GregMiscellaneous: zzogl-pg:

* use a size of 4 for the gsvertex array (allow compiler to optimize a % into an and)
* Reuse previous computed vertex for _strip & _fan primitives


git-svn-id: http://pcsx2.googlecode.com/svn/branches/GregMiscellaneous@3973 96395faa-99c1-11dd-bbfe-3dabce05a288

plugins/zzogl-pg/opengl/GS.h

@@ -471,14 +471,16 @@ typedef struct
 
 typedef struct
 {
-	Vertex gsvertex[3];
+	Vertex gsvertex[4]; // circular buffer that contains the vertex
-	u32 rgba;
+    Vertex gsTriFanVertex; // Base of triangle fan primitive vertex
+	u32 rgba; // global color for flat shading texture
 	float q;
-	Vertex vertexregs;
+	Vertex vertexregs; // accumulation buffer that collect current vertex data
 
 	int primC;		// number of verts current storing
 	int primIndex;	// current prim index
-	int nTriFanVert;
+	int nTriFanVert; // remember the index of the base of triangle fan
+    int new_tri_fan; // 1 if we process a new triangle fan primitive. 0 otherwise
 
 	int prac;
 	int dthe;
@@ -512,12 +514,14 @@ typedef struct
 	GSClut clut_buffer;
 	int primNext(int inc = 1)
 	{
+        // Note: ARRAY_SIZE(gsvertex) == 2^n => modulo is replaced by an and instruction
 		return ((primIndex + inc) % ARRAY_SIZE(gsvertex));
 	}
 	
     int primPrev(int dec = 1)
     {
         // Note: assert( dec <= ARRAY_SIZE(gsvertex) );
+        // Note: ARRAY_SIZE(gsvertex) == 2^n => modulo is replaced by an and instruction
 		return ((primIndex + (ARRAY_SIZE(gsvertex) - dec)) % ARRAY_SIZE(gsvertex));
     }
 	
@@ -535,7 +539,12 @@ typedef struct
 		vertexregs.y = y;
 		vertexregs.z = z;
 		vertexregs.f = f;
-		gsvertex[primIndex] = vertexregs;
+        if (likely(!new_tri_fan)) {
+            gsvertex[primIndex] = vertexregs;
+        } else {
+            gsTriFanVertex = vertexregs;
+            new_tri_fan = false;
+        }
 	}
 	
 	inline void add_vertex(u16 x, u16 y, u32 z)
@@ -543,7 +552,12 @@ typedef struct
 		vertexregs.x = x;
 		vertexregs.y = y;
 		vertexregs.z = z;
-		gsvertex[primIndex] = vertexregs;
+        if (likely(!new_tri_fan)) {
+            gsvertex[primIndex] = vertexregs;
+        } else {
+            gsTriFanVertex = vertexregs;
+            new_tri_fan = false;
+        }
 	}
 } GSinternal;
 

plugins/zzogl-pg/opengl/GSmain.cpp

@@ -309,7 +309,6 @@ void CALLBACK GSreset()
 
 	gs.prac = 1;
 	prim = &gs._prim[0];
-	gs.nTriFanVert = -1;
 	gs.imageTransfer = -1;
 	gs.q = 1;
 }
@@ -324,7 +323,6 @@ void CALLBACK GSgifSoftReset(u32 mask)
 
 	gs.imageTransfer = -1;
 	gs.q = 1;
-	gs.nTriFanVert = -1;
 }
 
 s32 CALLBACK GSinit()

plugins/zzogl-pg/opengl/Regs.cpp

@@ -305,12 +305,14 @@ void __gifCall GIFRegHandlerPRIM(const u32 *data)
 
 
 	gs.primC = 0;
-	prim->prim = (data[0]) & 0x7;
+    u16 prim_type = (data[0]) & 0x7;
-	gs._prim[0].prim = (data[0]) & 0x7;
+	prim->prim = prim_type;
-	gs._prim[1].prim = (data[0]) & 0x7;
+	gs._prim[0].prim = prim_type;
+	gs._prim[1].prim = prim_type;
 	gs._prim[1]._val = (data[0] >> 3) & 0xff;
 
-    gs.nTriFanVert = gs.primIndex;
+    gs.new_tri_fan = !(prim_type ^ PRIM_TRIANGLE_FAN);
+    ZZKick->DirtyValidPrevPrim();
 
 	Prim();
 }

plugins/zzogl-pg/opengl/ZZKick.cpp

@@ -57,18 +57,11 @@ void __forceinline Kick::KickVertex(bool adc)
 	if (++gs.primC >= (int)g_primmult[prim->prim])
 	{
 		if (!adc && NoHighlights(prim->ctxt)) DrawPrim(prim->prim);
+        else DirtyValidPrevPrim();
 		
 		gs.primC -= g_primsub[prim->prim];
-
-        gs.primIndex = gs.primNext();
-		if (prim->prim == 5)
-		{
-			/* tri fans need special processing */
-			if (gs.nTriFanVert == gs.primIndex) gs.primIndex = gs.primNext();
 		}
-	} else {
+    gs.primIndex = gs.primNext();
-        gs.primIndex = gs.primNext();
-	}
 }
 
 template<bool DO_Z_FOG>
@@ -111,7 +104,7 @@ void Kick::Set_Vertex(VertexGPU *p, Vertex & gsvertex)
 __forceinline void Kick::Output_Vertex(VertexGPU vert, u32 id)
 {
 #ifdef WRITE_PRIM_LOGS
-	ZZLog::Prim_Log("%c%d(%d): xyzf=(%4d,%4d,0x%x,%3d), rgba=0x%8.8x, stq = (%2.5f,%2.5f,%2.5f)\n",
+	ZZLog::Prim_Log("%c%d(%d): xyzf=(%4d,%4d,0x%x,%3d), rgba=0x%8.8x, stq = (%2.5f,%2.5f,%2.5f)",
 					id == 0 ? '*' : ' ', id, prim->prim, vert.x / 8, vert.y / 8, vert.z, vert.f / 128,
 					vert.rgba, Clamp(vert.s, -10, 10), Clamp(vert.t, -10, 10), Clamp(vert.q, -10, 10));
 #endif
@@ -140,35 +133,70 @@ void Kick::DrawPrim(u32 prim_type)
     u32 last;
     switch(prim_type) {
         case PRIM_POINT:
-            assert(gs.primC >= 1);
-
             Set_Vertex<true>(&p[0], gs.gsvertex[gs.primIndex]);
             curvb.nCount ++;
             break;
 
         case PRIM_LINE:
-        case PRIM_LINE_STRIP:
-            assert(gs.primC >= 2);
-
             Set_Vertex<true>(&p[0], gs.gsvertex[gs.primPrev()]);
             Set_Vertex<true>(&p[1], gs.gsvertex[gs.primIndex]);
             curvb.nCount += 2;
             break;
 
-        case PRIM_TRIANGLE:
+        case PRIM_LINE_STRIP:
-        case PRIM_TRIANGLE_STRIP:
+            if (likely(ValidPrevPrim)) {
-        case PRIM_TRIANGLE_FAN:
+                assert(curvb.nCount >= 1);
-            assert(gs.primC >= 3);
+                p[0] = p[-1];
+            } else {
+                Set_Vertex<true>(&p[0], gs.gsvertex[gs.primPrev()]);
+                ValidPrevPrim = true;
+            }
 
-            Set_Vertex<true>(&p[0], gs.gsvertex[0]);
+            Set_Vertex<true>(&p[1], gs.gsvertex[gs.primIndex]);
-            Set_Vertex<true>(&p[1], gs.gsvertex[1]);
+            curvb.nCount += 2;
-            Set_Vertex<true>(&p[2], gs.gsvertex[2]);
+            break;
+
+        case PRIM_TRIANGLE:
+            Set_Vertex<true>(&p[0], gs.gsvertex[gs.primPrev(2)]);
+            Set_Vertex<true>(&p[1], gs.gsvertex[gs.primPrev()]);
+            Set_Vertex<true>(&p[2], gs.gsvertex[gs.primIndex]);
+            curvb.nCount += 3;
+            break;
+
+        case PRIM_TRIANGLE_STRIP:
+            if (likely(ValidPrevPrim)) {
+                assert(curvb.nCount >= 2);
+                p[0] = p[-2];
+                p[1] = p[-1];
+            } else {
+                Set_Vertex<true>(&p[0], gs.gsvertex[gs.primPrev(2)]);
+                Set_Vertex<true>(&p[1], gs.gsvertex[gs.primPrev()]);
+                ValidPrevPrim = true;
+            }
+
+            Set_Vertex<true>(&p[2], gs.gsvertex[gs.primIndex]);
+            curvb.nCount += 3;
+            break;
+
+        case PRIM_TRIANGLE_FAN:
+            if (likely(ValidPrevPrim)) {
+                assert(curvb.nCount >= 2);
+                VertexGPU* TriFanVert = curvb.pBufferData + gs.nTriFanVert;
+                p[0] = TriFanVert[0];
+                p[1] = p[-1];
+            } else {
+                Set_Vertex<true>(&p[0], gs.gsTriFanVertex);
+                Set_Vertex<true>(&p[1], gs.gsvertex[gs.primPrev(1)]);
+                ValidPrevPrim = true;
+                // Remenber the base for future processing
+                gs.nTriFanVert = curvb.nCount;
+            }
+
+            Set_Vertex<true>(&p[2], gs.gsvertex[gs.primIndex]);
             curvb.nCount += 3;
             break;
 
         case PRIM_SPRITE:
-            assert(gs.primC >= 2);
-
             prev = gs.primPrev();
             last = gs.primIndex;
 
@@ -206,17 +234,20 @@ void Kick::DrawPrim(u32 prim_type)
         default: break;
     }
 
-    // Print DEBUG info
+    // Print DEBUG info and code assertion
     switch(prim_type) {
         case PRIM_TRIANGLE:
         case PRIM_TRIANGLE_STRIP:
         case PRIM_TRIANGLE_FAN:
+            assert(gs.primC >= 3);
             Output_Vertex(p[2],2);
         case PRIM_LINE:
         case PRIM_LINE_STRIP:
         case PRIM_SPRITE:
+            assert(gs.primC >= 2);
             Output_Vertex(p[1],1);
         case PRIM_POINT:
+            assert(gs.primC >= 1);
             Output_Vertex(p[0],0);
         default: break;
     }

plugins/zzogl-pg/opengl/ZZKick.h

@@ -42,12 +42,17 @@ class Kick
         // template<bool DO_Z_FOG> void Set_Vertex(VertexGPU *p, int i);
         template<bool DO_Z_FOG> void Set_Vertex(VertexGPU *p, Vertex &gsvertex);
 		void Output_Vertex(VertexGPU vert, u32 id);
+        bool ValidPrevPrim;
 	public:
 		Kick() { }
 		~Kick() { }
 		
 		void KickVertex(bool adc);
 		void DrawPrim(u32 i);
+
+        inline void DirtyValidPrevPrim() {
+            ValidPrevPrim = 0;
+        }
 };
 extern Kick* ZZKick;
 

plugins/zzogl-pg/opengl/targets.h

@@ -151,7 +151,7 @@ class CMemoryTarget
 			int ref;
 		};
 
-		inline CMemoryTarget() : ptex(NULL), starty(0), height(0), realy(0), realheight(0), usedstamp(0), psm(0), cpsm(0), channels(0), clearminy(0), clearmaxy(0), validatecount(0), clutsize(0), clut(NULL) {}
+		inline CMemoryTarget() : ptex(NULL), starty(0), height(0), realy(0), realheight(0), usedstamp(0), psm(0), cpsm(0), channels(0), clearminy(0), clearmaxy(0), validatecount(0), clut(NULL), clutsize(0) {}
 
 		inline CMemoryTarget(const CMemoryTarget& r)
 		{