Ripped out SwiftConfig

src/Device/BUILD.gn

@@ -42,8 +42,6 @@ swiftshader_source_set("Device") {
     "Renderer.hpp",
     "SetupProcessor.cpp",
     "SetupProcessor.hpp",
-    "SwiftConfig.cpp",
-    "SwiftConfig.hpp",
     "Vector.cpp",
     "Vector.hpp",
     "VertexProcessor.cpp",

src/Device/Config.cpp

@@ -30,39 +30,10 @@ namespace sw
 		framesSec = 0;
 		framesTotal = 0;
 		FPS = 0;
-
-		#if PERF_PROFILE
-			for(int i = 0; i < PERF_TIMERS; i++)
-			{
-				cycles[i] = 0;
-			}
-
-			ropOperations = 0;
-			ropOperationsTotal = 0;
-			ropOperationsFrame = 0;
-
-			texOperations = 0;
-			texOperationsTotal = 0;
-			texOperationsFrame = 0;
-
-			compressedTex = 0;
-			compressedTexTotal = 0;
-			compressedTexFrame = 0;
-		#endif
 	}
 
 	void Profiler::nextFrame()
 	{
-		#if PERF_PROFILE
-			ropOperationsFrame = sw::atomicExchange(&ropOperations, 0);
-			texOperationsFrame = sw::atomicExchange(&texOperations, 0);
-			compressedTexFrame = sw::atomicExchange(&compressedTex, 0);
-
-			ropOperationsTotal += ropOperationsFrame;
-			texOperationsTotal += texOperationsFrame;
-			compressedTexTotal += compressedTexFrame;
-		#endif
-
 		static double fpsTime = sw::Timer::seconds();
 
 		double time = sw::Timer::seconds();

src/Device/Config.hpp

@@ -17,18 +17,6 @@
 
 #include "System/Types.hpp"
 
-#define PERF_HUD 0       // Display time spent on vertex, setup and pixel processing for each thread
-#define PERF_PROFILE 0   // Profile various pipeline stages and display the timing in SwiftConfig
-
-#define ASTC_SUPPORT 0
-
-// Worker thread count when not set by SwiftConfig
-// 0 = process affinity count (recommended)
-// 1 = rendering on main thread (no worker threads), useful for debugging
-#ifndef DEFAULT_THREAD_COUNT
-#define DEFAULT_THREAD_COUNT 0
-#endif
-
 namespace sw
 {
 	enum
@@ -53,22 +41,6 @@ namespace sw
 		int framesSec;
 		int framesTotal;
 		double FPS;
-
-		#if PERF_PROFILE
-		double cycles[PERF_TIMERS];
-
-		int64_t ropOperations;
-		int64_t ropOperationsTotal;
-		int64_t ropOperationsFrame;
-
-		int64_t texOperations;
-		int64_t texOperationsTotal;
-		int64_t texOperationsFrame;
-
-		int64_t compressedTex;
-		int64_t compressedTexTotal;
-		int64_t compressedTexFrame;
-		#endif
 	};
 
 	extern Profiler profiler;

src/Device/Context.cpp

@@ -24,16 +24,6 @@
 
 namespace sw
 {
-	bool booleanFaceRegister = false;
-	bool fullPixelPositionRegister = false;
-
-	bool forceWindowed = false;
-	bool quadLayoutEnabled = false;
-	bool postBlendSRGB = false;
-	bool exactColorRounding = false;
-	TransparencyAntialiasing transparencyAntialiasing = TRANSPARENCY_NONE;
-	bool forceClearRegisters = false;
-
 	Context::Context()
 	{
 		init();
@@ -197,11 +187,6 @@ namespace sw
 		return depthWriteEnable;
 	}
 
-	bool Context::alphaTestActive() const
-	{
-		return transparencyAntialiasing != TRANSPARENCY_NONE;
-	}
-
 	bool Context::depthBufferActive() const
 	{
 		return depthBuffer && depthBufferEnable;
@@ -589,6 +574,6 @@ namespace sw
 
 	bool Context::colorUsed() const
 	{
-		return colorWriteActive() || alphaTestActive() || (pixelShader && pixelShader->getModes().ContainsKill);
+		return colorWriteActive() || (pixelShader && pixelShader->getModes().ContainsKill);
 	}
 }

src/Device/Context.hpp

@@ -97,7 +97,6 @@ namespace sw
 		bool setWriteSRGB(bool sRGB);
 
 		bool depthWriteActive() const;
-		bool alphaTestActive() const;
 		bool depthBufferActive() const;
 		bool stencilActive() const;
 

src/Device/PixelProcessor.cpp

@@ -24,10 +24,6 @@
 
 namespace sw
 {
-	extern TransparencyAntialiasing transparencyAntialiasing;
-
-	bool precachePixel = false;
-
 	unsigned int PixelProcessor::States::computeHash()
 	{
 		unsigned int *state = (unsigned int*)this;

src/Device/QuadRasterizer.cpp

@@ -22,10 +22,6 @@
 
 namespace sw
 {
-	extern bool fullPixelPositionRegister;
-
-	extern int clusterCount;
-
 	QuadRasterizer::QuadRasterizer(const PixelProcessor::State &state, SpirvShader const *spirvShader) : state(state), spirvShader{spirvShader}
 	{
 	}
@@ -36,15 +32,6 @@ namespace sw
 
 	void QuadRasterizer::generate()
 	{
-		#if PERF_PROFILE
-			for(int i = 0; i < PERF_TIMERS; i++)
-			{
-				cycles[i] = 0;
-			}
-
-			Long pixelTime = Ticks();
-		#endif
-
 		constants = *Pointer<Pointer<Byte>>(data + OFFSET(DrawData,constants));
 		occlusion = 0;
 		int clusterCount = Renderer::getClusterCount();
@@ -76,15 +63,6 @@ namespace sw
 			*Pointer<UInt>(data + OFFSET(DrawData,occlusion) + 4 * cluster) = clusterOcclusion;
 		}
 
-		#if PERF_PROFILE
-			cycles[PERF_PIXEL] = Ticks() - pixelTime;
-
-			for(int i = 0; i < PERF_TIMERS; i++)
-			{
-				*Pointer<Long>(data + OFFSET(DrawData,cycles[i]) + 8 * cluster) += cycles[i];
-			}
-		#endif
-
 		Return();
 	}
 

src/Device/QuadRasterizer.hpp

@@ -40,10 +40,6 @@ namespace sw
 
 		UInt occlusion;
 
-#if PERF_PROFILE
-		Long cycles[PERF_TIMERS];
-#endif
-
 		virtual void quad(Pointer<Byte> cBuffer[4], Pointer<Byte> &zBuffer, Pointer<Byte> &sBuffer, Int cMask[4], Int &x, Int &y) = 0;
 
 		bool interpolateZ() const;

src/Device/Renderer.cpp

@@ -17,7 +17,6 @@
 #include "Clipper.hpp"
 #include "Primitive.hpp"
 #include "Polygon.hpp"
-#include "Device/SwiftConfig.hpp"
 #include "Reactor/Reactor.hpp"
 #include "Pipeline/Constants.hpp"
 #include "System/CPUID.hpp"
@@ -35,8 +34,6 @@
 
 #undef max
 
-bool disableServer = true;
-
 #ifndef NDEBUG
 unsigned int minPrimitives = 1;
 unsigned int maxPrimitives = 1 << 21;
@@ -44,42 +41,11 @@ unsigned int maxPrimitives = 1 << 21;
 
 namespace sw
 {
-	extern bool booleanFaceRegister;
-	extern bool fullPixelPositionRegister;
-
-	extern bool forceWindowed;
-	extern bool postBlendSRGB;
-	extern bool exactColorRounding;
-	extern TransparencyAntialiasing transparencyAntialiasing;
-	extern bool forceClearRegisters;
-
-	extern bool precacheVertex;
-	extern bool precacheSetup;
-	extern bool precachePixel;
-
 	static const int batchSize = 128;
 	AtomicInt threadCount(1);
 	AtomicInt Renderer::unitCount(1);
 	AtomicInt Renderer::clusterCount(1);
 
-	TranscendentalPrecision logPrecision = ACCURATE;
-	TranscendentalPrecision expPrecision = ACCURATE;
-	TranscendentalPrecision rcpPrecision = ACCURATE;
-	TranscendentalPrecision rsqPrecision = ACCURATE;
-
-	static void setGlobalRenderingSettings(Conventions conventions, bool exactColorRounding)
-	{
-		static bool initialized = false;
-
-		if(!initialized)
-		{
-			sw::booleanFaceRegister = conventions.booleanFaceRegister;
-			sw::fullPixelPositionRegister = conventions.fullPixelPositionRegister;
-			sw::exactColorRounding = exactColorRounding;
-			initialized = true;
-		}
-	}
-
 	template<typename T>
 	inline bool setBatchIndices(unsigned int batch[128][3], VkPrimitiveTopology topology, T indices, unsigned int start, unsigned int triangleCount)
 	{
@@ -196,14 +162,8 @@ namespace sw
 		deallocate(data);
 	}
 
-	Renderer::Renderer(Conventions conventions, bool exactColorRounding)
+	Renderer::Renderer()
 	{
-		setGlobalRenderingSettings(conventions, exactColorRounding);
-
-		#if PERF_HUD
-			resetTimers();
-		#endif
-
 		for(int i = 0; i < 16; i++)
 		{
 			vertexTask[i] = nullptr;
@@ -246,7 +206,6 @@ namespace sw
 
 		clipFlags = 0;
 
-		swiftConfig = new SwiftConfig(disableServer);
 		updateConfiguration(true);
 	}
 
@@ -263,9 +222,6 @@ namespace sw
 			delete drawCall[draw];
 			drawCall[draw] = nullptr;
 		}
-
-		delete swiftConfig;
-		swiftConfig = nullptr;
 	}
 
 	// This object has to be mem aligned
@@ -298,10 +254,14 @@ namespace sw
 		if(count == 0) { return; }
 
 		#ifndef NDEBUG
+		{
+			unsigned int minPrimitives = 1;
+			unsigned int maxPrimitives = 1 << 21;
 			if(count < minPrimitives || count > maxPrimitives)
 			{
 				return;
 			}
+		}
 		#endif
 
 		updateConfiguration();
@@ -455,16 +415,6 @@ namespace sw
 			}
 		}
 
-		#if PERF_PROFILE
-			for(int cluster = 0; cluster < clusterCount; cluster++)
-			{
-				for(int i = 0; i < PERF_TIMERS; i++)
-				{
-					data->cycles[i][cluster] = 0;
-				}
-			}
-		#endif
-
 		// Viewport
 		{
 			float W = 0.5f * viewport.width;
@@ -574,11 +524,8 @@ namespace sw
 		Renderer *renderer = static_cast<Parameters*>(parameters)->renderer;
 		int threadIndex = static_cast<Parameters*>(parameters)->threadIndex;
 
-		if(logPrecision < IEEE)
-		{
-			CPUID::setFlushToZero(true);
-			CPUID::setDenormalsAreZero(true);
-		}
+		CPUID::setFlushToZero(true);
+		CPUID::setDenormalsAreZero(true);
 
 		renderer->threadLoop(threadIndex);
 	}
@@ -733,10 +680,6 @@ namespace sw
 
 	void Renderer::executeTask(int threadIndex)
 	{
-		#if PERF_HUD
-			int64_t startTick = Timer::ticks();
-		#endif
-
 		switch(task[threadIndex].type)
 		{
 		case Task::PRIMITIVES:
@@ -750,12 +693,6 @@ namespace sw
 
 				processPrimitiveVertices(unit, input, count, draw->count, threadIndex);
 
-				#if PERF_HUD
-					int64_t time = Timer::ticks();
-					vertexTime[threadIndex] += time - startTick;
-					startTick = time;
-				#endif
-
 				int visible = 0;
 
 				if(!draw->setupState.rasterizerDiscard)
@@ -765,10 +702,6 @@ namespace sw
 
 				primitiveProgress[unit].visible = visible;
 				primitiveProgress[unit].references = clusterCount;
-
-				#if PERF_HUD
-					setupTime[threadIndex] += Timer::ticks() - startTick;
-				#endif
 			}
 			break;
 		case Task::PIXELS:
@@ -788,10 +721,6 @@ namespace sw
 				}
 
 				finishRendering(task[threadIndex]);
-
-				#if PERF_HUD
-					pixelTime[threadIndex] += Timer::ticks() - startTick;
-				#endif
 			}
 			break;
 		case Task::RESUME:
@@ -835,16 +764,6 @@ namespace sw
 
 			if(ref == 0)
 			{
-				#if PERF_PROFILE
-					for(int cluster = 0; cluster < clusterCount; cluster++)
-					{
-						for(int i = 0; i < PERF_TIMERS; i++)
-						{
-							profiler.cycles[i] += data.cycles[i][cluster];
-						}
-					}
-				#endif
-
 				if(draw.queries)
 				{
 					for(auto &query : *(draw.queries))
@@ -1393,38 +1312,6 @@ namespace sw
 		}
 	}
 
-	#if PERF_HUD
-		int Renderer::getThreadCount()
-		{
-			return threadCount;
-		}
-
-		int64_t Renderer::getVertexTime(int thread)
-		{
-			return vertexTime[thread];
-		}
-
-		int64_t Renderer::getSetupTime(int thread)
-		{
-			return setupTime[thread];
-		}
-
-		int64_t Renderer::getPixelTime(int thread)
-		{
-			return pixelTime[thread];
-		}
-
-		void Renderer::resetTimers()
-		{
-			for(int thread = 0; thread < threadCount; thread++)
-			{
-				vertexTime[thread] = 0;
-				setupTime[thread] = 0;
-				pixelTime[thread] = 0;
-			}
-		}
-	#endif
-
 	void Renderer::setViewport(const VkViewport &viewport)
 	{
 		this->viewport = viewport;
@@ -1437,97 +1324,21 @@ namespace sw
 
 	void Renderer::updateConfiguration(bool initialUpdate)
 	{
-		bool newConfiguration = swiftConfig->hasNewConfiguration();
-
-		if(newConfiguration || initialUpdate)
+		if(initialUpdate)
 		{
 			terminateThreads();
 
-			SwiftConfig::Configuration configuration = {};
-			swiftConfig->getConfiguration(configuration);
+			VertexProcessor::setRoutineCacheSize(1024);
+			PixelProcessor::setRoutineCacheSize(1024);
+			SetupProcessor::setRoutineCacheSize(1024);
 
-			precacheVertex = !newConfiguration && configuration.precache;
-			precacheSetup = !newConfiguration && configuration.precache;
-			precachePixel = !newConfiguration && configuration.precache;
+			threadCount = CPUID::processAffinity();
 
-			VertexProcessor::setRoutineCacheSize(configuration.vertexRoutineCacheSize);
-			PixelProcessor::setRoutineCacheSize(configuration.pixelRoutineCacheSize);
-			SetupProcessor::setRoutineCacheSize(configuration.setupRoutineCacheSize);
-
-			switch(configuration.transcendentalPrecision)
-			{
-			case 0:
-				logPrecision = APPROXIMATE;
-				expPrecision = APPROXIMATE;
-				rcpPrecision = APPROXIMATE;
-				rsqPrecision = APPROXIMATE;
-				break;
-			case 1:
-				logPrecision = PARTIAL;
-				expPrecision = PARTIAL;
-				rcpPrecision = PARTIAL;
-				rsqPrecision = PARTIAL;
-				break;
-			case 2:
-				logPrecision = ACCURATE;
-				expPrecision = ACCURATE;
-				rcpPrecision = ACCURATE;
-				rsqPrecision = ACCURATE;
-				break;
-			case 3:
-				logPrecision = WHQL;
-				expPrecision = WHQL;
-				rcpPrecision = WHQL;
-				rsqPrecision = WHQL;
-				break;
-			case 4:
-				logPrecision = IEEE;
-				expPrecision = IEEE;
-				rcpPrecision = IEEE;
-				rsqPrecision = IEEE;
-				break;
-			default:
-				logPrecision = ACCURATE;
-				expPrecision = ACCURATE;
-				rcpPrecision = ACCURATE;
-				rsqPrecision = ACCURATE;
-				break;
-			}
-
-			switch(configuration.transparencyAntialiasing)
-			{
-			case 0:  transparencyAntialiasing = TRANSPARENCY_NONE;              break;
-			case 1:  transparencyAntialiasing = TRANSPARENCY_ALPHA_TO_COVERAGE; break;
-			default: transparencyAntialiasing = TRANSPARENCY_NONE;              break;
-			}
-
-			switch(configuration.threadCount)
-			{
-			case -1: threadCount = CPUID::coreCount();        break;
-			case 0:  threadCount = CPUID::processAffinity();  break;
-			default: threadCount = configuration.threadCount; break;
-			}
-
-			CPUID::setEnableSSE4_1(configuration.enableSSE4_1);
-			CPUID::setEnableSSSE3(configuration.enableSSSE3);
-			CPUID::setEnableSSE3(configuration.enableSSE3);
-			CPUID::setEnableSSE2(configuration.enableSSE2);
-			CPUID::setEnableSSE(configuration.enableSSE);
-
-			for(int pass = 0; pass < 10; pass++)
-			{
-				optimization[pass] = configuration.optimization[pass];
-			}
-
-			forceWindowed = configuration.forceWindowed;
-			postBlendSRGB = configuration.postBlendSRGB;
-			exactColorRounding = configuration.exactColorRounding;
-			forceClearRegisters = configuration.forceClearRegisters;
-
-		#ifndef NDEBUG
-			minPrimitives = configuration.minPrimitives;
-			maxPrimitives = configuration.maxPrimitives;
-		#endif
+			CPUID::setEnableSSE4_1(true);
+			CPUID::setEnableSSSE3(true);
+			CPUID::setEnableSSE3(true);
+			CPUID::setEnableSSE2(true);
+			CPUID::setEnableSSE(true);
 		}
 
 		if(!initialUpdate && !worker[0])

src/Device/Renderer.hpp

@@ -40,50 +40,11 @@ namespace sw
 	struct DrawCall;
 	class PixelShader;
 	class VertexShader;
-	class SwiftConfig;
 	struct Task;
 	class TaskEvents;
 	class Resource;
 	struct Constants;
 
-	enum TranscendentalPrecision
-	{
-		APPROXIMATE,
-		PARTIAL,	// 2^-10
-		ACCURATE,
-		WHQL,		// 2^-21
-		IEEE		// 2^-23
-	};
-
-	extern TranscendentalPrecision logPrecision;
-	extern TranscendentalPrecision expPrecision;
-	extern TranscendentalPrecision rcpPrecision;
-	extern TranscendentalPrecision rsqPrecision;
-
-	struct Conventions  // FIXME(capn): Eliminate. Only support Vulkan 1.1 conventions.
-	{
-		bool halfIntegerCoordinates;
-		bool symmetricNormalizedDepth;
-		bool booleanFaceRegister;
-		bool fullPixelPositionRegister;
-	};
-
-	static const Conventions OpenGL =
-	{
-		true,    // halfIntegerCoordinates
-		true,    // symmetricNormalizedDepth
-		true,    // booleanFaceRegister
-		true,    // fullPixelPositionRegister
-	};
-
-	static const Conventions Direct3D =
-	{
-		false,   // halfIntegerCoordinates
-		false,   // symmetricNormalizedDepth
-		false,   // booleanFaceRegister
-		false,   // fullPixelPositionRegister
-	};
-
 	struct DrawData
 	{
 		const Constants *constants;
@@ -103,10 +64,6 @@ namespace sw
 		PixelProcessor::Factor factor;
 		unsigned int occlusion[16];   // Number of pixels passing depth test
 
-		#if PERF_PROFILE
-			int64_t cycles[PERF_TIMERS][16];
-		#endif
-
 		float4 Wx16;
 		float4 Hx16;
 		float4 X0x16;
@@ -193,7 +150,7 @@ namespace sw
 		};
 
 	public:
-		Renderer(Conventions conventions, bool exactColorRounding);
+		Renderer();
 
 		virtual ~Renderer();
 
@@ -215,15 +172,6 @@ namespace sw
 
 		void synchronize();
 
-		#if PERF_HUD
-			// Performance timers
-			int getThreadCount();
-			int64_t getVertexTime(int thread);
-			int64_t getSetupTime(int thread);
-			int64_t getPixelTime(int thread);
-			void resetTimers();
-		#endif
-
 		static int getClusterCount() { return clusterCount; }
 
 	private:
@@ -289,16 +237,8 @@ namespace sw
 
 		std::mutex schedulerMutex;
 
-		#if PERF_HUD
-			int64_t vertexTime[16];
-			int64_t setupTime[16];
-			int64_t pixelTime[16];
-		#endif
-
 		VertexTask *vertexTask[16];
 
-		SwiftConfig *swiftConfig;
-
 		std::list<vk::Query*> queries;
 		WaitGroup sync;
 

src/Device/Sampler.hpp

@@ -157,10 +157,6 @@ namespace sw
 		VkSamplerYcbcrModelConversion ycbcrModel;
 		bool studioSwing;    // Narrow range
 		bool swappedChroma;  // Cb/Cr components in reverse order
-
-		#if PERF_PROFILE
-		bool compressedFormat;
-		#endif
 	};
 }
 

src/Device/SetupProcessor.cpp

@@ -25,10 +25,6 @@
 
 namespace sw
 {
-	extern bool fullPixelPositionRegister;
-
-	bool precacheSetup = false;
-
 	unsigned int SetupProcessor::States::computeHash()
 	{
 		unsigned int *state = (unsigned int*)this;

src/Device/SwiftConfig.hpp

-// Copyright 2016 The SwiftShader Authors. All Rights Reserved.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//    http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#ifndef sw_SwiftConfig_hpp
-#define sw_SwiftConfig_hpp
-
-#include "Reactor/Nucleus.hpp"
-
-#include "System/Socket.hpp"
-
-#include <mutex>
-#include <string>
-#include <thread>
-
-#ifdef Status
-#undef Status // b/127920555
-#endif
-
-namespace sw
-{
-	class SwiftConfig
-	{
-	public:
-		struct Configuration
-		{
-			int pixelShaderVersion;
-			int vertexShaderVersion;
-			int textureMemory;
-			int identifier;
-			int vertexRoutineCacheSize;
-			int pixelRoutineCacheSize;
-			int setupRoutineCacheSize;
-			int vertexCacheSize;
-			int textureSampleQuality;
-			int mipmapQuality;
-			int transcendentalPrecision;
-			int threadCount;
-			bool enableSSE;
-			bool enableSSE2;
-			bool enableSSE3;
-			bool enableSSSE3;
-			bool enableSSE4_1;
-			rr::Optimization optimization[10];
-			bool disableServer;
-			bool keepSystemCursor;
-			bool forceWindowed;
-			bool postBlendSRGB;
-			bool exactColorRounding;
-			bool disableAlphaMode;
-			bool disable10BitMode;
-			int transparencyAntialiasing;
-			int frameBufferAPI;
-			bool precache;
-			int shadowMapping;
-			bool forceClearRegisters;
-		#ifndef NDEBUG
-			unsigned int minPrimitives;
-			unsigned int maxPrimitives;
-		#endif
-		};
-
-		SwiftConfig(bool disableServerOverride);
-
-		~SwiftConfig();
-
-		bool hasNewConfiguration(bool reset = true);
-		void getConfiguration(Configuration &configuration);
-
-	private:
-		enum Status
-		{
-			OK = 200,
-			NotFound = 404
-		};
-
-		void createServer();
-		void destroyServer();
-
-		static void serverRoutine(void *parameters);
-
-		void serverLoop();
-		void respond(Socket *clientSocket, const char *request);
-		std::string page();
-		std::string profile();
-		void send(Socket *clientSocket, Status code, std::string body = "");
-		void parsePost(const char *post);
-
-		void readConfiguration(bool disableServerOverride = false);
-		void writeConfiguration();
-
-		Configuration config;
-
-		std::thread *serverThread;
-		volatile bool terminate;
-		std::mutex criticalSection;   // Protects reading and writing the configuration settings
-
-		bool newConfig;
-
-		Socket *listenSocket;
-
-		int bufferLength;
-		char *receiveBuffer;
-	};
-}
-
-#endif   // sw_SwiftConfig_hpp

src/Device/VertexProcessor.cpp

@@ -23,8 +23,6 @@
 
 namespace sw
 {
-	bool precacheVertex = false;
-
 	void VertexCache::clear()
 	{
 		for(int i = 0; i < 16; i++)

src/Pipeline/PixelProgram.cpp

@@ -20,8 +20,6 @@
 
 namespace sw
 {
-	extern bool postBlendSRGB;
-
 	void PixelProgram::setBuiltins(Int &x, Int &y, Float4(&z)[4], Float4 &w)
 	{
 		routine.windowSpacePosition[0] = x + SIMD::Int(0,1,0,1);

src/Pipeline/PixelRoutine.cpp

@@ -24,10 +24,6 @@
 
 namespace sw
 {
-	extern bool postBlendSRGB;
-	extern bool exactColorRounding;
-	extern bool forceClearRegisters;
-
 	PixelRoutine::PixelRoutine(
 			const PixelProcessor::State &state,
 			vk::PipelineLayout const *pipelineLayout,
@@ -40,12 +36,12 @@ namespace sw
 		if (spirvShader)
 		{
 			spirvShader->emitProlog(&routine);
-			if (forceClearRegisters)
+
+			// Clearing inputs to 0 is not demanded by the spec,
+			// but it makes the undefined behavior deterministic.
+			for(int i = 0; i < MAX_INTERFACE_COMPONENTS; i++)
 			{
-				for (int i = 0; i < MAX_INTERFACE_COMPONENTS; i++)
-				{
-					routine.inputs[i] = Float4(0.0f);
-				}
+				routine.inputs[i] = Float4(0.0f);
 			}
 		}
 	}
@@ -56,10 +52,6 @@ namespace sw
 
 	void PixelRoutine::quad(Pointer<Byte> cBuffer[RENDERTARGETS], Pointer<Byte> &zBuffer, Pointer<Byte> &sBuffer, Int cMask[4], Int &x, Int &y)
 	{
-		#if PERF_PROFILE
-			Long pipeTime = Ticks();
-		#endif
-
 		// TODO: consider shader which modifies sample mask in general
 		const bool earlyDepthTest = !spirvShader || (!spirvShader->getModes().DepthReplacing && !state.alphaToCoverage);
 
@@ -109,10 +101,6 @@ namespace sw
 
 		If(depthPass || Bool(!earlyDepthTest))
 		{
-			#if PERF_PROFILE
-				Long interpTime = Ticks();
-			#endif
-
 			Float4 yyyy = Float4(Float(y)) + *Pointer<Float4>(primitive + OFFSET(Primitive,yQuad), 16);
 
 			// Centroid locations
@@ -176,25 +164,13 @@ namespace sw
 				setBuiltins(x, y, z, w);
 			}
 
-			#if PERF_PROFILE
-				cycles[PERF_INTERP] += Ticks() - interpTime;
-			#endif
-
 			Bool alphaPass = true;
 
-			#if PERF_PROFILE
-				Long shaderTime = Ticks();
-			#endif
-
 			if (spirvShader)
 			{
 				applyShader(cMask);
 			}
 
-			#if PERF_PROFILE
-				cycles[PERF_SHADER] += Ticks() - shaderTime;
-			#endif
-
 			alphaPass = alphaTest(cMask);
 
 			if((spirvShader && spirvShader->getModes().ContainsKill) || state.alphaToCoverage)
@@ -216,10 +192,6 @@ namespace sw
 					}
 				}
 
-				#if PERF_PROFILE
-					Long ropTime = Ticks();
-				#endif
-
 				If(depthPass || Bool(earlyDepthTest))
 				{
 					for(unsigned int q = 0; q < state.multiSample; q++)
@@ -235,16 +207,8 @@ namespace sw
 						}
 					}
 
-					#if PERF_PROFILE
-						AddAtomic(Pointer<Long>(&profiler.ropOperations), 4);
-					#endif
-
 					rasterOperation(cBuffer, x, sMask, zMask, cMask);
 				}
-
-				#if PERF_PROFILE
-					cycles[PERF_ROP] += Ticks() - ropTime;
-				#endif
 			}
 		}
 
@@ -255,10 +219,6 @@ namespace sw
 				writeStencil(sBuffer, q, x, sMask[q], zMask[q], cMask[q]);
 			}
 		}
-
-		#if PERF_PROFILE
-			cycles[PERF_PIPE] += Ticks() - pipeTime;
-		#endif
 	}
 
 	Float4 PixelRoutine::interpolateCentroid(Float4 &x, Float4 &y, Float4 &rhw, Pointer<Byte> planeEquation, bool flat, bool perspective)
@@ -1102,7 +1062,7 @@ namespace sw
 			UNIMPLEMENTED("VkFormat %d", state.targetFormat[index]);
 		}
 
-		if(postBlendSRGB || isSRGB(index))
+		if(isSRGB(index))
 		{
 			sRGBtoLinear16_12_16(pixel);
 		}
@@ -1229,36 +1189,33 @@ namespace sw
 
 	void PixelRoutine::writeColor(int index, Pointer<Byte> &cBuffer, Int &x, Vector4s &current, Int &sMask, Int &zMask, Int &cMask)
 	{
-		if(postBlendSRGB || isSRGB(index))
+		if(isSRGB(index))
 		{
 			linearToSRGB16_12_16(current);
 		}
 
-		if(exactColorRounding)
+		switch(state.targetFormat[index])
 		{
-			switch(state.targetFormat[index])
-			{
-			case VK_FORMAT_R5G6B5_UNORM_PACK16:
-				current.x = AddSat(As<UShort4>(current.x), UShort4(0x0400));
-				current.y = AddSat(As<UShort4>(current.y), UShort4(0x0200));
-				current.z = AddSat(As<UShort4>(current.z), UShort4(0x0400));
-				break;
-			case VK_FORMAT_B8G8R8A8_UNORM:
-			case VK_FORMAT_B8G8R8A8_SRGB:
-			case VK_FORMAT_R8G8B8A8_UNORM:
-			case VK_FORMAT_R8G8B8A8_SRGB:
-			case VK_FORMAT_R8G8_UNORM:
-			case VK_FORMAT_R8_UNORM:
-			case VK_FORMAT_A8B8G8R8_UNORM_PACK32:
-			case VK_FORMAT_A8B8G8R8_SRGB_PACK32:
-				current.x = current.x - As<Short4>(As<UShort4>(current.x) >> 8) + Short4(0x0080);
-				current.y = current.y - As<Short4>(As<UShort4>(current.y) >> 8) + Short4(0x0080);
-				current.z = current.z - As<Short4>(As<UShort4>(current.z) >> 8) + Short4(0x0080);
-				current.w = current.w - As<Short4>(As<UShort4>(current.w) >> 8) + Short4(0x0080);
-				break;
-			default:
-				break;
-			}
+		case VK_FORMAT_R5G6B5_UNORM_PACK16:
+			current.x = AddSat(As<UShort4>(current.x), UShort4(0x0400));
+			current.y = AddSat(As<UShort4>(current.y), UShort4(0x0200));
+			current.z = AddSat(As<UShort4>(current.z), UShort4(0x0400));
+			break;
+		case VK_FORMAT_B8G8R8A8_UNORM:
+		case VK_FORMAT_B8G8R8A8_SRGB:
+		case VK_FORMAT_R8G8B8A8_UNORM:
+		case VK_FORMAT_R8G8B8A8_SRGB:
+		case VK_FORMAT_R8G8_UNORM:
+		case VK_FORMAT_R8_UNORM:
+		case VK_FORMAT_A8B8G8R8_UNORM_PACK32:
+		case VK_FORMAT_A8B8G8R8_SRGB_PACK32:
+			current.x = current.x - As<Short4>(As<UShort4>(current.x) >> 8) + Short4(0x0080);
+			current.y = current.y - As<Short4>(As<UShort4>(current.y) >> 8) + Short4(0x0080);
+			current.z = current.z - As<Short4>(As<UShort4>(current.z) >> 8) + Short4(0x0080);
+			current.w = current.w - As<Short4>(As<UShort4>(current.w) >> 8) + Short4(0x0080);
+			break;
+		default:
+			break;
 		}
 
 		int rgbaWriteMask = state.colorWriteActive(index);
@@ -1956,7 +1913,7 @@ namespace sw
 			UNIMPLEMENTED("VkFormat: %d", int(state.targetFormat[index]));
 		}
 
-		if(postBlendSRGB || isSRGB(index))
+		if(isSRGB(index))
 		{
 			sRGBtoLinear(pixel.x);
 			sRGBtoLinear(pixel.y);

src/Pipeline/SamplerCore.cpp

@@ -57,15 +57,6 @@ namespace sw
 	{
 		Vector4f c;
 
-		#if PERF_PROFILE
-			AddAtomic(Pointer<Long>(&profiler.texOperations), 4);
-
-			if(state.compressedFormat)
-			{
-				AddAtomic(Pointer<Long>(&profiler.compressedTex), 4);
-			}
-		#endif
-
 		Float4 uuuu = u;
 		Float4 vvvv = v;
 		Float4 wwww = w;

src/Pipeline/SetupRoutine.cpp

@@ -23,8 +23,6 @@
 
 namespace sw
 {
-	extern TranscendentalPrecision logPrecision;
-
 	SetupRoutine::SetupRoutine(const SetupProcessor::State &state) : state(state)
 	{
 		routine = 0;

src/Pipeline/ShaderCore.cpp

@@ -21,11 +21,6 @@
 
 namespace sw
 {
-	extern TranscendentalPrecision logPrecision;
-	extern TranscendentalPrecision expPrecision;
-	extern TranscendentalPrecision rcpPrecision;
-	extern TranscendentalPrecision rsqPrecision;
-
 	Vector4s::Vector4s()
 	{
 	}
@@ -186,20 +181,11 @@ namespace sw
 
 	Float4 reciprocal(RValue<Float4> x, bool pp, bool finite, bool exactAtPow2)
 	{
-		Float4 rcp;
+		Float4 rcp = Rcp_pp(x, exactAtPow2);
 
-		if(!pp && rcpPrecision >= WHQL)
-		{
-			rcp = Float4(1.0f) / x;
-		}
-		else
+		if(!pp)
 		{
-			rcp = Rcp_pp(x, exactAtPow2);
-
-			if(!pp)
-			{
-				rcp = (rcp + rcp) - (x * rcp * rcp);
-			}
+			rcp = (rcp + rcp) - (x * rcp * rcp);
 		}
 
 		if(finite)

src/Vulkan/VkQueue.cpp

@@ -74,7 +74,7 @@ VkSubmitInfo* DeepCopySubmitInfo(uint32_t submitCount, const VkSubmitInfo* pSubm
 namespace vk
 {
 
-Queue::Queue() : renderer(sw::OpenGL, true)
+Queue::Queue() : renderer()
 {
 	queueThread = std::thread(TaskLoop, this);
 }

src/Vulkan/vulkan.vcxproj

@@ -149,7 +149,6 @@ IF EXIST "$(SolutionDir)..\deqp\build\external\vulkancts\modules\vulkan\" (copy
     <ClCompile Include="..\Device\QuadRasterizer.cpp" />
     <ClCompile Include="..\Device\Renderer.cpp" />
     <ClCompile Include="..\Device\SetupProcessor.cpp" />
-    <ClCompile Include="..\Device\SwiftConfig.cpp" />
     <ClCompile Include="..\Device\Vector.cpp" />
     <ClCompile Include="..\Device\VertexProcessor.cpp" />
     <ClCompile Include="..\Pipeline\ComputeProgram.cpp" />
@@ -243,7 +242,6 @@ IF EXIST "$(SolutionDir)..\deqp\build\external\vulkancts\modules\vulkan\" (copy
     <ClInclude Include="..\Device\Sampler.hpp" />
     <ClInclude Include="..\Device\SetupProcessor.hpp" />
     <ClInclude Include="..\Device\Stream.hpp" />
-    <ClInclude Include="..\Device\SwiftConfig.hpp" />
     <ClInclude Include="..\Device\Triangle.hpp" />
     <ClInclude Include="..\Device\Vector.hpp" />
     <ClInclude Include="..\Device\Vertex.hpp" />

src/Vulkan/vulkan.vcxproj.filters

@@ -51,9 +51,6 @@
     <ClCompile Include="..\Device\Vector.cpp">
       <Filter>Source Files\Device</Filter>
     </ClCompile>
-    <ClCompile Include="..\Device\SwiftConfig.cpp">
-      <Filter>Source Files\Device</Filter>
-    </ClCompile>
     <ClCompile Include="..\Device\SetupProcessor.cpp">
       <Filter>Source Files\Device</Filter>
     </ClCompile>
@@ -362,9 +359,6 @@
     <ClInclude Include="..\Device\Triangle.hpp">
       <Filter>Header Files\Device</Filter>
     </ClInclude>
-    <ClInclude Include="..\Device\SwiftConfig.hpp">
-      <Filter>Header Files\Device</Filter>
-    </ClInclude>
     <ClInclude Include="..\Device\Stream.hpp">
       <Filter>Header Files\Device</Filter>
     </ClInclude>