November 2016 - ffmpeg-devel - lists.ffmpeg.org

[PATCH] ffprobe: add -show_compact_data option
by Stefano Sabatini 22 Jan '17

22 Jan '17

This is meant to slightly reduce the output size. --- ffprobe.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/ffprobe.c b/ffprobe.c index 42a8d8e..f2c542a 100644 --- a/ffprobe.c +++ b/ffprobe.c @@ -99,6 +99,7 @@ static int use_byte_value_binary_prefix = 0; static int use_value_sexagesimal_format = 0; static int show_private_data = 1; static int show_headers_first = 0; +static int show_compact_data = 0; static char *print_format; static char *stream_specifier; @@ -723,16 +724,19 @@ static void writer_print_data(WriterContext *wctx, const char *name, av_bprint_init(&bp, 0, AV_BPRINT_SIZE_UNLIMITED); av_bprintf(&bp, "\n"); while (size) { + if (!show_compact_data) av_bprintf(&bp, "%08x: ", offset); - l = FFMIN(size, 16); + l = FFMIN(size, show_compact_data ? 40 : 16); for (i = 0; i < l; i++) { av_bprintf(&bp, "%02x", data[i]); - if (i & 1) + if (!show_compact_data && (i & 1)) av_bprintf(&bp, " "); } + if (!show_compact_data) { av_bprint_chars(&bp, ' ', 41 - 2 * i - i / 2); for (i = 0; i < l; i++) av_bprint_chars(&bp, data[i] - 32U < 95 ? data[i] : '.', 1); + } av_bprintf(&bp, "\n"); offset += l; data += l; @@ -3235,6 +3239,7 @@ static const OptionDef real_options[] = { { "default", HAS_ARG | OPT_AUDIO | OPT_VIDEO | OPT_EXPERT, {.func_arg = opt_default}, "generic catch all option", "" }, { "i", HAS_ARG, {.func_arg = opt_input_file_i}, "read specified file", "input_file"}, { "show_headers_first", OPT_BOOL, {&show_headers_first}, "show headers before the packets/frames" }, + { "show_compact_data", OPT_BOOL, {&show_compact_data}, "show packet data in a compact format" }, { NULL, }, }; -- 1.9.1

3 2

[PATCH] configure: disable the new optimizer in Visual Studio 2015 Update 3
by Hendrik Leppkes 17 Jan '17

17 Jan '17

Visual Studio 2015 Update 3 introduced a new SSA optimizer, however it unfortunately causes miscompilations. Until it is fixed, the new optimizations are disabled and should be re-checked on subsequent compiler releases. Fixes recent FATE failure on VS2015 stations. --- configure | 3 +++ 1 file changed, 3 insertions(+) diff --git a/configure b/configure index 007c953..a1d46c3 100755 --- a/configure +++ b/configure @@ -6203,6 +6203,9 @@ __declspec($_restrict) void* foo(int); EOF fi check_func strtoll || add_cflags -Dstrtoll=_strtoi64 + # the new SSA optimzer in VS2015 U3 is still mis-optimizing some parts of the code + # this flag should be re-checked on newer compiler releases and put under a version check once its fixed + check_cflags -d2SSAOptimizer- fi for pfx in "" host_; do -- 2.7.2.windows.1

6 11

(no subject)
by Matthieu Bouron 12 Jan '17

12 Jan '17

Hello, The following patchset add an audio frame pool for each link of the filtergraph. It extends the FFVideoFramePool API to support audio (and renames it to FFFramePool). The performance gain on a rpi2 is very little. malloc+free goes from 2.50% to 1.84% cpu time with the following command line: perf record ./ffmpeg_g -f lavfi -i sine=440,asetnsamples=4096,aformat=sample_fmts=s16,aresample=48000 -t 500 -f null - For reference, most of the time (81%) is spend in request_frame and in the resample function. Given the performance gain, i'm not sure if it's really useful. However it could be a good idea to have this API to use it in both libavcodec (and replace what is being done in lavc/utils.c) and libavfilter for audio and video if made public or semi-public later on. Best regards, Matthieu

2 19

[PATCH] lavfi/af_ebur128: update filter to use new ebur128 API
by Kyle Swanson 04 Jan '17

04 Jan '17

Hi, Here's a couple of patches which update the ebur128 filter to use the recently added ebur128 API. This updated filter allows fine-tuned control over which EBU R128 parameters are measured, and provides modest speed increases over the previous ebur128 filter. Also noteworthy: this removes the video output option of the ebur128 filter. This is extraneous for an ebur128 measurement filter IMHO, but if we wanted to keep similar functionality in FFmpeg, we'd be better served by a new video source filter where custom meters could be created via exported frame metadata. The first patch adds true peak functionality to the ebur128 API using swresample (this was already discussed a little bit: http://ffmpeg.org/pipermail/ffmpeg-devel/2016-November/202583.html) The second patch is an update to the ebur128 filter. Kyle

6 12

[PATCH] lavf: add ffprobe demuxer
by Stefano Sabatini 02 Jan '17

02 Jan '17

From: Nicolas George <george(a)nsup.org> With several modifications and documentation by Stefano Sabatini <stefasab(a)gmail.com>. Signed-off-by: Nicolas George <george(a)nsup.org> --- doc/ffprobe-format.texi | 130 ++++++++++++++++ libavformat/Makefile | 1 + libavformat/allformats.c | 2 + libavformat/ffprobedec.c | 397 +++++++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 530 insertions(+) create mode 100644 doc/ffprobe-format.texi create mode 100644 libavformat/ffprobedec.c diff --git a/doc/ffprobe-format.texi b/doc/ffprobe-format.texi new file mode 100644 index 0000000..159f4c4 --- /dev/null +++ b/doc/ffprobe-format.texi @@ -0,0 +1,130 @@ +@chapter ffprobe demuxer format +@c man begin FFPROBE DEMUXER FORMAT + +The ffprobe demuxer format is inspired by the output generated by the +ffprobe default format. + +It consists of several sections (FORMAT, STREAM, PACKET). Each section +start with a single line in the format @samp{[SECTION]} and ends with +the corresponding line @samp{[/SECTION]}. + +where @samp{SECTION} is the name of the section. + +A well-formed file will consists of an initial @samp{FORMAT} section, +followed by several @samp{STREAM} sections (one per stream), and +several @samp{PACKET} sections. + +Each section contains a sequence of lines in the form +@sample{key=value}. In the case of data the key and the value must +stay on different lines. + +Unrecognized values are discarded (this allows the demuxer to accept +the output generated by @program{ffprobe} without further +modifications. + +The following sections document the fields accepted by each section. + +@section FORMAT + +@table +@item nb_streams +the number of supported streams +@end table + +@section STREAM +@table +@item index +the index number +@item codec_name +the codec name (the name must be an accepted FFmpeg codec name +@item time_base +Specify the time_base as @samp{NUM/DEN}, this is the time base used to +specify the timestamps in the corresponding packets +@end table + +@section PACKET + +@table +@item stream_index +the stream index of the stream (must have been specified in a stream +section). If not specified the first stream is assumed. + +@item pts +the PTS expressed in the corresponding time base stream units + +@item pts_time +the PTS expressed as an absolute time, using the FFmpeg time syntax + +@item dts +the DTS expressed in the corresponding time base stream units + +@item pts_time +the DTS expressed as an absolute time, using the FFmpeg time syntax + +@item duration +the duration expressed in the corresponding time base stream units + +@item duration_tim +the packet duration expressed as an absolute time, using the FFmpeg time syntax + +@item flags +If the value is equal to "K" mark the packet as a key packet + +@item data +specifies the data as a sequence of hexadecimal values (two +hexadecimal values specify a byte). The data specification starts on +the following line and can span over several lines, and must be +terminate with an empty line. +@end table + +@section Examples + +A ffprobe demuxer file might look like this: +@example +[FORMAT] +nb_streams=1 +format_name=ffprobe_default +[/FORMAT] +[STREAM] +index=0 +codec_name=timed_id3 +time_base=1/1000000 +[/STREAM] +[PACKET] +stream_index=0 +pts_time=0 +dts_time=0 +duration=N/A +flags=K +data= +f000 0000 0167 42c0 1ed9 81b1 fefc 0440 +57ff d7d1 dfff e11a 3d7e 6cbd 0000 0001 +ce8c 4d9d 108e 25e9 fe + +[/PACKET] +[PACKET] +stream_index=0 +pts_time=1.00 +duration=N/A +flags=K +data= +f000 0000 0141 9a38 3be5 ffff ffff fffa +ebc1 3782 7d5e 21e9 ffff abf2 ea4f ed66 +0000 0001 ce8c 4d9d 108e 25e9 fe + +[/PACKET] +@end example + +The output generated by the @program{ffprobe} default format is valid +if the following options are also specified: +@example +-show_packets -show_data -show_compact_data -show_format -show_streams -show_packets -show_headers_first +@end example + +Thus, to generate the corresponding output from an input file you can +do: +@example +ffprobe INPUT.ts -show_packets -show_data -show_compact_data -show_format -show_streams -show_packets -show_headers_first > OUTPUT.ffprobe +@end example + +@c man end FFPROBE DEMUXER FORMAT diff --git a/libavformat/Makefile b/libavformat/Makefile index b68c27c..d222ac2 100644 --- a/libavformat/Makefile +++ b/libavformat/Makefile @@ -162,6 +162,7 @@ OBJS-$(CONFIG_FFM_DEMUXER) += ffmdec.o OBJS-$(CONFIG_FFM_MUXER) += ffmenc.o OBJS-$(CONFIG_FFMETADATA_DEMUXER) += ffmetadec.o OBJS-$(CONFIG_FFMETADATA_MUXER) += ffmetaenc.o +OBJS-$(CONFIG_FFPROBE_DEFAULT_DEMUXER) += ffprobedec.o OBJS-$(CONFIG_FFTEXTDATA_DEMUXER) += fftextdatadec.o OBJS-$(CONFIG_FFTEXTDATA_MUXER) += fftextdataenc.o OBJS-$(CONFIG_FIFO_MUXER) += fifo.o diff --git a/libavformat/allformats.c b/libavformat/allformats.c index e58e41d..2e10e26 100644 --- a/libavformat/allformats.c +++ b/libavformat/allformats.c @@ -123,8 +123,10 @@ void av_register_all(void) REGISTER_MUXER (F4V, f4v); REGISTER_MUXDEMUX(FFM, ffm); REGISTER_MUXDEMUX(FFMETADATA, ffmetadata); + REGISTER_DEMUXER (FFPROBE_DEFAULT, ffprobe_default); REGISTER_MUXDEMUX(FFTEXTDATA, fftextdata); REGISTER_MUXER (FIFO, fifo); + REGISTER_MUXDEMUX(FFTEXTDATA, fftextdata); REGISTER_MUXDEMUX(FILMSTRIP, filmstrip); REGISTER_MUXDEMUX(FLAC, flac); REGISTER_DEMUXER (FLIC, flic); diff --git a/libavformat/ffprobedec.c b/libavformat/ffprobedec.c new file mode 100644 index 0000000..3a631c9 --- /dev/null +++ b/libavformat/ffprobedec.c @@ -0,0 +1,397 @@ +/* + * Copyright (c) 2013 Nicolas George + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public License + * as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public License + * along with FFmpeg; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/avassert.h" +#include "libavutil/avstring.h" +#include "libavutil/bprint.h" +#include "libavutil/parseutils.h" +#include "avformat.h" +#include "internal.h" + +enum SectionType { + SEC_NONE = 0, + SEC_FORMAT, + SEC_STREAM, + SEC_PACKET, +}; + +const char *const section_names[] = { + [SEC_NONE] = "NONE", + [SEC_FORMAT] = "FORMAT", + [SEC_STREAM] = "STREAM", + [SEC_PACKET] = "PACKET", +}; + +typedef struct { + AVClass *class; + enum SectionType section; + AVBPrint data; + int packet_nb; +} FFprobeContext; + +static int ffprobe_probe(AVProbeData *probe) +{ + unsigned score; + + if (!av_strstart(probe->buf, "[FORMAT]\n", NULL)) + return 0; + score = !!strstr(probe->buf, "\nnb_streams=") + + !!strstr(probe->buf, "\nnb_programs=") + + !!strstr(probe->buf, "\nformat_name=") + + !!strstr(probe->buf, "\nstart_time=") + + !!strstr(probe->buf, "\nsize="); + return score >= 3 ? AVPROBE_SCORE_MAX : AVPROBE_SCORE_MAX / 2; +} + +static int ffprobe_read_close(AVFormatContext *avf) +{ + FFprobeContext *ffp = avf->priv_data; + + av_bprint_finalize(&ffp->data, NULL); + return 0; +} + +/** + * Read a section start line ("[SECTION]"). + * Update FFprobeContext.section. + * @return SectionType (>0) for success, + * SEC_NONE if no section start, + * <0 for error + */ +static int read_section_start(AVFormatContext *avf) +{ + FFprobeContext *ffp = avf->priv_data; + uint8_t buf[4096]; + const char *rest; + int i, ret; + + ret = ff_get_line(avf->pb, buf, sizeof(buf)); + if (ret == 0 && avio_feof(avf->pb)) + ret = AVERROR_EOF; + if (ret <= 0) + return ret; + if (*buf != '[') + return 0; + for (i = 1; i < FF_ARRAY_ELEMS(section_names); i++) { + if (av_strstart(buf + 1, section_names[i], &rest) && + !strcmp(rest, "]\n")) { + ffp->section = i; + return i; + } + } + return SEC_NONE; +} + +/** + * Read a line from within a section. + * @return >0 for success, 0 if end of section, <0 for error + */ +static int read_section_line(AVFormatContext *avf, uint8_t *buf, size_t size) +{ + FFprobeContext *ffp = avf->priv_data; + const char *rest; + int ret; + size_t l; + + if ((ret = ff_get_line(avf->pb, buf, size)) <= 0) + return ret; + if (av_strstart(buf, "[/", &rest)) { + ffp->section = 0; + return 0; + } + if ((l = strlen(buf)) > 0 && buf[l - 1] == '\n') + buf[--l] = 0; + return 1; +} + +/** + * Read hexadecimal data + * Store it in FFprobeContext.data. + * @return >=0 for success, <0 for error + */ +static int read_data(AVFormatContext *avf, int *size) +{ + FFprobeContext *ffp = avf->priv_data; + uint8_t buf[4096], *cur; + int ret, pos, val; + + if (ffp->data.len) + return AVERROR_INVALIDDATA; + while ((ret = read_section_line(avf, buf, sizeof(buf)))) { + if (ret < 0) + return ret; + if (!buf[0]) + break; + cur = buf; + pos = 0; + cur += pos; + while (1) { + while (*cur == ' ') + cur++; + if (!*cur) + break; + if ((unsigned)(*cur - '0') >= 10 && + (unsigned)(*cur - 'a') >= 6 && + (unsigned)(*cur - 'A') >= 6) { + av_log(avf, AV_LOG_ERROR, + "Invalid character %c in packet number %d data\n", *cur, ffp->packet_nb); + return AVERROR_INVALIDDATA; + } + pos = 0; + if (sscanf(cur, " %02x%n", &val, &pos) < 1 || !pos) { + av_log(avf, AV_LOG_ERROR, + "Could not parse value in packet number %d data\n", ffp->packet_nb); + return AVERROR_INVALIDDATA; + } + cur += pos; + av_bprint_chars(&ffp->data, val, 1); + } + } + + if (size) + *size = ffp->data.len; + + return av_bprint_is_complete(&ffp->data) ? 0 : AVERROR(ENOMEM); +} + +static int read_section_format(AVFormatContext *avf) +{ + uint8_t buf[4096]; + int ret, val; + + while ((ret = read_section_line(avf, buf, sizeof(buf)))) { + if (ret < 0) + return ret; + if (sscanf(buf, "nb_streams=%d", &val) >= 1) { + while (avf->nb_streams < val) + if (!avformat_new_stream(avf, NULL)) + return AVERROR(ENOMEM); + } + /* TODO programs */ + /* TODO start_time duration bit_rate */ + /* TODO tags */ + } + return SEC_FORMAT; +} + +static int read_section_stream(AVFormatContext *avf) +{ + FFprobeContext *ffp = avf->priv_data; + uint8_t buf[4096]; + int ret, index, val1, val2; + AVStream *st = NULL; + const char *val; + + av_bprint_clear(&ffp->data); + while ((ret = read_section_line(avf, buf, sizeof(buf)))) { + if (ret < 0) + return ret; + if (!st) { + if (sscanf(buf, "index=%d", &index) >= 1) { + if (index == avf->nb_streams) { + if (!avformat_new_stream(avf, NULL)) + return AVERROR(ENOMEM); + } + if ((unsigned)index >= avf->nb_streams) { + av_log(avf, AV_LOG_ERROR, "Invalid stream index: %d\n", + index); + return AVERROR_INVALIDDATA; + } + st = avf->streams[index]; + } else { + av_log(avf, AV_LOG_ERROR, "Stream without index\n"); + return AVERROR_INVALIDDATA; + } + } + if (av_strstart(buf, "codec_name=", &val)) { + const AVCodecDescriptor *desc = avcodec_descriptor_get_by_name(val); + if (desc) { + st->codecpar->codec_id = desc->id; + st->codecpar->codec_type = desc->type; + } + if (!desc) { + av_log(avf, AV_LOG_WARNING, "Cannot recognize codec name '%s'", val); + } + } else if (!strcmp(buf, "extradata=")) { + if ((ret = read_data(avf, NULL)) < 0) + return ret; + if (ffp->data.len) { + if ((ret = ff_alloc_extradata(st->codecpar, ffp->data.len)) < 0) + return ret; + memcpy(st->codecpar->extradata, ffp->data.str, ffp->data.len); + } + } else if (sscanf(buf, "time_base=%d/%d", &val1, &val2) >= 2) { + st->time_base.num = val1; + st->time_base.den = val2; + } + } + return SEC_STREAM; +} + +static int read_section_packet(AVFormatContext *avf, AVPacket *pkt) +{ + FFprobeContext *ffp = avf->priv_data; + uint8_t buf[4096]; + int ret; + AVPacket p; + char flags; + + av_init_packet(&p); + p.pos = avio_tell(avf->pb); + p.stream_index = -1; + p.size = -1; + av_bprint_clear(&ffp->data); + while ((ret = read_section_line(avf, buf, sizeof(buf)))) { + int has_time = 0; + int64_t pts, dts, duration; + char timebuf[1024]; + + if (ret < 0) + return ret; + if (sscanf(buf, "stream_index=%d", &p.stream_index)) { + if ((unsigned)p.stream_index >= avf->nb_streams) { + av_log(avf, AV_LOG_ERROR, "Invalid stream number %d specified in packet number %d\n", + p.stream_index, ffp->packet_nb); + return AVERROR_INVALIDDATA; + } + } + +#define PARSE_TIME(name_, is_duration) \ + sscanf(buf, #name_ "=%"SCNi64, &p.name_); \ + has_time = sscanf(buf, #name_ "_time=%s", timebuf); \ + if (has_time) { \ + int stream_index = p.stream_index == -1 ? 0 : p.stream_index; \ + \ + if (!strcmp(timebuf, "N/A")) { \ + p.name_ = is_duration ? 0 : AV_NOPTS_VALUE; \ + } else { \ + ret = av_parse_time(&name_, timebuf, 1); \ + if (ret < 0) { \ + av_log(avf, AV_LOG_ERROR, "Invalid " #name_ " time specification '%s' for data packet\n", \ + timebuf); \ + return ret; \ + } \ + p.name_ = av_rescale_q(name_, AV_TIME_BASE_Q, avf->streams[stream_index]->time_base); \ + } \ + } \ + + PARSE_TIME(pts, 0); + PARSE_TIME(dts, 0); + PARSE_TIME(duration, 1); + + if (sscanf(buf, "flags=%c", &flags) >= 1) + p.flags = flags == 'K' ? AV_PKT_FLAG_KEY : 0; + if (!strcmp(buf, "data=")) + if ((ret = read_data(avf, &p.size)) < 0) + return ret; + } + if (p.size < 0 || (unsigned)p.stream_index >= avf->nb_streams) + return SEC_NONE; + if ((ret = av_new_packet(pkt, p.size)) < 0) + return ret; + p.data = pkt->data; + p.buf = pkt->buf; + *pkt = p; + if (ffp->data.len) { + ffp->data.len = FFMIN(ffp->data.len, pkt->size); + memcpy(pkt->data, ffp->data.str, ffp->data.len); + } + return SEC_PACKET; +} + +static int read_section(AVFormatContext *avf, AVPacket *pkt) +{ + FFprobeContext *ffp = avf->priv_data; + int ret, section; + + while (!ffp->section) + if ((ret = read_section_start(avf)) < 0) + return ret; + switch (section = ffp->section) { + case SEC_FORMAT: + return read_section_format(avf); + case SEC_STREAM: + return read_section_stream(avf); + case SEC_PACKET: + ret = read_section_packet(avf, pkt); + ffp->packet_nb++; + return ret; + default: + av_assert0(!"reached"); + return AVERROR_BUG; + } +} + +static int ffprobe_read_header(AVFormatContext *avf) +{ + FFprobeContext *ffp = avf->priv_data; + int ret; + + av_bprint_init(&ffp->data, 0, AV_BPRINT_SIZE_UNLIMITED); + if ((ret = read_section_start(avf)) < 0) + return ret; + if (ret != SEC_FORMAT) { + av_log(avf, AV_LOG_INFO, "Using noheader mode\n"); + avf->ctx_flags |= AVFMTCTX_NOHEADER; + return 0; + } + if ((ret = read_section_format(avf)) < 0) + return ret; + + /* read stream information */ + while (1) { + ret = read_section_start(avf); + if (ret != SEC_STREAM) + break; + if ((ret = read_section_stream(avf)) < 0) + return ret; + } + + return 0; +} + +static int ffprobe_read_packet(AVFormatContext *avf, AVPacket *pkt) +{ + int ret; + + while (1) { + if ((ret = read_section(avf, pkt)) < 0) + return ret; + if (ret == SEC_PACKET) + return 0; + } +} + +static const AVClass ffprobe_default_class = { + .class_name = "ffprobe_default demuxer", + .item_name = av_default_item_name, + .version = LIBAVUTIL_VERSION_INT, +}; + +AVInputFormat ff_ffprobe_default_demuxer = { + .name = "ffprobe_default", + .long_name = NULL_IF_CONFIG_SMALL("FFprobe output (default writer)"), + .priv_data_size = sizeof(FFprobeContext), + .read_probe = ffprobe_probe, + .read_header = ffprobe_read_header, + .read_packet = ffprobe_read_packet, + .read_close = ffprobe_read_close, + .priv_class = &ffprobe_default_class, +}; -- 1.9.1

14 78

[PATCH] lavf/mov.c: Set skip_samples according to first edit list, when -ignore_editlist is set.
by Sasi Inguva 29 Dec '16

29 Dec '16

Signed-off-by: Sasi Inguva <isasi(a)google.com> --- libavformat/isom.h | 2 ++ libavformat/mov.c | 22 +++++++++++++++++++++- 2 files changed, 23 insertions(+), 1 deletion(-) diff --git a/libavformat/isom.h b/libavformat/isom.h index d684502..6a8a4e3 100644 --- a/libavformat/isom.h +++ b/libavformat/isom.h @@ -164,6 +164,8 @@ typedef struct MOVStreamContext { uint32_t tmcd_flags; ///< tmcd track flags int64_t track_end; ///< used for dts generation in fragmented movie files int start_pad; ///< amount of samples to skip due to enc-dec delay + int first_elist_start_time; ///< first edit list start time. + int num_non_empty_elst; ///< number of non empty edit lists. unsigned int rap_group_count; MOVSbgp *rap_group; diff --git a/libavformat/mov.c b/libavformat/mov.c index 9ec7d03..676694e 100644 --- a/libavformat/mov.c +++ b/libavformat/mov.c @@ -3245,6 +3245,12 @@ static void mov_build_index(MOVContext *mov, AVStream *st) } } + /* Adjust skip_samples correctly when ignore_editlist is set, to support gapless playback */ + if (mov->ignore_editlist && sc->num_non_empty_elst <= 1 && + st->codecpar->codec_id == AV_CODEC_ID_AAC && sc->first_elist_start_time > 0) { + sc->start_pad = sc->first_elist_start_time; + } + /* only use old uncompressed audio chunk demuxing when stts specifies it */ if (!(st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO && sc->stts_count == 1 && sc->stts_data[0].duration == 1)) { @@ -4424,9 +4430,11 @@ static int mov_read_elst(MOVContext *c, AVIOContext *pb, MOVAtom atom) MOVStreamContext *sc; int i, edit_count, version; - if (c->fc->nb_streams < 1 || c->ignore_editlist) + if (c->fc->nb_streams < 1) return 0; sc = c->fc->streams[c->fc->nb_streams-1]->priv_data; + sc->first_elist_start_time = 0; + sc->num_non_empty_elst = 0; version = avio_r8(pb); /* version */ avio_rb24(pb); /* flags */ @@ -4463,9 +4471,21 @@ static int mov_read_elst(MOVContext *c, AVIOContext *pb, MOVAtom atom) c->fc->nb_streams-1, i, e->time); return AVERROR_INVALIDDATA; } + + if (!sc->first_elist_start_time && e->time > 0) { + sc->first_elist_start_time = e->time; + } + + if (e->time >= 0) { + sc->num_non_empty_elst++; + } } sc->elst_count = i; + if (c->ignore_editlist) { + av_freep(&sc->elst_data); + sc->elst_count = 0; + } return 0; } -- 2.8.0.rc3.226.g39d4020

4 6

[GSoC] [PATCH 2/2] mlpenc: Working MLP/TrueHD encoder
by Jai Luthra 29 Dec '16

29 Dec '16

* Multichannel support for TrueHD is experimental There should be downmix substreams present for 2+ channel bitstreams, but ffmpeg decoder doesn't need it. Will add support for this soon. * There might be lossless check failures on LFE channels * 32-bit sample support has been removed for now, will add it later While testing, some samples gave lossless check failures when enforcing s32. Probably this will also get solved with other lossless issues. Signed-off-by: Jai Luthra <me(a)jailuthra.in> --- libavcodec/Makefile | 2 + libavcodec/allcodecs.c | 4 +- libavcodec/mlp.c | 21 + libavcodec/mlp.h | 40 + libavcodec/mlpenc.c | 2416 ++++++++++++++++++++++++++++++++++++++++++++++++ 5 files changed, 2481 insertions(+), 2 deletions(-) create mode 100644 libavcodec/mlpenc.c diff --git a/libavcodec/Makefile b/libavcodec/Makefile index abef19e..b5a10a0 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -382,6 +382,7 @@ OBJS-$(CONFIG_MJPEG_ENCODER) += mjpegenc.o mjpegenc_common.o OBJS-$(CONFIG_MJPEGB_DECODER) += mjpegbdec.o OBJS-$(CONFIG_MJPEG_VAAPI_ENCODER) += vaapi_encode_mjpeg.o OBJS-$(CONFIG_MLP_DECODER) += mlpdec.o mlpdsp.o +OBJS-$(CONFIG_MLP_ENCODER) += mlpenc.o OBJS-$(CONFIG_MMVIDEO_DECODER) += mmvideo.o OBJS-$(CONFIG_MOTIONPIXELS_DECODER) += motionpixels.o OBJS-$(CONFIG_MOVTEXT_DECODER) += movtextdec.o ass.o @@ -544,6 +545,7 @@ OBJS-$(CONFIG_TIFF_DECODER) += tiff.o lzw.o faxcompr.o tiff_data.o ti OBJS-$(CONFIG_TIFF_ENCODER) += tiffenc.o rle.o lzwenc.o tiff_data.o OBJS-$(CONFIG_TMV_DECODER) += tmv.o cga_data.o OBJS-$(CONFIG_TRUEHD_DECODER) += mlpdec.o mlpdsp.o +OBJS-$(CONFIG_TRUEHD_ENCODER) += mlpenc.o OBJS-$(CONFIG_TRUEMOTION1_DECODER) += truemotion1.o OBJS-$(CONFIG_TRUEMOTION2_DECODER) += truemotion2.o OBJS-$(CONFIG_TRUEMOTION2RT_DECODER) += truemotion2rt.o diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index 951e199..e4ab8b3 100644 --- a/libavcodec/allcodecs.c +++ b/libavcodec/allcodecs.c @@ -419,7 +419,7 @@ void avcodec_register_all(void) REGISTER_DECODER(MACE3, mace3); REGISTER_DECODER(MACE6, mace6); REGISTER_DECODER(METASOUND, metasound); - REGISTER_DECODER(MLP, mlp); + REGISTER_ENCDEC (MLP, mlp); REGISTER_DECODER(MP1, mp1); REGISTER_DECODER(MP1FLOAT, mp1float); REGISTER_ENCDEC (MP2, mp2); @@ -448,7 +448,7 @@ void avcodec_register_all(void) REGISTER_ENCDEC (SONIC, sonic); REGISTER_ENCODER(SONIC_LS, sonic_ls); REGISTER_DECODER(TAK, tak); - REGISTER_DECODER(TRUEHD, truehd); + REGISTER_ENCDEC (TRUEHD, truehd); REGISTER_DECODER(TRUESPEECH, truespeech); REGISTER_ENCDEC (TTA, tta); REGISTER_DECODER(TWINVQ, twinvq); diff --git a/libavcodec/mlp.c b/libavcodec/mlp.c index 87f7c77..ddbab60 100644 --- a/libavcodec/mlp.c +++ b/libavcodec/mlp.c @@ -41,6 +41,27 @@ const uint8_t ff_mlp_huffman_tables[3][18][2] = { } }; +const ChannelInformation ff_mlp_ch_info[21] = { + { 0x01, 0x01, 0x00, 0x1f }, { 0x03, 0x02, 0x00, 0x1b }, + { 0x07, 0x02, 0x01, 0x1f }, { 0x0F, 0x02, 0x02, 0x19 }, + { 0x07, 0x02, 0x01, 0x03 }, { 0x0F, 0x02, 0x02, 0x1f }, + { 0x1F, 0x02, 0x03, 0x01 }, { 0x07, 0x02, 0x01, 0x1a }, + { 0x0F, 0x02, 0x02, 0x1f }, { 0x1F, 0x02, 0x03, 0x18 }, + { 0x0F, 0x02, 0x02, 0x02 }, { 0x1F, 0x02, 0x03, 0x1f }, + { 0x3F, 0x02, 0x04, 0x00 }, { 0x0F, 0x03, 0x01, 0x1f }, + { 0x1F, 0x03, 0x02, 0x18 }, { 0x0F, 0x03, 0x01, 0x02 }, + { 0x1F, 0x03, 0x02, 0x1f }, { 0x3F, 0x03, 0x03, 0x00 }, + { 0x1F, 0x04, 0x01, 0x01 }, { 0x1F, 0x04, 0x01, 0x18 }, + { 0x3F, 0x04, 0x02, 0x00 }, +}; + +const uint64_t ff_mlp_channel_layouts[12] = { + AV_CH_LAYOUT_MONO, AV_CH_LAYOUT_STEREO, AV_CH_LAYOUT_2_1, + AV_CH_LAYOUT_QUAD, AV_CH_LAYOUT_2POINT1, AV_CH_LAYOUT_SURROUND, + AV_CH_LAYOUT_4POINT0, AV_CH_LAYOUT_5POINT0_BACK, AV_CH_LAYOUT_3POINT1, + AV_CH_LAYOUT_4POINT1, AV_CH_LAYOUT_5POINT1_BACK, 0, +}; + static int crc_init = 0; #if CONFIG_SMALL #define CRC_TABLE_SIZE 257 diff --git a/libavcodec/mlp.h b/libavcodec/mlp.h index 05d8dba..41a45a3 100644 --- a/libavcodec/mlp.h +++ b/libavcodec/mlp.h @@ -76,6 +76,9 @@ typedef struct FilterParams { uint8_t shift; ///< Right shift to apply to output of filter. int32_t state[MAX_FIR_ORDER]; + + int coeff_bits; + int coeff_shift; } FilterParams; /** sample data coding information */ @@ -96,6 +99,43 @@ typedef struct ChannelParams { */ extern const uint8_t ff_mlp_huffman_tables[3][18][2]; +typedef struct { + uint8_t channel_occupancy; + uint8_t group1_channels; + uint8_t group2_channels; + uint8_t summary_info; +} ChannelInformation; + +/** Tables defining channel information. + * + * Possible channel arrangements are: + * + * (Group 1) C + * (Group 1) L, R + * (Group 1) Lf, Rf / (Group 2) S + * (Group 1) Lf, Rf / (Group 2) Ls, Rs + * (Group 1) Lf, Rf / (Group 2) LFE + * (Group 1) Lf, Rf / (Group 2) LFE, S + * (Group 1) Lf, Rf / (Group 2) LFE, Ls, Rs + * (Group 1) Lf, Rf / (Group 2) C + * (Group 1) Lf, Rf / (Group 2) C, S + * (Group 1) Lf, Rf / (Group 2) C, Ls, Rs + * (Group 1) Lf, Rf / (Group 2) C, LFE + * (Group 1) Lf, Rf / (Group 2) C, LFE, S + * (Group 1) Lf, Rf / (Group 2) C, LFE, Ls, Rs + * (Group 1) Lf, Rf C / (Group 2) S + * (Group 1) Lf, Rf C / (Group 2) Ls, Rs + * (Group 1) Lf, Rf C / (Group 2) LFE + * (Group 1) Lf, Rf C / (Group 2) LFE, S + * (Group 1) Lf, Rf C / (Group 2) LFE, Ls, Rs + * (Group 1) Lf, Rf Ls Rs / (Group 2) LFE + * (Group 1) Lf, Rf Ls Rs / (Group 2) C + * (Group 1) Lf, Rf, Ls, Rs / (Group 2) C, LFE + */ +extern const ChannelInformation ff_mlp_ch_info[21]; + +extern const uint64_t ff_mlp_channel_layouts[12]; + /** MLP uses checksums that seem to be based on the standard CRC algorithm, but * are not (in implementation terms, the table lookup and XOR are reversed). * We can implement this behavior using a standard av_crc on all but the diff --git a/libavcodec/mlpenc.c b/libavcodec/mlpenc.c new file mode 100644 index 0000000..20924a8 --- /dev/null +++ b/libavcodec/mlpenc.c @@ -0,0 +1,2416 @@ +/** + * MLP encoder + * Copyright (c) 2008 Ramiro Polla + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "avcodec.h" +#include "internal.h" +#include "put_bits.h" +#include "audio_frame_queue.h" +#include "libavutil/crc.h" +#include "libavutil/avstring.h" +#include "libavutil/samplefmt.h" +#include "mlp.h" +#include "lpc.h" + +#define MAJOR_HEADER_INTERVAL 16 + +#define MLP_MIN_LPC_ORDER 1 +#define MLP_MAX_LPC_ORDER 8 +#define MLP_MIN_LPC_SHIFT 8 +#define MLP_MAX_LPC_SHIFT 15 + +typedef struct { + uint8_t min_channel; ///< The index of the first channel coded in this substream. + uint8_t max_channel; ///< The index of the last channel coded in this substream. + uint8_t max_matrix_channel; ///< The number of channels input into the rematrix stage. + + uint8_t noise_shift; ///< The left shift applied to random noise in 0x31ea substreams. + uint32_t noisegen_seed; ///< The current seed value for the pseudorandom noise generator(s). + + int data_check_present; ///< Set if the substream contains extra info to check the size of VLC blocks. + + int32_t lossless_check_data; ///< XOR of all output samples + + uint8_t max_huff_lsbs; ///< largest huff_lsbs + uint8_t max_output_bits; ///< largest output bit-depth +} RestartHeader; + +typedef struct { + uint8_t count; ///< number of matrices to apply + + uint8_t outch[MAX_MATRICES]; ///< output channel for each matrix + int32_t forco[MAX_MATRICES][MAX_CHANNELS+2]; ///< forward coefficients + int32_t coeff[MAX_MATRICES][MAX_CHANNELS+2]; ///< decoding coefficients + uint8_t fbits[MAX_CHANNELS]; ///< fraction bits + + int8_t shift[MAX_CHANNELS]; ///< Left shift to apply to decoded PCM values to get final 24-bit output. +} MatrixParams; + +enum ParamFlags { + PARAMS_DEFAULT = 0xff, + PARAM_PRESENCE_FLAGS = 1 << 8, + PARAM_BLOCKSIZE = 1 << 7, + PARAM_MATRIX = 1 << 6, + PARAM_OUTSHIFT = 1 << 5, + PARAM_QUANTSTEP = 1 << 4, + PARAM_FIR = 1 << 3, + PARAM_IIR = 1 << 2, + PARAM_HUFFOFFSET = 1 << 1, + PARAM_PRESENT = 1 << 0, +}; + +typedef struct { + uint16_t blocksize; ///< number of PCM samples in current audio block + uint8_t quant_step_size[MAX_CHANNELS]; ///< left shift to apply to Huffman-decoded residuals + + MatrixParams matrix_params; + + uint8_t param_presence_flags; ///< Bitmask of which parameter sets are conveyed in a decoding parameter block. +} DecodingParams; + +typedef struct BestOffset { + int16_t offset; + int bitcount; + int lsb_bits; + int16_t min; + int16_t max; +} BestOffset; + +#define HUFF_OFFSET_MIN -16384 +#define HUFF_OFFSET_MAX 16383 + +/** Number of possible codebooks (counting "no codebooks") */ +#define NUM_CODEBOOKS 4 + +typedef struct { + AVCodecContext *avctx; + + int num_substreams; ///< Number of substreams contained within this stream. + + int num_channels; /**< Number of channels in major_scratch_buffer. + * Normal channels + noise channels. */ + + int coded_sample_fmt [2]; ///< sample format encoded for MLP + int coded_sample_rate[2]; ///< sample rate encoded for MLP + int coded_peak_bitrate; ///< peak bitrate for this major sync header + + int flags; ///< major sync info flags + + /* channel_meaning */ + int substream_info; + int fs; + int wordlength; + int channel_occupancy; + int summary_info; + + int32_t *inout_buffer; ///< Pointer to data currently being read from lavc or written to bitstream. + int32_t *major_inout_buffer; ///< Buffer with all in/out data for one entire major frame interval. + int32_t *write_buffer; ///< Pointer to data currently being written to bitstream. + int32_t *sample_buffer; ///< Pointer to current access unit samples. + int32_t *major_scratch_buffer; ///< Scratch buffer big enough to fit all data for one entire major frame interval. + int32_t *last_frame; ///< Pointer to last frame with data to encode. + + int32_t *lpc_sample_buffer; + + unsigned int major_number_of_frames; + unsigned int next_major_number_of_frames; + + unsigned int major_frame_size; ///< Number of samples in current major frame being encoded. + unsigned int next_major_frame_size; ///< Counter of number of samples for next major frame. + + int32_t *lossless_check_data; ///< Array with lossless_check_data for each access unit. + + unsigned int *max_output_bits; ///< largest output bit-depth + unsigned int *frame_size; ///< Array with number of samples/channel in each access unit. + unsigned int frame_index; ///< Index of current frame being encoded. + + unsigned int one_sample_buffer_size; ///< Number of samples*channel for one access unit. + + unsigned int max_restart_interval; ///< Max interval of access units in between two major frames. + unsigned int min_restart_interval; ///< Min interval of access units in between two major frames. + unsigned int restart_intervals; ///< Number of possible major frame sizes. + + uint16_t timestamp; ///< Timestamp of current access unit. + uint16_t dts; ///< Decoding timestamp of current access unit. + + uint8_t channel_arrangement; ///< channel arrangement for MLP streams + + uint8_t ch_modifier_thd0; ///< channel modifier for TrueHD stream 0 + uint8_t ch_modifier_thd1; ///< channel modifier for TrueHD stream 1 + uint8_t ch_modifier_thd2; ///< channel modifier for TrueHD stream 2 + + unsigned int seq_size [MAJOR_HEADER_INTERVAL]; + unsigned int seq_offset[MAJOR_HEADER_INTERVAL]; + unsigned int sequence_size; + + ChannelParams *channel_params; + + BestOffset best_offset[MAJOR_HEADER_INTERVAL+1][MAX_CHANNELS][NUM_CODEBOOKS]; + + DecodingParams *decoding_params; + RestartHeader restart_header [MAX_SUBSTREAMS]; + + ChannelParams major_channel_params[MAJOR_HEADER_INTERVAL+1][MAX_CHANNELS]; ///< ChannelParams to be written to bitstream. + DecodingParams major_decoding_params[MAJOR_HEADER_INTERVAL+1][MAX_SUBSTREAMS]; ///< DecodingParams to be written to bitstream. + int major_params_changed[MAJOR_HEADER_INTERVAL+1][MAX_SUBSTREAMS]; ///< params_changed to be written to bitstream. + + unsigned int major_cur_subblock_index; + unsigned int major_filter_state_subblock; + unsigned int major_number_of_subblocks; + + BestOffset (*cur_best_offset)[NUM_CODEBOOKS]; + ChannelParams *cur_channel_params; + DecodingParams *cur_decoding_params; + RestartHeader *cur_restart_header; + + AudioFrameQueue afq; + + /* Analysis stage. */ + unsigned int starting_frame_index; + unsigned int number_of_frames; + unsigned int number_of_samples; + unsigned int number_of_subblocks; + unsigned int seq_index; ///< Sequence index for high compression levels. + + ChannelParams *prev_channel_params; + DecodingParams *prev_decoding_params; + + ChannelParams *seq_channel_params; + DecodingParams *seq_decoding_params; + + unsigned int max_codebook_search; + + LPCContext lpc_ctx; +} MLPEncodeContext; + +static ChannelParams restart_channel_params[MAX_CHANNELS]; +static DecodingParams restart_decoding_params[MAX_SUBSTREAMS]; +static BestOffset restart_best_offset[NUM_CODEBOOKS] = {{0}}; + +#define SYNC_MAJOR 0xf8726f +#define MAJOR_SYNC_INFO_SIGNATURE 0xB752 + +#define SYNC_MLP 0xbb +#define SYNC_TRUEHD 0xba + +/* must be set for DVD-A */ +#define FLAGS_DVDA 0x4000 +/* FIFO delay must be constant */ +#define FLAGS_CONST 0x8000 + +#define SUBSTREAM_INFO_MAX_2_CHAN 0x01 +#define SUBSTREAM_INFO_HIGH_RATE 0x02 +#define SUBSTREAM_INFO_ALWAYS_SET 0x04 +#define SUBSTREAM_INFO_2_SUBSTREAMS 0x08 + +/**************************************************************************** + ************ Functions that copy, clear, or compare parameters ************* + ****************************************************************************/ + +/** Compares two FilterParams structures and returns 1 if anything has + * changed. Returns 0 if they are both equal. + */ +static int compare_filter_params(const ChannelParams *prev_cp, const ChannelParams *cp, int filter) +{ + const FilterParams *prev = &prev_cp->filter_params[filter]; + const FilterParams *fp = &cp->filter_params[filter]; + int i; + + if (prev->order != fp->order) + return 1; + + if (!prev->order) + return 0; + + if (prev->shift != fp->shift) + return 1; + + for (i = 0; i < fp->order; i++) + if (prev_cp->coeff[filter][i] != cp->coeff[filter][i]) + return 1; + + return 0; +} + +/** Compare two primitive matrices and returns 1 if anything has changed. + * Returns 0 if they are both equal. + */ +static int compare_matrix_params(MLPEncodeContext *ctx, const MatrixParams *prev, const MatrixParams *mp) +{ + RestartHeader *rh = ctx->cur_restart_header; + unsigned int channel, mat; + + if (prev->count != mp->count) + return 1; + + if (!prev->count) + return 0; + + for (channel = rh->min_channel; channel <= rh->max_channel; channel++) + if (prev->fbits[channel] != mp->fbits[channel]) + return 1; + + for (mat = 0; mat < mp->count; mat++) { + if (prev->outch[mat] != mp->outch[mat]) + return 1; + + for (channel = 0; channel < ctx->num_channels; channel++) + if (prev->coeff[mat][channel] != mp->coeff[mat][channel]) + return 1; + } + + return 0; +} + +/** Compares two DecodingParams and ChannelParams structures to decide if a + * new decoding params header has to be written. + */ +static int compare_decoding_params(MLPEncodeContext *ctx) +{ + DecodingParams *prev = ctx->prev_decoding_params; + DecodingParams *dp = ctx->cur_decoding_params; + MatrixParams *prev_mp = &prev->matrix_params; + MatrixParams *mp = &dp->matrix_params; + RestartHeader *rh = ctx->cur_restart_header; + unsigned int ch; + int retval = 0; + + if (prev->param_presence_flags != dp->param_presence_flags) + retval |= PARAM_PRESENCE_FLAGS; + + if (prev->blocksize != dp->blocksize) + retval |= PARAM_BLOCKSIZE; + + if (compare_matrix_params(ctx, prev_mp, mp)) + retval |= PARAM_MATRIX; + + for (ch = 0; ch <= rh->max_matrix_channel; ch++) + if (prev_mp->shift[ch] != mp->shift[ch]) { + retval |= PARAM_OUTSHIFT; + break; + } + + for (ch = 0; ch <= rh->max_channel; ch++) + if (prev->quant_step_size[ch] != dp->quant_step_size[ch]) { + retval |= PARAM_QUANTSTEP; + break; + } + + for (ch = rh->min_channel; ch <= rh->max_channel; ch++) { + ChannelParams *prev_cp = &ctx->prev_channel_params[ch]; + ChannelParams *cp = &ctx->cur_channel_params[ch]; + + if (!(retval & PARAM_FIR) && + compare_filter_params(prev_cp, cp, FIR)) + retval |= PARAM_FIR; + + if (!(retval & PARAM_IIR) && + compare_filter_params(prev_cp, cp, IIR)) + retval |= PARAM_IIR; + + if (prev_cp->huff_offset != cp->huff_offset) + retval |= PARAM_HUFFOFFSET; + + if (prev_cp->codebook != cp->codebook || + prev_cp->huff_lsbs != cp->huff_lsbs ) + retval |= 0x1; + } + + return retval; +} + +static void copy_filter_params(ChannelParams *dst_cp, ChannelParams *src_cp, int filter) +{ + FilterParams *dst = &dst_cp->filter_params[filter]; + FilterParams *src = &src_cp->filter_params[filter]; + unsigned int order; + + dst->order = src->order; + + if (dst->order) { + dst->shift = src->shift; + + dst->coeff_shift = src->coeff_shift; + dst->coeff_bits = src->coeff_bits; + } + + for (order = 0; order < dst->order; order++) + dst_cp->coeff[filter][order] = src_cp->coeff[filter][order]; +} + +static void copy_matrix_params(MatrixParams *dst, MatrixParams *src) +{ + dst->count = src->count; + + if (dst->count) { + unsigned int channel, count; + + for (channel = 0; channel < MAX_CHANNELS; channel++) { + + dst->fbits[channel] = src->fbits[channel]; + dst->shift[channel] = src->shift[channel]; + + for (count = 0; count < MAX_MATRICES; count++) + dst->coeff[count][channel] = src->coeff[count][channel]; + } + + for (count = 0; count < MAX_MATRICES; count++) + dst->outch[count] = src->outch[count]; + } +} + +static void copy_restart_frame_params(MLPEncodeContext *ctx, + unsigned int substr) +{ + unsigned int index; + + for (index = 0; index < ctx->number_of_subblocks; index++) { + DecodingParams *dp = ctx->seq_decoding_params + index*(ctx->num_substreams) + substr; + unsigned int channel; + + copy_matrix_params(&dp->matrix_params, &ctx->cur_decoding_params->matrix_params); + + for (channel = 0; channel < ctx->avctx->channels; channel++) { + ChannelParams *cp = ctx->seq_channel_params + index*(ctx->avctx->channels) + channel; + unsigned int filter; + + dp->quant_step_size[channel] = ctx->cur_decoding_params->quant_step_size[channel]; + dp->matrix_params.shift[channel] = ctx->cur_decoding_params->matrix_params.shift[channel]; + + if (index) + for (filter = 0; filter < NUM_FILTERS; filter++) + copy_filter_params(cp, &ctx->cur_channel_params[channel], filter); + } + } +} + +/** Clears a DecodingParams struct the way it should be after a restart header. */ +static void clear_decoding_params(MLPEncodeContext *ctx, DecodingParams decoding_params[MAX_SUBSTREAMS]) +{ + unsigned int substr; + + for (substr = 0; substr < ctx->num_substreams; substr++) { + DecodingParams *dp = &decoding_params[substr]; + + dp->param_presence_flags = 0xff; + dp->blocksize = 8; + + memset(&dp->matrix_params , 0, sizeof(MatrixParams )); + memset(dp->quant_step_size, 0, sizeof(dp->quant_step_size)); + } +} + +/** Clears a ChannelParams struct the way it should be after a restart header. */ +static void clear_channel_params(MLPEncodeContext *ctx, ChannelParams channel_params[MAX_CHANNELS]) +{ + unsigned int channel; + + for (channel = 0; channel < ctx->avctx->channels; channel++) { + ChannelParams *cp = &channel_params[channel]; + + memset(&cp->filter_params, 0, sizeof(cp->filter_params)); + + /* Default audio coding is 24-bit raw PCM. */ + cp->huff_offset = 0; + cp->codebook = 0; + cp->huff_lsbs = 24; + } +} + +/** Sets default vales in our encoder for a DecodingParams struct. */ +static void default_decoding_params(MLPEncodeContext *ctx, + DecodingParams decoding_params[MAX_SUBSTREAMS]) +{ + unsigned int substr; + + clear_decoding_params(ctx, decoding_params); + + for (substr = 0; substr < ctx->num_substreams; substr++) { + DecodingParams *dp = &decoding_params[substr]; + uint8_t param_presence_flags = 0; + + param_presence_flags |= PARAM_BLOCKSIZE; + param_presence_flags |= PARAM_MATRIX; + param_presence_flags |= PARAM_OUTSHIFT; + param_presence_flags |= PARAM_QUANTSTEP; + param_presence_flags |= PARAM_FIR; +/* param_presence_flags |= PARAM_IIR; */ + param_presence_flags |= PARAM_HUFFOFFSET; + param_presence_flags |= PARAM_PRESENT; + + dp->param_presence_flags = param_presence_flags; + } +} + +/****************************************************************************/ + +/** Calculates the smallest number of bits it takes to encode a given signed + * value in two's complement. + */ +static int inline number_sbits(int number) +{ + if (number < 0) + number++; + + return av_log2(FFABS(number)) + 1 + !!number; +} + +enum InputBitDepth { + BITS_16, + BITS_20, + BITS_24, +}; + +static int mlp_peak_bitrate(int peak_bitrate, int sample_rate) +{ + return ((peak_bitrate << 4) - 8) / sample_rate; +} + +static av_cold int mlp_encode_init(AVCodecContext *avctx) +{ + MLPEncodeContext *ctx = avctx->priv_data; + unsigned int substr, index; + unsigned int sum = 0; + unsigned int size; + + ctx->avctx = avctx; + + switch (avctx->sample_rate) { + case 44100 << 0: + avctx->frame_size = 40 << 0; + ctx->coded_sample_rate[0] = 0x08 + 0; + ctx->fs = 0x08 + 1; + break; + case 44100 << 1: + avctx->frame_size = 40 << 1; + ctx->coded_sample_rate[0] = 0x08 + 1; + ctx->fs = 0x0C + 1; + break; + case 44100 << 2: + ctx->substream_info |= SUBSTREAM_INFO_HIGH_RATE; + avctx->frame_size = 40 << 2; + ctx->coded_sample_rate[0] = 0x08 + 2; + ctx->fs = 0x10 + 1; + break; + case 48000 << 0: + avctx->frame_size = 40 << 0; + ctx->coded_sample_rate[0] = 0x00 + 0; + ctx->fs = 0x08 + 2; + break; + case 48000 << 1: + avctx->frame_size = 40 << 1; + ctx->coded_sample_rate[0] = 0x00 + 1; + ctx->fs = 0x0C + 2; + break; + case 48000 << 2: + ctx->substream_info |= SUBSTREAM_INFO_HIGH_RATE; + avctx->frame_size = 40 << 2; + ctx->coded_sample_rate[0] = 0x00 + 2; + ctx->fs = 0x10 + 2; + break; + default: + av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d. Supported " + "sample rates are 44100, 88200, 176400, 48000, " + "96000, and 192000.\n", avctx->sample_rate); + return -1; + } + ctx->coded_sample_rate[1] = -1 & 0xf; + + /* TODO Keep count of bitrate and calculate real value. */ + ctx->coded_peak_bitrate = mlp_peak_bitrate(9600000, avctx->sample_rate); + + /* TODO support more channels. */ + if (avctx->channels > 2) { + av_log(avctx, AV_LOG_WARNING, + "Only mono and stereo are supported at the moment.\n"); + } + + ctx->substream_info |= SUBSTREAM_INFO_ALWAYS_SET; + if (avctx->channels <= 2) { + ctx->substream_info |= SUBSTREAM_INFO_MAX_2_CHAN; + } + + switch (avctx->sample_fmt) { + case AV_SAMPLE_FMT_S16: + ctx->coded_sample_fmt[0] = BITS_16; + ctx->wordlength = 16; + avctx->bits_per_raw_sample = 16; + break; + /* TODO 20 bits: */ + case AV_SAMPLE_FMT_S32: + ctx->coded_sample_fmt[0] = BITS_24; + ctx->wordlength = 24; + avctx->bits_per_raw_sample = 24; + break; + default: + av_log(avctx, AV_LOG_ERROR, "Sample format not supported. " + "Only 16- and 24-bit samples are supported.\n"); + return -1; + } + ctx->coded_sample_fmt[1] = -1 & 0xf; + + ctx->dts = -avctx->frame_size; + + ctx->num_channels = avctx->channels + 2; /* +2 noise channels */ + ctx->one_sample_buffer_size = avctx->frame_size + * ctx->num_channels; + /* TODO Let user pass major header interval as parameter. */ + ctx->max_restart_interval = MAJOR_HEADER_INTERVAL; + + ctx->max_codebook_search = 3; + ctx->min_restart_interval = MAJOR_HEADER_INTERVAL; + ctx->restart_intervals = ctx->max_restart_interval / ctx->min_restart_interval; + + /* TODO Let user pass parameters for LPC filter. */ + + size = sizeof(int32_t) * avctx->frame_size * ctx->max_restart_interval; + + ctx->lpc_sample_buffer = av_malloc(size); + if (!ctx->lpc_sample_buffer) { + av_log(avctx, AV_LOG_ERROR, + "Not enough memory for buffering samples.\n"); + return -1; + } + + size = sizeof(int32_t) + * ctx->one_sample_buffer_size * ctx->max_restart_interval; + + ctx->major_scratch_buffer = av_malloc(size); + if (!ctx->major_scratch_buffer) { + av_log(avctx, AV_LOG_ERROR, + "Not enough memory for buffering samples.\n"); + return -1; + } + + ctx->major_inout_buffer = av_malloc(size); + if (!ctx->major_inout_buffer) { + av_log(avctx, AV_LOG_ERROR, + "Not enough memory for buffering samples.\n"); + return -1; + } + + ff_mlp_init_crc(); + + ctx->num_substreams = 1; // TODO: change this after adding multi-channel support for TrueHD + + if (ctx->avctx->codec_id == AV_CODEC_ID_MLP) { + /* MLP */ + switch(avctx->channel_layout) { + case AV_CH_LAYOUT_MONO: + ctx->channel_arrangement = 0; break; + case AV_CH_LAYOUT_STEREO: + ctx->channel_arrangement = 1; break; + case AV_CH_LAYOUT_2_1: + ctx->channel_arrangement = 2; break; + case AV_CH_LAYOUT_QUAD: + ctx->channel_arrangement = 3; break; + case AV_CH_LAYOUT_2POINT1: + ctx->channel_arrangement = 4; break; + case AV_CH_LAYOUT_SURROUND: + ctx->channel_arrangement = 7; break; + case AV_CH_LAYOUT_4POINT0: + ctx->channel_arrangement = 8; break; + case AV_CH_LAYOUT_5POINT0_BACK: + ctx->channel_arrangement = 9; break; + case AV_CH_LAYOUT_3POINT1: + ctx->channel_arrangement = 10; break; + case AV_CH_LAYOUT_4POINT1: + ctx->channel_arrangement = 11; break; + case AV_CH_LAYOUT_5POINT1_BACK: + ctx->channel_arrangement = 12; break; + default: + av_log(avctx, AV_LOG_ERROR, "Unsupported channel arrangement\n"); + return -1; + } + ctx->flags = FLAGS_DVDA; + ctx->channel_occupancy = ff_mlp_ch_info[ctx->channel_arrangement].channel_occupancy; + ctx->summary_info = ff_mlp_ch_info[ctx->channel_arrangement].summary_info ; + } else { + /* TrueHD */ + switch(avctx->channel_layout) { + case AV_CH_LAYOUT_STEREO: + ctx->ch_modifier_thd0 = 0; + ctx->ch_modifier_thd1 = 0; + ctx->ch_modifier_thd2 = 0; + ctx->channel_arrangement = 1; + break; + case AV_CH_LAYOUT_5POINT0_BACK: + ctx->ch_modifier_thd0 = 1; + ctx->ch_modifier_thd1 = 1; + ctx->ch_modifier_thd2 = 1; + ctx->channel_arrangement = 11; + break; + case AV_CH_LAYOUT_5POINT1_BACK: + ctx->ch_modifier_thd0 = 2; + ctx->ch_modifier_thd1 = 1; + ctx->ch_modifier_thd2 = 2; + ctx->channel_arrangement = 15; + break; + default: + av_log(avctx, AV_LOG_ERROR, "Unsupported channel arrangement\n"); + return -1; + } + ctx->flags = 0; + ctx->channel_occupancy = 0; + ctx->summary_info = 0; + } + + size = sizeof(unsigned int) * ctx->max_restart_interval; + + ctx->frame_size = av_malloc(size); + if (!ctx->frame_size) + return -1; + + ctx->max_output_bits = av_malloc(size); + if (!ctx->max_output_bits) + return -1; + + size = sizeof(int32_t) + * ctx->num_substreams * ctx->max_restart_interval; + + ctx->lossless_check_data = av_malloc(size); + if (!ctx->lossless_check_data) + return -1; + + for (index = 0; index < ctx->restart_intervals; index++) { + ctx->seq_offset[index] = sum; + ctx->seq_size [index] = ((index + 1) * ctx->min_restart_interval) + 1; + sum += ctx->seq_size[index]; + } + ctx->sequence_size = sum; + size = sizeof(ChannelParams) + * ctx->restart_intervals * ctx->sequence_size * ctx->avctx->channels; + ctx->channel_params = av_malloc(size); + if (!ctx->channel_params) { + av_log(avctx, AV_LOG_ERROR, + "Not enough memory for analysis context.\n"); + return -1; + } + + size = sizeof(DecodingParams) + * ctx->restart_intervals * ctx->sequence_size * ctx->num_substreams; + ctx->decoding_params = av_malloc(size); + if (!ctx->decoding_params) { + av_log(avctx, AV_LOG_ERROR, + "Not enough memory for analysis context.\n"); + return -1; + } + + for (substr = 0; substr < ctx->num_substreams; substr++) { + RestartHeader *rh = &ctx->restart_header [substr]; + + /* TODO see if noisegen_seed is really worth it. */ + rh->noisegen_seed = 0; + + rh->min_channel = 0; + rh->max_channel = avctx->channels - 1; + /* FIXME: this works for 1 and 2 channels, but check for more */ + rh->max_matrix_channel = rh->max_channel; + } + + clear_channel_params(ctx, restart_channel_params); + clear_decoding_params(ctx, restart_decoding_params); + + if (ff_lpc_init(&ctx->lpc_ctx, ctx->number_of_samples, + MLP_MAX_LPC_ORDER, FF_LPC_TYPE_LEVINSON) < 0) { + av_log(avctx, AV_LOG_ERROR, + "Not enough memory for LPC context.\n"); + return -1; + } + + ff_af_queue_init(avctx, &ctx->afq); + + return 0; +} + +/**************************************************************************** + ****************** Functions that write to the bitstream ******************* + ****************************************************************************/ + +/** Writes a major sync header to the bitstream. */ +static void write_major_sync(MLPEncodeContext *ctx, uint8_t *buf, int buf_size) +{ + PutBitContext pb; + + init_put_bits(&pb, buf, buf_size); + + put_bits(&pb, 24, SYNC_MAJOR ); + + if (ctx->avctx->codec_id == AV_CODEC_ID_MLP) { + put_bits(&pb, 8, SYNC_MLP ); + put_bits(&pb, 4, ctx->coded_sample_fmt [0]); + put_bits(&pb, 4, ctx->coded_sample_fmt [1]); + put_bits(&pb, 4, ctx->coded_sample_rate[0]); + put_bits(&pb, 4, ctx->coded_sample_rate[1]); + put_bits(&pb, 4, 0 ); /* ignored */ + put_bits(&pb, 4, 0 ); /* multi_channel_type */ + put_bits(&pb, 3, 0 ); /* ignored */ + put_bits(&pb, 5, ctx->channel_arrangement ); + } else if (ctx->avctx->codec_id == AV_CODEC_ID_TRUEHD) { + put_bits(&pb, 8, SYNC_TRUEHD ); + put_bits(&pb, 4, ctx->coded_sample_rate[0]); + put_bits(&pb, 4, 0 ); /* ignored */ + put_bits(&pb, 2, ctx->ch_modifier_thd0 ); + put_bits(&pb, 2, ctx->ch_modifier_thd1 ); + put_bits(&pb, 5, ctx->channel_arrangement ); + put_bits(&pb, 2, ctx->ch_modifier_thd2 ); + put_bits(&pb, 13, ctx->channel_arrangement ); + } + + put_bits(&pb, 16, MAJOR_SYNC_INFO_SIGNATURE); + put_bits(&pb, 16, ctx->flags ); + put_bits(&pb, 16, 0 ); /* ignored */ + put_bits(&pb, 1, 1 ); /* is_vbr */ + put_bits(&pb, 15, ctx->coded_peak_bitrate ); + put_bits(&pb, 4, 1 ); /* num_substreams */ + put_bits(&pb, 4, 0x1 ); /* ignored */ + + /* channel_meaning */ + put_bits(&pb, 8, ctx->substream_info ); + put_bits(&pb, 5, ctx->fs ); + put_bits(&pb, 5, ctx->wordlength ); + put_bits(&pb, 6, ctx->channel_occupancy ); + put_bits(&pb, 3, 0 ); /* ignored */ + put_bits(&pb, 10, 0 ); /* speaker_layout */ + put_bits(&pb, 3, 0 ); /* copy_protection */ + put_bits(&pb, 16, 0x8080 ); /* ignored */ + put_bits(&pb, 7, 0 ); /* ignored */ + put_bits(&pb, 4, 0 ); /* source_format */ + put_bits(&pb, 5, ctx->summary_info ); + + flush_put_bits(&pb); + + AV_WL16(buf+26, ff_mlp_checksum16(buf, 26)); +} + +/** Writes a restart header to the bitstream. Damaged streams can start being + * decoded losslessly again after such a header and the subsequent decoding + * params header. + */ +static void write_restart_header(MLPEncodeContext *ctx, PutBitContext *pb) +{ + RestartHeader *rh = ctx->cur_restart_header; + int32_t lossless_check = xor_32_to_8(rh->lossless_check_data); + unsigned int start_count = put_bits_count(pb); + PutBitContext tmpb; + uint8_t checksum; + unsigned int ch; + + put_bits(pb, 14, 0x31ea ); /* TODO 0x31eb */ + put_bits(pb, 16, ctx->timestamp ); + put_bits(pb, 4, rh->min_channel ); + put_bits(pb, 4, rh->max_channel ); + put_bits(pb, 4, rh->max_matrix_channel); + put_bits(pb, 4, rh->noise_shift ); + put_bits(pb, 23, rh->noisegen_seed ); + put_bits(pb, 4, 0 ); /* TODO max_shift */ + put_bits(pb, 5, rh->max_huff_lsbs ); + put_bits(pb, 5, rh->max_output_bits ); + put_bits(pb, 5, rh->max_output_bits ); + put_bits(pb, 1, rh->data_check_present); + put_bits(pb, 8, lossless_check ); + put_bits(pb, 16, 0 ); /* ignored */ + + for (ch = 0; ch <= rh->max_matrix_channel; ch++) + put_bits(pb, 6, ch); + + /* Data must be flushed for the checksum to be correct. */ + tmpb = *pb; + flush_put_bits(&tmpb); + + checksum = ff_mlp_restart_checksum(pb->buf, put_bits_count(pb) - start_count); + + put_bits(pb, 8, checksum); +} + +/** Writes matrix params for all primitive matrices to the bitstream. */ +static void write_matrix_params(MLPEncodeContext *ctx, PutBitContext *pb) +{ + DecodingParams *dp = ctx->cur_decoding_params; + MatrixParams *mp = &dp->matrix_params; + unsigned int mat; + + put_bits(pb, 4, mp->count); + + for (mat = 0; mat < mp->count; mat++) { + unsigned int channel; + + put_bits(pb, 4, mp->outch[mat]); /* matrix_out_ch */ + put_bits(pb, 4, mp->fbits[mat]); + put_bits(pb, 1, 0 ); /* lsb_bypass */ + + for (channel = 0; channel < ctx->num_channels; channel++) { + int32_t coeff = mp->coeff[mat][channel]; + + if (coeff) { + put_bits(pb, 1, 1); + + coeff >>= 14 - mp->fbits[mat]; + + put_sbits(pb, mp->fbits[mat] + 2, coeff); + } else { + put_bits(pb, 1, 0); + } + } + } +} + +/** Writes filter parameters for one filter to the bitstream. */ +static void write_filter_params(MLPEncodeContext *ctx, PutBitContext *pb, + unsigned int channel, unsigned int filter) +{ + FilterParams *fp = &ctx->cur_channel_params[channel].filter_params[filter]; + + put_bits(pb, 4, fp->order); + + if (fp->order > 0) { + int i; + int32_t *fcoeff = ctx->cur_channel_params[channel].coeff[filter]; + + put_bits(pb, 4, fp->shift ); + put_bits(pb, 5, fp->coeff_bits ); + put_bits(pb, 3, fp->coeff_shift); + + for (i = 0; i < fp->order; i++) { + put_sbits(pb, fp->coeff_bits, fcoeff[i] >> fp->coeff_shift); + } + + /* TODO state data for IIR filter. */ + put_bits(pb, 1, 0); + } +} + +/** Writes decoding parameters to the bitstream. These change very often, + * usually at almost every frame. + */ +static void write_decoding_params(MLPEncodeContext *ctx, PutBitContext *pb, + int params_changed) +{ + DecodingParams *dp = ctx->cur_decoding_params; + RestartHeader *rh = ctx->cur_restart_header; + MatrixParams *mp = &dp->matrix_params; + unsigned int ch; + + if (dp->param_presence_flags != PARAMS_DEFAULT && + params_changed & PARAM_PRESENCE_FLAGS) { + put_bits(pb, 1, 1); + put_bits(pb, 8, dp->param_presence_flags); + } else { + put_bits(pb, 1, 0); + } + + if (dp->param_presence_flags & PARAM_BLOCKSIZE) { + if (params_changed & PARAM_BLOCKSIZE) { + put_bits(pb, 1, 1); + put_bits(pb, 9, dp->blocksize); + } else { + put_bits(pb, 1, 0); + } + } + + if (dp->param_presence_flags & PARAM_MATRIX) { + if (params_changed & PARAM_MATRIX) { + put_bits(pb, 1, 1); + write_matrix_params(ctx, pb); + } else { + put_bits(pb, 1, 0); + } + } + + if (dp->param_presence_flags & PARAM_OUTSHIFT) { + if (params_changed & PARAM_OUTSHIFT) { + put_bits(pb, 1, 1); + for (ch = 0; ch <= rh->max_matrix_channel; ch++) + put_sbits(pb, 4, mp->shift[ch]); + } else { + put_bits(pb, 1, 0); + } + } + + if (dp->param_presence_flags & PARAM_QUANTSTEP) { + if (params_changed & PARAM_QUANTSTEP) { + put_bits(pb, 1, 1); + for (ch = 0; ch <= rh->max_channel; ch++) + put_bits(pb, 4, dp->quant_step_size[ch]); + } else { + put_bits(pb, 1, 0); + } + } + + for (ch = rh->min_channel; ch <= rh->max_channel; ch++) { + ChannelParams *cp = &ctx->cur_channel_params[ch]; + + if (dp->param_presence_flags & 0xF) { + put_bits(pb, 1, 1); + + if (dp->param_presence_flags & PARAM_FIR) { + if (params_changed & PARAM_FIR) { + put_bits(pb, 1, 1); + write_filter_params(ctx, pb, ch, FIR); + } else { + put_bits(pb, 1, 0); + } + } + + if (dp->param_presence_flags & PARAM_IIR) { + if (params_changed & PARAM_IIR) { + put_bits(pb, 1, 1); + write_filter_params(ctx, pb, ch, IIR); + } else { + put_bits(pb, 1, 0); + } + } + + if (dp->param_presence_flags & PARAM_HUFFOFFSET) { + if (params_changed & PARAM_HUFFOFFSET) { + put_bits (pb, 1, 1); + put_sbits(pb, 15, cp->huff_offset); + } else { + put_bits(pb, 1, 0); + } + } + + put_bits(pb, 2, cp->codebook ); + put_bits(pb, 5, cp->huff_lsbs); + } else { + put_bits(pb, 1, 0); + } + } +} + +/** Writes the residuals to the bitstream. That is, the VLC codes from the + * codebooks (if any is used), and then the residual. + */ +static void write_block_data(MLPEncodeContext *ctx, PutBitContext *pb) +{ + DecodingParams *dp = ctx->cur_decoding_params; + RestartHeader *rh = ctx->cur_restart_header; + int32_t *sample_buffer = ctx->write_buffer; + int32_t sign_huff_offset[MAX_CHANNELS]; + int codebook_index [MAX_CHANNELS]; + int lsb_bits [MAX_CHANNELS]; + unsigned int i, ch; + + for (ch = rh->min_channel; ch <= rh->max_channel; ch++) { + ChannelParams *cp = &ctx->cur_channel_params[ch]; + int sign_shift; + + lsb_bits [ch] = cp->huff_lsbs - dp->quant_step_size[ch]; + codebook_index [ch] = cp->codebook - 1; + sign_huff_offset[ch] = cp->huff_offset; + + sign_shift = lsb_bits[ch] - 1; + + if (cp->codebook > 0) { + sign_huff_offset[ch] -= 7 << lsb_bits[ch]; + sign_shift += 3 - cp->codebook; + } + + /* Unsign if needed. */ + if (sign_shift >= 0) + sign_huff_offset[ch] -= 1 << sign_shift; + } + + for (i = 0; i < dp->blocksize; i++) { + for (ch = rh->min_channel; ch <= rh->max_channel; ch++) { + int32_t sample = *sample_buffer++ >> dp->quant_step_size[ch]; + + sample -= sign_huff_offset[ch]; + + if (codebook_index[ch] >= 0) { + int vlc = sample >> lsb_bits[ch]; + put_bits(pb, ff_mlp_huffman_tables[codebook_index[ch]][vlc][1], + ff_mlp_huffman_tables[codebook_index[ch]][vlc][0]); + } + + put_sbits(pb, lsb_bits[ch], sample); + } + sample_buffer += 2; /* noise channels */ + } + + ctx->write_buffer = sample_buffer; +} + +/** Writes the substreams data to the bitstream. */ +static uint8_t *write_substrs(MLPEncodeContext *ctx, uint8_t *buf, int buf_size, + int restart_frame, + uint16_t substream_data_len[MAX_SUBSTREAMS]) +{ + int32_t *lossless_check_data = ctx->lossless_check_data; + unsigned int substr; + int end = 0; + + lossless_check_data += ctx->frame_index * ctx->num_substreams; + + for (substr = 0; substr < ctx->num_substreams; substr++) { + unsigned int cur_subblock_index = ctx->major_cur_subblock_index; + unsigned int num_subblocks = ctx->major_filter_state_subblock; + unsigned int subblock; + RestartHeader *rh = &ctx->restart_header [substr]; + int substr_restart_frame = restart_frame; + uint8_t parity, checksum; + PutBitContext pb, tmpb; + int params_changed; + + ctx->cur_restart_header = rh; + + init_put_bits(&pb, buf, buf_size); + + for (subblock = 0; subblock <= num_subblocks; subblock++) { + unsigned int subblock_index; + + subblock_index = cur_subblock_index++; + + ctx->cur_decoding_params = &ctx->major_decoding_params[subblock_index][substr]; + ctx->cur_channel_params = ctx->major_channel_params[subblock_index]; + + params_changed = ctx->major_params_changed[subblock_index][substr]; + + if (substr_restart_frame || params_changed) { + put_bits(&pb, 1, 1); + + if (substr_restart_frame) { + put_bits(&pb, 1, 1); + + write_restart_header(ctx, &pb); + rh->lossless_check_data = 0; + } else { + put_bits(&pb, 1, 0); + } + + write_decoding_params(ctx, &pb, params_changed); + } else { + put_bits(&pb, 1, 0); + } + + write_block_data(ctx, &pb); + + put_bits(&pb, 1, !substr_restart_frame); + + substr_restart_frame = 0; + } + + put_bits(&pb, (-put_bits_count(&pb)) & 15, 0); + + rh->lossless_check_data ^= *lossless_check_data++; + + if (ctx->last_frame == ctx->inout_buffer) { + /* TODO find a sample and implement shorten_by. */ + put_bits(&pb, 32, END_OF_STREAM); + } + + /* Data must be flushed for the checksum and parity to be correct. */ + tmpb = pb; + flush_put_bits(&tmpb); + + parity = ff_mlp_calculate_parity(buf, put_bits_count(&pb) >> 3) ^ 0xa9; + checksum = ff_mlp_checksum8 (buf, put_bits_count(&pb) >> 3); + + put_bits(&pb, 8, parity ); + put_bits(&pb, 8, checksum); + + flush_put_bits(&pb); + + end += put_bits_count(&pb) >> 3; + substream_data_len[substr] = end; + + buf += put_bits_count(&pb) >> 3; + } + + ctx->major_cur_subblock_index += ctx->major_filter_state_subblock + 1; + ctx->major_filter_state_subblock = 0; + + return buf; +} + +/** Writes the access unit and substream headers to the bitstream. */ +static void write_frame_headers(MLPEncodeContext *ctx, uint8_t *frame_header, + uint8_t *substream_headers, unsigned int length, + int restart_frame, + uint16_t substream_data_len[MAX_SUBSTREAMS]) +{ + uint16_t access_unit_header = 0; + uint16_t parity_nibble = 0; + unsigned int substr; + + parity_nibble = ctx->dts; + parity_nibble ^= length; + + for (substr = 0; substr < ctx->num_substreams; substr++) { + uint16_t substr_hdr = 0; + + substr_hdr |= (0 << 15); /* extraword */ + substr_hdr |= (!restart_frame << 14); /* !restart_frame */ + substr_hdr |= (1 << 13); /* checkdata */ + substr_hdr |= (0 << 12); /* ??? */ + substr_hdr |= (substream_data_len[substr] / 2) & 0x0FFF; + + AV_WB16(substream_headers, substr_hdr); + + parity_nibble ^= *substream_headers++; + parity_nibble ^= *substream_headers++; + } + + parity_nibble ^= parity_nibble >> 8; + parity_nibble ^= parity_nibble >> 4; + parity_nibble &= 0xF; + + access_unit_header |= (parity_nibble ^ 0xF) << 12; + access_unit_header |= length & 0xFFF; + + AV_WB16(frame_header , access_unit_header); + AV_WB16(frame_header+2, ctx->dts ); +} + +/** Writes an entire access unit to the bitstream. */ +static unsigned int write_access_unit(MLPEncodeContext *ctx, uint8_t *buf, + int buf_size, int restart_frame) +{ + uint16_t substream_data_len[MAX_SUBSTREAMS]; + uint8_t *buf1, *buf0 = buf; + unsigned int substr; + int total_length; + + if (buf_size < 4) + return -1; + + /* Frame header will be written at the end. */ + buf += 4; + buf_size -= 4; + + if (restart_frame) { + if (buf_size < 28) + return -1; + write_major_sync(ctx, buf, buf_size); + buf += 28; + buf_size -= 28; + } + + buf1 = buf; + + /* Substream headers will be written at the end. */ + for (substr = 0; substr < ctx->num_substreams; substr++) { + buf += 2; + buf_size -= 2; + } + + buf = write_substrs(ctx, buf, buf_size, restart_frame, substream_data_len); + + total_length = buf - buf0; + + write_frame_headers(ctx, buf0, buf1, total_length / 2, restart_frame, substream_data_len); + + return total_length; +} + +/**************************************************************************** + ****************** Functions that input data to context ******************** + ****************************************************************************/ + +/** Inputs data from the samples passed by lavc into the context, shifts them + * appropriately depending on the bit-depth, and calculates the + * lossless_check_data that will be written to the restart header. + */ +static void input_data_internal(MLPEncodeContext *ctx, const uint8_t *samples, + int is24) +{ + int32_t *lossless_check_data = ctx->lossless_check_data; + const int32_t *samples_32 = (const int32_t *) samples; + const int16_t *samples_16 = (const int16_t *) samples; + unsigned int substr; + + lossless_check_data += ctx->frame_index * ctx->num_substreams; + + for (substr = 0; substr < ctx->num_substreams; substr++) { + RestartHeader *rh = &ctx->restart_header [substr]; + int32_t *sample_buffer = ctx->inout_buffer; + int32_t temp_lossless_check_data = 0; + uint32_t greatest = 0; + unsigned int channel; + int i; + + for (i = 0; i < ctx->frame_size[ctx->frame_index]; i++) { + for (channel = 0; channel <= rh->max_channel; channel++) { + uint32_t abs_sample; + int32_t sample; + + sample = is24 ? *samples_32++ >> 8 : *samples_16++ << 8; + + /* TODO Find out if number_sbits can be used for negative values. */ + abs_sample = FFABS(sample); + if (greatest < abs_sample) + greatest = abs_sample; + + temp_lossless_check_data ^= (sample & 0x00ffffff) << channel; + *sample_buffer++ = sample; + } + + sample_buffer += 2; /* noise channels */ + } + + ctx->max_output_bits[ctx->frame_index] = number_sbits(greatest); + + *lossless_check_data++ = temp_lossless_check_data; + } +} + +/** Wrapper function for inputting data in two different bit-depths. */ +static void input_data(MLPEncodeContext *ctx, void *samples) +{ + if (ctx->avctx->sample_fmt == AV_SAMPLE_FMT_S32) + input_data_internal(ctx, samples, 1); + else + input_data_internal(ctx, samples, 0); +} + +static void input_to_sample_buffer(MLPEncodeContext *ctx) +{ + int32_t *sample_buffer = ctx->sample_buffer; + unsigned int index; + + for (index = 0; index < ctx->number_of_frames; index++) { + unsigned int cur_index = (ctx->starting_frame_index + index) % ctx->max_restart_interval; + int32_t *input_buffer = ctx->inout_buffer + cur_index * ctx->one_sample_buffer_size; + unsigned int i, channel; + + for (i = 0; i < ctx->frame_size[cur_index]; i++) { + for (channel = 0; channel < ctx->avctx->channels; channel++) + *sample_buffer++ = *input_buffer++; + sample_buffer += 2; /* noise_channels */ + input_buffer += 2; /* noise_channels */ + } + } +} + +/**************************************************************************** + ********* Functions that analyze the data and set the parameters *********** + ****************************************************************************/ + +/** Counts the number of trailing zeroes in a value */ +static int number_trailing_zeroes(int32_t sample) +{ + int bits; + + for (bits = 0; bits < 24 && !(sample & (1<<bits)); bits++); + + /* All samples are 0. TODO Return previous quant_step_size to avoid + * writing a new header. */ + if (bits == 24) + return 0; + + return bits; +} + +/** Determines how many bits are zero at the end of all samples so they can be + * shifted out. + */ +static void determine_quant_step_size(MLPEncodeContext *ctx) +{ + DecodingParams *dp = ctx->cur_decoding_params; + RestartHeader *rh = ctx->cur_restart_header; + MatrixParams *mp = &dp->matrix_params; + int32_t *sample_buffer = ctx->sample_buffer; + int32_t sample_mask[MAX_CHANNELS]; + unsigned int channel; + int i; + + memset(sample_mask, 0x00, sizeof(sample_mask)); + + for (i = 0; i < ctx->number_of_samples; i++) { + for (channel = 0; channel <= rh->max_channel; channel++) + sample_mask[channel] |= *sample_buffer++; + + sample_buffer += 2; /* noise channels */ + } + + for (channel = 0; channel <= rh->max_channel; channel++) + dp->quant_step_size[channel] = number_trailing_zeroes(sample_mask[channel]) - mp->shift[channel]; +} + +/** Determines the smallest number of bits needed to encode the filter + * coefficients, and if it's possible to right-shift their values without + * losing any precision. + */ +static void code_filter_coeffs(MLPEncodeContext *ctx, FilterParams *fp, int32_t *fcoeff) +{ + int min = INT_MAX, max = INT_MIN; + int bits, shift; + int coeff_mask = 0; + int order; + + for (order = 0; order < fp->order; order++) { + int coeff = fcoeff[order]; + + if (coeff < min) + min = coeff; + if (coeff > max) + max = coeff; + + coeff_mask |= coeff; + } + + bits = FFMAX(number_sbits(min), number_sbits(max)); + + for (shift = 0; shift < 7 && bits + shift < 16 && !(coeff_mask & (1<<shift)); shift++); + + fp->coeff_bits = bits; + fp->coeff_shift = shift; +} + +/** Determines the best filter parameters for the given data and writes the + * necessary information to the context. + * TODO Add IIR filter predictor! + */ +static void set_filter_params(MLPEncodeContext *ctx, + unsigned int channel, unsigned int filter, + int clear_filter) +{ + ChannelParams *cp = &ctx->cur_channel_params[channel]; + FilterParams *fp = &cp->filter_params[filter]; + + if ((filter == IIR && ctx->substream_info & SUBSTREAM_INFO_HIGH_RATE) || + clear_filter) { + fp->order = 0; + } else if (filter == IIR) { + fp->order = 0; + } else if (filter == FIR) { + const int max_order = (ctx->substream_info & SUBSTREAM_INFO_HIGH_RATE) + ? 4 : MLP_MAX_LPC_ORDER; + int32_t *sample_buffer = ctx->sample_buffer + channel; + int32_t coefs[MAX_LPC_ORDER][MAX_LPC_ORDER]; + int32_t *lpc_samples = ctx->lpc_sample_buffer; + int32_t *fcoeff = ctx->cur_channel_params[channel].coeff[filter]; + int shift[MLP_MAX_LPC_ORDER]; + unsigned int i; + int order; + + for (i = 0; i < ctx->number_of_samples; i++) { + *lpc_samples++ = *sample_buffer; + sample_buffer += ctx->num_channels; + } + + order = ff_lpc_calc_coefs(&ctx->lpc_ctx, ctx->lpc_sample_buffer, + ctx->number_of_samples, MLP_MIN_LPC_ORDER, + max_order, 11, coefs, shift, FF_LPC_TYPE_LEVINSON, 0, + ORDER_METHOD_EST, MLP_MIN_LPC_SHIFT, + MLP_MAX_LPC_SHIFT, MLP_MIN_LPC_SHIFT); + + fp->order = order; + fp->shift = shift[order-1]; + + for (i = 0; i < order; i++) + fcoeff[i] = coefs[order-1][i]; + + code_filter_coeffs(ctx, fp, fcoeff); + } +} + +/** Tries to determine a good prediction filter, and applies it to the samples + * buffer if the filter is good enough. Sets the filter data to be cleared if + * no good filter was found. + */ +static void determine_filters(MLPEncodeContext *ctx) +{ + RestartHeader *rh = ctx->cur_restart_header; + int channel, filter; + + for (channel = rh->min_channel; channel <= rh->max_channel; channel++) { + for (filter = 0; filter < NUM_FILTERS; filter++) + set_filter_params(ctx, channel, filter, 0); + } +} + +enum MLPChMode { + MLP_CHMODE_LEFT_RIGHT, + MLP_CHMODE_LEFT_SIDE, + MLP_CHMODE_RIGHT_SIDE, + MLP_CHMODE_MID_SIDE, +}; + +static enum MLPChMode estimate_stereo_mode(MLPEncodeContext *ctx) +{ + uint64_t score[4], sum[4] = { 0, 0, 0, 0, }; + int32_t *right_ch = ctx->sample_buffer + 1; + int32_t *left_ch = ctx->sample_buffer; + int i; + enum MLPChMode best = 0; + + for(i = 2; i < ctx->number_of_samples; i++) { + int32_t left = left_ch [i * ctx->num_channels] - 2 * left_ch [(i - 1) * ctx->num_channels] + left_ch [(i - 2) * ctx->num_channels]; + int32_t right = right_ch[i * ctx->num_channels] - 2 * right_ch[(i - 1) * ctx->num_channels] + right_ch[(i - 2) * ctx->num_channels]; + + sum[0] += FFABS( left ); + sum[1] += FFABS( right); + sum[2] += FFABS((left + right) >> 1); + sum[3] += FFABS( left - right); + } + + score[MLP_CHMODE_LEFT_RIGHT] = sum[0] + sum[1]; + score[MLP_CHMODE_LEFT_SIDE] = sum[0] + sum[3]; + score[MLP_CHMODE_RIGHT_SIDE] = sum[1] + sum[3]; + score[MLP_CHMODE_MID_SIDE] = sum[2] + sum[3]; + + for(i = 1; i < 3; i++) + if(score[i] < score[best]) + best = i; + + return best; +} + +/** Determines how many fractional bits are needed to encode matrix + * coefficients. Also shifts the coefficients to fit within 2.14 bits. + */ +static void code_matrix_coeffs(MLPEncodeContext *ctx, unsigned int mat) +{ + DecodingParams *dp = ctx->cur_decoding_params; + MatrixParams *mp = &dp->matrix_params; + int32_t coeff_mask = 0; + unsigned int channel; + unsigned int bits; + + for (channel = 0; channel < ctx->num_channels; channel++) { + int32_t coeff = mp->coeff[mat][channel]; + coeff_mask |= coeff; + } + + for (bits = 0; bits < 14 && !(coeff_mask & (1<<bits)); bits++); + + mp->fbits [mat] = 14 - bits; +} + +/** Determines best coefficients to use for the lossless matrix. */ +static void lossless_matrix_coeffs(MLPEncodeContext *ctx) +{ + DecodingParams *dp = ctx->cur_decoding_params; + MatrixParams *mp = &dp->matrix_params; + unsigned int shift = 0; + unsigned int channel; + int mat; + enum MLPChMode mode; + + /* No decorrelation for non-stereo. */ + if (ctx->num_channels - 2 != 2) { + mp->count = 0; + return; + } + + mode = estimate_stereo_mode(ctx); + + switch(mode) { + /* TODO: add matrix for MID_SIDE */ + case MLP_CHMODE_MID_SIDE: + case MLP_CHMODE_LEFT_RIGHT: + mp->count = 0; + break; + case MLP_CHMODE_LEFT_SIDE: + mp->count = 1; + mp->outch[0] = 1; + mp->coeff[0][0] = 1 << 14; mp->coeff[0][1] = -(1 << 14); + mp->coeff[0][2] = 0 << 14; mp->coeff[0][2] = 0 << 14; + mp->forco[0][0] = 1 << 14; mp->forco[0][1] = -(1 << 14); + mp->forco[0][2] = 0 << 14; mp->forco[0][2] = 0 << 14; + break; + case MLP_CHMODE_RIGHT_SIDE: + mp->count = 1; + mp->outch[0] = 0; + mp->coeff[0][0] = 1 << 14; mp->coeff[0][1] = 1 << 14; + mp->coeff[0][2] = 0 << 14; mp->coeff[0][2] = 0 << 14; + mp->forco[0][0] = 1 << 14; mp->forco[0][1] = -(1 << 14); + mp->forco[0][2] = 0 << 14; mp->forco[0][2] = 0 << 14; + break; + } + + for (mat = 0; mat < mp->count; mat++) + code_matrix_coeffs(ctx, mat); + + for (channel = 0; channel < ctx->num_channels; channel++) + mp->shift[channel] = shift; +} + +/** Min and max values that can be encoded with each codebook. The values for + * the third codebook take into account the fact that the sign shift for this + * codebook is outside the coded value, so it has one more bit of precision. + * It should actually be -7 -> 7, shifted down by 0.5. + */ +static const int codebook_extremes[3][2] = { + {-9, 8}, {-8, 7}, {-15, 14}, +}; + +/** Determines the amount of bits needed to encode the samples using no + * codebooks and a specified offset. + */ +static void no_codebook_bits_offset(MLPEncodeContext *ctx, + unsigned int channel, int16_t offset, + int32_t min, int32_t max, + BestOffset *bo) +{ + DecodingParams *dp = ctx->cur_decoding_params; + int32_t unsign; + int lsb_bits; + + min -= offset; + max -= offset; + + lsb_bits = FFMAX(number_sbits(min), number_sbits(max)) - 1; + + lsb_bits += !!lsb_bits; + + unsign = 1 << (lsb_bits - 1); + + bo->offset = offset; + bo->lsb_bits = lsb_bits; + bo->bitcount = lsb_bits * dp->blocksize; + bo->min = offset - unsign + 1; + bo->max = offset + unsign; +} + +/** Determines the least amount of bits needed to encode the samples using no + * codebooks. + */ +static void no_codebook_bits(MLPEncodeContext *ctx, + unsigned int channel, + int32_t min, int32_t max, + BestOffset *bo) +{ + DecodingParams *dp = ctx->cur_decoding_params; + int16_t offset; + int32_t unsign; + uint32_t diff; + int lsb_bits; + + /* Set offset inside huffoffset's boundaries by adjusting extremes + * so that more bits are used, thus shifting the offset. */ + if (min < HUFF_OFFSET_MIN) + max = FFMAX(max, 2 * HUFF_OFFSET_MIN - min + 1); + if (max > HUFF_OFFSET_MAX) + min = FFMIN(min, 2 * HUFF_OFFSET_MAX - max - 1); + + /* Determine offset and minimum number of bits. */ + diff = max - min; + + lsb_bits = number_sbits(diff) - 1; + + unsign = 1 << (lsb_bits - 1); + + /* If all samples are the same (lsb_bits == 0), offset must be + * adjusted because of sign_shift. */ + offset = min + diff / 2 + !!lsb_bits; + + bo->offset = offset; + bo->lsb_bits = lsb_bits; + bo->bitcount = lsb_bits * dp->blocksize; + bo->min = max - unsign + 1; + bo->max = min + unsign; +} + +/** Determines the least amount of bits needed to encode the samples using a + * given codebook and a given offset. + */ +static inline void codebook_bits_offset(MLPEncodeContext *ctx, + unsigned int channel, int codebook, + int32_t sample_min, int32_t sample_max, + int16_t offset, BestOffset *bo) +{ + int32_t codebook_min = codebook_extremes[codebook][0]; + int32_t codebook_max = codebook_extremes[codebook][1]; + int32_t *sample_buffer = ctx->sample_buffer + channel; + DecodingParams *dp = ctx->cur_decoding_params; + int codebook_offset = 7 + (2 - codebook); + int32_t unsign_offset = offset; + int lsb_bits = 0, bitcount = 0; + int offset_min = INT_MAX, offset_max = INT_MAX; + int unsign, mask; + int i; + + sample_min -= offset; + sample_max -= offset; + + while (sample_min < codebook_min || sample_max > codebook_max) { + lsb_bits++; + sample_min >>= 1; + sample_max >>= 1; + } + + unsign = 1 << lsb_bits; + mask = unsign - 1; + + if (codebook == 2) { + unsign_offset -= unsign; + lsb_bits++; + } + + for (i = 0; i < dp->blocksize; i++) { + int32_t sample = *sample_buffer >> dp->quant_step_size[channel]; + int temp_min, temp_max; + + sample -= unsign_offset; + + temp_min = sample & mask; + if (temp_min < offset_min) + offset_min = temp_min; + + temp_max = unsign - temp_min - 1; + if (temp_max < offset_max) + offset_max = temp_max; + + sample >>= lsb_bits; + + bitcount += ff_mlp_huffman_tables[codebook][sample + codebook_offset][1]; + + sample_buffer += ctx->num_channels; + } + + bo->offset = offset; + bo->lsb_bits = lsb_bits; + bo->bitcount = lsb_bits * dp->blocksize + bitcount; + bo->min = FFMAX(offset - offset_min, HUFF_OFFSET_MIN); + bo->max = FFMIN(offset + offset_max, HUFF_OFFSET_MAX); +} + +/** Determines the least amount of bits needed to encode the samples using a + * given codebook. Searches for the best offset to minimize the bits. + */ +static inline void codebook_bits(MLPEncodeContext *ctx, + unsigned int channel, int codebook, + int offset, int32_t min, int32_t max, + BestOffset *bo, int direction) +{ + int previous_count = INT_MAX; + int offset_min, offset_max; + int is_greater = 0; + + offset_min = FFMAX(min, HUFF_OFFSET_MIN); + offset_max = FFMIN(max, HUFF_OFFSET_MAX); + + for (;;) { + BestOffset temp_bo; + + codebook_bits_offset(ctx, channel, codebook, + min, max, offset, + &temp_bo); + + if (temp_bo.bitcount < previous_count) { + if (temp_bo.bitcount < bo->bitcount) + *bo = temp_bo; + + is_greater = 0; + } else if (++is_greater >= ctx->max_codebook_search) + break; + + previous_count = temp_bo.bitcount; + + if (direction) { + offset = temp_bo.max + 1; + if (offset > offset_max) + break; + } else { + offset = temp_bo.min - 1; + if (offset < offset_min) + break; + } + } +} + +/** Determines the least amount of bits needed to encode the samples using + * any or no codebook. + */ +static void determine_bits(MLPEncodeContext *ctx) +{ + DecodingParams *dp = ctx->cur_decoding_params; + RestartHeader *rh = ctx->cur_restart_header; + unsigned int channel; + + for (channel = 0; channel <= rh->max_channel; channel++) { + ChannelParams *cp = &ctx->cur_channel_params[channel]; + int32_t *sample_buffer = ctx->sample_buffer + channel; + int32_t min = INT32_MAX, max = INT32_MIN; + int no_filters_used = !cp->filter_params[FIR].order; + int average = 0; + int offset = 0; + int i; + + /* Determine extremes and average. */ + for (i = 0; i < dp->blocksize; i++) { + int32_t sample = *sample_buffer >> dp->quant_step_size[channel]; + if (sample < min) + min = sample; + if (sample > max) + max = sample; + average += sample; + sample_buffer += ctx->num_channels; + } + average /= dp->blocksize; + + /* If filtering is used, we always set the offset to zero, otherwise + * we search for the offset that minimizes the bitcount. */ + if (no_filters_used) { + no_codebook_bits(ctx, channel, min, max, &ctx->cur_best_offset[channel][0]); + offset = av_clip(average, HUFF_OFFSET_MIN, HUFF_OFFSET_MAX); + } else { + no_codebook_bits_offset(ctx, channel, offset, min, max, &ctx->cur_best_offset[channel][0]); + } + + for (i = 1; i < NUM_CODEBOOKS; i++) { + BestOffset temp_bo = { 0, INT_MAX, 0, 0, 0, }; + int16_t offset_max; + + codebook_bits_offset(ctx, channel, i - 1, + min, max, offset, + &temp_bo); + + if (no_filters_used) { + offset_max = temp_bo.max; + + codebook_bits(ctx, channel, i - 1, temp_bo.min - 1, + min, max, &temp_bo, 0); + codebook_bits(ctx, channel, i - 1, offset_max + 1, + min, max, &temp_bo, 1); + } + + ctx->cur_best_offset[channel][i] = temp_bo; + } + } +} + +/**************************************************************************** + *************** Functions that process the data in some way **************** + ****************************************************************************/ + +#define SAMPLE_MAX(bitdepth) ((1 << (bitdepth - 1)) - 1) +#define SAMPLE_MIN(bitdepth) (~SAMPLE_MAX(bitdepth)) + +#define MSB_MASK(bits) (-1u << bits) + +/** Applies the filter to the current samples, and saves the residual back + * into the samples buffer. If the filter is too bad and overflows the + * maximum amount of bits allowed (16 or 24), the samples buffer is left as is and + * the function returns -1. + */ +static int apply_filter(MLPEncodeContext *ctx, unsigned int channel) +{ + FilterParams *fp[NUM_FILTERS] = { &ctx->cur_channel_params[channel].filter_params[FIR], + &ctx->cur_channel_params[channel].filter_params[IIR], }; + int32_t *filter_state_buffer[NUM_FILTERS]; + int32_t mask = MSB_MASK(ctx->cur_decoding_params->quant_step_size[channel]); + int32_t *sample_buffer = ctx->sample_buffer + channel; + unsigned int number_of_samples = ctx->number_of_samples; + unsigned int filter_shift = fp[FIR]->shift; + int filter; + int i; + + for (i = 0; i < NUM_FILTERS; i++) { + unsigned int size = ctx->number_of_samples; + filter_state_buffer[i] = av_malloc(size*sizeof(int32_t)); + if (!filter_state_buffer[i]) { + av_log(ctx->avctx, AV_LOG_ERROR, + "Not enough memory for applying filters.\n"); + return -1; + } + } + + for (i = 0; i < 8; i++) { + filter_state_buffer[FIR][i] = *sample_buffer; + filter_state_buffer[IIR][i] = *sample_buffer; + + sample_buffer += ctx->num_channels; + } + + for (i = 8; i < number_of_samples; i++) { + int32_t sample = *sample_buffer; + unsigned int order; + int64_t accum = 0; + int32_t residual; + + for (filter = 0; filter < NUM_FILTERS; filter++) { + int32_t *fcoeff = ctx->cur_channel_params[channel].coeff[filter]; + for (order = 0; order < fp[filter]->order; order++) + accum += (int64_t)filter_state_buffer[filter][i - 1 - order] * + fcoeff[order]; + } + + accum >>= filter_shift; + residual = sample - (accum & mask); + + if (residual < SAMPLE_MIN(ctx->wordlength) || residual > SAMPLE_MAX(ctx->wordlength)) + return -1; + + filter_state_buffer[FIR][i] = sample; + filter_state_buffer[IIR][i] = residual; + + sample_buffer += ctx->num_channels; + } + + sample_buffer = ctx->sample_buffer + channel; + for (i = 0; i < number_of_samples; i++) { + *sample_buffer = filter_state_buffer[IIR][i]; + + sample_buffer += ctx->num_channels; + } + + for (i = 0; i < NUM_FILTERS; i++) { + av_freep(&filter_state_buffer[i]); + } + + return 0; +} + +static void apply_filters(MLPEncodeContext *ctx) +{ + RestartHeader *rh = ctx->cur_restart_header; + int channel; + + for (channel = rh->min_channel; channel <= rh->max_channel; channel++) { + if (apply_filter(ctx, channel) < 0) { + /* Filter is horribly wrong. + * Clear filter params and update state. */ + set_filter_params(ctx, channel, FIR, 1); + set_filter_params(ctx, channel, IIR, 1); + apply_filter(ctx, channel); + } + } +} + +/** Generates two noise channels worth of data. */ +static void generate_2_noise_channels(MLPEncodeContext *ctx) +{ + int32_t *sample_buffer = ctx->sample_buffer + ctx->num_channels - 2; + RestartHeader *rh = ctx->cur_restart_header; + unsigned int i; + uint32_t seed = rh->noisegen_seed; + + for (i = 0; i < ctx->number_of_samples; i++) { + uint16_t seed_shr7 = seed >> 7; + *sample_buffer++ = ((int8_t)(seed >> 15)) << rh->noise_shift; + *sample_buffer++ = ((int8_t) seed_shr7) << rh->noise_shift; + + seed = (seed << 16) ^ seed_shr7 ^ (seed_shr7 << 5); + + sample_buffer += ctx->num_channels - 2; + } + + rh->noisegen_seed = seed & ((1 << 24)-1); +} + +/** Rematrixes all channels using chosen coefficients. */ +static void rematrix_channels(MLPEncodeContext *ctx) +{ + DecodingParams *dp = ctx->cur_decoding_params; + MatrixParams *mp = &dp->matrix_params; + int32_t *sample_buffer = ctx->sample_buffer; + unsigned int mat, i, maxchan; + + maxchan = ctx->num_channels; + + for (mat = 0; mat < mp->count; mat++) { + unsigned int msb_mask_bits = (ctx->avctx->sample_fmt == AV_SAMPLE_FMT_S16 ? 8 : 0) - mp->shift[mat]; + int32_t mask = MSB_MASK(msb_mask_bits); + unsigned int outch = mp->outch[mat]; + + sample_buffer = ctx->sample_buffer; + for (i = 0; i < ctx->number_of_samples; i++) { + unsigned int src_ch; + int64_t accum = 0; + + for (src_ch = 0; src_ch < maxchan; src_ch++) { + int32_t sample = *(sample_buffer + src_ch); + accum += (int64_t) sample * mp->forco[mat][src_ch]; + } + sample_buffer[outch] = (accum >> 14) & mask; + + sample_buffer += ctx->num_channels; + } + } +} + +/**************************************************************************** + **** Functions that deal with determining the best parameters and output *** + ****************************************************************************/ + +typedef struct { + char path[MAJOR_HEADER_INTERVAL + 3]; + int bitcount; +} PathCounter; + +static const char *path_counter_codebook[] = { "0", "1", "2", "3", }; + +#define ZERO_PATH '0' +#define CODEBOOK_CHANGE_BITS 21 + +static void clear_path_counter(PathCounter *path_counter) +{ + unsigned int i; + + for (i = 0; i < NUM_CODEBOOKS + 1; i++) { + path_counter[i].path[0] = ZERO_PATH; + path_counter[i].path[1] = 0x00; + path_counter[i].bitcount = 0; + } +} + +static int compare_best_offset(BestOffset *prev, BestOffset *cur) +{ + if (prev->lsb_bits != cur->lsb_bits) + return 1; + + return 0; +} + +static int best_codebook_path_cost(MLPEncodeContext *ctx, unsigned int channel, + PathCounter *src, int cur_codebook) +{ + BestOffset *cur_bo, *prev_bo = restart_best_offset; + int bitcount = src->bitcount; + char *path = src->path + 1; + int prev_codebook; + int i; + + for (i = 0; path[i]; i++) + prev_bo = ctx->best_offset[i][channel]; + + prev_codebook = path[i - 1] - ZERO_PATH; + + cur_bo = ctx->best_offset[i][channel]; + + bitcount += cur_bo[cur_codebook].bitcount; + + if (prev_codebook != cur_codebook || + compare_best_offset(&prev_bo[prev_codebook], &cur_bo[cur_codebook])) + bitcount += CODEBOOK_CHANGE_BITS; + + return bitcount; +} + +static void set_best_codebook(MLPEncodeContext *ctx) +{ + DecodingParams *dp = ctx->cur_decoding_params; + RestartHeader *rh = ctx->cur_restart_header; + unsigned int channel; + + for (channel = rh->min_channel; channel <= rh->max_channel; channel++) { + BestOffset *cur_bo, *prev_bo = restart_best_offset; + PathCounter path_counter[NUM_CODEBOOKS + 1]; + unsigned int best_codebook; + unsigned int index; + char *best_path; + + clear_path_counter(path_counter); + + for (index = 0; index < ctx->number_of_subblocks; index++) { + unsigned int best_bitcount = INT_MAX; + unsigned int codebook; + + cur_bo = ctx->best_offset[index][channel]; + + for (codebook = 0; codebook < NUM_CODEBOOKS; codebook++) { + int prev_best_bitcount = INT_MAX; + int last_best; + + for (last_best = 0; last_best < 2; last_best++) { + PathCounter *dst_path = &path_counter[codebook]; + PathCounter *src_path; + int temp_bitcount; + + /* First test last path with same headers, + * then with last best. */ + if (last_best) { + src_path = &path_counter[NUM_CODEBOOKS]; + } else { + if (compare_best_offset(&prev_bo[codebook], &cur_bo[codebook])) + continue; + else + src_path = &path_counter[codebook]; + } + + temp_bitcount = best_codebook_path_cost(ctx, channel, src_path, codebook); + + if (temp_bitcount < best_bitcount) { + best_bitcount = temp_bitcount; + best_codebook = codebook; + } + + if (temp_bitcount < prev_best_bitcount) { + prev_best_bitcount = temp_bitcount; + if (src_path != dst_path) + memcpy(dst_path, src_path, sizeof(PathCounter)); + av_strlcat(dst_path->path, path_counter_codebook[codebook], sizeof(dst_path->path)); + dst_path->bitcount = temp_bitcount; + } + } + } + + prev_bo = cur_bo; + + memcpy(&path_counter[NUM_CODEBOOKS], &path_counter[best_codebook], sizeof(PathCounter)); + } + + best_path = path_counter[NUM_CODEBOOKS].path + 1; + + /* Update context. */ + for (index = 0; index < ctx->number_of_subblocks; index++) { + ChannelParams *cp = ctx->seq_channel_params + index*(ctx->avctx->channels) + channel; + + best_codebook = *best_path++ - ZERO_PATH; + cur_bo = &ctx->best_offset[index][channel][best_codebook]; + + cp->huff_offset = cur_bo->offset; + cp->huff_lsbs = cur_bo->lsb_bits + dp->quant_step_size[channel]; + cp->codebook = best_codebook; + } + } +} + +/** Analyzes all collected bitcounts and selects the best parameters for each + * individual access unit. + * TODO This is just a stub! + */ +static void set_major_params(MLPEncodeContext *ctx) +{ + RestartHeader *rh = ctx->cur_restart_header; + unsigned int index; + unsigned int substr; + uint8_t max_huff_lsbs = 0; + uint8_t max_output_bits = 0; + + for (substr = 0; substr < ctx->num_substreams; substr++) { + DecodingParams *seq_dp = (DecodingParams *) ctx->decoding_params+ + (ctx->restart_intervals - 1)*(ctx->sequence_size)*(ctx->avctx->channels) + + (ctx->seq_offset[ctx->restart_intervals - 1])*(ctx->avctx->channels); + + ChannelParams *seq_cp = (ChannelParams *) ctx->channel_params + + (ctx->restart_intervals - 1)*(ctx->sequence_size)*(ctx->avctx->channels) + + (ctx->seq_offset[ctx->restart_intervals - 1])*(ctx->avctx->channels); + unsigned int channel; + for (index = 0; index < ctx->seq_size[ctx->restart_intervals-1]; index++) { + memcpy(&ctx->major_decoding_params[index][substr], seq_dp + index*(ctx->num_substreams) + substr, sizeof(DecodingParams)); + for (channel = 0; channel < ctx->avctx->channels; channel++) { + uint8_t huff_lsbs = (seq_cp + index*(ctx->avctx->channels) + channel)->huff_lsbs; + if (max_huff_lsbs < huff_lsbs) + max_huff_lsbs = huff_lsbs; + memcpy(&ctx->major_channel_params[index][channel], + (seq_cp + index*(ctx->avctx->channels) + channel), + sizeof(ChannelParams)); + } + } + } + + rh->max_huff_lsbs = max_huff_lsbs; + + for (index = 0; index < ctx->number_of_frames; index++) + if (max_output_bits < ctx->max_output_bits[index]) + max_output_bits = ctx->max_output_bits[index]; + rh->max_output_bits = max_output_bits; + + for (substr = 0; substr < ctx->num_substreams; substr++) { + + ctx->cur_restart_header = &ctx->restart_header[substr]; + + ctx->prev_decoding_params = &restart_decoding_params[substr]; + ctx->prev_channel_params = restart_channel_params; + + for (index = 0; index < MAJOR_HEADER_INTERVAL + 1; index++) { + ctx->cur_decoding_params = &ctx->major_decoding_params[index][substr]; + ctx->cur_channel_params = ctx->major_channel_params[index]; + + ctx->major_params_changed[index][substr] = compare_decoding_params(ctx); + + ctx->prev_decoding_params = ctx->cur_decoding_params; + ctx->prev_channel_params = ctx->cur_channel_params; + } + } + + ctx->major_number_of_subblocks = ctx->number_of_subblocks; + ctx->major_filter_state_subblock = 1; + ctx->major_cur_subblock_index = 0; +} + +static void analyze_sample_buffer(MLPEncodeContext *ctx) +{ + ChannelParams *seq_cp = ctx->seq_channel_params; + DecodingParams *seq_dp = ctx->seq_decoding_params; + unsigned int index; + unsigned int substr; + + for (substr = 0; substr < ctx->num_substreams; substr++) { + + ctx->cur_restart_header = &ctx->restart_header[substr]; + ctx->cur_decoding_params = seq_dp + 1*(ctx->num_substreams) + substr; + ctx->cur_channel_params = seq_cp + 1*(ctx->avctx->channels); + + determine_quant_step_size(ctx); + generate_2_noise_channels(ctx); + lossless_matrix_coeffs (ctx); + rematrix_channels (ctx); + determine_filters (ctx); + apply_filters (ctx); + + copy_restart_frame_params(ctx, substr); + + /* Copy frame_size from frames 0...max to decoding_params 1...max + 1 + * decoding_params[0] is for the filter state subblock. + */ + for (index = 0; index < ctx->number_of_frames; index++) { + DecodingParams *dp = seq_dp + (index + 1)*(ctx->num_substreams) + substr; + dp->blocksize = ctx->frame_size[index]; + } + /* The official encoder seems to always encode a filter state subblock + * even if there are no filters. TODO check if it is possible to skip + * the filter state subblock for no filters. + */ + (seq_dp + substr)->blocksize = 8; + (seq_dp + 1*(ctx->num_substreams) + substr)->blocksize -= 8; + + for (index = 0; index < ctx->number_of_subblocks; index++) { + ctx->cur_decoding_params = seq_dp + index*(ctx->num_substreams) + substr; + ctx->cur_channel_params = seq_cp + index*(ctx->avctx->channels); + ctx->cur_best_offset = ctx->best_offset[index]; + determine_bits(ctx); + ctx->sample_buffer += ctx->cur_decoding_params->blocksize * ctx->num_channels; + } + + set_best_codebook(ctx); + } +} + +static void process_major_frame(MLPEncodeContext *ctx) +{ + unsigned int substr; + + ctx->sample_buffer = ctx->major_inout_buffer; + + ctx->starting_frame_index = 0; + ctx->number_of_frames = ctx->major_number_of_frames; + ctx->number_of_samples = ctx->major_frame_size; + + for (substr = 0; substr < ctx->num_substreams; substr++) { + RestartHeader *rh = ctx->cur_restart_header; + unsigned int channel; + + ctx->cur_restart_header = &ctx->restart_header[substr]; + + ctx->cur_decoding_params = &ctx->major_decoding_params[1][substr]; + ctx->cur_channel_params = ctx->major_channel_params[1]; + + generate_2_noise_channels(ctx); + rematrix_channels (ctx); + + for (channel = rh->min_channel; channel <= rh->max_channel; channel++) + apply_filter(ctx, channel); + } +} + +/****************************************************************************/ + +static int mlp_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, + const AVFrame *frame, int *got_packet) +{ + MLPEncodeContext *ctx = avctx->priv_data; + unsigned int bytes_written = 0; + int restart_frame, ret; + uint8_t *data; + + if ((ret = ff_alloc_packet2(avctx, avpkt, 87500 * avctx->channels, 0)) < 0) + return ret; + + if (!frame) + return 1; + + /* add current frame to queue */ + if (frame) { + if ((ret = ff_af_queue_add(&ctx->afq, frame)) < 0) + return ret; + } + + data = frame->data[0]; + + ctx->frame_index = avctx->frame_number % ctx->max_restart_interval; + + ctx->inout_buffer = ctx->major_inout_buffer + + ctx->frame_index * ctx->one_sample_buffer_size; + + if (ctx->last_frame == ctx->inout_buffer) { + return 0; + } + + ctx->sample_buffer = ctx->major_scratch_buffer + + ctx->frame_index * ctx->one_sample_buffer_size; + + ctx->write_buffer = ctx->inout_buffer; + + if (avctx->frame_number < ctx->max_restart_interval) { + if (data) { + goto input_and_return; + } else { + /* There are less frames than the requested major header interval. + * Update the context to reflect this. + */ + ctx->max_restart_interval = avctx->frame_number; + ctx->frame_index = 0; + + ctx->sample_buffer = ctx->major_scratch_buffer; + ctx->inout_buffer = ctx->major_inout_buffer; + } + } + + if (ctx->frame_size[ctx->frame_index] > MAX_BLOCKSIZE) { + av_log(avctx, AV_LOG_ERROR, "Invalid frame size (%d > %d)\n", + ctx->frame_size[ctx->frame_index], MAX_BLOCKSIZE); + return -1; + } + + restart_frame = !ctx->frame_index; + + if (restart_frame) { + set_major_params(ctx); + if (ctx->min_restart_interval != ctx->max_restart_interval) + process_major_frame(ctx); + } + + if (ctx->min_restart_interval == ctx->max_restart_interval) + ctx->write_buffer = ctx->sample_buffer; + + bytes_written = write_access_unit(ctx, avpkt->data, avpkt->size, restart_frame); + + ctx->timestamp += ctx->frame_size[ctx->frame_index]; + ctx->dts += ctx->frame_size[ctx->frame_index]; + +input_and_return: + + if (data) { + ctx->frame_size[ctx->frame_index] = avctx->frame_size; + ctx->next_major_frame_size += avctx->frame_size; + ctx->next_major_number_of_frames++; + input_data(ctx, data); + } else if (!ctx->last_frame) { + ctx->last_frame = ctx->inout_buffer; + } + + restart_frame = (ctx->frame_index + 1) % ctx->min_restart_interval; + + if (!restart_frame) { + int seq_index; + + for (seq_index = 0; + seq_index < ctx->restart_intervals && (seq_index * ctx->min_restart_interval) <= ctx->avctx->frame_number; + seq_index++) { + unsigned int number_of_samples = 0; + unsigned int index; + + ctx->sample_buffer = ctx->major_scratch_buffer; + ctx->inout_buffer = ctx->major_inout_buffer; + ctx->seq_index = seq_index; + + ctx->starting_frame_index = (ctx->avctx->frame_number - (ctx->avctx->frame_number % ctx->min_restart_interval) + - (seq_index * ctx->min_restart_interval)) % ctx->max_restart_interval; + ctx->number_of_frames = ctx->next_major_number_of_frames; + ctx->number_of_subblocks = ctx->next_major_number_of_frames + 1; + + ctx->seq_channel_params = (ChannelParams *) ctx->channel_params + + (ctx->frame_index / ctx->min_restart_interval)*(ctx->sequence_size)*(ctx->avctx->channels) + + (ctx->seq_offset[seq_index])*(ctx->avctx->channels); + + ctx->seq_decoding_params = (DecodingParams *) ctx->decoding_params + + (ctx->frame_index / ctx->min_restart_interval)*(ctx->sequence_size)*(ctx->num_substreams) + + (ctx->seq_offset[seq_index])*(ctx->num_substreams); + + for (index = 0; index < ctx->number_of_frames; index++) { + number_of_samples += ctx->frame_size[(ctx->starting_frame_index + index) % ctx->max_restart_interval]; + } + ctx->number_of_samples = number_of_samples; + + for (index = 0; index < ctx->seq_size[seq_index]; index++) { + clear_channel_params(ctx, ctx->seq_channel_params + index*(ctx->avctx->channels)); + default_decoding_params(ctx, ctx->seq_decoding_params + index*(ctx->num_substreams)); + } + + input_to_sample_buffer(ctx); + + analyze_sample_buffer(ctx); + } + + if (ctx->frame_index == (ctx->max_restart_interval - 1)) { + ctx->major_frame_size = ctx->next_major_frame_size; + ctx->next_major_frame_size = 0; + ctx->major_number_of_frames = ctx->next_major_number_of_frames; + ctx->next_major_number_of_frames = 0; + + if (!ctx->major_frame_size) + goto no_data_left; + } + } + +no_data_left: + + ff_af_queue_remove(&ctx->afq, avctx->frame_size, &avpkt->pts, + &avpkt->duration); + avpkt->size = bytes_written; + *got_packet = 1; + return 0; +} + +static av_cold int mlp_encode_close(AVCodecContext *avctx) +{ + MLPEncodeContext *ctx = avctx->priv_data; + + ff_lpc_end(&ctx->lpc_ctx); + + av_freep(&ctx->lossless_check_data); + av_freep(&ctx->major_scratch_buffer); + av_freep(&ctx->major_inout_buffer); + av_freep(&ctx->lpc_sample_buffer); + av_freep(&ctx->decoding_params); + av_freep(&ctx->channel_params); + av_freep(&ctx->frame_size); + ff_af_queue_close(&ctx->afq); + + return 0; +} + +#if CONFIG_MLP_ENCODER +AVCodec ff_mlp_encoder = { + .name ="mlp", + .long_name = NULL_IF_CONFIG_SMALL("MLP (Meridian Lossless Packing)"), + .type = AVMEDIA_TYPE_AUDIO, + .id = AV_CODEC_ID_MLP, + .priv_data_size = sizeof(MLPEncodeContext), + .init = mlp_encode_init, + .encode2 = mlp_encode_frame, + .close = mlp_encode_close, + .capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY | AV_CODEC_CAP_EXPERIMENTAL, + .sample_fmts = (const enum AVSampleFormat[]) {AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE}, + .supported_samplerates = (const int[]) {44100, 48000, 88200, 96000, 176400, 192000, 0}, + .channel_layouts = ff_mlp_channel_layouts, +}; +#endif +#if CONFIG_TRUEHD_ENCODER +AVCodec ff_truehd_encoder = { + .name ="truehd", + .long_name = NULL_IF_CONFIG_SMALL("TrueHD"), + .type = AVMEDIA_TYPE_AUDIO, + .id = AV_CODEC_ID_TRUEHD, + .priv_data_size = sizeof(MLPEncodeContext), + .init = mlp_encode_init, + .encode2 = mlp_encode_frame, + .close = mlp_encode_close, + .capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY | AV_CODEC_CAP_EXPERIMENTAL, + .sample_fmts = (const enum AVSampleFormat[]) {AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE}, + .supported_samplerates = (const int[]) {44100, 48000, 88200, 96000, 176400, 192000, 0}, + .channel_layouts = (const uint64_t[]) {AV_CH_LAYOUT_STEREO, AV_CH_LAYOUT_5POINT0_BACK, AV_CH_LAYOUT_5POINT1_BACK, 0}, +}; +#endif -- 2.9.2 -- Jai Luthra

4 7

[PATCH 1/2] avformat/matroskaenc: postpone writing the Tracks master
by James Almer 26 Dec '16

26 Dec '16

This will allow us to write updated stream information not available during write_header(). Signed-off-by: James Almer <jamrial(a)gmail.com> --- libavformat/matroskaenc.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-) diff --git a/libavformat/matroskaenc.c b/libavformat/matroskaenc.c index 78540fb..f41aabe 100644 --- a/libavformat/matroskaenc.c +++ b/libavformat/matroskaenc.c @@ -121,6 +121,8 @@ typedef struct MatroskaMuxContext { ebml_master tags; AVIOContext *info_bc; ebml_master info; + AVIOContext *tracks_bc; + ebml_master tracks_master; ebml_master segment; int64_t segment_offset; ebml_master cluster; @@ -1264,15 +1266,14 @@ static int mkv_write_track(AVFormatContext *s, MatroskaMuxContext *mkv, static int mkv_write_tracks(AVFormatContext *s) { MatroskaMuxContext *mkv = s->priv_data; - AVIOContext *dyn_cp, *pb = s->pb; - ebml_master tracks; + AVIOContext *pb = s->pb; int i, ret, default_stream_exists = 0; ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_TRACKS, avio_tell(pb)); if (ret < 0) return ret; - ret = start_ebml_master_crc32(pb, &dyn_cp, mkv, &tracks, MATROSKA_ID_TRACKS, 0); + ret = start_ebml_master_crc32(pb, &mkv->tracks_bc, mkv, &mkv->tracks_master, MATROSKA_ID_TRACKS, 0); if (ret < 0) return ret; @@ -1281,11 +1282,16 @@ static int mkv_write_tracks(AVFormatContext *s) default_stream_exists |= st->disposition & AV_DISPOSITION_DEFAULT; } for (i = 0; i < s->nb_streams; i++) { - ret = mkv_write_track(s, mkv, i, dyn_cp, default_stream_exists); + ret = mkv_write_track(s, mkv, i, mkv->tracks_bc, default_stream_exists); if (ret < 0) return ret; } - end_ebml_master_crc32(pb, &dyn_cp, mkv, tracks); + + if (pb->seekable && !mkv->is_live) + put_ebml_void(pb, avio_tell(mkv->tracks_bc)); + else + end_ebml_master_crc32(pb, &mkv->tracks_bc, mkv, mkv->tracks_master); + return 0; } @@ -2317,6 +2323,10 @@ static int mkv_write_trailer(AVFormatContext *s) avio_seek(pb, mkv->info.pos, SEEK_SET); end_ebml_master_crc32(pb, &mkv->info_bc, mkv, mkv->info); + // dump tracks master + avio_seek(pb, mkv->tracks_master.pos, SEEK_SET); + end_ebml_master_crc32(pb, &mkv->tracks_bc, mkv, mkv->tracks_master); + // update stream durations if (!mkv->is_live && mkv->stream_durations) { int i; -- 2.10.1

2 5

[PATCH 1/2] lavfi: add FFFrameQueue API.
by Nicolas George 23 Dec '16

23 Dec '16

Signed-off-by: Nicolas George <george(a)nsup.org> --- libavfilter/Makefile | 1 + libavfilter/framequeue.c | 123 +++++++++++++++++++++++++++++++++ libavfilter/framequeue.h | 173 +++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 297 insertions(+) create mode 100644 libavfilter/framequeue.c create mode 100644 libavfilter/framequeue.h Unchanged. diff --git a/libavfilter/Makefile b/libavfilter/Makefile index cdddb1b..35bf11b 100644 --- a/libavfilter/Makefile +++ b/libavfilter/Makefile @@ -18,6 +18,7 @@ OBJS = allfilters.o \ fifo.o \ formats.o \ framepool.o \ + framequeue.o \ graphdump.o \ graphparser.o \ opencl_allkernels.o \ diff --git a/libavfilter/framequeue.c b/libavfilter/framequeue.c new file mode 100644 index 0000000..debeab2 --- /dev/null +++ b/libavfilter/framequeue.c @@ -0,0 +1,123 @@ +/* + * Generic frame queue + * Copyright (c) 2016 Nicolas George + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public License + * as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public License + * along with FFmpeg; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/avassert.h" +#include "framequeue.h" + +static inline FFFrameBucket *bucket(FFFrameQueue *fq, size_t idx) +{ + return &fq->queue[(fq->tail + idx) & (fq->allocated - 1)]; +} + +void ff_framequeue_global_init(FFFrameQueueGlobal *fqg) +{ +} + +static void check_consistency(FFFrameQueue *fq) +{ +#if ASSERT_LEVEL >= 2 + uint64_t nb_samples = 0; + size_t i; + + av_assert0(fq->queued == fq->total_frames_head - fq->total_frames_tail); + for (i = 0; i < fq->queued; i++) + nb_samples += bucket(fq, i)->frame->nb_samples; + av_assert0(nb_samples == fq->total_samples_head - fq->total_samples_tail); +#endif +} + +void ff_framequeue_init(FFFrameQueue *fq, FFFrameQueueGlobal *fqg) +{ + fq->queue = &fq->first_bucket; + fq->allocated = 1; +} + +void ff_framequeue_free(FFFrameQueue *fq) +{ + while (fq->queued) { + AVFrame *frame = ff_framequeue_take(fq); + av_frame_free(&frame); + } + if (fq->queue != &fq->first_bucket) + av_freep(&fq->queue); +} + +int ff_framequeue_add(FFFrameQueue *fq, AVFrame *frame) +{ + FFFrameBucket *b; + + check_consistency(fq); + if (fq->queued == fq->allocated) { + if (fq->allocated == 1) { + size_t na = 8; + FFFrameBucket *nq = av_realloc_array(NULL, na, sizeof(*nq)); + if (!nq) + return AVERROR(ENOMEM); + nq[0] = fq->queue[0]; + fq->queue = nq; + fq->allocated = na; + } else { + size_t na = fq->allocated << 1; + FFFrameBucket *nq = av_realloc_array(fq->queue, na, sizeof(*nq)); + if (!nq) + return AVERROR(ENOMEM); + if (fq->tail + fq->queued > fq->allocated) + memmove(nq + fq->allocated, nq, + (fq->tail + fq->queued - fq->allocated) * sizeof(*nq)); + fq->queue = nq; + fq->allocated = na; + } + } + b = bucket(fq, fq->queued); + b->frame = frame; + fq->queued++; + fq->total_frames_head++; + fq->total_samples_head += frame->nb_samples; + check_consistency(fq); + return 0; +} + +AVFrame *ff_framequeue_take(FFFrameQueue *fq) +{ + FFFrameBucket *b; + + check_consistency(fq); + av_assert1(fq->queued); + b = bucket(fq, 0); + fq->queued--; + fq->tail++; + fq->tail &= fq->allocated - 1; + fq->total_frames_tail++; + fq->total_samples_tail += b->frame->nb_samples; + check_consistency(fq); + return b->frame; +} + +AVFrame *ff_framequeue_peek(FFFrameQueue *fq, size_t idx) +{ + FFFrameBucket *b; + + check_consistency(fq); + av_assert1(idx < fq->queued); + b = bucket(fq, idx); + check_consistency(fq); + return b->frame; +} diff --git a/libavfilter/framequeue.h b/libavfilter/framequeue.h new file mode 100644 index 0000000..558ea22 --- /dev/null +++ b/libavfilter/framequeue.h @@ -0,0 +1,173 @@ +/* + * Generic frame queue + * Copyright (c) 2016 Nicolas George + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public License + * as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public License + * along with FFmpeg; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVFILTER_FRAMEQUEUE_H +#define AVFILTER_FRAMEQUEUE_H + +/** + * FFFrameQueue: simple AVFrame queue API + * + * Note: this API is not thread-safe. Concurrent access to the same queue + * must be protected by a mutex or any synchronization mechanism. + */ + +#include "libavutil/frame.h" + +typedef struct FFFrameBucket { + AVFrame *frame; +} FFFrameBucket; + +/** + * Structure to hold global options and statistics for frame queues. + * + * This structure is intended to allow implementing global control of the + * frame queues, including memory consumption caps. + * + * It is currently empty. + */ +typedef struct FFFrameQueueGlobal { +} FFFrameQueueGlobal; + +/** + * Queue of AVFrame pointers. + */ +typedef struct FFFrameQueue { + + /** + * Array of allocated buckets, used as a circular buffer. + */ + FFFrameBucket *queue; + + /** + * Size of the array of buckets. + */ + size_t allocated; + + /** + * Tail of the queue. + * It is the index in the array of the next frame to take. + */ + size_t tail; + + /** + * Number of currently queued frames. + */ + size_t queued; + + /** + * Pre-allocated bucket for queues of size 1. + */ + FFFrameBucket first_bucket; + + /** + * Total number of frames entered in the queue. + */ + uint64_t total_frames_head; + + /** + * Total number of frames dequeued from the queue. + * queued = total_frames_head - total_frames_tail + */ + uint64_t total_frames_tail; + + /** + * Total number of samples entered in the queue. + */ + uint64_t total_samples_head; + + /** + * Total number of samples dequeued from the queue. + * queued_samples = total_samples_head - total_samples_tail + */ + uint64_t total_samples_tail; + +} FFFrameQueue; + +/** + * Init a global structure. + */ +void ff_framequeue_global_init(FFFrameQueueGlobal *fqg); + +/** + * Init a frame queue and attach it to a global structure. + */ +void ff_framequeue_init(FFFrameQueue *fq, FFFrameQueueGlobal *fqg); + +/** + * Free the queue and all queued frames. + */ +void ff_framequeue_free(FFFrameQueue *fq); + +/** + * Add a frame. + * @return >=0 or an AVERROR code. + */ +int ff_framequeue_add(FFFrameQueue *fq, AVFrame *frame); + +/** + * Take the first frame in the queue. + * Must not be used with empty queues. + */ +AVFrame *ff_framequeue_take(FFFrameQueue *fq); + +/** + * Access a frame in the queue, without removing it. + * The first frame is numbered 0; the designated frame must exist. + */ +AVFrame *ff_framequeue_peek(FFFrameQueue *fq, size_t idx); + +/** + * Get the number of queued frames. + */ +static inline size_t ff_framequeue_queued_frames(const FFFrameQueue *fq) +{ + return fq->queued; +} + +/** + * Get the number of queued samples. + */ +static inline uint64_t ff_framequeue_queued_samples(const FFFrameQueue *fq) +{ + return fq->total_samples_head - fq->total_samples_tail; +} + +/** + * Update the statistics after a frame accessed using ff_framequeue_peek() + * was modified. + * Currently used only as a marker. + */ +static inline void ff_framequeue_update_peeked(FFFrameQueue *fq, size_t idx) +{ +} + +/** + * Update the sample count in the queue. + * + * This function must be used when the first frame was accessed using + * ff_framequeue_peek() and samples were removed from it. + */ +static inline void ff_framequeue_skip_samples(FFFrameQueue *fq, size_t n) +{ + fq->total_samples_tail += n; +} + +#endif /* AVFILTER_FRAMEQUEUE_H */ -- 2.10.2

7 48

[PATCH v2] libavformat/tee: Add fifo support for tee
by sebechlebskyjan＠gmail.com 22 Dec '16

22 Dec '16

From: Jan Sebechlebsky <sebechlebskyjan(a)gmail.com> Signed-off-by: Jan Sebechlebsky <sebechlebskyjan(a)gmail.com> --- Thanks for noticing, I've fixed the patch (also some minor formatting issues I've noticed). doc/muxers.texi | 20 +++++++++++++ libavformat/tee.c | 87 ++++++++++++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 106 insertions(+), 1 deletion(-) diff --git a/doc/muxers.texi b/doc/muxers.texi index dbe53f5..7b4e165 100644 --- a/doc/muxers.texi +++ b/doc/muxers.texi @@ -1473,6 +1473,7 @@ Specify whether to remove all fragments when finished. Default 0 (do not remove) @end table +@anchor{fifo} @section fifo The fifo pseudo-muxer allows the separation of encoding and muxing by using @@ -1580,6 +1581,18 @@ with the tee muxer; encoding can be a very expensive process. It is not useful when using the libavformat API directly because it is then possible to feed the same packets to several muxers directly. +@table @option + +@item use_fifo @var{bool} +If set to 1, slave outputs will be processed in separate thread using @ref{fifo} +muxer. This allows to compensate for different speed/latency/reliability of +outputs and setup transparent recovery. By default this feature is turned off. + +@item fifo_options +Options to pass to fifo pseudo-muxer instances. See @ref{fifo}. + +@end table + The slave outputs are specified in the file name given to the muxer, separated by '|'. If any of the slave name contains the '|' separator, leading or trailing spaces or any special character, it must be @@ -1601,6 +1614,13 @@ output name suffix. Specify a list of bitstream filters to apply to the specified output. +@item use_fifo @var{bool} +This allows to override tee muxer use_fifo option for individual slave muxer. + +@item fifo_options +This allows to override tee muxer fifo_options for individual slave muxer. +See @ref{fifo}. + It is possible to specify to which streams a given bitstream filter applies, by appending a stream specifier to the option separated by @code{/}. @var{spec} must be a stream specifier (see @ref{Format diff --git a/libavformat/tee.c b/libavformat/tee.c index d59ad4d..c668e95 100644 --- a/libavformat/tee.c +++ b/libavformat/tee.c @@ -40,6 +40,8 @@ typedef struct { AVBSFContext **bsfs; ///< bitstream filters per stream SlaveFailurePolicy on_fail; + int use_fifo; + AVDictionary *fifo_options; /** map from input to output streams indexes, * disabled output streams are set to -1 */ @@ -52,15 +54,28 @@ typedef struct TeeContext { unsigned nb_slaves; unsigned nb_alive; TeeSlave *slaves; + int use_fifo; + AVDictionary *fifo_options; + char *fifo_options_str; } TeeContext; static const char *const slave_delim = "|"; static const char *const slave_bsfs_spec_sep = "/"; static const char *const slave_select_sep = ","; +#define OFFSET(x) offsetof(TeeContext, x) +static const AVOption options[] = { + {"use_fifo", "Use fifo pseudo-muxer to separate actual muxers from encoder", + OFFSET(use_fifo), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM}, + {"fifo_options", "fifo pseudo-muxer options", OFFSET(fifo_options_str), + AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, AV_OPT_FLAG_ENCODING_PARAM}, + {NULL} +}; + static const AVClass tee_muxer_class = { .class_name = "Tee muxer", .item_name = av_default_item_name, + .option = options, .version = LIBAVUTIL_VERSION_INT, }; @@ -81,6 +96,27 @@ static inline int parse_slave_failure_policy_option(const char *opt, TeeSlave *t return AVERROR(EINVAL); } +static int parse_slave_fifo_options(const char *use_fifo, + const char *fifo_options, TeeSlave *tee_slave) +{ + int ret = 0; + + if (use_fifo) { + if (av_match_name(use_fifo, "true,y,yes,enable,enabled,on,1")) { + tee_slave->use_fifo = 1; + } else if (av_match_name(use_fifo, "false,n,no,disable,disabled,off,0")) { + tee_slave->use_fifo = 0; + } else { + return AVERROR(EINVAL); + } + } + + if (fifo_options) + ret = av_dict_parse_string(&tee_slave->fifo_options, fifo_options, "=", ":", 0); + + return ret; +} + static int close_slave(TeeSlave *tee_slave) { AVFormatContext *avf; @@ -125,6 +161,7 @@ static int open_slave(AVFormatContext *avf, char *slave, TeeSlave *tee_slave) AVDictionaryEntry *entry; char *filename; char *format = NULL, *select = NULL, *on_fail = NULL; + char *use_fifo = NULL, *fifo_options_str = NULL; AVFormatContext *avf2 = NULL; AVStream *st, *st2; int stream_count; @@ -145,6 +182,8 @@ static int open_slave(AVFormatContext *avf, char *slave, TeeSlave *tee_slave) STEAL_OPTION("f", format); STEAL_OPTION("select", select); STEAL_OPTION("onfail", on_fail); + STEAL_OPTION("use_fifo", use_fifo); + STEAL_OPTION("fifo_options", fifo_options_str); ret = parse_slave_failure_policy_option(on_fail, tee_slave); if (ret < 0) { @@ -153,7 +192,41 @@ static int open_slave(AVFormatContext *avf, char *slave, TeeSlave *tee_slave) goto end; } - ret = avformat_alloc_output_context2(&avf2, NULL, format, filename); + ret = parse_slave_fifo_options(use_fifo, fifo_options_str, tee_slave); + if (ret < 0) { + av_log(avf, AV_LOG_ERROR, "Error parsing fifo options: %s\n", av_err2str(ret)); + goto end; + } + + if (tee_slave->use_fifo) { + + if (options) { + char *format_options_str = NULL; + ret = av_dict_get_string(options, &format_options_str, '=', ':'); + if (ret < 0) + goto end; + + ret = av_dict_set(&tee_slave->fifo_options, "format_options", format_options_str, + AV_DICT_DONT_STRDUP_VAL); + if (ret < 0) + goto end; + } + + if (format) { + ret = av_dict_set(&tee_slave->fifo_options, "fifo_format", format, + AV_DICT_DONT_STRDUP_VAL); + format = NULL; + if (ret < 0) + goto end; + } + + av_dict_free(&options); + options = tee_slave->fifo_options; + + ret = avformat_alloc_output_context2(&avf2, NULL, "fifo", filename); + } else { + ret = avformat_alloc_output_context2(&avf2, NULL, format, filename); + } if (ret < 0) goto end; tee_slave->avf = avf2; @@ -394,6 +467,12 @@ static int tee_write_header(AVFormatContext *avf) filename++; } + if (tee->fifo_options_str) { + ret = av_dict_parse_string(&tee->fifo_options, tee->fifo_options_str, "=", ":", 0); + if (ret < 0) + goto fail; + } + if (!(tee->slaves = av_mallocz_array(nb_slaves, sizeof(*tee->slaves)))) { ret = AVERROR(ENOMEM); goto fail; @@ -401,6 +480,12 @@ static int tee_write_header(AVFormatContext *avf) tee->nb_slaves = tee->nb_alive = nb_slaves; for (i = 0; i < nb_slaves; i++) { + + tee->slaves[i].use_fifo = tee->use_fifo; + ret = av_dict_copy(&tee->slaves[i].fifo_options, tee->fifo_options, 0); + if (ret < 0) + goto fail; + if ((ret = open_slave(avf, slaves[i], &tee->slaves[i])) < 0) { ret = tee_process_slave_failure(avf, i, ret); if (ret < 0) -- 1.9.1

3 6