From 575bdd9166330ee2cd77e72abcacf6dce2ed863b Mon Sep 17 00:00:00 2001
From: hrydgard <hrydgard@gmail.com>
Date: Tue, 12 Aug 2008 20:05:45 +0000
Subject: [PATCH] Lots of various changes. CPU detect fix. Maybe a minor speed
 increase. CPU bugs remain.

git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@180 8ced0084-cf51-0410-be5f-012b33b47a6e
---
 Source/Core/Common/Common.vcproj              |   8 +
 Source/Core/Common/Src/ABI.h                  |   7 +-
 Source/Core/Common/Src/CPUDetect.cpp          | 158 +++++++++---------
 Source/Core/Common/Src/CPUDetect.h            |   8 +-
 Source/Core/Common/Src/SConscript             |   1 +
 Source/Core/Common/Src/Thunk.cpp              | 147 ++++++++++++++++
 Source/Core/Common/Src/Thunk.h                |  39 +++++
 Source/Core/Common/Src/x64Emitter.cpp         |   4 +-
 Source/Core/Common/Src/x64Emitter.h           |  23 +++
 Source/Core/Core/Src/Core.cpp                 |   5 +-
 Source/Core/Core/Src/HW/HW.cpp                |   3 +
 Source/Core/Core/Src/HW/Memmap.cpp            |   6 +-
 .../Core/Core/Src/HW/PeripheralInterface.cpp  |   2 +-
 .../Interpreter/Interpreter_LoadStore.cpp     |  36 ++--
 Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp    |  10 +-
 .../Core/Src/PowerPC/Jit64/JitBackpatch.cpp   |   3 +-
 .../Core/Src/PowerPC/Jit64/JitRegCache.cpp    |   9 +-
 .../Core/Core/Src/PowerPC/Jit64/JitRegCache.h |   1 -
 .../Core/Src/PowerPC/Jit64/Jit_Branch.cpp     |   6 +-
 .../Src/PowerPC/Jit64/Jit_FloatingPoint.cpp   |   1 +
 .../Core/Src/PowerPC/Jit64/Jit_Integer.cpp    |  27 +--
 .../Core/Src/PowerPC/Jit64/Jit_LoadStore.cpp  |  33 ++--
 .../PowerPC/Jit64/Jit_LoadStoreFloating.cpp   |  35 ++--
 .../Src/PowerPC/Jit64/Jit_LoadStorePaired.cpp |  14 +-
 .../Core/Src/PowerPC/Jit64/Jit_Paired.cpp     |   2 -
 .../Src/PowerPC/Jit64/Jit_SystemRegisters.cpp |   2 +-
 Source/Core/DolphinWX/src/Main.cpp            |   2 +-
 27 files changed, 400 insertions(+), 192 deletions(-)
 create mode 100644 Source/Core/Common/Src/Thunk.cpp
 create mode 100644 Source/Core/Common/Src/Thunk.h

diff --git a/Source/Core/Common/Common.vcproj b/Source/Core/Common/Common.vcproj
index 03d5d9dbca..bdd518410d 100644
--- a/Source/Core/Common/Common.vcproj
+++ b/Source/Core/Common/Common.vcproj
@@ -659,6 +659,14 @@
 			RelativePath=".\Src\Thread.h"
 			>
 		</File>
+		<File
+			RelativePath=".\Src\Thunk.cpp"
+			>
+		</File>
+		<File
+			RelativePath=".\Src\Thunk.h"
+			>
+		</File>
 		<File
 			RelativePath=".\Src\Timer.cpp"
 			>
diff --git a/Source/Core/Common/Src/ABI.h b/Source/Core/Common/Src/ABI.h
index feb4431531..2c6f59fac6 100644
--- a/Source/Core/Common/Src/ABI.h
+++ b/Source/Core/Common/Src/ABI.h
@@ -28,7 +28,7 @@
 // * Caller fixes stack after call
 // * function subtract from stack for local storage only.
 // Scratch:      EAX ECX EDX
-// Callee-save:  EBX ESI EDI EBP 
+// Callee-save:  EBX ESI EDI EBP
 // Parameters:   -
 
 // Windows 64-bit
@@ -103,6 +103,11 @@ void ABI_PopAllCalleeSavedRegsAndAdjustStack();
 void ABI_PushAllCallerSavedRegsAndAdjustStack();
 void ABI_PopAllCallerSavedRegsAndAdjustStack();
 
+#ifdef _M_IX86
+inline int ABI_GetNumXMMRegs() { return 8; }
+#else
+inline int ABI_GetNumXMMRegs() { return 16; }
+#endif
 
 #endif  // _JIT_ABI_H
 
diff --git a/Source/Core/Common/Src/CPUDetect.cpp b/Source/Core/Common/Src/CPUDetect.cpp
index 344b5b670e..298a9dfc51 100644
--- a/Source/Core/Common/Src/CPUDetect.cpp
+++ b/Source/Core/Common/Src/CPUDetect.cpp
@@ -21,7 +21,7 @@
 
 //#include <config/i386/cpuid.h>
 #include <xmmintrin.h>
-void __cpuid(int info[4], int x) {}
+void __cpuid(int info[4], int x) {memset(info, 0, sizeof(info));}
 
 #endif
 
@@ -72,94 +72,94 @@ void CPUInfoStruct::Detect()
 		isAMD = true;
 	}
 
-	// Get the information associated with each valid Id
-	for (unsigned int i = 0; i <= nIds; ++i)
+	if (nIds >= 2)
 	{
-		__cpuid(CPUInfo, i);
+		// Get the information associated with each valid Id
+		__cpuid(CPUInfo, 1);
+		
+		nSteppingID = CPUInfo[0] & 0xf;
+		nModel  = (CPUInfo[0] >> 4) & 0xf;
+		nFamily = (CPUInfo[0] >> 8) & 0xf;
+		nProcessorType  = (CPUInfo[0] >> 12) & 0x3;
+		nExtendedmodel  = (CPUInfo[0] >> 16) & 0xf;
+		nExtendedfamily = (CPUInfo[0] >> 20) & 0xff;
+		nBrandIndex = CPUInfo[1] & 0xff;
+		nCLFLUSHcachelinesize = ((CPUInfo[1] >> 8) & 0xff) * 8;
+		nAPICPhysicalID = (CPUInfo[1] >> 24) & 0xff;
+		bSSE3 = (CPUInfo[2] & 0x1) || false;
+		bSSSE3 = (CPUInfo[2] & 0x200) || false;
+		bMONITOR_MWAIT = (CPUInfo[2] & 0x8) || false;
+		bCPLQualifiedDebugStore = (CPUInfo[2] & 0x10) || false;
+		bThermalMonitor2 = (CPUInfo[2] & 0x100) || false;
+		nFeatureInfo = CPUInfo[3];
 
-		// Interpret CPU feature information.
-		if  (i == 1)
+		if (CPUInfo[2] & (1 << 23))
 		{
-			nSteppingID = CPUInfo[0] & 0xf;
-			nModel  = (CPUInfo[0] >> 4) & 0xf;
-			nFamily = (CPUInfo[0] >> 8) & 0xf;
-			nProcessorType  = (CPUInfo[0] >> 12) & 0x3;
-			nExtendedmodel  = (CPUInfo[0] >> 16) & 0xf;
-			nExtendedfamily = (CPUInfo[0] >> 20) & 0xff;
-			nBrandIndex = CPUInfo[1] & 0xff;
-			nCLFLUSHcachelinesize = ((CPUInfo[1] >> 8) & 0xff) * 8;
-			nAPICPhysicalID = (CPUInfo[1] >> 24) & 0xff;
-			bSSE3NewInstructions = (CPUInfo[2] & 0x1) || false;
-			bSSSE3NewInstructions = (CPUInfo[2] & 0x200) || false;
-			bMONITOR_MWAIT = (CPUInfo[2] & 0x8) || false;
-			bCPLQualifiedDebugStore = (CPUInfo[2] & 0x10) || false;
-			bThermalMonitor2 = (CPUInfo[2] & 0x100) || false;
-			nFeatureInfo = CPUInfo[3];
+			bPOPCNT = true;
+		}
 
-			if (CPUInfo[2] & (1 << 23))
-			{
-				bPOPCNT = true;
-			}
+		if (CPUInfo[2] & (1 << 19))
+		{
+			bSSE4_1 = true;
+		}
 
-			if (CPUInfo[2] & (1 << 19))
-			{
-				bSSE4_1 = true;
-			}
-
-			if (CPUInfo[2] & (1 << 20))
-			{
-				bSSE4_2 = true;
-			}
+		if (CPUInfo[2] & (1 << 20))
+		{
+			bSSE4_2 = true;
 		}
 	}
 
-	// Calling __cpuid with 0x80000000 as the InfoType argument
-	// gets the number of valid extended IDs.
-	__cpuid(CPUInfo, 0x80000000);
-	nExIds = CPUInfo[0];
-	memset(CPUBrandString, 0, sizeof(CPUBrandString));
-
-	// Get the information associated with each extended ID.
-	for (unsigned int i = 0x80000000; i <= nExIds; ++i)
+	if (bSSE3)
 	{
-		__cpuid(CPUInfo, i);
+		// Only SSE3 CPU-s support extended infotypes
+		// Calling __cpuid with 0x80000000 as the InfoType argument
+		// gets the number of valid extended IDs.
+		__cpuid(CPUInfo, 0x80000000);
+		nExIds = CPUInfo[0];
+		memset(CPUBrandString, 0, sizeof(CPUBrandString));
 
-		// Interpret CPU brand string and cache information.
-		if (i == 0x80000001)
+		// Get the information associated with each extended ID.
+		for (unsigned int i = 0x80000000; i <= nExIds; ++i)
 		{
-			// This block seems bugged.
-			nFeatureInfo2 = CPUInfo[1]; // ECX
-			bSSE5  = (nFeatureInfo2 & (1 << 11)) ? true : false;
-			bLZCNT = (nFeatureInfo2 & (1 << 5)) ? true : false;
-			bSSE4A = (nFeatureInfo2 & (1 << 6)) ? true : false;
-			bLAHFSAHF64 = (nFeatureInfo2 & (1 << 0)) ? true : false;
+			__cpuid(CPUInfo, i);
 
-			CPU64bit = (CPUInfo[2] & (1 << 29)) ? true : false;
-		}
-		else if  (i == 0x80000002)
-		{
-			memcpy(CPUBrandString, CPUInfo, sizeof(CPUInfo));
-		}
-		else if  (i == 0x80000003)
-		{
-			memcpy(CPUBrandString + 16, CPUInfo, sizeof(CPUInfo));
-		}
-		else if  (i == 0x80000004)
-		{
-			memcpy(CPUBrandString + 32, CPUInfo, sizeof(CPUInfo));
-		}
-		else if  (i == 0x80000006)
-		{
-			nCacheLineSize   = CPUInfo[2] & 0xff;
-			nL2Associativity = (CPUInfo[2] >> 12) & 0xf;
-			nCacheSizeK = (CPUInfo[2] >> 16) & 0xffff;
-		}
-		else if (i == 0x80000008)
-		{
-			int numLSB = (CPUInfo[2] >> 12) & 0xF;
-			numCores = 1 << numLSB;
-			//int coresPerDie = CPUInfo[2] & 0xFF;
-			// numCores = coresPerDie;
+			// Interpret CPU brand string and cache information.
+			if (i == 0x80000001)
+			{
+				// This block seems bugged.
+				nFeatureInfo2 = CPUInfo[1]; // ECX
+				bSSE5  = (nFeatureInfo2 & (1 << 11)) ? true : false;
+				bLZCNT = (nFeatureInfo2 & (1 << 5)) ? true : false;
+				bSSE4A = (nFeatureInfo2 & (1 << 6)) ? true : false;
+				bLAHFSAHF64 = (nFeatureInfo2 & (1 << 0)) ? true : false;
+
+				CPU64bit = (CPUInfo[2] & (1 << 29)) ? true : false;
+			}
+			else if  (i == 0x80000002)
+			{
+				memcpy(CPUBrandString, CPUInfo, sizeof(CPUInfo));
+			}
+			else if  (i == 0x80000003)
+			{
+				memcpy(CPUBrandString + 16, CPUInfo, sizeof(CPUInfo));
+			}
+			else if  (i == 0x80000004)
+			{
+				memcpy(CPUBrandString + 32, CPUInfo, sizeof(CPUInfo));
+			}
+			else if  (i == 0x80000006)
+			{
+				nCacheLineSize   = CPUInfo[2] & 0xff;
+				nL2Associativity = (CPUInfo[2] >> 12) & 0xf;
+				nCacheSizeK = (CPUInfo[2] >> 16) & 0xffff;
+			}
+			else if (i == 0x80000008)
+			{
+				int numLSB = (CPUInfo[2] >> 12) & 0xF;
+				numCores = 1 << numLSB;
+				//int coresPerDie = CPUInfo[2] & 0xFF;
+				// numCores = coresPerDie;
+			}
 		}
 	}
 
@@ -222,9 +222,9 @@ void CPUInfoStruct::Detect()
 	nIds <<= 1;
 	bFXSAVE_FXRSTOR = (nFeatureInfo & nIds) ? true : false;
 	nIds <<= 1;
-	bSSEExtensions = (nFeatureInfo & nIds) ? true : false;
+	bSSE = (nFeatureInfo & nIds) ? true : false;
 	nIds <<= 1;
-	bSSE2Extensions = (nFeatureInfo & nIds) ? true : false;
+	bSSE2 = (nFeatureInfo & nIds) ? true : false;
 	nIds <<= 1;
 	bSelfSnoop = (nFeatureInfo & nIds) ? true : false;
 	nIds <<= 1;
diff --git a/Source/Core/Common/Src/CPUDetect.h b/Source/Core/Common/Src/CPUDetect.h
index 89ef2a7a71..389127261a 100644
--- a/Source/Core/Common/Src/CPUDetect.h
+++ b/Source/Core/Common/Src/CPUDetect.h
@@ -77,16 +77,16 @@ struct CPUInfoStruct
 	bool bThermalMonitorandClockCtrl;
 	bool bMMXTechnology;
 	bool bFXSAVE_FXRSTOR;
-	bool bSSEExtensions;
-	bool bSSE2Extensions;
-	bool bSSE3NewInstructions;
-	bool bSSSE3NewInstructions;
 	bool bSelfSnoop;
 	bool bHyper_threadingTechnology;
 	bool bThermalMonitor;
 	bool bUnknown4;
 	bool bPendBrkEN;
 
+	bool bSSE;
+	bool bSSE2;
+	bool bSSE3;
+	bool bSSSE3;
 	bool bPOPCNT;
 	bool bSSE4_1;
 	bool bSSE4_2;
diff --git a/Source/Core/Common/Src/SConscript b/Source/Core/Common/Src/SConscript
index 839e7382b1..5f215ab564 100644
--- a/Source/Core/Common/Src/SConscript
+++ b/Source/Core/Common/Src/SConscript
@@ -17,6 +17,7 @@ files = ["ABI.cpp",
 		 "PortableSockets.cpp",
 		 "StringUtil.cpp",
 		 "TestFramework.cpp",
+		 "Thunk.cpp",
 		 "Timer.cpp",
 		 "Thread.cpp",
 		 "x64Emitter.cpp", 
diff --git a/Source/Core/Common/Src/Thunk.cpp b/Source/Core/Common/Src/Thunk.cpp
new file mode 100644
index 0000000000..ed7d93bc99
--- /dev/null
+++ b/Source/Core/Common/Src/Thunk.cpp
@@ -0,0 +1,147 @@
+// Copyright (C) 2003-2008 Dolphin Project.
+
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, version 2.0.
+
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License 2.0 for more details.
+
+// A copy of the GPL 2.0 should have been included with the program.
+// If not, see http://www.gnu.org/licenses/
+
+// Official SVN repository and contact information can be found at
+// http://code.google.com/p/dolphin-emu/
+
+#include <map>
+
+#include "Common.h"
+#include "Thunk.h"
+#include "x64Emitter.h"
+#include "MemoryUtil.h"
+#include "ABI.h"
+
+using namespace Gen;
+
+#define THUNK_ARENA_SIZE 1024*1024*1
+
+namespace {
+static std::map<void *, const u8 *> thunks;
+u8 GC_ALIGNED32(saved_fp_state[16 * 4 * 4]);
+u8 GC_ALIGNED32(saved_gpr_state[16 * 8]);
+}
+
+static u8 *thunk_memory;
+static u8 *thunk_code;
+static const u8 *save_regs;
+static const u8 *load_regs;
+u32 saved_return;
+
+void Thunk_Init()
+{
+	thunk_memory = (u8 *)AllocateExecutableMemory(THUNK_ARENA_SIZE);
+	thunk_code = thunk_memory;
+
+	GenContext ctx(&thunk_code);
+	save_regs = GetCodePtr();
+	for (int i = 2; i < ABI_GetNumXMMRegs(); i++)
+		MOVAPS(M(saved_fp_state + i * 16), (X64Reg)(XMM0 + i));
+#ifdef _M_X64
+	MOV(64, M(saved_gpr_state + 0 ), R(RCX));
+	MOV(64, M(saved_gpr_state + 8 ), R(RDX));
+	MOV(64, M(saved_gpr_state + 16), R(R8) );
+	MOV(64, M(saved_gpr_state + 24), R(R9) );
+	MOV(64, M(saved_gpr_state + 32), R(R10));
+	MOV(64, M(saved_gpr_state + 40), R(R11));
+#ifndef _WIN32
+	MOV(64, M(saved_gpr_state + 48), R(RSI));
+	MOV(64, M(saved_gpr_state + 56), R(RDI));
+#endif
+#else
+	MOV(32, M(saved_gpr_state + 0 ), R(RCX));
+	MOV(32, M(saved_gpr_state + 4 ), R(RDX));
+#endif
+	RET();
+	load_regs = GetCodePtr();
+	for (int i = 2; i < ABI_GetNumXMMRegs(); i++)
+		MOVAPS((X64Reg)(XMM0 + i), M(saved_fp_state + i * 16));
+#ifdef _M_X64
+	MOV(64, R(RCX), M(saved_gpr_state + 0 ));
+	MOV(64, R(RDX), M(saved_gpr_state + 8 ));
+	MOV(64, R(R8) , M(saved_gpr_state + 16));
+	MOV(64, R(R9) , M(saved_gpr_state + 24));
+	MOV(64, R(R10), M(saved_gpr_state + 32));
+	MOV(64, R(R11), M(saved_gpr_state + 40));
+#ifndef _WIN32
+	MOV(64, R(RSI), M(saved_gpr_state + 48));
+	MOV(64, R(RDI), M(saved_gpr_state + 56));
+#endif
+#else
+	MOV(32, R(RCX), M(saved_gpr_state + 0 ));
+	MOV(32, R(RDX), M(saved_gpr_state + 4 ));
+#endif
+	RET();
+}
+
+void Thunk_Reset()
+{
+	thunks.clear();
+	thunk_code = thunk_memory;
+}
+
+void Thunk_Shutdown()
+{
+	Thunk_Reset();
+	FreeMemoryPages(thunk_memory, THUNK_ARENA_SIZE);
+	thunk_memory = 0;
+	thunk_code = 0;
+}
+
+void *ProtectFunction(void *function, int num_params)
+{
+	std::map<void *, const u8 *>::iterator iter;
+	iter = thunks.find(function);
+	if (iter != thunks.end())
+		return (void *)iter->second;
+
+	if (!thunk_memory)
+		PanicAlert("Trying to protect functions before the emu is started. Bad bad bad.");
+
+	GenContext gen(&thunk_code);
+	const u8 *call_point = GetCodePtr();
+	// Make sure to align stack.
+#ifdef _M_X64
+#ifdef _WIN32
+	SUB(64, R(ESP), Imm8(0x28));
+#else
+	SUB(64, R(ESP), Imm8(0x8));
+#endif
+	CALL((void*)save_regs);
+	CALL((void*)function);
+	CALL((void*)load_regs);
+#ifdef _WIN32
+	ADD(64, R(ESP), Imm8(0x28));
+#else
+	ADD(64, R(ESP), Imm8(0x8));
+#endif
+	RET();
+#else
+	//INT3();
+	CALL((void*)save_regs);
+	// Re-push parameters from previous stack frame
+	for (int i = 0; i < num_params; i++) {
+		// ESP is changing, so we do not need i
+		PUSH(32, MDisp(ESP, (num_params) * 4));
+	}
+	CALL(function);
+	if (num_params)
+		ADD(32, R(ESP), Imm8(num_params * 4));
+	CALL((void*)load_regs);
+	RET();
+#endif
+
+	thunks[function] = call_point;
+	return (void *)call_point;
+}
\ No newline at end of file
diff --git a/Source/Core/Common/Src/Thunk.h b/Source/Core/Common/Src/Thunk.h
new file mode 100644
index 0000000000..5ce19a9ef1
--- /dev/null
+++ b/Source/Core/Common/Src/Thunk.h
@@ -0,0 +1,39 @@
+// Copyright (C) 2003-2008 Dolphin Project.
+
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, version 2.0.
+
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License 2.0 for more details.
+
+// A copy of the GPL 2.0 should have been included with the program.
+// If not, see http://www.gnu.org/licenses/
+
+// Official SVN repository and contact information can be found at
+// http://code.google.com/p/dolphin-emu/
+
+#ifndef _THUNK_H
+#define _THUNK_H
+
+// This simple class creates a wrapper around a C/C++ function that saves all fp state
+// before entering it, and restores it upon exit. This is required to be able to selectively
+// call functions from generated code, without inflicting the performance hit and increase
+// of complexity that it means to protect the generated code from this problem.
+
+// This process is called thunking.
+
+// There will only ever be one level of thunking on the stack, plus,
+// we don't want to pollute the stack, so we store away regs somewhere global.
+// NOT THREAD SAFE. This may only be used from the CPU thread.
+// Any other thread using this stuff will be FATAL.
+
+void Thunk_Init();
+void Thunk_Reset();
+void Thunk_Shutdown();
+
+void *ProtectFunction(void *function, int num_params);
+
+#endif
diff --git a/Source/Core/Common/Src/x64Emitter.cpp b/Source/Core/Common/Src/x64Emitter.cpp
index 27c353b3a1..6611cf50a2 100644
--- a/Source/Core/Common/Src/x64Emitter.cpp
+++ b/Source/Core/Common/Src/x64Emitter.cpp
@@ -1156,7 +1156,7 @@ namespace Gen
 	void MOVDDUP(X64Reg regOp, OpArg arg) 
 	{
 		// TODO(ector): check SSE3 flag
-		if (cpu_info.bSSE3NewInstructions)
+		if (cpu_info.bSSE3)
 		{
 			WriteSSEOp(64, 0x12, false, regOp, arg); //SSE3
 		}
@@ -1205,7 +1205,7 @@ namespace Gen
 	}
 
 	void PSHUFB(X64Reg dest, OpArg arg) {
-		if (!cpu_info.bSSE3NewInstructions) {
+		if (!cpu_info.bSSSE3) {
 			PanicAlert("Trying to use PSHUFB on a system that doesn't support it. Bad programmer.");
 		}
 		Write8(0x66);
diff --git a/Source/Core/Common/Src/x64Emitter.h b/Source/Core/Common/Src/x64Emitter.h
index c7e2185412..5126aa1d66 100644
--- a/Source/Core/Common/Src/x64Emitter.h
+++ b/Source/Core/Common/Src/x64Emitter.h
@@ -14,6 +14,9 @@
 
 // Official SVN repository and contact information can be found at
 // http://code.google.com/p/dolphin-emu/
+
+// WARNING - THIS LIBRARY IS NOT THREAD SAFE!!!
+
 #ifndef _DOLPHIN_INTEL_CODEGEN
 #define _DOLPHIN_INTEL_CODEGEN
 
@@ -92,6 +95,26 @@ namespace Gen
 	const u8 *GetCodePtr();
 	u8 *GetWritableCodePtr();
 
+
+	// Safe way to temporarily redirect the code generator.
+	class GenContext 
+	{
+		u8 **code_ptr_ptr;
+		u8 *saved_ptr;
+	public:
+		GenContext(u8 **code_ptr_ptr_)
+		{
+			saved_ptr = GetWritableCodePtr();
+			code_ptr_ptr = code_ptr_ptr_;
+			SetCodePtr(*code_ptr_ptr);
+		}
+		~GenContext()
+		{
+			*code_ptr_ptr = GetWritableCodePtr();
+			SetCodePtr(saved_ptr);
+		}
+	};
+
 	enum NormalOp {
 		nrmADD,
 		nrmADC,
diff --git a/Source/Core/Core/Src/Core.cpp b/Source/Core/Core/Src/Core.cpp
index e2d3397b9f..8646260a82 100644
--- a/Source/Core/Core/Src/Core.cpp
+++ b/Source/Core/Core/Src/Core.cpp
@@ -26,6 +26,7 @@
 
 #include "Console.h"
 #include "Core.h"
+#include "CPUDetect.h"
 #include "CoreTiming.h"
 #include "Boot/Boot.h"
 #include "PatchEngine.h"
@@ -140,7 +141,9 @@ bool Init(const SCoreStartupParameter _CoreParameter)
 	// all right ... here we go
 	Host_SetWaitCursor(false);
 
-	DisplayMessage("Emulation started.", 3000);
+	DisplayMessage(cpu_info.CPUBrandString, 3000);
+	DisplayMessage(_CoreParameter.m_strFilename, 3000);
+
 
 	//RegisterPanicAlertHandler(PanicAlertToVideo);
 
diff --git a/Source/Core/Core/Src/HW/HW.cpp b/Source/Core/Core/Src/HW/HW.cpp
index ccd161e03f..401fd792eb 100644
--- a/Source/Core/Core/Src/HW/HW.cpp
+++ b/Source/Core/Core/Src/HW/HW.cpp
@@ -16,6 +16,7 @@
 // http://code.google.com/p/dolphin-emu/
 
 #include "Common.h"
+#include "Thunk.h"
 #include "../Core.h"
 #include "HW.h"
 #include "../PowerPC/PowerPC.h"
@@ -42,6 +43,7 @@ namespace HW
 {
 	void Init()
 	{
+		Thunk_Init(); // not really hw, but this way we know it's inited first :P
 		// Init the whole Hardware
 		PixelEngine::Init();
 		CommandProcessor::Init();
@@ -72,5 +74,6 @@ namespace HW
 
         WII_IPC_HLE_Interface::Shutdown();
         WII_IPCInterface::Shutdown();
+		Thunk_Shutdown();
 	}
 }
diff --git a/Source/Core/Core/Src/HW/Memmap.cpp b/Source/Core/Core/Src/HW/Memmap.cpp
index 6c9d92a00c..87eae61765 100644
--- a/Source/Core/Core/Src/HW/Memmap.cpp
+++ b/Source/Core/Core/Src/HW/Memmap.cpp
@@ -1054,10 +1054,14 @@ void SDRUpdated()
 u32 CheckDTLB(u32 _Address, XCheckTLBFlag _Flag)
 {
 	PanicAlert("TLB: %s unknown memory (0x%08x)\n"
+		       "This is either the game crashing randomly, or a TLB write."
 			   "Several games uses the TLB to map memory. This\n"
-			   "function is not support in dolphin. Cheers!",
+			   "function is not supported in Dolphin. "
+			   "Also, unfortunately there is no way to recover from this error,"
+			   "so Dolphin will now exit abruptly. Sorry!",
 			   _Flag == FLAG_WRITE ? "Write to" : "Read from", _Address);
 
+	exit(0);
 	u32 sr = PowerPC::ppcState.sr[EA_SR(_Address)]; 
 
 	u32 offset = EA_Offset(_Address);			// 12 bit 
diff --git a/Source/Core/Core/Src/HW/PeripheralInterface.cpp b/Source/Core/Core/Src/HW/PeripheralInterface.cpp
index d0fb2d9af7..94224c5ee8 100644
--- a/Source/Core/Core/Src/HW/PeripheralInterface.cpp
+++ b/Source/Core/Core/Src/HW/PeripheralInterface.cpp
@@ -132,7 +132,7 @@ void CPeripheralInterface::Write32(const u32 _uValue, const u32 _iAddress)
             if ((_uValue != 0x80000001) && (_uValue != 0x80000005)) // DVDLowReset 
             {
                 TCHAR szTemp[256];
-                sprintf(szTemp, "Unknown write to PI_RESET_CODE (%08x)", _uValue);
+				sprintf(szTemp, "Game wants to reset the machine. PI_RESET_CODE: (%08x)", _uValue);
                 PanicAlert(szTemp);
             }
         }
diff --git a/Source/Core/Core/Src/PowerPC/Interpreter/Interpreter_LoadStore.cpp b/Source/Core/Core/Src/PowerPC/Interpreter/Interpreter_LoadStore.cpp
index 952773c5bb..d638885ac1 100644
--- a/Source/Core/Core/Src/PowerPC/Interpreter/Interpreter_LoadStore.cpp
+++ b/Source/Core/Core/Src/PowerPC/Interpreter/Interpreter_LoadStore.cpp
@@ -145,7 +145,7 @@ void CInterpreter::lmw(UGeckoInstruction _inst)
 		u32 TempReg = Memory::Read_U32(uAddress);		
 		if (PowerPC::ppcState.Exceptions & EXCEPTION_DSI)
 		{
-			PanicAlert("DSI exception in lmv. This is very bad.");
+			PanicAlert("DSI exception in lmv.");
 			return;
 		}
 
@@ -500,9 +500,9 @@ void CInterpreter::lswi(UGeckoInstruction _inst)
 
 	u32 n;
 	if (_inst.NB == 0)
-		n=32;
+		n = 32;
 	else
-		n=_inst.NB;
+		n = _inst.NB;
 
 	int r = _inst.RD - 1;
 	int i = 0;
@@ -511,22 +511,22 @@ void CInterpreter::lswi(UGeckoInstruction _inst)
 		if (i==0)
 		{
 			r++;
-			r&=31;
+			r &= 31;
 			m_GPR[r] = 0;
 		}
 
-		u32 TempValue = Memory::Read_U8(EA) << (24-i);		
+		u32 TempValue = Memory::Read_U8(EA) << (24 - i);		
 		if (PowerPC::ppcState.Exceptions & EXCEPTION_DSI)
 		{
-			PanicAlert("DSI exception in lsw. This is very bad.");
+			PanicAlert("DSI exception in lsw.");
 			return;
 		}
 
 		m_GPR[r] |= TempValue;
 
-		i+=8;
-		if (i==32)
-			i=0;
+		i += 8;
+		if (i == 32)
+			i = 0;
 		EA++;
 		n--;
 	}
@@ -546,26 +546,26 @@ void CInterpreter::stswi(UGeckoInstruction _inst)
     
 	u32 n;
 	if (_inst.NB == 0)
-		n=32;
+		n = 32;
 	else
-		n=_inst.NB;
+		n = _inst.NB;
 
 	int r = _inst.RS - 1;
 	int i = 0;
-	while (n>0)
+	while (n > 0)
 	{
-		if (i==0)
+		if (i == 0)
 		{
 			r++;
-			r&=31;
+			r &= 31;
 		}
-		Memory::Write_U8((m_GPR[r] >> (24-i)) & 0xFF, EA);
+		Memory::Write_U8((m_GPR[r] >> (24 - i)) & 0xFF, EA);
 		if (PowerPC::ppcState.Exceptions & EXCEPTION_DSI)
 			return;
 
-		i+=8;
-		if (i==32)
-			i=0;
+		i += 8;
+		if (i == 32)
+			i = 0;
 		EA++;
 		n--;
 	}
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp
index ed8d94b3d0..1c66cfcae3 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp
@@ -220,7 +220,7 @@ subfex
 // Evil
 namespace CPUCompare
 {
-	extern  u32    m_BlockStart;
+	extern u32 m_BlockStart;
 }
 
 
@@ -231,9 +231,8 @@ namespace Jit64
 
 	void WriteCallInterpreter(UGeckoInstruction _inst)
 	{
-		gpr.Flush(js.op);
-		if (PPCTables::UsesFPU(_inst))
-			fpr.Flush(js.op);
+		gpr.Flush(FLUSH_ALL);
+		fpr.Flush(FLUSH_ALL);
 		if (js.isLastInstruction)
 		{
 			MOV(32, M(&PC), Imm32(js.compilerPC));
@@ -250,7 +249,8 @@ namespace Jit64
 
 	void HLEFunction(UGeckoInstruction _inst)
 	{
-		FlushRegCaches();
+		gpr.Flush(FLUSH_ALL);
+		fpr.Flush(FLUSH_ALL);
 		ABI_CallFunctionCC((void*)&HLE::Execute, js.compilerPC, _inst.hex);
 		MOV(32, R(EAX), M(&NPC));
 		WriteExitDestInEAX(0);
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/JitBackpatch.cpp b/Source/Core/Core/Src/PowerPC/Jit64/JitBackpatch.cpp
index 48c3ff56bb..2a22a4f29e 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/JitBackpatch.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/JitBackpatch.cpp
@@ -24,6 +24,7 @@
 
 #include "x64Emitter.h"
 #include "ABI.h"
+#include "Thunk.h"
 #include "x64Analyzer.h"
 
 #include "StringUtil.h"
@@ -109,7 +110,7 @@ void BackPatch(u8 *codePtr, int accessType, u32 emAddress)
 	//	break;
 	case 4:
 		// THIS FUNCTION CANNOT TOUCH FLOATING POINT REGISTERS.
-		CALL((void *)&Memory::Read_U32);
+		CALL(ProtectFunction((void *)&Memory::Read_U32, 1));
 		break;
 	default:
 		BackPatchError(StringFromFormat("We don't handle the size %i yet in backpatch", info.operandSize), codePtr, emAddress);
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/JitRegCache.cpp b/Source/Core/Core/Src/PowerPC/Jit64/JitRegCache.cpp
index 36021251d0..3cbee4af76 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/JitRegCache.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/JitRegCache.cpp
@@ -259,12 +259,12 @@ namespace Jit64
 		{
 #ifdef _M_X64
 #ifdef _WIN32
-			RSI, RDI, R12, R13, R14, R8, R9, RDX, R10, R11 //, RCX
+			RSI, RDI, R12, R13, R14, R8, R9, R10, R11 //, RCX
 #else
 			R12, R13, R14, R8, R9, R10, R11, RSI, RDI //, RCX
 #endif
 #elif _M_IX86
-			ESI, EDI, EBX, EBP, EDX
+			ESI, EDI, EBX, EBP, EDX, ECX,
 #endif
 		};
 		count = sizeof(allocationOrder) / sizeof(const int);
@@ -412,10 +412,7 @@ namespace Jit64
 				if (regs[i].location.IsSimpleReg())
 				{
 					X64Reg xr = RX(i);
-					if (mode != FLUSH_VOLATILE || IsXRegVolatile(xr))
-					{
-						StoreFromX64(i);
-					}
+					StoreFromX64(i);
 					xregs[xr].dirty = false;
 				}
 				else if (regs[i].location.IsImm())
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/JitRegCache.h b/Source/Core/Core/Src/PowerPC/Jit64/JitRegCache.h
index eb0e6b22dd..dabbaa2b57 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/JitRegCache.h
+++ b/Source/Core/Core/Src/PowerPC/Jit64/JitRegCache.h
@@ -25,7 +25,6 @@ namespace Jit64
 	using namespace Gen;
 	enum FlushMode
 	{
-		FLUSH_VOLATILE,
 		//		FLUSH_ALLNONSTATIC,
 		FLUSH_ALL
 	};
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_Branch.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_Branch.cpp
index 416f147c7f..d82eddd3e7 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_Branch.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_Branch.cpp
@@ -48,7 +48,7 @@ namespace Jit64
 	void rfi(UGeckoInstruction _inst)
 	{
 		FlushRegCaches();
-		//Bits SRR1[0,5-9,16�23, 25�27, 30�31] are placed into the corresponding bits of the MSR.
+		//Bits SRR1[0, 5-9, 16-23, 25-27, 30-31] are placed into the corresponding bits of the MSR.
 		//MSR[13] is set to 0.
 		const int mask = 0x87C0FF73;
 		// MSR = (MSR & ~mask) | (SRR1 & mask);
@@ -105,7 +105,7 @@ namespace Jit64
 		const bool only_counter_check = ((inst.BO >> 4) & 1);
 		const bool only_condition_check = ((inst.BO >> 2) & 1);
 		if (only_condition_check && only_counter_check)
-			PanicAlert("Stupid bcx encountered. Likely bad or corrupt code.");
+			PanicAlert("Bizarre bcx encountered. Likely bad or corrupt code.");
 		bool doFullTest = (inst.BO & 16) == 0 && (inst.BO & 4) == 0;
 		bool ctrDecremented = false;
 
@@ -182,7 +182,7 @@ namespace Jit64
 
 		bool fastway = true;
 
-		if((inst.BO & 16) == 0)				
+		if ((inst.BO & 16) == 0)				
 		{
 			PanicAlert("Bizarro bcctrx %08x, not supported.", inst.hex);
 			_assert_msg_(DYNA_REC, 0, "Bizarro bcctrx");
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp
index 9eca6538d0..b4f9ba1e59 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp
@@ -27,6 +27,7 @@
 
 #ifdef _WIN32
 #define INSTRUCTION_START
+//#define INSTRUCTION_START Default(inst); return;
 #else
 #define INSTRUCTION_START Default(inst); return;
 #endif
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_Integer.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_Integer.cpp
index ae09fc056b..297e8edeec 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_Integer.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_Integer.cpp
@@ -317,8 +317,7 @@ namespace Jit64
 	{
 		INSTRUCTION_START;
 		int a = inst.RA, d = inst.RD;
-		gpr.FlushR(ECX);
-		gpr.LockX(ECX);
+		gpr.FlushLockX(ECX);
 		gpr.Lock(a, d);
 		if (a != d)
 			gpr.LoadToX64(d, false, true);
@@ -376,8 +375,7 @@ namespace Jit64
 	{
 		INSTRUCTION_START;
 		int a = inst.RA, d = inst.RD;
-		gpr.FlushR(EDX);
-		gpr.LockX(EDX);
+		gpr.FlushLockX(EDX);
 		gpr.Lock(a, d);
 		if (d != a) {
 			gpr.LoadToX64(d, false, true);
@@ -396,8 +394,7 @@ namespace Jit64
 	{
 		INSTRUCTION_START;
 		int a = inst.RA, b = inst.RB, d = inst.RD;
-		gpr.FlushR(EDX);
-		gpr.LockX(EDX);
+		gpr.FlushLockX(EDX);
 		gpr.Lock(a, b, d);
 		if (d != a && d != b) {
 			gpr.LoadToX64(d, false, true);
@@ -420,8 +417,7 @@ namespace Jit64
 	{
 		INSTRUCTION_START;
 		int a = inst.RA, b = inst.RB, d = inst.RD;
-		gpr.FlushR(EDX);
-		gpr.LockX(EDX);
+		gpr.FlushLockX(EDX);
 		gpr.Lock(a, b, d);
 		if (d != a && d != b) {
 			gpr.LoadToX64(d, false, true);
@@ -450,8 +446,7 @@ namespace Jit64
 		Default(inst); return;
 
 		int a = inst.RA, b = inst.RB, d = inst.RD;
-		gpr.FlushR(EDX);
-		gpr.LockX(EDX);
+		gpr.FlushLockX(EDX);
 		gpr.Lock(a, b, d);
 		if (d != a && d != b) {
 			gpr.LoadToX64(d, false, true);
@@ -534,8 +529,7 @@ namespace Jit64
 	{
 		INSTRUCTION_START;
 		int a = inst.RA, b = inst.RB, d = inst.RD;
-		gpr.FlushR(ECX);
-		gpr.LockX(ECX);
+		gpr.FlushLockX(ECX);
 		gpr.Lock(a, b, d);
 		if (d != a && d != b)
 			gpr.LoadToX64(d, false);
@@ -649,8 +643,7 @@ namespace Jit64
 		}
 
 		u32 mask = Helper_Mask(inst.MB, inst.ME);
-		gpr.FlushR(ECX);
-		gpr.LockX(ECX);
+		gpr.FlushLockX(ECX);
 		gpr.Lock(a, b, s);
 		MOV(32, R(EAX), gpr.R(s));
 		MOV(32, R(ECX), gpr.R(b));
@@ -691,8 +684,7 @@ namespace Jit64
 		int a = inst.RA;
 		int b = inst.RB;
 		int s = inst.RS;
-		gpr.FlushR(ECX);
-		gpr.LockX(ECX);
+		gpr.FlushLockX(ECX);
 		gpr.Lock(a, b, s);
 		gpr.LoadToX64(a, a == s || a == b || s == b, true);
 		MOV(32, R(ECX), gpr.R(b));
@@ -719,8 +711,7 @@ namespace Jit64
 		int a = inst.RA;
 		int b = inst.RB;
 		int s = inst.RS;
-		gpr.FlushR(ECX);
-		gpr.LockX(ECX);
+		gpr.FlushLockX(ECX);
 		gpr.Lock(a, b, s);
 		gpr.LoadToX64(a, a == s || a == b || s == b, true);
 		MOV(32, R(ECX), gpr.R(b));
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStore.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStore.cpp
index b5935b9694..c10c5dde94 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStore.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStore.cpp
@@ -19,6 +19,7 @@
 // Should give a very noticable speed boost to paired single heavy code.
 
 #include "Common.h"
+#include "Thunk.h"
 
 #include "../PowerPC.h"
 #include "../../Core.h"
@@ -82,9 +83,9 @@ namespace Jit64
 		SetJumpTarget(argh);
 		switch (accessSize)
 		{
-		case 32: ABI_CallFunctionR((void *)&Memory::Read_U32, reg); break;
-		case 16: ABI_CallFunctionR((void *)&Memory::Read_U16, reg); break;
-		case 8:  ABI_CallFunctionR((void *)&Memory::Read_U8, reg);  break;
+		case 32: ABI_CallFunctionR(ProtectFunction((void *)&Memory::Read_U32, 1), reg); break;
+		case 16: ABI_CallFunctionR(ProtectFunction((void *)&Memory::Read_U16, 1), reg); break;
+		case 8:  ABI_CallFunctionR(ProtectFunction((void *)&Memory::Read_U8, 1), reg);  break;
 		}
 		SetJumpTarget(arg2);
 	}
@@ -97,9 +98,9 @@ namespace Jit64
 		BSWAP(32, reg_value);
 #ifdef _M_IX86
 		AND(32, R(reg_addr), Imm32(Memory::MEMVIEW32_MASK));
-		MOV(accessSize, MDisp(reg_addr, (u32)Memory::base), R(reg_value));
+		MOV(accessSize, MDisp(reg_addr, (u32)Memory::base + offset), R(reg_value));
 #else
-		MOV(accessSize, MComplex(RBX, reg_addr, SCALE_1, 0), R(reg_value));
+		MOV(accessSize, MComplex(RBX, reg_addr, SCALE_1, offset), R(reg_value));
 #endif
 	}
 
@@ -113,17 +114,16 @@ namespace Jit64
 		UnsafeWriteRegToReg(reg_value, reg_addr, accessSize, 0);
 		FixupBranch arg2 = J();
 		SetJumpTarget(argh);
-		ABI_CallFunctionRR((void *)&Memory::Write_U32, ABI_PARAM1, ABI_PARAM2); 
+		ABI_CallFunctionRR(ProtectFunction((void *)&Memory::Write_U32, 2), ABI_PARAM1, ABI_PARAM2); 
 		SetJumpTarget(arg2);
 	}
 
 	void lbzx(UGeckoInstruction inst)
 	{
 		INSTRUCTION_START;
-		gpr.Flush(FLUSH_VOLATILE);
-		fpr.Flush(FLUSH_VOLATILE);
 		int a = inst.RA, b = inst.RB, d = inst.RD;
 		gpr.Lock(a, b, d);
+		gpr.FlushLockX(ABI_PARAM1);
 		if (b == d || a == d)
 			gpr.LoadToX64(d, true, true);
 		else 
@@ -134,6 +134,7 @@ namespace Jit64
 		SafeLoadRegToEAX(ABI_PARAM1, 8, 0);
 		MOV(32, gpr.R(d), R(EAX));
 		gpr.UnlockAll();
+		gpr.UnlockAllX();
 	}
 
 	void lXz(UGeckoInstruction inst)
@@ -145,7 +146,6 @@ namespace Jit64
 		// TODO(ector): Make it dynamically enable/disable idle skipping where appropriate
 		// Will give nice boost to dual core mode
 		// if (CommandProcessor::AllowIdleSkipping() && PixelEngine::AllowIdleSkipping())
-
 		if (!Core::GetStartupParameter().bUseDualCore && 
 			inst.OPCD == 32 && 
 			(inst.hex & 0xFFFF0000) == 0x800D0000 &&
@@ -172,7 +172,7 @@ namespace Jit64
 		{
 		case 32: accessSize = 32; break; //lwz
 		case 40: accessSize = 16; break; //lhz
-		case 34: accessSize = 8; break;  //lbz
+		case 34: accessSize = 8;  break; //lbz
 		default: _assert_msg_(DYNA_REC, 0, "lXz: invalid access size"); return;
 		}
 
@@ -183,8 +183,6 @@ namespace Jit64
 		if (true) {
 #endif
 			// Safe and boring
-			gpr.Flush(FLUSH_VOLATILE);
-			fpr.Flush(FLUSH_VOLATILE);
 			gpr.FlushLockX(ABI_PARAM1);
 			gpr.Lock(d, a);
 			MOV(32, R(ABI_PARAM1), gpr.R(a));
@@ -221,8 +219,6 @@ namespace Jit64
 		int a = inst.RA;
 		s32 offset = (s32)(s16)inst.SIMM_16;
 		// Safe and boring
-		gpr.Flush(FLUSH_VOLATILE);
-		fpr.Flush(FLUSH_VOLATILE);
 		gpr.FlushLockX(ABI_PARAM1);
 		gpr.Lock(d, a);
 		MOV(32, R(ABI_PARAM1), gpr.R(a));
@@ -272,8 +268,6 @@ namespace Jit64
 		s32 offset = (s32)(s16)inst.SIMM_16;
 		if (a || update) 
 		{
-			gpr.Flush(FLUSH_VOLATILE);
-			fpr.Flush(FLUSH_VOLATILE);
 			int accessSize;
 			switch (inst.OPCD & ~1)
 			{
@@ -358,6 +352,7 @@ namespace Jit64
 				ADD(32, R(ABI_PARAM2), Imm32((u32)offset));
 			if (update && offset)
 			{
+				gpr.LoadToX64(a, true, true);
 				MOV(32, gpr.R(a), R(ABI_PARAM2));
 			}
 			TEST(32, R(ABI_PARAM2), Imm32(0x0C000000));
@@ -380,9 +375,9 @@ namespace Jit64
 			SetJumpTarget(argh);
 			switch (accessSize)
 			{
-			case 32: ABI_CallFunctionRR((void *)&Memory::Write_U32, ABI_PARAM1, ABI_PARAM2); break;
-			case 16: ABI_CallFunctionRR((void *)&Memory::Write_U16, ABI_PARAM1, ABI_PARAM2); break;
-			case 8:  ABI_CallFunctionRR((void *)&Memory::Write_U8,  ABI_PARAM1, ABI_PARAM2); break;
+			case 32: ABI_CallFunctionRR(ProtectFunction((void *)&Memory::Write_U32, 2), ABI_PARAM1, ABI_PARAM2); break;
+			case 16: ABI_CallFunctionRR(ProtectFunction((void *)&Memory::Write_U16, 2), ABI_PARAM1, ABI_PARAM2); break;
+			case 8:  ABI_CallFunctionRR(ProtectFunction((void *)&Memory::Write_U8,  2), ABI_PARAM1, ABI_PARAM2); break;
 			}
 			SetJumpTarget(arg2);
 			gpr.UnlockAll();
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStoreFloating.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStoreFloating.cpp
index 95580bf76b..89e23ccad1 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStoreFloating.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStoreFloating.cpp
@@ -65,7 +65,6 @@ static u32 GC_ALIGNED16(temp32);
 void lfs(UGeckoInstruction inst)
 {
 	INSTRUCTION_START;
-	DISABLE_32BIT;
 	int d = inst.RD;
 	int a = inst.RA;
 	if (!a) 
@@ -74,15 +73,8 @@ void lfs(UGeckoInstruction inst)
 		return;
 	}
 	s32 offset = (s32)(s16)inst.SIMM_16;
-
-	if (jo.noAssumeFPLoadFromMem) {
-		// We might call a function.
-		gpr.Flush(FLUSH_VOLATILE);
-		fpr.Flush(FLUSH_VOLATILE);
-		gpr.FlushLockX(ABI_PARAM1);
-	}
-	gpr.Lock(d, a);
-	
+	gpr.FlushLockX(ABI_PARAM1);
+	gpr.Lock(a);
 	MOV(32, R(ABI_PARAM1), gpr.R(a));
 	if (!jo.noAssumeFPLoadFromMem)
 	{
@@ -103,6 +95,7 @@ void lfs(UGeckoInstruction inst)
 	fpr.UnlockAll();
 }
 
+
 void lfd(UGeckoInstruction inst)
 {
 	INSTRUCTION_START;
@@ -115,11 +108,12 @@ void lfd(UGeckoInstruction inst)
 		return;
 	}
 	s32 offset = (s32)(s16)inst.SIMM_16;
+	gpr.FlushLockX(ABI_PARAM1);
 	gpr.Lock(a);
 	MOV(32, R(ABI_PARAM1), gpr.R(a));
 	fpr.LoadToX64(d, false);
 	fpr.Lock(d);
-	if (cpu_info.bSSE3NewInstructions) {
+	if (cpu_info.bSSSE3) {
 		X64Reg xd = fpr.RX(d);
 		MOVQ_xmm(xd, MComplex(RBX, ABI_PARAM1, SCALE_1, offset));
 		PSHUFB(xd, M((void *)bswapShuffle1x8Dupe));
@@ -130,13 +124,14 @@ void lfd(UGeckoInstruction inst)
 		MOVDDUP(fpr.RX(d), M(&temp64));
 	}
 	gpr.UnlockAll();
+	gpr.UnlockAllX();
 	fpr.UnlockAll();
 }
 
 void stfd(UGeckoInstruction inst)
 {
 	INSTRUCTION_START;
-	if (!cpu_info.bSSSE3NewInstructions)
+	if (!cpu_info.bSSSE3)
 	{
 		DISABLE_32BIT;
 	}
@@ -148,14 +143,14 @@ void stfd(UGeckoInstruction inst)
 		return;
 	}
 	s32 offset = (s32)(s16)inst.SIMM_16;
+	gpr.FlushLockX(ABI_PARAM1);
 	gpr.Lock(a);
 	fpr.Lock(s);
-	gpr.FlushLockX(ABI_PARAM1);
 	MOV(32, R(ABI_PARAM1), gpr.R(a));
 #ifdef _M_IX86
 	AND(32, R(ABI_PARAM1), Imm32(Memory::MEMVIEW32_MASK));
 #endif
-	if (cpu_info.bSSSE3NewInstructions) {
+	if (cpu_info.bSSSE3) {
 		MOVAPS(XMM0, fpr.R(s));
 		PSHUFB(XMM0, M((void *)bswapShuffle1x8));
 #ifdef _M_X64
@@ -175,6 +170,7 @@ void stfd(UGeckoInstruction inst)
 	fpr.UnlockAll();
 }
 
+
 void stfs(UGeckoInstruction inst)
 {
 	INSTRUCTION_START;
@@ -185,12 +181,11 @@ void stfs(UGeckoInstruction inst)
 	s32 offset = (s32)(s16)inst.SIMM_16;
 	if (a && !update)
 	{
-		gpr.Flush(FLUSH_VOLATILE);
-		fpr.Flush(FLUSH_VOLATILE);
+		gpr.FlushLockX(ABI_PARAM1, ABI_PARAM2);
 		gpr.Lock(a);
 		fpr.Lock(s);
-		gpr.FlushLockX(ABI_PARAM1, ABI_PARAM2);
 		MOV(32, R(ABI_PARAM2), gpr.R(a));
+		ADD(32, R(ABI_PARAM2), Imm32(offset));
 		if (update && offset)
 		{
 			MOV(32, gpr.R(a), R(ABI_PARAM2));
@@ -198,7 +193,7 @@ void stfs(UGeckoInstruction inst)
 		CVTSD2SS(XMM0, fpr.R(s));
 		MOVSS(M(&temp32), XMM0);
 		MOV(32, R(ABI_PARAM1), M(&temp32));
-		SafeWriteRegToReg(ABI_PARAM1, ABI_PARAM2, 32, offset);
+		SafeWriteRegToReg(ABI_PARAM1, ABI_PARAM2, 32, 0);
 		gpr.UnlockAll();
 		gpr.UnlockAllX();
 		fpr.UnlockAll();
@@ -209,6 +204,7 @@ void stfs(UGeckoInstruction inst)
 	}
 }
 
+
 void lfsx(UGeckoInstruction inst)
 {
 	INSTRUCTION_START;
@@ -217,8 +213,7 @@ void lfsx(UGeckoInstruction inst)
 	MOV(32, R(EAX), gpr.R(inst.RB));
 	if (inst.RA)
 		ADD(32, R(EAX), gpr.R(inst.RA));
-	if (cpu_info.bSSSE3NewInstructions) {
-		// PanicAlert("SSE3 supported!");
+	if (cpu_info.bSSSE3) {
 		X64Reg r = fpr.R(inst.RS).GetSimpleReg();
 #ifdef _M_IX86
 		AND(32, R(EAX), Imm32(Memory::MEMVIEW32_MASK));
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStorePaired.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStorePaired.cpp
index 281f2debcd..fa344cafe2 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStorePaired.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_LoadStorePaired.cpp
@@ -20,6 +20,7 @@
 
 #include "Common.h"
 
+#include "Thunk.h"
 #include "../PowerPC.h"
 #include "../../Core.h"
 #include "../../HW/GPFifo.h"
@@ -36,7 +37,7 @@
 #include "JitAsm.h"
 #include "JitRegCache.h"
 
-// #define INSTRUCTION_START Default(inst); return;
+//#define INSTRUCTION_START Default(inst); return;
 #define INSTRUCTION_START
 
 #ifdef _M_IX86
@@ -125,8 +126,6 @@ void psq_st(UGeckoInstruction inst)
 	if (stType == QUANTIZE_FLOAT)
 	{
 		DISABLE_32BIT;
-		gpr.Flush(FLUSH_VOLATILE);
-		fpr.Flush(FLUSH_VOLATILE);
 		gpr.FlushLockX(ABI_PARAM1, ABI_PARAM2);
 		gpr.Lock(a);
 		fpr.Lock(s);
@@ -147,7 +146,7 @@ void psq_st(UGeckoInstruction inst)
 		MOV(64, MComplex(RBX, ABI_PARAM2, SCALE_1, 0), R(ABI_PARAM1));
 		FixupBranch arg2 = J();
 		SetJumpTarget(argh);
-		CALL((void *)&WriteDual32); 
+		CALL(ProtectFunction((void *)&WriteDual32, 0));
 		SetJumpTarget(arg2);
 		gpr.UnlockAll();
 		gpr.UnlockAllX();
@@ -255,7 +254,7 @@ void psq_l(UGeckoInstruction inst)
 #ifdef _M_X64
 			gpr.LoadToX64(inst.RA, true, update);
 			fpr.LoadToX64(inst.RS, false);
-			if (cpu_info.bSSSE3NewInstructions) {
+			if (cpu_info.bSSSE3) {
 				X64Reg xd = fpr.R(inst.RS).GetSimpleReg();
 				MOVQ_xmm(xd, MComplex(RBX, gpr.R(inst.RA).GetSimpleReg(), 1, offset));
 				PSHUFB(xd, M((void *)pbswapShuffle2x4));
@@ -272,7 +271,7 @@ void psq_l(UGeckoInstruction inst)
 				ADD(32, gpr.R(inst.RA), Imm32(offset));
 			break;
 #else
-			if (cpu_info.bSSSE3NewInstructions) {
+			if (cpu_info.bSSSE3) {
 				gpr.LoadToX64(inst.RA, true, update);
 				fpr.LoadToX64(inst.RS, false);
 				X64Reg xd = fpr.R(inst.RS).GetSimpleReg();
@@ -282,8 +281,7 @@ void psq_l(UGeckoInstruction inst)
 				PSHUFB(xd, M((void *)pbswapShuffle2x4));
 				CVTPS2PD(xd, R(xd));
 			} else {
-				gpr.FlushR(ECX);
-				gpr.LockX(ECX);
+				gpr.FlushLockX(ECX);
 				gpr.LoadToX64(inst.RA);
 				// This can probably be optimized somewhat.
 				LEA(32, ECX, MDisp(gpr.R(inst.RA).GetSimpleReg(), offset));
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_Paired.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_Paired.cpp
index 862c7c1dae..90a93f9f3e 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_Paired.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_Paired.cpp
@@ -93,8 +93,6 @@ namespace Jit64
 		fpr.UnlockAll();
 	}
 
-
-
 	//add a, b, c
 	
 	//mov a, b
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_SystemRegisters.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_SystemRegisters.cpp
index 6acedc9650..5e27957d74 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_SystemRegisters.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_SystemRegisters.cpp
@@ -59,7 +59,7 @@ namespace Jit64
 		case SPR_GQR0 + 5:
 		case SPR_GQR0 + 6:
 		case SPR_GQR0 + 7:
-			js.blockSetsQuantizers = false;
+			js.blockSetsQuantizers = true;
 			// Prevent recompiler from compiling in old quantizer values.
 			// TODO - actually save the set state and use it in following quantizer ops.
 			break;
diff --git a/Source/Core/DolphinWX/src/Main.cpp b/Source/Core/DolphinWX/src/Main.cpp
index 612614e72e..ca6383db92 100644
--- a/Source/Core/DolphinWX/src/Main.cpp
+++ b/Source/Core/DolphinWX/src/Main.cpp
@@ -45,7 +45,7 @@ bool DolphinApp::OnInit()
 
 #ifdef _WIN32
 	// TODO: if First Boot
-	if (!cpu_info.bSSE2Extensions) 
+	if (!cpu_info.bSSE2) 
 	{
 		MessageBox(0, _T("Hi,\n\nDolphin requires that your CPU has support for SSE2 extensions.\n"
 			             "Unfortunately your CPU does not support them, so Dolphin will not run.\n\n"