From: Carl Hetherington Date: Thu, 15 Apr 2021 23:01:59 +0000 (+0200) Subject: Split audio analysis code off from the job. X-Git-Tag: v2.15.141~58 X-Git-Url: https://main.carlh.net/gitweb/?p=dcpomatic.git;a=commitdiff_plain;h=191adfe030803a49588afc4b9087da27654d946b;ds=sidebyside Split audio analysis code off from the job. --- diff --git a/src/lib/analyse_audio_job.cc b/src/lib/analyse_audio_job.cc index 448902e1e..ca0f49f57 100644 --- a/src/lib/analyse_audio_job.cc +++ b/src/lib/analyse_audio_job.cc @@ -1,5 +1,5 @@ /* - Copyright (C) 2012-2018 Carl Hetherington + Copyright (C) 2012-2021 Carl Hetherington This file is part of DCP-o-matic. @@ -18,150 +18,77 @@ */ -#include "audio_analysis.h" -#include "audio_buffers.h" + #include "analyse_audio_job.h" -#include "audio_content.h" +#include "audio_analysis.h" #include "compose.hpp" #include "dcpomatic_log.h" #include "film.h" +#include "filter.h" #include "player.h" #include "playlist.h" -#include "filter.h" -#include "audio_filter_graph.h" #include "config.h" -extern "C" { -#include -#include -#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG -#include -#endif -} #include #include "i18n.h" -using std::string; -using std::vector; + +using std::cout; +using std::dynamic_pointer_cast; +using std::make_shared; using std::max; using std::min; -using std::cout; using std::shared_ptr; -using std::dynamic_pointer_cast; +using std::string; +using std::vector; using namespace dcpomatic; #if BOOST_VERSION >= 106100 using namespace boost::placeholders; #endif -int const AnalyseAudioJob::_num_points = 1024; - -static void add_if_required(vector& v, size_t i, double db) -{ - if (v.size() > i) { - v[i] = pow(10, db / 20); - } -} /** @param from_zero true to analyse audio from time 0 in the playlist, otherwise begin at Playlist::start */ AnalyseAudioJob::AnalyseAudioJob (shared_ptr film, shared_ptr playlist, bool from_zero) : Job (film) + , _analyser (film, playlist, from_zero, boost::bind(&Job::set_progress, this, _1, false)) , _playlist (playlist) , _path (film->audio_analysis_path(playlist)) - , _from_zero (from_zero) - , _done (0) - , _samples_per_point (1) - , _current (0) - , _sample_peak (new float[film->audio_channels()]) - , _sample_peak_frame (new Frame[film->audio_channels()]) -#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG - , _ebur128 (new AudioFilterGraph (film->audio_frame_rate(), film->audio_channels())) -#endif { LOG_DEBUG_AUDIO_ANALYSIS_NC("AnalyseAudioJob::AnalyseAudioJob"); - -#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG - _filters.push_back (new Filter ("ebur128", "ebur128", "audio", "ebur128=peak=true")); - _ebur128->setup (_filters); -#endif - - for (int i = 0; i < film->audio_channels(); ++i) { - _sample_peak[i] = 0; - _sample_peak_frame[i] = 0; - } - - if (!_from_zero) { - _start = _playlist->start().get_value_or(DCPTime()); - } - - /* XXX: is this right? Especially for more than 5.1? */ - vector channel_corrections(film->audio_channels(), 1); - add_if_required (channel_corrections, 4, -3); // Ls - add_if_required (channel_corrections, 5, -3); // Rs - add_if_required (channel_corrections, 6, -144); // HI - add_if_required (channel_corrections, 7, -144); // VI - add_if_required (channel_corrections, 8, -3); // Lc - add_if_required (channel_corrections, 9, -3); // Rc - add_if_required (channel_corrections, 10, -3); // Lc - add_if_required (channel_corrections, 11, -3); // Rc - add_if_required (channel_corrections, 12, -144); // DBox - add_if_required (channel_corrections, 13, -144); // Sync - add_if_required (channel_corrections, 14, -144); // Sign Language - add_if_required (channel_corrections, 15, -144); // Unused - - _leqm.reset(new leqm_nrt::Calculator( - film->audio_channels(), - film->audio_frame_rate(), - 24, - channel_corrections, - 850, // suggested by leqm_nrt CLI source - 64, // suggested by leqm_nrt CLI source - boost::thread::hardware_concurrency() - )); } + AnalyseAudioJob::~AnalyseAudioJob () { stop_thread (); - for (auto i: _filters) { - delete const_cast (i); - } - delete[] _current; - delete[] _sample_peak; - delete[] _sample_peak_frame; } + string AnalyseAudioJob::name () const { return _("Analysing audio"); } + string AnalyseAudioJob::json_name () const { return N_("analyse_audio"); } + void AnalyseAudioJob::run () { LOG_DEBUG_AUDIO_ANALYSIS_NC("AnalyseAudioJob::run"); - shared_ptr player (new Player(_film, _playlist)); + auto player = make_shared(_film, _playlist); player->set_ignore_video (); player->set_ignore_text (); player->set_fast (); player->set_play_referenced (); - player->Audio.connect (bind (&AnalyseAudioJob::analyse, this, _1, _2)); - - DCPTime const length = _playlist->length (_film); - - Frame const len = DCPTime (length - _start).frames_round (_film->audio_frame_rate()); - _samples_per_point = max (int64_t (1), len / _num_points); - - delete[] _current; - _current = new AudioPoint[_film->audio_channels ()]; - _analysis.reset (new AudioAnalysis (_film->audio_channels ())); + player->Audio.connect (bind(&AudioAnalyser::analyse, &_analyser, _1, _2)); bool has_any_audio = false; for (auto c: _playlist->content()) { @@ -171,105 +98,17 @@ AnalyseAudioJob::run () } if (has_any_audio) { - LOG_DEBUG_AUDIO_ANALYSIS("Seeking to %1", to_string(_start)); - player->seek (_start, true); - _done = 0; - LOG_DEBUG_AUDIO_ANALYSIS("Starting loop for playlist of length %1", to_string(length)); + player->seek (_analyser.start(), true); while (!player->pass ()) {} } LOG_DEBUG_AUDIO_ANALYSIS_NC("Loop complete"); - vector sample_peak; - for (int i = 0; i < _film->audio_channels(); ++i) { - sample_peak.push_back ( - AudioAnalysis::PeakTime (_sample_peak[i], DCPTime::from_frames (_sample_peak_frame[i], _film->audio_frame_rate ())) - ); - } - _analysis->set_sample_peak (sample_peak); - -#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG - if (Config::instance()->analyse_ebur128 ()) { - void* eb = _ebur128->get("Parsed_ebur128_0")->priv; - vector true_peak; - for (int i = 0; i < _film->audio_channels(); ++i) { - true_peak.push_back (av_ebur128_get_true_peaks(eb)[i]); - } - _analysis->set_true_peak (true_peak); - _analysis->set_integrated_loudness (av_ebur128_get_integrated_loudness(eb)); - _analysis->set_loudness_range (av_ebur128_get_loudness_range(eb)); - } -#endif - - if (_playlist->content().size() == 1) { - /* If there was only one piece of content in this analysis we may later need to know what its - gain was when we analysed it. - */ - if (auto ac = _playlist->content().front()->audio) { - _analysis->set_analysis_gain (ac->gain()); - } - } - - _analysis->set_samples_per_point (_samples_per_point); - _analysis->set_sample_rate (_film->audio_frame_rate ()); - _analysis->set_leqm (_leqm->leq_m()); - _analysis->write (_path); + _analyser.finish (); + auto analysis = _analyser.get(); + analysis.write (_path); LOG_DEBUG_AUDIO_ANALYSIS_NC("Job finished"); set_progress (1); set_state (FINISHED_OK); } - -void -AnalyseAudioJob::analyse (shared_ptr b, DCPTime time) -{ - LOG_DEBUG_AUDIO_ANALYSIS("Received %1 frames at %2", b->frames(), to_string(time)); - DCPOMATIC_ASSERT (time >= _start); - -#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG - if (Config::instance()->analyse_ebur128 ()) { - _ebur128->process (b); - } -#endif - - int const frames = b->frames (); - int const channels = b->channels (); - vector interleaved(frames * channels); - - for (int j = 0; j < channels; ++j) { - float* data = b->data(j); - for (int i = 0; i < frames; ++i) { - float s = data[i]; - - interleaved[i * channels + j] = s; - - float as = fabsf (s); - if (as < 10e-7) { - /* We may struggle to serialise and recover inf or -inf, so prevent such - values by replacing with this (140dB down) */ - s = as = 10e-7; - } - _current[j][AudioPoint::RMS] += pow (s, 2); - _current[j][AudioPoint::PEAK] = max (_current[j][AudioPoint::PEAK], as); - - if (as > _sample_peak[j]) { - _sample_peak[j] = as; - _sample_peak_frame[j] = _done + i; - } - - if (((_done + i) % _samples_per_point) == 0) { - _current[j][AudioPoint::RMS] = sqrt (_current[j][AudioPoint::RMS] / _samples_per_point); - _analysis->add_point (j, _current[j]); - _current[j] = AudioPoint (); - } - } - } - - _leqm->add(interleaved); - - _done += frames; - - DCPTime const length = _playlist->length (_film); - set_progress ((time.seconds() - _start.seconds()) / (length.seconds() - _start.seconds())); - LOG_DEBUG_AUDIO_ANALYSIS_NC("Frames processed"); -} diff --git a/src/lib/analyse_audio_job.h b/src/lib/analyse_audio_job.h index 81287d2d2..864a6a7cd 100644 --- a/src/lib/analyse_audio_job.h +++ b/src/lib/analyse_audio_job.h @@ -18,10 +18,13 @@ */ + /** @file src/lib/analyse_audio_job.h * @brief AnalyseAudioJob class. */ + +#include "audio_analyser.h" #include "job.h" #include "audio_point.h" #include "types.h" @@ -29,6 +32,7 @@ #include #include + class AudioBuffers; class AudioAnalysis; class Playlist; @@ -36,6 +40,7 @@ class AudioPoint; class AudioFilterGraph; class Filter; + /** @class AnalyseAudioJob * @brief A job to analyse the audio of a film and make a note of its * broad peak and RMS levels. @@ -58,27 +63,11 @@ public: } private: - void analyse (std::shared_ptr, dcpomatic::DCPTime time); + AudioAnalyser _analyser; std::shared_ptr _playlist; /** playlist's audio analysis path when the job was created */ boost::filesystem::path _path; - dcpomatic::DCPTime _start; - bool _from_zero; - - Frame _done; - Frame _samples_per_point; - AudioPoint* _current; - - float* _sample_peak; - Frame* _sample_peak_frame; - - std::shared_ptr _analysis; - - std::shared_ptr _ebur128; - std::vector _filters; - - boost::scoped_ptr _leqm; static const int _num_points; }; diff --git a/src/lib/audio_analyser.cc b/src/lib/audio_analyser.cc new file mode 100644 index 000000000..3caa997df --- /dev/null +++ b/src/lib/audio_analyser.cc @@ -0,0 +1,221 @@ +/* + Copyright (C) 2021 Carl Hetherington + + This file is part of DCP-o-matic. + + DCP-o-matic is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2 of the License, or + (at your option) any later version. + + DCP-o-matic 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 General Public License for more details. + + You should have received a copy of the GNU General Public License + along with DCP-o-matic. If not, see . + +*/ + + +#include "audio_analyser.h" +#include "audio_analysis.h" +#include "audio_buffers.h" +#include "audio_content.h" +#include "audio_filter_graph.h" +#include "audio_point.h" +#include "config.h" +#include "dcpomatic_log.h" +#include "film.h" +#include "filter.h" +#include "playlist.h" +#include "types.h" +extern "C" { +#include +#include +#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG +#include +#endif +} + + +using std::make_shared; +using std::max; +using std::shared_ptr; +using std::vector; +using namespace dcpomatic; + + +static auto constexpr num_points = 1024; + + +AudioAnalyser::AudioAnalyser (shared_ptr film, shared_ptr playlist, bool from_zero, std::function set_progress) + : _film (film) + , _playlist (playlist) + , _set_progress (set_progress) +#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG + , _ebur128 (new AudioFilterGraph(film->audio_frame_rate(), film->audio_channels())) +#endif + , _sample_peak (new float[film->audio_channels()]) + , _sample_peak_frame (new Frame[film->audio_channels()]) + , _analysis (film->audio_channels()) +{ + +#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG + _filters.push_back (new Filter("ebur128", "ebur128", "audio", "ebur128=peak=true")); + _ebur128->setup (_filters); +#endif + + _current = new AudioPoint[_film->audio_channels()]; + + if (!from_zero) { + _start = _playlist->start().get_value_or(DCPTime()); + } + + for (int i = 0; i < film->audio_channels(); ++i) { + _sample_peak[i] = 0; + _sample_peak_frame[i] = 0; + } + + auto add_if_required = [](vector& v, size_t i, double db) { + if (v.size() > i) { + v[i] = pow(10, db / 20); + } + }; + + /* XXX: is this right? Especially for more than 5.1? */ + vector channel_corrections(film->audio_channels(), 1); + add_if_required (channel_corrections, 4, -3); // Ls + add_if_required (channel_corrections, 5, -3); // Rs + add_if_required (channel_corrections, 6, -144); // HI + add_if_required (channel_corrections, 7, -144); // VI + add_if_required (channel_corrections, 8, -3); // Lc + add_if_required (channel_corrections, 9, -3); // Rc + add_if_required (channel_corrections, 10, -3); // Lc + add_if_required (channel_corrections, 11, -3); // Rc + add_if_required (channel_corrections, 12, -144); // DBox + add_if_required (channel_corrections, 13, -144); // Sync + add_if_required (channel_corrections, 14, -144); // Sign Language + add_if_required (channel_corrections, 15, -144); // Unused + + _leqm.reset(new leqm_nrt::Calculator( + film->audio_channels(), + film->audio_frame_rate(), + 24, + channel_corrections, + 850, // suggested by leqm_nrt CLI source + 64, // suggested by leqm_nrt CLI source + boost::thread::hardware_concurrency() + )); + + DCPTime const length = _playlist->length (_film); + + Frame const len = DCPTime (length - _start).frames_round (film->audio_frame_rate()); + _samples_per_point = max (int64_t (1), len / num_points); +} + + +AudioAnalyser::~AudioAnalyser () +{ + delete[] _current; + for (auto i: _filters) { + delete const_cast (i); + } + delete[] _sample_peak; + delete[] _sample_peak_frame; +} + + +void +AudioAnalyser::analyse (shared_ptr b, DCPTime time) +{ + LOG_DEBUG_AUDIO_ANALYSIS("Received %1 frames at %2", b->frames(), to_string(time)); + DCPOMATIC_ASSERT (time >= _start); + +#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG + if (Config::instance()->analyse_ebur128 ()) { + _ebur128->process (b); + } +#endif + + int const frames = b->frames (); + int const channels = b->channels (); + vector interleaved(frames * channels); + + for (int j = 0; j < channels; ++j) { + float* data = b->data(j); + for (int i = 0; i < frames; ++i) { + float s = data[i]; + + interleaved[i * channels + j] = s; + + float as = fabsf (s); + if (as < 10e-7) { + /* We may struggle to serialise and recover inf or -inf, so prevent such + values by replacing with this (140dB down) */ + s = as = 10e-7; + } + _current[j][AudioPoint::RMS] += pow (s, 2); + _current[j][AudioPoint::PEAK] = max (_current[j][AudioPoint::PEAK], as); + + if (as > _sample_peak[j]) { + _sample_peak[j] = as; + _sample_peak_frame[j] = _done + i; + } + + if (((_done + i) % _samples_per_point) == 0) { + _current[j][AudioPoint::RMS] = sqrt (_current[j][AudioPoint::RMS] / _samples_per_point); + _analysis.add_point (j, _current[j]); + _current[j] = AudioPoint (); + } + } + } + + _leqm->add(interleaved); + + _done += frames; + + DCPTime const length = _playlist->length (_film); + _set_progress ((time.seconds() - _start.seconds()) / (length.seconds() - _start.seconds())); + LOG_DEBUG_AUDIO_ANALYSIS_NC("Frames processed"); +} + + +void +AudioAnalyser::finish () +{ + vector sample_peak; + for (int i = 0; i < _film->audio_channels(); ++i) { + sample_peak.push_back ( + AudioAnalysis::PeakTime (_sample_peak[i], DCPTime::from_frames (_sample_peak_frame[i], _film->audio_frame_rate ())) + ); + } + _analysis.set_sample_peak (sample_peak); + +#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG + if (Config::instance()->analyse_ebur128 ()) { + void* eb = _ebur128->get("Parsed_ebur128_0")->priv; + vector true_peak; + for (int i = 0; i < _film->audio_channels(); ++i) { + true_peak.push_back (av_ebur128_get_true_peaks(eb)[i]); + } + _analysis.set_true_peak (true_peak); + _analysis.set_integrated_loudness (av_ebur128_get_integrated_loudness(eb)); + _analysis.set_loudness_range (av_ebur128_get_loudness_range(eb)); + } +#endif + + if (_playlist->content().size() == 1) { + /* If there was only one piece of content in this analysis we may later need to know what its + gain was when we analysed it. + */ + if (auto ac = _playlist->content().front()->audio) { + _analysis.set_analysis_gain (ac->gain()); + } + } + + _analysis.set_samples_per_point (_samples_per_point); + _analysis.set_sample_rate (_film->audio_frame_rate ()); + _analysis.set_leqm (_leqm->leq_m()); +} diff --git a/src/lib/audio_analyser.h b/src/lib/audio_analyser.h new file mode 100644 index 000000000..e47ab94b4 --- /dev/null +++ b/src/lib/audio_analyser.h @@ -0,0 +1,81 @@ +/* + Copyright (C) 2021 Carl Hetherington + + This file is part of DCP-o-matic. + + DCP-o-matic is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2 of the License, or + (at your option) any later version. + + DCP-o-matic 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 General Public License for more details. + + You should have received a copy of the GNU General Public License + along with DCP-o-matic. If not, see . + +*/ + + +#include "audio_analysis.h" +#include "dcpomatic_time.h" +#include "types.h" +#include +#include +#include + + +class AudioAnalysis; +class AudioBuffers; +class AudioFilterGraph; +class AudioPoint; +class Film; +class Filter; +class Playlist; + + +class AudioAnalyser +{ +public: + AudioAnalyser (std::shared_ptr film, std::shared_ptr playlist, bool from_zero, std::function set_progress); + ~AudioAnalyser (); + + AudioAnalyser (AudioAnalyser const&) = delete; + AudioAnalyser& operator= (AudioAnalyser const&) = delete; + + void analyse (std::shared_ptr, dcpomatic::DCPTime time); + + dcpomatic::DCPTime start () const { + return _start; + } + + void finish (); + + AudioAnalysis get () const { + return _analysis; + } + +private: + std::shared_ptr _film; + std::shared_ptr _playlist; + + std::function _set_progress; + + dcpomatic::DCPTime _start; +#ifdef DCPOMATIC_HAVE_EBUR128_PATCHED_FFMPEG + std::shared_ptr _ebur128; +#endif + std::vector _filters; + Frame _samples_per_point = 1; + + boost::scoped_ptr _leqm; + Frame _done = 0; + float* _sample_peak = nullptr; + Frame* _sample_peak_frame = nullptr; + AudioPoint* _current = nullptr; + + AudioAnalysis _analysis; +}; + diff --git a/src/lib/audio_analysis.h b/src/lib/audio_analysis.h index 263e8a8a0..038059502 100644 --- a/src/lib/audio_analysis.h +++ b/src/lib/audio_analysis.h @@ -39,7 +39,7 @@ namespace xmlpp { class Playlist; -class AudioAnalysis : public boost::noncopyable +class AudioAnalysis { public: explicit AudioAnalysis (int c); diff --git a/src/lib/wscript b/src/lib/wscript index 7cbd85a4c..2965111ad 100644 --- a/src/lib/wscript +++ b/src/lib/wscript @@ -30,6 +30,7 @@ sources = """ atmos_decoder.cc atmos_metadata.cc atmos_mxf_decoder.cc + audio_analyser.cc audio_analysis.cc audio_buffers.cc audio_content.cc