE) { + ac->tags_mapped++; + return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][0]; + } else if (ac->oc[1].m4ac.chan_config == 2) { + return NULL; + } + case 1: + if (!ac->tags_mapped && type == TYPE_SCE) { + ac->tags_mapped++; + return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][0]; + } + default: + return NULL; + } } /** * Decode an array of 4 bit element IDs, optionally interleaved with a stereo/mono switching bit. * - * @param cpe_map Stereo (Channel Pair Element) map, NULL if stereo bit is not present. - * @param sce_map mono (Single Channel Element) map * @param type speaker type/position for these channels */ -static void decode_channel_map(enum ChannelPosition *cpe_map, - enum ChannelPosition *sce_map, +static void decode_channel_map(uint8_t layout_map[][3], enum ChannelPosition type, GetBitContext *gb, int n) { while (n--) { - enum ChannelPosition *map = cpe_map && get_bits1(gb) ? cpe_map : sce_map; // stereo or mono map - map[get_bits(gb, 4)] = type; + enum RawDataBlockType syn_ele; + switch (type) { + case AAC_CHANNEL_FRONT: + case AAC_CHANNEL_BACK: + case AAC_CHANNEL_SIDE: + syn_ele = get_bits1(gb); + break; + case AAC_CHANNEL_CC: + skip_bits1(gb); + syn_ele = TYPE_CCE; + break; + case AAC_CHANNEL_LFE: + syn_ele = TYPE_LFE; + break; + } + layout_map[0][0] = syn_ele; + layout_map[0][1] = get_bits(gb, 4); + layout_map[0][2] = type; + layout_map++; } } /** * Decode program configuration element; reference: table 4.2. * - * @param new_che_pos New channel position configuration - we only do something if it differs from the current one. - * * @return Returns error status. 0 - OK, !0 - error */ -static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID], +static int decode_pce(AVCodecContext *avctx, MPEG4AudioConfig *m4ac, + uint8_t (*layout_map)[3], GetBitContext *gb) { int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc, sampling_index; int comment_len; + int tags; skip_bits(gb, 2); // object_type sampling_index = get_bits(gb, 4); - if (ac->m4ac.sampling_index != sampling_index) - 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"); + if (m4ac->sampling_index != sampling_index) + av_log(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); @@ -327,86 +586,51 @@ static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_EL if (get_bits1(gb)) skip_bits(gb, 3); // mixdown_coeff_index and pseudo_surround - decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_FRONT, gb, num_front); - decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_SIDE, gb, num_side ); - decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_BACK, gb, num_back ); - decode_channel_map(NULL, new_che_pos[TYPE_LFE], AAC_CHANNEL_LFE, gb, num_lfe ); + decode_channel_map(layout_map , AAC_CHANNEL_FRONT, gb, num_front); + tags = num_front; + decode_channel_map(layout_map + tags, AAC_CHANNEL_SIDE, gb, num_side); + tags += num_side; + decode_channel_map(layout_map + tags, AAC_CHANNEL_BACK, gb, num_back); + tags += num_back; + decode_channel_map(layout_map + tags, AAC_CHANNEL_LFE, gb, num_lfe); + tags += num_lfe; skip_bits_long(gb, 4 * num_assoc_data); - decode_channel_map(new_che_pos[TYPE_CCE], new_che_pos[TYPE_CCE], AAC_CHANNEL_CC, gb, num_cc ); + decode_channel_map(layout_map + tags, AAC_CHANNEL_CC, gb, num_cc); + tags += num_cc; align_get_bits(gb); /* comment field, first byte is length */ comment_len = get_bits(gb, 8) * 8; if (get_bits_left(gb) < comment_len) { - av_log(ac->avctx, AV_LOG_ERROR, overread_err); + av_log(avctx, AV_LOG_ERROR, overread_err); return -1; } skip_bits_long(gb, comment_len); - return 0; -} - -/** - * Set up channel positions based on a default channel configuration - * as specified in table 1.17. - * - * @param new_che_pos New channel position configuration - we only do something if it differs from the current one. - * - * @return Returns error status. 0 - OK, !0 - error - */ -static av_cold int set_default_channel_config(AACContext *ac, - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID], - int channel_config) -{ - if (channel_config < 1 || channel_config > 7) { - av_log(ac->avctx, AV_LOG_ERROR, "invalid default channel configuration (%d)\n", - channel_config); - return -1; - } - - /* default channel configurations: - * - * 1ch : front center (mono) - * 2ch : L + R (stereo) - * 3ch : front center + L + R - * 4ch : front center + L + R + back center - * 5ch : front center + L + R + back stereo - * 6ch : front center + L + R + back stereo + LFE - * 7ch : front center + L + R + outer front left + outer front right + back stereo + LFE - */ - - if (channel_config != 2) - new_che_pos[TYPE_SCE][0] = AAC_CHANNEL_FRONT; // front center (or mono) - if (channel_config > 1) - new_che_pos[TYPE_CPE][0] = AAC_CHANNEL_FRONT; // L + R (or stereo) - if (channel_config == 4) - new_che_pos[TYPE_SCE][1] = AAC_CHANNEL_BACK; // back center - if (channel_config > 4) - new_che_pos[TYPE_CPE][(channel_config == 7) + 1] - = AAC_CHANNEL_BACK; // back stereo - if (channel_config > 5) - new_che_pos[TYPE_LFE][0] = AAC_CHANNEL_LFE; // LFE - if (channel_config == 7) - new_che_pos[TYPE_CPE][1] = AAC_CHANNEL_FRONT; // outer front left + outer front right - - return 0; + return tags; } /** * Decode GA "General Audio" specific configuration; reference: table 4.1. * + * @param ac pointer to AACContext, may be null + * @param avctx pointer to AVCCodecContext, used for logging + * * @return Returns error status. 0 - OK, !0 - error */ -static int decode_ga_specific_config(AACContext *ac, GetBitContext *gb, +static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx, + GetBitContext *gb, + MPEG4AudioConfig *m4ac, int channel_config) { - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID]; int extension_flag, ret; + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int tags = 0; if (get_bits1(gb)) { // frameLengthFlag - av_log_missing_feature(ac->avctx, "960/120 MDCT window is", 1); + av_log_missing_feature(avctx, "960/120 MDCT window is", 1); return -1; } @@ -414,24 +638,31 @@ static int decode_ga_specific_config(AACContext *ac, GetBitContext *gb, skip_bits(gb, 14); // coreCoderDelay extension_flag = get_bits1(gb); - if (ac->m4ac.object_type == AOT_AAC_SCALABLE || - ac->m4ac.object_type == AOT_ER_AAC_SCALABLE) + if (m4ac->object_type == AOT_AAC_SCALABLE || + m4ac->object_type == AOT_ER_AAC_SCALABLE) skip_bits(gb, 3); // layerNr - memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0])); if (channel_config == 0) { skip_bits(gb, 4); // element_instance_tag - if ((ret = decode_pce(ac, new_che_pos, gb))) - return ret; + tags = decode_pce(avctx, m4ac, layout_map, gb); + if (tags < 0) + return tags; } else { - if ((ret = set_default_channel_config(ac, new_che_pos, channel_config))) + if ((ret = set_default_channel_config(avctx, layout_map, &tags, channel_config))) return ret; } - if ((ret = output_configure(ac, ac->che_pos, new_che_pos, channel_config, OC_GLOBAL_HDR))) + + if (count_channels(layout_map, tags) > 1) { + m4ac->ps = 0; + } else if (m4ac->sbr == 1 && m4ac->ps == -1) + m4ac->ps = 1; + + if (ac && (ret = output_configure(ac, layout_map, tags, + channel_config, OC_GLOBAL_HDR))) return ret; if (extension_flag) { - switch (ac->m4ac.object_type) { + switch (m4ac->object_type) { case AOT_ER_BSAC: skip_bits(gb, 5); // numOfSubFrame skip_bits(gb, 11); // layer_length @@ -454,42 +685,58 @@ static int decode_ga_specific_config(AACContext *ac, GetBitContext *gb, /** * Decode audio specific configuration; reference: table 1.13. * - * @param data pointer to AVCodecContext extradata - * @param data_size size of AVCCodecContext extradata + * @param ac pointer to AACContext, may be null + * @param avctx pointer to AVCCodecContext, used for logging + * @param m4ac pointer to MPEG4AudioConfig, used for parsing + * @param data pointer to buffer holding an audio specific config + * @param bit_size size of audio specific config or data in bits + * @param sync_extension look for an appended sync extension * - * @return Returns error status. 0 - OK, !0 - error + * @return Returns error status or number of consumed bits. <0 - error */ -static int decode_audio_specific_config(AACContext *ac, void *data, - int data_size) +static int decode_audio_specific_config(AACContext *ac, + AVCodecContext *avctx, + MPEG4AudioConfig *m4ac, + const uint8_t *data, int bit_size, + int sync_extension) { GetBitContext gb; int i; - init_get_bits(&gb, data, data_size * 8); + av_dlog(avctx, "extradata size %d\n", avctx->extradata_size); + for (i = 0; i < avctx->extradata_size; i++) + av_dlog(avctx, "%02x ", avctx->extradata[i]); + av_dlog(avctx, "\n"); - if ((i = ff_mpeg4audio_get_config(&ac->m4ac, data, data_size)) < 0) + init_get_bits(&gb, data, bit_size); + + if ((i = avpriv_mpeg4audio_get_config(m4ac, data, bit_size, sync_extension)) < 0) return -1; - if (ac->m4ac.sampling_index > 12) { - av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index); + if (m4ac->sampling_index > 12) { + av_log(avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", m4ac->sampling_index); return -1; } - if (ac->m4ac.sbr == 1 && ac->m4ac.ps == -1) - ac->m4ac.ps = 1; skip_bits_long(&gb, i); - switch (ac->m4ac.object_type) { + switch (m4ac->object_type) { case AOT_AAC_MAIN: case AOT_AAC_LC: - if (decode_ga_specific_config(ac, &gb, ac->m4ac.chan_config)) + case AOT_AAC_LTP: + if (decode_ga_specific_config(ac, avctx, &gb, m4ac, m4ac->chan_config)) return -1; break; default: - 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); + av_log(avctx, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n", + m4ac->sbr == 1? "SBR+" : "", m4ac->object_type); return -1; } - return 0; + + av_dlog(avctx, "AOT %d chan config %d sampling index %d (%d) SBR %d PS %d\n", + m4ac->object_type, m4ac->chan_config, m4ac->sampling_index, + m4ac->sample_rate, m4ac->sbr, m4ac->ps); + + return get_bits_count(&gb); } /** @@ -521,6 +768,22 @@ static void reset_all_predictors(PredictorState *ps) reset_predict_state(&ps[i]); } +static int sample_rate_idx (int rate) +{ + if (92017 <= rate) return 0; + else if (75132 <= rate) return 1; + else if (55426 <= rate) return 2; + else if (46009 <= rate) return 3; + else if (37566 <= rate) return 4; + else if (27713 <= rate) return 5; + else if (23004 <= rate) return 6; + else if (18783 <= rate) return 7; + else if (13856 <= rate) return 8; + else if (11502 <= rate) return 9; + else if (9391 <= rate) return 10; + else return 11; +} + static void reset_predictor_group(PredictorState *ps, int group_num) { int i; @@ -537,17 +800,53 @@ static void reset_predictor_group(PredictorState *ps, int group_num) static av_cold int aac_decode_init(AVCodecContext *avctx) { AACContext *ac = avctx->priv_data; - int i; + float output_scale_factor; ac->avctx = avctx; - ac->m4ac.sample_rate = avctx->sample_rate; + ac->oc[1].m4ac.sample_rate = avctx->sample_rate; if (avctx->extradata_size > 0) { - if (decode_audio_specific_config(ac, avctx->extradata, avctx->extradata_size)) + if (decode_audio_specific_config(ac, ac->avctx, &ac->oc[1].m4ac, + avctx->extradata, + avctx->extradata_size*8, 1) < 0) return -1; + } else { + int sr, i; + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int layout_map_tags; + + sr = sample_rate_idx(avctx->sample_rate); + ac->oc[1].m4ac.sampling_index = sr; + ac->oc[1].m4ac.channels = avctx->channels; + ac->oc[1].m4ac.sbr = -1; + ac->oc[1].m4ac.ps = -1; + + for (i = 0; i < FF_ARRAY_ELEMS(ff_mpeg4audio_channels); i++) + if (ff_mpeg4audio_channels[i] == avctx->channels) + break; + if (i == FF_ARRAY_ELEMS(ff_mpeg4audio_channels)) { + i = 0; + } + ac->oc[1].m4ac.chan_config = i; + + if (ac->oc[1].m4ac.chan_config) { + int ret = set_default_channel_config(avctx, layout_map, + &layout_map_tags, ac->oc[1].m4ac.chan_config); + if (!ret) + output_configure(ac, layout_map, layout_map_tags, + ac->oc[1].m4ac.chan_config, OC_GLOBAL_HDR); + else if (avctx->err_recognition & AV_EF_EXPLODE) + return AVERROR_INVALIDDATA; + } } - avctx->sample_fmt = SAMPLE_FMT_S16; + if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) { + avctx->sample_fmt = AV_SAMPLE_FMT_FLT; + output_scale_factor = 1.0 / 32768.0; + } else { + avctx->sample_fmt = AV_SAMPLE_FMT_S16; + output_scale_factor = 1.0; + } AAC_INIT_VLC_STATIC( 0, 304); AAC_INIT_VLC_STATIC( 1, 270); @@ -563,36 +862,21 @@ static av_cold int aac_decode_init(AVCodecContext *avctx) ff_aac_sbr_init(); - dsputil_init(&ac->dsp, avctx); + ff_dsputil_init(&ac->dsp, avctx); + ff_fmt_convert_init(&ac->fmt_conv, avctx); ac->random_state = 0x1f2e3d4c; - // -1024 - Compensate wrong IMDCT method. - // 32768 - Required to scale values to the correct range for the bias method - // for float to int16 conversion. - - if (ac->dsp.float_to_int16_interleave == ff_float_to_int16_interleave_c) { - ac->add_bias = 385.0f; - ac->sf_scale = 1. / (-1024. * 32768.); - ac->sf_offset = 0; - } else { - ac->add_bias = 0.0f; - ac->sf_scale = 1. / -1024.; - 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]), ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]), 352); - ff_mdct_init(&ac->mdct, 11, 1, 1.0); - ff_mdct_init(&ac->mdct_small, 8, 1, 1.0); + ff_mdct_init(&ac->mdct, 11, 1, output_scale_factor/1024.0); + ff_mdct_init(&ac->mdct_small, 8, 1, output_scale_factor/128.0); + ff_mdct_init(&ac->mdct_ltp, 11, 0, -2.0/output_scale_factor); // window initialization ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024); ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128); @@ -601,6 +885,9 @@ static av_cold int aac_decode_init(AVCodecContext *avctx) cbrt_tableinit(); + avcodec_get_frame_defaults(&ac->frame); + avctx->coded_frame = &ac->frame; + return 0; } @@ -635,24 +922,35 @@ static int decode_prediction(AACContext *ac, IndividualChannelStream *ics, return -1; } } - for (sfb = 0; sfb < FFMIN(ics->max_sfb, ff_aac_pred_sfb_max[ac->m4ac.sampling_index]); sfb++) { + for (sfb = 0; sfb < FFMIN(ics->max_sfb, ff_aac_pred_sfb_max[ac->oc[1].m4ac.sampling_index]); sfb++) { ics->prediction_used[sfb] = get_bits1(gb); } return 0; } +/** + * Decode Long Term Prediction data; reference: table 4.xx. + */ +static void decode_ltp(AACContext *ac, LongTermPrediction *ltp, + GetBitContext *gb, uint8_t max_sfb) +{ + int sfb; + + ltp->lag = get_bits(gb, 11); + ltp->coef = ltp_coef[get_bits(gb, 3)]; + for (sfb = 0; sfb < FFMIN(max_sfb, MAX_LTP_LONG_SFB); sfb++) + ltp->used[sfb] = get_bits1(gb); +} + /** * Decode Individual Channel Stream info; reference: table 4.6. - * - * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information. */ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, - GetBitContext *gb, int common_window) + GetBitContext *gb) { if (get_bits1(gb)) { av_log(ac->avctx, AV_LOG_ERROR, "Reserved bit set.\n"); - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + return AVERROR_INVALIDDATA; } ics->window_sequence[1] = ics->window_sequence[0]; ics->window_sequence[0] = get_bits(gb, 2); @@ -672,32 +970,29 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, } } ics->num_windows = 8; - ics->swb_offset = ff_swb_offset_128[ac->m4ac.sampling_index]; - ics->num_swb = ff_aac_num_swb_128[ac->m4ac.sampling_index]; - ics->tns_max_bands = ff_tns_max_bands_128[ac->m4ac.sampling_index]; + ics->swb_offset = ff_swb_offset_128[ac->oc[1].m4ac.sampling_index]; + ics->num_swb = ff_aac_num_swb_128[ac->oc[1].m4ac.sampling_index]; + ics->tns_max_bands = ff_tns_max_bands_128[ac->oc[1].m4ac.sampling_index]; ics->predictor_present = 0; } else { ics->max_sfb = get_bits(gb, 6); ics->num_windows = 1; - ics->swb_offset = ff_swb_offset_1024[ac->m4ac.sampling_index]; - ics->num_swb = ff_aac_num_swb_1024[ac->m4ac.sampling_index]; - ics->tns_max_bands = ff_tns_max_bands_1024[ac->m4ac.sampling_index]; + ics->swb_offset = ff_swb_offset_1024[ac->oc[1].m4ac.sampling_index]; + ics->num_swb = ff_aac_num_swb_1024[ac->oc[1].m4ac.sampling_index]; + ics->tns_max_bands = ff_tns_max_bands_1024[ac->oc[1].m4ac.sampling_index]; ics->predictor_present = get_bits1(gb); ics->predictor_reset_group = 0; if (ics->predictor_present) { - if (ac->m4ac.object_type == AOT_AAC_MAIN) { + if (ac->oc[1].m4ac.object_type == AOT_AAC_MAIN) { if (decode_prediction(ac, ics, gb)) { - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + return AVERROR_INVALIDDATA; } - } else if (ac->m4ac.object_type == AOT_AAC_LC) { + } else if (ac->oc[1].m4ac.object_type == AOT_AAC_LC) { av_log(ac->avctx, AV_LOG_ERROR, "Prediction is not allowed in AAC-LC.\n"); - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + return AVERROR_INVALIDDATA; } else { - av_log_missing_feature(ac->avctx, "Predictor bit set but LTP is", 1); - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + if ((ics->ltp.present = get_bits(gb, 1))) + decode_ltp(ac, &ics->ltp, gb, ics->max_sfb); } } } @@ -706,8 +1001,7 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, 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)); - return -1; + return AVERROR_INVALIDDATA; } return 0; @@ -737,19 +1031,20 @@ static int decode_band_types(AACContext *ac, enum BandType band_type[120], av_log(ac->avctx, AV_LOG_ERROR, "invalid band type\n"); return -1; } - while ((sect_len_incr = get_bits(gb, bits)) == (1 << bits) - 1) + do { + sect_len_incr = get_bits(gb, bits); sect_end += sect_len_incr; - sect_end += sect_len_incr; - if (get_bits_left(gb) < 0) { - av_log(ac->avctx, AV_LOG_ERROR, overread_err); - return -1; - } - if (sect_end > ics->max_sfb) { - av_log(ac->avctx, AV_LOG_ERROR, - "Number of bands (%d) exceeds limit (%d).\n", - sect_end, ics->max_sfb); - return -1; - } + if (get_bits_left(gb) < 0) { + av_log(ac->avctx, AV_LOG_ERROR, overread_err); + return -1; + } + if (sect_end > ics->max_sfb) { + av_log(ac->avctx, AV_LOG_ERROR, + "Number of bands (%d) exceeds limit (%d).\n", + sect_end, ics->max_sfb); + return -1; + } + } while (sect_len_incr == (1 << bits) - 1); for (; k < sect_end; k++) { band_type [idx] = sect_band_type; band_type_run_end[idx++] = sect_end; @@ -775,11 +1070,10 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, enum BandType band_type[120], int band_type_run_end[120]) { - const int sf_offset = ac->sf_offset + (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE ? 12 : 0); int g, i, idx = 0; - int offset[3] = { global_gain, global_gain - 90, 100 }; + int offset[3] = { global_gain, global_gain - 90, 0 }; + int clipped_offset; int noise_flag = 1; - static const char *sf_str[3] = { "Global gain", "Noise gain", "Intensity stereo position" }; for (g = 0; g < ics->num_window_groups; g++) { for (i = 0; i < ics->max_sfb;) { int run_end = band_type_run_end[idx]; @@ -789,12 +1083,14 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, } else if ((band_type[idx] == INTENSITY_BT) || (band_type[idx] == INTENSITY_BT2)) { for (; i < run_end; i++, idx++) { offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60; - if (offset[2] > 255U) { - av_log(ac->avctx, AV_LOG_ERROR, - "%s (%d) out of range.\n", sf_str[2], offset[2]); - return -1; + clipped_offset = av_clip(offset[2], -155, 100); + if (offset[2] != clipped_offset) { + av_log_ask_for_sample(ac->avctx, "Intensity stereo " + "position clipped (%d -> %d).\nIf you heard an " + "audible artifact, there may be a bug in the " + "decoder. ", offset[2], clipped_offset); } - sf[idx] = ff_aac_pow2sf_tab[-offset[2] + 300]; + sf[idx] = ff_aac_pow2sf_tab[-clipped_offset + POW_SF2_ZERO]; } } else if (band_type[idx] == NOISE_BT) { for (; i < run_end; i++, idx++) { @@ -802,22 +1098,24 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, offset[1] += get_bits(gb, 9) - 256; else offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60; - if (offset[1] > 255U) { - av_log(ac->avctx, AV_LOG_ERROR, - "%s (%d) out of range.\n", sf_str[1], offset[1]); - return -1; + clipped_offset = av_clip(offset[1], -100, 155); + if (offset[1] != clipped_offset) { + av_log_ask_for_sample(ac->avctx, "Noise gain clipped " + "(%d -> %d).\nIf you heard an audible " + "artifact, there may be a bug in the decoder. ", + offset[1], clipped_offset); } - sf[idx] = -ff_aac_pow2sf_tab[offset[1] + sf_offset + 100]; + sf[idx] = -ff_aac_pow2sf_tab[clipped_offset + POW_SF2_ZERO]; } } else { for (; i < run_end; i++, idx++) { offset[0] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60; if (offset[0] > 255U) { av_log(ac->avctx, AV_LOG_ERROR, - "%s (%d) out of range.\n", sf_str[0], offset[0]); + "Scalefactor (%d) out of range.\n", offset[0]); return -1; } - sf[idx] = -ff_aac_pow2sf_tab[ offset[0] + sf_offset]; + sf[idx] = -ff_aac_pow2sf_tab[offset[0] - 100 + POW_SF2_ZERO]; } } } @@ -860,7 +1158,7 @@ static int decode_tns(AACContext *ac, TemporalNoiseShaping *tns, { int w, filt, i, coef_len, coef_res, coef_compress; const int is8 = ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE; - const int tns_max_order = is8 ? 7 : ac->m4ac.object_type == AOT_AAC_MAIN ? 20 : 12; + const int tns_max_order = is8 ? 7 : ac->oc[1].m4ac.object_type == AOT_AAC_MAIN ? 20 : 12; for (w = 0; w < ics->num_windows; w++) { if ((tns->n_filt[w] = get_bits(gb, 2 - is8))) { coef_res = get_bits1(gb); @@ -937,7 +1235,7 @@ static inline float *VMUL4(float *dst, const float *v, unsigned idx, static inline float *VMUL2S(float *dst, const float *v, unsigned idx, unsigned sign, const float *scale) { - union float754 s0, s1; + union av_intfloat32 s0, s1; s0.f = s1.f = *scale; s0.i ^= sign >> 1 << 31; @@ -955,22 +1253,22 @@ static inline float *VMUL4S(float *dst, const float *v, unsigned idx, unsigned sign, const float *scale) { unsigned nz = idx >> 12; - union float754 s = { .f = *scale }; - union float754 t; + union av_intfloat32 s = { .f = *scale }; + union av_intfloat32 t; - t.i = s.i ^ (sign & 1<<31); + t.i = s.i ^ (sign & 1U<<31); *dst++ = v[idx & 3] * t.f; sign <<= nz & 1; nz >>= 1; - t.i = s.i ^ (sign & 1<<31); + t.i = s.i ^ (sign & 1U<<31); *dst++ = v[idx>>2 & 3] * t.f; sign <<= nz & 1; nz >>= 1; - t.i = s.i ^ (sign & 1<<31); + t.i = s.i ^ (sign & 1U<<31); *dst++ = v[idx>>4 & 3] * t.f; sign <<= nz & 1; nz >>= 1; - t.i = s.i ^ (sign & 1<<31); + t.i = s.i ^ (sign & 1U<<31); *dst++ = v[idx>>6 & 3] * t.f; return dst; @@ -999,7 +1297,6 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], const int c = 1024 / ics->num_windows; const uint16_t *offsets = ics->swb_offset; float *coef_base = coef; - int err_idx; for (g = 0; g < ics->num_windows; g++) memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float) * (c - offsets[ics->max_sfb])); @@ -1035,7 +1332,6 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], const float *vq = ff_aac_codebook_vector_vals[cbt_m1]; const uint16_t *cb_vector_idx = ff_aac_codebook_vector_idx[cbt_m1]; VLC_TYPE (*vlc_tab)[2] = vlc_spectral[cbt_m1].table; - const int cb_size = ff_aac_spectral_sizes[cbt_m1]; OPEN_READER(re, gb); switch (cbt_m1 >> 1) { @@ -1050,12 +1346,6 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], UPDATE_CACHE(re, gb); GET_VLC(code, re, gb, vlc_tab, 8, 2); - - if (code >= cb_size) { - err_idx = code; - goto err_cb_overflow; - } - cb_idx = cb_vector_idx[code]; cf = VMUL4(cf, vq, cb_idx, sf + idx); } while (len -= 4); @@ -1075,18 +1365,9 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], UPDATE_CACHE(re, gb); GET_VLC(code, re, gb, vlc_tab, 8, 2); - - if (code >= cb_size) { - err_idx = code; - goto err_cb_overflow; - } - -#if MIN_CACHE_BITS < 20 - UPDATE_CACHE(re, gb); -#endif cb_idx = cb_vector_idx[code]; nnz = cb_idx >> 8 & 15; - bits = SHOW_UBITS(re, gb, nnz) << (32-nnz); + bits = nnz ? GET_CACHE(re, gb) : 0; LAST_SKIP_BITS(re, gb, nnz); cf = VMUL4S(cf, vq, cb_idx, bits, sf + idx); } while (len -= 4); @@ -1104,12 +1385,6 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], UPDATE_CACHE(re, gb); GET_VLC(code, re, gb, vlc_tab, 8, 2); - - if (code >= cb_size) { - err_idx = code; - goto err_cb_overflow; - } - cb_idx = cb_vector_idx[code]; cf = VMUL2(cf, vq, cb_idx, sf + idx); } while (len -= 2); @@ -1130,15 +1405,9 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], UPDATE_CACHE(re, gb); GET_VLC(code, re, gb, vlc_tab, 8, 2); - - if (code >= cb_size) { - err_idx = code; - goto err_cb_overflow; - } - cb_idx = cb_vector_idx[code]; nnz = cb_idx >> 8 & 15; - sign = SHOW_UBITS(re, gb, nnz) << (cb_idx >> 12); + sign = nnz ? SHOW_UBITS(re, gb, nnz) << (cb_idx >> 12) : 0; LAST_SKIP_BITS(re, gb, nnz); cf = VMUL2S(cf, vq, cb_idx, sign, sf + idx); } while (len -= 2); @@ -1167,11 +1436,6 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], continue; } - if (code >= cb_size) { - err_idx = code; - goto err_cb_overflow; - } - cb_idx = cb_vector_idx[code]; nnz = cb_idx >> 12; nzt = cb_idx >> 8; @@ -1193,20 +1457,15 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], return -1; } -#if MIN_CACHE_BITS < 21 - LAST_SKIP_BITS(re, gb, b + 1); - UPDATE_CACHE(re, gb); -#else SKIP_BITS(re, gb, b + 1); -#endif b += 4; n = (1 << b) + SHOW_UBITS(re, gb, b); LAST_SKIP_BITS(re, gb, b); - *icf++ = cbrt_tab[n] | (bits & 1<<31); + *icf++ = cbrt_tab[n] | (bits & 1U<<31); bits <<= 1; } else { unsigned v = ((const uint32_t*)vq)[cb_idx & 15]; - *icf++ = (bits & 1<<31) | v; + *icf++ = (bits & 1U<<31) | v; bits <<= !!v; } cb_idx >>= 4; @@ -1240,17 +1499,11 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], } } return 0; - -err_cb_overflow: - 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; } static av_always_inline float flt16_round(float pf) { - union float754 tmp; + union av_intfloat32 tmp; tmp.f = pf; tmp.i = (tmp.i + 0x00008000U) & 0xFFFF0000U; return tmp.f; @@ -1258,7 +1511,7 @@ static av_always_inline float flt16_round(float pf) static av_always_inline float flt16_even(float pf) { - union float754 tmp; + union av_intfloat32 tmp; tmp.f = pf; tmp.i = (tmp.i + 0x00007FFFU + (tmp.i & 0x00010000U >> 16)) & 0xFFFF0000U; return tmp.f; @@ -1266,37 +1519,40 @@ static av_always_inline float flt16_even(float pf) static av_always_inline float flt16_trunc(float pf) { - union float754 pun; + union av_intfloat32 pun; pun.f = pf; pun.i &= 0xFFFF0000U; return pun.f; } -static av_always_inline void predict(AACContext *ac, PredictorState *ps, float *coef, - int output_enable) +static av_always_inline void predict(PredictorState *ps, float *coef, + int output_enable) { const float a = 0.953125; // 61.0 / 64 const float alpha = 0.90625; // 29.0 / 32 float e0, e1; float pv; float k1, k2; + float r0 = ps->r0, r1 = ps->r1; + float cor0 = ps->cor0, cor1 = ps->cor1; + float var0 = ps->var0, var1 = ps->var1; - k1 = ps->var0 > 1 ? ps->cor0 * flt16_even(a / ps->var0) : 0; - k2 = ps->var1 > 1 ? ps->cor1 * flt16_even(a / ps->var1) : 0; + k1 = var0 > 1 ? cor0 * flt16_even(a / var0) : 0; + k2 = var1 > 1 ? cor1 * flt16_even(a / var1) : 0; - pv = flt16_round(k1 * ps->r0 + k2 * ps->r1); + pv = flt16_round(k1 * r0 + k2 * r1); if (output_enable) - *coef += pv * ac->sf_scale; + *coef += pv; - e0 = *coef / ac->sf_scale; - e1 = e0 - k1 * ps->r0; + e0 = *coef; + e1 = e0 - k1 * r0; - ps->cor1 = flt16_trunc(alpha * ps->cor1 + ps->r1 * e1); - ps->var1 = flt16_trunc(alpha * ps->var1 + 0.5 * (ps->r1 * ps->r1 + e1 * e1)); - ps->cor0 = flt16_trunc(alpha * ps->cor0 + ps->r0 * e0); - ps->var0 = flt16_trunc(alpha * ps->var0 + 0.5 * (ps->r0 * ps->r0 + e0 * e0)); + ps->cor1 = flt16_trunc(alpha * cor1 + r1 * e1); + ps->var1 = flt16_trunc(alpha * var1 + 0.5f * (r1 * r1 + e1 * e1)); + ps->cor0 = flt16_trunc(alpha * cor0 + r0 * e0); + ps->var0 = flt16_trunc(alpha * var0 + 0.5f * (r0 * r0 + e0 * e0)); - ps->r1 = flt16_trunc(a * (ps->r0 - k1 * e0)); + ps->r1 = flt16_trunc(a * (r0 - k1 * e0)); ps->r0 = flt16_trunc(a * e0); } @@ -1313,9 +1569,9 @@ static void apply_prediction(AACContext *ac, SingleChannelElement *sce) } if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) { - for (sfb = 0; sfb < ff_aac_pred_sfb_max[ac->m4ac.sampling_index]; sfb++) { + for (sfb = 0; sfb < ff_aac_pred_sfb_max[ac->oc[1].m4ac.sampling_index]; sfb++) { for (k = sce->ics.swb_offset[sfb]; k < sce->ics.swb_offset[sfb + 1]; k++) { - predict(ac, &sce->predictor_state[k], &sce->coeffs[k], + predict(&sce->predictor_state[k], &sce->coeffs[k], sce->ics.predictor_present && sce->ics.prediction_used[sfb]); } } @@ -1350,8 +1606,8 @@ static int decode_ics(AACContext *ac, SingleChannelElement *sce, global_gain = get_bits(gb, 8); if (!common_window && !scale_flag) { - if (decode_ics_info(ac, ics, gb, 0) < 0) - return -1; + if (decode_ics_info(ac, ics, gb) < 0) + return AVERROR_INVALIDDATA; } if (decode_band_types(ac, sce->band_type, sce->band_type_run_end, gb, ics) < 0) @@ -1382,7 +1638,7 @@ static int decode_ics(AACContext *ac, SingleChannelElement *sce, if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0) return -1; - if (ac->m4ac.object_type == AOT_AAC_MAIN && !common_window) + if (ac->oc[1].m4ac.object_type == AOT_AAC_MAIN && !common_window) apply_prediction(ac, sce); return 0; @@ -1421,13 +1677,13 @@ static void apply_mid_side_stereo(AACContext *ac, ChannelElement *cpe) * [1] mask is decoded from bitstream; [2] mask is all 1s; * [3] reserved for scalable AAC */ -static void apply_intensity_stereo(ChannelElement *cpe, int ms_present) +static void apply_intensity_stereo(AACContext *ac, ChannelElement *cpe, int ms_present) { const IndividualChannelStream *ics = &cpe->ch[1].ics; SingleChannelElement *sce1 = &cpe->ch[1]; float *coef0 = cpe->ch[0].coeffs, *coef1 = cpe->ch[1].coeffs; const uint16_t *offsets = ics->swb_offset; - int g, group, i, k, idx = 0; + int g, group, i, idx = 0; int c; float scale; for (g = 0; g < ics->num_window_groups; g++) { @@ -1440,8 +1696,10 @@ static void apply_intensity_stereo(ChannelElement *cpe, int ms_present) c *= 1 - 2 * cpe->ms_mask[idx]; scale = c * sce1->sf[idx]; for (group = 0; group < ics->group_len[g]; group++) - for (k = offsets[i]; k < offsets[i + 1]; k++) - coef1[group * 128 + k] = scale * coef0[group * 128 + k]; + ac->dsp.vector_fmul_scalar(coef1 + group * 128 + offsets[i], + coef0 + group * 128 + offsets[i], + scale, + offsets[i + 1] - offsets[i]); } } else { int bt_run_end = sce1->band_type_run_end[idx]; @@ -1457,8 +1715,6 @@ static void apply_intensity_stereo(ChannelElement *cpe, int ms_present) /** * Decode a channel_pair_element; reference: table 4.4. * - * @param elem_id Identifies the instance of a syntax element. - * * @return Returns error status. 0 - OK, !0 - error */ static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe) @@ -1467,11 +1723,14 @@ static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe) common_window = get_bits1(gb); if (common_window) { - if (decode_ics_info(ac, &cpe->ch[0].ics, gb, 1)) - return -1; + if (decode_ics_info(ac, &cpe->ch[0].ics, gb)) + return AVERROR_INVALIDDATA; i = cpe->ch[1].ics.use_kb_window[0]; cpe->ch[1].ics = cpe->ch[0].ics; cpe->ch[1].ics.use_kb_window[1] = i; + if (cpe->ch[1].ics.predictor_present && (ac->oc[1].m4ac.object_type != AOT_AAC_MAIN)) + if ((cpe->ch[1].ics.ltp.present = get_bits(gb, 1))) + decode_ltp(ac, &cpe->ch[1].ics.ltp, gb, cpe->ch[1].ics.max_sfb); ms_present = get_bits(gb, 2); if (ms_present == 3) { av_log(ac->avctx, AV_LOG_ERROR, "ms_present = 3 is reserved.\n"); @@ -1487,21 +1746,26 @@ static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe) if (common_window) { if (ms_present) apply_mid_side_stereo(ac, cpe); - if (ac->m4ac.object_type == AOT_AAC_MAIN) { + if (ac->oc[1].m4ac.object_type == AOT_AAC_MAIN) { apply_prediction(ac, &cpe->ch[0]); apply_prediction(ac, &cpe->ch[1]); } } - apply_intensity_stereo(cpe, ms_present); + apply_intensity_stereo(ac, cpe, ms_present); return 0; } +static const float cce_scale[] = { + 1.09050773266525765921, //2^(1/8) + 1.18920711500272106672, //2^(1/4) + M_SQRT2, + 2, +}; + /** * Decode coupling_channel_element; reference: table 4.8. * - * @param elem_id Identifies the instance of a syntax element. - * * @return Returns error status. 0 - OK, !0 - error */ static int decode_cce(AACContext *ac, GetBitContext *gb, ChannelElement *che) @@ -1529,7 +1793,7 @@ static int decode_cce(AACContext *ac, GetBitContext *gb, ChannelElement *che) coup->coupling_point += get_bits1(gb) || (coup->coupling_point >> 1); sign = get_bits(gb, 1); - scale = pow(2., pow(2., (int)get_bits(gb, 2) - 3)); + scale = cce_scale[get_bits(gb, 2)]; if ((ret = decode_ics(ac, sce, gb, 0, 0))) return ret; @@ -1542,7 +1806,7 @@ static int decode_cce(AACContext *ac, GetBitContext *gb, ChannelElement *che) if (c) { cge = coup->coupling_point == AFTER_IMDCT ? 1 : get_bits1(gb); gain = cge ? get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60: 0; - gain_cache = pow(scale, -gain); + gain_cache = powf(scale, -gain); } if (coup->coupling_point == AFTER_IMDCT) { coup->gain[c][0] = gain_cache; @@ -1559,7 +1823,7 @@ static int decode_cce(AACContext *ac, GetBitContext *gb, ChannelElement *che) s -= 2 * (t & 0x1); t >>= 1; } - gain_cache = pow(scale, -t) * s; + gain_cache = powf(scale, -t) * s; } } coup->gain[c][idx] = gain_cache; @@ -1663,20 +1927,21 @@ static int decode_extension_payload(AACContext *ac, GetBitContext *gb, int cnt, if (!che) { av_log(ac->avctx, AV_LOG_ERROR, "SBR was found before the first channel element.\n"); return res; - } else if (!ac->m4ac.sbr) { + } else if (!ac->oc[1].m4ac.sbr) { 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) { + } else if (ac->oc[1].m4ac.sbr == -1 && ac->oc[1].status == OC_LOCKED) { 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 if (ac->oc[1].m4ac.ps == -1 && ac->oc[1].status < OC_LOCKED && ac->avctx->channels == 1) { + ac->oc[1].m4ac.sbr = 1; + ac->oc[1].m4ac.ps = 1; + output_configure(ac, ac->oc[1].layout_map, ac->oc[1].layout_map_tags, + ac->oc[1].m4ac.chan_config, ac->oc[1].status); } else { - ac->m4ac.sbr = 1; + ac->oc[1].m4ac.sbr = 1; } res = ff_decode_sbr_extension(ac, &che->sbr, gb, crc_flag, cnt, elem_type); break; @@ -1706,6 +1971,7 @@ static void apply_tns(float coef[1024], TemporalNoiseShaping *tns, int w, filt, m, i; int bottom, top, order, start, end, size, inc; float lpc[TNS_MAX_ORDER]; + float tmp[TNS_MAX_ORDER]; for (w = 0; w < ics->num_windows; w++) { bottom = ics->num_swb; @@ -1731,18 +1997,123 @@ static void apply_tns(float coef[1024], TemporalNoiseShaping *tns, } start += w * 128; - // ar filter - for (m = 0; m < size; m++, start += inc) - for (i = 1; i <= FFMIN(m, order); i++) - coef[start] -= coef[start - i * inc] * lpc[i - 1]; + if (decode) { + // ar filter + for (m = 0; m < size; m++, start += inc) + for (i = 1; i <= FFMIN(m, order); i++) + coef[start] -= coef[start - i * inc] * lpc[i - 1]; + } else { + // ma filter + for (m = 0; m < size; m++, start += inc) { + tmp[0] = coef[start]; + for (i = 1; i <= FFMIN(m, order); i++) + coef[start] += tmp[i] * lpc[i - 1]; + for (i = order; i > 0; i--) + tmp[i] = tmp[i - 1]; + } + } } } } +/** + * Apply windowing and MDCT to obtain the spectral + * coefficient from the predicted sample by LTP. + */ +static void windowing_and_mdct_ltp(AACContext *ac, float *out, + float *in, IndividualChannelStream *ics) +{ + const float *lwindow = ics->use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024; + const float *swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128; + const float *lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024; + const float *swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128; + + if (ics->window_sequence[0] != LONG_STOP_SEQUENCE) { + ac->dsp.vector_fmul(in, in, lwindow_prev, 1024); + } else { + memset(in, 0, 448 * sizeof(float)); + ac->dsp.vector_fmul(in + 448, in + 448, swindow_prev, 128); + } + if (ics->window_sequence[0] != LONG_START_SEQUENCE) { + ac->dsp.vector_fmul_reverse(in + 1024, in + 1024, lwindow, 1024); + } else { + ac->dsp.vector_fmul_reverse(in + 1024 + 448, in + 1024 + 448, swindow, 128); + memset(in + 1024 + 576, 0, 448 * sizeof(float)); + } + ac->mdct_ltp.mdct_calc(&ac->mdct_ltp, out, in); +} + +/** + * Apply the long term prediction + */ +static void apply_ltp(AACContext *ac, SingleChannelElement *sce) +{ + const LongTermPrediction *ltp = &sce->ics.ltp; + const uint16_t *offsets = sce->ics.swb_offset; + int i, sfb; + + if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) { + float *predTime = sce->ret; + float *predFreq = ac->buf_mdct; + int16_t num_samples = 2048; + + if (ltp->lag < 1024) + num_samples = ltp->lag + 1024; + for (i = 0; i < num_samples; i++) + predTime[i] = sce->ltp_state[i + 2048 - ltp->lag] * ltp->coef; + memset(&predTime[i], 0, (2048 - i) * sizeof(float)); + + windowing_and_mdct_ltp(ac, predFreq, predTime, &sce->ics); + + if (sce->tns.present) + apply_tns(predFreq, &sce->tns, &sce->ics, 0); + + for (sfb = 0; sfb < FFMIN(sce->ics.max_sfb, MAX_LTP_LONG_SFB); sfb++) + if (ltp->used[sfb]) + for (i = offsets[sfb]; i < offsets[sfb + 1]; i++) + sce->coeffs[i] += predFreq[i]; + } +} + +/** + * Update the LTP buffer for next frame + */ +static void update_ltp(AACContext *ac, SingleChannelElement *sce) +{ + IndividualChannelStream *ics = &sce->ics; + float *saved = sce->saved; + float *saved_ltp = sce->coeffs; + const float *lwindow = ics->use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024; + const float *swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128; + int i; + + if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) { + memcpy(saved_ltp, saved, 512 * sizeof(float)); + memset(saved_ltp + 576, 0, 448 * sizeof(float)); + ac->dsp.vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960, &swindow[64], 64); + for (i = 0; i < 64; i++) + saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * swindow[63 - i]; + } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) { + memcpy(saved_ltp, ac->buf_mdct + 512, 448 * sizeof(float)); + memset(saved_ltp + 576, 0, 448 * sizeof(float)); + ac->dsp.vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960, &swindow[64], 64); + for (i = 0; i < 64; i++) + saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * swindow[63 - i]; + } else { // LONG_STOP or ONLY_LONG + ac->dsp.vector_fmul_reverse(saved_ltp, ac->buf_mdct + 512, &lwindow[512], 512); + for (i = 0; i < 512; i++) + saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * lwindow[511 - i]; + } + + memcpy(sce->ltp_state, sce->ltp_state+1024, 1024 * sizeof(*sce->ltp_state)); + memcpy(sce->ltp_state+1024, sce->ret, 1024 * sizeof(*sce->ltp_state)); + memcpy(sce->ltp_state+2048, saved_ltp, 1024 * sizeof(*sce->ltp_state)); +} + /** * Conduct IMDCT and windowing. */ -static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce, float bias) +static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce) { IndividualChannelStream *ics = &sce->ics; float *in = sce->coeffs; @@ -1757,14 +2128,10 @@ 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->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) - ff_imdct_half(&ac->mdct_small, buf + i, in + i); + ac->mdct_small.imdct_half(&ac->mdct_small, buf + i, in + i); } else - ff_imdct_half(&ac->mdct, buf, in); + ac->mdct.imdct_half(&ac->mdct, buf, in); /* window overlapping * NOTE: To simplify the overlapping code, all 'meaningless' short to long @@ -1774,32 +2141,29 @@ static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce, float */ if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) && (ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) { - ac->dsp.vector_fmul_window( out, saved, buf, lwindow_prev, bias, 512); + ac->dsp.vector_fmul_window( out, saved, buf, lwindow_prev, 512); } else { - for (i = 0; i < 448; i++) - out[i] = saved[i] + bias; + memcpy( out, saved, 448 * sizeof(float)); if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) { - ac->dsp.vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, bias, 64); - ac->dsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, bias, 64); - ac->dsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, bias, 64); - ac->dsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, bias, 64); - ac->dsp.vector_fmul_window(temp, buf + 3*128 + 64, buf + 4*128, swindow, bias, 64); + ac->dsp.vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, 64); + ac->dsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, 64); + ac->dsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, 64); + ac->dsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, 64); + ac->dsp.vector_fmul_window(temp, buf + 3*128 + 64, buf + 4*128, swindow, 64); memcpy( out + 448 + 4*128, temp, 64 * sizeof(float)); } else { - ac->dsp.vector_fmul_window(out + 448, saved + 448, buf, swindow_prev, bias, 64); - for (i = 576; i < 1024; i++) - out[i] = buf[i-512] + bias; + ac->dsp.vector_fmul_window(out + 448, saved + 448, buf, swindow_prev, 64); + memcpy( out + 576, buf + 64, 448 * sizeof(float)); } } // buffer update if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) { - for (i = 0; i < 64; i++) - saved[i] = temp[64 + i] - bias; - ac->dsp.vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 0, 64); - ac->dsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 0, 64); - ac->dsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 0, 64); + memcpy( saved, temp + 64, 64 * sizeof(float)); + ac->dsp.vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 64); + ac->dsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 64); + ac->dsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 64); memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(float)); } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) { memcpy( saved, buf + 512, 448 * sizeof(float)); @@ -1823,7 +2187,7 @@ static void apply_dependent_coupling(AACContext *ac, float *dest = target->coeffs; const float *src = cce->ch[0].coeffs; int g, i, group, k, idx = 0; - if (ac->m4ac.object_type == AOT_AAC_LTP) { + if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP) { av_log(ac->avctx, AV_LOG_ERROR, "Dependent coupling is not supported together with LTP\n"); return; @@ -1856,19 +2220,17 @@ static void apply_independent_coupling(AACContext *ac, { int i; const float gain = cce->coup.gain[index][0]; - const float bias = ac->add_bias; const float *src = cce->ch[0].ret; float *dest = target->ret; - const int len = 1024 << (ac->m4ac.sbr == 1); + const int len = 1024 << (ac->oc[1].m4ac.sbr == 1); for (i = 0; i < len; i++) - dest[i] += gain * (src[i] - bias); + dest[i] += gain * src[i]; } /** * channel coupling transformation interface * - * @param index index into coupling gain array * @param apply_coupling_method pointer to (in)dependent coupling function */ static void apply_channel_coupling(AACContext *ac, ChannelElement *cc, @@ -1907,13 +2269,20 @@ static void apply_channel_coupling(AACContext *ac, ChannelElement *cc, static void spectral_to_sample(AACContext *ac) { int i, type; - float imdct_bias = (ac->m4ac.sbr <= 0) ? ac->add_bias : 0.0f; for (type = 3; type >= 0; type--) { for (i = 0; i < MAX_ELEM_ID; i++) { ChannelElement *che = ac->che[type][i]; if (che) { if (type <= TYPE_CPE) apply_channel_coupling(ac, che, type, i, BEFORE_TNS, apply_dependent_coupling); + if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP) { + if (che->ch[0].ics.predictor_present) { + if (che->ch[0].ics.ltp.present) + apply_ltp(ac, &che->ch[0]); + if (che->ch[1].ics.ltp.present && type == TYPE_CPE) + apply_ltp(ac, &che->ch[1]); + } + } if (che->ch[0].tns.present) apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1); if (che->ch[1].tns.present) @@ -1921,11 +2290,15 @@ static void spectral_to_sample(AACContext *ac) if (type <= TYPE_CPE) apply_channel_coupling(ac, che, type, i, BETWEEN_TNS_AND_IMDCT, apply_dependent_coupling); if (type != TYPE_CCE || che->coup.coupling_point == AFTER_IMDCT) { - imdct_and_windowing(ac, &che->ch[0], imdct_bias); + imdct_and_windowing(ac, &che->ch[0]); + if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP) + update_ltp(ac, &che->ch[0]); if (type == TYPE_CPE) { - imdct_and_windowing(ac, &che->ch[1], imdct_bias); + imdct_and_windowing(ac, &che->ch[1]); + if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP) + update_ltp(ac, &che->ch[1]); } - if (ac->m4ac.sbr > 0) { + if (ac->oc[1].m4ac.sbr > 0) { ff_sbr_apply(ac, &che->sbr, type, che->ch[0].ret, che->ch[1].ret); } } @@ -1940,77 +2313,76 @@ static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb) { int size; AACADTSHeaderInfo hdr_info; + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int layout_map_tags; - size = ff_aac_parse_header(gb, &hdr_info); + size = avpriv_aac_parse_header(gb, &hdr_info); if (size > 0) { - if (ac->output_configured != OC_LOCKED && hdr_info.chan_config) { - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID]; - memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0])); - ac->m4ac.chan_config = hdr_info.chan_config; - if (set_default_channel_config(ac, new_che_pos, hdr_info.chan_config)) + if (hdr_info.num_aac_frames != 1) { + av_log_missing_feature(ac->avctx, "More than one AAC RDB per ADTS frame is", 0); + return -1; + } + push_output_configuration(ac); + if (hdr_info.chan_config) { + ac->oc[1].m4ac.chan_config = hdr_info.chan_config; + if (set_default_channel_config(ac->avctx, layout_map, + &layout_map_tags, hdr_info.chan_config)) return -7; - if (output_configure(ac, ac->che_pos, new_che_pos, hdr_info.chan_config, OC_TRIAL_FRAME)) + if (output_configure(ac, layout_map, layout_map_tags, + hdr_info.chan_config, + FFMAX(ac->oc[1].status, OC_TRIAL_FRAME))) return -7; - } else if (ac->output_configured != OC_LOCKED) { - ac->output_configured = OC_NONE; - } - 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->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->avctx, "More than one AAC RDB per ADTS frame is", 0); - return -1; + ac->oc[1].m4ac.chan_config = 0; } + ac->oc[1].m4ac.sample_rate = hdr_info.sample_rate; + ac->oc[1].m4ac.sampling_index = hdr_info.sampling_index; + ac->oc[1].m4ac.object_type = hdr_info.object_type; + if (ac->oc[0].status != OC_LOCKED || + ac->oc[0].m4ac.chan_config != hdr_info.chan_config || + ac->oc[0].m4ac.sample_rate != hdr_info.sample_rate) { + ac->oc[1].m4ac.sbr = -1; + ac->oc[1].m4ac.ps = -1; + } + if (!hdr_info.crc_absent) + skip_bits(gb, 16); } return size; } -static int aac_decode_frame(AVCodecContext *avctx, void *data, - int *data_size, AVPacket *avpkt) +static int aac_decode_frame_int(AVCodecContext *avctx, void *data, + int *got_frame_ptr, GetBitContext *gb) { - const uint8_t *buf = avpkt->data; - int buf_size = avpkt->size; 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 = 0, multiplier; - int buf_offset; + int err, elem_id; + int samples = 0, multiplier, audio_found = 0, pce_found = 0; - init_get_bits(&gb, buf, buf_size * 8); - - if (show_bits(&gb, 12) == 0xfff) { - if (parse_adts_frame_header(ac, &gb) < 0) { + if (show_bits(gb, 12) == 0xfff) { + if (parse_adts_frame_header(ac, gb) < 0) { av_log(avctx, AV_LOG_ERROR, "Error decoding AAC frame header.\n"); - return -1; + err = -1; + goto fail; } - if (ac->m4ac.sampling_index > 12) { - av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index); - return -1; + if (ac->oc[1].m4ac.sampling_index > 12) { + av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->oc[1].m4ac.sampling_index); + err = -1; + goto fail; } } - memset(ac->tags_seen_this_frame, 0, sizeof(ac->tags_seen_this_frame)); + ac->tags_mapped = 0; // parse - while ((elem_type = get_bits(&gb, 3)) != TYPE_END) { - elem_id = get_bits(&gb, 4); + while ((elem_type = get_bits(gb, 3)) != TYPE_END) { + elem_id = get_bits(gb, 4); 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; + err = -1; + goto fail; } samples = 1024; } @@ -2018,47 +2390,58 @@ static int aac_decode_frame(AVCodecContext *avctx, void *data, switch (elem_type) { case TYPE_SCE: - err = decode_ics(ac, &che->ch[0], &gb, 0, 0); + err = decode_ics(ac, &che->ch[0], gb, 0, 0); + audio_found = 1; break; case TYPE_CPE: - err = decode_cpe(ac, &gb, che); + err = decode_cpe(ac, gb, che); + audio_found = 1; break; case TYPE_CCE: - err = decode_cce(ac, &gb, che); + err = decode_cce(ac, gb, che); break; case TYPE_LFE: - err = decode_ics(ac, &che->ch[0], &gb, 0, 0); + err = decode_ics(ac, &che->ch[0], gb, 0, 0); + audio_found = 1; break; case TYPE_DSE: - err = skip_data_stream_element(ac, &gb); + err = skip_data_stream_element(ac, gb); break; case TYPE_PCE: { - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID]; - memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0])); - if ((err = decode_pce(ac, new_che_pos, &gb))) + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int tags; + push_output_configuration(ac); + tags = decode_pce(avctx, &ac->oc[1].m4ac, layout_map, gb); + if (tags < 0) { + err = tags; break; - if (ac->output_configured > OC_TRIAL_PCE) + } + if (pce_found) { 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); + pop_output_configuration(ac); + } else { + err = output_configure(ac, layout_map, tags, 0, OC_TRIAL_PCE); + pce_found = 1; + } break; } case TYPE_FIL: if (elem_id == 15) - elem_id += get_bits(&gb, 8) - 1; - if (get_bits_left(&gb) < 8 * elem_id) { + elem_id += get_bits(gb, 8) - 1; + if (get_bits_left(gb) < 8 * elem_id) { av_log(avctx, AV_LOG_ERROR, overread_err); - return -1; + err = -1; + goto fail; } while (elem_id > 0) - elem_id -= decode_extension_payload(ac, &gb, elem_id, che_prev, elem_type_prev); + elem_id -= decode_extension_payload(ac, gb, elem_id, che_prev, elem_type_prev); err = 0; /* FIXME */ break; @@ -2071,37 +2454,90 @@ static int aac_decode_frame(AVCodecContext *avctx, void *data, elem_type_prev = elem_type; if (err) - return err; + goto fail; - if (get_bits_left(&gb) < 3) { + if (get_bits_left(gb) < 3) { av_log(avctx, AV_LOG_ERROR, overread_err); - return -1; + err = -1; + goto fail; } } spectral_to_sample(ac); - multiplier = (ac->m4ac.sbr == 1) ? ac->m4ac.ext_sample_rate > ac->m4ac.sample_rate : 0; + multiplier = (ac->oc[1].m4ac.sbr == 1) ? ac->oc[1].m4ac.ext_sample_rate > ac->oc[1].m4ac.sample_rate : 0; samples <<= multiplier; - if (ac->output_configured < OC_LOCKED) { - avctx->sample_rate = ac->m4ac.sample_rate << multiplier; + + if (samples) { + /* get output buffer */ + ac->frame.nb_samples = samples; + if ((err = avctx->get_buffer(avctx, &ac->frame)) < 0) { + av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); + err = -1; + goto fail; + } + + if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) + ac->fmt_conv.float_interleave((float *)ac->frame.data[0], + (const float **)ac->output_data, + samples, avctx->channels); + else + ac->fmt_conv.float_to_int16_interleave((int16_t *)ac->frame.data[0], + (const float **)ac->output_data, + samples, avctx->channels); + + *(AVFrame *)data = ac->frame; + } + *got_frame_ptr = !!samples; + + if (ac->oc[1].status && audio_found) { + avctx->sample_rate = ac->oc[1].m4ac.sample_rate << multiplier; avctx->frame_size = samples; + ac->oc[1].status = OC_LOCKED; } - data_size_tmp = samples * avctx->channels * sizeof(int16_t); - if (*data_size < data_size_tmp) { - 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; + return 0; +fail: + pop_output_configuration(ac); + return err; +} + +static int aac_decode_frame(AVCodecContext *avctx, void *data, + int *got_frame_ptr, AVPacket *avpkt) +{ + AACContext *ac = avctx->priv_data; + const uint8_t *buf = avpkt->data; + int buf_size = avpkt->size; + GetBitContext gb; + int buf_consumed; + int buf_offset; + int err; + int new_extradata_size; + const uint8_t *new_extradata = av_packet_get_side_data(avpkt, + AV_PKT_DATA_NEW_EXTRADATA, + &new_extradata_size); + + if (new_extradata) { + av_free(avctx->extradata); + avctx->extradata = av_mallocz(new_extradata_size + + FF_INPUT_BUFFER_PADDING_SIZE); + if (!avctx->extradata) + return AVERROR(ENOMEM); + avctx->extradata_size = new_extradata_size; + memcpy(avctx->extradata, new_extradata, new_extradata_size); + push_output_configuration(ac); + if (decode_audio_specific_config(ac, ac->avctx, &ac->oc[1].m4ac, + avctx->extradata, + avctx->extradata_size*8, 1) < 0) { + pop_output_configuration(ac); + return AVERROR_INVALIDDATA; + } } - *data_size = data_size_tmp; - if (samples) - ac->dsp.float_to_int16_interleave(data, (const float **)ac->output_data, samples, avctx->channels); + init_get_bits(&gb, buf, buf_size * 8); - if (ac->output_configured) - ac->output_configured = OC_LOCKED; + if ((err = aac_decode_frame_int(avctx, data, got_frame_ptr, &gb)) < 0) + return err; buf_consumed = (get_bits_count(&gb) + 7) >> 3; for (buf_offset = buf_consumed; buf_offset < buf_size; buf_offset++) @@ -2126,21 +2562,311 @@ static av_cold int aac_decode_close(AVCodecContext *avctx) ff_mdct_end(&ac->mdct); ff_mdct_end(&ac->mdct_small); + ff_mdct_end(&ac->mdct_ltp); + return 0; +} + + +#define LOAS_SYNC_WORD 0x2b7 ///< 11 bits LOAS sync word + +struct LATMContext { + AACContext aac_ctx; ///< containing AACContext + int initialized; ///< initilized after a valid extradata was seen + + // parser data + int audio_mux_version_A; ///< LATM syntax version + int frame_length_type; ///< 0/1 variable/fixed frame length + int frame_length; ///< frame length for fixed frame length +}; + +static inline uint32_t latm_get_value(GetBitContext *b) +{ + int length = get_bits(b, 2); + + return get_bits_long(b, (length+1)*8); +} + +static int latm_decode_audio_specific_config(struct LATMContext *latmctx, + GetBitContext *gb, int asclen) +{ + AACContext *ac = &latmctx->aac_ctx; + AVCodecContext *avctx = ac->avctx; + MPEG4AudioConfig m4ac = { 0 }; + int config_start_bit = get_bits_count(gb); + int sync_extension = 0; + int bits_consumed, esize; + + if (asclen) { + sync_extension = 1; + asclen = FFMIN(asclen, get_bits_left(gb)); + } else + asclen = get_bits_left(gb); + + if (config_start_bit % 8) { + av_log_missing_feature(latmctx->aac_ctx.avctx, "audio specific " + "config not byte aligned.\n", 1); + return AVERROR_INVALIDDATA; + } + if (asclen <= 0) + return AVERROR_INVALIDDATA; + bits_consumed = decode_audio_specific_config(NULL, avctx, &m4ac, + gb->buffer + (config_start_bit / 8), + asclen, sync_extension); + + if (bits_consumed < 0) + return AVERROR_INVALIDDATA; + + if (ac->oc[1].m4ac.sample_rate != m4ac.sample_rate || + ac->oc[1].m4ac.chan_config != m4ac.chan_config) { + + av_log(avctx, AV_LOG_INFO, "audio config changed\n"); + latmctx->initialized = 0; + + esize = (bits_consumed+7) / 8; + + if (avctx->extradata_size < esize) { + av_free(avctx->extradata); + avctx->extradata = av_malloc(esize + FF_INPUT_BUFFER_PADDING_SIZE); + if (!avctx->extradata) + return AVERROR(ENOMEM); + } + + avctx->extradata_size = esize; + memcpy(avctx->extradata, gb->buffer + (config_start_bit/8), esize); + memset(avctx->extradata+esize, 0, FF_INPUT_BUFFER_PADDING_SIZE); + } + skip_bits_long(gb, bits_consumed); + + return bits_consumed; +} + +static int read_stream_mux_config(struct LATMContext *latmctx, + GetBitContext *gb) +{ + int ret, audio_mux_version = get_bits(gb, 1); + + latmctx->audio_mux_version_A = 0; + if (audio_mux_version) + latmctx->audio_mux_version_A = get_bits(gb, 1); + + if (!latmctx->audio_mux_version_A) { + + if (audio_mux_version) + latm_get_value(gb); // taraFullness + + skip_bits(gb, 1); // allStreamSameTimeFraming + skip_bits(gb, 6); // numSubFrames + // numPrograms + if (get_bits(gb, 4)) { // numPrograms + av_log_missing_feature(latmctx->aac_ctx.avctx, + "multiple programs are not supported\n", 1); + return AVERROR_PATCHWELCOME; + } + + // for each program (which there is only on in DVB) + + // for each layer (which there is only on in DVB) + if (get_bits(gb, 3)) { // numLayer + av_log_missing_feature(latmctx->aac_ctx.avctx, + "multiple layers are not supported\n", 1); + return AVERROR_PATCHWELCOME; + } + + // for all but first stream: use_same_config = get_bits(gb, 1); + if (!audio_mux_version) { + if ((ret = latm_decode_audio_specific_config(latmctx, gb, 0)) < 0) + return ret; + } else { + int ascLen = latm_get_value(gb); + if ((ret = latm_decode_audio_specific_config(latmctx, gb, ascLen)) < 0) + return ret; + ascLen -= ret; + skip_bits_long(gb, ascLen); + } + + latmctx->frame_length_type = get_bits(gb, 3); + switch (latmctx->frame_length_type) { + case 0: + skip_bits(gb, 8); // latmBufferFullness + break; + case 1: + latmctx->frame_length = get_bits(gb, 9); + break; + case 3: + case 4: + case 5: + skip_bits(gb, 6); // CELP frame length table index + break; + case 6: + case 7: + skip_bits(gb, 1); // HVXC frame length table index + break; + } + + if (get_bits(gb, 1)) { // other data + if (audio_mux_version) { + latm_get_value(gb); // other_data_bits + } else { + int esc; + do { + esc = get_bits(gb, 1); + skip_bits(gb, 8); + } while (esc); + } + } + + if (get_bits(gb, 1)) // crc present + skip_bits(gb, 8); // config_crc + } + + return 0; +} + +static int read_payload_length_info(struct LATMContext *ctx, GetBitContext *gb) +{ + uint8_t tmp; + + if (ctx->frame_length_type == 0) { + int mux_slot_length = 0; + do { + tmp = get_bits(gb, 8); + mux_slot_length += tmp; + } while (tmp == 255); + return mux_slot_length; + } else if (ctx->frame_length_type == 1) { + return ctx->frame_length; + } else if (ctx->frame_length_type == 3 || + ctx->frame_length_type == 5 || + ctx->frame_length_type == 7) { + skip_bits(gb, 2); // mux_slot_length_coded + } + return 0; +} + +static int read_audio_mux_element(struct LATMContext *latmctx, + GetBitContext *gb) +{ + int err; + uint8_t use_same_mux = get_bits(gb, 1); + if (!use_same_mux) { + if ((err = read_stream_mux_config(latmctx, gb)) < 0) + return err; + } else if (!latmctx->aac_ctx.avctx->extradata) { + av_log(latmctx->aac_ctx.avctx, AV_LOG_DEBUG, + "no decoder config found\n"); + return AVERROR(EAGAIN); + } + if (latmctx->audio_mux_version_A == 0) { + int mux_slot_length_bytes = read_payload_length_info(latmctx, gb); + if (mux_slot_length_bytes * 8 > get_bits_left(gb)) { + av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR, "incomplete frame\n"); + return AVERROR_INVALIDDATA; + } else if (mux_slot_length_bytes * 8 + 256 < get_bits_left(gb)) { + av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR, + "frame length mismatch %d << %d\n", + mux_slot_length_bytes * 8, get_bits_left(gb)); + return AVERROR_INVALIDDATA; + } + } return 0; } -AVCodec aac_decoder = { - "aac", - AVMEDIA_TYPE_AUDIO, - CODEC_ID_AAC, - sizeof(AACContext), - aac_decode_init, - NULL, - aac_decode_close, - aac_decode_frame, - .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"), - .sample_fmts = (const enum SampleFormat[]) { - SAMPLE_FMT_S16,SAMPLE_FMT_NONE + +static int latm_decode_frame(AVCodecContext *avctx, void *out, + int *got_frame_ptr, AVPacket *avpkt) +{ + struct LATMContext *latmctx = avctx->priv_data; + int muxlength, err; + GetBitContext gb; + + init_get_bits(&gb, avpkt->data, avpkt->size * 8); + + // check for LOAS sync word + if (get_bits(&gb, 11) != LOAS_SYNC_WORD) + return AVERROR_INVALIDDATA; + + muxlength = get_bits(&gb, 13) + 3; + // not enough data, the parser should have sorted this + if (muxlength > avpkt->size) + return AVERROR_INVALIDDATA; + + if ((err = read_audio_mux_element(latmctx, &gb)) < 0) + return err; + + if (!latmctx->initialized) { + if (!avctx->extradata) { + *got_frame_ptr = 0; + return avpkt->size; + } else { + push_output_configuration(&latmctx->aac_ctx); + if ((err = decode_audio_specific_config( + &latmctx->aac_ctx, avctx, &latmctx->aac_ctx.oc[1].m4ac, + avctx->extradata, avctx->extradata_size*8, 1)) < 0) { + pop_output_configuration(&latmctx->aac_ctx); + return err; + } + latmctx->initialized = 1; + } + } + + if (show_bits(&gb, 12) == 0xfff) { + av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR, + "ADTS header detected, probably as result of configuration " + "misparsing\n"); + return AVERROR_INVALIDDATA; + } + + if ((err = aac_decode_frame_int(avctx, out, got_frame_ptr, &gb)) < 0) + return err; + + return muxlength; +} + +static av_cold int latm_decode_init(AVCodecContext *avctx) +{ + struct LATMContext *latmctx = avctx->priv_data; + int ret = aac_decode_init(avctx); + + if (avctx->extradata_size > 0) + latmctx->initialized = !ret; + + return ret; +} + + +AVCodec ff_aac_decoder = { + .name = "aac", + .type = AVMEDIA_TYPE_AUDIO, + .id = CODEC_ID_AAC, + .priv_data_size = sizeof(AACContext), + .init = aac_decode_init, + .close = aac_decode_close, + .decode = aac_decode_frame, + .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"), + .sample_fmts = (const enum AVSampleFormat[]) { + AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE + }, + .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1, + .channel_layouts = aac_channel_layout, +}; + +/* + Note: This decoder filter is intended to decode LATM streams transferred + in MPEG transport streams which only contain one program. + To do a more complex LATM demuxing a separate LATM demuxer should be used. +*/ +AVCodec ff_aac_latm_decoder = { + .name = "aac_latm", + .type = AVMEDIA_TYPE_AUDIO, + .id = CODEC_ID_AAC_LATM, + .priv_data_size = sizeof(struct LATMContext), + .init = latm_decode_init, + .close = aac_decode_close, + .decode = latm_decode_frame, + .long_name = NULL_IF_CONFIG_SMALL("AAC LATM (Advanced Audio Codec LATM syntax)"), + .sample_fmts = (const enum AVSampleFormat[]) { + AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE }, + .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1, .channel_layouts = aac_channel_layout, };