Reduce latency with DXGI 1.3 GetFrameLatencyWaitableObject (#6435)

This pull request reduces input lag, especially with selection, by using
`IDXGISwapChain2::GetFrameLatencyWaitableObject`.

This is based on the [DXGI 1.3 documentation].

Excerpt from the [DXGI 1.3 improvement list]:
> The following functionality has been added in Microsoft DirectX
Graphics Infrastructure (DXGI) 1.3, which is included starting in
Windows 8.1.

Before, during rendering:
1. render frame
2. call `Present` on swap chain:
   2.a. blocks until it can present
   2.b. meanwhile, selection/text in terminal might have changed, but
     we're still using the frame that we rendered before blocking
   2.c. presents

After, during rendering:
1. block until we can present
2. render frame with latest data
3. call `Present` on swap chain:
  3.a. present without blocking

[DXGI 1.3 documentation]: https://docs.microsoft.com/en-us/windows/uwp/gaming/reduce-latency-with-dxgi-1-3-swap-chains
[DXGI 1.3 improvement list]: https://docs.microsoft.com/en-us/windows/win32/direct3ddxgi/dxgi-1-3-improvements:

src/renderer/base/RenderEngineBase.cpp

@@ -40,3 +40,10 @@ HRESULT RenderEngineBase::PrepareRenderInfo(const RenderFrameInfo& /*info*/) noe
 {
     return S_FALSE;
 }
+
+// Method Description:
+// - Blocks until the engine is able to render without blocking.
+void RenderEngineBase::WaitUntilCanRender() noexcept
+{
+    // do nothing by default
+}

src/renderer/base/renderer.cpp

@@ -1200,3 +1200,13 @@ void Renderer::ResetErrorStateAndResume()
     // because we're not stateful (we could be in the future), all we want to do is reenable painting.
     EnablePainting();
 }
+
+// Method Description:
+// - Blocks until the engines are able to render without blocking.
+void Renderer::WaitUntilCanRender()
+{
+    for (const auto pEngine : _rgpEngines)
+    {
+        pEngine->WaitUntilCanRender();
+    }
+}

src/renderer/base/renderer.hpp

@@ -73,6 +73,7 @@ namespace Microsoft::Console::Render
 
         void EnablePainting() override;
         void WaitForPaintCompletionAndDisable(const DWORD dwTimeoutMs) override;
+        void WaitUntilCanRender() override;
 
         void AddRenderEngine(_In_ IRenderEngine* const pEngine) override;
 

src/renderer/base/thread.cpp

@@ -201,6 +201,7 @@ DWORD WINAPI RenderThread::_ThreadProc()
 
         ResetEvent(_hPaintCompletedEvent);
 
+        _pRenderer->WaitUntilCanRender();
         LOG_IF_FAILED(_pRenderer->PaintFrame());
 
         SetEvent(_hPaintCompletedEvent);

src/renderer/dx/DxRenderer.cpp

@@ -83,6 +83,7 @@ DxEngine::DxEngine() :
     _glyphCell{},
     _boxDrawingEffect{},
     _haveDeviceResources{ false },
+    _swapChainFrameLatencyWaitableObject{ INVALID_HANDLE_VALUE },
     _retroTerminalEffects{ false },
     _forceFullRepaintRendering{ false },
     _softwareRendering{ false },
@@ -425,6 +426,11 @@ try
 
     if (createSwapChain)
     {
+        _swapChainFlags = 0;
+
+        // requires DXGI 1.3 which was introduced in Windows 8.1
+        WI_SetFlagIf(_swapChainFlags, DXGI_SWAP_CHAIN_FLAG_FRAME_LATENCY_WAITABLE_OBJECT, IsWindows8Point1OrGreater());
+
         DXGI_SWAP_CHAIN_DESC1 SwapChainDesc = { 0 };
         SwapChainDesc.Format = DXGI_FORMAT_B8G8R8A8_UNORM;
         SwapChainDesc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT;
@@ -433,6 +439,7 @@ try
         SwapChainDesc.SampleDesc.Count = 1;
         SwapChainDesc.AlphaMode = DXGI_ALPHA_MODE_UNSPECIFIED;
         SwapChainDesc.Scaling = DXGI_SCALING_NONE;
+        SwapChainDesc.Flags = _swapChainFlags;
 
         switch (_chainMode)
         {
@@ -487,6 +494,20 @@ try
             THROW_HR(E_NOTIMPL);
         }
 
+        if (IsWindows8Point1OrGreater())
+        {
+            ::Microsoft::WRL::ComPtr<IDXGISwapChain2> swapChain2;
+            const HRESULT asResult = _dxgiSwapChain.As(&swapChain2);
+            if (SUCCEEDED(asResult))
+            {
+                _swapChainFrameLatencyWaitableObject = wil::unique_handle{ swapChain2->GetFrameLatencyWaitableObject() };
+            }
+            else
+            {
+                LOG_HR_MSG(asResult, "Failed to obtain IDXGISwapChain2 from swap chain");
+            }
+        }
+
         if (_retroTerminalEffects)
         {
             const HRESULT hr = _SetupTerminalEffects();
@@ -612,6 +633,7 @@ void DxEngine::_ReleaseDeviceResources() noexcept
 
         _dxgiSurface.Reset();
         _dxgiSwapChain.Reset();
+        _swapChainFrameLatencyWaitableObject.reset();
 
         if (nullptr != _d3dDeviceContext.Get())
         {
@@ -960,7 +982,7 @@ try
             _d2dRenderTarget.Reset();
 
             // Change the buffer size and recreate the render target (and surface)
-            RETURN_IF_FAILED(_dxgiSwapChain->ResizeBuffers(2, clientSize.width<UINT>(), clientSize.height<UINT>(), DXGI_FORMAT_B8G8R8A8_UNORM, 0));
+            RETURN_IF_FAILED(_dxgiSwapChain->ResizeBuffers(2, clientSize.width<UINT>(), clientSize.height<UINT>(), DXGI_FORMAT_B8G8R8A8_UNORM, _swapChainFlags));
             RETURN_IF_FAILED(_PrepareRenderTarget());
 
             // OK we made it past the parts that can cause errors. We can release our failure handler.
@@ -1085,6 +1107,26 @@ CATCH_RETURN()
     return S_OK;
 }
 
+// Method Description:
+// - Blocks until the engine is able to render without blocking.
+// - See https://docs.microsoft.com/en-us/windows/uwp/gaming/reduce-latency-with-dxgi-1-3-swap-chains.
+void DxEngine::WaitUntilCanRender() noexcept
+{
+    if (!_swapChainFrameLatencyWaitableObject)
+    {
+        return;
+    }
+
+    const auto ret = WaitForSingleObjectEx(
+        _swapChainFrameLatencyWaitableObject.get(),
+        1000, // 1 second timeout (shouldn't ever occur)
+        true);
+    if (ret != WAIT_OBJECT_0)
+    {
+        LOG_WIN32_MSG(ret, "Waiting for swap chain frame latency waitable object returned error or timeout.");
+    }
+}
+
 // Routine Description:
 // - Takes queued drawing information and presents it to the screen.
 // - This is separated out so it can be done outside the lock as it's expensive.

src/renderer/dx/DxRenderer.hpp

@@ -9,6 +9,7 @@
 
 #include <dxgi.h>
 #include <dxgi1_2.h>
+#include <dxgi1_3.h>
 
 #include <d3d11.h>
 #include <d2d1.h>
@@ -73,6 +74,8 @@ namespace Microsoft::Console::Render
 
         [[nodiscard]] HRESULT StartPaint() noexcept override;
         [[nodiscard]] HRESULT EndPaint() noexcept override;
+
+        void WaitUntilCanRender() noexcept override;
         [[nodiscard]] HRESULT Present() noexcept override;
 
         [[nodiscard]] HRESULT ScrollFrame() noexcept override;
@@ -180,7 +183,9 @@ namespace Microsoft::Console::Render
         ::Microsoft::WRL::ComPtr<ID2D1RenderTarget> _d2dRenderTarget;
         ::Microsoft::WRL::ComPtr<ID2D1SolidColorBrush> _d2dBrushForeground;
         ::Microsoft::WRL::ComPtr<ID2D1SolidColorBrush> _d2dBrushBackground;
+        UINT _swapChainFlags;
         ::Microsoft::WRL::ComPtr<IDXGISwapChain1> _dxgiSwapChain;
+        wil::unique_handle _swapChainFrameLatencyWaitableObject;
 
         // Terminal effects resources.
         bool _retroTerminalEffects;

src/renderer/inc/IRenderEngine.hpp

@@ -49,6 +49,8 @@ namespace Microsoft::Console::Render
     public:
         [[nodiscard]] virtual HRESULT StartPaint() noexcept = 0;
         [[nodiscard]] virtual HRESULT EndPaint() noexcept = 0;
+
+        virtual void WaitUntilCanRender() noexcept = 0;
         [[nodiscard]] virtual HRESULT Present() noexcept = 0;
 
         [[nodiscard]] virtual HRESULT PrepareForTeardown(_Out_ bool* const pForcePaint) noexcept = 0;

src/renderer/inc/IRenderer.hpp

@@ -58,6 +58,7 @@ namespace Microsoft::Console::Render
 
         virtual void EnablePainting() = 0;
         virtual void WaitForPaintCompletionAndDisable(const DWORD dwTimeoutMs) = 0;
+        virtual void WaitUntilCanRender() = 0;
 
         virtual void AddRenderEngine(_In_ IRenderEngine* const pEngine) = 0;
 

src/renderer/inc/RenderEngineBase.hpp

@@ -40,6 +40,8 @@ namespace Microsoft::Console::Render
 
         [[nodiscard]] HRESULT PrepareRenderInfo(const RenderFrameInfo& info) noexcept override;
 
+        void WaitUntilCanRender() noexcept override;
+
     protected:
         [[nodiscard]] virtual HRESULT _DoUpdateTitle(const std::wstring& newTitle) noexcept = 0;