Graphics: (faked) bounding box support. Helps some Paper Mario effects although they're still a bit glitchy. Might also help other games? Enable with the #define in VideoCommon.h. Since there might be a speed hit it's off by default.

git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@3580 8ced0084-cf51-0410-be5f-012b33b47a6e

Source/Core/Common/Src/MathUtil.cpp

@@ -145,12 +145,15 @@ void SaveSSEState()
 	saved_sse_state = _mm_getcsr();
 }
 
-void MatrixMul(int n, const float *a, const float *b, float *result)
+inline void MatrixMul(int n, const float *a, const float *b, float *result)
 {    
-    for(int i = 0; i < n; ++i) {        
+    for (int i = 0; i < n; ++i)
-        for(int j= 0; j < n; ++j) {
+	{
+        for (int j = 0; j < n; ++j)
+		{
             float temp = 0;
-            for(int k = 0; k < n; ++k) {
+            for (int k = 0; k < n; ++k)
+			{
                 temp += a[i * n + k] * b[k * n + j];
             }
             result[i * n + j] = temp;
@@ -207,9 +210,9 @@ void Matrix33::Multiply(const Matrix33 &a, const Matrix33 &b, Matrix33 &result)
 
 void Matrix33::Multiply(const Matrix33 &a, const float vec[3], float result[3])
 {
-    for(int i = 0; i < 3; ++i) {
+    for (int i = 0; i < 3; ++i) {
         result[i] = 0;
-        for(int k = 0; k < 3; ++k) {
+        for (int k = 0; k < 3; ++k) {
             result[i] += a.data[i * 3 + k] * vec[k];
         }
     }
@@ -226,13 +229,16 @@ void Matrix44::LoadIdentity(Matrix44 &mtx)
 
 void Matrix44::LoadMatrix33(Matrix44 &mtx, const Matrix33 &m33)
 {
-    for(int i = 0; i < 3; ++i) {
+    for (int i = 0; i < 3; ++i)
-        for(int j = 0; j < 3; ++j) {
+	{
+        for (int j = 0; j < 3; ++j)
+		{
             mtx.data[i * 4 + j] = m33.data[i * 3 + j];
         }
     }
 
-    for(int i = 0; i < 3; ++i) {
+    for (int i = 0; i < 3; ++i)
+	{
         mtx.data[i * 4 + 3] = 0;
         mtx.data[i + 12] = 0;
     }

Source/Core/Common/Src/MathUtil.h

@@ -157,4 +157,5 @@ public:
     float data[16];
 };
 
+
 #endif // _MATH_UTIL_H_

Source/Core/Core/Src/Core.cpp

@@ -387,6 +387,10 @@ THREAD_RETURN EmuThread(void *pArg)
 	VideoInitialize.pKeyPress           = Callback_KeyPress;
 	VideoInitialize.bWii                = _CoreParameter.bWii;
 	VideoInitialize.bUseDualCore		= _CoreParameter.bUseDualCore;
+	VideoInitialize.pBBox               = &PixelEngine::bbox[0];
+	VideoInitialize.pBBoxActive         = &PixelEngine::bbox_active;
+
+
 	// May be needed for Stop and Start
 	#ifdef SETUP_FREE_VIDEO_PLUGIN_ON_BOOT
 		Plugins.FreeVideo();

Source/Core/Core/Src/HW/PixelEngine.cpp

@@ -128,6 +128,9 @@ static bool g_bSignalFinishInterrupt;
 static int et_SetTokenOnMainThread;
 static int et_SetFinishOnMainThread;
 
+u16 bbox[4];
+bool bbox_active;
+
 void DoState(PointerWrap &p)
 {
 	p.Do(m_ZConf);
@@ -140,6 +143,9 @@ void DoState(PointerWrap &p)
 
 	p.Do(g_bSignalTokenInterrupt);
 	p.Do(g_bSignalFinishInterrupt);
+	
+	p.Do(bbox);
+	p.Do(bbox_active);
 }
 
 void UpdateInterrupts();
@@ -153,6 +159,13 @@ void Init()
 
 	et_SetTokenOnMainThread = CoreTiming::RegisterEvent("SetToken", SetToken_OnMainThread);
 	et_SetFinishOnMainThread = CoreTiming::RegisterEvent("SetFinish", SetFinish_OnMainThread);
+
+	bbox[0] = 0x80;
+	bbox[1] = 0xA0;
+	bbox[2] = 0x80;
+	bbox[3] = 0xA0;
+
+	bbox_active = false;
 }
 
 void Read16(u16& _uReturnValue, const u32 _iAddress)
@@ -196,18 +209,12 @@ void Read16(u16& _uReturnValue, const u32 _iAddress)
 
 		// The return values for these BBOX registers need to be gotten from the bounding box of the object. 
 		// See http://code.google.com/p/dolphin-emu/issues/detail?id=360#c74 for more details.
-	case PE_BBOX_LEFT:
+
-		_uReturnValue = 0x80;
+	// 0x80, 0xa0, 0x80, 0xa0 makes Paper Mario happy.
-		break;
+	case PE_BBOX_LEFT:   _uReturnValue = bbox[0]; INFO_LOG(PIXELENGINE, "R: BBOX_LEFT   = %i", bbox[0]); bbox_active = false; break;
-	case PE_BBOX_RIGHT:
+	case PE_BBOX_RIGHT:  _uReturnValue = bbox[1]; INFO_LOG(PIXELENGINE, "R: BBOX_RIGHT  = %i", bbox[1]); bbox_active = false; break;
-		_uReturnValue = 0xA0;
+	case PE_BBOX_TOP:    _uReturnValue = bbox[2]; INFO_LOG(PIXELENGINE, "R: BBOX_TOP    = %i", bbox[2]); bbox_active = false; break;
-		break;
+	case PE_BBOX_BOTTOM: _uReturnValue = bbox[3]; INFO_LOG(PIXELENGINE, "R: BBOX_BOTTOM = %i", bbox[3]); bbox_active = false; break;
-	case PE_BBOX_TOP:
-		_uReturnValue = 0x80;
-		break;
-	case PE_BBOX_BOTTOM:
-		_uReturnValue = 0xA0;
-		break;
 
 	default:
 		WARN_LOG(PIXELENGINE, "(r16) unknown @ %08x", _iAddress);

Source/Core/Core/Src/HW/PixelEngine.h

@@ -55,6 +55,10 @@ void SetToken(const u16 _token, const int _bSetTokenAcknowledge);
 void SetFinish(void);
 bool AllowIdleSkipping();
 
+// Bounding box functionality. Paper Mario (both) are a couple of the few games that use it.
+extern u16 bbox[4];
+extern bool bbox_active;
+
 } // end of namespace PixelEngine
 
 #endif

Source/Core/VideoCommon/Src/BPMemory.h

@@ -62,6 +62,7 @@
 #define BPMEM_COPYFILTER1      0x54
 #define BPMEM_CLEARBBOX1       0x55 
 #define BPMEM_CLEARBBOX2       0x56
+// what about 0x57?
 #define BPMEM_UNKNOWN          0x58
 #define BPMEM_SCISSOROFFSET    0x59
 #define BPMEM_UNKNOWN1         0x60

Source/Core/VideoCommon/Src/BPStructs.cpp

@@ -211,7 +211,12 @@ void BPWritten(const Bypass& bp)
 			}
 			else
 			{
+				// 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
+				*g_VideoInitialize.pBBoxActive = false;
+#endif
 				const float yScale = bpmem.dispcopyyscale / 256.0f;
 				const float xfbLines = ((bpmem.copyTexSrcWH.y + 1.0f) * yScale);
 				RenderToXFB(bp, multirc, yScale, xfbLines, 
@@ -324,13 +329,35 @@ void BPWritten(const Bypass& bp)
 	case BPMEM_CLEAR_Z:  // Z Components (24-bit Zbuffer)
 		break;
 	// -------------------------
-	// Culling Occulsion, we don't support this
+	// Bounding Box support
-	// let's hope not many games use bboxes..
-	// TODO(ector): add something that watches bboxes
 	// -------------------------
 	case BPMEM_CLEARBBOX1:
-	case BPMEM_CLEARBBOX2:
+	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;
+			
+			// We should only set these if bbox is calculated properly.
+			g_VideoInitialize.pBBox[0] = left;
+			g_VideoInitialize.pBBox[1] = right;
+			*g_VideoInitialize.pBBoxActive = 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.
+			g_VideoInitialize.pBBox[2] = top;
+			g_VideoInitialize.pBBox[3] = bottom;
+			*g_VideoInitialize.pBBoxActive = true;
+			// WARN_LOG(VIDEO, "ClearBBox TB: %i, %08x - %i, %i", bp.address, bp.newvalue, top, bottom);
+		}
+#endif
 		break;
+						   }
 	case BPMEM_ZCOMPARE:      // Set the Z-Compare
 	case BPMEM_TEXINVALIDATE: // Used, if game has manual control the Texture Cache, which we don't allow
 	case BPMEM_MIPMAP_STRIDE: // MipMap Stride Channel
@@ -363,14 +390,14 @@ void BPWritten(const Bypass& bp)
 			PanicAlert("Unknown is not 0xF! val = 0x%08x", bp.newvalue);
 		break;
 
-	// Cases added due to: http://code.google.com/p/dolphin-emu/issues/detail?id=360#c90
-	// Are these related to BBox?
 	case BPMEM_UNKNOWN1:
 	case BPMEM_UNKNOWN2:
 	case BPMEM_UNKNOWN3:
 	case BPMEM_UNKNOWN4:
-
+		// Cases added due to: http://code.google.com/p/dolphin-emu/issues/detail?id=360#c90
+		// Are these related to BBox?
 		break;
+
 		// ------------------------------------------------
 		// On Default, we try to look for other things
 		// before we give up and say its an unknown opcode

Source/Core/VideoCommon/Src/VertexLoader.cpp

@@ -38,6 +38,10 @@
 #include "VertexLoader_Color.h"
 #include "VertexLoader_TextCoord.h"
 
+//BBox
+#include "XFMemory.h"
+extern float GC_ALIGNED16(g_fProjectionMatrix[16]);
+
 #define USE_JIT
 
 #define COMPILED_CODE_SIZE 4096
@@ -82,6 +86,50 @@ void LOADERDECL PosMtx_Write()
 	*VertexManager::s_pCurBufferPointer++ = 0;
 }
 
+void LOADERDECL UpdateBoundingBox() 
+{
+	if (!*g_VideoInitialize.pBBoxActive)
+		return;
+
+	// Truly evil hack, reading backwards from the write pointer. If we were writing to write-only
+	// memory like we might have been with a D3D vertex buffer, this would have been a bad idea.
+	float *data = (float *)(VertexManager::s_pCurBufferPointer - 12);
+	// We must transform the just loaded point by the current world and projection matrix - in software.
+	// Then convert to screen space and update the bounding box.
+	float p[3] = {data[0], data[1], data[2]};
+
+    const float *world_matrix  = (float*)xfmem + MatrixIndexA.PosNormalMtxIdx * 4;
+	const float *proj_matrix = &g_fProjectionMatrix[0];
+
+	float t[3];
+	t[0] = p[0] * world_matrix[0] + p[1] * world_matrix[1] + p[2] * world_matrix[2] + world_matrix[3];
+	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[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[2] = t[0] * proj_matrix[8]  + t[1] * proj_matrix[9]  + t[2] * proj_matrix[10] + proj_matrix[11];
+	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];
+
+	// should possibly adjust for viewport?
+	o[0] = (o[0] + 1.0f) * 320.0f;
+	o[1] = (o[1] + 1.0f) * 240.0f;
+	
+	if (o[0] < g_VideoInitialize.pBBox[0]) g_VideoInitialize.pBBox[0] = std::max(0.0f, o[0]);
+	if (o[0] > g_VideoInitialize.pBBox[1]) g_VideoInitialize.pBBox[1] = std::min(640.0f, o[0]);
+	if (o[1] < g_VideoInitialize.pBBox[2]) g_VideoInitialize.pBBox[2] = std::max(0.0f, o[1]);
+	if (o[1] > g_VideoInitialize.pBBox[3]) g_VideoInitialize.pBBox[3] = std::min(480.0f, o[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]);
+	}*/
+}
+
 void LOADERDECL TexMtx_ReadDirect_UByte()
 {
 	s_curtexmtx[s_texmtxread] = DataReadU8() & 0x3f;
@@ -247,6 +295,12 @@ void VertexLoader::CompileVertexTranslator()
 		break;
 	}
 
+	// OK, so we just got a point. Let's go back and read it for the bounding box.
+	
+#ifdef BBOX_SUPPORT
+	WriteCall(UpdateBoundingBox);
+#endif
+
 	// Normals
 	vtx_decl.num_normals = 0;
 	if (m_VtxDesc.Normal != NOT_PRESENT) {

Source/Core/VideoCommon/Src/VertexShaderManager.cpp

@@ -36,7 +36,8 @@ namespace VertexManager
 	void Flush();
 }
 
-static float s_fMaterials[16];
+static float GC_ALIGNED16(s_fMaterials[16]);
+float GC_ALIGNED16(g_fProjectionMatrix[16]);
 
 // track changes
 static bool bTexMatricesChanged[2], bPosNormalMatrixChanged, bProjectionChanged, bViewportChanged;
@@ -204,8 +205,6 @@ void VertexShaderManager::SetConstants(bool proj_hax_1,bool Hack_hack1 ,float Ha
 
     if (bProjectionChanged) {
         bProjectionChanged = false;
-		static float GC_ALIGNED16(g_fProjectionMatrix[16]);
-
 
         if (xfregs.rawProjection[6] == 0) { // Perspective
             g_fProjectionMatrix[0] = xfregs.rawProjection[0];

Source/Core/VideoCommon/Src/VideoCommon.h

@@ -45,7 +45,15 @@ enum
 	// TODO: figure out what to do with PAL
 };
 
+// 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).
+// May help Super Mario Galaxy?
+// Do testing to figure out if the speed hit is bad?
+// #define BBOX_SUPPORT
+
 extern SVideoInitialize g_VideoInitialize;
+
 // (mb2) for XFB update hack. TODO: find a static better place
 extern volatile u32 g_XFBUpdateRequested;
 

Source/PluginSpecs/pluginspecs_video.h

@@ -72,6 +72,11 @@ typedef struct
 	void *pMemoryBase;
 	bool bWii;
 	bool bUseDualCore;
+
+	unsigned short                  *pBBox;  // points to four shorts: left, top, right, bottom
+	// TODO:
+	bool                            *pBBoxActive;  // we guess that after a bbox reset, we only need to track bbox size until the corresponding read.
+
 } SVideoInitialize;
 
 /////////////////////////////////////////////////////////////////////////////////////////////////////