Merge pull request #19563 from hrydgard/even-more-misc-fixes

Vulkan: Fix potential crash from binding old CLUT textures

Author: hrydgard

Common/Data/Encoding/Shiftjis.h

@@ -20,6 +20,7 @@ struct ShiftJIS {
 				return INVALID;
 			}
 			// Intentional fall-through.
+			[[fallthrough]];
 		case 0x9:
 		case 0xE:
 			row = ((j & 0x3F) << 1) - 0x01;

Common/Log/LogManager.cpp

@@ -363,14 +363,15 @@ void RingbufferLogListener::Log(const LogMessage &message) {
 #ifdef _WIN32
 
 void OutputDebugStringUTF8(const char *p) {
-	wchar_t temp[16384*4];
+	wchar_t *temp = new wchar_t[65536];
 
 	int len = std::min(16383*4, (int)strlen(p));
 	int size = (int)MultiByteToWideChar(CP_UTF8, 0, p, len, NULL, 0);
 	MultiByteToWideChar(CP_UTF8, 0, p, len, temp, size);
 	temp[size] = 0;
 
 	OutputDebugString(temp);
+	delete[] temp;
 }
 
 #else

Common/UI/Tween.h

@@ -1,4 +1,4 @@
-#pragma once
+#pragma once
 
 #include <algorithm>
 #include <cstdint>
@@ -65,7 +65,7 @@ template <typename Value>
 class TweenBase: public Tween {
 public:
 	TweenBase(float duration, float (*curve)(float) = [](float f) { return f; })
-		: Tween(duration, curve) {
+		: Tween(duration, curve), from_{}, to_{} {
 	}
 	TweenBase(Value from, Value to, float duration, float (*curve)(float) = [](float f) { return f; })
 		: Tween(duration, curve), from_(from), to_(to) {
@@ -77,8 +77,9 @@ class TweenBase: public Tween {
 	void Divert(const Value &newTo, float newDuration = -1.0f) {
 		const Value newFrom = valid_ ? Current(Position()) : newTo;
 
+		double now = time_now_d();
 		// Are we already part way through another transition?
-		if (time_now_d() < start_ + delay_ + duration_ && valid_) {
+		if (now < start_ + delay_ + duration_ && valid_) {
 			if (newTo == to_) {
 				// Already on course.  Don't change.
 				return;
@@ -88,17 +89,17 @@ class TweenBase: public Tween {
 				if (newDuration >= 0.0f) {
 					newOffset *= newDuration / duration_;
 				}
-				start_ = time_now_d() - newOffset - delay_;
-			} else if (time_now_d() <= start_ + delay_) {
+				start_ = now - newOffset - delay_;
+			} else if (now <= start_ + delay_) {
 				// Start the delay over again.
-				start_ = time_now_d();
+				start_ = now;
 			} else {
 				// Since we've partially animated to the other value, skip delay.
-				start_ = time_now_d() - delay_;
+				start_ = now - delay_;
 			}
 		} else {
 			// Already finished, so restart.
-			start_ = time_now_d();
+			start_ = now;
 			finishApplied_ = false;
 		}
 

Core/Dialog/SavedataParam.cpp

@@ -883,7 +883,7 @@ std::set<std::string> SavedataParam::GetSecureFileNames(const std::string &dirPa
 
 	std::set<std::string> secureFileNames;
 	for (const auto &entry : entries) {
-		char temp[14];
+		char temp[14]{};
 		truncate_cpy(temp, entry.filename);
 		secureFileNames.insert(temp);
 	}

Core/HLE/sceFont.cpp

@@ -1208,7 +1208,7 @@ static int sceFontFindOptimumFont(u32 libHandle, u32 fontStylePtr, u32 errorCode
 
 	if (PSP_CoreParameter().compat.flags().Fontltn12Hack && requestedStyle->fontLanguage == 2) {
 		for (size_t j = 0; j < internalFonts.size(); j++) {
-			const auto tempmatchStyle = internalFonts[j]->GetFontStyle();
+			const auto &tempmatchStyle = internalFonts[j]->GetFontStyle();
 			const std::string str(tempmatchStyle.fontFileName);
 			if (str == "ltn12.pgf") {
 				optimumFont = internalFonts[j];

Core/MIPS/IR/IRInst.h

@@ -379,7 +379,7 @@ class IRWriter {
 		insts_ = w.insts_;
 		return *this;
 	}
-	IRWriter &operator =(IRWriter &&w) {
+	IRWriter &operator =(IRWriter &&w) noexcept {
 		insts_ = std::move(w.insts_);
 		return *this;
 	}

Core/MIPS/IR/IRJit.h

@@ -51,7 +51,7 @@ class IRBlock {
 	IRBlock() {}
 	IRBlock(u32 emAddr, u32 origSize, int instOffset, u32 numInstructions)
 		: origAddr_(emAddr), origSize_(origSize), arenaOffset_(instOffset), numIRInstructions_(numInstructions) {}
-	IRBlock(IRBlock &&b) {
+	IRBlock(IRBlock &&b) noexcept {
 		arenaOffset_ = b.arenaOffset_;
 		hash_ = b.hash_;
 		origAddr_ = b.origAddr_;

Core/MIPS/MIPS.cpp

@@ -207,6 +207,8 @@ void MIPSState::Init() {
 	nextPC = 0;
 	downcount = 0;
 
+	memset(vcmpResult, 0, sizeof(vcmpResult));
+
 	std::lock_guard<std::recursive_mutex> guard(MIPSComp::jitLock);
 	if (PSP_CoreParameter().cpuCore == CPUCore::JIT || PSP_CoreParameter().cpuCore == CPUCore::JIT_IR) {
 		MIPSComp::jit = MIPSComp::CreateNativeJit(this, PSP_CoreParameter().cpuCore == CPUCore::JIT_IR);

Core/MIPS/MIPSIntVFPU.cpp

@@ -961,9 +961,11 @@ namespace MIPSInt
 			case V_Triple:
 				sz = V_Pair;
 				// Intentional fallthrough.
+				[[fallthrough]];
 			case V_Pair:
 				oz = V_Quad;
 				// Intentional fallthrough.
+				[[fallthrough]];
 			case V_Single:
 				for (int i = 0; i < GetNumVectorElements(sz); i++) {
 					u32 value = s[i];
@@ -985,9 +987,11 @@ namespace MIPSInt
 			case V_Triple:
 				sz = V_Pair;
 				// Intentional fallthrough.
+				[[fallthrough]];
 			case V_Pair:
 				oz = V_Quad;
 				// Intentional fallthrough.
+				[[fallthrough]];
 			case V_Single:
 				for (int i = 0; i < GetNumVectorElements(sz); i++) {
 					u32 value = s[i];

Core/Util/BlockAllocator.cpp

@@ -27,10 +27,6 @@
 
 // Slow freaking thing but works (eventually) :)
 
-BlockAllocator::BlockAllocator(int grain) : bottom_(NULL), top_(NULL), grain_(grain)
-{
-}
-
 BlockAllocator::~BlockAllocator()
 {
 	Shutdown();

Core/Util/BlockAllocator.h

@@ -24,7 +24,7 @@ class PointerWrap;
 class BlockAllocator
 {
 public:
-	BlockAllocator(int grain = 16);  // 16 byte granularity by default.
+	BlockAllocator(int grain = 16) : grain_(grain) {}  // 16 byte granularity by default.
 	~BlockAllocator();
 
 	void Init(u32 _rangeStart, u32 _rangeSize, bool suballoc);
@@ -59,8 +59,7 @@ class BlockAllocator
 private:
 	void CheckBlocks() const;
 
-	struct Block
-	{
+	struct Block {
 		Block(u32 _start, u32 _size, bool _taken, Block *_prev, Block *_next);
 		void SetAllocated(const char *_tag, bool suballoc);
 		void DoState(PointerWrap &p);
@@ -72,13 +71,13 @@ class BlockAllocator
 		Block *next;
 	};
 
-	Block *bottom_;
-	Block *top_;
-	u32 rangeStart_;
-	u32 rangeSize_;
+	Block *bottom_ = nullptr;
+	Block *top_ = nullptr;
+	u32 rangeStart_ = 0;
+	u32 rangeSize_ = 0;
 
 	u32 grain_;
-	bool suballoc_;
+	bool suballoc_ = false;
 
 	void MergeFreeBlocks(Block *fromBlock);
 	Block *GetBlockFromAddress(u32 addr);

Core/Util/DisArm64.cpp

@@ -418,7 +418,7 @@ static void DataProcessingRegister(uint32_t w, uint64_t addr, Instruction *instr
 		int opcode2 = (w >> 16) & 0x1F;
 		int opcode = (w >> 10) & 0x3F;
 		// Data-processing (1 source)
-		const char *opname[8] = { "rbit", "rev16", "rev32", "(unk)", "clz", "cls" };
+		const char *opname[64] = { "rbit", "rev16", "rev32", "(unk)", "clz", "cls" };
 		const char *op = opcode2 >= 8 ? "unk" : opname[opcode];
 		snprintf(instr->text, sizeof(instr->text), "%s %c%d, %c%d", op, r, Rd, r, Rn);
 	} else if (((w >> 21) & 0x2FF) == 0x0D6) {
@@ -749,7 +749,7 @@ static void FPandASIMD1(uint32_t w, uint64_t addr, Instruction *instr) {
 				int dst_index = imm5 >> (size + 1);
 				int src_index = imm4 >> size;
 				int op = (w >> 29) & 1;
-				char s;
+				char s = '_';
 				switch (size) {
 				case 0x00: s = 'b'; break;
 				case 0x01: s = 'h'; break;

Core/Util/GameDB.h

@@ -12,8 +12,8 @@ class VFSInterface;
 struct GameDBInfo {
 	std::string title;
 	std::string foreignTitle;
-	uint32_t crc;
-	uint64_t size;
+	uint32_t crc = 0;
+	uint64_t size = 0;
 };
 
 class GameDB {

GPU/Common/DrawEngineCommon.cpp

@@ -44,6 +44,7 @@ DrawEngineCommon::DrawEngineCommon() : decoderMap_(16) {
 	transformedExpanded_ = (TransformedVertex *)AllocateMemoryPages(3 * TRANSFORMED_VERTEX_BUFFER_SIZE, MEM_PROT_READ | MEM_PROT_WRITE);
 	decoded_ = (u8 *)AllocateMemoryPages(DECODED_VERTEX_BUFFER_SIZE, MEM_PROT_READ | MEM_PROT_WRITE);
 	decIndex_ = (u16 *)AllocateMemoryPages(DECODED_INDEX_BUFFER_SIZE, MEM_PROT_READ | MEM_PROT_WRITE);
+	indexGen.Setup(decIndex_);
 }
 
 DrawEngineCommon::~DrawEngineCommon() {

GPU/Common/FramebufferManagerCommon.cpp

@@ -1911,8 +1911,8 @@ void FramebufferManagerCommon::ResizeFramebufFBO(VirtualFramebuffer *vfb, int w,
 
 struct CopyCandidate {
 	VirtualFramebuffer *vfb = nullptr;
-	int y;
-	int h;
+	int y = 0;
+	int h = 0;
 
 	std::string ToString(RasterChannel channel) const {
 		return StringFromFormat("%08x %s %dx%d y=%d h=%d", vfb->Address(channel), GeBufferFormatToString(vfb->Format(channel)), vfb->width, vfb->height, y, h);

GPU/Common/GPUStateUtils.cpp

@@ -91,7 +91,7 @@ bool IsAlphaTestTriviallyTrue() {
 			return false;
 		}
 		// Fallthrough on purpose
-
+		[[fallthrough]];
 	case GE_COMP_GREATER:
 	{
 		// If the texture and vertex only use 1.0 alpha, then the ref value doesn't matter.

GPU/Common/IndexGenerator.cpp

@@ -48,11 +48,6 @@ const u8 IndexGenerator::indexedPrimitiveType[7] = {
 	GE_PRIM_RECTANGLES,
 };
 
-void IndexGenerator::Setup(u16 *inds) {
-	this->indsBase_ = inds;
-	Reset();
-}
-
 void IndexGenerator::AddPrim(int prim, int vertexCount, int indexOffset, bool clockwise) {
 	switch (prim) {
 	case GE_PRIM_POINTS: AddPoints(vertexCount, indexOffset); break;

GPU/Common/IndexGenerator.h

@@ -24,9 +24,12 @@
 
 class IndexGenerator {
 public:
-	void Setup(u16 *indexptr);
+	void Setup(u16 *indexptr) {
+		indsBase_ = indexptr;
+		inds_ = indexptr;
+	}
 	void Reset() {
-		this->inds_ = indsBase_;
+		inds_ = indsBase_;
 	}
 
 	static bool PrimCompatible(int prim1, int prim2) {

GPU/Common/TextureShaderCommon.cpp

@@ -194,6 +194,7 @@ void TextureShaderCache::Decimate() {
 			++tex;
 		}
 	}
+	gpuStats.numClutTextures = (int)texCache_.size();
 }
 
 Draw2DPipeline *TextureShaderCache::GetDepalettizeShader(uint32_t clutMode, GETextureFormat textureFormat, GEBufferFormat bufferFormat, bool smoothedDepal, u32 depthUpperBits) {

GPU/Common/TextureShaderCommon.h

@@ -27,7 +27,6 @@
 #include "GPU/Common/Draw2D.h"
 #include "GPU/Common/ShaderCommon.h"
 
-
 class ClutTexture {
 public:
 	enum { MAX_RAMPS = 3 };

GPU/D3D11/DrawEngineD3D11.cpp

@@ -80,16 +80,7 @@ DrawEngineD3D11::DrawEngineD3D11(Draw::DrawContext *draw, ID3D11Device *device,
 	decOptions_.expandAllWeightsToFloat = true;
 	decOptions_.expand8BitNormalsToFloat = true;
 
-	// Allocate nicely aligned memory. Maybe graphics drivers will
-	// appreciate it.
-	// All this is a LOT of memory, need to see if we can cut down somehow.
-	indexGen.Setup(decIndex_);
-
 	InitDeviceObjects();
-
-	// Vertex pushing buffers. For uniforms we use short DISCARD buffers, but we could use
-	// this kind of buffer there as well with D3D11.1. We might be able to use the same buffer
-	// for both vertices and indices, and possibly all three data types.
 }
 
 DrawEngineD3D11::~DrawEngineD3D11() {

GPU/Directx9/DrawEngineDX9.cpp

@@ -89,8 +89,6 @@ DrawEngineDX9::DrawEngineDX9(Draw::DrawContext *draw) : draw_(draw), vertexDeclM
 	decOptions_.expandAllWeightsToFloat = true;
 	decOptions_.expand8BitNormalsToFloat = true;
 
-	indexGen.Setup(decIndex_);
-
 	InitDeviceObjects();
 
 	tessDataTransferDX9 = new TessellationDataTransferDX9();

GPU/GLES/DrawEngineGLES.cpp

@@ -67,8 +67,6 @@ DrawEngineGLES::DrawEngineGLES(Draw::DrawContext *draw) : inputLayoutMap_(16), d
 	decOptions_.expandAllWeightsToFloat = false;
 	decOptions_.expand8BitNormalsToFloat = false;
 
-	indexGen.Setup(decIndex_);
-
 	InitDeviceObjects();
 
 	tessDataTransferGLES = new TessellationDataTransferGLES(render_);

GPU/GPU.h

@@ -105,6 +105,7 @@ struct GPUStatistics {
 		numBlockTransfers = 0;
 		numReplacerTrackedTex = 0;
 		numCachedReplacedTextures = 0;
+		numClutTextures = 0;
 		msProcessingDisplayLists = 0;
 		vertexGPUCycles = 0;
 		otherGPUCycles = 0;
@@ -142,6 +143,7 @@ struct GPUStatistics {
 	int numBlockTransfers;
 	int numReplacerTrackedTex;
 	int numCachedReplacedTextures;
+	int numClutTextures;
 	double msProcessingDisplayLists;
 	int vertexGPUCycles;
 	int otherGPUCycles;

GPU/GPUCommonHW.cpp

@@ -1756,7 +1756,7 @@ size_t GPUCommonHW::FormatGPUStatsCommon(char *buffer, size_t size) {
 		"Draw: %d (%d dec, %d culled), flushes %d, clears %d, bbox jumps %d (%d updates)\n"
 		"Vertices: %d dec: %d drawn: %d\n"
 		"FBOs active: %d (evaluations: %d)\n"
-		"Textures: %d, dec: %d, invalidated: %d, hashed: %d kB\n"
+		"Textures: %d, dec: %d, invalidated: %d, hashed: %d kB, clut %d\n"
 		"readbacks %d (%d non-block), upload %d (cached %d), depal %d\n"
 		"block transfers: %d\n"
 		"replacer: tracks %d references, %d unique textures\n"
@@ -1781,6 +1781,7 @@ size_t GPUCommonHW::FormatGPUStatsCommon(char *buffer, size_t size) {
 		gpuStats.numTexturesDecoded,
 		gpuStats.numTextureInvalidations,
 		gpuStats.numTextureDataBytesHashed / 1024,
+		gpuStats.numClutTextures,
 		gpuStats.numBlockingReadbacks,
 		gpuStats.numReadbacks,
 		gpuStats.numUploads,

GPU/Vulkan/DrawEngineVulkan.cpp

@@ -62,7 +62,6 @@ DrawEngineVulkan::DrawEngineVulkan(Draw::DrawContext *draw)
 	: draw_(draw) {
 	decOptions_.expandAllWeightsToFloat = false;
 	decOptions_.expand8BitNormalsToFloat = false;
-	indexGen.Setup(decIndex_);
 }
 
 void DrawEngineVulkan::InitDeviceObjects() {
@@ -167,6 +166,10 @@ void DrawEngineVulkan::DeviceRestore(Draw::DrawContext *draw) {
 void DrawEngineVulkan::BeginFrame() {
 	lastPipeline_ = nullptr;
 
+	// These will be re-bound if needed, let's not let old bindings linger around too long.
+	boundDepal_ = VK_NULL_HANDLE;
+	boundSecondary_ = VK_NULL_HANDLE;
+
 	// pushUBO is the thin3d push pool, don't need to BeginFrame again.
 	pushVertex_->BeginFrame();
 	pushIndex_->BeginFrame();

ext/riscv-disas.cpp

@@ -2233,7 +2233,7 @@ static void decode_inst_operands(rv_decode *dec)
 
 static void decode_inst_decompress(rv_decode *dec, rv_isa isa)
 {
-    int decomp_op;
+    int decomp_op = 0;
     switch (isa) {
     case rv32: decomp_op = opcode_data[dec->op].decomp_rv32; break;
     case rv64: decomp_op = opcode_data[dec->op].decomp_rv64; break;