microVU:

- Added a new status-flag speedhack after learning how evil sVU is with status flags (mVU with both flag-speedhacks ON is less evil than sVU's default setting which does something even hackier)
- Minor Changes

git-svn-id: http://pcsx2.googlecode.com/svn/trunk@1246 96395faa-99c1-11dd-bbfe-3dabce05a288

pcsx2/x86/microVU_Analyze.inl

@@ -291,6 +291,7 @@ microVUt(void) mVUanalyzeSflag(int It) {
 	if (!It) { mVUinfo |= _isNOP; }
 	else {
 		mVUinfo |= _swapOps;
+		mVUsFlagHack = 0; // Don't Optimize Out Status Flags for this block
 		if (mVUcount < 4)	{ mVUpBlock->pState.needExactMatch |= 0xf /*<< mVUcount*/; }
 		if (mVUcount >= 1)	{ incPC2(-2); mVUinfo |= _isSflag; incPC2(2); }
 		// Note: _isSflag is used for status flag optimizations.

pcsx2/x86/microVU_Compile.inl

@@ -237,6 +237,7 @@ microVUt(void*) __fastcall mVUcompile(u32 startPC, uptr pState) {
 	mVUpBlock		= pBlock;
 	mVUregs.flags	= 0;
 	mVUflagInfo		= 0;
+	mVUsFlagHack	= CHECK_VU_FLAGHACK2 | mVUflagHack;
 	bool eBitBranch = 0; // E-bit Set on Branch
 
 	for (int branch = 0;  mVUcount < (vuIndex ? (0x3fff/8) : (0xfff/8)); ) {
@@ -252,7 +253,6 @@ microVUt(void*) __fastcall mVUcompile(u32 startPC, uptr pState) {
 		mVUsetCycles<vuIndex>();
 		if (mVU->p)			  { mVUinfo |= _readP; }
 		if (mVU->q)			  { mVUinfo |= _readQ; }
-		else				  { mVUinfo |= _writeQ; }
 		if		(branch >= 2) { mVUinfo |= _isEOB | ((branch == 3) ? _isBdelay : 0); mVUcount++; branchWarning(); break; }
 		else if (branch == 1) { branch = 2; }
 		if		(mVUbranch)   { mVUsetFlagInfo<vuIndex>(); branchEbit(); branch = 3; mVUbranch = 0; mVUinfo |= _isBranch; }

pcsx2/x86/microVU_Flags.inl

@@ -72,6 +72,8 @@ void sortFlag(int* fFlag, int* bFlag, int cycles) {
 	}
 }
 
+#define sFlagCond ((doStatus && !mVUsFlagHack) || isFSSET || doDivFlag)
+
 // Note: Flag handling is 'very' complex, it requires full knowledge of how microVU recs work, so don't touch!
 microVUt(int) mVUsetFlags(int* xStatus, int* xMac, int* xClip) {
 	microVU* mVU = mVUx;
@@ -132,9 +134,9 @@ microVUt(int) mVUsetFlags(int* xStatus, int* xMac, int* xClip) {
 		mVUinfo |= xM << 10; // _fmInstance
 		mVUinfo |= xC << 14; // _fcInstance
 
-		if (doStatus || isFSSET || doDivFlag) { xStatus[xS] = cycles + 4;  xS = (xS+1) & 3; }
+		if (sFlagCond)	{ xStatus[xS] = cycles + 4;  xS = (xS+1) & 3; }
-		if (doMac)							  { xMac   [xM] = cycles + 4;  xM = (xM+1) & 3; }
+		if (doMac)		{ xMac   [xM] = cycles + 4;  xM = (xM+1) & 3; }
-		if (doClip)							  { xClip  [xC] = cycles + 4;  xC = (xC+1) & 3; }
+		if (doClip)		{ xClip  [xC] = cycles + 4;  xC = (xC+1) & 3; }
 
 		cycles++;
 		incPC2(2);
@@ -183,7 +185,7 @@ microVUt(void) mVUsetupFlags(int* xStatus, int* xMac, int* xClip, int cycles) {
 	}
 }
 
-microVUt(void) mVUpass4(int startPC) {
+microVUt(void) mVUflagPass(int startPC) {
 
 	microVU* mVU  = mVUx;
 	int oldPC	  = iPC;
@@ -195,7 +197,7 @@ microVUt(void) mVUpass4(int startPC) {
 	for (int branch = 0; mVUcount < 4; mVUcount++) {
 		incPC(1);
 		if (  curI & _Ebit_   )	{ branch = 1; }
-		if (  curI & _MDTbit_ )	{ branch = 2; }
+		if (  curI & _MDTbit_ )	{ branch = 4; }
 		if (!(curI & _Ibit_)  )	{ incPC(-1); mVUopL<vuIndex, 3>(); incPC(1); }
 		if		(branch >= 2)	{ break; }
 		else if (branch == 1)	{ branch = 2; }
@@ -215,15 +217,15 @@ microVUt(void) mVUpass4(int startPC) {
 
 microVUt(void) mVUsetFlagInfo() {
 	microVU* mVU = mVUx;
-	branchType1 { incPC(-1); mVUpass4<vuIndex>(branchAddr); incPC(1); }
+	branchType1 { incPC(-1); mVUflagPass<vuIndex>(branchAddr); incPC(1); }
 	branchType2 { mVUflagInfo |= 0xfff; }
 	branchType3 {
 		incPC(-1); 
-		mVUpass4<vuIndex>(branchAddr);
+		mVUflagPass<vuIndex>(branchAddr);
 		int backupFlagInfo = mVUflagInfo;
 		mVUflagInfo = 0;
 		incPC(4); // Branch Not Taken
-		mVUpass4<vuIndex>(xPC);
+		mVUflagPass<vuIndex>(xPC);
 		incPC(-3);		
 		mVUflagInfo |= backupFlagInfo;
 	}

pcsx2/x86/microVU_Misc.h

@@ -189,8 +189,8 @@ declareAllVariables
 #define _isEOB		 (1<<2) // End of Block
 #define _isBdelay	 (1<<3) // Cur Instruction in Branch Delay slot
 #define _isSflag	 (1<<4) // Cur Instruction uses status flag
-#define _writeQ		 (1<<5)
+#define _writeQ		 (1<<6)
-#define _readQ		 (1<<6)
+#define _readQ		 (1<<6) // same as writeQ
 #define _writeP		 (1<<7)
 #define _readP		 (1<<7) // same as writeP
 #define _doFlags	 (3<<8)
@@ -220,10 +220,10 @@ declareAllVariables
 #define isEOB		 (mVUinfo & (1<<2))
 #define isBdelay	 (mVUinfo & (1<<3))
 #define isSflag		 (mVUinfo & (1<<4))
-#define writeQ		((mVUinfo >> 5) & 1)
+#define readQ		((mVUinfo >> 6) & 1) // same as writeQ
-#define readQ		((mVUinfo >> 6) & 1)
-#define writeP		(((mVUinfo >> 7) + 1) & 1)
 #define readP		((mVUinfo >> 7) & 1) // same as writeP
+#define writeQ	   (((mVUinfo >> 6) + 1) & 1)
+#define writeP	   (((mVUinfo >> 7) + 1) & 1)
 #define doFlags		 (mVUinfo & (3<<8))
 #define doMac		 (mVUinfo & (1<<8))
 #define doStatus	 (mVUinfo & (1<<9))
@@ -289,8 +289,9 @@ declareAllVariables
 #endif
 
 // Speed Hacks (Set to 1 to turn On)
-#define CHECK_VU_FLAGHACK	0 // Status Flag Speed Hack
+#define CHECK_VU_FLAGHACK	0 // Status Flag Speed Hack (Very-Compatible)
-#define CHECK_VU_MINMAXHACK	0 // Min/Max Speed Hack
+#define CHECK_VU_FLAGHACK2	0 // Status Flag Speed Hack (More-Hacky)
+#define CHECK_VU_MINMAXHACK	0 // Min/Max Speed Hack		(Semi-Compatible)
 
 // Cache Limit Check
 #define mVUcacheCheck(ptr, start, limit) {																\

pcsx2/x86/microVU_Upper.inl

@@ -35,7 +35,7 @@ microVUt(void) mVUupdateFlags(int reg, int regT1, int regT2, int xyzw, bool modX
 	static const u16 flipMask[16] = {0, 8, 4, 12, 2, 10, 6, 14, 1, 9, 5, 13, 3, 11, 7, 15};
 
 	//SysPrintf("doStatus = %d; doMac = %d\n", doStatus>>9, doMac>>8);
-	if (mVUflagHack) { mVUinfo &= ~_doStatus; }
+	if (mVUsFlagHack) { mVUinfo &= ~_doStatus; }
 	if (!doFlags) return;
 	if (!doMac || (_XYZW_SS && modXYZW)) { regT1 = reg; }
 	else { SSE2_PSHUFD_XMM_to_XMM(regT1, reg, 0x1B); } // Flip wzyx to xyzw