Merge branch 'master' into perfqueries.

Conflicts: Source/Core/VideoCommon/Src/VideoConfig.h Source/Core/VideoCommon/VideoCommon.vcxproj.filters Source/Plugins/Plugin_VideoDX11/Src/VertexManager.cpp Source/Plugins/Plugin_VideoSoftware/Src/Rasterizer.cpp Source/Plugins/Plugin_VideoSoftware/Src/Tev.cpp
2013-03-01 01:07:34 +01:00
parent 0f617183a8 bc15919e2e
commit 7682ed22c6
886 changed files with 69167 additions and 56567 deletions
@@ -157,9 +157,21 @@ void AVIDump::Stop()
 	NOTICE_LOG(VIDEO, "Stop");
 }

-void AVIDump::AddFrame(char *data)
+void AVIDump::AddFrame(const u8* data, int w, int h)
 {
-	AVIStreamWrite(m_streamCompressed, ++m_frameCount, 1, (LPVOID) data, m_bitmap.biSizeImage, AVIIF_KEYFRAME, NULL, &m_byteBuffer);
+	static bool shown_error = false;
+	if ((w != m_bitmap.biWidth || h != m_bitmap.biHeight) && !shown_error)
+	{
+		PanicAlert("You have resized the window while dumping frames.\n"
+			"Nothing sane can be done to handle this.\n"
+			"Your video will likely be broken.");
+		shown_error = true;
+
+		m_bitmap.biWidth = w;
+		m_bitmap.biHeight = h;
+	}
+
+	AVIStreamWrite(m_streamCompressed, ++m_frameCount, 1, const_cast<u8*>(data), m_bitmap.biSizeImage, AVIIF_KEYFRAME, NULL, &m_byteBuffer);
 	m_totalBytes += m_byteBuffer;
 	// Close the recording if the file is more than 2gb
 	// VfW can't properly save files over 2gb in size, but can keep writing to them up to 4gb.
@@ -252,7 +264,7 @@ bool AVIDump::CreateFile()
 	File::CreateFullPath(s_FormatContext->filename);

 	if (!(s_FormatContext->oformat = av_guess_format("avi", NULL, NULL)) ||
-			!(s_Stream = av_new_stream(s_FormatContext, 0)))
+			!(s_Stream = avformat_new_stream(s_FormatContext, codec)))
 	{
 		CloseFile();
 		return false;
@@ -269,7 +281,7 @@ bool AVIDump::CreateFile()
 	s_Stream->codec->pix_fmt = g_Config.bUseFFV1 ? PIX_FMT_BGRA : PIX_FMT_YUV420P;

 	if (!(codec = avcodec_find_encoder(s_Stream->codec->codec_id)) ||
-			(avcodec_open(s_Stream->codec, codec) < 0))
+			(avcodec_open2(s_Stream->codec, codec, NULL) < 0))
 	{
 		CloseFile();
 		return false;
@@ -298,9 +310,9 @@ bool AVIDump::CreateFile()
 	return true;
 }

-void AVIDump::AddFrame(uint8_t *data, int width, int height)
+void AVIDump::AddFrame(const u8* data, int width, int height)
 {
-	avpicture_fill((AVPicture *)s_BGRFrame, data, PIX_FMT_BGR24, width, height);
+	avpicture_fill((AVPicture *)s_BGRFrame, const_cast<u8*>(data), PIX_FMT_BGR24, width, height);

 	// Convert image from BGR24 to desired pixel format, and scale to initial
 	// width and height
@@ -24,6 +24,8 @@
 #include <stdint.h>
 #endif

+#include "CommonTypes.h"
+
 class AVIDump
 {
 	private:
@@ -36,11 +38,11 @@ class AVIDump
 	public:
 #ifdef _WIN32
 		static bool Start(HWND hWnd, int w, int h);
-		static void AddFrame(char *data);
 #else
 		static bool Start(int w, int h);
-		static void AddFrame(uint8_t *data, int width, int height);
 #endif
+		static void AddFrame(const u8* data, int width, int height);
+		
 		static void Stop();
 };

@@ -55,6 +55,11 @@ void GetBPRegInfo(const u8* data, char* name, size_t name_size, char* desc, size
 		snprintf(name, name_size, #reg); \
 		(void)(reg);

+	case BPMEM_GENMODE: // 0x00
+		SetRegName(BPMEM_GENMODE);
+		// TODO: Description
+		break;
+
 	case BPMEM_DISPLAYCOPYFILER: // 0x01
 		// TODO: This is actually the sample pattern used for copies from an antialiased EFB
 		SetRegName(BPMEM_DISPLAYCOPYFILER);
@@ -75,6 +80,28 @@ void GetBPRegInfo(const u8* data, char* name, size_t name_size, char* desc, size
 		}
 		break;

+	case BPMEM_BLENDMODE: // 0x41
+		{
+			SetRegName(BPMEM_BLENDMODE);
+			BlendMode mode; mode.hex = cmddata;
+			const char* dstfactors[] = { "0", "1", "src_color", "1-src_color", "src_alpha", "1-src_alpha", "dst_alpha", "1-dst_alpha" };
+			const char* srcfactors[] = { "0", "1", "dst_color", "1-dst_color", "src_alpha", "1-src_alpha", "dst_alpha", "1-dst_alpha" };
+			const char* logicmodes[] = { "0", "s & d", "s & ~d", "s", "~s & d", "d", "s ^ d", "s | d", "~(s | d)", "~(s ^ d)", "~d", "s | ~d", "~s", "~s | d", "~(s & d)", "1" };
+			snprintf(desc, desc_size, "Enable: %s\n"
+									"Logic ops: %s\n"
+									"Dither: %s\n"
+									"Color write: %s\n"
+									"Alpha write: %s\n"
+									"Dest factor: %s\n"
+									"Source factor: %s\n"
+									"Subtract: %s\n"
+									"Logic mode: %s\n",
+									no_yes[mode.blendenable], no_yes[mode.logicopenable], no_yes[mode.dither],
+									no_yes[mode.colorupdate], no_yes[mode.alphaupdate], dstfactors[mode.dstfactor],
+									srcfactors[mode.srcfactor], no_yes[mode.subtract], logicmodes[mode.logicmode]);
+		}
+		break;
+
 	case BPMEM_EFB_BR: // 0x4A
 		{
 			// TODO: Misleading name, should be BPMEM_EFB_WH instead
@@ -148,6 +175,147 @@ void GetBPRegInfo(const u8* data, char* name, size_t name_size, char* desc, size
 		// TODO: Description
 		break;

-#undef SET_REG_NAME
+	case BPMEM_TEV_COLOR_ENV: // 0xC0
+	case BPMEM_TEV_COLOR_ENV+2:
+	case BPMEM_TEV_COLOR_ENV+4:
+	case BPMEM_TEV_COLOR_ENV+8:
+	case BPMEM_TEV_COLOR_ENV+10:
+	case BPMEM_TEV_COLOR_ENV+12:
+	case BPMEM_TEV_COLOR_ENV+14:
+	case BPMEM_TEV_COLOR_ENV+16:
+	case BPMEM_TEV_COLOR_ENV+18:
+	case BPMEM_TEV_COLOR_ENV+20:
+	case BPMEM_TEV_COLOR_ENV+22:
+	case BPMEM_TEV_COLOR_ENV+24:
+	case BPMEM_TEV_COLOR_ENV+26:
+	case BPMEM_TEV_COLOR_ENV+28:
+	case BPMEM_TEV_COLOR_ENV+30:
+		{
+			SetRegName(BPMEM_TEV_COLOR_ENV);
+			TevStageCombiner::ColorCombiner cc; cc.hex = cmddata;
+			const char* tevin[] =
+			{
+				"prev.rgb", "prev.aaa",
+				"c0.rgb", "c0.aaa",
+				"c1.rgb", "c1.aaa",
+				"c2.rgb", "c2.aaa",
+				"tex.rgb", "tex.aaa",
+				"ras.rgb", "ras.aaa",
+				"ONE", "HALF", "konst.rgb", "ZERO",
+			};
+			const char* tevbias[] = { "0", "+0.5", "-0.5", "compare" };
+			const char* tevop[] = { "add", "sub" };
+			const char* tevscale[] = { "1", "2", "4", "0.5" };
+			const char* tevout[] = { "prev.rgb", "c0.rgb", "c1.rgb", "c2.rgb" };
+			snprintf(desc, desc_size, "tev stage: %d\n"
+									"a: %s\n"
+									"b: %s\n"
+									"c: %s\n"
+									"d: %s\n"
+									"bias: %s\n"
+									"op: %s\n"
+									"clamp: %s\n"
+									"scale factor: %s\n"
+									"dest: %s\n",
+									(data[0] - BPMEM_TEV_COLOR_ENV)/2, tevin[cc.a], tevin[cc.b], tevin[cc.c], tevin[cc.d],
+									tevbias[cc.bias], tevop[cc.op], no_yes[cc.clamp], tevscale[cc.shift], tevout[cc.dest]);
+			break;
+		}
+
+	case BPMEM_TEV_ALPHA_ENV: // 0xC1
+	case BPMEM_TEV_ALPHA_ENV+2:
+	case BPMEM_TEV_ALPHA_ENV+4:
+	case BPMEM_TEV_ALPHA_ENV+6:
+	case BPMEM_TEV_ALPHA_ENV+8:
+	case BPMEM_TEV_ALPHA_ENV+10:
+	case BPMEM_TEV_ALPHA_ENV+12:
+	case BPMEM_TEV_ALPHA_ENV+14:
+	case BPMEM_TEV_ALPHA_ENV+16:
+	case BPMEM_TEV_ALPHA_ENV+18:
+	case BPMEM_TEV_ALPHA_ENV+20:
+	case BPMEM_TEV_ALPHA_ENV+22:
+	case BPMEM_TEV_ALPHA_ENV+24:
+	case BPMEM_TEV_ALPHA_ENV+26:
+	case BPMEM_TEV_ALPHA_ENV+28:
+	case BPMEM_TEV_ALPHA_ENV+30:
+		{
+			SetRegName(BPMEM_TEV_ALPHA_ENV);
+			TevStageCombiner::AlphaCombiner ac; ac.hex = cmddata;
+			const char* tevin[] =
+			{
+				"prev", "c0", "c1", "c2",
+				"tex", "ras", "konst", "ZERO",
+			};
+			const char* tevbias[] = { "0", "+0.5", "-0.5", "compare" };
+			const char* tevop[] = { "add", "sub" };
+			const char* tevscale[] = { "1", "2", "4", "0.5" };
+			const char* tevout[] = { "prev", "c0", "c1", "c2" };
+			snprintf(desc, desc_size, "tev stage: %d\n"
+									"a: %s\n"
+									"b: %s\n"
+									"c: %s\n"
+									"d: %s\n"
+									"bias: %s\n"
+									"op: %s\n"
+									"clamp: %s\n"
+									"scale factor: %s\n"
+									"dest: %s\n"
+									"ras sel: %d\n"
+									"tex sel: %d\n",
+									(data[0] - BPMEM_TEV_ALPHA_ENV)/2, tevin[ac.a], tevin[ac.b], tevin[ac.c], tevin[ac.d],
+									tevbias[ac.bias], tevop[ac.op], no_yes[ac.clamp], tevscale[ac.shift], tevout[ac.dest],
+									ac.rswap, ac.tswap);
+			break;
+		}
+
+		case BPMEM_ALPHACOMPARE: // 0xF3
+			{
+				SetRegName(BPMEM_ALPHACOMPARE);
+				AlphaTest test; test.hex = cmddata;
+				const char* functions[] = { "NEVER", "LESS", "EQUAL", "LEQUAL", "GREATER", "NEQUAL", "GEQUAL", "ALWAYS" };
+				const char* logic[] = { "AND", "OR", "XOR", "XNOR" };
+				snprintf(desc, desc_size, "test 1: %s (ref: %#02x)\n"
+										"test 2: %s (ref: %#02x)\n"
+										"logic: %s\n",
+										functions[test.comp0], test.ref0, functions[test.comp1], test.ref1, logic[test.logic]);
+				break;
+			}
+
+#undef SetRegName
 	}
 }
+
+AlphaTest::TEST_RESULT AlphaTest::TestResult()
+{
+	switch(logic)
+	{
+	case 0: // AND
+		if (comp0 == ALPHACMP_ALWAYS && comp1 == ALPHACMP_ALWAYS)
+			return PASS;
+		if (comp0 == ALPHACMP_NEVER || comp1 == ALPHACMP_NEVER)
+			return FAIL;
+		break;
+
+	case 1: // OR
+		if (comp0 == ALPHACMP_ALWAYS || comp1 == ALPHACMP_ALWAYS)
+			return PASS;
+		if (comp0 == ALPHACMP_NEVER && comp1 == ALPHACMP_NEVER)
+			return FAIL;
+		break;
+
+	case 2: // XOR
+		if ((comp0 == ALPHACMP_ALWAYS && comp1 == ALPHACMP_NEVER) || (comp0 == ALPHACMP_NEVER && comp1 == ALPHACMP_ALWAYS))
+			return PASS;
+		if ((comp0 == ALPHACMP_ALWAYS && comp1 == ALPHACMP_ALWAYS) || (comp0 == ALPHACMP_NEVER && comp1 == ALPHACMP_NEVER))
+			return FAIL;
+		break;
+
+	case 3: // XNOR
+		if ((comp0 == ALPHACMP_ALWAYS && comp1 == ALPHACMP_NEVER) || (comp0 == ALPHACMP_NEVER && comp1 == ALPHACMP_ALWAYS))
+			return FAIL;
+		if ((comp0 == ALPHACMP_ALWAYS && comp1 == ALPHACMP_ALWAYS) || (comp0 == ALPHACMP_NEVER && comp1 == ALPHACMP_NEVER))
+			return PASS;
+		break;
+	}
+	return UNDETERMINED;
+}
@@ -793,7 +793,7 @@ union PE_CONTROL
    {
        u32 pixel_format : 3;	// PIXELFMT_X
        u32 zformat : 3;		// Z Compression for 16bit Z format
-        u32 zcomploc : 1;		// 1: before tex stage
+        u32 early_ztest : 1;	// 1: before tex stage
        u32 unused : 17;
        u32 rid : 8;
    };
@@ -857,7 +857,7 @@ union TevKSel
    int getKA(int i) {return i?kasel1:kasel0;}
 };

-union AlphaFunc
+union AlphaTest
 {
    struct
    {
@@ -868,6 +868,15 @@ union AlphaFunc
        u32 logic : 2;
    };
    u32 hex;
+
+	enum TEST_RESULT
+	{
+		UNDETERMINED = 0,
+		FAIL = 1,
+		PASS = 2,
+	};
+
+	TEST_RESULT TestResult();
 };

 union UPE_Copy
@@ -981,7 +990,7 @@ struct BPMemory
    TevReg tevregs[4];  //0xE0
    FogRangeParams fogRange;
    FogParams fog; //0xEE,0xEF,0xF0,0xF1,0xF2
-    AlphaFunc alphaFunc; //0xF3
+    AlphaTest alpha_test; //0xF3
    ZTex1 ztex1; //0xf4,0xf5
    ZTex2 ztex2;
    TevKSel tevksel[8];//0xf6,0xf7,f8,f9,fa,fb,fc,fd
@@ -35,9 +35,9 @@

 using namespace BPFunctions;

-u32 mapTexAddress;
-bool mapTexFound;
-int numWrites;
+static u32 mapTexAddress;
+static bool mapTexFound;
+static int numWrites;

 extern volatile bool g_bSkipCurrentFrame;

@@ -81,6 +81,9 @@ void BPWritten(const BPCmd& bp)
 		  just stuff geometry in them and don't put state changes there
 	----------------------------------------------------------------------------------------------------------------
 	*/
+	
+	// check for invalid state, else unneeded configuration are built
+	g_video_backend->CheckInvalidState();

 	// Debugging only, this lets you skip a bp update
 	//static int times = 0;
@@ -359,9 +362,9 @@ void BPWritten(const BPCmd& bp)
 			PixelShaderManager::SetFogColorChanged();
 		break;
 	case BPMEM_ALPHACOMPARE: // Compare Alpha Values
-		PRIM_LOG("alphacmp: ref0=%d, ref1=%d, comp0=%d, comp1=%d, logic=%d", bpmem.alphaFunc.ref0,
-				bpmem.alphaFunc.ref1, bpmem.alphaFunc.comp0, bpmem.alphaFunc.comp1, bpmem.alphaFunc.logic);
-		PixelShaderManager::SetAlpha(bpmem.alphaFunc);
+		PRIM_LOG("alphacmp: ref0=%d, ref1=%d, comp0=%d, comp1=%d, logic=%d", bpmem.alpha_test.ref0,
+				bpmem.alpha_test.ref1, bpmem.alpha_test.comp0, bpmem.alpha_test.comp1, bpmem.alpha_test.logic);
+		PixelShaderManager::SetAlpha(bpmem.alpha_test);
 		break;
 	case BPMEM_BIAS: // BIAS
 		PRIM_LOG("ztex bias=0x%x", bpmem.ztex1.bias);
@@ -458,6 +461,7 @@ void BPWritten(const BPCmd& bp)

 	case BPMEM_ZCOMPARE:      // Set the Z-Compare and EFB pixel format
 		g_renderer->SetColorMask(); // alpha writing needs to be disabled if the new pixel format doesn't have an alpha channel
+		g_renderer->SetBlendMode(true);
 		OnPixelFormatChange();
 		break;

@@ -500,18 +504,39 @@ void BPWritten(const BPCmd& bp)
 		// if this is different from 0, manual TMEM management is used (GX_PreloadEntireTexture).
 		if (bp.newvalue != 0)
 		{
-			// NOTE(neobrain): Apparently tmemodd doesn't affect hardware behavior at all (libogc uses it just as a buffer and switches its contents with tmemeven whenever this is called)
+			// TODO: Not quite sure if this is completely correct (likely not)
+			// NOTE: libogc's implementation of GX_PreloadEntireTexture seems flawed, so it's not necessarily a good reference for RE'ing this feature.
+
 			BPS_TmemConfig& tmem_cfg = bpmem.tmem_config;
-			u8* ram_ptr = Memory::GetPointer(tmem_cfg.preload_addr << 5);
-			u32 tmem_addr = tmem_cfg.preload_tmem_even * TMEM_LINE_SIZE;
-			u32 size = tmem_cfg.preload_tile_info.count * 32;
+			u8* src_ptr = Memory::GetPointer(tmem_cfg.preload_addr << 5); // TODO: Should we add mask here on GC?
+			u32 size = tmem_cfg.preload_tile_info.count * TMEM_LINE_SIZE;
+			u32 tmem_addr_even = tmem_cfg.preload_tmem_even * TMEM_LINE_SIZE;

-			// Check if the game has overflowed TMEM, and copy up to the limit.
-			// Paper Mario does this when entering the Great Boogly Tree (Chap 2)
-			if ((tmem_addr + size) > TMEM_SIZE)
-				size = TMEM_SIZE - tmem_addr;
+			if (tmem_cfg.preload_tile_info.type != 3)
+			{
+				if (tmem_addr_even + size > TMEM_SIZE)
+					size = TMEM_SIZE - tmem_addr_even;

-			memcpy(texMem + tmem_addr, ram_ptr, size);
+				memcpy(texMem + tmem_addr_even, src_ptr, size);
+			}
+			else // RGBA8 tiles (and CI14, but that might just be stupid libogc!)
+			{
+				// AR and GB tiles are stored in separate TMEM banks => can't use a single memcpy for everything
+				u32 tmem_addr_odd = tmem_cfg.preload_tmem_odd * TMEM_LINE_SIZE;
+
+				for (u32 i = 0; i < tmem_cfg.preload_tile_info.count; ++i)
+				{
+					if (tmem_addr_even + TMEM_LINE_SIZE > TMEM_SIZE ||
+						tmem_addr_odd  + TMEM_LINE_SIZE > TMEM_SIZE)
+						break;
+
+					memcpy(texMem + tmem_addr_even, src_ptr, TMEM_LINE_SIZE);
+					memcpy(texMem + tmem_addr_odd, src_ptr + TMEM_LINE_SIZE, TMEM_LINE_SIZE);
+					tmem_addr_even += TMEM_LINE_SIZE;
+					tmem_addr_odd += TMEM_LINE_SIZE;
+					src_ptr += TMEM_LINE_SIZE * 2;
+				}
+			}
 		}
 		break;

@@ -72,21 +72,9 @@ void GFXDebuggerCheckAndPause(bool update);
 void GFXDebuggerToPause(bool update);
 void GFXDebuggerUpdateScreen();

-#undef ENABLE_GFX_DEBUGGER
-#if defined(_DEBUG) || defined(DEBUGFAST)
-#define ENABLE_GFX_DEBUGGER
-#endif
-
-#ifdef ENABLE_GFX_DEBUGGER
 #define GFX_DEBUGGER_PAUSE_AT(event,update) {if (((GFXDebuggerToPauseAtNext & event) && --GFXDebuggerEventToPauseCount<=0) || GFXDebuggerPauseFlag) GFXDebuggerToPause(update);}
 #define GFX_DEBUGGER_PAUSE_LOG_AT(event,update,dumpfunc) {if (((GFXDebuggerToPauseAtNext & event) && --GFXDebuggerEventToPauseCount<=0) || GFXDebuggerPauseFlag) {{dumpfunc};GFXDebuggerToPause(update);}}
 #define GFX_DEBUGGER_LOG_AT(event,dumpfunc) {if (( GFXDebuggerToPauseAtNext & event ) ) {{dumpfunc};}}
-#else
-// Disable debugging calls in Release build
-#define GFX_DEBUGGER_PAUSE_AT(event,update)
-#define GFX_DEBUGGER_PAUSE_LOG_AT(event,update,dumpfunc)
-#define GFX_DEBUGGER_LOG_AT(event,dumpfunc)
-#endif // ENABLE_GFX_DEBUGGER


 #endif // _GFX_DEBUGGER_H_
@@ -205,10 +205,6 @@ LRESULT CALLBACK WndProc( HWND hWnd, UINT iMsg, WPARAM wParam, LPARAM lParam )
 			OnKeyDown(lParam);
 			FreeLookInput((u32)wParam, lParam);
 		}
-		else if (wParam == WIIMOTE_DISCONNECT)
-		{
-			PostMessage(m_hParent, WM_USER, wParam, lParam);
-		}
 		break;

 	// Called when a screensaver wants to show up while this window is active
@@ -0,0 +1,64 @@
+// Copyright (C) 2003 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 "FPSCounter.h"
+#include "FileUtil.h"
+#include "Timer.h"
+#include "VideoConfig.h"
+
+#define FPS_REFRESH_INTERVAL 1000
+
+static unsigned int s_counter = 0;
+static unsigned int s_fps = 0;
+static unsigned int s_fps_last_counter = 0;
+static unsigned long s_last_update_time = 0;
+static File::IOFile s_bench_file;
+
+void InitFPSCounter()
+{
+	s_counter = s_fps_last_counter = 0;
+	s_fps = 0;
+	s_last_update_time = Common::Timer::GetTimeMs();
+
+	if (s_bench_file.IsOpen())
+		s_bench_file.Close();
+}
+
+static void LogFPSToFile(unsigned long val)
+{
+	if (!s_bench_file.IsOpen())
+		s_bench_file.Open(File::GetUserPath(D_LOGS_IDX) + "fps.txt", "w");
+
+	char buffer[256];
+	snprintf(buffer, 256, "%ld\n", val);
+	s_bench_file.WriteArray(buffer, strlen(buffer));
+}
+
+int UpdateFPSCounter()
+{
+	if (Common::Timer::GetTimeMs() - s_last_update_time >= FPS_REFRESH_INTERVAL)
+	{
+		s_last_update_time = Common::Timer::GetTimeMs();
+		s_fps = s_counter - s_fps_last_counter;
+		s_fps_last_counter = s_counter;
+		if (g_ActiveConfig.bLogFPSToFile)
+			LogFPSToFile(s_fps);
+	}
+
+	s_counter++;
+	return s_fps;
+}
@@ -0,0 +1,28 @@
+// Copyright (C) 2003 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 _FPS_COUNTER_H_
+#define _FPS_COUNTER_H_
+
+// Initializes the FPS counter.
+void InitFPSCounter();
+
+// Called when a frame is rendered. Returns the value to be displayed on
+// screen as the FPS counter (updated every second).
+int UpdateFPSCounter();
+
+#endif // _FPS_COUNTER_H_
@@ -16,7 +16,6 @@
 // http://code.google.com/p/dolphin-emu/

 #include "VideoConfig.h"
-#include "Setup.h"
 #include "MemoryUtil.h"
 #include "Thread.h"
 #include "Atomic.h"
@@ -223,11 +222,11 @@ void RunGpu()
 		{
 			u8 *uData = Memory::GetPointer(fifo.CPReadPointer);			
 			
-			SaveSSEState();
-			LoadDefaultSSEState();
+			FPURoundMode::SaveSIMDState();
+			FPURoundMode::LoadDefaultSIMDState();
 			ReadDataFromFifo(uData, 32);				
 			OpcodeDecoder_Run(g_bSkipCurrentFrame);	
-			LoadSSEState();
+			FPURoundMode::LoadSIMDState();

 			//DEBUG_LOG(COMMANDPROCESSOR, "Fifo wraps to base");

@@ -10,6 +10,9 @@ XFBSourceBase *FramebufferManagerBase::m_realXFBSource; // Only used in Real XFB
 FramebufferManagerBase::VirtualXFBListType FramebufferManagerBase::m_virtualXFBList; // Only used in Virtual XFB mode
 const XFBSourceBase* FramebufferManagerBase::m_overlappingXFBArray[MAX_VIRTUAL_XFB];

+unsigned int FramebufferManagerBase::s_last_xfb_width = 1;
+unsigned int FramebufferManagerBase::s_last_xfb_height = 1;
+
 FramebufferManagerBase::FramebufferManagerBase()
 {
 	m_realXFBSource = NULL;
@@ -226,3 +229,31 @@ void FramebufferManagerBase::ReplaceVirtualXFB()
 		}
 	}
 }
+
+int FramebufferManagerBase::ScaleToVirtualXfbWidth(int x, unsigned int backbuffer_width)
+{
+	if (g_ActiveConfig.RealXFBEnabled())
+		return x;
+
+	if (g_ActiveConfig.b3DVision)
+	{
+		// This works, yet the version in the else doesn't. No idea why.
+		return x * (int)backbuffer_width / (int)FramebufferManagerBase::LastXfbWidth();
+	}
+	else
+		return x * (int)Renderer::GetTargetRectangle().GetWidth() / (int)FramebufferManagerBase::LastXfbWidth();
+}
+
+int FramebufferManagerBase::ScaleToVirtualXfbHeight(int y, unsigned int backbuffer_height)
+{
+	if (g_ActiveConfig.RealXFBEnabled())
+		return y;
+
+	if (g_ActiveConfig.b3DVision)
+	{
+		// This works, yet the version in the else doesn't. No idea why.
+		return y * (int)backbuffer_height / (int)FramebufferManagerBase::LastXfbHeight();
+	}
+	else
+		return y * (int)Renderer::GetTargetRectangle().GetHeight() / (int)FramebufferManagerBase::LastXfbHeight();
+}
@@ -50,6 +50,14 @@ public:
 	static void CopyToXFB(u32 xfbAddr, u32 fbWidth, u32 fbHeight, const EFBRectangle& sourceRc,float Gamma);
 	static const XFBSourceBase* const* GetXFBSource(u32 xfbAddr, u32 fbWidth, u32 fbHeight, u32 &xfbCount);

+	static void SetLastXfbWidth(unsigned int width) { s_last_xfb_width = width; }
+	static void SetLastXfbHeight(unsigned int height) { s_last_xfb_height = height; }
+	static unsigned int LastXfbWidth() { return s_last_xfb_width; }
+	static unsigned int LastXfbHeight() { return s_last_xfb_height; }
+
+	static int ScaleToVirtualXfbWidth(int x, unsigned int backbuffer_width);
+	static int ScaleToVirtualXfbHeight(int y, unsigned int backbuffer_height);
+
 protected:
 	struct VirtualXFB
 	{
@@ -85,6 +93,9 @@ private:
 	static VirtualXFBListType m_virtualXFBList; // Only used in Virtual XFB mode

 	static const XFBSourceBase* m_overlappingXFBArray[MAX_VIRTUAL_XFB];
+
+	static unsigned int s_last_xfb_width;
+	static unsigned int s_last_xfb_height;
 };

 extern FramebufferManagerBase *g_framebuffer_manager;
@@ -0,0 +1,52 @@
+// Copyright (C) 2003-2009 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/
+
+// TODO: Handle cache-is-full condition :p
+
+
+#include "Common.h"
+#include "DLCache.h"
+
+namespace DLCache
+{
+	
+void Init()
+{
+}
+
+void Shutdown()
+{
+}
+
+void Clear() 
+{	
+}
+
+void ProgressiveCleanup()
+{
+}
+}  // namespace
+
+// NOTE - outside the namespace on purpose.
+bool HandleDisplayList(u32 address, u32 size)
+{
+	return false;
+}
+
+void IncrementCheckContextId()
+{
+}
@@ -23,7 +23,7 @@

 int GetLightingShaderId(u32* out)
 {
-	for (int i = 0; i < xfregs.numChan.numColorChans; ++i)
+	for (u32 i = 0; i < xfregs.numChan.numColorChans; ++i)
 	{
 		out[i] = xfregs.color[i].enablelighting ?
 			(u32)xfregs.color[i].hex :
@@ -222,11 +222,17 @@ void VideoBackendHardware::InitializeShared()
 	memset((void*)&s_beginFieldArgs, 0, sizeof(s_beginFieldArgs));
 	memset(&s_accessEFBArgs, 0, sizeof(s_accessEFBArgs));
 	s_AccessEFBResult = 0;
+	m_invalid = false;
 }

 // Run from the CPU thread
 void VideoBackendHardware::DoState(PointerWrap& p)
 {
+	bool software = false;
+	p.Do(software);
+	if (p.GetMode() == PointerWrap::MODE_READ && software == true)
+		// change mode to abort load of incompatible save state.
+		p.SetMode(PointerWrap::MODE_VERIFY);
 	VideoCommon_DoState(p);
 	p.DoMarker("VideoCommon");

@@ -240,16 +246,25 @@ void VideoBackendHardware::DoState(PointerWrap& p)
 	// Refresh state.
 	if (p.GetMode() == PointerWrap::MODE_READ)
 	{
-		BPReload();
+		m_invalid = true;
 		RecomputeCachedArraybases();

 		// Clear all caches that touch RAM
 		// (? these don't appear to touch any emulation state that gets saved. moved to on load only.)
-		TextureCache::Invalidate();
 		VertexLoaderManager::MarkAllDirty();
 	}
 }

+void VideoBackendHardware::CheckInvalidState() {
+	if (m_invalid)
+	{
+		m_invalid = false;
+		
+		BPReload();
+		TextureCache::Invalidate();
+	}
+}
+
 void VideoBackendHardware::PauseAndLock(bool doLock, bool unpauseOnUnlock)
 {
 	Fifo_PauseAndLock(doLock, unpauseOnUnlock);
@@ -19,6 +19,7 @@

 #include "Common.h"

+#include "ConfigManager.h"
 #include "OnScreenDisplay.h"
 #include "RenderBase.h"
 #include "Timer.h"
@@ -47,6 +48,8 @@ void AddMessage(const char* pstr, u32 ms)

 void DrawMessages()
 {
+	if(!SConfig::GetInstance().m_LocalCoreStartupParameter.bOnScreenDisplayMessages) return;
+
 	if (s_listMsgs.size() > 0)
 	{
 		int left = 25, top = 15;
@@ -27,9 +27,8 @@
 #include "VideoConfig.h"
 #include "NativeVertexFormat.h"

-static int AlphaPreTest();

-static void StageHash(int stage, u32* out)
+static void StageHash(u32 stage, u32* out)
 {
 	out[0] |= bpmem.combiners[stage].colorC.hex & 0xFFFFFF; // 24
 	u32 alphaC = bpmem.combiners[stage].alphaC.hex & 0xFFFFF0; // 24, strip out tswap and rswap for now
@@ -109,35 +108,19 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode, u32 compo
 	uid->values[0] |= bpmem.genMode.numtexgens << 4; // 4
 	uid->values[0] |= dstAlphaMode << 8; // 2

-	bool DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth;
-
-	uid->values[0] |= DepthTextureEnable << 10; // 1
-
 	bool enablePL = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting;
-	uid->values[0] |= enablePL << 11; // 1
+	uid->values[0] |= enablePL << 10; // 1

-	if (!enablePL) uid->values[0] |= xfregs.numTexGen.numTexGens << 12; // 4
-	u32 alphaPreTest = AlphaPreTest()+1;
+	if (!enablePL) uid->values[0] |= xfregs.numTexGen.numTexGens << 11; // 4

-	uid->values[0] |= alphaPreTest << 16; // 2
-
-	if (alphaPreTest == 1 || (alphaPreTest && !DepthTextureEnable && dstAlphaMode == DSTALPHA_ALPHA_PASS))
-	{
-		// Courtesy of PreAlphaTest, we're done already ;)
-		// NOTE: The comment header of generated shaders depends on the value of bpmem.genmode.numindstages.. shouldnt really bother about that though.
-		uid->num_values = 1;
-		return;
-	}
+	AlphaTest::TEST_RESULT alphaPreTest = bpmem.alpha_test.TestResult();
+	uid->values[0] |= alphaPreTest << 15; // 2

+	// numtexgens should be <= 8
 	for (unsigned int i = 0; i < bpmem.genMode.numtexgens; ++i)
-	{
-		if (18+i < 32)
-			uid->values[0] |= xfregs.texMtxInfo[i].projection << (18+i); // 1
-		else
-			uid->values[1] |= xfregs.texMtxInfo[i].projection << (i - 14); // 1
-	}
+		uid->values[0] |= xfregs.texMtxInfo[i].projection << (17+i); // 1

-	uid->values[1] = bpmem.genMode.numindstages << 2; // 3
+	uid->values[1] = bpmem.genMode.numindstages; // 3
 	u32 indirectStagesUsed = 0;
 	for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i)
 		if (bpmem.tevind[i].IsActive() && bpmem.tevind[i].bt < bpmem.genMode.numindstages)
@@ -145,43 +128,37 @@ void GetPixelShaderId(PIXELSHADERUID *uid, DSTALPHA_MODE dstAlphaMode, u32 compo

 	assert(indirectStagesUsed == (indirectStagesUsed & 0xF));

-	uid->values[1] |= indirectStagesUsed << 5; // 4;
+	uid->values[1] |= indirectStagesUsed << 3; // 4;

 	for (unsigned int i = 0; i < bpmem.genMode.numindstages; ++i)
 	{
 		if (indirectStagesUsed & (1 << i))
 		{
-			uid->values[1] |= (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens) << (9 + 3*i); // 1
+			uid->values[1] |= (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens) << (7 + 3*i); // 1
 			if (bpmem.tevindref.getTexCoord(i) < bpmem.genMode.numtexgens)
-				uid->values[1] |= bpmem.tevindref.getTexCoord(i) << (10 + 3*i); // 2
+				uid->values[1] |= bpmem.tevindref.getTexCoord(i) << (8 + 3*i); // 2
 		}
 	}

 	u32* ptr = &uid->values[2];
-	for (unsigned int i = 0; i < bpmem.genMode.numtevstages+1; ++i)
+	for (unsigned int i = 0; i < bpmem.genMode.numtevstages+1u; ++i)
 	{
 		StageHash(i, ptr);
 		ptr += 4; // max: ptr = &uid->values[66]
 	}

-	ptr[0] |= bpmem.alphaFunc.comp0; // 3
-	ptr[0] |= bpmem.alphaFunc.comp1 << 3; // 3
-	ptr[0] |= bpmem.alphaFunc.logic << 6; // 2
+	ptr[0] |= bpmem.alpha_test.comp0; // 3
+	ptr[0] |= bpmem.alpha_test.comp1 << 3; // 3
+	ptr[0] |= bpmem.alpha_test.logic << 6; // 2

-	if (alphaPreTest == 0 || alphaPreTest == 2)
-	{
-		ptr[0] |= bpmem.fog.c_proj_fsel.fsel << 8; // 3
-		if (DepthTextureEnable)
-		{
-			ptr[0] |= bpmem.ztex2.op << 11; // 2
-			ptr[0] |= bpmem.zcontrol.zcomploc << 13; // 1
-			ptr[0] |= bpmem.zmode.testenable << 14; // 1
-			ptr[0] |= bpmem.zmode.updateenable << 15; // 1
-		}
-	}
+	ptr[0] |= bpmem.ztex2.op << 8; // 2
+	ptr[0] |= bpmem.zcontrol.early_ztest << 10; // 1
+	ptr[0] |= bpmem.zmode.testenable << 11; // 1
+	ptr[0] |= bpmem.zmode.updateenable << 12; // 1

 	if (dstAlphaMode != DSTALPHA_ALPHA_PASS)
 	{
+		ptr[0] |= bpmem.fog.c_proj_fsel.fsel << 13; // 3
 		if (bpmem.fog.c_proj_fsel.fsel != 0)
 		{
 			ptr[0] |= bpmem.fog.c_proj_fsel.proj << 16; // 1
@@ -208,9 +185,8 @@ void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode, u
 	*ptr++ = bpmem.ztex2.hex; // 2
 	*ptr++ = bpmem.zcontrol.hex; // 3
 	*ptr++ = bpmem.zmode.hex; // 4
-	*ptr++ = g_ActiveConfig.bEnablePerPixelDepth; // 5
-	*ptr++ = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; // 6
-	*ptr++ = xfregs.numTexGen.hex; // 7
+	*ptr++ = g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting; // 5
+	*ptr++ = xfregs.numTexGen.hex; // 6

 	if (g_ActiveConfig.bEnablePixelLighting && g_ActiveConfig.backend_info.bSupportsPixelLighting)
 	{
@@ -222,28 +198,28 @@ void GetSafePixelShaderId(PIXELSHADERUIDSAFE *uid, DSTALPHA_MODE dstAlphaMode, u
 	}

 	for (unsigned int i = 0; i < 8; ++i)
-		*ptr++ = xfregs.texMtxInfo[i].hex; // 8-15
+		*ptr++ = xfregs.texMtxInfo[i].hex; // 7-14

 	for (unsigned int i = 0; i < 16; ++i)
-		*ptr++ = bpmem.tevind[i].hex; // 16-31
+		*ptr++ = bpmem.tevind[i].hex; // 15-30

-	*ptr++ = bpmem.tevindref.hex; // 32
+	*ptr++ = bpmem.tevindref.hex; // 31

-	for (int i = 0; i < bpmem.genMode.numtevstages+1; ++i) // up to 16 times
+	for (unsigned int i = 0; i < bpmem.genMode.numtevstages+1u; ++i) // up to 16 times
 	{
-		*ptr++ = bpmem.combiners[i].colorC.hex; // 33+5*i
-		*ptr++ = bpmem.combiners[i].alphaC.hex; // 34+5*i
-		*ptr++ = bpmem.tevind[i].hex; // 35+5*i
-		*ptr++ = bpmem.tevksel[i/2].hex; // 36+5*i
-		*ptr++ = bpmem.tevorders[i/2].hex; // 37+5*i
+		*ptr++ = bpmem.combiners[i].colorC.hex; // 32+5*i
+		*ptr++ = bpmem.combiners[i].alphaC.hex; // 33+5*i
+		*ptr++ = bpmem.tevind[i].hex; // 34+5*i
+		*ptr++ = bpmem.tevksel[i/2].hex; // 35+5*i
+		*ptr++ = bpmem.tevorders[i/2].hex; // 36+5*i
 	}

-	ptr = &uid->values[113];
+	ptr = &uid->values[112];

-	*ptr++ = bpmem.alphaFunc.hex; // 113
+	*ptr++ = bpmem.alpha_test.hex; // 112

-	*ptr++ = bpmem.fog.c_proj_fsel.hex; // 114
-	*ptr++ = bpmem.fogRange.Base.hex; // 115
+	*ptr++ = bpmem.fog.c_proj_fsel.hex; // 113
+	*ptr++ = bpmem.fogRange.Base.hex; // 114

 	_assert_((ptr - uid->values) == uid->GetNumValues());
 }
@@ -312,10 +288,10 @@ static const char *tevKSelTableC[] = // KCSEL
 	"0.375f,0.375f,0.375f", // 3_8 = 0x05
 	"0.25f,0.25f,0.25f",    // 1_4 = 0x06
 	"0.125f,0.125f,0.125f", // 1_8 = 0x07
-	"ERROR", // 0x08
-	"ERROR", // 0x09
-	"ERROR", // 0x0a
-	"ERROR", // 0x0b
+	"ERROR1", // 0x08
+	"ERROR2", // 0x09
+	"ERROR3", // 0x0a
+	"ERROR4", // 0x0b
 	I_KCOLORS"[0].rgb", // K0 = 0x0C
 	I_KCOLORS"[1].rgb", // K1 = 0x0D
 	I_KCOLORS"[2].rgb", // K2 = 0x0E
@@ -348,14 +324,14 @@ static const char *tevKSelTableA[] = // KASEL
 	"0.375f",// 3_8 = 0x05
 	"0.25f", // 1_4 = 0x06
 	"0.125f",// 1_8 = 0x07
-	"ERROR", // 0x08
-	"ERROR", // 0x09
-	"ERROR", // 0x0a
-	"ERROR", // 0x0b
-	"ERROR", // 0x0c
-	"ERROR", // 0x0d
-	"ERROR", // 0x0e
-	"ERROR", // 0x0f
+	"ERROR5", // 0x08
+	"ERROR6", // 0x09
+	"ERROR7", // 0x0a
+	"ERROR8", // 0x0b
+	"ERROR9", // 0x0c
+	"ERROR10", // 0x0d
+	"ERROR11", // 0x0e
+	"ERROR12", // 0x0f
 	I_KCOLORS"[0].r", // K0_R = 0x10
 	I_KCOLORS"[1].r", // K1_R = 0x11
 	I_KCOLORS"[2].r", // K2_R = 0x12
@@ -430,7 +406,7 @@ static const char *tevCInputTable[] = // CC
 	"float3(0.5f, 0.5f, 0.5f)",                 // HALF
 	"(ckonsttemp.rgb)", //"konsttemp.rgb",        // KONST
 	"float3(0.0f, 0.0f, 0.0f)",              // ZERO
-	"PADERROR", "PADERROR", "PADERROR", "PADERROR"
+	"PADERROR1", "PADERROR2", "PADERROR3", "PADERROR4"
 };

 static const char *tevAInputTable[] = // CA
@@ -452,17 +428,17 @@ static const char *tevAInputTable[] = // CA
 	"crastemp",         // RASA,
 	"ckonsttemp",       // KONST,  (hw1 had quarter)
 	"float4(0.0f, 0.0f, 0.0f, 0.0f)", // ZERO
-	"PADERROR", "PADERROR", "PADERROR", "PADERROR",
-	"PADERROR", "PADERROR", "PADERROR", "PADERROR",
+	"PADERROR5", "PADERROR6", "PADERROR7", "PADERROR8",
+	"PADERROR9", "PADERROR10", "PADERROR11", "PADERROR12",
 };

 static const char *tevRasTable[] =
 {
 	"colors_0",
 	"colors_1",
-	"ERROR", //2
-	"ERROR", //3
-	"ERROR", //4
+	"ERROR13", //2
+	"ERROR14", //3
+	"ERROR15", //4
 	"alphabump", // use bump alpha
 	"(alphabump*(255.0f/248.0f))", //normalized
 	"float4(0.0f, 0.0f, 0.0f, 0.0f)", // zero
@@ -485,7 +461,15 @@ static const char *tevIndFmtScale[]   = {"255.0f", "31.0f", "15.0f", "7.0f" };
 static char swapModeTable[4][5];

 static char text[16384];
-static bool DepthTextureEnable;
+
+struct RegisterState
+{
+	bool ColorNeedOverflowControl;
+	bool AlphaNeedOverflowControl;
+	bool AuxStored;
+};
+
+static RegisterState RegisterStates[4];

 static void BuildSwapModeTable()
 {
@@ -523,7 +507,6 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType
 				nIndirectStagesUsed |= 1 << bpmem.tevind[i].bt;
 		}
 	}
-	DepthTextureEnable = (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable) || g_ActiveConfig.bEnablePerPixelDepth ;
 	// Declare samplers

 	if(ApiType != API_D3D11)
@@ -579,14 +562,14 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType
 	{
 		WRITE(p, "  out float4 ocol0 : COLOR0,%s%s\n  in float4 rawpos : %s,\n",
 			dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n  out float4 ocol1 : COLOR1," : "",
-			DepthTextureEnable ? "\n  out float depth : DEPTH," : "",
+			"\n  out float depth : DEPTH,",
 			ApiType & API_OPENGL ? "WPOS" : ApiType & API_D3D9_SM20 ? "POSITION" : "VPOS");
 	}
 	else
 	{
 		WRITE(p, "  out float4 ocol0 : SV_Target0,%s%s\n  in float4 rawpos : SV_Position,\n",
 			dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND ? "\n  out float4 ocol1 : SV_Target1," : "",
-			DepthTextureEnable ? "\n  out float depth : SV_Depth," : "");
+			"\n  out float depth : SV_Depth,");
 	}

 	WRITE(p, "  in float4 colors_0 : COLOR0,\n");
@@ -617,33 +600,6 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType
 	}
 	WRITE(p, "        ) {\n");

-	char* pmainstart = p;
-	int Pretest = AlphaPreTest();
-	if(Pretest >= 0 && !DepthTextureEnable)
-	{
-		if (!Pretest)
-		{
-			// alpha test will always fail, so restart the shader and just make it an empty function		
-			WRITE(p, "ocol0 = 0;\n");
-			if(DepthTextureEnable)
-				WRITE(p, "depth = 1.f;\n");
-			if(dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND)
-					WRITE(p, "ocol1 = 0;\n");
-			WRITE(p, "discard;\n");
-			if(ApiType != API_D3D11)
-				WRITE(p, "return;\n");
-		}
-		else if (dstAlphaMode == DSTALPHA_ALPHA_PASS)
-		{
-			WRITE(p, "  ocol0 = " I_ALPHA"[0].aaaa;\n");
-		}
-		if(!Pretest || dstAlphaMode == DSTALPHA_ALPHA_PASS)
-		{
-			WRITE(p, "}\n");
-			return text;
-		}
-	}
-
 	WRITE(p, "  float4 c0 = " I_COLORS"[1], c1 = " I_COLORS"[2], c2 = " I_COLORS"[3], prev = float4(0.0f, 0.0f, 0.0f, 0.0f), textemp = float4(0.0f, 0.0f, 0.0f, 0.0f), rastemp = float4(0.0f, 0.0f, 0.0f, 0.0f), konsttemp = float4(0.0f, 0.0f, 0.0f, 0.0f);\n"
 			"  float3 comp16 = float3(1.0f, 255.0f, 0.0f), comp24 = float3(1.0f, 255.0f, 255.0f*255.0f);\n"
 			"  float4 alphabump=float4(0.0f,0.0f,0.0f,0.0f);\n"
@@ -717,6 +673,16 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType
 		}
 	}

+	RegisterStates[0].AlphaNeedOverflowControl = false;
+	RegisterStates[0].ColorNeedOverflowControl = false;
+	RegisterStates[0].AuxStored = false;
+	for(int i = 1; i < 4; i++)
+	{
+		RegisterStates[i].AlphaNeedOverflowControl = true;
+		RegisterStates[i].ColorNeedOverflowControl = true;
+		RegisterStates[i].AuxStored = false;
+	}
+
 	for (int i = 0; i < numStages; i++)
 		WriteStage(p, i, ApiType); //build the equation for this stage

@@ -724,40 +690,43 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType
 	{
 		// The results of the last texenv stage are put onto the screen,
 		// regardless of the used destination register
-		WRITE(p, "prev.rgb = %s;\n",tevCOutputTable[bpmem.combiners[numStages-1].colorC.dest]);
-		WRITE(p, "prev.a = %s;\n",tevAOutputTable[bpmem.combiners[numStages-1].alphaC.dest]);
+		if(bpmem.combiners[numStages - 1].colorC.dest != 0)
+		{
+			bool retrieveFromAuxRegister = !RegisterStates[bpmem.combiners[numStages - 1].colorC.dest].ColorNeedOverflowControl && RegisterStates[bpmem.combiners[numStages - 1].colorC.dest].AuxStored;
+			WRITE(p, "prev.rgb = %s%s;\n", retrieveFromAuxRegister ? "c" : "" , tevCOutputTable[bpmem.combiners[numStages - 1].colorC.dest]);
+			RegisterStates[0].ColorNeedOverflowControl = RegisterStates[bpmem.combiners[numStages - 1].colorC.dest].ColorNeedOverflowControl;
+		}
+		if(bpmem.combiners[numStages - 1].alphaC.dest != 0)
+		{
+			bool retrieveFromAuxRegister = !RegisterStates[bpmem.combiners[numStages - 1].alphaC.dest].AlphaNeedOverflowControl && RegisterStates[bpmem.combiners[numStages - 1].alphaC.dest].AuxStored;
+			WRITE(p, "prev.a = %s%s;\n", retrieveFromAuxRegister ? "c" : "" , tevAOutputTable[bpmem.combiners[numStages - 1].alphaC.dest]);
+			RegisterStates[0].AlphaNeedOverflowControl = RegisterStates[bpmem.combiners[numStages - 1].alphaC.dest].AlphaNeedOverflowControl;
+		}
 	}
-	// emulation of unsigned 8 overflow when casting
-	WRITE(p, "prev = frac(4.0f + prev * (255.0f/256.0f)) * (256.0f/255.0f);\n");
+	// emulation of unsigned 8 overflow when casting if needed
+	if(RegisterStates[0].AlphaNeedOverflowControl || RegisterStates[0].ColorNeedOverflowControl)
+		WRITE(p, "prev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n");

-	if(Pretest ==  -1)
-	{
+	AlphaTest::TEST_RESULT Pretest = bpmem.alpha_test.TestResult();
+	if (Pretest == AlphaTest::UNDETERMINED)
 		WriteAlphaTest(p, ApiType, dstAlphaMode);
-	}

-	if((bpmem.fog.c_proj_fsel.fsel != 0) || DepthTextureEnable)
-	{
-		// the screen space depth value = far z + (clip z / clip w) * z range
-		WRITE(p, "float zCoord = " I_ZBIAS"[1].x + (clipPos.z / clipPos.w) * " I_ZBIAS"[1].y;\n");
-	}
+	// the screen space depth value = far z + (clip z / clip w) * z range
+	WRITE(p, "float zCoord = " I_ZBIAS"[1].x + (clipPos.z / clipPos.w) * " I_ZBIAS"[1].y;\n");

-	if (DepthTextureEnable)
+	// Note: depth textures are disabled if early depth test is enabled
+	if (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.early_ztest && bpmem.zmode.testenable)
 	{
 		// use the texture input of the last texture stage (textemp), hopefully this has been read and is in correct format...
-		if (bpmem.ztex2.op != ZTEXTURE_DISABLE && !bpmem.zcontrol.zcomploc && bpmem.zmode.testenable && bpmem.zmode.updateenable)
-		{
-			if (bpmem.ztex2.op == ZTEXTURE_ADD)
-				WRITE(p, "zCoord = dot(" I_ZBIAS"[0].xyzw, textemp.xyzw) + " I_ZBIAS"[1].w + zCoord;\n");
-			else
-				WRITE(p, "zCoord = dot(" I_ZBIAS"[0].xyzw, textemp.xyzw) + " I_ZBIAS"[1].w;\n");
+		WRITE(p, "zCoord = dot(" I_ZBIAS"[0].xyzw, textemp.xyzw) + " I_ZBIAS"[1].w %s;\n",
+									(bpmem.ztex2.op == ZTEXTURE_ADD) ? "+ zCoord" : "");

-			// scale to make result from frac correct
-			WRITE(p, "zCoord = zCoord * (16777215.0f/16777216.0f);\n");
-			WRITE(p, "zCoord = frac(zCoord);\n");
-			WRITE(p, "zCoord = zCoord * (16777216.0f/16777215.0f);\n");
-		}
-		WRITE(p, "depth = zCoord;\n");
+		// scale to make result from frac correct
+		WRITE(p, "zCoord = zCoord * (16777215.0f/16777216.0f);\n");
+		WRITE(p, "zCoord = frac(zCoord);\n");
+		WRITE(p, "zCoord = zCoord * (16777216.0f/16777215.0f);\n");
 	}
+	WRITE(p, "depth = zCoord;\n");

 	if (dstAlphaMode == DSTALPHA_ALPHA_PASS)
 		WRITE(p, "  ocol0 = float4(prev.rgb, " I_ALPHA"[0].a);\n");
@@ -917,7 +886,6 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType)
 	TevStageCombiner::ColorCombiner &cc = bpmem.combiners[n].colorC;
 	TevStageCombiner::AlphaCombiner &ac = bpmem.combiners[n].alphaC;

-	// blah1
 	if(cc.a == TEVCOLORARG_RASA || cc.a == TEVCOLORARG_RASC
 		|| cc.b == TEVCOLORARG_RASA || cc.b == TEVCOLORARG_RASC
 		|| cc.c == TEVCOLORARG_RASA || cc.c == TEVCOLORARG_RASC
@@ -950,7 +918,6 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType)
 		WRITE(p, "textemp = float4(1.0f, 1.0f, 1.0f, 1.0f);\n");


-	// blah2
 	if (cc.a == TEVCOLORARG_KONST || cc.b == TEVCOLORARG_KONST || cc.c == TEVCOLORARG_KONST || cc.d == TEVCOLORARG_KONST
 		|| ac.a == TEVALPHAARG_KONST || ac.b == TEVALPHAARG_KONST || ac.c == TEVALPHAARG_KONST || ac.d == TEVALPHAARG_KONST)
 	{
@@ -971,30 +938,78 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType)
 		|| cc.b == TEVCOLORARG_CPREV || cc.b == TEVCOLORARG_APREV
 		|| cc.c == TEVCOLORARG_CPREV || cc.c == TEVCOLORARG_APREV
 		|| ac.a == TEVALPHAARG_APREV || ac.b == TEVALPHAARG_APREV || ac.c == TEVALPHAARG_APREV)
-		WRITE(p, "cprev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n");
-
+	{
+		if(RegisterStates[0].AlphaNeedOverflowControl || RegisterStates[0].ColorNeedOverflowControl)
+		{
+			WRITE(p, "cprev = frac(prev * (255.0f/256.0f)) * (256.0f/255.0f);\n");
+			RegisterStates[0].AlphaNeedOverflowControl = false;
+			RegisterStates[0].ColorNeedOverflowControl = false;
+		}
+		else
+		{
+			WRITE(p, "cprev = prev;\n");
+		}
+		RegisterStates[0].AuxStored = true;
+	}

 	if(cc.a == TEVCOLORARG_C0 || cc.a == TEVCOLORARG_A0
 	|| cc.b == TEVCOLORARG_C0 || cc.b == TEVCOLORARG_A0
 	|| cc.c == TEVCOLORARG_C0 || cc.c == TEVCOLORARG_A0
 	|| ac.a == TEVALPHAARG_A0 || ac.b == TEVALPHAARG_A0 || ac.c == TEVALPHAARG_A0)
-		WRITE(p, "cc0 = frac(c0 * (255.0f/256.0f)) * (256.0f/255.0f);\n");
-
+	{
+		if(RegisterStates[1].AlphaNeedOverflowControl || RegisterStates[1].ColorNeedOverflowControl)
+		{
+			WRITE(p, "cc0 = frac(c0 * (255.0f/256.0f)) * (256.0f/255.0f);\n");
+			RegisterStates[1].AlphaNeedOverflowControl = false;
+			RegisterStates[1].ColorNeedOverflowControl = false;
+		}
+		else
+		{
+			WRITE(p, "cc0 = c0;\n");
+		}
+		RegisterStates[1].AuxStored = true;
+	}

 	if(cc.a == TEVCOLORARG_C1 || cc.a == TEVCOLORARG_A1
 	|| cc.b == TEVCOLORARG_C1 || cc.b == TEVCOLORARG_A1
 	|| cc.c == TEVCOLORARG_C1 || cc.c == TEVCOLORARG_A1
 	|| ac.a == TEVALPHAARG_A1 || ac.b == TEVALPHAARG_A1 || ac.c == TEVALPHAARG_A1)
-		WRITE(p, "cc1 = frac(c1 * (255.0f/256.0f)) * (256.0f/255.0f);\n");
-
+	{
+		if(RegisterStates[2].AlphaNeedOverflowControl || RegisterStates[2].ColorNeedOverflowControl)
+		{
+			WRITE(p, "cc1 = frac(c1 * (255.0f/256.0f)) * (256.0f/255.0f);\n");
+			RegisterStates[2].AlphaNeedOverflowControl = false;
+			RegisterStates[2].ColorNeedOverflowControl = false;
+		}
+		else
+		{
+			WRITE(p, "cc1 = c1;\n");
+		}
+		RegisterStates[2].AuxStored = true;
+	}

 	if(cc.a == TEVCOLORARG_C2 || cc.a == TEVCOLORARG_A2
 	|| cc.b == TEVCOLORARG_C2 || cc.b == TEVCOLORARG_A2
 	|| cc.c == TEVCOLORARG_C2 || cc.c == TEVCOLORARG_A2
 	|| ac.a == TEVALPHAARG_A2 || ac.b == TEVALPHAARG_A2 || ac.c == TEVALPHAARG_A2)
+	{
+		if(RegisterStates[3].AlphaNeedOverflowControl || RegisterStates[3].ColorNeedOverflowControl)
+		{
 			WRITE(p, "cc2 = frac(c2 * (255.0f/256.0f)) * (256.0f/255.0f);\n");
+			RegisterStates[3].AlphaNeedOverflowControl = false;
+			RegisterStates[3].ColorNeedOverflowControl = false;
+		}
+		else
+		{
+			WRITE(p, "cc2 = c2;\n");
+		}
+		RegisterStates[3].AuxStored = true;
+	}

+	RegisterStates[cc.dest].ColorNeedOverflowControl = (cc.clamp == 0);
+	RegisterStates[cc.dest].AuxStored = false;

+	// combine the color channel
 	WRITE(p, "// color combine\n");
 	if (cc.clamp)
 		WRITE(p, "%s = saturate(", tevCOutputTable[cc.dest]);
@@ -1042,8 +1057,11 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType)
 		WRITE(p, ")");
 	WRITE(p,";\n");

-	WRITE(p, "// alpha combine\n");
+	RegisterStates[ac.dest].AlphaNeedOverflowControl = (ac.clamp == 0);
+	RegisterStates[ac.dest].AuxStored = false;
+
 	// combine the alpha channel
+	WRITE(p, "// alpha combine\n");
 	if (ac.clamp)
 		WRITE(p, "%s = saturate(", tevAOutputTable[ac.dest]);
 	else
@@ -1118,39 +1136,6 @@ static const char *tevAlphaFunclogicTable[] =
 	" != ", // xor
 	" == "  // xnor
 };
-static int AlphaPreTest()
-{
-	u32 op = bpmem.alphaFunc.logic;
-	u32 comp[2] = {bpmem.alphaFunc.comp0, bpmem.alphaFunc.comp1};
-
-	// First kill all the simple cases
-	switch(op)
-	{
-	case 0: // AND
-		if (comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) return true;
-		if (comp[0] == ALPHACMP_NEVER || comp[1] == ALPHACMP_NEVER) return false;
-		break;
-	case 1: // OR
-		if (comp[0] == ALPHACMP_ALWAYS || comp[1] == ALPHACMP_ALWAYS) return true;
-		if (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER)return false;
-		break;
-	case 2: // XOR
-		if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS))
-			return true;
-		if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER))
-			return false;
-		break;
-	case 3: // XNOR
-		if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_NEVER) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_ALWAYS))
-			return false;
-		if ((comp[0] == ALPHACMP_ALWAYS && comp[1] == ALPHACMP_ALWAYS) || (comp[0] == ALPHACMP_NEVER && comp[1] == ALPHACMP_NEVER))
-			return true;
-		break;
-	default: PanicAlert("bad logic for alpha test? %08x", op);
-	}
-	return -1;
-}
-

 static void WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode)
 {
@@ -1163,34 +1148,31 @@ static void WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode
 	// using discard then return works the same in cg and dx9 but not in dx11
 	WRITE(p, "if(!( ");

-	int compindex = bpmem.alphaFunc.comp0 % 8;
+	int compindex = bpmem.alpha_test.comp0;
 	WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[0]);//lookup the first component from the alpha function table

-	WRITE(p, "%s", tevAlphaFunclogicTable[bpmem.alphaFunc.logic % 4]);//lookup the logic op
+	WRITE(p, "%s", tevAlphaFunclogicTable[bpmem.alpha_test.logic]);//lookup the logic op

-	compindex = bpmem.alphaFunc.comp1 % 8;
+	compindex = bpmem.alpha_test.comp1;
 	WRITE(p, tevAlphaFuncsTable[compindex],alphaRef[1]);//lookup the second component from the alpha function table
 	WRITE(p, ")) {\n");

 	WRITE(p, "ocol0 = 0;\n");
 	if (dstAlphaMode == DSTALPHA_DUAL_SOURCE_BLEND)
 		WRITE(p, "ocol1 = 0;\n");
-	if (DepthTextureEnable)
-		WRITE(p, "depth = 1.f;\n");
+	WRITE(p, "depth = 1.f;\n");

-	// HAXX: zcomploc is a way to control whether depth test is done before
-	// or after texturing and alpha test. PC GPU does depth test before texturing ONLY if depth value is
-	// not updated during shader execution.
+	// HAXX: zcomploc (aka early_ztest) is a way to control whether depth test is done before
+	// or after texturing and alpha test. PC GPUs have no way to support this
+	// feature properly as of 2012: depth buffer and depth test are not
+	// programmable and the depth test is always done after texturing.
+	// Most importantly, PC GPUs do not allow writing to the z buffer without
+	// writing a color value (unless color writing is disabled altogether).
 	// We implement "depth test before texturing" by discarding the fragment
 	// when the alpha test fail. This is not a correct implementation because
-	// even if the depth test fails the fragment could be alpha blended.
-	// this implemnetation is a trick to  keep speed.
-	// the correct, but slow, way to implement a correct zComploc is :
-	// 1 - if zcomplock is enebled make a first pass, with color channel write disabled updating only 
-	// depth channel.
-	// 2 - in the next pass disable depth chanel update, but proccess the color data normally
-	// this way is the only CORRECT way to emulate perfectly the zcomplock behaviour
-	if (!(bpmem.zcontrol.zcomploc && bpmem.zmode.updateenable))
+	// even if the depth test fails the fragment could be alpha blended, but
+	// we don't have a choice.
+	if (!(bpmem.zcontrol.early_ztest && bpmem.zmode.updateenable))
 	{
 		WRITE(p, "discard;\n");
 		if (ApiType != API_D3D11)
@@ -1198,7 +1180,6 @@ static void WriteAlphaTest(char *&p, API_TYPE ApiType,DSTALPHA_MODE dstAlphaMode
 	}

 	WRITE(p, "}\n");
-	
 }

 static const char *tevFogFuncsTable[] =
@@ -45,7 +45,7 @@
 #define C_PMATERIALS	(C_PLIGHTS + 40)
 #define C_PENVCONST_END (C_PMATERIALS + 4)
 #define PIXELSHADERUID_MAX_VALUES 70
-#define PIXELSHADERUID_MAX_VALUES_SAFE 120
+#define PIXELSHADERUID_MAX_VALUES_SAFE 115

 // DO NOT make anything in this class virtual.
 template<bool safe>
@@ -302,7 +302,7 @@ void PixelShaderManager::SetConstants()
 			float GC_ALIGNED16(material[4]);
 			float NormalizationCoef = 1 / 255.0f;

-			for (int i = 0; i < 4; ++i)
+			for (int i = 0; i < 2; ++i)
 			{
 				if (nMaterialsChanged & (1 << i))
 				{
@@ -316,6 +316,21 @@ void PixelShaderManager::SetConstants()
 					SetPSConstant4fv(C_PMATERIALS + i, material);
 				}
 			}
+			
+			for (int i = 0; i < 2; ++i)
+			{
+				if (nMaterialsChanged & (1 << (i + 2)))
+				{
+					u32 data = *(xfregs.matColor + i);
+
+					material[0] = ((data >> 24) & 0xFF) * NormalizationCoef;
+					material[1] = ((data >> 16) & 0xFF) * NormalizationCoef;
+					material[2] = ((data >>  8) & 0xFF) * NormalizationCoef;
+					material[3] = ( data        & 0xFF) * NormalizationCoef;
+
+					SetPSConstant4fv(C_PMATERIALS + i + 2, material);
+				}
+			}

 			nMaterialsChanged = 0;
 		}
@@ -358,7 +373,7 @@ void PixelShaderManager::SetColorChanged(int type, int num, bool high)
    PRIM_LOG("pixel %scolor%d: %f %f %f %f\n", type?"k":"", num, pf[0], pf[1], pf[2], pf[3]);
 }

-void PixelShaderManager::SetAlpha(const AlphaFunc& alpha)
+void PixelShaderManager::SetAlpha(const AlphaTest& alpha)
 {
    if ((alpha.hex & 0xffff) != lastAlpha)
 	{
@@ -38,7 +38,7 @@ public:

 	// constant management, should be called after memory is committed
 	static void SetColorChanged(int type, int index, bool high);
-	static void SetAlpha(const AlphaFunc& alpha);
+	static void SetAlpha(const AlphaTest& alpha);
 	static void SetDestAlpha(const ConstantAlpha& alpha);
 	static void SetTexDims(int texmapid, u32 width, u32 height, u32 wraps, u32 wrapt);
 	static void SetZTextureBias(u32 bias);
@@ -67,12 +67,7 @@ int Renderer::s_target_height;
 int Renderer::s_backbuffer_width;
 int Renderer::s_backbuffer_height;

-// ratio of backbuffer size and render area size
-float Renderer::xScale;
-float Renderer::yScale;
-
-unsigned int Renderer::s_XFB_width;
-unsigned int Renderer::s_XFB_height;
+TargetRectangle Renderer::target_rc;

 int Renderer::s_LastEFBScale;

@@ -81,9 +76,16 @@ bool Renderer::XFBWrited;
 bool Renderer::s_EnableDLCachingAfterRecording;

 unsigned int Renderer::prev_efb_format = (unsigned int)-1;
+unsigned int Renderer::efb_scale_numeratorX = 1;
+unsigned int Renderer::efb_scale_numeratorY = 1;
+unsigned int Renderer::efb_scale_denominatorX = 1;
+unsigned int Renderer::efb_scale_denominatorY = 1;
+unsigned int Renderer::ssaa_multiplier = 1;


-Renderer::Renderer() : frame_data(NULL), bLastFrameDumped(false)
+Renderer::Renderer()
+	: frame_data()
+	, bLastFrameDumped(false)
 {
 	UpdateActiveConfig();
 	TextureCache::OnConfigChanged(g_ActiveConfig);
@@ -91,6 +93,9 @@ Renderer::Renderer() : frame_data(NULL), bLastFrameDumped(false)
 #if defined _WIN32 || defined HAVE_LIBAV
 	bAVIDumping = false;
 #endif
+
+	OSDChoice = 0;
+	OSDTime = 0;
 }

 Renderer::~Renderer()
@@ -98,6 +103,8 @@ Renderer::~Renderer()
 	// invalidate previous efb format
 	prev_efb_format = (unsigned int)-1;

+	efb_scale_numeratorX = efb_scale_numeratorY = efb_scale_denominatorX = efb_scale_denominatorY = ssaa_multiplier = 1;
+
 #if defined _WIN32 || defined HAVE_LIBAV
 	if (g_ActiveConfig.bDumpFrames && bLastFrameDumped && bAVIDumping)
 		AVIDump::Stop();
@@ -105,7 +112,6 @@ Renderer::~Renderer()
 	if (pFrameDump.IsOpen())
 		pFrameDump.Close();
 #endif
-	delete[] frame_data;
 }

 void Renderer::RenderToXFB(u32 xfbAddr, u32 fbWidth, u32 fbHeight, const EFBRectangle& sourceRc, float Gamma)
@@ -121,64 +127,117 @@ void Renderer::RenderToXFB(u32 xfbAddr, u32 fbWidth, u32 fbHeight, const EFBRect
 	VideoFifo_CheckSwapRequestAt(xfbAddr, fbWidth, fbHeight);
 	XFBWrited = true;

-	// XXX: Without the VI, how would we know what kind of field this is? So
-	// just use progressive.
 	if (g_ActiveConfig.bUseXFB)
 	{
 		FramebufferManagerBase::CopyToXFB(xfbAddr, fbWidth, fbHeight, sourceRc,Gamma);
 	}
 	else
 	{
+		// XXX: Without the VI, how would we know what kind of field this is? So
+		// just use progressive.
 		g_renderer->Swap(xfbAddr, FIELD_PROGRESSIVE, fbWidth, fbHeight,sourceRc,Gamma);
 		Common::AtomicStoreRelease(s_swapRequested, false);
 	}
 }

-void Renderer::CalculateTargetScale(int x, int y, int &scaledX, int &scaledY)
+int Renderer::EFBToScaledX(int x)
 {
 	switch (g_ActiveConfig.iEFBScale)
 	{
-		case 3: // 1.5x
-			scaledX = (x / 2) * 3;
-			scaledY = (y / 2) * 3;
-			break;
-		case 4: // 2x
-			scaledX = x * 2;
-			scaledY = y * 2;
-			break;
-		case 5: // 2.5x
-			scaledX = (x / 2) * 5;
-			scaledY = (y / 2) * 5;
-			break;
-		case 6: // 3x
-			scaledX = x * 3;
-			scaledY = y * 3;
-			break;
-		case 7: // 4x
-			scaledX = x * 4;
-			scaledY = y * 4;
-			break;
+		case 0: // fractional
+			return (int)ssaa_multiplier * FramebufferManagerBase::ScaleToVirtualXfbWidth(x, s_backbuffer_width);
+
 		default:
-			scaledX = x;
-			scaledY = y;
-			break;
+			return x * (int)ssaa_multiplier * (int)efb_scale_numeratorX / (int)efb_scale_denominatorX;
 	};
 }

+int Renderer::EFBToScaledY(int y)
+{
+	switch (g_ActiveConfig.iEFBScale)
+	{
+		case 0: // fractional
+			return (int)ssaa_multiplier * FramebufferManagerBase::ScaleToVirtualXfbHeight(y, s_backbuffer_height);
+
+		default:
+			return y * (int)ssaa_multiplier * (int)efb_scale_numeratorY / (int)efb_scale_denominatorY;
+	};
+}
+
+void Renderer::CalculateTargetScale(int x, int y, int &scaledX, int &scaledY)
+{
+	if (g_ActiveConfig.iEFBScale == 0 || g_ActiveConfig.iEFBScale == 1)
+	{
+		scaledX = x;
+		scaledY = y;
+	}
+	else
+	{
+		scaledX = x * (int)efb_scale_numeratorX / (int)efb_scale_denominatorX;
+		scaledY = y * (int)efb_scale_numeratorY / (int)efb_scale_denominatorY;
+	}
+}
+
 // return true if target size changed
-bool Renderer::CalculateTargetSize(int multiplier)
+bool Renderer::CalculateTargetSize(unsigned int framebuffer_width, unsigned int framebuffer_height, int multiplier)
 {
 	int newEFBWidth, newEFBHeight;
+
+	// TODO: Ugly. Clean up
+	switch (s_LastEFBScale)
+	{
+		case 2: // 1x
+			efb_scale_numeratorX = efb_scale_numeratorY = 1;
+			efb_scale_denominatorX = efb_scale_denominatorY = 1;
+			break;
+
+		case 3: // 1.5x
+			efb_scale_numeratorX = efb_scale_numeratorY = 3;
+			efb_scale_denominatorX = efb_scale_denominatorY = 2;
+			break;
+
+		case 4: // 2x
+			efb_scale_numeratorX = efb_scale_numeratorY = 2;
+			efb_scale_denominatorX = efb_scale_denominatorY = 1;
+			break;
+
+		case 5: // 2.5x
+			efb_scale_numeratorX = efb_scale_numeratorY = 5;
+			efb_scale_denominatorX = efb_scale_denominatorY = 2;
+			break;
+
+		case 6: // 3x
+			efb_scale_numeratorX = efb_scale_numeratorY = 3;
+			efb_scale_denominatorX = efb_scale_denominatorY = 1;
+			break;
+
+		case 7: // 4x
+			efb_scale_numeratorX = efb_scale_numeratorY = 4;
+			efb_scale_denominatorX = efb_scale_denominatorY = 1;
+			break;
+
+		default: // fractional & integral handled later
+			break;
+	}
+
 	switch (s_LastEFBScale)
 	{
 		case 0: // fractional
-			newEFBWidth = (int)(EFB_WIDTH * xScale);
-			newEFBHeight = (int)(EFB_HEIGHT * yScale);
-			break;
 		case 1: // integral
-			newEFBWidth = EFB_WIDTH * (int)ceilf(xScale);
-			newEFBHeight = EFB_HEIGHT * (int)ceilf(yScale);
+			newEFBWidth = FramebufferManagerBase::ScaleToVirtualXfbWidth(EFB_WIDTH, framebuffer_width);
+			newEFBHeight = FramebufferManagerBase::ScaleToVirtualXfbHeight(EFB_HEIGHT, framebuffer_height);
+
+			if (s_LastEFBScale == 1)
+			{
+				newEFBWidth = ((newEFBWidth-1) / EFB_WIDTH + 1) * EFB_WIDTH;
+				newEFBHeight = ((newEFBHeight-1) / EFB_HEIGHT + 1) * EFB_HEIGHT;
+			}
+			efb_scale_numeratorX = newEFBWidth;
+			efb_scale_denominatorX = EFB_WIDTH;
+			efb_scale_numeratorY = newEFBHeight;
+			efb_scale_denominatorY = EFB_HEIGHT;
 			break;
+
 		default:
 			CalculateTargetScale(EFB_WIDTH, EFB_HEIGHT, newEFBWidth, newEFBHeight);
 			break;
@@ -186,6 +245,7 @@ bool Renderer::CalculateTargetSize(int multiplier)

 	newEFBWidth *= multiplier;
 	newEFBHeight *= multiplier;
+	ssaa_multiplier = multiplier;

 	if (newEFBWidth != s_target_width || newEFBHeight != s_target_height)
 	{
@@ -207,131 +267,230 @@ void Renderer::SetScreenshot(const char *filename)
 // Create On-Screen-Messages
 void Renderer::DrawDebugText()
 {
+	if (!g_Config.bOSDHotKey)
+		return;
+
 	// OSD Menu messages
-	if (g_ActiveConfig.bOSDHotKey)
+	if (OSDChoice > 0)
 	{
-		if (OSDChoice > 0)
-		{
-			OSDTime = Common::Timer::GetTimeMs() + 3000;
-			OSDChoice = -OSDChoice;
-		}
-		if ((u32)OSDTime > Common::Timer::GetTimeMs())
-		{
-			const char* res_text = "";
-			switch (g_ActiveConfig.iEFBScale)
-			{
-			case 0:
-				res_text = "Auto (fractional)";
-				break;
-			case 1:
-				res_text = "Auto (integral)";
-				break;
-			case 2:
-				res_text = "Native";
-				break;
-			case 3:
-				res_text = "1.5x";
-				break;
-			case 4:
-				res_text = "2x";
-				break;
-			case 5:
-				res_text = "2.5x";
-				break;
-			case 6:
-				res_text = "3x";
-				break;
-			case 7:
-				res_text = "4x";
-				break;
-			}
-
-			const char* ar_text = "";
-			switch(g_ActiveConfig.iAspectRatio)
-			{
-			case ASPECT_AUTO:
-				ar_text = "Auto";
-				break;
-			case ASPECT_FORCE_16_9:
-				ar_text = "16:9";
-				break;
-			case ASPECT_FORCE_4_3:
-				ar_text = "4:3";
-				break;
-			case ASPECT_STRETCH:
-				ar_text = "Stretch";
-				break;
-			}
-
-			const char* const efbcopy_text = g_ActiveConfig.bEFBCopyEnable ?
-				(g_ActiveConfig.bCopyEFBToTexture ? "to Texture" : "to RAM") : "Disabled";
-
-			// The rows
-			const std::string lines[] =
-			{
-				std::string("3: Internal Resolution: ") + res_text,
-				std::string("4: Aspect Ratio: ") + ar_text + (g_ActiveConfig.bCrop ? " (crop)" : ""),
-				std::string("5: Copy EFB: ") + efbcopy_text,
-				std::string("6: Fog: ") + (g_ActiveConfig.bDisableFog ? "Disabled" : "Enabled"),
-			};
-
-			enum { lines_count = sizeof(lines)/sizeof(*lines) };
-
-			std::string final_yellow, final_cyan;
-
-			// If there is more text than this we will have a collision
-			if (g_ActiveConfig.bShowFPS)
-			{
-				final_yellow = final_cyan = "\n\n";
-			}
-
-			// The latest changed setting in yellow
-			for (int i = 0; i != lines_count; ++i)
-			{
-				if (OSDChoice == -i - 1)
-					final_yellow += lines[i];
-				final_yellow += '\n';
-			}
-
-			// The other settings in cyan
-			for (int i = 0; i != lines_count; ++i)
-			{
-				if (OSDChoice != -i - 1)
-					final_cyan += lines[i];
-				final_cyan += '\n';
-			}
-
-			// Render a shadow
-			g_renderer->RenderText(final_cyan.c_str(), 21, 21, 0xDD000000);
-			g_renderer->RenderText(final_yellow.c_str(), 21, 21, 0xDD000000);
-			//and then the text
-			g_renderer->RenderText(final_cyan.c_str(), 20, 20, 0xFF00FFFF);
-			g_renderer->RenderText(final_yellow.c_str(), 20, 20, 0xFFFFFF00);
-		}
+		OSDTime = Common::Timer::GetTimeMs() + 3000;
+		OSDChoice = -OSDChoice;
 	}
+
+	if ((u32)OSDTime <= Common::Timer::GetTimeMs())
+		return;
+
+	const char* res_text = "";
+	switch (g_ActiveConfig.iEFBScale)
+	{
+	case 0:
+		res_text = "Auto (fractional)";
+		break;
+	case 1:
+		res_text = "Auto (integral)";
+		break;
+	case 2:
+		res_text = "Native";
+		break;
+	case 3:
+		res_text = "1.5x";
+		break;
+	case 4:
+		res_text = "2x";
+		break;
+	case 5:
+		res_text = "2.5x";
+		break;
+	case 6:
+		res_text = "3x";
+		break;
+	case 7:
+		res_text = "4x";
+		break;
+	}
+
+	const char* ar_text = "";
+	switch(g_ActiveConfig.iAspectRatio)
+	{
+	case ASPECT_AUTO:
+		ar_text = "Auto";
+		break;
+	case ASPECT_FORCE_16_9:
+		ar_text = "16:9";
+		break;
+	case ASPECT_FORCE_4_3:
+		ar_text = "4:3";
+		break;
+	case ASPECT_STRETCH:
+		ar_text = "Stretch";
+		break;
+	}
+
+	const char* const efbcopy_text = g_ActiveConfig.bEFBCopyEnable ?
+		(g_ActiveConfig.bCopyEFBToTexture ? "to Texture" : "to RAM") : "Disabled";
+
+	// The rows
+	const std::string lines[] =
+	{
+		std::string("3: Internal Resolution: ") + res_text,
+		std::string("4: Aspect Ratio: ") + ar_text + (g_ActiveConfig.bCrop ? " (crop)" : ""),
+		std::string("5: Copy EFB: ") + efbcopy_text,
+		std::string("6: Fog: ") + (g_ActiveConfig.bDisableFog ? "Disabled" : "Enabled"),
+	};
+
+	enum { lines_count = sizeof(lines)/sizeof(*lines) };
+
+	std::string final_yellow, final_cyan;
+
+	// If there is more text than this we will have a collision
+	if (g_ActiveConfig.bShowFPS)
+	{
+		final_yellow = final_cyan = "\n\n";
+	}
+
+	// The latest changed setting in yellow
+	for (int i = 0; i != lines_count; ++i)
+	{
+		if (OSDChoice == -i - 1)
+			final_yellow += lines[i];
+		final_yellow += '\n';
+	}
+
+	// The other settings in cyan
+	for (int i = 0; i != lines_count; ++i)
+	{
+		if (OSDChoice != -i - 1)
+			final_cyan += lines[i];
+		final_cyan += '\n';
+	}
+
+	// Render a shadow
+	g_renderer->RenderText(final_cyan.c_str(), 21, 21, 0xDD000000);
+	g_renderer->RenderText(final_yellow.c_str(), 21, 21, 0xDD000000);
+	//and then the text
+	g_renderer->RenderText(final_cyan.c_str(), 20, 20, 0xFF00FFFF);
+	g_renderer->RenderText(final_yellow.c_str(), 20, 20, 0xFFFFFF00);
 }

-void Renderer::CalculateXYScale(const TargetRectangle& dst_rect)
+// TODO: remove
+extern bool g_aspect_wide;
+
+void Renderer::UpdateDrawRectangle(int backbuffer_width, int backbuffer_height)
 {
-	if (g_ActiveConfig.bUseXFB && g_ActiveConfig.bUseRealXFB)
+	float FloatGLWidth = (float)backbuffer_width;
+	float FloatGLHeight = (float)backbuffer_height;
+	float FloatXOffset = 0;
+	float FloatYOffset = 0;
+
+	// The rendering window size
+	const float WinWidth = FloatGLWidth;
+	const float WinHeight = FloatGLHeight;
+
+	// Handle aspect ratio.
+	// Default to auto.
+	bool use16_9 = g_aspect_wide;
+
+	// Update aspect ratio hack values
+	// Won't take effect until next frame
+	// Don't know if there is a better place for this code so there isn't a 1 frame delay
+	if ( g_ActiveConfig.bWidescreenHack )
 	{
-		xScale = 1.0f;
-		yScale = 1.0f;
-	}
-	else
-	{
-		if (g_ActiveConfig.b3DVision)
+		float source_aspect = use16_9 ? (16.0f / 9.0f) : (4.0f / 3.0f);
+		float target_aspect;
+
+		switch ( g_ActiveConfig.iAspectRatio )
 		{
-			// This works, yet the version in the else doesn't. No idea why.
-			xScale = (float)(s_backbuffer_width-1) / (float)(s_XFB_width-1);
-			yScale = (float)(s_backbuffer_height-1) / (float)(s_XFB_height-1);
+		case ASPECT_FORCE_16_9 :
+			target_aspect = 16.0f / 9.0f;
+			break;
+		case ASPECT_FORCE_4_3 :
+			target_aspect = 4.0f / 3.0f;
+			break;
+		case ASPECT_STRETCH :
+			target_aspect = WinWidth / WinHeight;
+			break;
+		default :
+			// ASPECT_AUTO == no hacking
+			target_aspect = source_aspect;
+			break;
+		}
+
+		float adjust = source_aspect / target_aspect;
+		if ( adjust > 1 )
+		{
+			// Vert+
+			g_Config.fAspectRatioHackW = 1;
+			g_Config.fAspectRatioHackH = 1/adjust;
 		}
 		else
 		{
-			xScale = (float)(dst_rect.right - dst_rect.left - 1) / (float)(s_XFB_width-1);
-			yScale = (float)(dst_rect.bottom - dst_rect.top - 1) / (float)(s_XFB_height-1);
+			// Hor+
+			g_Config.fAspectRatioHackW = adjust;
+			g_Config.fAspectRatioHackH = 1;
 		}
 	}
+	else
+	{
+		// Hack is disabled
+		g_Config.fAspectRatioHackW = 1;
+		g_Config.fAspectRatioHackH = 1;
+	}
+
+	// Check for force-settings and override.
+	if (g_ActiveConfig.iAspectRatio == ASPECT_FORCE_16_9)
+		use16_9 = true;
+	else if (g_ActiveConfig.iAspectRatio == ASPECT_FORCE_4_3)
+		use16_9 = false;
+
+	if (g_ActiveConfig.iAspectRatio != ASPECT_STRETCH)
+	{
+		// The rendering window aspect ratio as a proportion of the 4:3 or 16:9 ratio
+		float Ratio = (WinWidth / WinHeight) / (!use16_9 ? (4.0f / 3.0f) : (16.0f / 9.0f));
+		// Check if height or width is the limiting factor. If ratio > 1 the picture is too wide and have to limit the width.
+		if (Ratio > 1.0f)
+		{
+			// Scale down and center in the X direction.
+			FloatGLWidth /= Ratio;
+			FloatXOffset = (WinWidth - FloatGLWidth) / 2.0f;
+		}
+		// The window is too high, we have to limit the height
+		else
+		{
+			// Scale down and center in the Y direction.
+			FloatGLHeight *= Ratio;
+			FloatYOffset = FloatYOffset + (WinHeight - FloatGLHeight) / 2.0f;
+		}
+	}
+
+	// -----------------------------------------------------------------------
+	// Crop the picture from 4:3 to 5:4 or from 16:9 to 16:10.
+	//		Output: FloatGLWidth, FloatGLHeight, FloatXOffset, FloatYOffset
+	// ------------------
+	if (g_ActiveConfig.iAspectRatio != ASPECT_STRETCH && g_ActiveConfig.bCrop)
+	{
+		float Ratio = !use16_9 ? ((4.0f / 3.0f) / (5.0f / 4.0f)) : (((16.0f / 9.0f) / (16.0f / 10.0f)));
+		// The width and height we will add (calculate this before FloatGLWidth and FloatGLHeight is adjusted)
+		float IncreasedWidth = (Ratio - 1.0f) * FloatGLWidth;
+		float IncreasedHeight = (Ratio - 1.0f) * FloatGLHeight;
+		// The new width and height
+		FloatGLWidth = FloatGLWidth * Ratio;
+		FloatGLHeight = FloatGLHeight * Ratio;
+		// Adjust the X and Y offset
+		FloatXOffset = FloatXOffset - (IncreasedWidth * 0.5f);
+		FloatYOffset = FloatYOffset - (IncreasedHeight * 0.5f);
+	}
+
+	int XOffset = (int)(FloatXOffset + 0.5f);
+	int YOffset = (int)(FloatYOffset + 0.5f);
+	int iWhidth = (int)ceil(FloatGLWidth);
+	int iHeight = (int)ceil(FloatGLHeight);
+	iWhidth -= iWhidth % 4; // ensure divisibility by 4 to make it compatible with all the video encoders
+	iHeight -= iHeight % 4;
+
+	target_rc.left = XOffset;
+	target_rc.top = YOffset;
+	target_rc.right = XOffset + iWhidth;
+	target_rc.bottom = YOffset + iHeight;
 }

 void Renderer::SetWindowSize(int width, int height)
@@ -39,7 +39,7 @@

 // TODO: Move these out of here.
 extern int frameCount;
-extern int OSDChoice, OSDTime;
+extern int OSDChoice;

 extern bool bLastFrameDumped;

@@ -83,10 +83,6 @@ public:
 	static int GetBackbufferWidth() { return s_backbuffer_width; }
 	static int GetBackbufferHeight() { return s_backbuffer_height; }

-	// XFB scale - TODO: Remove this and add two XFBToScaled functions instead
-	static float GetXFBScaleX() { return xScale; }
-	static float GetXFBScaleY() { return yScale; }
-
 	static void SetWindowSize(int width, int height);

 	// EFB coordinate conversion functions
@@ -94,9 +90,13 @@ public:
 	// Use this to convert a whole native EFB rect to backbuffer coordinates
 	virtual TargetRectangle ConvertEFBRectangle(const EFBRectangle& rc) = 0;

+	static const TargetRectangle& GetTargetRectangle() { return target_rc; }
+	static void UpdateDrawRectangle(int backbuffer_width, int backbuffer_height);
+
+
 	// Use this to upscale native EFB coordinates to IDEAL internal resolution
-	static unsigned int EFBToScaledX(int x) { return x * GetTargetWidth() / EFB_WIDTH; }
-	static unsigned int EFBToScaledY(int y) { return y * GetTargetHeight() / EFB_HEIGHT; }
+	static int EFBToScaledX(int x);
+	static int EFBToScaledY(int y);

 	// Floating point versions of the above - only use them if really necessary
 	static float EFBToScaledXf(float x) { return x * ((float)GetTargetWidth() / (float)EFB_WIDTH); }
@@ -147,8 +147,7 @@ public:
 protected:

 	static void CalculateTargetScale(int x, int y, int &scaledX, int &scaledY);
-	static bool CalculateTargetSize(int multiplier = 1);
-	static void CalculateXYScale(const TargetRectangle& dst_rect);
+	static bool CalculateTargetSize(unsigned int framebuffer_width, unsigned int framebuffer_height, int multiplier = 1);

 	static void CheckFifoRecording();
 	static void RecordVideoMemory();
@@ -162,7 +161,7 @@ protected:
 #else
 	File::IOFile pFrameDump;
 #endif
-	char* frame_data;
+	std::vector<u8> frame_data;
 	bool bLastFrameDumped;

 	// The framebuffer size
@@ -173,12 +172,7 @@ protected:
 	static int s_backbuffer_width;
 	static int s_backbuffer_height;

-	// ratio of backbuffer size and render area size - TODO: Remove these!
-	static float xScale;
-	static float yScale;
-
-	static unsigned int s_XFB_width;
-	static unsigned int s_XFB_height;
+	static TargetRectangle target_rc;

 	// can probably eliminate this static var
 	static int s_LastEFBScale;
@@ -190,6 +184,11 @@ protected:

 private:
 	static unsigned int prev_efb_format;
+	static unsigned int efb_scale_numeratorX;
+	static unsigned int efb_scale_numeratorY;
+	static unsigned int efb_scale_denominatorX;
+	static unsigned int efb_scale_denominatorY;
+	static unsigned int ssaa_multiplier;
 };

 extern Renderer *g_renderer;
@@ -43,7 +43,7 @@ struct Statistics

 	int numUniquePixelShaders;

-	float proj_0, proj_1, proj_2, proj_3, proj_4, proj_5, proj_6;
+	float proj_0, proj_1, proj_2, proj_3, proj_4, proj_5;
 	float gproj_0, gproj_1, gproj_2, gproj_3, gproj_4, gproj_5;
 	float gproj_6, gproj_7, gproj_8, gproj_9, gproj_10, gproj_11, gproj_12, gproj_13, gproj_14, gproj_15;

@@ -84,35 +84,34 @@ TextureCache::~TextureCache()

 void TextureCache::OnConfigChanged(VideoConfig& config)
 {
-	if (!g_texture_cache)
-		goto skip_checks;
-
-	// TODO: Invalidating texcache is really stupid in some of these cases
-	if (config.iSafeTextureCache_ColorSamples != backup_config.s_colorsamples ||
-		config.bTexFmtOverlayEnable != backup_config.s_texfmt_overlay ||
-		config.bTexFmtOverlayCenter != backup_config.s_texfmt_overlay_center ||
-		config.bHiresTextures != backup_config.s_hires_textures)
+	if (g_texture_cache)
 	{
-		g_texture_cache->Invalidate();
+		// TODO: Invalidating texcache is really stupid in some of these cases
+		if (config.iSafeTextureCache_ColorSamples != backup_config.s_colorsamples ||
+			config.bTexFmtOverlayEnable != backup_config.s_texfmt_overlay ||
+			config.bTexFmtOverlayCenter != backup_config.s_texfmt_overlay_center ||
+			config.bHiresTextures != backup_config.s_hires_textures)
+		{
+			g_texture_cache->Invalidate();

-		if(g_ActiveConfig.bHiresTextures)
-			HiresTextures::Init(SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
+			if(g_ActiveConfig.bHiresTextures)
+				HiresTextures::Init(SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());

-		SetHash64Function(g_ActiveConfig.bHiresTextures || g_ActiveConfig.bDumpTextures);
-		TexDecoder_SetTexFmtOverlayOptions(g_ActiveConfig.bTexFmtOverlayEnable, g_ActiveConfig.bTexFmtOverlayCenter);
+			SetHash64Function(g_ActiveConfig.bHiresTextures || g_ActiveConfig.bDumpTextures);
+			TexDecoder_SetTexFmtOverlayOptions(g_ActiveConfig.bTexFmtOverlayEnable, g_ActiveConfig.bTexFmtOverlayCenter);
+		}
+
+		// TODO: Probably shouldn't clear all render targets here, just mark them dirty or something.
+		if (config.bEFBCopyCacheEnable != backup_config.s_copy_cache_enable || // TODO: not sure if this is needed?
+			config.bCopyEFBToTexture != backup_config.s_copy_efb_to_texture ||
+			config.bCopyEFBScaled != backup_config.s_copy_efb_scaled ||
+			config.bEFBCopyEnable != backup_config.s_copy_efb ||
+			config.iEFBScale != backup_config.s_efb_scale)
+		{
+			g_texture_cache->ClearRenderTargets();
+		}
 	}
-
-	// TODO: Probably shouldn't clear all render targets here, just mark them dirty or something.
-	if (config.bEFBCopyCacheEnable != backup_config.s_copy_cache_enable || // TODO: not sure if this is needed?
-		config.bCopyEFBToTexture != backup_config.s_copy_efb_to_texture ||
-		config.bCopyEFBScaled != backup_config.s_copy_efb_scaled ||
-		config.bEFBCopyEnable != backup_config.s_copy_efb ||
-		config.iEFBScale != backup_config.s_efb_scale)
-	{
-		g_texture_cache->ClearRenderTargets();
-	}
-
-skip_checks:
+	
 	backup_config.s_colorsamples = config.iSafeTextureCache_ColorSamples;
 	backup_config.s_copy_efb_to_texture = config.bCopyEFBToTexture;
 	backup_config.s_copy_efb_scaled = config.bCopyEFBScaled;
@@ -130,7 +129,11 @@ void TextureCache::Cleanup()
 	TexCache::iterator tcend = textures.end();
 	while (iter != tcend)
 	{
-		if (frameCount > TEXTURE_KILL_THRESHOLD + iter->second->frameCount) // TODO: Deleting EFB copies might not be a good idea here...
+		if (	frameCount > TEXTURE_KILL_THRESHOLD + iter->second->frameCount
+			
+			// EFB copies living on the host GPU are unrecoverable and thus shouldn't be deleted
+			// TODO: encoding the texture back to RAM here might be a good idea
+			&& ! (g_ActiveConfig.bCopyEFBToTexture && iter->second->IsEfbCopy()) )
 		{
 			delete iter->second;
 			textures.erase(iter++);
@@ -204,13 +207,23 @@ void TextureCache::ClearRenderTargets()
 		iter = textures.begin(),
 		tcend = textures.end();

-	for (; iter!=tcend; ++iter)
-		if (iter->second->type != TCET_EC_DYNAMIC)
-			iter->second->type = TCET_NORMAL;
+	while (iter != tcend)
+	{
+		if (iter->second->type == TCET_EC_VRAM)
+		{
+			delete iter->second;
+			textures.erase(iter++);
+		}
+		else
+			++iter;
+	}
 }

 bool TextureCache::CheckForCustomTextureLODs(u64 tex_hash, int texformat, unsigned int levels)
 {
+	if (levels == 1)
+		return false;
+
 	// Just checking if the necessary files exist, if they can't be loaded or have incorrect dimensions LODs will be black
 	char texBasePathTemp[MAX_PATH];
 	char texPathTemp[MAX_PATH];
@@ -291,14 +304,30 @@ void TextureCache::DumpTexture(TCacheEntryBase* entry, unsigned int level)
 		entry->Save(szTemp, level);
 }

-TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage,
-	u32 address, unsigned int width, unsigned int height, int texformat,
-	unsigned int tlutaddr, int tlutfmt, bool UseNativeMips, unsigned int maxlevel, bool from_tmem)
+static u32 CalculateLevelSize(u32 level_0_size, u32 level)
+{
+	return (level_0_size + ((1 << level) - 1)) >> level;
+}
+
+// Used by TextureCache::Load
+static TextureCache::TCacheEntryBase* ReturnEntry(unsigned int stage, TextureCache::TCacheEntryBase* entry)
+{
+	entry->frameCount = frameCount;
+	entry->Bind(stage);
+
+	GFX_DEBUGGER_PAUSE_AT(NEXT_TEXTURE_CHANGE, true);
+
+	return entry;
+}
+
+TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int const stage,
+	u32 const address, unsigned int width, unsigned int height, int const texformat,
+	unsigned int const tlutaddr, int const tlutfmt, bool const use_mipmaps, unsigned int maxlevel, bool const from_tmem)
 {
 	if (0 == address)
 		return NULL;

-	// TexelSizeInNibbles(format)*width*height/16;
+	// TexelSizeInNibbles(format) * width * height / 16;
 	const unsigned int bsw = TexDecoder_GetBlockWidthInTexels(texformat) - 1;
 	const unsigned int bsh = TexDecoder_GetBlockHeightInTexels(texformat) - 1;

@@ -307,11 +336,9 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage,
 	const unsigned int nativeW = width;
 	const unsigned int nativeH = height;

-	bool using_custom_texture = false;
-	bool using_custom_lods = false;
-
 	u32 texID = address;
-	u64 tex_hash = TEXHASH_INVALID; // Hash assigned to texcache entry (also used to generate filenames used for texture dumping and custom texture lookup)
+	// Hash assigned to texcache entry (also used to generate filenames used for texture dumping and custom texture lookup)
+	u64 tex_hash = TEXHASH_INVALID;
 	u64 tlut_hash = TEXHASH_INVALID;

 	u32 full_format = texformat;
@@ -322,10 +349,14 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage,
 		full_format = texformat | (tlutfmt << 16);

 	const u32 texture_size = TexDecoder_GetTextureSizeInBytes(expandedWidth, expandedHeight, texformat);
-	u8* src_data;
-	if (from_tmem) src_data = &texMem[bpmem.tex[stage/4].texImage1[stage%4].tmem_even * TMEM_LINE_SIZE];
-	else src_data = Memory::GetPointer(address);

+	const u8* src_data;
+	if (from_tmem)
+		src_data = &texMem[bpmem.tex[stage / 4].texImage1[stage % 4].tmem_even * TMEM_LINE_SIZE];
+	else
+		src_data = Memory::GetPointer(address);
+
+	// TODO: This doesn't hash GB tiles for preloaded RGBA8 textures (instead, it's hashing more data from the low tmem bank than it should)
 	tex_hash = GetHash64(src_data, texture_size, g_ActiveConfig.iSafeTextureCache_ColorSamples);
 	if (isPaletteTexture)
 	{
@@ -345,6 +376,11 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage,
 		tex_hash ^= tlut_hash;
 	}

+	// D3D doesn't like when the specified mipmap count would require more than one 1x1-sized LOD in the mipmap chain
+	// e.g. 64x64 with 7 LODs would have the mipmap chain 64x64,32x32,16x16,8x8,4x4,2x2,1x1,1x1, so we limit the mipmap count to 6 there
+	while (g_ActiveConfig.backend_info.bUseMinimalMipCount && max(expandedWidth, expandedHeight) >> maxlevel == 0)
+		--maxlevel;
+
 	TCacheEntryBase *entry = textures[texID];
 	if (entry)
 	{
@@ -358,15 +394,17 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage,
 		{
 			entry->type = TCET_EC_VRAM;

-			// TODO: Print a warning if the format changes! In this case, we could reinterpret the internal texture object data to the new pixel format (similiar to what is already being done in Renderer::ReinterpretPixelFormat())
-			goto return_entry;
+			// TODO: Print a warning if the format changes! In this case,
+			// we could reinterpret the internal texture object data to the new pixel format
+			// (similiar to what is already being done in Renderer::ReinterpretPixelFormat())
+			return ReturnEntry(stage, entry);
 		}

 		// 2. b) For normal textures, all texture parameters need to match
 		if (address == entry->addr && tex_hash == entry->hash && full_format == entry->format &&
-			entry->num_mipmaps == maxlevel && entry->native_width == nativeW && entry->native_height == nativeH)
+			entry->num_mipmaps > maxlevel && entry->native_width == nativeW && entry->native_height == nativeH)
 		{
-			goto return_entry;
+			return ReturnEntry(stage, entry);
 		}

 		// 3. If we reach this line, we'll have to upload the new texture data to VRAM.
@@ -374,7 +412,8 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage,
 		//
 		// TODO: Don't we need to force texture decoding to RGBA8 for dynamic EFB copies?
 		// TODO: Actually, it should be enough if the internal texture format matches...
-		if ((entry->type == TCET_NORMAL && width == entry->native_width && height == entry->native_height && full_format == entry->format && entry->num_mipmaps == maxlevel)
+		if ((entry->type == TCET_NORMAL && width == entry->virtual_width && height == entry->virtual_height
+			&& full_format == entry->format && entry->num_mipmaps > maxlevel)
 			|| (entry->type == TCET_EC_DYNAMIC && entry->native_width == width && entry->native_height == height))
 		{
 			// reuse the texture
@@ -387,132 +426,135 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage,
 		}
 	}

+	bool using_custom_texture = false;

 	if (g_ActiveConfig.bHiresTextures)
 	{
+		// This function may modify width/height.
 		pcfmt = LoadCustomTexture(tex_hash, texformat, 0, width, height);
 		if (pcfmt != PC_TEX_FMT_NONE)
 		{
-			expandedWidth = width;
-			expandedHeight = height;
+			if (expandedWidth != width || expandedHeight != height)
+			{
+				expandedWidth = width;
+				expandedHeight = height;
+
+				// If we thought we could reuse the texture before, make sure to delete it now!
+				delete entry;
+				entry = NULL;
+			}
 			using_custom_texture = true;
 		}
 	}

-	// TODO: RGBA8 textures are stored non-continuously in tmem, that might cause problems when preloading is enabled
-	if (pcfmt == PC_TEX_FMT_NONE)
-		pcfmt = TexDecoder_Decode(temp, src_data, expandedWidth,
-					expandedHeight, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures);
+	if (!using_custom_texture)
+	{
+		if (!(texformat == GX_TF_RGBA8 && from_tmem))
+		{
+			pcfmt = TexDecoder_Decode(temp, src_data, expandedWidth,
+						expandedHeight, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures);
+		}
+		else
+		{
+			u8* src_data_gb = &texMem[bpmem.tex[stage/4].texImage2[stage%4].tmem_odd * TMEM_LINE_SIZE];
+			pcfmt = TexDecoder_DecodeRGBA8FromTmem(temp, src_data, src_data_gb, expandedWidth, expandedHeight);
+		}
+	}

-	// TODO: Cleanup. Plus, we still autogenerate mipmaps in certain cases (we shouldn't do that)
-	bool isPow2;
-	unsigned int texLevels;
-	isPow2 = !((width & (width - 1)) || (height & (height - 1)));
-	texLevels = (isPow2 && maxlevel) ? GetPow2(std::max(width, height)) : !isPow2;
-	texLevels = maxlevel ? std::min(texLevels, maxlevel + 1) : texLevels;
-	using_custom_lods = using_custom_texture && CheckForCustomTextureLODs(tex_hash, texformat, texLevels);
-	UseNativeMips = UseNativeMips && !using_custom_lods && (width == nativeW && height == nativeH); // Only load native mips if their dimensions fit to our virtual texture dimensions
-	texLevels = (UseNativeMips || using_custom_lods) ? texLevels : !isPow2;
+	u32 texLevels = use_mipmaps ? (maxlevel + 1) : 1;
+	const bool using_custom_lods = using_custom_texture && CheckForCustomTextureLODs(tex_hash, texformat, texLevels);
+	// Only load native mips if their dimensions fit to our virtual texture dimensions
+	const bool use_native_mips = use_mipmaps && !using_custom_lods && (width == nativeW && height == nativeH);
+	texLevels = (use_native_mips || using_custom_lods) ? texLevels : 1; // TODO: Should be forced to 1 for non-pow2 textures (e.g. efb copies with automatically adjusted IR)

 	// create the entry/texture
-	if (NULL == entry) {
+	if (NULL == entry)
+	{
 		textures[texID] = entry = g_texture_cache->CreateTexture(width, height, expandedWidth, texLevels, pcfmt);

 		// Sometimes, we can get around recreating a texture if only the number of mip levels changes
 		// e.g. if our texture cache entry got too many mipmap levels we can limit the number of used levels by setting the appropriate render states
 		// Thus, we don't update this member for every Load, but just whenever the texture gets recreated
-		//
-		// TODO: Won't we end up recreating textures all the time because maxlevel doesn't necessarily equal texLevels?
-		entry->num_mipmaps = maxlevel; // TODO: Does this actually work? We can't really adjust mipmap settings per-stage...
+		// TODO: D3D9 doesn't support min_lod. We should add a workaround for that here!
+		
+		// TODO: This is the wrong value. We should be storing the number of levels our actual texture has.
+		// But that will currently make the above "existing entry" tests fail as "texLevels" is not calculated until after.
+		// Currently, we might try to reuse a texture which appears to have more levels than actual, maybe..
+		entry->num_mipmaps = maxlevel + 1;
 		entry->type = TCET_NORMAL;

 		GFX_DEBUGGER_PAUSE_AT(NEXT_NEW_TEXTURE, true);
 	}
+	else
+	{
+		// load texture (CreateTexture also loads level 0)
+		entry->Load(width, height, expandedWidth, 0);
+	}

 	entry->SetGeneralParameters(address, texture_size, full_format, entry->num_mipmaps);
 	entry->SetDimensions(nativeW, nativeH, width, height);
 	entry->hash = tex_hash;
-	if (entry->IsEfbCopy() && !g_ActiveConfig.bCopyEFBToTexture) entry->type = TCET_EC_DYNAMIC;
-	else entry->type = TCET_NORMAL;
-
-	// load texture
-	entry->Load(width, height, expandedWidth, 0, (texLevels == 0));
+	
+	if (entry->IsEfbCopy() && !g_ActiveConfig.bCopyEFBToTexture)
+		entry->type = TCET_EC_DYNAMIC;
+	else
+		entry->type = TCET_NORMAL;

 	if (g_ActiveConfig.bDumpTextures && !using_custom_texture)
 		DumpTexture(entry, 0);

+	u32 level = 1;
 	// load mips - TODO: Loading mipmaps from tmem is untested!
-	if (texLevels > 1 && pcfmt != PC_TEX_FMT_NONE && UseNativeMips)
+	if (pcfmt != PC_TEX_FMT_NONE)
 	{
-		const unsigned int bsdepth = TexDecoder_GetTexelSizeInNibbles(texformat);
-
-		unsigned int level = 1;
-		unsigned int mipWidth = (width + 1) >> 1;
-		unsigned int mipHeight = (height + 1) >> 1;
-
-		u8* ptr_even = NULL, *ptr_odd = NULL;
-		if (from_tmem)
+		if (use_native_mips)
 		{
-			ptr_even = &texMem[bpmem.tex[stage/4].texImage1[stage%4].tmem_even * TMEM_LINE_SIZE + texture_size];
-			ptr_odd = &texMem[bpmem.tex[stage/4].texImage2[stage%4].tmem_odd * TMEM_LINE_SIZE];
+			src_data += texture_size;
+			
+			const u8* ptr_even = NULL;
+			const u8* ptr_odd = NULL;
+			if (from_tmem)
+			{
+				ptr_even = &texMem[bpmem.tex[stage/4].texImage1[stage%4].tmem_even * TMEM_LINE_SIZE + texture_size];
+				ptr_odd = &texMem[bpmem.tex[stage/4].texImage2[stage%4].tmem_odd * TMEM_LINE_SIZE];
+			}
+
+			for (; level != texLevels; ++level)
+			{
+				const u32 mip_width = CalculateLevelSize(width, level);
+				const u32 mip_height = CalculateLevelSize(height, level);
+				const u32 expanded_mip_width = (mip_width + bsw) & (~bsw);
+				const u32 expanded_mip_height = (mip_height + bsh) & (~bsh);
+				
+				const u8*& mip_src_data = from_tmem
+					? ((level % 2) ? ptr_odd : ptr_even)
+					: src_data;
+				TexDecoder_Decode(temp, mip_src_data, expanded_mip_width, expanded_mip_height, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures);
+				mip_src_data += TexDecoder_GetTextureSizeInBytes(expanded_mip_width, expanded_mip_height, texformat);
+				
+				entry->Load(mip_width, mip_height, expanded_mip_width, level);
+
+				if (g_ActiveConfig.bDumpTextures)
+					DumpTexture(entry, level);
+			}
 		}
-		src_data += texture_size;
-
-		while ((mipHeight || mipWidth) && (level < texLevels))
+		else if (using_custom_lods)
 		{
-			u8** ptr;
-			if (from_tmem) ptr = (level % 2) ? &ptr_odd : &ptr_even;
-			else ptr = &src_data;
+			for (; level != texLevels; ++level)
+			{
+				unsigned int mip_width = CalculateLevelSize(width, level);
+				unsigned int mip_height = CalculateLevelSize(height, level);

-			const unsigned int currentWidth = (mipWidth > 0) ? mipWidth : 1;
-			const unsigned int currentHeight = (mipHeight > 0) ? mipHeight : 1;
-
-			expandedWidth  = (currentWidth + bsw)  & (~bsw);
-			expandedHeight = (currentHeight + bsh) & (~bsh);
-
-			TexDecoder_Decode(temp, *ptr, expandedWidth, expandedHeight, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures);
-			entry->Load(currentWidth, currentHeight, expandedWidth, level, false);
-
-			if (g_ActiveConfig.bDumpTextures)
-				DumpTexture(entry, level);
-
-			*ptr += ((std::max(mipWidth, bsw) * std::max(mipHeight, bsh) * bsdepth) >> 1);
-			mipWidth >>= 1;
-			mipHeight >>= 1;
-			++level;
-		}
-	}
-	else if (texLevels > 1 && pcfmt != PC_TEX_FMT_NONE && using_custom_lods)
-	{
-		unsigned int level = 1;
-		unsigned int mipWidth = (width + 1) >> 1;
-		unsigned int mipHeight = (height + 1) >> 1;
-
-		while ((mipHeight || mipWidth) && (level < texLevels))
-		{
-			unsigned int currentWidth = (mipWidth > 0) ? mipWidth : 1;
-			unsigned int currentHeight = (mipHeight > 0) ? mipHeight : 1;
-
-			LoadCustomTexture(tex_hash, texformat, level, currentWidth, currentHeight);
-			entry->Load(currentWidth, currentHeight, currentWidth, level, false);
-
-			mipWidth >>= 1;
-			mipHeight >>= 1;
-			++level;
+				LoadCustomTexture(tex_hash, texformat, level, mip_width, mip_height);
+				entry->Load(mip_width, mip_height, mip_width, level);
+			}
 		}
 	}

 	INCSTAT(stats.numTexturesCreated);
 	SETSTAT(stats.numTexturesAlive, textures.size());

-return_entry:
-
-	entry->frameCount = frameCount;
-	entry->Bind(stage);
-
-	GFX_DEBUGGER_PAUSE_AT(NEXT_TEXTURE_CHANGE, true);
-
-	return entry;
+	return ReturnEntry(stage, entry);
 }

 void TextureCache::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat, unsigned int srcFormat,
@@ -782,7 +824,7 @@ void TextureCache::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat
 		textures[dstAddr] = entry = g_texture_cache->CreateRenderTargetTexture(scaled_tex_w, scaled_tex_h);

 		// TODO: Using the wrong dstFormat, dumb...
-		entry->SetGeneralParameters(dstAddr, 0, dstFormat, 0);
+		entry->SetGeneralParameters(dstAddr, 0, dstFormat, 1);
 		entry->SetDimensions(tex_w, tex_h, scaled_tex_w, scaled_tex_h);
 		entry->SetHashes(TEXHASH_INVALID);
 		entry->type = TCET_EC_VRAM;
@@ -60,26 +60,26 @@ public:
 		int frameCount;


-		void SetGeneralParameters(u32 addr, u32 size, u32 format, unsigned int num_mipmaps)
+		void SetGeneralParameters(u32 _addr, u32 _size, u32 _format, unsigned int _num_mipmaps)
 		{
-			this->addr = addr;
-			this->size_in_bytes = size;
-			this->format = format;
-			this->num_mipmaps = num_mipmaps;
+			addr = _addr;
+			size_in_bytes = _size;
+			format = _format;
+			num_mipmaps = _num_mipmaps;
 		}

-		void SetDimensions(unsigned int native_width, unsigned int native_height, unsigned int virtual_width, unsigned int virtual_height)
+		void SetDimensions(unsigned int _native_width, unsigned int _native_height, unsigned int _virtual_width, unsigned int _virtual_height)
 		{
-			this->native_width = native_width;
-			this->native_height = native_height;
-			this->virtual_width = virtual_width;
-			this->virtual_height = virtual_height;
+			native_width = _native_width;
+			native_height = _native_height;
+			virtual_width = _virtual_width;
+			virtual_height = _virtual_height;
 		}

-		void SetHashes(u64 hash/*, u32 pal_hash*/)
+		void SetHashes(u64 _hash/*, u32 _pal_hash*/)
 		{
-			this->hash = hash;
-			//this->pal_hash = pal_hash;
+			hash = _hash;
+			//pal_hash = _pal_hash;
 		}


@@ -89,7 +89,7 @@ public:
 		virtual bool Save(const char filename[], unsigned int level) = 0;

 		virtual void Load(unsigned int width, unsigned int height,
-			unsigned int expanded_width, unsigned int level, bool autogen_mips) = 0;
+			unsigned int expanded_width, unsigned int level) = 0;
 		virtual void FromRenderTarget(u32 dstAddr, unsigned int dstFormat,
 			unsigned int srcFormat, const EFBRectangle& srcRect,
 			bool isIntensity, bool scaleByHalf, unsigned int cbufid,
@@ -116,7 +116,7 @@ public:
 	virtual TCacheEntryBase* CreateRenderTargetTexture(unsigned int scaled_tex_w, unsigned int scaled_tex_h) = 0;

 	static TCacheEntryBase* Load(unsigned int stage, u32 address, unsigned int width, unsigned int height,
-		int format, unsigned int tlutaddr, int tlutfmt, bool UseNativeMips, unsigned int maxlevel, bool from_tmem);
+		int format, unsigned int tlutaddr, int tlutfmt, bool use_mipmaps, unsigned int maxlevel, bool from_tmem);
 	static void CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat, unsigned int srcFormat,
 		const EFBRectangle& srcRect, bool isIntensity, bool scaleByHalf);

@@ -86,6 +86,8 @@ enum PC_TexFormat
 PC_TexFormat TexDecoder_Decode(u8 *dst, const u8 *src, int width, int height, int texformat, int tlutaddr, int tlutfmt,bool rgbaOnly = false);
 PC_TexFormat GetPC_TexFormat(int texformat, int tlutfmt);
 void TexDecoder_DecodeTexel(u8 *dst, const u8 *src, int s, int t, int imageWidth, int texformat, int tlutaddr, int tlutfmt);
+void TexDecoder_DecodeTexelRGBA8FromTmem(u8 *dst, const u8 *src_ar, const u8* src_gb, int s, int t, int imageWidth);
+PC_TexFormat TexDecoder_DecodeRGBA8FromTmem(u8* dst, const u8 *src_ar, const u8 *src_gb, int width, int height);
 void TexDecoder_SetTexFmtOverlayOptions(bool enable, bool center);

 #endif
@@ -23,7 +23,7 @@
 #include "MemoryUtil.h"
 #include "StringUtil.h"
 #include "x64Emitter.h"
-#include "ABI.h"
+#include "x64ABI.h"
 #include "PixelEngine.h"
 #include "Host.h"

@@ -43,8 +43,9 @@
 //BBox
 #include "XFMemory.h"
 extern float GC_ALIGNED16(g_fProjectionMatrix[16]);
-
+#ifndef _M_GENERIC
 #define USE_JIT
+#endif

 #define COMPILED_CODE_SIZE 4096

@@ -82,8 +83,9 @@ static const float fractionTable[32] = {
 	1.0f / (1U << 24), 1.0f / (1U << 25), 1.0f / (1U << 26), 1.0f / (1U << 27),
 	1.0f / (1U << 28), 1.0f / (1U << 29), 1.0f / (1U << 30), 1.0f / (1U << 31),
 };
-
+#ifdef USE_JIT
 using namespace Gen;
+#endif

 void LOADERDECL PosMtx_ReadDirect_UByte()
 {
@@ -182,14 +184,19 @@ VertexLoader::VertexLoader(const TVtxDesc &vtx_desc, const VAT &vtx_attr)
 	m_VtxDesc = vtx_desc;
 	SetVAT(vtx_attr.g0.Hex, vtx_attr.g1.Hex, vtx_attr.g2.Hex);

+	#ifdef USE_JIT
 	AllocCodeSpace(COMPILED_CODE_SIZE);
 	CompileVertexTranslator();
 	WriteProtect();
+	#endif
+
 }

 VertexLoader::~VertexLoader() 
 {
+	#ifdef USE_JIT
 	FreeCodeSpace();
+	#endif
 	delete m_NativeFmt;
 }

@@ -224,14 +231,14 @@ void VertexLoader::CompileVertexTranslator()
 #endif

 	// Colors
-	const int col[2] = {m_VtxDesc.Color0, m_VtxDesc.Color1};
+	const u32 col[2] = {m_VtxDesc.Color0, m_VtxDesc.Color1};
 	// TextureCoord
 	// Since m_VtxDesc.Text7Coord is broken across a 32 bit word boundary, retrieve its value manually.
 	// If we didn't do this, the vertex format would be read as one bit offset from where it should be, making
 	// 01 become 00, and 10/11 become 01
-	const int tc[8] = {
+	const u32 tc[8] = {
 		m_VtxDesc.Tex0Coord, m_VtxDesc.Tex1Coord, m_VtxDesc.Tex2Coord, m_VtxDesc.Tex3Coord,
-		m_VtxDesc.Tex4Coord, m_VtxDesc.Tex5Coord, m_VtxDesc.Tex6Coord, (const int)((m_VtxDesc.Hex >> 31) & 3)
+		m_VtxDesc.Tex4Coord, m_VtxDesc.Tex5Coord, m_VtxDesc.Tex6Coord, (const u32)((m_VtxDesc.Hex >> 31) & 3)
 	};
 	
 	// Reset pipeline
@@ -474,7 +481,8 @@ void VertexLoader::WriteCall(TPipelineFunction func)
 	m_PipelineStages[m_numPipelineStages++] = func;
 #endif
 }
-
+// ARMTODO: This should be done in a better way
+#ifndef _M_GENERIC
 void VertexLoader::WriteGetVariable(int bits, OpArg dest, void *address)
 {
 #ifdef USE_JIT
@@ -498,7 +506,7 @@ void VertexLoader::WriteSetVariable(int bits, void *address, OpArg value)
 #endif
 #endif
 }
-
+#endif
 void VertexLoader::RunVertices(int vtx_attr_group, int primitive, int count)
 {
 	m_numLoadedVertices += count;
@@ -770,7 +778,7 @@ void VertexLoader::AppendToString(std::string *dest) const
 		dest->append(StringFromFormat("Nrm: %i %s-%s ",
 			m_VtxAttr.NormalElements, posMode[m_VtxDesc.Normal], posFormats[m_VtxAttr.NormalFormat]));
 	}
-	int color_mode[2] = {m_VtxDesc.Color0, m_VtxDesc.Color1};
+	u32 color_mode[2] = {m_VtxDesc.Color0, m_VtxDesc.Color1};
 	for (int i = 0; i < 2; i++)
 	{
 		if (color_mode[i])
@@ -778,7 +786,7 @@ void VertexLoader::AppendToString(std::string *dest) const
 			dest->append(StringFromFormat("C%i: %i %s-%s ", i, m_VtxAttr.color[i].Elements, posMode[color_mode[i]], colorFormat[m_VtxAttr.color[i].Comp]));
 		}
 	}
-	int tex_mode[8] = {
+	u32 tex_mode[8] = {
 		m_VtxDesc.Tex0Coord, m_VtxDesc.Tex1Coord, m_VtxDesc.Tex2Coord, m_VtxDesc.Tex3Coord, 
 		m_VtxDesc.Tex4Coord, m_VtxDesc.Tex5Coord, m_VtxDesc.Tex6Coord, m_VtxDesc.Tex7Coord
 	};
@@ -76,7 +76,12 @@ private:
 	}
 };

+// ARMTODO: This should be done in a better way
+#ifndef _M_GENERIC
 class VertexLoader : public Gen::XCodeBlock, NonCopyable
+#else
+class VertexLoader
+#endif
 {
 public:
 	VertexLoader(const TVtxDesc &vtx_desc, const VAT &vtx_attr);
@@ -122,8 +127,10 @@ private:

 	void WriteCall(TPipelineFunction);

+#ifndef _M_GENERIC
 	void WriteGetVariable(int bits, Gen::OpArg dest, void *address);
 	void WriteSetVariable(int bits, void *address, Gen::OpArg dest);
+#endif
 };									  

 #endif
@@ -19,9 +19,12 @@
 #ifdef _MSC_VER
 #include <hash_map>
 using stdext::hash_map;
-#else
+#elif defined __APPLE__
 #include <ext/hash_map>
 using __gnu_cxx::hash_map;
+#else
+#include <unordered_map>
+using std::unordered_map;
 #endif
 #include <map>
 #include <vector>
@@ -45,7 +48,12 @@ namespace stdext {
 	}
 }
 #else
-namespace __gnu_cxx {
+#ifdef __APPLE__
+namespace __gnu_cxx
+#else
+namespace std
+#endif
+{
 	template<> struct hash<VertexLoaderUID> {
 		size_t operator()(const VertexLoaderUID& uid) const {
 			return uid.GetHash();
@@ -54,10 +62,15 @@ namespace __gnu_cxx {
 }
 #endif

+#if defined _MSC_VER || defined __APPLE__
+typedef hash_map<VertexLoaderUID, VertexLoader*> VertexLoaderMap;
+#else
+typedef unordered_map<VertexLoaderUID, VertexLoader*> VertexLoaderMap;
+#endif
+
 namespace VertexLoaderManager
 {

-typedef hash_map<VertexLoaderUID, VertexLoader*> VertexLoaderMap;
 static VertexLoaderMap g_VertexLoaderMap;
 // TODO - change into array of pointers. Keep a map of all seen so far.

@@ -9,6 +9,7 @@
 #include "NativeVertexFormat.h"
 #include "TextureCacheBase.h"
 #include "RenderBase.h"
+#include "BPStructs.h"

 #include "VertexManagerBase.h"
 #include "VideoConfig.h"
@@ -159,6 +160,9 @@ void VertexManager::AddVertices(int primitive, int numVertices)

 void VertexManager::Flush()
 {
+	// loading a state will invalidate BP, so check for it
+	g_video_backend->CheckInvalidState();
+	
 	g_vertex_manager->vFlush();
 }

@@ -198,7 +202,7 @@ void VertexManager::Flush()
 	}

 	PRIM_LOG("pixel: tev=%d, ind=%d, texgen=%d, dstalpha=%d, alphafunc=0x%x", bpmem.genMode.numtevstages+1, bpmem.genMode.numindstages,
-		bpmem.genMode.numtexgens, (u32)bpmem.dstalpha.enable, (bpmem.alphaFunc.hex>>16)&0xff);
+		bpmem.genMode.numtexgens, (u32)bpmem.dstalpha.enable, (bpmem.alpha_test.hex>>16)&0xff);
 #endif

 	u32 usedtextures = 0;
@@ -21,7 +21,7 @@ public:

 		// values from DX11 backend
 		MAXVBUFFERSIZE = 0x50000,
-		MAXIBUFFERSIZE = 0x10000,
+		MAXIBUFFERSIZE = 0xFFFF,
 	};

 	VertexManager();
@@ -46,7 +46,8 @@ public:
 	static u8* GetVertexBuffer() { return LocalVBuffer; }

 	static void DoState(PointerWrap& p);
-
+	virtual void CreateDeviceObjects(){};
+	virtual void DestroyDeviceObjects(){};
 protected:
 	// TODO: make private after Flush() is merged
 	static void ResetBuffer();
@@ -465,6 +465,7 @@ const char *GenerateVertexShaderCode(u32 components, API_TYPE api_type)

 	//write the true depth value, if the game uses depth textures pixel shaders will override with the correct values
 	//if not early z culling will improve speed
+	// TODO: Can probably be dropped?
 	if (is_d3d)
 	{
 		WRITE(p, "o.pos.z = " I_DEPTHPARAMS".x * o.pos.w + o.pos.z * " I_DEPTHPARAMS".y;\n");
@@ -169,14 +169,24 @@ void VertexShaderManager::Shutdown()

 void VertexShaderManager::Dirty()
 {
-	nTransformMatricesChanged[0] = 0; nTransformMatricesChanged[1] = 256;
-	nNormalMatricesChanged[0] = 0; nNormalMatricesChanged[1] = 96;
-	nPostTransformMatricesChanged[0] = 0; nPostTransformMatricesChanged[1] = 256;
-	nLightsChanged[0] = 0; nLightsChanged[1] = 0x80;
+	nTransformMatricesChanged[0] = 0; 
+	nTransformMatricesChanged[1] = 256;
+	
+	nNormalMatricesChanged[0] = 0;
+	nNormalMatricesChanged[1] = 96;
+	
+	nPostTransformMatricesChanged[0] = 0; 
+	nPostTransformMatricesChanged[1] = 256;
+	
+	nLightsChanged[0] = 0; 
+	nLightsChanged[1] = 0x80;
+	
 	bPosNormalMatrixChanged = true;
-	bTexMatricesChanged[0] = bTexMatricesChanged[1] = true;
+	bTexMatricesChanged[0] = true;
+	bTexMatricesChanged[1] = true;
+	
 	bProjectionChanged = true;
-	bPosNormalMatrixChanged = bTexMatricesChanged[0] = bTexMatricesChanged[1] = true;
+	
 	nMaterialsChanged = 15;
 }

@@ -207,6 +217,7 @@ void VertexShaderManager::SetConstants()
 		int endn = (nPostTransformMatricesChanged[1] + 3 ) / 4;
 		const float* pstart = (const float*)&xfmem[XFMEM_POSTMATRICES + startn * 4];
 		SetMultiVSConstant4fv(C_POSTTRANSFORMMATRICES + startn, endn - startn, pstart);
+		nPostTransformMatricesChanged[0] = nPostTransformMatricesChanged[1] = -1;
 	}

 	if (nLightsChanged[0] >= 0)
@@ -250,7 +261,7 @@ void VertexShaderManager::SetConstants()
 		float GC_ALIGNED16(material[4]);
 		float NormalizationCoef = 1 / 255.0f;

-		for (int i = 0; i < 4; ++i)
+		for (int i = 0; i < 2; ++i)
 		{
 			if (nMaterialsChanged & (1 << i))
 			{
@@ -264,6 +275,21 @@ void VertexShaderManager::SetConstants()
 				SetVSConstant4fv(C_MATERIALS + i, material);
 			}
 		}
+		
+		for (int i = 0; i < 2; ++i)
+		{
+			if (nMaterialsChanged & (1 << (i + 2)))
+			{
+				u32 data = *(xfregs.matColor + i);
+
+				material[0] = ((data >> 24) & 0xFF) * NormalizationCoef;
+				material[1] = ((data >> 16) & 0xFF) * NormalizationCoef;
+				material[2] = ((data >>  8) & 0xFF) * NormalizationCoef;
+				material[3] = ( data        & 0xFF) * NormalizationCoef;
+
+				SetVSConstant4fv(C_MATERIALS + i + 2, material);
+			}
+		}

 		nMaterialsChanged = 0;
 	}
@@ -324,26 +350,28 @@ void VertexShaderManager::SetConstants()
 	if (bProjectionChanged)
 	{
 		bProjectionChanged = false;
+		
+		float *rawProjection = xfregs.projection.rawProjection;

-		if (xfregs.rawProjection[6] == 0)
+		switch(xfregs.projection.type)
 		{
-			// Perspective
+		case GX_PERSPECTIVE:
 			
-			g_fProjectionMatrix[0] = xfregs.rawProjection[0] * g_ActiveConfig.fAspectRatioHackW;
+			g_fProjectionMatrix[0] = rawProjection[0] * g_ActiveConfig.fAspectRatioHackW;
 			g_fProjectionMatrix[1] = 0.0f;
-			g_fProjectionMatrix[2] = xfregs.rawProjection[1];
+			g_fProjectionMatrix[2] = rawProjection[1];
 			g_fProjectionMatrix[3] = 0.0f;

 			g_fProjectionMatrix[4] = 0.0f;
-			g_fProjectionMatrix[5] = xfregs.rawProjection[2] * g_ActiveConfig.fAspectRatioHackH;
-			g_fProjectionMatrix[6] = xfregs.rawProjection[3];
+			g_fProjectionMatrix[5] = rawProjection[2] * g_ActiveConfig.fAspectRatioHackH;
+			g_fProjectionMatrix[6] = rawProjection[3];
 			g_fProjectionMatrix[7] = 0.0f;

 			g_fProjectionMatrix[8] = 0.0f;
 			g_fProjectionMatrix[9] = 0.0f;
-			g_fProjectionMatrix[10] = xfregs.rawProjection[4];
+			g_fProjectionMatrix[10] = rawProjection[4];

-			g_fProjectionMatrix[11] = xfregs.rawProjection[5];
+			g_fProjectionMatrix[11] = rawProjection[5];
 			
 			g_fProjectionMatrix[12] = 0.0f;
 			g_fProjectionMatrix[13] = 0.0f;
@@ -368,24 +396,24 @@ void VertexShaderManager::SetConstants()
 			SETSTAT_FT(stats.gproj_13, g_fProjectionMatrix[13]);
 			SETSTAT_FT(stats.gproj_14, g_fProjectionMatrix[14]);
 			SETSTAT_FT(stats.gproj_15, g_fProjectionMatrix[15]);
-		}
-		else
-		{ 
-			// Orthographic Projection
-			g_fProjectionMatrix[0] = xfregs.rawProjection[0];
+			break;
+			
+		case GX_ORTHOGRAPHIC:
+			
+			g_fProjectionMatrix[0] = rawProjection[0];
 			g_fProjectionMatrix[1] = 0.0f;
 			g_fProjectionMatrix[2] = 0.0f;
-			g_fProjectionMatrix[3] = xfregs.rawProjection[1];
+			g_fProjectionMatrix[3] = rawProjection[1];

 			g_fProjectionMatrix[4] = 0.0f;
-			g_fProjectionMatrix[5] = xfregs.rawProjection[2];
+			g_fProjectionMatrix[5] = rawProjection[2];
 			g_fProjectionMatrix[6] = 0.0f;
-			g_fProjectionMatrix[7] = xfregs.rawProjection[3];
+			g_fProjectionMatrix[7] = rawProjection[3];

 			g_fProjectionMatrix[8] = 0.0f;
 			g_fProjectionMatrix[9] = 0.0f;
-			g_fProjectionMatrix[10] = (g_ProjHack1.value + xfregs.rawProjection[4]) * ((g_ProjHack1.sign == 0) ? 1.0f : g_ProjHack1.sign);
-			g_fProjectionMatrix[11] = (g_ProjHack2.value + xfregs.rawProjection[5]) * ((g_ProjHack2.sign == 0) ? 1.0f : g_ProjHack2.sign);
+			g_fProjectionMatrix[10] = (g_ProjHack1.value + rawProjection[4]) * ((g_ProjHack1.sign == 0) ? 1.0f : g_ProjHack1.sign);
+			g_fProjectionMatrix[11] = (g_ProjHack2.value + rawProjection[5]) * ((g_ProjHack2.sign == 0) ? 1.0f : g_ProjHack2.sign);

 			g_fProjectionMatrix[12] = 0.0f;
 			g_fProjectionMatrix[13] = 0.0f;
@@ -398,7 +426,7 @@ void VertexShaderManager::SetConstants()
 			*/
 			
 			g_fProjectionMatrix[14] = 0.0f;
-			g_fProjectionMatrix[15] = (g_ProjHack3 && xfregs.rawProjection[0] == 2.0f ? 0.0f : 1.0f);  //causes either the efb copy or bloom layer not to show if proj hack enabled
+			g_fProjectionMatrix[15] = (g_ProjHack3 && rawProjection[0] == 2.0f ? 0.0f : 1.0f);  //causes either the efb copy or bloom layer not to show if proj hack enabled
 		
 			SETSTAT_FT(stats.g2proj_0, g_fProjectionMatrix[0]);
 			SETSTAT_FT(stats.g2proj_1, g_fProjectionMatrix[1]);
@@ -416,18 +444,21 @@ void VertexShaderManager::SetConstants()
 			SETSTAT_FT(stats.g2proj_13, g_fProjectionMatrix[13]);
 			SETSTAT_FT(stats.g2proj_14, g_fProjectionMatrix[14]);
 			SETSTAT_FT(stats.g2proj_15, g_fProjectionMatrix[15]);
-			SETSTAT_FT(stats.proj_0, xfregs.rawProjection[0]);
-			SETSTAT_FT(stats.proj_1, xfregs.rawProjection[1]);
-			SETSTAT_FT(stats.proj_2, xfregs.rawProjection[2]);
-			SETSTAT_FT(stats.proj_3, xfregs.rawProjection[3]);
-			SETSTAT_FT(stats.proj_4, xfregs.rawProjection[4]);
-			SETSTAT_FT(stats.proj_5, xfregs.rawProjection[5]);
-			SETSTAT_FT(stats.proj_6, xfregs.rawProjection[6]);
+			SETSTAT_FT(stats.proj_0, rawProjection[0]);
+			SETSTAT_FT(stats.proj_1, rawProjection[1]);
+			SETSTAT_FT(stats.proj_2, rawProjection[2]);
+			SETSTAT_FT(stats.proj_3, rawProjection[3]);
+			SETSTAT_FT(stats.proj_4, rawProjection[4]);
+			SETSTAT_FT(stats.proj_5, rawProjection[5]);
+			break;
+			
+		default:
+			ERROR_LOG(VIDEO, "unknown projection type: %d", xfregs.projection.type);
 		}

-		PRIM_LOG("Projection: %f %f %f %f %f %f\n", xfregs.rawProjection[0], xfregs.rawProjection[1], xfregs.rawProjection[2], xfregs.rawProjection[3], xfregs.rawProjection[4], xfregs.rawProjection[5]);
+		PRIM_LOG("Projection: %f %f %f %f %f %f\n", rawProjection[0], rawProjection[1], rawProjection[2], rawProjection[3], rawProjection[4], rawProjection[5]);

-		if ((g_ActiveConfig.bFreeLook || g_ActiveConfig.bAnaglyphStereo ) && xfregs.rawProjection[6] == 0)
+		if ((g_ActiveConfig.bFreeLook || g_ActiveConfig.bAnaglyphStereo ) && xfregs.projection.type == GX_PERSPECTIVE)
 		{
 			Matrix44 mtxA;
 			Matrix44 mtxB;
@@ -44,17 +44,14 @@ enum
 	EFB_HEIGHT = 528,
 };

-enum
-{
-	// XFB width is decided by EFB copy operation. The VI can do horizontal
-	// scaling (TODO: emulate).
-	MAX_XFB_WIDTH = EFB_WIDTH,
+// XFB width is decided by EFB copy operation. The VI can do horizontal
+// scaling (TODO: emulate).
+const u32 MAX_XFB_WIDTH = EFB_WIDTH;

-	// Although EFB height is 528, 574-line XFB's can be created either with
-	// vertical scaling by the EFB copy operation or copying to multiple XFB's
-	// that are next to each other in memory (TODO: handle that situation).
-	MAX_XFB_HEIGHT = 574
-};
+// Although EFB height is 528, 574-line XFB's can be created either with
+// vertical scaling by the EFB copy operation or copying to multiple XFB's
+// that are next to each other in memory (TODO: handle that situation).
+const u32 MAX_XFB_HEIGHT = 574;

 // Logging
 // ----------
@@ -42,6 +42,7 @@ VideoConfig::VideoConfig()
 	// disable all features by default
 	backend_info.APIType = API_NONE;
 	backend_info.bUseRGBATextures = false;
+	backend_info.bUseMinimalMipCount = false;
 	backend_info.bSupports3DVision = false;
 }

@@ -59,6 +60,7 @@ void VideoConfig::Load(const char *ini_file)
 	iniFile.Get("Settings", "UseRealXFB", &bUseRealXFB, 0);
 	iniFile.Get("Settings", "SafeTextureCacheColorSamples", &iSafeTextureCache_ColorSamples,128);
 	iniFile.Get("Settings", "ShowFPS", &bShowFPS, false); // Settings
+	iniFile.Get("Settings", "LogFPSToFile", &bLogFPSToFile, false);
 	iniFile.Get("Settings", "ShowInputDisplay", &bShowInputDisplay, false);
 	iniFile.Get("Settings", "OverlayStats", &bOverlayStats, false);
 	iniFile.Get("Settings", "OverlayProjStats", &bOverlayProjStats, false);
@@ -74,7 +76,6 @@ void VideoConfig::Load(const char *ini_file)
 	iniFile.Get("Settings", "AnaglyphStereoSeparation", &iAnaglyphStereoSeparation, 200);
 	iniFile.Get("Settings", "AnaglyphFocalAngle", &iAnaglyphFocalAngle, 0);
 	iniFile.Get("Settings", "EnablePixelLighting", &bEnablePixelLighting, 0);
-	iniFile.Get("Settings", "EnablePerPixelDepth", &bEnablePerPixelDepth, 0);
 	
 	iniFile.Get("Settings", "MSAA", &iMultisampleMode, 0);
 	iniFile.Get("Settings", "EFBScale", &iEFBScale, 2); // native
@@ -134,7 +135,6 @@ void VideoConfig::GameIniLoad(const char *ini_file)
 	iniFile.GetIfExists("Video_Settings", "AnaglyphStereoSeparation", &iAnaglyphStereoSeparation);
 	iniFile.GetIfExists("Video_Settings", "AnaglyphFocalAngle", &iAnaglyphFocalAngle);
 	iniFile.GetIfExists("Video_Settings", "EnablePixelLighting", &bEnablePixelLighting);
-	iniFile.GetIfExists("Video_Settings", "EnablePerPixelDepth", &bEnablePerPixelDepth);
 	iniFile.GetIfExists("Video_Settings", "MSAA", &iMultisampleMode);
 	iniFile.GetIfExists("Video_Settings", "EFBScale", &iEFBScale); // integral
 	iniFile.GetIfExists("Video_Settings", "DstAlphaPass", &bDstAlphaPass);
@@ -189,6 +189,7 @@ void VideoConfig::Save(const char *ini_file)
 	iniFile.Set("Settings", "UseRealXFB", bUseRealXFB);
 	iniFile.Set("Settings", "SafeTextureCacheColorSamples", iSafeTextureCache_ColorSamples);
 	iniFile.Set("Settings", "ShowFPS", bShowFPS);
+	iniFile.Set("Settings", "LogFPSToFile", bLogFPSToFile);
 	iniFile.Set("Settings", "ShowInputDisplay", bShowInputDisplay);
 	iniFile.Set("Settings", "OverlayStats", bOverlayStats);
 	iniFile.Set("Settings", "OverlayProjStats", bOverlayProjStats);
@@ -204,7 +205,6 @@ void VideoConfig::Save(const char *ini_file)
 	iniFile.Set("Settings", "AnaglyphStereoSeparation", iAnaglyphStereoSeparation);
 	iniFile.Set("Settings", "AnaglyphFocalAngle", iAnaglyphFocalAngle);
 	iniFile.Set("Settings", "EnablePixelLighting", bEnablePixelLighting);
-	iniFile.Set("Settings", "EnablePerPixelDepth", bEnablePerPixelDepth);
 	

 	iniFile.Set("Settings", "ShowEFBCopyRegions", bShowEFBCopyRegions);
@@ -271,7 +271,6 @@ void VideoConfig::GameIniSave(const char* default_ini, const char* game_ini)
 	SET_IF_DIFFERS("Video_Settings", "AnaglyphStereoSeparation", iAnaglyphStereoSeparation);
 	SET_IF_DIFFERS("Video_Settings", "AnaglyphFocalAngle", iAnaglyphFocalAngle);
 	SET_IF_DIFFERS("Video_Settings", "EnablePixelLighting", bEnablePixelLighting);
-	SET_IF_DIFFERS("Video_Settings", "EnablePerPixelDepth", bEnablePerPixelDepth);
 	SET_IF_DIFFERS("Video_Settings", "MSAA", iMultisampleMode);
 	SET_IF_DIFFERS("Video_Settings", "EFBScale", iEFBScale); // integral
 	SET_IF_DIFFERS("Video_Settings", "DstAlphaPass", bDstAlphaPass);
@@ -295,125 +294,3 @@ void VideoConfig::GameIniSave(const char* default_ini, const char* game_ini)

 	iniFile.Save(game_ini);
 }
-
-
-// TODO: remove
-extern bool g_aspect_wide;
-
-// TODO: Figure out a better place for this function.
-void ComputeDrawRectangle(int backbuffer_width, int backbuffer_height, bool flip, TargetRectangle *rc)
-{
-	float FloatGLWidth = (float)backbuffer_width;
-	float FloatGLHeight = (float)backbuffer_height;
-	float FloatXOffset = 0;
-	float FloatYOffset = 0;
-	
-	// The rendering window size
-	const float WinWidth = FloatGLWidth;
-	const float WinHeight = FloatGLHeight;
-	
-	// Handle aspect ratio.
-	// Default to auto.
-	bool use16_9 = g_aspect_wide;
-	
-	// Update aspect ratio hack values
-	// Won't take effect until next frame
-	// Don't know if there is a better place for this code so there isn't a 1 frame delay
-	if ( g_ActiveConfig.bWidescreenHack )
-	{
-		float source_aspect = use16_9 ? (16.0f / 9.0f) : (4.0f / 3.0f);
-		float target_aspect;
-		
-		switch ( g_ActiveConfig.iAspectRatio )
-		{
-		case ASPECT_FORCE_16_9 :
-			target_aspect = 16.0f / 9.0f;
-			break;
-		case ASPECT_FORCE_4_3 :
-			target_aspect = 4.0f / 3.0f;
-			break;
-		case ASPECT_STRETCH :
-			target_aspect = WinWidth / WinHeight;
-			break;
-		default :
-			// ASPECT_AUTO == no hacking
-			target_aspect = source_aspect;
-			break;
-		}
-		
-		float adjust = source_aspect / target_aspect;
-		if ( adjust > 1 )
-		{
-			// Vert+
-			g_Config.fAspectRatioHackW = 1;
-			g_Config.fAspectRatioHackH = 1/adjust;
-		}
-		else
-		{
-			// Hor+
-			g_Config.fAspectRatioHackW = adjust;
-			g_Config.fAspectRatioHackH = 1;
-		}
-	}
-	else
-	{
-		// Hack is disabled
-		g_Config.fAspectRatioHackW = 1;
-		g_Config.fAspectRatioHackH = 1;
-	}
-	
-	// Check for force-settings and override.
-	if (g_ActiveConfig.iAspectRatio == ASPECT_FORCE_16_9)
-		use16_9 = true;
-	else if (g_ActiveConfig.iAspectRatio == ASPECT_FORCE_4_3)
-		use16_9 = false;
-	
-	if (g_ActiveConfig.iAspectRatio != ASPECT_STRETCH)
-	{
-		// The rendering window aspect ratio as a proportion of the 4:3 or 16:9 ratio
-		float Ratio = (WinWidth / WinHeight) / (!use16_9 ? (4.0f / 3.0f) : (16.0f / 9.0f));
-		// Check if height or width is the limiting factor. If ratio > 1 the picture is too wide and have to limit the width.
-		if (Ratio > 1.0f)
-		{
-			// Scale down and center in the X direction.
-			FloatGLWidth /= Ratio;
-			FloatXOffset = (WinWidth - FloatGLWidth) / 2.0f;
-		}
-		// The window is too high, we have to limit the height
-		else
-		{
-			// Scale down and center in the Y direction.
-			FloatGLHeight *= Ratio;
-			FloatYOffset = FloatYOffset + (WinHeight - FloatGLHeight) / 2.0f;
-		}
-	}
-	
-	// -----------------------------------------------------------------------
-	// Crop the picture from 4:3 to 5:4 or from 16:9 to 16:10.
-	//		Output: FloatGLWidth, FloatGLHeight, FloatXOffset, FloatYOffset
-	// ------------------
-	if (g_ActiveConfig.iAspectRatio != ASPECT_STRETCH && g_ActiveConfig.bCrop)
-	{
-		float Ratio = !use16_9 ? ((4.0f / 3.0f) / (5.0f / 4.0f)) : (((16.0f / 9.0f) / (16.0f / 10.0f)));
-		// The width and height we will add (calculate this before FloatGLWidth and FloatGLHeight is adjusted)
-		float IncreasedWidth = (Ratio - 1.0f) * FloatGLWidth;
-		float IncreasedHeight = (Ratio - 1.0f) * FloatGLHeight;
-		// The new width and height
-		FloatGLWidth = FloatGLWidth * Ratio;
-		FloatGLHeight = FloatGLHeight * Ratio;
-		// Adjust the X and Y offset
-		FloatXOffset = FloatXOffset - (IncreasedWidth * 0.5f);
-		FloatYOffset = FloatYOffset - (IncreasedHeight * 0.5f);
-	}
-	
-	int XOffset = (int)(FloatXOffset + 0.5f);
-	int YOffset = (int)(FloatYOffset + 0.5f);
-	int iWhidth = (int)ceil(FloatGLWidth);
-	int iHeight = (int)ceil(FloatGLHeight);
-	iWhidth -= iWhidth % 4; // ensure divisibility by 4 to make it compatible with all the video encoders
-	iHeight -= iHeight % 4;
-	rc->left = XOffset;
-	rc->top = flip ? (int)(YOffset + iHeight) : YOffset;
-	rc->right = XOffset + iWhidth;
-	rc->bottom = flip ? YOffset : (int)(YOffset + iHeight);
-}
@@ -97,6 +97,7 @@ struct VideoConfig
 	bool bTexFmtOverlayEnable;
 	bool bTexFmtOverlayCenter;
 	bool bShowEFBCopyRegions;
+	bool bLogFPSToFile;
 	
 	// Render
 	bool bWireFrame;
@@ -132,7 +133,6 @@ struct VideoConfig
 	bool bZTPSpeedHack; // The Legend of Zelda: Twilight Princess
 	bool bUseBBox;
 	bool bEnablePixelLighting;
-	bool bEnablePerPixelDepth;
 	bool bDisablePixelPerf;

 	int iLog; // CONF_ bits
@@ -158,12 +158,19 @@ struct VideoConfig
 		std::vector<std::string> PPShaders; // post-processing shaders

 		bool bUseRGBATextures; // used for D3D11 in TextureCache
+		bool bUseMinimalMipCount;
 		bool bSupports3DVision;
 		bool bSupportsDualSourceBlend; // only supported by D3D11 and OpenGL
 		bool bSupportsFormatReinterpretation;
 		bool bSupportsPixelLighting;
 		bool bSupportsPixelPerfQuery;
 	} backend_info;
+
+    // Utility
+    bool RealXFBEnabled() const { return bUseXFB && bUseRealXFB; }
+    bool VirtualXFBEnabled() const { return bUseXFB && !bUseRealXFB; }
+    bool EFBCopiesToTextureEnabled() const { return bEFBCopyEnable && bCopyEFBToTexture; }
+    bool EFBCopiesToRamEnabled() const { return bEFBCopyEnable && !bCopyEFBToTexture; }
 };

 extern VideoConfig g_Config;
@@ -172,6 +179,4 @@ extern VideoConfig g_ActiveConfig;
 // Called every frame.
 void UpdateActiveConfig();

-void ComputeDrawRectangle(int backbuffer_width, int backbuffer_height, bool flip, TargetRectangle *rc);
-
 #endif  // _VIDEO_CONFIG_H_
@@ -61,6 +61,9 @@
 #define LIGHTATTN_NONE 2
 #define LIGHTATTN_DIR  3

+#define GX_PERSPECTIVE  0
+#define GX_ORTHOGRAPHIC 1
+
 #define XFMEM_SIZE               0x8000
 #define XFMEM_POSMATRICES        0x000
 #define XFMEM_POSMATRICES_END    0x100
@@ -235,6 +238,12 @@ struct Viewport
 	float farZ;
 };

+struct Projection
+{
+	float rawProjection[6];
+	u32 type;                      // only GX_PERSPECTIVE or GX_ORTHOGRAPHIC are allowed
+};
+
 struct XFRegisters
 {
    u32 error;                      // 0x1000
@@ -257,10 +266,10 @@ struct XFRegisters
    u32 unk5;                       // 0x1015
    u32 unk6;                       // 0x1016
    u32 unk7;                       // 0x1017
-    u32 MatrixIndexA;				// 0x1018
-    u32 MatrixIndexB;				// 0x1019
+    u32 MatrixIndexA;               // 0x1018
+    u32 MatrixIndexB;               // 0x1019
    Viewport viewport;              // 0x101a - 0x101f
-    float rawProjection[7];         // 0x1020 - 0x1026
+    Projection projection;          // 0x1020 - 0x1026
    u32 unk8[24];                   // 0x1027 - 0x103e
    NumTexGen numTexGen;            // 0x103f
    TexMtxInfo texMtxInfo[8];       // 0x1040 - 0x1047
@@ -121,18 +121,12 @@ void XFRegWritten(int transferSize, u32 baseAddress, u32 *pData)
 		case XFMEM_SETVIEWPORT+3:
 		case XFMEM_SETVIEWPORT+4:
 		case XFMEM_SETVIEWPORT+5:
-			{
-				u8 size = std::min(transferSize * 4, 6 * 4);
-				if (memcmp((u32*)&xfregs + (address - 0x1000), pData + dataIndex, size))
-				{
-					VertexManager::Flush();
-					VertexShaderManager::SetViewportChanged();
-					PixelShaderManager::SetViewportChanged();
-				}
+			VertexManager::Flush();
+			VertexShaderManager::SetViewportChanged();
+			PixelShaderManager::SetViewportChanged();

-				nextAddress = XFMEM_SETVIEWPORT + 6;
-				break;
-			}
+			nextAddress = XFMEM_SETVIEWPORT + 6;
+			break;

 		case XFMEM_SETPROJECTION:
 		case XFMEM_SETPROJECTION+1:
@@ -141,21 +135,15 @@ void XFRegWritten(int transferSize, u32 baseAddress, u32 *pData)
 		case XFMEM_SETPROJECTION+4:
 		case XFMEM_SETPROJECTION+5:
 		case XFMEM_SETPROJECTION+6:
-			{
-				u8 size = std::min(transferSize * 4, 7 * 4);
-				if (memcmp((u32*)&xfregs + (address - 0x1000), pData + dataIndex, size))
-				{
-					VertexManager::Flush();
-					VertexShaderManager::SetProjectionChanged();
-				}
+			VertexManager::Flush();			
+			VertexShaderManager::SetProjectionChanged();

-				nextAddress = XFMEM_SETPROJECTION + 7;
-				break;
-			}
+			nextAddress = XFMEM_SETPROJECTION + 7;
+			break;

        case XFMEM_SETNUMTEXGENS: // GXSetNumTexGens
 			if (xfregs.numTexGen.numTexGens != (newValue & 15))
-                VertexManager::Flush();
+				VertexManager::Flush();
            break;

 		case XFMEM_SETTEXMTXINFO:
@@ -166,16 +154,10 @@ void XFRegWritten(int transferSize, u32 baseAddress, u32 *pData)
 		case XFMEM_SETTEXMTXINFO+5:
 		case XFMEM_SETTEXMTXINFO+6:
 		case XFMEM_SETTEXMTXINFO+7:
-			{
-				u8 size = std::min(transferSize * 4, 8 * 4);
-				if (memcmp((u32*)&xfregs + (address - 0x1000), pData + dataIndex, size))
-				{
-					VertexManager::Flush();
-				}
+			VertexManager::Flush();

-				nextAddress = XFMEM_SETTEXMTXINFO + 8;
-				break;
-			}
+			nextAddress = XFMEM_SETTEXMTXINFO + 8;
+			break;

 		case XFMEM_SETPOSMTXINFO:
 		case XFMEM_SETPOSMTXINFO+1:
@@ -185,16 +167,10 @@ void XFRegWritten(int transferSize, u32 baseAddress, u32 *pData)
 		case XFMEM_SETPOSMTXINFO+5:
 		case XFMEM_SETPOSMTXINFO+6:
 		case XFMEM_SETPOSMTXINFO+7:
-			{
-				u8 size = std::min(transferSize * 4, 8 * 4);
-				if (memcmp((u32*)&xfregs + (address - 0x1000), pData + dataIndex, size))
-				{
-					VertexManager::Flush();
-				}
+			VertexManager::Flush();

-				nextAddress = XFMEM_SETPOSMTXINFO + 8;
-				break;
-			}
+			nextAddress = XFMEM_SETPOSMTXINFO + 8;
+			break;

 		// --------------
 		// Unknown Regs
@@ -264,15 +240,8 @@ void LoadXFReg(u32 transferSize, u32 baseAddress, u32 *pData)
 			transferSize = 0;
 		}
 		
-		for (u32 i = 0; i < xfMemTransferSize; ++i)
-		{
-			if (((u32*)&xfmem[xfMemBase])[i] != pData[i])
-			{
-				XFMemWritten(xfMemTransferSize, xfMemBase);
-				memcpy_gc(&xfmem[xfMemBase], pData, xfMemTransferSize * 4);
-				break;
-			}
-		}
+		XFMemWritten(xfMemTransferSize, xfMemBase);
+		memcpy_gc(&xfmem[xfMemBase], pData, xfMemTransferSize * 4);

 		pData += xfMemTransferSize;
 	}
@@ -35,7 +35,7 @@
 #include "VertexLoaderManager.h"
 #include "VertexManagerBase.h"
 #include "x64Emitter.h"
-#include "ABI.h"
+#include "x64ABI.h"

 #include "DLCache.h"
 #include "VideoConfig.h"
@@ -361,9 +361,9 @@ u32 AnalyzeAndRunDisplayList(u32 address, u32 size, CachedDisplayList *dl)
 							(cmd_byte & GX_PRIMITIVE_MASK) >> GX_PRIMITIVE_SHIFT,
 							numVertices);
 						num_draw_call++;
-						const int tc[12] = {
+						const u32 tc[12] = {
 							g_VtxDesc.Position, g_VtxDesc.Normal, g_VtxDesc.Color0, g_VtxDesc.Color1, g_VtxDesc.Tex0Coord, g_VtxDesc.Tex1Coord, 
-							g_VtxDesc.Tex2Coord, g_VtxDesc.Tex3Coord, g_VtxDesc.Tex4Coord, g_VtxDesc.Tex5Coord, g_VtxDesc.Tex6Coord, (const int)((g_VtxDesc.Hex >> 31) & 3)
+							g_VtxDesc.Tex2Coord, g_VtxDesc.Tex3Coord, g_VtxDesc.Tex4Coord, g_VtxDesc.Tex5Coord, g_VtxDesc.Tex6Coord, (const u32)((g_VtxDesc.Hex >> 31) & 3)
 						};
 						for(int i = 0; i < 12; i++)
 						{
@@ -563,9 +563,9 @@ void CompileAndRunDisplayList(u32 address, u32 size, CachedDisplayList *dl)
 							dl->InsertRegion(NewRegion);
 							memcpy(NewRegion->start_address, StartAddress, Vdatasize);
 							emitter.ABI_CallFunctionCCCP((void *)&VertexLoaderManager::RunCompiledVertices, cmd_byte & GX_VAT_MASK, (cmd_byte & GX_PRIMITIVE_MASK) >> GX_PRIMITIVE_SHIFT, numVertices, NewRegion->start_address);
-							const int tc[12] = {
+							const u32 tc[12] = {
 							g_VtxDesc.Position, g_VtxDesc.Normal, g_VtxDesc.Color0, g_VtxDesc.Color1, g_VtxDesc.Tex0Coord, g_VtxDesc.Tex1Coord, 
-							g_VtxDesc.Tex2Coord, g_VtxDesc.Tex3Coord, g_VtxDesc.Tex4Coord, g_VtxDesc.Tex5Coord, g_VtxDesc.Tex6Coord, (const int)((g_VtxDesc.Hex >> 31) & 3)
+							g_VtxDesc.Tex2Coord, g_VtxDesc.Tex3Coord, g_VtxDesc.Tex4Coord, g_VtxDesc.Tex5Coord, g_VtxDesc.Tex6Coord, (const u32)((g_VtxDesc.Hex >> 31) & 3)
 							};
 							for(int i = 0; i < 12; i++)
 							{
@@ -692,7 +692,7 @@ inline void SetOpenMPThreadCount(int width, int height)
 	if (g_ActiveConfig.bOMPDecoder && width > 127 && height > 127)
 	{
 		// don't span to many threads they will kill the rest of the emu :)
-		omp_set_num_threads((cpu_info.num_cores + 2) / 3);
+		omp_set_num_threads((omp_get_num_procs() + 2) / 3);
 	}
 	else
 	{
@@ -1119,20 +1119,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he
 							_mm_storeu_si128( (__m128i*)( dst+(y + iy+1) * width + x + 4 ), o4 );
 						}
 			}
-#if 0
-			// Reference C implementation:
-			for (int y = 0; y < height; y += 8)
-				for (int x = 0; x < width; x += 8)
-					for (int iy = 0; iy < 8; iy++, src += 4)
-						for (int ix = 0; ix < 4; ix++)
-						{
-							int val = src[ix];
-							u8 i1 = Convert4To8(val >> 4);
-							u8 i2 = Convert4To8(val & 0xF);
-							memset(dst+(y + iy) * width + x + ix * 2 , i1,4);
-							memset(dst+(y + iy) * width + x + ix * 2 + 1 , i2,4);
-						}
-#endif
 		}
 	   break;
 	case GX_TF_I8:  // speed critical
@@ -1248,26 +1234,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he
 						
 					}
 			}
-#if 0
-			// Reference C implementation
-			for (int y = 0; y < height; y += 4)
-				for (int x = 0; x < width; x += 8)
-					for (int iy = 0; iy < 4; ++iy, src += 8)
-					{
-						u32 *  newdst = dst + (y + iy)*width+x;
-						const u8 *  newsrc = src;
-						u8 srcval;
-
-						srcval = (newsrc++)[0]; (newdst++)[0] = srcval | (srcval << 8) | (srcval << 16) | (srcval << 24);
-						srcval = (newsrc++)[0]; (newdst++)[0] = srcval | (srcval << 8) | (srcval << 16) | (srcval << 24);
-						srcval = (newsrc++)[0]; (newdst++)[0] = srcval | (srcval << 8) | (srcval << 16) | (srcval << 24);
-						srcval = (newsrc++)[0]; (newdst++)[0] = srcval | (srcval << 8) | (srcval << 16) | (srcval << 24);
-						srcval = (newsrc++)[0]; (newdst++)[0] = srcval | (srcval << 8) | (srcval << 16) | (srcval << 24);
-						srcval = (newsrc++)[0]; (newdst++)[0] = srcval | (srcval << 8) | (srcval << 16) | (srcval << 24);
-						srcval = (newsrc++)[0]; (newdst++)[0] = srcval | (srcval << 8) | (srcval << 16) | (srcval << 24);
-						srcval = newsrc[0]; newdst[0] = srcval | (srcval << 8) | (srcval << 16) | (srcval << 24);
-					}
-#endif
 		}
 		break;
 	case GX_TF_C8:
@@ -1380,20 +1346,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he
 							_mm_storeu_si128( (__m128i*)(dst + (y + iy) * width + x), r1 );
 						}
 			}
-#if 0
-			// Reference C implementation:
-			for (int y = 0; y < height; y += 4)
-				for (int x = 0; x < width; x += 4)
-					for (int iy = 0; iy < 4; iy++, src += 8)
-					{
-						u32 *ptr = dst + (y + iy) * width + x;
-						u16 *s = (u16 *)src;
-						ptr[0] = decodeIA8Swapped(s[0]);
-						ptr[1] = decodeIA8Swapped(s[1]);
-						ptr[2] = decodeIA8Swapped(s[2]);
-						ptr[3] = decodeIA8Swapped(s[3]);
-					}
-#endif
 		}
 		break;
 	case GX_TF_C14X2:
@@ -1493,18 +1445,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he
 						__m128i *ptr = (__m128i *)(dst + (y + iy) * width + x);
 						_mm_storeu_si128(ptr, abgr888x4);
 					}
-#if 0
-			// Reference C implementation.
-			for (int y = 0; y < height; y += 4)
-				for (int x = 0; x < width; x += 4)
-					for (int iy = 0; iy < 4; iy++, src += 8)
-					{
-						u32 *ptr = dst + (y + iy) * width + x;
-						u16 *s = (u16 *)src;
-						for(int j = 0; j < 4; j++)
-							*ptr++ = decode565RGBA(Common::swap16(*s++));
-					}
-#endif
 		}
 		break;
 	case GX_TF_RGB5A3:
@@ -1718,13 +1658,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he
 							}
 						}
 				}
-#if 0
-			// Reference C implementation:
-			for (int y = 0; y < height; y += 4)
-				for (int x = 0; x < width; x += 4)
-					for (int iy = 0; iy < 4; iy++, src += 8)
-						decodebytesRGB5A3rgba(dst+(y+iy)*width+x, (u16*)src);
-#endif
 		}
 		break;
 	case GX_TF_RGBA8:  // speed critical
@@ -1860,16 +1793,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he
 						_mm_storeu_si128(dst128, rgba11);
 					}				
 			}
-#if 0
-			// Reference C implementation.
-			for (int y = 0; y < height; y += 4)
-				for (int x = 0; x < width; x += 4)
-				{
-					for (int iy = 0; iy < 4; iy++)
-						decodebytesARGB8_4ToRgba(dst + (y+iy)*width + x, (u16*)src + 4 * iy, (u16*)src + 4 * iy + 16);
-					src += 64;
-				}
-#endif
 		}
 		break;
 	case GX_TF_CMPR:  // speed critical
@@ -2104,22 +2027,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he
 					}					
 				}
 			}
-#if 0
-			for (int y = 0; y < height; y += 8)
-			{
-				for (int x = 0; x < width; x += 8)
-				{
-					decodeDXTBlockRGBA((u32*)dst + y * width + x, (DXTBlock*)src, width);
-										src += sizeof(DXTBlock);
-					decodeDXTBlockRGBA((u32*)dst + y * width + x + 4, (DXTBlock*)src, width);
-										src += sizeof(DXTBlock);
-					decodeDXTBlockRGBA((u32*)dst + (y + 4) * width + x, (DXTBlock*)src, width);
-										src += sizeof(DXTBlock);
-					decodeDXTBlockRGBA((u32*)dst + (y + 4) * width + x + 4, (DXTBlock*)src, width);
-										src += sizeof(DXTBlock);
-				}
-			}
-#endif
 			break;
 		}
 	}
@@ -2515,6 +2422,40 @@ void TexDecoder_DecodeTexel(u8 *dst, const u8 *src, int s, int t, int imageWidth
 	}
 }

+void TexDecoder_DecodeTexelRGBA8FromTmem(u8 *dst, const u8 *src_ar, const u8* src_gb, int s, int t, int imageWidth)
+{
+	u16 sBlk = s >> 2;
+	u16 tBlk = t >> 2;
+	u16 widthBlks = (imageWidth >> 2) + 1; // TODO: Looks wrong. Shouldn't this be ((imageWidth-1)>>2)+1 ?
+	u32 base_ar = (tBlk * widthBlks + sBlk) << 4;
+	u32 base_gb = (tBlk * widthBlks + sBlk) << 4;
+	u16 blkS = s & 3;
+	u16 blkT =  t & 3;
+	u32 blk_off = (blkT << 2) + blkS;
+
+	u32 offset_ar = (base_ar + blk_off) << 1;
+	u32 offset_gb = (base_gb + blk_off) << 1;
+	const u8* val_addr_ar = src_ar + offset_ar;
+	const u8* val_addr_gb = src_gb + offset_gb;
+
+	dst[3] = val_addr_ar[0]; // A
+	dst[0] = val_addr_ar[1]; // R
+	dst[1] = val_addr_gb[0]; // G
+	dst[2] = val_addr_gb[1]; // B
+}
+
+PC_TexFormat TexDecoder_DecodeRGBA8FromTmem(u8* dst, const u8 *src_ar, const u8 *src_gb, int width, int height)
+{
+	// TODO for someone who cares: Make this less slow!
+	for (int y = 0; y < height; ++y)
+		for (int x = 0; x < width; ++x)
+		{
+			TexDecoder_DecodeTexelRGBA8FromTmem(dst, src_ar, src_gb, x, y, width-1);
+			dst += 4;
+		}
+
+	return PC_TEX_FMT_RGBA32;
+}

 const char* texfmt[] = {
 	// pixel