Added proper Bounding Box support.

Should fix most graphical issues with Paper Mario: TTYD and Super Paper Mario. Fixes issue 360.

Since only those two games seem to require BBox support, and as per ector's suggestion, BBox is only enabled for those two games.

BBoxes and Display List Caches don't get along too well, causing Paper Mario: TTYD to hang during certain effects where BBoxes are used. For now, I disabled DList Cache for the Paper Mario games, hopefully both will be compatible in the future.

5 files changed, 63 insertions, 62 deletions

diff --git a/Source/Core/VideoCommon/Src/BPStructs.cpp b/Source/Core/VideoCommon/Src/BPStructs.cpp
index cee68327f5..d5909fdd95 100644
--- a/Source/Core/VideoCommon/Src/BPStructs.cpp
+++ b/Source/Core/VideoCommon/Src/BPStructs.cpp
@@ -257,9 +257,8 @@ void BPWritten(const BPCmd& bp)
 				// We should be able to get away with deactivating the current bbox tracking
 				// here. Not sure if there's a better spot to put this.
 				// the number of lines copied is determined by the y scale * source efb height
-#ifdef BBOX_SUPPORT
+			
 				PixelEngine::bbox_active = false;
-#endif
 
 				float yScale;
 				if (PE_copy.scale_invert)
@@ -400,28 +399,29 @@ void BPWritten(const BPCmd& bp)
 	case BPMEM_CLEARBBOX1:
 	case BPMEM_CLEARBBOX2:
 		{
-#ifdef BBOX_SUPPORT
-		// which is which? these are GUESSES!
-		if (bp.address == BPMEM_CLEARBBOX1) {
-			int right = bp.newvalue >> 10;
-			int left = bp.newvalue & 0x3ff;
+		if(g_ActiveConfig.bUseBBox)
+		{
+			// which is which? these are GUESSES!
+			if (bp.address == BPMEM_CLEARBBOX1) {
+				int right = bp.newvalue >> 10;
+				int left = bp.newvalue & 0x3ff;
 			
-			// We should only set these if bbox is calculated properly.
-			PixelEngine::bbox[0] = left;
-			PixelEngine::bbox[1] = right;
-			PixelEngine::bbox_active = true;
-			// WARN_LOG(VIDEO, "ClearBBox LR: %i, %08x - %i, %i", bp.address, bp.newvalue, left, right);
-		} else {
-			int bottom = bp.newvalue >> 10;
-			int top = bp.newvalue & 0x3ff;
-
-			// We should only set these if bbox is calculated properly.
-			PixelEngine::bbox[2] = top;
-			PixelEngine::bbox[3] = bottom;
-			PixelEngine::bbox_active = true;
-			// WARN_LOG(VIDEO, "ClearBBox TB: %i, %08x - %i, %i", bp.address, bp.newvalue, top, bottom);
+				// We should only set these if bbox is calculated properly.
+				PixelEngine::bbox[0] = left;
+				PixelEngine::bbox[1] = right;
+				PixelEngine::bbox_active = true;
+				// WARN_LOG(VIDEO, "ClearBBox LR: %i, %08x - %i, %i", bp.address, bp.newvalue, left, right);
+			} else {
+				int bottom = bp.newvalue >> 10;
+				int top = bp.newvalue & 0x3ff;
+
+				// We should only set these if bbox is calculated properly.
+				PixelEngine::bbox[2] = top;
+				PixelEngine::bbox[3] = bottom;
+				PixelEngine::bbox_active = true;
+				// WARN_LOG(VIDEO, "ClearBBox TB: %i, %08x - %i, %i", bp.address, bp.newvalue, top, bottom);
+			}
 		}
-#endif
 		}
 		break;
 	case BPMEM_TEXINVALIDATE: // Used, if game has manual control the Texture Cache, which we don't allow
diff --git a/Source/Core/VideoCommon/Src/VertexLoader.cpp b/Source/Core/VideoCommon/Src/VertexLoader.cpp
index ce9e81219d..ee0538751a 100644
--- a/Source/Core/VideoCommon/Src/VertexLoader.cpp
+++ b/Source/Core/VideoCommon/Src/VertexLoader.cpp
@@ -101,7 +101,7 @@ void LOADERDECL PosMtx_Write()
 
 void LOADERDECL UpdateBoundingBox() 
 {
-    if (!PixelEngine::bbox_active)
+	if (!PixelEngine::bbox_active)
 		return;
 
 	// Truly evil hack, reading backwards from the write pointer. If we were writing to write-only
@@ -119,40 +119,46 @@ void LOADERDECL UpdateBoundingBox()
 	t[1] = p[0] * world_matrix[4] + p[1] * world_matrix[5] + p[2] * world_matrix[6] + world_matrix[7];
 	t[2] = p[0] * world_matrix[8] + p[1] * world_matrix[9] + p[2] * world_matrix[10] + world_matrix[11];
 
-	float o[4];
-	o[2] = t[0] * proj_matrix[8]  + t[1] * proj_matrix[9]  + t[2] * proj_matrix[10] + proj_matrix[11];
-	// Depth culling
-	if (o[2] < 0.0) {
-		// No pixels are likely to be drawn - don't update bounding box.
-		return;
-	}
+	float o[3];
 	o[0] = t[0] * proj_matrix[0]  + t[1] * proj_matrix[1]  + t[2] * proj_matrix[2] + proj_matrix[3];
 	o[1] = t[0] * proj_matrix[4]  + t[1] * proj_matrix[5]  + t[2] * proj_matrix[6] + proj_matrix[7];
-	o[3] = t[0] * proj_matrix[12] + t[1] * proj_matrix[13] + t[2] * proj_matrix[14] + proj_matrix[15];
-	
-	o[0] /= o[3];
-	o[1] /= o[3];
+	o[2] = t[0] * proj_matrix[12] + t[1] * proj_matrix[13] + t[2] * proj_matrix[14] + proj_matrix[15];
+
+
+	o[0] /= o[2];
+	o[1] /= o[2];
 
 	// should possibly adjust for viewport?
-	o[0] = (o[0] + 1.0f) * 320.0f;
-	o[1] = (o[1] + 1.0f) * 240.0f;
-
-    if (o[0] < PixelEngine::bbox[0]) PixelEngine::bbox[0] = (u16)std::max(0.0f, o[0]);
-	if (o[0] > PixelEngine::bbox[1]) PixelEngine::bbox[1] = (u16)std::min(640.0f, o[0]);
-	if (o[1] < PixelEngine::bbox[2]) PixelEngine::bbox[2] = (u16)std::max(0.0f, o[1]);
-	if (o[1] > PixelEngine::bbox[3]) PixelEngine::bbox[3] = (u16)std::min(480.0f, o[1]);
-
-	// Hardware tests bounding boxes in 2x2 blocks => left and top are even, right and bottom are odd
-	PixelEngine::bbox[0] &= ~1;
-	PixelEngine::bbox[1] |= 1;
-	PixelEngine::bbox[2] &= ~1;
-	PixelEngine::bbox[3] |= 1;
-
-	/*
-	if (GetAsyncKeyState(VK_LSHIFT)) {
-		ERROR_LOG(VIDEO, "XForm: %f %f %f to %f %f", p[0], p[1], p[2], o[0], o[1]);
-		ERROR_LOG(VIDEO, "%i %i %i %i", g_VideoInitialize.pBBox[0], g_VideoInitialize.pBBox[1], g_VideoInitialize.pBBox[2], g_VideoInitialize.pBBox[3]);
-	}*/
+	o[0] = (o[0] + 1.0f) * 304.0f;
+	o[1] = (1.0f - o[1]) * 240.0f;
+
+    if (o[0] < PixelEngine::bbox[0])
+	{
+		PixelEngine::bbox[0] = (u16) std::max(0.0f, o[0]);
+
+		// Hardware tests bounding boxes in 2x2 blocks => left and top are even, right and bottom are odd
+		PixelEngine::bbox[0] &= ~1;
+	}
+
+	if (o[0] > PixelEngine::bbox[1])
+	{
+		PixelEngine::bbox[1] = (u16) std::min(608.0f, o[0]);
+		if(!(PixelEngine::bbox[1] & 1) && PixelEngine::bbox[1] != 0)
+			PixelEngine::bbox[1]--;
+	}
+
+	if (o[1] < PixelEngine::bbox[2])
+	{
+		PixelEngine::bbox[2] = (u16) std::max(0.0f, o[1]);
+		PixelEngine::bbox[2] &= ~1;
+	}
+
+	if (o[1] > PixelEngine::bbox[3])
+	{
+		PixelEngine::bbox[3] = (u16) std::min(480.0f, o[1]);
+		if(!(PixelEngine::bbox[3] & 1) && PixelEngine::bbox[3] != 0)
+			PixelEngine::bbox[3]--;
+	}
 }
 
 void LOADERDECL TexMtx_ReadDirect_UByte()
@@ -290,9 +296,8 @@ void VertexLoader::CompileVertexTranslator()
 
 	// OK, so we just got a point. Let's go back and read it for the bounding box.
 	
-#ifdef BBOX_SUPPORT
-	WriteCall(UpdateBoundingBox);
-#endif
+	if(g_ActiveConfig.bUseBBox)
+		WriteCall(UpdateBoundingBox);
 
 	// Normals
 	vtx_decl.num_normals = 0;
diff --git a/Source/Core/VideoCommon/Src/VideoCommon.h b/Source/Core/VideoCommon/Src/VideoCommon.h
index 80c0937884..9f6dd59cd3 100644
--- a/Source/Core/VideoCommon/Src/VideoCommon.h
+++ b/Source/Core/VideoCommon/Src/VideoCommon.h
@@ -56,12 +56,6 @@ enum
 	MAX_XFB_HEIGHT = 574
 };
 
-// If this is enabled, bounding boxes will be computed for everything drawn.
-// This can theoretically have a big speed hit in some geom heavy games. Needs more work.
-// Helps some effects in Paper Mario (but they aren't quite right yet).
-// Do testing to figure out if the speed hit is bad?
-// #define BBOX_SUPPORT
-
 // Logging
 // ----------
 void HandleGLError();
diff --git a/Source/Core/VideoCommon/Src/VideoConfig.cpp b/Source/Core/VideoCommon/Src/VideoConfig.cpp
index 2c7788c9cc..b6ba76f45e 100644
--- a/Source/Core/VideoCommon/Src/VideoConfig.cpp
+++ b/Source/Core/VideoCommon/Src/VideoConfig.cpp
@@ -172,6 +172,7 @@ void VideoConfig::GameIniLoad(const char *ini_file)
 	iniFile.GetIfExists("Video", "PH_ZNear", &sPhackvalue[0]);
 	iniFile.GetIfExists("Video", "PH_ZFar", &sPhackvalue[1]);
 	iniFile.GetIfExists("Video", "ZTPSpeedupHack", &bZTPSpeedHack);
+	iniFile.GetIfExists("Video", "UseBBox", &bUseBBox);
 }
 
 void VideoConfig::VerifyValidity()
diff --git a/Source/Core/VideoCommon/Src/VideoConfig.h b/Source/Core/VideoCommon/Src/VideoConfig.h
index 83ec08784d..3bd833cc73 100644
--- a/Source/Core/VideoCommon/Src/VideoConfig.h
+++ b/Source/Core/VideoCommon/Src/VideoConfig.h
@@ -135,6 +135,7 @@ struct VideoConfig
 	std::string sPhackvalue[2];
 	float fAspectRatioHackW, fAspectRatioHackH;
 	bool bZTPSpeedHack; // The Legend of Zelda: Twilight Princess
+	bool bUseBBox;
 	bool bEnablePixelLighting;
 	bool bEnablePerPixelDepth;