diff --git a/.gitmodules b/.gitmodules
index 98e42059f7..2be86e3e76 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -22,6 +22,7 @@
 [submodule "libpng"]
 	path = 3rdparty/libpng
 	url = https://github.com/RPCS3/libpng
+	ignore = dirty
 [submodule "Vulkan/glslang"]
 	path = Vulkan/glslang
 	url = https://github.com/KhronosGroup/glslang.git
diff --git a/3rdparty/XAudio2_7/XAudio2.h b/3rdparty/XAudio2_7/XAudio2.h
index 885b92f919..e4ac16786f 100644
--- a/3rdparty/XAudio2_7/XAudio2.h
+++ b/3rdparty/XAudio2_7/XAudio2.h
@@ -62,6 +62,19 @@ DEFINE_IID(IXAudio2, 8bcf1f58, 9fe7, 4583, 8a, c6, e2, ad, c4, 65, c8, bb);
     #include <objbase.h>    // Windows COM declarations
 #endif
 
+#ifndef _MSC_VER
+#define __in
+#define __in_opt
+#define __in_bcount(x)
+#define __in_ecount(x)
+#define __deref_out
+#define __out
+#define __out_ecount(x)
+#define __out_bcount(x)
+#define __inout
+#define __reserved
+#endif
+
 #include <sal.h>            // Markers for documenting API semantics
 #include <audiodefs.h>      // Basic audio data types and constants
 #include <xma2defs.h>       // Data types and constants for XMA2 audio
diff --git a/3rdparty/XAudio2_7/comdecl.h b/3rdparty/XAudio2_7/comdecl.h
index 2ae9a961e9..e0605892e0 100644
--- a/3rdparty/XAudio2_7/comdecl.h
+++ b/3rdparty/XAudio2_7/comdecl.h
@@ -21,7 +21,13 @@
 // GUID definitions by defining the INITGUID preprocessor constant or by linking
 // to a GUID library.  This works in either C or C++.
 
-#ifdef __cplusplus
+#ifndef _MSC_VER
+
+#define DEFINE_UUID(name, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) __CRT_UUID_DECL(name, 0x##l, 0x##w1, 0x##w2, 0x##b1, 0x##b2, 0x##b3, 0x##b4, 0x##b5, 0x##b6, 0x##b7, 0x##b8)
+#define DEFINE_CLSID(className, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) class className; DEFINE_UUID(className, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8)
+#define DEFINE_IID(interfaceName, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) struct interfaceName; DEFINE_UUID(interfaceName, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8)
+
+#elif __cplusplus
 
     #define DECLSPEC_UUID_WRAPPER(x) __declspec(uuid(#x))
     #ifdef INITGUID
diff --git a/CMakeLists.txt b/CMakeLists.txt
index cc5da37646..e9694ce636 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -17,7 +17,6 @@ set(PNG_TESTS  OFF CACHE BOOL "Build tests." FORCE)
 
 add_definitions(-DCMAKE_BUILD)
 
-add_subdirectory( asmjit )
 add_subdirectory( 3rdparty/libpng )
 # TODO: do real installation, including copying directory structure
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY_RELEASE "${PROJECT_BINARY_DIR}/bin")
diff --git a/Utilities/Atomic.h b/Utilities/Atomic.h
index b7591b556a..6790d56400 100644
--- a/Utilities/Atomic.h
+++ b/Utilities/Atomic.h
@@ -976,4 +976,17 @@ public:
 	{
 		return atomic_op(atomic_test_and_complement<type, T2>{}, rhs);
 	}
+
+	// Minimal pointer support (TODO: must forward operator ->())
+	type operator ->() const
+	{
+		return load();
+	}
+
+	// Minimal array support
+	template<typename I = std::size_t>
+	auto operator [](const I& index) const -> decltype(std::declval<const type>()[std::declval<I>()])
+	{
+		return load()[index];
+	}
 };
diff --git a/Utilities/AtomicPtr.h b/Utilities/AtomicPtr.h
new file mode 100644
index 0000000000..a2e811bf08
--- /dev/null
+++ b/Utilities/AtomicPtr.h
@@ -0,0 +1,150 @@
+#pragma once
+
+#include "Atomic.h"
+#include <memory>
+
+// Unfinished. Only std::default_delete will work as expected.
+template<typename T, typename D>
+class atomic_ptr_base : D
+{
+protected:
+	atomic_t<T*> m_ptr;
+
+	constexpr atomic_ptr_base(T* ptr)
+		: m_ptr(ptr)
+	{
+	}
+
+public:
+	~atomic_ptr_base()
+	{
+		if (m_ptr)
+		{
+			(*this)(m_ptr.load());
+		}
+	}
+
+	D& get_deleter()
+	{
+		return *this;
+	}
+
+	const D& get_deleter() const
+	{
+		return *this;
+	}
+};
+
+// Simple atomic pointer with unique ownership. Draft, unfinished.
+template<typename T, typename D = std::default_delete<T>>
+class atomic_ptr final : atomic_ptr_base<T, D>
+{
+	using base = atomic_ptr_base<T, D>;
+
+	static_assert(sizeof(T*) == sizeof(base), "atomic_ptr<> error: invalid deleter (empty class expected)");
+
+public:
+	constexpr atomic_ptr()
+		: base(nullptr)
+	{
+	}
+
+	constexpr atomic_ptr(std::nullptr_t)
+		: base(nullptr)
+	{
+	}
+
+	explicit atomic_ptr(T* ptr)
+		: base(ptr)
+	{
+	}
+
+	template<typename T2, typename = std::enable_if_t<std::is_convertible<T2, T>::value>>
+	atomic_ptr(std::unique_ptr<T2, D>&& ptr)
+		: base(ptr.release())
+	{
+	}
+
+	atomic_ptr& operator =(std::nullptr_t)
+	{
+		if (T* old = base::m_ptr.exchange(nullptr))
+		{
+			this->get_deleter()(old);
+		}
+
+		return *this;
+	}
+
+	template<typename T2, typename = std::enable_if_t<std::is_convertible<T2, T>::value>>
+	atomic_ptr& operator =(std::unique_ptr<T2, D>&& ptr)
+	{
+		if (T* old = base::m_ptr.exchange(ptr.release()))
+		{
+			this->get_deleter()(old);
+		}
+
+		return *this;
+	}
+
+	void swap(std::unique_ptr<T, D>& ptr)
+	{
+		ptr.reset(base::m_ptr.exchange(ptr.release()));
+	}
+
+	std::add_lvalue_reference_t<T> operator *() const
+	{
+		return *base::m_ptr;
+	}
+
+	T* operator ->() const
+	{
+		return base::m_ptr;
+	}
+
+	T* get() const
+	{
+		return base::m_ptr;
+	}
+
+	explicit operator bool() const
+	{
+		return base::m_ptr != nullptr;
+	}
+
+	T* release() const
+	{
+		return base::m_ptr.exchange(0);
+	}
+
+	void reset(T* ptr = nullptr)
+	{
+		if (T* old = base::m_ptr.exchange(ptr))
+		{
+			this->get_deleter()(old);
+		}
+	}
+
+	// Steal the pointer from `ptr`, convert old value to unique_ptr
+	std::unique_ptr<T, D> exchange(std::unique_ptr<T, D>&& ptr)
+	{
+		return std::unique_ptr<T, D>(base::m_ptr.exchange(ptr.release()));
+	}
+
+	// If pointer is null, steal it from `ptr`
+	bool test_and_swap(std::unique_ptr<T, D>&& ptr)
+	{
+		if (base::m_ptr.compare_and_swap_test(nullptr, ptr.get()))
+		{
+			ptr.release();
+			return true;
+		}
+
+		return false;
+	}
+};
+
+template<typename T, typename D>
+class atomic_ptr<T[], D> final : atomic_ptr_base<T[], D>
+{
+	// TODO
+};
diff --git a/Utilities/BEType.h b/Utilities/BEType.h
index d5b51b99c7..be6f012c70 100644
--- a/Utilities/BEType.h
+++ b/Utilities/BEType.h
@@ -910,20 +910,15 @@ template<typename T> using atomic_be_t = atomic_t<be_t<T>>;
 template<typename T> using atomic_le_t = atomic_t<le_t<T>>;
 #endif
 
-namespace fmt
+// Formatting for BE/LE data
+template<typename T, bool Se>
+struct unveil<se_t<T, Se>, void>
 {
-	// Formatting for BE/LE data
-	template<typename T, bool Se>
-	struct unveil<se_t<T, Se>, void>
+	static inline auto get(const se_t<T, Se>& arg)
 	{
-		using result_type = typename unveil<T>::result_type;
-
-		static inline result_type get_value(const se_t<T, Se>& arg)
-		{
-			return unveil<T>::get_value(arg);
-		}
-	};
-}
+		return unveil<T>::get(arg);
+	}
+};
 
 #undef IS_BINARY_COMPARABLE
 #undef IS_INTEGER
diff --git a/Utilities/Config.h b/Utilities/Config.h
index 203e959a30..6d81580d60 100644
--- a/Utilities/Config.h
+++ b/Utilities/Config.h
@@ -1,8 +1,17 @@
 #pragma once
 
+#include "Utilities/types.h"
 #include "Utilities/Atomic.h"
+
+#include <initializer_list>
+#include <exception>
+#include <utility>
+#include <string>
+#include <vector>
 #include <set>
+#include <unordered_map>
 #include <map>
+#include <mutex>
 
 namespace cfg
 {
diff --git a/Utilities/File.cpp b/Utilities/File.cpp
index 7581ccbde4..246735f950 100644
--- a/Utilities/File.cpp
+++ b/Utilities/File.cpp
@@ -2,13 +2,13 @@
 #include "StrFmt.h"
 #include "Macro.h"
 #include "SharedMutex.h"
+
 #include <unordered_map>
+#include <algorithm>
 
 #ifdef _WIN32
 
 #include <cwchar>
-#undef _WIN32_WINNT
-#define _WIN32_WINNT 0x0601
 #include <Windows.h>
 
 static std::unique_ptr<wchar_t[]> to_wchar(const std::string& source)
@@ -97,7 +97,7 @@ namespace fs
 		std::unordered_map<std::string, std::shared_ptr<device_base>> m_map;
 
 	public:
-		std::shared_ptr<device_base> get_device(const std::string& name);
+		std::shared_ptr<device_base> get_device(const std::string& path);
 		std::shared_ptr<device_base> set_device(const std::string& name, const std::shared_ptr<device_base>&);
 	};
 
@@ -109,11 +109,11 @@ namespace fs
 	}
 }
 
-safe_buffers std::shared_ptr<fs::device_base> fs::device_manager::get_device(const std::string& name)
+std::shared_ptr<fs::device_base> fs::device_manager::get_device(const std::string& path)
 {
 	reader_lock lock(m_mutex);
 
-	const auto found = m_map.find(name);
+	const auto found = m_map.find(path.substr(0, path.find_first_of('/', 2)));
 
 	if (found == m_map.end())
 	{
@@ -123,29 +123,27 @@ safe_buffers std::shared_ptr<fs::device_base> fs::device_manager::get_device(con
 	return found->second;
 }
 
-safe_buffers std::shared_ptr<fs::device_base> fs::device_manager::set_device(const std::string& name, const std::shared_ptr<device_base>& device)
+std::shared_ptr<fs::device_base> fs::device_manager::set_device(const std::string& name, const std::shared_ptr<device_base>& device)
 {
-	std::lock_guard<shared_mutex> lock(m_mutex);
+	writer_lock lock(m_mutex);
 
 	return m_map[name] = device;
 }
 
-safe_buffers std::shared_ptr<fs::device_base> fs::get_virtual_device(const std::string& path)
+std::shared_ptr<fs::device_base> fs::get_virtual_device(const std::string& path)
 {
 	// Every virtual device path must have "//" at the beginning
-	if (path.size() > 2 && reinterpret_cast<const u16&>(path.front()) == '//')
+	if (path.size() > 2 && reinterpret_cast<const u16&>(path.front()) == "//"_u16)
 	{
-		return get_device_manager().get_device(path.substr(0, path.find_first_of('/', 2)));
+		return get_device_manager().get_device(path);
 	}
 
 	return nullptr;
 }
 
-safe_buffers std::shared_ptr<fs::device_base> fs::set_virtual_device(const std::string& name, const std::shared_ptr<device_base>& device)
+std::shared_ptr<fs::device_base> fs::set_virtual_device(const std::string& name, const std::shared_ptr<device_base>& device)
 {
-	Expects(name.size() > 2);
-	Expects(name[0] == '/');
-	Expects(name[1] == '/');
+	Expects(name.size() > 2 && name[0] == '/' && name[1] == '/' && name.find('/', 2) == -1);
 
 	return get_device_manager().set_device(name, device);
 }
@@ -1005,17 +1003,17 @@ bool fs::file::open(const std::string& path, mset<open_mode> mode)
 
 fs::file::file(const void* ptr, std::size_t size)
 {
-	class memory_stream final : public file_base
+	class memory_stream : public file_base
 	{
-		u64 m_pos = 0;
-		u64 m_size;
+		u64 m_pos{}; // TODO: read/seek could modify m_pos atomically
+
+		const char* const m_ptr;
+		const u64 m_size;
 
 	public:
-		const char* const ptr;
-
-		memory_stream(const void* ptr, std::size_t size)
-			: m_size(size)
-			, ptr(static_cast<const char*>(ptr))
+		memory_stream(const void* ptr, u64 size)
+			: m_ptr(static_cast<const char*>(ptr))
+			, m_size(size)
 		{
 		}
 
@@ -1034,7 +1032,7 @@ fs::file::file(const void* ptr, std::size_t size)
 			const u64 start = m_pos;
 			const u64 end = seek(count, fs::seek_cur);
 			const u64 read_size = end >= start ? end - start : throw std::logic_error("memory_stream::read(): overflow");
-			std::memcpy(buffer, ptr + start, read_size);
+			std::memcpy(buffer, m_ptr + start, read_size);
 			return read_size;
 		}
 
@@ -1045,10 +1043,10 @@ fs::file::file(const void* ptr, std::size_t size)
 
 		u64 seek(s64 offset, fs::seek_mode whence) override
 		{
-			return m_pos =
-				whence == fs::seek_set ? std::min<u64>(offset, m_size) :
-				whence == fs::seek_cur ? std::min<u64>(offset + m_pos, m_size) :
-				whence == fs::seek_end ? std::min<u64>(offset + m_size, m_size) :
+			return
+				whence == fs::seek_set ? m_pos = std::min<u64>(offset, m_size) :
+				whence == fs::seek_cur ? m_pos = std::min<u64>(offset + m_pos, m_size) :
+				whence == fs::seek_end ? m_pos = std::min<u64>(offset + m_size, m_size) :
 				throw std::logic_error("memory_stream::seek(): invalid whence");
 		}
 
@@ -1061,93 +1059,6 @@ fs::file::file(const void* ptr, std::size_t size)
 	m_file = std::make_unique<memory_stream>(ptr, size);
 }
 
-fs::file::file(std::vector<char>& vec)
-{
-	class vector_stream final : public file_base
-	{
-		u64 m_pos = 0;
-
-	public:
-		std::vector<char>& vec;
-
-		vector_stream(std::vector<char>& vec)
-			: vec(vec)
-		{
-		}
-
-		fs::stat_t stat() override
-		{
-			throw std::logic_error("vector_stream doesn't support stat()");
-		}
-
-		bool trunc(u64 length) override
-		{
-			vec.resize(length);
-			return true;
-		}
-
-		u64 read(void* buffer, u64 count) override
-		{
-			const u64 start = m_pos;
-			const u64 end = seek(count, fs::seek_cur);
-			const u64 read_size = end >= start ? end - start : throw std::logic_error("vector_stream::read(): overflow");
-			std::memcpy(buffer, vec.data() + start, read_size);
-			return read_size;
-		}
-
-		u64 write(const void* buffer, u64 count) override
-		{
-			throw std::logic_error("TODO: vector_stream doesn't support write()");
-		}
-
-		u64 seek(s64 offset, fs::seek_mode whence) override
-		{
-			return m_pos =
-				whence == fs::seek_set ? std::min<u64>(offset, vec.size()) :
-				whence == fs::seek_cur ? std::min<u64>(offset + m_pos, vec.size()) :
-				whence == fs::seek_end ? std::min<u64>(offset + vec.size(), vec.size()) :
-				throw std::logic_error("vector_stream::seek(): invalid whence");
-		}
-
-		u64 size() override
-		{
-			return vec.size();
-		}
-	};
-
-	m_file = std::make_unique<vector_stream>(vec);
-}
-
-//void fs::file_read_map::reset(const file& f)
-//{
-//	reset();
-//
-//	if (f)
-//	{
-//#ifdef _WIN32
-//		const HANDLE handle = ::CreateFileMapping((HANDLE)f.m_fd, NULL, PAGE_READONLY, 0, 0, NULL);
-//		m_ptr = (char*)::MapViewOfFile(handle, FILE_MAP_READ, 0, 0, 0);
-//		m_size = f.size();
-//		::CloseHandle(handle);
-//#else
-//		m_ptr = (char*)::mmap(nullptr, m_size = f.size(), PROT_READ, MAP_SHARED, f.m_fd, 0);
-//		if (m_ptr == (void*)-1) m_ptr = nullptr;
-//#endif
-//	}
-//}
-//
-//void fs::file_read_map::reset()
-//{
-//	if (m_ptr)
-//	{
-//#ifdef _WIN32
-//		::UnmapViewOfFile(m_ptr);
-//#else
-//		::munmap(m_ptr, m_size);
-//#endif
-//	}
-//}
-
 void fs::dir::xnull() const
 {
 	throw std::logic_error("fs::dir is null");
diff --git a/Utilities/File.h b/Utilities/File.h
index 7a0605b872..52d82f800a 100644
--- a/Utilities/File.h
+++ b/Utilities/File.h
@@ -161,9 +161,6 @@ namespace fs
 		// Open memory for read
 		explicit file(const void* ptr, std::size_t size);
 
-		// Open vector
-		explicit file(std::vector<char>& vec);
-
 		// Check whether the handle is valid (opened file)
 		explicit operator bool() const
 		{
@@ -214,7 +211,7 @@ namespace fs
 			return m_file->write(buffer, count);
 		}
 
-		// Change current position, returns previous position
+		// Change current position, returns resulting position
 		u64 seek(s64 offset, seek_mode whence = seek_set) const
 		{
 			if (!m_file) xnull();
diff --git a/Utilities/Log.cpp b/Utilities/Log.cpp
index 08d6d7197e..66600ac33c 100644
--- a/Utilities/Log.cpp
+++ b/Utilities/Log.cpp
@@ -1,4 +1,5 @@
 #include "Log.h"
+#include <cstdarg>
 
 namespace _log
 {
@@ -23,9 +24,12 @@ namespace _log
 	thread_local std::string(*g_tls_make_prefix)(const channel&, level, const std::string&) = nullptr;
 }
 
-void _log::broadcast(const _log::channel& ch, _log::level sev, const std::string& text)
+void _log::channel::broadcast(const _log::channel& ch, _log::level sev, const char* fmt...)
 {
-	get_logger().log(ch, sev, text);
+	va_list args;
+	va_start(args, fmt);
+	get_logger().log(ch, sev, fmt::_vformat(fmt, args));
+	va_end(args);
 }
 
 [[noreturn]] extern void catch_all_exceptions();
diff --git a/Utilities/Log.h b/Utilities/Log.h
index 5f319ae480..8e662cb42e 100644
--- a/Utilities/Log.h
+++ b/Utilities/Log.h
@@ -22,9 +22,6 @@ namespace _log
 	struct channel;
 	struct listener;
 
-	// Send log message to global logger instance
-	void broadcast(const channel& ch, level sev, const std::string& text);
-
 	// Log channel
 	struct channel
 	{
@@ -42,23 +39,23 @@ namespace _log
 		}
 
 		// Log without formatting
-		force_inline void log(level sev, const std::string& text) const
+		void log(level sev, const std::string& text) const
 		{
 			if (sev <= enabled)
-				broadcast(*this, sev, text);
+				broadcast(*this, sev, "%s", text.c_str());
 		}
 
 		// Log with formatting
 		template<typename... Args>
-		force_inline safe_buffers void format(level sev, const char* fmt, const Args&... args) const
+		void format(level sev, const char* fmt, const Args&... args) const
 		{
 			if (sev <= enabled)
-				broadcast(*this, sev, fmt::format(fmt, fmt::do_unveil(args)...));
+				broadcast(*this, sev, fmt, ::unveil<Args>::get(args)...);
 		}
 
 #define GEN_LOG_METHOD(_sev)\
 		template<typename... Args>\
-		force_inline void _sev(const char* fmt, const Args&... args) const\
+		void _sev(const char* fmt, const Args&... args) const\
 		{\
 			return format<Args...>(level::_sev, fmt, args...);\
 		}
@@ -72,6 +69,10 @@ namespace _log
 		GEN_LOG_METHOD(trace)
 
 #undef GEN_LOG_METHOD
+
+	private:
+		// Send log message to global logger instance
+		static void broadcast(const channel& ch, level sev, const char* fmt...);
 	};
 
 	// Log listener (destination)
diff --git a/Utilities/Platform.h b/Utilities/Platform.h
index 18d9297e24..557b2eb0c7 100644
--- a/Utilities/Platform.h
+++ b/Utilities/Platform.h
@@ -13,25 +13,28 @@
 #include <x86intrin.h>
 #endif
 
+#ifdef _MSC_VER
+#define ASSUME(cond) __assume(cond)
+#define LIKELY(cond) (cond)
+#define UNLIKELY(cond) (cond)
+#else
+#define ASSUME(cond) do { if (!(cond)) __builtin_unreachable(); } while (0)
+#define LIKELY(cond) __builtin_expect(!!(cond), 1)
+#define UNLIKELY(cond) __builtin_expect(!!(cond), 0)
+#endif
+
 // Some platforms don't support thread_local well yet.
 #ifndef _MSC_VER
 #define thread_local __thread
-#define __assume(cond) do { if (!(cond)) __builtin_unreachable(); } while (0)
 #endif
 
-#if defined(_MSC_VER)
-#define safe_buffers __declspec(safebuffers)
-#else
-#define safe_buffers
-#endif
-
-#if defined(_MSC_VER)
+#ifdef _MSC_VER
 #define never_inline __declspec(noinline)
 #else
 #define never_inline __attribute__((noinline))
 #endif
 
-#if defined(_MSC_VER)
+#ifdef _MSC_VER
 #define force_inline __forceinline
 #else
 #define force_inline __attribute__((always_inline)) inline
diff --git a/Utilities/Semaphore.h b/Utilities/Semaphore.h
index ea3c2c4e5c..c5d3550dd7 100644
--- a/Utilities/Semaphore.h
+++ b/Utilities/Semaphore.h
@@ -1,5 +1,8 @@
 #pragma once
 
+#include "types.h"
+#include "Atomic.h"
+
 class semaphore_t
 {
 	// semaphore mutex
@@ -34,4 +37,4 @@ public:
 	void wait();
 
 	bool post_and_wait();
-};
\ No newline at end of file
+};
diff --git a/Utilities/SharedMutex.cpp b/Utilities/SharedMutex.cpp
new file mode 100644
index 0000000000..7b0e20bcd6
--- /dev/null
+++ b/Utilities/SharedMutex.cpp
@@ -0,0 +1,157 @@
+#include "SharedMutex.h"
+
+#include <mutex>
+#include <condition_variable>
+
+struct shared_mutex::internal
+{
+	std::mutex mutex;
+
+	std::size_t rq_size{}; // Reader queue size (threads waiting on m_rcv)
+	std::size_t wq_size{}; // Writer queue size (threads waiting on m_wcv and m_ocv)
+
+	std::condition_variable rcv; // Reader queue
+	std::condition_variable wcv; // Writer queue
+	std::condition_variable ocv; // For current exclusive owner
+};
+
+shared_mutex::~shared_mutex()
+{
+	delete m_data;
+}
+
+void shared_mutex::initialize_once()
+{
+	if (!m_data)
+	{
+		auto ptr = new shared_mutex::internal;
+
+		if (!m_data.compare_and_swap_test(nullptr, ptr))
+		{
+			delete ptr;
+		}
+	}
+}
+
+void shared_mutex::lock_shared_hard()
+{
+	initialize_once();
+
+	std::unique_lock<std::mutex> lock(m_data->mutex);
+
+	// Validate
+	if ((m_ctrl & SM_INVALID_BIT) != 0) throw std::runtime_error("shared_mutex::lock_shared(): Invalid bit");
+	if ((m_ctrl & SM_READER_MASK) == 0) throw std::runtime_error("shared_mutex::lock_shared(): No readers");
+
+	// Notify non-zero reader queue size
+	m_ctrl |= SM_WAITERS_BIT, m_data->rq_size++;
+
+	// Fix excess reader count
+	if ((--m_ctrl & SM_READER_MASK) == 0 && m_data->wq_size)
+	{
+		// Notify exclusive owner
+		m_data->ocv.notify_one();
+	}
+
+	// Obtain the reader lock
+	while (true)
+	{
+		const auto ctrl = m_ctrl.load();
+
+		// Check writers and reader limit
+		if (m_data->wq_size || (ctrl & ~SM_WAITERS_BIT) >= SM_READER_MAX)
+		{
+			m_data->rcv.wait(lock);
+			continue;
+		}
+
+		if (m_ctrl.compare_and_swap_test(ctrl, ctrl + 1))
+		{
+			break;
+		}
+	}
+
+	if (!--m_data->rq_size && !m_data->wq_size)
+	{
+		m_ctrl &= ~SM_WAITERS_BIT;
+	}
+}
+
+void shared_mutex::unlock_shared_notify()
+{
+	initialize_once();
+
+	// Mutex is locked for reliable notification because m_ctrl has been changed outside
+	std::lock_guard<std::mutex> lock(m_data->mutex);
+
+	if ((m_ctrl & SM_READER_MASK) == 0 && m_data->wq_size)
+	{
+		// Notify exclusive owner
+		m_data->ocv.notify_one();
+	}
+	else if (m_data->rq_size)
+	{
+		// Notify other readers
+		m_data->rcv.notify_one();
+	}
+}
+
+void shared_mutex::lock_hard()
+{
+	initialize_once();
+
+	std::unique_lock<std::mutex> lock(m_data->mutex);
+
+	// Validate
+	if ((m_ctrl & SM_INVALID_BIT) != 0) throw std::runtime_error("shared_mutex::lock(): Invalid bit");
+
+	// Notify non-zero writer queue size
+	m_ctrl |= SM_WAITERS_BIT, m_data->wq_size++;
+
+	// Obtain the writer lock
+	while (true)
+	{
+		const auto ctrl = m_ctrl.load();
+
+		if (ctrl & SM_WRITER_LOCK)
+		{
+			m_data->wcv.wait(lock);
+			continue;
+		}
+
+		if (m_ctrl.compare_and_swap_test(ctrl, ctrl | SM_WRITER_LOCK))
+		{
+			break;
+		}
+	}
+
+	// Wait for remaining readers
+	while ((m_ctrl & SM_READER_MASK) != 0)
+	{
+		m_data->ocv.wait(lock);
+	}
+
+	if (!--m_data->wq_size && !m_data->rq_size)
+	{
+		m_ctrl &= ~SM_WAITERS_BIT;
+	}
+}
+
+void shared_mutex::unlock_notify()
+{
+	initialize_once();
+
+	// Mutex is locked for reliable notification because m_ctrl has been changed outside
+	std::lock_guard<std::mutex> lock(m_data->mutex);
+
+	if (m_data->wq_size)
+	{
+		// Notify next exclusive owner
+		m_data->wcv.notify_one();
+	}
+	else if (m_data->rq_size)
+	{
+		// Notify all readers
+		m_data->rcv.notify_all();
+	}
+}
diff --git a/Utilities/SharedMutex.h b/Utilities/SharedMutex.h
index 361f94f2ee..be66f75a94 100644
--- a/Utilities/SharedMutex.h
+++ b/Utilities/SharedMutex.h
@@ -1,22 +1,16 @@
 #pragma once
 
-#include <cstdint>
-#include <exception>
-#include <thread>
-#include <mutex>
-#include <condition_variable>
-
+#include "types.h"
 #include "Atomic.h"
+#include "Platform.h"
 
 //! An attempt to create effective implementation of "shared mutex", lock-free in optimistic case.
-//! All locking and unlocking may be done by single LOCK XADD or LOCK CMPXCHG instructions.
+//! All locking and unlocking may be done by a single LOCK XADD or LOCK CMPXCHG instruction.
 //! MSVC implementation of std::shared_timed_mutex seems suboptimal.
 //! std::shared_mutex is not available until C++17.
 class shared_mutex final
 {
-	using ctrl_type = u32;
-
-	enum : ctrl_type
+	enum : u32
 	{
 		SM_WRITER_LOCK = 1u << 31, // Exclusive lock flag, must be MSB
 		SM_WAITERS_BIT = 1u << 30, // Flag set if m_wq_size or m_rq_size is non-zero
@@ -26,186 +20,79 @@ class shared_mutex final
 		SM_READER_MAX  = 1u << 24, // Max reader count
 	};
 
-	atomic_t<ctrl_type> m_ctrl{}; // Control atomic variable: reader count | SM_* flags
+	atomic_t<u32> m_ctrl{}; // Control variable: reader count | SM_* flags
 
-	std::mutex m_mutex;
+	struct internal;
 
-	std::size_t m_rq_size{}; // Reader queue size (threads waiting on m_rcv)
-	std::size_t m_wq_size{}; // Writer queue size (threads waiting on m_wcv and m_ocv)
-	
-	std::condition_variable m_rcv; // Reader queue
-	std::condition_variable m_wcv; // Writer queue
-	std::condition_variable m_ocv; // For current exclusive owner
+	atomic_t<internal*> m_data{}; // Internal data
 
-	void lock_shared_hard()
-	{
-		std::unique_lock<std::mutex> lock(m_mutex);
+	void lock_shared_hard();
+	void unlock_shared_notify();
 
-		// Validate
-		if ((m_ctrl & SM_INVALID_BIT) != 0) throw std::runtime_error("shared_mutex::lock_shared(): Invalid bit");
-		if ((m_ctrl & SM_READER_MASK) == 0) throw std::runtime_error("shared_mutex::lock_shared(): No readers");
-
-		// Notify non-zero reader queue size
-		m_ctrl |= SM_WAITERS_BIT, m_rq_size++;
-
-		// Fix excess reader count
-		if ((--m_ctrl & SM_READER_MASK) == 0 && m_wq_size)
-		{
-			// Notify exclusive owner
-			m_ocv.notify_one();
-		}
-
-		// Obtain the reader lock
-		while (true)
-		{
-			const auto ctrl = m_ctrl.load();
-
-			// Check writers and reader limit
-			if (m_wq_size || (ctrl & ~SM_WAITERS_BIT) >= SM_READER_MAX)
-			{
-				m_rcv.wait(lock);
-				continue;
-			}
-
-			if (m_ctrl.compare_and_swap_test(ctrl, ctrl + 1))
-			{
-				break;
-			}
-		}
-
-		if (!--m_rq_size && !m_wq_size)
-		{
-			m_ctrl &= ~SM_WAITERS_BIT;
-		}
-	}
-
-	void unlock_shared_notify()
-	{
-		// Mutex is locked for reliable notification because m_ctrl has been changed outside
-		std::lock_guard<std::mutex> lock(m_mutex);
-
-		if ((m_ctrl & SM_READER_MASK) == 0 && m_wq_size)
-		{
-			// Notify exclusive owner
-			m_ocv.notify_one();
-		}
-		else if (m_rq_size)
-		{
-			// Notify other readers
-			m_rcv.notify_one();
-		}
-	}
-
-	void lock_hard()
-	{
-		std::unique_lock<std::mutex> lock(m_mutex);
-
-		// Validate
-		if ((m_ctrl & SM_INVALID_BIT) != 0) throw std::runtime_error("shared_mutex::lock(): Invalid bit");
-
-		// Notify non-zero writer queue size
-		m_ctrl |= SM_WAITERS_BIT, m_wq_size++;
-
-		// Obtain the writer lock
-		while (true)
-		{
-			const auto ctrl = m_ctrl.load();
-
-			if (ctrl & SM_WRITER_LOCK)
-			{
-				m_wcv.wait(lock);
-				continue;
-			}
-
-			if (m_ctrl.compare_and_swap_test(ctrl, ctrl | SM_WRITER_LOCK))
-			{
-				break;
-			}
-		}
-
-		// Wait for remaining readers
-		while ((m_ctrl & SM_READER_MASK) != 0)
-		{
-			m_ocv.wait(lock);
-		}
-
-		if (!--m_wq_size && !m_rq_size)
-		{
-			m_ctrl &= ~SM_WAITERS_BIT;
-		}
-	}
-
-	void unlock_notify()
-	{
-		// Mutex is locked for reliable notification because m_ctrl has been changed outside
-		std::lock_guard<std::mutex> lock(m_mutex);
-
-		if (m_wq_size)
-		{
-			// Notify next exclusive owner
-			m_wcv.notify_one();
-		}
-		else if (m_rq_size)
-		{
-			// Notify all readers
-			m_rcv.notify_all();
-		}
-	}
+	void lock_hard();
+	void unlock_notify();
 
 public:
-	shared_mutex() = default;
+	constexpr shared_mutex() = default;
+
+	~shared_mutex();
+
+	// Initialize internal data
+	void initialize_once();
+
+	bool try_lock_shared()
+	{
+		auto ctrl = m_ctrl.load();
+
+		if (UNLIKELY(ctrl >= SM_READER_MAX))
+		{
+			ctrl = 0;
+		}
+
+		// Weak attempt
+		return LIKELY(m_ctrl.compare_and_swap_test(ctrl, ctrl + 1));
+	}
 
-	// Lock in shared mode
 	void lock_shared()
 	{
-		if (m_ctrl++ >= SM_READER_MAX)
+		// Optimization: unconditional increment, compensated later
+		if (UNLIKELY(m_ctrl++ >= SM_READER_MAX))
 		{
 			lock_shared_hard();
 		}
 	}
 
-	// Try to lock in shared mode
-	bool try_lock_shared()
-	{
-		auto ctrl = m_ctrl.load();
-
-		return ctrl < SM_READER_MAX && m_ctrl.compare_and_swap_test(ctrl, ctrl + 1);
-	}
-
-	// Unlock in shared mode
 	void unlock_shared()
 	{
-		if (m_ctrl-- >= SM_READER_MAX)
+		if (UNLIKELY(m_ctrl-- >= SM_READER_MAX))
 		{
 			unlock_shared_notify();
 		}
 	}
 
-	// Try to lock exclusively
 	bool try_lock()
 	{
-		return m_ctrl.compare_and_swap_test(0, SM_WRITER_LOCK);
+		return LIKELY(m_ctrl.compare_and_swap_test(0, SM_WRITER_LOCK));
 	}
 
-	// Lock exclusively
 	void lock()
 	{
-		if (m_ctrl.compare_and_swap_test(0, SM_WRITER_LOCK)) return;
-
-		lock_hard();		
+		if (UNLIKELY(!try_lock()))
+		{
+			lock_hard();
+		}
 	}
 
-	// Unlock exclusively
 	void unlock()
 	{
-		if (m_ctrl.fetch_sub(SM_WRITER_LOCK) != SM_WRITER_LOCK)
+		if (UNLIKELY(m_ctrl.fetch_sub(SM_WRITER_LOCK) != SM_WRITER_LOCK))
 		{
 			unlock_notify();
 		}
 	}
 };
 
-//! Simplified shared (reader) lock implementation, similar to std::lock_guard.
+//! Simplified shared (reader) lock implementation.
 //! std::shared_lock may be used instead if necessary.
 class reader_lock final
 {
@@ -225,3 +112,24 @@ public:
 		m_mutex.unlock_shared();
 	}
 };
+
+//! Simplified exclusive (writer) lock implementation.
+//! std::lock_guard may or std::unique_lock be used instead if necessary.
+class writer_lock final
+{
+	shared_mutex& m_mutex;
+
+public:
+	writer_lock(const writer_lock&) = delete;
+
+	writer_lock(shared_mutex& mutex)
+		: m_mutex(mutex)
+	{
+		m_mutex.lock();
+	}
+
+	~writer_lock()
+	{
+		m_mutex.unlock();
+	}
+};
diff --git a/Utilities/StrFmt.cpp b/Utilities/StrFmt.cpp
index 1b7df25a4d..3e9b8c1719 100644
--- a/Utilities/StrFmt.cpp
+++ b/Utilities/StrFmt.cpp
@@ -1,6 +1,10 @@
 #include "StrFmt.h"
 #include "BEType.h"
 
+#include <cassert>
+#include <array>
+#include <memory>
+
 std::string v128::to_hex() const
 {
 	return fmt::format("%016llx%016llx", _u64[1], _u64[0]);
@@ -74,6 +78,59 @@ std::string fmt::to_sdec(s64 svalue)
 	return std::string(&res[first], sizeof(res) - first);
 }
 
+std::string fmt::_vformat(const char* fmt, va_list _args) noexcept
+{
+	// Fixed stack buffer for the first attempt
+	std::array<char, 4096> fixed_buf;
+
+	// Possibly dynamically allocated buffer for the second attempt
+	std::unique_ptr<char[]> buf;
+
+	// Pointer to the current buffer
+	char* buf_addr = fixed_buf.data();
+
+	for (std::size_t buf_size = fixed_buf.size();;)
+	{
+		va_list args;
+		va_copy(args, _args);
+
+#ifndef _MSC_VER
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wformat-security"
+#endif
+		const std::size_t len = std::vsnprintf(buf_addr, buf_size, fmt, args);
+#ifndef _MSC_VER
+#pragma GCC diagnostic pop
+#endif
+		va_end(args);
+
+		assert(len <= INT_MAX);
+
+		if (len < buf_size)
+		{
+			return{ buf_addr, len };
+		}
+
+		buf.reset(buf_addr = new char[buf_size = len + 1]);
+	}
+}
+
+std::string fmt::_format(const char* fmt...) noexcept
+{
+	va_list args;
+	va_start(args, fmt);
+	auto result = fmt::_vformat(fmt, args);
+	va_end(args);
+
+	return result;
+}
+
+fmt::exception_base::exception_base(const char* fmt...)
+	: std::runtime_error((va_start(m_args, fmt), _vformat(fmt, m_args)))
+{
+	va_end(m_args);
+}
+
 std::string fmt::replace_first(const std::string& src, const std::string& from, const std::string& to)
 {
 	auto pos = src.find(from);
diff --git a/Utilities/StrFmt.h b/Utilities/StrFmt.h
index d88a4a7cef..a44c73b300 100644
--- a/Utilities/StrFmt.h
+++ b/Utilities/StrFmt.h
@@ -1,21 +1,16 @@
 #pragma once
 
-#include <array>
+#include <cstdarg>
 #include <string>
 #include <vector>
 #include <functional>
-#include <memory>
 
 #include "Platform.h"
 #include "types.h"
 
-#if defined(_MSC_VER) && _MSC_VER <= 1800
-#define snprintf _snprintf
-#endif
-
 // Copy null-terminated string from std::string to char array with truncation
 template<std::size_t N>
-inline void strcpy_trunc(char(&dst)[N], const std::string& src)
+force_inline void strcpy_trunc(char(&dst)[N], const std::string& src)
 {
 	const std::size_t count = src.size() >= N ? N - 1 : src.size();
 	std::memcpy(dst, src.c_str(), count);
@@ -24,13 +19,33 @@ inline void strcpy_trunc(char(&dst)[N], const std::string& src)
 
 // Copy null-terminated string from char array to another char array with truncation
 template<std::size_t N, std::size_t N2>
-inline void strcpy_trunc(char(&dst)[N], const char(&src)[N2])
+force_inline void strcpy_trunc(char(&dst)[N], const char(&src)[N2])
 {
 	const std::size_t count = N2 >= N ? N - 1 : N2;
 	std::memcpy(dst, src, count);
 	dst[count] = '\0';
 }
 
+// Formatting helper, type-specific preprocessing for improving safety and functionality
+template<typename T, typename>
+struct unveil
+{
+	// TODO
+	static inline const T& get(const T& arg)
+	{
+		return arg;
+	}
+};
+
+template<>
+struct unveil<std::string, void>
+{
+	static inline const char* get(const std::string& arg)
+	{
+		return arg.c_str();
+	}
+};
+
 namespace fmt
 {
 	std::string replace_first(const std::string& src, const std::string& from, const std::string& to);
@@ -87,125 +102,45 @@ namespace fmt
 	std::string to_hex(u64 value, u64 count = 1);
 	std::string to_udec(u64 value);
 	std::string to_sdec(s64 value);
-
-	template<typename T, typename>
-	struct unveil
-	{
-		using result_type = T;
-
-		force_inline static result_type get_value(const T& arg)
-		{
-			return arg;
-		}
-	};
-
-	template<>
-	struct unveil<const char*, void>
-	{
-		using result_type = const char* const;
-
-		static result_type get_value(const char* const& arg)
-		{
-			return arg;
-		}
-	};
-
-	template<std::size_t N>
-	struct unveil<char[N], void>
-	{
-		using result_type = const char* const;
-
-		static result_type get_value(const char(&arg)[N])
-		{
-			return arg;
-		}
-	};
-
-	template<>
-	struct unveil<std::string, void>
-	{
-		using result_type = const char*;
-
-		static result_type get_value(const std::string& arg)
-		{
-			return arg.c_str();
-		}
-	};
-
-	template<typename T>
-	struct unveil<T, std::enable_if_t<std::is_enum<T>::value>>
-	{
-		using result_type = std::underlying_type_t<T>;
-
-		force_inline static result_type get_value(const T& arg)
-		{
-			return static_cast<result_type>(arg);
-		}
-	};
-
-	template<typename T>
-	force_inline typename unveil<T>::result_type do_unveil(const T& arg)
-	{
-		return unveil<T>::get_value(arg);
-	}
+	std::string _format(const char* fmt...) noexcept;
+	std::string _vformat(const char*, va_list) noexcept;
 
 	// Formatting function with special functionality (fmt::unveil)
 	template<typename... Args>
-	safe_buffers std::string format(const char* fmt, const Args&... args)
+	force_inline std::string format(const char* fmt, const Args&... args) noexcept
 	{
-		// fixed stack buffer for the first attempt
-		std::array<char, 4096> fixed_buf;
+		return _format(fmt, ::unveil<Args>::get(args)...);
+	}
 
-		// possibly dynamically allocated buffer for the second attempt
-		std::unique_ptr<char[]> buf;
+	// Helper class
+	class exception_base : public std::runtime_error
+	{
+		// Helper (there is no other room)
+		va_list m_args;
 
-		// pointer to the current buffer
-		char* buf_addr = fixed_buf.data();
+	protected:
+		// Internal formatting constructor
+		exception_base(const char* fmt...);
+	};
 
-		for (std::size_t buf_size = fixed_buf.size();;)
+	// Exception type derived from std::runtime_error with formatting constructor
+	class exception : public exception_base
+	{
+	public:
+		// Formatting constructor
+		template<typename... Args>
+		exception(const char* fmt, const Args&... args)
+			: exception_base(fmt, ::unveil<Args>::get(args)...)
 		{
-#ifndef _MSC_VER
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wformat-security"
-#endif
-			const std::size_t len = std::snprintf(buf_addr, buf_size, fmt, do_unveil(args)...);
-#ifndef _MSC_VER
-#pragma GCC diagnostic pop
-#endif
-			if (len > INT_MAX)
-			{
-				throw std::runtime_error("std::snprintf() failed");
-			}
-
-			if (len < buf_size)
-			{
-				return{ buf_addr, len };
-			}
-
-			buf.reset(buf_addr = new char[buf_size = len + 1]);
 		}
-	}
-
-	// Create exception of type T (std::runtime_error by default) with formatting
-	template<typename T = std::runtime_error, typename... Args>
-	never_inline safe_buffers T exception(const char* fmt, const Args&... args) noexcept(noexcept(T{ fmt }))
-	{
-		return T{ format(fmt, do_unveil(args)...).c_str() };
-	}
-
-	// Create exception of type T (std::runtime_error by default) without formatting
-	template<typename T = std::runtime_error>
-	safe_buffers T exception(const char* msg) noexcept(noexcept(T{ msg }))
-	{
-		return T{ msg };
-	}
+	};
 
 	// Narrow cast (similar to gsl::narrow) with exception message formatting
 	template<typename To, typename From, typename... Args>
 	inline auto narrow(const char* format_str, const From& value, const Args&... args) -> decltype(static_cast<To>(static_cast<From>(std::declval<To>())))
 	{
 		const auto result = static_cast<To>(value);
-		if (static_cast<From>(result) != value) throw fmt::exception(format_str, fmt::do_unveil(value), fmt::do_unveil(args)...);
+		if (static_cast<From>(result) != value) throw fmt::exception(format_str, value, args...);
 		return result;
 	}
 
diff --git a/Utilities/Thread.cpp b/Utilities/Thread.cpp
index 1fd83ba5b0..ac9cf21b48 100644
--- a/Utilities/Thread.cpp
+++ b/Utilities/Thread.cpp
@@ -52,38 +52,6 @@ static void report_fatal_error(const std::string& msg)
 	std::abort();
 }
 
-void SetCurrentThreadDebugName(const char* threadName)
-{
-#if defined(_MSC_VER) // this is VS-specific way to set thread names for the debugger
-
-	#pragma pack(push,8)
-
-	struct THREADNAME_INFO
-	{
-		DWORD dwType;
-		LPCSTR szName;
-		DWORD dwThreadID;
-		DWORD dwFlags;
-	} info;
-
-	#pragma pack(pop)
-
-	info.dwType = 0x1000;
-	info.szName = threadName;
-	info.dwThreadID = -1;
-	info.dwFlags = 0;
-
-	__try
-	{
-		RaiseException(0x406D1388, 0, sizeof(info) / sizeof(ULONG_PTR), (ULONG_PTR*)&info);
-	}
-	__except (EXCEPTION_EXECUTE_HANDLER)
-	{
-	}
-
-#endif
-}
-
 enum x64_reg_t : u32
 {
 	X64R_RAX = 0,
@@ -1295,34 +1263,204 @@ const bool s_self_test = []() -> bool
 	return true;
 }();
 
+#include <mutex>
+#include <condition_variable>
+#include <exception>
+
 thread_local DECLARE(thread_ctrl::g_tls_this_thread) = nullptr;
 
+struct thread_ctrl::internal
+{
+	std::mutex mutex;
+	std::condition_variable cond;
+	std::condition_variable join; // Allows simultaneous joining
+
+	task_stack atexit;
+
+	std::exception_ptr exception; // Caught exception
+};
+
+// Temporarily until better interface is implemented
+extern std::condition_variable& get_current_thread_cv()
+{
+	return thread_ctrl::get_current()->get_data()->cond;
+}
+
+extern std::mutex& get_current_thread_mutex()
+{
+	return thread_ctrl::get_current()->get_data()->mutex;
+}
+
 // TODO
-atomic_t<u32> g_thread_count{ 0 };
+extern atomic_t<u32> g_thread_count(0);
 
 void thread_ctrl::initialize()
 {
-	SetCurrentThreadDebugName(g_tls_this_thread->m_name.c_str());
+	// Initialize TLS variable
+	g_tls_this_thread = this;
+
+#if defined(_MSC_VER)
+
+	struct THREADNAME_INFO
+	{
+		DWORD dwType;
+		LPCSTR szName;
+		DWORD dwThreadID;
+		DWORD dwFlags;
+	};
+
+	// Set thread name for VS debugger
+	if (IsDebuggerPresent())
+	{
+		THREADNAME_INFO info;
+		info.dwType = 0x1000;
+		info.szName = m_name.c_str();
+		info.dwThreadID = -1;
+		info.dwFlags = 0;
+
+		__try
+		{
+			RaiseException(0x406D1388, 0, sizeof(info) / sizeof(ULONG_PTR), (ULONG_PTR*)&info);
+		}
+		__except (EXCEPTION_EXECUTE_HANDLER)
+		{
+		}
+	}
+
+#endif
 
 	_log::g_tls_make_prefix = [](const auto&, auto, const auto&)
 	{
 		return g_tls_this_thread->m_name;
 	};
 
-	// TODO
 	++g_thread_count;
 }
 
+void thread_ctrl::set_exception() noexcept
+{
+	initialize_once();
+	m_data->exception = std::current_exception();
+}
+
 void thread_ctrl::finalize() noexcept
 {
 	// TODO
 	vm::reservation_free();
 
-	// TODO
+	// Call atexit functions
+	if (m_data) m_data->atexit.exec();
+
 	--g_thread_count;
 
-	// Call atexit functions
-	g_tls_this_thread->m_atexit.exec();
+#ifdef _MSC_VER
+	ULONG64 time;
+	QueryThreadCycleTime(m_thread.native_handle(), &time);
+	LOG_NOTICE(GENERAL, "Thread time: %f Gc", time / 1000000000.);
+#endif
+}
+
+task_stack& thread_ctrl::get_atexit() const
+{
+	initialize_once();
+	return m_data->atexit;
+}
+
+thread_ctrl::~thread_ctrl()
+{
+	if (m_thread.joinable())
+	{
+		m_thread.detach();
+	}
+
+	delete m_data;
+}
+
+void thread_ctrl::initialize_once() const
+{
+	if (!m_data)
+	{
+		auto ptr = new thread_ctrl::internal;
+
+		if (!m_data.compare_and_swap_test(nullptr, ptr))
+		{
+			delete ptr;
+		}
+	}
+}
+
+void thread_ctrl::join()
+{
+	if (m_thread.joinable())
+	{
+		// Increase contention counter
+		if (m_joining++)
+		{
+			// Hard way
+			initialize_once();
+
+			std::unique_lock<std::mutex> lock(m_data->mutex);
+			m_data->join.wait(lock, WRAP_EXPR(!m_thread.joinable()));
+		}
+		else
+		{
+			// Winner joins the thread
+			m_thread.join();
+
+			// Notify others if necessary
+			if (m_joining > 1)
+			{
+				initialize_once();
+
+				// Serialize for reliable notification
+				m_data->mutex.lock();
+				m_data->mutex.unlock();
+				m_data->join.notify_all();
+			}
+		}
+	}
+
+	if (m_data && m_data->exception)
+	{
+		std::rethrow_exception(m_data->exception);
+	}
+}
+
+void thread_ctrl::lock_notify() const
+{
+	if (UNLIKELY(g_tls_this_thread == this))
+	{
+		return;
+	}
+
+	initialize_once();
+
+	// Serialize for reliable notification, condition is assumed to be changed externally
+	m_data->mutex.lock();
+	m_data->mutex.unlock();
+	m_data->cond.notify_one();
+}
+
+void thread_ctrl::notify() const
+{
+	initialize_once();
+	m_data->cond.notify_one();
+}
+
+thread_ctrl::internal* thread_ctrl::get_data() const
+{
+	initialize_once();
+	return m_data;
+}
+
+
+named_thread::named_thread()
+{
+}
+
+named_thread::~named_thread()
+{
+	LOG_TRACE(GENERAL, "%s", __func__);
 }
 
 std::string named_thread::get_name() const
@@ -1332,8 +1470,6 @@ std::string named_thread::get_name() const
 
 void named_thread::start()
 {
-	Expects(!m_thread);
-
 	// Get shared_ptr instance (will throw if called from the constructor or the object has been created incorrectly)
 	auto&& ptr = shared_from_this();
 
@@ -1359,19 +1495,3 @@ void named_thread::start()
 		thread->on_exit();
 	});
 }
-
-void named_thread::join()
-{
-	Expects(m_thread);
-
-	try
-	{
-		m_thread->join();
-		m_thread.reset();
-	}
-	catch (...)
-	{
-		m_thread.reset();
-		throw;
-	}
-}
diff --git a/Utilities/Thread.h b/Utilities/Thread.h
index f8d098dfda..2bf7b63566 100644
--- a/Utilities/Thread.h
+++ b/Utilities/Thread.h
@@ -1,13 +1,11 @@
 #pragma once
 
-#include <exception>
 #include <string>
 #include <memory>
 #include <thread>
-#include <mutex>
-#include <condition_variable>
 
 #include "Platform.h"
+#include "Atomic.h"
 
 // Will report exception and call std::abort() if put in catch(...)
 [[noreturn]] void catch_all_exceptions();
@@ -32,12 +30,6 @@ class task_stack
 
 	std::unique_ptr<task_base> m_stack;
 
-	never_inline void push(std::unique_ptr<task_base> task)
-	{
-		m_stack.swap(task->next);
-		m_stack.swap(task);
-	}
-
 public:
 	template<typename F>
 	void push(F&& func)
@@ -58,7 +50,16 @@ public:
 			}
 		};
 
-		return push(std::unique_ptr<task_base>{ new task_t(std::forward<F>(func)) });
+		auto _top = new task_t(std::forward<F>(func));
+		auto _next = m_stack.release();
+		m_stack.reset(_top);
+#ifndef _MSC_VER
+		_top->next.reset(_next);
+#else
+		auto& next = _top->next;
+		next.release();
+		next.reset(_next);
+#endif
 	}
 
 	void reset()
@@ -78,25 +79,33 @@ public:
 // Thread control class
 class thread_ctrl final
 {
+	struct internal;
+
 	static thread_local thread_ctrl* g_tls_this_thread;
 
+	// Thread handle
+	std::thread m_thread;
+
+	// Thread join contention counter
+	atomic_t<uint> m_joining{};
+
 	// Fixed name
 	std::string m_name;
 
-	// Thread handle (be careful)
-	std::thread m_thread;
-
-	// Thread result (exception)
-	std::exception_ptr m_exception;
-
-	// Functions scheduled at thread exit
-	task_stack m_atexit;
+	// Thread internals
+	mutable atomic_t<internal*> m_data{};
 
 	// Called at the thread start
-	static void initialize();
+	void initialize();
+
+	// Set std::current_exception
+	void set_exception() noexcept;
 
 	// Called at the thread end
-	static void finalize() noexcept;
+	void finalize() noexcept;
+
+	// Get atexit function
+	task_stack& get_atexit() const;
 
 public:
 	template<typename N>
@@ -108,13 +117,7 @@ public:
 	// Disable copy/move constructors and operators
 	thread_ctrl(const thread_ctrl&) = delete;
 
-	~thread_ctrl()
-	{
-		if (m_thread.joinable())
-		{
-			m_thread.detach();
-		}
-	}
+	~thread_ctrl();
 
 	// Get thread name
 	const std::string& get_name() const
@@ -122,19 +125,20 @@ public:
 		return m_name;
 	}
 
-	// Get thread result (may throw)
-	void join()
-	{
-		if (m_thread.joinable())
-		{
-			m_thread.join();
-		}
+	// Initialize internal data
+	void initialize_once() const;
 
-		if (auto&& e = std::move(m_exception))
-		{
-			std::rethrow_exception(e);
-		}
-	}
+	// Get thread result (may throw, simultaneous joining allowed)
+	void join();
+
+	// Lock, unlock, notify the thread (required if the condition changed locklessly)
+	void lock_notify() const;
+
+	// Notify the thread, beware the condition change
+	void notify() const;
+
+	//
+	internal* get_data() const;
 
 	// Get current thread (may be nullptr)
 	static const thread_ctrl* get_current()
@@ -146,34 +150,30 @@ public:
 	template<typename F>
 	static inline void at_exit(F&& func)
 	{
-		return g_tls_this_thread->m_atexit.push(std::forward<F>(func));
+		return g_tls_this_thread->get_atexit().push(std::forward<F>(func));
 	}
 
 	// Named thread factory
-	template<typename N, typename F>
-	static inline std::shared_ptr<thread_ctrl> spawn(N&& name, F&& func)
+	template<typename N, typename F, typename... Args>
+	static inline std::shared_ptr<thread_ctrl> spawn(N&& name, F&& func, Args&&... args)
 	{
 		auto ctrl = std::make_shared<thread_ctrl>(std::forward<N>(name));
 
-		ctrl->m_thread = std::thread([ctrl, task = std::forward<F>(func)]()
+		ctrl->m_thread = std::thread([ctrl, task = std::forward<F>(func)](Args&&... args)
 		{
-			// Initialize TLS variable
-			g_tls_this_thread = ctrl.get();
-
 			try
 			{
-				initialize();
-				task();
-				finalize();
+				ctrl->initialize();
+				task(std::forward<Args>(args)...);
 			}
 			catch (...)
 			{
-				finalize();
-
-				// Set exception
-				ctrl->m_exception = std::current_exception();
+				ctrl->set_exception();
 			}
-		});
+
+			ctrl->finalize();
+
+		}, std::forward<Args>(args)...);
 
 		return ctrl;
 	}
@@ -185,21 +185,27 @@ class named_thread : public std::enable_shared_from_this<named_thread>
 	std::shared_ptr<thread_ctrl> m_thread;
 
 public:
-	// Thread condition variable for external use (this thread waits on it, other threads may notify)
-	std::condition_variable cv;
+	named_thread();
 
-	// Thread mutex for external use (can be used with `cv`)
-	std::mutex mutex;
+	virtual ~named_thread();
 
-	// Lock mutex, notify condition variable
-	void safe_notify()
-	{
-		// Lock for reliable notification, condition is assumed to be changed externally
-		std::unique_lock<std::mutex> lock(mutex);
+	// Deleted copy/move constructors + copy/move operators
+	named_thread(const named_thread&) = delete;
 
-		cv.notify_one();
-	}
+	// Get thread name
+	virtual std::string get_name() const;
 
+protected:
+	// Start thread (cannot be called from the constructor: should throw bad_weak_ptr in such case)
+	void start();
+
+	// Thread task (called in the thread)
+	virtual void on_task() = 0;
+
+	// Thread finalization (called after on_task)
+	virtual void on_exit() {}
+
+public:
 	// ID initialization
 	virtual void on_init()
 	{
@@ -209,43 +215,25 @@ public:
 	// ID finalization
 	virtual void on_stop()
 	{
-		join();
+		m_thread->join();
 	}
 
-protected:
-	// Thread task (called in the thread)
-	virtual void on_task() = 0;
-
-	// Thread finalization (called after on_task)
-	virtual void on_exit() {}
-
-public:
-	named_thread() = default;
-
-	virtual ~named_thread() = default;
-
-	// Deleted copy/move constructors + copy/move operators
-	named_thread(const named_thread&) = delete;
-
-	// Get thread name
-	virtual std::string get_name() const;
-
-	// Start thread (cannot be called from the constructor: should throw bad_weak_ptr in such case)
-	void start();
-
-	// Join thread (get thread result)
-	void join();
-
 	// Get thread_ctrl
-	const thread_ctrl* get_thread_ctrl() const
+	const thread_ctrl* operator->() const
 	{
 		return m_thread.get();
 	}
 
-	// Compare with the current thread
-	bool is_current() const
+	// Lock mutex, notify condition variable
+	void lock_notify() const
 	{
-		return m_thread && thread_ctrl::get_current() == m_thread.get();
+		m_thread->lock_notify();
+	}
+
+	// Notify condition variable
+	void notify() const
+	{
+		m_thread->notify();
 	}
 };
 
diff --git a/Utilities/types.h b/Utilities/types.h
index 78facc95b7..d8d32f1778 100644
--- a/Utilities/types.h
+++ b/Utilities/types.h
@@ -79,11 +79,8 @@ struct atomic_test_and_complement;
 template<typename T>
 class atomic_t;
 
-namespace fmt
-{
-	template<typename T, typename = void>
-	struct unveil;
-}
+template<typename T, typename = void>
+struct unveil;
 
 // TODO: replace with std::void_t when available
 namespace void_details
@@ -409,6 +406,38 @@ struct ignore
 	}
 };
 
+// Contains value of any POD type with fixed size and alignment. TT<> is the type converter applied.
+// For example, `simple_t` may be used to remove endianness.
+template<template<typename> class TT, std::size_t S, std::size_t A = S>
+struct alignas(A) any_pod
+{
+	enum class byte : char {} data[S];
+
+	any_pod() = default;
+
+	template<typename T, typename T2 = TT<T>, typename = std::enable_if_t<std::is_pod<T2>::value && sizeof(T2) == S && alignof(T2) <= A>>
+	any_pod(const T& value)
+	{
+		reinterpret_cast<T2&>(data) = value;
+	}
+
+	template<typename T, typename T2 = TT<T>, typename = std::enable_if_t<std::is_pod<T2>::value && sizeof(T2) == S && alignof(T2) <= A>>
+	T2& as()
+	{
+		return reinterpret_cast<T2&>(data);
+	}
+
+	template<typename T, typename T2 = TT<T>, typename = std::enable_if_t<std::is_pod<T2>::value && sizeof(T2) == S && alignof(T2) <= A>>
+	const T2& as() const
+	{
+		return reinterpret_cast<const T2&>(data);
+	}
+};
+
+using any16 = any_pod<simple_t, sizeof(u16)>;
+using any32 = any_pod<simple_t, sizeof(u32)>;
+using any64 = any_pod<simple_t, sizeof(u64)>;
+
 // Allows to define integer convertible to multiple enum types
 template<typename T = void, typename... Ts>
 struct multicast : multicast<Ts...>
diff --git a/Vulkan/CMakeLists.txt b/Vulkan/CMakeLists.txt
index d305f033c7..099905a96a 100644
--- a/Vulkan/CMakeLists.txt
+++ b/Vulkan/CMakeLists.txt
@@ -1,4 +1,5 @@
-if (NOT APPLE)
+if(APPLE OR WIN32 AND NOT MSVC)
+else()
 add_subdirectory( glslang )
 set(BUILD_TESTS OFF CACHE BOOL "Build tests" FORCE)
 set(BUILD_DEMOS OFF CACHE BOOL "Build demos" FORCE)
@@ -6,4 +7,4 @@ set(BUILD_DEMOS OFF CACHE BOOL "Build demos" FORCE)
 set(BUILD_LAYERS OFF CACHE BOOL "Build demos" FORCE)
 set(BUILD_VKJSON OFF CACHE BOOL "Build demos" FORCE)
 add_subdirectory( Vulkan-LoaderAndValidationLayers )
-endif()
\ No newline at end of file
+endif()
diff --git a/rpcs3/CMakeLists.txt b/rpcs3/CMakeLists.txt
index a84ea142f2..bd631662f2 100644
--- a/rpcs3/CMakeLists.txt
+++ b/rpcs3/CMakeLists.txt
@@ -23,6 +23,15 @@ elseif("${CMAKE_CXX_COMPILER_ID}" MATCHES "Clang")
 	if(APPLE)
 		add_compile_options(-stdlib=libc++)
 	endif()
+	if(WIN32)
+		add_compile_options(-pthread)
+		add_compile_options(-D__GXX_ABI_VERSION=1009)
+	endif()
+endif()
+
+if(WIN32)
+	add_definitions(-DUNICODE)
+	add_definitions(-D_WIN32_WINNT=0x0601)
 endif()
 
 if(NOT MSVC)
@@ -32,11 +41,8 @@ if(NOT MSVC)
 	endif()
 	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14 -fexceptions")
 	add_compile_options(-msse -msse2 -mcx16 -mssse3)
-	if(WIN32)
-		add_compile_options(-municode -static -mwindows)
-	endif()
 else()
-	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /EHa /Zc:throwingNew /D _CRT_SECURE_NO_DEPRECATE=1 /D _CRT_NON_CONFORMING_SWPRINTFS=1 /D _SCL_SECURE_NO_WARNINGS=1")
+	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zc:throwingNew /D _CRT_SECURE_NO_DEPRECATE=1 /D _CRT_NON_CONFORMING_SWPRINTFS=1 /D _SCL_SECURE_NO_WARNINGS=1")
 endif()
 
 if(APPLE)
@@ -88,7 +94,7 @@ if(NOT MSVC)
 endif()
 find_package(OpenGL REQUIRED)
 find_package(OpenAL REQUIRED)
-find_package(LLVM CONFIG)
+#find_package(LLVM CONFIG) # TODO
 
 include("${wxWidgets_USE_FILE}")
 
@@ -141,15 +147,9 @@ else()
 	else()
 		llvm_map_components_to_libnames(LLVM_LIBS mcjit vectorize x86codegen x86disassembler mcdisassembler)
 	endif()
-	if(MSVC)
-		set_source_files_properties(${RPCS3_SRC_DIR}/Emu/Cell/PPULLVMRecompiler.cpp PROPERTIES COMPILE_FLAGS /GR-)
-	else()
-		set_source_files_properties(${RPCS3_SRC_DIR}/Emu/Cell/PPULLVMRecompiler.cpp PROPERTIES COMPILE_FLAGS -fno-rtti)
-	endif()
 endif()
 
 link_directories(
-"${RPCS3_SRC_DIR}/../asmjit/"
 "${RPCS3_SRC_DIR}/../3rdparty/minidx12/"
 )
 
@@ -169,9 +169,10 @@ RPCS3_SRC
 "${RPCS3_SRC_DIR}/../Utilities/*.cpp"
 "${RPCS3_SRC_DIR}/../rsx_program_decompiler/rsx_decompiler/*.cpp"
 "${RPCS3_SRC_DIR}/../rsx_program_decompiler/shader_code/*.cpp"
+"${RPCS3_SRC_DIR}/../asmjit/src/asmjit/*.cpp"
 )
 
-if(APPLE)
+if(APPLE OR WIN32 AND NOT MSVC)
 	set (EXCLUDE_DIR "/RSX/VK/")
 	foreach (TMP_PATH ${RPCS3_SRC})
 		string (FIND ${TMP_PATH} ${EXCLUDE_DIR} EXCLUDE_DIR_FOUND)
@@ -184,36 +185,26 @@ endif()
 add_executable(rpcs3 ${RPCS3_SRC})
 
 
-if(NOT MSVC)
-	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -L${CMAKE_CURRENT_BINARY_DIR}/../asmjit/") #hack because the asmjit cmake file force fno exceptions upd: not sure if vs2015 build is affected
-else()
+if(MSVC)
 	set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} /SUBSYSTEM:WINDOWS /NODEFAULTLIB:libc.lib /NODEFAULTLIB:libcmt.lib /NODEFAULTLIB:libcd.lib /NODEFAULTLIB:libcmtd.lib /NODEFAULTLIB:msvcrtd.lib")
 endif()
-if(WIN32) # I'm not sure we need all of these libs, but we link them in vs
-	target_link_libraries(rpcs3  odbc32.lib odbccp32.lib comctl32.lib ws2_32.lib shlwapi.lib winmm.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib Iphlpapi.lib)
-	if(LLVM_FOUND)
-		target_link_libraries(rpcs3  ${LLVM_LIBS})
-	endif()
+if(WIN32)
+	target_link_libraries(rpcs3  ws2_32.lib Winmm.lib)
 	if(NOT MSVC)
-		target_link_libraries(rpcs3  ${OPENGL_LIBRARIES} ${GLEW_LIBRARY} opengl32.lib glu32.lib)
+		target_link_libraries(rpcs3  ${OPENGL_LIBRARIES} ${GLEW_LIBRARY} opengl32.lib glu32.lib -lstdc++.dll -lpthread.dll)
 	else()
-		target_link_libraries(rpcs3  dxgi.lib d2d1.lib dwrite.lib)
+		target_link_libraries(rpcs3  dxgi.lib d2d1.lib dwrite.lib VKstatic.1 glslang OSDependent OGLCompiler SPIRV)
 	endif()
-	target_link_libraries(rpcs3  asmjit.lib avformat.lib avcodec.lib avutil.lib swresample.lib swscale.lib png16_static ${wxWidgets_LIBRARIES} ${OPENAL_LIBRARY} ${ADDITIONAL_LIBS} VKstatic.1 glslang OSDependent OGLCompiler SPIRV)
+	target_link_libraries(rpcs3  avformat.lib avcodec.lib avutil.lib swresample.lib swscale.lib png16_static ${wxWidgets_LIBRARIES} ${OPENAL_LIBRARY} ${ADDITIONAL_LIBS})
 else()
-	if(LLVM_FOUND)
-		target_link_libraries(rpcs3  asmjit.a ${wxWidgets_LIBRARIES} ${OPENAL_LIBRARY} ${GLEW_LIBRARY} ${OPENGL_LIBRARIES})
-		target_link_libraries(rpcs3  libavformat.a libavcodec.a libavutil.a libswresample.a libswscale.a png16_static ${ZLIB_LIBRARIES} ${LLVM_LIBS} ${ADDITIONAL_LIBS})
-		if (NOT APPLE)
-			target_link_libraries(rpcs3  vulkan glslang OSDependent OGLCompiler SPIRV)
-		endif()
-	else()
-		target_link_libraries(rpcs3  asmjit.a ${wxWidgets_LIBRARIES} ${OPENAL_LIBRARY} ${GLEW_LIBRARY} ${OPENGL_LIBRARIES})
-		target_link_libraries(rpcs3  libavformat.a libavcodec.a libavutil.a libswresample.a libswscale.a png16_static ${ZLIB_LIBRARIES} ${ADDITIONAL_LIBS})
-		if (NOT APPLE)
-			target_link_libraries(rpcs3  vulkan glslang OSDependent OGLCompiler SPIRV)
-		endif()
-    endif()
+	target_link_libraries(rpcs3  ${wxWidgets_LIBRARIES} ${OPENAL_LIBRARY} ${GLEW_LIBRARY} ${OPENGL_LIBRARIES})
+	target_link_libraries(rpcs3  libavformat.a libavcodec.a libavutil.a libswresample.a libswscale.a png16_static ${ZLIB_LIBRARIES} ${ADDITIONAL_LIBS})
+	if (NOT APPLE)
+		target_link_libraries(rpcs3  vulkan glslang OSDependent OGLCompiler SPIRV)
+	endif()
+endif()
+if(LLVM_FOUND)
+	target_link_libraries(rpcs3  ${LLVM_LIBS})
 endif()
 
 set_target_properties(rpcs3 PROPERTIES COTIRE_CXX_PREFIX_HEADER_INIT "${RPCS3_SRC_DIR}/stdafx.h")
diff --git a/rpcs3/Crypto/unpkg.cpp b/rpcs3/Crypto/unpkg.cpp
index 79f7fc59b5..6f7e6f3c7c 100644
--- a/rpcs3/Crypto/unpkg.cpp
+++ b/rpcs3/Crypto/unpkg.cpp
@@ -5,59 +5,7 @@
 #include "key_vault.h"
 #include "unpkg.h"
 
-static bool CheckHeader(const fs::file& pkg_f, PKGHeader& header)
-{
-	if (header.pkg_magic != 0x7F504B47)
-	{
-		LOG_ERROR(LOADER, "PKG: Not a package file!");
-		return false;
-	}
-
-	switch (const u16 type = header.pkg_type)
-	{
-	case PKG_RELEASE_TYPE_DEBUG:   break;
-	case PKG_RELEASE_TYPE_RELEASE: break;
-	default:
-	{
-		LOG_ERROR(LOADER, "PKG: Unknown PKG type (0x%x)", type);
-		return false;
-	}
-	}
-
-	switch (const u16 platform = header.pkg_platform)
-	{
-	case PKG_PLATFORM_TYPE_PS3: break;
-	case PKG_PLATFORM_TYPE_PSP: break;
-	default:
-	{
-		LOG_ERROR(LOADER, "PKG: Unknown PKG platform (0x%x)", platform);
-		return false;
-	}
-	}
-
-	if (header.header_size != PKG_HEADER_SIZE && header.header_size != PKG_HEADER_SIZE2)
-	{
-		LOG_ERROR(LOADER, "PKG: Wrong header size (0x%x)", header.header_size);
-		return false;
-	}
-
-	if (header.pkg_size > pkg_f.size())
-	{
-		LOG_ERROR(LOADER, "PKG: File size mismatch (pkg_size=0x%llx)", header.pkg_size);
-		return false;
-	}
-
-	if (header.data_size + header.data_offset > header.pkg_size)
-	{
-		LOG_ERROR(LOADER, "PKG: Data size mismatch (data_size=0x%llx, data_offset=0x%llx, file_size=0x%llx)", header.data_size, header.data_offset, header.pkg_size);
-		return false;
-	}
-
-	return true;
-}
-
-// PKG Decryption
-bool pkg_install(const fs::file& pkg_f, const std::string& dir, volatile f64& progress)
+bool pkg_install(const fs::file& pkg_f, const std::string& dir, atomic_t<double>& sync)
 {
 	const std::size_t BUF_SIZE = 8192 * 1024; // 8 MB
 
@@ -69,12 +17,53 @@ bool pkg_install(const fs::file& pkg_f, const std::string& dir, volatile f64& pr
 
 	if (!pkg_f.read(header))
 	{
-		LOG_ERROR(LOADER, "PKG: Package file is too short!");
+		LOG_ERROR(LOADER, "PKG file is too short!");
 		return false;
 	}
 
-	if (!CheckHeader(pkg_f, header))
+	if (header.pkg_magic != "\x7FPKG"_u32)
 	{
+		LOG_ERROR(LOADER, "Not a PKG file!");
+		return false;
+	}
+
+	switch (const u16 type = header.pkg_type)
+	{
+	case PKG_RELEASE_TYPE_DEBUG:   break;
+	case PKG_RELEASE_TYPE_RELEASE: break;
+	default:
+	{
+		LOG_ERROR(LOADER, "Unknown PKG type (0x%x)", type);
+		return false;
+	}
+	}
+
+	switch (const u16 platform = header.pkg_platform)
+	{
+	case PKG_PLATFORM_TYPE_PS3: break;
+	case PKG_PLATFORM_TYPE_PSP: break;
+	default:
+	{
+		LOG_ERROR(LOADER, "Unknown PKG platform (0x%x)", platform);
+		return false;
+	}
+	}
+
+	if (header.header_size != PKG_HEADER_SIZE && header.header_size != PKG_HEADER_SIZE2)
+	{
+		LOG_ERROR(LOADER, "Wrong PKG header size (0x%x)", header.header_size);
+		return false;
+	}
+
+	if (header.pkg_size > pkg_f.size())
+	{
+		LOG_ERROR(LOADER, "PKG file size mismatch (pkg_size=0x%llx)", header.pkg_size);
+		return false;
+	}
+
+	if (header.data_size + header.data_offset > header.pkg_size)
+	{
+		LOG_ERROR(LOADER, "PKG data size mismatch (data_size=0x%llx, data_offset=0x%llx, file_size=0x%llx)", header.data_size, header.data_offset, header.pkg_size);
 		return false;
 	}
 
@@ -141,8 +130,6 @@ bool pkg_install(const fs::file& pkg_f, const std::string& dir, volatile f64& pr
 		return read;
 	};
 
-	LOG_SUCCESS(LOADER, "PKG: Installing in %s (%d entries)...", dir, header.file_count);
-
 	decrypt(0, header.file_count * sizeof(PKGEntry), header.pkg_platform == PKG_PLATFORM_TYPE_PSP);
 
 	std::vector<PKGEntry> entries(header.file_count);
@@ -155,7 +142,7 @@ bool pkg_install(const fs::file& pkg_f, const std::string& dir, volatile f64& pr
 
 		if (entry.name_size > 256)
 		{
-			LOG_ERROR(LOADER, "PKG: Name size is too big (0x%x)", entry.name_size);
+			LOG_ERROR(LOADER, "PKG name size is too big (0x%x)", entry.name_size);
 			continue;
 		}
 
@@ -183,31 +170,35 @@ bool pkg_install(const fs::file& pkg_f, const std::string& dir, volatile f64& pr
 
 					if (decrypt(entry.file_offset + pos, block_size, is_psp) != block_size)
 					{
-						LOG_ERROR(LOADER, "PKG: Failed to extract file %s", path);
+						LOG_ERROR(LOADER, "Failed to extract file %s", path);
 						break;
 					}
 
 					if (out.write(buf.get(), block_size) != block_size)
 					{
-						LOG_ERROR(LOADER, "PKG: Failed to write file %s", path);
+						LOG_ERROR(LOADER, "Failed to write file %s", path);
 						break;
 					}
 
-					progress += (block_size + 0.0) / header.data_size;
+					if (sync.fetch_add((block_size + 0.0) / header.data_size) < 0.)
+					{
+						LOG_ERROR(LOADER, "Package installation cancelled: %s", dir);
+						return false;
+					}
 				}
 
 				if (did_overwrite)
 				{
-					LOG_SUCCESS(LOADER, "PKG: %s file overwritten", name);
+					LOG_WARNING(LOADER, "Overwritten file %s", name);
 				}
 				else
 				{
-					LOG_SUCCESS(LOADER, "PKG: %s file created", name);
+					LOG_NOTICE(LOADER, "Created file %s", name);
 				}
 			}
 			else
 			{
-				LOG_ERROR(LOADER, "PKG: Could not create file %s", path);
+				LOG_ERROR(LOADER, "Failed to create file %s", path);
 			}
 
 			break;
@@ -219,15 +210,15 @@ bool pkg_install(const fs::file& pkg_f, const std::string& dir, volatile f64& pr
 
 			if (fs::create_dir(path))
 			{
-				LOG_SUCCESS(LOADER, "PKG: %s directory created", name);
+				LOG_NOTICE(LOADER, "Created directory %s", name);
 			}
 			else if (fs::is_dir(path))
 			{
-				LOG_SUCCESS(LOADER, "PKG: %s directory already exists", name);
+				LOG_WARNING(LOADER, "Reused existing directory %s", name);
 			}
 			else
 			{
-				LOG_ERROR(LOADER, "PKG: Could not create directory %s", path);
+				LOG_ERROR(LOADER, "Failed to create directory %s", path);
 			}
 
 			break;
@@ -235,10 +226,11 @@ bool pkg_install(const fs::file& pkg_f, const std::string& dir, volatile f64& pr
 
 		default:
 		{
-			LOG_ERROR(LOADER, "PKG: Unknown PKG entry type (0x%x) %s", entry.type, name);
+			LOG_ERROR(LOADER, "Unknown PKG entry type (0x%x) %s", entry.type, name);
 		}
 		}
 	}
 
+	LOG_SUCCESS(LOADER, "Package successfully installed to %s", dir);
 	return true;
 }
diff --git a/rpcs3/Crypto/unpkg.h b/rpcs3/Crypto/unpkg.h
index fe39e9aed0..f288b2c095 100644
--- a/rpcs3/Crypto/unpkg.h
+++ b/rpcs3/Crypto/unpkg.h
@@ -32,7 +32,7 @@ enum : u32
 // Structs
 struct PKGHeader
 {
-	be_t<u32> pkg_magic;    // Magic (0x7f504b47)
+	nse_t<u32> pkg_magic;   // Magic (0x7f504b47)
 	be_t<u16> pkg_type;     // Release type (Retail:0x8000, Debug:0x0000)
 	be_t<u16> pkg_platform; // Platform type (PS3:0x0001, PSP:0x0002)
 	be_t<u32> header_size;  // Header size (0xc0)
@@ -57,4 +57,4 @@ struct PKGEntry
 	be_t<u32> pad;          // Padding (zeros)
 };
 
-bool pkg_install(const class fs::file& pkg_f, const std::string& dir, volatile f64& progress);
+bool pkg_install(const class fs::file& pkg_f, const std::string& dir, atomic_t<double>&);
diff --git a/rpcs3/Emu/ARMv7/ARMv7Thread.cpp b/rpcs3/Emu/ARMv7/ARMv7Thread.cpp
index a240c216a3..98828c1c81 100644
--- a/rpcs3/Emu/ARMv7/ARMv7Thread.cpp
+++ b/rpcs3/Emu/ARMv7/ARMv7Thread.cpp
@@ -182,13 +182,13 @@ ARMv7Thread::~ARMv7Thread()
 	}
 }
 
+ARMv7Thread::ARMv7Thread(const std::string& name)
+	: cpu_thread(cpu_type::arm, name)
+{
+}
+
 void ARMv7Thread::fast_call(u32 addr)
 {
-	if (!is_current())
-	{
-		throw EXCEPTION("Called from the wrong thread");
-	}
-
 	auto old_PC = PC;
 	auto old_stack = SP;
 	auto old_LR = LR;
diff --git a/rpcs3/Emu/ARMv7/ARMv7Thread.h b/rpcs3/Emu/ARMv7/ARMv7Thread.h
index 1b6743c2da..7b9e619a08 100644
--- a/rpcs3/Emu/ARMv7/ARMv7Thread.h
+++ b/rpcs3/Emu/ARMv7/ARMv7Thread.h
@@ -18,14 +18,10 @@ public:
 	virtual std::string dump() const override;
 	virtual void cpu_init() override;
 	virtual void cpu_task() override;
-
-	ARMv7Thread(const std::string& name)
-		: cpu_thread(cpu_type::arm, name)
-	{
-	}
-
 	virtual ~ARMv7Thread() override;
 
+	ARMv7Thread(const std::string& name);
+
 	union
 	{
 		u32 GPR[15];
@@ -209,7 +205,7 @@ template<typename T>
 struct arm_gpr_cast_impl<T, std::enable_if_t<std::is_integral<T>::value || std::is_enum<T>::value>>
 {
 	static_assert(sizeof(T) <= 4, "Too big integral type for arm_gpr_cast<>()");
-	static_assert(std::is_same<const T, const bool>::value == false, "bool type is deprecated in arm_gpr_cast<>(), use b8 instead");
+	static_assert(std::is_same<CV T, CV bool>::value == false, "bool type is deprecated in arm_gpr_cast<>(), use b8 instead");
 
 	static inline u32 to(const T& value)
 	{
diff --git a/rpcs3/Emu/ARMv7/Modules/sceLibKernel.cpp b/rpcs3/Emu/ARMv7/Modules/sceLibKernel.cpp
index 2a00e9c5c1..57066f455c 100644
--- a/rpcs3/Emu/ARMv7/Modules/sceLibKernel.cpp
+++ b/rpcs3/Emu/ARMv7/Modules/sceLibKernel.cpp
@@ -9,6 +9,8 @@ LOG_CHANNEL(sceLibKernel);
 
 extern u64 get_system_time();
 
+extern std::condition_variable& get_current_thread_cv();
+
 s32 sceKernelAllocMemBlock(vm::cptr<char> name, s32 type, u32 vsize, vm::ptr<SceKernelAllocMemBlockOpt> pOpt)
 {
 	throw EXCEPTION("");
@@ -71,7 +73,7 @@ arm_error_code sceKernelStartThread(s32 threadId, u32 argSize, vm::cptr<void> pA
 	thread->GPR[1] = pos;
 
 	thread->state -= cpu_state::stop;
-	thread->safe_notify();
+	thread->lock_notify();
 	return SCE_OK;
 }
 
@@ -484,11 +486,11 @@ arm_error_code sceKernelWaitEventFlag(ARMv7Thread& cpu, s32 evfId, u32 bitPatter
 				break;
 			}
 
-			cpu.cv.wait_for(lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			cpu.cv.wait(lock);
+			get_current_thread_cv().wait(lock);
 		}
 	}
 
@@ -563,7 +565,7 @@ arm_error_code sceKernelSetEventFlag(s32 evfId, u32 bitPattern)
 			cpu.GPR[1] = result;
 
 			thread->state += cpu_state::signal;
-			thread->cv.notify_one();
+			thread->notify();
 			return true;
 		}
 
@@ -612,7 +614,7 @@ arm_error_code sceKernelCancelEventFlag(s32 evfId, u32 setPattern, vm::ptr<s32>
 		static_cast<ARMv7Thread&>(*thread).GPR[0] = SCE_KERNEL_ERROR_WAIT_CANCEL;
 		static_cast<ARMv7Thread&>(*thread).GPR[1] = setPattern;
 		thread->state += cpu_state::signal;
-		thread->cv.notify_one();
+		thread->notify();
 	}
 
 	*pNumWaitThreads = static_cast<u32>(evf->sq.size());
diff --git a/rpcs3/Emu/ARMv7/Modules/sceLibKernel.h b/rpcs3/Emu/ARMv7/Modules/sceLibKernel.h
index a94a7ba760..443b2f5df1 100644
--- a/rpcs3/Emu/ARMv7/Modules/sceLibKernel.h
+++ b/rpcs3/Emu/ARMv7/Modules/sceLibKernel.h
@@ -762,7 +762,7 @@ struct psv_event_flag_t
 			static_cast<ARMv7Thread&>(*thread).GPR[0] = SCE_KERNEL_ERROR_WAIT_DELETE;
 			static_cast<ARMv7Thread&>(*thread).GPR[1] = pattern;
 			thread->state += cpu_state::signal;
-			thread->cv.notify_one();
+			thread->notify();
 		}
 	}
 };
diff --git a/rpcs3/Emu/Audio/XAudio2/XAudio2Thread.cpp b/rpcs3/Emu/Audio/XAudio2/XAudio2Thread.cpp
index 3611783d2e..65cb3130ad 100644
--- a/rpcs3/Emu/Audio/XAudio2/XAudio2Thread.cpp
+++ b/rpcs3/Emu/Audio/XAudio2/XAudio2Thread.cpp
@@ -1,5 +1,5 @@
-#include "stdafx.h"
-#ifdef _MSC_VER
+#ifdef _WIN32
+#include "Utilities/Log.h"
 #include "Utilities/Config.h"
 #include "Emu/System.h"
 
@@ -143,4 +143,5 @@ void XAudio2Thread::AddData(const void* src, int size)
 		Emu.Pause();
 	}
 }
+
 #endif
diff --git a/rpcs3/Emu/Audio/XAudio2/XAudio2Thread.h b/rpcs3/Emu/Audio/XAudio2/XAudio2Thread.h
index 25729c450e..7addcf4c69 100644
--- a/rpcs3/Emu/Audio/XAudio2/XAudio2Thread.h
+++ b/rpcs3/Emu/Audio/XAudio2/XAudio2Thread.h
@@ -1,13 +1,10 @@
 #pragma once
-#ifdef _MSC_VER
+
+#ifdef _WIN32
 
 #include "Emu/Audio/AudioThread.h"
 
-#pragma push_macro("_WIN32_WINNT")
-#undef _WIN32_WINNT
-#define _WIN32_WINNT 0x0601 // This is to be sure that correct (2.7) header is included
-#include "3rdparty/XAudio2_7/XAudio2.h" // XAudio2 2.8 available only on Win8+, used XAudio2 2.7 from dxsdk
-#pragma pop_macro("_WIN32_WINNT")
+#include "3rdparty/XAudio2_7/XAudio2.h"
 
 class XAudio2Thread : public AudioThread
 {
@@ -25,4 +22,5 @@ public:
 	virtual void Stop() override;
 	virtual void AddData(const void* src, int size) override;
 };
+
 #endif
diff --git a/rpcs3/Emu/CPU/CPUThread.cpp b/rpcs3/Emu/CPU/CPUThread.cpp
index b80cc30c93..9cab30f773 100644
--- a/rpcs3/Emu/CPU/CPUThread.cpp
+++ b/rpcs3/Emu/CPU/CPUThread.cpp
@@ -12,11 +12,13 @@ thread_local cpu_thread* g_tls_current_cpu_thread = nullptr;
 
 void cpu_thread::on_task()
 {
+	state -= cpu_state::exit;
+
 	g_tls_current_cpu_thread = this;
 
 	Emu.SendDbgCommand(DID_CREATE_THREAD, this);
 
-	std::unique_lock<std::mutex> lock(mutex);
+	std::unique_lock<std::mutex> lock(get_current_thread_mutex());
 
 	// Check thread status
 	while (!(state & cpu_state::exit))
@@ -52,13 +54,29 @@ void cpu_thread::on_task()
 			continue;
 		}
 
-		cv.wait(lock);
+		get_current_thread_cv().wait(lock);
 	}
 }
 
+void cpu_thread::on_stop()
+{
+	state += cpu_state::exit;
+	lock_notify();
+}
+
+cpu_thread::~cpu_thread()
+{
+}
+
+cpu_thread::cpu_thread(cpu_type type, const std::string& name)
+	: type(type)
+	, name(name)
+{
+}
+
 bool cpu_thread::check_status()
 {
-	std::unique_lock<std::mutex> lock(mutex, std::defer_lock);
+	std::unique_lock<std::mutex> lock(get_current_thread_mutex(), std::defer_lock);
 
 	while (true)
 	{
@@ -85,7 +103,7 @@ bool cpu_thread::check_status()
 			continue;
 		}
 
-		cv.wait(lock);
+		get_current_thread_cv().wait(lock);
 	}
 
 	const auto state_ = state.load();
@@ -104,6 +122,11 @@ bool cpu_thread::check_status()
 	return false;
 }
 
+[[noreturn]] void cpu_thread::xsleep()
+{
+	throw std::runtime_error("cpu_thread: sleep()/awake() inconsistency");
+}
+
 std::vector<std::shared_ptr<cpu_thread>> get_all_cpu_threads()
 {
 	std::vector<std::shared_ptr<cpu_thread>> result;
diff --git a/rpcs3/Emu/CPU/CPUThread.h b/rpcs3/Emu/CPU/CPUThread.h
index 8baa65b03e..42623087e2 100644
--- a/rpcs3/Emu/CPU/CPUThread.h
+++ b/rpcs3/Emu/CPU/CPUThread.h
@@ -34,26 +34,14 @@ class cpu_thread : public named_thread
 	void on_task() override;
 
 public:
-	virtual void on_init() override
-	{
-		named_thread::on_init();
-	}
-
-	virtual void on_stop() override
-	{
-		state += cpu_state::exit;
-		safe_notify();
-	}
+	virtual void on_stop() override;
+	virtual ~cpu_thread() override;
 
 	const std::string name;
-	const u32 id{};
+	const u32 id = -1;
 	const cpu_type type;
 
-	cpu_thread(cpu_type type, const std::string& name)
-		: type(type)
-		, name(name)
-	{
-	}
+	cpu_thread(cpu_type type, const std::string& name);
 
 	// Public thread state
 	atomic_t<mset<cpu_state>> state{ cpu_state::stop };
@@ -79,15 +67,15 @@ public:
 	virtual bool handle_interrupt() { return false; }
 
 private:
-	[[noreturn]] void xsleep()
-	{
-		throw std::runtime_error("cpu_thread: sleep()/awake() inconsistency");
-	}
+	[[noreturn]] void xsleep();
 
 	// Sleep/Awake counter
 	atomic_t<u32> m_sleep{};
 };
 
+extern std::mutex& get_current_thread_mutex();
+extern std::condition_variable& get_current_thread_cv();
+
 inline cpu_thread* get_current_cpu_thread() noexcept
 {
 	extern thread_local cpu_thread* g_tls_current_cpu_thread;
diff --git a/rpcs3/Emu/Cell/Modules/cellAdec.cpp b/rpcs3/Emu/Cell/Modules/cellAdec.cpp
index ba9b76fe19..b6acc1f23c 100644
--- a/rpcs3/Emu/Cell/Modules/cellAdec.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellAdec.cpp
@@ -468,7 +468,7 @@ void adecOpen(u32 adec_id) // TODO: call from the constructor
 
 	adec.adecCb->cpu_init();
 	adec.adecCb->state -= cpu_state::stop;
-	adec.adecCb->safe_notify();
+	adec.adecCb->lock_notify();
 }
 
 bool adecCheckType(s32 type)
diff --git a/rpcs3/Emu/Cell/Modules/cellAudio.cpp b/rpcs3/Emu/Cell/Modules/cellAudio.cpp
index d5239f2557..9a119a0865 100644
--- a/rpcs3/Emu/Cell/Modules/cellAudio.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellAudio.cpp
@@ -280,8 +280,6 @@ void audio_config::on_task()
 
 			LV2_LOCK;
 
-			std::lock_guard<std::mutex> lock(mutex);
-
 			for (u64 key : keys)
 			{
 				if (auto&& queue = lv2_event_queue_t::find(key))
@@ -699,7 +697,7 @@ s32 cellAudioSetNotifyEventQueue(u64 key)
 		return CELL_AUDIO_ERROR_NOT_INIT;
 	}
 
-	std::lock_guard<std::mutex> lock(g_audio->mutex);
+	LV2_LOCK;
 
 	for (auto k : g_audio->keys) // check for duplicates
 	{
@@ -734,7 +732,7 @@ s32 cellAudioRemoveNotifyEventQueue(u64 key)
 		return CELL_AUDIO_ERROR_NOT_INIT;
 	}
 
-	std::lock_guard<std::mutex> lock(g_audio->mutex);
+	LV2_LOCK;
 
 	for (auto i = g_audio->keys.begin(); i != g_audio->keys.end(); i++)
 	{
diff --git a/rpcs3/Emu/Cell/Modules/cellAudioOut.cpp b/rpcs3/Emu/Cell/Modules/cellAudioOut.cpp
index d294bbd776..8383a7657f 100644
--- a/rpcs3/Emu/Cell/Modules/cellAudioOut.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellAudioOut.cpp
@@ -9,8 +9,6 @@ s32 cellAudioOutGetSoundAvailability(u32 audioOut, u32 type, u32 fs, u32 option)
 {
 	cellSysutil.warning("cellAudioOutGetSoundAvailability(audioOut=%d, type=%d, fs=0x%x, option=%d)", audioOut, type, fs, option);
 
-	option = 0;
-
 	s32 available = 8; // should be at least 2
 
 	switch (fs)
@@ -49,8 +47,6 @@ s32 cellAudioOutGetSoundAvailability2(u32 audioOut, u32 type, u32 fs, u32 ch, u3
 {
 	cellSysutil.warning("cellAudioOutGetSoundAvailability2(audioOut=%d, type=%d, fs=0x%x, ch=%d, option=%d)", audioOut, type, fs, ch, option);
 
-	option = 0;
-
 	s32 available = 8; // should be at least 2
 
 	switch (fs)
diff --git a/rpcs3/Emu/Cell/Modules/cellDmux.cpp b/rpcs3/Emu/Cell/Modules/cellDmux.cpp
index 0c755b197d..cb86cb76bb 100644
--- a/rpcs3/Emu/Cell/Modules/cellDmux.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellDmux.cpp
@@ -759,7 +759,7 @@ void dmuxOpen(u32 dmux_id) // TODO: call from the constructor
 
 	dmux.dmuxCb->cpu_init();
 	dmux.dmuxCb->state -= cpu_state::stop;
-	dmux.dmuxCb->safe_notify();
+	dmux.dmuxCb->lock_notify();
 }
 
 s32 cellDmuxQueryAttr(vm::cptr<CellDmuxType> type, vm::ptr<CellDmuxAttr> attr)
diff --git a/rpcs3/Emu/Cell/Modules/cellFs.cpp b/rpcs3/Emu/Cell/Modules/cellFs.cpp
index ee85879eff..2374bdd421 100644
--- a/rpcs3/Emu/Cell/Modules/cellFs.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellFs.cpp
@@ -263,7 +263,7 @@ s32 cellFsReadWithOffset(u32 fd, u64 offset, vm::ptr<void> buf, u64 buffer_size,
 
 	const auto read = file->file.read(buf.get_ptr(), buffer_size);
 
-	ASSERT(file->file.seek(old_pos) == old_pos);
+	file->file.seek(old_pos);
 
 	if (nread)
 	{
@@ -292,7 +292,7 @@ s32 cellFsWriteWithOffset(u32 fd, u64 offset, vm::cptr<void> buf, u64 data_size,
 
 	const auto written = file->file.write(buf.get_ptr(), data_size);
 
-	ASSERT(file->file.seek(old_pos) == old_pos);
+	file->file.seek(old_pos);
 
 	if (nwrite)
 	{
@@ -502,7 +502,7 @@ s32 cellFsStReadStart(u32 fd, u64 offset, u64 size)
 				auto old = file->file.pos();
 				file->file.seek(offset + file->st_total_read);
 				auto res = file->file.read(vm::base(position), file->st_block_size);
-				ASSERT(file->file.seek(old) == old);
+				file->file.seek(old);
 
 				// notify
 				file->st_total_read += res;
@@ -796,11 +796,11 @@ s32 sdata_unpack(const std::string& packed_file, const std::string& unpacked_fil
 
 		if (flags & 0x20)
 		{
-			ASSERT(packed_stream.seek(0x100) == 0x100);
+			packed_stream.seek(0x100);
 		}
 		else
 		{
-			ASSERT(packed_stream.seek(startOffset) == startOffset);
+			packed_stream.seek(startOffset);
 		}
 
 		for (u32 i = 0; i < blockCount; i++)
@@ -887,7 +887,7 @@ void fsAio(vm::ptr<CellFsAio> aio, bool write, s32 xid, fs_aio_cb_t func)
 			? file->file.write(aio->buf.get_ptr(), aio->size)
 			: file->file.read(aio->buf.get_ptr(), aio->size);
 
-		ASSERT(file->file.seek(old_pos) == old_pos);
+		file->file.seek(old_pos);
 	}
 
 	// should be executed directly by FS AIO thread
diff --git a/rpcs3/Emu/Cell/Modules/cellGame.cpp b/rpcs3/Emu/Cell/Modules/cellGame.cpp
index 0230547d47..3d45e3d13f 100644
--- a/rpcs3/Emu/Cell/Modules/cellGame.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellGame.cpp
@@ -7,7 +7,7 @@
 #include "cellMsgDialog.h"
 #include "cellGame.h"
 
-#include <future>
+#include "Loader/PSF.h"
 
 LOG_CHANNEL(cellGame);
 
@@ -20,21 +20,25 @@ LOG_CHANNEL(cellGame);
 // Usual (persistent) content directory (if is_temporary):
 // contentInfo = "/dev_hdd0/game/" + dir
 // usrdir = "/dev_hdd0/game/" + dir + "/USRDIR"
-struct content_permission_t final
+struct content_permission final
 {
-	// content directory name or path
+	// Content directory name or path
 	const std::string dir;
 
-	// true if temporary directory is created and must be moved or deleted
+	// SFO file
+	psf::registry sfo;
+
+	// True if temporary directory is created and must be moved or deleted
 	bool is_temporary = false;
 
-	content_permission_t(const std::string& dir, bool is_temp)
-		: dir(dir)
+	content_permission(std::string&& dir, psf::registry&& sfo, bool is_temp)
+		: dir(std::move(dir))
+		, sfo(std::move(sfo))
 		, is_temporary(is_temp)
 	{
 	}
 
-	~content_permission_t()
+	~content_permission()
 	{
 		if (is_temporary)
 		{
@@ -145,7 +149,7 @@ s32 cellGameDataExitBroken()
 }
 
 
-s32 cellGameBootCheck(vm::ptr<u32> type, vm::ptr<u32> attributes, vm::ptr<CellGameContentSize> size, vm::ptr<char[CELL_GAME_DIRNAME_SIZE]> dirName)
+ppu_error_code cellGameBootCheck(vm::ptr<u32> type, vm::ptr<u32> attributes, vm::ptr<CellGameContentSize> size, vm::ptr<char[CELL_GAME_DIRNAME_SIZE]> dirName)
 {
 	cellGame.warning("cellGameBootCheck(type=*0x%x, attributes=*0x%x, size=*0x%x, dirName=*0x%x)", type, attributes, size, dirName);
 
@@ -160,7 +164,9 @@ s32 cellGameBootCheck(vm::ptr<u32> type, vm::ptr<u32> attributes, vm::ptr<CellGa
 	}
 
 	// According to testing (in debug mode) cellGameBootCheck doesn't return an error code, when PARAM.SFO doesn't exist.
-	const std::string& category = psf::get_string(Emu.GetPSF(), "CATEGORY");
+	psf::registry&& sfo = psf::load_object(fs::file(vfs::get("/app_home/../PARAM.SFO")));
+
+	const std::string& category = psf::get_string(sfo, "CATEGORY");
 
 	if (category == "DG")
 	{
@@ -168,7 +174,7 @@ s32 cellGameBootCheck(vm::ptr<u32> type, vm::ptr<u32> attributes, vm::ptr<CellGa
 		*attributes = 0; // TODO
 		if (dirName) strcpy_trunc(*dirName, ""); // ???
 
-		if (!fxm::make<content_permission_t>("/dev_bdvd/PS3_GAME", false))
+		if (!fxm::make<content_permission>("/dev_bdvd/PS3_GAME", std::move(sfo), false))
 		{
 			return CELL_GAME_ERROR_BUSY;
 		}
@@ -179,7 +185,7 @@ s32 cellGameBootCheck(vm::ptr<u32> type, vm::ptr<u32> attributes, vm::ptr<CellGa
 		*attributes = 0; // TODO
 		if (dirName) strcpy_trunc(*dirName, Emu.GetTitleID()); 
 
-		if (!fxm::make<content_permission_t>("/dev_hdd0/game/" + Emu.GetTitleID(), false))
+		if (!fxm::make<content_permission>("/dev_hdd0/game/" + Emu.GetTitleID(), std::move(sfo), false))
 		{
 			return CELL_GAME_ERROR_BUSY;
 		}
@@ -190,20 +196,20 @@ s32 cellGameBootCheck(vm::ptr<u32> type, vm::ptr<u32> attributes, vm::ptr<CellGa
 		*attributes = CELL_GAME_ATTRIBUTE_PATCH; // TODO
 		if (dirName) strcpy_trunc(*dirName, Emu.GetTitleID()); // ???
 
-		if (!fxm::make<content_permission_t>("/dev_bdvd/PS3_GAME", false))
+		if (!fxm::make<content_permission>("/dev_bdvd/PS3_GAME", std::move(sfo), false))
 		{
 			return CELL_GAME_ERROR_BUSY;
 		}
 	}
 	else
 	{
-		cellGame.error("cellGameBootCheck(): Unknown CATEGORY value: %s", category);
+		throw fmt::exception("cellGameBootCheck(): Unknown CATEGORY: %s", category);
 	}
 
 	return CELL_GAME_RET_OK;
 }
 
-s32 cellGamePatchCheck(vm::ptr<CellGameContentSize> size, vm::ptr<void> reserved)
+ppu_error_code cellGamePatchCheck(vm::ptr<CellGameContentSize> size, vm::ptr<void> reserved)
 {
 	cellGame.warning("cellGamePatchCheck(size=*0x%x, reserved=*0x%x)", size, reserved);
 
@@ -217,13 +223,14 @@ s32 cellGamePatchCheck(vm::ptr<CellGameContentSize> size, vm::ptr<void> reserved
 		size->sysSizeKB = 0;
 	}
 
-	if (psf::get_string(Emu.GetPSF(), "CATEGORY") != "GD")
+	psf::registry&& sfo = psf::load_object(fs::file(vfs::get("/app_home/../PARAM.SFO")));
+
+	if (psf::get_string(sfo, "CATEGORY") != "GD")
 	{
-		cellGame.error("cellGamePatchCheck(): CELL_GAME_ERROR_NOTPATCH");
 		return CELL_GAME_ERROR_NOTPATCH;
 	}
 
-	if (!fxm::make<content_permission_t>("/dev_hdd0/game/" + Emu.GetTitleID(), false))
+	if (!fxm::make<content_permission>("/dev_hdd0/game/" + Emu.GetTitleID(), std::move(sfo), false))
 	{
 		return CELL_GAME_ERROR_BUSY;
 	}
@@ -231,13 +238,12 @@ s32 cellGamePatchCheck(vm::ptr<CellGameContentSize> size, vm::ptr<void> reserved
 	return CELL_OK;
 }
 
-s32 cellGameDataCheck(u32 type, vm::cptr<char> dirName, vm::ptr<CellGameContentSize> size)
+ppu_error_code cellGameDataCheck(u32 type, vm::cptr<char> dirName, vm::ptr<CellGameContentSize> size)
 {
 	cellGame.warning("cellGameDataCheck(type=%d, dirName=*0x%x, size=*0x%x)", type, dirName, size);
 
 	if ((type - 1) >= 3)
 	{
-		cellGame.error("cellGameDataCheck(): CELL_GAME_ERROR_PARAM");
 		return CELL_GAME_ERROR_PARAM;
 	}
 
@@ -252,7 +258,7 @@ s32 cellGameDataCheck(u32 type, vm::cptr<char> dirName, vm::ptr<CellGameContentS
 	}
 
 	// TODO: not sure what should be checked there
-	const std::string& dir = type == CELL_GAME_GAMETYPE_DISC ? "/dev_bdvd/PS3_GAME"s : "/dev_hdd0/game/"s + dirName.get_ptr();
+	std::string&& dir = type == CELL_GAME_GAMETYPE_DISC ? "/dev_bdvd/PS3_GAME"s : "/dev_hdd0/game/"s + dirName.get_ptr();
 
 	if (!fs::is_dir(vfs::get(dir)))
 	{
@@ -260,7 +266,7 @@ s32 cellGameDataCheck(u32 type, vm::cptr<char> dirName, vm::ptr<CellGameContentS
 		return CELL_GAME_RET_NONE;
 	}
 
-	if (!fxm::make<content_permission_t>(dir, false))
+	if (!fxm::make<content_permission>(std::move(dir), psf::load_object(fs::file(vfs::get(dir + "/PARAM.SFO"))), false))
 	{
 		return CELL_GAME_ERROR_BUSY;
 	}
@@ -268,7 +274,7 @@ s32 cellGameDataCheck(u32 type, vm::cptr<char> dirName, vm::ptr<CellGameContentS
 	return CELL_GAME_RET_OK;
 }
 
-s32 cellGameContentPermit(vm::ptr<char[CELL_GAME_PATH_MAX]> contentInfoPath, vm::ptr<char[CELL_GAME_PATH_MAX]> usrdirPath)
+ppu_error_code cellGameContentPermit(vm::ptr<char[CELL_GAME_PATH_MAX]> contentInfoPath, vm::ptr<char[CELL_GAME_PATH_MAX]> usrdirPath)
 {
 	cellGame.warning("cellGameContentPermit(contentInfoPath=*0x%x, usrdirPath=*0x%x)", contentInfoPath, usrdirPath);
 
@@ -277,21 +283,21 @@ s32 cellGameContentPermit(vm::ptr<char[CELL_GAME_PATH_MAX]> contentInfoPath, vm:
 		return CELL_GAME_ERROR_PARAM;
 	}
 
-	const auto path_set = fxm::withdraw<content_permission_t>();
+	const auto prm = fxm::withdraw<content_permission>();
 	
-	if (!path_set)
+	if (!prm)
 	{
 		return CELL_GAME_ERROR_FAILURE;
 	}
 
-	if (path_set->is_temporary)
+	if (prm->is_temporary)
 	{
-		const std::string& dir = "/dev_hdd0/game/" + path_set->dir;
+		const std::string& dir = "/dev_hdd0/game/" + prm->dir;
 
 		// Make temporary directory persistent
 		fs::remove_all(vfs::get(dir));
 
-		if (fs::rename(vfs::get("/dev_hdd1/game/" + path_set->dir), vfs::get(dir)))
+		if (fs::rename(vfs::get("/dev_hdd1/game/" + prm->dir), vfs::get(dir)))
 		{
 			cellGame.success("cellGameContentPermit(): created directory %s", dir);
 		}
@@ -300,28 +306,30 @@ s32 cellGameContentPermit(vm::ptr<char[CELL_GAME_PATH_MAX]> contentInfoPath, vm:
 			throw fmt::exception("cellGameContentPermit(): failed to rename to %s", dir);
 		}
 
+		// Create PARAM.SFO
+		fs::file(dir + "/PARAM.SFO", fs::rewrite).write(psf::save_object(prm->sfo));
+
 		// Disable deletion
-		path_set->is_temporary = false;
+		prm->is_temporary = false;
 
 		strcpy_trunc(*contentInfoPath, dir);
 		strcpy_trunc(*usrdirPath, dir + "/USRDIR");
 	}
 	else
 	{
-		strcpy_trunc(*contentInfoPath, path_set->dir);
-		strcpy_trunc(*usrdirPath, path_set->dir + "/USRDIR");
+		strcpy_trunc(*contentInfoPath, prm->dir);
+		strcpy_trunc(*usrdirPath, prm->dir + "/USRDIR");
 	}
 	
 	return CELL_OK;
 }
 
-s32 cellGameDataCheckCreate2(PPUThread& ppu, u32 version, vm::cptr<char> dirName, u32 errDialog, vm::ptr<CellGameDataStatCallback> funcStat, u32 container)
+ppu_error_code cellGameDataCheckCreate2(PPUThread& ppu, u32 version, vm::cptr<char> dirName, u32 errDialog, vm::ptr<CellGameDataStatCallback> funcStat, u32 container)
 {
 	cellGame.error("cellGameDataCheckCreate2(version=0x%x, dirName=*0x%x, errDialog=0x%x, funcStat=*0x%x, container=%d)", version, dirName, errDialog, funcStat, container);
 
 	if (version != CELL_GAMEDATA_VERSION_CURRENT || errDialog > 1)
 	{
-		cellGame.error("cellGameDataCheckCreate2(): CELL_GAMEDATA_ERROR_PARAM");
 		return CELL_GAMEDATA_ERROR_PARAM;
 	}
 
@@ -356,11 +364,13 @@ s32 cellGameDataCheckCreate2(PPUThread& ppu, u32 version, vm::cptr<char> dirName
 	cbGet->sizeKB = CELL_GAMEDATA_SIZEKB_NOTCALC;
 	cbGet->sysSizeKB = 0;
 
+	psf::registry&& sfo = psf::load_object(fs::file(vfs::get("/app_home/../PARAM.SFO")));
+
 	cbGet->getParam.attribute = CELL_GAMEDATA_ATTR_NORMAL;
-	cbGet->getParam.parentalLevel = Emu.GetPSF().at("PARENTAL_LEVEL").as_integer();
-	strcpy_trunc(cbGet->getParam.dataVersion, Emu.GetPSF().at("APP_VER").as_string());
-	strcpy_trunc(cbGet->getParam.titleId, Emu.GetPSF().at("TITLE_ID").as_string());
-	strcpy_trunc(cbGet->getParam.title, Emu.GetPSF().at("TITLE").as_string());
+	cbGet->getParam.parentalLevel = psf::get_integer(sfo, "PARENTAL_LEVEL", 0);
+	strcpy_trunc(cbGet->getParam.dataVersion, psf::get_string(sfo, "APP_VER", ""));
+	strcpy_trunc(cbGet->getParam.titleId, psf::get_string(sfo, "TITLE_ID", ""));
+	strcpy_trunc(cbGet->getParam.title, psf::get_string(sfo, "TITLE", ""));
 	// TODO: write lang titles
 
 	funcStat(ppu, cbResult, cbGet, cbSet);
@@ -410,11 +420,12 @@ s32 cellGameDataCheckCreate(PPUThread& ppu, u32 version, vm::cptr<char> dirName,
 	return cellGameDataCheckCreate2(ppu, version, dirName, errDialog, funcStat, container);
 }
 
-s32 cellGameCreateGameData(vm::ptr<CellGameSetInitParams> init, vm::ptr<char[CELL_GAME_PATH_MAX]> tmp_contentInfoPath, vm::ptr<char[CELL_GAME_PATH_MAX]> tmp_usrdirPath)
+ppu_error_code cellGameCreateGameData(vm::ptr<CellGameSetInitParams> init, vm::ptr<char[CELL_GAME_PATH_MAX]> tmp_contentInfoPath, vm::ptr<char[CELL_GAME_PATH_MAX]> tmp_usrdirPath)
 {
 	cellGame.error("cellGameCreateGameData(init=*0x%x, tmp_contentInfoPath=*0x%x, tmp_usrdirPath=*0x%x)", init, tmp_contentInfoPath, tmp_usrdirPath);
 
-	std::string dir = init->titleId;
+	std::string&& dir = init->titleId;
+
 	std::string tmp_contentInfo = "/dev_hdd1/game/" + dir;
 	std::string tmp_usrdir = "/dev_hdd1/game/" + dir + "/USRDIR";
 
@@ -430,7 +441,14 @@ s32 cellGameCreateGameData(vm::ptr<CellGameSetInitParams> init, vm::ptr<char[CEL
 		return CELL_GAME_ERROR_ACCESS_ERROR; // ???
 	}
 
-	if (!fxm::make<content_permission_t>(dir, true))
+	psf::registry sfo
+	{
+		{ "TITLE_ID", psf::string(CELL_GAME_SYSP_TITLEID_SIZE, init->titleId) },
+		{ "TITLE", psf::string(CELL_GAME_SYSP_TITLE_SIZE, init->title) },
+		{ "VERSION", psf::string(CELL_GAME_SYSP_VERSION_SIZE, init->version) },
+	};
+
+	if (!fxm::make<content_permission>(std::move(dir), std::move(sfo), true))
 	{
 		return CELL_GAME_ERROR_BUSY;
 	}
@@ -452,10 +470,17 @@ s32 cellGameDeleteGameData()
 	return CELL_OK;
 }
 
-s32 cellGameGetParamInt(u32 id, vm::ptr<u32> value)
+ppu_error_code cellGameGetParamInt(s32 id, vm::ptr<s32> value)
 {
 	cellGame.warning("cellGameGetParamInt(id=%d, value=*0x%x)", id, value);
 
+	const auto prm = fxm::get<content_permission>();
+
+	if (!prm)
+	{
+		return CELL_GAME_ERROR_FAILURE;
+	}
+
 	std::string key;
 
 	switch(id)
@@ -464,63 +489,106 @@ s32 cellGameGetParamInt(u32 id, vm::ptr<u32> value)
 	case CELL_GAME_PARAMID_RESOLUTION:      key = "RESOLUTION";     break;
 	case CELL_GAME_PARAMID_SOUND_FORMAT:    key = "SOUND_FORMAT";   break;
 	default:
-		cellGame.error("cellGameGetParamInt(): Unimplemented parameter (%d)", id);
+	{
 		return CELL_GAME_ERROR_INVALID_ID;
 	}
+	}
 
-	*value = Emu.GetPSF().at(key).as_integer();
+	*value = psf::get_integer(prm->sfo, key, 0);
 	return CELL_OK;
 }
 
-s32 cellGameGetParamString(u32 id, vm::ptr<char> buf, u32 bufsize)
+static const char* get_param_string_key(s32 id)
+{
+	switch (id)
+	{
+	case CELL_GAME_PARAMID_TITLE:                    return "TITLE"; // TODO: Is this value correct?
+	case CELL_GAME_PARAMID_TITLE_DEFAULT:            return "TITLE";
+	case CELL_GAME_PARAMID_TITLE_JAPANESE:           return "TITLE_00";
+	case CELL_GAME_PARAMID_TITLE_ENGLISH:            return "TITLE_01";
+	case CELL_GAME_PARAMID_TITLE_FRENCH:             return "TITLE_02";
+	case CELL_GAME_PARAMID_TITLE_SPANISH:            return "TITLE_03";
+	case CELL_GAME_PARAMID_TITLE_GERMAN:             return "TITLE_04";
+	case CELL_GAME_PARAMID_TITLE_ITALIAN:            return "TITLE_05";
+	case CELL_GAME_PARAMID_TITLE_DUTCH:              return "TITLE_06";
+	case CELL_GAME_PARAMID_TITLE_PORTUGUESE:         return "TITLE_07";
+	case CELL_GAME_PARAMID_TITLE_RUSSIAN:            return "TITLE_08";
+	case CELL_GAME_PARAMID_TITLE_KOREAN:             return "TITLE_09";
+	case CELL_GAME_PARAMID_TITLE_CHINESE_T:          return "TITLE_10";
+	case CELL_GAME_PARAMID_TITLE_CHINESE_S:          return "TITLE_11";
+	case CELL_GAME_PARAMID_TITLE_FINNISH:            return "TITLE_12";
+	case CELL_GAME_PARAMID_TITLE_SWEDISH:            return "TITLE_13";
+	case CELL_GAME_PARAMID_TITLE_DANISH:             return "TITLE_14";
+	case CELL_GAME_PARAMID_TITLE_NORWEGIAN:          return "TITLE_15";
+	case CELL_GAME_PARAMID_TITLE_POLISH:             return "TITLE_16";
+	case CELL_GAME_PARAMID_TITLE_PORTUGUESE_BRAZIL:  return "TITLE_17";
+	case CELL_GAME_PARAMID_TITLE_ENGLISH_UK:         return "TITLE_18";
+
+	case CELL_GAME_PARAMID_TITLE_ID:                 return "TITLE_ID";
+	case CELL_GAME_PARAMID_VERSION:                  return "VERSION";
+	case CELL_GAME_PARAMID_PS3_SYSTEM_VER:           return "PS3_SYSTEM_VER";
+	case CELL_GAME_PARAMID_APP_VER:                  return "APP_VER";
+	}
+
+	return nullptr;
+}
+
+ppu_error_code cellGameGetParamString(s32 id, vm::ptr<char> buf, u32 bufsize)
 {
 	cellGame.warning("cellGameGetParamString(id=%d, buf=*0x%x, bufsize=%d)", id, buf, bufsize);
 
-	std::string key;
+	const auto prm = fxm::get<content_permission>();
 
-	switch(id)
+	if (!prm)
 	{
-	case CELL_GAME_PARAMID_TITLE:                    key = "TITLE";    break; // TODO: Is this value correct?
-	case CELL_GAME_PARAMID_TITLE_DEFAULT:            key = "TITLE";    break;
-	case CELL_GAME_PARAMID_TITLE_JAPANESE:           key = "TITLE_00"; break;
-	case CELL_GAME_PARAMID_TITLE_ENGLISH:            key = "TITLE_01"; break;
-	case CELL_GAME_PARAMID_TITLE_FRENCH:             key = "TITLE_02"; break;
-	case CELL_GAME_PARAMID_TITLE_SPANISH:            key = "TITLE_03"; break;
-	case CELL_GAME_PARAMID_TITLE_GERMAN:             key = "TITLE_04"; break;
-	case CELL_GAME_PARAMID_TITLE_ITALIAN:            key = "TITLE_05"; break;
-	case CELL_GAME_PARAMID_TITLE_DUTCH:              key = "TITLE_06"; break;
-	case CELL_GAME_PARAMID_TITLE_PORTUGUESE:         key = "TITLE_07"; break;
-	case CELL_GAME_PARAMID_TITLE_RUSSIAN:            key = "TITLE_08"; break;
-	case CELL_GAME_PARAMID_TITLE_KOREAN:             key = "TITLE_09"; break;
-	case CELL_GAME_PARAMID_TITLE_CHINESE_T:          key = "TITLE_10"; break;
-	case CELL_GAME_PARAMID_TITLE_CHINESE_S:          key = "TITLE_11"; break;
-	case CELL_GAME_PARAMID_TITLE_FINNISH:            key = "TITLE_12"; break;
-	case CELL_GAME_PARAMID_TITLE_SWEDISH:            key = "TITLE_13"; break;
-	case CELL_GAME_PARAMID_TITLE_DANISH:             key = "TITLE_14"; break;
-	case CELL_GAME_PARAMID_TITLE_NORWEGIAN:          key = "TITLE_15"; break;
-	case CELL_GAME_PARAMID_TITLE_POLISH:             key = "TITLE_16"; break;
-	case CELL_GAME_PARAMID_TITLE_PORTUGUESE_BRAZIL:  key = "TITLE_17"; break;
-	case CELL_GAME_PARAMID_TITLE_ENGLISH_UK:         key = "TITLE_18"; break;
+		return CELL_GAME_ERROR_FAILURE;
+	}
 
-	case CELL_GAME_PARAMID_TITLE_ID:                 key = "TITLE_ID";       break;
-	case CELL_GAME_PARAMID_VERSION:                  key = "PS3_SYSTEM_VER"; break;
-	case CELL_GAME_PARAMID_APP_VER:                  key = "APP_VER";        break;
+	const auto key = get_param_string_key(id);
 
-	default:
-		cellGame.error("cellGameGetParamString(): Unimplemented parameter (%d)", id);
+	if (!key)
+	{
 		return CELL_GAME_ERROR_INVALID_ID;
 	}
 
-	const std::string& value = Emu.GetPSF().at(key).as_string().substr(0, bufsize - 1);
+	std::string&& value = psf::get_string(prm->sfo, key);
+	value.resize(bufsize - 1);
 
 	std::copy_n(value.c_str(), value.size() + 1, buf.get_ptr());
 
 	return CELL_OK;
 }
 
-s32 cellGameSetParamString()
+ppu_error_code cellGameSetParamString(s32 id, vm::cptr<char> buf)
 {
-	UNIMPLEMENTED_FUNC(cellGame);
+	cellGame.warning("cellGameSetParamString(id=%d, buf=*0x%x)", id, buf);
+
+	const auto prm = fxm::get<content_permission>();
+
+	if (!prm)
+	{
+		return CELL_GAME_ERROR_FAILURE;
+	}
+
+	const auto key = get_param_string_key(id);
+
+	if (!key)
+	{
+		return CELL_GAME_ERROR_INVALID_ID;
+	}
+
+	u32 max_size = CELL_GAME_SYSP_TITLE_SIZE;
+
+	switch (id)
+	{
+	case CELL_GAME_PARAMID_TITLE_ID:       max_size = CELL_GAME_SYSP_TITLEID_SIZE; break;
+	case CELL_GAME_PARAMID_VERSION:        max_size = CELL_GAME_SYSP_VERSION_SIZE; break;
+	case CELL_GAME_PARAMID_PS3_SYSTEM_VER: max_size = CELL_GAME_SYSP_PS3_SYSTEM_VER_SIZE; break;
+	case CELL_GAME_PARAMID_APP_VER:        max_size = CELL_GAME_SYSP_APP_VER_SIZE; break;
+	}
+
+	prm->sfo.emplace(key, psf::string(max_size, buf.get_ptr()));
+
 	return CELL_OK;
 }
 
@@ -542,7 +610,7 @@ s32 cellGameGetLocalWebContentPath()
 	return CELL_OK;
 }
 
-s32 cellGameContentErrorDialog(s32 type, s32 errNeedSizeKB, vm::cptr<char> dirName)
+ppu_error_code cellGameContentErrorDialog(s32 type, s32 errNeedSizeKB, vm::cptr<char> dirName)
 {
 	cellGame.warning("cellGameContentErrorDialog(type=%d, errNeedSizeKB=%d, dirName=*0x%x)", type, errNeedSizeKB, dirName);
 
@@ -579,17 +647,20 @@ s32 cellGameContentErrorDialog(s32 type, s32 errNeedSizeKB, vm::cptr<char> dirNa
 	dlg->type.button_type = 2; // OK
 	dlg->type.disable_cancel = true;
 
-	const auto p = std::make_shared<std::promise<void>>();
-	std::future<void> future = p->get_future();
+	atomic_t<bool> result(false);
 
-	dlg->on_close = [=](s32 status)
+	dlg->on_close = [&](s32 status)
 	{
-		p->set_value();
+		result = true;
 	};
 
 	dlg->Create(errorMsg);
 
-	future.get();
+	while (!result)
+	{
+		CHECK_EMU_STATUS;
+		std::this_thread::sleep_for(1ms);
+	}
 
 	return CELL_OK;
 }
diff --git a/rpcs3/Emu/Cell/Modules/cellGame.h b/rpcs3/Emu/Cell/Modules/cellGame.h
index fbe714d5c4..b546c6bf70 100644
--- a/rpcs3/Emu/Cell/Modules/cellGame.h
+++ b/rpcs3/Emu/Cell/Modules/cellGame.h
@@ -1,5 +1,7 @@
 #pragma once
 
+#include "Emu/Cell/ErrorCodes.h"
+
 namespace vm { using namespace ps3; }
 
 // Return Codes
@@ -8,24 +10,54 @@ enum
 	CELL_GAME_RET_OK                   = 0,
 	CELL_GAME_RET_CANCEL               = 1,
 	CELL_GAME_RET_NONE                 = 2,
-	CELL_GAME_ERROR_NOTFOUND           = 0x8002cb04,
-	CELL_GAME_ERROR_BROKEN             = 0x8002cb05,
-	CELL_GAME_ERROR_INTERNAL           = 0x8002cb06,
-	CELL_GAME_ERROR_PARAM              = 0x8002cb07,
-	CELL_GAME_ERROR_NOAPP              = 0x8002cb08,
-	CELL_GAME_ERROR_ACCESS_ERROR       = 0x8002cb09,
-	CELL_GAME_ERROR_NOSPACE            = 0x8002cb20,
-	CELL_GAME_ERROR_NOTSUPPORTED       = 0x8002cb21,
-	CELL_GAME_ERROR_FAILURE            = 0x8002cb22,
-	CELL_GAME_ERROR_BUSY               = 0x8002cb23,
-	CELL_GAME_ERROR_IN_SHUTDOWN        = 0x8002cb24,
-	CELL_GAME_ERROR_INVALID_ID         = 0x8002cb25,
-	CELL_GAME_ERROR_EXIST              = 0x8002cb26,
-	CELL_GAME_ERROR_NOTPATCH           = 0x8002cb27,
-	CELL_GAME_ERROR_INVALID_THEME_FILE = 0x8002cb28,
-	CELL_GAME_ERROR_BOOTPATH           = 0x8002cb50,
 };
 
+enum CellGameError : s32
+{
+	CELL_GAME_ERROR_NOTFOUND           = ERROR_CODE(0x8002cb04),
+	CELL_GAME_ERROR_BROKEN             = ERROR_CODE(0x8002cb05),
+	CELL_GAME_ERROR_INTERNAL           = ERROR_CODE(0x8002cb06),
+	CELL_GAME_ERROR_PARAM              = ERROR_CODE(0x8002cb07),
+	CELL_GAME_ERROR_NOAPP              = ERROR_CODE(0x8002cb08),
+	CELL_GAME_ERROR_ACCESS_ERROR       = ERROR_CODE(0x8002cb09),
+	CELL_GAME_ERROR_NOSPACE            = ERROR_CODE(0x8002cb20),
+	CELL_GAME_ERROR_NOTSUPPORTED       = ERROR_CODE(0x8002cb21),
+	CELL_GAME_ERROR_FAILURE            = ERROR_CODE(0x8002cb22),
+	CELL_GAME_ERROR_BUSY               = ERROR_CODE(0x8002cb23),
+	CELL_GAME_ERROR_IN_SHUTDOWN        = ERROR_CODE(0x8002cb24),
+	CELL_GAME_ERROR_INVALID_ID         = ERROR_CODE(0x8002cb25),
+	CELL_GAME_ERROR_EXIST              = ERROR_CODE(0x8002cb26),
+	CELL_GAME_ERROR_NOTPATCH           = ERROR_CODE(0x8002cb27),
+	CELL_GAME_ERROR_INVALID_THEME_FILE = ERROR_CODE(0x8002cb28),
+	CELL_GAME_ERROR_BOOTPATH           = ERROR_CODE(0x8002cb50),
+};
+
+template<>
+inline const char* ppu_error_code::print(CellGameError error)
+{
+	switch (error)
+	{
+		STR_CASE(CELL_GAME_ERROR_NOTFOUND);
+		STR_CASE(CELL_GAME_ERROR_BROKEN);
+		STR_CASE(CELL_GAME_ERROR_INTERNAL);
+		STR_CASE(CELL_GAME_ERROR_PARAM);
+		STR_CASE(CELL_GAME_ERROR_NOAPP);
+		STR_CASE(CELL_GAME_ERROR_ACCESS_ERROR);
+		STR_CASE(CELL_GAME_ERROR_NOSPACE);
+		STR_CASE(CELL_GAME_ERROR_NOTSUPPORTED);
+		STR_CASE(CELL_GAME_ERROR_FAILURE);
+		STR_CASE(CELL_GAME_ERROR_BUSY);
+		STR_CASE(CELL_GAME_ERROR_IN_SHUTDOWN);
+		STR_CASE(CELL_GAME_ERROR_INVALID_ID);
+		STR_CASE(CELL_GAME_ERROR_EXIST);
+		STR_CASE(CELL_GAME_ERROR_NOTPATCH);
+		STR_CASE(CELL_GAME_ERROR_INVALID_THEME_FILE);
+		STR_CASE(CELL_GAME_ERROR_BOOTPATH);
+	}
+
+	return nullptr;
+}
+
 // Definitions
 enum
 {
@@ -37,6 +69,7 @@ enum
 	CELL_GAME_SYSP_TITLE_SIZE    = 128,
 	CELL_GAME_SYSP_TITLEID_SIZE  = 10,
 	CELL_GAME_SYSP_VERSION_SIZE  = 6,
+	CELL_GAME_SYSP_PS3_SYSTEM_VER_SIZE = 8,
 	CELL_GAME_SYSP_APP_VER_SIZE  = 6,
 
 	CELL_GAME_GAMETYPE_SYS  = 0,
@@ -90,6 +123,7 @@ enum
 	CELL_GAME_PARAMID_TITLE_ENGLISH_UK        = 20,
 	CELL_GAME_PARAMID_TITLE_ID                = 100,
 	CELL_GAME_PARAMID_VERSION                 = 101,
+	CELL_GAME_PARAMID_PS3_SYSTEM_VER          = 105,
 	CELL_GAME_PARAMID_APP_VER                 = 106,
 };
 
diff --git a/rpcs3/Emu/Cell/Modules/cellGcmSys.cpp b/rpcs3/Emu/Cell/Modules/cellGcmSys.cpp
index f1e856df33..cb290a71a1 100644
--- a/rpcs3/Emu/Cell/Modules/cellGcmSys.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellGcmSys.cpp
@@ -510,23 +510,13 @@ s32 cellGcmSetPrepareFlip(PPUThread& ppu, vm::ptr<CellGcmContextData> ctxt, u32
 			return res;
 		}
 	}
-#if 1
-	*ctxt->current++ = (GCM_FLIP_COMMAND << 2) | (1 << 18);
-	*ctxt->current++ = id;
+
+	const u32 cmd_size = rsx::make_command(ctxt->current, GCM_FLIP_COMMAND, { id });
 
 	if (ctxt.addr() == gcm_info.context_addr)
 	{
-		vm::_ref<CellGcmControl>(gcm_info.control_addr).put += 2 * sizeof(u32);
+		vm::_ref<CellGcmControl>(gcm_info.control_addr).put += cmd_size;
 	}
-#else
-	// internal compiler error, try to avoid it for now
-	u32 command_size = rsx::make_command(ctxt->current, GCM_FLIP_COMMAND, id);
-
-	if (ctxt.addr() == gcm_info.context_addr)
-	{
-		vm::_ref<CellGcmControl>(gcm_info.control_addr).put += command_size * sizeof(u32);
-	}
-#endif
 
 	return id;
 }
diff --git a/rpcs3/Emu/Cell/Modules/cellMsgDialog.cpp b/rpcs3/Emu/Cell/Modules/cellMsgDialog.cpp
index 9718fa9a95..880a126d1c 100644
--- a/rpcs3/Emu/Cell/Modules/cellMsgDialog.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellMsgDialog.cpp
@@ -6,8 +6,6 @@
 #include "cellSysutil.h"
 #include "cellMsgDialog.h"
 
-#include <future>
-
 extern _log::channel cellSysutil;
 
 s32 cellMsgDialogOpen()
@@ -103,21 +101,20 @@ s32 cellMsgDialogOpen2(u32 type, vm::cptr<char> msgString, vm::ptr<CellMsgDialog
 		}
 	};
 
-	// Make "shared" promise to workaround std::function limitation
-	auto spr = std::make_shared<std::promise<void>>();
-
-	// Get future
-	std::future<void> future = spr->get_future();
+	atomic_t<bool> result(false);
 
 	// Run asynchronously in GUI thread
-	Emu.CallAfter([&, spr = std::move(spr)]()
+	Emu.CallAfter([&]()
 	{
 		dlg->Create(msgString.get_ptr());
-		spr->set_value();
+		result = true;
 	});
 
-	// Wait for the "result"
-	future.get();
+	while (!result)
+	{
+		CHECK_EMU_STATUS;
+		std::this_thread::sleep_for(1ms);
+	}
 
 	return CELL_OK;
 }
@@ -219,7 +216,7 @@ s32 cellMsgDialogClose(f32 delay)
 
 	const u64 wait_until = get_system_time() + static_cast<s64>(std::max<float>(delay, 0.0f) * 1000);
 
-	thread_ctrl::spawn("MsgDialog Thread", [=]()
+	thread_ctrl::spawn("cellMsgDialogClose() Thread", [=]()
 	{
 		while (dlg->state == MsgDialogState::Open && get_system_time() < wait_until)
 		{
diff --git a/rpcs3/Emu/Cell/Modules/cellSaveData.cpp b/rpcs3/Emu/Cell/Modules/cellSaveData.cpp
index 333b4e86c6..c4a5d45132 100644
--- a/rpcs3/Emu/Cell/Modules/cellSaveData.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellSaveData.cpp
@@ -4,6 +4,8 @@
 
 #include "cellSaveData.h"
 
+#include "Loader/PSF.h"
+
 LOG_CHANNEL(cellSaveData);
 
 // cellSaveData aliases (only for cellSaveData.cpp)
diff --git a/rpcs3/Emu/Cell/Modules/cellSpursSpu.cpp b/rpcs3/Emu/Cell/Modules/cellSpursSpu.cpp
index 30c13c6242..7739b1efc3 100644
--- a/rpcs3/Emu/Cell/Modules/cellSpursSpu.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellSpursSpu.cpp
@@ -770,7 +770,7 @@ void spursSysServiceIdleHandler(SPUThread& spu, SpursKernelContext* ctxt)
 {
 	bool shouldExit;
 
-	std::unique_lock<std::mutex> lock(spu.mutex, std::defer_lock);
+	std::unique_lock<std::mutex> lock(get_current_thread_mutex(), std::defer_lock);
 
 	while (true)
 	{
@@ -861,7 +861,7 @@ void spursSysServiceIdleHandler(SPUThread& spu, SpursKernelContext* ctxt)
 			// The system service blocks by making a reservation and waiting on the lock line reservation lost event.
 			CHECK_EMU_STATUS;
 			if (!lock) lock.lock();
-			spu.cv.wait_for(lock, std::chrono::milliseconds(1));
+			get_current_thread_cv().wait_for(lock, std::chrono::milliseconds(1));
 			continue;
 		}
 
diff --git a/rpcs3/Emu/Cell/Modules/cellSync.cpp b/rpcs3/Emu/Cell/Modules/cellSync.cpp
index 83ada37380..eb1f4fc716 100644
--- a/rpcs3/Emu/Cell/Modules/cellSync.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellSync.cpp
@@ -6,6 +6,8 @@
 #include "Emu/Cell/lv2/sys_process.h"
 #include "cellSync.h"
 
+#include "Emu/Memory/wait_engine.h"
+
 LOG_CHANNEL(cellSync);
 
 namespace _sync
@@ -45,7 +47,7 @@ ppu_error_code cellSyncMutexInitialize(vm::ptr<CellSyncMutex> mutex)
 	return CELL_OK;
 }
 
-ppu_error_code cellSyncMutexLock(PPUThread& ppu, vm::ptr<CellSyncMutex> mutex)
+ppu_error_code cellSyncMutexLock(vm::ptr<CellSyncMutex> mutex)
 {
 	cellSync.trace("cellSyncMutexLock(mutex=*0x%x)", mutex);
 
@@ -63,7 +65,7 @@ ppu_error_code cellSyncMutexLock(PPUThread& ppu, vm::ptr<CellSyncMutex> mutex)
 	const auto order = mutex->ctrl.atomic_op(_sync::mutex_acquire);
 
 	// wait until rel value is equal to old acq value
-	vm::wait_op(ppu, mutex.addr(), 4, WRAP_EXPR(mutex->ctrl.load().rel == order));
+	vm::wait_op(mutex.addr(), 4, WRAP_EXPR(mutex->ctrl.load().rel == order));
 
 	_mm_mfence();
 
@@ -108,7 +110,7 @@ ppu_error_code cellSyncMutexUnlock(vm::ptr<CellSyncMutex> mutex)
 
 	mutex->ctrl.atomic_op(_sync::mutex_unlock);
 
-	vm::notify_at(mutex);
+	vm::notify_at(mutex.addr(), 4);
 
 	return CELL_OK;
 }
@@ -138,7 +140,7 @@ ppu_error_code cellSyncBarrierInitialize(vm::ptr<CellSyncBarrier> barrier, u16 t
 	return CELL_OK;
 }
 
-ppu_error_code cellSyncBarrierNotify(PPUThread& ppu, vm::ptr<CellSyncBarrier> barrier)
+ppu_error_code cellSyncBarrierNotify(vm::ptr<CellSyncBarrier> barrier)
 {
 	cellSync.trace("cellSyncBarrierNotify(barrier=*0x%x)", barrier);
 
@@ -152,9 +154,9 @@ ppu_error_code cellSyncBarrierNotify(PPUThread& ppu, vm::ptr<CellSyncBarrier> ba
 		return CELL_SYNC_ERROR_ALIGN;
 	}
 
-	vm::wait_op(ppu, barrier.addr(), 4, WRAP_EXPR(barrier->ctrl.atomic_op(_sync::barrier::try_notify)));
+	vm::wait_op(barrier.addr(), 4, WRAP_EXPR(barrier->ctrl.atomic_op(_sync::barrier::try_notify)));
 
-	vm::notify_at(barrier);
+	vm::notify_at(barrier.addr(), 4);
 
 	return CELL_OK;
 }
@@ -180,12 +182,12 @@ ppu_error_code cellSyncBarrierTryNotify(vm::ptr<CellSyncBarrier> barrier)
 		return NOT_AN_ERROR(CELL_SYNC_ERROR_BUSY);
 	}
 
-	vm::notify_at(barrier);
+	vm::notify_at(barrier.addr(), 4);
 
 	return CELL_OK;
 }
 
-ppu_error_code cellSyncBarrierWait(PPUThread& ppu, vm::ptr<CellSyncBarrier> barrier)
+ppu_error_code cellSyncBarrierWait(vm::ptr<CellSyncBarrier> barrier)
 {
 	cellSync.trace("cellSyncBarrierWait(barrier=*0x%x)", barrier);
 
@@ -201,9 +203,9 @@ ppu_error_code cellSyncBarrierWait(PPUThread& ppu, vm::ptr<CellSyncBarrier> barr
 
 	_mm_mfence();
 
-	vm::wait_op(ppu, barrier.addr(), 4, WRAP_EXPR(barrier->ctrl.atomic_op(_sync::barrier::try_wait)));
+	vm::wait_op(barrier.addr(), 4, WRAP_EXPR(barrier->ctrl.atomic_op(_sync::barrier::try_wait)));
 
-	vm::notify_at(barrier);
+	vm::notify_at(barrier.addr(), 4);
 
 	return CELL_OK;
 }
@@ -229,7 +231,7 @@ ppu_error_code cellSyncBarrierTryWait(vm::ptr<CellSyncBarrier> barrier)
 		return NOT_AN_ERROR(CELL_SYNC_ERROR_BUSY);
 	}
 
-	vm::notify_at(barrier);
+	vm::notify_at(barrier.addr(), 4);
 
 	return CELL_OK;
 }
@@ -263,7 +265,7 @@ ppu_error_code cellSyncRwmInitialize(vm::ptr<CellSyncRwm> rwm, vm::ptr<void> buf
 	return CELL_OK;
 }
 
-ppu_error_code cellSyncRwmRead(PPUThread& ppu, vm::ptr<CellSyncRwm> rwm, vm::ptr<void> buffer)
+ppu_error_code cellSyncRwmRead(vm::ptr<CellSyncRwm> rwm, vm::ptr<void> buffer)
 {
 	cellSync.trace("cellSyncRwmRead(rwm=*0x%x, buffer=*0x%x)", rwm, buffer);
 
@@ -278,7 +280,7 @@ ppu_error_code cellSyncRwmRead(PPUThread& ppu, vm::ptr<CellSyncRwm> rwm, vm::ptr
 	}
 
 	// wait until `writers` is zero, increase `readers`
-	vm::wait_op(ppu, rwm.addr(), 8, WRAP_EXPR(rwm->ctrl.atomic_op(_sync::rwlock::try_read_begin)));
+	vm::wait_op(rwm.addr(), 8, WRAP_EXPR(rwm->ctrl.atomic_op(_sync::rwlock::try_read_begin)));
 
 	// copy data to buffer
 	std::memcpy(buffer.get_ptr(), rwm->buffer.get_ptr(), rwm->size);
@@ -289,7 +291,7 @@ ppu_error_code cellSyncRwmRead(PPUThread& ppu, vm::ptr<CellSyncRwm> rwm, vm::ptr
 		return CELL_SYNC_ERROR_ABORT;
 	}
 
-	vm::notify_at(rwm.ptr(&CellSyncRwm::ctrl));
+	vm::notify_at(rwm.addr(), 8);
 
 	return CELL_OK;
 }
@@ -323,12 +325,12 @@ ppu_error_code cellSyncRwmTryRead(vm::ptr<CellSyncRwm> rwm, vm::ptr<void> buffer
 		return CELL_SYNC_ERROR_ABORT;
 	}
 
-	vm::notify_at(rwm.ptr(&CellSyncRwm::ctrl));
+	vm::notify_at(rwm.addr(), 8);
 
 	return CELL_OK;
 }
 
-ppu_error_code cellSyncRwmWrite(PPUThread& ppu, vm::ptr<CellSyncRwm> rwm, vm::cptr<void> buffer)
+ppu_error_code cellSyncRwmWrite(vm::ptr<CellSyncRwm> rwm, vm::cptr<void> buffer)
 {
 	cellSync.trace("cellSyncRwmWrite(rwm=*0x%x, buffer=*0x%x)", rwm, buffer);
 
@@ -343,10 +345,10 @@ ppu_error_code cellSyncRwmWrite(PPUThread& ppu, vm::ptr<CellSyncRwm> rwm, vm::cp
 	}
 
 	// wait until `writers` is zero, set to 1
-	vm::wait_op(ppu, rwm.addr(), 8, WRAP_EXPR(rwm->ctrl.atomic_op(_sync::rwlock::try_write_begin)));
+	vm::wait_op(rwm.addr(), 8, WRAP_EXPR(rwm->ctrl.atomic_op(_sync::rwlock::try_write_begin)));
 
 	// wait until `readers` is zero
-	vm::wait_op(ppu, rwm.addr(), 8, WRAP_EXPR(!rwm->ctrl.load().readers));
+	vm::wait_op(rwm.addr(), 8, WRAP_EXPR(!rwm->ctrl.load().readers));
 
 	// copy data from buffer
 	std::memcpy(rwm->buffer.get_ptr(), buffer.get_ptr(), rwm->size);
@@ -354,7 +356,7 @@ ppu_error_code cellSyncRwmWrite(PPUThread& ppu, vm::ptr<CellSyncRwm> rwm, vm::cp
 	// sync and clear `readers` and `writers`
 	rwm->ctrl.exchange({ 0, 0 });
 
-	vm::notify_at(rwm.ptr(&CellSyncRwm::ctrl));
+	vm::notify_at(rwm.addr(), 8);
 
 	return CELL_OK;
 }
@@ -385,7 +387,7 @@ ppu_error_code cellSyncRwmTryWrite(vm::ptr<CellSyncRwm> rwm, vm::cptr<void> buff
 	// sync and clear `readers` and `writers`
 	rwm->ctrl.exchange({ 0, 0 });
 
-	vm::notify_at(rwm.ptr(&CellSyncRwm::ctrl));
+	vm::notify_at(rwm.addr(), 8);
 
 	return CELL_OK;
 }
@@ -425,7 +427,7 @@ ppu_error_code cellSyncQueueInitialize(vm::ptr<CellSyncQueue> queue, vm::ptr<u8>
 	return CELL_OK;
 }
 
-ppu_error_code cellSyncQueuePush(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, vm::cptr<void> buffer)
+ppu_error_code cellSyncQueuePush(vm::ptr<CellSyncQueue> queue, vm::cptr<void> buffer)
 {
 	cellSync.trace("cellSyncQueuePush(queue=*0x%x, buffer=*0x%x)", queue, buffer);
 
@@ -443,7 +445,7 @@ ppu_error_code cellSyncQueuePush(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, v
 
 	u32 position;
 
-	vm::wait_op(ppu, queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_push_begin, depth, &position)));
+	vm::wait_op(queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_push_begin, depth, &position)));
 
 	// copy data from the buffer at the position
 	std::memcpy(&queue->buffer[position * queue->size], buffer.get_ptr(), queue->size);
@@ -451,7 +453,7 @@ ppu_error_code cellSyncQueuePush(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, v
 	// ...push_end
 	queue->ctrl.atomic_op([](_sync::queue& ctrl) { ctrl._push = 0; });
 
-	vm::notify_at(queue.ptr(&CellSyncQueue::ctrl));
+	vm::notify_at(queue.addr(), 8);
 
 	return CELL_OK;
 }
@@ -485,12 +487,12 @@ ppu_error_code cellSyncQueueTryPush(vm::ptr<CellSyncQueue> queue, vm::cptr<void>
 	// ...push_end
 	queue->ctrl.atomic_op([](_sync::queue& ctrl) { ctrl._push = 0; });
 
-	vm::notify_at(queue.ptr(&CellSyncQueue::ctrl));
+	vm::notify_at(queue.addr(), 8);
 
 	return CELL_OK;
 }
 
-ppu_error_code cellSyncQueuePop(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, vm::ptr<void> buffer)
+ppu_error_code cellSyncQueuePop(vm::ptr<CellSyncQueue> queue, vm::ptr<void> buffer)
 {
 	cellSync.trace("cellSyncQueuePop(queue=*0x%x, buffer=*0x%x)", queue, buffer);
 
@@ -508,7 +510,7 @@ ppu_error_code cellSyncQueuePop(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, vm
 	
 	u32 position;
 
-	vm::wait_op(ppu, queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_pop_begin, depth, &position)));
+	vm::wait_op(queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_pop_begin, depth, &position)));
 
 	// copy data at the position to the buffer
 	std::memcpy(buffer.get_ptr(), &queue->buffer[position % depth * queue->size], queue->size);
@@ -516,7 +518,7 @@ ppu_error_code cellSyncQueuePop(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, vm
 	// ...pop_end
 	queue->ctrl.atomic_op([](_sync::queue& ctrl) { ctrl._pop = 0; });
 
-	vm::notify_at(queue.ptr(&CellSyncQueue::ctrl));
+	vm::notify_at(queue.addr(), 8);
 
 	return CELL_OK;
 }
@@ -550,12 +552,12 @@ ppu_error_code cellSyncQueueTryPop(vm::ptr<CellSyncQueue> queue, vm::ptr<void> b
 	// ...pop_end
 	queue->ctrl.atomic_op([](_sync::queue& ctrl) { ctrl._pop = 0; });
 
-	vm::notify_at(queue.ptr(&CellSyncQueue::ctrl));
+	vm::notify_at(queue.addr(), 8);
 
 	return CELL_OK;
 }
 
-ppu_error_code cellSyncQueuePeek(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, vm::ptr<void> buffer)
+ppu_error_code cellSyncQueuePeek(vm::ptr<CellSyncQueue> queue, vm::ptr<void> buffer)
 {
 	cellSync.trace("cellSyncQueuePeek(queue=*0x%x, buffer=*0x%x)", queue, buffer);
 
@@ -573,7 +575,7 @@ ppu_error_code cellSyncQueuePeek(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, v
 
 	u32 position;
 
-	vm::wait_op(ppu, queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_peek_begin, depth, &position)));
+	vm::wait_op(queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_peek_begin, depth, &position)));
 
 	// copy data at the position to the buffer
 	std::memcpy(buffer.get_ptr(), &queue->buffer[position % depth * queue->size], queue->size);
@@ -581,7 +583,7 @@ ppu_error_code cellSyncQueuePeek(PPUThread& ppu, vm::ptr<CellSyncQueue> queue, v
 	// ...peek_end
 	queue->ctrl.atomic_op([](_sync::queue& ctrl) { ctrl._pop = 0; });
 
-	vm::notify_at(queue.ptr(&CellSyncQueue::ctrl));
+	vm::notify_at(queue.addr(), 8);
 
 	return CELL_OK;
 }
@@ -615,7 +617,7 @@ ppu_error_code cellSyncQueueTryPeek(vm::ptr<CellSyncQueue> queue, vm::ptr<void>
 	// ...peek_end
 	queue->ctrl.atomic_op([](_sync::queue& ctrl) { ctrl._pop = 0; });
 
-	vm::notify_at(queue.ptr(&CellSyncQueue::ctrl));
+	vm::notify_at(queue.addr(), 8);
 
 	return CELL_OK;
 }
@@ -639,7 +641,7 @@ ppu_error_code cellSyncQueueSize(vm::ptr<CellSyncQueue> queue)
 	return NOT_AN_ERROR(queue->ctrl.load().count & 0xffffff);
 }
 
-ppu_error_code cellSyncQueueClear(PPUThread& ppu, vm::ptr<CellSyncQueue> queue)
+ppu_error_code cellSyncQueueClear(vm::ptr<CellSyncQueue> queue)
 {
 	cellSync.trace("cellSyncQueueClear(queue=*0x%x)", queue);
 
@@ -655,12 +657,12 @@ ppu_error_code cellSyncQueueClear(PPUThread& ppu, vm::ptr<CellSyncQueue> queue)
 
 	const u32 depth = queue->check_depth();
 
-	vm::wait_op(ppu, queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_clear_begin_1)));
-	vm::wait_op(ppu, queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_clear_begin_2)));
+	vm::wait_op(queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_clear_begin_1)));
+	vm::wait_op(queue.addr(), 8, WRAP_EXPR(queue->ctrl.atomic_op(_sync::queue::try_clear_begin_2)));
 
 	queue->ctrl.exchange({ 0, 0 });
 
-	vm::notify_at(queue.ptr(&CellSyncQueue::ctrl));
+	vm::notify_at(queue.addr(), 8);
 
 	return CELL_OK;
 }
diff --git a/rpcs3/Emu/Cell/Modules/cellSysutil.cpp b/rpcs3/Emu/Cell/Modules/cellSysutil.cpp
index d5a9f94c12..0653f9fa8d 100644
--- a/rpcs3/Emu/Cell/Modules/cellSysutil.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellSysutil.cpp
@@ -80,9 +80,20 @@ static const char* get_systemparam_id_name(s32 id)
 	return tls_id_name;
 }
 
+enum class systemparam_id_name : s32 {};
+
+template<>
+struct unveil<systemparam_id_name, void>
+{
+	static inline const char* get(systemparam_id_name arg)
+	{
+		return get_systemparam_id_name((s32)arg);
+	}
+};
+
 s32 cellSysutilGetSystemParamInt(s32 id, vm::ptr<s32> value)
 {
-	cellSysutil.warning("cellSysutilGetSystemParamInt(id=%s, value=*0x%x)", get_systemparam_id_name(id), value);
+	cellSysutil.warning("cellSysutilGetSystemParamInt(id=%s, value=*0x%x)", systemparam_id_name(id), value);
 
 	// TODO: load this information from config (preferably "sys/" group)
 
diff --git a/rpcs3/Emu/Cell/Modules/cellVdec.cpp b/rpcs3/Emu/Cell/Modules/cellVdec.cpp
index 92f33a2d66..2192d6f4c9 100644
--- a/rpcs3/Emu/Cell/Modules/cellVdec.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellVdec.cpp
@@ -541,7 +541,7 @@ void vdecOpen(u32 vdec_id) // TODO: call from the constructor
 
 	vdec.vdecCb->cpu_init();
 	vdec.vdecCb->state -= cpu_state::stop;
-	vdec.vdecCb->safe_notify();
+	vdec.vdecCb->lock_notify();
 }
 
 s32 cellVdecQueryAttr(vm::cptr<CellVdecType> type, vm::ptr<CellVdecAttr> attr)
diff --git a/rpcs3/Emu/Cell/Modules/libmixer.cpp b/rpcs3/Emu/Cell/Modules/libmixer.cpp
index ac5b0ed831..e40ce2ffa7 100644
--- a/rpcs3/Emu/Cell/Modules/libmixer.cpp
+++ b/rpcs3/Emu/Cell/Modules/libmixer.cpp
@@ -447,7 +447,7 @@ s32 cellSurMixerCreate(vm::cptr<CellSurMixerConfig> config)
 
 	ppu->cpu_init();
 	ppu->state -= cpu_state::stop;
-	ppu->safe_notify();
+	ppu->lock_notify();
 
 	return CELL_OK;
 }
diff --git a/rpcs3/Emu/Cell/Modules/sys_net.cpp b/rpcs3/Emu/Cell/Modules/sys_net.cpp
index 9ebd8d7e4b..abe038f6c4 100644
--- a/rpcs3/Emu/Cell/Modules/sys_net.cpp
+++ b/rpcs3/Emu/Cell/Modules/sys_net.cpp
@@ -4,8 +4,6 @@
 #include "sys_net.h"
 
 #ifdef _WIN32
-#undef _WIN32_WINNT
-#define _WIN32_WINNT 0x0601
 #include <winsock2.h>
 #include <WS2tcpip.h>
 #else
@@ -122,7 +120,7 @@ namespace sys_net
 	// TODO
 	thread_local vm::ptr<_tls_data_t> g_tls_net_data{};
 
-	static void initialize_tls()
+	static never_inline void initialize_tls()
 	{
 		// allocate if not initialized
 		if (!g_tls_net_data)
diff --git a/rpcs3/Emu/Cell/Modules/sys_spinlock.cpp b/rpcs3/Emu/Cell/Modules/sys_spinlock.cpp
index 42397477b7..9ec0c65ff8 100644
--- a/rpcs3/Emu/Cell/Modules/sys_spinlock.cpp
+++ b/rpcs3/Emu/Cell/Modules/sys_spinlock.cpp
@@ -4,6 +4,8 @@
 
 #include "sysPrxForUser.h"
 
+#include "Emu/Memory/wait_engine.h"
+
 extern _log::channel sysPrxForUser;
 
 void sys_spinlock_initialize(vm::ptr<atomic_be_t<u32>> lock)
@@ -13,12 +15,12 @@ void sys_spinlock_initialize(vm::ptr<atomic_be_t<u32>> lock)
 	lock->exchange(0);
 }
 
-void sys_spinlock_lock(PPUThread& ppu, vm::ptr<atomic_be_t<u32>> lock)
+void sys_spinlock_lock(vm::ptr<atomic_be_t<u32>> lock)
 {
 	sysPrxForUser.trace("sys_spinlock_lock(lock=*0x%x)", lock);
 
 	// prx: exchange with 0xabadcafe, repeat until exchanged with 0
-	vm::wait_op(ppu, lock.addr(), 4, WRAP_EXPR(!lock->exchange(0xabadcafe)));
+	vm::wait_op(lock.addr(), 4, WRAP_EXPR(!lock->exchange(0xabadcafe)));
 }
 
 s32 sys_spinlock_trylock(vm::ptr<atomic_be_t<u32>> lock)
diff --git a/rpcs3/Emu/Cell/PPUCallback.cpp b/rpcs3/Emu/Cell/PPUCallback.cpp
index 8ca6a675e0..f1652f4d66 100644
--- a/rpcs3/Emu/Cell/PPUCallback.cpp
+++ b/rpcs3/Emu/Cell/PPUCallback.cpp
@@ -24,7 +24,7 @@ void CallbackManager::Async(async_cb_t func)
 
 	m_async_cb.emplace(std::move(func));
 
-	m_cb_thread->cv.notify_one();
+	m_cb_thread->notify();
 }
 
 CallbackManager::check_cb_t CallbackManager::Check()
@@ -74,7 +74,7 @@ void CallbackManager::Init()
 				continue;
 			}
 
-			ppu.cv.wait(lock);
+			get_current_thread_cv().wait(lock);
 		}
 	};
 
diff --git a/rpcs3/Emu/Cell/PPUFunction.cpp b/rpcs3/Emu/Cell/PPUFunction.cpp
index 8ad14ff3db..dfeaf54b00 100644
--- a/rpcs3/Emu/Cell/PPUFunction.cpp
+++ b/rpcs3/Emu/Cell/PPUFunction.cpp
@@ -2374,6 +2374,6 @@ s32 ppu_error_code::report(s32 error, const char* text)
 		}
 	}
 
-	LOG_ERROR(PPU, "Illegal call to ppu_report_error(0x%x, '%s')!");
+	LOG_ERROR(PPU, "Illegal call to ppu_error_code::report(0x%x, '%s')!");
 	return error;
 }
diff --git a/rpcs3/Emu/Cell/PPUInterpreter.cpp b/rpcs3/Emu/Cell/PPUInterpreter.cpp
index 1a33bad5a6..29aa118d61 100644
--- a/rpcs3/Emu/Cell/PPUInterpreter.cpp
+++ b/rpcs3/Emu/Cell/PPUInterpreter.cpp
@@ -56,7 +56,6 @@ public:
 }
 const g_ppu_scale_table;
 
-
 void ppu_interpreter::TDI(PPUThread& ppu, ppu_opcode_t op)
 {
 	const s64 a = ppu.GPR[op.ra], b = op.simm16;
@@ -242,79 +241,79 @@ void ppu_interpreter::VCMPBFP(PPUThread& ppu, ppu_opcode_t op)
 	const auto cmp1 = _mm_cmpnle_ps(a, b);
 	const auto cmp2 = _mm_cmpnge_ps(a, _mm_xor_ps(b, sign));
 	ppu.VR[op.vd].vf = _mm_or_ps(_mm_and_ps(cmp1, sign), _mm_and_ps(cmp2, _mm_castsi128_ps(_mm_set1_epi32(0x40000000))));
-	if (op.oe) ppu.SetCR(6, false, false, _mm_movemask_ps(_mm_or_ps(cmp1, cmp2)) == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, false, false, _mm_movemask_ps(_mm_or_ps(cmp1, cmp2)) == 0, false);
 }
 
 void ppu_interpreter::VCMPEQFP(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_ps(ppu.VR[op.vd].vf = _mm_cmpeq_ps(ppu.VR[op.va].vf, ppu.VR[op.vb].vf));
-	if (op.oe) ppu.SetCR(6, rmask == 0xf, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xf, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPEQUB(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8((ppu.VR[op.vd] = v128::eq8(ppu.VR[op.va], ppu.VR[op.vb])).vi);
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPEQUH(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8((ppu.VR[op.vd] = v128::eq16(ppu.VR[op.va], ppu.VR[op.vb])).vi);
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPEQUW(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8((ppu.VR[op.vd] = v128::eq32(ppu.VR[op.va], ppu.VR[op.vb])).vi);
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPGEFP(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_ps(ppu.VR[op.vd].vf = _mm_cmpge_ps(ppu.VR[op.va].vf, ppu.VR[op.vb].vf));
-	if (op.oe) ppu.SetCR(6, rmask == 0xf, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xf, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPGTFP(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_ps(ppu.VR[op.vd].vf = _mm_cmpgt_ps(ppu.VR[op.va].vf, ppu.VR[op.vb].vf));
-	if (op.oe) ppu.SetCR(6, rmask == 0xf, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xf, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPGTSB(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8(ppu.VR[op.vd].vi = _mm_cmpgt_epi8(ppu.VR[op.va].vi, ppu.VR[op.vb].vi));
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPGTSH(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8(ppu.VR[op.vd].vi = _mm_cmpgt_epi16(ppu.VR[op.va].vi, ppu.VR[op.vb].vi));
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPGTSW(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8(ppu.VR[op.vd].vi = _mm_cmpgt_epi32(ppu.VR[op.va].vi, ppu.VR[op.vb].vi));
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPGTUB(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8(ppu.VR[op.vd].vi = sse_cmpgt_epu8(ppu.VR[op.va].vi, ppu.VR[op.vb].vi));
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPGTUH(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8(ppu.VR[op.vd].vi = sse_cmpgt_epu16(ppu.VR[op.va].vi, ppu.VR[op.vb].vi));
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCMPGTUW(PPUThread& ppu, ppu_opcode_t op)
 {
 	const auto rmask = _mm_movemask_epi8(ppu.VR[op.vd].vi = sse_cmpgt_epu32(ppu.VR[op.va].vi, ppu.VR[op.vb].vi));
-	if (op.oe) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
+	if (UNLIKELY(op.oe)) ppu.SetCR(6, rmask == 0xffff, false, rmask == 0, false);
 }
 
 void ppu_interpreter::VCTSXS(PPUThread& ppu, ppu_opcode_t op)
@@ -1510,7 +1509,7 @@ void ppu_interpreter::ADDIC(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(a, i);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.main & 1) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.main & 1)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::ADDI(PPUThread& ppu, ppu_opcode_t op)
@@ -1649,19 +1648,19 @@ void ppu_interpreter::RLWIMI(PPUThread& ppu, ppu_opcode_t op)
 {
 	const u64 mask = ppu_rotate_mask(32 + op.mb32, 32 + op.me32);
 	ppu.GPR[op.ra] = (ppu.GPR[op.ra] & ~mask) | (dup32(rol32(u32(ppu.GPR[op.rs]), op.sh32)) & mask);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::RLWINM(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = dup32(rol32(u32(ppu.GPR[op.rs]), op.sh32)) & ppu_rotate_mask(32 + op.mb32, 32 + op.me32);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::RLWNM(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = dup32(rol32(u32(ppu.GPR[op.rs]), ppu.GPR[op.rb] & 0x1f)) & ppu_rotate_mask(32 + op.mb32, 32 + op.me32);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::ORI(PPUThread& ppu, ppu_opcode_t op)
@@ -1698,39 +1697,39 @@ void ppu_interpreter::ANDIS(PPUThread& ppu, ppu_opcode_t op)
 
 void ppu_interpreter::RLDICL(PPUThread& ppu, ppu_opcode_t op)
 {
-	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], op.sh64) & ppu_rotate_mask(op.mbe64, 63);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], op.sh64) & (~0ull >> op.mbe64);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::RLDICR(PPUThread& ppu, ppu_opcode_t op)
 {
-	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], op.sh64) & ppu_rotate_mask(0, op.mbe64);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], op.sh64) & (~0ull << (op.mbe64 ^ 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::RLDIC(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], op.sh64) & ppu_rotate_mask(op.mbe64, op.sh64 ^ 63);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::RLDIMI(PPUThread& ppu, ppu_opcode_t op)
 {
 	const u64 mask = ppu_rotate_mask(op.mbe64, op.sh64 ^ 63);
 	ppu.GPR[op.ra] = (ppu.GPR[op.ra] & ~mask) | (rol64(ppu.GPR[op.rs], op.sh64) & mask);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::RLDCL(PPUThread& ppu, ppu_opcode_t op)
 {
-	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], ppu.GPR[op.rb] & 0x3f) & ppu_rotate_mask(op.mbe64, 63);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], ppu.GPR[op.rb] & 0x3f) & (~0ull >> op.mbe64);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::RLDCR(PPUThread& ppu, ppu_opcode_t op)
 {
-	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], ppu.GPR[op.rb] & 0x3f) & ppu_rotate_mask(0, op.mbe64);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	ppu.GPR[op.ra] = rol64(ppu.GPR[op.rs], ppu.GPR[op.rb] & 0x3f) & (~0ull << (op.mbe64 ^ 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::CMP(PPUThread& ppu, ppu_opcode_t op)
@@ -1801,14 +1800,14 @@ void ppu_interpreter::SUBFC(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(~RA, RB, 1);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.oe) ppu.SetOV((~RA >> 63 == RB >> 63) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((~RA >> 63 == RB >> 63) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::MULHDU(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.rd] = UMULH64(ppu.GPR[op.ra], ppu.GPR[op.rb]);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
 }
 
 void ppu_interpreter::ADDC(PPUThread& ppu, ppu_opcode_t op)
@@ -1818,8 +1817,8 @@ void ppu_interpreter::ADDC(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(RA, RB);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.oe) ppu.SetOV((RA >> 63 == RB >> 63) && (RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((RA >> 63 == RB >> 63) && (RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::MULHWU(PPUThread& ppu, ppu_opcode_t op)
@@ -1827,7 +1826,7 @@ void ppu_interpreter::MULHWU(PPUThread& ppu, ppu_opcode_t op)
 	u32 a = (u32)ppu.GPR[op.ra];
 	u32 b = (u32)ppu.GPR[op.rb];
 	ppu.GPR[op.rd] = ((u64)a * (u64)b) >> 32;
-	if (op.rc) ppu.SetCR(0, false, false, false, ppu.SO);
+	if (UNLIKELY(op.rc)) ppu.SetCR(0, false, false, false, ppu.SO);
 }
 
 void ppu_interpreter::MFOCRF(PPUThread& ppu, ppu_opcode_t op)
@@ -1877,30 +1876,26 @@ void ppu_interpreter::LWZX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::SLW(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = u32(ppu.GPR[op.rs] << (ppu.GPR[op.rb] & 0x3f));
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::CNTLZW(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = cntlz32(u32(ppu.GPR[op.rs]));
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::SLD(PPUThread& ppu, ppu_opcode_t op)
 {
-	u32 n = ppu.GPR[op.rb] & 0x3f;
-	u64 r = rol64(ppu.GPR[op.rs], n);
-	u64 m = (ppu.GPR[op.rb] & 0x40) ? 0 : ppu_rotate_mask(0, 63 - n);
-
-	ppu.GPR[op.ra] = r & m;
-
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	const u32 n = ppu.GPR[op.rb];
+	ppu.GPR[op.ra] = UNLIKELY(n & 0x40) ? 0 : ppu.GPR[op.rs] << n;
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::AND(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = ppu.GPR[op.rs] & ppu.GPR[op.rb];
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::CMPL(PPUThread& ppu, ppu_opcode_t op)
@@ -1954,8 +1949,8 @@ void ppu_interpreter::SUBF(PPUThread& ppu, ppu_opcode_t op)
 	const u64 RA = ppu.GPR[op.ra];
 	const u64 RB = ppu.GPR[op.rb];
 	ppu.GPR[op.rd] = RB - RA;
-	if (op.oe) ppu.SetOV((~RA >> 63 == RB >> 63) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((~RA >> 63 == RB >> 63) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
 }
 
 void ppu_interpreter::LDUX(PPUThread& ppu, ppu_opcode_t op)
@@ -1979,13 +1974,13 @@ void ppu_interpreter::LWZUX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::CNTLZD(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = cntlz64(ppu.GPR[op.rs]);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::ANDC(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = ppu.GPR[op.rs] & ~ppu.GPR[op.rb];
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::TD(PPUThread& ppu, ppu_opcode_t op)
@@ -2012,7 +2007,7 @@ void ppu_interpreter::LVEWX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::MULHD(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.rd] = MULH64(ppu.GPR[op.ra], ppu.GPR[op.rb]);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
 }
 
 void ppu_interpreter::MULHW(PPUThread& ppu, ppu_opcode_t op)
@@ -2020,7 +2015,7 @@ void ppu_interpreter::MULHW(PPUThread& ppu, ppu_opcode_t op)
 	s32 a = (s32)ppu.GPR[op.ra];
 	s32 b = (s32)ppu.GPR[op.rb];
 	ppu.GPR[op.rd] = ((s64)a * (s64)b) >> 32;
-	if (op.rc) ppu.SetCR(0, false, false, false, ppu.SO);
+	if (UNLIKELY(op.rc)) ppu.SetCR(0, false, false, false, ppu.SO);
 }
 
 void ppu_interpreter::LDARX(PPUThread& ppu, ppu_opcode_t op)
@@ -2053,8 +2048,8 @@ void ppu_interpreter::NEG(PPUThread& ppu, ppu_opcode_t op)
 {
 	const u64 RA = ppu.GPR[op.ra];
 	ppu.GPR[op.rd] = 0 - RA;
-	if (op.oe) ppu.SetOV((~RA >> 63 == 0) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((~RA >> 63 == 0) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
 }
 
 void ppu_interpreter::LBZUX(PPUThread& ppu, ppu_opcode_t op)
@@ -2067,7 +2062,7 @@ void ppu_interpreter::LBZUX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::NOR(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = ~(ppu.GPR[op.rs] | ppu.GPR[op.rb]);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::STVEBX(PPUThread& ppu, ppu_opcode_t op)
@@ -2084,8 +2079,8 @@ void ppu_interpreter::SUBFE(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(~RA, RB, ppu.CA);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.oe) ppu.SetOV((~RA >> 63 == RB >> 63) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((~RA >> 63 == RB >> 63) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::ADDE(PPUThread& ppu, ppu_opcode_t op)
@@ -2095,8 +2090,8 @@ void ppu_interpreter::ADDE(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(RA, RB, ppu.CA);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.oe) ppu.SetOV((RA >> 63 == RB >> 63) && (RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((RA >> 63 == RB >> 63) && (RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::MTOCRF(PPUThread& ppu, ppu_opcode_t op)
@@ -2189,8 +2184,8 @@ void ppu_interpreter::SUBFZE(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(~RA, 0, ppu.CA);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.oe) ppu.SetOV((~RA >> 63 == 0) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((~RA >> 63 == 0) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::ADDZE(PPUThread& ppu, ppu_opcode_t op)
@@ -2199,8 +2194,8 @@ void ppu_interpreter::ADDZE(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(RA, 0, ppu.CA);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.oe) ppu.SetOV((RA >> 63 == 0) && (RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((RA >> 63 == 0) && (RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::STDCX(PPUThread& ppu, ppu_opcode_t op)
@@ -2228,12 +2223,12 @@ void ppu_interpreter::MULLD(PPUThread& ppu, ppu_opcode_t op)
 	const s64 RA = ppu.GPR[op.ra];
 	const s64 RB = ppu.GPR[op.rb];
 	ppu.GPR[op.rd] = (s64)(RA * RB);
-	if (op.oe)
+	if (UNLIKELY(op.oe))
 	{
 		const s64 high = MULH64(RA, RB);
 		ppu.SetOV(high != s64(ppu.GPR[op.rd]) >> 63);
 	}
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
 }
 
 void ppu_interpreter::SUBFME(PPUThread& ppu, ppu_opcode_t op)
@@ -2242,8 +2237,8 @@ void ppu_interpreter::SUBFME(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(~RA, ~0ull, ppu.CA);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.oe) ppu.SetOV((~RA >> 63 == 1) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((~RA >> 63 == 1) && (~RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::ADDME(PPUThread& ppu, ppu_opcode_t op)
@@ -2252,15 +2247,15 @@ void ppu_interpreter::ADDME(PPUThread& ppu, ppu_opcode_t op)
 	const auto r = add64_flags(RA, ~0ull, ppu.CA);
 	ppu.GPR[op.rd] = r.result;
 	ppu.CA = r.carry;
-	if (op.oe) ppu.SetOV((u64(RA) >> 63 == 1) && (u64(RA) >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, r.result, 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((u64(RA) >> 63 == 1) && (u64(RA) >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, r.result, 0);
 }
 
 void ppu_interpreter::MULLW(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.rd] = (s64)((s64)(s32)ppu.GPR[op.ra] * (s64)(s32)ppu.GPR[op.rb]);
-	if (op.oe) ppu.SetOV(s64(ppu.GPR[op.rd]) < s64(-1) << 31 || s64(ppu.GPR[op.rd]) >= s64(1) << 31);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV(s64(ppu.GPR[op.rd]) < s64(-1) << 31 || s64(ppu.GPR[op.rd]) >= s64(1) << 31);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::DCBTST(PPUThread& ppu, ppu_opcode_t op)
@@ -2279,8 +2274,8 @@ void ppu_interpreter::ADD(PPUThread& ppu, ppu_opcode_t op)
 	const u64 RA = ppu.GPR[op.ra];
 	const u64 RB = ppu.GPR[op.rb];
 	ppu.GPR[op.rd] = RA + RB;
-	if (op.oe) ppu.SetOV((RA >> 63 == RB >> 63) && (RA >> 63 != ppu.GPR[op.rd] >> 63));
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV((RA >> 63 == RB >> 63) && (RA >> 63 != ppu.GPR[op.rd] >> 63));
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
 }
 
 void ppu_interpreter::DCBT(PPUThread& ppu, ppu_opcode_t op)
@@ -2296,7 +2291,7 @@ void ppu_interpreter::LHZX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::EQV(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = ~(ppu.GPR[op.rs] ^ ppu.GPR[op.rb]);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::ECIWX(PPUThread& ppu, ppu_opcode_t op)
@@ -2314,7 +2309,7 @@ void ppu_interpreter::LHZUX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::XOR(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = ppu.GPR[op.rs] ^ ppu.GPR[op.rb];
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::MFSPR(PPUThread& ppu, ppu_opcode_t op)
@@ -2396,7 +2391,7 @@ void ppu_interpreter::STHX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::ORC(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = ppu.GPR[op.rs] | ~ppu.GPR[op.rb];
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::ECOWX(PPUThread& ppu, ppu_opcode_t op)
@@ -2414,7 +2409,7 @@ void ppu_interpreter::STHUX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::OR(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = ppu.GPR[op.rs] | ppu.GPR[op.rb];
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::DIVDU(PPUThread& ppu, ppu_opcode_t op)
@@ -2422,8 +2417,8 @@ void ppu_interpreter::DIVDU(PPUThread& ppu, ppu_opcode_t op)
 	const u64 RA = ppu.GPR[op.ra];
 	const u64 RB = ppu.GPR[op.rb];
 	ppu.GPR[op.rd] = RB == 0 ? 0 : RA / RB;
-	if (op.oe) ppu.SetOV(RB == 0);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV(RB == 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
 }
 
 void ppu_interpreter::DIVWU(PPUThread& ppu, ppu_opcode_t op)
@@ -2431,8 +2426,8 @@ void ppu_interpreter::DIVWU(PPUThread& ppu, ppu_opcode_t op)
 	const u32 RA = (u32)ppu.GPR[op.ra];
 	const u32 RB = (u32)ppu.GPR[op.rb];
 	ppu.GPR[op.rd] = RB == 0 ? 0 : RA / RB;
-	if (op.oe) ppu.SetOV(RB == 0);
-	if (op.rc) ppu.SetCR(0, false, false, false, ppu.SO);
+	if (UNLIKELY(op.oe)) ppu.SetOV(RB == 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR(0, false, false, false, ppu.SO);
 }
 
 void ppu_interpreter::MTSPR(PPUThread& ppu, ppu_opcode_t op)
@@ -2465,7 +2460,7 @@ void ppu_interpreter::DCBI(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::NAND(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = ~(ppu.GPR[op.rs] & ppu.GPR[op.rb]);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::STVXL(PPUThread& ppu, ppu_opcode_t op)
@@ -2480,8 +2475,8 @@ void ppu_interpreter::DIVD(PPUThread& ppu, ppu_opcode_t op)
 	const s64 RB = ppu.GPR[op.rb];
 	const bool o = RB == 0 || ((u64)RA == (1ULL << 63) && RB == -1);
 	ppu.GPR[op.rd] = o ? 0 : RA / RB;
-	if (op.oe) ppu.SetOV(o);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
+	if (UNLIKELY(op.oe)) ppu.SetOV(o);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.rd], 0);
 }
 
 void ppu_interpreter::DIVW(PPUThread& ppu, ppu_opcode_t op)
@@ -2490,8 +2485,8 @@ void ppu_interpreter::DIVW(PPUThread& ppu, ppu_opcode_t op)
 	const s32 RB = (s32)ppu.GPR[op.rb];
 	const bool o = RB == 0 || ((u32)RA == (1 << 31) && RB == -1);
 	ppu.GPR[op.rd] = o ? 0 : u32(RA / RB);
-	if (op.oe) ppu.SetOV(o);
-	if (op.rc) ppu.SetCR(0, false, false, false, ppu.SO);
+	if (UNLIKELY(op.oe)) ppu.SetOV(o);
+	if (UNLIKELY(op.rc)) ppu.SetCR(0, false, false, false, ppu.SO);
 }
 
 void ppu_interpreter::LVLX(PPUThread& ppu, ppu_opcode_t op)
@@ -2544,17 +2539,14 @@ void ppu_interpreter::LFSX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::SRW(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = (ppu.GPR[op.rs] & 0xffffffff) >> (ppu.GPR[op.rb] & 0x3f);
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::SRD(PPUThread& ppu, ppu_opcode_t op)
 {
-	u32 n = ppu.GPR[op.rb] & 0x3f;
-	u64 r = rol64(ppu.GPR[op.rs], 64 - n);
-	u64 m = (ppu.GPR[op.rb] & 0x40) ? 0 : ppu_rotate_mask(n, 63);
-	ppu.GPR[op.ra] = r & m;
-
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	const u32 n = ppu.GPR[op.rb];
+	ppu.GPR[op.ra] = UNLIKELY(n & 0x40) ? 0 : ppu.GPR[op.rs] >> n;
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::LVRX(PPUThread& ppu, ppu_opcode_t op)
@@ -2754,7 +2746,7 @@ void ppu_interpreter::SRAW(PPUThread& ppu, ppu_opcode_t op)
 		ppu.CA = (RS < 0) && ((ppu.GPR[op.ra] << shift) != RS);
 	}
 
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::SRAD(PPUThread& ppu, ppu_opcode_t op)
@@ -2772,7 +2764,7 @@ void ppu_interpreter::SRAD(PPUThread& ppu, ppu_opcode_t op)
 		ppu.CA = (RS < 0) && ((ppu.GPR[op.ra] << shift) != RS);
 	}
 
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::LVRXL(PPUThread& ppu, ppu_opcode_t op)
@@ -2794,7 +2786,7 @@ void ppu_interpreter::SRAWI(PPUThread& ppu, ppu_opcode_t op)
 	ppu.GPR[op.ra] = RS >> op.sh32;
 	ppu.CA = (RS < 0) && ((u32)(ppu.GPR[op.ra] << op.sh32) != RS);
 
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::SRADI(PPUThread& ppu, ppu_opcode_t op)
@@ -2804,7 +2796,7 @@ void ppu_interpreter::SRADI(PPUThread& ppu, ppu_opcode_t op)
 	ppu.GPR[op.ra] = RS >> sh;
 	ppu.CA = (RS < 0) && ((ppu.GPR[op.ra] << sh) != RS);
 
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::EIEIO(PPUThread& ppu, ppu_opcode_t op)
@@ -2829,7 +2821,7 @@ void ppu_interpreter::STHBRX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::EXTSH(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = (s64)(s16)ppu.GPR[op.rs];
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::STVRXL(PPUThread& ppu, ppu_opcode_t op)
@@ -2843,7 +2835,7 @@ void ppu_interpreter::STVRXL(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::EXTSB(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = (s64)(s8)ppu.GPR[op.rs];
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::STFIWX(PPUThread& ppu, ppu_opcode_t op)
@@ -2855,7 +2847,7 @@ void ppu_interpreter::STFIWX(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::EXTSW(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.GPR[op.ra] = (s64)(s32)ppu.GPR[op.rs];
-	if (op.rc) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
+	if (UNLIKELY(op.rc)) ppu.SetCR<s64>(0, ppu.GPR[op.ra], 0);
 }
 
 void ppu_interpreter::ICBI(PPUThread& ppu, ppu_opcode_t op)
@@ -3049,66 +3041,6 @@ void ppu_interpreter::LWA(PPUThread& ppu, ppu_opcode_t op)
 	ppu.GPR[op.rd] = (s64)(s32)vm::read32(vm::cast(addr, HERE));
 }
 
-void ppu_interpreter::FDIVS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] / ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FSUBS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] - ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FADDS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] + ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FSQRTS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(sqrt(ppu.FPR[op.frb]));
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FRES(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(1.0 / ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FMULS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] * ppu.FPR[op.frc]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FMADDS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] * ppu.FPR[op.frc] + ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FMSUBS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] * ppu.FPR[op.frc] - ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FNMSUBS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(-(ppu.FPR[op.fra] * ppu.FPR[op.frc]) + ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
-void ppu_interpreter::FNMADDS(PPUThread& ppu, ppu_opcode_t op)
-{
-	ppu.FPR[op.frd] = f32(-(ppu.FPR[op.fra] * ppu.FPR[op.frc]) - ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
-}
-
 void ppu_interpreter::STD(PPUThread& ppu, ppu_opcode_t op)
 {
 	const u64 addr = (op.simm16 & ~3) + (op.ra ? ppu.GPR[op.ra] : 0);
@@ -3122,10 +3054,70 @@ void ppu_interpreter::STDU(PPUThread& ppu, ppu_opcode_t op)
 	ppu.GPR[op.ra] = addr;
 }
 
+void ppu_interpreter::FDIVS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] / ppu.FPR[op.frb]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FSUBS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] - ppu.FPR[op.frb]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FADDS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] + ppu.FPR[op.frb]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FSQRTS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(sqrt(ppu.FPR[op.frb]));
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FRES(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(1.0 / ppu.FPR[op.frb]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FMULS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] * ppu.FPR[op.frc]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FMADDS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] * ppu.FPR[op.frc] + ppu.FPR[op.frb]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FMSUBS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(ppu.FPR[op.fra] * ppu.FPR[op.frc] - ppu.FPR[op.frb]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FNMSUBS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(-(ppu.FPR[op.fra] * ppu.FPR[op.frc]) + ppu.FPR[op.frb]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
+void ppu_interpreter::FNMADDS(PPUThread& ppu, ppu_opcode_t op)
+{
+	ppu.FPR[op.frd] = f32(-(ppu.FPR[op.fra] * ppu.FPR[op.frc]) - ppu.FPR[op.frb]);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+}
+
 void ppu_interpreter::MTFSB1(PPUThread& ppu, ppu_opcode_t op)
 {
 	LOG_WARNING(PPU, "MTFSB1");
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::MCRFS(PPUThread& ppu, ppu_opcode_t op)
@@ -3137,26 +3129,26 @@ void ppu_interpreter::MCRFS(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::MTFSB0(PPUThread& ppu, ppu_opcode_t op)
 {
 	LOG_WARNING(PPU, "MTFSB0");
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::MTFSFI(PPUThread& ppu, ppu_opcode_t op)
 {
 	LOG_WARNING(PPU, "MTFSFI");
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::MFFS(PPUThread& ppu, ppu_opcode_t op)
 {
 	LOG_WARNING(PPU, "MFFS");
 	ppu.FPR[op.frd] = 0.0;
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::MTFSF(PPUThread& ppu, ppu_opcode_t op)
 {
 	LOG_WARNING(PPU, "MTFSF");
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FCMPU(PPUThread& ppu, ppu_opcode_t op)
@@ -3173,85 +3165,85 @@ void ppu_interpreter::FCMPU(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::FRSP(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = f32(ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FCTIW(PPUThread& ppu, ppu_opcode_t op)
 {
 	(s32&)ppu.FPR[op.frd] = lrint(ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FCTIWZ(PPUThread& ppu, ppu_opcode_t op)
 {
 	(s32&)ppu.FPR[op.frd] = static_cast<s32>(ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FDIV(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = ppu.FPR[op.fra] / ppu.FPR[op.frb];
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FSUB(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = ppu.FPR[op.fra] - ppu.FPR[op.frb];
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FADD(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = ppu.FPR[op.fra] + ppu.FPR[op.frb];
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FSQRT(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = sqrt(ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FSEL(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = ppu.FPR[op.fra] >= 0.0 ? ppu.FPR[op.frc] : ppu.FPR[op.frb];
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FMUL(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = ppu.FPR[op.fra] * ppu.FPR[op.frc];
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FRSQRTE(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = 1.0 / sqrt(ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FMSUB(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = ppu.FPR[op.fra] * ppu.FPR[op.frc] - ppu.FPR[op.frb];
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FMADD(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = ppu.FPR[op.fra] * ppu.FPR[op.frc] + ppu.FPR[op.frb];
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FNMSUB(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = -(ppu.FPR[op.fra] * ppu.FPR[op.frc]) + ppu.FPR[op.frb];
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FNMADD(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = -(ppu.FPR[op.fra] * ppu.FPR[op.frc]) - ppu.FPR[op.frb];
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FCMPO(PPUThread& ppu, ppu_opcode_t op)
@@ -3262,43 +3254,43 @@ void ppu_interpreter::FCMPO(PPUThread& ppu, ppu_opcode_t op)
 void ppu_interpreter::FNEG(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = -ppu.FPR[op.frb];
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FMR(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = ppu.FPR[op.frb];
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FNABS(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = -fabs(ppu.FPR[op.frb]);
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FABS(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = fabs(ppu.FPR[op.frb]);
-	if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FCTID(PPUThread& ppu, ppu_opcode_t op)
 {
 	(s64&)ppu.FPR[op.frd] = llrint(ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FCTIDZ(PPUThread& ppu, ppu_opcode_t op)
 {
 	(s64&)ppu.FPR[op.frd] = static_cast<s64>(ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 void ppu_interpreter::FCFID(PPUThread& ppu, ppu_opcode_t op)
 {
 	ppu.FPR[op.frd] = static_cast<double>((s64&)ppu.FPR[op.frb]);
-	ASSERT(!op.rc); //if (op.rc) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
+	ASSERT(!op.rc); //if (UNLIKELY(op.rc)) ppu.SetCR(1, ppu.FG, ppu.FL, ppu.FE, ppu.FU);
 }
 
 
diff --git a/rpcs3/Emu/Cell/PPUThread.cpp b/rpcs3/Emu/Cell/PPUThread.cpp
index ccce38e9e6..2e287190aa 100644
--- a/rpcs3/Emu/Cell/PPUThread.cpp
+++ b/rpcs3/Emu/Cell/PPUThread.cpp
@@ -108,15 +108,15 @@ void PPUThread::cpu_task()
 
 	while (true)
 	{
-		if (_pc == PC && !state.load())
+		if (LIKELY(_pc == PC && !state.load()))
 		{
 			func0(*this, { op0 });
-			
-			if ((_pc += 4) == (PC += 4) && !state.load())
+
+			if (LIKELY((_pc += 4) == (PC += 4) && !state.load()))
 			{
 				func1(*this, { op1 });
 
-				if ((_pc += 4) == (PC += 4))
+				if (LIKELY((_pc += 4) == (PC += 4)))
 				{
 					op0 = op2;
 					func0 = func2;
@@ -135,7 +135,7 @@ void PPUThread::cpu_task()
 		func1 = table[ppu_decode(op1 = ops[1])];
 		func2 = table[ppu_decode(op2 = ops[2])];
 
-		if (check_status()) return;
+		if (UNLIKELY(check_status())) return;
 	}
 }
 
@@ -152,6 +152,11 @@ PPUThread::~PPUThread()
 	}
 }
 
+PPUThread::PPUThread(const std::string& name)
+	: cpu_thread(cpu_type::ppu, name)
+{
+}
+
 be_t<u64>* PPUThread::get_stack_arg(s32 i, u64 align)
 {
 	if (align != 1 && align != 2 && align != 4 && align != 8 && align != 16) throw fmt::exception("Unsupported alignment: 0x%llx" HERE, align);
diff --git a/rpcs3/Emu/Cell/PPUThread.h b/rpcs3/Emu/Cell/PPUThread.h
index 4c007110e8..8507f54c44 100644
--- a/rpcs3/Emu/Cell/PPUThread.h
+++ b/rpcs3/Emu/Cell/PPUThread.h
@@ -12,14 +12,10 @@ public:
 	virtual void cpu_init() override;
 	virtual void cpu_task() override;
 	virtual bool handle_interrupt() override;
-
-	PPUThread(const std::string& name)
-		: cpu_thread(cpu_type::ppu, name)
-	{
-	}
-
 	virtual ~PPUThread() override;
 
+	PPUThread(const std::string& name);
+
 	u64 GPR[32]{}; // General-Purpose Registers
 	f64 FPR[32]{}; // Floating Point Registers
 	v128 VR[32]{}; // Vector Registers
diff --git a/rpcs3/Emu/Cell/RawSPUThread.cpp b/rpcs3/Emu/Cell/RawSPUThread.cpp
index 70294626e9..60033cfa2f 100644
--- a/rpcs3/Emu/Cell/RawSPUThread.cpp
+++ b/rpcs3/Emu/Cell/RawSPUThread.cpp
@@ -24,6 +24,24 @@ void RawSPUThread::cpu_task()
 	npc = pc | ((ch_event_stat & SPU_EVENT_INTR_ENABLED) != 0);
 }
 
+void RawSPUThread::on_init()
+{
+	if (!offset)
+	{
+		// Install correct SPU index and LS address
+		const_cast<u32&>(index) = id;
+		const_cast<u32&>(offset) = vm::falloc(RAW_SPU_BASE_ADDR + RAW_SPU_OFFSET * index, 0x40000);
+		ASSERT(offset);
+
+		SPUThread::on_init();
+	}
+}
+
+RawSPUThread::RawSPUThread(const std::string& name)
+	: SPUThread(name)
+{
+}
+
 bool RawSPUThread::read_reg(const u32 addr, u32& value)
 {
 	const u32 offset = addr - RAW_SPU_BASE_ADDR - index * RAW_SPU_OFFSET - RAW_SPU_PROB_OFFSET;
@@ -83,7 +101,7 @@ bool RawSPUThread::write_reg(const u32 addr, const u32 value)
 		}))
 		{
 			state -= cpu_state::stop;
-			safe_notify();
+			lock_notify();
 		}
 	};
 
diff --git a/rpcs3/Emu/Cell/RawSPUThread.h b/rpcs3/Emu/Cell/RawSPUThread.h
index c3a27aa46f..5e37357e0d 100644
--- a/rpcs3/Emu/Cell/RawSPUThread.h
+++ b/rpcs3/Emu/Cell/RawSPUThread.h
@@ -14,23 +14,9 @@ public:
 	static constexpr u32 id_min = 0;
 	static constexpr u32 id_max = 4;
 
-	void on_init() override
-	{
-		if (!offset)
-		{
-			// Install correct SPU index and LS address
-			const_cast<u32&>(index) = id;
-			const_cast<u32&>(offset) = vm::falloc(RAW_SPU_BASE_ADDR + RAW_SPU_OFFSET * index, 0x40000);
-			ASSERT(offset);
+	void on_init() override;
 
-			SPUThread::on_init();
-		}
-	}
-
-	RawSPUThread(const std::string& name)
-		: SPUThread(name)
-	{
-	}
+	RawSPUThread(const std::string& name);
 
 	bool read_reg(const u32 addr, u32& value);
 	bool write_reg(const u32 addr, const u32 value);
diff --git a/rpcs3/Emu/Cell/SPUAnalyser.cpp b/rpcs3/Emu/Cell/SPUAnalyser.cpp
index 57cc2d4b14..6a616519c2 100644
--- a/rpcs3/Emu/Cell/SPUAnalyser.cpp
+++ b/rpcs3/Emu/Cell/SPUAnalyser.cpp
@@ -58,7 +58,7 @@ std::shared_ptr<spu_function_t> SPUDatabase::analyse(const be_t<u32>* ls, u32 en
 		}
 	}
 
-	std::lock_guard<shared_mutex> lock(m_mutex);
+	writer_lock lock(m_mutex);
 
 	// Double-check
 	if (auto func = find(ls + entry / 4, key, max_limit - entry))
diff --git a/rpcs3/Emu/Cell/SPUThread.cpp b/rpcs3/Emu/Cell/SPUThread.cpp
index 1d8c1bf3d1..2325e22dfd 100644
--- a/rpcs3/Emu/Cell/SPUThread.cpp
+++ b/rpcs3/Emu/Cell/SPUThread.cpp
@@ -16,6 +16,8 @@
 #include "Emu/Cell/SPUInterpreter.h"
 #include "Emu/Cell/SPURecompiler.h"
 
+#include "Emu/Memory/wait_engine.h"
+
 #include <cfenv>
 
 extern u64 get_timebased_time();
@@ -52,8 +54,7 @@ void spu_int_ctrl_t::set(u64 ints)
 		if (tag && tag->handler)
 		{
 			tag->handler->signal++;
-
-			tag->handler->thread->cv.notify_one();
+			tag->handler->thread->notify();
 		}
 	}
 }
@@ -167,7 +168,20 @@ void SPUThread::cpu_task()
 	}
 }
 
-SPUThread::SPUThread(const std::string & name, u32 index)
+SPUThread::~SPUThread()
+{
+	// Deallocate Local Storage
+	vm::dealloc_verbose_nothrow(offset);
+}
+
+SPUThread::SPUThread(const std::string& name)
+	: cpu_thread(cpu_type::spu, name)
+	, index(0)
+	, offset(0)
+{
+}
+
+SPUThread::SPUThread(const std::string& name, u32 index)
 	: cpu_thread(cpu_type::spu, name)
 	, index(index)
 	, offset(vm::alloc(0x40000, vm::main))
@@ -175,12 +189,6 @@ SPUThread::SPUThread(const std::string & name, u32 index)
 	Ensures(offset);
 }
 
-SPUThread::~SPUThread()
-{
-	// Deallocate Local Storage
-	vm::dealloc_verbose_nothrow(offset);
-}
-
 void SPUThread::push_snr(u32 number, u32 value)
 {
 	// Get channel
@@ -430,7 +438,7 @@ void SPUThread::process_mfc_cmd(u32 cmd)
 u32 SPUThread::get_events(bool waiting)
 {
 	// check reservation status and set SPU_EVENT_LR if lost
-	if (last_raddr != 0 && !vm::reservation_test(get_thread_ctrl()))
+	if (last_raddr != 0 && !vm::reservation_test(operator->()))
 	{
 		ch_event_stat |= SPU_EVENT_LR;
 
@@ -473,11 +481,11 @@ void SPUThread::set_events(u32 mask)
 	// notify if some events were set
 	if (~old_stat & mask && old_stat & SPU_EVENT_WAITING)
 	{
-		std::lock_guard<std::mutex> lock(mutex);
+		std::lock_guard<std::mutex> lock(get_current_thread_mutex());
 
 		if (ch_event_stat & SPU_EVENT_WAITING)
 		{
-			cv.notify_one();
+			notify();
 		}
 	}
 }
@@ -530,7 +538,7 @@ bool SPUThread::get_ch_value(u32 ch, u32& out)
 
 	auto read_channel = [&](spu_channel_t& channel)
 	{
-		std::unique_lock<std::mutex> lock(mutex, std::defer_lock);
+		std::unique_lock<std::mutex> lock(get_current_thread_mutex(), std::defer_lock);
 
 		while (true)
 		{
@@ -552,7 +560,7 @@ bool SPUThread::get_ch_value(u32 ch, u32& out)
 				continue;
 			}
 
-			cv.wait(lock);
+			get_current_thread_cv().wait(lock);
 		}
 	};
 
@@ -563,7 +571,7 @@ bool SPUThread::get_ch_value(u32 ch, u32& out)
 	//	break;
 	case SPU_RdInMbox:
 	{
-		std::unique_lock<std::mutex> lock(mutex, std::defer_lock);
+		std::unique_lock<std::mutex> lock(get_current_thread_mutex(), std::defer_lock);
 
 		while (true)
 		{
@@ -590,7 +598,7 @@ bool SPUThread::get_ch_value(u32 ch, u32& out)
 				continue;
 			}
 
-			cv.wait(lock);
+			get_current_thread_cv().wait(lock);
 		}
 	}
 
@@ -639,7 +647,7 @@ bool SPUThread::get_ch_value(u32 ch, u32& out)
 
 	case SPU_RdEventStat:
 	{
-		std::unique_lock<std::mutex> lock(mutex, std::defer_lock);
+		std::unique_lock<std::mutex> lock(get_current_thread_mutex(), std::defer_lock);
 
 		// start waiting or return immediately
 		if (u32 res = get_events(true))
@@ -651,7 +659,7 @@ bool SPUThread::get_ch_value(u32 ch, u32& out)
 		if (ch_event_mask & SPU_EVENT_LR)
 		{
 			// register waiter if polling reservation status is required
-			vm::wait_op(*this, last_raddr, 128, WRAP_EXPR(get_events(true) || state & cpu_state::stop));
+			vm::wait_op(last_raddr, 128, WRAP_EXPR(get_events(true) || state & cpu_state::stop));
 		}
 		else
 		{
@@ -662,7 +670,7 @@ bool SPUThread::get_ch_value(u32 ch, u32& out)
 			{
 				CHECK_EMU_STATUS;
 
-				cv.wait(lock);
+				get_current_thread_cv().wait(lock);
 			}
 		}
 
@@ -702,7 +710,7 @@ bool SPUThread::set_ch_value(u32 ch, u32 value)
 	{
 		if (offset >= RAW_SPU_BASE_ADDR)
 		{
-			std::unique_lock<std::mutex> lock(mutex, std::defer_lock);
+			std::unique_lock<std::mutex> lock(get_current_thread_mutex(), std::defer_lock);
 
 			while (!ch_out_intr_mbox.try_push(value))
 			{
@@ -719,7 +727,7 @@ bool SPUThread::set_ch_value(u32 ch, u32 value)
 					continue;
 				}
 
-				cv.wait(lock);
+				get_current_thread_cv().wait(lock);
 			}
 
 			int_ctrl[2].set(SPU_INT2_STAT_MAILBOX_INT);
@@ -909,7 +917,7 @@ bool SPUThread::set_ch_value(u32 ch, u32 value)
 
 	case SPU_WrOutMbox:
 	{
-		std::unique_lock<std::mutex> lock(mutex, std::defer_lock);
+		std::unique_lock<std::mutex> lock(get_current_thread_mutex(), std::defer_lock);
 
 		while (!ch_out_mbox.try_push(value))
 		{
@@ -926,7 +934,7 @@ bool SPUThread::set_ch_value(u32 ch, u32 value)
 				continue;
 			}
 
-			cv.wait(lock);
+			get_current_thread_cv().wait(lock);
 		}
 
 		return true;
@@ -1226,7 +1234,7 @@ bool SPUThread::stop_and_signal(u32 code)
 					return false;
 				}
 
-				cv.wait(lv2_lock);
+				get_current_thread_cv().wait(lv2_lock);
 			}
 
 			// event data must be set by push()
@@ -1251,7 +1259,7 @@ bool SPUThread::stop_and_signal(u32 code)
 			if (thread && thread.get() != this)
 			{
 				thread->state -= cpu_state::suspend;
-				thread->safe_notify();
+				thread->lock_notify();
 			}
 		}
 
@@ -1290,7 +1298,7 @@ bool SPUThread::stop_and_signal(u32 code)
 			if (thread && thread.get() != this)
 			{
 				thread->state += cpu_state::stop;
-				thread->safe_notify();
+				thread->lock_notify();
 			}
 		}
 
@@ -1393,3 +1401,29 @@ void SPUThread::fast_call(u32 ls_addr)
 	gpr[1]._u32[3] = old_stack;
 	custom_task = std::move(old_task);
 }
+
+void SPUThread::RegisterHleFunction(u32 addr, std::function<bool(SPUThread&SPU)> function)
+{
+	m_addr_to_hle_function_map[addr] = function;
+	_ref<u32>(addr) = 0x00000003; // STOP 3
+}
+
+void SPUThread::UnregisterHleFunction(u32 addr)
+{
+	m_addr_to_hle_function_map.erase(addr);
+}
+
+void SPUThread::UnregisterHleFunctions(u32 start_addr, u32 end_addr)
+{
+	for (auto iter = m_addr_to_hle_function_map.begin(); iter != m_addr_to_hle_function_map.end();)
+	{
+		if (iter->first >= start_addr && iter->first <= end_addr)
+		{
+			m_addr_to_hle_function_map.erase(iter++);
+		}
+		else
+		{
+			iter++;
+		}
+	}
+}
diff --git a/rpcs3/Emu/Cell/SPUThread.h b/rpcs3/Emu/Cell/SPUThread.h
index deb1a15605..187c3a6aa6 100644
--- a/rpcs3/Emu/Cell/SPUThread.h
+++ b/rpcs3/Emu/Cell/SPUThread.h
@@ -180,7 +180,7 @@ public:
 			data.value |= value;
 		});
 
-		if (old.wait) spu.safe_notify();
+		if (old.wait) spu->lock_notify();
 	}
 
 	// push unconditionally (overwriting previous value), may require notification
@@ -193,7 +193,7 @@ public:
 			data.value = value;
 		});
 
-		if (old.wait) spu.safe_notify();
+		if (old.wait) spu->lock_notify();
 	}
 
 	// returns true on success
@@ -228,7 +228,7 @@ public:
 			// value is not cleared and may be read again
 		});
 
-		if (old.wait) spu.safe_notify();
+		if (old.wait) spu->lock_notify();
 
 		return old.value;
 	}
@@ -299,7 +299,7 @@ public:
 			return false;
 		}))
 		{
-			spu.safe_notify();
+			spu->lock_notify();
 		}
 	}
 
@@ -554,20 +554,14 @@ public:
 	virtual std::string dump() const override;
 	virtual void cpu_init() override;
 	virtual void cpu_task() override;
+	virtual ~SPUThread() override;
 
 protected:
-	SPUThread(const std::string& name)
-		: cpu_thread(cpu_type::spu, name)
-		, index(0)
-		, offset(0)
-	{
-	}
+	SPUThread(const std::string& name);
 
 public:
 	SPUThread(const std::string& name, u32 index);
 
-	virtual ~SPUThread() override;
-
 	std::array<v128, 128> gpr; // General-Purpose Registers
 	SPU_FPSCR fpscr;
 
@@ -651,29 +645,7 @@ public:
 		return *_ptr<T>(lsa);
 	}
 
-	void RegisterHleFunction(u32 addr, std::function<bool(SPUThread & SPU)> function)
-	{
-		m_addr_to_hle_function_map[addr] = function;
-		_ref<u32>(addr) = 0x00000003; // STOP 3
-	}
-
-	void UnregisterHleFunction(u32 addr)
-	{
-		m_addr_to_hle_function_map.erase(addr);
-	}
-
-	void UnregisterHleFunctions(u32 start_addr, u32 end_addr)
-	{
-		for (auto iter = m_addr_to_hle_function_map.begin(); iter != m_addr_to_hle_function_map.end();)
-		{
-			if (iter->first >= start_addr && iter->first <= end_addr)
-			{
-				m_addr_to_hle_function_map.erase(iter++);
-			}
-			else
-			{
-				iter++;
-			}
-		}
-	}
+	void RegisterHleFunction(u32 addr, std::function<bool(SPUThread & SPU)> function);
+	void UnregisterHleFunction(u32 addr);
+	void UnregisterHleFunctions(u32 start_addr, u32 end_addr);
 };
diff --git a/rpcs3/Emu/Cell/lv2/IPC.h b/rpcs3/Emu/Cell/lv2/IPC.h
new file mode 100644
index 0000000000..777409307a
--- /dev/null
+++ b/rpcs3/Emu/Cell/lv2/IPC.h
@@ -0,0 +1,54 @@
+#pragma once
+
+#include <memory>
+#include <unordered_map>
+
+#include "Utilities/SharedMutex.h"
+
+// IPC manager for lv2 objects of type T and 64-bit IPC keys.
+// External declaration of g_ipc is required.
+template<typename T>
+class ipc_manager final
+{
+	std::unordered_map<u64, std::weak_ptr<T>> m_map;
+
+	shared_mutex m_mutex;
+
+	static ipc_manager g_ipc;
+
+public:
+	// Add new object if specified ipc_key is not used
+	template<typename F>
+	static auto add(u64 ipc_key, F&& provider) -> decltype(static_cast<std::shared_ptr<T>>(provider()))
+	{
+		writer_lock lock(g_ipc.m_mutex);
+
+		// Get object location
+		std::weak_ptr<T>& wptr = g_ipc.m_map[ipc_key];
+
+		if (wptr.expired())
+		{
+			// Call a function which must return the object
+			std::shared_ptr<T> result = provider();
+			wptr = result;
+			return result;
+		}
+
+		return{};
+	}
+
+	// Get existing object with specified ipc_key
+	static std::shared_ptr<T> get(u64 ipc_key)
+	{
+		reader_lock lock(g_ipc.m_mutex);
+
+		const auto found = g_ipc.m_map.find(ipc_key);
+
+		if (found != g_ipc.m_map.end())
+		{
+			return found->second.lock();
+		}
+
+		return{};
+	}
+};
diff --git a/rpcs3/Emu/Cell/lv2/sys_cond.cpp b/rpcs3/Emu/Cell/lv2/sys_cond.cpp
index 845a3ec097..82853dac81 100644
--- a/rpcs3/Emu/Cell/lv2/sys_cond.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_cond.cpp
@@ -25,7 +25,7 @@ void lv2_cond_t::notify(lv2_lock_t, cpu_thread* thread)
 		mutex->owner = std::static_pointer_cast<cpu_thread>(thread->shared_from_this());
 
 		ASSERT(!thread->state.test_and_set(cpu_state::signal));
-		thread->cv.notify_one();
+		thread->notify();
 	}
 }
 
@@ -221,11 +221,11 @@ s32 sys_cond_wait(PPUThread& ppu, u32 cond_id, u64 timeout)
 				continue;
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_event.cpp b/rpcs3/Emu/Cell/lv2/sys_event.cpp
index 81a71bdc5d..ac105aabb5 100644
--- a/rpcs3/Emu/Cell/lv2/sys_event.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_event.cpp
@@ -9,17 +9,13 @@
 #include "Emu/Cell/SPUThread.h"
 #include "sys_process.h"
 #include "sys_event.h"
+#include "IPC.h"
 
 LOG_CHANNEL(sys_event);
 
-extern u64 get_system_time();
+template<> DECLARE(ipc_manager<lv2_event_queue_t>::g_ipc) {};
 
-static ipc_manager<lv2_event_queue_t>& get_ipc_manager()
-{
-	// Use magic static
-	static ipc_manager<lv2_event_queue_t> instance;
-	return instance;
-}
+extern u64 get_system_time();
 
 std::shared_ptr<lv2_event_queue_t> lv2_event_queue_t::make(u32 protocol, s32 type, u64 name, u64 ipc_key, s32 size)
 {
@@ -32,7 +28,7 @@ std::shared_ptr<lv2_event_queue_t> lv2_event_queue_t::make(u32 protocol, s32 typ
 	}
 
 	// IPC queue
-	return get_ipc_manager().add(ipc_key, make_expr);
+	return ipc_manager<lv2_event_queue_t>::add(ipc_key, make_expr);
 }
 
 std::shared_ptr<lv2_event_queue_t> lv2_event_queue_t::find(u64 ipc_key)
@@ -43,7 +39,7 @@ std::shared_ptr<lv2_event_queue_t> lv2_event_queue_t::find(u64 ipc_key)
 		return{};
 	}
 
-	return get_ipc_manager().get(ipc_key);
+	return ipc_manager<lv2_event_queue_t>::get(ipc_key);
 }
 
 void lv2_event_queue_t::push(lv2_lock_t, u64 source, u64 data1, u64 data2, u64 data3)
@@ -82,7 +78,7 @@ void lv2_event_queue_t::push(lv2_lock_t, u64 source, u64 data1, u64 data2, u64 d
 	}
 
 	ASSERT(!thread->state.test_and_set(cpu_state::signal));
-	thread->cv.notify_one();
+	thread->notify();
 
 	return m_sq.pop_front();
 }
@@ -175,7 +171,7 @@ s32 sys_event_queue_destroy(u32 equeue_id, s32 mode)
 		}
 
 		thread->state += cpu_state::signal;
-		thread->cv.notify_one();
+		thread->notify();
 	}
 
 	return CELL_OK;
@@ -264,11 +260,11 @@ s32 sys_event_queue_receive(PPUThread& ppu, u32 equeue_id, vm::ptr<sys_event_t>
 				return CELL_ETIMEDOUT;
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_event_flag.cpp b/rpcs3/Emu/Cell/lv2/sys_event_flag.cpp
index 443cc55de3..87814cf8e4 100644
--- a/rpcs3/Emu/Cell/lv2/sys_event_flag.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_event_flag.cpp
@@ -29,7 +29,7 @@ void lv2_event_flag_t::notify_all(lv2_lock_t)
 			ppu.GPR[4] = clear_pattern(bitptn, mode);
 
 			ASSERT(!thread->state.test_and_set(cpu_state::signal));
-			thread->cv.notify_one();
+			thread->notify();
 
 			return true;
 		}
@@ -160,11 +160,11 @@ s32 sys_event_flag_wait(PPUThread& ppu, u32 id, u64 bitptn, u32 mode, vm::ptr<u6
 				return CELL_ETIMEDOUT;
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 	
@@ -292,7 +292,7 @@ s32 sys_event_flag_cancel(u32 id, vm::ptr<u32> num)
 		ppu.GPR[5] = 0;
 
 		ASSERT(!thread->state.test_and_set(cpu_state::signal));
-		thread->cv.notify_one();
+		thread->notify();
 	}
 
 	eflag->sq.clear();
diff --git a/rpcs3/Emu/Cell/lv2/sys_interrupt.cpp b/rpcs3/Emu/Cell/lv2/sys_interrupt.cpp
index 21c873cc3c..ca2fe1f9d9 100644
--- a/rpcs3/Emu/Cell/lv2/sys_interrupt.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_interrupt.cpp
@@ -14,15 +14,14 @@ void lv2_int_serv_t::join(PPUThread& ppu, lv2_lock_t lv2_lock)
 {
 	// Use is_joining to stop interrupt thread and signal
 	thread->is_joining = true;
-
-	thread->cv.notify_one();
+	thread->notify();
 
 	// Start joining
 	while (!(thread->state & cpu_state::exit))
 	{
 		CHECK_EMU_STATUS;
 
-		ppu.cv.wait_for(lv2_lock, std::chrono::milliseconds(1));
+		get_current_thread_cv().wait_for(lv2_lock, std::chrono::milliseconds(1));
 	}
 
 	// Cleanup
@@ -122,14 +121,14 @@ s32 _sys_interrupt_thread_establish(vm::ptr<u32> ih, u32 intrtag, u32 intrthread
 				continue;
 			}
 
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 
 		ppu.state += cpu_state::exit;
 	};
 
 	it->state -= cpu_state::stop;
-	it->safe_notify();
+	it->lock_notify();
 
 	*ih = handler->id;
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_lwcond.cpp b/rpcs3/Emu/Cell/lv2/sys_lwcond.cpp
index cb77b4b479..e56a69f78d 100644
--- a/rpcs3/Emu/Cell/lv2/sys_lwcond.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_lwcond.cpp
@@ -30,7 +30,7 @@ void lv2_lwcond_t::notify(lv2_lock_t, cpu_thread* thread, const std::shared_ptr<
 	}
 
 	ASSERT(!thread->state.test_and_set(cpu_state::signal));
-	thread->cv.notify_one();
+	thread->notify();
 }
 
 s32 _sys_lwcond_create(vm::ptr<u32> lwcond_id, u32 lwmutex_id, vm::ptr<sys_lwcond_t> control, u64 name, u32 arg5)
@@ -202,11 +202,11 @@ s32 _sys_lwcond_queue_wait(PPUThread& ppu, u32 lwcond_id, u32 lwmutex_id, u64 ti
 				}
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_lwmutex.cpp b/rpcs3/Emu/Cell/lv2/sys_lwmutex.cpp
index 76833e1a93..479030acf5 100644
--- a/rpcs3/Emu/Cell/lv2/sys_lwmutex.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_lwmutex.cpp
@@ -23,7 +23,7 @@ void lv2_lwmutex_t::unlock(lv2_lock_t)
 	{
 		auto& thread = sq.front();
 		ASSERT(!thread->state.test_and_set(cpu_state::signal));
-		thread->cv.notify_one();
+		thread->notify();
 
 		sq.pop_front();
 	}
@@ -114,11 +114,11 @@ s32 _sys_lwmutex_lock(PPUThread& ppu, u32 lwmutex_id, u64 timeout)
 				return CELL_ETIMEDOUT;
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_mutex.cpp b/rpcs3/Emu/Cell/lv2/sys_mutex.cpp
index faf3f21b99..8e32dfb99e 100644
--- a/rpcs3/Emu/Cell/lv2/sys_mutex.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_mutex.cpp
@@ -22,7 +22,7 @@ void lv2_mutex_t::unlock(lv2_lock_t)
 		owner = std::static_pointer_cast<cpu_thread>(sq.front()->shared_from_this());
 
 		ASSERT(!owner->state.test_and_set(cpu_state::signal));
-		owner->cv.notify_one();
+		owner->notify();
 	}
 }
 
@@ -149,11 +149,11 @@ s32 sys_mutex_lock(PPUThread& ppu, u32 mutex_id, u64 timeout)
 				return CELL_ETIMEDOUT;
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_ppu_thread.cpp b/rpcs3/Emu/Cell/lv2/sys_ppu_thread.cpp
index f895072977..a77a242c23 100644
--- a/rpcs3/Emu/Cell/lv2/sys_ppu_thread.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_ppu_thread.cpp
@@ -6,7 +6,6 @@
 
 #include "Emu/Cell/ErrorCodes.h"
 #include "Emu/Cell/PPUThread.h"
-#include "sys_mutex.h"
 #include "sys_ppu_thread.h"
 
 LOG_CHANNEL(sys_ppu_thread);
@@ -17,24 +16,25 @@ void _sys_ppu_thread_exit(PPUThread& ppu, u64 errorcode)
 
 	LV2_LOCK;
 
-	// get all sys_mutex objects
-	for (auto& mutex : idm::get_all<lv2_mutex_t>())
-	{
-		// unlock mutex if locked by this thread
-		if (mutex->owner.get() == &ppu)
-		{
-			mutex->unlock(lv2_lock);
-		}
-	}
+	// TODO: Should we really unlock mutexes?
 
+	//// get all sys_mutex objects
+	//for (auto& mutex : idm::get_all<lv2_mutex_t>())
+	//{
+	//	// unlock mutex if locked by this thread
+	//	if (mutex->owner.get() == &ppu)
+	//	{
+	//		mutex->unlock(lv2_lock);
+	//	}
+	//}
+
+	ppu.state += cpu_state::exit;
+
+	// Delete detached thread
 	if (!ppu.is_joinable)
 	{
 		idm::remove<PPUThread>(ppu.id);
 	}
-	else
-	{
-		ppu.state += cpu_state::exit;
-	}
 
 	// Throw if this syscall was not called directly by the SC instruction (hack)
 	if (ppu.GPR[11] != 41 || ppu.custom_task)
@@ -81,7 +81,7 @@ s32 sys_ppu_thread_join(PPUThread& ppu, u32 thread_id, vm::ptr<u64> vptr)
 	{
 		CHECK_EMU_STATUS;
 
-		ppu.cv.wait_for(lv2_lock, std::chrono::milliseconds(1));
+		get_current_thread_cv().wait_for(lv2_lock, std::chrono::milliseconds(1));
 	}
 
 	// get exit status from the register
@@ -236,7 +236,7 @@ u32 ppu_thread_create(u32 entry, u64 arg, s32 prio, u32 stacksize, const std::st
 
 	ppu->GPR[3] = arg;
 	ppu->state -= cpu_state::stop;
-	ppu->safe_notify();
+	ppu->lock_notify();
 
 	return ppu->id;
 }
@@ -294,7 +294,7 @@ s32 sys_ppu_thread_start(u32 thread_id)
 	}
 
 	thread->state -= cpu_state::stop;
-	thread->safe_notify();
+	thread->lock_notify();
 
 	return CELL_OK;
 }
diff --git a/rpcs3/Emu/Cell/lv2/sys_ppu_thread.h b/rpcs3/Emu/Cell/lv2/sys_ppu_thread.h
index 1075e4b543..8a103ae7bf 100644
--- a/rpcs3/Emu/Cell/lv2/sys_ppu_thread.h
+++ b/rpcs3/Emu/Cell/lv2/sys_ppu_thread.h
@@ -1,6 +1,6 @@
 #pragma once
 
-namespace vm { using namespace ps3; }
+#include "sys_sync.h"
 
 class PPUThread;
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_rwlock.cpp b/rpcs3/Emu/Cell/lv2/sys_rwlock.cpp
index 0188cf7ac6..351d9ca7bb 100644
--- a/rpcs3/Emu/Cell/lv2/sys_rwlock.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_rwlock.cpp
@@ -20,7 +20,7 @@ void lv2_rwlock_t::notify_all(lv2_lock_t)
 		writer = std::static_pointer_cast<cpu_thread>(wsq.front()->shared_from_this());
 
 		ASSERT(!writer->state.test_and_set(cpu_state::signal));
-		writer->cv.notify_one();
+		writer->notify();
 
 		return wsq.pop_front();
 	}
@@ -33,7 +33,7 @@ void lv2_rwlock_t::notify_all(lv2_lock_t)
 		for (auto& thread : rsq)
 		{
 			ASSERT(!thread->state.test_and_set(cpu_state::signal));
-			thread->cv.notify_one();
+			thread->notify();
 		}
 
 		return rsq.clear();
@@ -132,11 +132,11 @@ s32 sys_rwlock_rlock(PPUThread& ppu, u32 rw_lock_id, u64 timeout)
 				return CELL_ETIMEDOUT;
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 
@@ -255,11 +255,11 @@ s32 sys_rwlock_wlock(PPUThread& ppu, u32 rw_lock_id, u64 timeout)
 				return CELL_ETIMEDOUT;
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_semaphore.cpp b/rpcs3/Emu/Cell/lv2/sys_semaphore.cpp
index 04c4a59c9c..ec49e8c24d 100644
--- a/rpcs3/Emu/Cell/lv2/sys_semaphore.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_semaphore.cpp
@@ -107,11 +107,11 @@ s32 sys_semaphore_wait(PPUThread& ppu, u32 sem_id, u64 timeout)
 				return CELL_ETIMEDOUT;
 			}
 
-			ppu.cv.wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
+			get_current_thread_cv().wait_for(lv2_lock, std::chrono::microseconds(timeout - passed));
 		}
 		else
 		{
-			ppu.cv.wait(lv2_lock);
+			get_current_thread_cv().wait(lv2_lock);
 		}
 	}
 
@@ -175,7 +175,7 @@ s32 sys_semaphore_post(u32 sem_id, s32 count)
 
 		auto& thread = sem->sq.front();
 		ASSERT(!thread->state.test_and_set(cpu_state::signal));
-		thread->cv.notify_one();
+		thread->notify();
 
 		sem->sq.pop_front();
 	}
diff --git a/rpcs3/Emu/Cell/lv2/sys_spu.cpp b/rpcs3/Emu/Cell/lv2/sys_spu.cpp
index b4a29f0a1b..cdccda8e14 100644
--- a/rpcs3/Emu/Cell/lv2/sys_spu.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_spu.cpp
@@ -323,7 +323,7 @@ s32 sys_spu_thread_group_start(u32 id)
 		if (thread)
 		{
 			thread->state -= cpu_state::stop;
-			thread->safe_notify();
+			thread->lock_notify();
 		}
 	}
 
@@ -421,7 +421,7 @@ s32 sys_spu_thread_group_resume(u32 id)
 		if (thread)
 		{
 			thread->state -= cpu_state::suspend;
-			thread->safe_notify();
+			thread->lock_notify();
 		}
 	}
 
@@ -504,7 +504,7 @@ s32 sys_spu_thread_group_terminate(u32 id, s32 value)
 		if (thread)
 		{
 			thread->state += cpu_state::stop;
-			thread->safe_notify();
+			thread->lock_notify();
 		}
 	}
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_sync.h b/rpcs3/Emu/Cell/lv2/sys_sync.h
index 37b147e472..e96c05f9f8 100644
--- a/rpcs3/Emu/Cell/lv2/sys_sync.h
+++ b/rpcs3/Emu/Cell/lv2/sys_sync.h
@@ -1,6 +1,5 @@
 #pragma once
 
-#include "Utilities/SharedMutex.h"
 #include "Utilities/SleepQueue.h"
 
 namespace vm { using namespace ps3; }
@@ -44,49 +43,8 @@ enum
 	SYS_SYNC_NOT_ADAPTIVE = 0x2000,
 };
 
-// IPC manager collection for lv2 objects of type T
-template<typename T>
-class ipc_manager final
-{
-	mutable shared_mutex m_mutex;
-	std::unordered_map<u64, std::weak_ptr<T>> m_map;
-
-public:
-	// Add new object if specified ipc_key is not used
-	template<typename F>
-	auto add(u64 ipc_key, F&& provider) -> decltype(static_cast<std::shared_ptr<T>>(provider()))
-	{
-		std::lock_guard<shared_mutex> lock(m_mutex);
-		
-		// Get object location
-		std::weak_ptr<T>& wptr = m_map[ipc_key];
-
-		if (wptr.expired())
-		{
-			// Call a function which must return the object
-			std::shared_ptr<T>&& result = provider();
-			wptr = result;
-			return result;
-		}
-
-		return{};
-	}
-
-	// Get existing object with specified ipc_key
-	std::shared_ptr<T> get(u64 ipc_key) const
-	{
-		reader_lock lock(m_mutex);
-
-		const auto found = m_map.find(ipc_key);
-
-		if (found != m_map.end())
-		{
-			return found->second.lock();
-		}
-
-		return{};
-	}
-};
+extern std::mutex& get_current_thread_mutex();
+extern std::condition_variable& get_current_thread_cv();
 
 // Simple class for global mutex to pass unique_lock and check it
 struct lv2_lock_t
diff --git a/rpcs3/Emu/Cell/lv2/sys_timer.cpp b/rpcs3/Emu/Cell/lv2/sys_timer.cpp
index 3f8c8f78fc..5944157ad2 100644
--- a/rpcs3/Emu/Cell/lv2/sys_timer.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_timer.cpp
@@ -15,7 +15,7 @@ extern u64 get_system_time();
 
 void lv2_timer_t::on_task()
 {
-	std::unique_lock<std::mutex> lock(mutex);
+	std::unique_lock<std::mutex> lock(get_current_thread_mutex());
 
 	while (state <= SYS_TIMER_STATE_RUN)
 	{
@@ -23,11 +23,9 @@ void lv2_timer_t::on_task()
 
 		if (state == SYS_TIMER_STATE_RUN)
 		{
-			if (lock) lock.unlock();
-
 			LV2_LOCK;
 
-			if (get_system_time() >= expire)
+			while (get_system_time() >= expire)
 			{
 				const auto queue = port.lock();
 
@@ -45,20 +43,31 @@ void lv2_timer_t::on_task()
 				else
 				{
 					state = SYS_TIMER_STATE_STOP; // stop if oneshot or the event port was disconnected (TODO: is it correct?)
+
+					break;
 				}
 			}
-		}
 
-		if (!lock)
-		{
-			lock.lock();
 			continue;
 		}
 
-		cv.wait_for(lock, std::chrono::milliseconds(1));
+		get_current_thread_cv().wait_for(lock, std::chrono::milliseconds(1));
 	}
 }
 
+std::string lv2_timer_t::get_name() const
+{
+	return fmt::format("Timer Thread[0x%x]", id);
+}
+
+void lv2_timer_t::on_stop()
+{
+	// Signal thread using invalid state and join
+	state = -1;
+	lock_notify();
+	named_thread::on_stop();
+}
+
 s32 sys_timer_create(vm::ptr<u32> timer_id)
 {
 	sys_timer.warning("sys_timer_create(timer_id=*0x%x)", timer_id);
@@ -157,15 +166,11 @@ s32 _sys_timer_start(u32 timer_id, u64 base_time, u64 period)
 	}
 
 	// sys_timer_start_periodic() will use current time (TODO: is it correct?)
-
-	// lock for reliable notification
-	std::lock_guard<std::mutex> lock(timer->mutex);
-
 	timer->expire = base_time ? base_time : start_time + period;
 	timer->period = period;
 	timer->state  = SYS_TIMER_STATE_RUN;
 
-	timer->cv.notify_one();
+	timer->lock_notify();
 
 	return CELL_OK;
 }
diff --git a/rpcs3/Emu/Cell/lv2/sys_timer.h b/rpcs3/Emu/Cell/lv2/sys_timer.h
index a0bcbba701..185d5f56fe 100644
--- a/rpcs3/Emu/Cell/lv2/sys_timer.h
+++ b/rpcs3/Emu/Cell/lv2/sys_timer.h
@@ -24,19 +24,9 @@ class lv2_timer_t final : public named_thread
 	void on_task() override;
 
 public:
-	std::string get_name() const override
-	{
-		return fmt::format("Timer Thread[0x%x]", id);
-	}
+	std::string get_name() const override;
 
-	void on_stop() override
-	{
-		// Signal thread using invalid state and join
-		std::lock_guard<std::mutex>{ mutex }, state = -1;
-
-		cv.notify_one();
-		join();
-	}
+	void on_stop() override;
 
 	const u32 id{}; // Timer id
 
diff --git a/rpcs3/Emu/IdManager.h b/rpcs3/Emu/IdManager.h
index 25604293ee..01b8843bac 100644
--- a/rpcs3/Emu/IdManager.h
+++ b/rpcs3/Emu/IdManager.h
@@ -225,7 +225,7 @@ class idm
 	template<typename T, typename F, typename = std::result_of_t<F()>>
 	static map_type::pointer create_id(F&& provider)
 	{
-		std::lock_guard<shared_mutex> lock(g_mutex);
+		writer_lock lock(g_mutex);
 
 		if (auto place = allocate_id(get_tag<T>(), id_manager::id_traits<T>::min, id_manager::id_traits<T>::max))
 		{
@@ -255,7 +255,7 @@ class idm
 	// Remove ID and return object
 	static std::shared_ptr<void> delete_id(u32 type, u32 tag, u32 id)
 	{
-		std::lock_guard<shared_mutex> lock(g_mutex);
+		writer_lock lock(g_mutex);
 
 		auto&& ptr = deallocate_id(tag, id);
 
@@ -471,7 +471,7 @@ class fxm
 
 	static std::shared_ptr<void> remove(u32 type)
 	{
-		std::lock_guard<shared_mutex> lock(g_mutex);
+		writer_lock lock(g_mutex);
 
 		return std::move(g_map[type]);
 	}
@@ -504,7 +504,7 @@ public:
 	{
 		std::shared_ptr<T> ptr;
 		{
-			std::lock_guard<shared_mutex> lock(g_mutex);
+			writer_lock lock(g_mutex);
 
 			if (!g_map[get_type<T>()])
 			{
@@ -529,7 +529,7 @@ public:
 		std::shared_ptr<T> ptr;
 		std::shared_ptr<void> old;
 		{
-			std::lock_guard<shared_mutex> lock(g_mutex);
+			writer_lock lock(g_mutex);
 
 			old = std::move(g_map[get_type<T>()]);
 			ptr = std::make_shared<Make>(std::forward<Args>(args)...);
@@ -552,7 +552,7 @@ public:
 	{
 		std::shared_ptr<T> ptr;
 		{
-			std::lock_guard<shared_mutex> lock(g_mutex);
+			writer_lock lock(g_mutex);
 
 			if (!g_map[get_type<T>()])
 			{
@@ -577,7 +577,7 @@ public:
 		std::shared_ptr<T> ptr;
 		std::shared_ptr<void> old;
 		{
-			std::lock_guard<shared_mutex> lock(g_mutex);
+			writer_lock lock(g_mutex);
 
 			old = std::move(g_map[get_type<T>()]);
 			ptr = provider();
@@ -600,7 +600,7 @@ public:
 	{
 		std::shared_ptr<T> ptr;
 		{
-			std::lock_guard<shared_mutex> lock(g_mutex);
+			writer_lock lock(g_mutex);
 
 			if (auto& value = g_map[get_type<T>()])
 			{
diff --git a/rpcs3/Emu/Memory/vm.cpp b/rpcs3/Emu/Memory/vm.cpp
index 3986166b12..c45a731c10 100644
--- a/rpcs3/Emu/Memory/vm.cpp
+++ b/rpcs3/Emu/Memory/vm.cpp
@@ -22,6 +22,8 @@
 #endif
 #endif
 
+#include "wait_engine.h"
+
 namespace vm
 {
 	thread_local u64 g_tls_fault_count{};
@@ -159,160 +161,7 @@ namespace vm
 
 	reservation_mutex_t g_reservation_mutex;
 
-	std::array<waiter_t, 1024> g_waiter_list;
 
-	std::size_t g_waiter_max = 0; // min unused position
-	std::size_t g_waiter_nil = 0; // min search position
-
-	std::mutex g_waiter_list_mutex;
-
-	waiter_t* _add_waiter(named_thread& thread, u32 addr, u32 size)
-	{
-		std::lock_guard<std::mutex> lock(g_waiter_list_mutex);
-
-		const u64 align = 0x80000000ull >> cntlz32(size);
-
-		if (!size || !addr || size > 4096 || size != align || addr & (align - 1))
-		{
-			throw EXCEPTION("Invalid arguments (addr=0x%x, size=0x%x)", addr, size);
-		}
-
-		thread.mutex.lock();
-
-		// look for empty position
-		for (; g_waiter_nil < g_waiter_max; g_waiter_nil++)
-		{
-			waiter_t& waiter = g_waiter_list[g_waiter_nil];
-
-			if (!waiter.thread)
-			{
-				// store next position for further addition
-				g_waiter_nil++;
-
-				return waiter.reset(addr, size, thread);
-			}
-		}
-
-		if (g_waiter_max >= g_waiter_list.size())
-		{
-			throw EXCEPTION("Waiter list limit broken (%lld)", g_waiter_max);
-		}
-
-		waiter_t& waiter = g_waiter_list[g_waiter_max++];
-
-		g_waiter_nil = g_waiter_max;
-		
-		return waiter.reset(addr, size, thread);
-	}
-
-	void _remove_waiter(waiter_t* waiter)
-	{
-		std::lock_guard<std::mutex> lock(g_waiter_list_mutex);
-
-		// mark as deleted
-		waiter->thread = nullptr;
-
-		// amortize adding new element
-		g_waiter_nil = std::min<std::size_t>(g_waiter_nil, waiter - g_waiter_list.data());
-
-		// amortize polling
-		while (g_waiter_max && !g_waiter_list[g_waiter_max - 1].thread)
-		{
-			g_waiter_max--;
-		}
-	}
-
-	bool waiter_t::try_notify()
-	{
-		std::lock_guard<std::mutex> lock(thread->mutex);
-
-		try
-		{
-			// test predicate
-			if (!pred || !pred())
-			{
-				return false;
-			}
-
-			// clear predicate
-			pred = nullptr;
-		}
-		catch (...)
-		{
-			// capture any exception possibly thrown by predicate
-			pred = [exception = std::current_exception()]() -> bool
-			{
-				// new predicate will throw the captured exception from the original thread
-				std::rethrow_exception(exception);
-			};
-		}
-
-		// set addr and mask to invalid values to prevent further polling
-		addr = 0;
-		mask = ~0;
-
-		// signal thread
-		thread->cv.notify_one();
-
-		return true;
-	}
-
-	waiter_lock_t::waiter_lock_t(named_thread& thread, u32 addr, u32 size)
-		: m_waiter(_add_waiter(thread, addr, size))
-		, m_lock(thread.mutex, std::adopt_lock) // must be locked in _add_waiter
-	{
-	}
-
-	void waiter_lock_t::wait()
-	{
-		// if another thread successfully called pred(), it must be set to null
-		while (m_waiter->pred)
-		{
-			// if pred() called by another thread threw an exception, it'll be rethrown
-			if (m_waiter->pred())
-			{
-				return;
-			}
-
-			CHECK_EMU_STATUS;
-
-			m_waiter->thread->cv.wait(m_lock);
-		}
-	}	
-
-	waiter_lock_t::~waiter_lock_t()
-	{
-		// reset some data to avoid excessive signaling
-		m_waiter->addr = 0;
-		m_waiter->mask = ~0;
-		m_waiter->pred = nullptr;
-
-		// unlock thread's mutex to avoid deadlock with g_waiter_list_mutex
-		m_lock.unlock();
-
-		_remove_waiter(m_waiter);
-	}
-
-	void _notify_at(u32 addr, u32 size)
-	{
-		// skip notification if no waiters available
-		if (_mm_mfence(), !g_waiter_max) return;
-
-		std::lock_guard<std::mutex> lock(g_waiter_list_mutex);
-
-		const u32 mask = ~(size - 1);
-
-		for (std::size_t i = 0; i < g_waiter_max; i++)
-		{
-			waiter_t& waiter = g_waiter_list[i];
-
-			// check address range overlapping using masks generated from size (power of 2)
-			if (waiter.thread && ((waiter.addr ^ addr) & (mask & waiter.mask)) == 0)
-			{
-				waiter.try_notify();
-			}
-		}
-	}
 
 	access_violation::access_violation(u64 addr, const char* cause)
 		: std::runtime_error(fmt::exception("Access violation %s address 0x%llx", cause, addr))
@@ -320,62 +169,6 @@ namespace vm
 		g_tls_fault_count &= ~(1ull << 63);
 	}
 
-	void notify_at(u32 addr, u32 size)
-	{
-		const u64 align = 0x80000000ull >> cntlz32(size);
-
-		if (!size || !addr || size > 4096 || size != align || addr & (align - 1))
-		{
-			throw EXCEPTION("Invalid arguments (addr=0x%x, size=0x%x)", addr, size);
-		}
-
-		_notify_at(addr, size);
-	}
-
-	bool notify_all()
-	{
-		std::unique_lock<std::mutex> lock(g_waiter_list_mutex);
-
-		std::size_t waiters = 0;
-		std::size_t signaled = 0;
-
-		for (std::size_t i = 0; i < g_waiter_max; i++)
-		{
-			waiter_t& waiter = g_waiter_list[i];
-
-			if (waiter.thread && waiter.addr)
-			{
-				waiters++;
-
-				if (waiter.try_notify())
-				{
-					signaled++;
-				}
-			}
-		}
-
-		// return true if waiter list is empty or all available waiters were signaled
-		return waiters == signaled;
-	}
-
-	void start()
-	{
-		// start notification thread
-		thread_ctrl::spawn("vm::start thread", []()
-		{
-			while (!Emu.IsStopped())
-			{
-				// poll waiters periodically (TODO)
-				while (!notify_all() && !Emu.IsPaused())
-				{
-					std::this_thread::yield();
-				}
-
-				std::this_thread::sleep_for(std::chrono::milliseconds(1));
-			}
-		});
-	}
-
 	void _reservation_set(u32 addr, bool no_access = false)
 	{
 #ifdef _WIN32
@@ -422,7 +215,7 @@ namespace vm
 
 		if ((g_tls_did_break_reservation = _reservation_break(addr)))
 		{
-			lock.unlock(), _notify_at(raddr, rsize);
+			lock.unlock(), vm::notify_at(raddr, rsize);
 		}
 	}
 
@@ -489,7 +282,7 @@ namespace vm
 		_reservation_break(addr);
 
 		// notify waiter
-		lock.unlock(), _notify_at(addr, size);
+		lock.unlock(), vm::notify_at(addr, size);
 
 		// atomic update succeeded
 		return true;
@@ -517,7 +310,7 @@ namespace vm
 				// break the reservation if overlap
 				if ((g_tls_did_break_reservation = _reservation_break(addr)))
 				{
-					lock.unlock(), _notify_at(raddr, rsize);
+					lock.unlock(), vm::notify_at(raddr, rsize);
 				}
 			}
 			
@@ -536,11 +329,13 @@ namespace vm
 
 	void reservation_free()
 	{
-		if (reservation_test())
+		auto thread = thread_ctrl::get_current();
+
+		if (reservation_test(thread))
 		{
 			std::lock_guard<reservation_mutex_t> lock(g_reservation_mutex);
 
-			if (g_reservation_owner && g_reservation_owner == thread_ctrl::get_current())
+			if (g_reservation_owner && g_reservation_owner == thread)
 			{
 				g_tls_did_break_reservation = _reservation_break(g_reservation_addr);
 			}
@@ -589,7 +384,7 @@ namespace vm
 		_reservation_break(addr);
 
 		// notify waiter
-		lock.unlock(), _notify_at(addr, size);
+		lock.unlock(), vm::notify_at(addr, size);
 	}
 
 	void _page_map(u32 addr, u32 size, u8 flags)
@@ -1014,6 +809,8 @@ namespace vm
 		return nullptr;
 	}
 
+	extern void start();
+
 	namespace ps3
 	{
 		void init()
diff --git a/rpcs3/Emu/Memory/vm.h b/rpcs3/Emu/Memory/vm.h
index 7b61048b54..dda271ebcc 100644
--- a/rpcs3/Emu/Memory/vm.h
+++ b/rpcs3/Emu/Memory/vm.h
@@ -1,8 +1,9 @@
 #pragma once
 
-#include "Utilities/Thread.h"
 #include <map>
 
+class thread_ctrl;
+
 namespace vm
 {
 	extern u8* const g_base_addr;
@@ -38,71 +39,6 @@ namespace vm
 
 	[[noreturn]] void throw_access_violation(u64 addr, const char* cause);
 
-	struct waiter_t
-	{
-		u32 addr = 0;
-		u32 mask = ~0;
-		named_thread* thread = nullptr;
-
-		std::function<bool()> pred;
-
-		waiter_t() = default;
-
-		waiter_t* reset(u32 addr, u32 size, named_thread& thread)
-		{
-			this->addr = addr;
-			this->mask = ~(size - 1);
-			this->thread = &thread;
-
-			// must be null at this point
-			Ensures(!pred);
-			return this;
-		}
-
-		bool try_notify();
-	};
-
-	class waiter_lock_t
-	{
-		waiter_t* m_waiter;
-		std::unique_lock<std::mutex> m_lock;
-
-	public:
-		waiter_lock_t(named_thread& thread, u32 addr, u32 size);
-
-		waiter_t* operator ->() const
-		{
-			return m_waiter;
-		}
-
-		void wait();
-
-		~waiter_lock_t();
-	};
-
-	// Wait until pred() returns true, addr must be aligned to size which must be a power of 2, pred() may be called by any thread
-	template<typename F, typename... Args>
-	auto wait_op(named_thread& thread, u32 addr, u32 size, F pred, Args&&... args) -> decltype(static_cast<void>(pred(args...)))
-	{
-		// return immediately if condition passed (optimistic case)
-		if (pred(args...)) return;
-
-		// initialize waiter and locker
-		waiter_lock_t lock(thread, addr, size);
-
-		// initialize predicate
-		lock->pred = WRAP_EXPR(pred(args...));
-
-		// start waiting
-		lock.wait();
-	}
-
-	// Notify waiters on specific addr, addr must be aligned to size which must be a power of 2
-	void notify_at(u32 addr, u32 size);
-
-	// Try to poll each waiter's condition (false if try_lock failed)
-	bool notify_all();
-
 	// This flag is changed by various reservation functions and may have different meaning.
 	// reservation_break() - true if the reservation was successfully broken.
 	// reservation_acquire() - true if another existing reservation was broken.
@@ -124,7 +60,7 @@ namespace vm
 	bool reservation_query(u32 addr, u32 size, bool is_writing, std::function<bool()> callback);
 
 	// Returns true if the current thread owns reservation
-	bool reservation_test(const thread_ctrl* current = thread_ctrl::get_current());
+	bool reservation_test(const thread_ctrl* current);
 
 	// Break all reservations created by the current thread
 	void reservation_free();
@@ -406,39 +342,6 @@ namespace vm
 	u32 stack_push(u32 size, u32 align_v);
 	void stack_pop_verbose(u32 addr, u32 size) noexcept;
 
-	class stack
-	{
-		u32 m_begin;
-		u32 m_size;
-		int m_page_size;
-		int m_position;
-		u8 m_align;
-
-	public:
-		void init(u32 begin, u32 size, u32 page_size = 180, u8 align = 0x10)
-		{
-			m_begin = begin;
-			m_size = size;
-			m_page_size = page_size;
-			m_position = 0;
-			m_align = align;
-		}
-
-		u32 alloc_new_page()
-		{
-			Expects(m_position + m_page_size < (int)m_size);
-			m_position += (int)m_page_size;
-			return m_begin + m_position;
-		}
-
-		u32 dealloc_new_page()
-		{
-			Expects(m_position - m_page_size > 0);
-			m_position -= (int)m_page_size;
-			return m_begin + m_position;
-		}
-	};
-
 	extern thread_local u64 g_tls_fault_count;
 }
 
diff --git a/rpcs3/Emu/Memory/vm_ptr.h b/rpcs3/Emu/Memory/vm_ptr.h
index f43e9e0df8..e30a8f3b04 100644
--- a/rpcs3/Emu/Memory/vm_ptr.h
+++ b/rpcs3/Emu/Memory/vm_ptr.h
@@ -437,20 +437,6 @@ namespace vm
 
 	// Null pointer convertible to any vm::ptr* type
 	static null_t null;
-
-	// Call wait_op() for specified vm pointer
-	template<typename T, typename AT, typename F, typename... Args>
-	static inline auto wait_op(named_thread& thread, const _ptr_base<T, AT>& ptr, F pred, Args&&... args) -> decltype(static_cast<void>(pred(args...)))
-	{
-		return wait_op(thread, ptr.addr(), SIZE_32(T), std::move(pred), std::forward<Args>(args)...);
-	}
-
-	// Call notify_at() for specified vm pointer
-	template<typename T, typename AT>
-	inline void notify_at(const vm::_ptr_base<T, AT>& ptr)
-	{
-		return notify_at(ptr.addr(), SIZE_32(T));
-	}
 }
 
 template<typename T1, typename AT1, typename T2, typename AT2>
@@ -496,17 +482,12 @@ struct to_se<vm::_ptr_base<T, AT>, Se>
 	using type = vm::_ptr_base<T, typename to_se<AT, Se>::type>;
 };
 
-namespace fmt
+// Format pointer
+template<typename T, typename AT>
+struct unveil<vm::_ptr_base<T, AT>, void>
 {
-	// Format pointer
-	template<typename T, typename AT>
-	struct unveil<vm::_ptr_base<T, AT>, void>
+	static inline auto get(const vm::_ptr_base<T, AT>& arg)
 	{
-		using result_type = typename unveil<AT>::result_type;
-
-		static inline result_type get_value(const vm::_ptr_base<T, AT>& arg)
-		{
-			return unveil<AT>::get_value(arg.addr());
-		}
-	};
-}
+		return unveil<AT>::get(arg.addr());
+	}
+};
diff --git a/rpcs3/Emu/Memory/vm_ref.h b/rpcs3/Emu/Memory/vm_ref.h
index 9b5d1bcc23..216dff79fc 100644
--- a/rpcs3/Emu/Memory/vm_ref.h
+++ b/rpcs3/Emu/Memory/vm_ref.h
@@ -192,12 +192,9 @@ struct to_se<vm::_ref_base<T, AT>, Se>
 	using type = vm::_ref_base<T, typename to_se<AT, Se>::type>;
 };
 
-namespace fmt
+// Forbid formatting
+template<typename T, typename AT>
+struct unveil<vm::_ref_base<T, AT>, void>
 {
-	// Forbid formatting
-	template<typename T, typename AT>
-	struct unveil<vm::_ref_base<T, AT>, void>
-	{
-		static_assert(!sizeof(T), "vm::_ref_base<>: ambiguous format argument");
-	};
-}
+	static_assert(!sizeof(T), "vm::_ref_base<>: ambiguous format argument");
+};
diff --git a/rpcs3/Emu/Memory/wait_engine.cpp b/rpcs3/Emu/Memory/wait_engine.cpp
new file mode 100644
index 0000000000..ffe441e39e
--- /dev/null
+++ b/rpcs3/Emu/Memory/wait_engine.cpp
@@ -0,0 +1,158 @@
+#include "stdafx.h"
+#include "Emu/System.h"
+#include "vm.h"
+#include "wait_engine.h"
+
+#include "Utilities/Thread.h"
+#include "Utilities/SharedMutex.h"
+
+extern std::condition_variable& get_current_thread_cv();
+extern std::mutex& get_current_thread_mutex();
+
+namespace vm
+{
+	static shared_mutex s_mutex;
+
+	static std::unordered_set<waiter*> s_waiters(256);
+
+	bool waiter::try_notify()
+	{
+		{
+			std::lock_guard<mutex_t> lock(*mutex);
+
+			try
+			{
+				// Test predicate
+				if (!pred || !pred())
+				{
+					return false;
+				}
+
+				// Clear predicate
+				pred = nullptr;
+			}
+			catch (...)
+			{
+				// Capture any exception possibly thrown by predicate
+				pred = [exception = std::current_exception()]() -> bool
+				{
+					// New predicate will throw the captured exception from the original thread
+					std::rethrow_exception(exception);
+				};
+			}
+
+			// Set addr and mask to invalid values to prevent further polling
+			addr = 0;
+			mask = ~0;
+		}
+		
+		// Signal thread
+		cond->notify_one();
+		return true;
+	}
+
+	waiter::~waiter()
+	{
+	}
+
+	waiter_lock::waiter_lock(u32 addr, u32 size)
+		: m_lock(get_current_thread_mutex(), std::defer_lock)
+	{
+		Expects(addr && (size & (~size + 1)) == size && (addr & (size - 1)) == 0);
+
+		m_waiter.mutex = m_lock.mutex();
+		m_waiter.cond = &get_current_thread_cv();
+		m_waiter.addr = addr;
+		m_waiter.mask = ~(size - 1);
+
+		{
+			writer_lock lock(s_mutex);
+
+			s_waiters.emplace(&m_waiter);
+		}
+		
+		m_lock.lock();
+	}
+
+	void waiter_lock::wait()
+	{
+		// If another thread successfully called pred(), it must be set to null
+		while (m_waiter.pred)
+		{
+			// If pred() called by another thread threw an exception, it'll be rethrown
+			if (m_waiter.pred())
+			{
+				return;
+			}
+
+			CHECK_EMU_STATUS;
+
+			m_waiter.cond->wait(m_lock);
+		}
+	}
+
+	waiter_lock::~waiter_lock()
+	{
+		if (m_lock) m_lock.unlock();
+
+		writer_lock lock(s_mutex);
+
+		s_waiters.erase(&m_waiter);
+	}
+
+	void notify_at(u32 addr, u32 size)
+	{
+		reader_lock lock(s_mutex);
+
+		for (const auto _w : s_waiters)
+		{
+			// Check address range overlapping using masks generated from size (power of 2)
+			if (((_w->addr ^ addr) & (_w->mask & ~(size - 1))) == 0)
+			{
+				_w->try_notify();
+			}
+		}
+	}
+
+	static bool notify_all()
+	{
+		reader_lock lock(s_mutex);
+
+		std::size_t waiters = 0;
+		std::size_t signaled = 0;
+
+		for (const auto _w : s_waiters)
+		{
+			if (_w->addr)
+			{
+				waiters++;
+
+				if (_w->try_notify())
+				{
+					signaled++;
+				}
+			}
+		}
+
+		// return true if waiter list is empty or all available waiters were signaled
+		return waiters == signaled;
+	}
+
+	void start()
+	{
+		// start notification thread
+		thread_ctrl::spawn("vm::start thread", []()
+		{
+			while (!Emu.IsStopped())
+			{
+				// poll waiters periodically (TODO)
+				while (!notify_all() && !Emu.IsPaused())
+				{
+					std::this_thread::yield();
+				}
+
+				std::this_thread::sleep_for(std::chrono::milliseconds(1));
+			}
+		});
+	}
+}
diff --git a/rpcs3/Emu/Memory/wait_engine.h b/rpcs3/Emu/Memory/wait_engine.h
new file mode 100644
index 0000000000..433884aee7
--- /dev/null
+++ b/rpcs3/Emu/Memory/wait_engine.h
@@ -0,0 +1,63 @@
+#pragma once
+
+#include <mutex>
+#include <condition_variable>
+#include <functional>
+
+class named_thread;
+
+namespace vm
+{
+	using mutex_t = std::mutex;
+	using cond_t = std::condition_variable;
+
+	struct waiter
+	{
+		u32 addr;
+		u32 mask;
+		mutex_t* mutex;
+		cond_t* cond;
+
+		std::function<bool()> pred;
+
+		~waiter();
+
+		bool try_notify();
+	};
+
+	class waiter_lock
+	{
+		waiter m_waiter;
+		std::unique_lock<mutex_t> m_lock;
+
+	public:
+		waiter_lock(u32 addr, u32 size);
+
+		waiter* operator ->()
+		{
+			return &m_waiter;
+		}
+
+		void wait();
+
+		~waiter_lock();
+	};
+
+	// Wait until pred() returns true, addr must be aligned to size which must be a power of 2, pred() may be called by any thread
+	template<typename F, typename... Args>
+	auto wait_op(u32 addr, u32 size, F&& pred, Args&&... args) -> decltype(static_cast<void>(pred(args...)))
+	{
+		// Return immediately if condition passed (optimistic case)
+		if (pred(args...)) return;
+
+		waiter_lock lock(addr, size);
+
+		// Initialize predicate
+		lock->pred = WRAP_EXPR(pred(args...));
+
+		lock.wait();
+	}
+
+	// Notify waiters on specific addr, addr must be aligned to size which must be a power of 2
+	void notify_at(u32 addr, u32 size);
+}
diff --git a/rpcs3/Emu/RSX/GCM.h b/rpcs3/Emu/RSX/GCM.h
index 29362c2df2..f42f07fa04 100644
--- a/rpcs3/Emu/RSX/GCM.h
+++ b/rpcs3/Emu/RSX/GCM.h
@@ -1041,33 +1041,25 @@ enum Method
 
 namespace rsx
 {
-	template<typename ...T>
-	static auto make_command(u32 start_register, T... values) -> std::array<u32, sizeof...(values) + 1>
+	template<typename AT>
+	static inline u32 make_command(vm::_ptr_base<be_t<u32>, AT>& dst, u32 start_register, std::initializer_list<any32> values)
 	{
-		return{ (start_register << 2) | u32(sizeof...(values) << 18), u32(values)... };
-	}
+		*dst++ = start_register << 2 | static_cast<u32>(values.size()) << 18;
 
-	static u32 make_jump(u32 offset)
-	{
-		return CELL_GCM_METHOD_FLAG_JUMP | offset;
-	}
-
-	template<typename AT, typename ...T>
-	static size_t make_command(vm::ps3::ptr<u32, AT> &dst, u32 start_register, T... values)
-	{
-		for (u32 command : { (start_register << 2) | u32(sizeof...(values) << 18), static_cast<u32>(values)... })
+		for (const any32& cmd : values)
 		{
-			*dst++ = command;
+			*dst++ = cmd.as<u32>();
 		}
 
-		return sizeof...(values) + 1;
+		return SIZE_32(u32) * (static_cast<u32>(values.size()) + 1);
 	}
 
 	template<typename AT>
-	static size_t make_jump(vm::ps3::ptr<u32, AT> &dst, u32 offset)
+	static inline u32 make_jump(vm::_ptr_base<be_t<u32>, AT>& dst, u32 offset)
 	{
-		*dst++ = make_jump(offset);
-		return 1;
+		*dst++ = CELL_GCM_METHOD_FLAG_JUMP | offset;
+
+		return SIZE_32(u32);
 	}
 
 	std::string get_method_name(const u32 id);
diff --git a/rpcs3/Emu/RSX/GL/GLGSRender.cpp b/rpcs3/Emu/RSX/GL/GLGSRender.cpp
index 1bd3f7c505..e712e1447c 100644
--- a/rpcs3/Emu/RSX/GL/GLGSRender.cpp
+++ b/rpcs3/Emu/RSX/GL/GLGSRender.cpp
@@ -665,7 +665,7 @@ void GLGSRender::flip(int buffer)
 
 		if (buffer_region.tile)
 		{
-			std::unique_ptr<u8> temp(new u8[buffer_height * buffer_pitch]);
+			std::unique_ptr<u8[]> temp(new u8[buffer_height * buffer_pitch]);
 			buffer_region.read(temp.get(), buffer_width, buffer_height, buffer_pitch);
 			__glcheck m_flip_tex_color.copy_from(temp.get(), gl::texture::format::bgra, gl::texture::type::uint_8_8_8_8);
 		}
diff --git a/rpcs3/Emu/RSX/RSXThread.cpp b/rpcs3/Emu/RSX/RSXThread.cpp
index 492f405459..c610b57c7d 100644
--- a/rpcs3/Emu/RSX/RSXThread.cpp
+++ b/rpcs3/Emu/RSX/RSXThread.cpp
@@ -560,43 +560,44 @@ namespace rsx
 		}
 		else
 		{
-			std::lock_guard<std::mutex> lock{ m_mtx_task };
+			Expects(0);
+			//std::lock_guard<std::mutex> lock{ m_mtx_task };
 
-			internal_task_entry &front = m_internal_tasks.front();
+			//internal_task_entry &front = m_internal_tasks.front();
 
-			if (front.callback())
-			{
-				front.promise.set_value();
-				m_internal_tasks.pop_front();
-			}
+			//if (front.callback())
+			//{
+			//	front.promise.set_value();
+			//	m_internal_tasks.pop_front();
+			//}
 		}
 	}
 
-	std::future<void> thread::add_internal_task(std::function<bool()> callback)
-	{
-		std::lock_guard<std::mutex> lock{ m_mtx_task };
-		m_internal_tasks.emplace_back(callback);
+	//std::future<void> thread::add_internal_task(std::function<bool()> callback)
+	//{
+	//	std::lock_guard<std::mutex> lock{ m_mtx_task };
+	//	m_internal_tasks.emplace_back(callback);
 
-		return m_internal_tasks.back().promise.get_future();
-	}
+	//	return m_internal_tasks.back().promise.get_future();
+	//}
 
-	void thread::invoke(std::function<bool()> callback)
-	{
-		if (get_thread_ctrl() == thread_ctrl::get_current())
-		{
-			while (true)
-			{
-				if (callback())
-				{
-					break;
-				}
-			}
-		}
-		else
-		{
-			add_internal_task(callback).wait();
-		}
-	}
+	//void thread::invoke(std::function<bool()> callback)
+	//{
+	//	if (operator->() == thread_ctrl::get_current())
+	//	{
+	//		while (true)
+	//		{
+	//			if (callback())
+	//			{
+	//				break;
+	//			}
+	//		}
+	//	}
+	//	else
+	//	{
+	//		add_internal_task(callback).wait();
+	//	}
+	//}
 
 	namespace
 	{
diff --git a/rpcs3/Emu/RSX/RSXThread.h b/rpcs3/Emu/RSX/RSXThread.h
index d449f4c4a5..1b481edd85 100644
--- a/rpcs3/Emu/RSX/RSXThread.h
+++ b/rpcs3/Emu/RSX/RSXThread.h
@@ -3,7 +3,6 @@
 #include <stack>
 #include <deque>
 #include <set>
-#include <future>
 #include "GCM.h"
 #include "RSXTexture.h"
 #include "RSXVertexProgram.h"
@@ -320,7 +319,7 @@ namespace rsx
 		struct internal_task_entry
 		{
 			std::function<bool()> callback;
-			std::promise<void> promise;
+			//std::promise<void> promise;
 
 			internal_task_entry(std::function<bool()> callback) : callback(callback)
 			{
@@ -331,8 +330,8 @@ namespace rsx
 		void do_internal_task();
 
 	public:
-		std::future<void> add_internal_task(std::function<bool()> callback);
-		void invoke(std::function<bool()> callback);
+		//std::future<void> add_internal_task(std::function<bool()> callback);
+		//void invoke(std::function<bool()> callback);
 
 		/**
 		 * Fill buffer with 4x4 scale offset matrix.
diff --git a/rpcs3/Emu/RSX/rsx_methods.cpp b/rpcs3/Emu/RSX/rsx_methods.cpp
index 8193f66324..dee9342949 100644
--- a/rpcs3/Emu/RSX/rsx_methods.cpp
+++ b/rpcs3/Emu/RSX/rsx_methods.cpp
@@ -451,7 +451,7 @@ namespace rsx
 			//	method_registers[NV3089_IMAGE_IN_SIZE], in_pitch, src_offset, double(1 << 20) / (method_registers[NV3089_DS_DX]), double(1 << 20) / (method_registers[NV3089_DT_DY]),
 			//	method_registers[NV3089_CLIP_SIZE], method_registers[NV3089_IMAGE_OUT_SIZE]);
 
-			std::unique_ptr<u8[]> temp1, temp2;
+			std::unique_ptr<u8[]> temp1, temp2, sw_temp;
 
 			AVPixelFormat in_format = src_color_format == CELL_GCM_TRANSFER_SCALE_FORMAT_R5G6B5 ? AV_PIX_FMT_RGB565BE : AV_PIX_FMT_ARGB;
 			AVPixelFormat out_format = dst_color_format == CELL_GCM_TRANSFER_SURFACE_FORMAT_R5G6B5 ? AV_PIX_FMT_RGB565BE : AV_PIX_FMT_ARGB;
@@ -573,7 +573,7 @@ namespace rsx
 				// Check and pad texture out if we are given non square texture for swizzle to be correct
 				if (sw_width != out_w || sw_height != out_h)
 				{
-					std::unique_ptr<u8[]> sw_temp(new u8[out_bpp * sw_width * sw_height]);
+					sw_temp.reset(new u8[out_bpp * sw_width * sw_height]);
 
 					switch (out_bpp)
 					{
diff --git a/rpcs3/Emu/System.cpp b/rpcs3/Emu/System.cpp
index cad2d7f48c..710bb82596 100644
--- a/rpcs3/Emu/System.cpp
+++ b/rpcs3/Emu/System.cpp
@@ -178,9 +178,10 @@ void Emulator::Load()
 			}
 
 			// Load PARAM.SFO
-			m_psf = psf::load_object(fs::file(elf_dir + "/../PARAM.SFO"));
-			m_title = psf::get_string(m_psf, "TITLE", m_path);
-			m_title_id = psf::get_string(m_psf, "TITLE_ID");
+			const auto _psf = psf::load_object(fs::file(elf_dir + "/../PARAM.SFO"));
+			m_title = psf::get_string(_psf, "TITLE", m_path);
+			m_title_id = psf::get_string(_psf, "TITLE_ID");
+
 			LOG_NOTICE(LOADER, "Title: %s", GetTitle());
 			LOG_NOTICE(LOADER, "Serial: %s", GetTitleID());
 			LOG_NOTICE(LOADER, "");
@@ -287,7 +288,7 @@ void Emulator::Run()
 	for (auto& thread : get_all_cpu_threads())
 	{
 		thread->state -= cpu_state::stop;
-		thread->safe_notify();
+		thread->lock_notify();
 	}
 
 	SendDbgCommand(DID_STARTED_EMU);
@@ -350,7 +351,7 @@ void Emulator::Resume()
 	for (auto& thread : get_all_cpu_threads())
 	{
 		thread->state -= cpu_state::dbg_global_pause;
-		thread->safe_notify();
+		thread->lock_notify();
 	}
 
 	rpcs3::on_resume()();
@@ -376,7 +377,7 @@ void Emulator::Stop()
 		for (auto& thread : get_all_cpu_threads())
 		{
 			thread->state += cpu_state::dbg_global_stop;
-			thread->safe_notify();
+			thread->lock_notify();
 		}
 	}
 
diff --git a/rpcs3/Emu/System.h b/rpcs3/Emu/System.h
index ce13e581b8..5a655b4c36 100644
--- a/rpcs3/Emu/System.h
+++ b/rpcs3/Emu/System.h
@@ -2,7 +2,6 @@
 
 #include "VFS.h"
 #include "DbgCommand.h"
-#include "Loader/PSF.h"
 
 enum class frame_type;
 
@@ -73,7 +72,6 @@ class Emulator final
 	std::string m_elf_path;
 	std::string m_title_id;
 	std::string m_title;
-	psf::registry m_psf;
 
 public:
 	Emulator();
@@ -125,11 +123,6 @@ public:
 		return m_title;
 	}
 
-	const psf::registry& GetPSF() const
-	{
-		return m_psf;
-	}
-
 	u64 GetPauseTime()
 	{
 		return m_pause_amend_time;
diff --git a/rpcs3/Gui/InterpreterDisAsm.cpp b/rpcs3/Gui/InterpreterDisAsm.cpp
index 6c5877fea5..7f218da646 100644
--- a/rpcs3/Gui/InterpreterDisAsm.cpp
+++ b/rpcs3/Gui/InterpreterDisAsm.cpp
@@ -437,7 +437,7 @@ void InterpreterDisAsmFrame::DoRun(wxCommandEvent& WXUNUSED(event))
 	if (cpu && cpu->state.test(cpu_state_pause))
 	{
 		cpu->state -= cpu_state::dbg_pause;
-		cpu->safe_notify();
+		cpu->lock_notify();
 	}
 }
 
@@ -459,7 +459,7 @@ void InterpreterDisAsmFrame::DoStep(wxCommandEvent& WXUNUSED(event))
 			return state.test_and_reset(cpu_state::dbg_pause);
 		}))
 		{
-			cpu->safe_notify();
+			cpu->lock_notify();
 		}
 	}
 }
diff --git a/rpcs3/Gui/MainFrame.cpp b/rpcs3/Gui/MainFrame.cpp
index 5eb6c61f85..d9cdf5709e 100644
--- a/rpcs3/Gui/MainFrame.cpp
+++ b/rpcs3/Gui/MainFrame.cpp
@@ -19,7 +19,7 @@
 #include "Gui/CgDisasm.h"
 #include "Crypto/unpkg.h"
 
-#include <future>
+#include "Utilities/Thread.h"
 
 #ifndef _WIN32
 #include "frame_icon.xpm"
@@ -263,30 +263,39 @@ void MainFrame::InstallPkg(wxCommandEvent& WXUNUSED(event))
 		}
 	}
 
-	wxProgressDialog pdlg("PKG Decrypter / Installer", "Please wait, unpacking...", 1000, this, wxPD_AUTO_HIDE | wxPD_APP_MODAL);
+	wxProgressDialog pdlg("PKG Installer", "Please wait, unpacking...", 1000, this, wxPD_AUTO_HIDE | wxPD_APP_MODAL | wxPD_CAN_ABORT);
 
-	volatile f64 progress = 0.0;
-
-	// Run PKG unpacking asynchronously
-	auto result = std::async(std::launch::async, WRAP_EXPR(pkg_install(pkg_f, local_path + '/', progress)));
-
-	// Wait for the completion
-	while (result.wait_for(15ms) != std::future_status::ready)
+	// Synchronization variable
+	atomic_t<double> progress(0.);
 	{
-		// Update progress window
-		pdlg.Update(progress * pdlg.GetRange());
+		// Run PKG unpacking asynchronously
+		scope_thread worker("PKG Installer", [&]
+		{
+			if (pkg_install(pkg_f, local_path + '/', progress))
+			{
+				progress = 1.;
+			}
 
-		// Update main frame
-		Update();
-		wxGetApp().ProcessPendingEvents();
+			// TODO: Ask user to delete files on cancellation/failure?
+		});
+
+		// Wait for the completion
+		while (std::this_thread::sleep_for(5ms), progress < 1.)
+		{
+			// Update progress window
+			if (!pdlg.Update(static_cast<int>(progress * pdlg.GetRange())))
+			{
+				// Installation cancelled (signal with negative value)
+				progress -= 1.;
+				break;
+			}
+		}
 	}
 
 	pdlg.Close();
 
-	if (result.get())
+	if (progress >= 1.)
 	{
-		LOG_SUCCESS(LOADER, "PKG: Package successfully installed in %s", local_path);
-
 		// Refresh game list
 		m_game_viewer->Refresh();
 	}
diff --git a/rpcs3/emucore.vcxproj b/rpcs3/emucore.vcxproj
index 7101a3d7d2..91dda91d51 100644
--- a/rpcs3/emucore.vcxproj
+++ b/rpcs3/emucore.vcxproj
@@ -75,6 +75,9 @@
     <ClCompile Include="..\Utilities\Config.cpp" />
     <ClCompile Include="..\Utilities\rXml.cpp" />
     <ClCompile Include="..\Utilities\Semaphore.cpp" />
+    <ClCompile Include="..\Utilities\SharedMutex.cpp">
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+    </ClCompile>
     <ClCompile Include="..\Utilities\StrFmt.cpp">
       <PrecompiledHeader>NotUsing</PrecompiledHeader>
     </ClCompile>
@@ -224,6 +227,7 @@
     <ClCompile Include="Emu\Cell\SPUASMJITRecompiler.cpp" />
     <ClCompile Include="Emu\Cell\SPUDisAsm.cpp" />
     <ClCompile Include="Emu\Cell\SPUInterpreter.cpp" />
+    <ClCompile Include="Emu\Memory\wait_engine.cpp" />
     <ClCompile Include="Emu\RSX\CgBinaryFragmentProgram.cpp" />
     <ClCompile Include="Emu\RSX\CgBinaryVertexProgram.cpp" />
     <ClCompile Include="Emu\RSX\Common\BufferUtils.cpp" />
@@ -349,6 +353,7 @@
   <ItemGroup>
     <ClInclude Include="..\3rdparty\stblib\stb_image.h" />
     <ClInclude Include="..\Utilities\Atomic.h" />
+    <ClInclude Include="..\Utilities\AtomicPtr.h" />
     <ClInclude Include="..\Utilities\AutoPause.h" />
     <ClInclude Include="..\Utilities\BEType.h" />
     <ClInclude Include="..\Utilities\BitField.h" />
@@ -449,6 +454,7 @@
     <ClInclude Include="Emu\Audio\Null\NullAudioThread.h" />
     <ClInclude Include="Emu\Cell\Common.h" />
     <ClInclude Include="Emu\Cell\ErrorCodes.h" />
+    <ClInclude Include="Emu\Cell\lv2\IPC.h" />
     <ClInclude Include="Emu\Cell\lv2\sys_cond.h" />
     <ClInclude Include="Emu\Cell\lv2\sys_dbg.h" />
     <ClInclude Include="Emu\Cell\lv2\sys_event.h" />
@@ -555,6 +561,7 @@
     <ClInclude Include="Emu\CPU\CPUDisAsm.h" />
     <ClInclude Include="Emu\CPU\CPUThread.h" />
     <ClInclude Include="Emu\DbgCommand.h" />
+    <ClInclude Include="Emu\Memory\wait_engine.h" />
     <ClInclude Include="Emu\VFS.h" />
     <ClInclude Include="Emu\GameInfo.h" />
     <ClInclude Include="Emu\IdManager.h" />
diff --git a/rpcs3/emucore.vcxproj.filters b/rpcs3/emucore.vcxproj.filters
index c85fb0186e..4766e6cb2a 100644
--- a/rpcs3/emucore.vcxproj.filters
+++ b/rpcs3/emucore.vcxproj.filters
@@ -845,6 +845,12 @@
     <ClCompile Include="Emu\VFS.cpp">
       <Filter>Emu</Filter>
     </ClCompile>
+    <ClCompile Include="..\Utilities\SharedMutex.cpp">
+      <Filter>Utilities</Filter>
+    </ClCompile>
+    <ClCompile Include="Emu\Memory\wait_engine.cpp">
+      <Filter>Emu\Memory</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="Crypto\aes.h">
@@ -1591,5 +1597,14 @@
     <ClInclude Include="..\Utilities\GSL.h">
       <Filter>Utilities</Filter>
     </ClInclude>
+    <ClInclude Include="..\Utilities\AtomicPtr.h">
+      <Filter>Utilities</Filter>
+    </ClInclude>
+    <ClInclude Include="Emu\Memory\wait_engine.h">
+      <Filter>Emu\Memory</Filter>
+    </ClInclude>
+    <ClInclude Include="Emu\Cell\lv2\IPC.h">
+      <Filter>Emu\Cell\lv2</Filter>
+    </ClInclude>
   </ItemGroup>
 </Project>
\ No newline at end of file
diff --git a/rpcs3/rpcs3.cpp b/rpcs3/rpcs3.cpp
index 9342ae6163..3151972d08 100644
--- a/rpcs3/rpcs3.cpp
+++ b/rpcs3/rpcs3.cpp
@@ -36,10 +36,10 @@
 #ifdef _MSC_VER
 #include "Emu/RSX/VK/VKGSRender.h"
 #include "Emu/RSX/D3D12/D3D12GSRender.h"
-#include "Emu/Audio/XAudio2/XAudio2Thread.h"
 #endif
 
 #ifdef _WIN32
+#include "Emu/Audio/XAudio2/XAudio2Thread.h"
 #include <wx/msw/wrapwin.h>
 #endif
 
@@ -105,7 +105,7 @@ cfg::map_entry<std::function<std::shared_ptr<AudioThread>()>> g_cfg_audio_render
 {
 	{ "Null", PURE_EXPR(std::make_shared<NullAudioThread>()) },
 	{ "OpenAL", PURE_EXPR(std::make_shared<OpenALThread>()) },
-#ifdef _MSC_VER
+#ifdef _WIN32
 	{ "XAudio2", PURE_EXPR(std::make_shared<XAudio2Thread>()) },
 #endif
 });
diff --git a/rpcs3_default.props b/rpcs3_default.props
index 18c630eb11..5466d413c5 100644
--- a/rpcs3_default.props
+++ b/rpcs3_default.props
@@ -18,6 +18,7 @@
       <SDLCheck>false</SDLCheck>
       <RuntimeTypeInfo>true</RuntimeTypeInfo>
       <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <AdditionalOptions>/Zc:throwingNew</AdditionalOptions>
     </ClCompile>
     <Link>
       <AdditionalDependencies>ws2_32.lib;avcodec.lib;avformat.lib;avutil.lib;swresample.lib;swscale.lib</AdditionalDependencies>