19 files changed, 727 insertions, 51 deletions
diff --git a/libpcsxcore/gpu.c b/libpcsxcore/gpu.c
index 324c2a00..61a7b3dc 100755
--- a/libpcsxcore/gpu.c
+++ b/libpcsxcore/gpu.c
@@ -19,6 +19,7 @@
 #include "psxhw.h"
 #include "gpu.h"
 #include "psxdma.h"
+#include "pgxp_gte.h"
 
 #define GPUSTATUS_ODDLINES            0x80000000
 #define GPUSTATUS_DMABITS             0x60000000 // Two bits
@@ -142,6 +143,7 @@ void psxDma2(u32 madr, u32 bcr, u32 chcr) { // GPU
 #endif
 				break;
 			}
+			GPU_pgxpMemory(madr, PGXP_GetMem());
 			GPU_writeDataMem(ptr, size);
 
 #if 0
diff --git a/libpcsxcore/gte.c b/libpcsxcore/gte.c
index 599672c9..04a3968c 100755
--- a/libpcsxcore/gte.c
+++ b/libpcsxcore/gte.c
@@ -7,6 +7,7 @@
 
 #include "gte.h"
 #include "psxmem.h"
+#include "pgxp_gte.h"
 
 #define GTE_SF(op) ((op >> 19) & 1)
 #define GTE_MX(op) ((op >> 17) & 3)
@@ -251,6 +252,7 @@ static void CTC2(u32 value, int reg) {
 void gteMFC2() {
 	if (!_Rt_) return;
 	psxRegs.GPR.r[_Rt_] = MFC2(_Rd_);
+	PGXP_MFC2(_Rt_, _Rd_, psxRegs.CP2D.p[_Rd_].d);
 }
 
 void gteCFC2() {
@@ -259,6 +261,7 @@ void gteCFC2() {
 }
 
 void gteMTC2() {
+	PGXP_MTC2(_Rt_, _Rd_, psxRegs.GPR.p[_Rt_].d);
 	MTC2(psxRegs.GPR.r[_Rt_], _Rd_);
 }
 
@@ -269,11 +272,15 @@ void gteCTC2() {
 #define _oB_ (psxRegs.GPR.r[_Rs_] + _Imm_)
 
 void gteLWC2() {
-	MTC2(psxMemRead32(_oB_), _Rt_);
+	u32 val = psxMemRead32(_oB_);
+	PGXP_LWC2(_oB_, _Rt_, val);
+	MTC2(val, _Rt_);
 }
 
 void gteSWC2() {
-	psxMemWrite32(_oB_, MFC2(_Rt_));
+	u32 val = MFC2(_Rt_);
+	PGXP_SWC2(_oB_, _Rt_, val);
+	psxMemWrite32(_oB_, val);
 }
 
 inline s64 gte_shift(s64 a, int sf) {
@@ -490,10 +497,9 @@ int docop2(int op) {
 		SX2 = Lm_G1(F((s64) OFX + ((s64) IR1 * h_over_sz3) * (Config.Widescreen ? 0.75 : 1)) >> 16);
 		SY2 = Lm_G2(F((s64) OFY + ((s64) IR2 * h_over_sz3)) >> 16);
 		
-		GPU_addVertex(SX2, SY2,
-                  Lm_G1_ia((s64) OFX + (s64)(IR1 * h_over_sz3) * (Config.Widescreen ? 0.75 : 1)),
-                  Lm_G2_ia((s64) OFY + (s64)(IR2 * h_over_sz3)),
-				  ((s64)SZ3));
+		PGXP_pushSXYZ2s(Lm_G1_ia((s64)OFX + (s64)(IR1 * h_over_sz3) * (Config.Widescreen ? 0.75 : 1)), 
+			Lm_G2_ia((s64)OFY + (s64)(IR2 * h_over_sz3)),
+			SZ3);
 		
 		MAC0 = F((s64) DQB + ((s64) DQA * h_over_sz3));
 		IR0 = Lm_H(m_mac0, 1);
@@ -503,8 +509,10 @@ int docop2(int op) {
 #ifdef GTE_LOG
 		GTELOG("%08x NCLIP", op);
 #endif
-
-		MAC0 = F((s64) (SX0 * SY1) + (SX1 * SY2) + (SX2 * SY0) - (SX0 * SY2) - (SX1 * SY0) - (SX2 * SY1));
+		if (PGXP_NLCIP_valid())
+			MAC0 = F(PGXP_NCLIP());
+		else
+			MAC0 = F((s64) (SX0 * SY1) + (SX1 * SY2) + (SX2 * SY0) - (SX0 * SY2) - (SX1 * SY0) - (SX2 * SY1));
 		return 1;
 
 	case 0x0c:
@@ -879,10 +887,9 @@ int docop2(int op) {
 			SX2 = Lm_G1(F((s64) OFX + ((s64) IR1 * h_over_sz3) * (Config.Widescreen ? 0.75 : 1)) >> 16);
 			SY2 = Lm_G2(F((s64) OFY + ((s64) IR2 * h_over_sz3)) >> 16);
 			
-			GPU_addVertex(SX2, SY2,
-                  Lm_G1_ia((s64) OFX + (s64)(IR1 * h_over_sz3) * (Config.Widescreen ? 0.75 : 1)),
-                  Lm_G2_ia((s64) OFY + (s64)(IR2 * h_over_sz3)),
-				  ((s64)SZ3));
+			PGXP_pushSXYZ2s(Lm_G1_ia((s64)OFX + (s64)(IR1 * h_over_sz3) * (Config.Widescreen ? 0.75 : 1)),
+				Lm_G2_ia((s64)OFY + (s64)(IR2 * h_over_sz3)), 
+				SZ3);
 		}
 
 		MAC0 = F((s64) DQB + ((s64) DQA * h_over_sz3));
diff --git a/libpcsxcore/ix86/iR3000A.c b/libpcsxcore/ix86/iR3000A.c
index d7f68890..d18c47ec 100755
--- a/libpcsxcore/ix86/iR3000A.c
+++ b/libpcsxcore/ix86/iR3000A.c
@@ -25,6 +25,7 @@
 
 #include "ix86.h"
 #include <sys/mman.h>
+#include "pgxp_gte.h"
 
 #ifndef MAP_ANONYMOUS
 #define MAP_ANONYMOUS MAP_ANON
@@ -1546,14 +1547,15 @@ static void recLW() {
 //		SysPrintf("unhandled r32 %x\n", addr);
 	}
 
+	PUSH32I(psxRegs.code);	// iCB: Needed to extract reg and opcode
 	iPushOfB();
-	CALLFunc((u32)psxMemRead32);
+	CALLFunc((u32)PGXP_psxMemRead32Trace);
 	if (_Rt_) {
 		iRegs[_Rt_].state = ST_UNK;
 		MOV32RtoM((u32)&psxRegs.GPR.r[_Rt_], EAX);
 	}
 //	ADD32ItoR(ESP, 4);
-	resp+= 4;
+	resp+= 8;
 }
 
 extern u32 LWL_MASK[4];
@@ -1601,12 +1603,13 @@ void recLWL() {
 		if (_Imm_) ADD32ItoR(EAX, _Imm_);
 	}
 	PUSH32R  (EAX);
+	PUSH32I(psxRegs.code);	// iCB: Needed to extract reg and opcode
 	AND32ItoR(EAX, ~3);
 	PUSH32R  (EAX);
-	CALLFunc((u32)psxMemRead32);
+	CALLFunc((u32)PGXP_psxMemRead32Trace);
 
 	if (_Rt_) {
-		ADD32ItoR(ESP, 4);
+		ADD32ItoR(ESP, 8);
 		POP32R   (EDX);
 		AND32ItoR(EDX, 0x3); // shift = addr & 3;
 
@@ -1629,7 +1632,7 @@ void recLWL() {
 		MOV32RtoM((u32)&psxRegs.GPR.r[_Rt_], EAX);
 	} else {
 //		ADD32ItoR(ESP, 8);
-		resp+= 8;
+		resp+= 12;
 	}
 }
 
@@ -1756,12 +1759,13 @@ void recLWR() {
 		if (_Imm_) ADD32ItoR(EAX, _Imm_);
 	}
 	PUSH32R  (EAX);
+	PUSH32I(psxRegs.code);	// iCB: Needed to extract reg and opcode
 	AND32ItoR(EAX, ~3);
 	PUSH32R  (EAX);
-	CALLFunc((u32)psxMemRead32);
+	CALLFunc((u32)PGXP_psxMemRead32Trace);
 
 	if (_Rt_) {
-		ADD32ItoR(ESP, 4);
+		ADD32ItoR(ESP, 8);
 		POP32R   (EDX);
 		AND32ItoR(EDX, 0x3); // shift = addr & 3;
 
@@ -1785,7 +1789,7 @@ void recLWR() {
 		MOV32RtoM((u32)&psxRegs.GPR.r[_Rt_], EAX);
 	} else {
 //		ADD32ItoR(ESP, 8);
-		resp+= 8;
+		resp+= 12;
 	}
 }
 
@@ -1978,15 +1982,16 @@ static void recSW() {
 //		SysPrintf("unhandled w32 %x\n", addr);
 	}
 
+	PUSH32I(psxRegs.code);	// iCB: Needed to extract reg and opcode
 	if (IsConst(_Rt_)) {
 		PUSH32I  (iRegs[_Rt_].k);
 	} else {
 		PUSH32M  ((u32)&psxRegs.GPR.r[_Rt_]);
 	}
 	iPushOfB();
-	CALLFunc((u32)psxMemWrite32);
+	CALLFunc((u32)PGXP_psxMemWrite32Trace);
 //	ADD32ItoR(ESP, 8);
-	resp+= 8;
+	resp+= 12;
 }
 //#endif
 
@@ -2104,12 +2109,13 @@ void recSWL() {
 		if (_Imm_) ADD32ItoR(EAX, _Imm_);
 	}
 	PUSH32R  (EAX);
+	PUSH32I(psxRegs.code);	// iCB: Needed to extract reg and opcode
 	AND32ItoR(EAX, ~3);
 	PUSH32R  (EAX);
 
-	CALLFunc((u32)psxMemRead32);
+	CALLFunc((u32)PGXP_psxMemRead32Trace);
 
-	ADD32ItoR(ESP, 4);
+	ADD32ItoR(ESP, 8);
 	POP32R   (EDX);
 	AND32ItoR(EDX, 0x3); // shift = addr & 3;
 
@@ -2126,6 +2132,8 @@ void recSWL() {
 	}
 	SHR32CLtoR(EDX); // _rRt_ >> SWL_SHIFT[shift]
 
+	PUSH32I(psxRegs.code);	// iCB: Needed to extract reg and opcode
+
 	OR32RtoR (EAX, EDX);
 	PUSH32R  (EAX);
 
@@ -2137,9 +2145,9 @@ void recSWL() {
 	AND32ItoR(EAX, ~3);
 	PUSH32R  (EAX);
 
-	CALLFunc((u32)psxMemWrite32);
+	CALLFunc((u32)PGXP_psxMemWrite32Trace);
 //	ADD32ItoR(ESP, 8);
-	resp+= 8;
+	resp+= 12;
 }
 
 extern u32 SWR_MASK[4];
@@ -2186,12 +2194,13 @@ void recSWR() {
 		if (_Imm_) ADD32ItoR(EAX, _Imm_);
 	}
 	PUSH32R  (EAX);
+	PUSH32I(psxRegs.code);	// iCB: Needed to extract reg and opcode
 	AND32ItoR(EAX, ~3);
 	PUSH32R  (EAX);
 
-	CALLFunc((u32)psxMemRead32);
+	CALLFunc((u32)PGXP_psxMemRead32Trace);
 
-	ADD32ItoR(ESP, 4);
+	ADD32ItoR(ESP, 8);
 	POP32R   (EDX);
 	AND32ItoR(EDX, 0x3); // shift = addr & 3;
 
@@ -2208,6 +2217,8 @@ void recSWR() {
 	}
 	SHL32CLtoR(EDX); // _rRt_ << SWR_SHIFT[shift]
 
+	PUSH32I(psxRegs.code);	// iCB: Needed to extract reg and opcode
+
 	OR32RtoR (EAX, EDX);
 	PUSH32R  (EAX);
 
@@ -2219,9 +2230,9 @@ void recSWR() {
 	AND32ItoR(EAX, ~3);
 	PUSH32R  (EAX);
 
-	CALLFunc((u32)psxMemWrite32);
+	CALLFunc((u32)PGXP_psxMemWrite32Trace);
 //	ADD32ItoR(ESP, 8);
-	resp += 8;
+	resp += 12;
 }
 
 /*REC_FUNC(SLL);
diff --git a/libpcsxcore/pgxp_gte.c b/libpcsxcore/pgxp_gte.c
new file mode 100644
index 00000000..4b22c778
--- /dev/null
+++ b/libpcsxcore/pgxp_gte.c
@@ -0,0 +1,437 @@
+/***************************************************************************
+*   Copyright (C) 2016 by iCatButler                                      *
+*                                                                         *
+*   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; either version 2 of the License, or     *
+*   (at your option) any later version.                                   *
+*                                                                         *
+*   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 for more details.                          *
+*                                                                         *
+*   You should have received a copy of the GNU General Public License     *
+*   along with this program; if not, write to the                         *
+*   Free Software Foundation, Inc.,                                       *
+*   51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.           *
+***************************************************************************/
+
+/**************************************************************************
+*	pgxp_gte.c
+*	PGXP - Parallel/Precision Geometry Xform Pipeline
+*
+*	Created on: 12 Mar 2016
+*      Author: iCatButler
+***************************************************************************/
+
+#include "pgxp_gte.h"
+#include "psxmem.h"
+
+typedef struct 
+{
+	float			x;
+	float			y;
+	float			z;
+	unsigned int	valid;
+	unsigned int	count;
+} precise_value;
+
+typedef union
+{
+	struct
+	{
+		s16 x;
+		s16 y;
+	};
+	u32 word;
+} low_value;
+
+#define tolerance 1.0f
+
+precise_value GTE_reg[32];
+precise_value CPU_reg[32];
+
+precise_value Mem[2048 * 1024 / 4];		// mirror 2MB in 32-bit words
+precise_value Scratch[2048 * 1024 / 4]; // mirror 2MB in 32-bit words
+
+void PGXP_Init()
+{
+	memset(Mem, 0, sizeof(Mem));
+}
+
+char* PGXP_GetMem()
+{
+	return (char*)(Mem);
+}
+
+#define VRAM 0
+#define SCRATCH 1
+
+precise_value* ReadMem(u32 addr)
+{
+	u32 memType;
+	uint32_t paddr = addr;
+	int* ip = NULL;
+
+	switch (paddr >> 20)
+	{
+	case 0x800:
+	case 0x801:
+	case 0xa00:
+	case 0xa01:
+	case 0x000:
+	case 0x001:
+		memType = VRAM;
+		break;
+	case 0x1f8:
+		memType = SCRATCH;
+		break;
+	default:
+		return NULL;
+	}
+	
+#ifdef GTE_LOG
+	//GTE_LOG("PGXP_Read %x [%x] |", addr, paddr);
+#endif
+	if (memType == VRAM)
+	{
+		paddr = (paddr & 0x1FFFFF) >> 2;
+		return &Mem[paddr];
+	}
+	else if (memType == SCRATCH)
+	{
+		paddr = (paddr & 0x1FFFFF) >> 2;
+		return &Scratch[paddr];
+	}
+
+	return NULL;
+}
+
+void WriteMem(precise_value value, u32 addr)
+{
+	u32 memType;
+	uint32_t paddr = addr;
+	int* ip = NULL;
+
+	switch (paddr >> 20)
+	{
+	case 0x800:
+	case 0x801:
+	case 0xa00:
+	case 0xa01:
+	case 0x000:
+	case 0x001:
+		memType = VRAM;
+		break;
+	case 0x1f8:
+		memType = SCRATCH;
+		break;
+	default:
+		if (value.valid)
+			*ip = 5;
+		return;
+	}
+
+#ifdef GTE_LOG
+	GTE_LOG("PGXP_Write %x [%x] |", addr, paddr);
+#endif
+
+	// Store to RAM
+	if (memType == VRAM)
+	{
+		paddr = (paddr & 0x1FFFFF) >> 2;
+		Mem[paddr] = value;
+	}
+	else if (memType == SCRATCH)
+	{
+		paddr = (paddr & 0x1FFFFF) >> 2;// (paddr & 0x3FFF) >> 2;
+		Scratch[paddr] = value;
+	}
+}
+
+#define SX0 (GTE_reg[ 12 ].x)
+#define SY0 (GTE_reg[ 12 ].y)
+#define SX1 (GTE_reg[ 13 ].x)
+#define SY1 (GTE_reg[ 13 ].y)
+#define SX2 (GTE_reg[ 14 ].x)
+#define SY2 (GTE_reg[ 14 ].y)
+
+#define SXY0 (GTE_reg[ 12 ])
+#define SXY1 (GTE_reg[ 13 ])
+#define SXY2 (GTE_reg[ 14 ])
+#define SXYP (GTE_reg[ 15 ])
+
+unsigned int PGXP_validate(float high, s16 low)
+{
+	if (fabs(high - (float)(low)) < tolerance)
+		return 1;
+	return 0;
+}
+
+// Check that value is still within tolerance of low precision value and invalidate if not
+precise_value PGXP_validateXY(precise_value *high, u32 low)
+{
+	low_value temp;
+	precise_value ret;
+
+	ret.valid = 0;
+	temp.word = low;
+
+	if (!high)
+		return ret;
+
+	high->valid = (high->valid && PGXP_validate(high->x, temp.x) && PGXP_validate(high->y, temp.y));
+
+	// Cheat
+	//if (!high->valid)
+	//{
+	//	high->x = temp.x;
+	//	high->y = temp.y;
+	//	high->valid = 1;
+	//}
+
+	return *high;
+}
+
+u32 PGXP_compareXY(precise_value high, u32 low)
+{
+	low_value temp;
+	temp.word = low;
+
+	if (PGXP_validate(high.x, temp.x) && PGXP_validate(high.y, temp.y))
+		return 1;
+	return 0;
+}
+
+precise_value PGXP_copyXY(u32 low)
+{
+	low_value temp;
+	precise_value ret;
+
+	ret.valid = 0;
+	temp.word = low;
+
+	ret.x = temp.x;
+	ret.y = temp.y;
+	ret.count = 0;
+	ret.valid = 1;
+
+	return ret;
+}
+
+void PGXP_pushSXYZ2f(float _x, float _y, float _z)
+{
+	static unsigned int uCount = 0;
+	// push values down FIFO
+	SXY0 = SXY1;
+	SXY1 = SXY2;
+	
+	SXY2.x		= _x;
+	SXY2.y		= _y;
+	SXY2.z		= _z;
+	SXY2.valid	= 1;
+	SXY2.count	= uCount++;
+
+#ifdef GTE_LOG
+	GTE_LOG("PGPR_PUSH (%f, %f) %u %u|", SXY2.x, SXY2.y, SXY2.valid, SXY2.count);
+#endif
+}
+
+void PGXP_pushSXYZ2s(s64 _x, s64 _y, s64 _z)
+{
+	float fx = (float)(_x) / (float)(1 << 16);
+	float fy = (float)(_y) / (float)(1 << 16);
+	float fz = (float)(_z);
+
+	PGXP_pushSXYZ2f(fx, fy, fz);
+}
+
+int PGXP_NLCIP_valid()
+{
+	if (SXY0.valid && SXY1.valid && SXY2.valid)
+		return 1;
+	return 0;
+}
+
+float PGXP_NCLIP()
+{
+	float nclip = ((SX0 * SY1) + (SX1 * SY2) + (SX2 * SY0) - (SX0 * SY2) - (SX1 * SY0) - (SX2 * SY1));
+
+	// ensure fractional values are not incorrectly rounded to 0
+	if (fabs(nclip) < 1.0f)
+		nclip += (nclip < 0.f ? -2 : 2);
+
+	return nclip;
+}
+
+static precise_value PGXP_MFC2_int(u32 reg)
+{
+	switch (reg) 
+	{
+	case 15:
+		GTE_reg[reg] = SXYP = SXY2;
+		break;
+	}
+
+	return GTE_reg[reg];
+}
+
+
+static void PGXP_MTC2_int(precise_value value, u32 reg)
+{
+	switch(reg)
+	{
+		case 15:
+			// push FIFO
+			SXY0 = SXY1;
+			SXY1 = SXY2;
+			SXY2 = value;
+			SXYP = SXY2;
+			break;
+
+		case 31:
+			return;
+	}
+
+	GTE_reg[reg] = value;
+}
+
+// copy GTE data reg to GPR reg (MFC2)
+void PGXP_MFC2(u32 gpr, u32 gtr, u32 value)
+{
+	if (!gpr) return;
+#ifdef GTE_LOG
+	GTE_LOG("PGXP_MFC2 [%x] [%x] %x (%u %u)|", gpr, gtr, value, GTE_reg[gtr].valid, GTE_reg[gtr].count);
+#endif
+
+	CPU_reg[gpr] = PGXP_validateXY(&GTE_reg[gtr], value);
+}
+
+// copy GPR reg to GTE data reg (MTC2)
+void PGXP_MTC2(u32 gpr, u32 gtr, u32 value)
+{
+#ifdef GTE_LOG
+	GTE_LOG("PGXP_MTC2 [%x] [%x] %x (%u %u)|", gpr, gtr, value, CPU_reg[gtr].valid, CPU_reg[gtr].count);
+#endif
+	PGXP_MTC2_int(PGXP_validateXY(&CPU_reg[gpr], value), gtr);
+}
+
+// copy memory to GTE reg
+void PGXP_LWC2(u32 addr, u32 gtr, u32 value)
+{
+#ifdef GTE_LOG
+	precise_value* pp = ReadMem(addr);
+	precise_value p;
+	low_value temp;
+	temp.word = value;
+
+	p.x = p.y = p.valid = 0;
+
+	if (pp)
+		p = *pp;
+
+	GTE_LOG("PGXP_LWC2 %x [%x] %x (%d, %d) (%f, %f) %u %u|", addr, gtr, value, temp.x, temp.y, p.x, p.y, p.valid, p.count);
+#endif
+	PGXP_MTC2_int(PGXP_validateXY(ReadMem(addr), value), gtr);
+}
+
+//copy GTE reg to memory
+void PGXP_SWC2(u32 addr, u32 gtr, u32 value)
+{
+#ifdef GTE_LOG
+	low_value temp;
+	temp.word = value;
+
+	if (PGXP_compareXY(GTE_reg[gtr], value))
+		GTE_LOG("PGPR_SWC2 %x [%x] %x (%d, %d) (%f, %f) %u %u|", addr, gtr, value, temp.x, temp.y, GTE_reg[gtr].x, GTE_reg[gtr].y, GTE_reg[gtr].valid, GTE_reg[gtr].count);
+#endif
+	WriteMem(PGXP_validateXY(&GTE_reg[gtr], value), addr);
+}
+
+// ltore 32bit word
+void PGPR_L32(u32 addr, u32 code, u32 value)
+{
+	u32 reg = ((code >> 16) & 0x1F); // The rt part of the instruction register 
+	u32 op = ((code >> 26));
+	precise_value p;
+
+	low_value temp;
+	temp.word = value;
+
+	p.x = p.y = p.valid = p.count = 0;
+
+	switch (op)
+	{
+	case 34:	//LWL
+		CPU_reg[reg] = PGXP_validateXY(ReadMem(addr), value);
+		break;
+	case 35:	//LW
+		CPU_reg[reg] = PGXP_validateXY(ReadMem(addr), value);
+		break;
+	case 37:	//LWR
+		CPU_reg[reg] = PGXP_validateXY(ReadMem(addr), value);
+		break;
+	case 50:	//LWC2 (GTE vertex reads)
+		GTE_reg[reg] = PGXP_validateXY(ReadMem(addr), value);
+		break;
+	default:
+		// invalidate register
+		//	WriteMem(p, addr);
+		break;
+	}
+
+#ifdef GTE_LOG
+	GTE_LOG("PGPR_L32 %u: %x %x[%x %x] (%d, %d) (%f, %f) %x %u|", op, addr, value, code, reg, temp.x, temp.y, CPU_reg[reg].x, CPU_reg[reg].y, CPU_reg[reg].valid, CPU_reg[reg].count);
+#endif
+}
+
+// store 32bit word
+void PGPR_S32(u32 addr, u32 code, u32 value)
+{
+	u32 reg = ((code >> 16) & 0x1F); // The rt part of the instruction register 
+	u32 op = ((code >> 26));
+	precise_value p;
+
+	low_value temp;
+	temp.word = value;
+
+	p.x = p.y = p.valid = p.count = 0;
+
+#ifdef GTE_LOG
+	GTE_LOG("PGPR_S32 %u: %x %x[%x %x] (%d, %d) (%f, %f) %x %u|", op, addr, value, code, reg, temp.x, temp.y, CPU_reg[reg].x, CPU_reg[reg].y, CPU_reg[reg].valid, CPU_reg[reg].count);
+#endif
+
+	switch (op)
+	{
+	case 42:	//SWL
+		WriteMem(PGXP_validateXY(&CPU_reg[reg], value), addr);
+		break;
+	case 43:	//SW
+		WriteMem(PGXP_validateXY(&CPU_reg[reg], value), addr);
+		break;
+	case 46:	//SWR
+		WriteMem(PGXP_validateXY(&CPU_reg[reg], value), addr);
+		break;
+	case 58:	//SWC2 (GTE vertex writes)
+		WriteMem(PGXP_validateXY(&GTE_reg[reg], value), addr);
+		break;
+	default:
+		// invalidate memory
+	//	WriteMem(p, addr);
+		break;
+	}
+}
+
+u32 PGXP_psxMemRead32Trace(u32 mem, u32 code)
+{
+	u32 value = psxMemRead32(mem);
+	PGPR_L32(mem, code, value);
+	return value;
+}
+
+void PGXP_psxMemWrite32Trace(u32 mem, u32 value, u32 code)
+{
+	PGPR_S32(mem, code, value);
+	psxMemWrite32(mem, value);
+}
+\ No newline at end of file
diff --git a/libpcsxcore/pgxp_gte.h b/libpcsxcore/pgxp_gte.h
new file mode 100644
index 00000000..c97c5b29
--- /dev/null
+++ b/libpcsxcore/pgxp_gte.h
@@ -0,0 +1,58 @@
+/***************************************************************************
+*   Copyright (C) 2016 by iCatButler                                      *
+*                                                                         *
+*   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; either version 2 of the License, or     *
+*   (at your option) any later version.                                   *
+*                                                                         *
+*   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 for more details.                          *
+*                                                                         *
+*   You should have received a copy of the GNU General Public License     *
+*   along with this program; if not, write to the                         *
+*   Free Software Foundation, Inc.,                                       *
+*   51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.           *
+***************************************************************************/
+
+/**************************************************************************
+*	pgxp_gte.h
+*	PGXP - Parallel/Precision Geometry Xform Pipeline
+*
+*	Created on: 12 Mar 2016
+*      Author: iCatButler
+***************************************************************************/
+
+#ifndef _PGXP_GTE_H_
+#define _PGXP_GTE_H_
+
+#include "psxcommon.h"
+
+void	PGXP_Init();	// initialise memory
+char*	PGXP_GetMem();	// return pointer to precision memory
+
+// -- GTE functions
+// Transforms
+void	PGXP_pushSXYZ2f(float _x, float _y, float _z);
+void	PGXP_pushSXYZ2s(s64 _x, s64 _y, s64 _z);
+int		PGXP_NLCIP_valid();
+float	PGXP_NCLIP();
+
+// Data transfer tracking
+void	PGXP_MFC2(u32 gpr, u32 gtr, u32 value);		// copy GTE reg to GPR reg (MFC2)
+void	PGXP_MTC2(u32 gpr, u32 gtr, u32 value);		// copy GPR reg to GTR reg (MTC2)
+void	PGXP_LWC2(u32 addr, u32 gtr, u32 value);	// copy memory to GTE reg
+void	PGXP_SWC2(u32 addr, u32 gtr, u32 value);	// copy GTE reg to memory
+
+// -- CPU functions
+// Data transfer tracking
+void	PGPR_L32(u32 addr, u32 code, u32 value);	// load 32bit word
+void	PGPR_S32(u32 addr, u32 code, u32 value);	// store 32bit word
+
+// Memory Read/Write hooks
+u32		PGXP_psxMemRead32Trace(u32 mem, u32 code);
+void	PGXP_psxMemWrite32Trace(u32 mem, u32 value, u32 code);
+
+#endif /* _PGXP_GTE_H_ */
diff --git a/libpcsxcore/plugins.c b/libpcsxcore/plugins.c
index 05770154..54c7e468 100755
--- a/libpcsxcore/plugins.c
+++ b/libpcsxcore/plugins.c
@@ -54,6 +54,7 @@ GPUvBlank             GPU_vBlank;
 GPUvisualVibration    GPU_visualVibration;
 GPUcursor             GPU_cursor;
 GPUaddVertex          GPU_addVertex;
+GPUpgxpMemory		  GPU_pgxpMemory;
 
 CDRinit               CDR_init;
 CDRshutdown           CDR_shutdown;
@@ -214,6 +215,7 @@ void CALLBACK GPU__vBlank(int val) {}
 void CALLBACK GPU__visualVibration(unsigned long iSmall, unsigned long iBig) {}
 void CALLBACK GPU__cursor(int player, int x, int y) {}
 void CALLBACK GPU__addVertex(short sx,short sy,s64 fx,s64 fy,s64 fz) {}
+void CALLBACK GPU__pgxpMemory(unsigned char* pMem, unsigned int count, unsigned int addr, unsigned char* pVRAM) {}
 
 #define LoadGpuSym1(dest, name) \
 	LoadSym(GPU_##dest, GPU##dest, name, TRUE);
@@ -259,6 +261,7 @@ static int LoadGPUplugin(const char *GPUdll) {
     LoadGpuSym0(visualVibration, "GPUvisualVibration");
     LoadGpuSym0(cursor, "GPUcursor");
 	LoadGpuSym0(addVertex, "GPUaddVertex");
+	LoadGpuSym0(pgxpMemory, "GPUpgxpMemory");
 	LoadGpuSym0(configure, "GPUconfigure");
 	LoadGpuSym0(test, "GPUtest");
 	LoadGpuSym0(about, "GPUabout");
diff --git a/libpcsxcore/plugins.h b/libpcsxcore/plugins.h
index f8cfdf44..8f2cc189 100755
--- a/libpcsxcore/plugins.h
+++ b/libpcsxcore/plugins.h
@@ -96,6 +96,7 @@ typedef void (CALLBACK* GPUvBlank)(int);
 typedef void (CALLBACK* GPUvisualVibration)(uint32_t, uint32_t);
 typedef void (CALLBACK* GPUcursor)(int, int, int);
 typedef void (CALLBACK* GPUaddVertex)(short,short,s64,s64,s64);
+typedef void (CALLBACK* GPUpgxpMemory)(unsigned int, unsigned char*);
 
 // GPU function pointers
 extern GPUupdateLace    GPU_updateLace;
@@ -125,6 +126,7 @@ extern GPUvBlank        GPU_vBlank;
 extern GPUvisualVibration GPU_visualVibration;
 extern GPUcursor        GPU_cursor;
 extern GPUaddVertex     GPU_addVertex;
+extern GPUpgxpMemory	GPU_pgxpMemory;
 
 // CD-ROM Functions
 typedef long (CALLBACK* CDRinit)(void);
diff --git a/libpcsxcore/r3000a.c b/libpcsxcore/r3000a.c
index fe908ecb..cb2f4025 100755
--- a/libpcsxcore/r3000a.c
+++ b/libpcsxcore/r3000a.c
@@ -44,6 +44,7 @@ int psxInit() {
 	Log = 0;
 
 	if (psxMemInit() == -1) return -1;
+	PGXP_Init();
 
 	return psxCpu->Init();
 }
diff --git a/plugins/peopsxgl/draw.c b/plugins/peopsxgl/draw.c
index cb838d33..c3fed1cf 100755
--- a/plugins/peopsxgl/draw.c
+++ b/plugins/peopsxgl/draw.c
@@ -28,6 +28,7 @@
 #include "menu.h"
 
 #include "gte_accuracy.h"
+#include "pgxp_gpu.h"
 
 #if defined(_MACGL)
 // if you use it, you must include it
@@ -943,7 +944,7 @@ void offsetline(void)
 #define VERTEX_OFFX 0.2f
 #define VERTEX_OFFY 0.2f
 
-BOOL offsetline(void)           
+BOOL offsetline(unsigned int* addr)
 {
  short x0,x1,y0,y1,dx,dy;float px,py;
 
@@ -1004,6 +1005,8 @@ BOOL offsetline(void)
      else           py= 0.0f;
     }
   } 
+
+ PGXP_GetVertices(addr, vertex);
  
  vertex[0].x=(short)((float)x0-px);
  vertex[3].x=(short)((float)x0+py);
@@ -1040,7 +1043,7 @@ BOOL offsetline(void)
 
 ///////////////////////////////////////////////////////// 
 
-BOOL offset2(void)
+BOOL offset2(unsigned int* addr)
 {
  if(bDisplayNotSet)
   SetOGLDisplaySettings(1);
@@ -1066,6 +1069,8 @@ BOOL offset2(void)
 	vertex[1].y=ly1;
  }
 
+ PGXP_GetVertices(addr, vertex);
+
  vertex[0].x+=PSXDisplay.CumulOffset.x;
  vertex[1].x+=PSXDisplay.CumulOffset.x;
  vertex[0].y+=PSXDisplay.CumulOffset.y;
@@ -1076,7 +1081,7 @@ BOOL offset2(void)
 
 ///////////////////////////////////////////////////////// 
 
-BOOL offset3(void)
+BOOL offset3(unsigned int* addr)
 {
  if(bDisplayNotSet)
   SetOGLDisplaySettings(1);
@@ -1109,6 +1114,8 @@ BOOL offset3(void)
 	vertex[2].y=ly2;
  }
 
+ PGXP_GetVertices(addr, vertex);
+
  vertex[0].x+=PSXDisplay.CumulOffset.x;
  vertex[1].x+=PSXDisplay.CumulOffset.x;
  vertex[2].x+=PSXDisplay.CumulOffset.x;
@@ -1121,7 +1128,7 @@ BOOL offset3(void)
 
 ///////////////////////////////////////////////////////// 
 
-BOOL offset4(void)
+BOOL offset4(unsigned int* addr)
 {
  if(bDisplayNotSet)
   SetOGLDisplaySettings(1);
@@ -1161,6 +1168,8 @@ BOOL offset4(void)
 	vertex[3].y=ly3;
  }
  
+ PGXP_GetVertices(addr, vertex);
+
  vertex[0].x+=PSXDisplay.CumulOffset.x;
  vertex[1].x+=PSXDisplay.CumulOffset.x;
  vertex[2].x+=PSXDisplay.CumulOffset.x;
diff --git a/plugins/peopsxgl/draw.h b/plugins/peopsxgl/draw.h
index 10501d37..2f524904 100755
--- a/plugins/peopsxgl/draw.h
+++ b/plugins/peopsxgl/draw.h
@@ -35,10 +35,10 @@ BOOL bSetupPixelFormat(HDC hDC);
 
 int  GLinitialize();
 void GLcleanup();
-BOOL offset2(void);
-BOOL offset3(void);
-BOOL offset4(void);
-BOOL offsetline(void);
+BOOL offset2(unsigned int* addr);
+BOOL offset3(unsigned int* addr);
+BOOL offset4(unsigned int* addr);
+BOOL offsetline(unsigned int* addr);
 void offsetST(void);
 void offsetBlk(void);
 void offsetScreenUpload(int Position);
diff --git a/plugins/peopsxgl/gpu.c b/plugins/peopsxgl/gpu.c
index d5af3330..65e074d5 100755
--- a/plugins/peopsxgl/gpu.c
+++ b/plugins/peopsxgl/gpu.c
@@ -48,6 +48,7 @@ static int iOldMode=0;
 #include "fps.h"
 #include "key.h"
 #include "gte_accuracy.h"
+#include "PGXP_gpu.h"
 #ifdef _WINDOWS
 #include "resource.h"
 #include "ssave.h"
@@ -3145,7 +3146,11 @@ long CALLBACK GPUdmaChain(uint32_t *baseAddrL, uint32_t addr)
 
    dmaMem=addr+4;
 
-   if(count>0) GPUwriteDataMem(&baseAddrL[dmaMem>>2],count);
+   if (count > 0)
+   {
+	   PGXP_SetAddress(dmaMem >> 2);
+	   GPUwriteDataMem(&baseAddrL[dmaMem >> 2], count);
+   }
 
    addr = baseAddrL[addr>>2]&0xffffff;
   }
diff --git a/plugins/peopsxgl/prim.c b/plugins/peopsxgl/prim.c
index 6a06a9be..0c78145d 100755
--- a/plugins/peopsxgl/prim.c
+++ b/plugins/peopsxgl/prim.c
@@ -3105,7 +3105,7 @@ void primPolyF4(unsigned char *baseAddr)
  lx3 = sgpuData[8];
  ly3 = sgpuData[9];
 
- if(offset4()) return;
+ if(offset4(baseAddr)) return;
 
  bDrawTextured = FALSE;
  bDrawSmoothShaded = FALSE;
@@ -3214,7 +3214,7 @@ void primPolyG4(unsigned char * baseAddr)
  lx3 = sgpuData[14];
  ly3 = sgpuData[15];
 
- if(offset4()) return;
+ if(offset4(baseAddr)) return;
 
  bDrawTextured = FALSE;
  bDrawSmoothShaded = TRUE;
@@ -3430,7 +3430,7 @@ void primPolyFT3(unsigned char * baseAddr)
  lx2 = sgpuData[10];
  ly2 = sgpuData[11];
 
- if(offset3()) return;
+ if(offset3(baseAddr)) return;
     
  // do texture UV coordinates stuff
  gl_ux[0]=gl_ux[3]=baseAddr[8];//gpuData[2]&0xff;
@@ -3868,7 +3868,7 @@ void primPolyFT4(unsigned char * baseAddr)
  lx3 = sgpuData[14];
  ly3 = sgpuData[15];
 
- if(offset4()) return;
+ if(offset4(baseAddr)) return;
 
  gl_vy[0]=baseAddr[9];//((gpuData[2]>>8)&0xff);
  gl_vy[1]=baseAddr[17];//((gpuData[4]>>8)&0xff);
@@ -3953,7 +3953,7 @@ void primPolyGT3(unsigned char *baseAddr)
  lx2 = sgpuData[14];
  ly2 = sgpuData[15];
 
- if(offset3()) return;
+ if(offset3(baseAddr)) return;
 
  // do texture stuff
  gl_ux[0]=gl_ux[3]=baseAddr[8];//gpuData[2]&0xff;
@@ -4062,7 +4062,7 @@ void primPolyG3(unsigned char *baseAddr)
  lx2 = sgpuData[10];
  ly2 = sgpuData[11];
 
- if(offset3()) return;
+ if(offset3(baseAddr)) return;
 
  bDrawTextured = FALSE;
  bDrawSmoothShaded = TRUE;
@@ -4109,7 +4109,7 @@ void primPolyGT4(unsigned char *baseAddr)
  lx3 = sgpuData[20];
  ly3 = sgpuData[21];
 
- if(offset4()) return;
+ if(offset4(baseAddr)) return;
 
  // do texture stuff
  gl_ux[0]=baseAddr[8];//gpuData[2]&0xff;
@@ -4227,7 +4227,7 @@ void primPolyF3(unsigned char *baseAddr)
  lx2 = sgpuData[6];
  ly2 = sgpuData[7];
 
- if(offset3()) return;
+ if(offset3(baseAddr)) return;
 
  bDrawTextured     = FALSE;
  bDrawSmoothShaded = FALSE;
@@ -4318,7 +4318,7 @@ void primLineGEx(unsigned char *baseAddr)
    ly1 = (short)((gpuData[i]>>16) & 0xffff);
    lx1 = (short)(gpuData[i] & 0xffff);
 
-   if(offsetline()) bDraw=FALSE; else bDraw=TRUE;
+   if(offsetline(baseAddr)) bDraw=FALSE; else bDraw=TRUE;
   
    if (bDraw && ((lx0 != lx1) || (ly0 != ly1)))
     {
@@ -4367,7 +4367,7 @@ void primLineG2(unsigned char *baseAddr)
 
  if((lx0 == lx1) && (ly0 == ly1)) return;
     
- if(offsetline()) return;
+ if(offsetline(baseAddr)) return;
     
  SetRenderState(gpuData[0]);
  SetRenderMode(gpuData[0], FALSE);
@@ -4445,7 +4445,7 @@ void primLineFEx(unsigned char *baseAddr)
    ly1 = (short)((gpuData[i]>>16) & 0xffff);
    lx1 = (short)(gpuData[i] & 0xffff);
 
-   if(!offsetline())
+   if(!offsetline(baseAddr))
     {
      if(iOffscreenDrawing)
       {
@@ -4481,7 +4481,7 @@ void primLineF2(unsigned char *baseAddr)
  lx1 = sgpuData[4];
  ly1 = sgpuData[5];
 
- if(offsetline()) return;
+ if(offsetline(baseAddr)) return;
 
  bDrawTextured = FALSE;
  bDrawSmoothShaded = FALSE;
diff --git a/win32/pcsxr.vcxproj b/win32/pcsxr.vcxproj
index 6a273ff5..81893bf4 100755
--- a/win32/pcsxr.vcxproj
+++ b/win32/pcsxr.vcxproj
@@ -199,6 +199,7 @@
       <AdditionalIncludeDirectories Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <PreprocessorDefinitions Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">%(PreprocessorDefinitions)</PreprocessorDefinitions>
     </ClCompile>
+    <ClCompile Include="..\libpcsxcore\pgxp_gte.c" />
     <ClCompile Include="..\libpcsxcore\plugins.c">
       <AdditionalIncludeDirectories Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <PreprocessorDefinitions Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">%(PreprocessorDefinitions)</PreprocessorDefinitions>
@@ -418,6 +419,7 @@
     <ClInclude Include="..\libpcsxcore\gte_divider.h" />
     <ClInclude Include="..\libpcsxcore\mdec.h" />
     <ClInclude Include="..\libpcsxcore\misc.h" />
+    <ClInclude Include="..\libpcsxcore\pgxp_gte.h" />
     <ClInclude Include="..\libpcsxcore\plugins.h" />
     <ClInclude Include="..\libpcsxcore\ppf.h" />
     <ClInclude Include="..\libpcsxcore\psemu_plugin_defs.h" />
diff --git a/win32/pcsxr.vcxproj.filters b/win32/pcsxr.vcxproj.filters
index 09c5b72b..5c714f63 100755
--- a/win32/pcsxr.vcxproj.filters
+++ b/win32/pcsxr.vcxproj.filters
@@ -192,6 +192,9 @@
     <ClCompile Include="zlib\zutil.c">
       <Filter>zlib</Filter>
     </ClCompile>
+    <ClCompile Include="..\libpcsxcore\pgxp_gte.c">
+      <Filter>libpcsxcore</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="..\libpcsxcore\cdriso.h">
@@ -355,6 +358,9 @@
     <ClInclude Include="zlib\zutil.h">
       <Filter>zlib</Filter>
     </ClInclude>
+    <ClInclude Include="..\libpcsxcore\pgxp_gte.h">
+      <Filter>libpcsxcore</Filter>
+    </ClInclude>
   </ItemGroup>
   <ItemGroup>
     <None Include="gui\about.bmp">
diff --git a/win32/plugins/peopsxgl/gpuPeopsOpenGL.def b/win32/plugins/peopsxgl/gpuPeopsOpenGL.def
index 17d2e66c..83d853a7 100755
--- a/win32/plugins/peopsxgl/gpuPeopsOpenGL.def
+++ b/win32/plugins/peopsxgl/gpuPeopsOpenGL.def
@@ -34,3 +34,4 @@ EXPORTS
 	GPUsetframelimit    @27
 	GPUvisualVibration  @28
 	GPUaddVertex        @29
+	GPUpgxpMemory		@30
diff --git a/win32/plugins/peopsxgl/gpuPeopsOpenGL.vcxproj b/win32/plugins/peopsxgl/gpuPeopsOpenGL.vcxproj
index b72348bd..f65ce2cb 100755
--- a/win32/plugins/peopsxgl/gpuPeopsOpenGL.vcxproj
+++ b/win32/plugins/peopsxgl/gpuPeopsOpenGL.vcxproj
@@ -152,6 +152,7 @@
     </PostBuildEvent>
   </ItemDefinitionGroup>
   <ItemGroup>
+    <ClCompile Include="pgxp_gpu.c" />
     <ClCompile Include="winsrc\cfg.c">
       <AdditionalIncludeDirectories Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <PreprocessorDefinitions Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">%(PreprocessorDefinitions)</PreprocessorDefinitions>
@@ -221,6 +222,7 @@
     </ClCompile>
   </ItemGroup>
   <ItemGroup>
+    <ClInclude Include="pgxp_gpu.h" />
     <ClInclude Include="winsrc\ssave.h" />
     <ClInclude Include="..\..\..\plugins\peopsxgl\cfg.h" />
     <ClInclude Include="..\..\..\plugins\peopsxgl\draw.h" />
diff --git a/win32/plugins/peopsxgl/gpuPeopsOpenGL.vcxproj.filters b/win32/plugins/peopsxgl/gpuPeopsOpenGL.vcxproj.filters
index 69b2b9b9..d34a9a34 100755
--- a/win32/plugins/peopsxgl/gpuPeopsOpenGL.vcxproj.filters
+++ b/win32/plugins/peopsxgl/gpuPeopsOpenGL.vcxproj.filters
@@ -45,6 +45,9 @@
     <ClCompile Include="..\..\..\plugins\peopsxgl\texture.c">
       <Filter>peopsxgl</Filter>
     </ClCompile>
+    <ClCompile Include="pgxp_gpu.c">
+      <Filter>peopsxgl</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="winsrc\ssave.h">
@@ -90,6 +93,9 @@
       <Filter>peopsxgl</Filter>
     </ClInclude>
     <ClInclude Include="resource.h" />
+    <ClInclude Include="pgxp_gpu.h">
+      <Filter>peopsxgl</Filter>
+    </ClInclude>
   </ItemGroup>
   <ItemGroup>
     <None Include="res\gpu.bmp" />
diff --git a/win32/plugins/peopsxgl/pgxp_gpu.c b/win32/plugins/peopsxgl/pgxp_gpu.c
new file mode 100644
index 00000000..5b64382a
--- /dev/null
+++ b/win32/plugins/peopsxgl/pgxp_gpu.c
@@ -0,0 +1,90 @@
+/***************************************************************************
+*   Copyright (C) 2016 by iCatButler                                      *
+*                                                                         *
+*   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; either version 2 of the License, or     *
+*   (at your option) any later version.                                   *
+*                                                                         *
+*   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 for more details.                          *
+*                                                                         *
+*   You should have received a copy of the GNU General Public License     *
+*   along with this program; if not, write to the                         *
+*   Free Software Foundation, Inc.,                                       *
+*   51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.           *
+***************************************************************************/
+
+/**************************************************************************
+*	pgxp_gpu.c
+*	PGXP - Parallel/Precision Geometry Xform Pipeline
+*
+*	Created on: 25 Mar 2016
+*      Author: iCatButler
+***************************************************************************/
+
+#include "pgxp_gpu.h"
+#include "stdafx.h"
+#include "externals.h"
+
+#include <math.h>
+
+typedef struct
+{
+	float	x;
+	float	y;
+	float	z;
+	unsigned int	valid;
+	unsigned int	count;
+} PGXP_vertex;
+
+const unsigned int primStrideTable[] = { 1, 2, 1, 2, 2, 3, 2, 3, 0 };
+const unsigned int primCountTable[] = { 3, 3, 4, 4, 3, 3, 4, 4, 0 };
+
+PGXP_vertex*	PGXP_Mem = NULL;	// pointer to parallel memory
+unsigned int	currentAddr = 0;	// address of current DMA
+
+// Set current DMA address and pointer to parallel memory
+void CALLBACK GPUpgxpMemory(unsigned int addr, unsigned char* pVRAM)
+{
+	if (pVRAM)
+		PGXP_Mem = (PGXP_vertex*)(pVRAM);
+	currentAddr = addr;
+}
+
+// Set current DMA address
+void PGXP_SetAddress(unsigned int addr)
+{
+	currentAddr = addr;
+}
+
+// Get parallel vertex values
+int PGXP_GetVertices(unsigned int* addr, void* pOutput)
+{
+	unsigned int	primCmd		= ((*addr >> 24) & 0xff);		// primitive command
+	unsigned int	primIdx		= (primCmd - 0x20) >> 2;		// index to primitive lookup
+	OGLVertex*		pVertex		= (OGLVertex*)pOutput;			// pointer to output vertices
+	unsigned int	stride		= primStrideTable[primIdx];		// stride between vertices
+	unsigned int	count		= primCountTable[primIdx];		// number of vertices
+	PGXP_vertex*	primStart	= NULL;							// pointer to first vertex
+
+	if (PGXP_Mem == NULL)
+		return 0;
+
+	// Offset to start of primitive
+	primStart = &PGXP_Mem[currentAddr + 1];
+
+	for (unsigned i = 0; i < count; ++i)
+	{
+		if (primStart[stride * i].valid)
+		{
+			pVertex[i].x = primStart[stride * i].x;
+			pVertex[i].y = primStart[stride * i].y;
+		}
+	}
+
+	return 1;
+}
+
diff --git a/win32/plugins/peopsxgl/pgxp_gpu.h b/win32/plugins/peopsxgl/pgxp_gpu.h
new file mode 100644
index 00000000..1777dc19
--- /dev/null
+++ b/win32/plugins/peopsxgl/pgxp_gpu.h
@@ -0,0 +1,34 @@
+/***************************************************************************
+*   Copyright (C) 2016 by iCatButler                                      *
+*                                                                         *
+*   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; either version 2 of the License, or     *
+*   (at your option) any later version.                                   *
+*                                                                         *
+*   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 for more details.                          *
+*                                                                         *
+*   You should have received a copy of the GNU General Public License     *
+*   along with this program; if not, write to the                         *
+*   Free Software Foundation, Inc.,                                       *
+*   51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.           *
+***************************************************************************/
+
+/**************************************************************************
+*	pgxp_gpu.h
+*	PGXP - Parallel/Precision Geometry Xform Pipeline
+*
+*	Created on: 25 Mar 2016
+*      Author: iCatButler
+***************************************************************************/
+
+#ifndef _PGXP_GPU_H_
+#define _PGXP_GPU_H_
+
+void	PGXP_SetAddress(unsigned int addr);
+int		PGXP_GetVertices(unsigned int* addr, void* pOutput);
+
+#endif // _PGXP_GPU_H_