- Nneeve fixed the Tri-Ace gamefix so Gradius5 doesn't crash with it enabled anymore.

- Moved one global variable for the VU interpreters, which surprisingly speeds up Star Ocean 3 for me.
- Set Flush to Zero for FPU and VU back to on. Let's see how long it lasts this time :p
- Removed the FFX hack from pcsx2! It's still toggled in the GS plugins, the correct behaviour is having it always on.

git-svn-id: http://pcsx2.googlecode.com/svn/trunk@848 96395faa-99c1-11dd-bbfe-3dabce05a288

common/include/Pcsx2Config.h

@@ -81,8 +81,8 @@ extern SessionOverrideFlags g_Session;
 #define DEFAULT_eeOptions	0x01
 #define DEFAULT_vuOptions	0x01
 //------------ DEFAULT sseMXCSR VALUES!!! ---------------
-#define DEFAULT_sseMXCSR	0x7fc0 //FPU rounding, DaZ, "chop" - Note: Dont enable FtZ by default, it breaks games! E.g. Enthusia (Refraction)
+#define DEFAULT_sseMXCSR	0xffc0 //FPU rounding > DaZ, FtZ, "chop"
-#define DEFAULT_sseVUMXCSR	0x7fc0 //VU rounding, DaZ, "chop"
+#define DEFAULT_sseVUMXCSR	0xffc0 //VU  rounding > DaZ, FtZ, "chop"
 
 #define CHECK_FRAMELIMIT (Config.Options&PCSX2_FRAMELIMIT_MASK)
 

pcsx2/Elfheader.cpp

@@ -571,7 +571,6 @@ int loadElfFile(const char *filename)
 }
 
 #include "VU.h"
-extern int g_FFXHack;
 extern int path3hack;
 int g_VUGameFixes = 0;
 
@@ -580,7 +579,6 @@ void LoadGameSpecificSettings()
 {
 	// default
 	g_VUGameFixes = 0;
-	g_FFXHack = 0;
 
 	switch(ElfCRC) {
 		case 0xb99379b7: // erementar gerad (discolored chars)
@@ -589,33 +587,6 @@ void LoadGameSpecificSettings()
 		case 0xa08c4057:  //Sprint Cars (SLUS)
 		case 0x8b0725d5:  //Flinstones Bedrock Racing (SLES)
 			path3hack = 1; // We can move this to patch files right now
-			break;
-
-		case 0xb4414ea1: // ffx(rus)
-		case 0xee97db5b: // ffx(rus)
-		case 0xaec495cc: // ffx(rus)
-		case 0x6a4efe60: // ffx(j)
-		case 0xA39517AB: // ffx(e)
-		case 0xBB3D833A: // ffx(u)
-		case 0x941bb7d9: // ffx(g)
-		case 0xD9FC6310: // ffx int(j)
-		case 0xa39517ae: // ffx(f)
-		case 0xa39517a9: // ffx(i)
-		case 0x658597e2: // ffx int
-		case 0x941BB7DE: // ffx(s)
-		case 0x3866CA7E: // ffx(asia)
-		case 0x48FE0C71: // ffx2 (u)
-		case 0x9aac530d: // ffx2 (g)
-		case 0x9AAC5309: // ffx2 (e)
-		case 0x8A6D7F14: // ffx2 (j)
-		case 0x9AAC530B: // ffx2 (i)
-		case 0x9AAC530A: // ffx2 (f)
-		case 0x9aac530c: // ffx2 (f)
-		case 0xe1fd9a2d: // ffx2 last mission (?)
-		case 0x93f9b89a: // ffx2 demo (g)
-		case 0x304C115C: // harvest moon - awl
-		case 0xF0A6D880: // harvest moon - sth
-			g_FFXHack = 1;
 			break;		
 	}
 }

pcsx2/GS.cpp

@@ -41,8 +41,6 @@ using namespace R5900;
 
 static bool m_gsOpened = false;
 
-int g_FFXHack=0;
-
 #ifdef PCSX2_DEVBUILD
 
 // GS Playback

pcsx2/MTGS.cpp

@@ -327,7 +327,7 @@ __forceinline u32 mtgsThreadObject::_gifTransferDummy( GIF_PATH pathidx, const u
 			}
 			else if(path.tag.nloop == 0)
 			{
-				if(pathidx == 0 && g_FFXHack)
+				if(pathidx == 0)
 					continue;
 
 				eop = true;

pcsx2/Patch.cpp

@@ -653,7 +653,8 @@ int AddPatch(int Mode, int Place, int Address, int Size, u64 data)
 	
 void patchFunc_ffxhack( char * cmd, char * param )
 {
-	 g_FFXHack = 1;
+	 //Keeping this as a dummy a while :p
+	 //g_FFXHack = 1;
 }
 
 void patchFunc_xkickdelay( char * cmd, char * param )

pcsx2/Patch.h

@@ -110,7 +110,6 @@ int AddPatch(int Mode, int Place, int Address, int Size, u64 data);
 extern void SetFastMemory(int); // iR5900LoadStore.c
 
 extern int path3hack;
-extern int g_FFXHack;
 //extern int g_VUGameFixes;
 extern int g_ZeroGSOptions;
 extern u32 g_sseMXCSR;

pcsx2/VUops.cpp

@@ -316,7 +316,6 @@ void _vuAddLowerStalls(VURegs * VU, _VURegsNum *VUregsn) {
 /*   VU Upper instructions    */
 /******************************/
 #ifndef INT_VUDOUBLEHACK
-static u32 d;
 float vuDouble(u32 f)
 {
 	switch(f & 0x7f800000){
@@ -324,10 +323,13 @@ float vuDouble(u32 f)
 			f &= 0x80000000;
 			return *(float*)&f;
 			break;
-		case 0x7f800000:
+		case 0x7f800000: 
+		{
+			u32 d;
 			d = (f & 0x80000000)|0x7f7fffff;
 			return *(float*)&d;
 			break;
+		}
 		default:
 			return *(float*)&f;
 			break;
@@ -2718,7 +2720,8 @@ void _vuRegsFSSET(VURegs * VU, _VURegsNum *VUregsn) {
     VUregsn->VFread0 = 0;
     VUregsn->VFread1 = 0;
     VUregsn->VIwrite = 1 << REG_STATUS_FLAG;
-    VUregsn->VIread  = 0;//1 << REG_STATUS_FLAG; this kills speed
+	//VUregsn->VIread  = 0; // 1 << REG_STATUS_FLAG; this kills speed. Todo: Orly? (rama)
+    VUregsn->VIread  = 1 << REG_STATUS_FLAG;
 }
 
 void _vuRegsFMAND(VURegs * VU, _VURegsNum *VUregsn) {

pcsx2/x86/iVUmicroUpper.cpp

@@ -333,12 +333,23 @@ void recUpdateFlags(VURegs * VU, int reg, int info)
 static PCSX2_ALIGNED16(u32 VU_addsuband[2][4]);
 static PCSX2_ALIGNED16(u32 VU_addsub_reg[2][4]);
 
+static u32 tempECX;
+
 void VU_ADD_SUB(u32 regd, u32 regt, int is_sub, int info)  
 {
 	u8 *localptr[4][8];
-	int temp1 = _allocX86reg(ECX, X86TYPE_TEMP, 0, 0); //receives regd
+
+	MOV32RtoM((uptr)&tempECX, ECX);
+
+	int temp1 = ECX; //receives regd
 	int temp2 = ALLOCTEMPX86(0);
 
+	if (temp2 == ECX) 
+	{
+		temp2 = ALLOCTEMPX86(0);
+		_freeX86reg(ECX);
+	}
+
 	SSE_MOVAPS_XMM_to_M128((uptr)&VU_addsub_reg[0][0], regd);
 	SSE_MOVAPS_XMM_to_M128((uptr)&VU_addsub_reg[1][0], regt);
 
@@ -413,98 +424,26 @@ void VU_ADD_SUB(u32 regd, u32 regt, int is_sub, int info)
 
 	SSE_MOVAPS_M128_to_XMM(regt, (uptr)&VU_addsub_reg[1][0]);
 
-	_freeX86reg(temp1);
 	_freeX86reg(temp2);
-}
 
-void VU_ADD_SUB_SSE4(u32 regd, u32 regt, int is_sub, int info)  
+	MOV32MtoR(ECX, (uptr)&tempECX);
-{
-	u8 *localptr[4][8];
-	int temp1 = _allocX86reg(ECX, X86TYPE_TEMP, 0, 0); //receives regd
-	int temp2 = ALLOCTEMPX86(0);
-
-	SSE_MOVAPS_XMM_to_M128((uptr)&VU_addsub_reg[0][0], regd);
-	SSE_MOVAPS_XMM_to_M128((uptr)&VU_addsub_reg[1][0], regt);
-
-	SSE2_PSLLD_I8_to_XMM(regd, 1);
-	SSE2_PSLLD_I8_to_XMM(regt, 1);
-
-	SSE2_PSRLD_I8_to_XMM(regd, 24);
-	SSE2_PSRLD_I8_to_XMM(regt, 24);
-
-	SSE2_PSUBD_XMM_to_XMM(regd, regt);
-
-#define PERFORM_SSE4(i) \
-	\
-	SSE_PEXTRW_XMM_to_R32(temp1, regd, i*2); \
-	MOVSX32R16toR(temp1, temp1); \
-	CMP32ItoR(temp1, 25);\
-	localptr[i][0] = JGE8(0);\
-	CMP32ItoR(temp1, 0);\
-	localptr[i][1] = JG8(0);\
-	localptr[i][2] = JE8(0);\
-	CMP32ItoR(temp1, -25);\
-	localptr[i][3] = JLE8(0);\
-	\
-	NEG32R(temp1); \
-	DEC32R(temp1);\
-	MOV32ItoR(temp2, 0xffffffff); \
-	SHL32CLtoR(temp2); \
-	SSE4_PINSRD_R32_to_XMM(regd, temp2, i); \
-	localptr[i][4] = JMP8(0);\
-	\
-	x86SetJ8(localptr[i][0]);\
-	MOV32ItoR(temp2, 0xffffffff); \
-	SSE4_PINSRD_R32_to_XMM(regd, temp2, i); \
-	SHL32ItoR(temp2, 31); \
-	SSE4_PINSRD_R32_to_XMM(regt, temp2, i); \
-	localptr[i][5] = JMP8(0);\
-	\
-	x86SetJ8(localptr[i][1]);\
-	DEC32R(temp1);\
-	MOV32ItoR(temp2, 0xffffffff);\
-	SSE4_PINSRD_R32_to_XMM(regd, temp2, i); \
-	SHL32CLtoR(temp2); \
-	SSE4_PINSRD_R32_to_XMM(regt, temp2, i); \
-	localptr[i][6] = JMP8(0);\
-	\
-	x86SetJ8(localptr[i][3]);\
-	MOV32ItoR(temp2, 0x80000000); \
-	SSE4_PINSRD_R32_to_XMM(regd, temp2, i); \
-	localptr[i][7] = JMP8(0);\
-	\
-	x86SetJ8(localptr[i][2]);\
-	\
-	x86SetJ8(localptr[i][4]);\
-	x86SetJ8(localptr[i][5]);\
-	x86SetJ8(localptr[i][6]);\
-	x86SetJ8(localptr[i][7]);
-
-	SSE2_PCMPEQB_XMM_to_XMM(regt, regt);
-	PERFORM_SSE4(0);
-	PERFORM_SSE4(1);
-	PERFORM_SSE4(2);
-	PERFORM_SSE4(3);
-#undef PERFORM_SSE4
-
-	SSE_ANDPS_M128_to_XMM(regd, (uptr)&VU_addsub_reg[0][0]); //regd contains mask
-	SSE_ANDPS_M128_to_XMM(regt, (uptr)&VU_addsub_reg[1][0]); //regt contains mask
-
-	if (is_sub)	SSE_SUBPS_XMM_to_XMM(regd, regt);
-	else		SSE_ADDPS_XMM_to_XMM(regd, regt);
-
-	SSE_MOVAPS_M128_to_XMM(regt, (uptr)&VU_addsub_reg[1][0]);
-
-	_freeX86reg(temp1);
-	_freeX86reg(temp2);
 }
 
 void VU_ADD_SUB_SS(u32 regd, u32 regt, int is_sub, int is_mem, int info)  
 {
 	u8 *localptr[8];
 	u32 addrt = regt; //for case is_mem
-	int temp1 = _allocX86reg(ECX, X86TYPE_TEMP, 0, 0); //receives regd //_allocX86reg(ECX, X86TYPE_TEMP, 0, ((info&PROCESS_VU_SUPER)?0:MODE_NOFRAME)|mode);
+
+	MOV32RtoM((uptr)&tempECX, ECX);
+
+	int temp1 = ECX; //receives regd 
 	int temp2 = ALLOCTEMPX86(0);
+
+	if (temp2 == ECX) 
+	{
+		temp2 = ALLOCTEMPX86(0);
+		_freeX86reg(ECX);
+	}
 	
 	SSE_MOVAPS_XMM_to_M128((uptr)&VU_addsub_reg[0][0], regd);
 	if (!is_mem) SSE_MOVAPS_XMM_to_M128((uptr)&VU_addsub_reg[1][0], regt);
@@ -617,167 +556,44 @@ void VU_ADD_SUB_SS(u32 regd, u32 regt, int is_sub, int is_mem, int info)
 		SSE_MOVAPS_M128_to_XMM(regt, (uptr)&VU_addsub_reg[1][0]);
 	}
 
-	_freeX86reg(temp1);
 	_freeX86reg(temp2);
-}
 
-void VU_ADD_SUB_SS_SSE4(u32 regd, u32 regt, int is_sub, int is_mem, int info)  
+	MOV32MtoR(ECX, (uptr)&tempECX);
-{
-	u8 *localptr[8];
-	u32 addrt = regt; //for case is_mem
-	int temp1 = _allocX86reg(ECX, X86TYPE_TEMP, 0, 0); //receives regd //_allocX86reg(ECX, X86TYPE_TEMP, 0, ((info&PROCESS_VU_SUPER)?0:MODE_NOFRAME)|mode);
-	int temp2 = ALLOCTEMPX86(0);
-	
-	SSE_MOVAPS_XMM_to_M128((uptr)&VU_addsub_reg[0][0], regd);
-	if (!is_mem) SSE_MOVAPS_XMM_to_M128((uptr)&VU_addsub_reg[1][0], regt);
-
-	SSE2_MOVD_XMM_to_R(temp1, regd);
-	SHR32ItoR(temp1, 23); 
-
-	if (is_mem) {
-		MOV32MtoR(temp2, addrt);
-		MOV32RtoM((uptr)&VU_addsub_reg[1][0], temp2);
-		SHR32ItoR(temp2, 23);
-	}
-	else {
-		SSE2_MOVD_XMM_to_R(temp2, regt);
-		SHR32ItoR(temp2, 23);
-	}
-
-	AND32ItoR(temp1, 0xff);
-	AND32ItoR(temp2, 0xff); 
-
-	SUB32RtoR(temp1, temp2); //temp1 = exponent difference
-
-	CMP32ItoR(temp1, 25);
-	localptr[0] = JGE8(0);
-	CMP32ItoR(temp1, 0);
-	localptr[1] = JG8(0);
-	localptr[2] = JE8(0);
-	CMP32ItoR(temp1, -25);
-	localptr[3] = JLE8(0);
-
-	NEG32R(temp1); 
-	DEC32R(temp1);
-	MOV32ItoR(temp2, 0xffffffff);
-	SHL32CLtoR(temp2);
-	SSE2_PCMPEQB_XMM_to_XMM(regd, regd);
-	SSE4_PINSRD_R32_to_XMM(regd, temp2, 0);
-	if (!is_mem)
-		SSE2_PCMPEQB_XMM_to_XMM(regt, regt);
-	localptr[4] = JMP8(0);
-
-	x86SetJ8(localptr[0]);
-	MOV32ItoR(temp2, 0x80000000);
-	if (is_mem)
-		AND32RtoM((uptr)&VU_addsub_reg[1][0], temp2);
-	else {
-		SSE2_PCMPEQB_XMM_to_XMM(regt, regt);
-		SSE4_PINSRD_R32_to_XMM(regt, temp2, 0);
-	}
-	SSE2_PCMPEQB_XMM_to_XMM(regd, regd);
-	localptr[5] = JMP8(0);
-
-	x86SetJ8(localptr[1]);
-	DEC32R(temp1);
-	MOV32ItoR(temp2, 0xffffffff);
-	SHL32CLtoR(temp2); 
-	if (is_mem) 
-		AND32RtoM((uptr)&VU_addsub_reg[1][0], temp2);
-	else {
-		SSE2_PCMPEQB_XMM_to_XMM(regt, regt);
-		SSE4_PINSRD_R32_to_XMM(regt, temp2, 0);
-	}
-	SSE2_PCMPEQB_XMM_to_XMM(regd, regd);
-	localptr[6] = JMP8(0);
-
-	x86SetJ8(localptr[3]);
-	MOV32ItoR(temp2, 0x80000000);
-	SSE2_PCMPEQB_XMM_to_XMM(regd, regd);
-	SSE4_PINSRD_R32_to_XMM(regd, temp2, 0);
-	if (!is_mem)
-		SSE2_PCMPEQB_XMM_to_XMM(regt, regt);
-	localptr[7] = JMP8(0);
-
-	x86SetJ8(localptr[2]);
-	x86SetJ8(localptr[4]);
-	x86SetJ8(localptr[5]);
-	x86SetJ8(localptr[6]);
-	x86SetJ8(localptr[7]);
-
-	if (is_mem) 
-	{
-		SSE_ANDPS_M128_to_XMM(regd, (uptr)&VU_addsub_reg[0][0]); //regd contains mask
-
-		if (is_sub)	SSE_SUBSS_M32_to_XMM(regd, (uptr)&VU_addsub_reg[1][0]);
-		else		SSE_ADDSS_M32_to_XMM(regd, (uptr)&VU_addsub_reg[1][0]);
-	}
-	else
-	{
-		SSE_ANDPS_M128_to_XMM(regd, (uptr)&VU_addsub_reg[0][0]); //regd contains mask
-		SSE_ANDPS_M128_to_XMM(regt, (uptr)&VU_addsub_reg[1][0]); //regt contains mask
-
-		if (is_sub)	SSE_SUBSS_XMM_to_XMM(regd, regt);
-		else		SSE_ADDSS_XMM_to_XMM(regd, regt);
-
-		SSE_MOVAPS_M128_to_XMM(regt, (uptr)&VU_addsub_reg[1][0]);
-	}
-
-	_freeX86reg(temp1);
-	_freeX86reg(temp2);
 }
 
 void SSE_ADDPS_XMM_to_XMM_custom(int info, int regd, int regt) {
 	if (CHECK_VUADDSUBHACK) {
-		if ( cpucaps.hasStreamingSIMD4Extensions )
+		VU_ADD_SUB(regd, regt, 0, info);
-			VU_ADD_SUB_SSE4(regd, regt, 0, info);
-		else
-			VU_ADD_SUB(regd, regt, 0, info);
 	}
 	else SSE_ADDPS_XMM_to_XMM(regd, regt);
 }
 void SSE_SUBPS_XMM_to_XMM_custom(int info, int regd, int regt) {
 	if (CHECK_VUADDSUBHACK) {
-		if ( cpucaps.hasStreamingSIMD4Extensions )
+		VU_ADD_SUB(regd, regt, 1, info);
-			VU_ADD_SUB_SSE4(regd, regt, 1, info);
-		else
-			VU_ADD_SUB(regd, regt, 1, info);
 	}
 	else SSE_SUBPS_XMM_to_XMM(regd, regt);
 }
 void SSE_ADDSS_XMM_to_XMM_custom(int info, int regd, int regt) {
 	if (CHECK_VUADDSUBHACK) {
-		if ( cpucaps.hasStreamingSIMD4Extensions )
+		VU_ADD_SUB_SS(regd, regt, 0, 0, info);
-			VU_ADD_SUB_SS_SSE4(regd, regt, 0, 0, info);
-		else
-			VU_ADD_SUB_SS(regd, regt, 0, 0, info);
 	}
 	else SSE_ADDSS_XMM_to_XMM(regd, regt);
 }
 void SSE_SUBSS_XMM_to_XMM_custom(int info, int regd, int regt) {
 	if (CHECK_VUADDSUBHACK) {
-		if ( cpucaps.hasStreamingSIMD4Extensions )
+		VU_ADD_SUB_SS(regd, regt, 1, 0, info);
-			VU_ADD_SUB_SS_SSE4(regd, regt, 1, 0, info);
-		else
-			VU_ADD_SUB_SS(regd, regt, 1, 0, info);
 	}
 	else SSE_SUBSS_XMM_to_XMM(regd, regt);
 }
 void SSE_ADDSS_M32_to_XMM_custom(int info, int regd, int regt) {
 	if (CHECK_VUADDSUBHACK) {
-		if ( cpucaps.hasStreamingSIMD4Extensions )
+		VU_ADD_SUB_SS(regd, regt, 0, 1, info);
-			VU_ADD_SUB_SS_SSE4(regd, regt, 0, 1, info);
-		else
-			VU_ADD_SUB_SS(regd, regt, 0, 1, info);
 	}
 	else SSE_ADDSS_M32_to_XMM(regd, regt);
 }
 void SSE_SUBSS_M32_to_XMM_custom(int info, int regd, int regt) {
 	if (CHECK_VUADDSUBHACK) {
-		if ( cpucaps.hasStreamingSIMD4Extensions )
+		VU_ADD_SUB_SS(regd, regt, 1, 1, info);
-			VU_ADD_SUB_SS_SSE4(regd, regt, 1, 1, info);
-		else
-			VU_ADD_SUB_SS(regd, regt, 1, 1, info);
 	}
 	else SSE_SUBSS_M32_to_XMM(regd, regt);
 }