Vif: Fix for Unpacks when WL = 0. (KH2 + Tests) Removes a long standing hack.

Note: CL = 0 behaviour is still not completely accurate, the first vector is incorrect, but will look at that another day, need a game that does it first really so we can see if it helps :)

pcsx2/MTVU.cpp

@@ -40,7 +40,8 @@ enum MTVU_EVENT {
 // Calls the vif unpack functions from the MTVU thread
 static void MTVU_Unpack(void* data, VIFregisters& vifRegs)
 {
-	bool isFill = vifRegs.cycle.cl < vifRegs.cycle.wl;
+	u16 wl = vifRegs.cycle.wl > 0 ? vifRegs.cycle.wl : 256;
+	bool isFill = vifRegs.cycle.cl < wl;
 	if (newVifDynaRec) dVifUnpack<1>((u8*)data, isFill);
 	else              _nVifUnpack(1, (u8*)data, vifRegs.mode, isFill);
 }

pcsx2/Vif_Unpack.cpp

@@ -188,11 +188,6 @@ _vifT void vifUnpackSetup(const u32 *data) {
 
 	vifStruct& vifX = GetVifX;
 
-	if ((vifXRegs.cycle.wl == 0) && (vifXRegs.cycle.wl < vifXRegs.cycle.cl)) {
-        //DevCon.WriteLn("Vif%d CL %d, WL %d Mode %x Mask %x Num %x", idx, vifXRegs.cycle.cl, vifXRegs.cycle.wl, vifXRegs.mode, vifXRegs.mask, (vifXRegs.code >> 16) & 0xff);
-		vifX.cmd = 0;
-        return; // Skipping write and 0 write-cycles, so do nothing!
-	}
 	GetVifX.unpackcalls++;
 	
 	if (GetVifX.unpackcalls > 3)
@@ -216,12 +211,14 @@ _vifT void vifUnpackSetup(const u32 *data) {
 
 	const u8& gsize = nVifT[vifX.cmd & 0x0f];
 
-	if (vifXRegs.cycle.wl <= vifXRegs.cycle.cl) {
+	uint wl = vifXRegs.cycle.wl ? vifXRegs.cycle.wl : 256;
+
+	if (wl <= vifXRegs.cycle.cl) { //Skipping write
 		vifX.tag.size = ((vifNum * gsize) + 3) / 4;
 	}
-	else {
+	else { //Filling write
-		int n = vifXRegs.cycle.cl * (vifNum / vifXRegs.cycle.wl) +
+		int n = vifXRegs.cycle.cl * (vifNum / wl) +
-		        _limit(vifNum % vifXRegs.cycle.wl, vifXRegs.cycle.cl);
+		        _limit(vifNum % wl, vifXRegs.cycle.cl);
 
 		vifX.tag.size = ((n * gsize) + 3) >> 2;
 	}

pcsx2/x86/newVif.h

@@ -61,9 +61,9 @@ struct __aligned16 nVifBlock {
 	u8   num;		// [00] Num  Field
 	u8   upkType;	// [01] Unpack Type [usn*1:mask*1:upk*4]
 	u8   mode;		// [02] Mode Field
-	u8   cl;		// [03] CL   Field
+	u16   cl;		// [03] CL   Field
 	u32  mask;		// [04] Mask Field
-	u8   wl;		// [08] WL   Field
+	u16   wl;		// [08] WL   Field
 	u8   aligned;   // [09] Packet Alignment
 	u8	 padding[2];// [10] through [11]
 	uptr startPtr;	// [12] Start Ptr of RecGen Code

pcsx2/x86/newVif_Dynarec.cpp

@@ -236,11 +236,6 @@ void VifUnpackSSE_Dynarec::CompileRoutine() {
 		else {
 			dstIndirect += (16 * skipSize);
 			vCL = 0;
-			// FIXME
-			// temporary code to avoid an infinite loop until the code is correctly fixed
-			if (blockSize == cycleSize && cycleSize == 0)
-				break;
-			// END FIXME
 		}
 	}
 
@@ -253,18 +248,14 @@ _vifT static __fi u8* dVifsetVUptr(uint cl, uint wl, bool isFill) {
 	vifStruct&    vif        = MTVU_VifX;
 	const VURegs& VU         = vuRegs[idx];
 	const uint    vuMemLimit = idx ? 0x4000 : 0x1000;
-
+	
 	u8*  startmem = VU.Mem + (vif.tag.addr & (vuMemLimit-0x10));
 	u8*  endmem   = VU.Mem + vuMemLimit;
 	uint length   = (v.block.num > 0) ? (v.block.num * 16) : 4096; // 0 = 256
 
+	//wl = wl ? wl : 256; //0 is taken as 256 (KH2)
+	//if (wl == 256) isFill = true;
 	if (!isFill) {
-		// Accounting for skipping mode: Subtract the last skip cycle, since the skipped part of the run
-		// shouldn't count as wrapped data.  Otherwise, a trailing skip can cause the emu to drop back
-		// to the interpreter. -- Refraction (test with MGS3)
-		// FIXME temporary solution to avoid a crash
-		if (wl == 0) wl = 1;
-		// END FIXME
 		uint skipSize  = (cl - wl) * 16;
 		uint blocks    = v.block.num / wl;
 		length += (blocks-1) * skipSize;
@@ -321,7 +312,7 @@ _vifT __fi void dVifUnpack(const u8* data, bool isFill) {
 	v.block.num     = (u8&)vifRegs.num;
 	v.block.mode    = (u8&)vifRegs.mode;
 	v.block.cl      = vifRegs.cycle.cl;
-	v.block.wl      = vifRegs.cycle.wl;
+	v.block.wl      = vifRegs.cycle.wl ? vifRegs.cycle.wl : 256;
 	v.block.aligned = vif.start_aligned;  //MTVU doesn't have a packet size!
 
 	if ((upkType & 0xf) != 9)

pcsx2/x86/newVif_Unpack.cpp

@@ -111,10 +111,11 @@ _vifT int nVifUnpack(const u8* data) {
 	vifStruct&    vif     = GetVifX;
 	VIFregisters& vifRegs = vifXRegs;
 
+	const uint wl     = vifRegs.cycle.wl ? vifRegs.cycle.wl : 256;
 	const uint ret    = aMin(vif.vifpacketsize, vif.tag.size);
-	const bool isFill = (vifRegs.cycle.cl < vifRegs.cycle.wl);
+	const bool isFill = (vifRegs.cycle.cl < wl);
 	s32		   size   = ret << 2;
-
+	
 	if (ret == vif.tag.size) { // Full Transfer
 		if (v.bSize) { // Last transfer was partial
 			memcpy(&v.buffer[v.bSize], data, size);
@@ -252,7 +253,7 @@ __ri void __fastcall _nVifUnpackLoop(const u8* data) {
 	const UNPACKFUNCTYPE ft = VIFfuncTable[idx][doMode ? vifRegs.mode : 0][ ((usn*2*16) + upkNum) ];
 
 	pxAssume (vif.cl == 0);
-	pxAssume (vifRegs.cycle.wl > 0);
+	//pxAssume (vifRegs.cycle.wl > 0);
 
 	do {
 		u8* dest = getVUptr(idx, vif.tag.addr);