Factor the basic metrics (memory, uptime, etc.) into a separate class, so that Kaeru...

[nageru] / mixer.cpp

#undef Success

#include "mixer.h"

#include <assert.h>
#include <epoxy/egl.h>
#include <movit/effect_chain.h>
#include <movit/effect_util.h>
#include <movit/flat_input.h>
#include <movit/image_format.h>
#include <movit/init.h>
#include <movit/resource_pool.h>
#include <pthread.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <algorithm>
#include <chrono>
#include <condition_variable>
#include <cstddef>
#include <cstdint>
#include <memory>
#include <mutex>
#include <ratio>
#include <string>
#include <thread>
#include <utility>
#include <vector>

#include "DeckLinkAPI.h"
#include "LinuxCOM.h"
#include "alsa_output.h"
#include "basic_stats.h"
#include "bmusb/bmusb.h"
#include "bmusb/fake_capture.h"
#include "chroma_subsampler.h"
#include "context.h"
#include "decklink_capture.h"
#include "decklink_output.h"
#include "defs.h"
#include "disk_space_estimator.h"
#include "ffmpeg_capture.h"
#include "flags.h"
#include "input_mapping.h"
#include "metrics.h"
#include "pbo_frame_allocator.h"
#include "ref_counted_gl_sync.h"
#include "resampling_queue.h"
#include "timebase.h"
#include "timecode_renderer.h"
#include "v210_converter.h"
#include "video_encoder.h"

class IDeckLink;
class QOpenGLContext;

using namespace movit;
using namespace std;
using namespace std::chrono;
using namespace std::placeholders;
using namespace bmusb;

Mixer *global_mixer = nullptr;

namespace {

void insert_new_frame(RefCountedFrame frame, unsigned field_num, bool interlaced, unsigned card_index, InputState *input_state)
{
        if (interlaced) {
                for (unsigned frame_num = FRAME_HISTORY_LENGTH; frame_num --> 1; ) {  // :-)
                        input_state->buffered_frames[card_index][frame_num] =
                                input_state->buffered_frames[card_index][frame_num - 1];
                }
                input_state->buffered_frames[card_index][0] = { frame, field_num };
        } else {
                for (unsigned frame_num = 0; frame_num < FRAME_HISTORY_LENGTH; ++frame_num) {
                        input_state->buffered_frames[card_index][frame_num] = { frame, field_num };
                }
        }
}

void ensure_texture_resolution(PBOFrameAllocator::Userdata *userdata, unsigned field, unsigned width, unsigned height, unsigned cbcr_width, unsigned cbcr_height, unsigned v210_width)
{
        bool first;
        switch (userdata->pixel_format) {
        case PixelFormat_10BitYCbCr:
                first = userdata->tex_v210[field] == 0 || userdata->tex_444[field] == 0;
                break;
        case PixelFormat_8BitYCbCr:
                first = userdata->tex_y[field] == 0 || userdata->tex_cbcr[field] == 0;
                break;
        case PixelFormat_8BitBGRA:
                first = userdata->tex_rgba[field] == 0;
                break;
        case PixelFormat_8BitYCbCrPlanar:
                first = userdata->tex_y[field] == 0 || userdata->tex_cb[field] == 0 || userdata->tex_cr[field] == 0;
                break;
        default:
                assert(false);
        }

        if (first ||
            width != userdata->last_width[field] ||
            height != userdata->last_height[field] ||
            cbcr_width != userdata->last_cbcr_width[field] ||
            cbcr_height != userdata->last_cbcr_height[field]) {
                // We changed resolution since last use of this texture, so we need to create
                // a new object. Note that this each card has its own PBOFrameAllocator,
                // we don't need to worry about these flip-flopping between resolutions.
                switch (userdata->pixel_format) {
                case PixelFormat_10BitYCbCr:
                        glBindTexture(GL_TEXTURE_2D, userdata->tex_444[field]);
                        check_error();
                        glTexImage2D(GL_TEXTURE_2D, 0, GL_RGB10_A2, width, height, 0, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, nullptr);
                        check_error();
                        break;
                case PixelFormat_8BitYCbCr: {
                        glBindTexture(GL_TEXTURE_2D, userdata->tex_cbcr[field]);
                        check_error();
                        glTexImage2D(GL_TEXTURE_2D, 0, GL_RG8, cbcr_width, height, 0, GL_RG, GL_UNSIGNED_BYTE, nullptr);
                        check_error();
                        glBindTexture(GL_TEXTURE_2D, userdata->tex_y[field]);
                        check_error();
                        glTexImage2D(GL_TEXTURE_2D, 0, GL_R8, width, height, 0, GL_RED, GL_UNSIGNED_BYTE, nullptr);
                        check_error();
                        break;
                }
                case PixelFormat_8BitYCbCrPlanar: {
                        glBindTexture(GL_TEXTURE_2D, userdata->tex_y[field]);
                        check_error();
                        glTexImage2D(GL_TEXTURE_2D, 0, GL_R8, width, height, 0, GL_RED, GL_UNSIGNED_BYTE, nullptr);
                        check_error();
                        glBindTexture(GL_TEXTURE_2D, userdata->tex_cb[field]);
                        check_error();
                        glTexImage2D(GL_TEXTURE_2D, 0, GL_R8, cbcr_width, cbcr_height, 0, GL_RED, GL_UNSIGNED_BYTE, nullptr);
                        check_error();
                        glBindTexture(GL_TEXTURE_2D, userdata->tex_cr[field]);
                        check_error();
                        glTexImage2D(GL_TEXTURE_2D, 0, GL_R8, cbcr_width, cbcr_height, 0, GL_RED, GL_UNSIGNED_BYTE, nullptr);
                        check_error();
                        break;
                }
                case PixelFormat_8BitBGRA:
                        glBindTexture(GL_TEXTURE_2D, userdata->tex_rgba[field]);
                        check_error();
                        if (global_flags.can_disable_srgb_decoder) {  // See the comments in tweaked_inputs.h.
                                glTexImage2D(GL_TEXTURE_2D, 0, GL_SRGB8_ALPHA8, width, height, 0, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, nullptr);
                        } else {
                                glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, width, height, 0, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, nullptr);
                        }
                        check_error();
                        break;
                }
                userdata->last_width[field] = width;
                userdata->last_height[field] = height;
                userdata->last_cbcr_width[field] = cbcr_width;
                userdata->last_cbcr_height[field] = cbcr_height;
        }
        if (global_flags.ten_bit_input &&
            (first || v210_width != userdata->last_v210_width[field])) {
                // Same as above; we need to recreate the texture.
                glBindTexture(GL_TEXTURE_2D, userdata->tex_v210[field]);
                check_error();
                glTexImage2D(GL_TEXTURE_2D, 0, GL_RGB10_A2, v210_width, height, 0, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, nullptr);
                check_error();
                userdata->last_v210_width[field] = v210_width;
        }
}

void upload_texture(GLuint tex, GLuint width, GLuint height, GLuint stride, bool interlaced_stride, GLenum format, GLenum type, GLintptr offset)
{
        if (interlaced_stride) {
                stride *= 2;
        }
        if (global_flags.flush_pbos) {
                glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, offset, stride * height);
                check_error();
        }

        glBindTexture(GL_TEXTURE_2D, tex);
        check_error();
        if (interlaced_stride) {
                glPixelStorei(GL_UNPACK_ROW_LENGTH, width * 2);
                check_error();
        } else {
                glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
                check_error();
        }

        glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, width, height, format, type, BUFFER_OFFSET(offset));
        check_error();
        glBindTexture(GL_TEXTURE_2D, 0);
        check_error();
        glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
        check_error();
}

}  // namespace

void JitterHistory::register_metrics(const vector<pair<string, string>> &labels)
{
        global_metrics.add("input_underestimated_jitter_frames", labels, &metric_input_underestimated_jitter_frames);
        global_metrics.add("input_estimated_max_jitter_seconds", labels, &metric_input_estimated_max_jitter_seconds, Metrics::TYPE_GAUGE);
}

void JitterHistory::unregister_metrics(const vector<pair<string, string>> &labels)
{
        global_metrics.remove("input_underestimated_jitter_frames", labels);
        global_metrics.remove("input_estimated_max_jitter_seconds", labels);
}

void JitterHistory::frame_arrived(steady_clock::time_point now, int64_t frame_duration, size_t dropped_frames)
{
        if (expected_timestamp > steady_clock::time_point::min()) {
                expected_timestamp += dropped_frames * nanoseconds(frame_duration * 1000000000 / TIMEBASE);
                double jitter_seconds = fabs(duration<double>(expected_timestamp - now).count());
                history.push_back(orders.insert(jitter_seconds));
                if (jitter_seconds > estimate_max_jitter()) {
                        ++metric_input_underestimated_jitter_frames;
                }

                metric_input_estimated_max_jitter_seconds = estimate_max_jitter();

                if (history.size() > history_length) {
                        orders.erase(history.front());
                        history.pop_front();
                }
                assert(history.size() <= history_length);
        }
        expected_timestamp = now + nanoseconds(frame_duration * 1000000000 / TIMEBASE);
}

double JitterHistory::estimate_max_jitter() const
{
        if (orders.empty()) {
                return 0.0;
        }
        size_t elem_idx = lrint((orders.size() - 1) * percentile);
        if (percentile <= 0.5) {
                return *next(orders.begin(), elem_idx) * multiplier;
        } else {
                return *prev(orders.end(), elem_idx + 1) * multiplier;
        }
}

void QueueLengthPolicy::register_metrics(const vector<pair<string, string>> &labels)
{
        global_metrics.add("input_queue_safe_length_frames", labels, &metric_input_queue_safe_length_frames, Metrics::TYPE_GAUGE);
}

void QueueLengthPolicy::unregister_metrics(const vector<pair<string, string>> &labels)
{
        global_metrics.remove("input_queue_safe_length_frames", labels);
}

void QueueLengthPolicy::update_policy(steady_clock::time_point now,
                                      steady_clock::time_point expected_next_frame,
                                      int64_t input_frame_duration,
                                      int64_t master_frame_duration,
                                      double max_input_card_jitter_seconds,
                                      double max_master_card_jitter_seconds)
{
        double input_frame_duration_seconds = input_frame_duration / double(TIMEBASE);
        double master_frame_duration_seconds = master_frame_duration / double(TIMEBASE);

        // Figure out when we can expect the next frame for this card, assuming
        // worst-case jitter (ie., the frame is maximally late).
        double seconds_until_next_frame = max(duration<double>(expected_next_frame - now).count() + max_input_card_jitter_seconds, 0.0);

        // How many times are the master card expected to tick in that time?
        // We assume the master clock has worst-case jitter but not any rate
        // discrepancy, ie., it ticks as early as possible every time, but not
        // cumulatively.
        double frames_needed = (seconds_until_next_frame + max_master_card_jitter_seconds) / master_frame_duration_seconds;

        // As a special case, if the master card ticks faster than the input card,
        // we expect the queue to drain by itself even without dropping. But if
        // the difference is small (e.g. 60 Hz master and 59.94 input), it would
        // go slowly enough that the effect wouldn't really be appreciable.
        // We account for this by looking at the situation five frames ahead,
        // assuming everything else is the same.
        double frames_allowed;
        if (master_frame_duration < input_frame_duration) {
                frames_allowed = frames_needed + 5 * (input_frame_duration_seconds - master_frame_duration_seconds) / master_frame_duration_seconds;
        } else {
                frames_allowed = frames_needed;
        }

        safe_queue_length = max<int>(floor(frames_allowed), 0);
        metric_input_queue_safe_length_frames = safe_queue_length;
}

Mixer::Mixer(const QSurfaceFormat &format, unsigned num_cards)
        : httpd(),
          num_cards(num_cards),
          mixer_surface(create_surface(format)),
          h264_encoder_surface(create_surface(format)),
          decklink_output_surface(create_surface(format)),
          audio_mixer(num_cards)
{
        memcpy(ycbcr_interpretation, global_flags.ycbcr_interpretation, sizeof(ycbcr_interpretation));
        CHECK(init_movit(MOVIT_SHADER_DIR, MOVIT_DEBUG_OFF));
        check_error();

        // This nearly always should be true.
        global_flags.can_disable_srgb_decoder =
                epoxy_has_gl_extension("GL_EXT_texture_sRGB_decode") &&
                epoxy_has_gl_extension("GL_ARB_sampler_objects");

        // Since we allow non-bouncing 4:2:2 YCbCrInputs, effective subpixel precision
        // will be halved when sampling them, and we need to compensate here.
        movit_texel_subpixel_precision /= 2.0;

        resource_pool.reset(new ResourcePool);
        for (unsigned i = 0; i < NUM_OUTPUTS; ++i) {
                output_channel[i].parent = this;
                output_channel[i].channel = i;
        }

        ImageFormat inout_format;
        inout_format.color_space = COLORSPACE_sRGB;
        inout_format.gamma_curve = GAMMA_sRGB;

        // Matches the 4:2:0 format created by the main chain.
        YCbCrFormat ycbcr_format;
        ycbcr_format.chroma_subsampling_x = 2;
        ycbcr_format.chroma_subsampling_y = 2;
        if (global_flags.ycbcr_rec709_coefficients) {
                ycbcr_format.luma_coefficients = YCBCR_REC_709;
        } else {
                ycbcr_format.luma_coefficients = YCBCR_REC_601;
        }
        ycbcr_format.full_range = false;
        ycbcr_format.num_levels = 1 << global_flags.x264_bit_depth;
        ycbcr_format.cb_x_position = 0.0f;
        ycbcr_format.cr_x_position = 0.0f;
        ycbcr_format.cb_y_position = 0.5f;
        ycbcr_format.cr_y_position = 0.5f;

        // Display chain; shows the live output produced by the main chain (or rather, a copy of it).
        display_chain.reset(new EffectChain(global_flags.width, global_flags.height, resource_pool.get()));
        check_error();
        GLenum type = global_flags.x264_bit_depth > 8 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_BYTE;
        display_input = new YCbCrInput(inout_format, ycbcr_format, global_flags.width, global_flags.height, YCBCR_INPUT_SPLIT_Y_AND_CBCR, type);
        display_chain->add_input(display_input);
        display_chain->add_output(inout_format, OUTPUT_ALPHA_FORMAT_POSTMULTIPLIED);
        display_chain->set_dither_bits(0);  // Don't bother.
        display_chain->finalize();

        video_encoder.reset(new VideoEncoder(resource_pool.get(), h264_encoder_surface, global_flags.va_display, global_flags.width, global_flags.height, &httpd, global_disk_space_estimator));

        // Must be instantiated after VideoEncoder has initialized global_flags.use_zerocopy.
        theme.reset(new Theme(global_flags.theme_filename, global_flags.theme_dirs, resource_pool.get(), num_cards));

        // Start listening for clients only once VideoEncoder has written its header, if any.
        httpd.start(9095);

        // First try initializing the then PCI devices, then USB, then
        // fill up with fake cards until we have the desired number of cards.
        unsigned num_pci_devices = 0;
        unsigned card_index = 0;

        {
                IDeckLinkIterator *decklink_iterator = CreateDeckLinkIteratorInstance();
                if (decklink_iterator != nullptr) {
                        for ( ; card_index < num_cards; ++card_index) {
                                IDeckLink *decklink;
                                if (decklink_iterator->Next(&decklink) != S_OK) {
                                        break;
                                }

                                DeckLinkCapture *capture = new DeckLinkCapture(decklink, card_index);
                                DeckLinkOutput *output = new DeckLinkOutput(resource_pool.get(), decklink_output_surface, global_flags.width, global_flags.height, card_index);
                                output->set_device(decklink);
                                configure_card(card_index, capture, CardType::LIVE_CARD, output);
                                ++num_pci_devices;
                        }
                        decklink_iterator->Release();
                        fprintf(stderr, "Found %u DeckLink PCI card(s).\n", num_pci_devices);
                } else {
                        fprintf(stderr, "DeckLink drivers not found. Probing for USB cards only.\n");
                }
        }

        unsigned num_usb_devices = BMUSBCapture::num_cards();
        for (unsigned usb_card_index = 0; usb_card_index < num_usb_devices && card_index < num_cards; ++usb_card_index, ++card_index) {
                BMUSBCapture *capture = new BMUSBCapture(usb_card_index);
                capture->set_card_disconnected_callback(bind(&Mixer::bm_hotplug_remove, this, card_index));
                configure_card(card_index, capture, CardType::LIVE_CARD, /*output=*/nullptr);
        }
        fprintf(stderr, "Found %u USB card(s).\n", num_usb_devices);

        unsigned num_fake_cards = 0;
        for ( ; card_index < num_cards; ++card_index, ++num_fake_cards) {
                FakeCapture *capture = new FakeCapture(global_flags.width, global_flags.height, FAKE_FPS, OUTPUT_FREQUENCY, card_index, global_flags.fake_cards_audio);
                configure_card(card_index, capture, CardType::FAKE_CAPTURE, /*output=*/nullptr);
        }

        if (num_fake_cards > 0) {
                fprintf(stderr, "Initialized %u fake cards.\n", num_fake_cards);
        }

        // Initialize all video inputs the theme asked for. Note that these are
        // all put _after_ the regular cards, which stop at <num_cards> - 1.
        std::vector<FFmpegCapture *> video_inputs = theme->get_video_inputs();
        for (unsigned video_card_index = 0; video_card_index < video_inputs.size(); ++card_index, ++video_card_index) {
                if (card_index >= MAX_VIDEO_CARDS) {
                        fprintf(stderr, "ERROR: Not enough card slots available for the videos the theme requested.\n");
                        exit(1);
                }
                configure_card(card_index, video_inputs[video_card_index], CardType::FFMPEG_INPUT, /*output=*/nullptr);
                video_inputs[video_card_index]->set_card_index(card_index);
        }
        num_video_inputs = video_inputs.size();

        BMUSBCapture::set_card_connected_callback(bind(&Mixer::bm_hotplug_add, this, _1));
        BMUSBCapture::start_bm_thread();

        for (unsigned card_index = 0; card_index < num_cards + num_video_inputs; ++card_index) {
                cards[card_index].queue_length_policy.reset(card_index);
        }

        chroma_subsampler.reset(new ChromaSubsampler(resource_pool.get()));

        if (global_flags.ten_bit_input) {
                if (!v210Converter::has_hardware_support()) {
                        fprintf(stderr, "ERROR: --ten-bit-input requires support for OpenGL compute shaders\n");
                        fprintf(stderr, "       (OpenGL 4.3, or GL_ARB_compute_shader + GL_ARB_shader_image_load_store).\n");
                        exit(1);
                }
                v210_converter.reset(new v210Converter());

                // These are all the widths listed in the Blackmagic SDK documentation
                // (section 2.7.3, “Display Modes”).
                v210_converter->precompile_shader(720);
                v210_converter->precompile_shader(1280);
                v210_converter->precompile_shader(1920);
                v210_converter->precompile_shader(2048);
                v210_converter->precompile_shader(3840);
                v210_converter->precompile_shader(4096);
        }
        if (global_flags.ten_bit_output) {
                if (!v210Converter::has_hardware_support()) {
                        fprintf(stderr, "ERROR: --ten-bit-output requires support for OpenGL compute shaders\n");
                        fprintf(stderr, "       (OpenGL 4.3, or GL_ARB_compute_shader + GL_ARB_shader_image_load_store).\n");
                        exit(1);
                }
        }

        timecode_renderer.reset(new TimecodeRenderer(resource_pool.get(), global_flags.width, global_flags.height));
        display_timecode_in_stream = global_flags.display_timecode_in_stream;
        display_timecode_on_stdout = global_flags.display_timecode_on_stdout;

        if (global_flags.enable_alsa_output) {
                alsa.reset(new ALSAOutput(OUTPUT_FREQUENCY, /*num_channels=*/2));
        }
        if (global_flags.output_card != -1) {
                desired_output_card_index = global_flags.output_card;
                set_output_card_internal(global_flags.output_card);
        }

        output_jitter_history.register_metrics({{ "card", "output" }});
}

Mixer::~Mixer()
{
        BMUSBCapture::stop_bm_thread();

        for (unsigned card_index = 0; card_index < num_cards + num_video_inputs; ++card_index) {
                {
                        unique_lock<mutex> lock(card_mutex);
                        cards[card_index].should_quit = true;  // Unblock thread.
                        cards[card_index].new_frames_changed.notify_all();
                }
                cards[card_index].capture->stop_dequeue_thread();
                if (cards[card_index].output) {
                        cards[card_index].output->end_output();
                        cards[card_index].output.reset();
                }
        }

        video_encoder.reset(nullptr);
}

void Mixer::configure_card(unsigned card_index, CaptureInterface *capture, CardType card_type, DeckLinkOutput *output)
{
        printf("Configuring card %d...\n", card_index);

        CaptureCard *card = &cards[card_index];
        if (card->capture != nullptr) {
                card->capture->stop_dequeue_thread();
        }
        card->capture.reset(capture);
        card->is_fake_capture = (card_type == CardType::FAKE_CAPTURE);
        card->type = card_type;
        if (card->output.get() != output) {
                card->output.reset(output);
        }

        PixelFormat pixel_format;
        if (card_type == CardType::FFMPEG_INPUT) {
                pixel_format = capture->get_current_pixel_format();
        } else if (global_flags.ten_bit_input) {
                pixel_format = PixelFormat_10BitYCbCr;
        } else {
                pixel_format = PixelFormat_8BitYCbCr;
        }

        card->capture->set_frame_callback(bind(&Mixer::bm_frame, this, card_index, _1, _2, _3, _4, _5, _6, _7));
        if (card->frame_allocator == nullptr) {
                card->frame_allocator.reset(new PBOFrameAllocator(pixel_format, 8 << 20, global_flags.width, global_flags.height));  // 8 MB.
        }
        card->capture->set_video_frame_allocator(card->frame_allocator.get());
        if (card->surface == nullptr) {
                card->surface = create_surface_with_same_format(mixer_surface);
        }
        while (!card->new_frames.empty()) card->new_frames.pop_front();
        card->last_timecode = -1;
        card->capture->set_pixel_format(pixel_format);
        card->capture->configure_card();

        // NOTE: start_bm_capture() happens in thread_func().

        DeviceSpec device{InputSourceType::CAPTURE_CARD, card_index};
        audio_mixer.reset_resampler(device);
        audio_mixer.set_display_name(device, card->capture->get_description());
        audio_mixer.trigger_state_changed_callback();

        // Unregister old metrics, if any.
        if (!card->labels.empty()) {
                const vector<pair<string, string>> &labels = card->labels;
                card->jitter_history.unregister_metrics(labels);
                card->queue_length_policy.unregister_metrics(labels);
                global_metrics.remove("input_received_frames", labels);
                global_metrics.remove("input_dropped_frames_jitter", labels);
                global_metrics.remove("input_dropped_frames_error", labels);
                global_metrics.remove("input_dropped_frames_resets", labels);
                global_metrics.remove("input_queue_length_frames", labels);
                global_metrics.remove("input_queue_duped_frames", labels);

                global_metrics.remove("input_has_signal_bool", labels);
                global_metrics.remove("input_is_connected_bool", labels);
                global_metrics.remove("input_interlaced_bool", labels);
                global_metrics.remove("input_width_pixels", labels);
                global_metrics.remove("input_height_pixels", labels);
                global_metrics.remove("input_frame_rate_nom", labels);
                global_metrics.remove("input_frame_rate_den", labels);
                global_metrics.remove("input_sample_rate_hz", labels);
        }

        // Register metrics.
        vector<pair<string, string>> labels;
        char card_name[64];
        snprintf(card_name, sizeof(card_name), "%d", card_index);
        labels.emplace_back("card", card_name);

        switch (card_type) {
        case CardType::LIVE_CARD:
                labels.emplace_back("cardtype", "live");
                break;
        case CardType::FAKE_CAPTURE:
                labels.emplace_back("cardtype", "fake");
                break;
        case CardType::FFMPEG_INPUT:
                labels.emplace_back("cardtype", "ffmpeg");
                break;
        default:
                assert(false);
        }
        card->jitter_history.register_metrics(labels);
        card->queue_length_policy.register_metrics(labels);
        global_metrics.add("input_received_frames", labels, &card->metric_input_received_frames);
        global_metrics.add("input_dropped_frames_jitter", labels, &card->metric_input_dropped_frames_jitter);
        global_metrics.add("input_dropped_frames_error", labels, &card->metric_input_dropped_frames_error);
        global_metrics.add("input_dropped_frames_resets", labels, &card->metric_input_resets);
        global_metrics.add("input_queue_length_frames", labels, &card->metric_input_queue_length_frames, Metrics::TYPE_GAUGE);
        global_metrics.add("input_queue_duped_frames", labels, &card->metric_input_duped_frames);

        global_metrics.add("input_has_signal_bool", labels, &card->metric_input_has_signal_bool, Metrics::TYPE_GAUGE);
        global_metrics.add("input_is_connected_bool", labels, &card->metric_input_is_connected_bool, Metrics::TYPE_GAUGE);
        global_metrics.add("input_interlaced_bool", labels, &card->metric_input_interlaced_bool, Metrics::TYPE_GAUGE);
        global_metrics.add("input_width_pixels", labels, &card->metric_input_width_pixels, Metrics::TYPE_GAUGE);
        global_metrics.add("input_height_pixels", labels, &card->metric_input_height_pixels, Metrics::TYPE_GAUGE);
        global_metrics.add("input_frame_rate_nom", labels, &card->metric_input_frame_rate_nom, Metrics::TYPE_GAUGE);
        global_metrics.add("input_frame_rate_den", labels, &card->metric_input_frame_rate_den, Metrics::TYPE_GAUGE);
        global_metrics.add("input_sample_rate_hz", labels, &card->metric_input_sample_rate_hz, Metrics::TYPE_GAUGE);
        card->labels = labels;
}

void Mixer::set_output_card_internal(int card_index)
{
        // We don't really need to take card_mutex, since we're in the mixer
        // thread and don't mess with any queues (which is the only thing that happens
        // from other threads), but it's probably the safest in the long run.
        unique_lock<mutex> lock(card_mutex);
        if (output_card_index != -1) {
                // Switch the old card from output to input.
                CaptureCard *old_card = &cards[output_card_index];
                old_card->output->end_output();

                // Stop the fake card that we put into place.
                // This needs to _not_ happen under the mutex, to avoid deadlock
                // (delivering the last frame needs to take the mutex).
                CaptureInterface *fake_capture = old_card->capture.get();
                lock.unlock();
                fake_capture->stop_dequeue_thread();
                lock.lock();
                old_card->capture = move(old_card->parked_capture);  // TODO: reset the metrics
                old_card->is_fake_capture = false;
                old_card->capture->start_bm_capture();
        }
        if (card_index != -1) {
                CaptureCard *card = &cards[card_index];
                CaptureInterface *capture = card->capture.get();
                // TODO: DeckLinkCapture::stop_dequeue_thread can actually take
                // several seconds to complete (blocking on DisableVideoInput);
                // see if we can maybe do it asynchronously.
                lock.unlock();
                capture->stop_dequeue_thread();
                lock.lock();
                card->parked_capture = move(card->capture);
                CaptureInterface *fake_capture = new FakeCapture(global_flags.width, global_flags.height, FAKE_FPS, OUTPUT_FREQUENCY, card_index, global_flags.fake_cards_audio);
                configure_card(card_index, fake_capture, CardType::FAKE_CAPTURE, card->output.release());
                card->queue_length_policy.reset(card_index);
                card->capture->start_bm_capture();
                desired_output_video_mode = output_video_mode = card->output->pick_video_mode(desired_output_video_mode);
                card->output->start_output(desired_output_video_mode, pts_int);
        }
        output_card_index = card_index;
        output_jitter_history.clear();
}

namespace {

int unwrap_timecode(uint16_t current_wrapped, int last)
{
        uint16_t last_wrapped = last & 0xffff;
        if (current_wrapped > last_wrapped) {
                return (last & ~0xffff) | current_wrapped;
        } else {
                return 0x10000 + ((last & ~0xffff) | current_wrapped);
        }
}

}  // namespace

void Mixer::bm_frame(unsigned card_index, uint16_t timecode,
                     FrameAllocator::Frame video_frame, size_t video_offset, VideoFormat video_format,
                     FrameAllocator::Frame audio_frame, size_t audio_offset, AudioFormat audio_format)
{
        DeviceSpec device{InputSourceType::CAPTURE_CARD, card_index};
        CaptureCard *card = &cards[card_index];

        ++card->metric_input_received_frames;
        card->metric_input_has_signal_bool = video_format.has_signal;
        card->metric_input_is_connected_bool = video_format.is_connected;
        card->metric_input_interlaced_bool = video_format.interlaced;
        card->metric_input_width_pixels = video_format.width;
        card->metric_input_height_pixels = video_format.height;
        card->metric_input_frame_rate_nom = video_format.frame_rate_nom;
        card->metric_input_frame_rate_den = video_format.frame_rate_den;
        card->metric_input_sample_rate_hz = audio_format.sample_rate;

        if (is_mode_scanning[card_index]) {
                if (video_format.has_signal) {
                        // Found a stable signal, so stop scanning.
                        is_mode_scanning[card_index] = false;
                } else {
                        static constexpr double switch_time_s = 0.1;  // Should be enough time for the signal to stabilize.
                        steady_clock::time_point now = steady_clock::now();
                        double sec_since_last_switch = duration<double>(steady_clock::now() - last_mode_scan_change[card_index]).count();
                        if (sec_since_last_switch > switch_time_s) {
                                // It isn't this mode; try the next one.
                                mode_scanlist_index[card_index]++;
                                mode_scanlist_index[card_index] %= mode_scanlist[card_index].size();
                                cards[card_index].capture->set_video_mode(mode_scanlist[card_index][mode_scanlist_index[card_index]]);
                                last_mode_scan_change[card_index] = now;
                        }
                }
        }

        int64_t frame_length = int64_t(TIMEBASE) * video_format.frame_rate_den / video_format.frame_rate_nom;
        assert(frame_length > 0);

        size_t num_samples = (audio_frame.len > audio_offset) ? (audio_frame.len - audio_offset) / audio_format.num_channels / (audio_format.bits_per_sample / 8) : 0;
        if (num_samples > OUTPUT_FREQUENCY / 10) {
                printf("Card %d: Dropping frame with implausible audio length (len=%d, offset=%d) [timecode=0x%04x video_len=%d video_offset=%d video_format=%x)\n",
                        card_index, int(audio_frame.len), int(audio_offset),
                        timecode, int(video_frame.len), int(video_offset), video_format.id);
                if (video_frame.owner) {
                        video_frame.owner->release_frame(video_frame);
                }
                if (audio_frame.owner) {
                        audio_frame.owner->release_frame(audio_frame);
                }
                return;
        }

        int dropped_frames = 0;
        if (card->last_timecode != -1) {
                dropped_frames = unwrap_timecode(timecode, card->last_timecode) - card->last_timecode - 1;
        }

        // Number of samples per frame if we need to insert silence.
        // (Could be nonintegral, but resampling will save us then.)
        const int silence_samples = OUTPUT_FREQUENCY * video_format.frame_rate_den / video_format.frame_rate_nom;

        if (dropped_frames > MAX_FPS * 2) {
                fprintf(stderr, "Card %d lost more than two seconds (or time code jumping around; from 0x%04x to 0x%04x), resetting resampler\n",
                        card_index, card->last_timecode, timecode);
                audio_mixer.reset_resampler(device);
                dropped_frames = 0;
                ++card->metric_input_resets;
        } else if (dropped_frames > 0) {
                // Insert silence as needed.
                fprintf(stderr, "Card %d dropped %d frame(s) (before timecode 0x%04x), inserting silence.\n",
                        card_index, dropped_frames, timecode);
                card->metric_input_dropped_frames_error += dropped_frames;

                bool success;
                do {
                        success = audio_mixer.add_silence(device, silence_samples, dropped_frames, frame_length);
                } while (!success);
        }

        if (num_samples > 0) {
                audio_mixer.add_audio(device, audio_frame.data + audio_offset, num_samples, audio_format, frame_length, audio_frame.received_timestamp);
        }

        // Done with the audio, so release it.
        if (audio_frame.owner) {
                audio_frame.owner->release_frame(audio_frame);
        }

        card->last_timecode = timecode;

        PBOFrameAllocator::Userdata *userdata = (PBOFrameAllocator::Userdata *)video_frame.userdata;

        size_t cbcr_width, cbcr_height, cbcr_offset, y_offset;
        size_t expected_length = video_format.stride * (video_format.height + video_format.extra_lines_top + video_format.extra_lines_bottom);
        if (userdata != nullptr && userdata->pixel_format == PixelFormat_8BitYCbCrPlanar) {
                // The calculation above is wrong for planar Y'CbCr, so just override it.
                assert(card->type == CardType::FFMPEG_INPUT);
                assert(video_offset == 0);
                expected_length = video_frame.len;

                userdata->ycbcr_format = (static_cast<FFmpegCapture *>(card->capture.get()))->get_current_frame_ycbcr_format();
                cbcr_width = video_format.width / userdata->ycbcr_format.chroma_subsampling_x;
                cbcr_height = video_format.height / userdata->ycbcr_format.chroma_subsampling_y;
                cbcr_offset = video_format.width * video_format.height;
                y_offset = 0;
        } else {
                // All the other Y'CbCr formats are 4:2:2.
                cbcr_width = video_format.width / 2;
                cbcr_height = video_format.height;
                cbcr_offset = video_offset / 2;
                y_offset = video_frame.size / 2 + video_offset / 2;
        }
        if (video_frame.len - video_offset == 0 ||
            video_frame.len - video_offset != expected_length) {
                if (video_frame.len != 0) {
                        printf("Card %d: Dropping video frame with wrong length (%ld; expected %ld)\n",
                                card_index, video_frame.len - video_offset, expected_length);
                }
                if (video_frame.owner) {
                        video_frame.owner->release_frame(video_frame);
                }

                // Still send on the information that we _had_ a frame, even though it's corrupted,
                // so that pts can go up accordingly.
                {
                        unique_lock<mutex> lock(card_mutex);
                        CaptureCard::NewFrame new_frame;
                        new_frame.frame = RefCountedFrame(FrameAllocator::Frame());
                        new_frame.length = frame_length;
                        new_frame.interlaced = false;
                        new_frame.dropped_frames = dropped_frames;
                        new_frame.received_timestamp = video_frame.received_timestamp;
                        card->new_frames.push_back(move(new_frame));
                        card->jitter_history.frame_arrived(video_frame.received_timestamp, frame_length, dropped_frames);
                }
                card->new_frames_changed.notify_all();
                return;
        }

        unsigned num_fields = video_format.interlaced ? 2 : 1;
        steady_clock::time_point frame_upload_start;
        bool interlaced_stride = false;
        if (video_format.interlaced) {
                // Send the two fields along as separate frames; the other side will need to add
                // a deinterlacer to actually get this right.
                assert(video_format.height % 2 == 0);
                video_format.height /= 2;
                cbcr_height /= 2;
                assert(frame_length % 2 == 0);
                frame_length /= 2;
                num_fields = 2;
                if (video_format.second_field_start == 1) {
                        interlaced_stride = true;
                }
                frame_upload_start = steady_clock::now();
        }
        userdata->last_interlaced = video_format.interlaced;
        userdata->last_has_signal = video_format.has_signal;
        userdata->last_is_connected = video_format.is_connected;
        userdata->last_frame_rate_nom = video_format.frame_rate_nom;
        userdata->last_frame_rate_den = video_format.frame_rate_den;
        RefCountedFrame frame(video_frame);

        // Upload the textures.
        for (unsigned field = 0; field < num_fields; ++field) {
                // Put the actual texture upload in a lambda that is executed in the main thread.
                // It is entirely possible to do this in the same thread (and it might even be
                // faster, depending on the GPU and driver), but it appears to be trickling
                // driver bugs very easily.
                //
                // Note that this means we must hold on to the actual frame data in <userdata>
                // until the upload command is run, but we hold on to <frame> much longer than that
                // (in fact, all the way until we no longer use the texture in rendering).
                auto upload_func = [this, field, video_format, y_offset, video_offset, cbcr_offset, cbcr_width, cbcr_height, interlaced_stride, userdata]() {
                        unsigned field_start_line;
                        if (field == 1) {
                                field_start_line = video_format.second_field_start;
                        } else {
                                field_start_line = video_format.extra_lines_top;
                        }

                        // For anything not FRAME_FORMAT_YCBCR_10BIT, v210_width will be nonsensical but not used.
                        size_t v210_width = video_format.stride / sizeof(uint32_t);
                        ensure_texture_resolution(userdata, field, video_format.width, video_format.height, cbcr_width, cbcr_height, v210_width);

                        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, userdata->pbo);
                        check_error();

                        switch (userdata->pixel_format) {
                        case PixelFormat_10BitYCbCr: {
                                size_t field_start = video_offset + video_format.stride * field_start_line;
                                upload_texture(userdata->tex_v210[field], v210_width, video_format.height, video_format.stride, interlaced_stride, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, field_start);
                                v210_converter->convert(userdata->tex_v210[field], userdata->tex_444[field], video_format.width, video_format.height);
                                break;
                        }
                        case PixelFormat_8BitYCbCr: {
                                size_t field_y_start = y_offset + video_format.width * field_start_line;
                                size_t field_cbcr_start = cbcr_offset + cbcr_width * field_start_line * sizeof(uint16_t);

                                // Make up our own strides, since we are interleaving.
                                upload_texture(userdata->tex_y[field], video_format.width, video_format.height, video_format.width, interlaced_stride, GL_RED, GL_UNSIGNED_BYTE, field_y_start);
                                upload_texture(userdata->tex_cbcr[field], cbcr_width, cbcr_height, cbcr_width * sizeof(uint16_t), interlaced_stride, GL_RG, GL_UNSIGNED_BYTE, field_cbcr_start);
                                break;
                        }
                        case PixelFormat_8BitYCbCrPlanar: {
                                assert(field_start_line == 0);  // We don't really support interlaced here.
                                size_t field_y_start = y_offset;
                                size_t field_cb_start = cbcr_offset;
                                size_t field_cr_start = cbcr_offset + cbcr_width * cbcr_height;

                                // Make up our own strides, since we are interleaving.
                                upload_texture(userdata->tex_y[field], video_format.width, video_format.height, video_format.width, interlaced_stride, GL_RED, GL_UNSIGNED_BYTE, field_y_start);
                                upload_texture(userdata->tex_cb[field], cbcr_width, cbcr_height, cbcr_width, interlaced_stride, GL_RED, GL_UNSIGNED_BYTE, field_cb_start);
                                upload_texture(userdata->tex_cr[field], cbcr_width, cbcr_height, cbcr_width, interlaced_stride, GL_RED, GL_UNSIGNED_BYTE, field_cr_start);
                                break;
                        }
                        case PixelFormat_8BitBGRA: {
                                size_t field_start = video_offset + video_format.stride * field_start_line;
                                upload_texture(userdata->tex_rgba[field], video_format.width, video_format.height, video_format.stride, interlaced_stride, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, field_start);
                                // These could be asked to deliver mipmaps at any time.
                                glBindTexture(GL_TEXTURE_2D, userdata->tex_rgba[field]);
                                check_error();
                                glGenerateMipmap(GL_TEXTURE_2D);
                                check_error();
                                glBindTexture(GL_TEXTURE_2D, 0);
                                check_error();
                                break;
                        }
                        default:
                                assert(false);
                        }

                        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
                        check_error();
                };

                if (field == 1) {
                        // Don't upload the second field as fast as we can; wait until
                        // the field time has approximately passed. (Otherwise, we could
                        // get timing jitter against the other sources, and possibly also
                        // against the video display, although the latter is not as critical.)
                        // This requires our system clock to be reasonably close to the
                        // video clock, but that's not an unreasonable assumption.
                        steady_clock::time_point second_field_start = frame_upload_start +
                                nanoseconds(frame_length * 1000000000 / TIMEBASE);
                        this_thread::sleep_until(second_field_start);
                }

                {
                        unique_lock<mutex> lock(card_mutex);
                        CaptureCard::NewFrame new_frame;
                        new_frame.frame = frame;
                        new_frame.length = frame_length;
                        new_frame.field = field;
                        new_frame.interlaced = video_format.interlaced;
                        new_frame.upload_func = upload_func;
                        new_frame.dropped_frames = dropped_frames;
                        new_frame.received_timestamp = video_frame.received_timestamp;  // Ignore the audio timestamp.
                        card->new_frames.push_back(move(new_frame));
                        card->jitter_history.frame_arrived(video_frame.received_timestamp, frame_length, dropped_frames);
                }
                card->new_frames_changed.notify_all();
        }
}

void Mixer::bm_hotplug_add(libusb_device *dev)
{
        lock_guard<mutex> lock(hotplug_mutex);
        hotplugged_cards.push_back(dev);
}

void Mixer::bm_hotplug_remove(unsigned card_index)
{
        cards[card_index].new_frames_changed.notify_all();
}

void Mixer::thread_func()
{
        pthread_setname_np(pthread_self(), "Mixer_OpenGL");

        eglBindAPI(EGL_OPENGL_API);
        QOpenGLContext *context = create_context(mixer_surface);
        if (!make_current(context, mixer_surface)) {
                printf("oops\n");
                exit(1);
        }

        // Start the actual capture. (We don't want to do it before we're actually ready
        // to process output frames.)
        for (unsigned card_index = 0; card_index < num_cards + num_video_inputs; ++card_index) {
                if (int(card_index) != output_card_index) {
                        cards[card_index].capture->start_bm_capture();
                }
        }

        BasicStats basic_stats(/*verbose=*/true);
        int stats_dropped_frames = 0;

        while (!should_quit) {
                if (desired_output_card_index != output_card_index) {
                        set_output_card_internal(desired_output_card_index);
                }
                if (output_card_index != -1 &&
                    desired_output_video_mode != output_video_mode) {
                        DeckLinkOutput *output = cards[output_card_index].output.get();
                        output->end_output();
                        desired_output_video_mode = output_video_mode = output->pick_video_mode(desired_output_video_mode);
                        output->start_output(desired_output_video_mode, pts_int);
                }

                CaptureCard::NewFrame new_frames[MAX_VIDEO_CARDS];
                bool has_new_frame[MAX_VIDEO_CARDS] = { false };

                bool master_card_is_output;
                unsigned master_card_index;
                if (output_card_index != -1) {
                        master_card_is_output = true;
                        master_card_index = output_card_index;
                } else {
                        master_card_is_output = false;
                        master_card_index = theme->map_signal(master_clock_channel);
                        assert(master_card_index < num_cards);
                }

                OutputFrameInfo output_frame_info = get_one_frame_from_each_card(master_card_index, master_card_is_output, new_frames, has_new_frame);
                schedule_audio_resampling_tasks(output_frame_info.dropped_frames, output_frame_info.num_samples, output_frame_info.frame_duration, output_frame_info.is_preroll, output_frame_info.frame_timestamp);
                stats_dropped_frames += output_frame_info.dropped_frames;

                handle_hotplugged_cards();

                for (unsigned card_index = 0; card_index < num_cards + num_video_inputs; ++card_index) {
                        if (card_index == master_card_index || !has_new_frame[card_index]) {
                                continue;
                        }
                        if (new_frames[card_index].frame->len == 0) {
                                ++new_frames[card_index].dropped_frames;
                        }
                        if (new_frames[card_index].dropped_frames > 0) {
                                printf("Card %u dropped %d frames before this\n",
                                        card_index, int(new_frames[card_index].dropped_frames));
                        }
                }

                // If the first card is reporting a corrupted or otherwise dropped frame,
                // just increase the pts (skipping over this frame) and don't try to compute anything new.
                if (!master_card_is_output && new_frames[master_card_index].frame->len == 0) {
                        ++stats_dropped_frames;
                        pts_int += new_frames[master_card_index].length;
                        continue;
                }

                for (unsigned card_index = 0; card_index < num_cards + num_video_inputs; ++card_index) {
                        if (!has_new_frame[card_index] || new_frames[card_index].frame->len == 0)
                                continue;

                        CaptureCard::NewFrame *new_frame = &new_frames[card_index];
                        assert(new_frame->frame != nullptr);
                        insert_new_frame(new_frame->frame, new_frame->field, new_frame->interlaced, card_index, &input_state);
                        check_error();

                        // The new texture might need uploading before use.
                        if (new_frame->upload_func) {
                                new_frame->upload_func();
                                new_frame->upload_func = nullptr;
                        }
                }

                int64_t frame_duration = output_frame_info.frame_duration;
                render_one_frame(frame_duration);
                ++frame_num;
                pts_int += frame_duration;

                basic_stats.update(frame_num, stats_dropped_frames);
                // if (frame_num % 100 == 0) chain->print_phase_timing();

                if (should_cut.exchange(false)) {  // Test and clear.
                        video_encoder->do_cut(frame_num);
                }

#if 0
                // Reset every 100 frames, so that local variations in frame times
                // (especially for the first few frames, when the shaders are
                // compiled etc.) don't make it hard to measure for the entire
                // remaining duration of the program.
                if (frame == 10000) {
                        frame = 0;
                        start = now;
                }
#endif
                check_error();
        }

        resource_pool->clean_context();
}

bool Mixer::input_card_is_master_clock(unsigned card_index, unsigned master_card_index) const
{
        if (output_card_index != -1) {
                // The output card (ie., cards[output_card_index].output) is the master clock,
                // so no input card (ie., cards[card_index].capture) is.
                return false;
        }
        return (card_index == master_card_index);
}

void Mixer::trim_queue(CaptureCard *card, size_t safe_queue_length)
{
        // Count the number of frames in the queue, including any frames
        // we dropped. It's hard to know exactly how we should deal with
        // dropped (corrupted) input frames; they don't help our goal of
        // avoiding starvation, but they still add to the problem of latency.
        // Since dropped frames is going to mean a bump in the signal anyway,
        // we err on the side of having more stable latency instead.
        unsigned queue_length = 0;
        for (const CaptureCard::NewFrame &frame : card->new_frames) {
                queue_length += frame.dropped_frames + 1;
        }

        // If needed, drop frames until the queue is below the safe limit.
        // We prefer to drop from the head, because all else being equal,
        // we'd like more recent frames (less latency).
        unsigned dropped_frames = 0;
        while (queue_length > safe_queue_length) {
                assert(!card->new_frames.empty());
                assert(queue_length > card->new_frames.front().dropped_frames);
                queue_length -= card->new_frames.front().dropped_frames;

                if (queue_length <= safe_queue_length) {
                        // No need to drop anything.
                        break;
                }

                card->new_frames.pop_front();
                card->new_frames_changed.notify_all();
                --queue_length;
                ++dropped_frames;
        }

        card->metric_input_dropped_frames_jitter += dropped_frames;
        card->metric_input_queue_length_frames = queue_length;

#if 0
        if (dropped_frames > 0) {
                fprintf(stderr, "Card %u dropped %u frame(s) to keep latency down.\n",
                        card_index, dropped_frames);
        }
#endif
}


Mixer::OutputFrameInfo Mixer::get_one_frame_from_each_card(unsigned master_card_index, bool master_card_is_output, CaptureCard::NewFrame new_frames[MAX_VIDEO_CARDS], bool has_new_frame[MAX_VIDEO_CARDS])
{
        OutputFrameInfo output_frame_info;
start:
        unique_lock<mutex> lock(card_mutex, defer_lock);
        if (master_card_is_output) {
                // Clocked to the output, so wait for it to be ready for the next frame.
                cards[master_card_index].output->wait_for_frame(pts_int, &output_frame_info.dropped_frames, &output_frame_info.frame_duration, &output_frame_info.is_preroll, &output_frame_info.frame_timestamp);
                lock.lock();
        } else {
                // Wait for the master card to have a new frame.
                // TODO: Add a timeout.
                output_frame_info.is_preroll = false;
                lock.lock();
                cards[master_card_index].new_frames_changed.wait(lock, [this, master_card_index]{ return !cards[master_card_index].new_frames.empty() || cards[master_card_index].capture->get_disconnected(); });
        }

        if (master_card_is_output) {
                handle_hotplugged_cards();
        } else if (cards[master_card_index].new_frames.empty()) {
                // We were woken up, but not due to a new frame. Deal with it
                // and then restart.
                assert(cards[master_card_index].capture->get_disconnected());
                handle_hotplugged_cards();
                lock.unlock();
                goto start;
        }

        for (unsigned card_index = 0; card_index < num_cards + num_video_inputs; ++card_index) {
                CaptureCard *card = &cards[card_index];
                if (card->new_frames.empty()) {  // Starvation.
                        ++card->metric_input_duped_frames;
                } else {
                        new_frames[card_index] = move(card->new_frames.front());
                        has_new_frame[card_index] = true;
                        card->new_frames.pop_front();
                        card->new_frames_changed.notify_all();
                }
        }

        if (!master_card_is_output) {
                output_frame_info.frame_timestamp = new_frames[master_card_index].received_timestamp;
                output_frame_info.dropped_frames = new_frames[master_card_index].dropped_frames;
                output_frame_info.frame_duration = new_frames[master_card_index].length;
        }

        if (!output_frame_info.is_preroll) {
                output_jitter_history.frame_arrived(output_frame_info.frame_timestamp, output_frame_info.frame_duration, output_frame_info.dropped_frames);
        }

        for (unsigned card_index = 0; card_index < num_cards + num_video_inputs; ++card_index) {
                CaptureCard *card = &cards[card_index];
                if (has_new_frame[card_index] &&
                    !input_card_is_master_clock(card_index, master_card_index) &&
                    !output_frame_info.is_preroll) {
                        card->queue_length_policy.update_policy(
                                output_frame_info.frame_timestamp,
                                card->jitter_history.get_expected_next_frame(),
                                new_frames[master_card_index].length,
                                output_frame_info.frame_duration,
                                card->jitter_history.estimate_max_jitter(),
                                output_jitter_history.estimate_max_jitter());
                        trim_queue(card, min<int>(global_flags.max_input_queue_frames,
                                                  card->queue_length_policy.get_safe_queue_length()));
                }
        }

        // This might get off by a fractional sample when changing master card
        // between ones with different frame rates, but that's fine.
        int num_samples_times_timebase = OUTPUT_FREQUENCY * output_frame_info.frame_duration + fractional_samples;
        output_frame_info.num_samples = num_samples_times_timebase / TIMEBASE;
        fractional_samples = num_samples_times_timebase % TIMEBASE;
        assert(output_frame_info.num_samples >= 0);

        return output_frame_info;
}

void Mixer::handle_hotplugged_cards()
{
        // Check for cards that have been disconnected since last frame.
        for (unsigned card_index = 0; card_index < num_cards; ++card_index) {
                CaptureCard *card = &cards[card_index];
                if (card->capture->get_disconnected()) {
                        fprintf(stderr, "Card %u went away, replacing with a fake card.\n", card_index);
                        FakeCapture *capture = new FakeCapture(global_flags.width, global_flags.height, FAKE_FPS, OUTPUT_FREQUENCY, card_index, global_flags.fake_cards_audio);
                        configure_card(card_index, capture, CardType::FAKE_CAPTURE, /*output=*/nullptr);
                        card->queue_length_policy.reset(card_index);
                        card->capture->start_bm_capture();
                }
        }

        // Check for cards that have been connected since last frame.
        vector<libusb_device *> hotplugged_cards_copy;
        {
                lock_guard<mutex> lock(hotplug_mutex);
                swap(hotplugged_cards, hotplugged_cards_copy);
        }
        for (libusb_device *new_dev : hotplugged_cards_copy) {
                // Look for a fake capture card where we can stick this in.
                int free_card_index = -1;
                for (unsigned card_index = 0; card_index < num_cards; ++card_index) {
                        if (cards[card_index].is_fake_capture) {
                                free_card_index = card_index;
                                break;
                        }
                }

                if (free_card_index == -1) {
                        fprintf(stderr, "New card plugged in, but no free slots -- ignoring.\n");
                        libusb_unref_device(new_dev);
                } else {
                        // BMUSBCapture takes ownership.
                        fprintf(stderr, "New card plugged in, choosing slot %d.\n", free_card_index);
                        CaptureCard *card = &cards[free_card_index];
                        BMUSBCapture *capture = new BMUSBCapture(free_card_index, new_dev);
                        configure_card(free_card_index, capture, CardType::LIVE_CARD, /*output=*/nullptr);
                        card->queue_length_policy.reset(free_card_index);
                        capture->set_card_disconnected_callback(bind(&Mixer::bm_hotplug_remove, this, free_card_index));
                        capture->start_bm_capture();
                }
        }
}


void Mixer::schedule_audio_resampling_tasks(unsigned dropped_frames, int num_samples_per_frame, int length_per_frame, bool is_preroll, steady_clock::time_point frame_timestamp)
{
        // Resample the audio as needed, including from previously dropped frames.
        assert(num_cards > 0);
        for (unsigned frame_num = 0; frame_num < dropped_frames + 1; ++frame_num) {
                const bool dropped_frame = (frame_num != dropped_frames);
                {
                        // Signal to the audio thread to process this frame.
                        // Note that if the frame is a dropped frame, we signal that
                        // we don't want to use this frame as base for adjusting
                        // the resampler rate. The reason for this is that the timing
                        // of these frames is often way too late; they typically don't
                        // “arrive” before we synthesize them. Thus, we could end up
                        // in a situation where we have inserted e.g. five audio frames
                        // into the queue before we then start pulling five of them
                        // back out. This makes ResamplingQueue overestimate the delay,
                        // causing undue resampler changes. (We _do_ use the last,
                        // non-dropped frame; perhaps we should just discard that as well,
                        // since dropped frames are expected to be rare, and it might be
                        // better to just wait until we have a slightly more normal situation).
                        unique_lock<mutex> lock(audio_mutex);
                        bool adjust_rate = !dropped_frame && !is_preroll;
                        audio_task_queue.push(AudioTask{pts_int, num_samples_per_frame, adjust_rate, frame_timestamp});
                        audio_task_queue_changed.notify_one();
                }
                if (dropped_frame) {
                        // For dropped frames, increase the pts. 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.
                        pts_int += length_per_frame;
                }
        }
}

void Mixer::render_one_frame(int64_t duration)
{
        // Determine the time code for this frame before we start rendering.
        string timecode_text = timecode_renderer->get_timecode_text(double(pts_int) / TIMEBASE, frame_num);
        if (display_timecode_on_stdout) {
                printf("Timecode: '%s'\n", timecode_text.c_str());
        }

        // Update Y'CbCr settings for all cards.
        {
                unique_lock<mutex> lock(card_mutex);
                for (unsigned card_index = 0; card_index < num_cards; ++card_index) {
                        YCbCrInterpretation *interpretation = &ycbcr_interpretation[card_index];
                        input_state.ycbcr_coefficients_auto[card_index] = interpretation->ycbcr_coefficients_auto;
                        input_state.ycbcr_coefficients[card_index] = interpretation->ycbcr_coefficients;
                        input_state.full_range[card_index] = interpretation->full_range;
                }
        }

        // Get the main chain from the theme, and set its state immediately.
        Theme::Chain theme_main_chain = theme->get_chain(0, pts(), global_flags.width, global_flags.height, input_state);
        EffectChain *chain = theme_main_chain.chain;
        theme_main_chain.setup_chain();
        //theme_main_chain.chain->enable_phase_timing(true);

        // The theme can't (or at least shouldn't!) call connect_signal() on
        // each FFmpeg input, so we'll do it here.
        for (const pair<LiveInputWrapper *, FFmpegCapture *> &conn : theme->get_signal_connections()) {
                conn.first->connect_signal_raw(conn.second->get_card_index(), input_state);
        }

        // If HDMI/SDI output is active and the user has requested auto mode,
        // its mode overrides the existing Y'CbCr setting for the chain.
        YCbCrLumaCoefficients ycbcr_output_coefficients;
        if (global_flags.ycbcr_auto_coefficients && output_card_index != -1) {
                ycbcr_output_coefficients = cards[output_card_index].output->preferred_ycbcr_coefficients();
        } else {
                ycbcr_output_coefficients = global_flags.ycbcr_rec709_coefficients ? YCBCR_REC_709 : YCBCR_REC_601;
        }

        // TODO: Reduce the duplication against theme.cpp.
        YCbCrFormat output_ycbcr_format;
        output_ycbcr_format.chroma_subsampling_x = 1;
        output_ycbcr_format.chroma_subsampling_y = 1;
        output_ycbcr_format.luma_coefficients = ycbcr_output_coefficients;
        output_ycbcr_format.full_range = false;
        output_ycbcr_format.num_levels = 1 << global_flags.x264_bit_depth;
        chain->change_ycbcr_output_format(output_ycbcr_format);

        // Render main chain. If we're using zerocopy Quick Sync encoding
        // (the default case), we take an extra copy of the created outputs,
        // so that we can display it back to the screen later (it's less memory
        // bandwidth than writing and reading back an RGBA texture, even at 16-bit).
        // Ideally, we'd like to avoid taking copies and just use the main textures
        // for display as well, but they're just views into VA-API memory and must be
        // unmapped during encoding, so we can't use them for display, unfortunately.
        GLuint y_tex, cbcr_full_tex, cbcr_tex;
        GLuint y_copy_tex, cbcr_copy_tex = 0;
        GLuint y_display_tex, cbcr_display_tex;
        GLenum y_type = (global_flags.x264_bit_depth > 8) ? GL_R16 : GL_R8;
        GLenum cbcr_type = (global_flags.x264_bit_depth > 8) ? GL_RG16 : GL_RG8;
        const bool is_zerocopy = video_encoder->is_zerocopy();
        if (is_zerocopy) {
                cbcr_full_tex = resource_pool->create_2d_texture(cbcr_type, global_flags.width, global_flags.height);
                y_copy_tex = resource_pool->create_2d_texture(y_type, global_flags.width, global_flags.height);
                cbcr_copy_tex = resource_pool->create_2d_texture(cbcr_type, global_flags.width / 2, global_flags.height / 2);

                y_display_tex = y_copy_tex;
                cbcr_display_tex = cbcr_copy_tex;

                // y_tex and cbcr_tex will be given by VideoEncoder.
        } else {
                cbcr_full_tex = resource_pool->create_2d_texture(cbcr_type, global_flags.width, global_flags.height);
                y_tex = resource_pool->create_2d_texture(y_type, global_flags.width, global_flags.height);
                cbcr_tex = resource_pool->create_2d_texture(cbcr_type, global_flags.width / 2, global_flags.height / 2);

                y_display_tex = y_tex;
                cbcr_display_tex = cbcr_tex;
        }

        const int64_t av_delay = lrint(global_flags.audio_queue_length_ms * 0.001 * TIMEBASE);  // Corresponds to the delay in ResamplingQueue.
        bool got_frame = video_encoder->begin_frame(pts_int + av_delay, duration, ycbcr_output_coefficients, theme_main_chain.input_frames, &y_tex, &cbcr_tex);
        assert(got_frame);

        GLuint fbo;
        if (is_zerocopy) {
                fbo = resource_pool->create_fbo(y_tex, cbcr_full_tex, y_copy_tex);
        } else {
                fbo = resource_pool->create_fbo(y_tex, cbcr_full_tex);
        }
        check_error();
        chain->render_to_fbo(fbo, global_flags.width, global_flags.height);

        if (display_timecode_in_stream) {
                // Render the timecode on top.
                timecode_renderer->render_timecode(fbo, timecode_text);
        }

        resource_pool->release_fbo(fbo);

        if (is_zerocopy) {
                chroma_subsampler->subsample_chroma(cbcr_full_tex, global_flags.width, global_flags.height, cbcr_tex, cbcr_copy_tex);
        } else {
                chroma_subsampler->subsample_chroma(cbcr_full_tex, global_flags.width, global_flags.height, cbcr_tex);
        }
        if (output_card_index != -1) {
                cards[output_card_index].output->send_frame(y_tex, cbcr_full_tex, ycbcr_output_coefficients, theme_main_chain.input_frames, pts_int, duration);
        }
        resource_pool->release_2d_texture(cbcr_full_tex);

        // Set the right state for the Y' and CbCr textures we use for display.
        glBindFramebuffer(GL_FRAMEBUFFER, 0);
        glBindTexture(GL_TEXTURE_2D, y_display_tex);
        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);

        glBindTexture(GL_TEXTURE_2D, cbcr_display_tex);
        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);

        RefCountedGLsync fence = video_encoder->end_frame();

        // The live frame pieces the Y'CbCr texture copies back into RGB and displays them.
        // It owns y_display_tex and cbcr_display_tex now (whichever textures they are).
        DisplayFrame live_frame;
        live_frame.chain = display_chain.get();
        live_frame.setup_chain = [this, y_display_tex, cbcr_display_tex]{
                display_input->set_texture_num(0, y_display_tex);
                display_input->set_texture_num(1, cbcr_display_tex);
        };
        live_frame.ready_fence = fence;
        live_frame.input_frames = {};
        live_frame.temp_textures = { y_display_tex, cbcr_display_tex };
        output_channel[OUTPUT_LIVE].output_frame(live_frame);

        // Set up preview and any additional channels.
        for (int i = 1; i < theme->get_num_channels() + 2; ++i) {
                DisplayFrame display_frame;
                Theme::Chain chain = theme->get_chain(i, pts(), global_flags.width, global_flags.height, input_state);  // FIXME: dimensions
                display_frame.chain = chain.chain;
                display_frame.setup_chain = chain.setup_chain;
                display_frame.ready_fence = fence;
                display_frame.input_frames = chain.input_frames;
                display_frame.temp_textures = {};
                output_channel[i].output_frame(display_frame);
        }
}

void Mixer::audio_thread_func()
{
        pthread_setname_np(pthread_self(), "Mixer_Audio");

        while (!should_quit) {
                AudioTask task;

                {
                        unique_lock<mutex> lock(audio_mutex);
                        audio_task_queue_changed.wait(lock, [this]{ return should_quit || !audio_task_queue.empty(); });
                        if (should_quit) {
                                return;
                        }
                        task = audio_task_queue.front();
                        audio_task_queue.pop();
                }

                ResamplingQueue::RateAdjustmentPolicy rate_adjustment_policy =
                        task.adjust_rate ? ResamplingQueue::ADJUST_RATE : ResamplingQueue::DO_NOT_ADJUST_RATE;
                vector<float> samples_out = audio_mixer.get_output(
                        task.frame_timestamp,
                        task.num_samples,
                        rate_adjustment_policy);

                // Send the samples to the sound card, then add them to the output.
                if (alsa) {
                        alsa->write(samples_out);
                }
                if (output_card_index != -1) {
                        const int64_t av_delay = lrint(global_flags.audio_queue_length_ms * 0.001 * TIMEBASE);  // Corresponds to the delay in ResamplingQueue.
                        cards[output_card_index].output->send_audio(task.pts_int + av_delay, samples_out);
                }
                video_encoder->add_audio(task.pts_int, move(samples_out));
        }
}

void Mixer::release_display_frame(DisplayFrame *frame)
{
        for (GLuint texnum : frame->temp_textures) {
                resource_pool->release_2d_texture(texnum);
        }
        frame->temp_textures.clear();
        frame->ready_fence.reset();
        frame->input_frames.clear();
}

void Mixer::start()
{
        mixer_thread = thread(&Mixer::thread_func, this);
        audio_thread = thread(&Mixer::audio_thread_func, this);
}

void Mixer::quit()
{
        should_quit = true;
        audio_task_queue_changed.notify_one();
        mixer_thread.join();
        audio_thread.join();
}

void Mixer::transition_clicked(int transition_num)
{
        theme->transition_clicked(transition_num, pts());
}

void Mixer::channel_clicked(int preview_num)
{
        theme->channel_clicked(preview_num);
}

YCbCrInterpretation Mixer::get_input_ycbcr_interpretation(unsigned card_index) const
{
        unique_lock<mutex> lock(card_mutex);
        return ycbcr_interpretation[card_index];
}

void Mixer::set_input_ycbcr_interpretation(unsigned card_index, const YCbCrInterpretation &interpretation)
{
        unique_lock<mutex> lock(card_mutex);
        ycbcr_interpretation[card_index] = interpretation;
}

void Mixer::start_mode_scanning(unsigned card_index)
{
        assert(card_index < num_cards);
        if (is_mode_scanning[card_index]) {
                return;
        }
        is_mode_scanning[card_index] = true;
        mode_scanlist[card_index].clear();
        for (const auto &mode : cards[card_index].capture->get_available_video_modes()) {
                mode_scanlist[card_index].push_back(mode.first);
        }
        assert(!mode_scanlist[card_index].empty());
        mode_scanlist_index[card_index] = 0;
        cards[card_index].capture->set_video_mode(mode_scanlist[card_index][0]);
        last_mode_scan_change[card_index] = steady_clock::now();
}

map<uint32_t, VideoMode> Mixer::get_available_output_video_modes() const
{
        assert(desired_output_card_index != -1);
        unique_lock<mutex> lock(card_mutex);
        return cards[desired_output_card_index].output->get_available_video_modes();
}

Mixer::OutputChannel::~OutputChannel()
{
        if (has_current_frame) {
                parent->release_display_frame(&current_frame);
        }
        if (has_ready_frame) {
                parent->release_display_frame(&ready_frame);
        }
}

void Mixer::OutputChannel::output_frame(DisplayFrame frame)
{
        // Store this frame for display. Remove the ready frame if any
        // (it was seemingly never used).
        {
                unique_lock<mutex> lock(frame_mutex);
                if (has_ready_frame) {
                        parent->release_display_frame(&ready_frame);
                }
                ready_frame = frame;
                has_ready_frame = true;

                // Call the callbacks under the mutex (they should be short),
                // so that we don't race against a callback removal.
                for (const auto &key_and_callback : new_frame_ready_callbacks) {
                        key_and_callback.second();
                }
        }

        // Reduce the number of callbacks by filtering duplicates. The reason
        // why we bother doing this is that Qt seemingly can get into a state
        // where its builds up an essentially unbounded queue of signals,
        // consuming more and more memory, and there's no good way of collapsing
        // user-defined signals or limiting the length of the queue.
        if (transition_names_updated_callback) {
                vector<string> transition_names = global_mixer->get_transition_names();
                bool changed = false;
                if (transition_names.size() != last_transition_names.size()) {
                        changed = true;
                } else {
                        for (unsigned i = 0; i < transition_names.size(); ++i) {
                                if (transition_names[i] != last_transition_names[i]) {
                                        changed = true;
                                        break;
                                }
                        }
                }
                if (changed) {
                        transition_names_updated_callback(transition_names);
                        last_transition_names = transition_names;
                }
        }
        if (name_updated_callback) {
                string name = global_mixer->get_channel_name(channel);
                if (name != last_name) {
                        name_updated_callback(name);
                        last_name = name;
                }
        }
        if (color_updated_callback) {
                string color = global_mixer->get_channel_color(channel);
                if (color != last_color) {
                        color_updated_callback(color);
                        last_color = color;
                }
        }
}

bool Mixer::OutputChannel::get_display_frame(DisplayFrame *frame)
{
        unique_lock<mutex> lock(frame_mutex);
        if (!has_current_frame && !has_ready_frame) {
                return false;
        }

        if (has_current_frame && has_ready_frame) {
                // We have a new ready frame. Toss the current one.
                parent->release_display_frame(&current_frame);
                has_current_frame = false;
        }
        if (has_ready_frame) {
                assert(!has_current_frame);
                current_frame = ready_frame;
                ready_frame.ready_fence.reset();  // Drop the refcount.
                ready_frame.input_frames.clear();  // Drop the refcounts.
                has_current_frame = true;
                has_ready_frame = false;
        }

        *frame = current_frame;
        return true;
}

void Mixer::OutputChannel::add_frame_ready_callback(void *key, Mixer::new_frame_ready_callback_t callback)
{
        unique_lock<mutex> lock(frame_mutex);
        new_frame_ready_callbacks[key] = callback;
}

void Mixer::OutputChannel::remove_frame_ready_callback(void *key)
{
        unique_lock<mutex> lock(frame_mutex);
        new_frame_ready_callbacks.erase(key);
}

void Mixer::OutputChannel::set_transition_names_updated_callback(Mixer::transition_names_updated_callback_t callback)
{
        transition_names_updated_callback = callback;
}

void Mixer::OutputChannel::set_name_updated_callback(Mixer::name_updated_callback_t callback)
{
        name_updated_callback = callback;
}

void Mixer::OutputChannel::set_color_updated_callback(Mixer::color_updated_callback_t callback)
{
        color_updated_callback = callback;
}

mutex RefCountedGLsync::fence_lock;