42 files changed, 2449 insertions, 1665 deletions

diff --git a/src/core/hle/kernel/address_arbiter.cpp b/src/core/hle/kernel/address_arbiter.cpp
index b882eaa0f..20ffa7d47 100644
--- a/src/core/hle/kernel/address_arbiter.cpp
+++ b/src/core/hle/kernel/address_arbiter.cpp
@@ -12,8 +12,9 @@
 #include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/time_manager.h"
 #include "core/hle/result.h"
@@ -58,7 +59,7 @@ ResultCode AddressArbiter::SignalToAddress(VAddr address, SignalType type, s32 v
 }
 
 ResultCode AddressArbiter::SignalToAddressOnly(VAddr address, s32 num_to_wake) {
-    SchedulerLock lock(system.Kernel());
+    KScopedSchedulerLock lock(system.Kernel());
     const std::vector<std::shared_ptr<Thread>> waiting_threads =
         GetThreadsWaitingOnAddress(address);
     WakeThreads(waiting_threads, num_to_wake);
@@ -67,7 +68,7 @@ ResultCode AddressArbiter::SignalToAddressOnly(VAddr address, s32 num_to_wake) {
 
 ResultCode AddressArbiter::IncrementAndSignalToAddressIfEqual(VAddr address, s32 value,
                                                               s32 num_to_wake) {
-    SchedulerLock lock(system.Kernel());
+    KScopedSchedulerLock lock(system.Kernel());
     auto& memory = system.Memory();
 
     // Ensure that we can write to the address.
@@ -92,7 +93,7 @@ ResultCode AddressArbiter::IncrementAndSignalToAddressIfEqual(VAddr address, s32
 
 ResultCode AddressArbiter::ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
                                                                          s32 num_to_wake) {
-    SchedulerLock lock(system.Kernel());
+    KScopedSchedulerLock lock(system.Kernel());
     auto& memory = system.Memory();
 
     // Ensure that we can write to the address.
@@ -153,11 +154,11 @@ ResultCode AddressArbiter::WaitForAddressIfLessThan(VAddr address, s32 value, s6
                                                     bool should_decrement) {
     auto& memory = system.Memory();
     auto& kernel = system.Kernel();
-    Thread* current_thread = system.CurrentScheduler().GetCurrentThread();
+    Thread* current_thread = kernel.CurrentScheduler()->GetCurrentThread();
 
     Handle event_handle = InvalidHandle;
     {
-        SchedulerLockAndSleep lock(kernel, event_handle, current_thread, timeout);
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, current_thread, timeout);
 
         if (current_thread->IsPendingTermination()) {
             lock.CancelSleep();
@@ -210,7 +211,7 @@ ResultCode AddressArbiter::WaitForAddressIfLessThan(VAddr address, s32 value, s6
     }
 
     {
-        SchedulerLock lock(kernel);
+        KScopedSchedulerLock lock(kernel);
         if (current_thread->IsWaitingForArbitration()) {
             RemoveThread(SharedFrom(current_thread));
             current_thread->WaitForArbitration(false);
@@ -223,11 +224,11 @@ ResultCode AddressArbiter::WaitForAddressIfLessThan(VAddr address, s32 value, s6
 ResultCode AddressArbiter::WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout) {
     auto& memory = system.Memory();
     auto& kernel = system.Kernel();
-    Thread* current_thread = system.CurrentScheduler().GetCurrentThread();
+    Thread* current_thread = kernel.CurrentScheduler()->GetCurrentThread();
 
     Handle event_handle = InvalidHandle;
     {
-        SchedulerLockAndSleep lock(kernel, event_handle, current_thread, timeout);
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, current_thread, timeout);
 
         if (current_thread->IsPendingTermination()) {
             lock.CancelSleep();
@@ -265,7 +266,7 @@ ResultCode AddressArbiter::WaitForAddressIfEqual(VAddr address, s32 value, s64 t
     }
 
     {
-        SchedulerLock lock(kernel);
+        KScopedSchedulerLock lock(kernel);
         if (current_thread->IsWaitingForArbitration()) {
             RemoveThread(SharedFrom(current_thread));
             current_thread->WaitForArbitration(false);
@@ -275,12 +276,6 @@ ResultCode AddressArbiter::WaitForAddressIfEqual(VAddr address, s32 value, s64 t
     return current_thread->GetSignalingResult();
 }
 
-void AddressArbiter::HandleWakeupThread(std::shared_ptr<Thread> thread) {
-    ASSERT(thread->GetStatus() == ThreadStatus::WaitArb);
-    RemoveThread(thread);
-    thread->SetArbiterWaitAddress(0);
-}
-
 void AddressArbiter::InsertThread(std::shared_ptr<Thread> thread) {
     const VAddr arb_addr = thread->GetArbiterWaitAddress();
     std::list<std::shared_ptr<Thread>>& thread_list = arb_threads[arb_addr];
diff --git a/src/core/hle/kernel/address_arbiter.h b/src/core/hle/kernel/address_arbiter.h
index 0b05d533c..b91edc67d 100644
--- a/src/core/hle/kernel/address_arbiter.h
+++ b/src/core/hle/kernel/address_arbiter.h
@@ -50,9 +50,6 @@ public:
     /// Waits on an address with a particular arbitration type.
     ResultCode WaitForAddress(VAddr address, ArbitrationType type, s32 value, s64 timeout_ns);
 
-    /// Removes a thread from the container and resets its address arbiter adress to 0
-    void HandleWakeupThread(std::shared_ptr<Thread> thread);
-
 private:
     /// Signals an address being waited on.
     ResultCode SignalToAddressOnly(VAddr address, s32 num_to_wake);
diff --git a/src/core/hle/kernel/global_scheduler_context.cpp b/src/core/hle/kernel/global_scheduler_context.cpp
new file mode 100644
index 000000000..a133e8ed0
--- /dev/null
+++ b/src/core/hle/kernel/global_scheduler_context.cpp
@@ -0,0 +1,52 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <mutex>
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "core/hle/kernel/global_scheduler_context.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
+
+namespace Kernel {
+
+GlobalSchedulerContext::GlobalSchedulerContext(KernelCore& kernel)
+    : kernel{kernel}, scheduler_lock{kernel} {}
+
+GlobalSchedulerContext::~GlobalSchedulerContext() = default;
+
+void GlobalSchedulerContext::AddThread(std::shared_ptr<Thread> thread) {
+    std::scoped_lock lock{global_list_guard};
+    thread_list.push_back(std::move(thread));
+}
+
+void GlobalSchedulerContext::RemoveThread(std::shared_ptr<Thread> thread) {
+    std::scoped_lock lock{global_list_guard};
+    thread_list.erase(std::remove(thread_list.begin(), thread_list.end(), thread),
+                      thread_list.end());
+}
+
+void GlobalSchedulerContext::PreemptThreads() {
+    // The priority levels at which the global scheduler preempts threads every 10 ms. They are
+    // ordered from Core 0 to Core 3.
+    static constexpr std::array<u32, Core::Hardware::NUM_CPU_CORES> preemption_priorities{
+        59,
+        59,
+        59,
+        63,
+    };
+
+    ASSERT(IsLocked());
+    for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
+        const u32 priority = preemption_priorities[core_id];
+        kernel.Scheduler(core_id).RotateScheduledQueue(core_id, priority);
+    }
+}
+
+bool GlobalSchedulerContext::IsLocked() const {
+    return scheduler_lock.IsLockedByCurrentThread();
+}
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/global_scheduler_context.h b/src/core/hle/kernel/global_scheduler_context.h
new file mode 100644
index 000000000..5c7b89290
--- /dev/null
+++ b/src/core/hle/kernel/global_scheduler_context.h
@@ -0,0 +1,81 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+#include <vector>
+
+#include "common/common_types.h"
+#include "common/spin_lock.h"
+#include "core/hardware_properties.h"
+#include "core/hle/kernel/k_priority_queue.h"
+#include "core/hle/kernel/k_scheduler_lock.h"
+#include "core/hle/kernel/thread.h"
+
+namespace Kernel {
+
+class KernelCore;
+class SchedulerLock;
+
+using KSchedulerPriorityQueue =
+    KPriorityQueue<Thread, Core::Hardware::NUM_CPU_CORES, THREADPRIO_LOWEST, THREADPRIO_HIGHEST>;
+constexpr s32 HighestCoreMigrationAllowedPriority = 2;
+
+class GlobalSchedulerContext final {
+    friend class KScheduler;
+
+public:
+    using LockType = KAbstractSchedulerLock<KScheduler>;
+
+    explicit GlobalSchedulerContext(KernelCore& kernel);
+    ~GlobalSchedulerContext();
+
+    /// Adds a new thread to the scheduler
+    void AddThread(std::shared_ptr<Thread> thread);
+
+    /// Removes a thread from the scheduler
+    void RemoveThread(std::shared_ptr<Thread> thread);
+
+    /// Returns a list of all threads managed by the scheduler
+    [[nodiscard]] const std::vector<std::shared_ptr<Thread>>& GetThreadList() const {
+        return thread_list;
+    }
+
+    /**
+     * Rotates the scheduling queues of threads at a preemption priority and then does
+     * some core rebalancing. Preemption priorities can be found in the array
+     * 'preemption_priorities'.
+     *
+     * @note This operation happens every 10ms.
+     */
+    void PreemptThreads();
+
+    /// Returns true if the global scheduler lock is acquired
+    bool IsLocked() const;
+
+    [[nodiscard]] LockType& SchedulerLock() {
+        return scheduler_lock;
+    }
+
+    [[nodiscard]] const LockType& SchedulerLock() const {
+        return scheduler_lock;
+    }
+
+private:
+    friend class KScopedSchedulerLock;
+    friend class KScopedSchedulerLockAndSleep;
+
+    KernelCore& kernel;
+
+    std::atomic_bool scheduler_update_needed{};
+    KSchedulerPriorityQueue priority_queue;
+    LockType scheduler_lock;
+
+    /// Lists all thread ids that aren't deleted/etc.
+    std::vector<std::shared_ptr<Thread>> thread_list;
+    Common::SpinLock global_list_guard{};
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/handle_table.cpp b/src/core/hle/kernel/handle_table.cpp
index fb30b6f8b..40988b0fd 100644
--- a/src/core/hle/kernel/handle_table.cpp
+++ b/src/core/hle/kernel/handle_table.cpp
@@ -8,9 +8,9 @@
 #include "core/core.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 
 namespace Kernel {
@@ -105,7 +105,7 @@ bool HandleTable::IsValid(Handle handle) const {
 
 std::shared_ptr<Object> HandleTable::GetGeneric(Handle handle) const {
     if (handle == CurrentThread) {
-        return SharedFrom(kernel.CurrentScheduler().GetCurrentThread());
+        return SharedFrom(kernel.CurrentScheduler()->GetCurrentThread());
     } else if (handle == CurrentProcess) {
         return SharedFrom(kernel.CurrentProcess());
     }
@@ -118,7 +118,7 @@ std::shared_ptr<Object> HandleTable::GetGeneric(Handle handle) const {
 
 void HandleTable::Clear() {
     for (u16 i = 0; i < table_size; ++i) {
-        generations[i] = i + 1;
+        generations[i] = static_cast<u16>(i + 1);
         objects[i] = nullptr;
     }
     next_free_slot = 0;
diff --git a/src/core/hle/kernel/hle_ipc.cpp b/src/core/hle/kernel/hle_ipc.cpp
index 81f85643b..83decf6cf 100644
--- a/src/core/hle/kernel/hle_ipc.cpp
+++ b/src/core/hle/kernel/hle_ipc.cpp
@@ -17,11 +17,12 @@
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/readable_event.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/server_session.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/time_manager.h"
@@ -45,44 +46,6 @@ void SessionRequestHandler::ClientDisconnected(
     boost::range::remove_erase(connected_sessions, server_session);
 }
 
-std::shared_ptr<WritableEvent> HLERequestContext::SleepClientThread(
-    const std::string& reason, u64 timeout, WakeupCallback&& callback,
-    std::shared_ptr<WritableEvent> writable_event) {
-    // Put the client thread to sleep until the wait event is signaled or the timeout expires.
-
-    if (!writable_event) {
-        // Create event if not provided
-        const auto pair = WritableEvent::CreateEventPair(kernel, "HLE Pause Event: " + reason);
-        writable_event = pair.writable;
-    }
-
-    {
-        Handle event_handle = InvalidHandle;
-        SchedulerLockAndSleep lock(kernel, event_handle, thread.get(), timeout);
-        thread->SetHLECallback(
-            [context = *this, callback](std::shared_ptr<Thread> thread) mutable -> bool {
-                ThreadWakeupReason reason = thread->GetSignalingResult() == RESULT_TIMEOUT
-                                                ? ThreadWakeupReason::Timeout
-                                                : ThreadWakeupReason::Signal;
-                callback(thread, context, reason);
-                context.WriteToOutgoingCommandBuffer(*thread);
-                return true;
-            });
-        const auto readable_event{writable_event->GetReadableEvent()};
-        writable_event->Clear();
-        thread->SetHLESyncObject(readable_event.get());
-        thread->SetStatus(ThreadStatus::WaitHLEEvent);
-        thread->SetSynchronizationResults(nullptr, RESULT_TIMEOUT);
-        readable_event->AddWaitingThread(thread);
-        lock.Release();
-        thread->SetHLETimeEvent(event_handle);
-    }
-
-    is_thread_waiting = true;
-
-    return writable_event;
-}
-
 HLERequestContext::HLERequestContext(KernelCore& kernel, Core::Memory::Memory& memory,
                                      std::shared_ptr<ServerSession> server_session,
                                      std::shared_ptr<Thread> thread)
diff --git a/src/core/hle/kernel/hle_ipc.h b/src/core/hle/kernel/hle_ipc.h
index f3277b766..b112e1ebd 100644
--- a/src/core/hle/kernel/hle_ipc.h
+++ b/src/core/hle/kernel/hle_ipc.h
@@ -24,6 +24,10 @@ namespace Core::Memory {
 class Memory;
 }
 
+namespace IPC {
+class ResponseBuilder;
+}
+
 namespace Service {
 class ServiceFrameworkBase;
 }
@@ -125,23 +129,6 @@ public:
     using WakeupCallback = std::function<void(
         std::shared_ptr<Thread> thread, HLERequestContext& context, ThreadWakeupReason reason)>;
 
-    /**
-     * Puts the specified guest thread to sleep until the returned event is signaled or until the
-     * specified timeout expires.
-     * @param reason Reason for pausing the thread, to be used for debugging purposes.
-     * @param timeout Timeout in nanoseconds after which the thread will be awoken and the callback
-     * invoked with a Timeout reason.
-     * @param callback Callback to be invoked when the thread is resumed. This callback must write
-     * the entire command response once again, regardless of the state of it before this function
-     * was called.
-     * @param writable_event Event to use to wake up the thread. If unspecified, an event will be
-     * created.
-     * @returns Event that when signaled will resume the thread and call the callback function.
-     */
-    std::shared_ptr<WritableEvent> SleepClientThread(
-        const std::string& reason, u64 timeout, WakeupCallback&& callback,
-        std::shared_ptr<WritableEvent> writable_event = nullptr);
-
     /// Populates this context with data from the requesting process/thread.
     ResultCode PopulateFromIncomingCommandBuffer(const HandleTable& handle_table,
                                                  u32_le* src_cmdbuf);
@@ -287,6 +274,8 @@ public:
     }
 
 private:
+    friend class IPC::ResponseBuilder;
+
     void ParseCommandBuffer(const HandleTable& handle_table, u32_le* src_cmdbuf, bool incoming);
 
     std::array<u32, IPC::COMMAND_BUFFER_LENGTH> cmd_buf;
diff --git a/src/core/hle/kernel/k_affinity_mask.h b/src/core/hle/kernel/k_affinity_mask.h
new file mode 100644
index 000000000..dd73781cd
--- /dev/null
+++ b/src/core/hle/kernel/k_affinity_mask.h
@@ -0,0 +1,58 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "core/hardware_properties.h"
+
+namespace Kernel {
+
+class KAffinityMask {
+public:
+    constexpr KAffinityMask() = default;
+
+    [[nodiscard]] constexpr u64 GetAffinityMask() const {
+        return this->mask;
+    }
+
+    constexpr void SetAffinityMask(u64 new_mask) {
+        ASSERT((new_mask & ~AllowedAffinityMask) == 0);
+        this->mask = new_mask;
+    }
+
+    [[nodiscard]] constexpr bool GetAffinity(s32 core) const {
+        return this->mask & GetCoreBit(core);
+    }
+
+    constexpr void SetAffinity(s32 core, bool set) {
+        ASSERT(0 <= core && core < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
+
+        if (set) {
+            this->mask |= GetCoreBit(core);
+        } else {
+            this->mask &= ~GetCoreBit(core);
+        }
+    }
+
+    constexpr void SetAll() {
+        this->mask = AllowedAffinityMask;
+    }
+
+private:
+    [[nodiscard]] static constexpr u64 GetCoreBit(s32 core) {
+        ASSERT(0 <= core && core < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
+        return (1ULL << core);
+    }
+
+    static constexpr u64 AllowedAffinityMask = (1ULL << Core::Hardware::NUM_CPU_CORES) - 1;
+
+    u64 mask{};
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/k_priority_queue.h b/src/core/hle/kernel/k_priority_queue.h
new file mode 100644
index 000000000..99fb8fe93
--- /dev/null
+++ b/src/core/hle/kernel/k_priority_queue.h
@@ -0,0 +1,451 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include <array>
+#include <concepts>
+
+#include "common/assert.h"
+#include "common/bit_set.h"
+#include "common/bit_util.h"
+#include "common/common_types.h"
+#include "common/concepts.h"
+
+namespace Kernel {
+
+class Thread;
+
+template <typename T>
+concept KPriorityQueueAffinityMask = !std::is_reference_v<T> && requires(T & t) {
+    { t.GetAffinityMask() }
+    ->Common::ConvertibleTo<u64>;
+    {t.SetAffinityMask(std::declval<u64>())};
+
+    { t.GetAffinity(std::declval<int32_t>()) }
+    ->std::same_as<bool>;
+    {t.SetAffinity(std::declval<int32_t>(), std::declval<bool>())};
+    {t.SetAll()};
+};
+
+template <typename T>
+concept KPriorityQueueMember = !std::is_reference_v<T> && requires(T & t) {
+    {typename T::QueueEntry()};
+    {(typename T::QueueEntry()).Initialize()};
+    {(typename T::QueueEntry()).SetPrev(std::addressof(t))};
+    {(typename T::QueueEntry()).SetNext(std::addressof(t))};
+    { (typename T::QueueEntry()).GetNext() }
+    ->std::same_as<T*>;
+    { (typename T::QueueEntry()).GetPrev() }
+    ->std::same_as<T*>;
+    { t.GetPriorityQueueEntry(std::declval<s32>()) }
+    ->std::same_as<typename T::QueueEntry&>;
+
+    {t.GetAffinityMask()};
+    { typename std::remove_cvref<decltype(t.GetAffinityMask())>::type() }
+    ->KPriorityQueueAffinityMask;
+
+    { t.GetActiveCore() }
+    ->Common::ConvertibleTo<s32>;
+    { t.GetPriority() }
+    ->Common::ConvertibleTo<s32>;
+};
+
+template <typename Member, size_t _NumCores, int LowestPriority, int HighestPriority>
+requires KPriorityQueueMember<Member> class KPriorityQueue {
+public:
+    using AffinityMaskType = typename std::remove_cv_t<
+        typename std::remove_reference<decltype(std::declval<Member>().GetAffinityMask())>::type>;
+
+    static_assert(LowestPriority >= 0);
+    static_assert(HighestPriority >= 0);
+    static_assert(LowestPriority >= HighestPriority);
+    static constexpr size_t NumPriority = LowestPriority - HighestPriority + 1;
+    static constexpr size_t NumCores = _NumCores;
+
+    static constexpr bool IsValidCore(s32 core) {
+        return 0 <= core && core < static_cast<s32>(NumCores);
+    }
+
+    static constexpr bool IsValidPriority(s32 priority) {
+        return HighestPriority <= priority && priority <= LowestPriority + 1;
+    }
+
+private:
+    using Entry = typename Member::QueueEntry;
+
+public:
+    class KPerCoreQueue {
+    private:
+        std::array<Entry, NumCores> root{};
+
+    public:
+        constexpr KPerCoreQueue() {
+            for (auto& per_core_root : root) {
+                per_core_root.Initialize();
+            }
+        }
+
+        constexpr bool PushBack(s32 core, Member* member) {
+            // Get the entry associated with the member.
+            Entry& member_entry = member->GetPriorityQueueEntry(core);
+
+            // Get the entry associated with the end of the queue.
+            Member* tail = this->root[core].GetPrev();
+            Entry& tail_entry =
+                (tail != nullptr) ? tail->GetPriorityQueueEntry(core) : this->root[core];
+
+            // Link the entries.
+            member_entry.SetPrev(tail);
+            member_entry.SetNext(nullptr);
+            tail_entry.SetNext(member);
+            this->root[core].SetPrev(member);
+
+            return tail == nullptr;
+        }
+
+        constexpr bool PushFront(s32 core, Member* member) {
+            // Get the entry associated with the member.
+            Entry& member_entry = member->GetPriorityQueueEntry(core);
+
+            // Get the entry associated with the front of the queue.
+            Member* head = this->root[core].GetNext();
+            Entry& head_entry =
+                (head != nullptr) ? head->GetPriorityQueueEntry(core) : this->root[core];
+
+            // Link the entries.
+            member_entry.SetPrev(nullptr);
+            member_entry.SetNext(head);
+            head_entry.SetPrev(member);
+            this->root[core].SetNext(member);
+
+            return (head == nullptr);
+        }
+
+        constexpr bool Remove(s32 core, Member* member) {
+            // Get the entry associated with the member.
+            Entry& member_entry = member->GetPriorityQueueEntry(core);
+
+            // Get the entries associated with next and prev.
+            Member* prev = member_entry.GetPrev();
+            Member* next = member_entry.GetNext();
+            Entry& prev_entry =
+                (prev != nullptr) ? prev->GetPriorityQueueEntry(core) : this->root[core];
+            Entry& next_entry =
+                (next != nullptr) ? next->GetPriorityQueueEntry(core) : this->root[core];
+
+            // Unlink.
+            prev_entry.SetNext(next);
+            next_entry.SetPrev(prev);
+
+            return (this->GetFront(core) == nullptr);
+        }
+
+        constexpr Member* GetFront(s32 core) const {
+            return this->root[core].GetNext();
+        }
+    };
+
+    class KPriorityQueueImpl {
+    public:
+        constexpr KPriorityQueueImpl() = default;
+
+        constexpr void PushBack(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority > LowestPriority) {
+                return;
+            }
+
+            if (this->queues[priority].PushBack(core, member)) {
+                this->available_priorities[core].SetBit(priority);
+            }
+        }
+
+        constexpr void PushFront(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority > LowestPriority) {
+                return;
+            }
+
+            if (this->queues[priority].PushFront(core, member)) {
+                this->available_priorities[core].SetBit(priority);
+            }
+        }
+
+        constexpr void Remove(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority > LowestPriority) {
+                return;
+            }
+
+            if (this->queues[priority].Remove(core, member)) {
+                this->available_priorities[core].ClearBit(priority);
+            }
+        }
+
+        constexpr Member* GetFront(s32 core) const {
+            ASSERT(IsValidCore(core));
+
+            const s32 priority =
+                static_cast<s32>(this->available_priorities[core].CountLeadingZero());
+            if (priority <= LowestPriority) {
+                return this->queues[priority].GetFront(core);
+            } else {
+                return nullptr;
+            }
+        }
+
+        constexpr Member* GetFront(s32 priority, s32 core) const {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority <= LowestPriority) {
+                return this->queues[priority].GetFront(core);
+            } else {
+                return nullptr;
+            }
+        }
+
+        constexpr Member* GetNext(s32 core, const Member* member) const {
+            ASSERT(IsValidCore(core));
+
+            Member* next = member->GetPriorityQueueEntry(core).GetNext();
+            if (next == nullptr) {
+                const s32 priority = static_cast<s32>(
+                    this->available_priorities[core].GetNextSet(member->GetPriority()));
+                if (priority <= LowestPriority) {
+                    next = this->queues[priority].GetFront(core);
+                }
+            }
+            return next;
+        }
+
+        constexpr void MoveToFront(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority <= LowestPriority) {
+                this->queues[priority].Remove(core, member);
+                this->queues[priority].PushFront(core, member);
+            }
+        }
+
+        constexpr Member* MoveToBack(s32 priority, s32 core, Member* member) {
+            ASSERT(IsValidCore(core));
+            ASSERT(IsValidPriority(priority));
+
+            if (priority <= LowestPriority) {
+                this->queues[priority].Remove(core, member);
+                this->queues[priority].PushBack(core, member);
+                return this->queues[priority].GetFront(core);
+            } else {
+                return nullptr;
+            }
+        }
+
+    private:
+        std::array<KPerCoreQueue, NumPriority> queues{};
+        std::array<Common::BitSet64<NumPriority>, NumCores> available_priorities{};
+    };
+
+private:
+    KPriorityQueueImpl scheduled_queue;
+    KPriorityQueueImpl suggested_queue;
+
+private:
+    constexpr void ClearAffinityBit(u64& affinity, s32 core) {
+        affinity &= ~(u64(1) << core);
+    }
+
+    constexpr s32 GetNextCore(u64& affinity) {
+        const s32 core = Common::CountTrailingZeroes64(affinity);
+        ClearAffinityBit(affinity, core);
+        return core;
+    }
+
+    constexpr void PushBack(s32 priority, Member* member) {
+        ASSERT(IsValidPriority(priority));
+
+        // Push onto the scheduled queue for its core, if we can.
+        u64 affinity = member->GetAffinityMask().GetAffinityMask();
+        if (const s32 core = member->GetActiveCore(); core >= 0) {
+            this->scheduled_queue.PushBack(priority, core, member);
+            ClearAffinityBit(affinity, core);
+        }
+
+        // And suggest the thread for all other cores.
+        while (affinity) {
+            this->suggested_queue.PushBack(priority, GetNextCore(affinity), member);
+        }
+    }
+
+    constexpr void PushFront(s32 priority, Member* member) {
+        ASSERT(IsValidPriority(priority));
+
+        // Push onto the scheduled queue for its core, if we can.
+        u64 affinity = member->GetAffinityMask().GetAffinityMask();
+        if (const s32 core = member->GetActiveCore(); core >= 0) {
+            this->scheduled_queue.PushFront(priority, core, member);
+            ClearAffinityBit(affinity, core);
+        }
+
+        // And suggest the thread for all other cores.
+        // Note: Nintendo pushes onto the back of the suggested queue, not the front.
+        while (affinity) {
+            this->suggested_queue.PushBack(priority, GetNextCore(affinity), member);
+        }
+    }
+
+    constexpr void Remove(s32 priority, Member* member) {
+        ASSERT(IsValidPriority(priority));
+
+        // Remove from the scheduled queue for its core.
+        u64 affinity = member->GetAffinityMask().GetAffinityMask();
+        if (const s32 core = member->GetActiveCore(); core >= 0) {
+            this->scheduled_queue.Remove(priority, core, member);
+            ClearAffinityBit(affinity, core);
+        }
+
+        // Remove from the suggested queue for all other cores.
+        while (affinity) {
+            this->suggested_queue.Remove(priority, GetNextCore(affinity), member);
+        }
+    }
+
+public:
+    constexpr KPriorityQueue() = default;
+
+    // Getters.
+    constexpr Member* GetScheduledFront(s32 core) const {
+        return this->scheduled_queue.GetFront(core);
+    }
+
+    constexpr Member* GetScheduledFront(s32 core, s32 priority) const {
+        return this->scheduled_queue.GetFront(priority, core);
+    }
+
+    constexpr Member* GetSuggestedFront(s32 core) const {
+        return this->suggested_queue.GetFront(core);
+    }
+
+    constexpr Member* GetSuggestedFront(s32 core, s32 priority) const {
+        return this->suggested_queue.GetFront(priority, core);
+    }
+
+    constexpr Member* GetScheduledNext(s32 core, const Member* member) const {
+        return this->scheduled_queue.GetNext(core, member);
+    }
+
+    constexpr Member* GetSuggestedNext(s32 core, const Member* member) const {
+        return this->suggested_queue.GetNext(core, member);
+    }
+
+    constexpr Member* GetSamePriorityNext(s32 core, const Member* member) const {
+        return member->GetPriorityQueueEntry(core).GetNext();
+    }
+
+    // Mutators.
+    constexpr void PushBack(Member* member) {
+        this->PushBack(member->GetPriority(), member);
+    }
+
+    constexpr void Remove(Member* member) {
+        this->Remove(member->GetPriority(), member);
+    }
+
+    constexpr void MoveToScheduledFront(Member* member) {
+        this->scheduled_queue.MoveToFront(member->GetPriority(), member->GetActiveCore(), member);
+    }
+
+    constexpr Thread* MoveToScheduledBack(Member* member) {
+        return this->scheduled_queue.MoveToBack(member->GetPriority(), member->GetActiveCore(),
+                                                member);
+    }
+
+    // First class fancy operations.
+    constexpr void ChangePriority(s32 prev_priority, bool is_running, Member* member) {
+        ASSERT(IsValidPriority(prev_priority));
+
+        // Remove the member from the queues.
+        const s32 new_priority = member->GetPriority();
+        this->Remove(prev_priority, member);
+
+        // And enqueue. If the member is running, we want to keep it running.
+        if (is_running) {
+            this->PushFront(new_priority, member);
+        } else {
+            this->PushBack(new_priority, member);
+        }
+    }
+
+    constexpr void ChangeAffinityMask(s32 prev_core, const AffinityMaskType& prev_affinity,
+                                      Member* member) {
+        // Get the new information.
+        const s32 priority = member->GetPriority();
+        const AffinityMaskType& new_affinity = member->GetAffinityMask();
+        const s32 new_core = member->GetActiveCore();
+
+        // Remove the member from all queues it was in before.
+        for (s32 core = 0; core < static_cast<s32>(NumCores); core++) {
+            if (prev_affinity.GetAffinity(core)) {
+                if (core == prev_core) {
+                    this->scheduled_queue.Remove(priority, core, member);
+                } else {
+                    this->suggested_queue.Remove(priority, core, member);
+                }
+            }
+        }
+
+        // And add the member to all queues it should be in now.
+        for (s32 core = 0; core < static_cast<s32>(NumCores); core++) {
+            if (new_affinity.GetAffinity(core)) {
+                if (core == new_core) {
+                    this->scheduled_queue.PushBack(priority, core, member);
+                } else {
+                    this->suggested_queue.PushBack(priority, core, member);
+                }
+            }
+        }
+    }
+
+    constexpr void ChangeCore(s32 prev_core, Member* member, bool to_front = false) {
+        // Get the new information.
+        const s32 new_core = member->GetActiveCore();
+        const s32 priority = member->GetPriority();
+
+        // We don't need to do anything if the core is the same.
+        if (prev_core != new_core) {
+            // Remove from the scheduled queue for the previous core.
+            if (prev_core >= 0) {
+                this->scheduled_queue.Remove(priority, prev_core, member);
+            }
+
+            // Remove from the suggested queue and add to the scheduled queue for the new core.
+            if (new_core >= 0) {
+                this->suggested_queue.Remove(priority, new_core, member);
+                if (to_front) {
+                    this->scheduled_queue.PushFront(priority, new_core, member);
+                } else {
+                    this->scheduled_queue.PushBack(priority, new_core, member);
+                }
+            }
+
+            // Add to the suggested queue for the previous core.
+            if (prev_core >= 0) {
+                this->suggested_queue.PushBack(priority, prev_core, member);
+            }
+        }
+    }
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/k_scheduler.cpp b/src/core/hle/kernel/k_scheduler.cpp
new file mode 100644
index 000000000..c5fd82a6b
--- /dev/null
+++ b/src/core/hle/kernel/k_scheduler.cpp
@@ -0,0 +1,784 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#include "common/assert.h"
+#include "common/bit_util.h"
+#include "common/fiber.h"
+#include "common/logging/log.h"
+#include "core/arm/arm_interface.h"
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/cpu_manager.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/physical_core.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+
+namespace Kernel {
+
+static void IncrementScheduledCount(Kernel::Thread* thread) {
+    if (auto process = thread->GetOwnerProcess(); process) {
+        process->IncrementScheduledCount();
+    }
+}
+
+void KScheduler::RescheduleCores(KernelCore& kernel, u64 cores_pending_reschedule,
+                                 Core::EmuThreadHandle global_thread) {
+    u32 current_core = global_thread.host_handle;
+    bool must_context_switch = global_thread.guest_handle != InvalidHandle &&
+                               (current_core < Core::Hardware::NUM_CPU_CORES);
+
+    while (cores_pending_reschedule != 0) {
+        u32 core = Common::CountTrailingZeroes64(cores_pending_reschedule);
+        ASSERT(core < Core::Hardware::NUM_CPU_CORES);
+        if (!must_context_switch || core != current_core) {
+            auto& phys_core = kernel.PhysicalCore(core);
+            phys_core.Interrupt();
+        } else {
+            must_context_switch = true;
+        }
+        cores_pending_reschedule &= ~(1ULL << core);
+    }
+    if (must_context_switch) {
+        auto core_scheduler = kernel.CurrentScheduler();
+        kernel.ExitSVCProfile();
+        core_scheduler->RescheduleCurrentCore();
+        kernel.EnterSVCProfile();
+    }
+}
+
+u64 KScheduler::UpdateHighestPriorityThread(Thread* highest_thread) {
+    std::scoped_lock lock{guard};
+    if (Thread* prev_highest_thread = this->state.highest_priority_thread;
+        prev_highest_thread != highest_thread) {
+        if (prev_highest_thread != nullptr) {
+            IncrementScheduledCount(prev_highest_thread);
+            prev_highest_thread->SetLastScheduledTick(system.CoreTiming().GetCPUTicks());
+        }
+        if (this->state.should_count_idle) {
+            if (highest_thread != nullptr) {
+                // if (Process* process = highest_thread->GetOwnerProcess(); process != nullptr) {
+                //    process->SetRunningThread(this->core_id, highest_thread,
+                //                              this->state.idle_count);
+                //}
+            } else {
+                this->state.idle_count++;
+            }
+        }
+
+        this->state.highest_priority_thread = highest_thread;
+        this->state.needs_scheduling = true;
+        return (1ULL << this->core_id);
+    } else {
+        return 0;
+    }
+}
+
+u64 KScheduler::UpdateHighestPriorityThreadsImpl(KernelCore& kernel) {
+    ASSERT(kernel.GlobalSchedulerContext().IsLocked());
+
+    // Clear that we need to update.
+    ClearSchedulerUpdateNeeded(kernel);
+
+    u64 cores_needing_scheduling = 0, idle_cores = 0;
+    Thread* top_threads[Core::Hardware::NUM_CPU_CORES];
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    /// We want to go over all cores, finding the highest priority thread and determining if
+    /// scheduling is needed for that core.
+    for (size_t core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
+        Thread* top_thread = priority_queue.GetScheduledFront(static_cast<s32>(core_id));
+        if (top_thread != nullptr) {
+            // If the thread has no waiters, we need to check if the process has a thread pinned.
+            // TODO(bunnei): Implement thread pinning
+        } else {
+            idle_cores |= (1ULL << core_id);
+        }
+
+        top_threads[core_id] = top_thread;
+        cores_needing_scheduling |=
+            kernel.Scheduler(core_id).UpdateHighestPriorityThread(top_threads[core_id]);
+    }
+
+    // Idle cores are bad. We're going to try to migrate threads to each idle core in turn.
+    while (idle_cores != 0) {
+        u32 core_id = Common::CountTrailingZeroes64(idle_cores);
+        if (Thread* suggested = priority_queue.GetSuggestedFront(core_id); suggested != nullptr) {
+            s32 migration_candidates[Core::Hardware::NUM_CPU_CORES];
+            size_t num_candidates = 0;
+
+            // While we have a suggested thread, try to migrate it!
+            while (suggested != nullptr) {
+                // Check if the suggested thread is the top thread on its core.
+                const s32 suggested_core = suggested->GetActiveCore();
+                if (Thread* top_thread =
+                        (suggested_core >= 0) ? top_threads[suggested_core] : nullptr;
+                    top_thread != suggested) {
+                    // Make sure we're not dealing with threads too high priority for migration.
+                    if (top_thread != nullptr &&
+                        top_thread->GetPriority() < HighestCoreMigrationAllowedPriority) {
+                        break;
+                    }
+
+                    // The suggested thread isn't bound to its core, so we can migrate it!
+                    suggested->SetActiveCore(core_id);
+                    priority_queue.ChangeCore(suggested_core, suggested);
+
+                    top_threads[core_id] = suggested;
+                    cores_needing_scheduling |=
+                        kernel.Scheduler(core_id).UpdateHighestPriorityThread(top_threads[core_id]);
+                    break;
+                }
+
+                // Note this core as a candidate for migration.
+                ASSERT(num_candidates < Core::Hardware::NUM_CPU_CORES);
+                migration_candidates[num_candidates++] = suggested_core;
+                suggested = priority_queue.GetSuggestedNext(core_id, suggested);
+            }
+
+            // If suggested is nullptr, we failed to migrate a specific thread. So let's try all our
+            // candidate cores' top threads.
+            if (suggested == nullptr) {
+                for (size_t i = 0; i < num_candidates; i++) {
+                    // Check if there's some other thread that can run on the candidate core.
+                    const s32 candidate_core = migration_candidates[i];
+                    suggested = top_threads[candidate_core];
+                    if (Thread* next_on_candidate_core =
+                            priority_queue.GetScheduledNext(candidate_core, suggested);
+                        next_on_candidate_core != nullptr) {
+                        // The candidate core can run some other thread! We'll migrate its current
+                        // top thread to us.
+                        top_threads[candidate_core] = next_on_candidate_core;
+                        cores_needing_scheduling |=
+                            kernel.Scheduler(candidate_core)
+                                .UpdateHighestPriorityThread(top_threads[candidate_core]);
+
+                        // Perform the migration.
+                        suggested->SetActiveCore(core_id);
+                        priority_queue.ChangeCore(candidate_core, suggested);
+
+                        top_threads[core_id] = suggested;
+                        cores_needing_scheduling |=
+                            kernel.Scheduler(core_id).UpdateHighestPriorityThread(
+                                top_threads[core_id]);
+                        break;
+                    }
+                }
+            }
+        }
+
+        idle_cores &= ~(1ULL << core_id);
+    }
+
+    return cores_needing_scheduling;
+}
+
+void KScheduler::OnThreadStateChanged(KernelCore& kernel, Thread* thread, u32 old_state) {
+    ASSERT(kernel.GlobalSchedulerContext().IsLocked());
+
+    // Check if the state has changed, because if it hasn't there's nothing to do.
+    const auto cur_state = thread->scheduling_state;
+    if (cur_state == old_state) {
+        return;
+    }
+
+    // Update the priority queues.
+    if (old_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+        // If we were previously runnable, then we're not runnable now, and we should remove.
+        GetPriorityQueue(kernel).Remove(thread);
+        IncrementScheduledCount(thread);
+        SetSchedulerUpdateNeeded(kernel);
+    } else if (cur_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+        // If we're now runnable, then we weren't previously, and we should add.
+        GetPriorityQueue(kernel).PushBack(thread);
+        IncrementScheduledCount(thread);
+        SetSchedulerUpdateNeeded(kernel);
+    }
+}
+
+void KScheduler::OnThreadPriorityChanged(KernelCore& kernel, Thread* thread, Thread* current_thread,
+                                         u32 old_priority) {
+
+    ASSERT(kernel.GlobalSchedulerContext().IsLocked());
+
+    // If the thread is runnable, we want to change its priority in the queue.
+    if (thread->scheduling_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+        GetPriorityQueue(kernel).ChangePriority(
+            old_priority, thread == kernel.CurrentScheduler()->GetCurrentThread(), thread);
+        IncrementScheduledCount(thread);
+        SetSchedulerUpdateNeeded(kernel);
+    }
+}
+
+void KScheduler::OnThreadAffinityMaskChanged(KernelCore& kernel, Thread* thread,
+                                             const KAffinityMask& old_affinity, s32 old_core) {
+    ASSERT(kernel.GlobalSchedulerContext().IsLocked());
+
+    // If the thread is runnable, we want to change its affinity in the queue.
+    if (thread->scheduling_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+        GetPriorityQueue(kernel).ChangeAffinityMask(old_core, old_affinity, thread);
+        IncrementScheduledCount(thread);
+        SetSchedulerUpdateNeeded(kernel);
+    }
+}
+
+void KScheduler::RotateScheduledQueue(s32 core_id, s32 priority) {
+    ASSERT(system.GlobalSchedulerContext().IsLocked());
+
+    // Get a reference to the priority queue.
+    auto& kernel = system.Kernel();
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    // Rotate the front of the queue to the end.
+    Thread* top_thread = priority_queue.GetScheduledFront(core_id, priority);
+    Thread* next_thread = nullptr;
+    if (top_thread != nullptr) {
+        next_thread = priority_queue.MoveToScheduledBack(top_thread);
+        if (next_thread != top_thread) {
+            IncrementScheduledCount(top_thread);
+            IncrementScheduledCount(next_thread);
+        }
+    }
+
+    // While we have a suggested thread, try to migrate it!
+    {
+        Thread* suggested = priority_queue.GetSuggestedFront(core_id, priority);
+        while (suggested != nullptr) {
+            // Check if the suggested thread is the top thread on its core.
+            const s32 suggested_core = suggested->GetActiveCore();
+            if (Thread* top_on_suggested_core =
+                    (suggested_core >= 0) ? priority_queue.GetScheduledFront(suggested_core)
+                                          : nullptr;
+                top_on_suggested_core != suggested) {
+                // If the next thread is a new thread that has been waiting longer than our
+                // suggestion, we prefer it to our suggestion.
+                if (top_thread != next_thread && next_thread != nullptr &&
+                    next_thread->GetLastScheduledTick() < suggested->GetLastScheduledTick()) {
+                    suggested = nullptr;
+                    break;
+                }
+
+                // If we're allowed to do a migration, do one.
+                // NOTE: Unlike migrations in UpdateHighestPriorityThread, this moves the suggestion
+                // to the front of the queue.
+                if (top_on_suggested_core == nullptr ||
+                    top_on_suggested_core->GetPriority() >= HighestCoreMigrationAllowedPriority) {
+                    suggested->SetActiveCore(core_id);
+                    priority_queue.ChangeCore(suggested_core, suggested, true);
+                    IncrementScheduledCount(suggested);
+                    break;
+                }
+            }
+
+            // Get the next suggestion.
+            suggested = priority_queue.GetSamePriorityNext(core_id, suggested);
+        }
+    }
+
+    // Now that we might have migrated a thread with the same priority, check if we can do better.
+
+    {
+        Thread* best_thread = priority_queue.GetScheduledFront(core_id);
+        if (best_thread == GetCurrentThread()) {
+            best_thread = priority_queue.GetScheduledNext(core_id, best_thread);
+        }
+
+        // If the best thread we can choose has a priority the same or worse than ours, try to
+        // migrate a higher priority thread.
+        if (best_thread != nullptr && best_thread->GetPriority() >= static_cast<u32>(priority)) {
+            Thread* suggested = priority_queue.GetSuggestedFront(core_id);
+            while (suggested != nullptr) {
+                // If the suggestion's priority is the same as ours, don't bother.
+                if (suggested->GetPriority() >= best_thread->GetPriority()) {
+                    break;
+                }
+
+                // Check if the suggested thread is the top thread on its core.
+                const s32 suggested_core = suggested->GetActiveCore();
+                if (Thread* top_on_suggested_core =
+                        (suggested_core >= 0) ? priority_queue.GetScheduledFront(suggested_core)
+                                              : nullptr;
+                    top_on_suggested_core != suggested) {
+                    // If we're allowed to do a migration, do one.
+                    // NOTE: Unlike migrations in UpdateHighestPriorityThread, this moves the
+                    // suggestion to the front of the queue.
+                    if (top_on_suggested_core == nullptr ||
+                        top_on_suggested_core->GetPriority() >=
+                            HighestCoreMigrationAllowedPriority) {
+                        suggested->SetActiveCore(core_id);
+                        priority_queue.ChangeCore(suggested_core, suggested, true);
+                        IncrementScheduledCount(suggested);
+                        break;
+                    }
+                }
+
+                // Get the next suggestion.
+                suggested = priority_queue.GetSuggestedNext(core_id, suggested);
+            }
+        }
+    }
+
+    // After a rotation, we need a scheduler update.
+    SetSchedulerUpdateNeeded(kernel);
+}
+
+bool KScheduler::CanSchedule(KernelCore& kernel) {
+    return kernel.CurrentScheduler()->GetCurrentThread()->GetDisableDispatchCount() <= 1;
+}
+
+bool KScheduler::IsSchedulerUpdateNeeded(const KernelCore& kernel) {
+    return kernel.GlobalSchedulerContext().scheduler_update_needed.load(std::memory_order_acquire);
+}
+
+void KScheduler::SetSchedulerUpdateNeeded(KernelCore& kernel) {
+    kernel.GlobalSchedulerContext().scheduler_update_needed.store(true, std::memory_order_release);
+}
+
+void KScheduler::ClearSchedulerUpdateNeeded(KernelCore& kernel) {
+    kernel.GlobalSchedulerContext().scheduler_update_needed.store(false, std::memory_order_release);
+}
+
+void KScheduler::DisableScheduling(KernelCore& kernel) {
+    if (auto* scheduler = kernel.CurrentScheduler(); scheduler) {
+        ASSERT(scheduler->GetCurrentThread()->GetDisableDispatchCount() >= 0);
+        scheduler->GetCurrentThread()->DisableDispatch();
+    }
+}
+
+void KScheduler::EnableScheduling(KernelCore& kernel, u64 cores_needing_scheduling,
+                                  Core::EmuThreadHandle global_thread) {
+    if (auto* scheduler = kernel.CurrentScheduler(); scheduler) {
+        scheduler->GetCurrentThread()->EnableDispatch();
+    }
+    RescheduleCores(kernel, cores_needing_scheduling, global_thread);
+}
+
+u64 KScheduler::UpdateHighestPriorityThreads(KernelCore& kernel) {
+    if (IsSchedulerUpdateNeeded(kernel)) {
+        return UpdateHighestPriorityThreadsImpl(kernel);
+    } else {
+        return 0;
+    }
+}
+
+KSchedulerPriorityQueue& KScheduler::GetPriorityQueue(KernelCore& kernel) {
+    return kernel.GlobalSchedulerContext().priority_queue;
+}
+
+void KScheduler::YieldWithoutCoreMigration() {
+    auto& kernel = system.Kernel();
+
+    // Validate preconditions.
+    ASSERT(CanSchedule(kernel));
+    ASSERT(kernel.CurrentProcess() != nullptr);
+
+    // Get the current thread and process.
+    Thread& cur_thread = *GetCurrentThread();
+    Process& cur_process = *kernel.CurrentProcess();
+
+    // If the thread's yield count matches, there's nothing for us to do.
+    if (cur_thread.GetYieldScheduleCount() == cur_process.GetScheduledCount()) {
+        return;
+    }
+
+    // Get a reference to the priority queue.
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    // Perform the yield.
+    {
+        KScopedSchedulerLock lock(kernel);
+
+        const auto cur_state = cur_thread.scheduling_state;
+        if (cur_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+            // Put the current thread at the back of the queue.
+            Thread* next_thread = priority_queue.MoveToScheduledBack(std::addressof(cur_thread));
+            IncrementScheduledCount(std::addressof(cur_thread));
+
+            // If the next thread is different, we have an update to perform.
+            if (next_thread != std::addressof(cur_thread)) {
+                SetSchedulerUpdateNeeded(kernel);
+            } else {
+                // Otherwise, set the thread's yield count so that we won't waste work until the
+                // process is scheduled again.
+                cur_thread.SetYieldScheduleCount(cur_process.GetScheduledCount());
+            }
+        }
+    }
+}
+
+void KScheduler::YieldWithCoreMigration() {
+    auto& kernel = system.Kernel();
+
+    // Validate preconditions.
+    ASSERT(CanSchedule(kernel));
+    ASSERT(kernel.CurrentProcess() != nullptr);
+
+    // Get the current thread and process.
+    Thread& cur_thread = *GetCurrentThread();
+    Process& cur_process = *kernel.CurrentProcess();
+
+    // If the thread's yield count matches, there's nothing for us to do.
+    if (cur_thread.GetYieldScheduleCount() == cur_process.GetScheduledCount()) {
+        return;
+    }
+
+    // Get a reference to the priority queue.
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    // Perform the yield.
+    {
+        KScopedSchedulerLock lock(kernel);
+
+        const auto cur_state = cur_thread.scheduling_state;
+        if (cur_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+            // Get the current active core.
+            const s32 core_id = cur_thread.GetActiveCore();
+
+            // Put the current thread at the back of the queue.
+            Thread* next_thread = priority_queue.MoveToScheduledBack(std::addressof(cur_thread));
+            IncrementScheduledCount(std::addressof(cur_thread));
+
+            // While we have a suggested thread, try to migrate it!
+            bool recheck = false;
+            Thread* suggested = priority_queue.GetSuggestedFront(core_id);
+            while (suggested != nullptr) {
+                // Check if the suggested thread is the thread running on its core.
+                const s32 suggested_core = suggested->GetActiveCore();
+
+                if (Thread* running_on_suggested_core =
+                        (suggested_core >= 0)
+                            ? kernel.Scheduler(suggested_core).state.highest_priority_thread
+                            : nullptr;
+                    running_on_suggested_core != suggested) {
+                    // If the current thread's priority is higher than our suggestion's we prefer
+                    // the next thread to the suggestion. We also prefer the next thread when the
+                    // current thread's priority is equal to the suggestions, but the next thread
+                    // has been waiting longer.
+                    if ((suggested->GetPriority() > cur_thread.GetPriority()) ||
+                        (suggested->GetPriority() == cur_thread.GetPriority() &&
+                         next_thread != std::addressof(cur_thread) &&
+                         next_thread->GetLastScheduledTick() < suggested->GetLastScheduledTick())) {
+                        suggested = nullptr;
+                        break;
+                    }
+
+                    // If we're allowed to do a migration, do one.
+                    // NOTE: Unlike migrations in UpdateHighestPriorityThread, this moves the
+                    // suggestion to the front of the queue.
+                    if (running_on_suggested_core == nullptr ||
+                        running_on_suggested_core->GetPriority() >=
+                            HighestCoreMigrationAllowedPriority) {
+                        suggested->SetActiveCore(core_id);
+                        priority_queue.ChangeCore(suggested_core, suggested, true);
+                        IncrementScheduledCount(suggested);
+                        break;
+                    } else {
+                        // We couldn't perform a migration, but we should check again on a future
+                        // yield.
+                        recheck = true;
+                    }
+                }
+
+                // Get the next suggestion.
+                suggested = priority_queue.GetSuggestedNext(core_id, suggested);
+            }
+
+            // If we still have a suggestion or the next thread is different, we have an update to
+            // perform.
+            if (suggested != nullptr || next_thread != std::addressof(cur_thread)) {
+                SetSchedulerUpdateNeeded(kernel);
+            } else if (!recheck) {
+                // Otherwise if we don't need to re-check, set the thread's yield count so that we
+                // won't waste work until the process is scheduled again.
+                cur_thread.SetYieldScheduleCount(cur_process.GetScheduledCount());
+            }
+        }
+    }
+}
+
+void KScheduler::YieldToAnyThread() {
+    auto& kernel = system.Kernel();
+
+    // Validate preconditions.
+    ASSERT(CanSchedule(kernel));
+    ASSERT(kernel.CurrentProcess() != nullptr);
+
+    // Get the current thread and process.
+    Thread& cur_thread = *GetCurrentThread();
+    Process& cur_process = *kernel.CurrentProcess();
+
+    // If the thread's yield count matches, there's nothing for us to do.
+    if (cur_thread.GetYieldScheduleCount() == cur_process.GetScheduledCount()) {
+        return;
+    }
+
+    // Get a reference to the priority queue.
+    auto& priority_queue = GetPriorityQueue(kernel);
+
+    // Perform the yield.
+    {
+        KScopedSchedulerLock lock(kernel);
+
+        const auto cur_state = cur_thread.scheduling_state;
+        if (cur_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
+            // Get the current active core.
+            const s32 core_id = cur_thread.GetActiveCore();
+
+            // Migrate the current thread to core -1.
+            cur_thread.SetActiveCore(-1);
+            priority_queue.ChangeCore(core_id, std::addressof(cur_thread));
+            IncrementScheduledCount(std::addressof(cur_thread));
+
+            // If there's nothing scheduled, we can try to perform a migration.
+            if (priority_queue.GetScheduledFront(core_id) == nullptr) {
+                // While we have a suggested thread, try to migrate it!
+                Thread* suggested = priority_queue.GetSuggestedFront(core_id);
+                while (suggested != nullptr) {
+                    // Check if the suggested thread is the top thread on its core.
+                    const s32 suggested_core = suggested->GetActiveCore();
+                    if (Thread* top_on_suggested_core =
+                            (suggested_core >= 0) ? priority_queue.GetScheduledFront(suggested_core)
+                                                  : nullptr;
+                        top_on_suggested_core != suggested) {
+                        // If we're allowed to do a migration, do one.
+                        if (top_on_suggested_core == nullptr ||
+                            top_on_suggested_core->GetPriority() >=
+                                HighestCoreMigrationAllowedPriority) {
+                            suggested->SetActiveCore(core_id);
+                            priority_queue.ChangeCore(suggested_core, suggested);
+                            IncrementScheduledCount(suggested);
+                        }
+
+                        // Regardless of whether we migrated, we had a candidate, so we're done.
+                        break;
+                    }
+
+                    // Get the next suggestion.
+                    suggested = priority_queue.GetSuggestedNext(core_id, suggested);
+                }
+
+                // If the suggestion is different from the current thread, we need to perform an
+                // update.
+                if (suggested != std::addressof(cur_thread)) {
+                    SetSchedulerUpdateNeeded(kernel);
+                } else {
+                    // Otherwise, set the thread's yield count so that we won't waste work until the
+                    // process is scheduled again.
+                    cur_thread.SetYieldScheduleCount(cur_process.GetScheduledCount());
+                }
+            } else {
+                // Otherwise, we have an update to perform.
+                SetSchedulerUpdateNeeded(kernel);
+            }
+        }
+    }
+}
+
+KScheduler::KScheduler(Core::System& system, std::size_t core_id)
+    : system(system), core_id(core_id) {
+    switch_fiber = std::make_shared<Common::Fiber>(OnSwitch, this);
+    this->state.needs_scheduling = true;
+    this->state.interrupt_task_thread_runnable = false;
+    this->state.should_count_idle = false;
+    this->state.idle_count = 0;
+    this->state.idle_thread_stack = nullptr;
+    this->state.highest_priority_thread = nullptr;
+}
+
+KScheduler::~KScheduler() = default;
+
+Thread* KScheduler::GetCurrentThread() const {
+    if (current_thread) {
+        return current_thread;
+    }
+    return idle_thread;
+}
+
+u64 KScheduler::GetLastContextSwitchTicks() const {
+    return last_context_switch_time;
+}
+
+void KScheduler::RescheduleCurrentCore() {
+    ASSERT(GetCurrentThread()->GetDisableDispatchCount() == 1);
+
+    auto& phys_core = system.Kernel().PhysicalCore(core_id);
+    if (phys_core.IsInterrupted()) {
+        phys_core.ClearInterrupt();
+    }
+    guard.lock();
+    if (this->state.needs_scheduling) {
+        Schedule();
+    } else {
+        guard.unlock();
+    }
+}
+
+void KScheduler::OnThreadStart() {
+    SwitchContextStep2();
+}
+
+void KScheduler::Unload(Thread* thread) {
+    if (thread) {
+        thread->SetIsRunning(false);
+        if (thread->IsContinuousOnSVC() && !thread->IsHLEThread()) {
+            system.ArmInterface(core_id).ExceptionalExit();
+            thread->SetContinuousOnSVC(false);
+        }
+        if (!thread->IsHLEThread() && !thread->HasExited()) {
+            Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
+            cpu_core.SaveContext(thread->GetContext32());
+            cpu_core.SaveContext(thread->GetContext64());
+            // Save the TPIDR_EL0 system register in case it was modified.
+            thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+            cpu_core.ClearExclusiveState();
+        }
+        thread->context_guard.unlock();
+    }
+}
+
+void KScheduler::Reload(Thread* thread) {
+    if (thread) {
+        ASSERT_MSG(thread->GetSchedulingStatus() == ThreadSchedStatus::Runnable,
+                   "Thread must be runnable.");
+
+        // Cancel any outstanding wakeup events for this thread
+        thread->SetIsRunning(true);
+        thread->SetWasRunning(false);
+
+        auto* const thread_owner_process = thread->GetOwnerProcess();
+        if (thread_owner_process != nullptr) {
+            system.Kernel().MakeCurrentProcess(thread_owner_process);
+        }
+        if (!thread->IsHLEThread()) {
+            Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
+            cpu_core.LoadContext(thread->GetContext32());
+            cpu_core.LoadContext(thread->GetContext64());
+            cpu_core.SetTlsAddress(thread->GetTLSAddress());
+            cpu_core.SetTPIDR_EL0(thread->GetTPIDR_EL0());
+            cpu_core.ClearExclusiveState();
+        }
+    }
+}
+
+void KScheduler::SwitchContextStep2() {
+    // Load context of new thread
+    Reload(current_thread);
+
+    RescheduleCurrentCore();
+}
+
+void KScheduler::ScheduleImpl() {
+    Thread* previous_thread = current_thread;
+    current_thread = state.highest_priority_thread;
+
+    this->state.needs_scheduling = false;
+
+    if (current_thread == previous_thread) {
+        guard.unlock();
+        return;
+    }
+
+    Process* const previous_process = system.Kernel().CurrentProcess();
+
+    UpdateLastContextSwitchTime(previous_thread, previous_process);
+
+    // Save context for previous thread
+    Unload(previous_thread);
+
+    std::shared_ptr<Common::Fiber>* old_context;
+    if (previous_thread != nullptr) {
+        old_context = &previous_thread->GetHostContext();
+    } else {
+        old_context = &idle_thread->GetHostContext();
+    }
+    guard.unlock();
+
+    Common::Fiber::YieldTo(*old_context, switch_fiber);
+    /// When a thread wakes up, the scheduler may have changed to other in another core.
+    auto& next_scheduler = *system.Kernel().CurrentScheduler();
+    next_scheduler.SwitchContextStep2();
+}
+
+void KScheduler::OnSwitch(void* this_scheduler) {
+    KScheduler* sched = static_cast<KScheduler*>(this_scheduler);
+    sched->SwitchToCurrent();
+}
+
+void KScheduler::SwitchToCurrent() {
+    while (true) {
+        {
+            std::scoped_lock lock{guard};
+            current_thread = state.highest_priority_thread;
+            this->state.needs_scheduling = false;
+        }
+        const auto is_switch_pending = [this] {
+            std::scoped_lock lock{guard};
+            return state.needs_scheduling.load(std::memory_order_relaxed);
+        };
+        do {
+            if (current_thread != nullptr && !current_thread->IsHLEThread()) {
+                current_thread->context_guard.lock();
+                if (!current_thread->IsRunnable()) {
+                    current_thread->context_guard.unlock();
+                    break;
+                }
+                if (static_cast<u32>(current_thread->GetProcessorID()) != core_id) {
+                    current_thread->context_guard.unlock();
+                    break;
+                }
+            }
+            std::shared_ptr<Common::Fiber>* next_context;
+            if (current_thread != nullptr) {
+                next_context = &current_thread->GetHostContext();
+            } else {
+                next_context = &idle_thread->GetHostContext();
+            }
+            Common::Fiber::YieldTo(switch_fiber, *next_context);
+        } while (!is_switch_pending());
+    }
+}
+
+void KScheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
+    const u64 prev_switch_ticks = last_context_switch_time;
+    const u64 most_recent_switch_ticks = system.CoreTiming().GetCPUTicks();
+    const u64 update_ticks = most_recent_switch_ticks - prev_switch_ticks;
+
+    if (thread != nullptr) {
+        thread->UpdateCPUTimeTicks(update_ticks);
+    }
+
+    if (process != nullptr) {
+        process->UpdateCPUTimeTicks(update_ticks);
+    }
+
+    last_context_switch_time = most_recent_switch_ticks;
+}
+
+void KScheduler::Initialize() {
+    std::string name = "Idle Thread Id:" + std::to_string(core_id);
+    std::function<void(void*)> init_func = Core::CpuManager::GetIdleThreadStartFunc();
+    void* init_func_parameter = system.GetCpuManager().GetStartFuncParamater();
+    ThreadType type = static_cast<ThreadType>(THREADTYPE_KERNEL | THREADTYPE_HLE | THREADTYPE_IDLE);
+    auto thread_res = Thread::Create(system, type, name, 0, 64, 0, static_cast<u32>(core_id), 0,
+                                     nullptr, std::move(init_func), init_func_parameter);
+    idle_thread = thread_res.Unwrap().get();
+
+    {
+        KScopedSchedulerLock lock{system.Kernel()};
+        idle_thread->SetStatus(ThreadStatus::Ready);
+    }
+}
+
+KScopedSchedulerLock::KScopedSchedulerLock(KernelCore& kernel)
+    : KScopedLock(kernel.GlobalSchedulerContext().SchedulerLock()) {}
+
+KScopedSchedulerLock::~KScopedSchedulerLock() = default;
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/k_scheduler.h b/src/core/hle/kernel/k_scheduler.h
new file mode 100644
index 000000000..e84abc84c
--- /dev/null
+++ b/src/core/hle/kernel/k_scheduler.h
@@ -0,0 +1,201 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include <atomic>
+
+#include "common/common_types.h"
+#include "common/spin_lock.h"
+#include "core/hle/kernel/global_scheduler_context.h"
+#include "core/hle/kernel/k_priority_queue.h"
+#include "core/hle/kernel/k_scheduler_lock.h"
+#include "core/hle/kernel/k_scoped_lock.h"
+
+namespace Common {
+class Fiber;
+}
+
+namespace Core {
+class System;
+}
+
+namespace Kernel {
+
+class KernelCore;
+class Process;
+class SchedulerLock;
+class Thread;
+
+class KScheduler final {
+public:
+    explicit KScheduler(Core::System& system, std::size_t core_id);
+    ~KScheduler();
+
+    /// Reschedules to the next available thread (call after current thread is suspended)
+    void RescheduleCurrentCore();
+
+    /// Reschedules cores pending reschedule, to be called on EnableScheduling.
+    static void RescheduleCores(KernelCore& kernel, u64 cores_pending_reschedule,
+                                Core::EmuThreadHandle global_thread);
+
+    /// The next two are for SingleCore Only.
+    /// Unload current thread before preempting core.
+    void Unload(Thread* thread);
+
+    /// Reload current thread after core preemption.
+    void Reload(Thread* thread);
+
+    /// Gets the current running thread
+    [[nodiscard]] Thread* GetCurrentThread() const;
+
+    /// Gets the timestamp for the last context switch in ticks.
+    [[nodiscard]] u64 GetLastContextSwitchTicks() const;
+
+    [[nodiscard]] bool ContextSwitchPending() const {
+        return state.needs_scheduling.load(std::memory_order_relaxed);
+    }
+
+    void Initialize();
+
+    void OnThreadStart();
+
+    [[nodiscard]] std::shared_ptr<Common::Fiber>& ControlContext() {
+        return switch_fiber;
+    }
+
+    [[nodiscard]] const std::shared_ptr<Common::Fiber>& ControlContext() const {
+        return switch_fiber;
+    }
+
+    [[nodiscard]] u64 UpdateHighestPriorityThread(Thread* highest_thread);
+
+    /**
+     * Takes a thread and moves it to the back of the it's priority list.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldWithoutCoreMigration();
+
+    /**
+     * Takes a thread and moves it to the back of the it's priority list.
+     * Afterwards, tries to pick a suggested thread from the suggested queue that has worse time or
+     * a better priority than the next thread in the core.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldWithCoreMigration();
+
+    /**
+     * Takes a thread and moves it out of the scheduling queue.
+     * and into the suggested queue. If no thread can be scheduled afterwards in that core,
+     * a suggested thread is obtained instead.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    void YieldToAnyThread();
+
+    /// Notify the scheduler a thread's status has changed.
+    static void OnThreadStateChanged(KernelCore& kernel, Thread* thread, u32 old_state);
+
+    /// Notify the scheduler a thread's priority has changed.
+    static void OnThreadPriorityChanged(KernelCore& kernel, Thread* thread, Thread* current_thread,
+                                        u32 old_priority);
+
+    /// Notify the scheduler a thread's core and/or affinity mask has changed.
+    static void OnThreadAffinityMaskChanged(KernelCore& kernel, Thread* thread,
+                                            const KAffinityMask& old_affinity, s32 old_core);
+
+    static bool CanSchedule(KernelCore& kernel);
+    static bool IsSchedulerUpdateNeeded(const KernelCore& kernel);
+    static void SetSchedulerUpdateNeeded(KernelCore& kernel);
+    static void ClearSchedulerUpdateNeeded(KernelCore& kernel);
+    static void DisableScheduling(KernelCore& kernel);
+    static void EnableScheduling(KernelCore& kernel, u64 cores_needing_scheduling,
+                                 Core::EmuThreadHandle global_thread);
+    [[nodiscard]] static u64 UpdateHighestPriorityThreads(KernelCore& kernel);
+
+private:
+    friend class GlobalSchedulerContext;
+
+    /**
+     * Takes care of selecting the new scheduled threads in three steps:
+     *
+     * 1. First a thread is selected from the top of the priority queue. If no thread
+     *    is obtained then we move to step two, else we are done.
+     *
+     * 2. Second we try to get a suggested thread that's not assigned to any core or
+     *    that is not the top thread in that core.
+     *
+     * 3. Third is no suggested thread is found, we do a second pass and pick a running
+     *    thread in another core and swap it with its current thread.
+     *
+     * returns the cores needing scheduling.
+     */
+    [[nodiscard]] static u64 UpdateHighestPriorityThreadsImpl(KernelCore& kernel);
+
+    [[nodiscard]] static KSchedulerPriorityQueue& GetPriorityQueue(KernelCore& kernel);
+
+    void RotateScheduledQueue(s32 core_id, s32 priority);
+
+    void Schedule() {
+        ASSERT(GetCurrentThread()->GetDisableDispatchCount() == 1);
+        this->ScheduleImpl();
+    }
+
+    /// Switches the CPU's active thread context to that of the specified thread
+    void ScheduleImpl();
+
+    /// When a thread wakes up, it must run this through it's new scheduler
+    void SwitchContextStep2();
+
+    /**
+     * Called on every context switch to update the internal timestamp
+     * This also updates the running time ticks for the given thread and
+     * process using the following difference:
+     *
+     * ticks += most_recent_ticks - last_context_switch_ticks
+     *
+     * The internal tick timestamp for the scheduler is simply the
+     * most recent tick count retrieved. No special arithmetic is
+     * applied to it.
+     */
+    void UpdateLastContextSwitchTime(Thread* thread, Process* process);
+
+    static void OnSwitch(void* this_scheduler);
+    void SwitchToCurrent();
+
+    Thread* current_thread{};
+    Thread* idle_thread{};
+
+    std::shared_ptr<Common::Fiber> switch_fiber{};
+
+    struct SchedulingState {
+        std::atomic<bool> needs_scheduling;
+        bool interrupt_task_thread_runnable{};
+        bool should_count_idle{};
+        u64 idle_count{};
+        Thread* highest_priority_thread{};
+        void* idle_thread_stack{};
+    };
+
+    SchedulingState state;
+
+    Core::System& system;
+    u64 last_context_switch_time{};
+    const std::size_t core_id;
+
+    Common::SpinLock guard{};
+};
+
+class KScopedSchedulerLock : KScopedLock<GlobalSchedulerContext::LockType> {
+public:
+    explicit KScopedSchedulerLock(KernelCore& kernel);
+    ~KScopedSchedulerLock();
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/k_scheduler_lock.h b/src/core/hle/kernel/k_scheduler_lock.h
new file mode 100644
index 000000000..2f1c1f691
--- /dev/null
+++ b/src/core/hle/kernel/k_scheduler_lock.h
@@ -0,0 +1,75 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/assert.h"
+#include "common/spin_lock.h"
+#include "core/hardware_properties.h"
+#include "core/hle/kernel/kernel.h"
+
+namespace Kernel {
+
+class KernelCore;
+
+template <typename SchedulerType>
+class KAbstractSchedulerLock {
+public:
+    explicit KAbstractSchedulerLock(KernelCore& kernel) : kernel{kernel} {}
+
+    bool IsLockedByCurrentThread() const {
+        return this->owner_thread == kernel.GetCurrentEmuThreadID();
+    }
+
+    void Lock() {
+        if (this->IsLockedByCurrentThread()) {
+            // If we already own the lock, we can just increment the count.
+            ASSERT(this->lock_count > 0);
+            this->lock_count++;
+        } else {
+            // Otherwise, we want to disable scheduling and acquire the spinlock.
+            SchedulerType::DisableScheduling(kernel);
+            this->spin_lock.lock();
+
+            // For debug, ensure that our state is valid.
+            ASSERT(this->lock_count == 0);
+            ASSERT(this->owner_thread == Core::EmuThreadHandle::InvalidHandle());
+
+            // Increment count, take ownership.
+            this->lock_count = 1;
+            this->owner_thread = kernel.GetCurrentEmuThreadID();
+        }
+    }
+
+    void Unlock() {
+        ASSERT(this->IsLockedByCurrentThread());
+        ASSERT(this->lock_count > 0);
+
+        // Release an instance of the lock.
+        if ((--this->lock_count) == 0) {
+            // We're no longer going to hold the lock. Take note of what cores need scheduling.
+            const u64 cores_needing_scheduling =
+                SchedulerType::UpdateHighestPriorityThreads(kernel);
+            Core::EmuThreadHandle leaving_thread = owner_thread;
+
+            // Note that we no longer hold the lock, and unlock the spinlock.
+            this->owner_thread = Core::EmuThreadHandle::InvalidHandle();
+            this->spin_lock.unlock();
+
+            // Enable scheduling, and perform a rescheduling operation.
+            SchedulerType::EnableScheduling(kernel, cores_needing_scheduling, leaving_thread);
+        }
+    }
+
+private:
+    KernelCore& kernel;
+    Common::SpinLock spin_lock{};
+    s32 lock_count{};
+    Core::EmuThreadHandle owner_thread{Core::EmuThreadHandle::InvalidHandle()};
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/k_scoped_lock.h b/src/core/hle/kernel/k_scoped_lock.h
new file mode 100644
index 000000000..d7cc557b2
--- /dev/null
+++ b/src/core/hle/kernel/k_scoped_lock.h
@@ -0,0 +1,41 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Kernel {
+
+template <typename T>
+concept KLockable = !std::is_reference_v<T> && requires(T & t) {
+    { t.Lock() }
+    ->std::same_as<void>;
+    { t.Unlock() }
+    ->std::same_as<void>;
+};
+
+template <typename T>
+requires KLockable<T> class KScopedLock {
+public:
+    explicit KScopedLock(T* l) : lock_ptr(l) {
+        this->lock_ptr->Lock();
+    }
+    explicit KScopedLock(T& l) : KScopedLock(std::addressof(l)) { /* ... */
+    }
+    ~KScopedLock() {
+        this->lock_ptr->Unlock();
+    }
+
+    KScopedLock(const KScopedLock&) = delete;
+    KScopedLock(KScopedLock&&) = delete;
+
+private:
+    T* lock_ptr;
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h b/src/core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h
new file mode 100644
index 000000000..2bb3817fa
--- /dev/null
+++ b/src/core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h
@@ -0,0 +1,50 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+// This file references various implementation details from Atmosphere, an open-source firmware for
+// the Nintendo Switch. Copyright 2018-2020 Atmosphere-NX.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+
+namespace Kernel {
+
+class KScopedSchedulerLockAndSleep {
+public:
+    explicit KScopedSchedulerLockAndSleep(KernelCore& kernel, Handle& event_handle, Thread* t,
+                                          s64 timeout)
+        : kernel(kernel), event_handle(event_handle), thread(t), timeout_tick(timeout) {
+        event_handle = InvalidHandle;
+
+        // Lock the scheduler.
+        kernel.GlobalSchedulerContext().scheduler_lock.Lock();
+    }
+
+    ~KScopedSchedulerLockAndSleep() {
+        // Register the sleep.
+        if (this->timeout_tick > 0) {
+            kernel.TimeManager().ScheduleTimeEvent(event_handle, this->thread, this->timeout_tick);
+        }
+
+        // Unlock the scheduler.
+        kernel.GlobalSchedulerContext().scheduler_lock.Unlock();
+    }
+
+    void CancelSleep() {
+        this->timeout_tick = 0;
+    }
+
+private:
+    KernelCore& kernel;
+    Handle& event_handle;
+    Thread* thread{};
+    s64 timeout_tick{};
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index f2b0fe2fd..e8ece8164 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -7,15 +7,15 @@
 #include <bitset>
 #include <functional>
 #include <memory>
-#include <mutex>
 #include <thread>
-#include <unordered_map>
+#include <unordered_set>
 #include <utility>
 
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/microprofile.h"
 #include "common/thread.h"
+#include "common/thread_worker.h"
 #include "core/arm/arm_interface.h"
 #include "core/arm/cpu_interrupt_handler.h"
 #include "core/arm/exclusive_monitor.h"
@@ -28,6 +28,7 @@
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/memory/memory_layout.h"
 #include "core/hle/kernel/memory/memory_manager.h"
@@ -35,7 +36,7 @@
 #include "core/hle/kernel/physical_core.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
-#include "core/hle/kernel/scheduler.h"
+#include "core/hle/kernel/service_thread.h"
 #include "core/hle/kernel/shared_memory.h"
 #include "core/hle/kernel/synchronization.h"
 #include "core/hle/kernel/thread.h"
@@ -50,17 +51,20 @@ namespace Kernel {
 
 struct KernelCore::Impl {
     explicit Impl(Core::System& system, KernelCore& kernel)
-        : global_scheduler{kernel}, synchronization{system}, time_manager{system},
-          global_handle_table{kernel}, system{system} {}
+        : synchronization{system}, time_manager{system}, global_handle_table{kernel}, system{
+                                                                                          system} {}
 
     void SetMulticore(bool is_multicore) {
         this->is_multicore = is_multicore;
     }
 
     void Initialize(KernelCore& kernel) {
-        Shutdown();
         RegisterHostThread();
 
+        global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
+        service_thread_manager =
+            std::make_unique<Common::ThreadWorker>(1, "yuzu:ServiceThreadManager");
+
         InitializePhysicalCores();
         InitializeSystemResourceLimit(kernel);
         InitializeMemoryLayout();
@@ -69,7 +73,19 @@ struct KernelCore::Impl {
         InitializeSuspendThreads();
     }
 
+    void InitializeCores() {
+        for (auto& core : cores) {
+            core.Initialize(current_process->Is64BitProcess());
+        }
+    }
+
     void Shutdown() {
+        process_list.clear();
+
+        // Ensures all service threads gracefully shutdown
+        service_thread_manager.reset();
+        service_threads.clear();
+
         next_object_id = 0;
         next_kernel_process_id = Process::InitialKIPIDMin;
         next_user_process_id = Process::ProcessIDMin;
@@ -81,41 +97,30 @@ struct KernelCore::Impl {
             }
         }
 
-        for (std::size_t i = 0; i < cores.size(); i++) {
-            cores[i].Shutdown();
-            schedulers[i].reset();
-        }
         cores.clear();
 
-        registered_core_threads.reset();
-
-        process_list.clear();
         current_process = nullptr;
 
         system_resource_limit = nullptr;
 
         global_handle_table.Clear();
-        preemption_event = nullptr;
 
-        global_scheduler.Shutdown();
+        preemption_event = nullptr;
 
         named_ports.clear();
 
-        for (auto& core : cores) {
-            core.Shutdown();
-        }
-        cores.clear();
-
         exclusive_monitor.reset();
-        host_thread_ids.clear();
+
+        // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
+        next_host_thread_id = Core::Hardware::NUM_CPU_CORES;
     }
 
     void InitializePhysicalCores() {
         exclusive_monitor =
             Core::MakeExclusiveMonitor(system.Memory(), Core::Hardware::NUM_CPU_CORES);
         for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
-            schedulers[i] = std::make_unique<Kernel::Scheduler>(system, i);
-            cores.emplace_back(system, i, *schedulers[i], interrupts[i]);
+            schedulers[i] = std::make_unique<Kernel::KScheduler>(system, i);
+            cores.emplace_back(i, system, *schedulers[i], interrupts);
         }
     }
 
@@ -147,8 +152,8 @@ struct KernelCore::Impl {
         preemption_event = Core::Timing::CreateEvent(
             "PreemptionCallback", [this, &kernel](std::uintptr_t, std::chrono::nanoseconds) {
                 {
-                    SchedulerLock lock(kernel);
-                    global_scheduler.PreemptThreads();
+                    KScopedSchedulerLock lock(kernel);
+                    global_scheduler_context->PreemptThreads();
                 }
                 const auto time_interval = std::chrono::nanoseconds{
                     Core::Timing::msToCycles(std::chrono::milliseconds(10))};
@@ -177,63 +182,62 @@ struct KernelCore::Impl {
 
     void MakeCurrentProcess(Process* process) {
         current_process = process;
-
         if (process == nullptr) {
             return;
         }
 
-        u32 core_id = GetCurrentHostThreadID();
+        const u32 core_id = GetCurrentHostThreadID();
         if (core_id < Core::Hardware::NUM_CPU_CORES) {
             system.Memory().SetCurrentPageTable(*process, core_id);
         }
     }
 
+    /// Creates a new host thread ID, should only be called by GetHostThreadId
+    u32 AllocateHostThreadId(std::optional<std::size_t> core_id) {
+        if (core_id) {
+            // The first for slots are reserved for CPU core threads
+            ASSERT(*core_id < Core::Hardware::NUM_CPU_CORES);
+            return static_cast<u32>(*core_id);
+        } else {
+            return next_host_thread_id++;
+        }
+    }
+
+    /// Gets the host thread ID for the caller, allocating a new one if this is the first time
+    u32 GetHostThreadId(std::optional<std::size_t> core_id = std::nullopt) {
+        const thread_local auto host_thread_id{AllocateHostThreadId(core_id)};
+        return host_thread_id;
+    }
+
+    /// Registers a CPU core thread by allocating a host thread ID for it
     void RegisterCoreThread(std::size_t core_id) {
-        std::unique_lock lock{register_thread_mutex};
+        ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+        const auto this_id = GetHostThreadId(core_id);
         if (!is_multicore) {
-            single_core_thread_id = std::this_thread::get_id();
+            single_core_thread_id = this_id;
         }
-        const std::thread::id this_id = std::this_thread::get_id();
-        const auto it = host_thread_ids.find(this_id);
-        ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
-        ASSERT(it == host_thread_ids.end());
-        ASSERT(!registered_core_threads[core_id]);
-        host_thread_ids[this_id] = static_cast<u32>(core_id);
-        registered_core_threads.set(core_id);
     }
 
+    /// Registers a new host thread by allocating a host thread ID for it
     void RegisterHostThread() {
-        std::unique_lock lock{register_thread_mutex};
-        const std::thread::id this_id = std::this_thread::get_id();
-        const auto it = host_thread_ids.find(this_id);
-        if (it != host_thread_ids.end()) {
-            return;
-        }
-        host_thread_ids[this_id] = registered_thread_ids++;
+        [[maybe_unused]] const auto this_id = GetHostThreadId();
     }
 
-    u32 GetCurrentHostThreadID() const {
-        const std::thread::id this_id = std::this_thread::get_id();
-        if (!is_multicore) {
-            if (single_core_thread_id == this_id) {
-                return static_cast<u32>(system.GetCpuManager().CurrentCore());
-            }
-        }
-        std::unique_lock lock{register_thread_mutex};
-        const auto it = host_thread_ids.find(this_id);
-        if (it == host_thread_ids.end()) {
-            return Core::INVALID_HOST_THREAD_ID;
+    [[nodiscard]] u32 GetCurrentHostThreadID() {
+        const auto this_id = GetHostThreadId();
+        if (!is_multicore && single_core_thread_id == this_id) {
+            return static_cast<u32>(system.GetCpuManager().CurrentCore());
         }
-        return it->second;
+        return this_id;
     }
 
-    Core::EmuThreadHandle GetCurrentEmuThreadID() const {
+    [[nodiscard]] Core::EmuThreadHandle GetCurrentEmuThreadID() {
         Core::EmuThreadHandle result = Core::EmuThreadHandle::InvalidHandle();
         result.host_handle = GetCurrentHostThreadID();
         if (result.host_handle >= Core::Hardware::NUM_CPU_CORES) {
             return result;
         }
-        const Kernel::Scheduler& sched = cores[result.host_handle].Scheduler();
+        const Kernel::KScheduler& sched = cores[result.host_handle].Scheduler();
         const Kernel::Thread* current = sched.GetCurrentThread();
         if (current != nullptr && !current->IsPhantomMode()) {
             result.guest_handle = current->GetGlobalHandle();
@@ -302,7 +306,7 @@ struct KernelCore::Impl {
     // Lists all processes that exist in the current session.
     std::vector<std::shared_ptr<Process>> process_list;
     Process* current_process = nullptr;
-    Kernel::GlobalScheduler global_scheduler;
+    std::unique_ptr<Kernel::GlobalSchedulerContext> global_scheduler_context;
     Kernel::Synchronization synchronization;
     Kernel::TimeManager time_manager;
 
@@ -321,11 +325,8 @@ struct KernelCore::Impl {
     std::unique_ptr<Core::ExclusiveMonitor> exclusive_monitor;
     std::vector<Kernel::PhysicalCore> cores;
 
-    // 0-3 IDs represent core threads, >3 represent others
-    std::unordered_map<std::thread::id, u32> host_thread_ids;
-    u32 registered_thread_ids{Core::Hardware::NUM_CPU_CORES};
-    std::bitset<Core::Hardware::NUM_CPU_CORES> registered_core_threads;
-    mutable std::mutex register_thread_mutex;
+    // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
+    std::atomic<u32> next_host_thread_id{Core::Hardware::NUM_CPU_CORES};
 
     // Kernel memory management
     std::unique_ptr<Memory::MemoryManager> memory_manager;
@@ -337,12 +338,19 @@ struct KernelCore::Impl {
     std::shared_ptr<Kernel::SharedMemory> irs_shared_mem;
     std::shared_ptr<Kernel::SharedMemory> time_shared_mem;
 
+    // Threads used for services
+    std::unordered_set<std::shared_ptr<Kernel::ServiceThread>> service_threads;
+
+    // Service threads are managed by a worker thread, so that a calling service thread can queue up
+    // the release of itself
+    std::unique_ptr<Common::ThreadWorker> service_thread_manager;
+
     std::array<std::shared_ptr<Thread>, Core::Hardware::NUM_CPU_CORES> suspend_threads{};
     std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES> interrupts{};
-    std::array<std::unique_ptr<Kernel::Scheduler>, Core::Hardware::NUM_CPU_CORES> schedulers{};
+    std::array<std::unique_ptr<Kernel::KScheduler>, Core::Hardware::NUM_CPU_CORES> schedulers{};
 
     bool is_multicore{};
-    std::thread::id single_core_thread_id{};
+    u32 single_core_thread_id{};
 
     std::array<u64, Core::Hardware::NUM_CPU_CORES> svc_ticks{};
 
@@ -363,6 +371,10 @@ void KernelCore::Initialize() {
     impl->Initialize(*this);
 }
 
+void KernelCore::InitializeCores() {
+    impl->InitializeCores();
+}
+
 void KernelCore::Shutdown() {
     impl->Shutdown();
 }
@@ -395,19 +407,19 @@ const std::vector<std::shared_ptr<Process>>& KernelCore::GetProcessList() const
     return impl->process_list;
 }
 
-Kernel::GlobalScheduler& KernelCore::GlobalScheduler() {
-    return impl->global_scheduler;
+Kernel::GlobalSchedulerContext& KernelCore::GlobalSchedulerContext() {
+    return *impl->global_scheduler_context;
 }
 
-const Kernel::GlobalScheduler& KernelCore::GlobalScheduler() const {
-    return impl->global_scheduler;
+const Kernel::GlobalSchedulerContext& KernelCore::GlobalSchedulerContext() const {
+    return *impl->global_scheduler_context;
 }
 
-Kernel::Scheduler& KernelCore::Scheduler(std::size_t id) {
+Kernel::KScheduler& KernelCore::Scheduler(std::size_t id) {
     return *impl->schedulers[id];
 }
 
-const Kernel::Scheduler& KernelCore::Scheduler(std::size_t id) const {
+const Kernel::KScheduler& KernelCore::Scheduler(std::size_t id) const {
     return *impl->schedulers[id];
 }
 
@@ -431,16 +443,13 @@ const Kernel::PhysicalCore& KernelCore::CurrentPhysicalCore() const {
     return impl->cores[core_id];
 }
 
-Kernel::Scheduler& KernelCore::CurrentScheduler() {
+Kernel::KScheduler* KernelCore::CurrentScheduler() {
     u32 core_id = impl->GetCurrentHostThreadID();
-    ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
-    return *impl->schedulers[core_id];
-}
-
-const Kernel::Scheduler& KernelCore::CurrentScheduler() const {
-    u32 core_id = impl->GetCurrentHostThreadID();
-    ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
-    return *impl->schedulers[core_id];
+    if (core_id >= Core::Hardware::NUM_CPU_CORES) {
+        // This is expected when called from not a guest thread
+        return {};
+    }
+    return impl->schedulers[core_id].get();
 }
 
 std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES>& KernelCore::Interrupts() {
@@ -477,12 +486,17 @@ const Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() const {
 }
 
 void KernelCore::InvalidateAllInstructionCaches() {
-    auto& threads = GlobalScheduler().GetThreadList();
-    for (auto& thread : threads) {
-        if (!thread->IsHLEThread()) {
-            auto& arm_interface = thread->ArmInterface();
-            arm_interface.ClearInstructionCache();
+    for (auto& physical_core : impl->cores) {
+        physical_core.ArmInterface().ClearInstructionCache();
+    }
+}
+
+void KernelCore::InvalidateCpuInstructionCacheRange(VAddr addr, std::size_t size) {
+    for (auto& physical_core : impl->cores) {
+        if (!physical_core.IsInitialized()) {
+            continue;
         }
+        physical_core.ArmInterface().InvalidateCacheRange(addr, size);
     }
 }
 
@@ -598,7 +612,7 @@ const Kernel::SharedMemory& KernelCore::GetTimeSharedMem() const {
 void KernelCore::Suspend(bool in_suspention) {
     const bool should_suspend = exception_exited || in_suspention;
     {
-        SchedulerLock lock(*this);
+        KScopedSchedulerLock lock(*this);
         ThreadStatus status = should_suspend ? ThreadStatus::Ready : ThreadStatus::WaitSleep;
         for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             impl->suspend_threads[i]->SetStatus(status);
@@ -625,4 +639,19 @@ void KernelCore::ExitSVCProfile() {
     MicroProfileLeave(MICROPROFILE_TOKEN(Kernel_SVC), impl->svc_ticks[core]);
 }
 
+std::weak_ptr<Kernel::ServiceThread> KernelCore::CreateServiceThread(const std::string& name) {
+    auto service_thread = std::make_shared<Kernel::ServiceThread>(*this, 1, name);
+    impl->service_thread_manager->QueueWork(
+        [this, service_thread] { impl->service_threads.emplace(service_thread); });
+    return service_thread;
+}
+
+void KernelCore::ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread) {
+    impl->service_thread_manager->QueueWork([this, service_thread] {
+        if (auto strong_ptr = service_thread.lock()) {
+            impl->service_threads.erase(strong_ptr);
+        }
+    });
+}
+
 } // namespace Kernel
diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h
index 16285c3f0..e3169f5a7 100644
--- a/src/core/hle/kernel/kernel.h
+++ b/src/core/hle/kernel/kernel.h
@@ -35,13 +35,14 @@ class SlabHeap;
 
 class AddressArbiter;
 class ClientPort;
-class GlobalScheduler;
+class GlobalSchedulerContext;
 class HandleTable;
 class PhysicalCore;
 class Process;
 class ResourceLimit;
-class Scheduler;
+class KScheduler;
 class SharedMemory;
+class ServiceThread;
 class Synchronization;
 class Thread;
 class TimeManager;
@@ -74,6 +75,9 @@ public:
     /// Resets the kernel to a clean slate for use.
     void Initialize();
 
+    /// Initializes the CPU cores.
+    void InitializeCores();
+
     /// Clears all resources in use by the kernel instance.
     void Shutdown();
 
@@ -99,16 +103,16 @@ public:
     const std::vector<std::shared_ptr<Process>>& GetProcessList() const;
 
     /// Gets the sole instance of the global scheduler
-    Kernel::GlobalScheduler& GlobalScheduler();
+    Kernel::GlobalSchedulerContext& GlobalSchedulerContext();
 
     /// Gets the sole instance of the global scheduler
-    const Kernel::GlobalScheduler& GlobalScheduler() const;
+    const Kernel::GlobalSchedulerContext& GlobalSchedulerContext() const;
 
     /// Gets the sole instance of the Scheduler assoviated with cpu core 'id'
-    Kernel::Scheduler& Scheduler(std::size_t id);
+    Kernel::KScheduler& Scheduler(std::size_t id);
 
     /// Gets the sole instance of the Scheduler assoviated with cpu core 'id'
-    const Kernel::Scheduler& Scheduler(std::size_t id) const;
+    const Kernel::KScheduler& Scheduler(std::size_t id) const;
 
     /// Gets the an instance of the respective physical CPU core.
     Kernel::PhysicalCore& PhysicalCore(std::size_t id);
@@ -117,10 +121,7 @@ public:
     const Kernel::PhysicalCore& PhysicalCore(std::size_t id) const;
 
     /// Gets the sole instance of the Scheduler at the current running core.
-    Kernel::Scheduler& CurrentScheduler();
-
-    /// Gets the sole instance of the Scheduler at the current running core.
-    const Kernel::Scheduler& CurrentScheduler() const;
+    Kernel::KScheduler* CurrentScheduler();
 
     /// Gets the an instance of the current physical CPU core.
     Kernel::PhysicalCore& CurrentPhysicalCore();
@@ -153,6 +154,8 @@ public:
 
     void InvalidateAllInstructionCaches();
 
+    void InvalidateCpuInstructionCacheRange(VAddr addr, std::size_t size);
+
     /// Adds a port to the named port table
     void AddNamedPort(std::string name, std::shared_ptr<ClientPort> port);
 
@@ -225,6 +228,22 @@ public:
 
     void ExitSVCProfile();
 
+    /**
+     * Creates an HLE service thread, which are used to execute service routines asynchronously.
+     * While these are allocated per ServerSession, these need to be owned and managed outside of
+     * ServerSession to avoid a circular dependency.
+     * @param name String name for the ServerSession creating this thread, used for debug purposes.
+     * @returns The a weak pointer newly created service thread.
+     */
+    std::weak_ptr<Kernel::ServiceThread> CreateServiceThread(const std::string& name);
+
+    /**
+     * Releases a HLE service thread, instructing KernelCore to free it. This should be called when
+     * the ServerSession associated with the thread is destroyed.
+     * @param service_thread Service thread to release.
+     */
+    void ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread);
+
 private:
     friend class Object;
     friend class Process;
diff --git a/src/core/hle/kernel/memory/address_space_info.cpp b/src/core/hle/kernel/memory/address_space_info.cpp
index e4288cab4..6cf43ba24 100644
--- a/src/core/hle/kernel/memory/address_space_info.cpp
+++ b/src/core/hle/kernel/memory/address_space_info.cpp
@@ -96,6 +96,7 @@ u64 AddressSpaceInfo::GetAddressSpaceStart(std::size_t width, Type type) {
         return AddressSpaceInfos[AddressSpaceIndices39Bit[index]].address;
     }
     UNREACHABLE();
+    return 0;
 }
 
 std::size_t AddressSpaceInfo::GetAddressSpaceSize(std::size_t width, Type type) {
@@ -112,6 +113,7 @@ std::size_t AddressSpaceInfo::GetAddressSpaceSize(std::size_t width, Type type)
         return AddressSpaceInfos[AddressSpaceIndices39Bit[index]].size;
     }
     UNREACHABLE();
+    return 0;
 }
 
 } // namespace Kernel::Memory
diff --git a/src/core/hle/kernel/memory/memory_block.h b/src/core/hle/kernel/memory/memory_block.h
index 9d7839d08..83acece1e 100644
--- a/src/core/hle/kernel/memory/memory_block.h
+++ b/src/core/hle/kernel/memory/memory_block.h
@@ -73,12 +73,12 @@ enum class MemoryState : u32 {
     ThreadLocal =
         static_cast<u32>(Svc::MemoryState::ThreadLocal) | FlagMapped | FlagReferenceCounted,
 
-    Transfered = static_cast<u32>(Svc::MemoryState::Transfered) | FlagsMisc |
-                 FlagCanAlignedDeviceMap | FlagCanChangeAttribute | FlagCanUseIpc |
-                 FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+    Transferred = static_cast<u32>(Svc::MemoryState::Transferred) | FlagsMisc |
+                  FlagCanAlignedDeviceMap | FlagCanChangeAttribute | FlagCanUseIpc |
+                  FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
 
-    SharedTransfered = static_cast<u32>(Svc::MemoryState::SharedTransfered) | FlagsMisc |
-                       FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+    SharedTransferred = static_cast<u32>(Svc::MemoryState::SharedTransferred) | FlagsMisc |
+                        FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
 
     SharedCode = static_cast<u32>(Svc::MemoryState::SharedCode) | FlagMapped |
                  FlagReferenceCounted | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
@@ -111,8 +111,8 @@ static_assert(static_cast<u32>(MemoryState::AliasCodeData) == 0x03FFBD09);
 static_assert(static_cast<u32>(MemoryState::Ipc) == 0x005C3C0A);
 static_assert(static_cast<u32>(MemoryState::Stack) == 0x005C3C0B);
 static_assert(static_cast<u32>(MemoryState::ThreadLocal) == 0x0040200C);
-static_assert(static_cast<u32>(MemoryState::Transfered) == 0x015C3C0D);
-static_assert(static_cast<u32>(MemoryState::SharedTransfered) == 0x005C380E);
+static_assert(static_cast<u32>(MemoryState::Transferred) == 0x015C3C0D);
+static_assert(static_cast<u32>(MemoryState::SharedTransferred) == 0x005C380E);
 static_assert(static_cast<u32>(MemoryState::SharedCode) == 0x0040380F);
 static_assert(static_cast<u32>(MemoryState::Inaccessible) == 0x00000010);
 static_assert(static_cast<u32>(MemoryState::NonSecureIpc) == 0x005C3811);
@@ -222,9 +222,9 @@ public:
 
 public:
     constexpr MemoryBlock() = default;
-    constexpr MemoryBlock(VAddr addr, std::size_t num_pages, MemoryState state,
-                          MemoryPermission perm, MemoryAttribute attribute)
-        : addr{addr}, num_pages(num_pages), state{state}, perm{perm}, attribute{attribute} {}
+    constexpr MemoryBlock(VAddr addr_, std::size_t num_pages_, MemoryState state_,
+                          MemoryPermission perm_, MemoryAttribute attribute_)
+        : addr{addr_}, num_pages(num_pages_), state{state_}, perm{perm_}, attribute{attribute_} {}
 
     constexpr VAddr GetAddress() const {
         return addr;
diff --git a/src/core/hle/kernel/memory/memory_block_manager.h b/src/core/hle/kernel/memory/memory_block_manager.h
index 6e1d41075..f57d1bbcc 100644
--- a/src/core/hle/kernel/memory/memory_block_manager.h
+++ b/src/core/hle/kernel/memory/memory_block_manager.h
@@ -57,8 +57,8 @@ public:
 private:
     void MergeAdjacent(iterator it, iterator& next_it);
 
-    const VAddr start_addr;
-    const VAddr end_addr;
+    [[maybe_unused]] const VAddr start_addr;
+    [[maybe_unused]] const VAddr end_addr;
 
     MemoryBlockTree memory_block_tree;
 };
diff --git a/src/core/hle/kernel/memory/page_table.cpp b/src/core/hle/kernel/memory/page_table.cpp
index a3fadb533..080886554 100644
--- a/src/core/hle/kernel/memory/page_table.cpp
+++ b/src/core/hle/kernel/memory/page_table.cpp
@@ -265,7 +265,7 @@ ResultCode PageTable::InitializeForProcess(FileSys::ProgramAddressSpaceType as_t
     physical_memory_usage = 0;
     memory_pool = pool;
 
-    page_table_impl.Resize(address_space_width, PageBits, true);
+    page_table_impl.Resize(address_space_width, PageBits);
 
     return InitializeMemoryLayout(start, end);
 }
@@ -670,6 +670,11 @@ ResultCode PageTable::SetCodeMemoryPermission(VAddr addr, std::size_t size, Memo
         return RESULT_SUCCESS;
     }
 
+    if ((prev_perm & MemoryPermission::Execute) != (perm & MemoryPermission::Execute)) {
+        // Memory execution state is changing, invalidate CPU cache range
+        system.InvalidateCpuInstructionCacheRange(addr, size);
+    }
+
     const std::size_t num_pages{size / PageSize};
     const OperationType operation{(perm & MemoryPermission::Execute) != MemoryPermission::None
                                       ? OperationType::ChangePermissionsAndRefresh
@@ -1002,8 +1007,8 @@ constexpr VAddr PageTable::GetRegionAddress(MemoryState state) const {
     case MemoryState::Shared:
     case MemoryState::AliasCode:
     case MemoryState::AliasCodeData:
-    case MemoryState::Transfered:
-    case MemoryState::SharedTransfered:
+    case MemoryState::Transferred:
+    case MemoryState::SharedTransferred:
     case MemoryState::SharedCode:
     case MemoryState::GeneratedCode:
     case MemoryState::CodeOut:
@@ -1037,8 +1042,8 @@ constexpr std::size_t PageTable::GetRegionSize(MemoryState state) const {
     case MemoryState::Shared:
     case MemoryState::AliasCode:
     case MemoryState::AliasCodeData:
-    case MemoryState::Transfered:
-    case MemoryState::SharedTransfered:
+    case MemoryState::Transferred:
+    case MemoryState::SharedTransferred:
     case MemoryState::SharedCode:
     case MemoryState::GeneratedCode:
     case MemoryState::CodeOut:
@@ -1075,8 +1080,8 @@ constexpr bool PageTable::CanContain(VAddr addr, std::size_t size, MemoryState s
     case MemoryState::AliasCodeData:
     case MemoryState::Stack:
     case MemoryState::ThreadLocal:
-    case MemoryState::Transfered:
-    case MemoryState::SharedTransfered:
+    case MemoryState::Transferred:
+    case MemoryState::SharedTransferred:
     case MemoryState::SharedCode:
     case MemoryState::GeneratedCode:
     case MemoryState::CodeOut:
diff --git a/src/core/hle/kernel/mutex.cpp b/src/core/hle/kernel/mutex.cpp
index 8f6c944d1..4f8075e0e 100644
--- a/src/core/hle/kernel/mutex.cpp
+++ b/src/core/hle/kernel/mutex.cpp
@@ -11,11 +11,11 @@
 #include "core/core.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/mutex.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/result.h"
 #include "core/memory.h"
@@ -73,9 +73,9 @@ ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
 
     auto& kernel = system.Kernel();
     std::shared_ptr<Thread> current_thread =
-        SharedFrom(kernel.CurrentScheduler().GetCurrentThread());
+        SharedFrom(kernel.CurrentScheduler()->GetCurrentThread());
     {
-        SchedulerLock lock(kernel);
+        KScopedSchedulerLock lock(kernel);
         // The mutex address must be 4-byte aligned
         if ((address % sizeof(u32)) != 0) {
             return ERR_INVALID_ADDRESS;
@@ -114,7 +114,7 @@ ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
     }
 
     {
-        SchedulerLock lock(kernel);
+        KScopedSchedulerLock lock(kernel);
         auto* owner = current_thread->GetLockOwner();
         if (owner != nullptr) {
             owner->RemoveMutexWaiter(current_thread);
@@ -153,10 +153,10 @@ std::pair<ResultCode, std::shared_ptr<Thread>> Mutex::Unlock(std::shared_ptr<Thr
 
 ResultCode Mutex::Release(VAddr address) {
     auto& kernel = system.Kernel();
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
 
     std::shared_ptr<Thread> current_thread =
-        SharedFrom(kernel.CurrentScheduler().GetCurrentThread());
+        SharedFrom(kernel.CurrentScheduler()->GetCurrentThread());
 
     auto [result, new_owner] = Unlock(current_thread, address);
 
diff --git a/src/core/hle/kernel/physical_core.cpp b/src/core/hle/kernel/physical_core.cpp
index c6bbdb080..7fea45f96 100644
--- a/src/core/hle/kernel/physical_core.cpp
+++ b/src/core/hle/kernel/physical_core.cpp
@@ -2,54 +2,60 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "common/assert.h"
-#include "common/logging/log.h"
 #include "common/spin_lock.h"
-#include "core/arm/arm_interface.h"
-#ifdef ARCHITECTURE_x86_64
+#include "core/arm/cpu_interrupt_handler.h"
 #include "core/arm/dynarmic/arm_dynarmic_32.h"
 #include "core/arm/dynarmic/arm_dynarmic_64.h"
-#endif
-#include "core/arm/cpu_interrupt_handler.h"
-#include "core/arm/exclusive_monitor.h"
-#include "core/arm/unicorn/arm_unicorn.h"
 #include "core/core.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/physical_core.h"
-#include "core/hle/kernel/scheduler.h"
-#include "core/hle/kernel/thread.h"
 
 namespace Kernel {
 
-PhysicalCore::PhysicalCore(Core::System& system, std::size_t id, Kernel::Scheduler& scheduler,
-                           Core::CPUInterruptHandler& interrupt_handler)
-    : interrupt_handler{interrupt_handler}, core_index{id}, scheduler{scheduler} {
-
-    guard = std::make_unique<Common::SpinLock>();
-}
+PhysicalCore::PhysicalCore(std::size_t core_index, Core::System& system,
+                           Kernel::KScheduler& scheduler, Core::CPUInterrupts& interrupts)
+    : core_index{core_index}, system{system}, scheduler{scheduler},
+      interrupts{interrupts}, guard{std::make_unique<Common::SpinLock>()} {}
 
 PhysicalCore::~PhysicalCore() = default;
 
-void PhysicalCore::Idle() {
-    interrupt_handler.AwaitInterrupt();
+void PhysicalCore::Initialize([[maybe_unused]] bool is_64_bit) {
+#ifdef ARCHITECTURE_x86_64
+    auto& kernel = system.Kernel();
+    if (is_64_bit) {
+        arm_interface = std::make_unique<Core::ARM_Dynarmic_64>(
+            system, interrupts, kernel.IsMulticore(), kernel.GetExclusiveMonitor(), core_index);
+    } else {
+        arm_interface = std::make_unique<Core::ARM_Dynarmic_32>(
+            system, interrupts, kernel.IsMulticore(), kernel.GetExclusiveMonitor(), core_index);
+    }
+#else
+#error Platform not supported yet.
+#endif
 }
 
-void PhysicalCore::Shutdown() {
-    scheduler.Shutdown();
+void PhysicalCore::Run() {
+    arm_interface->Run();
+}
+
+void PhysicalCore::Idle() {
+    interrupts[core_index].AwaitInterrupt();
 }
 
 bool PhysicalCore::IsInterrupted() const {
-    return interrupt_handler.IsInterrupted();
+    return interrupts[core_index].IsInterrupted();
 }
 
 void PhysicalCore::Interrupt() {
     guard->lock();
-    interrupt_handler.SetInterrupt(true);
+    interrupts[core_index].SetInterrupt(true);
     guard->unlock();
 }
 
 void PhysicalCore::ClearInterrupt() {
     guard->lock();
-    interrupt_handler.SetInterrupt(false);
+    interrupts[core_index].SetInterrupt(false);
     guard->unlock();
 }
 
diff --git a/src/core/hle/kernel/physical_core.h b/src/core/hle/kernel/physical_core.h
index d7a7a951c..f2b0911aa 100644
--- a/src/core/hle/kernel/physical_core.h
+++ b/src/core/hle/kernel/physical_core.h
@@ -4,19 +4,21 @@
 
 #pragma once
 
+#include <array>
 #include <cstddef>
 #include <memory>
 
+#include "core/arm/arm_interface.h"
+
 namespace Common {
 class SpinLock;
 }
 
 namespace Kernel {
-class Scheduler;
+class KScheduler;
 } // namespace Kernel
 
 namespace Core {
-class ARM_Interface;
 class CPUInterruptHandler;
 class ExclusiveMonitor;
 class System;
@@ -26,17 +28,24 @@ namespace Kernel {
 
 class PhysicalCore {
 public:
-    PhysicalCore(Core::System& system, std::size_t id, Kernel::Scheduler& scheduler,
-                 Core::CPUInterruptHandler& interrupt_handler);
+    PhysicalCore(std::size_t core_index, Core::System& system, Kernel::KScheduler& scheduler,
+                 Core::CPUInterrupts& interrupts);
     ~PhysicalCore();
 
     PhysicalCore(const PhysicalCore&) = delete;
     PhysicalCore& operator=(const PhysicalCore&) = delete;
 
     PhysicalCore(PhysicalCore&&) = default;
-    PhysicalCore& operator=(PhysicalCore&&) = default;
+    PhysicalCore& operator=(PhysicalCore&&) = delete;
+
+    /// Initialize the core for the specified parameters.
+    void Initialize(bool is_64_bit);
+
+    /// Execute current jit state
+    void Run();
 
     void Idle();
+
     /// Interrupt this physical core.
     void Interrupt();
 
@@ -46,8 +55,17 @@ public:
     /// Check if this core is interrupted
     bool IsInterrupted() const;
 
-    // Shutdown this physical core.
-    void Shutdown();
+    bool IsInitialized() const {
+        return arm_interface != nullptr;
+    }
+
+    Core::ARM_Interface& ArmInterface() {
+        return *arm_interface;
+    }
+
+    const Core::ARM_Interface& ArmInterface() const {
+        return *arm_interface;
+    }
 
     bool IsMainCore() const {
         return core_index == 0;
@@ -61,19 +79,21 @@ public:
         return core_index;
     }
 
-    Kernel::Scheduler& Scheduler() {
+    Kernel::KScheduler& Scheduler() {
         return scheduler;
     }
 
-    const Kernel::Scheduler& Scheduler() const {
+    const Kernel::KScheduler& Scheduler() const {
         return scheduler;
     }
 
 private:
-    Core::CPUInterruptHandler& interrupt_handler;
-    std::size_t core_index;
-    Kernel::Scheduler& scheduler;
+    const std::size_t core_index;
+    Core::System& system;
+    Kernel::KScheduler& scheduler;
+    Core::CPUInterrupts& interrupts;
     std::unique_ptr<Common::SpinLock> guard;
+    std::unique_ptr<Core::ARM_Interface> arm_interface;
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index ff9d9248b..b905b486a 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -4,6 +4,7 @@
 
 #include <algorithm>
 #include <bitset>
+#include <ctime>
 #include <memory>
 #include <random>
 #include "common/alignment.h"
@@ -14,13 +15,13 @@
 #include "core/file_sys/program_metadata.h"
 #include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/memory/memory_block_manager.h"
 #include "core/hle/kernel/memory/page_table.h"
 #include "core/hle/kernel/memory/slab_heap.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/lock.h"
 #include "core/memory.h"
@@ -53,7 +54,7 @@ void SetupMainThread(Core::System& system, Process& owner_process, u32 priority,
     auto& kernel = system.Kernel();
     // Threads by default are dormant, wake up the main thread so it runs when the scheduler fires
     {
-        SchedulerLock lock{kernel};
+        KScopedSchedulerLock lock{kernel};
         thread->SetStatus(ThreadStatus::Ready);
     }
 }
@@ -123,7 +124,7 @@ std::shared_ptr<Process> Process::Create(Core::System& system, std::string name,
                                                               : kernel.CreateNewUserProcessID();
     process->capabilities.InitializeForMetadatalessProcess();
 
-    std::mt19937 rng(Settings::values.rng_seed.GetValue().value_or(0));
+    std::mt19937 rng(Settings::values.rng_seed.GetValue().value_or(std::time(nullptr)));
     std::uniform_int_distribution<u64> distribution;
     std::generate(process->random_entropy.begin(), process->random_entropy.end(),
                   [&] { return distribution(rng); });
@@ -212,7 +213,7 @@ void Process::UnregisterThread(const Thread* thread) {
 }
 
 ResultCode Process::ClearSignalState() {
-    SchedulerLock lock(system.Kernel());
+    KScopedSchedulerLock lock(system.Kernel());
     if (status == ProcessStatus::Exited) {
         LOG_ERROR(Kernel, "called on a terminated process instance.");
         return ERR_INVALID_STATE;
@@ -313,7 +314,7 @@ void Process::PrepareForTermination() {
             if (thread->GetOwnerProcess() != this)
                 continue;
 
-            if (thread.get() == system.CurrentScheduler().GetCurrentThread())
+            if (thread.get() == kernel.CurrentScheduler()->GetCurrentThread())
                 continue;
 
             // TODO(Subv): When are the other running/ready threads terminated?
@@ -324,7 +325,7 @@ void Process::PrepareForTermination() {
         }
     };
 
-    stop_threads(system.GlobalScheduler().GetThreadList());
+    stop_threads(system.GlobalSchedulerContext().GetThreadList());
 
     FreeTLSRegion(tls_region_address);
     tls_region_address = 0;
@@ -346,7 +347,7 @@ static auto FindTLSPageWithAvailableSlots(std::vector<TLSPage>& tls_pages) {
 }
 
 VAddr Process::CreateTLSRegion() {
-    SchedulerLock lock(system.Kernel());
+    KScopedSchedulerLock lock(system.Kernel());
     if (auto tls_page_iter{FindTLSPageWithAvailableSlots(tls_pages)};
         tls_page_iter != tls_pages.cend()) {
         return *tls_page_iter->ReserveSlot();
@@ -377,7 +378,7 @@ VAddr Process::CreateTLSRegion() {
 }
 
 void Process::FreeTLSRegion(VAddr tls_address) {
-    SchedulerLock lock(system.Kernel());
+    KScopedSchedulerLock lock(system.Kernel());
     const VAddr aligned_address = Common::AlignDown(tls_address, Core::Memory::PAGE_SIZE);
     auto iter =
         std::find_if(tls_pages.begin(), tls_pages.end(), [aligned_address](const auto& page) {
diff --git a/src/core/hle/kernel/process.h b/src/core/hle/kernel/process.h
index f45cb5674..e412e58aa 100644
--- a/src/core/hle/kernel/process.h
+++ b/src/core/hle/kernel/process.h
@@ -216,6 +216,16 @@ public:
         total_process_running_time_ticks += ticks;
     }
 
+    /// Gets the process schedule count, used for thread yelding
+    s64 GetScheduledCount() const {
+        return schedule_count;
+    }
+
+    /// Increments the process schedule count, used for thread yielding.
+    void IncrementScheduledCount() {
+        ++schedule_count;
+    }
+
     /// Gets 8 bytes of random data for svcGetInfo RandomEntropy
     u64 GetRandomEntropy(std::size_t index) const {
         return random_entropy.at(index);
@@ -397,6 +407,9 @@ private:
     /// Name of this process
     std::string name;
 
+    /// Schedule count of this process
+    s64 schedule_count{};
+
     /// System context
     Core::System& system;
 };
diff --git a/src/core/hle/kernel/process_capability.cpp b/src/core/hle/kernel/process_capability.cpp
index 63880f13d..0f128c586 100644
--- a/src/core/hle/kernel/process_capability.cpp
+++ b/src/core/hle/kernel/process_capability.cpp
@@ -199,7 +199,7 @@ ResultCode ProcessCapabilities::ParseSingleFlagCapability(u32& set_flags, u32& s
         break;
     }
 
-    LOG_ERROR(Kernel, "Invalid capability type! type={}", static_cast<u32>(type));
+    LOG_ERROR(Kernel, "Invalid capability type! type={}", type);
     return ERR_INVALID_CAPABILITY_DESCRIPTOR;
 }
 
diff --git a/src/core/hle/kernel/readable_event.cpp b/src/core/hle/kernel/readable_event.cpp
index 6e286419e..cea262ce0 100644
--- a/src/core/hle/kernel/readable_event.cpp
+++ b/src/core/hle/kernel/readable_event.cpp
@@ -6,10 +6,10 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/readable_event.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 
 namespace Kernel {
@@ -39,7 +39,7 @@ void ReadableEvent::Clear() {
 }
 
 ResultCode ReadableEvent::Reset() {
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
     if (!is_signaled) {
         LOG_TRACE(Kernel, "Handle is not signaled! object_id={}, object_type={}, object_name={}",
                   GetObjectId(), GetTypeName(), GetName());
diff --git a/src/core/hle/kernel/resource_limit.cpp b/src/core/hle/kernel/resource_limit.cpp
index 212e442f4..7bf50339d 100644
--- a/src/core/hle/kernel/resource_limit.cpp
+++ b/src/core/hle/kernel/resource_limit.cpp
@@ -65,8 +65,8 @@ ResultCode ResourceLimit::SetLimitValue(ResourceType resource, s64 value) {
         limit[index] = value;
         return RESULT_SUCCESS;
     } else {
-        LOG_ERROR(Kernel, "Limit value is too large! resource={}, value={}, index={}",
-                  static_cast<u32>(resource), value, index);
+        LOG_ERROR(Kernel, "Limit value is too large! resource={}, value={}, index={}", resource,
+                  value, index);
         return ERR_INVALID_STATE;
     }
 }
diff --git a/src/core/hle/kernel/scheduler.cpp b/src/core/hle/kernel/scheduler.cpp
deleted file mode 100644
index 5cbd3b912..000000000
--- a/src/core/hle/kernel/scheduler.cpp
+++ /dev/null
@@ -1,849 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-//
-// SelectThreads, Yield functions originally by TuxSH.
-// licensed under GPLv2 or later under exception provided by the author.
-
-#include <algorithm>
-#include <mutex>
-#include <set>
-#include <unordered_set>
-#include <utility>
-
-#include "common/assert.h"
-#include "common/bit_util.h"
-#include "common/fiber.h"
-#include "common/logging/log.h"
-#include "core/arm/arm_interface.h"
-#include "core/core.h"
-#include "core/core_timing.h"
-#include "core/cpu_manager.h"
-#include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/physical_core.h"
-#include "core/hle/kernel/process.h"
-#include "core/hle/kernel/scheduler.h"
-#include "core/hle/kernel/time_manager.h"
-
-namespace Kernel {
-
-GlobalScheduler::GlobalScheduler(KernelCore& kernel) : kernel{kernel} {}
-
-GlobalScheduler::~GlobalScheduler() = default;
-
-void GlobalScheduler::AddThread(std::shared_ptr<Thread> thread) {
-    std::scoped_lock lock{global_list_guard};
-    thread_list.push_back(std::move(thread));
-}
-
-void GlobalScheduler::RemoveThread(std::shared_ptr<Thread> thread) {
-    std::scoped_lock lock{global_list_guard};
-    thread_list.erase(std::remove(thread_list.begin(), thread_list.end(), thread),
-                      thread_list.end());
-}
-
-u32 GlobalScheduler::SelectThreads() {
-    ASSERT(is_locked);
-    const auto update_thread = [](Thread* thread, Scheduler& sched) {
-        std::scoped_lock lock{sched.guard};
-        if (thread != sched.selected_thread_set.get()) {
-            if (thread == nullptr) {
-                ++sched.idle_selection_count;
-            }
-            sched.selected_thread_set = SharedFrom(thread);
-        }
-        const bool reschedule_pending =
-            sched.is_context_switch_pending || (sched.selected_thread_set != sched.current_thread);
-        sched.is_context_switch_pending = reschedule_pending;
-        std::atomic_thread_fence(std::memory_order_seq_cst);
-        return reschedule_pending;
-    };
-    if (!is_reselection_pending.load()) {
-        return 0;
-    }
-    std::array<Thread*, Core::Hardware::NUM_CPU_CORES> top_threads{};
-
-    u32 idle_cores{};
-
-    // Step 1: Get top thread in schedule queue.
-    for (u32 core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-        Thread* top_thread =
-            scheduled_queue[core].empty() ? nullptr : scheduled_queue[core].front();
-        if (top_thread != nullptr) {
-            // TODO(Blinkhawk): Implement Thread Pinning
-        } else {
-            idle_cores |= (1ul << core);
-        }
-        top_threads[core] = top_thread;
-    }
-
-    while (idle_cores != 0) {
-        u32 core_id = Common::CountTrailingZeroes32(idle_cores);
-
-        if (!suggested_queue[core_id].empty()) {
-            std::array<s32, Core::Hardware::NUM_CPU_CORES> migration_candidates{};
-            std::size_t num_candidates = 0;
-            auto iter = suggested_queue[core_id].begin();
-            Thread* suggested = nullptr;
-            // Step 2: Try selecting a suggested thread.
-            while (iter != suggested_queue[core_id].end()) {
-                suggested = *iter;
-                iter++;
-                s32 suggested_core_id = suggested->GetProcessorID();
-                Thread* top_thread =
-                    suggested_core_id >= 0 ? top_threads[suggested_core_id] : nullptr;
-                if (top_thread != suggested) {
-                    if (top_thread != nullptr &&
-                        top_thread->GetPriority() < THREADPRIO_MAX_CORE_MIGRATION) {
-                        suggested = nullptr;
-                        break;
-                        // There's a too high thread to do core migration, cancel
-                    }
-                    TransferToCore(suggested->GetPriority(), static_cast<s32>(core_id), suggested);
-                    break;
-                }
-                suggested = nullptr;
-                migration_candidates[num_candidates++] = suggested_core_id;
-            }
-            // Step 3: Select a suggested thread from another core
-            if (suggested == nullptr) {
-                for (std::size_t i = 0; i < num_candidates; i++) {
-                    s32 candidate_core = migration_candidates[i];
-                    suggested = top_threads[candidate_core];
-                    auto it = scheduled_queue[candidate_core].begin();
-                    it++;
-                    Thread* next = it != scheduled_queue[candidate_core].end() ? *it : nullptr;
-                    if (next != nullptr) {
-                        TransferToCore(suggested->GetPriority(), static_cast<s32>(core_id),
-                                       suggested);
-                        top_threads[candidate_core] = next;
-                        break;
-                    } else {
-                        suggested = nullptr;
-                    }
-                }
-            }
-            top_threads[core_id] = suggested;
-        }
-
-        idle_cores &= ~(1ul << core_id);
-    }
-    u32 cores_needing_context_switch{};
-    for (u32 core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-        Scheduler& sched = kernel.Scheduler(core);
-        ASSERT(top_threads[core] == nullptr ||
-               static_cast<u32>(top_threads[core]->GetProcessorID()) == core);
-        if (update_thread(top_threads[core], sched)) {
-            cores_needing_context_switch |= (1ul << core);
-        }
-    }
-    return cores_needing_context_switch;
-}
-
-bool GlobalScheduler::YieldThread(Thread* yielding_thread) {
-    ASSERT(is_locked);
-    // Note: caller should use critical section, etc.
-    if (!yielding_thread->IsRunnable()) {
-        // Normally this case shouldn't happen except for SetThreadActivity.
-        is_reselection_pending.store(true, std::memory_order_release);
-        return false;
-    }
-    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
-    const u32 priority = yielding_thread->GetPriority();
-
-    // Yield the thread
-    Reschedule(priority, core_id, yielding_thread);
-    const Thread* const winner = scheduled_queue[core_id].front();
-    if (kernel.GetCurrentHostThreadID() != core_id) {
-        is_reselection_pending.store(true, std::memory_order_release);
-    }
-
-    return AskForReselectionOrMarkRedundant(yielding_thread, winner);
-}
-
-bool GlobalScheduler::YieldThreadAndBalanceLoad(Thread* yielding_thread) {
-    ASSERT(is_locked);
-    // Note: caller should check if !thread.IsSchedulerOperationRedundant and use critical section,
-    // etc.
-    if (!yielding_thread->IsRunnable()) {
-        // Normally this case shouldn't happen except for SetThreadActivity.
-        is_reselection_pending.store(true, std::memory_order_release);
-        return false;
-    }
-    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
-    const u32 priority = yielding_thread->GetPriority();
-
-    // Yield the thread
-    Reschedule(priority, core_id, yielding_thread);
-
-    std::array<Thread*, Core::Hardware::NUM_CPU_CORES> current_threads;
-    for (std::size_t i = 0; i < current_threads.size(); i++) {
-        current_threads[i] = scheduled_queue[i].empty() ? nullptr : scheduled_queue[i].front();
-    }
-
-    Thread* next_thread = scheduled_queue[core_id].front(priority);
-    Thread* winner = nullptr;
-    for (auto& thread : suggested_queue[core_id]) {
-        const s32 source_core = thread->GetProcessorID();
-        if (source_core >= 0) {
-            if (current_threads[source_core] != nullptr) {
-                if (thread == current_threads[source_core] ||
-                    current_threads[source_core]->GetPriority() < min_regular_priority) {
-                    continue;
-                }
-            }
-        }
-        if (next_thread->GetLastRunningTicks() >= thread->GetLastRunningTicks() ||
-            next_thread->GetPriority() < thread->GetPriority()) {
-            if (thread->GetPriority() <= priority) {
-                winner = thread;
-                break;
-            }
-        }
-    }
-
-    if (winner != nullptr) {
-        if (winner != yielding_thread) {
-            TransferToCore(winner->GetPriority(), s32(core_id), winner);
-        }
-    } else {
-        winner = next_thread;
-    }
-
-    if (kernel.GetCurrentHostThreadID() != core_id) {
-        is_reselection_pending.store(true, std::memory_order_release);
-    }
-
-    return AskForReselectionOrMarkRedundant(yielding_thread, winner);
-}
-
-bool GlobalScheduler::YieldThreadAndWaitForLoadBalancing(Thread* yielding_thread) {
-    ASSERT(is_locked);
-    // Note: caller should check if !thread.IsSchedulerOperationRedundant and use critical section,
-    // etc.
-    if (!yielding_thread->IsRunnable()) {
-        // Normally this case shouldn't happen except for SetThreadActivity.
-        is_reselection_pending.store(true, std::memory_order_release);
-        return false;
-    }
-    Thread* winner = nullptr;
-    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
-
-    // Remove the thread from its scheduled mlq, put it on the corresponding "suggested" one instead
-    TransferToCore(yielding_thread->GetPriority(), -1, yielding_thread);
-
-    // If the core is idle, perform load balancing, excluding the threads that have just used this
-    // function...
-    if (scheduled_queue[core_id].empty()) {
-        // Here, "current_threads" is calculated after the ""yield"", unlike yield -1
-        std::array<Thread*, Core::Hardware::NUM_CPU_CORES> current_threads;
-        for (std::size_t i = 0; i < current_threads.size(); i++) {
-            current_threads[i] = scheduled_queue[i].empty() ? nullptr : scheduled_queue[i].front();
-        }
-        for (auto& thread : suggested_queue[core_id]) {
-            const s32 source_core = thread->GetProcessorID();
-            if (source_core < 0 || thread == current_threads[source_core]) {
-                continue;
-            }
-            if (current_threads[source_core] == nullptr ||
-                current_threads[source_core]->GetPriority() >= min_regular_priority) {
-                winner = thread;
-            }
-            break;
-        }
-        if (winner != nullptr) {
-            if (winner != yielding_thread) {
-                TransferToCore(winner->GetPriority(), static_cast<s32>(core_id), winner);
-            }
-        } else {
-            winner = yielding_thread;
-        }
-    } else {
-        winner = scheduled_queue[core_id].front();
-    }
-
-    if (kernel.GetCurrentHostThreadID() != core_id) {
-        is_reselection_pending.store(true, std::memory_order_release);
-    }
-
-    return AskForReselectionOrMarkRedundant(yielding_thread, winner);
-}
-
-void GlobalScheduler::PreemptThreads() {
-    ASSERT(is_locked);
-    for (std::size_t core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-        const u32 priority = preemption_priorities[core_id];
-
-        if (scheduled_queue[core_id].size(priority) > 0) {
-            if (scheduled_queue[core_id].size(priority) > 1) {
-                scheduled_queue[core_id].front(priority)->IncrementYieldCount();
-            }
-            scheduled_queue[core_id].yield(priority);
-            if (scheduled_queue[core_id].size(priority) > 1) {
-                scheduled_queue[core_id].front(priority)->IncrementYieldCount();
-            }
-        }
-
-        Thread* current_thread =
-            scheduled_queue[core_id].empty() ? nullptr : scheduled_queue[core_id].front();
-        Thread* winner = nullptr;
-        for (auto& thread : suggested_queue[core_id]) {
-            const s32 source_core = thread->GetProcessorID();
-            if (thread->GetPriority() != priority) {
-                continue;
-            }
-            if (source_core >= 0) {
-                Thread* next_thread = scheduled_queue[source_core].empty()
-                                          ? nullptr
-                                          : scheduled_queue[source_core].front();
-                if (next_thread != nullptr && next_thread->GetPriority() < 2) {
-                    break;
-                }
-                if (next_thread == thread) {
-                    continue;
-                }
-            }
-            if (current_thread != nullptr &&
-                current_thread->GetLastRunningTicks() >= thread->GetLastRunningTicks()) {
-                winner = thread;
-                break;
-            }
-        }
-
-        if (winner != nullptr) {
-            TransferToCore(winner->GetPriority(), s32(core_id), winner);
-            current_thread =
-                winner->GetPriority() <= current_thread->GetPriority() ? winner : current_thread;
-        }
-
-        if (current_thread != nullptr && current_thread->GetPriority() > priority) {
-            for (auto& thread : suggested_queue[core_id]) {
-                const s32 source_core = thread->GetProcessorID();
-                if (thread->GetPriority() < priority) {
-                    continue;
-                }
-                if (source_core >= 0) {
-                    Thread* next_thread = scheduled_queue[source_core].empty()
-                                              ? nullptr
-                                              : scheduled_queue[source_core].front();
-                    if (next_thread != nullptr && next_thread->GetPriority() < 2) {
-                        break;
-                    }
-                    if (next_thread == thread) {
-                        continue;
-                    }
-                }
-                if (current_thread != nullptr &&
-                    current_thread->GetLastRunningTicks() >= thread->GetLastRunningTicks()) {
-                    winner = thread;
-                    break;
-                }
-            }
-
-            if (winner != nullptr) {
-                TransferToCore(winner->GetPriority(), s32(core_id), winner);
-                current_thread = winner;
-            }
-        }
-
-        is_reselection_pending.store(true, std::memory_order_release);
-    }
-}
-
-void GlobalScheduler::EnableInterruptAndSchedule(u32 cores_pending_reschedule,
-                                                 Core::EmuThreadHandle global_thread) {
-    u32 current_core = global_thread.host_handle;
-    bool must_context_switch = global_thread.guest_handle != InvalidHandle &&
-                               (current_core < Core::Hardware::NUM_CPU_CORES);
-    while (cores_pending_reschedule != 0) {
-        u32 core = Common::CountTrailingZeroes32(cores_pending_reschedule);
-        ASSERT(core < Core::Hardware::NUM_CPU_CORES);
-        if (!must_context_switch || core != current_core) {
-            auto& phys_core = kernel.PhysicalCore(core);
-            phys_core.Interrupt();
-        } else {
-            must_context_switch = true;
-        }
-        cores_pending_reschedule &= ~(1ul << core);
-    }
-    if (must_context_switch) {
-        auto& core_scheduler = kernel.CurrentScheduler();
-        kernel.ExitSVCProfile();
-        core_scheduler.TryDoContextSwitch();
-        kernel.EnterSVCProfile();
-    }
-}
-
-void GlobalScheduler::Suggest(u32 priority, std::size_t core, Thread* thread) {
-    ASSERT(is_locked);
-    suggested_queue[core].add(thread, priority);
-}
-
-void GlobalScheduler::Unsuggest(u32 priority, std::size_t core, Thread* thread) {
-    ASSERT(is_locked);
-    suggested_queue[core].remove(thread, priority);
-}
-
-void GlobalScheduler::Schedule(u32 priority, std::size_t core, Thread* thread) {
-    ASSERT(is_locked);
-    ASSERT_MSG(thread->GetProcessorID() == s32(core), "Thread must be assigned to this core.");
-    scheduled_queue[core].add(thread, priority);
-}
-
-void GlobalScheduler::SchedulePrepend(u32 priority, std::size_t core, Thread* thread) {
-    ASSERT(is_locked);
-    ASSERT_MSG(thread->GetProcessorID() == s32(core), "Thread must be assigned to this core.");
-    scheduled_queue[core].add(thread, priority, false);
-}
-
-void GlobalScheduler::Reschedule(u32 priority, std::size_t core, Thread* thread) {
-    ASSERT(is_locked);
-    scheduled_queue[core].remove(thread, priority);
-    scheduled_queue[core].add(thread, priority);
-}
-
-void GlobalScheduler::Unschedule(u32 priority, std::size_t core, Thread* thread) {
-    ASSERT(is_locked);
-    scheduled_queue[core].remove(thread, priority);
-}
-
-void GlobalScheduler::TransferToCore(u32 priority, s32 destination_core, Thread* thread) {
-    ASSERT(is_locked);
-    const bool schedulable = thread->GetPriority() < THREADPRIO_COUNT;
-    const s32 source_core = thread->GetProcessorID();
-    if (source_core == destination_core || !schedulable) {
-        return;
-    }
-    thread->SetProcessorID(destination_core);
-    if (source_core >= 0) {
-        Unschedule(priority, static_cast<u32>(source_core), thread);
-    }
-    if (destination_core >= 0) {
-        Unsuggest(priority, static_cast<u32>(destination_core), thread);
-        Schedule(priority, static_cast<u32>(destination_core), thread);
-    }
-    if (source_core >= 0) {
-        Suggest(priority, static_cast<u32>(source_core), thread);
-    }
-}
-
-bool GlobalScheduler::AskForReselectionOrMarkRedundant(Thread* current_thread,
-                                                       const Thread* winner) {
-    if (current_thread == winner) {
-        current_thread->IncrementYieldCount();
-        return true;
-    } else {
-        is_reselection_pending.store(true, std::memory_order_release);
-        return false;
-    }
-}
-
-void GlobalScheduler::AdjustSchedulingOnStatus(Thread* thread, u32 old_flags) {
-    if (old_flags == thread->scheduling_state) {
-        return;
-    }
-    ASSERT(is_locked);
-
-    if (old_flags == static_cast<u32>(ThreadSchedStatus::Runnable)) {
-        // In this case the thread was running, now it's pausing/exitting
-        if (thread->processor_id >= 0) {
-            Unschedule(thread->current_priority, static_cast<u32>(thread->processor_id), thread);
-        }
-
-        for (u32 core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-            if (core != static_cast<u32>(thread->processor_id) &&
-                ((thread->affinity_mask >> core) & 1) != 0) {
-                Unsuggest(thread->current_priority, core, thread);
-            }
-        }
-    } else if (thread->scheduling_state == static_cast<u32>(ThreadSchedStatus::Runnable)) {
-        // The thread is now set to running from being stopped
-        if (thread->processor_id >= 0) {
-            Schedule(thread->current_priority, static_cast<u32>(thread->processor_id), thread);
-        }
-
-        for (u32 core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-            if (core != static_cast<u32>(thread->processor_id) &&
-                ((thread->affinity_mask >> core) & 1) != 0) {
-                Suggest(thread->current_priority, core, thread);
-            }
-        }
-    }
-
-    SetReselectionPending();
-}
-
-void GlobalScheduler::AdjustSchedulingOnPriority(Thread* thread, u32 old_priority) {
-    if (thread->scheduling_state != static_cast<u32>(ThreadSchedStatus::Runnable)) {
-        return;
-    }
-    ASSERT(is_locked);
-    if (thread->processor_id >= 0) {
-        Unschedule(old_priority, static_cast<u32>(thread->processor_id), thread);
-    }
-
-    for (u32 core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-        if (core != static_cast<u32>(thread->processor_id) &&
-            ((thread->affinity_mask >> core) & 1) != 0) {
-            Unsuggest(old_priority, core, thread);
-        }
-    }
-
-    if (thread->processor_id >= 0) {
-        if (thread == kernel.CurrentScheduler().GetCurrentThread()) {
-            SchedulePrepend(thread->current_priority, static_cast<u32>(thread->processor_id),
-                            thread);
-        } else {
-            Schedule(thread->current_priority, static_cast<u32>(thread->processor_id), thread);
-        }
-    }
-
-    for (u32 core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-        if (core != static_cast<u32>(thread->processor_id) &&
-            ((thread->affinity_mask >> core) & 1) != 0) {
-            Suggest(thread->current_priority, core, thread);
-        }
-    }
-    thread->IncrementYieldCount();
-    SetReselectionPending();
-}
-
-void GlobalScheduler::AdjustSchedulingOnAffinity(Thread* thread, u64 old_affinity_mask,
-                                                 s32 old_core) {
-    if (thread->scheduling_state != static_cast<u32>(ThreadSchedStatus::Runnable) ||
-        thread->current_priority >= THREADPRIO_COUNT) {
-        return;
-    }
-    ASSERT(is_locked);
-
-    for (u32 core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-        if (((old_affinity_mask >> core) & 1) != 0) {
-            if (core == static_cast<u32>(old_core)) {
-                Unschedule(thread->current_priority, core, thread);
-            } else {
-                Unsuggest(thread->current_priority, core, thread);
-            }
-        }
-    }
-
-    for (u32 core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-        if (((thread->affinity_mask >> core) & 1) != 0) {
-            if (core == static_cast<u32>(thread->processor_id)) {
-                Schedule(thread->current_priority, core, thread);
-            } else {
-                Suggest(thread->current_priority, core, thread);
-            }
-        }
-    }
-
-    thread->IncrementYieldCount();
-    SetReselectionPending();
-}
-
-void GlobalScheduler::Shutdown() {
-    for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-        scheduled_queue[core].clear();
-        suggested_queue[core].clear();
-    }
-    thread_list.clear();
-}
-
-void GlobalScheduler::Lock() {
-    Core::EmuThreadHandle current_thread = kernel.GetCurrentEmuThreadID();
-    ASSERT(!current_thread.IsInvalid());
-    if (current_thread == current_owner) {
-        ++scope_lock;
-    } else {
-        inner_lock.lock();
-        is_locked = true;
-        current_owner = current_thread;
-        ASSERT(current_owner != Core::EmuThreadHandle::InvalidHandle());
-        scope_lock = 1;
-    }
-}
-
-void GlobalScheduler::Unlock() {
-    if (--scope_lock != 0) {
-        ASSERT(scope_lock > 0);
-        return;
-    }
-    u32 cores_pending_reschedule = SelectThreads();
-    Core::EmuThreadHandle leaving_thread = current_owner;
-    current_owner = Core::EmuThreadHandle::InvalidHandle();
-    scope_lock = 1;
-    is_locked = false;
-    inner_lock.unlock();
-    EnableInterruptAndSchedule(cores_pending_reschedule, leaving_thread);
-}
-
-Scheduler::Scheduler(Core::System& system, std::size_t core_id) : system(system), core_id(core_id) {
-    switch_fiber = std::make_shared<Common::Fiber>(std::function<void(void*)>(OnSwitch), this);
-}
-
-Scheduler::~Scheduler() = default;
-
-bool Scheduler::HaveReadyThreads() const {
-    return system.GlobalScheduler().HaveReadyThreads(core_id);
-}
-
-Thread* Scheduler::GetCurrentThread() const {
-    if (current_thread) {
-        return current_thread.get();
-    }
-    return idle_thread.get();
-}
-
-Thread* Scheduler::GetSelectedThread() const {
-    return selected_thread.get();
-}
-
-u64 Scheduler::GetLastContextSwitchTicks() const {
-    return last_context_switch_time;
-}
-
-void Scheduler::TryDoContextSwitch() {
-    auto& phys_core = system.Kernel().CurrentPhysicalCore();
-    if (phys_core.IsInterrupted()) {
-        phys_core.ClearInterrupt();
-    }
-    guard.lock();
-    if (is_context_switch_pending) {
-        SwitchContext();
-    } else {
-        guard.unlock();
-    }
-}
-
-void Scheduler::OnThreadStart() {
-    SwitchContextStep2();
-}
-
-void Scheduler::Unload() {
-    Thread* thread = current_thread.get();
-    if (thread) {
-        thread->SetContinuousOnSVC(false);
-        thread->last_running_ticks = system.CoreTiming().GetCPUTicks();
-        thread->SetIsRunning(false);
-        if (!thread->IsHLEThread() && !thread->HasExited()) {
-            Core::ARM_Interface& cpu_core = thread->ArmInterface();
-            cpu_core.SaveContext(thread->GetContext32());
-            cpu_core.SaveContext(thread->GetContext64());
-            // Save the TPIDR_EL0 system register in case it was modified.
-            thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
-            cpu_core.ClearExclusiveState();
-        }
-        thread->context_guard.unlock();
-    }
-}
-
-void Scheduler::Reload() {
-    Thread* thread = current_thread.get();
-    if (thread) {
-        ASSERT_MSG(thread->GetSchedulingStatus() == ThreadSchedStatus::Runnable,
-                   "Thread must be runnable.");
-
-        // Cancel any outstanding wakeup events for this thread
-        thread->SetIsRunning(true);
-        thread->SetWasRunning(false);
-        thread->last_running_ticks = system.CoreTiming().GetCPUTicks();
-
-        auto* const thread_owner_process = thread->GetOwnerProcess();
-        if (thread_owner_process != nullptr) {
-            system.Kernel().MakeCurrentProcess(thread_owner_process);
-        }
-        if (!thread->IsHLEThread()) {
-            Core::ARM_Interface& cpu_core = thread->ArmInterface();
-            cpu_core.LoadContext(thread->GetContext32());
-            cpu_core.LoadContext(thread->GetContext64());
-            cpu_core.SetTlsAddress(thread->GetTLSAddress());
-            cpu_core.SetTPIDR_EL0(thread->GetTPIDR_EL0());
-            cpu_core.ChangeProcessorID(this->core_id);
-            cpu_core.ClearExclusiveState();
-        }
-    }
-}
-
-void Scheduler::SwitchContextStep2() {
-    // Load context of new thread
-    if (selected_thread) {
-        ASSERT_MSG(selected_thread->GetSchedulingStatus() == ThreadSchedStatus::Runnable,
-                   "Thread must be runnable.");
-
-        // Cancel any outstanding wakeup events for this thread
-        selected_thread->SetIsRunning(true);
-        selected_thread->last_running_ticks = system.CoreTiming().GetCPUTicks();
-        selected_thread->SetWasRunning(false);
-
-        auto* const thread_owner_process = current_thread->GetOwnerProcess();
-        if (thread_owner_process != nullptr) {
-            system.Kernel().MakeCurrentProcess(thread_owner_process);
-        }
-        if (!selected_thread->IsHLEThread()) {
-            Core::ARM_Interface& cpu_core = selected_thread->ArmInterface();
-            cpu_core.LoadContext(selected_thread->GetContext32());
-            cpu_core.LoadContext(selected_thread->GetContext64());
-            cpu_core.SetTlsAddress(selected_thread->GetTLSAddress());
-            cpu_core.SetTPIDR_EL0(selected_thread->GetTPIDR_EL0());
-            cpu_core.ChangeProcessorID(this->core_id);
-            cpu_core.ClearExclusiveState();
-        }
-    }
-
-    TryDoContextSwitch();
-}
-
-void Scheduler::SwitchContext() {
-    current_thread_prev = current_thread;
-    selected_thread = selected_thread_set;
-    Thread* previous_thread = current_thread_prev.get();
-    Thread* new_thread = selected_thread.get();
-    current_thread = selected_thread;
-
-    is_context_switch_pending = false;
-
-    if (new_thread == previous_thread) {
-        guard.unlock();
-        return;
-    }
-
-    Process* const previous_process = system.Kernel().CurrentProcess();
-
-    UpdateLastContextSwitchTime(previous_thread, previous_process);
-
-    // Save context for previous thread
-    if (previous_thread) {
-        if (new_thread != nullptr && new_thread->IsSuspendThread()) {
-            previous_thread->SetWasRunning(true);
-        }
-        previous_thread->SetContinuousOnSVC(false);
-        previous_thread->last_running_ticks = system.CoreTiming().GetCPUTicks();
-        previous_thread->SetIsRunning(false);
-        if (!previous_thread->IsHLEThread() && !previous_thread->HasExited()) {
-            Core::ARM_Interface& cpu_core = previous_thread->ArmInterface();
-            cpu_core.SaveContext(previous_thread->GetContext32());
-            cpu_core.SaveContext(previous_thread->GetContext64());
-            // Save the TPIDR_EL0 system register in case it was modified.
-            previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
-            cpu_core.ClearExclusiveState();
-        }
-        previous_thread->context_guard.unlock();
-    }
-
-    std::shared_ptr<Common::Fiber>* old_context;
-    if (previous_thread != nullptr) {
-        old_context = &previous_thread->GetHostContext();
-    } else {
-        old_context = &idle_thread->GetHostContext();
-    }
-    guard.unlock();
-
-    Common::Fiber::YieldTo(*old_context, switch_fiber);
-    /// When a thread wakes up, the scheduler may have changed to other in another core.
-    auto& next_scheduler = system.Kernel().CurrentScheduler();
-    next_scheduler.SwitchContextStep2();
-}
-
-void Scheduler::OnSwitch(void* this_scheduler) {
-    Scheduler* sched = static_cast<Scheduler*>(this_scheduler);
-    sched->SwitchToCurrent();
-}
-
-void Scheduler::SwitchToCurrent() {
-    while (true) {
-        {
-            std::scoped_lock lock{guard};
-            selected_thread = selected_thread_set;
-            current_thread = selected_thread;
-            is_context_switch_pending = false;
-        }
-        const auto is_switch_pending = [this] {
-            std::scoped_lock lock{guard};
-            return is_context_switch_pending;
-        };
-        do {
-            if (current_thread != nullptr && !current_thread->IsHLEThread()) {
-                current_thread->context_guard.lock();
-                if (!current_thread->IsRunnable()) {
-                    current_thread->context_guard.unlock();
-                    break;
-                }
-                if (current_thread->GetProcessorID() != core_id) {
-                    current_thread->context_guard.unlock();
-                    break;
-                }
-            }
-            std::shared_ptr<Common::Fiber>* next_context;
-            if (current_thread != nullptr) {
-                next_context = &current_thread->GetHostContext();
-            } else {
-                next_context = &idle_thread->GetHostContext();
-            }
-            Common::Fiber::YieldTo(switch_fiber, *next_context);
-        } while (!is_switch_pending());
-    }
-}
-
-void Scheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
-    const u64 prev_switch_ticks = last_context_switch_time;
-    const u64 most_recent_switch_ticks = system.CoreTiming().GetCPUTicks();
-    const u64 update_ticks = most_recent_switch_ticks - prev_switch_ticks;
-
-    if (thread != nullptr) {
-        thread->UpdateCPUTimeTicks(update_ticks);
-    }
-
-    if (process != nullptr) {
-        process->UpdateCPUTimeTicks(update_ticks);
-    }
-
-    last_context_switch_time = most_recent_switch_ticks;
-}
-
-void Scheduler::Initialize() {
-    std::string name = "Idle Thread Id:" + std::to_string(core_id);
-    std::function<void(void*)> init_func = Core::CpuManager::GetIdleThreadStartFunc();
-    void* init_func_parameter = system.GetCpuManager().GetStartFuncParamater();
-    ThreadType type = static_cast<ThreadType>(THREADTYPE_KERNEL | THREADTYPE_HLE | THREADTYPE_IDLE);
-    auto thread_res = Thread::Create(system, type, name, 0, 64, 0, static_cast<u32>(core_id), 0,
-                                     nullptr, std::move(init_func), init_func_parameter);
-    idle_thread = std::move(thread_res).Unwrap();
-}
-
-void Scheduler::Shutdown() {
-    current_thread = nullptr;
-    selected_thread = nullptr;
-}
-
-SchedulerLock::SchedulerLock(KernelCore& kernel) : kernel{kernel} {
-    kernel.GlobalScheduler().Lock();
-}
-
-SchedulerLock::~SchedulerLock() {
-    kernel.GlobalScheduler().Unlock();
-}
-
-SchedulerLockAndSleep::SchedulerLockAndSleep(KernelCore& kernel, Handle& event_handle,
-                                             Thread* time_task, s64 nanoseconds)
-    : SchedulerLock{kernel}, event_handle{event_handle}, time_task{time_task}, nanoseconds{
-                                                                                   nanoseconds} {
-    event_handle = InvalidHandle;
-}
-
-SchedulerLockAndSleep::~SchedulerLockAndSleep() {
-    if (sleep_cancelled) {
-        return;
-    }
-    auto& time_manager = kernel.TimeManager();
-    time_manager.ScheduleTimeEvent(event_handle, time_task, nanoseconds);
-}
-
-void SchedulerLockAndSleep::Release() {
-    if (sleep_cancelled) {
-        return;
-    }
-    auto& time_manager = kernel.TimeManager();
-    time_manager.ScheduleTimeEvent(event_handle, time_task, nanoseconds);
-    sleep_cancelled = true;
-}
-
-} // namespace Kernel
diff --git a/src/core/hle/kernel/scheduler.h b/src/core/hle/kernel/scheduler.h
deleted file mode 100644
index b6f04dcea..000000000
--- a/src/core/hle/kernel/scheduler.h
+++ /dev/null
@@ -1,318 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <atomic>
-#include <memory>
-#include <mutex>
-#include <vector>
-
-#include "common/common_types.h"
-#include "common/multi_level_queue.h"
-#include "common/spin_lock.h"
-#include "core/hardware_properties.h"
-#include "core/hle/kernel/thread.h"
-
-namespace Common {
-class Fiber;
-}
-
-namespace Core {
-class ARM_Interface;
-class System;
-} // namespace Core
-
-namespace Kernel {
-
-class KernelCore;
-class Process;
-class SchedulerLock;
-
-class GlobalScheduler final {
-public:
-    explicit GlobalScheduler(KernelCore& kernel);
-    ~GlobalScheduler();
-
-    /// Adds a new thread to the scheduler
-    void AddThread(std::shared_ptr<Thread> thread);
-
-    /// Removes a thread from the scheduler
-    void RemoveThread(std::shared_ptr<Thread> thread);
-
-    /// Returns a list of all threads managed by the scheduler
-    const std::vector<std::shared_ptr<Thread>>& GetThreadList() const {
-        return thread_list;
-    }
-
-    /// Notify the scheduler a thread's status has changed.
-    void AdjustSchedulingOnStatus(Thread* thread, u32 old_flags);
-
-    /// Notify the scheduler a thread's priority has changed.
-    void AdjustSchedulingOnPriority(Thread* thread, u32 old_priority);
-
-    /// Notify the scheduler a thread's core and/or affinity mask has changed.
-    void AdjustSchedulingOnAffinity(Thread* thread, u64 old_affinity_mask, s32 old_core);
-
-    /**
-     * Takes care of selecting the new scheduled threads in three steps:
-     *
-     * 1. First a thread is selected from the top of the priority queue. If no thread
-     *    is obtained then we move to step two, else we are done.
-     *
-     * 2. Second we try to get a suggested thread that's not assigned to any core or
-     *    that is not the top thread in that core.
-     *
-     * 3. Third is no suggested thread is found, we do a second pass and pick a running
-     *    thread in another core and swap it with its current thread.
-     *
-     * returns the cores needing scheduling.
-     */
-    u32 SelectThreads();
-
-    bool HaveReadyThreads(std::size_t core_id) const {
-        return !scheduled_queue[core_id].empty();
-    }
-
-    /**
-     * Takes a thread and moves it to the back of the it's priority list.
-     *
-     * @note This operation can be redundant and no scheduling is changed if marked as so.
-     */
-    bool YieldThread(Thread* thread);
-
-    /**
-     * Takes a thread and moves it to the back of the it's priority list.
-     * Afterwards, tries to pick a suggested thread from the suggested queue that has worse time or
-     * a better priority than the next thread in the core.
-     *
-     * @note This operation can be redundant and no scheduling is changed if marked as so.
-     */
-    bool YieldThreadAndBalanceLoad(Thread* thread);
-
-    /**
-     * Takes a thread and moves it out of the scheduling queue.
-     * and into the suggested queue. If no thread can be scheduled afterwards in that core,
-     * a suggested thread is obtained instead.
-     *
-     * @note This operation can be redundant and no scheduling is changed if marked as so.
-     */
-    bool YieldThreadAndWaitForLoadBalancing(Thread* thread);
-
-    /**
-     * Rotates the scheduling queues of threads at a preemption priority and then does
-     * some core rebalancing. Preemption priorities can be found in the array
-     * 'preemption_priorities'.
-     *
-     * @note This operation happens every 10ms.
-     */
-    void PreemptThreads();
-
-    u32 CpuCoresCount() const {
-        return Core::Hardware::NUM_CPU_CORES;
-    }
-
-    void SetReselectionPending() {
-        is_reselection_pending.store(true, std::memory_order_release);
-    }
-
-    bool IsReselectionPending() const {
-        return is_reselection_pending.load(std::memory_order_acquire);
-    }
-
-    void Shutdown();
-
-private:
-    friend class SchedulerLock;
-
-    /// Lock the scheduler to the current thread.
-    void Lock();
-
-    /// Unlocks the scheduler, reselects threads, interrupts cores for rescheduling
-    /// and reschedules current core if needed.
-    void Unlock();
-
-    void EnableInterruptAndSchedule(u32 cores_pending_reschedule,
-                                    Core::EmuThreadHandle global_thread);
-
-    /**
-     * Add a thread to the suggested queue of a cpu core. Suggested threads may be
-     * picked if no thread is scheduled to run on the core.
-     */
-    void Suggest(u32 priority, std::size_t core, Thread* thread);
-
-    /**
-     * Remove a thread to the suggested queue of a cpu core. Suggested threads may be
-     * picked if no thread is scheduled to run on the core.
-     */
-    void Unsuggest(u32 priority, std::size_t core, Thread* thread);
-
-    /**
-     * Add a thread to the scheduling queue of a cpu core. The thread is added at the
-     * back the queue in its priority level.
-     */
-    void Schedule(u32 priority, std::size_t core, Thread* thread);
-
-    /**
-     * Add a thread to the scheduling queue of a cpu core. The thread is added at the
-     * front the queue in its priority level.
-     */
-    void SchedulePrepend(u32 priority, std::size_t core, Thread* thread);
-
-    /// Reschedule an already scheduled thread based on a new priority
-    void Reschedule(u32 priority, std::size_t core, Thread* thread);
-
-    /// Unschedules a thread.
-    void Unschedule(u32 priority, std::size_t core, Thread* thread);
-
-    /**
-     * Transfers a thread into an specific core. If the destination_core is -1
-     * it will be unscheduled from its source code and added into its suggested
-     * queue.
-     */
-    void TransferToCore(u32 priority, s32 destination_core, Thread* thread);
-
-    bool AskForReselectionOrMarkRedundant(Thread* current_thread, const Thread* winner);
-
-    static constexpr u32 min_regular_priority = 2;
-    std::array<Common::MultiLevelQueue<Thread*, THREADPRIO_COUNT>, Core::Hardware::NUM_CPU_CORES>
-        scheduled_queue;
-    std::array<Common::MultiLevelQueue<Thread*, THREADPRIO_COUNT>, Core::Hardware::NUM_CPU_CORES>
-        suggested_queue;
-    std::atomic<bool> is_reselection_pending{false};
-
-    // The priority levels at which the global scheduler preempts threads every 10 ms. They are
-    // ordered from Core 0 to Core 3.
-    std::array<u32, Core::Hardware::NUM_CPU_CORES> preemption_priorities = {59, 59, 59, 62};
-
-    /// Scheduler lock mechanisms.
-    bool is_locked{};
-    std::mutex inner_lock;
-    std::atomic<s64> scope_lock{};
-    Core::EmuThreadHandle current_owner{Core::EmuThreadHandle::InvalidHandle()};
-
-    Common::SpinLock global_list_guard{};
-
-    /// Lists all thread ids that aren't deleted/etc.
-    std::vector<std::shared_ptr<Thread>> thread_list;
-    KernelCore& kernel;
-};
-
-class Scheduler final {
-public:
-    explicit Scheduler(Core::System& system, std::size_t core_id);
-    ~Scheduler();
-
-    /// Returns whether there are any threads that are ready to run.
-    bool HaveReadyThreads() const;
-
-    /// Reschedules to the next available thread (call after current thread is suspended)
-    void TryDoContextSwitch();
-
-    /// The next two are for SingleCore Only.
-    /// Unload current thread before preempting core.
-    void Unload();
-    /// Reload current thread after core preemption.
-    void Reload();
-
-    /// Gets the current running thread
-    Thread* GetCurrentThread() const;
-
-    /// Gets the currently selected thread from the top of the multilevel queue
-    Thread* GetSelectedThread() const;
-
-    /// Gets the timestamp for the last context switch in ticks.
-    u64 GetLastContextSwitchTicks() const;
-
-    bool ContextSwitchPending() const {
-        return is_context_switch_pending;
-    }
-
-    void Initialize();
-
-    /// Shutdowns the scheduler.
-    void Shutdown();
-
-    void OnThreadStart();
-
-    std::shared_ptr<Common::Fiber>& ControlContext() {
-        return switch_fiber;
-    }
-
-    const std::shared_ptr<Common::Fiber>& ControlContext() const {
-        return switch_fiber;
-    }
-
-private:
-    friend class GlobalScheduler;
-
-    /// Switches the CPU's active thread context to that of the specified thread
-    void SwitchContext();
-
-    /// When a thread wakes up, it must run this through it's new scheduler
-    void SwitchContextStep2();
-
-    /**
-     * Called on every context switch to update the internal timestamp
-     * This also updates the running time ticks for the given thread and
-     * process using the following difference:
-     *
-     * ticks += most_recent_ticks - last_context_switch_ticks
-     *
-     * The internal tick timestamp for the scheduler is simply the
-     * most recent tick count retrieved. No special arithmetic is
-     * applied to it.
-     */
-    void UpdateLastContextSwitchTime(Thread* thread, Process* process);
-
-    static void OnSwitch(void* this_scheduler);
-    void SwitchToCurrent();
-
-    std::shared_ptr<Thread> current_thread = nullptr;
-    std::shared_ptr<Thread> selected_thread = nullptr;
-    std::shared_ptr<Thread> current_thread_prev = nullptr;
-    std::shared_ptr<Thread> selected_thread_set = nullptr;
-    std::shared_ptr<Thread> idle_thread = nullptr;
-
-    std::shared_ptr<Common::Fiber> switch_fiber = nullptr;
-
-    Core::System& system;
-    u64 last_context_switch_time = 0;
-    u64 idle_selection_count = 0;
-    const std::size_t core_id;
-
-    Common::SpinLock guard{};
-
-    bool is_context_switch_pending = false;
-};
-
-class SchedulerLock {
-public:
-    [[nodiscard]] explicit SchedulerLock(KernelCore& kernel);
-    ~SchedulerLock();
-
-protected:
-    KernelCore& kernel;
-};
-
-class SchedulerLockAndSleep : public SchedulerLock {
-public:
-    explicit SchedulerLockAndSleep(KernelCore& kernel, Handle& event_handle, Thread* time_task,
-                                   s64 nanoseconds);
-    ~SchedulerLockAndSleep();
-
-    void CancelSleep() {
-        sleep_cancelled = true;
-    }
-
-    void Release();
-
-private:
-    Handle& event_handle;
-    Thread* time_task;
-    s64 nanoseconds;
-    bool sleep_cancelled{};
-};
-
-} // namespace Kernel
diff --git a/src/core/hle/kernel/server_session.cpp b/src/core/hle/kernel/server_session.cpp
index 8c19f2534..b40fe3916 100644
--- a/src/core/hle/kernel/server_session.cpp
+++ b/src/core/hle/kernel/server_session.cpp
@@ -14,9 +14,9 @@
 #include "core/hle/kernel/client_session.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/server_session.h"
 #include "core/hle/kernel/session.h"
 #include "core/hle/kernel/thread.h"
@@ -25,19 +25,19 @@
 namespace Kernel {
 
 ServerSession::ServerSession(KernelCore& kernel) : SynchronizationObject{kernel} {}
-ServerSession::~ServerSession() = default;
+
+ServerSession::~ServerSession() {
+    kernel.ReleaseServiceThread(service_thread);
+}
 
 ResultVal<std::shared_ptr<ServerSession>> ServerSession::Create(KernelCore& kernel,
                                                                 std::shared_ptr<Session> parent,
                                                                 std::string name) {
     std::shared_ptr<ServerSession> session{std::make_shared<ServerSession>(kernel)};
 
-    session->request_event =
-        Core::Timing::CreateEvent(name, [session](std::uintptr_t, std::chrono::nanoseconds) {
-            session->CompleteSyncRequest();
-        });
     session->name = std::move(name);
     session->parent = std::move(parent);
+    session->service_thread = kernel.CreateServiceThread(session->name);
 
     return MakeResult(std::move(session));
 }
@@ -130,8 +130,7 @@ ResultCode ServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& con
     }
     }
 
-    LOG_CRITICAL(IPC, "Unknown domain command={}",
-                 static_cast<int>(domain_message_header.command.Value()));
+    LOG_CRITICAL(IPC, "Unknown domain command={}", domain_message_header.command.Value());
     ASSERT(false);
     return RESULT_SUCCESS;
 }
@@ -143,16 +142,16 @@ ResultCode ServerSession::QueueSyncRequest(std::shared_ptr<Thread> thread,
         std::make_shared<HLERequestContext>(kernel, memory, SharedFrom(this), std::move(thread));
 
     context->PopulateFromIncomingCommandBuffer(kernel.CurrentProcess()->GetHandleTable(), cmd_buf);
-    request_queue.Push(std::move(context));
+
+    if (auto strong_ptr = service_thread.lock()) {
+        strong_ptr->QueueSyncRequest(*this, std::move(context));
+        return RESULT_SUCCESS;
+    }
 
     return RESULT_SUCCESS;
 }
 
-ResultCode ServerSession::CompleteSyncRequest() {
-    ASSERT(!request_queue.Empty());
-
-    auto& context = *request_queue.Front();
-
+ResultCode ServerSession::CompleteSyncRequest(HLERequestContext& context) {
     ResultCode result = RESULT_SUCCESS;
     // If the session has been converted to a domain, handle the domain request
     if (IsDomain() && context.HasDomainMessageHeader()) {
@@ -171,25 +170,20 @@ ResultCode ServerSession::CompleteSyncRequest() {
 
     // Some service requests require the thread to block
     {
-        SchedulerLock lock(kernel);
+        KScopedSchedulerLock lock(kernel);
         if (!context.IsThreadWaiting()) {
             context.GetThread().ResumeFromWait();
             context.GetThread().SetSynchronizationResults(nullptr, result);
         }
     }
 
-    request_queue.Pop();
-
     return result;
 }
 
 ResultCode ServerSession::HandleSyncRequest(std::shared_ptr<Thread> thread,
                                             Core::Memory::Memory& memory,
                                             Core::Timing::CoreTiming& core_timing) {
-    const ResultCode result = QueueSyncRequest(std::move(thread), memory);
-    const auto delay = std::chrono::nanoseconds{kernel.IsMulticore() ? 0 : 20000};
-    core_timing.ScheduleEvent(delay, request_event, {});
-    return result;
+    return QueueSyncRequest(std::move(thread), memory);
 }
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/server_session.h b/src/core/hle/kernel/server_session.h
index d23e9ec68..e8d1d99ea 100644
--- a/src/core/hle/kernel/server_session.h
+++ b/src/core/hle/kernel/server_session.h
@@ -10,6 +10,7 @@
 #include <vector>
 
 #include "common/threadsafe_queue.h"
+#include "core/hle/kernel/service_thread.h"
 #include "core/hle/kernel/synchronization_object.h"
 #include "core/hle/result.h"
 
@@ -43,6 +44,8 @@ class Thread;
  * TLS buffer and control is transferred back to it.
  */
 class ServerSession final : public SynchronizationObject {
+    friend class ServiceThread;
+
 public:
     explicit ServerSession(KernelCore& kernel);
     ~ServerSession() override;
@@ -132,7 +135,7 @@ private:
     ResultCode QueueSyncRequest(std::shared_ptr<Thread> thread, Core::Memory::Memory& memory);
 
     /// Completes a sync request from the emulated application.
-    ResultCode CompleteSyncRequest();
+    ResultCode CompleteSyncRequest(HLERequestContext& context);
 
     /// Handles a SyncRequest to a domain, forwarding the request to the proper object or closing an
     /// object handle.
@@ -163,11 +166,8 @@ private:
     /// The name of this session (optional)
     std::string name;
 
-    /// Core timing event used to schedule the service request at some point in the future
-    std::shared_ptr<Core::Timing::EventType> request_event;
-
-    /// Queue of scheduled service requests
-    Common::MPSCQueue<std::shared_ptr<Kernel::HLERequestContext>> request_queue;
+    /// Thread to dispatch service requests
+    std::weak_ptr<ServiceThread> service_thread;
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/service_thread.cpp b/src/core/hle/kernel/service_thread.cpp
new file mode 100644
index 000000000..ee46f3e21
--- /dev/null
+++ b/src/core/hle/kernel/service_thread.cpp
@@ -0,0 +1,110 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <condition_variable>
+#include <functional>
+#include <mutex>
+#include <thread>
+#include <vector>
+#include <queue>
+
+#include "common/assert.h"
+#include "common/scope_exit.h"
+#include "common/thread.h"
+#include "core/core.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/service_thread.h"
+#include "core/hle/lock.h"
+#include "video_core/renderer_base.h"
+
+namespace Kernel {
+
+class ServiceThread::Impl final {
+public:
+    explicit Impl(KernelCore& kernel, std::size_t num_threads, const std::string& name);
+    ~Impl();
+
+    void QueueSyncRequest(ServerSession& session, std::shared_ptr<HLERequestContext>&& context);
+
+private:
+    std::vector<std::thread> threads;
+    std::queue<std::function<void()>> requests;
+    std::mutex queue_mutex;
+    std::condition_variable condition;
+    const std::string service_name;
+    bool stop{};
+};
+
+ServiceThread::Impl::Impl(KernelCore& kernel, std::size_t num_threads, const std::string& name)
+    : service_name{name} {
+    for (std::size_t i = 0; i < num_threads; ++i)
+        threads.emplace_back([this, &kernel] {
+            Common::SetCurrentThreadName(std::string{"yuzu:HleService:" + service_name}.c_str());
+
+            // Wait for first request before trying to acquire a render context
+            {
+                std::unique_lock lock{queue_mutex};
+                condition.wait(lock, [this] { return stop || !requests.empty(); });
+            }
+
+            kernel.RegisterHostThread();
+
+            while (true) {
+                std::function<void()> task;
+
+                {
+                    std::unique_lock lock{queue_mutex};
+                    condition.wait(lock, [this] { return stop || !requests.empty(); });
+                    if (stop || requests.empty()) {
+                        return;
+                    }
+                    task = std::move(requests.front());
+                    requests.pop();
+                }
+
+                task();
+            }
+        });
+}
+
+void ServiceThread::Impl::QueueSyncRequest(ServerSession& session,
+                                           std::shared_ptr<HLERequestContext>&& context) {
+    {
+        std::unique_lock lock{queue_mutex};
+
+        // ServerSession owns the service thread, so we cannot caption a strong pointer here in the
+        // event that the ServerSession is terminated.
+        std::weak_ptr<ServerSession> weak_ptr{SharedFrom(&session)};
+        requests.emplace([weak_ptr, context{std::move(context)}]() {
+            if (auto strong_ptr = weak_ptr.lock()) {
+                strong_ptr->CompleteSyncRequest(*context);
+            }
+        });
+    }
+    condition.notify_one();
+}
+
+ServiceThread::Impl::~Impl() {
+    {
+        std::unique_lock lock{queue_mutex};
+        stop = true;
+    }
+    condition.notify_all();
+    for (std::thread& thread : threads) {
+        thread.join();
+    }
+}
+
+ServiceThread::ServiceThread(KernelCore& kernel, std::size_t num_threads, const std::string& name)
+    : impl{std::make_unique<Impl>(kernel, num_threads, name)} {}
+
+ServiceThread::~ServiceThread() = default;
+
+void ServiceThread::QueueSyncRequest(ServerSession& session,
+                                     std::shared_ptr<HLERequestContext>&& context) {
+    impl->QueueSyncRequest(session, std::move(context));
+}
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/service_thread.h b/src/core/hle/kernel/service_thread.h
new file mode 100644
index 000000000..025ab8fb5
--- /dev/null
+++ b/src/core/hle/kernel/service_thread.h
@@ -0,0 +1,28 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string>
+
+namespace Kernel {
+
+class HLERequestContext;
+class KernelCore;
+class ServerSession;
+
+class ServiceThread final {
+public:
+    explicit ServiceThread(KernelCore& kernel, std::size_t num_threads, const std::string& name);
+    ~ServiceThread();
+
+    void QueueSyncRequest(ServerSession& session, std::shared_ptr<HLERequestContext>&& context);
+
+private:
+    class Impl;
+    std::unique_ptr<Impl> impl;
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp
index bafd1ced7..de3ed25da 100644
--- a/src/core/hle/kernel/svc.cpp
+++ b/src/core/hle/kernel/svc.cpp
@@ -24,6 +24,8 @@
 #include "core/hle/kernel/client_session.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/memory/memory_block.h"
 #include "core/hle/kernel/memory/page_table.h"
@@ -32,7 +34,6 @@
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/readable_event.h"
 #include "core/hle/kernel/resource_limit.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/shared_memory.h"
 #include "core/hle/kernel/svc.h"
 #include "core/hle/kernel/svc_types.h"
@@ -234,8 +235,7 @@ static ResultCode SetMemoryAttribute(Core::System& system, VAddr address, u64 si
 
 static ResultCode SetMemoryAttribute32(Core::System& system, u32 address, u32 size, u32 mask,
                                        u32 attribute) {
-    return SetMemoryAttribute(system, static_cast<VAddr>(address), static_cast<std::size_t>(size),
-                              mask, attribute);
+    return SetMemoryAttribute(system, address, size, mask, attribute);
 }
 
 /// Maps a memory range into a different range.
@@ -255,8 +255,7 @@ static ResultCode MapMemory(Core::System& system, VAddr dst_addr, VAddr src_addr
 }
 
 static ResultCode MapMemory32(Core::System& system, u32 dst_addr, u32 src_addr, u32 size) {
-    return MapMemory(system, static_cast<VAddr>(dst_addr), static_cast<VAddr>(src_addr),
-                     static_cast<std::size_t>(size));
+    return MapMemory(system, dst_addr, src_addr, size);
 }
 
 /// Unmaps a region that was previously mapped with svcMapMemory
@@ -276,8 +275,7 @@ static ResultCode UnmapMemory(Core::System& system, VAddr dst_addr, VAddr src_ad
 }
 
 static ResultCode UnmapMemory32(Core::System& system, u32 dst_addr, u32 src_addr, u32 size) {
-    return UnmapMemory(system, static_cast<VAddr>(dst_addr), static_cast<VAddr>(src_addr),
-                       static_cast<std::size_t>(size));
+    return UnmapMemory(system, dst_addr, src_addr, size);
 }
 
 /// Connect to an OS service given the port name, returns the handle to the port to out
@@ -332,7 +330,8 @@ static ResultCode ConnectToNamedPort32(Core::System& system, Handle* out_handle,
 
 /// Makes a blocking IPC call to an OS service.
 static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
-    const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
+    auto& kernel = system.Kernel();
+    const auto& handle_table = kernel.CurrentProcess()->GetHandleTable();
     std::shared_ptr<ClientSession> session = handle_table.Get<ClientSession>(handle);
     if (!session) {
         LOG_ERROR(Kernel_SVC, "called with invalid handle=0x{:08X}", handle);
@@ -341,9 +340,9 @@ static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
 
     LOG_TRACE(Kernel_SVC, "called handle=0x{:08X}({})", handle, session->GetName());
 
-    auto thread = system.CurrentScheduler().GetCurrentThread();
+    auto thread = kernel.CurrentScheduler()->GetCurrentThread();
     {
-        SchedulerLock lock(system.Kernel());
+        KScopedSchedulerLock lock(kernel);
         thread->InvalidateHLECallback();
         thread->SetStatus(ThreadStatus::WaitIPC);
         session->SendSyncRequest(SharedFrom(thread), system.Memory(), system.CoreTiming());
@@ -352,12 +351,12 @@ static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
     if (thread->HasHLECallback()) {
         Handle event_handle = thread->GetHLETimeEvent();
         if (event_handle != InvalidHandle) {
-            auto& time_manager = system.Kernel().TimeManager();
+            auto& time_manager = kernel.TimeManager();
             time_manager.UnscheduleTimeEvent(event_handle);
         }
 
         {
-            SchedulerLock lock(system.Kernel());
+            KScopedSchedulerLock lock(kernel);
             auto* sync_object = thread->GetHLESyncObject();
             sync_object->RemoveWaitingThread(SharedFrom(thread));
         }
@@ -531,8 +530,7 @@ static ResultCode ArbitrateLock(Core::System& system, Handle holding_thread_hand
 
 static ResultCode ArbitrateLock32(Core::System& system, Handle holding_thread_handle,
                                   u32 mutex_addr, Handle requesting_thread_handle) {
-    return ArbitrateLock(system, holding_thread_handle, static_cast<VAddr>(mutex_addr),
-                         requesting_thread_handle);
+    return ArbitrateLock(system, holding_thread_handle, mutex_addr, requesting_thread_handle);
 }
 
 /// Unlock a mutex
@@ -555,7 +553,7 @@ static ResultCode ArbitrateUnlock(Core::System& system, VAddr mutex_addr) {
 }
 
 static ResultCode ArbitrateUnlock32(Core::System& system, u32 mutex_addr) {
-    return ArbitrateUnlock(system, static_cast<VAddr>(mutex_addr));
+    return ArbitrateUnlock(system, mutex_addr);
 }
 
 enum class BreakType : u32 {
@@ -658,7 +656,6 @@ static void Break(Core::System& system, u32 reason, u64 info1, u64 info2) {
         info2, has_dumped_buffer ? std::make_optional(debug_buffer) : std::nullopt);
 
     if (!break_reason.signal_debugger) {
-        SchedulerLock lock(system.Kernel());
         LOG_CRITICAL(
             Debug_Emulated,
             "Emulated program broke execution! reason=0x{:016X}, info1=0x{:016X}, info2=0x{:016X}",
@@ -666,22 +663,18 @@ static void Break(Core::System& system, u32 reason, u64 info1, u64 info2) {
 
         handle_debug_buffer(info1, info2);
 
-        auto* const current_thread = system.CurrentScheduler().GetCurrentThread();
+        auto* const current_thread = system.Kernel().CurrentScheduler()->GetCurrentThread();
         const auto thread_processor_id = current_thread->GetProcessorID();
         system.ArmInterface(static_cast<std::size_t>(thread_processor_id)).LogBacktrace();
-
-        // Kill the current thread
-        system.Kernel().ExceptionalExit();
-        current_thread->Stop();
     }
 }
 
 static void Break32(Core::System& system, u32 reason, u32 info1, u32 info2) {
-    Break(system, reason, static_cast<u64>(info1), static_cast<u64>(info2));
+    Break(system, reason, info1, info2);
 }
 
 /// Used to output a message on a debug hardware unit - does nothing on a retail unit
-static void OutputDebugString([[maybe_unused]] Core::System& system, VAddr address, u64 len) {
+static void OutputDebugString(Core::System& system, VAddr address, u64 len) {
     if (len == 0) {
         return;
     }
@@ -922,7 +915,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
         }
 
         const auto& core_timing = system.CoreTiming();
-        const auto& scheduler = system.CurrentScheduler();
+        const auto& scheduler = *system.Kernel().CurrentScheduler();
         const auto* const current_thread = scheduler.GetCurrentThread();
         const bool same_thread = current_thread == thread.get();
 
@@ -948,7 +941,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
 
 static ResultCode GetInfo32(Core::System& system, u32* result_low, u32* result_high, u32 sub_id_low,
                             u32 info_id, u32 handle, u32 sub_id_high) {
-    const u64 sub_id{static_cast<u64>(sub_id_low | (static_cast<u64>(sub_id_high) << 32))};
+    const u64 sub_id{u64{sub_id_low} | (u64{sub_id_high} << 32)};
     u64 res_value{};
 
     const ResultCode result{GetInfo(system, &res_value, info_id, handle, sub_id)};
@@ -1009,7 +1002,7 @@ static ResultCode MapPhysicalMemory(Core::System& system, VAddr addr, u64 size)
 }
 
 static ResultCode MapPhysicalMemory32(Core::System& system, u32 addr, u32 size) {
-    return MapPhysicalMemory(system, static_cast<VAddr>(addr), static_cast<std::size_t>(size));
+    return MapPhysicalMemory(system, addr, size);
 }
 
 /// Unmaps memory previously mapped via MapPhysicalMemory
@@ -1063,7 +1056,7 @@ static ResultCode UnmapPhysicalMemory(Core::System& system, VAddr addr, u64 size
 }
 
 static ResultCode UnmapPhysicalMemory32(Core::System& system, u32 addr, u32 size) {
-    return UnmapPhysicalMemory(system, static_cast<VAddr>(addr), static_cast<std::size_t>(size));
+    return UnmapPhysicalMemory(system, addr, size);
 }
 
 /// Sets the thread activity
@@ -1090,7 +1083,7 @@ static ResultCode SetThreadActivity(Core::System& system, Handle handle, u32 act
         return ERR_INVALID_HANDLE;
     }
 
-    if (thread.get() == system.CurrentScheduler().GetCurrentThread()) {
+    if (thread.get() == system.Kernel().CurrentScheduler()->GetCurrentThread()) {
         LOG_ERROR(Kernel_SVC, "The thread handle specified is the current running thread");
         return ERR_BUSY;
     }
@@ -1123,7 +1116,7 @@ static ResultCode GetThreadContext(Core::System& system, VAddr thread_context, H
         return ERR_INVALID_HANDLE;
     }
 
-    if (thread.get() == system.CurrentScheduler().GetCurrentThread()) {
+    if (thread.get() == system.Kernel().CurrentScheduler()->GetCurrentThread()) {
         LOG_ERROR(Kernel_SVC, "The thread handle specified is the current running thread");
         return ERR_BUSY;
     }
@@ -1144,7 +1137,7 @@ static ResultCode GetThreadContext(Core::System& system, VAddr thread_context, H
 }
 
 static ResultCode GetThreadContext32(Core::System& system, u32 thread_context, Handle handle) {
-    return GetThreadContext(system, static_cast<VAddr>(thread_context), handle);
+    return GetThreadContext(system, thread_context, handle);
 }
 
 /// Gets the priority for the specified thread
@@ -1281,8 +1274,7 @@ static ResultCode MapSharedMemory(Core::System& system, Handle shared_memory_han
 
 static ResultCode MapSharedMemory32(Core::System& system, Handle shared_memory_handle, u32 addr,
                                     u32 size, u32 permissions) {
-    return MapSharedMemory(system, shared_memory_handle, static_cast<VAddr>(addr),
-                           static_cast<std::size_t>(size), permissions);
+    return MapSharedMemory(system, shared_memory_handle, addr, size, permissions);
 }
 
 static ResultCode QueryProcessMemory(Core::System& system, VAddr memory_info_address,
@@ -1480,7 +1472,7 @@ static void ExitProcess(Core::System& system) {
     current_process->PrepareForTermination();
 
     // Kill the current thread
-    system.CurrentScheduler().GetCurrentThread()->Stop();
+    system.Kernel().CurrentScheduler()->GetCurrentThread()->Stop();
 }
 
 static void ExitProcess32(Core::System& system) {
@@ -1552,8 +1544,7 @@ static ResultCode CreateThread(Core::System& system, Handle* out_handle, VAddr e
 
 static ResultCode CreateThread32(Core::System& system, Handle* out_handle, u32 priority,
                                  u32 entry_point, u32 arg, u32 stack_top, s32 processor_id) {
-    return CreateThread(system, out_handle, static_cast<VAddr>(entry_point), static_cast<u64>(arg),
-                        static_cast<VAddr>(stack_top), priority, processor_id);
+    return CreateThread(system, out_handle, entry_point, arg, stack_top, priority, processor_id);
 }
 
 /// Starts the thread for the provided handle
@@ -1581,8 +1572,8 @@ static ResultCode StartThread32(Core::System& system, Handle thread_handle) {
 static void ExitThread(Core::System& system) {
     LOG_DEBUG(Kernel_SVC, "called, pc=0x{:08X}", system.CurrentArmInterface().GetPC());
 
-    auto* const current_thread = system.CurrentScheduler().GetCurrentThread();
-    system.GlobalScheduler().RemoveThread(SharedFrom(current_thread));
+    auto* const current_thread = system.Kernel().CurrentScheduler()->GetCurrentThread();
+    system.GlobalSchedulerContext().RemoveThread(SharedFrom(current_thread));
     current_thread->Stop();
 }
 
@@ -1592,53 +1583,39 @@ static void ExitThread32(Core::System& system) {
 
 /// Sleep the current thread
 static void SleepThread(Core::System& system, s64 nanoseconds) {
-    LOG_DEBUG(Kernel_SVC, "called nanoseconds={}", nanoseconds);
+    LOG_TRACE(Kernel_SVC, "called nanoseconds={}", nanoseconds);
 
     enum class SleepType : s64 {
-        YieldWithoutLoadBalancing = 0,
-        YieldWithLoadBalancing = -1,
+        YieldWithoutCoreMigration = 0,
+        YieldWithCoreMigration = -1,
         YieldAndWaitForLoadBalancing = -2,
     };
 
-    auto& scheduler = system.CurrentScheduler();
-    auto* const current_thread = scheduler.GetCurrentThread();
-    bool is_redundant = false;
-
+    auto& scheduler = *system.Kernel().CurrentScheduler();
     if (nanoseconds <= 0) {
         switch (static_cast<SleepType>(nanoseconds)) {
-        case SleepType::YieldWithoutLoadBalancing: {
-            auto pair = current_thread->YieldSimple();
-            is_redundant = pair.second;
+        case SleepType::YieldWithoutCoreMigration: {
+            scheduler.YieldWithoutCoreMigration();
             break;
         }
-        case SleepType::YieldWithLoadBalancing: {
-            auto pair = current_thread->YieldAndBalanceLoad();
-            is_redundant = pair.second;
+        case SleepType::YieldWithCoreMigration: {
+            scheduler.YieldWithCoreMigration();
             break;
         }
         case SleepType::YieldAndWaitForLoadBalancing: {
-            auto pair = current_thread->YieldAndWaitForLoadBalancing();
-            is_redundant = pair.second;
+            scheduler.YieldToAnyThread();
             break;
         }
         default:
             UNREACHABLE_MSG("Unimplemented sleep yield type '{:016X}'!", nanoseconds);
         }
     } else {
-        current_thread->Sleep(nanoseconds);
-    }
-
-    if (is_redundant && !system.Kernel().IsMulticore()) {
-        system.Kernel().ExitSVCProfile();
-        system.CoreTiming().AddTicks(1000U);
-        system.GetCpuManager().PreemptSingleCore();
-        system.Kernel().EnterSVCProfile();
+        scheduler.GetCurrentThread()->Sleep(nanoseconds);
     }
 }
 
 static void SleepThread32(Core::System& system, u32 nanoseconds_low, u32 nanoseconds_high) {
-    const s64 nanoseconds = static_cast<s64>(static_cast<u64>(nanoseconds_low) |
-                                             (static_cast<u64>(nanoseconds_high) << 32));
+    const auto nanoseconds = static_cast<s64>(u64{nanoseconds_low} | (u64{nanoseconds_high} << 32));
     SleepThread(system, nanoseconds);
 }
 
@@ -1668,10 +1645,10 @@ static ResultCode WaitProcessWideKeyAtomic(Core::System& system, VAddr mutex_add
     ASSERT(condition_variable_addr == Common::AlignDown(condition_variable_addr, 4));
     auto& kernel = system.Kernel();
     Handle event_handle;
-    Thread* current_thread = system.CurrentScheduler().GetCurrentThread();
-    auto* const current_process = system.Kernel().CurrentProcess();
+    Thread* current_thread = kernel.CurrentScheduler()->GetCurrentThread();
+    auto* const current_process = kernel.CurrentProcess();
     {
-        SchedulerLockAndSleep lock(kernel, event_handle, current_thread, nano_seconds);
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, current_thread, nano_seconds);
         const auto& handle_table = current_process->GetHandleTable();
         std::shared_ptr<Thread> thread = handle_table.Get<Thread>(thread_handle);
         ASSERT(thread);
@@ -1707,7 +1684,7 @@ static ResultCode WaitProcessWideKeyAtomic(Core::System& system, VAddr mutex_add
     }
 
     {
-        SchedulerLock lock(kernel);
+        KScopedSchedulerLock lock(kernel);
 
         auto* owner = current_thread->GetLockOwner();
         if (owner != nullptr) {
@@ -1724,10 +1701,8 @@ static ResultCode WaitProcessWideKeyAtomic(Core::System& system, VAddr mutex_add
 static ResultCode WaitProcessWideKeyAtomic32(Core::System& system, u32 mutex_addr,
                                              u32 condition_variable_addr, Handle thread_handle,
                                              u32 nanoseconds_low, u32 nanoseconds_high) {
-    const s64 nanoseconds =
-        static_cast<s64>(nanoseconds_low | (static_cast<u64>(nanoseconds_high) << 32));
-    return WaitProcessWideKeyAtomic(system, static_cast<VAddr>(mutex_addr),
-                                    static_cast<VAddr>(condition_variable_addr), thread_handle,
+    const auto nanoseconds = static_cast<s64>(nanoseconds_low | (u64{nanoseconds_high} << 32));
+    return WaitProcessWideKeyAtomic(system, mutex_addr, condition_variable_addr, thread_handle,
                                     nanoseconds);
 }
 
@@ -1740,7 +1715,7 @@ static void SignalProcessWideKey(Core::System& system, VAddr condition_variable_
 
     // Retrieve a list of all threads that are waiting for this condition variable.
     auto& kernel = system.Kernel();
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
     auto* const current_process = kernel.CurrentProcess();
     std::vector<std::shared_ptr<Thread>> waiting_threads =
         current_process->GetConditionVariableThreads(condition_variable_addr);
@@ -1833,8 +1808,8 @@ static ResultCode WaitForAddress(Core::System& system, VAddr address, u32 type,
 
 static ResultCode WaitForAddress32(Core::System& system, u32 address, u32 type, s32 value,
                                    u32 timeout_low, u32 timeout_high) {
-    s64 timeout = static_cast<s64>(timeout_low | (static_cast<u64>(timeout_high) << 32));
-    return WaitForAddress(system, static_cast<VAddr>(address), type, value, timeout);
+    const auto timeout = static_cast<s64>(timeout_low | (u64{timeout_high} << 32));
+    return WaitForAddress(system, address, type, value, timeout);
 }
 
 // Signals to an address (via Address Arbiter)
@@ -1862,7 +1837,7 @@ static ResultCode SignalToAddress(Core::System& system, VAddr address, u32 type,
 
 static ResultCode SignalToAddress32(Core::System& system, u32 address, u32 type, s32 value,
                                     s32 num_to_wake) {
-    return SignalToAddress(system, static_cast<VAddr>(address), type, value, num_to_wake);
+    return SignalToAddress(system, address, type, value, num_to_wake);
 }
 
 static void KernelDebug([[maybe_unused]] Core::System& system,
@@ -1893,7 +1868,7 @@ static u64 GetSystemTick(Core::System& system) {
 }
 
 static void GetSystemTick32(Core::System& system, u32* time_low, u32* time_high) {
-    u64 time = GetSystemTick(system);
+    const auto time = GetSystemTick(system);
     *time_low = static_cast<u32>(time);
     *time_high = static_cast<u32>(time >> 32);
 }
@@ -1984,8 +1959,7 @@ static ResultCode CreateTransferMemory(Core::System& system, Handle* handle, VAd
 
 static ResultCode CreateTransferMemory32(Core::System& system, Handle* handle, u32 addr, u32 size,
                                          u32 permissions) {
-    return CreateTransferMemory(system, handle, static_cast<VAddr>(addr),
-                                static_cast<std::size_t>(size), permissions);
+    return CreateTransferMemory(system, handle, addr, size, permissions);
 }
 
 static ResultCode GetThreadCoreMask(Core::System& system, Handle thread_handle, u32* core,
@@ -2003,7 +1977,7 @@ static ResultCode GetThreadCoreMask(Core::System& system, Handle thread_handle,
     }
 
     *core = thread->GetIdealCore();
-    *mask = thread->GetAffinityMask();
+    *mask = thread->GetAffinityMask().GetAffinityMask();
 
     return RESULT_SUCCESS;
 }
@@ -2075,8 +2049,7 @@ static ResultCode SetThreadCoreMask(Core::System& system, Handle thread_handle,
 
 static ResultCode SetThreadCoreMask32(Core::System& system, Handle thread_handle, u32 core,
                                       u32 affinity_mask_low, u32 affinity_mask_high) {
-    const u64 affinity_mask =
-        static_cast<u64>(affinity_mask_low) | (static_cast<u64>(affinity_mask_high) << 32);
+    const auto affinity_mask = u64{affinity_mask_low} | (u64{affinity_mask_high} << 32);
     return SetThreadCoreMask(system, thread_handle, core, affinity_mask);
 }
 
@@ -2341,9 +2314,10 @@ static ResultCode GetThreadList(Core::System& system, u32* out_num_threads, VAdd
     return RESULT_SUCCESS;
 }
 
-static ResultCode FlushProcessDataCache32(Core::System& system, Handle handle, u32 address,
-                                          u32 size) {
-    // Note(Blinkhawk): For emulation purposes of the data cache this is mostly a nope
+static ResultCode FlushProcessDataCache32([[maybe_unused]] Core::System& system,
+                                          [[maybe_unused]] Handle handle,
+                                          [[maybe_unused]] u32 address, [[maybe_unused]] u32 size) {
+    // Note(Blinkhawk): For emulation purposes of the data cache this is mostly a no-op,
     // as all emulation is done in the same cache level in host architecture, thus data cache
     // does not need flushing.
     LOG_DEBUG(Kernel_SVC, "called");
@@ -2639,6 +2613,9 @@ void Call(Core::System& system, u32 immediate) {
     auto& kernel = system.Kernel();
     kernel.EnterSVCProfile();
 
+    auto* thread = kernel.CurrentScheduler()->GetCurrentThread();
+    thread->SetContinuousOnSVC(true);
+
     const FunctionDef* info = system.CurrentProcess()->Is64BitProcess() ? GetSVCInfo64(immediate)
                                                                         : GetSVCInfo32(immediate);
     if (info) {
@@ -2652,6 +2629,12 @@ void Call(Core::System& system, u32 immediate) {
     }
 
     kernel.ExitSVCProfile();
+
+    if (!thread->IsContinuousOnSVC()) {
+        auto* host_context = thread->GetHostContext().get();
+        host_context->Rewind();
+    }
+
     system.EnterDynarmicProfile();
 }
 
diff --git a/src/core/hle/kernel/svc_types.h b/src/core/hle/kernel/svc_types.h
index 986724beb..11e1d8e2d 100644
--- a/src/core/hle/kernel/svc_types.h
+++ b/src/core/hle/kernel/svc_types.h
@@ -23,8 +23,8 @@ enum class MemoryState : u32 {
     Ipc = 0x0A,
     Stack = 0x0B,
     ThreadLocal = 0x0C,
-    Transfered = 0x0D,
-    SharedTransfered = 0x0E,
+    Transferred = 0x0D,
+    SharedTransferred = 0x0E,
     SharedCode = 0x0F,
     Inaccessible = 0x10,
     NonSecureIpc = 0x11,
diff --git a/src/core/hle/kernel/synchronization.cpp b/src/core/hle/kernel/synchronization.cpp
index 8b875d853..d3f520ea2 100644
--- a/src/core/hle/kernel/synchronization.cpp
+++ b/src/core/hle/kernel/synchronization.cpp
@@ -5,8 +5,9 @@
 #include "core/core.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/synchronization.h"
 #include "core/hle/kernel/synchronization_object.h"
 #include "core/hle/kernel/thread.h"
@@ -18,7 +19,7 @@ Synchronization::Synchronization(Core::System& system) : system{system} {}
 
 void Synchronization::SignalObject(SynchronizationObject& obj) const {
     auto& kernel = system.Kernel();
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
     if (obj.IsSignaled()) {
         for (auto thread : obj.GetWaitingThreads()) {
             if (thread->GetSchedulingStatus() == ThreadSchedStatus::Paused) {
@@ -37,10 +38,10 @@ void Synchronization::SignalObject(SynchronizationObject& obj) const {
 std::pair<ResultCode, Handle> Synchronization::WaitFor(
     std::vector<std::shared_ptr<SynchronizationObject>>& sync_objects, s64 nano_seconds) {
     auto& kernel = system.Kernel();
-    auto* const thread = system.CurrentScheduler().GetCurrentThread();
+    auto* const thread = kernel.CurrentScheduler()->GetCurrentThread();
     Handle event_handle = InvalidHandle;
     {
-        SchedulerLockAndSleep lock(kernel, event_handle, thread, nano_seconds);
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, thread, nano_seconds);
         const auto itr =
             std::find_if(sync_objects.begin(), sync_objects.end(),
                          [thread](const std::shared_ptr<SynchronizationObject>& object) {
@@ -89,7 +90,7 @@ std::pair<ResultCode, Handle> Synchronization::WaitFor(
     }
 
     {
-        SchedulerLock lock(kernel);
+        KScopedSchedulerLock lock(kernel);
         ResultCode signaling_result = thread->GetSignalingResult();
         SynchronizationObject* signaling_object = thread->GetSignalingObject();
         thread->SetSynchronizationObjects(nullptr);
diff --git a/src/core/hle/kernel/synchronization_object.h b/src/core/hle/kernel/synchronization_object.h
index f89b24204..7408ed51f 100644
--- a/src/core/hle/kernel/synchronization_object.h
+++ b/src/core/hle/kernel/synchronization_object.h
@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <atomic>
 #include <memory>
 #include <vector>
 
@@ -56,7 +57,7 @@ public:
     void ClearWaitingThreads();
 
 protected:
-    bool is_signaled{}; // Tells if this sync object is signalled;
+    std::atomic_bool is_signaled{}; // Tells if this sync object is signaled
 
 private:
     /// Threads waiting for this object to become available
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index d132aba34..a4f9e0d97 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -12,17 +12,16 @@
 #include "common/fiber.h"
 #include "common/logging/log.h"
 #include "common/thread_queue_list.h"
-#include "core/arm/arm_interface.h"
-#include "core/arm/unicorn/arm_unicorn.h"
 #include "core/core.h"
 #include "core/cpu_manager.h"
 #include "core/hardware_properties.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
+#include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/time_manager.h"
 #include "core/hle/result.h"
@@ -52,7 +51,7 @@ Thread::~Thread() = default;
 
 void Thread::Stop() {
     {
-        SchedulerLock lock(kernel);
+        KScopedSchedulerLock lock(kernel);
         SetStatus(ThreadStatus::Dead);
         Signal();
         kernel.GlobalHandleTable().Close(global_handle);
@@ -63,14 +62,13 @@ void Thread::Stop() {
             // Mark the TLS slot in the thread's page as free.
             owner_process->FreeTLSRegion(tls_address);
         }
-        arm_interface.reset();
         has_exited = true;
     }
     global_handle = 0;
 }
 
 void Thread::ResumeFromWait() {
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
     switch (status) {
     case ThreadStatus::Paused:
     case ThreadStatus::WaitSynch:
@@ -91,10 +89,6 @@ void Thread::ResumeFromWait() {
         // before actually resuming. We can ignore subsequent wakeups if the thread status has
         // already been set to ThreadStatus::Ready.
         return;
-
-    case ThreadStatus::Running:
-        DEBUG_ASSERT_MSG(false, "Thread with object id {} has already resumed.", GetObjectId());
-        return;
     case ThreadStatus::Dead:
         // This should never happen, as threads must complete before being stopped.
         DEBUG_ASSERT_MSG(false, "Thread with object id {} cannot be resumed because it's DEAD.",
@@ -106,19 +100,18 @@ void Thread::ResumeFromWait() {
 }
 
 void Thread::OnWakeUp() {
-    SchedulerLock lock(kernel);
-
+    KScopedSchedulerLock lock(kernel);
     SetStatus(ThreadStatus::Ready);
 }
 
 ResultCode Thread::Start() {
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
     SetStatus(ThreadStatus::Ready);
     return RESULT_SUCCESS;
 }
 
 void Thread::CancelWait() {
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
     if (GetSchedulingStatus() != ThreadSchedStatus::Paused || !is_waiting_on_sync) {
         is_sync_cancelled = true;
         return;
@@ -193,12 +186,14 @@ ResultVal<std::shared_ptr<Thread>> Thread::Create(Core::System& system, ThreadTy
     thread->status = ThreadStatus::Dormant;
     thread->entry_point = entry_point;
     thread->stack_top = stack_top;
+    thread->disable_count = 1;
     thread->tpidr_el0 = 0;
     thread->nominal_priority = thread->current_priority = priority;
-    thread->last_running_ticks = 0;
+    thread->schedule_count = -1;
+    thread->last_scheduled_tick = 0;
     thread->processor_id = processor_id;
     thread->ideal_core = processor_id;
-    thread->affinity_mask = 1ULL << processor_id;
+    thread->affinity_mask.SetAffinity(processor_id, true);
     thread->wait_objects = nullptr;
     thread->mutex_wait_address = 0;
     thread->condvar_wait_address = 0;
@@ -208,7 +203,7 @@ ResultVal<std::shared_ptr<Thread>> Thread::Create(Core::System& system, ThreadTy
     thread->owner_process = owner_process;
     thread->type = type_flags;
     if ((type_flags & THREADTYPE_IDLE) == 0) {
-        auto& scheduler = kernel.GlobalScheduler();
+        auto& scheduler = kernel.GlobalSchedulerContext();
         scheduler.AddThread(thread);
     }
     if (owner_process) {
@@ -217,33 +212,10 @@ ResultVal<std::shared_ptr<Thread>> Thread::Create(Core::System& system, ThreadTy
     } else {
         thread->tls_address = 0;
     }
+
     // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used
     // to initialize the context
-    thread->arm_interface.reset();
     if ((type_flags & THREADTYPE_HLE) == 0) {
-#ifdef ARCHITECTURE_x86_64
-        if (owner_process && !owner_process->Is64BitProcess()) {
-            thread->arm_interface = std::make_unique<Core::ARM_Dynarmic_32>(
-                system, kernel.Interrupts(), kernel.IsMulticore(), kernel.GetExclusiveMonitor(),
-                processor_id);
-        } else {
-            thread->arm_interface = std::make_unique<Core::ARM_Dynarmic_64>(
-                system, kernel.Interrupts(), kernel.IsMulticore(), kernel.GetExclusiveMonitor(),
-                processor_id);
-        }
-
-#else
-        if (owner_process && !owner_process->Is64BitProcess()) {
-            thread->arm_interface = std::make_shared<Core::ARM_Unicorn>(
-                system, kernel.Interrupts(), kernel.IsMulticore(), ARM_Unicorn::Arch::AArch32,
-                processor_id);
-        } else {
-            thread->arm_interface = std::make_shared<Core::ARM_Unicorn>(
-                system, kernel.Interrupts(), kernel.IsMulticore(), ARM_Unicorn::Arch::AArch64,
-                processor_id);
-        }
-        LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
-#endif
         ResetThreadContext32(thread->context_32, static_cast<u32>(stack_top),
                              static_cast<u32>(entry_point), static_cast<u32>(arg));
         ResetThreadContext64(thread->context_64, stack_top, entry_point, arg);
@@ -255,7 +227,7 @@ ResultVal<std::shared_ptr<Thread>> Thread::Create(Core::System& system, ThreadTy
 }
 
 void Thread::SetPriority(u32 priority) {
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
     ASSERT_MSG(priority <= THREADPRIO_LOWEST && priority >= THREADPRIO_HIGHEST,
                "Invalid priority value.");
     nominal_priority = priority;
@@ -279,14 +251,6 @@ VAddr Thread::GetCommandBufferAddress() const {
     return GetTLSAddress() + command_header_offset;
 }
 
-Core::ARM_Interface& Thread::ArmInterface() {
-    return *arm_interface;
-}
-
-const Core::ARM_Interface& Thread::ArmInterface() const {
-    return *arm_interface;
-}
-
 void Thread::SetStatus(ThreadStatus new_status) {
     if (new_status == status) {
         return;
@@ -294,7 +258,6 @@ void Thread::SetStatus(ThreadStatus new_status) {
 
     switch (new_status) {
     case ThreadStatus::Ready:
-    case ThreadStatus::Running:
         SetSchedulingStatus(ThreadSchedStatus::Runnable);
         break;
     case ThreadStatus::Dormant:
@@ -401,7 +364,7 @@ bool Thread::InvokeHLECallback(std::shared_ptr<Thread> thread) {
 }
 
 ResultCode Thread::SetActivity(ThreadActivity value) {
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
 
     auto sched_status = GetSchedulingStatus();
 
@@ -430,7 +393,7 @@ ResultCode Thread::SetActivity(ThreadActivity value) {
 ResultCode Thread::Sleep(s64 nanoseconds) {
     Handle event_handle{};
     {
-        SchedulerLockAndSleep lock(kernel, event_handle, this, nanoseconds);
+        KScopedSchedulerLockAndSleep lock(kernel, event_handle, this, nanoseconds);
         SetStatus(ThreadStatus::WaitSleep);
     }
 
@@ -441,39 +404,12 @@ ResultCode Thread::Sleep(s64 nanoseconds) {
     return RESULT_SUCCESS;
 }
 
-std::pair<ResultCode, bool> Thread::YieldSimple() {
-    bool is_redundant = false;
-    {
-        SchedulerLock lock(kernel);
-        is_redundant = kernel.GlobalScheduler().YieldThread(this);
-    }
-    return {RESULT_SUCCESS, is_redundant};
-}
-
-std::pair<ResultCode, bool> Thread::YieldAndBalanceLoad() {
-    bool is_redundant = false;
-    {
-        SchedulerLock lock(kernel);
-        is_redundant = kernel.GlobalScheduler().YieldThreadAndBalanceLoad(this);
-    }
-    return {RESULT_SUCCESS, is_redundant};
-}
-
-std::pair<ResultCode, bool> Thread::YieldAndWaitForLoadBalancing() {
-    bool is_redundant = false;
-    {
-        SchedulerLock lock(kernel);
-        is_redundant = kernel.GlobalScheduler().YieldThreadAndWaitForLoadBalancing(this);
-    }
-    return {RESULT_SUCCESS, is_redundant};
-}
-
 void Thread::AddSchedulingFlag(ThreadSchedFlags flag) {
     const u32 old_state = scheduling_state;
     pausing_state |= static_cast<u32>(flag);
     const u32 base_scheduling = static_cast<u32>(GetSchedulingStatus());
     scheduling_state = base_scheduling | pausing_state;
-    kernel.GlobalScheduler().AdjustSchedulingOnStatus(this, old_state);
+    KScheduler::OnThreadStateChanged(kernel, this, old_state);
 }
 
 void Thread::RemoveSchedulingFlag(ThreadSchedFlags flag) {
@@ -481,23 +417,24 @@ void Thread::RemoveSchedulingFlag(ThreadSchedFlags flag) {
     pausing_state &= ~static_cast<u32>(flag);
     const u32 base_scheduling = static_cast<u32>(GetSchedulingStatus());
     scheduling_state = base_scheduling | pausing_state;
-    kernel.GlobalScheduler().AdjustSchedulingOnStatus(this, old_state);
+    KScheduler::OnThreadStateChanged(kernel, this, old_state);
 }
 
 void Thread::SetSchedulingStatus(ThreadSchedStatus new_status) {
     const u32 old_state = scheduling_state;
     scheduling_state = (scheduling_state & static_cast<u32>(ThreadSchedMasks::HighMask)) |
                        static_cast<u32>(new_status);
-    kernel.GlobalScheduler().AdjustSchedulingOnStatus(this, old_state);
+    KScheduler::OnThreadStateChanged(kernel, this, old_state);
 }
 
 void Thread::SetCurrentPriority(u32 new_priority) {
     const u32 old_priority = std::exchange(current_priority, new_priority);
-    kernel.GlobalScheduler().AdjustSchedulingOnPriority(this, old_priority);
+    KScheduler::OnThreadPriorityChanged(kernel, this, kernel.CurrentScheduler()->GetCurrentThread(),
+                                        old_priority);
 }
 
 ResultCode Thread::SetCoreAndAffinityMask(s32 new_core, u64 new_affinity_mask) {
-    SchedulerLock lock(kernel);
+    KScopedSchedulerLock lock(kernel);
     const auto HighestSetCore = [](u64 mask, u32 max_cores) {
         for (s32 core = static_cast<s32>(max_cores - 1); core >= 0; core--) {
             if (((mask >> core) & 1) != 0) {
@@ -518,20 +455,21 @@ ResultCode Thread::SetCoreAndAffinityMask(s32 new_core, u64 new_affinity_mask) {
     }
     if (use_override) {
         ideal_core_override = new_core;
-        affinity_mask_override = new_affinity_mask;
     } else {
-        const u64 old_affinity_mask = std::exchange(affinity_mask, new_affinity_mask);
+        const auto old_affinity_mask = affinity_mask;
+        affinity_mask.SetAffinityMask(new_affinity_mask);
         ideal_core = new_core;
-        if (old_affinity_mask != new_affinity_mask) {
+        if (old_affinity_mask.GetAffinityMask() != new_affinity_mask) {
             const s32 old_core = processor_id;
-            if (processor_id >= 0 && ((affinity_mask >> processor_id) & 1) == 0) {
+            if (processor_id >= 0 && !affinity_mask.GetAffinity(processor_id)) {
                 if (static_cast<s32>(ideal_core) < 0) {
-                    processor_id = HighestSetCore(affinity_mask, Core::Hardware::NUM_CPU_CORES);
+                    processor_id = HighestSetCore(affinity_mask.GetAffinityMask(),
+                                                  Core::Hardware::NUM_CPU_CORES);
                 } else {
                     processor_id = ideal_core;
                 }
             }
-            kernel.GlobalScheduler().AdjustSchedulingOnAffinity(this, old_affinity_mask, old_core);
+            KScheduler::OnThreadAffinityMaskChanged(kernel, this, old_affinity_mask, old_core);
         }
     }
     return RESULT_SUCCESS;
diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h
index 8daf79fac..11ef29888 100644
--- a/src/core/hle/kernel/thread.h
+++ b/src/core/hle/kernel/thread.h
@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <array>
 #include <functional>
 #include <string>
 #include <utility>
@@ -12,6 +13,7 @@
 #include "common/common_types.h"
 #include "common/spin_lock.h"
 #include "core/arm/arm_interface.h"
+#include "core/hle/kernel/k_affinity_mask.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/synchronization_object.h"
 #include "core/hle/result.h"
@@ -27,10 +29,10 @@ class System;
 
 namespace Kernel {
 
-class GlobalScheduler;
+class GlobalSchedulerContext;
 class KernelCore;
 class Process;
-class Scheduler;
+class KScheduler;
 
 enum ThreadPriority : u32 {
     THREADPRIO_HIGHEST = 0,            ///< Highest thread priority
@@ -72,7 +74,6 @@ enum ThreadProcessorId : s32 {
 };
 
 enum class ThreadStatus {
-    Running,      ///< Currently running
     Ready,        ///< Ready to run
     Paused,       ///< Paused by SetThreadActivity or debug
     WaitHLEEvent, ///< Waiting for hle event to finish
@@ -248,10 +249,6 @@ public:
 
     void SetSynchronizationResults(SynchronizationObject* object, ResultCode result);
 
-    Core::ARM_Interface& ArmInterface();
-
-    const Core::ARM_Interface& ArmInterface() const;
-
     SynchronizationObject* GetSignalingObject() const {
         return signaling_object;
     }
@@ -350,8 +347,12 @@ public:
 
     void SetStatus(ThreadStatus new_status);
 
-    u64 GetLastRunningTicks() const {
-        return last_running_ticks;
+    s64 GetLastScheduledTick() const {
+        return this->last_scheduled_tick;
+    }
+
+    void SetLastScheduledTick(s64 tick) {
+        this->last_scheduled_tick = tick;
     }
 
     u64 GetTotalCPUTimeTicks() const {
@@ -366,10 +367,18 @@ public:
         return processor_id;
     }
 
+    s32 GetActiveCore() const {
+        return GetProcessorID();
+    }
+
     void SetProcessorID(s32 new_core) {
         processor_id = new_core;
     }
 
+    void SetActiveCore(s32 new_core) {
+        processor_id = new_core;
+    }
+
     Process* GetOwnerProcess() {
         return owner_process;
     }
@@ -474,7 +483,7 @@ public:
         return ideal_core;
     }
 
-    u64 GetAffinityMask() const {
+    const KAffinityMask& GetAffinityMask() const {
         return affinity_mask;
     }
 
@@ -483,21 +492,12 @@ public:
     /// Sleeps this thread for the given amount of nanoseconds.
     ResultCode Sleep(s64 nanoseconds);
 
-    /// Yields this thread without rebalancing loads.
-    std::pair<ResultCode, bool> YieldSimple();
-
-    /// Yields this thread and does a load rebalancing.
-    std::pair<ResultCode, bool> YieldAndBalanceLoad();
-
-    /// Yields this thread and if the core is left idle, loads are rebalanced
-    std::pair<ResultCode, bool> YieldAndWaitForLoadBalancing();
-
-    void IncrementYieldCount() {
-        yield_count++;
+    s64 GetYieldScheduleCount() const {
+        return this->schedule_count;
     }
 
-    u64 GetYieldCount() const {
-        return yield_count;
+    void SetYieldScheduleCount(s64 count) {
+        this->schedule_count = count;
     }
 
     ThreadSchedStatus GetSchedulingStatus() const {
@@ -573,9 +573,59 @@ public:
         return has_exited;
     }
 
+    class QueueEntry {
+    public:
+        constexpr QueueEntry() = default;
+
+        constexpr void Initialize() {
+            this->prev = nullptr;
+            this->next = nullptr;
+        }
+
+        constexpr Thread* GetPrev() const {
+            return this->prev;
+        }
+        constexpr Thread* GetNext() const {
+            return this->next;
+        }
+        constexpr void SetPrev(Thread* thread) {
+            this->prev = thread;
+        }
+        constexpr void SetNext(Thread* thread) {
+            this->next = thread;
+        }
+
+    private:
+        Thread* prev{};
+        Thread* next{};
+    };
+
+    QueueEntry& GetPriorityQueueEntry(s32 core) {
+        return this->per_core_priority_queue_entry[core];
+    }
+
+    const QueueEntry& GetPriorityQueueEntry(s32 core) const {
+        return this->per_core_priority_queue_entry[core];
+    }
+
+    s32 GetDisableDispatchCount() const {
+        return disable_count;
+    }
+
+    void DisableDispatch() {
+        ASSERT(GetDisableDispatchCount() >= 0);
+        disable_count++;
+    }
+
+    void EnableDispatch() {
+        ASSERT(GetDisableDispatchCount() > 0);
+        disable_count--;
+    }
+
 private:
-    friend class GlobalScheduler;
-    friend class Scheduler;
+    friend class GlobalSchedulerContext;
+    friend class KScheduler;
+    friend class Process;
 
     void SetSchedulingStatus(ThreadSchedStatus new_status);
     void AddSchedulingFlag(ThreadSchedFlags flag);
@@ -586,15 +636,16 @@ private:
     Common::SpinLock context_guard{};
     ThreadContext32 context_32{};
     ThreadContext64 context_64{};
-    std::unique_ptr<Core::ARM_Interface> arm_interface{};
     std::shared_ptr<Common::Fiber> host_context{};
 
-    u64 thread_id = 0;
-
     ThreadStatus status = ThreadStatus::Dormant;
+    u32 scheduling_state = 0;
+
+    u64 thread_id = 0;
 
     VAddr entry_point = 0;
     VAddr stack_top = 0;
+    std::atomic_int disable_count = 0;
 
     ThreadType type;
 
@@ -608,9 +659,8 @@ private:
     u32 current_priority = 0;
 
     u64 total_cpu_time_ticks = 0; ///< Total CPU running ticks.
-    u64 last_running_ticks = 0;   ///< CPU tick when thread was last running
-    u64 yield_count = 0;          ///< Number of redundant yields carried by this thread.
-                                  ///< a redundant yield is one where no scheduling is changed
+    s64 schedule_count{};
+    s64 last_scheduled_tick{};
 
     s32 processor_id = 0;
 
@@ -652,16 +702,16 @@ private:
     Handle hle_time_event;
     SynchronizationObject* hle_object;
 
-    Scheduler* scheduler = nullptr;
+    KScheduler* scheduler = nullptr;
+
+    std::array<QueueEntry, Core::Hardware::NUM_CPU_CORES> per_core_priority_queue_entry{};
 
     u32 ideal_core{0xFFFFFFFF};
-    u64 affinity_mask{0x1};
+    KAffinityMask affinity_mask{};
 
     s32 ideal_core_override = -1;
-    u64 affinity_mask_override = 0x1;
     u32 affinity_override_count = 0;
 
-    u32 scheduling_state = 0;
     u32 pausing_state = 0;
     bool is_running = false;
     bool is_waiting_on_sync = false;
diff --git a/src/core/hle/kernel/time_manager.cpp b/src/core/hle/kernel/time_manager.cpp
index 95f2446c9..79628e2b4 100644
--- a/src/core/hle/kernel/time_manager.cpp
+++ b/src/core/hle/kernel/time_manager.cpp
@@ -7,8 +7,8 @@
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/time_manager.h"
 
@@ -18,17 +18,27 @@ TimeManager::TimeManager(Core::System& system_) : system{system_} {
     time_manager_event_type = Core::Timing::CreateEvent(
         "Kernel::TimeManagerCallback",
         [this](std::uintptr_t thread_handle, std::chrono::nanoseconds) {
-            const SchedulerLock lock(system.Kernel());
+            const KScopedSchedulerLock lock(system.Kernel());
             const auto proper_handle = static_cast<Handle>(thread_handle);
-            if (cancelled_events[proper_handle]) {
-                return;
+
+            std::shared_ptr<Thread> thread;
+            {
+                std::lock_guard lock{mutex};
+                if (cancelled_events[proper_handle]) {
+                    return;
+                }
+                thread = system.Kernel().RetrieveThreadFromGlobalHandleTable(proper_handle);
+            }
+
+            if (thread) {
+                // Thread can be null if process has exited
+                thread->OnWakeUp();
             }
-            auto thread = this->system.Kernel().RetrieveThreadFromGlobalHandleTable(proper_handle);
-            thread->OnWakeUp();
         });
 }
 
 void TimeManager::ScheduleTimeEvent(Handle& event_handle, Thread* timetask, s64 nanoseconds) {
+    std::lock_guard lock{mutex};
     event_handle = timetask->GetGlobalHandle();
     if (nanoseconds > 0) {
         ASSERT(timetask);
@@ -43,6 +53,7 @@ void TimeManager::ScheduleTimeEvent(Handle& event_handle, Thread* timetask, s64
 }
 
 void TimeManager::UnscheduleTimeEvent(Handle event_handle) {
+    std::lock_guard lock{mutex};
     if (event_handle == InvalidHandle) {
         return;
     }
@@ -51,7 +62,8 @@ void TimeManager::UnscheduleTimeEvent(Handle event_handle) {
 }
 
 void TimeManager::CancelTimeEvent(Thread* time_task) {
-    Handle event_handle = time_task->GetGlobalHandle();
+    std::lock_guard lock{mutex};
+    const Handle event_handle = time_task->GetGlobalHandle();
     UnscheduleTimeEvent(event_handle);
 }
 
diff --git a/src/core/hle/kernel/time_manager.h b/src/core/hle/kernel/time_manager.h
index 307a18765..f39df39a0 100644
--- a/src/core/hle/kernel/time_manager.h
+++ b/src/core/hle/kernel/time_manager.h
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <memory>
+#include <mutex>
 #include <unordered_map>
 
 #include "core/hle/kernel/object.h"
@@ -42,6 +43,7 @@ private:
     Core::System& system;
     std::shared_ptr<Core::Timing::EventType> time_manager_event_type;
     std::unordered_map<Handle, bool> cancelled_events;
+    std::mutex mutex;
 };
 
 } // namespace Kernel