]> git.sesse.net Git - nageru/blobdiff - audio_mixer.h
Release Nageru 1.7.2.
[nageru] / audio_mixer.h
index 821171037f6faf70b11e7d8518484380ecbe587f..ebe142a74cbc9e6f81d125505022d88588cb00cd 100644 (file)
@@ -8,19 +8,20 @@
 //
 // All operations on AudioMixer (except destruction) are thread-safe.
 
-#include <math.h>
+#include <assert.h>
 #include <stdint.h>
+#include <zita-resampler/resampler.h>
 #include <atomic>
+#include <chrono>
+#include <functional>
 #include <map>
 #include <memory>
 #include <mutex>
 #include <set>
+#include <string>
 #include <vector>
-#include <zita-resampler/resampler.h>
 
-#include "alsa_input.h"
 #include "alsa_pool.h"
-#include "bmusb/bmusb.h"
 #include "correlation_measurer.h"
 #include "db.h"
 #include "defs.h"
@@ -30,6 +31,8 @@
 #include "resampling_queue.h"
 #include "stereocompressor.h"
 
+class DeviceSpecProto;
+
 namespace bmusb {
 struct AudioFormat;
 }  // namespace bmusb
@@ -43,7 +46,7 @@ enum EQBand {
 
 class AudioMixer {
 public:
-       AudioMixer(unsigned num_cards);
+       AudioMixer(unsigned num_capture_cards, unsigned num_ffmpeg_inputs);
        void reset_resampler(DeviceSpec device_spec);
        void reset_meters();
 
@@ -52,7 +55,7 @@ public:
        // (This is to avoid a deadlock where a card hangs on the mutex in add_audio()
        // while we are trying to shut it down from another thread that also holds
        // the mutex.) frame_length is in TIMEBASE units.
-       bool add_audio(DeviceSpec device_spec, const uint8_t *data, unsigned num_samples, bmusb::AudioFormat audio_format, int64_t frame_length);
+       bool add_audio(DeviceSpec device_spec, const uint8_t *data, unsigned num_samples, bmusb::AudioFormat audio_format, int64_t frame_length, std::chrono::steady_clock::time_point frame_time);
        bool add_silence(DeviceSpec device_spec, unsigned samples_per_frame, unsigned num_frames, int64_t frame_length);
 
        // If a given device is offline for whatever reason and cannot deliver audio
@@ -62,7 +65,7 @@ public:
        // affect it. Same true/false behavior as add_audio().
        bool silence_card(DeviceSpec device_spec, bool silence);
 
-       std::vector<float> get_output(double pts, unsigned num_samples, ResamplingQueue::RateAdjustmentPolicy rate_adjustment_policy);
+       std::vector<float> get_output(std::chrono::steady_clock::time_point ts, unsigned num_samples, ResamplingQueue::RateAdjustmentPolicy rate_adjustment_policy);
 
        float get_fader_volume(unsigned bus_index) const { return fader_volume_db[bus_index]; }
        void set_fader_volume(unsigned bus_index, float level_db) { fader_volume_db[bus_index] = level_db; }
@@ -299,7 +302,6 @@ public:
 private:
        struct AudioDevice {
                std::unique_ptr<ResamplingQueue> resampling_queue;
-               int64_t next_local_pts = 0;
                std::string display_name;
                unsigned capture_frequency = OUTPUT_FREQUENCY;
                // Which channels we consider interesting (ie., are part of some input_mapping).
@@ -325,13 +327,14 @@ private:
        std::vector<DeviceSpec> get_active_devices() const;
        void set_input_mapping_lock_held(const InputMapping &input_mapping);
 
-       unsigned num_cards;
+       unsigned num_capture_cards, num_ffmpeg_inputs;
 
        mutable std::timed_mutex audio_mutex;
 
        ALSAPool alsa_pool;
        AudioDevice video_cards[MAX_VIDEO_CARDS];  // Under audio_mutex.
        AudioDevice alsa_inputs[MAX_ALSA_CARDS];  // Under audio_mutex.
+       std::unique_ptr<AudioDevice[]> ffmpeg_inputs;  // Under audio_mutex.
 
        std::atomic<float> locut_cutoff_hz{120};
        StereoFilter locut[MAX_BUSES];  // Default cutoff 120 Hz, 24 dB/oct.
@@ -383,6 +386,28 @@ private:
        CorrelationMeasurer correlation;  // Under audio_measure_mutex.
        Resampler peak_resampler;  // Under audio_measure_mutex.
        std::atomic<float> peak{0.0f};
+
+       // Metrics.
+       std::atomic<double> metric_audio_loudness_short_lufs{0.0 / 0.0};
+       std::atomic<double> metric_audio_loudness_integrated_lufs{0.0 / 0.0};
+       std::atomic<double> metric_audio_loudness_range_low_lufs{0.0 / 0.0};
+       std::atomic<double> metric_audio_loudness_range_high_lufs{0.0 / 0.0};
+       std::atomic<double> metric_audio_peak_dbfs{0.0 / 0.0};
+       std::atomic<double> metric_audio_final_makeup_gain_db{0.0};
+       std::atomic<double> metric_audio_correlation{0.0};
+
+       // These are all gauges corresponding to the elements of BusLevel.
+       // In a sense, they'd probably do better as histograms, but that's an
+       // awful lot of time series when you have many buses.
+       struct BusMetrics {
+               std::vector<std::pair<std::string, std::string>> labels;
+               std::atomic<double> current_level_dbfs[2]{{0.0/0.0},{0.0/0.0}};
+               std::atomic<double> peak_level_dbfs[2]{{0.0/0.0},{0.0/0.0}};
+               std::atomic<double> historic_peak_dbfs{0.0/0.0};
+               std::atomic<double> gain_staging_db{0.0/0.0};
+               std::atomic<double> compressor_attenuation_db{0.0/0.0};
+       };
+       std::unique_ptr<BusMetrics[]> bus_metrics;  // One for each bus in <input_mapping>.
 };
 
 extern AudioMixer *global_audio_mixer;