Further centralization of EstimatePerVertexCost, now moved to GPUCommon

GPU/Common/DrawEngineCommon.h

@@ -58,35 +58,6 @@ public:
 	std::vector<std::string> DebugGetVertexLoaderIDs();
 	std::string DebugGetVertexLoaderString(std::string id, DebugShaderStringType stringType);
 
-	int EstimatePerVertexCost() {
-		// TODO: This is transform cost, also account for rasterization cost somehow... although it probably
-		// runs in parallel with transform.
-
-		// Also, this is all pure guesswork. If we can find a way to do measurements, that would be great.
-
-		// GTA wants a low value to run smooth, GoW wants a high value (otherwise it thinks things
-		// went too fast and starts doing all the work over again).
-
-		int cost = 20;
-		if (gstate.isLightingEnabled()) {
-			cost += 10;
-
-			for (int i = 0; i < 4; i++) {
-				if (gstate.isLightChanEnabled(i))
-					cost += 10;
-			}
-		}
-
-		if (gstate.getUVGenMode() != GE_TEXMAP_TEXTURE_COORDS) {
-			cost += 20;
-		}
-		int morphCount = gstate.getNumMorphWeights();
-		if (morphCount > 1) {
-			cost += 5 * morphCount;
-		}
-		return cost;
-	}
-
 protected:
 	// Preprocessing for spline/bezier
 	u32 NormalizeVertices(u8 *outPtr, u8 *bufPtr, const u8 *inPtr, int lowerBound, int upperBound, u32 vertType);

GPU/Directx9/GPU_DX9.cpp

@@ -756,8 +756,7 @@ void GPU_DX9::Execute_Prim(u32 op, u32 diff) {
 	if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB)) {
 		drawEngine_.SetupVertexDecoder(gstate.vertType);
 		// Rough estimate, not sure what's correct.
-		int vertexCost = drawEngine_.EstimatePerVertexCost();
-		cyclesExecuted += vertexCost * count;
+		cyclesExecuted += EstimatePerVertexCost() * count;
 		return;
 	}
 
@@ -788,7 +787,7 @@ void GPU_DX9::Execute_Prim(u32 op, u32 diff) {
 	int bytesRead = 0;
 	drawEngine_.SubmitPrim(verts, inds, prim, count, vertexType, &bytesRead);
 
-	int vertexCost = drawEngine_.EstimatePerVertexCost() * count;
+	int vertexCost = EstimatePerVertexCost() * count;
 	gpuStats.vertexGPUCycles += vertexCost;
 	cyclesExecuted += vertexCost;
 

GPU/GLES/GPU_GLES.cpp

@@ -924,8 +924,7 @@ void GPU_GLES::Execute_Prim(u32 op, u32 diff) {
 	if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB))	{
 		drawEngine_.SetupVertexDecoder(gstate.vertType);
 		// Rough estimate, not sure what's correct.
-		int vertexCost = drawEngine_.EstimatePerVertexCost();
-		cyclesExecuted += vertexCost * count;
+		cyclesExecuted += EstimatePerVertexCost() * count;
 		return;
 	}
 
@@ -953,7 +952,7 @@ void GPU_GLES::Execute_Prim(u32 op, u32 diff) {
 	int bytesRead = 0;
 	drawEngine_.SubmitPrim(verts, inds, prim, count, gstate.vertType, &bytesRead);
 
-	int vertexCost = drawEngine_.EstimatePerVertexCost();
+	int vertexCost = EstimatePerVertexCost();
 	gpuStats.vertexGPUCycles += vertexCost * count;
 	cyclesExecuted += vertexCost * count;
 

GPU/GPUCommon.cpp

@@ -88,6 +88,35 @@ void GPUCommon::Reinitialize() {
 	ScheduleEvent(GPU_EVENT_REINITIALIZE);
 }
 
+int GPUCommon::EstimatePerVertexCost() {
+	// TODO: This is transform cost, also account for rasterization cost somehow... although it probably
+	// runs in parallel with transform.
+
+	// Also, this is all pure guesswork. If we can find a way to do measurements, that would be great.
+
+	// GTA wants a low value to run smooth, GoW wants a high value (otherwise it thinks things
+	// went too fast and starts doing all the work over again).
+
+	int cost = 20;
+	if (gstate.isLightingEnabled()) {
+		cost += 10;
+
+		for (int i = 0; i < 4; i++) {
+			if (gstate.isLightChanEnabled(i))
+				cost += 10;
+		}
+	}
+
+	if (gstate.getUVGenMode() != GE_TEXMAP_TEXTURE_COORDS) {
+		cost += 20;
+	}
+	int morphCount = gstate.getNumMorphWeights();
+	if (morphCount > 1) {
+		cost += 5 * morphCount;
+	}
+	return cost;
+}
+
 void GPUCommon::PopDLQueue() {
 	easy_guard guard(listLock);
 	if(!dlQueue.empty()) {

GPU/GPUCommon.h

@@ -96,6 +96,8 @@ public:
 	void Execute_BoneMtxNum(u32 op, u32 diff);
 	void Execute_BoneMtxData(u32 op, u32 diff);
 
+	int EstimatePerVertexCost();
+
 	// Note: Not virtual!
 	inline void Flush();
 

GPU/Software/SoftGpu.cpp

@@ -283,40 +283,12 @@ void SoftGPU::FastRunLoop(DisplayList &list) {
 	}
 }
 
-int EstimatePerVertexCost() {
-	// TODO: This is transform cost, also account for rasterization cost somehow... although it probably
-	// runs in parallel with transform.
-
-	// Also, this is all pure guesswork. If we can find a way to do measurements, that would be great.
-
-	// GTA wants a low value to run smooth, GoW wants a high value (otherwise it thinks things
-	// went too fast and starts doing all the work over again).
-
-	int cost = 20;
-	if (gstate.isLightingEnabled()) {
-		cost += 10;
-	}
-
-	for (int i = 0; i < 4; i++) {
-		if (gstate.isLightChanEnabled(i))
-			cost += 10;
-	}
-	if (gstate.getUVGenMode() != GE_TEXMAP_TEXTURE_COORDS) {
-		cost += 20;
-	}
-	// TODO: morphcount
-
-	return cost;
-}
-
-void SoftGPU::ExecuteOp(u32 op, u32 diff)
-{
+void SoftGPU::ExecuteOp(u32 op, u32 diff) {
 	u32 cmd = op >> 24;
 	u32 data = op & 0xFFFFFF;
 
 	// Handle control and drawing commands here directly. The others we delegate.
-	switch (cmd)
-	{
+	switch (cmd) {
 	case GE_CMD_BASE:
 		break;
 

GPU/Vulkan/GPU_Vulkan.cpp

@@ -784,8 +784,7 @@ void GPU_Vulkan::Execute_Prim(u32 op, u32 diff) {
 	if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB)) {
 		drawEngine_.SetupVertexDecoder(gstate.vertType);
 		// Rough estimate, not sure what's correct.
-		int vertexCost = drawEngine_.EstimatePerVertexCost();
-		cyclesExecuted += vertexCost * count;
+		cyclesExecuted += EstimatePerVertexCost() * count;
 		return;
 	}
 
@@ -813,7 +812,7 @@ void GPU_Vulkan::Execute_Prim(u32 op, u32 diff) {
 	int bytesRead = 0;
 	drawEngine_.SubmitPrim(verts, inds, prim, count, gstate.vertType, &bytesRead);
 
-	int vertexCost = drawEngine_.EstimatePerVertexCost();
+	int vertexCost = EstimatePerVertexCost();
 	gpuStats.vertexGPUCycles += vertexCost * count;
 	cyclesExecuted += vertexCost * count;