| /* |
| * Copyright (c) 2020 The WebRTC project authors. All Rights Reserved. |
| * |
| * Use of this source code is governed by a BSD-style license |
| * that can be found in the LICENSE file in the root of the source |
| * tree. An additional intellectual property rights grant can be found |
| * in the file PATENTS. All contributing project authors may |
| * be found in the AUTHORS file in the root of the source tree. |
| */ |
| |
| #include "rtc_tools/rtc_event_log_visualizer/analyze_audio.h" |
| |
| #include <cstdint> |
| #include <map> |
| #include <memory> |
| #include <optional> |
| #include <set> |
| #include <string> |
| #include <utility> |
| #include <vector> |
| |
| #include "absl/strings/string_view.h" |
| #include "api/audio_codecs/audio_codec_pair_id.h" |
| #include "api/audio_codecs/audio_decoder.h" |
| #include "api/audio_codecs/audio_decoder_factory.h" |
| #include "api/audio_codecs/audio_format.h" |
| #include "api/environment/environment.h" |
| #include "api/function_view.h" |
| #include "api/make_ref_counted.h" |
| #include "api/neteq/neteq.h" |
| #include "api/scoped_refptr.h" |
| #include "api/units/timestamp.h" |
| #include "logging/rtc_event_log/events/rtc_event_audio_network_adaptation.h" |
| #include "logging/rtc_event_log/rtc_event_log_parser.h" |
| #include "modules/audio_coding/neteq/tools/audio_sink.h" |
| #include "modules/audio_coding/neteq/tools/fake_decode_from_file.h" |
| #include "modules/audio_coding/neteq/tools/neteq_delay_analyzer.h" |
| #include "modules/audio_coding/neteq/tools/neteq_event_log_input.h" |
| #include "modules/audio_coding/neteq/tools/neteq_input.h" |
| #include "modules/audio_coding/neteq/tools/neteq_replacement_input.h" |
| #include "modules/audio_coding/neteq/tools/neteq_stats_getter.h" |
| #include "modules/audio_coding/neteq/tools/neteq_test.h" |
| #include "modules/audio_coding/neteq/tools/resample_input_audio_file.h" |
| #include "rtc_base/checks.h" |
| #include "rtc_tools/rtc_event_log_visualizer/analyzer_common.h" |
| #include "rtc_tools/rtc_event_log_visualizer/plot_base.h" |
| #include "system_wrappers/include/field_trial.h" |
| |
| namespace webrtc { |
| |
| void CreateAudioEncoderTargetBitrateGraph(const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| Plot* plot) { |
| TimeSeries time_series("Audio encoder target bitrate", LineStyle::kLine, |
| PointStyle::kHighlight); |
| auto GetAnaBitrateBps = [](const LoggedAudioNetworkAdaptationEvent& ana_event) |
| -> std::optional<float> { |
| if (ana_event.config.bitrate_bps) |
| return std::optional<float>( |
| static_cast<float>(*ana_event.config.bitrate_bps)); |
| return std::nullopt; |
| }; |
| auto ToCallTime = [config](const LoggedAudioNetworkAdaptationEvent& packet) { |
| return config.GetCallTimeSec(packet.log_time()); |
| }; |
| ProcessPoints<LoggedAudioNetworkAdaptationEvent>( |
| ToCallTime, GetAnaBitrateBps, |
| parsed_log.audio_network_adaptation_events(), &time_series); |
| plot->AppendTimeSeries(std::move(time_series)); |
| plot->SetXAxis(config.CallBeginTimeSec(), config.CallEndTimeSec(), "Time (s)", |
| kLeftMargin, kRightMargin); |
| plot->SetSuggestedYAxis(0, 1, "Bitrate (bps)", kBottomMargin, kTopMargin); |
| plot->SetTitle("Reported audio encoder target bitrate"); |
| } |
| |
| void CreateAudioEncoderFrameLengthGraph(const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| Plot* plot) { |
| TimeSeries time_series("Audio encoder frame length", LineStyle::kLine, |
| PointStyle::kHighlight); |
| auto GetAnaFrameLengthMs = |
| [](const LoggedAudioNetworkAdaptationEvent& ana_event) { |
| if (ana_event.config.frame_length_ms) |
| return std::optional<float>( |
| static_cast<float>(*ana_event.config.frame_length_ms)); |
| return std::optional<float>(); |
| }; |
| auto ToCallTime = [config](const LoggedAudioNetworkAdaptationEvent& packet) { |
| return config.GetCallTimeSec(packet.log_time()); |
| }; |
| ProcessPoints<LoggedAudioNetworkAdaptationEvent>( |
| ToCallTime, GetAnaFrameLengthMs, |
| parsed_log.audio_network_adaptation_events(), &time_series); |
| plot->AppendTimeSeries(std::move(time_series)); |
| plot->SetXAxis(config.CallBeginTimeSec(), config.CallEndTimeSec(), "Time (s)", |
| kLeftMargin, kRightMargin); |
| plot->SetSuggestedYAxis(0, 1, "Frame length (ms)", kBottomMargin, kTopMargin); |
| plot->SetTitle("Reported audio encoder frame length"); |
| } |
| |
| void CreateAudioEncoderPacketLossGraph(const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| Plot* plot) { |
| TimeSeries time_series("Audio encoder uplink packet loss fraction", |
| LineStyle::kLine, PointStyle::kHighlight); |
| auto GetAnaPacketLoss = |
| [](const LoggedAudioNetworkAdaptationEvent& ana_event) { |
| if (ana_event.config.uplink_packet_loss_fraction) |
| return std::optional<float>(static_cast<float>( |
| *ana_event.config.uplink_packet_loss_fraction)); |
| return std::optional<float>(); |
| }; |
| auto ToCallTime = [config](const LoggedAudioNetworkAdaptationEvent& packet) { |
| return config.GetCallTimeSec(packet.log_time()); |
| }; |
| ProcessPoints<LoggedAudioNetworkAdaptationEvent>( |
| ToCallTime, GetAnaPacketLoss, |
| parsed_log.audio_network_adaptation_events(), &time_series); |
| plot->AppendTimeSeries(std::move(time_series)); |
| plot->SetXAxis(config.CallBeginTimeSec(), config.CallEndTimeSec(), "Time (s)", |
| kLeftMargin, kRightMargin); |
| plot->SetSuggestedYAxis(0, 10, "Percent lost packets", kBottomMargin, |
| kTopMargin); |
| plot->SetTitle("Reported audio encoder lost packets"); |
| } |
| |
| void CreateAudioEncoderEnableFecGraph(const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| Plot* plot) { |
| TimeSeries time_series("Audio encoder FEC", LineStyle::kLine, |
| PointStyle::kHighlight); |
| auto GetAnaFecEnabled = |
| [](const LoggedAudioNetworkAdaptationEvent& ana_event) { |
| if (ana_event.config.enable_fec) |
| return std::optional<float>( |
| static_cast<float>(*ana_event.config.enable_fec)); |
| return std::optional<float>(); |
| }; |
| auto ToCallTime = [config](const LoggedAudioNetworkAdaptationEvent& packet) { |
| return config.GetCallTimeSec(packet.log_time()); |
| }; |
| ProcessPoints<LoggedAudioNetworkAdaptationEvent>( |
| ToCallTime, GetAnaFecEnabled, |
| parsed_log.audio_network_adaptation_events(), &time_series); |
| plot->AppendTimeSeries(std::move(time_series)); |
| plot->SetXAxis(config.CallBeginTimeSec(), config.CallEndTimeSec(), "Time (s)", |
| kLeftMargin, kRightMargin); |
| plot->SetSuggestedYAxis(0, 1, "FEC (false/true)", kBottomMargin, kTopMargin); |
| plot->SetTitle("Reported audio encoder FEC"); |
| } |
| |
| void CreateAudioEncoderEnableDtxGraph(const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| Plot* plot) { |
| TimeSeries time_series("Audio encoder DTX", LineStyle::kLine, |
| PointStyle::kHighlight); |
| auto GetAnaDtxEnabled = |
| [](const LoggedAudioNetworkAdaptationEvent& ana_event) { |
| if (ana_event.config.enable_dtx) |
| return std::optional<float>( |
| static_cast<float>(*ana_event.config.enable_dtx)); |
| return std::optional<float>(); |
| }; |
| auto ToCallTime = [config](const LoggedAudioNetworkAdaptationEvent& packet) { |
| return config.GetCallTimeSec(packet.log_time()); |
| }; |
| ProcessPoints<LoggedAudioNetworkAdaptationEvent>( |
| ToCallTime, GetAnaDtxEnabled, |
| parsed_log.audio_network_adaptation_events(), &time_series); |
| plot->AppendTimeSeries(std::move(time_series)); |
| plot->SetXAxis(config.CallBeginTimeSec(), config.CallEndTimeSec(), "Time (s)", |
| kLeftMargin, kRightMargin); |
| plot->SetSuggestedYAxis(0, 1, "DTX (false/true)", kBottomMargin, kTopMargin); |
| plot->SetTitle("Reported audio encoder DTX"); |
| } |
| |
| void CreateAudioEncoderNumChannelsGraph(const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| Plot* plot) { |
| TimeSeries time_series("Audio encoder number of channels", LineStyle::kLine, |
| PointStyle::kHighlight); |
| auto GetAnaNumChannels = |
| [](const LoggedAudioNetworkAdaptationEvent& ana_event) { |
| if (ana_event.config.num_channels) |
| return std::optional<float>( |
| static_cast<float>(*ana_event.config.num_channels)); |
| return std::optional<float>(); |
| }; |
| auto ToCallTime = [config](const LoggedAudioNetworkAdaptationEvent& packet) { |
| return config.GetCallTimeSec(packet.log_time()); |
| }; |
| ProcessPoints<LoggedAudioNetworkAdaptationEvent>( |
| ToCallTime, GetAnaNumChannels, |
| parsed_log.audio_network_adaptation_events(), &time_series); |
| plot->AppendTimeSeries(std::move(time_series)); |
| plot->SetXAxis(config.CallBeginTimeSec(), config.CallEndTimeSec(), "Time (s)", |
| kLeftMargin, kRightMargin); |
| plot->SetSuggestedYAxis(0, 1, "Number of channels (1 (mono)/2 (stereo))", |
| kBottomMargin, kTopMargin); |
| plot->SetTitle("Reported audio encoder number of channels"); |
| } |
| |
| namespace { |
| |
| // Factory to create a "replacement decoder" that produces the decoded audio |
| // by reading from a file rather than from the encoded payloads. |
| class ReplacementAudioDecoderFactory : public AudioDecoderFactory { |
| public: |
| ReplacementAudioDecoderFactory(const absl::string_view replacement_file_name, |
| int file_sample_rate_hz) |
| : replacement_file_name_(replacement_file_name), |
| file_sample_rate_hz_(file_sample_rate_hz) {} |
| |
| std::vector<AudioCodecSpec> GetSupportedDecoders() override { |
| RTC_DCHECK_NOTREACHED(); |
| return {}; |
| } |
| |
| bool IsSupportedDecoder(const SdpAudioFormat& format) override { |
| return true; |
| } |
| |
| std::unique_ptr<AudioDecoder> Create( |
| const Environment& env, |
| const SdpAudioFormat& format, |
| std::optional<AudioCodecPairId> codec_pair_id) override { |
| auto replacement_file = std::make_unique<test::ResampleInputAudioFile>( |
| replacement_file_name_, file_sample_rate_hz_); |
| replacement_file->set_output_rate_hz(48000); |
| return std::make_unique<test::FakeDecodeFromFile>( |
| std::move(replacement_file), 48000, false); |
| } |
| |
| private: |
| const std::string replacement_file_name_; |
| const int file_sample_rate_hz_; |
| }; |
| |
| // Creates a NetEq test object and all necessary input and output helpers. Runs |
| // the test and returns the NetEqDelayAnalyzer object that was used to |
| // instrument the test. |
| std::unique_ptr<test::NetEqStatsGetter> CreateNetEqTestAndRun( |
| ParsedRtcEventLog parsed_log, |
| uint32_t ssrc, |
| const std::string& replacement_file_name, |
| int file_sample_rate_hz) { |
| std::unique_ptr<test::NetEqInput> input = |
| test::CreateNetEqEventLogInput(parsed_log, ssrc); |
| if (!input) { |
| return nullptr; |
| } |
| |
| constexpr int kReplacementPt = 127; |
| std::set<uint8_t> cn_types; |
| std::set<uint8_t> forbidden_types; |
| input.reset(new test::NetEqReplacementInput(std::move(input), kReplacementPt, |
| cn_types, forbidden_types)); |
| |
| std::unique_ptr<test::VoidAudioSink> output(new test::VoidAudioSink()); |
| |
| rtc::scoped_refptr<AudioDecoderFactory> decoder_factory = |
| rtc::make_ref_counted<ReplacementAudioDecoderFactory>( |
| replacement_file_name, file_sample_rate_hz); |
| |
| test::NetEqTest::DecoderMap codecs = { |
| {kReplacementPt, SdpAudioFormat("l16", 48000, 1)}}; |
| |
| std::unique_ptr<test::NetEqDelayAnalyzer> delay_cb( |
| new test::NetEqDelayAnalyzer); |
| std::unique_ptr<test::NetEqStatsGetter> neteq_stats_getter( |
| new test::NetEqStatsGetter(std::move(delay_cb))); |
| test::DefaultNetEqTestErrorCallback error_cb; |
| test::NetEqTest::Callbacks callbacks; |
| callbacks.error_callback = &error_cb; |
| callbacks.post_insert_packet = neteq_stats_getter->delay_analyzer(); |
| callbacks.get_audio_callback = neteq_stats_getter.get(); |
| |
| NetEq::Config config; |
| test::NetEqTest test(config, decoder_factory, codecs, /*text_log=*/nullptr, |
| /*factory=*/nullptr, std::move(input), std::move(output), |
| callbacks, field_trial::GetFieldTrialString()); |
| test.Run(); |
| return neteq_stats_getter; |
| } |
| } // namespace |
| |
| NetEqStatsGetterMap SimulateNetEq(const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| const std::string& replacement_file_name, |
| int file_sample_rate_hz) { |
| NetEqStatsGetterMap neteq_stats; |
| for (uint32_t ssrc : parsed_log.incoming_audio_ssrcs()) { |
| std::unique_ptr<test::NetEqStatsGetter> stats = CreateNetEqTestAndRun( |
| parsed_log, ssrc, replacement_file_name, file_sample_rate_hz); |
| if (stats) { |
| neteq_stats[ssrc] = std::move(stats); |
| } |
| } |
| return neteq_stats; |
| } |
| |
| // Given a NetEqStatsGetter and the SSRC that the NetEqStatsGetter was created |
| // for, this method generates a plot for the jitter buffer delay profile. |
| void CreateAudioJitterBufferGraph(const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| uint32_t ssrc, |
| const test::NetEqStatsGetter* stats_getter, |
| Plot* plot) { |
| test::NetEqDelayAnalyzer::Delays arrival_delay_ms; |
| test::NetEqDelayAnalyzer::Delays corrected_arrival_delay_ms; |
| test::NetEqDelayAnalyzer::Delays playout_delay_ms; |
| test::NetEqDelayAnalyzer::Delays target_delay_ms; |
| |
| stats_getter->delay_analyzer()->CreateGraphs( |
| &arrival_delay_ms, &corrected_arrival_delay_ms, &playout_delay_ms, |
| &target_delay_ms); |
| |
| TimeSeries time_series_packet_arrival("packet arrival delay", |
| LineStyle::kLine); |
| TimeSeries time_series_relative_packet_arrival( |
| "Relative packet arrival delay", LineStyle::kLine); |
| TimeSeries time_series_play_time("Playout delay", LineStyle::kLine); |
| TimeSeries time_series_target_time("Target delay", LineStyle::kLine, |
| PointStyle::kHighlight); |
| |
| for (const auto& data : arrival_delay_ms) { |
| const float x = config.GetCallTimeSec(Timestamp::Millis(data.first)); |
| const float y = data.second; |
| time_series_packet_arrival.points.emplace_back(TimeSeriesPoint(x, y)); |
| } |
| for (const auto& data : corrected_arrival_delay_ms) { |
| const float x = config.GetCallTimeSec(Timestamp::Millis(data.first)); |
| const float y = data.second; |
| time_series_relative_packet_arrival.points.emplace_back( |
| TimeSeriesPoint(x, y)); |
| } |
| for (const auto& data : playout_delay_ms) { |
| const float x = config.GetCallTimeSec(Timestamp::Millis(data.first)); |
| const float y = data.second; |
| time_series_play_time.points.emplace_back(TimeSeriesPoint(x, y)); |
| } |
| for (const auto& data : target_delay_ms) { |
| const float x = config.GetCallTimeSec(Timestamp::Millis(data.first)); |
| const float y = data.second; |
| time_series_target_time.points.emplace_back(TimeSeriesPoint(x, y)); |
| } |
| |
| plot->AppendTimeSeries(std::move(time_series_packet_arrival)); |
| plot->AppendTimeSeries(std::move(time_series_relative_packet_arrival)); |
| plot->AppendTimeSeries(std::move(time_series_play_time)); |
| plot->AppendTimeSeries(std::move(time_series_target_time)); |
| |
| plot->SetXAxis(config.CallBeginTimeSec(), config.CallEndTimeSec(), "Time (s)", |
| kLeftMargin, kRightMargin); |
| plot->SetSuggestedYAxis(0, 1, "Relative delay (ms)", kBottomMargin, |
| kTopMargin); |
| plot->SetTitle("NetEq timing for " + |
| GetStreamName(parsed_log, kIncomingPacket, ssrc)); |
| } |
| |
| template <typename NetEqStatsType> |
| void CreateNetEqStatsGraphInternal( |
| const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| const NetEqStatsGetterMap& neteq_stats, |
| rtc::FunctionView<const std::vector<std::pair<int64_t, NetEqStatsType>>*( |
| const test::NetEqStatsGetter*)> data_extractor, |
| rtc::FunctionView<float(const NetEqStatsType&)> stats_extractor, |
| const std::string& plot_name, |
| Plot* plot) { |
| std::map<uint32_t, TimeSeries> time_series; |
| |
| for (const auto& st : neteq_stats) { |
| const uint32_t ssrc = st.first; |
| const std::vector<std::pair<int64_t, NetEqStatsType>>* data_vector = |
| data_extractor(st.second.get()); |
| for (const auto& data : *data_vector) { |
| const float time = config.GetCallTimeSec(Timestamp::Millis(data.first)); |
| const float value = stats_extractor(data.second); |
| time_series[ssrc].points.emplace_back(TimeSeriesPoint(time, value)); |
| } |
| } |
| |
| for (auto& series : time_series) { |
| series.second.label = |
| GetStreamName(parsed_log, kIncomingPacket, series.first); |
| series.second.line_style = LineStyle::kLine; |
| plot->AppendTimeSeries(std::move(series.second)); |
| } |
| |
| plot->SetXAxis(config.CallBeginTimeSec(), config.CallEndTimeSec(), "Time (s)", |
| kLeftMargin, kRightMargin); |
| plot->SetSuggestedYAxis(0, 1, plot_name, kBottomMargin, kTopMargin); |
| plot->SetTitle(plot_name); |
| } |
| |
| void CreateNetEqNetworkStatsGraph( |
| const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| const NetEqStatsGetterMap& neteq_stats, |
| rtc::FunctionView<float(const NetEqNetworkStatistics&)> stats_extractor, |
| const std::string& plot_name, |
| Plot* plot) { |
| CreateNetEqStatsGraphInternal<NetEqNetworkStatistics>( |
| parsed_log, config, neteq_stats, |
| [](const test::NetEqStatsGetter* stats_getter) { |
| return stats_getter->stats(); |
| }, |
| stats_extractor, plot_name, plot); |
| } |
| |
| void CreateNetEqLifetimeStatsGraph( |
| const ParsedRtcEventLog& parsed_log, |
| const AnalyzerConfig& config, |
| const NetEqStatsGetterMap& neteq_stats, |
| rtc::FunctionView<float(const NetEqLifetimeStatistics&)> stats_extractor, |
| const std::string& plot_name, |
| Plot* plot) { |
| CreateNetEqStatsGraphInternal<NetEqLifetimeStatistics>( |
| parsed_log, config, neteq_stats, |
| [](const test::NetEqStatsGetter* stats_getter) { |
| return stats_getter->lifetime_stats(); |
| }, |
| stats_extractor, plot_name, plot); |
| } |
| |
| } // namespace webrtc |