diff --git a/Utilities/Log.cpp b/Utilities/Log.cpp
index ee76f7d307..322277ee8c 100644
--- a/Utilities/Log.cpp
+++ b/Utilities/Log.cpp
@@ -75,7 +75,7 @@ namespace logs
 #endif
 		uchar* m_fptr{};
 		z_stream m_zs{};
-		semaphore<> m_m;
+		shared_mutex m_m;
 
 		alignas(128) atomic_t<u64> m_buf{0}; // MSB (40 bit): push begin, LSB (24 bis): push size
 		alignas(128) atomic_t<u64> m_out{0}; // Amount of bytes written to file
@@ -183,7 +183,7 @@ namespace logs
 	channel SPU("SPU");
 
 	// Channel registry mutex
-	semaphore<> g_mutex;
+	shared_mutex g_mutex;
 
 	// Must be set to true in main()
 	atomic_t<bool> g_init{false};
diff --git a/Utilities/asm.h b/Utilities/asm.h
index 835b70bb60..f1c81e37bb 100644
--- a/Utilities/asm.h
+++ b/Utilities/asm.h
@@ -44,6 +44,20 @@ namespace utils
 #endif
 	}
 
+	inline u8 popcnt16(u16 arg)
+	{
+		const u32 a1 = arg & 0x5555;
+		const u32 a2 = (arg >> 1) & 0x5555;
+		const u32 a3 = a1 + a2;
+		const u32 b1 = a3 & 0x3333;
+		const u32 b2 = (a3 >> 2) & 0x3333;
+		const u32 b3 = b1 + b2;
+		const u32 c1 = b3 & 0x0f0f;
+		const u32 c2 = (b3 >> 4) & 0x0f0f;
+		const u32 c3 = c1 + c2;
+		return static_cast<u8>(c3 + (c3 >> 8));
+	}
+
 // Rotate helpers
 #if defined(__GNUG__)
 
diff --git a/Utilities/cond.cpp b/Utilities/cond.cpp
index a502210ff0..f15d30cdac 100644
--- a/Utilities/cond.cpp
+++ b/Utilities/cond.cpp
@@ -9,110 +9,47 @@
 
 bool cond_variable::imp_wait(u32 _old, u64 _timeout) noexcept
 {
-	verify(HERE), _old != -1; // Very unlikely: it requires 2^32 distinct threads to wait simultaneously
-	const bool is_inf = _timeout > max_timeout;
+	verify("cond_variable overflow" HERE), (_old & 0xffff) == 0; // Very unlikely: it requires 65535 distinct threads to wait simultaneously
+
+	return balanced_wait_until(m_value, _timeout, [&](u32& value, auto... ret) -> int
+	{
+		if (value >> 16)
+		{
+			// Success
+			value -= 0x10001;
+			return +1;
+		}
+
+		if constexpr (sizeof...(ret))
+		{
+			// Retire
+			value -= 1;
+			return -1;
+		}
+
+		return 0;
+	});
 
 #ifdef _WIN32
-	LARGE_INTEGER timeout;
-	timeout.QuadPart = _timeout * -10;
-
-	if (HRESULT rc = _timeout ? NtWaitForKeyedEvent(nullptr, &m_value, false, is_inf ? nullptr : &timeout) : WAIT_TIMEOUT)
+	if (_old >= 0x10000 && !OptWaitOnAddress && m_value)
 	{
-		verify(HERE), rc == WAIT_TIMEOUT;
-
-		// Retire
-		while (!m_value.try_dec())
-		{
-			timeout.QuadPart = 0;
-
-			if (HRESULT rc2 = NtWaitForKeyedEvent(nullptr, &m_value, false, &timeout))
-			{
-				verify(HERE), rc2 == WAIT_TIMEOUT;
-				SwitchToThread();
-				continue;
-			}
-
-			return true;
-		}
-
-		return false;
-	}
-
-	return true;
-#else
-	timespec timeout;
-	timeout.tv_sec  = _timeout / 1000000;
-	timeout.tv_nsec = (_timeout % 1000000) * 1000;
-
-	for (u32 value = _old + 1;; value = m_value)
-	{
-		const int err = futex(&m_value, FUTEX_WAIT_PRIVATE, value, is_inf ? nullptr : &timeout) == 0
-			? 0
-			: errno;
-
-		// Normal or timeout wakeup
-		if (!err || (!is_inf && err == ETIMEDOUT))
-		{
-			// Cleanup (remove waiter)
-			verify(HERE), m_value--;
-			return !err;
-		}
-
-		// Not a wakeup
-		verify(HERE), err == EAGAIN;
+		// Workaround possibly stolen signal
+		imp_wake(1);
 	}
 #endif
 }
 
 void cond_variable::imp_wake(u32 _count) noexcept
 {
-#ifdef _WIN32
-	// Try to subtract required amount of waiters
-	const u32 count = m_value.atomic_op([=](u32& value)
+	balanced_awaken(m_value, m_value.atomic_op([&](u32& value) -> u32
 	{
-		if (value > _count)
-		{
-			value -= _count;
-			return _count;
-		}
+		// Subtract already signaled number from total amount of waiters
+		const u32 can_sig = (value & 0xffff) - (value >> 16);
+		const u32 num_sig = std::min<u32>(can_sig, _count);
 
-		return std::exchange(value, 0);
-	});
-
-	for (u32 i = count; i > 0; i--)
-	{
-		NtReleaseKeyedEvent(nullptr, &m_value, false, nullptr);
-	}
-#else
-	for (u32 i = _count; i > 0; std::this_thread::yield())
-	{
-		const u32 value = m_value;
-
-		// Constrain remaining amount with imaginary waiter count
-		if (i > value)
-		{
-			i = value;
-		}
-
-		if (!value || i == 0)
-		{
-			// Nothing to do
-			return;
-		}
-
-		if (const int res = futex(&m_value, FUTEX_WAKE_PRIVATE, i > INT_MAX ? INT_MAX : i))
-		{
-			verify(HERE), res >= 0 && (u32)res <= i;
-			i -= res;
-		}
-
-		if (!m_value || i == 0)
-		{
-			// Escape
-			return;
-		}
-	}
-#endif
+		value += num_sig << 16;
+		return num_sig;
+	}));
 }
 
 bool notifier::imp_try_lock(u32 count)
@@ -209,62 +146,29 @@ bool notifier::wait(u64 usec_timeout)
 	return res;
 }
 
-bool cond_one::imp_wait(u32 _old, u64 _timeout) noexcept
+bool cond_one::imp_wait(u64 _timeout) noexcept
 {
-	verify(HERE), _old == c_lock;
+	// State transition: c_sig -> c_lock \ c_lock -> c_wait
+	const u32 _old = m_value.fetch_sub(1);
+	if (LIKELY(_old == c_sig))
+		return true;
 
-	const bool is_inf = _timeout > cond_variable::max_timeout;
-
-#ifdef _WIN32
-	LARGE_INTEGER timeout;
-	timeout.QuadPart = _timeout * -10;
-
-	if (HRESULT rc = _timeout ? NtWaitForKeyedEvent(nullptr, &m_value, false, is_inf ? nullptr : &timeout) : WAIT_TIMEOUT)
+	return balanced_wait_until(m_value, _timeout, [&](u32& value, auto... ret) -> int
 	{
-		verify(HERE), rc == WAIT_TIMEOUT;
-
-		// Retire
-		const bool signaled = m_value.exchange(c_lock) == c_sig;
-		while (signaled)
+		if (value == c_sig)
 		{
-			timeout.QuadPart = 0;
-
-			if (HRESULT rc2 = NtWaitForKeyedEvent(nullptr, &m_value, false, &timeout))
-			{
-				verify(HERE), rc2 == WAIT_TIMEOUT;
-				SwitchToThread();
-				continue;
-			}
-
-			return true;
+			value = c_lock;
+			return +1;
 		}
 
-		return false;
-	}
-#else
-	timespec timeout;
-	timeout.tv_sec  = _timeout / 1000000;
-	timeout.tv_nsec = (_timeout % 1000000) * 1000;
-
-	for (u32 value = _old - 1; value != c_sig; value = m_value)
-	{
-		const int err = futex(&m_value, FUTEX_WAIT_PRIVATE, value, is_inf ? nullptr : &timeout) == 0
-			? 0
-			: errno;
-
-		// Normal or timeout wakeup
-		if (!err || (!is_inf && err == ETIMEDOUT))
+		if constexpr (sizeof...(ret))
 		{
-			return m_value.exchange(c_lock) == c_sig;
+			value = c_lock;
+			return -1;
 		}
 
-		// Not a wakeup
-		verify(HERE), err == EAGAIN;
-	}
-#endif
-
-	verify(HERE), m_value.exchange(c_lock) == c_sig;
-	return true;
+		return 0;
+	});
 }
 
 void cond_one::imp_notify() noexcept
@@ -287,79 +191,54 @@ void cond_one::imp_notify() noexcept
 		return;
 	}
 
-#ifdef _WIN32
-	NtReleaseKeyedEvent(nullptr, &m_value, false, nullptr);
-#else
-	futex(&m_value, FUTEX_WAKE_PRIVATE, 1);
-#endif
+	balanced_awaken(m_value, 1);
 }
 
-bool cond_x16::imp_wait(u32 _new, u32 slot, u64 _timeout) noexcept
+bool cond_x16::imp_wait(u32 slot, u64 _timeout) noexcept
 {
 	const u32 wait_bit = c_wait << slot;
 	const u32 lock_bit = c_lock << slot;
 
-	const bool is_inf = _timeout > cond_variable::max_timeout;
-
-#ifdef _WIN32
-	LARGE_INTEGER timeout;
-	timeout.QuadPart = _timeout * -10;
-
-	if (HRESULT rc = _timeout ? NtWaitForKeyedEvent(nullptr, &m_cvx16, false, is_inf ? nullptr : &timeout) : WAIT_TIMEOUT)
+	// Change state from c_lock to c_wait
+	const u32 old_ = m_cvx16.fetch_op([=](u32& cvx16)
 	{
-		verify(HERE), rc == WAIT_TIMEOUT;
-
-		// Retire
-		const bool signaled = this->retire(slot);
-
-		while (signaled)
+		if (cvx16 & wait_bit)
 		{
-			timeout.QuadPart = 0;
+			// c_sig -> c_lock
+			cvx16 &= ~wait_bit;
+		}
+		else
+		{
+			cvx16 |= wait_bit;
+			cvx16 &= ~lock_bit;
+		}
+	});
 
-			if (HRESULT rc2 = NtWaitForKeyedEvent(nullptr, &m_cvx16, false, &timeout))
-			{
-				verify(HERE), rc2 == WAIT_TIMEOUT;
-				SwitchToThread();
-				continue;
-			}
+	if (old_ & wait_bit)
+	{
+		// Already signaled, return without waiting
+		return true;
+	}
 
-			return true;
+	return balanced_wait_until(m_cvx16, _timeout, [&](u32& cvx16, auto... ret) -> int
+	{
+		if (cvx16 & lock_bit)
+		{
+			// c_sig -> c_lock
+			cvx16 &= ~wait_bit;
+			return +1;
 		}
 
-		return false;
-	}
-
-	if (!this->retire(slot))
-	{
-		// Stolen notification: restore balance
-		NtReleaseKeyedEvent(nullptr, &m_cvx16, false, nullptr);
-	}
-#else
-	timespec timeout;
-	timeout.tv_sec  = _timeout / 1000000;
-	timeout.tv_nsec = (_timeout % 1000000) * 1000;
-
-	for (u32 value = _new; ((value >> slot) & c_sig) != c_sig; value = m_cvx16)
-	{
-		const int err = futex(&m_cvx16, FUTEX_WAIT_PRIVATE, value, is_inf ? nullptr : &timeout) == 0
-			? 0
-			: errno;
-
-		// Normal or timeout wakeup
-		if (!err || (!is_inf && err == ETIMEDOUT))
+		if constexpr (sizeof...(ret))
 		{
-			return this->retire(slot);
+			// Retire
+			cvx16 |= lock_bit;
+			cvx16 &= ~wait_bit;
+			return -1;
 		}
 
-		// Not a wakeup
-		verify(HERE), err == EAGAIN;
-	}
-
-	// Convert c_sig to c_lock
-	m_cvx16 &= ~wait_bit;
-#endif
-
-	return true;
+		return 0;
+	});
 }
 
 void cond_x16::imp_notify() noexcept
@@ -386,13 +265,5 @@ void cond_x16::imp_notify() noexcept
 		return;
 	}
 
-#ifdef _WIN32
-	for (u32 i = 0; i < 16; i++)
-	{
-		if ((wait_mask >> i) & 1)
-			NtReleaseKeyedEvent(nullptr, &m_cvx16, false, nullptr);
-	}
-#else
-	futex(&m_cvx16, FUTEX_WAKE_PRIVATE, INT_MAX);
-#endif
+	balanced_awaken(m_cvx16, utils::popcnt16(wait_mask));
 }
diff --git a/Utilities/cond.h b/Utilities/cond.h
index b17442dc1a..6e43a95922 100644
--- a/Utilities/cond.h
+++ b/Utilities/cond.h
@@ -57,7 +57,7 @@ public:
 	{
 		if (m_value)
 		{
-			imp_wake(-1);
+			imp_wake(65535);
 		}
 	}
 
@@ -140,7 +140,7 @@ class cond_one
 
 	atomic_t<u32> m_value{0};
 
-	bool imp_wait(u32 _old, u64 _timeout) noexcept;
+	bool imp_wait(u64 _timeout) noexcept;
 	void imp_notify() noexcept;
 
 public:
@@ -162,13 +162,7 @@ public:
 	{
 		AUDIT(lock.owns_lock());
 		AUDIT(lock.mutex() == this);
-
-		// State transition: c_sig -> c_lock, c_lock -> c_wait
-		const u32 _old = m_value.fetch_sub(1);
-		if (LIKELY(_old == c_sig))
-			return true;
-
-		return imp_wait(_old, usec_timeout);
+		return imp_wait(usec_timeout);
 	}
 
 	void notify() noexcept
@@ -244,28 +238,9 @@ class cond_x16
 		}
 	};
 
-	bool imp_wait(u32 _new, u32 slot, u64 _timeout) noexcept;
+	bool imp_wait(u32 slot, u64 _timeout) noexcept;
 	void imp_notify() noexcept;
 
-	bool retire(u32 slot) noexcept
-	{
-		const u32 wait_bit = c_wait << slot;
-		const u32 lock_bit = c_lock << slot;
-
-		return m_cvx16.atomic_op([=](u32& cvx16)
-		{
-			if (cvx16 & lock_bit)
-			{
-				cvx16 &= ~wait_bit;
-				return true;
-			}
-
-			cvx16 |= lock_bit;
-			cvx16 &= ~wait_bit;
-			return false;
-		});
-	}
-
 public:
 	constexpr cond_x16() = default;
 
@@ -277,33 +252,7 @@ public:
 	bool wait(lock_x16 const& lock, u64 usec_timeout = -1) noexcept
 	{
 		AUDIT(lock.m_this == this);
-
-		const u32 wait_bit = c_wait << lock.m_slot;
-		const u32 lock_bit = c_lock << lock.m_slot;
-
-		// Change state from c_lock to c_wait
-		const u32 new_ = m_cvx16.atomic_op([=](u32& cvx16)
-		{
-			if (cvx16 & wait_bit)
-			{
-				cvx16 &= ~wait_bit;
-			}
-			else
-			{
-				cvx16 |= wait_bit;
-				cvx16 &= ~lock_bit;
-			}
-
-			return cvx16;
-		});
-
-		if (new_ & lock_bit)
-		{
-			// Already signaled, return without waiting
-			return true;
-		}
-
-		return imp_wait(new_, lock.m_slot, usec_timeout);
+		return imp_wait(lock.m_slot, usec_timeout);
 	}
 
 	void notify_all() noexcept
diff --git a/Utilities/mutex.cpp b/Utilities/mutex.cpp
index 59b50d87bf..3ac8a1688b 100644
--- a/Utilities/mutex.cpp
+++ b/Utilities/mutex.cpp
@@ -44,42 +44,24 @@ void shared_mutex::imp_unlock_shared(u32 old)
 
 void shared_mutex::imp_wait()
 {
-#ifdef _WIN32
-	NtWaitForKeyedEvent(nullptr, &m_value, false, nullptr);
-#else
-	while (true)
+	while (!balanced_wait_until(m_value, -1, [&](u32& value, auto...)
 	{
-		// Load new value, try to acquire c_sig
-		auto [value, ok] = m_value.fetch_op([](u32& value)
+		if (value >= c_sig)
 		{
-			if (value >= c_sig)
-			{
-				value -= c_sig;
-				return true;
-			}
-
-			return false;
-		});
-
-		if (ok)
-		{
-			return;
+			value -= c_sig;
+			return true;
 		}
 
-		futex(&m_value, FUTEX_WAIT_BITSET_PRIVATE, value, nullptr, c_sig);
+		return false;
+	}))
+	{
 	}
-#endif
 }
 
 void shared_mutex::imp_signal()
 {
-#ifdef _WIN32
-	NtReleaseKeyedEvent(nullptr, &m_value, false, nullptr);
-#else
 	m_value += c_sig;
-	futex(&m_value, FUTEX_WAKE_BITSET_PRIVATE, 1, nullptr, c_sig);
-	//futex(&m_value, FUTEX_WAKE_BITSET_PRIVATE, c_one, nullptr, c_sig - 1);
-#endif
+	balanced_awaken(m_value, 1);
 }
 
 void shared_mutex::imp_lock(u32 val)
@@ -166,29 +148,6 @@ void shared_mutex::imp_lock_unlock()
 		busy_wait(1500);
 	}
 
-#ifndef _WIN32
-	while (false)
-	{
-		const u32 val = m_value;
-
-		if (val % c_one == 0 && (val / c_one < _max || val >= c_sig))
-		{
-			return;
-		}
-
-		if (val <= c_one)
-		{
-			// Can't expect a signal
-			break;
-		}
-
-		_max = val / c_one;
-
-		// Monitor all bits except c_sig
-		futex(&m_value, FUTEX_WAIT_BITSET_PRIVATE, val, nullptr, c_sig - 1);
-	}
-#endif
-
 	// Lock and unlock
 	if (!m_value.fetch_add(c_one))
 	{
diff --git a/Utilities/sync.h b/Utilities/sync.h
index d82aa833ee..6f9c1eeb71 100644
--- a/Utilities/sync.h
+++ b/Utilities/sync.h
@@ -72,12 +72,12 @@ inline int futex(volatile void* uaddr, int futex_op, uint val, const timespec* t
 
 			switch (futex_op)
 			{
-			case FUTEX_WAIT:
+			case FUTEX_WAIT_PRIVATE:
 			{
 				mask = -1;
 				[[fallthrough]];
 			}
-			case FUTEX_WAIT_BITSET:
+			case FUTEX_WAIT_BITSET_PRIVATE:
 			{
 				if (*reinterpret_cast<volatile uint*>(uaddr) != val)
 				{
@@ -115,12 +115,12 @@ inline int futex(volatile void* uaddr, int futex_op, uint val, const timespec* t
 				return res;
 			}
 
-			case FUTEX_WAKE:
+			case FUTEX_WAKE_PRIVATE:
 			{
 				mask = -1;
 				[[fallthrough]];
 			}
-			case FUTEX_WAKE_BITSET:
+			case FUTEX_WAKE_BITSET_PRIVATE:
 			{
 				int res = 0;
 
@@ -149,3 +149,144 @@ inline int futex(volatile void* uaddr, int futex_op, uint val, const timespec* t
 	return g_futex(uaddr, futex_op, val, timeout, mask);
 #endif
 }
+
+template <typename T, typename Pred>
+bool balanced_wait_until(atomic_t<T>& var, u64 usec_timeout, Pred&& pred)
+{
+	static_assert(sizeof(T) == 4);
+
+	const bool is_inf = usec_timeout > u64{UINT32_MAX / 1000} * 1000000;
+
+	// Optional second argument indicates that the predicate should try to retire
+	auto test_pred = [&](T& _new, auto... args)
+	{
+		T old = var.load();
+
+		while (true)
+		{
+			_new = old;
+
+			// Zero indicates failure without modifying the value
+			// Negative indicates failure but modifies the value
+			auto ret = std::invoke(std::forward<Pred>(pred), _new, args...);
+
+			if (LIKELY(!ret || var.compare_exchange(old, _new)))
+			{
+				return ret > 0;
+			}
+		}
+	};
+
+	T value;
+
+#ifdef _WIN32
+	if (OptWaitOnAddress)
+	{
+		while (!test_pred(value))
+		{
+			if (OptWaitOnAddress(&var, &value, sizeof(u32), is_inf ? INFINITE : usec_timeout / 1000))
+			{
+				if (!test_pred(value) && !test_pred(value, nullptr))
+				{
+					return false;
+				}
+
+				break;
+			}
+
+			if (GetLastError() == ERROR_TIMEOUT)
+			{
+				// Retire
+				return test_pred(value, nullptr);
+			}
+		}
+
+		return true;
+	}
+
+	LARGE_INTEGER timeout;
+	timeout.QuadPart = usec_timeout * -10;
+
+	if (!usec_timeout || NtWaitForKeyedEvent(nullptr, &var, false, is_inf ? nullptr : &timeout))
+	{
+		// Timed out: retire
+		if (!test_pred(value, nullptr))
+		{
+			return false;
+		}
+
+		// Signaled in the last moment: restore balance
+		NtWaitForKeyedEvent(nullptr, &var, false, nullptr);
+		return true;
+	}
+
+	if (!test_pred(value) && !test_pred(value, nullptr))
+	{
+		// Stolen notification: restore balance
+		NtReleaseKeyedEvent(nullptr, &var, false, nullptr);
+		return false;
+	}
+
+	return true;
+#else
+	struct timespec timeout;
+	timeout.tv_sec  = usec_timeout / 1000000;
+	timeout.tv_nsec = (usec_timeout % 1000000) * 1000;
+
+	while (!test_pred(value))
+	{
+		if (futex(&var, FUTEX_WAIT_PRIVATE, static_cast<u32>(value), is_inf ? nullptr : &timeout) == 0)
+		{
+			if (!test_pred(value) && !test_pred(value, nullptr))
+			{
+				return false;
+			}
+
+			break;
+		}
+
+		switch (errno)
+		{
+		case EAGAIN: break;
+		case ETIMEDOUT: return test_pred(value, nullptr);
+		default: verify("Unknown futex error" HERE), 0;
+		}
+	}
+
+	return true;
+#endif
+}
+
+template <typename T>
+void balanced_awaken(atomic_t<T>& var, u32 weight)
+{
+	static_assert(sizeof(T) == 4);
+
+#ifdef _WIN32
+	if (OptWaitOnAddress)
+	{
+		if (weight > 1)
+		{
+			OptWakeByAddressAll(&var);
+		}
+		else if (weight == 1)
+		{
+			OptWakeByAddressSingle(&var);
+		}
+
+		return;
+	}
+
+	for (u32 i = 0; i < weight; i++)
+	{
+		NtReleaseKeyedEvent(nullptr, &var, false, nullptr);
+	}
+#else
+	if (weight)
+	{
+		futex(&var, FUTEX_WAKE_PRIVATE, std::min<u32>(INT_MAX, weight));
+	}
+
+	return;
+#endif
+}
diff --git a/rpcs3/Emu/Cell/Modules/cellCamera.h b/rpcs3/Emu/Cell/Modules/cellCamera.h
index 9af50a5707..5008dd279d 100644
--- a/rpcs3/Emu/Cell/Modules/cellCamera.h
+++ b/rpcs3/Emu/Cell/Modules/cellCamera.h
@@ -2,7 +2,6 @@
 
 #include "Utilities/Timer.h"
 #include "Emu/Cell/lv2/sys_memory.h"
-#include "Utilities/sema.h"
 #include "Utilities/Thread.h"
 
 #include <map>
diff --git a/rpcs3/Emu/Cell/Modules/cellMouse.cpp b/rpcs3/Emu/Cell/Modules/cellMouse.cpp
index 1ece2e3446..141118bfe5 100644
--- a/rpcs3/Emu/Cell/Modules/cellMouse.cpp
+++ b/rpcs3/Emu/Cell/Modules/cellMouse.cpp
@@ -5,8 +5,6 @@
 
 #include "Emu/Io/MouseHandler.h"
 
-#include "Utilities/sema.h"
-
 #include "cellMouse.h"
 
 extern logs::channel sys_io;
diff --git a/rpcs3/Emu/Cell/PPUThread.cpp b/rpcs3/Emu/Cell/PPUThread.cpp
index d16c11abc1..dbe8178e39 100644
--- a/rpcs3/Emu/Cell/PPUThread.cpp
+++ b/rpcs3/Emu/Cell/PPUThread.cpp
@@ -1383,7 +1383,7 @@ extern void ppu_initialize(const ppu_module& info)
 	std::shared_ptr<jit_compiler> jit;
 
 	// Compiler mutex (global)
-	static semaphore<> jmutex;
+	static shared_mutex jmutex;
 
 	// Initialize global semaphore with the max number of threads
 	u32 max_threads = static_cast<u32>(g_cfg.core.llvm_threads);
diff --git a/rpcs3/Emu/Cell/lv2/sys_event.h b/rpcs3/Emu/Cell/lv2/sys_event.h
index 6bbe7b2fb9..2384c41133 100644
--- a/rpcs3/Emu/Cell/lv2/sys_event.h
+++ b/rpcs3/Emu/Cell/lv2/sys_event.h
@@ -83,7 +83,7 @@ struct lv2_event_queue final : public lv2_obj
 	const u64 key;
 	const s32 size;
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 	std::deque<lv2_event> events;
 	std::deque<cpu_thread*> sq;
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_event_flag.h b/rpcs3/Emu/Cell/lv2/sys_event_flag.h
index a7541c0319..ebcc3a839a 100644
--- a/rpcs3/Emu/Cell/lv2/sys_event_flag.h
+++ b/rpcs3/Emu/Cell/lv2/sys_event_flag.h
@@ -40,7 +40,7 @@ struct lv2_event_flag final : lv2_obj
 	const s32 type;
 	const u64 name;
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 	atomic_t<u32> waiters{0};
 	atomic_t<u64> pattern;
 	std::deque<cpu_thread*> sq;
diff --git a/rpcs3/Emu/Cell/lv2/sys_lwcond.h b/rpcs3/Emu/Cell/lv2/sys_lwcond.h
index a8fe686f27..db9c16fd10 100644
--- a/rpcs3/Emu/Cell/lv2/sys_lwcond.h
+++ b/rpcs3/Emu/Cell/lv2/sys_lwcond.h
@@ -27,7 +27,7 @@ struct lv2_lwcond final : lv2_obj
 	const u32 lwid;
 	vm::ptr<sys_lwcond_t> control;
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 	atomic_t<u32> waiters{0};
 	std::deque<cpu_thread*> sq;
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_lwmutex.h b/rpcs3/Emu/Cell/lv2/sys_lwmutex.h
index e3d66f50ed..d0c41f6416 100644
--- a/rpcs3/Emu/Cell/lv2/sys_lwmutex.h
+++ b/rpcs3/Emu/Cell/lv2/sys_lwmutex.h
@@ -57,7 +57,7 @@ struct lv2_lwmutex final : lv2_obj
 	const vm::ptr<sys_lwmutex_t> control;
 	const u64 name;
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 	atomic_t<u32> signaled{0};
 	std::deque<cpu_thread*> sq;
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_mmapper.h b/rpcs3/Emu/Cell/lv2/sys_mmapper.h
index 2ce67b8a3b..65021cddb0 100644
--- a/rpcs3/Emu/Cell/lv2/sys_mmapper.h
+++ b/rpcs3/Emu/Cell/lv2/sys_mmapper.h
@@ -56,7 +56,7 @@ struct page_fault_event
 struct page_fault_event_entries
 {
 	std::vector<page_fault_event> events;
-	semaphore<> pf_mutex;
+	shared_mutex pf_mutex;
 };
 
 // SysCalls
diff --git a/rpcs3/Emu/Cell/lv2/sys_mutex.h b/rpcs3/Emu/Cell/lv2/sys_mutex.h
index 28ce95d430..2d60d31f7a 100644
--- a/rpcs3/Emu/Cell/lv2/sys_mutex.h
+++ b/rpcs3/Emu/Cell/lv2/sys_mutex.h
@@ -31,7 +31,7 @@ struct lv2_mutex final : lv2_obj
 	const u64 name;
 	const s32 flags;
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 	atomic_t<u32> owner{0}; // Owner Thread ID
 	atomic_t<u32> lock_count{0}; // Recursive Locks
 	atomic_t<u32> cond_count{0}; // Condition Variables
diff --git a/rpcs3/Emu/Cell/lv2/sys_net.cpp b/rpcs3/Emu/Cell/lv2/sys_net.cpp
index 10076edd8e..427e752992 100644
--- a/rpcs3/Emu/Cell/lv2/sys_net.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_net.cpp
@@ -30,7 +30,7 @@ LOG_CHANNEL(sys_net);
 
 static std::vector<ppu_thread*> s_to_awake;
 
-static semaphore<> s_nw_mutex;
+static shared_mutex s_nw_mutex;
 
 extern u64 get_system_time();
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_net.h b/rpcs3/Emu/Cell/lv2/sys_net.h
index bc02c27734..191677c513 100644
--- a/rpcs3/Emu/Cell/lv2/sys_net.h
+++ b/rpcs3/Emu/Cell/lv2/sys_net.h
@@ -1,7 +1,7 @@
 #pragma once
 
 #include "Utilities/bit_set.h"
-#include "Utilities/sema.h"
+#include "Utilities/mutex.h"
 
 #include <vector>
 #include <utility>
@@ -324,7 +324,7 @@ struct lv2_socket final
 	lv2_socket(socket_type s);
 	~lv2_socket();
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 
 #ifdef _WIN32
 	// Remember events (WSAEnumNetworkEvents)
diff --git a/rpcs3/Emu/Cell/lv2/sys_rwlock.h b/rpcs3/Emu/Cell/lv2/sys_rwlock.h
index de4759bb83..86b8772858 100644
--- a/rpcs3/Emu/Cell/lv2/sys_rwlock.h
+++ b/rpcs3/Emu/Cell/lv2/sys_rwlock.h
@@ -27,7 +27,7 @@ struct lv2_rwlock final : lv2_obj
 	const u64 name;
 	const s32 flags;
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 	atomic_t<s64> owner{0};
 	std::deque<cpu_thread*> rq;
 	std::deque<cpu_thread*> wq;
diff --git a/rpcs3/Emu/Cell/lv2/sys_semaphore.h b/rpcs3/Emu/Cell/lv2/sys_semaphore.h
index 580ef572b9..3c3862fbb2 100644
--- a/rpcs3/Emu/Cell/lv2/sys_semaphore.h
+++ b/rpcs3/Emu/Cell/lv2/sys_semaphore.h
@@ -28,7 +28,7 @@ struct lv2_sema final : lv2_obj
 	const s32 flags;
 	const s32 max;
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 	atomic_t<s32> val;
 	std::deque<cpu_thread*> sq;
 
diff --git a/rpcs3/Emu/Cell/lv2/sys_sync.h b/rpcs3/Emu/Cell/lv2/sys_sync.h
index f9f53ec5fb..429b1c74ed 100644
--- a/rpcs3/Emu/Cell/lv2/sys_sync.h
+++ b/rpcs3/Emu/Cell/lv2/sys_sync.h
@@ -215,7 +215,7 @@ struct lv2_obj
 
 private:
 	// Scheduler mutex
-	static semaphore<> g_mutex;
+	static shared_mutex g_mutex;
 
 	// Scheduler queue for active PPU threads
 	static std::deque<class ppu_thread*> g_ppu;
diff --git a/rpcs3/Emu/Cell/lv2/sys_timer.h b/rpcs3/Emu/Cell/lv2/sys_timer.h
index b24a4d16ae..a13ba6a0a6 100644
--- a/rpcs3/Emu/Cell/lv2/sys_timer.h
+++ b/rpcs3/Emu/Cell/lv2/sys_timer.h
@@ -24,7 +24,7 @@ struct lv2_timer_context : lv2_obj
 	void operator()();
 	void on_abort();
 
-	semaphore<> mutex;
+	shared_mutex mutex;
 	atomic_t<u32> state{SYS_TIMER_STATE_STOP};
 
 	std::weak_ptr<lv2_event_queue> port;
diff --git a/rpcs3/Emu/Io/MouseHandler.h b/rpcs3/Emu/Io/MouseHandler.h
index c4437bb338..77ea3a7824 100644
--- a/rpcs3/Emu/Io/MouseHandler.h
+++ b/rpcs3/Emu/Io/MouseHandler.h
@@ -1,7 +1,7 @@
 #pragma once
 
 #include <list>
-#include "Utilities/sema.h"
+#include "Utilities/mutex.h"
 
 // TODO: HLE info (constants, structs, etc.) should not be available here
 
@@ -145,7 +145,7 @@ protected:
 	}
 
 public:
-	semaphore<> mutex;
+	shared_mutex mutex;
 
 	virtual void Init(const u32 max_connect) = 0;
 	virtual ~MouseHandlerBase() = default;
diff --git a/rpcs3/rpcs3qt/log_frame.cpp b/rpcs3/rpcs3qt/log_frame.cpp
index afa81290dc..c894d62d4d 100644
--- a/rpcs3/rpcs3qt/log_frame.cpp
+++ b/rpcs3/rpcs3qt/log_frame.cpp
@@ -13,7 +13,7 @@
 
 #include <deque>
 #include <mutex>
-#include "Utilities/sema.h"
+#include "Utilities/mutex.h"
 
 extern fs::file g_tty;
 extern atomic_t<s64> g_tty_size;