X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Faacdec.c;h=e436b4f2f74851ffad04d5dc3c88407c272afeb7;hb=83678dbbae64ad8c501e0c732c1117e642c25dae;hp=1015030b9a89491534d0206b4e7d243e229570ee;hpb=963f6855356fa527a27b08b55e026f683a12cebc;p=ffmpeg diff --git a/libavcodec/aacdec.c b/libavcodec/aacdec.c index 1015030b9a8..e436b4f2f74 100644 --- a/libavcodec/aacdec.c +++ b/libavcodec/aacdec.c @@ -2,10 +2,11 @@ * AAC decoder * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org ) * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com ) + * Copyright (c) 2008-2013 Alex Converse * * AAC LATM decoder * Copyright (c) 2008-2010 Paul Kendall - * Copyright (c) 2010 Janne Grunau + * Copyright (c) 2010 Janne Grunau * * This file is part of Libav. * @@ -79,13 +80,12 @@ Parametric Stereo. */ - +#include "libavutil/float_dsp.h" #include "avcodec.h" #include "internal.h" #include "get_bits.h" -#include "dsputil.h" #include "fft.h" -#include "fmtconvert.h" +#include "imdct15.h" #include "lpc.h" #include "kbdwin.h" #include "sinewin.h" @@ -93,78 +93,40 @@ #include "aac.h" #include "aactab.h" #include "aacdectab.h" +#include "adts_header.h" #include "cbrt_tablegen.h" #include "sbr.h" #include "aacsbr.h" #include "mpeg4audio.h" -#include "aacadtsdec.h" +#include "libavutil/intfloat.h" #include #include #include +#include #include #if ARCH_ARM # include "arm/aac.h" #endif -union float754 { - float f; - uint32_t i; -}; +#include "libavutil/thread.h" 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) { - // For PCE based channel configurations map the channels solely based on tags. - if (!ac->m4ac.chan_config) { - return ac->tag_che_map[type][elem_id]; - } - // For indexed channel configurations map the channels solely based on position. - 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; } /** @@ -180,20 +142,22 @@ 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], + enum ChannelPosition che_pos, int type, int id, int *channels) { - if (che_pos[type][id]) { + 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 (*channels >= MAX_CHANNELS - 2) + return AVERROR_INVALIDDATA; + ac->output_element[(*channels)++] = &ac->che[type][id]->ch[0]; if (type == TYPE_CPE || - (type == TYPE_SCE && ac->m4ac.ps == 1)) { - ac->output_data[(*channels)++] = ac->che[type][id]->ch[1].ret; + (type == TYPE_SCE && ac->oc[1].m4ac.ps == 1)) { + ac->output_element[(*channels)++] = &ac->che[type][id]->ch[1]; } } } else { @@ -204,103 +168,517 @@ static av_cold int che_configure(AACContext *ac, return 0; } -/** - * 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 frame_configure_elements(AVCodecContext *avctx) { - 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])); + AACContext *ac = avctx->priv_data; + int type, id, ch, ret; - 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; + /* set channel pointers to internal buffers by default */ + for (type = 0; type < 4; type++) { + for (id = 0; id < MAX_ELEM_ID; id++) { + ChannelElement *che = ac->che[type][id]; + if (che) { + che->ch[0].ret = che->ch[0].ret_buf; + che->ch[1].ret = che->ch[1].ret_buf; + } } + } + + /* get output buffer */ + av_frame_unref(ac->frame); + ac->frame->nb_samples = 2048; + if ((ret = ff_get_buffer(avctx, ac->frame, 0)) < 0) { + av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); + return ret; + } + + /* map output channel pointers to AVFrame data */ + for (ch = 0; ch < avctx->channels; ch++) { + if (ac->output_element[ch]) + ac->output_element[ch]->ret = (float *)ac->frame->extended_data[ch]; + } + + return 0; +} - memset(ac->tag_che_map, 0, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0])); +struct elem_to_channel { + uint64_t av_position; + uint8_t syn_ele; + uint8_t elem_id; + uint8_t aac_position; +}; - avctx->channel_layout = aac_channel_layout[channel_config - 1]; +static int assign_pair(struct elem_to_channel e2c_vec[MAX_ELEM_ID], + uint8_t (*layout_map)[3], int offset, 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 { - /* 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 ] - */ + 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; + } +} - 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; +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; + + if (num_side_channels == 0 && num_back_channels >= 4) { + num_side_channels = 2; + num_back_channels -= 2; + } + + 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, + 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, + 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, + UINT64_MAX, + UINT64_MAX, + AAC_CHANNEL_FRONT); + num_front_channels -= 2; + } + + if (num_side_channels >= 2) { + i += assign_pair(e2c_vec, layout_map, i, + 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, + UINT64_MAX, + UINT64_MAX, + AAC_CHANNEL_SIDE); + num_side_channels -= 2; + } + + while (num_back_channels >= 4) { + i += assign_pair(e2c_vec, layout_map, i, + UINT64_MAX, + UINT64_MAX, + AAC_CHANNEL_BACK); + num_back_channels -= 2; + } + if (num_back_channels >= 2) { + i += assign_pair(e2c_vec, layout_map, i, + 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; } + } - memcpy(ac->tag_che_map, ac->che, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0])); + return layout; +} - avctx->channel_layout = 0; +/** + * 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; +} - avctx->channels = channels; +/** + * 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[0].status != OC_NONE) { + ac->oc[1] = ac->oc[0]; + ac->avctx->channels = ac->oc[1].channels; + ac->avctx->channel_layout = ac->oc[1].channel_layout; + } +} - ac->output_configured = oc_type; +/** + * Configure output channel order based on the current program + * configuration element. + * + * @return Returns error status. 0 - OK, !0 - error + */ +static int output_configure(AACContext *ac, + uint8_t layout_map[MAX_ELEM_ID * 4][3], int tags, + enum OCStatus oc_type, int get_new_frame) +{ + AVCodecContext *avctx = ac->avctx; + int i, channels = 0, ret; + uint64_t layout = 0; + uint8_t id_map[TYPE_END][MAX_ELEM_ID] = {{ 0 }}; + uint8_t type_counts[TYPE_END] = { 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; + } + for (i = 0; i < tags; i++) { + int type = layout_map[i][0]; + int id = layout_map[i][1]; + id_map[type][id] = type_counts[type]++; + } + // 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 iid = id_map[type][id]; + 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, iid, &channels); + if (ret < 0) + return ret; + ac->tag_che_map[type][id] = ac->che[type][iid]; + } + 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; + } + } + + avctx->channel_layout = ac->oc[1].channel_layout = layout; + avctx->channels = ac->oc[1].channels = channels; + ac->oc[1].status = oc_type; + + if (get_new_frame) { + if ((ret = frame_configure_elements(ac->avctx)) < 0) + return ret; + } return 0; } /** - * Decode an array of 4 bit element IDs, optionally interleaved with a stereo/mono switching bit. + * 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 && channel_config < 11) || + channel_config > 12) { + av_log(avctx, AV_LOG_ERROR, + "invalid default channel configuration (%d)\n", + channel_config); + return AVERROR_INVALIDDATA; + } + *tags = tags_per_config[channel_config]; + memcpy(layout_map, aac_channel_layout_map[channel_config - 1], + *tags * sizeof(*layout_map)); + return 0; +} + +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, + OC_TRIAL_FRAME, 1) < 0) + return NULL; + + ac->oc[1].m4ac.chan_config = 2; + ac->oc[1].m4ac.ps = 0; + } + // 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, + OC_TRIAL_FRAME, 1) < 0) + return NULL; + + ac->oc[1].m4ac.chan_config = 1; + if (ac->oc[1].m4ac.sbr) + ac->oc[1].m4ac.ps = -1; + } + /* For indexed channel configurations map the channels solely based + * on position. */ + switch (ac->oc[1].m4ac.chan_config) { + case 12: + 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 11: + if (ac->tags_mapped == 2 && + ac->oc[1].m4ac.chan_config == 11 && + type == TYPE_SCE) { + ac->tags_mapped++; + return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][1]; + } + 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; + default: + // AAC_CHANNEL_OFF has no channel map + return; + } + 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(AVCodecContext *avctx, MPEG4AudioConfig *m4ac, - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID], + uint8_t (*layout_map)[3], GetBitContext *gb) { - int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc, sampling_index; + int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc; + int sampling_index; int comment_len; + int tags; skip_bits(gb, 2); // object_type sampling_index = get_bits(gb, 4); 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"); + 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); @@ -317,14 +695,19 @@ static int decode_pce(AVCodecContext *avctx, MPEG4AudioConfig *m4ac, 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); @@ -332,56 +715,10 @@ static int decode_pce(AVCodecContext *avctx, MPEG4AudioConfig *m4ac, comment_len = get_bits(gb, 8) * 8; if (get_bits_left(gb) < comment_len) { av_log(avctx, AV_LOG_ERROR, overread_err); - return -1; + return AVERROR_INVALIDDATA; } 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(AVCodecContext *avctx, - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID], - 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; - } - - /* 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; } /** @@ -397,13 +734,15 @@ static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx, MPEG4AudioConfig *m4ac, int channel_config) { - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID]; - int extension_flag, ret; + int extension_flag, ret, ep_config, res_flags; + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int tags = 0; if (get_bits1(gb)) { // frameLengthFlag - av_log_missing_feature(avctx, "960/120 MDCT window is", 1); - return -1; + avpriv_request_sample(avctx, "960/120 MDCT window"); + return AVERROR_PATCHWELCOME; } + m4ac->frame_length_short = 0; if (get_bits1(gb)) // dependsOnCoreCoder skip_bits(gb, 14); // coreCoderDelay @@ -413,16 +752,23 @@ static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx, 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(avctx, m4ac, 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(avctx, new_che_pos, channel_config))) + if ((ret = set_default_channel_config(avctx, layout_map, + &tags, channel_config))) return ret; } - if (ac && (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, OC_GLOBAL_HDR, 0))) return ret; if (extension_flag) { @@ -435,14 +781,86 @@ static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx, case AOT_ER_AAC_LTP: case AOT_ER_AAC_SCALABLE: case AOT_ER_AAC_LD: - skip_bits(gb, 3); /* aacSectionDataResilienceFlag - * aacScalefactorDataResilienceFlag - * aacSpectralDataResilienceFlag - */ + res_flags = get_bits(gb, 3); + if (res_flags) { + avpriv_report_missing_feature(avctx, + "AAC data resilience (flags %x)", + res_flags); + return AVERROR_PATCHWELCOME; + } break; } skip_bits1(gb); // extensionFlag3 (TBD in version 3) } + switch (m4ac->object_type) { + case AOT_ER_AAC_LC: + case AOT_ER_AAC_LTP: + case AOT_ER_AAC_SCALABLE: + case AOT_ER_AAC_LD: + ep_config = get_bits(gb, 2); + if (ep_config) { + avpriv_report_missing_feature(avctx, + "epConfig %d", ep_config); + return AVERROR_PATCHWELCOME; + } + } + return 0; +} + +static int decode_eld_specific_config(AACContext *ac, AVCodecContext *avctx, + GetBitContext *gb, + MPEG4AudioConfig *m4ac, + int channel_config) +{ + int ret, ep_config, res_flags; + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int tags = 0; + const int ELDEXT_TERM = 0; + + m4ac->ps = 0; + m4ac->sbr = 0; + + m4ac->frame_length_short = get_bits1(gb); + res_flags = get_bits(gb, 3); + if (res_flags) { + avpriv_report_missing_feature(avctx, + "AAC data resilience (flags %x)", + res_flags); + return AVERROR_PATCHWELCOME; + } + + if (get_bits1(gb)) { // ldSbrPresentFlag + avpriv_report_missing_feature(avctx, + "Low Delay SBR"); + return AVERROR_PATCHWELCOME; + } + + while (get_bits(gb, 4) != ELDEXT_TERM) { + int len = get_bits(gb, 4); + if (len == 15) + len += get_bits(gb, 8); + if (len == 15 + 255) + len += get_bits(gb, 16); + if (get_bits_left(gb) < len * 8 + 4) { + av_log(avctx, AV_LOG_ERROR, overread_err); + return AVERROR_INVALIDDATA; + } + skip_bits_long(gb, 8 * len); + } + + if ((ret = set_default_channel_config(avctx, layout_map, + &tags, channel_config))) + return ret; + + if (ac && (ret = output_configure(ac, layout_map, tags, OC_GLOBAL_HDR, 0))) + return ret; + + ep_config = get_bits(gb, 2); + if (ep_config) { + avpriv_report_missing_feature(avctx, + "epConfig %d", ep_config); + return AVERROR_PATCHWELCOME; + } return 0; } @@ -452,34 +870,45 @@ static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx, * @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 AVCodecContext extradata - * @param data_size size of AVCCodecContext extradata + * @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 or number of consumed bits. <0 - error */ static int decode_audio_specific_config(AACContext *ac, AVCodecContext *avctx, MPEG4AudioConfig *m4ac, - const uint8_t *data, int data_size) + const uint8_t *data, int bit_size, + int sync_extension) { GetBitContext gb; - int i; + int i, ret; - av_dlog(avctx, "extradata size %d\n", avctx->extradata_size); + ff_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"); + ff_dlog(avctx, "%02x ", avctx->extradata[i]); + ff_dlog(avctx, "\n"); - init_get_bits(&gb, data, data_size * 8); + if ((ret = init_get_bits(&gb, data, bit_size)) < 0) + return ret; - if ((i = avpriv_mpeg4audio_get_config(m4ac, data, data_size)) < 0) - return -1; + if ((i = avpriv_mpeg4audio_get_config(m4ac, data, bit_size, + sync_extension)) < 0) + return AVERROR_INVALIDDATA; if (m4ac->sampling_index > 12) { - av_log(avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", m4ac->sampling_index); - return -1; + av_log(avctx, AV_LOG_ERROR, + "invalid sampling rate index %d\n", + m4ac->sampling_index); + return AVERROR_INVALIDDATA; + } + if (m4ac->object_type == AOT_ER_AAC_LD && + (m4ac->sampling_index < 3 || m4ac->sampling_index > 7)) { + av_log(avctx, AV_LOG_ERROR, + "invalid low delay sampling rate index %d\n", + m4ac->sampling_index); + return AVERROR_INVALIDDATA; } - if (m4ac->sbr == 1 && m4ac->ps == -1) - m4ac->ps = 1; skip_bits_long(&gb, i); @@ -487,18 +916,30 @@ static int decode_audio_specific_config(AACContext *ac, case AOT_AAC_MAIN: case AOT_AAC_LC: case AOT_AAC_LTP: - if (decode_ga_specific_config(ac, avctx, &gb, m4ac, m4ac->chan_config)) - return -1; + case AOT_ER_AAC_LC: + case AOT_ER_AAC_LD: + if ((ret = decode_ga_specific_config(ac, avctx, &gb, + m4ac, m4ac->chan_config)) < 0) + return ret; + break; + case AOT_ER_AAC_ELD: + if ((ret = decode_eld_specific_config(ac, avctx, &gb, + m4ac, m4ac->chan_config)) < 0) + return ret; break; default: - av_log(avctx, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n", - m4ac->sbr == 1? "SBR+" : "", m4ac->object_type); - return -1; + avpriv_report_missing_feature(avctx, + "Audio object type %s%d", + m4ac->sbr == 1 ? "SBR+" : "", + m4ac->object_type); + return AVERROR(ENOSYS); } - av_dlog(avctx, "AOT %d chan config %d sampling index %d (%d) SBR %d PS %d\n", + ff_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); + m4ac->sample_rate, m4ac->sbr, + m4ac->ps); return get_bits_count(&gb); } @@ -512,7 +953,8 @@ static int decode_audio_specific_config(AACContext *ac, */ static av_always_inline int lcg_random(int previous_val) { - return previous_val * 1664525 + 1013904223; + union { unsigned u; int s; } v = { previous_val * 1664525u + 1013904223 }; + return v.s; } static av_always_inline void reset_predict_state(PredictorState *ps) @@ -555,34 +997,85 @@ static void reset_predictor_group(PredictorState *ps, int group_num) reset_predict_state(&ps[i]); } -#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]), \ +#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 void aac_static_table_init(void) +{ + AAC_INIT_VLC_STATIC( 0, 304); + AAC_INIT_VLC_STATIC( 1, 270); + AAC_INIT_VLC_STATIC( 2, 550); + AAC_INIT_VLC_STATIC( 3, 300); + AAC_INIT_VLC_STATIC( 4, 328); + AAC_INIT_VLC_STATIC( 5, 294); + AAC_INIT_VLC_STATIC( 6, 306); + AAC_INIT_VLC_STATIC( 7, 268); + AAC_INIT_VLC_STATIC( 8, 510); + AAC_INIT_VLC_STATIC( 9, 366); + AAC_INIT_VLC_STATIC(10, 462); + + ff_aac_sbr_init(); + + 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); + + + // window initialization + ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024); + ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128); + ff_init_ff_sine_windows(10); + ff_init_ff_sine_windows( 9); + ff_init_ff_sine_windows( 7); + + cbrt_tableinit(); +} + +static AVOnce aac_init = AV_ONCE_INIT; + static av_cold int aac_decode_init(AVCodecContext *avctx) { AACContext *ac = avctx->priv_data; - float output_scale_factor; + int ret; + + ret = ff_thread_once(&aac_init, &aac_static_table_init); + if (ret != 0) + return AVERROR_UNKNOWN; ac->avctx = avctx; - ac->m4ac.sample_rate = avctx->sample_rate; + ac->oc[1].m4ac.sample_rate = avctx->sample_rate; + + avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; if (avctx->extradata_size > 0) { - if (decode_audio_specific_config(ac, ac->avctx, &ac->m4ac, - avctx->extradata, - avctx->extradata_size) < 0) - return -1; + if ((ret = decode_audio_specific_config(ac, ac->avctx, &ac->oc[1].m4ac, + avctx->extradata, + avctx->extradata_size * 8, + 1)) < 0) + return ret; } else { int sr, i; - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID]; + uint8_t layout_map[MAX_ELEM_ID*4][3]; + int layout_map_tags; sr = sample_rate_idx(avctx->sample_rate); - ac->m4ac.sampling_index = sr; - ac->m4ac.channels = avctx->channels; - ac->m4ac.sbr = -1; - ac->m4ac.ps = -1; + 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) @@ -590,61 +1083,30 @@ static av_cold int aac_decode_init(AVCodecContext *avctx) if (i == FF_ARRAY_ELEMS(ff_mpeg4audio_channels)) { i = 0; } - ac->m4ac.chan_config = i; + ac->oc[1].m4ac.chan_config = i; - if (ac->m4ac.chan_config) { - int ret = set_default_channel_config(avctx, new_che_pos, ac->m4ac.chan_config); + 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, ac->che_pos, new_che_pos, ac->m4ac.chan_config, OC_GLOBAL_HDR); + output_configure(ac, layout_map, layout_map_tags, + OC_GLOBAL_HDR, 0); else if (avctx->err_recognition & AV_EF_EXPLODE) return AVERROR_INVALIDDATA; } } - 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); - AAC_INIT_VLC_STATIC( 2, 550); - AAC_INIT_VLC_STATIC( 3, 300); - AAC_INIT_VLC_STATIC( 4, 328); - AAC_INIT_VLC_STATIC( 5, 294); - AAC_INIT_VLC_STATIC( 6, 306); - AAC_INIT_VLC_STATIC( 7, 268); - AAC_INIT_VLC_STATIC( 8, 510); - AAC_INIT_VLC_STATIC( 9, 366); - AAC_INIT_VLC_STATIC(10, 462); - - ff_aac_sbr_init(); - - dsputil_init(&ac->dsp, avctx); - ff_fmt_convert_init(&ac->fmt_conv, avctx); + avpriv_float_dsp_init(&ac->fdsp, avctx->flags & AV_CODEC_FLAG_BITEXACT); ac->random_state = 0x1f2e3d4c; - 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, 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); - ff_init_ff_sine_windows(10); - ff_init_ff_sine_windows( 7); - - cbrt_tableinit(); + ff_mdct_init(&ac->mdct, 11, 1, 1.0 / (32768.0 * 1024.0)); + ff_mdct_init(&ac->mdct_ld, 10, 1, 1.0 / (32768.0 * 512.0)); + ff_mdct_init(&ac->mdct_small, 8, 1, 1.0 / (32768.0 * 128.0)); + ff_mdct_init(&ac->mdct_ltp, 11, 0, -2.0 * 32768.0); + ret = ff_imdct15_init(&ac->mdct480, 5); + if (ret < 0) + return ret; return 0; } @@ -663,7 +1125,7 @@ static int skip_data_stream_element(AACContext *ac, GetBitContext *gb) if (get_bits_left(gb) < 8 * count) { av_log(ac->avctx, AV_LOG_ERROR, overread_err); - return -1; + return AVERROR_INVALIDDATA; } skip_bits_long(gb, 8 * count); return 0; @@ -675,12 +1137,14 @@ static int decode_prediction(AACContext *ac, IndividualChannelStream *ics, int sfb; if (get_bits1(gb)) { ics->predictor_reset_group = get_bits(gb, 5); - if (ics->predictor_reset_group == 0 || ics->predictor_reset_group > 30) { - av_log(ac->avctx, AV_LOG_ERROR, "Invalid Predictor Reset Group.\n"); - return -1; + if (ics->predictor_reset_group == 0 || + ics->predictor_reset_group > 30) { + av_log(ac->avctx, AV_LOG_ERROR, + "Invalid Predictor Reset Group.\n"); + return AVERROR_INVALIDDATA; } } - 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; @@ -689,7 +1153,7 @@ static int decode_prediction(AACContext *ac, IndividualChannelStream *ics, /** * Decode Long Term Prediction data; reference: table 4.xx. */ -static void decode_ltp(AACContext *ac, LongTermPrediction *ltp, +static void decode_ltp(LongTermPrediction *ltp, GetBitContext *gb, uint8_t max_sfb) { int sfb; @@ -702,21 +1166,32 @@ static void decode_ltp(AACContext *ac, LongTermPrediction *ltp, /** * 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; + const MPEG4AudioConfig *const m4ac = &ac->oc[1].m4ac; + const int aot = m4ac->object_type; + const int sampling_index = m4ac->sampling_index; + if (aot != AOT_ER_AAC_ELD) { + if (get_bits1(gb)) { + av_log(ac->avctx, AV_LOG_ERROR, "Reserved bit set.\n"); + if (ac->avctx->err_recognition & AV_EF_BITSTREAM) + return AVERROR_INVALIDDATA; + } + ics->window_sequence[1] = ics->window_sequence[0]; + ics->window_sequence[0] = get_bits(gb, 2); + if (aot == AOT_ER_AAC_LD && + ics->window_sequence[0] != ONLY_LONG_SEQUENCE) { + av_log(ac->avctx, AV_LOG_ERROR, + "AAC LD is only defined for ONLY_LONG_SEQUENCE but " + "window sequence %d found.\n", ics->window_sequence[0]); + ics->window_sequence[0] = ONLY_LONG_SEQUENCE; + return AVERROR_INVALIDDATA; + } + ics->use_kb_window[1] = ics->use_kb_window[0]; + ics->use_kb_window[0] = get_bits1(gb); } - ics->window_sequence[1] = ics->window_sequence[0]; - ics->window_sequence[0] = get_bits(gb, 2); - ics->use_kb_window[1] = ics->use_kb_window[0]; - ics->use_kb_window[0] = get_bits1(gb); ics->num_window_groups = 1; ics->group_len[0] = 1; if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) { @@ -731,41 +1206,61 @@ 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[sampling_index]; + ics->num_swb = ff_aac_num_swb_128[sampling_index]; + ics->tns_max_bands = ff_tns_max_bands_128[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->predictor_present = get_bits1(gb); - ics->predictor_reset_group = 0; + ics->max_sfb = get_bits(gb, 6); + ics->num_windows = 1; + if (aot == AOT_ER_AAC_LD || aot == AOT_ER_AAC_ELD) { + if (m4ac->frame_length_short) { + ics->swb_offset = ff_swb_offset_480[sampling_index]; + ics->num_swb = ff_aac_num_swb_480[sampling_index]; + ics->tns_max_bands = ff_tns_max_bands_480[sampling_index]; + } else { + ics->swb_offset = ff_swb_offset_512[sampling_index]; + ics->num_swb = ff_aac_num_swb_512[sampling_index]; + ics->tns_max_bands = ff_tns_max_bands_512[sampling_index]; + } + if (!ics->num_swb || !ics->swb_offset) + return AVERROR_BUG; + } else { + ics->swb_offset = ff_swb_offset_1024[sampling_index]; + ics->num_swb = ff_aac_num_swb_1024[sampling_index]; + ics->tns_max_bands = ff_tns_max_bands_1024[sampling_index]; + } + if (aot != AOT_ER_AAC_ELD) { + ics->predictor_present = get_bits1(gb); + ics->predictor_reset_group = 0; + } if (ics->predictor_present) { - if (ac->m4ac.object_type == AOT_AAC_MAIN) { + if (aot == 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) { - av_log(ac->avctx, AV_LOG_ERROR, "Prediction is not allowed in AAC-LC.\n"); - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + } else if (aot == AOT_AAC_LC || + aot == AOT_ER_AAC_LC) { + av_log(ac->avctx, AV_LOG_ERROR, + "Prediction is not allowed in AAC-LC.\n"); + return AVERROR_INVALIDDATA; } else { + if (aot == AOT_ER_AAC_LD) { + avpriv_report_missing_feature(ac->avctx, "LTP in ER AAC LD"); + return AVERROR_PATCHWELCOME; + } if ((ics->ltp.present = get_bits(gb, 1))) - decode_ltp(ac, &ics->ltp, gb, ics->max_sfb); + decode_ltp(&ics->ltp, gb, ics->max_sfb); } } } if (ics->max_sfb > ics->num_swb) { av_log(ac->avctx, AV_LOG_ERROR, - "Number of scalefactor bands in group (%d) exceeds limit (%d).\n", + "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; @@ -793,21 +1288,22 @@ static int decode_band_types(AACContext *ac, enum BandType band_type[120], int sect_band_type = get_bits(gb, 4); if (sect_band_type == 12) { av_log(ac->avctx, AV_LOG_ERROR, "invalid band type\n"); - return -1; + return AVERROR_INVALIDDATA; } - 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 AVERROR_INVALIDDATA; + } + 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 AVERROR_INVALIDDATA; + } + } 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; @@ -837,22 +1333,22 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, 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]; if (band_type[idx] == ZERO_BT) { for (; i < run_end; i++, idx++) - sf[idx] = 0.; - } else if ((band_type[idx] == INTENSITY_BT) || (band_type[idx] == INTENSITY_BT2)) { + sf[idx] = 0.0; + } 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; 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); + avpriv_request_sample(ac->avctx, + "If you heard an audible artifact, there may be a bug in the decoder. " + "Clipped intensity stereo position (%d -> %d)", + offset[2], clipped_offset); } sf[idx] = ff_aac_pow2sf_tab[-clipped_offset + POW_SF2_ZERO]; } @@ -864,10 +1360,10 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60; 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); + avpriv_request_sample(ac->avctx, + "If you heard an audible artifact, there may be a bug in the decoder. " + "Clipped noise gain (%d -> %d)", + offset[1], clipped_offset); } sf[idx] = -ff_aac_pow2sf_tab[clipped_offset + POW_SF2_ZERO]; } @@ -876,8 +1372,8 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb, 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]); - return -1; + "Scalefactor (%d) out of range.\n", offset[0]); + return AVERROR_INVALIDDATA; } sf[idx] = -ff_aac_pow2sf_tab[offset[0] - 100 + POW_SF2_ZERO]; } @@ -922,7 +1418,7 @@ 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); @@ -932,10 +1428,11 @@ 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->avctx, 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; + return AVERROR_INVALIDDATA; } if (tns->order[w][filt]) { tns->direction[w][filt] = get_bits1(gb); @@ -963,11 +1460,12 @@ static void decode_mid_side_stereo(ChannelElement *cpe, GetBitContext *gb, int ms_present) { int idx; + int max_idx = cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb; if (ms_present == 1) { - for (idx = 0; idx < cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb; idx++) + for (idx = 0; idx < max_idx; idx++) cpe->ms_mask[idx] = get_bits1(gb); } else if (ms_present == 2) { - memset(cpe->ms_mask, 1, cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb * sizeof(cpe->ms_mask[0])); + memset(cpe->ms_mask, 1, max_idx * sizeof(cpe->ms_mask[0])); } } @@ -999,7 +1497,7 @@ 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; @@ -1017,8 +1515,8 @@ 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 & 1U<<31); *dst++ = v[idx & 3] * t.f; @@ -1031,7 +1529,7 @@ static inline float *VMUL4S(float *dst, const float *v, unsigned idx, t.i = s.i ^ (sign & 1U<<31); *dst++ = v[idx>>4 & 3] * t.f; - sign <<= nz & 1; nz >>= 1; + sign <<= nz & 1; t.i = s.i ^ (sign & 1U<<31); *dst++ = v[idx>>6 & 3] * t.f; @@ -1063,7 +1561,8 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], float *coef_base = coef; for (g = 0; g < ics->num_windows; g++) - memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float) * (c - offsets[ics->max_sfb])); + memset(coef + g * 128 + offsets[ics->max_sfb], 0, + sizeof(float) * (c - offsets[ics->max_sfb])); for (g = 0; g < ics->num_window_groups; g++) { unsigned g_len = ics->group_len[g]; @@ -1088,9 +1587,9 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], cfo[k] = ac->random_state; } - band_energy = ac->dsp.scalarproduct_float(cfo, cfo, off_len); + band_energy = ac->fdsp.scalarproduct_float(cfo, cfo, off_len); scale = sf[idx] / sqrtf(band_energy); - ac->dsp.vector_fmul_scalar(cfo, cfo, scale, off_len); + ac->fdsp.vector_fmul_scalar(cfo, cfo, scale, off_len); } } else { const float *vq = ff_aac_codebook_vector_vals[cbt_m1]; @@ -1218,7 +1717,7 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], if (b > 8) { av_log(ac->avctx, AV_LOG_ERROR, "error in spectral data, ESC overflow\n"); - return -1; + return AVERROR_INVALIDDATA; } SKIP_BITS(re, gb, b + 1); @@ -1236,7 +1735,7 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], } } while (len -= 2); - ac->dsp.vector_fmul_scalar(cfo, cfo, sf[idx], off_len); + ac->fdsp.vector_fmul_scalar(cfo, cfo, sf[idx], off_len); } } @@ -1267,7 +1766,7 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024], 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; @@ -1275,7 +1774,7 @@ 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; @@ -1283,7 +1782,7 @@ 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; @@ -1333,14 +1832,20 @@ 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 (k = sce->ics.swb_offset[sfb]; k < sce->ics.swb_offset[sfb + 1]; k++) { + 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(&sce->predictor_state[k], &sce->coeffs[k], - sce->ics.predictor_present && sce->ics.prediction_used[sfb]); + sce->ics.predictor_present && + sce->ics.prediction_used[sfb]); } } if (sce->ics.predictor_reset_group) - reset_predictor_group(sce->predictor_state, sce->ics.predictor_reset_group); + reset_predictor_group(sce->predictor_state, + sce->ics.predictor_reset_group); } else reset_all_predictors(sce->predictor_state); } @@ -1360,7 +1865,14 @@ static int decode_ics(AACContext *ac, SingleChannelElement *sce, TemporalNoiseShaping *tns = &sce->tns; IndividualChannelStream *ics = &sce->ics; float *out = sce->coeffs; - int global_gain, pulse_present = 0; + int global_gain, eld_syntax, er_syntax, pulse_present = 0; + int ret; + + eld_syntax = ac->oc[1].m4ac.object_type == AOT_ER_AAC_ELD; + er_syntax = ac->oc[1].m4ac.object_type == AOT_ER_AAC_LC || + ac->oc[1].m4ac.object_type == AOT_ER_AAC_LTP || + ac->oc[1].m4ac.object_type == AOT_ER_AAC_LD || + ac->oc[1].m4ac.object_type == AOT_ER_AAC_ELD; /* This assignment is to silence a GCC warning about the variable being used * uninitialized when in fact it always is. @@ -1370,39 +1882,51 @@ 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) - return -1; - if (decode_scalefactors(ac, sce->sf, gb, global_gain, ics, sce->band_type, sce->band_type_run_end) < 0) - return -1; + if ((ret = decode_band_types(ac, sce->band_type, + sce->band_type_run_end, gb, ics)) < 0) + return ret; + if ((ret = decode_scalefactors(ac, sce->sf, gb, global_gain, ics, + sce->band_type, sce->band_type_run_end)) < 0) + return ret; pulse_present = 0; if (!scale_flag) { - if ((pulse_present = get_bits1(gb))) { + if (!eld_syntax && (pulse_present = get_bits1(gb))) { if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) { - av_log(ac->avctx, AV_LOG_ERROR, "Pulse tool not allowed in eight short sequence.\n"); - return -1; + av_log(ac->avctx, AV_LOG_ERROR, + "Pulse tool not allowed in eight short sequence.\n"); + return AVERROR_INVALIDDATA; } if (decode_pulses(&pulse, gb, ics->swb_offset, ics->num_swb)) { - av_log(ac->avctx, AV_LOG_ERROR, "Pulse data corrupt or invalid.\n"); - return -1; + av_log(ac->avctx, AV_LOG_ERROR, + "Pulse data corrupt or invalid.\n"); + return AVERROR_INVALIDDATA; } } - if ((tns->present = get_bits1(gb)) && decode_tns(ac, tns, gb, ics)) - return -1; - if (get_bits1(gb)) { - av_log_missing_feature(ac->avctx, "SSR", 1); - return -1; + tns->present = get_bits1(gb); + if (tns->present && !er_syntax) + if (decode_tns(ac, tns, gb, ics) < 0) + return AVERROR_INVALIDDATA; + if (!eld_syntax && get_bits1(gb)) { + avpriv_request_sample(ac->avctx, "SSR"); + return AVERROR_PATCHWELCOME; } + // I see no textual basis in the spec for this occurring after SSR gain + // control, but this is what both reference and real implementations do + if (tns->present && er_syntax) + if (decode_tns(ac, tns, gb, ics) < 0) + return AVERROR_INVALIDDATA; } - if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0) - return -1; + if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, + &pulse, ics, sce->band_type) < 0) + return AVERROR_INVALIDDATA; - 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; @@ -1421,11 +1945,12 @@ static void apply_mid_side_stereo(AACContext *ac, ChannelElement *cpe) for (g = 0; g < ics->num_window_groups; g++) { for (i = 0; i < ics->max_sfb; i++, idx++) { if (cpe->ms_mask[idx] && - cpe->ch[0].band_type[idx] < NOISE_BT && cpe->ch[1].band_type[idx] < NOISE_BT) { + cpe->ch[0].band_type[idx] < NOISE_BT && + cpe->ch[1].band_type[idx] < NOISE_BT) { for (group = 0; group < ics->group_len[g]; group++) { - ac->dsp.butterflies_float(ch0 + group * 128 + offsets[i], - ch1 + group * 128 + offsets[i], - offsets[i+1] - offsets[i]); + ac->fdsp.butterflies_float(ch0 + group * 128 + offsets[i], + ch1 + group * 128 + offsets[i], + offsets[i+1] - offsets[i]); } } } @@ -1441,7 +1966,8 @@ 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(AACContext *ac, 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]; @@ -1452,7 +1978,8 @@ static void apply_intensity_stereo(AACContext *ac, ChannelElement *cpe, int ms_p float scale; for (g = 0; g < ics->num_window_groups; g++) { for (i = 0; i < ics->max_sfb;) { - if (sce1->band_type[idx] == INTENSITY_BT || sce1->band_type[idx] == INTENSITY_BT2) { + if (sce1->band_type[idx] == INTENSITY_BT || + sce1->band_type[idx] == INTENSITY_BT2) { const int bt_run_end = sce1->band_type_run_end[idx]; for (; i < bt_run_end; i++, idx++) { c = -1 + 2 * (sce1->band_type[idx] - 14); @@ -1460,10 +1987,10 @@ static void apply_intensity_stereo(AACContext *ac, ChannelElement *cpe, int ms_p c *= 1 - 2 * cpe->ms_mask[idx]; scale = c * sce1->sf[idx]; for (group = 0; group < ics->group_len[g]; group++) - ac->dsp.vector_fmul_scalar(coef1 + group * 128 + offsets[i], - coef0 + group * 128 + offsets[i], - scale, - offsets[i + 1] - offsets[i]); + ac->fdsp.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]; @@ -1484,21 +2011,23 @@ static void apply_intensity_stereo(AACContext *ac, ChannelElement *cpe, int ms_p static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe) { int i, ret, common_window, ms_present = 0; + int eld_syntax = ac->oc[1].m4ac.object_type == AOT_ER_AAC_ELD; - common_window = get_bits1(gb); + common_window = eld_syntax || 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->m4ac.object_type != AOT_AAC_MAIN)) + 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); + decode_ltp(&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"); - return -1; + return AVERROR_INVALIDDATA; } else if (ms_present) decode_mid_side_stereo(cpe, gb, ms_present); } @@ -1510,7 +2039,7 @@ 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]); } @@ -1566,7 +2095,7 @@ static int decode_cce(AACContext *ac, GetBitContext *gb, ChannelElement *che) int idx = 0; int cge = 1; int gain = 0; - float gain_cache = 1.; + float gain_cache = 1.0; if (c) { cge = coup->coupling_point == AFTER_IMDCT ? 1 : get_bits1(gb); gain = cge ? get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60: 0; @@ -1621,12 +2150,10 @@ static int decode_drc_channel_exclusions(DynamicRangeControl *che_drc, /** * Decode dynamic range information; reference: table 4.52. * - * @param cnt length of TYPE_FIL syntactic element in bytes - * * @return Returns number of bytes consumed. */ static int decode_dynamic_range(DynamicRangeControl *che_drc, - GetBitContext *gb, int cnt) + GetBitContext *gb) { int n = 1; int drc_num_bands = 1; @@ -1691,25 +2218,28 @@ 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; + ac->avctx->profile = FF_PROFILE_AAC_HE_V2; + output_configure(ac, ac->oc[1].layout_map, ac->oc[1].layout_map_tags, + ac->oc[1].status, 1); } else { - ac->m4ac.sbr = 1; + ac->oc[1].m4ac.sbr = 1; + ac->avctx->profile = FF_PROFILE_AAC_HE; } res = ff_decode_sbr_extension(ac, &che->sbr, gb, crc_flag, cnt, elem_type); break; case EXT_DYNAMIC_RANGE: - res = decode_dynamic_range(&ac->che_drc, gb, cnt); + res = decode_dynamic_range(&ac->che_drc, gb); break; case EXT_FILL: case EXT_FILL_DATA: @@ -1734,7 +2264,7 @@ 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]; + float tmp[TNS_MAX_ORDER + 1]; for (w = 0; w < ics->num_windows; w++) { bottom = ics->num_swb; @@ -1792,15 +2322,15 @@ static void windowing_and_mdct_ltp(AACContext *ac, float *out, 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); + ac->fdsp.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); + ac->fdsp.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); + ac->fdsp.vector_fmul_reverse(in + 1024, in + 1024, lwindow, 1024); } else { - ac->dsp.vector_fmul_reverse(in + 1024 + 448, in + 1024 + 448, swindow, 128); + ac->fdsp.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); @@ -1853,17 +2383,17 @@ static void update_ltp(AACContext *ac, SingleChannelElement *sce) 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); + ac->fdsp.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); + ac->fdsp.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); + ac->fdsp.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]; } @@ -1904,36 +2434,121 @@ static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce) */ 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, 512); + ac->fdsp.vector_fmul_window( out, saved, buf, lwindow_prev, 512); } else { - memcpy( out, saved, 448 * sizeof(float)); + 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, 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)); + ac->fdsp.vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, 64); + ac->fdsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, 64); + ac->fdsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, 64); + ac->fdsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, 64); + ac->fdsp.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, 64); - memcpy( out + 576, buf + 64, 448 * sizeof(float)); + ac->fdsp.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) { - 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)); + memcpy( saved, temp + 64, 64 * sizeof(float)); + ac->fdsp.vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 64); + ac->fdsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 64); + ac->fdsp.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)); - memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(float)); + memcpy( saved, buf + 512, 448 * sizeof(float)); + memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(float)); } else { // LONG_STOP or ONLY_LONG - memcpy( saved, buf + 512, 512 * sizeof(float)); + memcpy( saved, buf + 512, 512 * sizeof(float)); + } +} + +static void imdct_and_windowing_ld(AACContext *ac, SingleChannelElement *sce) +{ + IndividualChannelStream *ics = &sce->ics; + float *in = sce->coeffs; + float *out = sce->ret; + float *saved = sce->saved; + float *buf = ac->buf_mdct; + + // imdct + ac->mdct.imdct_half(&ac->mdct_ld, buf, in); + + // window overlapping + if (ics->use_kb_window[1]) { + // AAC LD uses a low overlap sine window instead of a KBD window + memcpy(out, saved, 192 * sizeof(float)); + ac->fdsp.vector_fmul_window(out + 192, saved + 192, buf, ff_sine_128, 64); + memcpy( out + 320, buf + 64, 192 * sizeof(float)); + } else { + ac->fdsp.vector_fmul_window(out, saved, buf, ff_sine_512, 256); } + + // buffer update + memcpy(saved, buf + 256, 256 * sizeof(float)); +} + +static void imdct_and_windowing_eld(AACContext *ac, SingleChannelElement *sce) +{ + float *in = sce->coeffs; + float *out = sce->ret; + float *saved = sce->saved; + float *buf = ac->buf_mdct; + int i; + const int n = ac->oc[1].m4ac.frame_length_short ? 480 : 512; + const int n2 = n >> 1; + const int n4 = n >> 2; + const float *const window = n == 480 ? ff_aac_eld_window_480 : + ff_aac_eld_window_512; + + // Inverse transform, mapped to the conventional IMDCT by + // Chivukula, R.K.; Reznik, Y.A.; Devarajan, V., + // "Efficient algorithms for MPEG-4 AAC-ELD, AAC-LD and AAC-LC filterbanks," + // Audio, Language and Image Processing, 2008. ICALIP 2008. International Conference on + // URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4590245&isnumber=4589950 + for (i = 0; i < n2; i+=2) { + float temp; + temp = in[i ]; in[i ] = -in[n - 1 - i]; in[n - 1 - i] = temp; + temp = -in[i + 1]; in[i + 1] = in[n - 2 - i]; in[n - 2 - i] = temp; + } + if (n == 480) + ac->mdct480->imdct_half(ac->mdct480, buf, in, 1, -1.f/(16*1024*960)); + else + ac->mdct.imdct_half(&ac->mdct_ld, buf, in); + for (i = 0; i < n; i+=2) { + buf[i] = -buf[i]; + } + // Like with the regular IMDCT at this point we still have the middle half + // of a transform but with even symmetry on the left and odd symmetry on + // the right + + // window overlapping + // The spec says to use samples [0..511] but the reference decoder uses + // samples [128..639]. + for (i = n4; i < n2; i ++) { + out[i - n4] = buf[n2 - 1 - i] * window[i - n4] + + saved[ i + n2] * window[i + n - n4] + + -saved[ n + n2 - 1 - i] * window[i + 2*n - n4] + + -saved[2*n + n2 + i] * window[i + 3*n - n4]; + } + for (i = 0; i < n2; i ++) { + out[n4 + i] = buf[i] * window[i + n2 - n4] + + -saved[ n - 1 - i] * window[i + n2 + n - n4] + + -saved[ n + i] * window[i + n2 + 2*n - n4] + + saved[2*n + n - 1 - i] * window[i + n2 + 3*n - n4]; + } + for (i = 0; i < n4; i ++) { + out[n2 + n4 + i] = buf[ i + n2] * window[i + n - n4] + + -saved[ n2 - 1 - i] * window[i + 2*n - n4] + + -saved[ n + n2 + i] * window[i + 3*n - n4]; + } + + // buffer update + memmove(saved + n, saved, 2 * n * sizeof(float)); + memcpy( saved, buf, n * sizeof(float)); } /** @@ -1950,7 +2565,7 @@ 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; @@ -1961,7 +2576,7 @@ static void apply_dependent_coupling(AACContext *ac, const float gain = cce->coup.gain[index][idx]; for (group = 0; group < ics->group_len[g]; group++) { for (k = offsets[i]; k < offsets[i + 1]; k++) { - // XXX dsputil-ize + // FIXME: SIMDify dest[group * 128 + k] += gain * src[group * 128 + k]; } } @@ -1985,7 +2600,7 @@ static void apply_independent_coupling(AACContext *ac, const float gain = cce->coup.gain[index][0]; 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]; @@ -2032,13 +2647,24 @@ static void apply_channel_coupling(AACContext *ac, ChannelElement *cc, static void spectral_to_sample(AACContext *ac) { int i, type; + void (*imdct_and_window)(AACContext *ac, SingleChannelElement *sce); + switch (ac->oc[1].m4ac.object_type) { + case AOT_ER_AAC_LD: + imdct_and_window = imdct_and_windowing_ld; + break; + case AOT_ER_AAC_ELD: + imdct_and_window = imdct_and_windowing_eld; + break; + default: + imdct_and_window = imdct_and_windowing; + } 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->m4ac.object_type == AOT_AAC_LTP) { + 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]); @@ -2053,15 +2679,15 @@ 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]); - if (ac->m4ac.object_type == AOT_AAC_LTP) + imdct_and_window(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]); - if (ac->m4ac.object_type == AOT_AAC_LTP) + imdct_and_window(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); } } @@ -2076,72 +2702,160 @@ 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, ret; - size = avpriv_aac_parse_header(gb, &hdr_info); + size = ff_adts_header_parse(gb, &hdr_info); if (size > 0) { + if (hdr_info.num_aac_frames != 1) { + avpriv_report_missing_feature(ac->avctx, + "More than one AAC RDB per ADTS frame"); + return AVERROR_PATCHWELCOME; + } + push_output_configuration(ac); if (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->avctx, new_che_pos, hdr_info.chan_config)) - return -7; - if (output_configure(ac, ac->che_pos, new_che_pos, hdr_info.chan_config, - FFMAX(ac->output_configured, OC_TRIAL_FRAME))) - return -7; - } else if (ac->output_configured != OC_LOCKED) { - ac->m4ac.chan_config = 0; - 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); + ac->oc[1].m4ac.chan_config = hdr_info.chan_config; + if ((ret = set_default_channel_config(ac->avctx, + layout_map, + &layout_map_tags, + hdr_info.chan_config)) < 0) + return ret; + if ((ret = output_configure(ac, layout_map, layout_map_tags, + FFMAX(ac->oc[1].status, + OC_TRIAL_FRAME), 0)) < 0) + return ret; } 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; + ac->oc[1].m4ac.frame_length_short = 0; + 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_er_frame(AVCodecContext *avctx, void *data, + int *got_frame_ptr, GetBitContext *gb) +{ + AACContext *ac = avctx->priv_data; + const MPEG4AudioConfig *const m4ac = &ac->oc[1].m4ac; + ChannelElement *che; + int err, i; + int samples = m4ac->frame_length_short ? 960 : 1024; + int chan_config = m4ac->chan_config; + int aot = m4ac->object_type; + + if (aot == AOT_ER_AAC_LD || aot == AOT_ER_AAC_ELD) + samples >>= 1; + + ac->frame = data; + + if ((err = frame_configure_elements(avctx)) < 0) + return err; + + // The FF_PROFILE_AAC_* defines are all object_type - 1 + // This may lead to an undefined profile being signaled + ac->avctx->profile = aot - 1; + + ac->tags_mapped = 0; + + if (chan_config < 0 || (chan_config >= 8 && chan_config < 11) || chan_config >= 13) { + avpriv_request_sample(avctx, "Unknown ER channel configuration %d", + chan_config); + return AVERROR_INVALIDDATA; + } + for (i = 0; i < tags_per_config[chan_config]; i++) { + const int elem_type = aac_channel_layout_map[chan_config-1][i][0]; + const int elem_id = aac_channel_layout_map[chan_config-1][i][1]; + 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 AVERROR_INVALIDDATA; + } + if (aot != AOT_ER_AAC_ELD) + skip_bits(gb, 4); + switch (elem_type) { + case TYPE_SCE: + err = decode_ics(ac, &che->ch[0], gb, 0, 0); + break; + case TYPE_CPE: + err = decode_cpe(ac, gb, che); + break; + case TYPE_LFE: + err = decode_ics(ac, &che->ch[0], gb, 0, 0); + break; + } + if (err < 0) + return err; + } + + spectral_to_sample(ac); + + ac->frame->nb_samples = samples; + ac->frame->sample_rate = avctx->sample_rate; + *got_frame_ptr = 1; + + skip_bits_long(gb, get_bits_left(gb)); + return 0; +} + static int aac_decode_frame_int(AVCodecContext *avctx, void *data, - int *data_size, GetBitContext *gb) + int *got_frame_ptr, GetBitContext *gb) { AACContext *ac = avctx->priv_data; ChannelElement *che = NULL, *che_prev = NULL; enum RawDataBlockType elem_type, elem_type_prev = TYPE_END; - int err, elem_id, data_size_tmp; - int samples = 0, multiplier, audio_found = 0; + int err, elem_id; + int samples = 0, multiplier, audio_found = 0, pce_found = 0; + + ac->frame = data; if (show_bits(gb, 12) == 0xfff) { - if (parse_adts_frame_header(ac, gb) < 0) { + if ((err = parse_adts_frame_header(ac, gb)) < 0) { av_log(avctx, AV_LOG_ERROR, "Error decoding AAC frame header.\n"); - return -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 = AVERROR_INVALIDDATA; + goto fail; } } + if (avctx->channels) + if ((err = frame_configure_elements(avctx)) < 0) + goto fail; + + // The FF_PROFILE_AAC_* defines are all object_type - 1 + // This may lead to an undefined profile being signaled + ac->avctx->profile = ac->oc[1].m4ac.object_type - 1; + ac->tags_mapped = 0; // parse while ((elem_type = get_bits(gb, 3)) != TYPE_END) { elem_id = get_bits(gb, 4); + if (!avctx->channels && elem_type != TYPE_PCE) { + err = AVERROR_INVALIDDATA; + goto fail; + } + 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 = AVERROR_INVALIDDATA; + goto fail; } samples = 1024; } @@ -2172,15 +2886,22 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data, 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(avctx, &ac->m4ac, 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, OC_TRIAL_PCE, 1); + pce_found = 1; + } break; } @@ -2189,7 +2910,8 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data, 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 = AVERROR_INVALIDDATA; + goto fail; } while (elem_id > 0) elem_id -= decode_extension_payload(ac, gb, elem_id, che_prev, elem_type_prev); @@ -2197,7 +2919,7 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data, break; default: - err = -1; /* should not happen, but keeps compiler happy */ + err = AVERROR_BUG; /* should not happen, but keeps compiler happy */ break; } @@ -2205,61 +2927,89 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data, elem_type_prev = elem_type; if (err) - return err; + goto fail; if (get_bits_left(gb) < 3) { av_log(avctx, AV_LOG_ERROR, overread_err); - return -1; + err = AVERROR_INVALIDDATA; + goto fail; } } + if (!avctx->channels) { + *got_frame_ptr = 0; + return 0; + } + 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; - avctx->frame_size = samples; - } - data_size_tmp = samples * avctx->channels * - av_get_bytes_per_sample(avctx->sample_fmt); - 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; + 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 = data_size_tmp; if (samples) { - if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) - ac->fmt_conv.float_interleave(data, (const float **)ac->output_data, - samples, avctx->channels); - else - ac->fmt_conv.float_to_int16_interleave(data, (const float **)ac->output_data, - samples, avctx->channels); + ac->frame->nb_samples = samples; + ac->frame->sample_rate = avctx->sample_rate; } - - if (ac->output_configured && audio_found) - ac->output_configured = OC_LOCKED; + *got_frame_ptr = !!samples; return 0; +fail: + pop_output_configuration(ac); + return err; } static int aac_decode_frame(AVCodecContext *avctx, void *data, - int *data_size, AVPacket *avpkt) + 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 + + AV_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; + } + } - init_get_bits(&gb, buf, buf_size * 8); + if ((err = init_get_bits(&gb, buf, buf_size * 8)) < 0) + return err; - if ((err = aac_decode_frame_int(avctx, data, data_size, &gb)) < 0) + switch (ac->oc[1].m4ac.object_type) { + case AOT_ER_AAC_LC: + case AOT_ER_AAC_LTP: + case AOT_ER_AAC_LD: + case AOT_ER_AAC_ELD: + err = aac_decode_er_frame(avctx, data, got_frame_ptr, &gb); + break; + default: + err = aac_decode_frame_int(avctx, data, got_frame_ptr, &gb); + } + if (err < 0) return err; buf_consumed = (get_bits_count(&gb) + 7) >> 3; @@ -2285,7 +3035,9 @@ 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_ld); ff_mdct_end(&ac->mdct_ltp); + ff_imdct15_uninit(&ac->mdct480); return 0; } @@ -2293,13 +3045,13 @@ static av_cold int aac_decode_close(AVCodecContext *avctx) #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 + AACContext aac_ctx; ///< containing AACContext + int initialized; ///< initialized 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 + 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) @@ -2310,41 +3062,56 @@ static inline uint32_t latm_get_value(GetBitContext *b) } static int latm_decode_audio_specific_config(struct LATMContext *latmctx, - GetBitContext *gb) + GetBitContext *gb, int asclen) { - AVCodecContext *avctx = latmctx->aac_ctx.avctx; - MPEG4AudioConfig m4ac; - int config_start_bit = get_bits_count(gb); - int bits_consumed, esize; + 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); + avpriv_request_sample(latmctx->aac_ctx.avctx, + "Non-byte-aligned audio-specific config"); + return AVERROR_PATCHWELCOME; + } + if (asclen <= 0) return AVERROR_INVALIDDATA; - } else { - bits_consumed = - decode_audio_specific_config(NULL, avctx, &m4ac, + bits_consumed = decode_audio_specific_config(NULL, avctx, &m4ac, gb->buffer + (config_start_bit / 8), - get_bits_left(gb) / 8); + asclen, sync_extension); - if (bits_consumed < 0) - return AVERROR_INVALIDDATA; + if (bits_consumed < 0) + return AVERROR_INVALIDDATA; + + if (!latmctx->initialized || + 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) { + if (avctx->extradata_size < esize) { av_free(avctx->extradata); - avctx->extradata = av_malloc(esize + FF_INPUT_BUFFER_PADDING_SIZE); + avctx->extradata = av_malloc(esize + AV_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); + memset(avctx->extradata+esize, 0, AV_INPUT_BUFFER_PADDING_SIZE); } + skip_bits_long(gb, bits_consumed); return bits_consumed; } @@ -2367,8 +3134,7 @@ static int read_stream_mux_config(struct LATMContext *latmctx, 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); + avpriv_request_sample(latmctx->aac_ctx.avctx, "Multiple programs"); return AVERROR_PATCHWELCOME; } @@ -2376,18 +3142,17 @@ static int read_stream_mux_config(struct LATMContext *latmctx, // 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); + avpriv_request_sample(latmctx->aac_ctx.avctx, "Multiple layers"); 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) + 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)) < 0) + if ((ret = latm_decode_audio_specific_config(latmctx, gb, ascLen)) < 0) return ret; ascLen -= ret; skip_bits_long(gb, ascLen); @@ -2463,7 +3228,7 @@ static int read_audio_mux_element(struct LATMContext *latmctx, } else if (!latmctx->aac_ctx.avctx->extradata) { av_log(latmctx->aac_ctx.avctx, AV_LOG_DEBUG, "no decoder config found\n"); - return AVERROR(EAGAIN); + return 1; } if (latmctx->audio_mux_version_A == 0) { int mux_slot_length_bytes = read_payload_length_info(latmctx, gb); @@ -2481,14 +3246,15 @@ static int read_audio_mux_element(struct LATMContext *latmctx, } -static int latm_decode_frame(AVCodecContext *avctx, void *out, int *out_size, - AVPacket *avpkt) +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); + if ((err = init_get_bits(&gb, avpkt->data, avpkt->size * 8)) < 0) + return err; // check for LOAS sync word if (get_bits(&gb, 11) != LOAS_SYNC_WORD) @@ -2499,18 +3265,21 @@ static int latm_decode_frame(AVCodecContext *avctx, void *out, int *out_size, if (muxlength > avpkt->size) return AVERROR_INVALIDDATA; - if ((err = read_audio_mux_element(latmctx, &gb)) < 0) - return err; + if ((err = read_audio_mux_element(latmctx, &gb))) + return (err < 0) ? err : avpkt->size; if (!latmctx->initialized) { if (!avctx->extradata) { - *out_size = 0; + *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.m4ac, - avctx->extradata, avctx->extradata_size)) < 0) + &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; } } @@ -2522,13 +3291,23 @@ static int latm_decode_frame(AVCodecContext *avctx, void *out, int *out_size, return AVERROR_INVALIDDATA; } - if ((err = aac_decode_frame_int(avctx, out, out_size, &gb)) < 0) + switch (latmctx->aac_ctx.oc[1].m4ac.object_type) { + case AOT_ER_AAC_LC: + case AOT_ER_AAC_LTP: + case AOT_ER_AAC_LD: + case AOT_ER_AAC_ELD: + err = aac_decode_er_frame(avctx, out, got_frame_ptr, &gb); + break; + default: + err = aac_decode_frame_int(avctx, out, got_frame_ptr, &gb); + } + if (err < 0) return err; return muxlength; } -av_cold static int latm_decode_init(AVCodecContext *avctx) +static av_cold int latm_decode_init(AVCodecContext *avctx) { struct LATMContext *latmctx = avctx->priv_data; int ret = aac_decode_init(avctx); @@ -2541,18 +3320,19 @@ av_cold static int latm_decode_init(AVCodecContext *avctx) 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 + .name = "aac", + .long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"), + .type = AVMEDIA_TYPE_AUDIO, + .id = AV_CODEC_ID_AAC, + .priv_data_size = sizeof(AACContext), + .init = aac_decode_init, + .close = aac_decode_close, + .decode = aac_decode_frame, + .sample_fmts = (const enum AVSampleFormat[]) { + AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE }, - .capabilities = CODEC_CAP_CHANNEL_CONF, + .capabilities = AV_CODEC_CAP_CHANNEL_CONF | AV_CODEC_CAP_DR1, + .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE, .channel_layouts = aac_channel_layout, }; @@ -2562,17 +3342,18 @@ AVCodec ff_aac_decoder = { 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 + .name = "aac_latm", + .long_name = NULL_IF_CONFIG_SMALL("AAC LATM (Advanced Audio Coding LATM syntax)"), + .type = AVMEDIA_TYPE_AUDIO, + .id = AV_CODEC_ID_AAC_LATM, + .priv_data_size = sizeof(struct LATMContext), + .init = latm_decode_init, + .close = aac_decode_close, + .decode = latm_decode_frame, + .sample_fmts = (const enum AVSampleFormat[]) { + AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE }, - .capabilities = CODEC_CAP_CHANNEL_CONF, + .capabilities = AV_CODEC_CAP_CHANNEL_CONF | AV_CODEC_CAP_DR1, + .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE, .channel_layouts = aac_channel_layout, };