Merge pull request #6319 from unknownbrackets/debugger

Skip stepping time in debug stats

Core/HLE/HLE.cpp

@@ -510,6 +510,12 @@ void *GetQuickSyscallFunc(MIPSOpcode op)
 	return (void *)&CallSyscallWithoutFlags;
 }
 
+static double hleSteppingTime = 0.0;
+void hleSetSteppingTime(double t)
+{
+	hleSteppingTime += t;
+}
+
 void CallSyscall(MIPSOpcode op)
 {
 	double start = 0.0;  // need to initialize to fix the race condition where g_Config.bShowDebugStats is enabled in the middle of this func.
@@ -540,6 +546,8 @@ void CallSyscall(MIPSOpcode op)
 		u32 callno = (op >> 6) & 0xFFFFF; //20 bits
 		int funcnum = callno & 0xFFF;
 		int modulenum = (callno & 0xFF000) >> 12;
-		updateSyscallStats(modulenum, funcnum, time_now_d() - start);
+		double total = time_now_d() - start - hleSteppingTime;
+		hleSteppingTime = 0.0;
+		updateSyscallStats(modulenum, funcnum, total);
 	}
 }

Core/HLE/HLE.h

@@ -91,6 +91,8 @@ void hleRunInterrupts();
 void hleDebugBreak();
 // Don't set temp regs to 0xDEADBEEF.
 void hleSkipDeadbeef();
+// Set time spent in debugger (for more useful debug stats while debugging.)
+void hleSetSteppingTime(double t);
 
 // Delays the result for usec microseconds, allowing other threads to run during this time.
 u32 hleDelayResult(u32 result, const char *reason, int usec);

GPU/Common/GPUDebugInterface.h

@@ -200,6 +200,10 @@ public:
 	virtual GPUDebugOp DissassembleOp(u32 pc, u32 op) = 0;
 	virtual std::vector<GPUDebugOp> DissassembleOpRange(u32 startpc, u32 endpc) = 0;
 
+	// Enter/exit stepping mode.  Mainly for better debug stats on time taken.
+	virtual void NotifySteppingEnter() = 0;
+	virtual void NotifySteppingExit() = 0;
+
 	virtual u32 GetRelativeAddress(u32 data) = 0;
 	virtual u32 GetVertexAddress() = 0;
 	virtual u32 GetIndexAddress() = 0;

GPU/Debugger/Stepping.cpp

@@ -118,6 +118,8 @@ bool EnterStepping(std::function<void()> callback) {
 		return false;
 	}
 
+	gpuDebug->NotifySteppingEnter();
+
 	// Just to be sure.
 	if (pauseAction == PAUSE_CONTINUE) {
 		pauseAction = PAUSE_BREAK;
@@ -131,6 +133,7 @@ bool EnterStepping(std::function<void()> callback) {
 		pauseWait.wait(pauseLock);
 	} while (pauseAction != PAUSE_CONTINUE);
 
+	gpuDebug->NotifySteppingExit();
 	isStepping = false;
 	return true;
 }

GPU/GPUCommon.cpp

@@ -10,6 +10,7 @@
 #include "Core/MemMap.h"
 #include "Core/Host.h"
 #include "Core/Reporting.h"
+#include "Core/HLE/HLE.h"
 #include "Core/HLE/sceKernelMemory.h"
 #include "Core/HLE/sceKernelInterrupt.h"
 #include "Core/HLE/sceKernelThread.h"
@@ -36,6 +37,7 @@ void GPUCommon::Reinitialize() {
 	isbreak = false;
 	drawCompleteTicks = 0;
 	busyTicks = 0;
+	timeSpentStepping_ = 0.0;
 	interruptsEnabled_ = true;
 	UpdateTickEstimate(0);
 }
@@ -446,6 +448,23 @@ u32 GPUCommon::Break(int mode) {
 	return currentList->id;
 }
 
+void GPUCommon::NotifySteppingEnter() {
+	if (g_Config.bShowDebugStats) {
+		time_update();
+		timeSteppingStarted_ = time_now_d();
+	}
+}
+void GPUCommon::NotifySteppingExit() {
+	if (g_Config.bShowDebugStats) {
+		if (timeSteppingStarted_ <= 0.0) {
+			ERROR_LOG(G3D, "Mismatched stepping enter/exit.");
+		}
+		time_update();
+		timeSpentStepping_ += time_now_d() - timeSteppingStarted_;
+		timeSteppingStarted_ = 0.0;
+	}
+}
+
 bool GPUCommon::InterpretList(DisplayList &list) {
 	// Initialized to avoid a race condition with bShowDebugStats changing.
 	double start = 0.0;
@@ -518,7 +537,10 @@ bool GPUCommon::InterpretList(DisplayList &list) {
 
 	if (g_Config.bShowDebugStats) {
 		time_update();
-		gpuStats.msProcessingDisplayLists += time_now_d() - start;
+		double total = time_now_d() - start - timeSpentStepping_;
+		hleSetSteppingTime(timeSpentStepping_);
+		timeSpentStepping_ = 0.0;
+		gpuStats.msProcessingDisplayLists += total;
 	}
 	return gpuState == GPUSTATE_DONE || gpuState == GPUSTATE_ERROR;
 }

GPU/GPUCommon.h

@@ -85,6 +85,40 @@ public:
 		return false;
 	}
 
+	// From GPUDebugInterface.
+	virtual bool GetCurrentDisplayList(DisplayList &list);
+	virtual std::vector<DisplayList> ActiveDisplayLists();
+	virtual void ResetListPC(int listID, u32 pc);
+	virtual void ResetListStall(int listID, u32 stall);
+	virtual void ResetListState(int listID, DisplayListState state);
+
+	virtual GPUDebugOp DissassembleOp(u32 pc, u32 op);
+	virtual std::vector<GPUDebugOp> DissassembleOpRange(u32 startpc, u32 endpc);
+
+	virtual void NotifySteppingEnter();
+	virtual void NotifySteppingExit();
+
+	virtual u32 GetRelativeAddress(u32 data);
+	virtual u32 GetVertexAddress();
+	virtual u32 GetIndexAddress();
+	virtual GPUgstate GetGState();
+	virtual void SetCmdValue(u32 op);
+
+	virtual DisplayList* getList(int listid) {
+		return &dls[listid];
+	}
+
+	const std::list<int>& GetDisplayLists() {
+		return dlQueue;
+	}
+	virtual bool DecodeTexture(u8* dest, GPUgstate state) {
+		return false;
+	}
+	std::vector<FramebufferInfo> GetFramebufferList() {
+		return std::vector<FramebufferInfo>();
+	}
+	virtual void ClearShaderCache() {}
+
 protected:
 	// To avoid virtual calls to PreExecuteOp().
 	virtual void FastRunLoop(DisplayList &list) = 0;
@@ -162,39 +196,7 @@ private:
 #endif
 	}
 
-public:
-	// From GPUDebugInterface.
-	virtual bool GetCurrentDisplayList(DisplayList &list);
-	virtual std::vector<DisplayList> ActiveDisplayLists();
-	virtual void ResetListPC(int listID, u32 pc);
-	virtual void ResetListStall(int listID, u32 stall);
-	virtual void ResetListState(int listID, DisplayListState state);
-
-	virtual GPUDebugOp DissassembleOp(u32 pc, u32 op);
-	virtual std::vector<GPUDebugOp> DissassembleOpRange(u32 startpc, u32 endpc);
-
-	virtual u32 GetRelativeAddress(u32 data);
-	virtual u32 GetVertexAddress();
-	virtual u32 GetIndexAddress();
-	virtual GPUgstate GetGState();
-	virtual void SetCmdValue(u32 op);
-
-	virtual DisplayList* getList(int listid)
-	{
-		return &dls[listid];
-	}
-
-	const std::list<int>& GetDisplayLists()
-	{
-		return dlQueue;
-	}
-	virtual bool DecodeTexture(u8* dest, GPUgstate state)
-	{
-		return false;
-	}
-	std::vector<FramebufferInfo> GetFramebufferList()
-	{
-		return std::vector<FramebufferInfo>();
-	}
-	virtual void ClearShaderCache() {}
+	// Debug stats.
+	double timeSteppingStarted_;
+	double timeSpentStepping_;
 };

headless/Headless.cpp

@@ -321,7 +321,7 @@ int main(int argc, const char* argv[])
 	// Never report from tests.
 	g_Config.sReportHost = "";
 	g_Config.bAutoSaveSymbolMap = false;
-	g_Config.iRenderingMode = 0;
+	g_Config.iRenderingMode = 1;
 	g_Config.bHardwareTransform = true;
 #ifdef USING_GLES2
 	g_Config.iAnisotropyLevel = 0;
@@ -346,6 +346,7 @@ int main(int argc, const char* argv[])
 	g_Config.iSFXVolume = MAX_CONFIG_VOLUME;
 	g_Config.bSoftwareSkinning = true;
 	g_Config.bVertexDecoderJit = true;
+	g_Config.bBlockTransferGPU = true;
 
 #ifdef _WIN32
 	InitSysDirectories();

headless/WindowsHeadlessHost.cpp

@@ -102,6 +102,11 @@ void WindowsHeadlessHost::SendOrCollectDebugOutput(const std::string &data)
 
 void WindowsHeadlessHost::SendDebugScreenshot(const u8 *pixbuf, u32 w, u32 h)
 {
+	// Only if we're actually comparing.
+	if (comparisonScreenshot.empty()) {
+		return;
+	}
+
 	// We ignore the current framebuffer parameters and just grab the full screen.
 	const static int FRAME_WIDTH = 512;
 	const static int FRAME_HEIGHT = 272;