5 files changed, 69 insertions, 118 deletions

diff --git a/src/common/uint128.h b/src/common/uint128.h
index 199d0f55e..f890ffec2 100644
--- a/src/common/uint128.h
+++ b/src/common/uint128.h
@@ -31,17 +31,12 @@ namespace Common {
     return _udiv128(r[1], r[0], d, &remainder);
 #endif
 #else
-#ifdef __SIZEOF_INT128__
-    const auto product = static_cast<unsigned __int128>(a) * static_cast<unsigned __int128>(b);
-    return static_cast<u64>(product / d);
-#else
     const u64 diva = a / d;
     const u64 moda = a % d;
     const u64 divb = b / d;
     const u64 modb = b % d;
     return diva * b + moda * divb + moda * modb / d;
 #endif
-#endif
 }
 
 // This function multiplies 2 u64 values and produces a u128 value;
diff --git a/src/common/x64/native_clock.cpp b/src/common/x64/native_clock.cpp
index 6aaa8cdf9..8b08332ab 100644
--- a/src/common/x64/native_clock.cpp
+++ b/src/common/x64/native_clock.cpp
@@ -75,8 +75,8 @@ NativeClock::NativeClock(u64 emulated_cpu_frequency_, u64 emulated_clock_frequen
 }
 
 u64 NativeClock::GetRTSC() {
-    TimePoint current_time_point{};
     TimePoint new_time_point{};
+    TimePoint current_time_point{};
 
     current_time_point.pack = Common::AtomicLoad128(time_point.pack.data());
     do {
diff --git a/src/core/core_timing.cpp b/src/core/core_timing.cpp
index 5425637f5..2dbb99c8b 100644
--- a/src/core/core_timing.cpp
+++ b/src/core/core_timing.cpp
@@ -6,9 +6,7 @@
 #include <string>
 #include <tuple>
 
-#include "common/logging/log.h"
 #include "common/microprofile.h"
-#include "common/thread.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hardware_properties.h"
@@ -44,10 +42,10 @@ CoreTiming::CoreTiming()
 
 CoreTiming::~CoreTiming() = default;
 
-void CoreTiming::ThreadEntry(CoreTiming& instance, size_t id) {
-    const std::string name = "yuzu:HostTiming_" + std::to_string(id);
-    MicroProfileOnThreadCreate(name.c_str());
-    Common::SetCurrentThreadName(name.c_str());
+void CoreTiming::ThreadEntry(CoreTiming& instance) {
+    constexpr char name[] = "yuzu:HostTiming";
+    MicroProfileOnThreadCreate(name);
+    Common::SetCurrentThreadName(name);
     Common::SetCurrentThreadPriority(Common::ThreadPriority::Critical);
     instance.on_thread_init();
     instance.ThreadLoop();
@@ -63,127 +61,100 @@ void CoreTiming::Initialize(std::function<void()>&& on_thread_init_) {
         -> std::optional<std::chrono::nanoseconds> { return std::nullopt; };
     ev_lost = CreateEvent("_lost_event", empty_timed_callback);
     if (is_multicore) {
-        worker_threads.emplace_back(ThreadEntry, std::ref(*this), 0);
+        timer_thread = std::make_unique<std::thread>(ThreadEntry, std::ref(*this));
     }
 }
 
 void CoreTiming::Shutdown() {
-    is_paused = true;
+    paused = true;
     shutting_down = true;
-    std::atomic_thread_fence(std::memory_order_release);
-
-    event_cv.notify_all();
-    wait_pause_cv.notify_all();
-    for (auto& thread : worker_threads) {
-        thread.join();
+    pause_event.Set();
+    event.Set();
+    if (timer_thread) {
+        timer_thread->join();
     }
-    worker_threads.clear();
     pause_callbacks.clear();
     ClearPendingEvents();
+    timer_thread.reset();
     has_started = false;
 }
 
-void CoreTiming::Pause(bool is_paused_) {
-    std::unique_lock main_lock(event_mutex);
-    if (is_paused_ == paused_state.load(std::memory_order_relaxed)) {
-        return;
-    }
-    if (is_multicore) {
-        is_paused = is_paused_;
-        event_cv.notify_all();
-        if (!is_paused_) {
-            wait_pause_cv.notify_all();
-        }
-    }
-    paused_state.store(is_paused_, std::memory_order_relaxed);
+void CoreTiming::Pause(bool is_paused) {
+    paused = is_paused;
+    pause_event.Set();
 
-    if (!is_paused_) {
+    if (!is_paused) {
         pause_end_time = GetGlobalTimeNs().count();
     }
 
     for (auto& cb : pause_callbacks) {
-        cb(is_paused_);
+        cb(is_paused);
     }
 }
 
-void CoreTiming::SyncPause(bool is_paused_) {
-    std::unique_lock main_lock(event_mutex);
-    if (is_paused_ == paused_state.load(std::memory_order_relaxed)) {
+void CoreTiming::SyncPause(bool is_paused) {
+    if (is_paused == paused && paused_set == paused) {
         return;
     }
 
-    if (is_multicore) {
-        is_paused = is_paused_;
-        event_cv.notify_all();
-        if (!is_paused_) {
-            wait_pause_cv.notify_all();
-        }
-    }
-    paused_state.store(is_paused_, std::memory_order_relaxed);
-    if (is_multicore) {
-        if (is_paused_) {
-            wait_signal_cv.wait(main_lock, [this] { return pause_count == worker_threads.size(); });
-        } else {
-            wait_signal_cv.wait(main_lock, [this] { return pause_count == 0; });
+    Pause(is_paused);
+    if (timer_thread) {
+        if (!is_paused) {
+            pause_event.Set();
         }
+        event.Set();
+        while (paused_set != is_paused)
+            ;
     }
 
-    if (!is_paused_) {
+    if (!is_paused) {
         pause_end_time = GetGlobalTimeNs().count();
     }
 
     for (auto& cb : pause_callbacks) {
-        cb(is_paused_);
+        cb(is_paused);
     }
 }
 
 bool CoreTiming::IsRunning() const {
-    return !paused_state.load(std::memory_order_acquire);
+    return !paused_set;
 }
 
 bool CoreTiming::HasPendingEvents() const {
-    std::unique_lock main_lock(event_mutex);
-    return !event_queue.empty() || pending_events.load(std::memory_order_relaxed) != 0;
+    return !(wait_set && event_queue.empty());
 }
 
 void CoreTiming::ScheduleEvent(std::chrono::nanoseconds ns_into_future,
                                const std::shared_ptr<EventType>& event_type,
                                std::uintptr_t user_data, bool absolute_time) {
+    {
+        std::scoped_lock scope{basic_lock};
+        const auto next_time{absolute_time ? ns_into_future : GetGlobalTimeNs() + ns_into_future};
 
-    std::unique_lock main_lock(event_mutex);
-    const auto next_time{absolute_time ? ns_into_future : GetGlobalTimeNs() + ns_into_future};
-
-    event_queue.emplace_back(Event{next_time.count(), event_fifo_id++, user_data, event_type, 0});
-    pending_events.fetch_add(1, std::memory_order_relaxed);
-
-    std::push_heap(event_queue.begin(), event_queue.end(), std::greater<>());
-
-    if (is_multicore) {
-        event_cv.notify_one();
+        event_queue.emplace_back(
+            Event{next_time.count(), event_fifo_id++, user_data, event_type, 0});
+        std::push_heap(event_queue.begin(), event_queue.end(), std::greater<>());
     }
+
+    event.Set();
 }
 
 void CoreTiming::ScheduleLoopingEvent(std::chrono::nanoseconds start_time,
                                       std::chrono::nanoseconds resched_time,
                                       const std::shared_ptr<EventType>& event_type,
                                       std::uintptr_t user_data, bool absolute_time) {
-    std::unique_lock main_lock(event_mutex);
+    std::scoped_lock scope{basic_lock};
     const auto next_time{absolute_time ? start_time : GetGlobalTimeNs() + start_time};
 
     event_queue.emplace_back(
         Event{next_time.count(), event_fifo_id++, user_data, event_type, resched_time.count()});
-    pending_events.fetch_add(1, std::memory_order_relaxed);
 
     std::push_heap(event_queue.begin(), event_queue.end(), std::greater<>());
-
-    if (is_multicore) {
-        event_cv.notify_one();
-    }
 }
 
 void CoreTiming::UnscheduleEvent(const std::shared_ptr<EventType>& event_type,
                                  std::uintptr_t user_data) {
-    std::unique_lock main_lock(event_mutex);
+    std::scoped_lock scope{basic_lock};
     const auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
         return e.type.lock().get() == event_type.get() && e.user_data == user_data;
     });
@@ -192,7 +163,6 @@ void CoreTiming::UnscheduleEvent(const std::shared_ptr<EventType>& event_type,
     if (itr != event_queue.end()) {
         event_queue.erase(itr, event_queue.end());
         std::make_heap(event_queue.begin(), event_queue.end(), std::greater<>());
-        pending_events.fetch_sub(1, std::memory_order_relaxed);
     }
 }
 
@@ -232,12 +202,11 @@ u64 CoreTiming::GetClockTicks() const {
 }
 
 void CoreTiming::ClearPendingEvents() {
-    std::unique_lock main_lock(event_mutex);
     event_queue.clear();
 }
 
 void CoreTiming::RemoveEvent(const std::shared_ptr<EventType>& event_type) {
-    std::unique_lock main_lock(event_mutex);
+    std::scoped_lock lock{basic_lock};
 
     const auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
         return e.type.lock().get() == event_type.get();
@@ -251,28 +220,27 @@ void CoreTiming::RemoveEvent(const std::shared_ptr<EventType>& event_type) {
 }
 
 void CoreTiming::RegisterPauseCallback(PauseCallback&& callback) {
-    std::unique_lock main_lock(event_mutex);
+    std::scoped_lock lock{basic_lock};
     pause_callbacks.emplace_back(std::move(callback));
 }
 
 std::optional<s64> CoreTiming::Advance() {
+    std::scoped_lock lock{advance_lock, basic_lock};
     global_timer = GetGlobalTimeNs().count();
 
-    std::unique_lock main_lock(event_mutex);
     while (!event_queue.empty() && event_queue.front().time <= global_timer) {
         Event evt = std::move(event_queue.front());
         std::pop_heap(event_queue.begin(), event_queue.end(), std::greater<>());
         event_queue.pop_back();
 
         if (const auto event_type{evt.type.lock()}) {
-            event_mutex.unlock();
+            basic_lock.unlock();
 
             const auto new_schedule_time{event_type->callback(
                 evt.user_data, evt.time,
                 std::chrono::nanoseconds{GetGlobalTimeNs().count() - evt.time})};
 
-            event_mutex.lock();
-            pending_events.fetch_sub(1, std::memory_order_relaxed);
+            basic_lock.lock();
 
             if (evt.reschedule_time != 0) {
                 // If this event was scheduled into a pause, its time now is going to be way behind.
@@ -285,9 +253,9 @@ std::optional<s64> CoreTiming::Advance() {
                 const auto next_schedule_time{new_schedule_time.has_value()
                                                   ? new_schedule_time.value().count()
                                                   : evt.reschedule_time};
+
                 event_queue.emplace_back(
                     Event{next_time, event_fifo_id++, evt.user_data, evt.type, next_schedule_time});
-                pending_events.fetch_add(1, std::memory_order_relaxed);
                 std::push_heap(event_queue.begin(), event_queue.end(), std::greater<>());
             }
         }
@@ -304,34 +272,27 @@ std::optional<s64> CoreTiming::Advance() {
 }
 
 void CoreTiming::ThreadLoop() {
-    const auto predicate = [this] { return !event_queue.empty() || is_paused; };
     has_started = true;
     while (!shutting_down) {
-        while (!is_paused && !shutting_down) {
+        while (!paused) {
+            paused_set = false;
             const auto next_time = Advance();
             if (next_time) {
                 if (*next_time > 0) {
                     std::chrono::nanoseconds next_time_ns = std::chrono::nanoseconds(*next_time);
-                    std::unique_lock main_lock(event_mutex);
-                    event_cv.wait_for(main_lock, next_time_ns, predicate);
+                    event.WaitFor(next_time_ns);
                 }
             } else {
-                std::unique_lock main_lock(event_mutex);
-                event_cv.wait(main_lock, predicate);
+                wait_set = true;
+                event.Wait();
             }
+            wait_set = false;
         }
-        std::unique_lock main_lock(event_mutex);
-        pause_count++;
-        if (pause_count == worker_threads.size()) {
-            clock->Pause(true);
-            wait_signal_cv.notify_all();
-        }
-        wait_pause_cv.wait(main_lock, [this] { return !is_paused || shutting_down; });
-        pause_count--;
-        if (pause_count == 0) {
-            clock->Pause(false);
-            wait_signal_cv.notify_all();
-        }
+
+        paused_set = true;
+        clock->Pause(true);
+        pause_event.Wait();
+        clock->Pause(false);
     }
 }
 
diff --git a/src/core/core_timing.h b/src/core/core_timing.h
index 09b6ed81a..6aa3ae923 100644
--- a/src/core/core_timing.h
+++ b/src/core/core_timing.h
@@ -5,7 +5,6 @@
 
 #include <atomic>
 #include <chrono>
-#include <condition_variable>
 #include <functional>
 #include <memory>
 #include <mutex>
@@ -15,6 +14,7 @@
 #include <vector>
 
 #include "common/common_types.h"
+#include "common/thread.h"
 #include "common/wall_clock.h"
 
 namespace Core::Timing {
@@ -143,7 +143,7 @@ private:
     /// Clear all pending events. This should ONLY be done on exit.
     void ClearPendingEvents();
 
-    static void ThreadEntry(CoreTiming& instance, size_t id);
+    static void ThreadEntry(CoreTiming& instance);
     void ThreadLoop();
 
     std::unique_ptr<Common::WallClock> clock;
@@ -156,24 +156,21 @@ private:
     // accomodated by the standard adaptor class.
     std::vector<Event> event_queue;
     u64 event_fifo_id = 0;
-    std::atomic<size_t> pending_events{};
 
     std::shared_ptr<EventType> ev_lost;
+    Common::Event event{};
+    Common::Event pause_event{};
+    std::mutex basic_lock;
+    std::mutex advance_lock;
+    std::unique_ptr<std::thread> timer_thread;
+    std::atomic<bool> paused{};
+    std::atomic<bool> paused_set{};
+    std::atomic<bool> wait_set{};
+    std::atomic<bool> shutting_down{};
     std::atomic<bool> has_started{};
     std::function<void()> on_thread_init{};
 
-    std::vector<std::thread> worker_threads;
-
-    std::condition_variable event_cv;
-    std::condition_variable wait_pause_cv;
-    std::condition_variable wait_signal_cv;
-    mutable std::mutex event_mutex;
-
-    std::atomic<bool> paused_state{};
-    bool is_paused{};
-    bool shutting_down{};
     bool is_multicore{};
-    size_t pause_count{};
     s64 pause_end_time{};
 
     /// Cycle timing
diff --git a/src/tests/core/core_timing.cpp b/src/tests/core/core_timing.cpp
index 894975e6f..7c432a63c 100644
--- a/src/tests/core/core_timing.cpp
+++ b/src/tests/core/core_timing.cpp
@@ -8,7 +8,6 @@
 #include <chrono>
 #include <cstdlib>
 #include <memory>
-#include <mutex>
 #include <optional>
 #include <string>
 
@@ -23,15 +22,14 @@ std::array<s64, 5> delays{};
 
 std::bitset<CB_IDS.size()> callbacks_ran_flags;
 u64 expected_callback = 0;
-std::mutex control_mutex;
 
 template <unsigned int IDX>
 std::optional<std::chrono::nanoseconds> HostCallbackTemplate(std::uintptr_t user_data, s64 time,
                                                              std::chrono::nanoseconds ns_late) {
-    std::unique_lock<std::mutex> lk(control_mutex);
     static_assert(IDX < CB_IDS.size(), "IDX out of range");
     callbacks_ran_flags.set(IDX);
     REQUIRE(CB_IDS[IDX] == user_data);
+    REQUIRE(CB_IDS[IDX] == CB_IDS[calls_order[expected_callback]]);
     delays[IDX] = ns_late.count();
     ++expected_callback;
     return std::nullopt;