perf(gui): Significantly reduce cost of 2d render elements (TheSuperHackers#2514)

This increases the overall game performance by 20 to 50%

Author: githubawn

Core/GameEngineDevice/Source/W3DDevice/GameClient/W3DView.cpp

@@ -1942,7 +1942,10 @@ void W3DView::draw()
 	/// @todo we might want to consider wiping this iterate out if there is nothing to post draw
 	//
 	TheGameClient->resetRenderedObjectCount();
+
+	TheDisplay->beginBatch();
 	TheGameClient->iterateDrawablesInRegion( &axisAlignedRegion, drawablePostDraw, this );
+	TheDisplay->endBatch();
 
 	TheGameClient->flushTextBearingDrawables();
 

Generals/Code/GameEngine/Include/GameClient/Display.h

@@ -113,6 +113,12 @@ class Display : public SubsystemInterface
 	virtual	Bool isClippingEnabled() = 0;
 	virtual	void enableClipping( Bool onoff ) = 0;
 
+	// TheSuperHackers @performance Batching 2D draw operations to reduce state changes and draw call overhead.
+	virtual void beginBatch(); 									///< start batching 2D draw operations.
+	virtual void endBatch();   									///< stop batching and flush pending 2D draw operations.
+	virtual void flush();      									///< flush pending 2D draw operations without ending the batch.
+	virtual Bool isBatching() const { return m_isBatching; }	///< returns true if currently batching 2D draw operations.
+
 	virtual void step() {}; ///< Do one fixed time step
 	virtual void draw() override;																		///< Redraw the entire display
 	virtual void setTimeOfDay( TimeOfDay tod ) = 0;								///< Set the time of day for this display
@@ -182,11 +188,15 @@ class Display : public SubsystemInterface
 	virtual Int getLastFrameDrawCalls() = 0;  ///< returns the number of draw calls issued in the previous frame
 
 protected:
+	virtual void onBeginBatch() { }
+	virtual void onEndBatch() { }
+	virtual void onFlush() { }
 
 	virtual void deleteViews();   ///< delete all views
 	UnsignedInt m_width, m_height;			///< Dimensions of the display
 	UnsignedInt m_bitDepth;							///< bit depth of the display
 	Bool m_windowed;										///< TRUE when windowed, FALSE when fullscreen
+	Bool m_isBatching;
 	View *m_viewList;										///< All of the views into the world
 
 	// Cinematic text data

Generals/Code/GameEngine/Source/GameClient/Display.cpp

@@ -66,6 +66,7 @@ Display::Display()
 
 	m_currentlyPlayingMovie.clear();
 	m_letterBoxFadeStartTime = 0;
+	m_isBatching = FALSE;
 }
 
 /**
@@ -387,3 +388,32 @@ Display::DebugDisplayCallback *Display::getDebugDisplayCallback()
 {
 	return m_debugDisplayCallback;
 }
+
+void Display::beginBatch()
+{
+	if (m_isBatching)
+	{
+		return;
+	}
+	m_isBatching = TRUE;
+	onBeginBatch();
+}
+
+void Display::endBatch()
+{
+	if (!m_isBatching)
+	{
+		return;
+	}
+	onFlush();
+	m_isBatching = FALSE;
+	onEndBatch();
+}
+
+void Display::flush()
+{
+	if (m_isBatching)
+	{
+		onFlush();
+	}
+}

Generals/Code/GameEngineDevice/Include/W3DDevice/GameClient/W3DDisplay.h

@@ -45,6 +45,7 @@ class Render2DClass;
 class RTS3DScene;
 class RTS2DScene;
 class RTS3DInterfaceScene;
+class TextureClass;
 
 
 //=============================================================================
@@ -160,6 +161,10 @@ class W3DDisplay : public Display
 	void calculateTerrainLOD();						///< Calculate terrain LOD.
 	void renderLetterBox(UnsignedInt time);							///< draw letter box border
 	void updateAverageFPS();	///< calculate the average fps over the last 30 frames.
+	void setup2DRenderState(TextureClass *tex, DrawImageMode mode, Bool grayscale);
+	virtual void onBeginBatch() override;
+	virtual void onEndBatch() override;
+	virtual void onFlush() override;
 
 	Byte m_initialized;												///< TRUE when system is initialized
 	LightClass *m_myLight[LightEnvironmentClass::MAX_LIGHTS];										///< light hack for now
@@ -168,6 +173,12 @@ class W3DDisplay : public Display
 	Bool m_isClippedEnabled;	///<used by 2D drawing operations to define clip re
 	Real m_averageFPS;		///<average fps over the last 30 frames.
 	Real m_currentFPS;		///<current fps value.
+
+	TextureClass *m_batchTexture;
+	DrawImageMode m_batchMode;
+	Bool m_batchGrayscale;
+	Bool m_batchNeedsInit;
+
 #if defined(RTS_DEBUG)
 	Int64 m_timerAtCumuFPSStart;
 #endif