Latte: Instance count minimum is 1

Previously we skipped drawcalls when the instance count was set to zero. But the hardware register enforces a minimum of 1.

Fixes black screen in "Cubit The Hardcore Platformer Robot" which does all it's drawcalls with an hardcoded instance count of 0.

src/Cafe/HW/Latte/Core/Latte.h

@@ -27,11 +27,6 @@ struct LatteGPUState_t
 	uint32 contextControl1;
 	// optional features
 	bool allowFramebufferSizeOptimization{false}; // allow using scissor box as size hint to determine non-padded rendertarget size
-	// draw context
-	struct  
-	{
-		uint32 numInstances;
-	}drawContext;
 	// stats
 	uint32 frameCounter;
 	uint32 flipCounter; // increased by one everytime a vsync + flip happens

src/Cafe/HW/Latte/Core/LatteCommandProcessor.cpp

@@ -62,16 +62,7 @@ public:
 	{
 		uint32 baseVertex = LatteGPUState.contextRegister[mmSQ_VTX_BASE_VTX_LOC];
 		uint32 baseInstance = LatteGPUState.contextRegister[mmSQ_VTX_START_INST_LOC];
-		uint32 numInstances = LatteGPUState.drawContext.numInstances;
+		uint32 numInstances = LatteGPUState.contextNew.VGT_DMA_NUM_INSTANCES.get_NUM_INSTANCES();
-		if (numInstances == 0)
-			return;
-
-		/*
-		if (GetAsyncKeyState('B'))
-		{
-			cemuLog_force("[executeDraw] {} Count {} BaseVertex {} BaseInstance {}", m_isFirstDraw?"Init":"Fast", count, baseVertex, baseInstance);
-		}
-		*/
 
 		if (!isAutoIndex)
 		{
@@ -391,7 +382,10 @@ LatteCMDPtr LatteCP_itIndexType(LatteCMDPtr cmd, uint32 nWords)
 LatteCMDPtr LatteCP_itNumInstances(LatteCMDPtr cmd, uint32 nWords)
 {
 	cemu_assert_debug(nWords == 1);
-	LatteGPUState.drawContext.numInstances = LatteReadCMD();
+	uint32 numInstances = LatteReadCMD();
+	if (numInstances == 0)
+		numInstances = 1;
+	LatteGPUState.contextNew.VGT_DMA_NUM_INSTANCES.set_NUM_INSTANCES(numInstances);
 	return cmd;
 }
 
@@ -690,9 +684,6 @@ LatteCMDPtr LatteCP_itDrawIndexAuto(LatteCMDPtr cmd, uint32 nWords, DrawPassCont
 	cemu_assert_debug(nWords == 2);
 	uint32 count = LatteReadCMD();
 	uint32 ukn = LatteReadCMD();
-
-	if (LatteGPUState.drawContext.numInstances == 0)
-		return cmd;
 	LatteGPUState.currentDrawCallTick = GetTickCount();
 	// todo - better way to identify compute drawcalls
 	if ((LatteGPUState.contextRegister[mmSQ_CONFIG] >> 24) == 0xE4)
@@ -750,13 +741,7 @@ LatteCMDPtr LatteCP_itDrawImmediate(LatteCMDPtr cmd, uint32 nWords, DrawPassCont
 		cemuLog_log(LogType::Force, "itDrawImmediate - Unsupported index type");
 		return cmd;
 	}
-	// verify packet size
+	cemu_assert_debug(nWords == (2 + numIndexU32s)); // verify packet size
-	if (nWords != (2 + numIndexU32s))
-		debugBreakpoint();
-
-	uint32 baseVertex = LatteGPUState.contextRegister[mmSQ_VTX_BASE_VTX_LOC];
-	uint32 baseInstance = LatteGPUState.contextRegister[mmSQ_VTX_START_INST_LOC];
-	uint32 numInstances = LatteGPUState.drawContext.numInstances;
 
 	drawPassCtx.executeDraw(count, false, _tempIndexArrayMPTR);
 	return cmd;

src/Cafe/HW/Latte/Core/LatteThread.cpp

@@ -30,6 +30,7 @@ void Latte_LoadInitialRegisters()
 {
 	LatteGPUState.contextNew.CB_TARGET_MASK.set_MASK(0xFFFFFFFF);
 	LatteGPUState.contextNew.VGT_MULTI_PRIM_IB_RESET_INDX.set_RESTART_INDEX(0xFFFFFFFF);
+	LatteGPUState.contextNew.VGT_DMA_NUM_INSTANCES.set_NUM_INSTANCES(1);
 	LatteGPUState.contextRegister[Latte::REGADDR::PA_CL_CLIP_CNTL] = 0;
 	*(float*)&LatteGPUState.contextRegister[mmDB_DEPTH_CLEAR] = 1.0f;
 }

src/Cafe/HW/Latte/ISA/LatteReg.h

@@ -445,6 +445,7 @@ namespace Latte
 		VGT_DMA_INDEX_TYPE					= 0xA29F, // todo - verify offset
 
 		VGT_PRIMITIVEID_EN					= 0xA2A1,
+		VGT_DMA_NUM_INSTANCES				= 0xA2A2,
 
 		VGT_MULTI_PRIM_IB_RESET_EN 			= 0xA2A5,
 
@@ -977,6 +978,11 @@ float get_##__regname() const \
 		LATTE_BITFIELD_BOOL(PRIMITIVEID_EN, 0);
 	};
 
+	struct LATTE_VGT_DMA_NUM_INSTANCES : LATTEREG // 0xA2A2
+	{
+		LATTE_BITFIELD_FULL_TYPED(NUM_INSTANCES, uint32);
+	};
+
 	struct LATTE_VGT_MULTI_PRIM_IB_RESET_EN : LATTEREG // 0xA2A5
 	{
 		LATTE_BITFIELD_BOOL(RESET_EN, 0);
@@ -1541,7 +1547,7 @@ struct LatteContextRegister
 	/* +0x28A7C */ Latte::LATTE_VGT_DMA_INDEX_TYPE VGT_DMA_INDEX_TYPE;
 	/* +0x28A80 */ uint32 ukn28A80;
 	/* +0x28A84 */ Latte::LATTE_VGT_PRIMITIVEID_EN VGT_PRIMITIVEID_EN;
-	/* +0x28A88 */ uint32 ukn28A88;
+	/* +0x28A88 */ Latte::LATTE_VGT_DMA_NUM_INSTANCES VGT_DMA_NUM_INSTANCES;
 	/* +0x28A8C */ uint32 ukn28A8C;
 	/* +0x28A90 */ uint32 ukn28A90;
 	/* +0x28A94 */ Latte::LATTE_VGT_MULTI_PRIM_IB_RESET_EN VGT_MULTI_PRIM_IB_RESET_EN;
@@ -1611,6 +1617,7 @@ static_assert(offsetof(LatteContextRegister, PA_SC_GENERIC_SCISSOR_TL) == Latte:
 static_assert(offsetof(LatteContextRegister, PA_SC_GENERIC_SCISSOR_BR) == Latte::REGADDR::PA_SC_GENERIC_SCISSOR_BR * 4);
 static_assert(offsetof(LatteContextRegister, VGT_MULTI_PRIM_IB_RESET_INDX) == Latte::REGADDR::VGT_MULTI_PRIM_IB_RESET_INDX * 4);
 static_assert(offsetof(LatteContextRegister, VGT_PRIMITIVEID_EN) == Latte::REGADDR::VGT_PRIMITIVEID_EN * 4);
+static_assert(offsetof(LatteContextRegister, VGT_DMA_NUM_INSTANCES) == Latte::REGADDR::VGT_DMA_NUM_INSTANCES * 4);
 static_assert(offsetof(LatteContextRegister, VGT_MULTI_PRIM_IB_RESET_EN) == Latte::REGADDR::VGT_MULTI_PRIM_IB_RESET_EN * 4);
 static_assert(offsetof(LatteContextRegister, VGT_INSTANCE_STEP_RATE_0) == Latte::REGADDR::VGT_INSTANCE_STEP_RATE_0 * 4);
 static_assert(offsetof(LatteContextRegister, VGT_INSTANCE_STEP_RATE_1) == Latte::REGADDR::VGT_INSTANCE_STEP_RATE_1 * 4);

src/gui/wxgui/components/wxGameList.cpp

@@ -1331,7 +1331,6 @@ void wxGameList::AsyncWorkerThread()
 		else
 		{
 			cemuLog_log(LogType::Force, "Failed to load icon for title {:016x}", titleId);
-			cemu_assert_debug(false);
 		}
 		titleInfo.Unmount(tempMountPath);
 		// notify UI about loaded icon