diff --git a/Common/MemoryUtil.cpp b/Common/MemoryUtil.cpp
index 36c7c74088..4fef8c6f3e 100644
--- a/Common/MemoryUtil.cpp
+++ b/Common/MemoryUtil.cpp
@@ -117,7 +117,6 @@ static void *SearchForFreeMem(size_t size) {
 
 // This is purposely not a full wrapper for virtualalloc/mmap, but it
 // provides exactly the primitive operations that PPSSPP needs.
-
 void *AllocateExecutableMemory(size_t size) {
 #if defined(_WIN32)
 	void *ptr = nullptr;
@@ -157,7 +156,7 @@ void *AllocateExecutableMemory(size_t size) {
 	}
 #else
 	static char *map_hint = 0;
-#if defined(_M_X64) && !defined(MAP_32BIT)
+#if defined(_M_X64)
 	// Try to request one that is close to our memory location if we're in high memory.
 	// We use a dummy global variable to give us a good location to start from.
 	if (!map_hint) {
@@ -165,9 +164,7 @@ void *AllocateExecutableMemory(size_t size) {
 			map_hint = (char*)round_page(&hint_location) - 0x20000000; // 0.5gb lower than our approximate location
 		else
 			map_hint = (char*)0x20000000; // 0.5GB mark in memory
-	}
-	else if ((uintptr_t) map_hint > 0xFFFFFFFFULL)
-	{
+	} else if ((uintptr_t) map_hint > 0xFFFFFFFFULL) {
 		map_hint -= round_page(size); /* round down to the next page if we're in high memory */
 	}
 #endif
@@ -176,13 +173,7 @@ void *AllocateExecutableMemory(size_t size) {
 	if (PlatformIsWXExclusive())
 		prot = PROT_READ | PROT_WRITE;  // POST_EXEC is added later in this case.
 
-	void* ptr = mmap(map_hint, size, prot,
-		MAP_ANON | MAP_PRIVATE
-#if defined(_M_X64) && defined(MAP_32BIT)
-		| MAP_32BIT
-#endif
-		, -1, 0);
-
+	void* ptr = mmap(map_hint, size, prot, MAP_ANON | MAP_PRIVATE, -1, 0);
 #endif /* defined(_WIN32) */
 
 #if !defined(_WIN32)
@@ -193,10 +184,11 @@ void *AllocateExecutableMemory(size_t size) {
 
 	if (ptr == failed_result) {
 		ptr = nullptr;
-		ERROR_LOG(MEMMAP, "Failed to allocate executable memory (%d)", (int)size);
+		ERROR_LOG(MEMMAP, "Failed to allocate executable memory (%d) errno=%d", (int)size, errno);
 		PanicAlert("Failed to allocate executable memory\n%s", GetLastErrorMsg());
 	}
-#if defined(_M_X64) && !defined(_WIN32) && !defined(MAP_32BIT)
+
+#if defined(_M_X64) && !defined(_WIN32)
 	else if ((uintptr_t)map_hint <= 0xFFFFFFFF) {
 		// Round up if we're below 32-bit mark, probably allocating sequentially.
 		map_hint += round_page(size);
diff --git a/Core/Compatibility.cpp b/Core/Compatibility.cpp
index 49c5527a08..d98611bcc8 100644
--- a/Core/Compatibility.cpp
+++ b/Core/Compatibility.cpp
@@ -68,7 +68,6 @@ void Compatibility::CheckSettings(IniFile &iniFile, const std::string &gameID) {
 	CheckSetting(iniFile, gameID, "DisableAccurateDepth", &flags_.DisableAccurateDepth);
 	CheckSetting(iniFile, gameID, "MGS2AcidHack", &flags_.MGS2AcidHack);
 	CheckSetting(iniFile, gameID, "SonicRivalsHack", &flags_.SonicRivalsHack);
-	CheckSetting(iniFile, gameID, "RenderPassMerge", &flags_.RenderPassMerge);
 	CheckSetting(iniFile, gameID, "AdhocHackDisableNBSwitch", &flags_.AdhocHackDisableNBSwitch);
 	CheckSetting(iniFile, gameID, "AdhocHackStopFriendClear", &flags_.AdhocHackStopFriendClear);
 	CheckSetting(iniFile, gameID, "AdhocHackPhantasyStarOffset", &flags_.AdhocHackPhantasyStarOffset);
diff --git a/Core/Compatibility.h b/Core/Compatibility.h
index a13ffddcda..0b9a3403f1 100644
--- a/Core/Compatibility.h
+++ b/Core/Compatibility.h
@@ -68,7 +68,6 @@ struct CompatFlags {
 	bool DisableAccurateDepth;
 	bool MGS2AcidHack;
 	bool SonicRivalsHack;
-	bool RenderPassMerge;
 	bool AdhocHackDisableNBSwitch;
 	bool AdhocHackStopFriendClear;
 	bool AdhocHackPhantasyStarOffset;
diff --git a/Core/Config.cpp b/Core/Config.cpp
index aac5b2fa72..4d09454d5f 100644
--- a/Core/Config.cpp
+++ b/Core/Config.cpp
@@ -870,6 +870,10 @@ static ConfigSetting controlSettings[] = {
 	ConfigSetting("TouchButtonHideSeconds", &g_Config.iTouchButtonHideSeconds, 20, true, true),
 	ConfigSetting("AutoCenterTouchAnalog", &g_Config.bAutoCenterTouchAnalog, false, true, true),
 
+	// Snap touch control position
+	ConfigSetting("TouchSnapToGrid", &g_Config.bTouchSnapToGrid, false, true, true),
+	ConfigSetting("TouchSnapGridSize", &g_Config.iTouchSnapGridSize, 64, true, true),
+
 	// -1.0f means uninitialized, set in GamepadEmu::CreatePadLayout().
 	ConfigSetting("ActionButtonSpacing2", &g_Config.fActionButtonSpacing, 1.0f, true, true),
 	ConfigSetting("ActionButtonCenterX", "ActionButtonCenterY", "ActionButtonScale", nullptr, &g_Config.touchActionButtonCenter, defaultTouchPosShow, true, true),
diff --git a/Core/Config.h b/Core/Config.h
index 92f66d3bad..e33eb7ad1d 100644
--- a/Core/Config.h
+++ b/Core/Config.h
@@ -285,6 +285,11 @@ public:
 	int iTouchButtonStyle;
 	int iTouchButtonOpacity;
 	int iTouchButtonHideSeconds;
+
+	// Snap touch control position
+	bool bTouchSnapToGrid;
+	int iTouchSnapGridSize;
+
 	// Floating analog stick (recenters on thumb on press).
 	bool bAutoCenterTouchAnalog;
 
diff --git a/Core/FileSystems/MetaFileSystem.cpp b/Core/FileSystems/MetaFileSystem.cpp
index 933f56f5b1..19bddf251e 100644
--- a/Core/FileSystems/MetaFileSystem.cpp
+++ b/Core/FileSystems/MetaFileSystem.cpp
@@ -226,7 +226,7 @@ int MetaFileSystem::MapFilePath(const std::string &_inpath, std::string &outpath
 		currentDirectory = &(it->second);
 	}
 
-	if ( RealPath(*currentDirectory, inpath, realpath) )
+	if (RealPath(*currentDirectory, inpath, realpath))
 	{
 		std::string prefix = realpath;
 		size_t prefixPos = realpath.find(':');
@@ -243,7 +243,7 @@ int MetaFileSystem::MapFilePath(const std::string &_inpath, std::string &outpath
 
 				VERBOSE_LOG(FILESYS, "MapFilePath: mapped \"%s\" to prefix: \"%s\", path: \"%s\"", inpath.c_str(), fileSystems[i].prefix.c_str(), outpath.c_str());
 
-				return 0;
+				return error == SCE_KERNEL_ERROR_NOCWD ? error : 0;
 			}
 		}
 	}
@@ -270,8 +270,7 @@ std::string MetaFileSystem::NormalizePrefix(std::string prefix) const {
 	return prefix;
 }
 
-void MetaFileSystem::Mount(std::string prefix, IFileSystem *system)
-{
+void MetaFileSystem::Mount(std::string prefix, IFileSystem *system) {
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	MountPoint x;
 	x.prefix = prefix;
@@ -279,8 +278,7 @@ void MetaFileSystem::Mount(std::string prefix, IFileSystem *system)
 	fileSystems.push_back(x);
 }
 
-void MetaFileSystem::Unmount(std::string prefix, IFileSystem *system)
-{
+void MetaFileSystem::Unmount(std::string prefix, IFileSystem *system) {
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	MountPoint x;
 	x.prefix = prefix;
@@ -288,12 +286,25 @@ void MetaFileSystem::Unmount(std::string prefix, IFileSystem *system)
 	fileSystems.erase(std::remove(fileSystems.begin(), fileSystems.end(), x), fileSystems.end());
 }
 
-void MetaFileSystem::Remount(IFileSystem *oldSystem, IFileSystem *newSystem) {
-	for (auto it = fileSystems.begin(); it != fileSystems.end(); ++it) {
-		if (it->system == oldSystem) {
-			it->system = newSystem;
+void MetaFileSystem::Remount(std::string prefix, IFileSystem *newSystem) {
+	std::lock_guard<std::recursive_mutex> guard(lock);
+	IFileSystem *oldSystem = nullptr;
+	for (auto &it : fileSystems) {
+		if (it.prefix == prefix) {
+			oldSystem = it.system;
+			it.system = newSystem;
 		}
 	}
+
+	bool delOldSystem = true;
+	for (auto &it : fileSystems) {
+		if (it.system == oldSystem) {
+			delOldSystem = false;
+		}
+	}
+
+	if (delOldSystem)
+		delete oldSystem;
 }
 
 IFileSystem *MetaFileSystem::GetSystemFromFilename(const std::string &filename) {
@@ -350,7 +361,8 @@ PSPFileInfo MetaFileSystem::GetFileInfo(std::string filename)
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	std::string of;
 	IFileSystem *system;
-	if (MapFilePath(filename, of, &system) == 0)
+	int error = MapFilePath(filename, of, &system);
+	if (error == 0)
 	{
 		return system->GetFileInfo(of);
 	}
@@ -366,7 +378,8 @@ bool MetaFileSystem::GetHostPath(const std::string &inpath, std::string &outpath
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	std::string of;
 	IFileSystem *system;
-	if (MapFilePath(inpath, of, &system) == 0) {
+	int error = MapFilePath(inpath, of, &system);
+	if (error == 0) {
 		return system->GetHostPath(of, outpath);
 	} else {
 		return false;
@@ -378,7 +391,8 @@ std::vector<PSPFileInfo> MetaFileSystem::GetDirListing(std::string path)
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	std::string of;
 	IFileSystem *system;
-	if (MapFilePath(path, of, &system) == 0)
+	int error = MapFilePath(path, of, &system);
+	if (error == 0)
 	{
 		return system->GetDirListing(of);
 	}
@@ -406,7 +420,8 @@ int MetaFileSystem::ChDir(const std::string &dir)
 	
 	std::string of;
 	MountPoint *mountPoint;
-	if (MapFilePath(dir, of, &mountPoint) == 0)
+	int error = MapFilePath(dir, of, &mountPoint);
+	if (error == 0)
 	{
 		currentDir[curThread] = mountPoint->prefix + of;
 		return 0;
@@ -435,7 +450,8 @@ bool MetaFileSystem::MkDir(const std::string &dirname)
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	std::string of;
 	IFileSystem *system;
-	if (MapFilePath(dirname, of, &system) == 0)
+	int error = MapFilePath(dirname, of, &system);
+	if (error == 0)
 	{
 		return system->MkDir(of);
 	}
@@ -450,7 +466,8 @@ bool MetaFileSystem::RmDir(const std::string &dirname)
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	std::string of;
 	IFileSystem *system;
-	if (MapFilePath(dirname, of, &system) == 0)
+	int error = MapFilePath(dirname, of, &system);
+	if (error == 0)
 	{
 		return system->RmDir(of);
 	}
@@ -467,12 +484,14 @@ int MetaFileSystem::RenameFile(const std::string &from, const std::string &to)
 	std::string rf;
 	IFileSystem *osystem;
 	IFileSystem *rsystem = NULL;
-	if (MapFilePath(from, of, &osystem) == 0)
+	int error = MapFilePath(from, of, &osystem);
+	if (error == 0)
 	{
 		// If it's a relative path, it seems to always use from's filesystem.
 		if (to.find(":/") != to.npos)
 		{
-			if (!MapFilePath(to, rf, &rsystem))
+			error = MapFilePath(to, rf, &rsystem);
+			if (error < 0)
 				return -1;
 		}
 		else
@@ -497,7 +516,8 @@ bool MetaFileSystem::RemoveFile(const std::string &filename)
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	std::string of;
 	IFileSystem *system;
-	if (MapFilePath(filename, of, &system) == 0)
+	int error = MapFilePath(filename, of, &system);
+	if (error == 0)
 	{
 		return system->RemoveFile(of);
 	}
@@ -604,7 +624,8 @@ u64 MetaFileSystem::FreeSpace(const std::string &path)
 	std::lock_guard<std::recursive_mutex> guard(lock);
 	std::string of;
 	IFileSystem *system;
-	if (MapFilePath(path, of, &system) == 0)
+	int error = MapFilePath(path, of, &system);
+	if (error == 0)
 		return system->FreeSpace(of);
 	else
 		return 0;
diff --git a/Core/FileSystems/MetaFileSystem.h b/Core/FileSystems/MetaFileSystem.h
index cfc202b0d1..c4bfd8d37f 100644
--- a/Core/FileSystems/MetaFileSystem.h
+++ b/Core/FileSystems/MetaFileSystem.h
@@ -52,7 +52,7 @@ public:
 
 	void Mount(std::string prefix, IFileSystem *system);
 	void Unmount(std::string prefix, IFileSystem *system);
-	void Remount(IFileSystem *oldSystem, IFileSystem *newSystem);
+	void Remount(std::string prefix, IFileSystem *newSystem);
 
 	IFileSystem *GetSystem(const std::string &prefix);
 	IFileSystem *GetSystemFromFilename(const std::string &filename);
diff --git a/Core/HLE/HLEHelperThread.cpp b/Core/HLE/HLEHelperThread.cpp
index 7f932c4a9c..36f9166813 100644
--- a/Core/HLE/HLEHelperThread.cpp
+++ b/Core/HLE/HLEHelperThread.cpp
@@ -16,9 +16,11 @@
 // https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
 
 #include "Common/ChunkFile.h"
+#include "Core/CoreTiming.h"
 #include "Core/MemMapHelpers.h"
 #include "Core/HLE/HLE.h"
 #include "Core/HLE/HLEHelperThread.h"
+#include "Core/HLE/KernelWaitHelpers.h"
 #include "Core/HLE/sceKernelThread.h"
 #include "Core/HLE/sceKernelMemory.h"
 #include "Core/MIPS/MIPSCodeUtils.h"
@@ -91,6 +93,13 @@ void HLEHelperThread::ChangePriority(u32 prio) {
 	KernelChangeThreadPriority(id_, prio);
 }
 
+void HLEHelperThread::Resume(WaitType waitType, SceUID uid, int result) {
+	bool res = HLEKernel::ResumeFromWait(id_, waitType, uid, result);
+	if (!res) {
+		ERROR_LOG(HLE, "Failed to wake helper thread from wait");
+	}
+}
+
 void HLEHelperThread::Forget() {
 	id_ = 0;
 	entry_ = 0;
diff --git a/Core/HLE/HLEHelperThread.h b/Core/HLE/HLEHelperThread.h
index 1868d309b0..39df4c7402 100644
--- a/Core/HLE/HLEHelperThread.h
+++ b/Core/HLE/HLEHelperThread.h
@@ -19,6 +19,7 @@
 #include "Core/HLE/sceKernel.h"
 
 class PointerWrap;
+enum WaitType : int;
 
 class HLEHelperThread {
 public:
@@ -33,10 +34,15 @@ public:
 	void Terminate();
 	bool Stopped();
 	void ChangePriority(u32 prio);
+	void Resume(WaitType waitType, SceUID uid, int result);
 
 	// For savestates.
 	void Forget();
 
+	u32 Entry() {
+		return entry_;
+	}
+
 private:
 	void AllocEntry(u32 size);
 	void Create(const char *threadName, u32 prio, int stacksize);
diff --git a/Core/HLE/sceIo.cpp b/Core/HLE/sceIo.cpp
index d81740095e..f1de847b12 100644
--- a/Core/HLE/sceIo.cpp
+++ b/Core/HLE/sceIo.cpp
@@ -199,6 +199,7 @@ struct IoAsyncParams {
 
 static IoAsyncParams asyncParams[PSP_COUNT_FDS];
 static HLEHelperThread *asyncThreads[PSP_COUNT_FDS]{};
+static int asyncDefaultPriority = -1;
 
 class FileNode : public KernelObject {
 public:
@@ -294,8 +295,11 @@ public:
 
 u64 __IoCompleteAsyncIO(FileNode *f);
 
-static void IoAsyncCleanupThread(int fd, bool force = false) {
-	if (asyncThreads[fd] && (asyncThreads[fd]->Stopped() || force)) {
+static void IoAsyncCleanupThread(int fd) {
+	if (asyncThreads[fd]) {
+		if (!asyncThreads[fd]->Stopped()) {
+			asyncThreads[fd]->Terminate();
+		}
 		delete asyncThreads[fd];
 		asyncThreads[fd] = nullptr;
 	}
@@ -437,8 +441,6 @@ static void __IoAsyncNotify(u64 userdata, int cyclesLate) {
 		if (f->closePending) {
 			__IoFreeFd(fd, error);
 		}
-
-		IoAsyncCleanupThread(fd);
 	}
 }
 
@@ -487,8 +489,6 @@ static void __IoSyncNotify(u64 userdata, int cyclesLate) {
 
 	HLEKernel::ResumeFromWait(threadID, WAITTYPE_IO, fd, result);
 	f->waitingSyncThreads.erase(std::remove(f->waitingSyncThreads.begin(), f->waitingSyncThreads.end(), threadID), f->waitingSyncThreads.end());
-
-	IoAsyncCleanupThread(fd);
 }
 
 static void __IoAsyncBeginCallback(SceUID threadID, SceUID prevCallbackId) {
@@ -662,7 +662,7 @@ void __IoInit() {
 }
 
 void __IoDoState(PointerWrap &p) {
-	auto s = p.Section("sceIo", 1, 4);
+	auto s = p.Section("sceIo", 1, 5);
 	if (!s)
 		return;
 
@@ -697,31 +697,36 @@ void __IoDoState(PointerWrap &p) {
 	}
 
 	for (int i = 0; i < PSP_COUNT_FDS; ++i) {
+		auto clearThread = [&]() {
+			if (asyncThreads[i])
+				asyncThreads[i]->Forget();
+			delete asyncThreads[i];
+			asyncThreads[i] = nullptr;
+		};
+
 		if (s >= 4) {
 			p.DoVoid(&asyncParams[i], (int)sizeof(IoAsyncParams));
 			bool hasThread = asyncThreads[i] != nullptr;
 			p.Do(hasThread);
 			if (hasThread) {
-				if (asyncThreads[i])
-					asyncThreads[i]->Forget();
-				delete asyncThreads[i];
-				asyncThreads[i] = nullptr;
+				if (p.GetMode() == p.MODE_READ)
+					clearThread();
 				p.DoClass(asyncThreads[i]);
 			} else if (!hasThread) {
-				if (asyncThreads[i])
-					asyncThreads[i]->Forget();
-				delete asyncThreads[i];
-				asyncThreads[i] = nullptr;
+				clearThread();
 			}
 		} else {
 			asyncParams[i].op = IoAsyncOp::NONE;
 			asyncParams[i].priority = -1;
-			if (asyncThreads[i])
-				asyncThreads[i]->Forget();
-			delete asyncThreads[i];
-			asyncThreads[i] = nullptr;
+			clearThread();
 		}
 	}
+
+	if (s >= 5) {
+		p.Do(asyncDefaultPriority);
+	} else {
+		asyncDefaultPriority = -1;
+	}
 }
 
 void __IoShutdown() {
@@ -739,6 +744,7 @@ void __IoShutdown() {
 		delete asyncThreads[i];
 		asyncThreads[i] = nullptr;
 	}
+	asyncDefaultPriority = -1;
 
 	pspFileSystem.Unmount("ms0:", memstickSystem);
 	pspFileSystem.Unmount("fatms0:", memstickSystem);
@@ -771,10 +777,20 @@ u32 __IoGetFileHandleFromId(u32 id, u32 &outError)
 }
 
 static void IoStartAsyncThread(int id, FileNode *f) {
-	IoAsyncCleanupThread(id, true);
-	int priority = asyncParams[id].priority == -1 ? KernelCurThreadPriority() : asyncParams[id].priority;
-	asyncThreads[id] = new HLEHelperThread("SceIoAsync", "IoFileMgrForUser", "__IoAsyncFinish", priority, 0x200);
-	asyncThreads[id]->Start(id, 0);
+	if (asyncThreads[id] && !asyncThreads[id]->Stopped()) {
+		// Wake the thread up.
+		if (asyncParams[id].priority == -1 && sceKernelGetCompiledSdkVersion() >= 0x04020000) {
+			asyncThreads[id]->ChangePriority(KernelCurThreadPriority());
+		}
+		asyncThreads[id]->Resume(WAITTYPE_ASYNCIO, id, 0);
+	} else {
+		IoAsyncCleanupThread(id);
+		int priority = asyncParams[id].priority;
+		if (priority == -1)
+			priority = KernelCurThreadPriority();
+		asyncThreads[id] = new HLEHelperThread("SceIoAsync", "IoFileMgrForUser", "__IoAsyncFinish", priority, 0x200);
+		asyncThreads[id]->Start(id, 0);
+	}
 	f->pendingAsyncResult = true;
 }
 
@@ -1249,17 +1265,14 @@ static u32 sceIoGetDevType(int id) {
 
 static u32 sceIoCancel(int id)
 {
-	ERROR_LOG_REPORT(SCEIO, "UNIMPL sceIoCancel(%d)", id);
 	u32 error;
 	FileNode *f = __IoGetFd(id, error);
 	if (f) {
-		// TODO: Cancel the async operation if possible?
+		// It seems like this is unsupported for UMDs and memory sticks, based on tests.
+		return hleReportError(SCEIO, SCE_KERNEL_ERROR_UNSUP, "unimplemented or unsupported");
 	} else {
-		ERROR_LOG(SCEIO, "sceIoCancel: unknown id %d", id);
-		error = SCE_KERNEL_ERROR_BADF;
+		return hleLogError(SCEIO, SCE_KERNEL_ERROR_BADF, "invalid fd");
 	}
-
-	return error;
 }
 
 static u32 npdrmLseek(FileNode *f, s32 where, FileMove whence)
@@ -1482,7 +1495,7 @@ static u32 sceIoOpen(const char *filename, int flags, int mode) {
 		return id;
 	} else {
 		DEBUG_LOG(SCEIO, "%i=sceIoOpen(%s, %08x, %08x)", id, filename, flags, mode);
-		asyncParams[id].priority = -1;
+		asyncParams[id].priority = asyncDefaultPriority;
 		// Timing is not accurate, aiming low for now.
 		return hleDelayResult(id, "file opened", 100);
 	}
@@ -1958,11 +1971,16 @@ static u32 sceIoChdir(const char *dirname) {
 }
 
 static int sceIoChangeAsyncPriority(int id, int priority) {
-	// priority = -1 is valid
+	// priority = -1 is valid,means the current thread'priority
 	if (priority != -1 && (priority < 0x08 || priority > 0x77)) {
 		return hleLogError(SCEIO, SCE_KERNEL_ERROR_ILLEGAL_PRIORITY, "illegal priority %d", priority);
 	}
 
+	if (id == -1) {
+		asyncDefaultPriority = priority;
+		return hleLogSuccessI(SCEIO, 0);
+	}
+
 	u32 error;
 	FileNode *f = __IoGetFd(id, error);
 	if (!f) {
@@ -1970,6 +1988,9 @@ static int sceIoChangeAsyncPriority(int id, int priority) {
 	}
 
 	if (asyncThreads[id] && !asyncThreads[id]->Stopped()) {
+		if (priority == -1) {
+			priority = KernelCurThreadPriority();
+		}
 		asyncThreads[id]->ChangePriority(priority);
 	}
 
@@ -2041,7 +2062,7 @@ static u32 sceIoOpenAsync(const char *filename, int flags, int mode)
 
 	auto &params = asyncParams[fd];
 	params.op = IoAsyncOp::OPEN;
-	params.priority = -1;
+	params.priority = asyncDefaultPriority;
 	params.open.filenameAddr = PARAM(0);
 	params.open.flags = flags;
 	params.open.mode = mode;
@@ -2087,7 +2108,6 @@ static u32 sceIoGetAsyncStat(int id, u32 poll, u32 address) {
 			DEBUG_LOG(SCEIO, "%lli = sceIoGetAsyncStat(%i, %i, %08x)", f->asyncResult, id, poll, address);
 			Memory::Write_U64((u64) f->asyncResult, address);
 			f->hasAsyncResult = false;
-			IoAsyncCleanupThread(id);
 
 			if (f->closePending) {
 				__IoFreeFd(id, error);
@@ -2125,7 +2145,6 @@ static int sceIoWaitAsync(int id, u32 address) {
 			}
 			Memory::Write_U64((u64) f->asyncResult, address);
 			f->hasAsyncResult = false;
-			IoAsyncCleanupThread(id);
 
 			if (f->closePending) {
 				__IoFreeFd(id, error);
@@ -2159,7 +2178,6 @@ static int sceIoWaitAsyncCB(int id, u32 address) {
 		} else if (f->hasAsyncResult) {
 			Memory::Write_U64((u64) f->asyncResult, address);
 			f->hasAsyncResult = false;
-			IoAsyncCleanupThread(id);
 
 			if (f->closePending) {
 				__IoFreeFd(id, error);
@@ -2183,7 +2201,6 @@ static u32 sceIoPollAsync(int id, u32 address) {
 		} else if (f->hasAsyncResult) {
 			Memory::Write_U64((u64) f->asyncResult, address);
 			f->hasAsyncResult = false;
-			IoAsyncCleanupThread(id);
 
 			if (f->closePending) {
 				__IoFreeFd(id, error);
@@ -2531,6 +2548,10 @@ static int __IoIoctl(u32 id, u32 cmd, u32 indataPtr, u32 inlen, u32 outdataPtr,
 		// TODO: Should probably move this to something common between ISOFileSystem and VirtualDiscSystem.
 		INFO_LOG(SCEIO, "sceIoIoctl: Sector seek for file %i", id);
 		// Even if the size is 4, it still actually reads a 16 byte struct, it seems.
+
+		//if (GetIOTimingMethod() == IOTIMING_REALISTIC) // Need a check for io timing method?
+		usec = 15000;// Fantasy Golf Pangya Portable(KS) needs a delay over 15000us.
+
 		if (Memory::IsValidAddress(indataPtr) && inlen >= 4) {
 			struct SeekInfo {
 				u64 offset;
@@ -2653,6 +2674,8 @@ static int IoAsyncFinish(int id) {
 	if (f) {
 		// Reset this so the Io funcs don't reject the request.
 		f->pendingAsyncResult = false;
+		// Reset the PC back so we will run again on resume.
+		currentMIPS->pc = asyncThreads[id]->Entry();
 
 		auto &params = asyncParams[id];
 
@@ -2714,6 +2737,7 @@ static int IoAsyncFinish(int id) {
 
 		__IoSchedAsync(f, id, us);
 		__KernelWaitCurThread(WAITTYPE_ASYNCIO, id, 0, 0, false, "async io");
+		hleSkipDeadbeef();
 
 		params.op = IoAsyncOp::NONE;
 		return 0;
diff --git a/Core/HLE/sceKernelThread.h b/Core/HLE/sceKernelThread.h
index c89a373ee9..8b95396b97 100644
--- a/Core/HLE/sceKernelThread.h
+++ b/Core/HLE/sceKernelThread.h
@@ -80,7 +80,7 @@ struct SceKernelSysClock {
 
 // TODO: Map these to PSP wait types.  Most of these are wrong.
 // remember to update the waitTypeNames array in sceKernelThread.cpp when changing these
-enum WaitType
+enum WaitType : int
 {
 	WAITTYPE_NONE         = 0,
 	WAITTYPE_SLEEP        = 1,
diff --git a/Core/HLE/sceMpeg.cpp b/Core/HLE/sceMpeg.cpp
index 43f30b9ce6..5243c76f03 100644
--- a/Core/HLE/sceMpeg.cpp
+++ b/Core/HLE/sceMpeg.cpp
@@ -1898,6 +1898,11 @@ static u32 sceMpegAtracDecode(u32 mpeg, u32 auAddr, u32 bufferAddr, int init)
 		return -1;
 	}
 
+	if (!Memory::IsValidAddress(bufferAddr)) {
+		WARN_LOG(ME, "sceMpegAtracDecode(%08x, %08x, %08x, %i): invalid addresses", mpeg, auAddr, bufferAddr, init);
+		return -1;
+	}
+
 	DEBUG_LOG(ME, "sceMpegAtracDecode(%08x, %08x, %08x, %i)", mpeg, auAddr, bufferAddr, init);
 
 	SceMpegAu atracAu;
diff --git a/Core/HLE/sceUmd.cpp b/Core/HLE/sceUmd.cpp
index bd4cb56d7b..ab2d62a490 100644
--- a/Core/HLE/sceUmd.cpp
+++ b/Core/HLE/sceUmd.cpp
@@ -425,7 +425,7 @@ static int sceUmdWaitDriveStatWithTimer(u32 stat, u32 timeout)
 		DEBUG_LOG(SCEIO, "sceUmdWaitDriveStatWithTimer(stat = %08x, timeout = %d): waiting", stat, timeout);
 		__UmdWaitStat(timeout);
 		umdWaitingThreads.push_back(__KernelGetCurThread());
-		__KernelWaitCurThread(WAITTYPE_UMD, 1, stat, 0, 0, "umd stat waited with timer");
+		__KernelWaitCurThread(WAITTYPE_UMD, 1, stat, 0, false, "umd stat waited with timer");
 		return 0;
 	} else {
 		hleReSchedule("umd stat checked");
@@ -495,45 +495,17 @@ static u32 sceUmdGetErrorStat()
 }
 
 void __UmdReplace(std::string filepath) {
-	// TODO: This should really go through Loaders, no?  What if it's an invalid file?
-
-	// Only get system from disc0 seems have been enough.
-	IFileSystem* currentUMD = pspFileSystem.GetSystem("disc0:");
-	IFileSystem* currentISOBlock = pspFileSystem.GetSystem("umd0:");
-	if (!currentUMD)
-		return;
-
-	FileLoader *loadedFile = ConstructFileLoader(filepath);
-
-	IFileSystem* umd2;
-	if (!loadedFile->Exists()) {
-		delete loadedFile;
+	std::string error = "";
+	if (!UmdReplace(filepath, error)) {
+		ERROR_LOG(SCEIO, "UMD Replace failed: %s", error.c_str());
 		return;
 	}
-	UpdateLoadedFile(loadedFile);
 
-	if (loadedFile->IsDirectory()) {
-		umd2 = new VirtualDiscFileSystem(&pspFileSystem, filepath);
-	} else {
-		auto bd = constructBlockDevice(loadedFile);
-		if (!bd)
-			return;
-		umd2 = new ISOFileSystem(&pspFileSystem, bd);
-		pspFileSystem.Remount(currentUMD, umd2);
-
-		if (currentUMD != currentISOBlock) {
-			// We mounted an ISO block system separately.
-			IFileSystem *iso = new ISOBlockSystem(static_cast<ISOFileSystem *>(umd2));
-			pspFileSystem.Remount(currentISOBlock, iso);
-			delete currentISOBlock;
-		}
-	}
-	delete currentUMD;
 	UMDInserted = false;
 	CoreTiming::ScheduleEvent(usToCycles(200*1000), umdInsertChangeEvent, 0); // Wait sceUmdCheckMedium call
 	// TODO Is this always correct if UMD was not activated?
 	u32 notifyArg = PSP_UMD_PRESENT | PSP_UMD_READABLE | PSP_UMD_CHANGED;
-	if (driveCBId != -1)
+	if (driveCBId != 0)
 		__KernelNotifyCallback(driveCBId, notifyArg);
 }
 
diff --git a/Core/Loaders.cpp b/Core/Loaders.cpp
index 14bc31973d..a4d102e122 100644
--- a/Core/Loaders.cpp
+++ b/Core/Loaders.cpp
@@ -355,3 +355,41 @@ bool LoadFile(FileLoader **fileLoaderPtr, std::string *error_string) {
 	coreState = CORE_ERROR;
 	return false;
 }
+
+bool UmdReplace(std::string filepath, std::string &error) {
+	IFileSystem* currentUMD = pspFileSystem.GetSystem("disc0:");
+
+	if (!currentUMD) {
+		error = "has no disc";
+		return false;
+	}
+
+	FileLoader *loadedFile = ConstructFileLoader(filepath);
+
+	if (!loadedFile->Exists()) {
+		delete loadedFile;
+		error = loadedFile->Path() + " doesn't exist";
+		return false;
+	}
+	UpdateLoadedFile(loadedFile);
+
+	loadedFile = ResolveFileLoaderTarget(loadedFile);
+	IdentifiedFileType type = Identify_File(loadedFile);
+
+	switch (type) {
+	case IdentifiedFileType::PSP_ISO:
+	case IdentifiedFileType::PSP_ISO_NP:
+	case IdentifiedFileType::PSP_DISC_DIRECTORY:
+		if (!ReInitMemoryForGameISO(loadedFile)) {
+			error = "reinit memory failed";
+			return false;
+		}
+
+		break;
+	default:
+		error = "Unsupported file type:" + std::to_string((int)type);
+		return false;
+		break;
+	}
+	return true;
+}
diff --git a/Core/Loaders.h b/Core/Loaders.h
index 690376c73b..a076cfa675 100644
--- a/Core/Loaders.h
+++ b/Core/Loaders.h
@@ -158,3 +158,5 @@ void RegisterFileLoaderFactory(std::string name, std::unique_ptr<FileLoaderFacto
 
 // Can modify the string filename, as it calls IdentifyFile above.
 bool LoadFile(FileLoader **fileLoaderPtr, std::string *error_string);
+
+bool UmdReplace(std::string filepath, std::string &error);
diff --git a/Core/PSPLoaders.cpp b/Core/PSPLoaders.cpp
index 8baf5c13fa..801a49485c 100644
--- a/Core/PSPLoaders.cpp
+++ b/Core/PSPLoaders.cpp
@@ -83,7 +83,7 @@ void InitMemoryForGameISO(FileLoader *fileLoader) {
 
 	IFileSystem *fileSystem = nullptr;
 	IFileSystem *blockSystem = nullptr;
-	bool actualIso = false;
+
 	if (fileLoader->IsDirectory()) {
 		fileSystem = new VirtualDiscFileSystem(&pspFileSystem, fileLoader->Path());
 		blockSystem = fileSystem;
@@ -144,6 +144,35 @@ void InitMemoryForGameISO(FileLoader *fileLoader) {
 	}
 }
 
+bool ReInitMemoryForGameISO(FileLoader *fileLoader) {
+	if (!fileLoader->Exists()) {
+		return false;
+	}
+
+	IFileSystem *fileSystem = nullptr;
+	IFileSystem *blockSystem = nullptr;
+
+	if (fileLoader->IsDirectory()) {
+		fileSystem = new VirtualDiscFileSystem(&pspFileSystem, fileLoader->Path());
+		blockSystem = fileSystem;
+	} else {
+		auto bd = constructBlockDevice(fileLoader);
+		if (!bd)
+			return false;
+
+		ISOFileSystem *iso = new ISOFileSystem(&pspFileSystem, bd);
+		fileSystem = iso;
+		blockSystem = new ISOBlockSystem(iso);
+	}
+
+	pspFileSystem.Remount("umd0:", blockSystem);
+	pspFileSystem.Remount("umd1:", blockSystem);
+	pspFileSystem.Remount("umd:", blockSystem);
+	pspFileSystem.Remount("disc0:", fileSystem);
+
+	return true;
+}
+
 void InitMemoryForGamePBP(FileLoader *fileLoader) {
 	if (!fileLoader->Exists()) {
 		return;
diff --git a/Core/PSPLoaders.h b/Core/PSPLoaders.h
index f344008460..e9c089494e 100644
--- a/Core/PSPLoaders.h
+++ b/Core/PSPLoaders.h
@@ -25,5 +25,6 @@ bool Load_PSP_ISO(FileLoader *fileLoader, std::string *error_string);
 bool Load_PSP_ELF_PBP(FileLoader *fileLoader, std::string *error_string);
 bool Load_PSP_GE_Dump(FileLoader *fileLoader, std::string *error_string);
 void InitMemoryForGameISO(FileLoader *fileLoader);
+bool ReInitMemoryForGameISO(FileLoader *fileLoader);
 void InitMemoryForGamePBP(FileLoader *fileLoader);
 void PSPLoaders_Shutdown();
diff --git a/GPU/Software/SoftGpu.cpp b/GPU/Software/SoftGpu.cpp
index 7254c2ce11..691a8756aa 100644
--- a/GPU/Software/SoftGpu.cpp
+++ b/GPU/Software/SoftGpu.cpp
@@ -158,6 +158,38 @@ void SoftGPU::SetDisplayFramebuffer(u32 framebuf, u32 stride, GEBufferFormat for
 
 bool g_DarkStalkerStretch;
 
+void SoftGPU::ConvertTextureDescFrom16(Draw::TextureDesc &desc, int srcwidth, int srcheight) {
+	// TODO: This should probably be converted in a shader instead..
+	fbTexBuffer_.resize(srcwidth * srcheight);
+	FormatBuffer displayBuffer;
+	displayBuffer.data = Memory::GetPointer(displayFramebuf_);
+	for (int y = 0; y < srcheight; ++y) {
+		u32 *buf_line = &fbTexBuffer_[y * srcwidth];
+		const u16 *fb_line = &displayBuffer.as16[y * displayStride_];
+
+		switch (displayFormat_) {
+		case GE_FORMAT_565:
+			ConvertRGBA565ToRGBA8888(buf_line, fb_line, srcwidth);
+			break;
+
+		case GE_FORMAT_5551:
+			ConvertRGBA5551ToRGBA8888(buf_line, fb_line, srcwidth);
+			break;
+
+		case GE_FORMAT_4444:
+			ConvertRGBA4444ToRGBA8888(buf_line, fb_line, srcwidth);
+			break;
+
+		default:
+			ERROR_LOG_REPORT(G3D, "Software: Unexpected framebuffer format: %d", displayFormat_);
+		}
+	}
+
+	desc.width = srcwidth;
+	desc.height = srcheight;
+	desc.initData.push_back((uint8_t *)fbTexBuffer_.data());
+}
+
 // Copies RGBA8 data from RAM to the currently bound render target.
 void SoftGPU::CopyToCurrentFboFromDisplayRam(int srcwidth, int srcheight) {
 	if (!draw_)
@@ -191,6 +223,7 @@ void SoftGPU::CopyToCurrentFboFromDisplayRam(int srcwidth, int srcheight) {
 	Draw::Pipeline *pipeline = texColor;
 	if (PSP_CoreParameter().compat.flags().DarkStalkersPresentHack && displayFormat_ == GE_FORMAT_5551 && g_DarkStalkerStretch) {
 		u8 *data = Memory::GetPointer(0x04088000);
+		bool fillDesc = true;
 		if (draw_->GetDataFormatSupport(Draw::DataFormat::A1B5G5R5_UNORM_PACK16) & Draw::FMT_TEXTURE) {
 			// The perfect one.
 			desc.format = Draw::DataFormat::A1B5G5R5_UNORM_PACK16;
@@ -198,10 +231,15 @@ void SoftGPU::CopyToCurrentFboFromDisplayRam(int srcwidth, int srcheight) {
 			// RB swapped, compensate with a shader.
 			desc.format = Draw::DataFormat::A1R5G5B5_UNORM_PACK16;
 			pipeline = texColorRBSwizzle;
+		} else {
+			ConvertTextureDescFrom16(desc, srcwidth, srcheight);
+			fillDesc = false;
+		}
+		if (fillDesc) {
+			desc.width = displayStride_ == 0 ? srcwidth : displayStride_;
+			desc.height = srcheight;
+			desc.initData.push_back(data);
 		}
-		desc.width = displayStride_ == 0 ? srcwidth : displayStride_;
-		desc.height = srcheight;
-		desc.initData.push_back(data);
 		u0 = 64.5f / 512.0f;
 		u1 = 447.5f / 512.0f;
 		v1 = 16.0f / 272.0f;
@@ -217,6 +255,7 @@ void SoftGPU::CopyToCurrentFboFromDisplayRam(int srcwidth, int srcheight) {
 		desc.format = Draw::DataFormat::R8G8B8A8_UNORM;
 	} else if (displayFormat_ == GE_FORMAT_5551) {
 		u8 *data = Memory::GetPointer(displayFramebuf_);
+		bool fillDesc = true;
 		desc.format = Draw::DataFormat::A1R5G5B5_UNORM_PACK16;
 		if (draw_->GetDataFormatSupport(Draw::DataFormat::A1B5G5R5_UNORM_PACK16) & Draw::FMT_TEXTURE) {
 			// The perfect one.
@@ -225,40 +264,17 @@ void SoftGPU::CopyToCurrentFboFromDisplayRam(int srcwidth, int srcheight) {
 			// RB swapped, compensate with a shader.
 			desc.format = Draw::DataFormat::A1R5G5B5_UNORM_PACK16;
 			pipeline = texColorRBSwizzle;
+		} else {
+			ConvertTextureDescFrom16(desc, srcwidth, srcheight);
+			fillDesc = false;
+		}
+		if (fillDesc) {
+			desc.width = displayStride_ == 0 ? srcwidth : displayStride_;
+			desc.height = srcheight;
+			desc.initData.push_back(data);
 		}
-		desc.width = displayStride_ == 0 ? srcwidth : displayStride_;
-		desc.height = srcheight;
-		desc.initData.push_back(data);
 	} else {
-		// TODO: This should probably be converted in a shader instead..
-		fbTexBuffer.resize(srcwidth * srcheight);
-		FormatBuffer displayBuffer;
-		displayBuffer.data = Memory::GetPointer(displayFramebuf_);
-		for (int y = 0; y < srcheight; ++y) {
-			u32 *buf_line = &fbTexBuffer[y * srcwidth];
-			const u16 *fb_line = &displayBuffer.as16[y * displayStride_];
-
-			switch (displayFormat_) {
-			case GE_FORMAT_565:
-				ConvertRGBA565ToRGBA8888(buf_line, fb_line, srcwidth);
-				break;
-
-			case GE_FORMAT_5551:
-				ConvertRGBA5551ToRGBA8888(buf_line, fb_line, srcwidth);
-				break;
-
-			case GE_FORMAT_4444:
-				ConvertRGBA4444ToRGBA8888(buf_line, fb_line, srcwidth);
-				break;
-
-			default:
-				ERROR_LOG_REPORT(G3D, "Software: Unexpected framebuffer format: %d", displayFormat_);
-			}
-		}
-
-		desc.width = srcwidth;
-		desc.height = srcheight;
-		desc.initData.push_back((uint8_t *)fbTexBuffer.data());
+		ConvertTextureDescFrom16(desc, srcwidth, srcheight);
 		u1 = 1.0f;
 	}
 	if (!hasImage) {
diff --git a/GPU/Software/SoftGpu.h b/GPU/Software/SoftGpu.h
index 02660b5576..84626b31c2 100644
--- a/GPU/Software/SoftGpu.h
+++ b/GPU/Software/SoftGpu.h
@@ -101,6 +101,7 @@ public:
 protected:
 	void FastRunLoop(DisplayList &list) override;
 	void CopyToCurrentFboFromDisplayRam(int srcwidth, int srcheight);
+	void ConvertTextureDescFrom16(Draw::TextureDesc &desc, int srcwidth, int srcheight);
 
 private:
 	bool framebufferDirty_;
@@ -113,7 +114,7 @@ private:
 	Draw::Texture *fbTex;
 	Draw::Pipeline *texColor;
 	Draw::Pipeline *texColorRBSwizzle;
-	std::vector<u32> fbTexBuffer;
+	std::vector<u32> fbTexBuffer_;
 
 	Draw::SamplerState *samplerNearest = nullptr;
 	Draw::SamplerState *samplerLinear = nullptr;
diff --git a/GPU/Vulkan/GPU_Vulkan.cpp b/GPU/Vulkan/GPU_Vulkan.cpp
index 1a5e254e27..3f20ec0642 100644
--- a/GPU/Vulkan/GPU_Vulkan.cpp
+++ b/GPU/Vulkan/GPU_Vulkan.cpp
@@ -485,8 +485,9 @@ void GPU_Vulkan::InitDeviceObjects() {
 		hacks |= QUEUE_HACK_MGS2_ACID;
 	if (PSP_CoreParameter().compat.flags().SonicRivalsHack)
 		hacks |= QUEUE_HACK_SONIC;
-	if (PSP_CoreParameter().compat.flags().RenderPassMerge)
-		hacks |= QUEUE_HACK_RENDERPASS_MERGE;
+
+	// Always on.
+	hacks |= QUEUE_HACK_RENDERPASS_MERGE;
 
 	if (hacks) {
 		rm->GetQueueRunner()->EnableHacks(hacks);
diff --git a/Qt/QtHost.cpp b/Qt/QtHost.cpp
index c2ef7d2eac..e13f916c69 100644
--- a/Qt/QtHost.cpp
+++ b/Qt/QtHost.cpp
@@ -18,12 +18,12 @@
 #include <string>
 #include "Qt/QtHost.h"
 
-const char* QtHost::SymbolMapFilename(std::string currentFilename) {
-	std::string result = currentFilename;
-	size_t dot = result.rfind('.');
-	if (dot == result.npos)
-		return (result + ".map").c_str();
+std::string QtHost::SymbolMapFilename(std::string currentFilename) {
+	size_t dot = currentFilename.rfind('.');
+	if (dot == std::string::npos)
+		currentFilename.append(".map");
+	else
+		currentFilename.replace(dot, -1, ".map");
 
-	result.replace(dot, result.npos, ".map");
-	return result.c_str();
+	return currentFilename;
 }
diff --git a/Qt/QtHost.h b/Qt/QtHost.h
index 846da1058f..1a15740806 100644
--- a/Qt/QtHost.h
+++ b/Qt/QtHost.h
@@ -64,13 +64,12 @@ public:
 #endif
 	}
 	virtual bool AttemptLoadSymbolMap() override {
-		return false;
-		// TODO: Make this work with Qt and threaded GL... not sure what's so broken.
-		// auto fn = SymbolMapFilename(PSP_CoreParameter().fileToStart);
-		// return g_symbolMap->LoadSymbolMap(fn);
+		auto fn = SymbolMapFilename(PSP_CoreParameter().fileToStart);
+		return g_symbolMap->LoadSymbolMap(fn.c_str());
 	}
 	void PrepareShutdown() {
-		g_symbolMap->SaveSymbolMap(SymbolMapFilename(PSP_CoreParameter().fileToStart));
+		auto fn = SymbolMapFilename(PSP_CoreParameter().fileToStart);
+		g_symbolMap->SaveSymbolMap(fn.c_str());
 	}
 	void SetWindowTitle(const char *message) override {
 		std::string title = std::string("PPSSPP ") + PPSSPP_GIT_VERSION;
@@ -91,6 +90,6 @@ public:
 	}
 
 private:
-	const char* SymbolMapFilename(std::string currentFilename);
+	std::string SymbolMapFilename(std::string currentFilename);
 	MainWindow* mainWindow;
 };
diff --git a/Qt/mainwindow.cpp b/Qt/mainwindow.cpp
index 502e22a0fd..4b4a6aaaf7 100644
--- a/Qt/mainwindow.cpp
+++ b/Qt/mainwindow.cpp
@@ -29,6 +29,8 @@ MainWindow::MainWindow(QWidget *parent, bool fullscreen) :
 	QRect rect = desktop->screenGeometry(screenNum);
 	move((rect.width()-frameGeometry().width()) / 4, (rect.height()-frameGeometry().height()) / 4);
 
+	setWindowIcon(QIcon(qApp->applicationDirPath() + "/assets/icon_regular_72.png"));
+
 	SetGameTitle("");
 	emugl = new MainUI(this);
 
diff --git a/SDL/SDLMain.cpp b/SDL/SDLMain.cpp
index f868592bd2..90db5d4195 100644
--- a/SDL/SDLMain.cpp
+++ b/SDL/SDLMain.cpp
@@ -910,7 +910,9 @@ int main(int argc, char *argv[]) {
 					if (!name) {
 						break;
 					}
-					if (g_Config.bAutoAudioDevice || g_Config.sAudioDevice == name) {
+					// Don't start auto switching for a second, because some devices init on start.
+					bool doAutoSwitch = g_Config.bAutoAudioDevice && framecount > 60;
+					if (doAutoSwitch || g_Config.sAudioDevice == name) {
 						StopSDLAudioDevice();
 						InitSDLAudioDevice(name ? name : "");
 					}
diff --git a/UI/DiscordIntegration.cpp b/UI/DiscordIntegration.cpp
index fea814924d..b88680f048 100644
--- a/UI/DiscordIntegration.cpp
+++ b/UI/DiscordIntegration.cpp
@@ -64,11 +64,12 @@ void Discord::Init() {
 }
 
 void Discord::Shutdown() {
-	assert(initialized_);
+	if (initialized_) {
 #ifdef ENABLE_DISCORD
-	Discord_Shutdown();
+		Discord_Shutdown();
 #endif
-	initialized_ = false;
+		initialized_ = false;
+	}
 }
 
 void Discord::Update() {
diff --git a/UI/TouchControlLayoutScreen.cpp b/UI/TouchControlLayoutScreen.cpp
index 24574ed951..65ebb654cf 100644
--- a/UI/TouchControlLayoutScreen.cpp
+++ b/UI/TouchControlLayoutScreen.cpp
@@ -243,9 +243,15 @@ bool TouchControlLayoutScreen::touch(const TouchInput &touch) {
 				// if the leftmost point of the control is ahead of the margin,
 				// move it. Otherwise, don't.
 				newX = touch.x;
+				// Snap to grid
+				if (g_Config.bTouchSnapToGrid)
+					newX -= (int)(touch.x - bounds.w) % g_Config.iTouchSnapGridSize;
 			}
 			if (touch.y > minTouchY && touch.y < maxTouchY) {
 				newY = touch.y;
+				// Snap to grid
+				if (g_Config.bTouchSnapToGrid)
+					newY -= (int)(touch.y - bounds.h) % g_Config.iTouchSnapGridSize;
 			}
 			pickedControl_->ReplaceLayoutParams(new UI::AnchorLayoutParams(newX, newY, NONE, NONE, true));
 		} else if (mode == 1) {
@@ -324,11 +330,15 @@ void TouchControlLayoutScreen::CreateViews() {
 
 	Choice *reset = new Choice(di->T("Reset"), "", false, new AnchorLayoutParams(leftColumnWidth, WRAP_CONTENT, 10, NONE, NONE, 84));
 	Choice *back = new Choice(di->T("Back"), "", false, new AnchorLayoutParams(leftColumnWidth, WRAP_CONTENT, 10, NONE, NONE, 10));
-	Choice *visibility = new Choice(co->T("Visibility"), "", false, new AnchorLayoutParams(leftColumnWidth, WRAP_CONTENT, 10, NONE, NONE, 158));
+	Choice *visibility = new Choice(co->T("Visibility"), "", false, new AnchorLayoutParams(leftColumnWidth, WRAP_CONTENT, 10, NONE, NONE, 298));
 	// controlsSettings->Add(new PopupSliderChoiceFloat(&g_Config.fButtonScale, 0.80, 2.0, co->T("Button Scaling"), screenManager()))
 	// 	->OnChange.Handle(this, &GameSettingsScreen::OnChangeControlScaling);
 
-	mode_ = new ChoiceStrip(ORIENT_VERTICAL, new AnchorLayoutParams(leftColumnWidth, WRAP_CONTENT, 10, NONE, NONE, 158 + 64 + 10));
+	CheckBox *snap = new CheckBox(&g_Config.bTouchSnapToGrid, di->T("Snap"), "", new AnchorLayoutParams(leftColumnWidth, WRAP_CONTENT, 10, NONE, NONE, 228));
+	PopupSliderChoice *gridSize = new PopupSliderChoice(&g_Config.iTouchSnapGridSize, 1, 256, di->T("Grid"), screenManager(), "", new AnchorLayoutParams(leftColumnWidth, WRAP_CONTENT, 10, NONE, NONE, 158));
+	gridSize->SetEnabledPtr(&g_Config.bTouchSnapToGrid);
+
+	mode_ = new ChoiceStrip(ORIENT_VERTICAL, new AnchorLayoutParams(leftColumnWidth, WRAP_CONTENT, 10, NONE, NONE, 140 + 158 + 64 + 10));
 	mode_->AddChoice(di->T("Move"));
 	mode_->AddChoice(di->T("Resize"));
 	mode_->SetSelection(0);
@@ -338,6 +348,8 @@ void TouchControlLayoutScreen::CreateViews() {
 	visibility->OnClick.Handle(this, &TouchControlLayoutScreen::OnVisibility);
 	root_->Add(mode_);
 	root_->Add(visibility);
+	root_->Add(snap);
+	root_->Add(gridSize);
 	root_->Add(reset);
 	root_->Add(back);
 
@@ -352,7 +364,7 @@ void TouchControlLayoutScreen::CreateViews() {
 
 	I18NCategory *ms = GetI18NCategory("MainSettings");
 
-	tabHolder->AddTab(ms->T("Controls"), controlsHolder);
+	//tabHolder->AddTab(ms->T("Controls"), controlsHolder);
 
 	if (!g_Config.bShowTouchControls) {
 		// Shouldn't even be able to get here as the way into this dialog should be closed.
diff --git a/assets/compat.ini b/assets/compat.ini
index 4d97b71450..df1bea3498 100644
--- a/assets/compat.ini
+++ b/assets/compat.ini
@@ -764,48 +764,14 @@ ULAS42214 = true
 ULJS19054 = true
 NPJH50184 = true
 
-[RenderPassMerge]
-UCJS10114 = true
-UCKS45084 = true
-# GOW : Ghost of Sparta
-UCUS98737 = true
-UCAS40323 = true
-NPHG00092 = true
-NPEG00044 = true
-NPEG00045 = true
-NPJG00120 = true
-NPUG80508 = true
-UCJS10114 = true
-UCES01401 = true
-UCES01473 = true
-# GOW : Ghost of Sparta Demo
-NPEG90035 = true
-NPUG70125 = true
-NPJG90095 = true
-# GOW : Chains Of Olympus
-UCAS40198 = true
-UCUS98653 = true
-UCES00842 = true
-ULJM05438 = true
-ULJM05348 = true
-UCKS45084 = true
-NPUG80325 = true
-NPEG00023 = true
-NPHG00027 = true
-NPHG00028 = true
-NPJH50170 = true
-UCET00844 = true
-# GOW: Chains of Olympus Demo
-UCUS98705 = true
-UCED00971 = true
-UCUS98713 = true
-
 [ForceSoftwareRenderer]
 # Darkstalkers
 ULES00016 = true
 ULUS10005 = true
+ULJM05005 = true
 
 [DarkStalkersPresentHack]
 # Darkstalkers
 ULES00016 = true
 ULUS10005 = true
+ULJM05005 = true
diff --git a/assets/lang b/assets/lang
index a95fd4974c..6189eb455e 160000
--- a/assets/lang
+++ b/assets/lang
@@ -1 +1 @@
-Subproject commit a95fd4974c436a1a07455c5beef4a12f9f47b811
+Subproject commit 6189eb455e4b184ede558f7354e6ba31b485fd58
diff --git a/build.gradle b/build.gradle
index c651082644..a9deb8ae96 100644
--- a/build.gradle
+++ b/build.gradle
@@ -5,7 +5,7 @@ buildscript {
         jcenter()
     }
     dependencies {
-        classpath 'com.android.tools.build:gradle:3.5.2'
+        classpath 'com.android.tools.build:gradle:3.5.3'
     }
 }
 
diff --git a/ext/native/gfx_es2/draw_buffer.cpp b/ext/native/gfx_es2/draw_buffer.cpp
index 5d1bb88153..b329534a93 100644
--- a/ext/native/gfx_es2/draw_buffer.cpp
+++ b/ext/native/gfx_es2/draw_buffer.cpp
@@ -495,8 +495,12 @@ void DrawBuffer::DrawTextRect(int font, const char *text, float x, float y, floa
 // ROTATE_* doesn't yet work right.
 void DrawBuffer::DrawText(int font, const char *text, float x, float y, Color color, int align) {
 	// rough estimate
-	if (count_ + strlen(text) * 6 > MAX_VERTS) {
+	size_t textLen = strlen(text);
+	if (count_ + textLen * 6 > MAX_VERTS) {
 		Flush(true);
+		if (textLen * 6 >= MAX_VERTS) {
+			textLen = std::min(MAX_VERTS / 6 - 10, (int)textLen);
+		}
 	}
 
 	const AtlasFont &atlasfont = *atlas->fonts[font];
@@ -508,14 +512,14 @@ void DrawBuffer::DrawText(int font, const char *text, float x, float y, Color co
 	}
 
 	if (align & ROTATE_90DEG_LEFT) {
-		x -= atlasfont.ascend*fontscaley;
+		x -= atlasfont.ascend * fontscaley;
 		// y += h;
+	} else {
+		y += atlasfont.ascend * fontscaley;
 	}
-	else
-		y += atlasfont.ascend*fontscaley;
 	float sx = x;
 	UTF8 utf(text);
-	while (true) {
+	for (size_t i = 0; i < textLen; i++) {
 		if (utf.end())
 			break;
 		cval = utf.next();
diff --git a/ext/native/thin3d/VulkanQueueRunner.cpp b/ext/native/thin3d/VulkanQueueRunner.cpp
index 3785b7d9b9..8a7dcd8f4a 100644
--- a/ext/native/thin3d/VulkanQueueRunner.cpp
+++ b/ext/native/thin3d/VulkanQueueRunner.cpp
@@ -469,7 +469,7 @@ void VulkanQueueRunner::RunSteps(VkCommandBuffer cmd, std::vector<VKRStep *> &st
 			break;
 		}
 
-		if (profile) {
+		if (profile && profile->timestampDescriptions.size() + 1 < MAX_TIMESTAMP_QUERIES) {
 			vkCmdWriteTimestamp(cmd, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, profile->queryPool, (uint32_t)profile->timestampDescriptions.size());
 			profile->timestampDescriptions.push_back(StepToString(step));
 		}
@@ -789,11 +789,13 @@ void VulkanQueueRunner::ApplyRenderPassMerge(std::vector<VKRStep *> &steps) {
 					if (steps[j]->copy.src == fb || steps[j]->copy.dst == fb) {
 						goto done_fb;
 					}
+					touchedFramebuffers.insert(steps[j]->copy.dst);
 					break;
 				case VKRStepType::BLIT:
 					if (steps[j]->blit.src == fb || steps[j]->blit.dst == fb) {
 						goto done_fb;
 					}
+					touchedFramebuffers.insert(steps[j]->blit.dst);
 					break;
 				case VKRStepType::READBACK:
 					// Not sure this has much effect, when executed READBACK is always the last step
diff --git a/ext/native/thin3d/VulkanRenderManager.h b/ext/native/thin3d/VulkanRenderManager.h
index 4201d912ad..c14c894a72 100644
--- a/ext/native/thin3d/VulkanRenderManager.h
+++ b/ext/native/thin3d/VulkanRenderManager.h
@@ -85,6 +85,10 @@ enum class VKRRunType {
 	SYNC,
 };
 
+enum {
+	MAX_TIMESTAMP_QUERIES = 128,
+};
+
 class VulkanRenderManager {
 public:
 	VulkanRenderManager(VulkanContext *vulkan);
@@ -299,10 +303,6 @@ private:
 
 	FrameData frameData_[VulkanContext::MAX_INFLIGHT_FRAMES];
 
-	enum {
-		MAX_TIMESTAMP_QUERIES = 256,
-	};
-
 	// Submission time state
 	int curWidth_ = -1;
 	int curHeight_ = -1;
diff --git a/ffmpeg b/ffmpeg
index 90701640c7..55147e5f33 160000
--- a/ffmpeg
+++ b/ffmpeg
@@ -1 +1 @@
-Subproject commit 90701640c7f458461310b54e7d4041230e2d5d5a
+Subproject commit 55147e5f33f5ae4904f75ec082af809267122b94