Phase 4

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLComposite.cpp

@@ -23,10 +23,8 @@
 #include <vector>
 
 OpenGLComposite::OpenGLComposite(HWND hWnd, HDC hDC, HGLRC hRC) :
-	hGLWnd(hWnd), hGLDC(hDC), hGLRC(hRC),
-	mScreenshotRequested(false)
+	hGLWnd(hWnd), hGLDC(hDC), hGLRC(hRC)
 {
-	InitializeCriticalSection(&pMutex);
 	mRuntimeStore = std::make_unique<RuntimeStore>();
 	mRuntimeEventDispatcher = std::make_unique<RuntimeEventDispatcher>();
 	mRuntimeSnapshotProvider = std::make_unique<RuntimeSnapshotProvider>(mRuntimeStore->GetRenderSnapshotBuilder(), *mRuntimeEventDispatcher);
@@ -34,11 +32,10 @@ OpenGLComposite::OpenGLComposite(HWND hWnd, HDC hDC, HGLRC hRC) :
 	mRenderEngine = std::make_unique<RenderEngine>(
 		*mRuntimeSnapshotProvider,
 		mRuntimeStore->GetHealthTelemetry(),
-		pMutex,
 		hGLDC,
 		hGLRC,
 		[this]() { renderEffect(); },
-		[this]() { ProcessScreenshotRequest(); },
+		[]() {},
 		[this]() { paintGL(false); });
 	mVideoBackend = std::make_unique<VideoBackend>(*mRenderEngine, mRuntimeStore->GetHealthTelemetry(), *mRuntimeEventDispatcher);
 	mShaderBuildQueue = std::make_unique<ShaderBuildQueue>(*mRuntimeSnapshotProvider, *mRuntimeEventDispatcher);
@@ -61,8 +58,6 @@ OpenGLComposite::~OpenGLComposite()
 		mShaderBuildQueue->Stop();
 	if (mVideoBackend)
 		mVideoBackend->ReleaseResources();
-
-	DeleteCriticalSection(&pMutex);
 }
 
 bool OpenGLComposite::InitDeckLink()
@@ -294,8 +289,50 @@ bool OpenGLComposite::ReloadShader(bool preserveFeedbackState)
 
 bool OpenGLComposite::RequestScreenshot(std::string& error)
 {
-	(void)error;
-	mScreenshotRequested.store(true);
+	if (!mRenderEngine || !mVideoBackend)
+	{
+		error = "The render engine is not ready.";
+		return false;
+	}
+
+	const unsigned width = mVideoBackend->OutputFrameWidth();
+	const unsigned height = mVideoBackend->OutputFrameHeight();
+	if (width == 0 || height == 0)
+	{
+		error = "The output frame size is not available.";
+		return false;
+	}
+
+	std::filesystem::path outputPath;
+	try
+	{
+		outputPath = BuildScreenshotPath();
+		std::filesystem::create_directories(outputPath.parent_path());
+	}
+	catch (const std::exception& exception)
+	{
+		error = exception.what();
+		return false;
+	}
+
+	if (!mRenderEngine->RequestScreenshotCapture(
+		width,
+		height,
+		[outputPath](unsigned captureWidth, unsigned captureHeight, std::vector<unsigned char> topDownPixels) {
+			try
+			{
+				WritePngFileAsync(outputPath, captureWidth, captureHeight, std::move(topDownPixels));
+			}
+			catch (const std::exception& exception)
+			{
+				OutputDebugStringA((std::string("Screenshot request failed: ") + exception.what() + "\n").c_str());
+			}
+		}))
+	{
+		error = "Screenshot capture request failed.";
+		return false;
+	}
+
 	return true;
 }
 
@@ -342,32 +379,6 @@ void OpenGLComposite::RenderFrame(const RenderFrameInput& frameInput)
 	mRenderEngine->RenderPreparedFrame(frameState);
 }
 
-void OpenGLComposite::ProcessScreenshotRequest()
-{
-	if (!mScreenshotRequested.exchange(false))
-		return;
-
-	const unsigned width = mVideoBackend ? mVideoBackend->OutputFrameWidth() : 0;
-	const unsigned height = mVideoBackend ? mVideoBackend->OutputFrameHeight() : 0;
-	if (width == 0 || height == 0)
-		return;
-
-	std::vector<unsigned char> topDownPixels;
-	if (!mRenderEngine->CaptureOutputFrameRgbaTopDown(width, height, topDownPixels))
-		return;
-
-	try
-	{
-		const std::filesystem::path outputPath = BuildScreenshotPath();
-		std::filesystem::create_directories(outputPath.parent_path());
-		WritePngFileAsync(outputPath, width, height, std::move(topDownPixels));
-	}
-	catch (const std::exception& exception)
-	{
-		OutputDebugStringA((std::string("Screenshot request failed: ") + exception.what() + "\n").c_str());
-	}
-}
-
 std::filesystem::path OpenGLComposite::BuildScreenshotPath() const
 {
 	const std::filesystem::path root = mRuntimeStore && !mRuntimeStore->GetRuntimeDataRoot().empty()

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLComposite.h

@@ -15,7 +15,6 @@
 #include "RenderFrameState.h"
 
 #include <functional>
-#include <atomic>
 #include <filesystem>
 #include <memory>
 #include <string>
@@ -73,7 +72,6 @@ private:
 	HWND									hGLWnd;
 	HDC										hGLDC;
 	HGLRC									hGLRC;
-	CRITICAL_SECTION						pMutex;
 
 	std::unique_ptr<RuntimeStore>			mRuntimeStore;
 	std::unique_ptr<RuntimeCoordinator>		mRuntimeCoordinator;
@@ -84,13 +82,11 @@ private:
 	std::unique_ptr<RuntimeServices>		mRuntimeServices;
 	std::unique_ptr<RuntimeUpdateController> mRuntimeUpdateController;
 	std::unique_ptr<VideoBackend>			mVideoBackend;
-	std::atomic<bool>						mScreenshotRequested;
 
 	bool InitOpenGLState();
 	void renderEffect();
 	RenderFrameInput BuildRenderFrameInput() const;
 	void RenderFrame(const RenderFrameInput& frameInput);
-	void ProcessScreenshotRequest();
 	std::filesystem::path BuildScreenshotPath() const;
 };
 

apps/LoopThroughWithOpenGLCompositing/gl/RenderCommandQueue.h

@@ -6,6 +6,7 @@
 #include <cstdint>
 #include <deque>
 #include <mutex>
+#include <vector>
 
 enum class RenderCommandResetScope
 {
@@ -31,6 +32,7 @@ struct RenderInputUploadRequest
 {
 	VideoIOFrame inputFrame;
 	VideoIOState videoState;
+	std::vector<unsigned char> ownedBytes;
 };
 
 struct RenderOutputFrameRequest

apps/LoopThroughWithOpenGLCompositing/gl/RenderEngine.cpp

@@ -3,11 +3,11 @@
 #include <gl/gl.h>
 
 #include <algorithm>
+#include <cstring>
 
 RenderEngine::RenderEngine(
 	RuntimeSnapshotProvider& runtimeSnapshotProvider,
 	HealthTelemetry& healthTelemetry,
-	CRITICAL_SECTION& mutex,
 	HDC hdc,
 	HGLRC hglrc,
 	RenderEffectCallback renderEffect,
@@ -17,7 +17,6 @@ RenderEngine::RenderEngine(
 	mRenderPass(mRenderer),
 	mRenderPipeline(mRenderer, runtimeSnapshotProvider, healthTelemetry, std::move(renderEffect), std::move(screenshotReady), std::move(previewPaint)),
 	mShaderPrograms(mRenderer, runtimeSnapshotProvider),
-	mMutex(mutex),
 	mHdc(hdc),
 	mHglrc(hglrc),
 	mFrameStateResolver(runtimeSnapshotProvider)
@@ -136,6 +135,24 @@ void RenderEngine::ReportRenderThreadRequestFailure(const char* operationName, c
 	OutputDebugStringA(message.str().c_str());
 }
 
+bool RenderEngine::IsRenderThreadAccessExpected() const
+{
+	return !mRenderThreadRunning || GetCurrentThreadId() == mRenderThreadId;
+}
+
+void RenderEngine::ReportWrongThreadRenderAccess(const char* operationName) const
+{
+	if (IsRenderThreadAccessExpected())
+		return;
+
+	std::ostringstream message;
+	message << "Wrong-thread render access detected";
+	if (operationName && operationName[0] != '\0')
+		message << " [" << operationName << "]";
+	message << ".\n";
+	OutputDebugStringA(message.str().c_str());
+}
+
 bool RenderEngine::CompileDecodeShader(int errorMessageSize, char* errorMessage)
 {
 	return InvokeOnRenderThread([this, errorMessageSize, errorMessage]() {
@@ -241,11 +258,13 @@ void RenderEngine::ApplyRuntimeCoordinatorRenderReset(RuntimeCoordinatorRenderRe
 
 void RenderEngine::ResetTemporalHistoryStateOnRenderThread()
 {
+	ReportWrongThreadRenderAccess("reset-temporal-history");
 	mShaderPrograms.ResetTemporalHistoryState();
 }
 
 void RenderEngine::ResetShaderFeedbackStateOnRenderThread()
 {
+	ReportWrongThreadRenderAccess("reset-shader-feedback");
 	mShaderPrograms.ResetShaderFeedbackState();
 }
 
@@ -276,6 +295,124 @@ void RenderEngine::ProcessRenderResetCommandsOnRenderThread()
 		ApplyRenderResetOnRenderThread(resetScope);
 }
 
+void RenderEngine::EnqueuePreviewPresentWake()
+{
+	if (!mRenderThreadRunning || GetCurrentThreadId() == mRenderThreadId)
+		return;
+
+	bool shouldNotify = false;
+	{
+		std::lock_guard<std::mutex> lock(mRenderThreadMutex);
+		if (!mRenderThreadStopping && !mPreviewPresentWakePending)
+		{
+			mPreviewPresentWakePending = true;
+			mRenderThreadTasks.push([this]() {
+				{
+					std::lock_guard<std::mutex> lock(mRenderThreadMutex);
+					mPreviewPresentWakePending = false;
+				}
+				ProcessPreviewPresentCommandsOnRenderThread();
+			});
+			shouldNotify = true;
+		}
+	}
+
+	if (shouldNotify)
+		mRenderThreadCondition.notify_one();
+}
+
+void RenderEngine::ProcessPreviewPresentCommandsOnRenderThread()
+{
+	RenderPreviewPresentRequest request;
+	if (mRenderCommandQueue.TryTakePreviewPresent(request))
+		PresentPreviewOnRenderThread(request.outputFrameWidth, request.outputFrameHeight);
+}
+
+void RenderEngine::EnqueueInputUploadWake()
+{
+	if (!mRenderThreadRunning || GetCurrentThreadId() == mRenderThreadId)
+		return;
+
+	bool shouldNotify = false;
+	{
+		std::lock_guard<std::mutex> lock(mRenderThreadMutex);
+		if (!mRenderThreadStopping && !mInputUploadWakePending)
+		{
+			mInputUploadWakePending = true;
+			mRenderThreadTasks.push([this]() {
+				{
+					std::lock_guard<std::mutex> lock(mRenderThreadMutex);
+					mInputUploadWakePending = false;
+				}
+				ProcessInputUploadCommandsOnRenderThread();
+			});
+			shouldNotify = true;
+		}
+	}
+
+	if (shouldNotify)
+		mRenderThreadCondition.notify_one();
+}
+
+void RenderEngine::ProcessInputUploadCommandsOnRenderThread()
+{
+	RenderInputUploadRequest request;
+	while (mRenderCommandQueue.TryTakeInputUpload(request))
+	{
+		if (request.ownedBytes.empty())
+			continue;
+
+		request.inputFrame.bytes = request.ownedBytes.data();
+		UploadInputFrameOnRenderThread(request.inputFrame, request.videoState);
+	}
+}
+
+void RenderEngine::EnqueueScreenshotCaptureWake()
+{
+	if (!mRenderThreadRunning || GetCurrentThreadId() == mRenderThreadId)
+		return;
+
+	bool shouldNotify = false;
+	{
+		std::lock_guard<std::mutex> lock(mRenderThreadMutex);
+		if (!mRenderThreadStopping && !mScreenshotCaptureWakePending)
+		{
+			mScreenshotCaptureWakePending = true;
+			mRenderThreadTasks.push([this]() {
+				{
+					std::lock_guard<std::mutex> lock(mRenderThreadMutex);
+					mScreenshotCaptureWakePending = false;
+				}
+				ProcessScreenshotCaptureCommandsOnRenderThread();
+			});
+			shouldNotify = true;
+		}
+	}
+
+	if (shouldNotify)
+		mRenderThreadCondition.notify_one();
+}
+
+void RenderEngine::ProcessScreenshotCaptureCommandsOnRenderThread()
+{
+	RenderScreenshotCaptureRequest request;
+	ScreenshotCaptureCallback completion;
+	{
+		std::lock_guard<std::mutex> lock(mRenderThreadMutex);
+		completion = mScreenshotCaptureCompletion;
+	}
+
+	while (mRenderCommandQueue.TryTakeScreenshotCapture(request))
+	{
+		if (!completion)
+			continue;
+
+		std::vector<unsigned char> topDownPixels;
+		if (CaptureOutputFrameRgbaTopDownOnRenderThread(request.width, request.height, topDownPixels))
+			completion(request.width, request.height, std::move(topDownPixels));
+	}
+}
+
 void RenderEngine::ClearOscOverlayState()
 {
 	mRuntimeLiveState.Clear();
@@ -323,32 +460,62 @@ bool RenderEngine::TryPresentPreview(bool force, unsigned previewFps, unsigned o
 
 	if (mRenderThreadRunning)
 	{
-		return TryInvokeOnRenderThread("preview-present", [this, outputFrameWidth, outputFrameHeight]() {
-			mRenderCommandQueue.RequestPreviewPresent({ outputFrameWidth, outputFrameHeight });
-			RenderPreviewPresentRequest request;
-			return mRenderCommandQueue.TryTakePreviewPresent(request) &&
-				PresentPreviewOnRenderThread(request.outputFrameWidth, request.outputFrameHeight);
-		});
+		mRenderCommandQueue.RequestPreviewPresent({ outputFrameWidth, outputFrameHeight });
+		if (GetCurrentThreadId() == mRenderThreadId)
+			ProcessPreviewPresentCommandsOnRenderThread();
+		else
+			EnqueuePreviewPresentWake();
+		return true;
 	}
 
-	if (!TryEnterCriticalSection(&mMutex))
-		return false;
-
-	mRenderCommandQueue.RequestPreviewPresent({ outputFrameWidth, outputFrameHeight });
-	RenderPreviewPresentRequest request;
-	const bool presented = mRenderCommandQueue.TryTakePreviewPresent(request) &&
-		PresentPreviewOnRenderThread(request.outputFrameWidth, request.outputFrameHeight);
-	LeaveCriticalSection(&mMutex);
-	return presented;
+	ReportRenderThreadRequestFailure("preview-present", "render thread is not running");
+	return false;
 }
 
 bool RenderEngine::PresentPreviewOnRenderThread(unsigned outputFrameWidth, unsigned outputFrameHeight)
 {
+	ReportWrongThreadRenderAccess("preview-present");
 	mRenderer.PresentToWindow(mHdc, outputFrameWidth, outputFrameHeight);
 	mLastPreviewPresentTime = std::chrono::steady_clock::now();
 	return true;
 }
 
+bool RenderEngine::RequestScreenshotCapture(unsigned width, unsigned height, ScreenshotCaptureCallback completion)
+{
+	if (width == 0 || height == 0 || !completion)
+		return false;
+	if (!mRenderThreadRunning)
+		return false;
+
+	{
+		std::lock_guard<std::mutex> lock(mRenderThreadMutex);
+		mScreenshotCaptureCompletion = std::move(completion);
+	}
+	mRenderCommandQueue.RequestScreenshotCapture({ width, height });
+	EnqueueScreenshotCaptureWake();
+	return true;
+}
+
+bool RenderEngine::QueueInputFrame(const VideoIOFrame& inputFrame, const VideoIOState& videoState)
+{
+	if (inputFrame.hasNoInputSource || inputFrame.bytes == nullptr)
+		return true;
+	if (inputFrame.rowBytes <= 0 || inputFrame.height == 0)
+		return false;
+
+	const std::size_t byteCount = static_cast<std::size_t>(inputFrame.rowBytes) * inputFrame.height;
+	RenderInputUploadRequest request;
+	request.inputFrame = inputFrame;
+	request.videoState = videoState;
+	request.ownedBytes.resize(byteCount);
+	std::memcpy(request.ownedBytes.data(), inputFrame.bytes, byteCount);
+	request.inputFrame.bytes = nullptr;
+
+	mRenderCommandQueue.RequestInputUpload(request);
+	EnqueueInputUploadWake();
+	return true;
+}
+
 bool RenderEngine::TryUploadInputFrame(const VideoIOFrame& inputFrame, const VideoIOState& videoState)
 {
 	if (inputFrame.hasNoInputSource || inputFrame.bytes == nullptr)
@@ -364,21 +531,13 @@ bool RenderEngine::TryUploadInputFrame(const VideoIOFrame& inputFrame, const Vid
 		});
 	}
 
-	if (!TryEnterCriticalSection(&mMutex))
-		return false;
-
-	wglMakeCurrent(mHdc, mHglrc);
-	mRenderCommandQueue.RequestInputUpload({ inputFrame, videoState });
-	RenderInputUploadRequest request;
-	const bool uploaded = mRenderCommandQueue.TryTakeInputUpload(request) &&
-		UploadInputFrameOnRenderThread(request.inputFrame, request.videoState);
-	wglMakeCurrent(NULL, NULL);
-	LeaveCriticalSection(&mMutex);
-	return uploaded;
+	ReportRenderThreadRequestFailure("input-upload", "render thread is not running");
+	return false;
 }
 
 bool RenderEngine::UploadInputFrameOnRenderThread(const VideoIOFrame& inputFrame, const VideoIOState& videoState)
 {
+	ReportWrongThreadRenderAccess("input-upload");
 	const long textureSize = inputFrame.rowBytes * static_cast<long>(inputFrame.height);
 	glPixelStorei(GL_UNPACK_ALIGNMENT, 4);
 	glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
@@ -395,7 +554,7 @@ bool RenderEngine::UploadInputFrameOnRenderThread(const VideoIOFrame& inputFrame
 	return true;
 }
 
-bool RenderEngine::RenderOutputFrame(const RenderPipelineFrameContext& context, VideoIOOutputFrame& outputFrame)
+bool RenderEngine::RequestOutputFrame(const RenderPipelineFrameContext& context, VideoIOOutputFrame& outputFrame)
 {
 	if (mRenderThreadRunning)
 	{
@@ -407,20 +566,20 @@ bool RenderEngine::RenderOutputFrame(const RenderPipelineFrameContext& context,
 		});
 	}
 
-	EnterCriticalSection(&mMutex);
-	wglMakeCurrent(mHdc, mHglrc);
-	mRenderCommandQueue.RequestOutputFrame({ context.videoState, context.completion });
-	RenderOutputFrameRequest request;
-	const bool rendered = mRenderCommandQueue.TryTakeOutputFrame(request) &&
-		RenderOutputFrameOnRenderThread({ request.videoState, request.completion }, outputFrame);
-	wglMakeCurrent(NULL, NULL);
-	LeaveCriticalSection(&mMutex);
-	return rendered;
+	ReportRenderThreadRequestFailure("output-render", "render thread is not running");
+	return false;
+}
+
+bool RenderEngine::RenderOutputFrame(const RenderPipelineFrameContext& context, VideoIOOutputFrame& outputFrame)
+{
+	return RequestOutputFrame(context, outputFrame);
 }
 
 bool RenderEngine::RenderOutputFrameOnRenderThread(const RenderPipelineFrameContext& context, VideoIOOutputFrame& outputFrame)
 {
+	ReportWrongThreadRenderAccess("output-render");
 	ProcessRenderResetCommandsOnRenderThread();
+	ProcessInputUploadCommandsOnRenderThread();
 	return mRenderPipeline.RenderFrame(context, outputFrame);
 }
 
@@ -466,6 +625,7 @@ void RenderEngine::RenderLayerStack(
 	VideoIOPixelFormat inputPixelFormat,
 	unsigned historyCap)
 {
+	ReportWrongThreadRenderAccess("render-layer-stack");
 	mRenderPass.Render(
 		hasInputSource,
 		layerStates,
@@ -484,6 +644,7 @@ void RenderEngine::RenderLayerStack(
 
 bool RenderEngine::ReadOutputFrameRgbaOnRenderThread(unsigned width, unsigned height, std::vector<unsigned char>& bottomUpPixels)
 {
+	ReportWrongThreadRenderAccess("read-output-frame-rgba");
 	if (width == 0 || height == 0)
 		return false;
 
@@ -510,15 +671,8 @@ bool RenderEngine::CaptureOutputFrameRgbaTopDown(unsigned width, unsigned height
 		});
 	}
 
-	EnterCriticalSection(&mMutex);
-	wglMakeCurrent(mHdc, mHglrc);
-	mRenderCommandQueue.RequestScreenshotCapture({ width, height });
-	RenderScreenshotCaptureRequest request;
-	const bool captured = mRenderCommandQueue.TryTakeScreenshotCapture(request) &&
-		CaptureOutputFrameRgbaTopDownOnRenderThread(request.width, request.height, topDownPixels);
-	wglMakeCurrent(NULL, NULL);
-	LeaveCriticalSection(&mMutex);
-	return captured;
+	ReportRenderThreadRequestFailure("screenshot-capture", "render thread is not running");
+	return false;
 }
 
 bool RenderEngine::CaptureOutputFrameRgbaTopDownOnRenderThread(unsigned width, unsigned height, std::vector<unsigned char>& topDownPixels)

apps/LoopThroughWithOpenGLCompositing/gl/RenderEngine.h

@@ -33,6 +33,7 @@ class RenderEngine
 public:
 	using RenderEffectCallback = std::function<void()>;
 	using ScreenshotCallback = std::function<void()>;
+	using ScreenshotCaptureCallback = std::function<void(unsigned, unsigned, std::vector<unsigned char>)>;
 	using PreviewPaintCallback = std::function<void()>;
 
 	struct OscOverlayUpdate
@@ -61,7 +62,6 @@ public:
 	RenderEngine(
 		RuntimeSnapshotProvider& runtimeSnapshotProvider,
 		HealthTelemetry& healthTelemetry,
-		CRITICAL_SECTION& mutex,
 		HDC hdc,
 		HGLRC hglrc,
 		RenderEffectCallback renderEffect,
@@ -100,7 +100,10 @@ public:
 		const std::vector<OscOverlayCommitCompletion>& completedCommits);
 	void ResizeView(int width, int height);
 	bool TryPresentPreview(bool force, unsigned previewFps, unsigned outputFrameWidth, unsigned outputFrameHeight);
+	bool RequestScreenshotCapture(unsigned width, unsigned height, ScreenshotCaptureCallback completion);
+	bool QueueInputFrame(const VideoIOFrame& inputFrame, const VideoIOState& videoState);
 	bool TryUploadInputFrame(const VideoIOFrame& inputFrame, const VideoIOState& videoState);
+	bool RequestOutputFrame(const RenderPipelineFrameContext& context, VideoIOOutputFrame& outputFrame);
 	bool RenderOutputFrame(const RenderPipelineFrameContext& context, VideoIOOutputFrame& outputFrame);
 	bool ResolveRenderFrameState(
 		const RenderFrameInput& input,
@@ -178,6 +181,8 @@ private:
 
 	void RenderThreadMain(std::promise<bool> ready);
 	void ReportRenderThreadRequestFailure(const char* operationName, const char* reason);
+	bool IsRenderThreadAccessExpected() const;
+	void ReportWrongThreadRenderAccess(const char* operationName) const;
 	bool CommitPreparedLayerPrograms(const PreparedShaderBuild& preparedBuild, unsigned inputFrameWidth, unsigned inputFrameHeight, int errorMessageSize, char* errorMessage);
 	void RenderLayerStack(
 		bool hasInputSource,
@@ -191,6 +196,12 @@ private:
 	void ResetShaderFeedbackStateOnRenderThread();
 	void ApplyRenderResetOnRenderThread(RenderCommandResetScope resetScope);
 	void ProcessRenderResetCommandsOnRenderThread();
+	void EnqueuePreviewPresentWake();
+	void ProcessPreviewPresentCommandsOnRenderThread();
+	void EnqueueInputUploadWake();
+	void ProcessInputUploadCommandsOnRenderThread();
+	void EnqueueScreenshotCaptureWake();
+	void ProcessScreenshotCaptureCommandsOnRenderThread();
 	bool PresentPreviewOnRenderThread(unsigned outputFrameWidth, unsigned outputFrameHeight);
 	bool UploadInputFrameOnRenderThread(const VideoIOFrame& inputFrame, const VideoIOState& videoState);
 	bool RenderOutputFrameOnRenderThread(const RenderPipelineFrameContext& context, VideoIOOutputFrame& outputFrame);
@@ -201,7 +212,6 @@ private:
 	OpenGLRenderPass mRenderPass;
 	OpenGLRenderPipeline mRenderPipeline;
 	OpenGLShaderPrograms mShaderPrograms;
-	CRITICAL_SECTION& mMutex;
 	HDC mHdc;
 	HGLRC mHglrc;
 
@@ -216,5 +226,9 @@ private:
 	std::queue<std::function<void()>> mRenderThreadTasks;
 	std::atomic<bool> mRenderThreadRunning = false;
 	bool mRenderThreadStopping = false;
+	bool mPreviewPresentWakePending = false;
+	bool mInputUploadWakePending = false;
+	bool mScreenshotCaptureWakePending = false;
+	ScreenshotCaptureCallback mScreenshotCaptureCompletion;
 	bool mResourcesDestroyed = false;
 };

apps/LoopThroughWithOpenGLCompositing/gl/pipeline/OpenGLVideoIOBridge.cpp

@@ -12,14 +12,14 @@ void OpenGLVideoIOBridge::UploadInputFrame(const VideoIOFrame& inputFrame, const
 	if (inputFrame.hasNoInputSource || inputFrame.bytes == nullptr)
 		return;							// don't transfer texture when there's no input
 
-	mRenderEngine.TryUploadInputFrame(inputFrame, state);
+	mRenderEngine.QueueInputFrame(inputFrame, state);
 }
 
-void OpenGLVideoIOBridge::RenderScheduledFrame(const VideoIOState& state, const VideoIOCompletion& completion, VideoIOOutputFrame& outputFrame)
+bool OpenGLVideoIOBridge::RenderScheduledFrame(const VideoIOState& state, const VideoIOCompletion& completion, VideoIOOutputFrame& outputFrame)
 {
 	RenderPipelineFrameContext frameContext;
 	frameContext.videoState = state;
 	frameContext.completion = completion;
 
-	mRenderEngine.RenderOutputFrame(frameContext, outputFrame);
+	return mRenderEngine.RequestOutputFrame(frameContext, outputFrame);
 }

apps/LoopThroughWithOpenGLCompositing/gl/pipeline/OpenGLVideoIOBridge.h

@@ -10,7 +10,7 @@ public:
 	explicit OpenGLVideoIOBridge(RenderEngine& renderEngine);
 
 	void UploadInputFrame(const VideoIOFrame& inputFrame, const VideoIOState& state);
-	void RenderScheduledFrame(const VideoIOState& state, const VideoIOCompletion& completion, VideoIOOutputFrame& outputFrame);
+	bool RenderScheduledFrame(const VideoIOState& state, const VideoIOCompletion& completion, VideoIOOutputFrame& outputFrame);
 
 private:
 	RenderEngine& mRenderEngine;

apps/LoopThroughWithOpenGLCompositing/videoio/VideoBackend.cpp

@@ -232,11 +232,17 @@ void VideoBackend::HandleOutputFrameCompletion(const VideoIOCompletion& completi
 		return;
 
 	const VideoIOState& state = mVideoIODevice->State();
+	bool rendered = true;
 	if (mBridge)
-		mBridge->RenderScheduledFrame(state, completion, outputFrame);
+		rendered = mBridge->RenderScheduledFrame(state, completion, outputFrame);
 
 	EndOutputFrame(outputFrame);
 	AccountForCompletionResult(completion.result);
+	if (!rendered)
+	{
+		PublishBackendStateChanged("output-render-failed", "Output frame render request failed; skipping schedule for this frame.");
+		return;
+	}
 
 	// Schedule the next frame after render work is complete so device-side
 	// bookkeeping stays with the backend seam and the bridge stays render-only.

docs/PHASE_4_RENDER_THREAD_OWNERSHIP_DESIGN.md

@@ -7,16 +7,16 @@ Phase 1 named the subsystems. Phase 2 added the typed event substrate. Phase 3 m
 ## Status
 
 - Phase 4 design package: proposed.
-- Phase 4 implementation: Step 3 started. The existing synchronous `RenderEngine` entrypoints delegate their GL bodies to named `...OnRenderThread(...)` helpers, preview/screenshot/render-reset/input-upload/output-render requests pass through a small `RenderCommandQueue` compatibility mailbox, and `RenderEngine` now starts a dedicated render thread for normal runtime GL work.
+- Phase 4 implementation: Step 7 started. The existing synchronous `RenderEngine` entrypoints delegate their GL bodies to named `...OnRenderThread(...)` helpers, preview/screenshot/render-reset/input-upload/output-render requests pass through a small `RenderCommandQueue` compatibility mailbox, and `RenderEngine` now starts a dedicated render thread for normal runtime GL work.
 - Current alignment: the repo has a named frame-state contract and cleaner render-state preparation. Normal runtime GL work is routed through the render thread after startup, while startup initialization still runs before the render thread is started.
 
 Current GL ownership footholds:
 
-- `RenderEngine` owns GL resources, a dedicated render thread, the current synchronous compatibility shims, a small render command mailbox, and named render-thread helper methods.
+- `RenderEngine` owns GL resources, a dedicated render thread, the current synchronous compatibility shims, a small render command mailbox, named render-thread helper methods, and wrong-thread diagnostics for those helpers.
 - `RenderFrameInput` / `RenderFrameState` provide the frame-state contract that a render thread can consume.
 - `RenderFrameStateResolver` prepares the render-facing layer state before drawing.
-- `OpenGLVideoIOBridge` still calls `RenderEngine::TryUploadInputFrame(...)` from the input path and `RenderEngine::RenderOutputFrame(...)` from the output path.
-- `OpenGLComposite::paintGL(...)`, screenshot capture, input upload, and output rendering still call synchronous `RenderEngine` methods, but those methods now invoke render-thread work once `OpenGLComposite::Start()` has started the render thread.
+- `OpenGLVideoIOBridge` calls `RenderEngine::QueueInputFrame(...)` from the input path and `RenderEngine::RequestOutputFrame(...)` from the output path.
+- `OpenGLComposite::paintGL(...)`, screenshot capture, input upload, and output rendering enter render work through explicit `RenderEngine` requests. After `OpenGLComposite::Start()` starts the render thread, those requests do not bind the GL context on the caller thread.
 
 ## Why Phase 4 Exists
 
@@ -65,11 +65,11 @@ The current code paths that matter most are:
 | Entry point | Current behavior | Phase 4 direction |
 | --- | --- | --- |
 | `RenderEngine::TryUploadInputFrame(...)` | synchronous compatibility shim; after render-thread startup it queues input upload work and waits for render-thread completion | enqueue latest input frame; render thread uploads without callback-owned GL |
-| `RenderEngine::RenderOutputFrame(...)` | synchronous compatibility shim; after render-thread startup it queues output render work and waits for render-thread completion | render thread executes output frame production |
-| `RenderEngine::TryPresentPreview(...)` | synchronous compatibility shim; after render-thread startup it queues preview presentation and waits for render-thread completion | render thread or preview presenter consumes latest completed frame |
+| `RenderEngine::RequestOutputFrame(...)` | synchronous output request; after render-thread startup it queues output render work and waits for render-thread completion with timeout/failure reporting | render thread executes output frame production |
+| `RenderEngine::TryPresentPreview(...)` | best-effort compatibility shim; after render-thread startup non-render callers queue preview presentation and return | render thread or preview presenter consumes latest completed frame |
 | `RenderEngine::CaptureOutputFrameRgbaTopDown(...)` | synchronous compatibility shim; after render-thread startup it queues screenshot readback and waits for render-thread completion | screenshot request becomes render-thread command |
-| `OpenGLVideoIOBridge::UploadInputFrame(...)` | calls render upload directly | push input frame into render queue/mailbox |
-| `OpenGLVideoIOBridge::RenderScheduledFrame(...)` | calls render output directly from backend path | request/consume render-produced output without callback-owned GL |
+| `OpenGLVideoIOBridge::UploadInputFrame(...)` | copies the latest input frame into the render mailbox and returns without waiting for GL | render thread uploads the latest queued input frame |
+| `OpenGLVideoIOBridge::RenderScheduledFrame(...)` | requests render-thread output production and reports success/failure to the backend | consume render-produced output without callback-owned GL |
 
 ## Target Ownership Model
 
@@ -260,10 +260,12 @@ Change `OpenGLVideoIOBridge::UploadInputFrame(...)` so it enqueues or replaces t
 
 Policy targets:
 
-- bounded memory
-- latest-frame wins under load
-- input upload skip count is observable
-- input callback never waits for GL
+- [x] bounded memory
+- [x] latest-frame wins under load
+- [x] input upload skip count is observable through render command coalescing metrics
+- [x] input callback never waits for GL
+
+Current implementation: `OpenGLVideoIOBridge::UploadInputFrame(...)` calls `RenderEngine::QueueInputFrame(...)`, which copies the input bytes into the latest-value render mailbox and schedules one bounded render-thread wakeup to upload the newest pending frame.
 
 ### Step 5. Move Output Rendering To The Render Thread
 
@@ -271,32 +273,38 @@ Change `OpenGLVideoIOBridge::RenderScheduledFrame(...)` so it requests render-th
 
 Transitional option:
 
-- synchronous request/response through the render thread
+- [x] synchronous request/response through the render thread
 
 Better follow-up:
 
 - render ahead into a bounded output queue and let backend callbacks consume ready frames
 
+Current implementation: `OpenGLVideoIOBridge::RenderScheduledFrame(...)` calls `RenderEngine::RequestOutputFrame(...)` and returns whether the render-thread request produced an output frame. `VideoBackend` skips scheduling that frame when render production fails or times out.
+
 ### Step 6. Decouple Preview And Screenshot Requests
 
 Preview should become best-effort:
 
-- request preview presentation from the render thread
-- skip when render is busy or output deadline pressure is high
-- record preview skips
+- [x] request preview presentation from the render thread
+- [x] skip/coalesce when render is busy or output deadline pressure is high
+- [x] record preview skips through render command coalescing metrics
 
 Screenshot should become:
 
-- queued render-thread capture request
-- async disk write remains outside render thread
+- [x] queued render-thread capture request
+- [x] async disk write remains outside render thread
+
+Current implementation: `OpenGLComposite::RequestScreenshot(...)` builds the output path, queues `RenderEngine::RequestScreenshotCapture(...)`, and the render thread captures pixels before handing them to the existing async PNG writer. Preview presentation is a latest-value best-effort render command; non-render callers enqueue and return, while render-thread callers drain the latest preview command inline.
 
 ### Step 7. Remove Shared GL Lock From Normal Paths
 
 Once all GL entrypoints are render-thread-owned:
 
-- remove normal dependence on `pMutex` for render correctness
-- keep assertions or diagnostics that detect wrong-thread GL calls
-- leave only lifecycle synchronization where needed
+- [x] remove normal dependence on `pMutex` for render correctness
+- [x] keep diagnostics that detect wrong-thread render-thread helper calls
+- [x] leave only lifecycle context binding where needed
+
+Current implementation: `OpenGLComposite` no longer owns or passes a shared `CRITICAL_SECTION`, and `RenderEngine` no longer has caller-thread GL fallback paths for preview, input upload, output render, or screenshot capture. Runtime callers must go through the render thread; pre-start direct GL fallback is limited to startup initialization while the app explicitly owns the context.
 
 ## Testing Strategy
 
@@ -366,12 +374,12 @@ Preview can remain a hidden budget consumer if it stays in the output frame path
 
 Phase 4 can be considered complete once the project can say:
 
-- [ ] one render thread owns the GL context during normal operation
-- [ ] input callbacks do not bind GL or wait on GL upload
-- [ ] output callbacks do not bind GL directly
-- [ ] preview and screenshot requests enter render through explicit render-thread requests
+- [x] one render thread owns the GL context during normal operation
+- [x] input callbacks do not bind GL or wait on GL upload
+- [x] output callbacks do not bind GL directly
+- [x] preview and screenshot requests enter render through explicit render-thread requests
 - [ ] `RenderFrameInput` / `RenderFrameState` remain the frame-state contract
-- [ ] normal frame production no longer depends on a shared GL `CRITICAL_SECTION`
+- [x] normal frame production no longer depends on a shared GL `CRITICAL_SECTION`
 - [ ] render-thread queue/mailbox behavior has non-GL tests
 - [ ] shutdown order is explicit and tested or manually verified
 

tests/RenderCommandQueueTests.cpp

@@ -70,12 +70,14 @@ void TestInputUploadRequestUsesLatestValue()
 	firstRequest.inputFrame.bytes = &firstPixel;
 	firstRequest.inputFrame.width = 1920;
 	firstRequest.videoState.captureTextureWidth = 1920;
+	firstRequest.ownedBytes = { 1, 2, 3, 4 };
 	queue.RequestInputUpload(firstRequest);
 
 	RenderInputUploadRequest secondRequest;
 	secondRequest.inputFrame.bytes = &secondPixel;
 	secondRequest.inputFrame.width = 1280;
 	secondRequest.videoState.captureTextureWidth = 1280;
+	secondRequest.ownedBytes = { 5, 6 };
 	queue.RequestInputUpload(secondRequest);
 
 	const RenderCommandQueueMetrics metrics = queue.GetMetrics();
@@ -88,6 +90,7 @@ void TestInputUploadRequestUsesLatestValue()
 	Expect(request.inputFrame.bytes == &secondPixel, "latest input upload bytes pointer wins");
 	Expect(request.inputFrame.width == 1280, "latest input upload frame wins");
 	Expect(request.videoState.captureTextureWidth == 1280, "latest input upload state wins");
+	Expect(request.ownedBytes.size() == 2 && request.ownedBytes[0] == 5 && request.ownedBytes[1] == 6, "latest input upload owned bytes win");
 	Expect(!queue.TryTakeInputUpload(request), "input upload request is removed after consume");
 }