More fixes for persective correct build

- Switch back to fixed point GTE
- Only clamp w to h/2 after divide
- Fix vertex function for multi-pass (was overwriting vertexon first pass)

3 files changed, 56 insertions, 23 deletions

diff --git a/libpcsxcore/gte.c b/libpcsxcore/gte.c
index abe3c205..2938d170 100755
--- a/libpcsxcore/gte.c
+++ b/libpcsxcore/gte.c
@@ -497,11 +497,11 @@ 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);
 		
-		//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, SXY2);
+		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)),
+			max(SZ3, H/2), SXY2);
 
-		PGXP_RTPS(0, SXY2);
+		//PGXP_RTPS(0, SXY2);
 		
 		MAC0 = F((s64) DQB + ((s64) DQA * h_over_sz3));
 		IR0 = Lm_H(m_mac0, 1);
@@ -888,12 +888,19 @@ int docop2(int op) {
 			SXY1 = SXY2;
 			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);
-			
-			//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, SXY2);
 
-			PGXP_RTPS(v, SXY2);
+			//float tempMx = MAC1;
+			//float tempx = IR1;
+			//float temphow = (float)h_over_sz3 / (float)(1 << 16);
+
+			//float tempMz = MAC3;
+			//float tempZ = 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)), 
+				max(SZ3, H/2), SXY2);
+
+			//PGXP_RTPS(v, SXY2);
 		}
 
 		MAC0 = F((s64) DQB + ((s64) DQA * h_over_sz3));
diff --git a/libpcsxcore/pgxp_gte.c b/libpcsxcore/pgxp_gte.c
index 3ee5db22..3844e6be 100644
--- a/libpcsxcore/pgxp_gte.c
+++ b/libpcsxcore/pgxp_gte.c
@@ -306,19 +306,42 @@ void PGXP_RTPS(u32 _n, u32 _v)
 	float MAC2 = TRY + (R21 * VX(_n)) + (R22 * VY(_n)) + (R23 * VZ(_n));

 	float MAC3 = TRZ + (R31 * VX(_n)) + (R32 * VY(_n)) + (R33 * VZ(_n));

 

+	float IR1 = max(min(MAC1, 0x7fff), -0x8000);

+	float IR2 = max(min(MAC2, 0x7fff), -0x8000);

+	float IR3 = max(min(MAC3, 0x7fff), -0x8000);

+

 	float H = psxRegs.CP2C.p[26].sw.l;	// Near plane

 	float F = 0xFFFF;					// Far plane?

-	float SZ3 = max(MAC3, H);			// Clamp SZ3 to near plane because we have no clipping (no proper Z)

+	float SZ3 = max(min(MAC3, 0xffff), 0x0000);			// Clamp SZ3 to near plane because we have no clipping (no proper Z)

 	//	float h_over_sz3 = H / SZ3;

 

 	// Offsets with 16-bit shift

 	float OFX = (float)psxRegs.CP2C.p[24].sd / (float)(1 << 16);

 	float OFY = (float)psxRegs.CP2C.p[25].sd / (float)(1 << 16);

 

+	float h_over_w = min(H / SZ3, (float)0x1ffff / (float)0xffff);

+	h_over_w = (SZ3 == 0) ? ((float)0x1ffff / (float)0xffff) : h_over_w;

+

 	// PSX Screen space X,Y,W components

-	float sx = OFX + (MAC1 * (H / SZ3)) * (Config.Widescreen ? 0.75 : 1);

-	float sy = OFY + (MAC2 * (H / SZ3));

-	float sw = SZ3;

+	float sx = OFX + (IR1 * h_over_w) * (Config.Widescreen ? 0.75 : 1);

+	float sy = OFY + (IR2 * h_over_w);

+	float sw = SZ3;// max(SZ3, 0.1);

+

+	sx = max(min(sx, 1024.f), -1024.f);

+	sy = max(min(sy, 1024.f), -1024.f);

+

+	//float sx2 = SX2;

+	//float sy2 = SY2;

+	//float sz2 = SXY2.z;

+

+	//float ftolerance = 5.f;

+

+	//if ((fabs(sx - sx2) > ftolerance) ||

+	//	(fabs(sy - sy2) > ftolerance) ||

+	//	(fabs(sw - sz2) > ftolerance))

+	//{

+	//	float r = 5;

+	//}

 

 	PGXP_pushSXYZ2f(sx , sy , sw, _v);

 

diff --git a/plugins/peopsxgl/pgxp_gpu.c b/plugins/peopsxgl/pgxp_gpu.c
index 67b381d8..e8191e48 100644
--- a/plugins/peopsxgl/pgxp_gpu.c
+++ b/plugins/peopsxgl/pgxp_gpu.c
@@ -106,22 +106,25 @@ void PGXP_SetMatrix(float left, float right, float bottom, float top, float zNea
 void PGXP_glVertexfv(GLfloat* pV)

 {

 	// If there are PGXP vertices expected

-	if (vertexIdx < numVertices)

-	{

+	//if (vertexIdx < numVertices)

+	//{

+	float temp[4];

+	memcpy(temp, pV, sizeof(float) * 4);

+

 		// pre-multiply each element by w (to negate perspective divide)

 		for (unsigned int i = 0; i < 3; i++)

-			pV[i] *= pV[3];

+			temp[i] *= temp[3];

 

 		// pass complete vertex to OpenGL

-		glVertex4fv(pV);

+		glVertex4fv(temp);

 		vertexIdx++;

 

-		pV[3] = 1.f;

-	}

-	else

-	{

-		glVertex3fv(pV);

-	}

+	//	pV[3] = 1.f;

+	//}

+	//else

+	//{

+	//	glVertex3fv(pV);

+	//}

 }

 

 // Get parallel vertex values