From e493d806716f9f440dc55235da990826b398320a Mon Sep 17 00:00:00 2001
From: UnknownShadow200 <unknownshadow200@gmail.com>
Date: Sun, 23 Jun 2024 13:00:46 +1000
Subject: [PATCH] PS2: Draw coloured vertices in same way as textured ones

---
 misc/ps2/VertexTransform.S | 108 +++++++++++++++++++++++++++----------
 src/Graphics_PS2.c         |  69 +++---------------------
 src/Platform_Dreamcast.c   |   1 +
 3 files changed, 87 insertions(+), 91 deletions(-)

diff --git a/misc/ps2/VertexTransform.S b/misc/ps2/VertexTransform.S
index b24dd70f2..8ae49a2fa 100644
--- a/misc/ps2/VertexTransform.S
+++ b/misc/ps2/VertexTransform.S
@@ -15,6 +15,8 @@
 .type   LoadClipScaleFactors,%function
 .global TransformTexturedQuad
 .type   TransformTexturedQuad,%function
+.global TransformColouredQuad
+.type   TransformColouredQuad,%function
 
 # Loads matrix into VU0 registers
 #	$a0 = addresss of mvp
@@ -36,6 +38,7 @@ LoadClipScaleFactors:
 
 
 .macro TransformVertex1
+	# TRANSFORM VERTEX 1
 	lqc2	$vf10, 0x00($a2)   # IN = tmp
 	vmulaw	$ACC,  $vf4,  $vf0 # ACC[xyzw] = mvp.row3[xyzw] * 1.0f; (vf0.w is 1)
 	vmaddax	$ACC,  $vf1, $vf10 # ACC[xyzw] = ACC[xyzw] + mvp.row0[xyzw] * IN.x
@@ -47,6 +50,11 @@ LoadClipScaleFactors:
 .endm
 
 .macro TransformVertex2
+	# STORE CLIP FLAGS 1 RESULT
+	cfc2	$t0, $18	  # t0 = VP0_REGS[CLIP_FLAGS]
+	sw		$t0,0x00($a3) # clip_flags[0] = t0
+
+	# TRANSFORM VERTEX 2
 	lqc2	$vf12, 0x00($a2)   # IN = tmp
 	vmulaw	$ACC,  $vf4,  $vf0 # ACC[xyzw] = mvp.row3[xyzw] * 1.0f; (vf0.w is 1)
 	vmaddax	$ACC,  $vf1, $vf12 # ACC[xyzw] = ACC[xyzw] + mvp.row0[xyzw] * IN.x
@@ -58,6 +66,11 @@ LoadClipScaleFactors:
 .endm
 
 .macro TransformVertex3
+	# STORE CLIP FLAGS 2 RESULT
+	cfc2	$t0, $18	  # t0 = VP0_REGS[CLIP_FLAGS]
+	sw		$t0,0x04($a3) # clip_flags[1] = t0
+
+	# TRANSFORM VERTEX 3
 	lqc2	$vf14, 0x00($a2)   # IN = tmp
 	vmulaw	$ACC,  $vf4,  $vf0 # ACC[xyzw] = mvp.row3[xyzw] * 1.0f; (vf0.w is 1)
 	vmaddax	$ACC,  $vf1, $vf14 # ACC[xyzw] = ACC[xyzw] + mvp.row0[xyzw] * IN.x
@@ -69,6 +82,11 @@ LoadClipScaleFactors:
 .endm
 
 .macro TransformVertex4
+	# STORE CLIP FLAGS 3 RESULT
+	cfc2	$t0, $18	  # t0 = VP0_REGS[CLIP_FLAGS]
+	sw		$t0,0x08($a3) # clip_flags[2] = t0
+
+	# TRANSFORM VERTEX 4
 	lqc2	$vf16, 0x00($a2)   # IN = tmp
 	vmulaw	$ACC,  $vf4,  $vf0 # ACC[xyzw] = mvp.row3[xyzw] * 1.0f; (vf0.w is 1)
 	vmaddax	$ACC,  $vf1, $vf16 # ACC[xyzw] = ACC[xyzw] + mvp.row0[xyzw] * IN.x
@@ -78,6 +96,25 @@ LoadClipScaleFactors:
 	vclipw.xyz $vf16, $vf16    # CLIP_FLAGS.append(CLIP(TMP.xyz, TMP.w))
 .endm
 
+.macro TransformFinish
+	# Vertex output
+	# dst[0] = V0 (done by TransformVertex1)
+	# dst[1] = V1 (done by TransformVertex2)
+	# dst[2] = V2 (done by TransformVertex3)
+	# dst[3] = V2
+	# dst[4] = V3
+	# dst[5] = V0
+	sqc2	$vf15, 0x30($a1)   # dst[3] = TRANSFORMED(V2)
+	sqc2	$vf17, 0x40($a1)   # dst[4] = TRANSFORMED(V3)
+	sqc2	$vf11, 0x50($a1)   # dst[5] = TRANSFORMED(V0)
+	vnop					   # adjust for delay
+
+	# STORE CLIP FLAGS 4 RESULT
+	cfc2	$t0, $18	  	   # t0 = VP0_REGS[CLIP_FLAGS]
+	sw		$t0,0x0C($a3)      # clip_flags[3] = t0
+.endm
+
+
 # Transforms 4 vertices with size of 24 bytes
 #	$a0 = addresss of src  vertices
 #	$a1 = addresss of dst  vertices
@@ -93,7 +130,6 @@ TransformTexturedQuad:
 	sd		$t0,0x00($a2) # tmp.x,y = t0
 	lw		$t0,0x08($a0) # t0 = src[0].z
 	sw		$t0,0x08($a2) # tmp.z = t0
-	# TRANSFORM VERTEX 1
 	TransformVertex1
 
 	# LOAD VERTEX 2
@@ -101,10 +137,6 @@ TransformTexturedQuad:
 	sd		$t0,0x00($a2) # tmp.x,y = t0
 	lw		$t0,0x20($a0) # t0 = src[1].z
 	sw		$t0,0x08($a2) # tmp.z = t0
-	# STORE CLIP FLAGS 1 RESULT
-	cfc2	$t0, $18	  # t0 = VP0_REGS[CLIP_FLAGS]
-	sw		$t0,0x00($a3) # clip_flags[0] = t0
-	# TRANSFORM VERTEX 2
 	TransformVertex2
 
 	# LOAD VERTEX 3
@@ -112,10 +144,6 @@ TransformTexturedQuad:
 	sd		$t0,0x00($a2) # tmp.x,y = t0
 	lw		$t0,0x38($a0) # t0 = src[2].z
 	sw		$t0,0x08($a2) # tmp.z = t0
-	# STORE CLIP FLAGS 2 RESULT
-	cfc2	$t0, $18	  # t0 = VP0_REGS[CLIP_FLAGS]
-	sw		$t0,0x04($a3) # clip_flags[1] = t0
-	# TRANSFORM VERTEX 3
 	TransformVertex3
 
 	# LOAD VERTEX 4
@@ -123,31 +151,53 @@ TransformTexturedQuad:
 	sd		$t0,0x00($a2) # tmp.x,y = t0
 	lw		$t0,0x50($a0) # t0 = src[3].z
 	sw		$t0,0x08($a2) # tmp.z = t0
-	# STORE CLIP FLAGS 3 RESULT
-	cfc2	$t0, $18	  # t0 = VP0_REGS[CLIP_FLAGS]
-	sw		$t0,0x08($a3) # clip_flags[2] = t0
-	# TRANSFORM VERTEX 4
 	TransformVertex4
 
-	# Desired output
-	# dst[0] = V0
-	# dst[1] = V1
-	# dst[2] = V2
-	# dst[3] = V2
-	# dst[4] = V3
-	# dst[5] = V0
-	sqc2	$vf15, 0x30($a1)   # dst[3] = TRANSFORMED(V2)
-	sqc2	$vf17, 0x40($a1)   # dst[4] = TRANSFORMED(V3)
-	sqc2	$vf11, 0x50($a1)   # dst[5] = TRANSFORMED(V0)
-	vnop					   # adjust for delay
-	# STORE CLIP FLAGS 4 RESULT
-	cfc2	$t0, $18	  	   # t0 = VP0_REGS[CLIP_FLAGS]
-	sw		$t0,0x0C($a3)      # clip_flags[3] = t0
-
+	TransformFinish
 	jr		$ra
 	nop
 
-.align 4
+# Transforms 4 vertices with size of 16 bytes
+#	$a0 = addresss of src  vertices
+#	$a1 = addresss of dst  vertices
+#   $a2 = address of  tmp  vertex
+#   $a3 = address of clip flags
+TransformColouredQuad:
+	# LOAD 1.0 into W
+	lw		$t0,ONE_VALUE # t0 = 1.0f
+	sw		$t0,0x0C($a2) # tmp.w = f5
+
+	# LOAD VERTEX 1
+	ld		$t0,0x00($a0) # t0 = src[0].x,y
+	sd		$t0,0x00($a2) # tmp.x,y = t0
+	lw		$t0,0x08($a0) # t0 = src[0].z
+	sw		$t0,0x08($a2) # tmp.z = t0
+	TransformVertex1
+
+	# LOAD VERTEX 2
+	ld		$t0,0x10($a0) # t0 = src[1].x,y
+	sd		$t0,0x00($a2) # tmp.x,y = t0
+	lw		$t0,0x18($a0) # t0 = src[1].z
+	sw		$t0,0x08($a2) # tmp.z = t0
+	TransformVertex2
+
+	# LOAD VERTEX 3
+	ld		$t0,0x20($a0) # t0 = src[2].x,y
+	sd		$t0,0x00($a2) # tmp.x,y = t0
+	lw		$t0,0x28($a0) # t0 = src[2].z
+	sw		$t0,0x08($a2) # tmp.z = t0
+	TransformVertex3
+
+	# LOAD VERTEX 4
+	ld		$t0,0x30($a0) # t0 = src[3].x,y
+	sd		$t0,0x00($a2) # tmp.x,y = t0
+	lw		$t0,0x38($a0) # t0 = src[3].z
+	sw		$t0,0x08($a2) # tmp.z = t0
+	TransformVertex4
+
+	TransformFinish
+	jr		$ra
+	nop
 
 .global ONE_VALUE
 ONE_VALUE:  .float 1.0
diff --git a/src/Graphics_PS2.c b/src/Graphics_PS2.c
index 00887bf32..a4df0821b 100644
--- a/src/Graphics_PS2.c
+++ b/src/Graphics_PS2.c
@@ -521,53 +521,6 @@ void Gfx_SetVertexFormat(VertexFormat fmt) {
 	formatDirty = true;
 }
 
-static cc_bool NotClipped(VU0_VECTOR pos) {
-	// The code below clips to the viewport clip planes
-	//  For e.g. X this is [2048 - vp_width / 2, 2048 + vp_width / 2]
-	//  However the guard band itself ranges from 0 to 4096
-	// To reduce need to clip, clip against guard band on X/Y axes instead
-	/*return
-		xAdj  >= -pos.w && xAdj  <= pos.w &&
-		yAdj  >= -pos.w && yAdj  <= pos.w &&
-		pos.z >= -pos.w && pos.z <= pos.w;*/	
-		
-	// Rescale clip planes to guard band extent:
-	//  X/W * vp_hwidth <= vp_hwidth -- clipping against viewport
-	//              X/W <= 1
-	//              X   <= W
-	//  X/W * vp_hwidth <= 2048      -- clipping against guard band
-	//              X/W <= 2048 / vp_hwidth
-	//              X * vp_hwidth / 2048 <= W
-	float xAdj = pos.x * (vp_hwidth/2048);
-	float yAdj = pos.y * (vp_hheight/2048);
-	
-	// X/W * vp_hwidth <= 2048
-	// 
-		
-	// Clip X/Y to INSIDE the guard band regions
-	return
-		xAdj > -pos.w && xAdj < pos.w &&
-		yAdj > -pos.w && yAdj < pos.w &&
-		pos.z >= -pos.w && pos.z <= pos.w;
-}
-
-static void TransformVertex(void* raw, VU0_VECTOR* dst) {
-	Vec3* pos = raw;
-
-	VU0_VECTOR coord; coord.x = pos->x; coord.y = pos->y; coord.z = pos->z; coord.w = 1.0f;
-	asm __volatile__ (
-		"lqc2		$vf8, 0x00(%0)	\n" // vf8 = coord
-		"vmulaw		$ACC, $vf4, $vf0\n" // ACC[xyzw] = mvp.row3[xyzw] * 1.0f; (vf0.w is 1)
-		"vmaddax	$ACC, $vf1, $vf8\n" // ACC[xyzw] = ACC[xyzw] + mvp.row0[xyzw] * coord.x
-		"vmadday	$ACC, $vf2, $vf8\n" // ACC[xyzw] = ACC[xyzw] + mvp.row2[xyzw] * coord.y
-		"vmaddz		$vf9, $vf3, $vf8\n" // vf9[xyzw] = ACC[xyzw] + mvp.row3[xyzw] * coord.z
-		"sqc2		$vf9, 0x00(%1)	\n" // dst = vf9
-	: 
-	: "r" (&coord), "r" (dst)
-	: "memory"
-	);
-}
-
 //#define VCopy(dst, src) dst.x = vp_hwidth  * (1 + src.x / src.w); dst.y = vp_hheight * (1 - src.y / src.w); dst.z = src.z / src.w; dst.w = src.w;
 static xyz_t FinishVertex(VU0_VECTOR* src, float invW) {
 	float x = vp_hwidth  * (src->x * invW);
@@ -640,6 +593,7 @@ static u64* DrawTexturedTriangle(u64* dw, VU0_VECTOR* coords,
 }
 
 extern void TransformTexturedQuad(void* src, VU0_VECTOR* dst, VU0_VECTOR* tmp, int* clip_flags);
+extern void TransformColouredQuad(void* src, VU0_VECTOR* dst, VU0_VECTOR* tmp, int* clip_flags);
 
 static void DrawTexturedTriangles(int verticesCount, int startVertex) {
 	struct VertexTextured* v = (struct VertexTextured*)gfx_vertices + startVertex;
@@ -685,29 +639,20 @@ static void DrawColouredTriangles(int verticesCount, int startVertex) {
 	u64* dw = (u64*)q;
 
 	unsigned numVerts = 0;
-	VU0_VECTOR V[4];
+	VU0_VECTOR V[6], tmp;
+	int clip[4];
 
 	for (int i = 0; i < verticesCount / 4; i++, v += 4)
 	{
-		TransformVertex(v + 0, &V[0]);
-		TransformVertex(v + 1, &V[1]);
-		TransformVertex(v + 2, &V[2]);
-		TransformVertex(v + 3, &V[3]);
+		TransformColouredQuad(v, V, &tmp, clip);
 		
-		// V0, V1, V2
-		if (NotClipped(V[0]) && NotClipped(V[1]) && NotClipped(V[2])) {
+		if (((clip[0] | clip[1] | clip[2]) & 0x3F) == 0) {
 			dw = DrawColouredTriangle(dw, V, v + 0, v + 1, v + 2);
 			numVerts += 3;
 		}
-
-		VU0_VECTOR v0 = V[0];
-		V[0] = V[2];
-		V[1] = V[3];
-		V[2] = v0;
 		
-		// V2, V3, V0
-		if (NotClipped(V[0]) && NotClipped(V[1]) && NotClipped(V[2])) {
-			dw = DrawColouredTriangle(dw, V, v + 2, v + 3, v + 0);
+		if (((clip[2] | clip[3] | clip[0]) & 0x3F) == 0) {
+			dw = DrawColouredTriangle(dw, V + 3, v + 2, v + 3, v + 0);
 			numVerts += 3;
 		}
 	}
diff --git a/src/Platform_Dreamcast.c b/src/Platform_Dreamcast.c
index 962f701f5..efd590aa3 100644
--- a/src/Platform_Dreamcast.c
+++ b/src/Platform_Dreamcast.c
@@ -541,6 +541,7 @@ static void InitSDCard(void) {
 	if (sd_blockdev_for_partition(0, &sd_dev, &partition_type)) {
 		Platform_LogConst("Unable to find first partition on SD card"); return;
 	}
+	Platform_Log1("Found SD card (partitioned using: %b)", &partition_type);
 	
 	if (fs_fat_init()) {
 		Platform_LogConst("Failed to init FAT filesystem"); return;