[nageru] / audio_mixer.cpp

#include "audio_mixer.h"

#include <assert.h>
#include <endian.h>
#include <bmusb/bmusb.h>
#include <stdio.h>
#include <endian.h>
#include <cmath>

#include "db.h"
#include "flags.h"
#include "timebase.h"

using namespace bmusb;
using namespace std;
using namespace std::placeholders;

namespace {

// TODO: If these prove to be a bottleneck, they can be SSSE3-optimized
// (usually including multiple channels at a time).

void convert_fixed16_to_fp32(float *dst, size_t out_channel, size_t out_num_channels,
                             const uint8_t *src, size_t in_channel, size_t in_num_channels,
                             size_t num_samples)
{
        assert(in_channel < in_num_channels);
        assert(out_channel < out_num_channels);
        src += in_channel * 2;
        dst += out_channel;

        for (size_t i = 0; i < num_samples; ++i) {
                int16_t s = le16toh(*(int16_t *)src);
                *dst = s * (1.0f / 32768.0f);

                src += 2 * in_num_channels;
                dst += out_num_channels;
        }
}

void convert_fixed24_to_fp32(float *dst, size_t out_channel, size_t out_num_channels,
                             const uint8_t *src, size_t in_channel, size_t in_num_channels,
                             size_t num_samples)
{
        assert(in_channel < in_num_channels);
        assert(out_channel < out_num_channels);
        src += in_channel * 3;
        dst += out_channel;

        for (size_t i = 0; i < num_samples; ++i) {
                uint32_t s1 = src[0];
                uint32_t s2 = src[1];
                uint32_t s3 = src[2];
                uint32_t s = s1 | (s1 << 8) | (s2 << 16) | (s3 << 24);
                *dst = int(s) * (1.0f / 2147483648.0f);

                src += 3 * in_num_channels;
                dst += out_num_channels;
        }
}

void convert_fixed32_to_fp32(float *dst, size_t out_channel, size_t out_num_channels,
                             const uint8_t *src, size_t in_channel, size_t in_num_channels,
                             size_t num_samples)
{
        assert(in_channel < in_num_channels);
        assert(out_channel < out_num_channels);
        src += in_channel * 4;
        dst += out_channel;

        for (size_t i = 0; i < num_samples; ++i) {
                int32_t s = le32toh(*(int32_t *)src);
                *dst = s * (1.0f / 2147483648.0f);

                src += 4 * in_num_channels;
                dst += out_num_channels;
        }
}

float find_peak(const float *samples, size_t num_samples)
{
        float m = fabs(samples[0]);
        for (size_t i = 1; i < num_samples; ++i) {
                m = max(m, fabs(samples[i]));
        }
        return m;
}

void deinterleave_samples(const vector<float> &in, vector<float> *out_l, vector<float> *out_r)
{
        size_t num_samples = in.size() / 2;
        out_l->resize(num_samples);
        out_r->resize(num_samples);

        const float *inptr = in.data();
        float *lptr = &(*out_l)[0];
        float *rptr = &(*out_r)[0];
        for (size_t i = 0; i < num_samples; ++i) {
                *lptr++ = *inptr++;
                *rptr++ = *inptr++;
        }
}

}  // namespace

AudioMixer::AudioMixer(unsigned num_cards)
        : num_cards(num_cards),
          level_compressor(OUTPUT_FREQUENCY),
          limiter(OUTPUT_FREQUENCY),
          compressor(OUTPUT_FREQUENCY),
          correlation(OUTPUT_FREQUENCY)
{
        locut.init(FILTER_HPF, 2);

        set_locut_enabled(global_flags.locut_enabled);
        set_gain_staging_db(global_flags.initial_gain_staging_db);
        set_gain_staging_auto(global_flags.gain_staging_auto);
        set_compressor_enabled(global_flags.compressor_enabled);
        set_limiter_enabled(global_flags.limiter_enabled);
        set_final_makeup_gain_auto(global_flags.final_makeup_gain_auto);

        // Generate a very simple, default input mapping.
        InputMapping::Bus input;
        input.name = "Main";
        input.device.type = InputSourceType::CAPTURE_CARD;
        input.device.index = 0;
        input.source_channel[0] = 0;
        input.source_channel[1] = 1;

        InputMapping new_input_mapping;
        new_input_mapping.buses.push_back(input);
        set_input_mapping(new_input_mapping);

        // Look for ALSA cards.
        available_alsa_cards = ALSAInput::enumerate_devices();

        r128.init(2, OUTPUT_FREQUENCY);
        r128.integr_start();

        // hlen=16 is pretty low quality, but we use quite a bit of CPU otherwise,
        // and there's a limit to how important the peak meter is.
        peak_resampler.setup(OUTPUT_FREQUENCY, OUTPUT_FREQUENCY * 4, /*num_channels=*/2, /*hlen=*/16, /*frel=*/1.0);
}

AudioMixer::~AudioMixer()
{
        for (unsigned card_index = 0; card_index < available_alsa_cards.size(); ++card_index) {
                const AudioDevice &device = alsa_inputs[card_index];
                if (device.alsa_device != nullptr) {
                        device.alsa_device->stop_capture_thread();
                }
        }
}


void AudioMixer::reset_resampler(DeviceSpec device_spec)
{
        lock_guard<timed_mutex> lock(audio_mutex);
        reset_resampler_mutex_held(device_spec);
}

void AudioMixer::reset_resampler_mutex_held(DeviceSpec device_spec)
{
        AudioDevice *device = find_audio_device(device_spec);

        if (device->interesting_channels.empty()) {
                device->resampling_queue.reset();
        } else {
                // TODO: ResamplingQueue should probably take the full device spec.
                // (It's only used for console output, though.)
                device->resampling_queue.reset(new ResamplingQueue(device_spec.index, device->capture_frequency, OUTPUT_FREQUENCY, device->interesting_channels.size()));
        }
        device->next_local_pts = 0;
}

void AudioMixer::reset_alsa_mutex_held(DeviceSpec device_spec)
{
        assert(device_spec.type == InputSourceType::ALSA_INPUT);
        unsigned card_index = device_spec.index;
        AudioDevice *device = find_audio_device(device_spec);

        if (device->alsa_device != nullptr) {
                device->alsa_device->stop_capture_thread();
        }
        if (device->interesting_channels.empty()) {
                device->alsa_device.reset();
        } else {
                device->alsa_device.reset(new ALSAInput(available_alsa_cards[card_index].address.c_str(), OUTPUT_FREQUENCY, 2, bind(&AudioMixer::add_audio, this, device_spec, _1, _2, _3, _4)));
                device->capture_frequency = device->alsa_device->get_sample_rate();
                device->alsa_device->start_capture_thread();
        }
}

bool AudioMixer::add_audio(DeviceSpec device_spec, const uint8_t *data, unsigned num_samples, AudioFormat audio_format, int64_t frame_length)
{
        AudioDevice *device = find_audio_device(device_spec);

        unique_lock<timed_mutex> lock(audio_mutex, defer_lock);
        if (!lock.try_lock_for(chrono::milliseconds(10))) {
                return false;
        }
        if (device->resampling_queue == nullptr) {
                // No buses use this device; throw it away.
                return true;
        }

        unsigned num_channels = device->interesting_channels.size();
        assert(num_channels > 0);

        // Convert the audio to fp32.
        vector<float> audio;
        audio.resize(num_samples * num_channels);
        unsigned channel_index = 0;
        for (auto channel_it = device->interesting_channels.cbegin(); channel_it != device->interesting_channels.end(); ++channel_it, ++channel_index) {
                switch (audio_format.bits_per_sample) {
                case 0:
                        assert(num_samples == 0);
                        break;
                case 16:
                        convert_fixed16_to_fp32(&audio[0], channel_index, num_channels, data, *channel_it, audio_format.num_channels, num_samples);
                        break;
                case 24:
                        convert_fixed24_to_fp32(&audio[0], channel_index, num_channels, data, *channel_it, audio_format.num_channels, num_samples);
                        break;
                case 32:
                        convert_fixed32_to_fp32(&audio[0], channel_index, num_channels, data, *channel_it, audio_format.num_channels, num_samples);
                        break;
                default:
                        fprintf(stderr, "Cannot handle audio with %u bits per sample\n", audio_format.bits_per_sample);
                        assert(false);
                }
        }

        // Now add it.
        int64_t local_pts = device->next_local_pts;
        device->resampling_queue->add_input_samples(local_pts / double(TIMEBASE), audio.data(), num_samples);
        device->next_local_pts = local_pts + frame_length;
        return true;
}

bool AudioMixer::add_silence(DeviceSpec device_spec, unsigned samples_per_frame, unsigned num_frames, int64_t frame_length)
{
        AudioDevice *device = find_audio_device(device_spec);

        unique_lock<timed_mutex> lock(audio_mutex, defer_lock);
        if (!lock.try_lock_for(chrono::milliseconds(10))) {
                return false;
        }
        if (device->resampling_queue == nullptr) {
                // No buses use this device; throw it away.
                return true;
        }

        unsigned num_channels = device->interesting_channels.size();
        assert(num_channels > 0);

        vector<float> silence(samples_per_frame * num_channels, 0.0f);
        for (unsigned i = 0; i < num_frames; ++i) {
                device->resampling_queue->add_input_samples(device->next_local_pts / double(TIMEBASE), silence.data(), samples_per_frame);
                // Note that if the format changed in the meantime, we have
                // no way of detecting that; we just have to assume the frame length
                // is always the same.
                device->next_local_pts += frame_length;
        }
        return true;
}

AudioMixer::AudioDevice *AudioMixer::find_audio_device(DeviceSpec device)
{
        switch (device.type) {
        case InputSourceType::CAPTURE_CARD:
                return &video_cards[device.index];
        case InputSourceType::ALSA_INPUT:
                return &alsa_inputs[device.index];
        case InputSourceType::SILENCE:
        default:
                assert(false);
        }
        return nullptr;
}

// Get a pointer to the given channel from the given device.
// The channel must be picked out earlier and resampled.
void AudioMixer::find_sample_src_from_device(const map<DeviceSpec, vector<float>> &samples_card, DeviceSpec device_spec, int source_channel, const float **srcptr, unsigned *stride)
{
        static float zero = 0.0f;
        if (source_channel == -1 || device_spec.type == InputSourceType::SILENCE) {
                *srcptr = &zero;
                *stride = 0;
                return;
        }
        AudioDevice *device = find_audio_device(device_spec);
        assert(device->interesting_channels.count(source_channel) != 0);
        unsigned channel_index = 0;
        for (int channel : device->interesting_channels) {
                if (channel == source_channel) break;
                ++channel_index;
        }
        assert(channel_index < device->interesting_channels.size());
        const auto it = samples_card.find(device_spec);
        assert(it != samples_card.end());
        *srcptr = &(it->second)[channel_index];
        *stride = device->interesting_channels.size();
}

// TODO: Can be SSSE3-optimized if need be.
void AudioMixer::fill_audio_bus(const map<DeviceSpec, vector<float>> &samples_card, const InputMapping::Bus &bus, unsigned num_samples, float *output)
{
        if (bus.device.type == InputSourceType::SILENCE) {
                memset(output, 0, num_samples * sizeof(*output));
        } else {
                assert(bus.device.type == InputSourceType::CAPTURE_CARD ||
                       bus.device.type == InputSourceType::ALSA_INPUT);
                const float *lsrc, *rsrc;
                unsigned lstride, rstride;
                float *dptr = output;
                find_sample_src_from_device(samples_card, bus.device, bus.source_channel[0], &lsrc, &lstride);
                find_sample_src_from_device(samples_card, bus.device, bus.source_channel[1], &rsrc, &rstride);
                for (unsigned i = 0; i < num_samples; ++i) {
                        *dptr++ = *lsrc;
                        *dptr++ = *rsrc;
                        lsrc += lstride;
                        rsrc += rstride;
                }
        }
}

vector<float> AudioMixer::get_output(double pts, unsigned num_samples, ResamplingQueue::RateAdjustmentPolicy rate_adjustment_policy)
{
        map<DeviceSpec, vector<float>> samples_card;
        vector<float> samples_bus;

        lock_guard<timed_mutex> lock(audio_mutex);

        // Pick out all the interesting channels from all the cards.
        // TODO: If the card has been hotswapped, the number of channels
        // might have changed; if so, we need to do some sort of remapping
        // to silence.
        for (const auto &spec_and_info : get_devices_mutex_held()) {
                const DeviceSpec &device_spec = spec_and_info.first;
                AudioDevice *device = find_audio_device(device_spec);
                if (!device->interesting_channels.empty()) {
                        samples_card[device_spec].resize(num_samples * device->interesting_channels.size());
                        device->resampling_queue->get_output_samples(
                                pts,
                                &samples_card[device_spec][0],
                                num_samples,
                                rate_adjustment_policy);
                }
        }

        // TODO: Move lo-cut etc. into each bus.
        vector<float> samples_out, left, right;
        samples_out.resize(num_samples * 2);
        samples_bus.resize(num_samples * 2);
        for (unsigned bus_index = 0; bus_index < input_mapping.buses.size(); ++bus_index) {
                fill_audio_bus(samples_card, input_mapping.buses[bus_index], num_samples, &samples_bus[0]);

                // TODO: We should measure post-fader.
                deinterleave_samples(samples_bus, &left, &right);
                measure_bus_levels(bus_index, left, right);

                float volume = from_db(fader_volume_db[bus_index]);
                if (bus_index == 0) {
                        for (unsigned i = 0; i < num_samples * 2; ++i) {
                                samples_out[i] = samples_bus[i] * volume;
                        }
                } else {
                        for (unsigned i = 0; i < num_samples * 2; ++i) {
                                samples_out[i] += samples_bus[i] * volume;
                        }
                }
        }

        // Cut away everything under 120 Hz (or whatever the cutoff is);
        // we don't need it for voice, and it will reduce headroom
        // and confuse the compressor. (In particular, any hums at 50 or 60 Hz
        // should be dampened.)
        if (locut_enabled) {
                locut.render(samples_out.data(), samples_out.size() / 2, locut_cutoff_hz * 2.0 * M_PI / OUTPUT_FREQUENCY, 0.5f);
        }

        {
                lock_guard<mutex> lock(compressor_mutex);

                // Apply a level compressor to get the general level right.
                // Basically, if it's over about -40 dBFS, we squeeze it down to that level
                // (or more precisely, near it, since we don't use infinite ratio),
                // then apply a makeup gain to get it to -14 dBFS. -14 dBFS is, of course,
                // entirely arbitrary, but from practical tests with speech, it seems to
                // put ut around -23 LUFS, so it's a reasonable starting point for later use.
                {
                        if (level_compressor_enabled) {
                                float threshold = 0.01f;   // -40 dBFS.
                                float ratio = 20.0f;
                                float attack_time = 0.5f;
                                float release_time = 20.0f;
                                float makeup_gain = from_db(ref_level_dbfs - (-40.0f));  // +26 dB.
                                level_compressor.process(samples_out.data(), samples_out.size() / 2, threshold, ratio, attack_time, release_time, makeup_gain);
                                gain_staging_db = to_db(level_compressor.get_attenuation() * makeup_gain);
                        } else {
                                // Just apply the gain we already had.
                                float g = from_db(gain_staging_db);
                                for (size_t i = 0; i < samples_out.size(); ++i) {
                                        samples_out[i] *= g;
                                }
                        }
                }

        #if 0
                printf("level=%f (%+5.2f dBFS) attenuation=%f (%+5.2f dB) end_result=%+5.2f dB\n",
                        level_compressor.get_level(), to_db(level_compressor.get_level()),
                        level_compressor.get_attenuation(), to_db(level_compressor.get_attenuation()),
                        to_db(level_compressor.get_level() * level_compressor.get_attenuation() * makeup_gain));
        #endif

        //      float limiter_att, compressor_att;

                // The real compressor.
                if (compressor_enabled) {
                        float threshold = from_db(compressor_threshold_dbfs);
                        float ratio = 20.0f;
                        float attack_time = 0.005f;
                        float release_time = 0.040f;
                        float makeup_gain = 2.0f;  // +6 dB.
                        compressor.process(samples_out.data(), samples_out.size() / 2, threshold, ratio, attack_time, release_time, makeup_gain);
        //              compressor_att = compressor.get_attenuation();
                }

                // Finally a limiter at -4 dB (so, -10 dBFS) to take out the worst peaks only.
                // Note that since ratio is not infinite, we could go slightly higher than this.
                if (limiter_enabled) {
                        float threshold = from_db(limiter_threshold_dbfs);
                        float ratio = 30.0f;
                        float attack_time = 0.0f;  // Instant.
                        float release_time = 0.020f;
                        float makeup_gain = 1.0f;  // 0 dB.
                        limiter.process(samples_out.data(), samples_out.size() / 2, threshold, ratio, attack_time, release_time, makeup_gain);
        //              limiter_att = limiter.get_attenuation();
                }

        //      printf("limiter=%+5.1f  compressor=%+5.1f\n", to_db(limiter_att), to_db(compressor_att));
        }

        // At this point, we are most likely close to +0 LU, but all of our
        // measurements have been on raw sample values, not R128 values.
        // So we have a final makeup gain to get us to +0 LU; the gain
        // adjustments required should be relatively small, and also, the
        // offset shouldn't change much (only if the type of audio changes
        // significantly). Thus, we shoot for updating this value basically
        // “whenever we process buffers”, since the R128 calculation isn't exactly
        // something we get out per-sample.
        //
        // Note that there's a feedback loop here, so we choose a very slow filter
        // (half-time of 30 seconds).
        double target_loudness_factor, alpha;
        double loudness_lu = r128.loudness_M() - ref_level_lufs;
        double current_makeup_lu = to_db(final_makeup_gain);
        target_loudness_factor = final_makeup_gain * from_db(-loudness_lu);

        // If we're outside +/- 5 LU uncorrected, we don't count it as
        // a normal signal (probably silence) and don't change the
        // correction factor; just apply what we already have.
        if (fabs(loudness_lu - current_makeup_lu) >= 5.0 || !final_makeup_gain_auto) {
                alpha = 0.0;
        } else {
                // Formula adapted from
                // https://en.wikipedia.org/wiki/Low-pass_filter#Simple_infinite_impulse_response_filter.
                const double half_time_s = 30.0;
                const double fc_mul_2pi_delta_t = 1.0 / (half_time_s * OUTPUT_FREQUENCY);
                alpha = fc_mul_2pi_delta_t / (fc_mul_2pi_delta_t + 1.0);
        }

        {
                lock_guard<mutex> lock(compressor_mutex);
                double m = final_makeup_gain;
                for (size_t i = 0; i < samples_out.size(); i += 2) {
                        samples_out[i + 0] *= m;
                        samples_out[i + 1] *= m;
                        m += (target_loudness_factor - m) * alpha;
                }
                final_makeup_gain = m;
        }

        update_meters(samples_out);

        return samples_out;
}

void AudioMixer::measure_bus_levels(unsigned bus_index, const vector<float> &left, const vector<float> &right)
{
        const float *ptrs[] = { left.data(), right.data() };
        {
                lock_guard<mutex> lock(audio_measure_mutex);
                bus_r128[bus_index]->process(left.size(), const_cast<float **>(ptrs));
        }
}

void AudioMixer::update_meters(const vector<float> &samples)
{
        // Upsample 4x to find interpolated peak.
        peak_resampler.inp_data = const_cast<float *>(samples.data());
        peak_resampler.inp_count = samples.size() / 2;

        vector<float> interpolated_samples;
        interpolated_samples.resize(samples.size());
        {
                lock_guard<mutex> lock(audio_measure_mutex);

                while (peak_resampler.inp_count > 0) {  // About four iterations.
                        peak_resampler.out_data = &interpolated_samples[0];
                        peak_resampler.out_count = interpolated_samples.size() / 2;
                        peak_resampler.process();
                        size_t out_stereo_samples = interpolated_samples.size() / 2 - peak_resampler.out_count;
                        peak = max<float>(peak, find_peak(interpolated_samples.data(), out_stereo_samples * 2));
                        peak_resampler.out_data = nullptr;
                }
        }

        // Find R128 levels and L/R correlation.
        vector<float> left, right;
        deinterleave_samples(samples, &left, &right);
        float *ptrs[] = { left.data(), right.data() };
        {
                lock_guard<mutex> lock(audio_measure_mutex);
                r128.process(left.size(), ptrs);
                correlation.process_samples(samples);
        }

        send_audio_level_callback();
}

void AudioMixer::reset_meters()
{
        lock_guard<mutex> lock(audio_measure_mutex);
        peak_resampler.reset();
        peak = 0.0f;
        r128.reset();
        r128.integr_start();
        correlation.reset();
}

void AudioMixer::send_audio_level_callback()
{
        if (audio_level_callback == nullptr) {
                return;
        }

        lock_guard<mutex> lock(audio_measure_mutex);
        double loudness_s = r128.loudness_S();
        double loudness_i = r128.integrated();
        double loudness_range_low = r128.range_min();
        double loudness_range_high = r128.range_max();

        vector<float> bus_loudness;
        bus_loudness.resize(input_mapping.buses.size());
        for (unsigned bus_index = 0; bus_index < bus_r128.size(); ++bus_index) {
                bus_loudness[bus_index] = bus_r128[bus_index]->loudness_S();
        }

        audio_level_callback(loudness_s, to_db(peak), bus_loudness,
                loudness_i, loudness_range_low, loudness_range_high,
                gain_staging_db,
                to_db(final_makeup_gain),
                correlation.get_correlation());
}

map<DeviceSpec, DeviceInfo> AudioMixer::get_devices() const
{
        lock_guard<timed_mutex> lock(audio_mutex);
        return get_devices_mutex_held();
}

map<DeviceSpec, DeviceInfo> AudioMixer::get_devices_mutex_held() const
{
        map<DeviceSpec, DeviceInfo> devices;
        for (unsigned card_index = 0; card_index < num_cards; ++card_index) {
                const DeviceSpec spec{ InputSourceType::CAPTURE_CARD, card_index };
                const AudioDevice *device = &video_cards[card_index];
                DeviceInfo info;
                info.name = device->name;
                info.num_channels = 8;  // FIXME: This is wrong for fake cards.
                devices.insert(make_pair(spec, info));
        }
        for (unsigned card_index = 0; card_index < available_alsa_cards.size(); ++card_index) {
                const DeviceSpec spec{ InputSourceType::ALSA_INPUT, card_index };
                const ALSAInput::Device &device = available_alsa_cards[card_index];
                DeviceInfo info;
                info.name = device.name + " (" + device.info + ")";
                info.num_channels = device.num_channels;
                devices.insert(make_pair(spec, info));
        }
        return devices;
}

void AudioMixer::set_name(DeviceSpec device_spec, const string &name)
{
        AudioDevice *device = find_audio_device(device_spec);

        lock_guard<timed_mutex> lock(audio_mutex);
        device->name = name;
}

void AudioMixer::set_input_mapping(const InputMapping &new_input_mapping)
{
        lock_guard<timed_mutex> lock(audio_mutex);

        map<DeviceSpec, set<unsigned>> interesting_channels;
        for (const InputMapping::Bus &bus : new_input_mapping.buses) {
                if (bus.device.type == InputSourceType::CAPTURE_CARD ||
                    bus.device.type == InputSourceType::ALSA_INPUT) {
                        for (unsigned channel = 0; channel < 2; ++channel) {
                                if (bus.source_channel[channel] != -1) {
                                        interesting_channels[bus.device].insert(bus.source_channel[channel]);
                                }
                        }
                }
        }

        // Reset resamplers for all cards that don't have the exact same state as before.
        for (const auto &spec_and_info : get_devices_mutex_held()) {
                const DeviceSpec &device_spec = spec_and_info.first;
                AudioDevice *device = find_audio_device(device_spec);
                if (device->interesting_channels != interesting_channels[device_spec]) {
                        device->interesting_channels = interesting_channels[device_spec];
                        if (device_spec.type == InputSourceType::ALSA_INPUT) {
                                reset_alsa_mutex_held(device_spec);
                        }
                        reset_resampler_mutex_held(device_spec);
                }
        }

        {
                lock_guard<mutex> lock(audio_measure_mutex);
                bus_r128.resize(new_input_mapping.buses.size());
                for (unsigned bus_index = 0; bus_index < bus_r128.size(); ++bus_index) {
                        if (bus_r128[bus_index] == nullptr) {
                                bus_r128[bus_index].reset(new Ebu_r128_proc);
                        }
                        bus_r128[bus_index]->init(2, OUTPUT_FREQUENCY);
                }
        }

        input_mapping = new_input_mapping;
}

InputMapping AudioMixer::get_input_mapping() const
{
        lock_guard<timed_mutex> lock(audio_mutex);
        return input_mapping;
}