Clean up pass

apps/RenderCadenceCompositor/telemetry/TelemetryHealthMonitor.h

@@ -0,0 +1,116 @@
+#pragma once
+
+#include "CadenceTelemetry.h"
+#include "../logging/Logger.h"
+
+#include <atomic>
+#include <chrono>
+#include <sstream>
+#include <thread>
+
+namespace RenderCadenceCompositor
+{
+struct TelemetryHealthMonitorConfig
+{
+	std::chrono::milliseconds interval = std::chrono::seconds(1);
+	std::size_t scheduledStarvationThreshold = 0;
+};
+
+class TelemetryHealthMonitor
+{
+public:
+	explicit TelemetryHealthMonitor(TelemetryHealthMonitorConfig config = TelemetryHealthMonitorConfig()) :
+		mConfig(config)
+	{
+	}
+
+	TelemetryHealthMonitor(const TelemetryHealthMonitor&) = delete;
+	TelemetryHealthMonitor& operator=(const TelemetryHealthMonitor&) = delete;
+
+	~TelemetryHealthMonitor()
+	{
+		Stop();
+	}
+
+	template <typename SystemFrameExchange, typename Output, typename OutputThread, typename RenderThread>
+	void Start(const SystemFrameExchange& exchange, const Output& output, const OutputThread& outputThread, const RenderThread& renderThread)
+	{
+		if (mRunning)
+			return;
+		mStopping = false;
+		mThread = std::thread([this, &exchange, &output, &outputThread, &renderThread]() {
+			CadenceTelemetry telemetry;
+			CadenceTelemetrySnapshot previous;
+			bool hasPrevious = false;
+			while (!mStopping)
+			{
+				std::this_thread::sleep_for(mConfig.interval);
+				const CadenceTelemetrySnapshot snapshot = telemetry.Sample(exchange, output, outputThread, renderThread);
+				ReportHealth(snapshot, hasPrevious ? &previous : nullptr);
+				previous = snapshot;
+				hasPrevious = true;
+			}
+		});
+		mRunning = true;
+	}
+
+	void Stop()
+	{
+		mStopping = true;
+		if (mThread.joinable())
+			mThread.join();
+		mRunning = false;
+	}
+
+private:
+	void ReportHealth(const CadenceTelemetrySnapshot& snapshot, const CadenceTelemetrySnapshot* previous) const
+	{
+		if (!previous)
+			return;
+
+		const uint64_t lateDelta = snapshot.displayedLate - previous->displayedLate;
+		const uint64_t droppedDelta = snapshot.dropped - previous->dropped;
+		const uint64_t scheduleFailureDelta = snapshot.scheduleFailures - previous->scheduleFailures;
+
+		if (droppedDelta > 0 || lateDelta > 0)
+		{
+			std::ostringstream message;
+			message << "DeckLink reported frame timing issue: lateDelta=" << lateDelta
+				<< " droppedDelta=" << droppedDelta
+				<< " totalLate=" << snapshot.displayedLate
+				<< " totalDropped=" << snapshot.dropped;
+			LogWarning("telemetry", message.str());
+		}
+
+		if (scheduleFailureDelta > 0)
+		{
+			std::ostringstream message;
+			message << "DeckLink schedule failures increased: delta=" << scheduleFailureDelta
+				<< " total=" << snapshot.scheduleFailures;
+			LogWarning("telemetry", message.str());
+		}
+
+		const bool appScheduledStarved = snapshot.scheduledFrames <= mConfig.scheduledStarvationThreshold
+			&& snapshot.scheduledTotal > 0;
+		const bool deckLinkStarved = snapshot.deckLinkBufferedAvailable && snapshot.deckLinkBuffered == 0;
+		if (appScheduledStarved || deckLinkStarved)
+		{
+			std::ostringstream message;
+			message << "Output buffer starvation detected: scheduled=" << snapshot.scheduledFrames
+				<< " decklinkBuffered=";
+			if (snapshot.deckLinkBufferedAvailable)
+				message << snapshot.deckLinkBuffered;
+			else
+				message << "n/a";
+			message << " renderFps=" << snapshot.renderFps
+				<< " scheduleFps=" << snapshot.scheduleFps;
+			LogError("telemetry", message.str());
+		}
+	}
+
+	TelemetryHealthMonitorConfig mConfig;
+	std::thread mThread;
+	std::atomic<bool> mStopping{ false };
+	std::atomic<bool> mRunning{ false };
+};
+}