Kernel: Implemented mutex priority inheritance.

Verified with a hwtest and implemented based on reverse engineering.

Thread A's priority will get bumped to the highest priority among all the threads that are waiting for a mutex that A holds.
Once A releases the mutex and ownership is transferred to B, A's priority will return to normal and B's priority will be bumped.

src/core/CMakeLists.txt

@@ -42,8 +42,6 @@ add_library(core STATIC
     hle/kernel/client_port.h
     hle/kernel/client_session.cpp
     hle/kernel/client_session.h
-    hle/kernel/condition_variable.cpp
-    hle/kernel/condition_variable.h
     hle/kernel/errors.h
     hle/kernel/event.cpp
     hle/kernel/event.h

src/core/core.cpp

@@ -12,13 +12,10 @@
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/gdbstub/gdbstub.h"
-#include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/service/service.h"
-#include "core/hle/service/sm/controller.h"
-#include "core/hle/service/sm/sm.h"
 #include "core/hw/hw.h"
 #include "core/loader/loader.h"
 #include "core/memory_setup.h"
@@ -29,8 +26,6 @@ namespace Core {
 
 /*static*/ System System::s_instance;
 
-System::~System() = default;
-
 System::ResultStatus System::RunLoop(bool tight_loop) {
     status = ResultStatus::Success;
     if (!cpu_core) {
@@ -172,12 +167,10 @@ System::ResultStatus System::Init(EmuWindow* emu_window, u32 system_mode) {
 
     telemetry_session = std::make_unique<Core::TelemetrySession>();
 
-    service_manager = std::make_shared<Service::SM::ServiceManager>();
-
     HW::Init();
     Kernel::Init(system_mode);
     scheduler = std::make_unique<Kernel::Scheduler>(cpu_core.get());
-    Service::Init(service_manager);
+    Service::Init();
     GDBStub::Init();
 
     if (!VideoCore::Init(emu_window)) {
@@ -207,26 +200,17 @@ void System::Shutdown() {
     VideoCore::Shutdown();
     GDBStub::Shutdown();
     Service::Shutdown();
-    scheduler.reset();
+    scheduler = nullptr;
     Kernel::Shutdown();
     HW::Shutdown();
-    service_manager.reset();
-    telemetry_session.reset();
-    gpu_core.reset();
-    cpu_core.reset();
+    telemetry_session = nullptr;
+    gpu_core = nullptr;
+    cpu_core = nullptr;
     CoreTiming::Shutdown();
 
-    app_loader.reset();
+    app_loader = nullptr;
 
     LOG_DEBUG(Core, "Shutdown OK");
 }
 
-Service::SM::ServiceManager& System::ServiceManager() {
-    return *service_manager;
-}
-
-const Service::SM::ServiceManager& System::ServiceManager() const {
-    return *service_manager;
-}
-
 } // namespace Core

src/core/core.h

@@ -19,16 +19,10 @@
 class EmuWindow;
 class ARM_Interface;
 
-namespace Service::SM {
-class ServiceManager;
-}
-
 namespace Core {
 
 class System {
 public:
-    ~System();
-
     /**
      * Gets the instance of the System singleton class.
      * @returns Reference to the instance of the System singleton class.
@@ -143,9 +137,6 @@ public:
         return *app_loader;
     }
 
-    Service::SM::ServiceManager& ServiceManager();
-    const Service::SM::ServiceManager& ServiceManager() const;
-
     void SetGPUDebugContext(std::shared_ptr<Tegra::DebugContext> context) {
         debug_context = std::move(context);
     }
@@ -180,9 +171,6 @@ private:
     /// When true, signals that a reschedule should happen
     bool reschedule_pending{};
 
-    /// Service manager
-    std::shared_ptr<Service::SM::ServiceManager> service_manager;
-
     /// Telemetry session for this emulation session
     std::unique_ptr<Core::TelemetrySession> telemetry_session;
 

src/core/hle/kernel/condition_variable.cpp

@@ -1,64 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "common/assert.h"
-#include "core/hle/kernel/condition_variable.h"
-#include "core/hle/kernel/errors.h"
-#include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/object_address_table.h"
-#include "core/hle/kernel/thread.h"
-
-namespace Kernel {
-
-ConditionVariable::ConditionVariable() {}
-ConditionVariable::~ConditionVariable() {}
-
-ResultVal<SharedPtr<ConditionVariable>> ConditionVariable::Create(VAddr guest_addr,
-                                                                  std::string name) {
-    SharedPtr<ConditionVariable> condition_variable(new ConditionVariable);
-
-    condition_variable->name = std::move(name);
-    condition_variable->guest_addr = guest_addr;
-    condition_variable->mutex_addr = 0;
-
-    // Condition variables are referenced by guest address, so track this in the kernel
-    g_object_address_table.Insert(guest_addr, condition_variable);
-
-    return MakeResult<SharedPtr<ConditionVariable>>(std::move(condition_variable));
-}
-
-bool ConditionVariable::ShouldWait(Thread* thread) const {
-    return GetAvailableCount() <= 0;
-}
-
-void ConditionVariable::Acquire(Thread* thread) {
-    if (GetAvailableCount() <= 0)
-        return;
-
-    SetAvailableCount(GetAvailableCount() - 1);
-}
-
-ResultCode ConditionVariable::Release(s32 target) {
-    if (target == -1) {
-        // When -1, wake up all waiting threads
-        SetAvailableCount(static_cast<s32>(GetWaitingThreads().size()));
-        WakeupAllWaitingThreads();
-    } else {
-        // Otherwise, wake up just a single thread
-        SetAvailableCount(target);
-        WakeupWaitingThread(GetHighestPriorityReadyThread());
-    }
-
-    return RESULT_SUCCESS;
-}
-
-s32 ConditionVariable::GetAvailableCount() const {
-    return Memory::Read32(guest_addr);
-}
-
-void ConditionVariable::SetAvailableCount(s32 value) const {
-    Memory::Write32(guest_addr, value);
-}
-
-} // namespace Kernel

src/core/hle/kernel/condition_variable.h

@@ -1,63 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <string>
-#include <queue>
-#include "common/common_types.h"
-#include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/wait_object.h"
-#include "core/hle/result.h"
-
-namespace Kernel {
-
-class ConditionVariable final : public WaitObject {
-public:
-    /**
-     * Creates a condition variable.
-     * @param guest_addr Address of the object tracking the condition variable in guest memory. If
-     * specified, this condition variable will update the guest object when its state changes.
-     * @param name Optional name of condition variable.
-     * @return The created condition variable.
-     */
-    static ResultVal<SharedPtr<ConditionVariable>> Create(VAddr guest_addr,
-                                                          std::string name = "Unknown");
-
-    std::string GetTypeName() const override {
-        return "ConditionVariable";
-    }
-    std::string GetName() const override {
-        return name;
-    }
-
-    static const HandleType HANDLE_TYPE = HandleType::ConditionVariable;
-    HandleType GetHandleType() const override {
-        return HANDLE_TYPE;
-    }
-
-    s32 GetAvailableCount() const;
-    void SetAvailableCount(s32 value) const;
-
-    std::string name; ///< Name of condition variable (optional)
-    VAddr guest_addr; ///< Address of the guest condition variable value
-    VAddr mutex_addr; ///< (optional) Address of guest mutex value associated with this condition
-                      ///< variable, used for implementing events
-
-    bool ShouldWait(Thread* thread) const override;
-    void Acquire(Thread* thread) override;
-
-    /**
-     * Releases a slot from a condition variable.
-     * @param target The number of threads to wakeup, -1 is all.
-     * @return ResultCode indicating if the operation succeeded.
-     */
-    ResultCode Release(s32 target);
-
-private:
-    ConditionVariable();
-    ~ConditionVariable() override;
-};
-
-} // namespace Kernel

src/core/hle/kernel/errors.h

@@ -20,6 +20,7 @@ enum {
     MaxConnectionsReached = 52,
 
     // Confirmed Switch OS error codes
+    MisalignedAddress = 102,
     InvalidHandle = 114,
     Timeout = 117,
     SynchronizationCanceled = 118,

src/core/hle/kernel/kernel.h

@@ -18,12 +18,10 @@ using Handle = u32;
 enum class HandleType : u32 {
     Unknown,
     Event,
-    Mutex,
     SharedMemory,
     Thread,
     Process,
     AddressArbiter,
-    ConditionVariable,
     Timer,
     ResourceLimit,
     CodeSet,
@@ -63,9 +61,7 @@ public:
     bool IsWaitable() const {
         switch (GetHandleType()) {
         case HandleType::Event:
-        case HandleType::Mutex:
         case HandleType::Thread:
-        case HandleType::ConditionVariable:
         case HandleType::Timer:
         case HandleType::ServerPort:
         case HandleType::ServerSession:

src/core/hle/kernel/mutex.cpp

@@ -7,6 +7,7 @@
 #include <boost/range/algorithm_ext/erase.hpp>
 #include "common/assert.h"
 #include "core/core.h"
+#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/mutex.h"
@@ -15,124 +16,120 @@
 
 namespace Kernel {
 
-void ReleaseThreadMutexes(Thread* thread) {
-    for (auto& mtx : thread->held_mutexes) {
-        mtx->SetHasWaiters(false);
-        mtx->SetHoldingThread(nullptr);
-        mtx->WakeupAllWaitingThreads();
-    }
-    thread->held_mutexes.clear();
-}
+/// Returns the number of threads that are waiting for a mutex, and the highest priority one among
+/// those.
+static std::pair<SharedPtr<Thread>, u32> GetHighestPriorityMutexWaitingThread(
+    SharedPtr<Thread> current_thread, VAddr mutex_addr) {
 
-Mutex::Mutex() {}
-Mutex::~Mutex() {}
+    SharedPtr<Thread> highest_priority_thread;
+    u32 num_waiters = 0;
 
-SharedPtr<Mutex> Mutex::Create(SharedPtr<Kernel::Thread> holding_thread, VAddr guest_addr,
-                               std::string name) {
-    SharedPtr<Mutex> mutex(new Mutex);
+    for (auto& thread : current_thread->wait_mutex_threads) {
+        if (thread->mutex_wait_address != mutex_addr)
+            continue;
 
-    mutex->guest_addr = guest_addr;
-    mutex->name = std::move(name);
+        ASSERT(thread->status == THREADSTATUS_WAIT_MUTEX);
 
-    // If mutex was initialized with a holding thread, acquire it by the holding thread
-    if (holding_thread) {
-        mutex->Acquire(holding_thread.get());
+        ++num_waiters;
+        if (highest_priority_thread == nullptr ||
+            thread->GetPriority() < highest_priority_thread->GetPriority()) {
+            highest_priority_thread = thread;
+        }
     }
 
-    // Mutexes are referenced by guest address, so track this in the kernel
-    g_object_address_table.Insert(guest_addr, mutex);
-
-    return mutex;
+    return {highest_priority_thread, num_waiters};
 }
 
-bool Mutex::ShouldWait(Thread* thread) const {
-    auto holding_thread = GetHoldingThread();
-    return holding_thread != nullptr && thread != holding_thread;
+/// Update the mutex owner field of all threads waiting on the mutex to point to the new owner.
+static void TransferMutexOwnership(VAddr mutex_addr, SharedPtr<Thread> current_thread,
+                                   SharedPtr<Thread> new_owner) {
+    auto threads = current_thread->wait_mutex_threads;
+    for (auto& thread : threads) {
+        if (thread->mutex_wait_address != mutex_addr)
+            continue;
+
+        ASSERT(thread->lock_owner == current_thread);
+        current_thread->RemoveMutexWaiter(thread);
+        if (new_owner != thread)
+            new_owner->AddMutexWaiter(thread);
+    }
 }
 
-void Mutex::Acquire(Thread* thread) {
-    ASSERT_MSG(!ShouldWait(thread), "object unavailable!");
+ResultCode Mutex::TryAcquire(VAddr address, Handle holding_thread_handle,
+                             Handle requesting_thread_handle) {
+    // The mutex address must be 4-byte aligned
+    if ((address % sizeof(u32)) != 0) {
+        return ResultCode(ErrorModule::Kernel, ErrCodes::MisalignedAddress);
+    }
 
-    priority = thread->current_priority;
-    thread->held_mutexes.insert(this);
-    SetHoldingThread(thread);
-    thread->UpdatePriority();
-    Core::System::GetInstance().PrepareReschedule();
-}
+    SharedPtr<Thread> holding_thread = g_handle_table.Get<Thread>(holding_thread_handle);
+    SharedPtr<Thread> requesting_thread = g_handle_table.Get<Thread>(requesting_thread_handle);
 
-ResultCode Mutex::Release(Thread* thread) {
-    auto holding_thread = GetHoldingThread();
-    ASSERT(holding_thread);
+    // TODO(Subv): It is currently unknown if it is possible to lock a mutex in behalf of another
+    // thread.
+    ASSERT(requesting_thread == GetCurrentThread());
 
-    // We can only release the mutex if it's held by the calling thread.
-    ASSERT(thread == holding_thread);
+    u32 addr_value = Memory::Read32(address);
+
+    // If the mutex isn't being held, just return success.
+    if (addr_value != (holding_thread_handle | Mutex::MutexHasWaitersFlag)) {
+        return RESULT_SUCCESS;
+    }
+
+    if (holding_thread == nullptr)
+        return ERR_INVALID_HANDLE;
+
+    // Wait until the mutex is released
+    GetCurrentThread()->mutex_wait_address = address;
+    GetCurrentThread()->wait_handle = requesting_thread_handle;
+
+    GetCurrentThread()->status = THREADSTATUS_WAIT_MUTEX;
+    GetCurrentThread()->wakeup_callback = nullptr;
+
+    // Update the lock holder thread's priority to prevent priority inversion.
+    holding_thread->AddMutexWaiter(GetCurrentThread());
 
-    holding_thread->held_mutexes.erase(this);
-    holding_thread->UpdatePriority();
-    SetHoldingThread(nullptr);
-    SetHasWaiters(!GetWaitingThreads().empty());
-    WakeupAllWaitingThreads();
     Core::System::GetInstance().PrepareReschedule();
 
     return RESULT_SUCCESS;
 }
 
-void Mutex::AddWaitingThread(SharedPtr<Thread> thread) {
-    WaitObject::AddWaitingThread(thread);
-    thread->pending_mutexes.insert(this);
-    SetHasWaiters(true);
-    UpdatePriority();
-}
-
-void Mutex::RemoveWaitingThread(Thread* thread) {
-    WaitObject::RemoveWaitingThread(thread);
-    thread->pending_mutexes.erase(this);
-    if (!GetHasWaiters())
-        SetHasWaiters(!GetWaitingThreads().empty());
-    UpdatePriority();
-}
-
-void Mutex::UpdatePriority() {
-    if (!GetHoldingThread())
-        return;
-
-    u32 best_priority = THREADPRIO_LOWEST;
-    for (auto& waiter : GetWaitingThreads()) {
-        if (waiter->current_priority < best_priority)
-            best_priority = waiter->current_priority;
+ResultCode Mutex::Release(VAddr address) {
+    // The mutex address must be 4-byte aligned
+    if ((address % sizeof(u32)) != 0) {
+        return ResultCode(ErrorModule::Kernel, ErrCodes::MisalignedAddress);
     }
 
-    if (best_priority != priority) {
-        priority = best_priority;
-        GetHoldingThread()->UpdatePriority();
+    auto [thread, num_waiters] = GetHighestPriorityMutexWaitingThread(GetCurrentThread(), address);
+
+    // There are no more threads waiting for the mutex, release it completely.
+    if (thread == nullptr) {
+        ASSERT(GetCurrentThread()->wait_mutex_threads.empty());
+        Memory::Write32(address, 0);
+        return RESULT_SUCCESS;
     }
-}
 
-Handle Mutex::GetOwnerHandle() const {
-    GuestState guest_state{Memory::Read32(guest_addr)};
-    return guest_state.holding_thread_handle;
-}
+    // Transfer the ownership of the mutex from the previous owner to the new one.
+    TransferMutexOwnership(address, GetCurrentThread(), thread);
 
-SharedPtr<Thread> Mutex::GetHoldingThread() const {
-    GuestState guest_state{Memory::Read32(guest_addr)};
-    return g_handle_table.Get<Thread>(guest_state.holding_thread_handle);
-}
+    u32 mutex_value = thread->wait_handle;
 
-void Mutex::SetHoldingThread(SharedPtr<Thread> thread) {
-    GuestState guest_state{Memory::Read32(guest_addr)};
-    guest_state.holding_thread_handle.Assign(thread ? thread->guest_handle : 0);
-    Memory::Write32(guest_addr, guest_state.raw);
-}
+    if (num_waiters >= 2) {
+        // Notify the guest that there are still some threads waiting for the mutex
+        mutex_value |= Mutex::MutexHasWaitersFlag;
+    }
 
-bool Mutex::GetHasWaiters() const {
-    GuestState guest_state{Memory::Read32(guest_addr)};
-    return guest_state.has_waiters != 0;
-}
+    // Grant the mutex to the next waiting thread and resume it.
+    Memory::Write32(address, mutex_value);
 
-void Mutex::SetHasWaiters(bool has_waiters) {
-    GuestState guest_state{Memory::Read32(guest_addr)};
-    guest_state.has_waiters.Assign(has_waiters ? 1 : 0);
-    Memory::Write32(guest_addr, guest_state.raw);
-}
+    ASSERT(thread->status == THREADSTATUS_WAIT_MUTEX);
+    thread->ResumeFromWait();
 
+    thread->lock_owner = nullptr;
+    thread->condvar_wait_address = 0;
+    thread->mutex_wait_address = 0;
+    thread->wait_handle = 0;
+
+    return RESULT_SUCCESS;
+}
 } // namespace Kernel

src/core/hle/kernel/mutex.h

@@ -15,87 +15,23 @@ namespace Kernel {
 
 class Thread;
 
-class Mutex final : public WaitObject {
+class Mutex final {
 public:
-    /**
-     * Creates a mutex.
-     * @param holding_thread Specifies a thread already holding the mutex. If not nullptr, this
-     * thread will acquire the mutex.
-     * @param guest_addr Address of the object tracking the mutex in guest memory. If specified,
-     * this mutex will update the guest object when its state changes.
-     * @param name Optional name of mutex
-     * @return Pointer to new Mutex object
-     */
-    static SharedPtr<Mutex> Create(SharedPtr<Kernel::Thread> holding_thread, VAddr guest_addr = 0,
-                                   std::string name = "Unknown");
+    /// Flag that indicates that a mutex still has threads waiting for it.
+    static constexpr u32 MutexHasWaitersFlag = 0x40000000;
+    /// Mask of the bits in a mutex address value that contain the mutex owner.
+    static constexpr u32 MutexOwnerMask = 0xBFFFFFFF;
 
-    std::string GetTypeName() const override {
-        return "Mutex";
-    }
-    std::string GetName() const override {
-        return name;
-    }
+    /// Attempts to acquire a mutex at the specified address.
+    static ResultCode TryAcquire(VAddr address, Handle holding_thread_handle,
+                                 Handle requesting_thread_handle);
 
-    static const HandleType HANDLE_TYPE = HandleType::Mutex;
-    HandleType GetHandleType() const override {
-        return HANDLE_TYPE;
-    }
-
-    u32 priority;     ///< The priority of the mutex, used for priority inheritance.
-    std::string name; ///< Name of mutex (optional)
-    VAddr guest_addr; ///< Address of the guest mutex value
-
-    /**
-     * Elevate the mutex priority to the best priority
-     * among the priorities of all its waiting threads.
-     */
-    void UpdatePriority();
-
-    bool ShouldWait(Thread* thread) const override;
-    void Acquire(Thread* thread) override;
-
-    void AddWaitingThread(SharedPtr<Thread> thread) override;
-    void RemoveWaitingThread(Thread* thread) override;
-
-    /**
-     * Attempts to release the mutex from the specified thread.
-     * @param thread Thread that wants to release the mutex.
-     * @returns The result code of the operation.
-     */
-    ResultCode Release(Thread* thread);
-
-    /// Gets the handle to the holding process stored in the guest state.
-    Handle GetOwnerHandle() const;
-
-    /// Gets the Thread pointed to by the owner handle
-    SharedPtr<Thread> GetHoldingThread() const;
-    /// Sets the holding process handle in the guest state.
-    void SetHoldingThread(SharedPtr<Thread> thread);
-
-    /// Returns the has_waiters bit in the guest state.
-    bool GetHasWaiters() const;
-    /// Sets the has_waiters bit in the guest state.
-    void SetHasWaiters(bool has_waiters);
+    /// Releases the mutex at the specified address.
+    static ResultCode Release(VAddr address);
 
 private:
-    Mutex();
-    ~Mutex() override;
-
-    /// Object in guest memory used to track the mutex state
-    union GuestState {
-        u32_le raw;
-        /// Handle of the thread that currently holds the mutex, 0 if available
-        BitField<0, 30, u32_le> holding_thread_handle;
-        /// 1 when there are threads waiting for this mutex, otherwise 0
-        BitField<30, 1, u32_le> has_waiters;
-    };
-    static_assert(sizeof(GuestState) == 4, "GuestState size is incorrect");
+    Mutex() = default;
+    ~Mutex() = default;
 };
 
-/**
- * Releases all the mutexes held by the specified thread
- * @param thread Thread that is holding the mutexes
- */
-void ReleaseThreadMutexes(Thread* thread);
-
 } // namespace Kernel

src/core/hle/kernel/resource_limit.cpp

@@ -34,57 +34,57 @@ SharedPtr<ResourceLimit> ResourceLimit::GetForCategory(ResourceLimitCategory cat
     }
 }
 
-s32 ResourceLimit::GetCurrentResourceValue(ResourceType resource) const {
+s32 ResourceLimit::GetCurrentResourceValue(u32 resource) const {
     switch (resource) {
-    case ResourceType::Commit:
+    case COMMIT:
         return current_commit;
-    case ResourceType::Thread:
+    case THREAD:
         return current_threads;
-    case ResourceType::Event:
+    case EVENT:
         return current_events;
-    case ResourceType::Mutex:
+    case MUTEX:
         return current_mutexes;
-    case ResourceType::Semaphore:
+    case SEMAPHORE:
         return current_semaphores;
-    case ResourceType::Timer:
+    case TIMER:
         return current_timers;
-    case ResourceType::SharedMemory:
+    case SHARED_MEMORY:
         return current_shared_mems;
-    case ResourceType::AddressArbiter:
+    case ADDRESS_ARBITER:
         return current_address_arbiters;
-    case ResourceType::CPUTime:
+    case CPU_TIME:
         return current_cpu_time;
     default:
-        LOG_ERROR(Kernel, "Unknown resource type=%08X", static_cast<u32>(resource));
+        LOG_ERROR(Kernel, "Unknown resource type=%08X", resource);
         UNIMPLEMENTED();
         return 0;
     }
 }
 
-u32 ResourceLimit::GetMaxResourceValue(ResourceType resource) const {
+u32 ResourceLimit::GetMaxResourceValue(u32 resource) const {
     switch (resource) {
-    case ResourceType::Priority:
+    case PRIORITY:
         return max_priority;
-    case ResourceType::Commit:
+    case COMMIT:
         return max_commit;
-    case ResourceType::Thread:
+    case THREAD:
         return max_threads;
-    case ResourceType::Event:
+    case EVENT:
         return max_events;
-    case ResourceType::Mutex:
+    case MUTEX:
         return max_mutexes;
-    case ResourceType::Semaphore:
+    case SEMAPHORE:
         return max_semaphores;
-    case ResourceType::Timer:
+    case TIMER:
         return max_timers;
-    case ResourceType::SharedMemory:
+    case SHARED_MEMORY:
         return max_shared_mems;
-    case ResourceType::AddressArbiter:
+    case ADDRESS_ARBITER:
         return max_address_arbiters;
-    case ResourceType::CPUTime:
+    case CPU_TIME:
         return max_cpu_time;
     default:
-        LOG_ERROR(Kernel, "Unknown resource type=%08X", static_cast<u32>(resource));
+        LOG_ERROR(Kernel, "Unknown resource type=%08X", resource);
         UNIMPLEMENTED();
         return 0;
     }

src/core/hle/kernel/resource_limit.h

@@ -16,17 +16,17 @@ enum class ResourceLimitCategory : u8 {
     OTHER = 3
 };
 
-enum class ResourceType {
-    Priority = 0,
-    Commit = 1,
-    Thread = 2,
-    Event = 3,
-    Mutex = 4,
-    Semaphore = 5,
-    Timer = 6,
-    SharedMemory = 7,
-    AddressArbiter = 8,
-    CPUTime = 9,
+enum ResourceTypes {
+    PRIORITY = 0,
+    COMMIT = 1,
+    THREAD = 2,
+    EVENT = 3,
+    MUTEX = 4,
+    SEMAPHORE = 5,
+    TIMER = 6,
+    SHARED_MEMORY = 7,
+    ADDRESS_ARBITER = 8,
+    CPU_TIME = 9,
 };
 
 class ResourceLimit final : public Object {
@@ -60,14 +60,14 @@ public:
      * @param resource Requested resource type
      * @returns The current value of the resource type
      */
-    s32 GetCurrentResourceValue(ResourceType resource) const;
+    s32 GetCurrentResourceValue(u32 resource) const;
 
     /**
      * Gets the max value for the specified resource.
      * @param resource Requested resource type
      * @returns The max value of the resource type
      */
-    u32 GetMaxResourceValue(ResourceType resource) const;
+    u32 GetMaxResourceValue(u32 resource) const;
 
     /// Name of resource limit object.
     std::string name;

src/core/hle/kernel/svc.cpp

@@ -13,7 +13,6 @@
 #include "core/core_timing.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
-#include "core/hle/kernel/condition_variable.h"
 #include "core/hle/kernel/event.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/mutex.h"
@@ -262,32 +261,14 @@ static ResultCode ArbitrateLock(Handle holding_thread_handle, VAddr mutex_addr,
               "requesting_current_thread_handle=0x%08X",
               holding_thread_handle, mutex_addr, requesting_thread_handle);
 
-    SharedPtr<Thread> holding_thread = g_handle_table.Get<Thread>(holding_thread_handle);
-    SharedPtr<Thread> requesting_thread = g_handle_table.Get<Thread>(requesting_thread_handle);
-
-    ASSERT(requesting_thread);
-    ASSERT(requesting_thread == GetCurrentThread());
-
-    SharedPtr<Mutex> mutex = g_object_address_table.Get<Mutex>(mutex_addr);
-    if (!mutex) {
-        // Create a new mutex for the specified address if one does not already exist
-        mutex = Mutex::Create(holding_thread, mutex_addr);
-        mutex->name = Common::StringFromFormat("mutex-%llx", mutex_addr);
-    }
-
-    ASSERT(holding_thread == mutex->GetHoldingThread());
-
-    return WaitSynchronization1(mutex, requesting_thread.get());
+    return Mutex::TryAcquire(mutex_addr, holding_thread_handle, requesting_thread_handle);
 }
 
 /// Unlock a mutex
 static ResultCode ArbitrateUnlock(VAddr mutex_addr) {
     LOG_TRACE(Kernel_SVC, "called mutex_addr=0x%llx", mutex_addr);
 
-    SharedPtr<Mutex> mutex = g_object_address_table.Get<Mutex>(mutex_addr);
-    ASSERT(mutex);
-
-    return mutex->Release(GetCurrentThread());
+    return Mutex::Release(mutex_addr);
 }
 
 /// Break program execution
@@ -407,16 +388,11 @@ static ResultCode SetThreadPriority(Handle handle, u32 priority) {
     // Note: The kernel uses the current process's resource limit instead of
     // the one from the thread owner's resource limit.
     SharedPtr<ResourceLimit>& resource_limit = Core::CurrentProcess()->resource_limit;
-    if (resource_limit->GetMaxResourceValue(ResourceType::Priority) > priority) {
+    if (resource_limit->GetMaxResourceValue(ResourceTypes::PRIORITY) > priority) {
         return ERR_NOT_AUTHORIZED;
     }
 
     thread->SetPriority(priority);
-    thread->UpdatePriority();
-
-    // Update the mutexes that this thread is waiting for
-    for (auto& mutex : thread->pending_mutexes)
-        mutex->UpdatePriority();
 
     Core::System::GetInstance().PrepareReschedule();
     return RESULT_SUCCESS;
@@ -541,7 +517,7 @@ static ResultCode CreateThread(Handle* out_handle, VAddr entry_point, u64 arg, V
     }
 
     SharedPtr<ResourceLimit>& resource_limit = Core::CurrentProcess()->resource_limit;
-    if (resource_limit->GetMaxResourceValue(ResourceType::Priority) > priority) {
+    if (resource_limit->GetMaxResourceValue(ResourceTypes::PRIORITY) > priority) {
         return ERR_NOT_AUTHORIZED;
     }
 
@@ -634,77 +610,20 @@ static ResultCode WaitProcessWideKeyAtomic(VAddr mutex_addr, VAddr condition_var
     SharedPtr<Thread> thread = g_handle_table.Get<Thread>(thread_handle);
     ASSERT(thread);
 
-    SharedPtr<Mutex> mutex = g_object_address_table.Get<Mutex>(mutex_addr);
-    if (!mutex) {
-        // Create a new mutex for the specified address if one does not already exist
-        mutex = Mutex::Create(thread, mutex_addr);
-        mutex->name = Common::StringFromFormat("mutex-%llx", mutex_addr);
-    }
+    CASCADE_CODE(Mutex::Release(mutex_addr));
 
-    SharedPtr<ConditionVariable> condition_variable =
-        g_object_address_table.Get<ConditionVariable>(condition_variable_addr);
-    if (!condition_variable) {
-        // Create a new condition_variable for the specified address if one does not already exist
-        condition_variable = ConditionVariable::Create(condition_variable_addr).Unwrap();
-        condition_variable->name =
-            Common::StringFromFormat("condition-variable-%llx", condition_variable_addr);
-    }
+    SharedPtr<Thread> current_thread = GetCurrentThread();
+    current_thread->condvar_wait_address = condition_variable_addr;
+    current_thread->mutex_wait_address = mutex_addr;
+    current_thread->wait_handle = thread_handle;
+    current_thread->status = THREADSTATUS_WAIT_MUTEX;
+    current_thread->wakeup_callback = nullptr;
 
-    if (condition_variable->mutex_addr) {
-        // Previously created the ConditionVariable using WaitProcessWideKeyAtomic, verify
-        // everything is correct
-        ASSERT(condition_variable->mutex_addr == mutex_addr);
-    } else {
-        // Previously created the ConditionVariable using SignalProcessWideKey, set the mutex
-        // associated with it
-        condition_variable->mutex_addr = mutex_addr;
-    }
+    current_thread->WakeAfterDelay(nano_seconds);
 
-    if (mutex->GetOwnerHandle()) {
-        // Release the mutex if the current thread is holding it
-        mutex->Release(thread.get());
-    }
-
-    auto wakeup_callback = [mutex, nano_seconds](ThreadWakeupReason reason,
-                                                 SharedPtr<Thread> thread,
-                                                 SharedPtr<WaitObject> object, size_t index) {
-        ASSERT(thread->status == THREADSTATUS_WAIT_SYNCH_ANY);
-
-        if (reason == ThreadWakeupReason::Timeout) {
-            thread->SetWaitSynchronizationResult(RESULT_TIMEOUT);
-            return true;
-        }
-
-        ASSERT(reason == ThreadWakeupReason::Signal);
-
-        // Now try to acquire the mutex and don't resume if it's not available.
-        if (!mutex->ShouldWait(thread.get())) {
-            mutex->Acquire(thread.get());
-            thread->SetWaitSynchronizationResult(RESULT_SUCCESS);
-            return true;
-        }
-
-        if (nano_seconds == 0) {
-            thread->SetWaitSynchronizationResult(RESULT_TIMEOUT);
-            return true;
-        }
-
-        thread->wait_objects = {mutex};
-        mutex->AddWaitingThread(thread);
-        thread->status = THREADSTATUS_WAIT_SYNCH_ANY;
-
-        // Create an event to wake the thread up after the
-        // specified nanosecond delay has passed
-        thread->WakeAfterDelay(nano_seconds);
-        thread->wakeup_callback = DefaultThreadWakeupCallback;
-
-        Core::System::GetInstance().PrepareReschedule();
-
-        return false;
-    };
-    CASCADE_CODE(
-        WaitSynchronization1(condition_variable, thread.get(), nano_seconds, wakeup_callback));
+    // Note: Deliberately don't attempt to inherit the lock owner's priority.
 
+    Core::System::GetInstance().PrepareReschedule();
     return RESULT_SUCCESS;
 }
 
@@ -713,24 +632,53 @@ static ResultCode SignalProcessWideKey(VAddr condition_variable_addr, s32 target
     LOG_TRACE(Kernel_SVC, "called, condition_variable_addr=0x%llx, target=0x%08x",
               condition_variable_addr, target);
 
-    // Wakeup all or one thread - Any other value is unimplemented
-    ASSERT(target == -1 || target == 1);
+    u32 processed = 0;
+    auto& thread_list = Core::System::GetInstance().Scheduler().GetThreadList();
 
-    SharedPtr<ConditionVariable> condition_variable =
-        g_object_address_table.Get<ConditionVariable>(condition_variable_addr);
-    if (!condition_variable) {
-        // Create a new condition_variable for the specified address if one does not already exist
-        condition_variable = ConditionVariable::Create(condition_variable_addr).Unwrap();
-        condition_variable->name =
-            Common::StringFromFormat("condition-variable-%llx", condition_variable_addr);
-    }
+    for (auto& thread : thread_list) {
+        if (thread->condvar_wait_address != condition_variable_addr)
+            continue;
 
-    CASCADE_CODE(condition_variable->Release(target));
+        // Only process up to 'target' threads, unless 'target' is -1, in which case process
+        // them all.
+        if (target != -1 && processed >= target)
+            break;
 
-    if (condition_variable->mutex_addr) {
-        // If a mutex was created for this condition_variable, wait the current thread on it
-        SharedPtr<Mutex> mutex = g_object_address_table.Get<Mutex>(condition_variable->mutex_addr);
-        return WaitSynchronization1(mutex, GetCurrentThread());
+        // If the mutex is not yet acquired, acquire it.
+        u32 mutex_val = Memory::Read32(thread->mutex_wait_address);
+
+        if (mutex_val == 0) {
+            // We were able to acquire the mutex, resume this thread.
+            Memory::Write32(thread->mutex_wait_address, thread->wait_handle);
+            ASSERT(thread->status == THREADSTATUS_WAIT_MUTEX);
+            thread->ResumeFromWait();
+
+            auto lock_owner = thread->lock_owner;
+            if (lock_owner)
+                lock_owner->RemoveMutexWaiter(thread);
+
+            thread->lock_owner = nullptr;
+            thread->mutex_wait_address = 0;
+            thread->condvar_wait_address = 0;
+            thread->wait_handle = 0;
+        } else {
+            // Couldn't acquire the mutex, block the thread.
+            Handle owner_handle = static_cast<Handle>(mutex_val & Mutex::MutexOwnerMask);
+            auto owner = g_handle_table.Get<Thread>(owner_handle);
+            ASSERT(owner);
+            ASSERT(thread->status != THREADSTATUS_RUNNING);
+            thread->status = THREADSTATUS_WAIT_MUTEX;
+            thread->wakeup_callback = nullptr;
+
+            // Signal that the mutex now has a waiting thread.
+            Memory::Write32(thread->mutex_wait_address, mutex_val | Mutex::MutexHasWaitersFlag);
+
+            owner->AddMutexWaiter(thread);
+
+            Core::System::GetInstance().PrepareReschedule();
+        }
+
+        ++processed;
     }
 
     return RESULT_SUCCESS;

src/core/hle/kernel/thread.cpp

@@ -77,9 +77,6 @@ void Thread::Stop() {
     }
     wait_objects.clear();
 
-    // Release all the mutexes that this thread holds
-    ReleaseThreadMutexes(this);
-
     // Mark the TLS slot in the thread's page as free.
     u64 tls_page = (tls_address - Memory::TLS_AREA_VADDR) / Memory::PAGE_SIZE;
     u64 tls_slot =
@@ -126,6 +123,19 @@ static void ThreadWakeupCallback(u64 thread_handle, int cycles_late) {
             resume = thread->wakeup_callback(ThreadWakeupReason::Timeout, thread, nullptr, 0);
     }
 
+    if (thread->mutex_wait_address != 0 || thread->condvar_wait_address != 0 ||
+        thread->wait_handle) {
+        ASSERT(thread->status == THREADSTATUS_WAIT_MUTEX);
+        thread->mutex_wait_address = 0;
+        thread->condvar_wait_address = 0;
+        thread->wait_handle = 0;
+
+        auto lock_owner = thread->lock_owner;
+        // Threads waking up by timeout from WaitProcessWideKey do not perform priority inheritance
+        // and don't have a lock owner.
+        ASSERT(lock_owner == nullptr);
+    }
+
     if (resume)
         thread->ResumeFromWait();
 }
@@ -151,6 +161,7 @@ void Thread::ResumeFromWait() {
     case THREADSTATUS_WAIT_HLE_EVENT:
     case THREADSTATUS_WAIT_SLEEP:
     case THREADSTATUS_WAIT_IPC:
+    case THREADSTATUS_WAIT_MUTEX:
         break;
 
     case THREADSTATUS_READY:
@@ -256,7 +267,9 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point,
     thread->last_running_ticks = CoreTiming::GetTicks();
     thread->processor_id = processor_id;
     thread->wait_objects.clear();
-    thread->wait_address = 0;
+    thread->mutex_wait_address = 0;
+    thread->condvar_wait_address = 0;
+    thread->wait_handle = 0;
     thread->name = std::move(name);
     thread->callback_handle = wakeup_callback_handle_table.Create(thread).Unwrap();
     thread->owner_process = owner_process;
@@ -317,17 +330,8 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point,
 void Thread::SetPriority(u32 priority) {
     ASSERT_MSG(priority <= THREADPRIO_LOWEST && priority >= THREADPRIO_HIGHEST,
                "Invalid priority value.");
-    Core::System::GetInstance().Scheduler().SetThreadPriority(this, priority);
-    nominal_priority = current_priority = priority;
-}
-
-void Thread::UpdatePriority() {
-    u32 best_priority = nominal_priority;
-    for (auto& mutex : held_mutexes) {
-        if (mutex->priority < best_priority)
-            best_priority = mutex->priority;
-    }
-    BoostPriority(best_priority);
+    nominal_priority = priority;
+    UpdatePriority();
 }
 
 void Thread::BoostPriority(u32 priority) {
@@ -377,6 +381,38 @@ VAddr Thread::GetCommandBufferAddress() const {
     return GetTLSAddress() + CommandHeaderOffset;
 }
 
+void Thread::AddMutexWaiter(SharedPtr<Thread> thread) {
+    thread->lock_owner = this;
+    wait_mutex_threads.emplace_back(std::move(thread));
+    UpdatePriority();
+}
+
+void Thread::RemoveMutexWaiter(SharedPtr<Thread> thread) {
+    boost::remove_erase(wait_mutex_threads, thread);
+    thread->lock_owner = nullptr;
+    UpdatePriority();
+}
+
+void Thread::UpdatePriority() {
+    // Find the highest priority among all the threads that are waiting for this thread's lock
+    u32 new_priority = nominal_priority;
+    for (const auto& thread : wait_mutex_threads) {
+        if (thread->nominal_priority < new_priority)
+            new_priority = thread->nominal_priority;
+    }
+
+    if (new_priority == current_priority)
+        return;
+
+    Core::System::GetInstance().Scheduler().SetThreadPriority(this, new_priority);
+
+    current_priority = new_priority;
+
+    // Recursively update the priority of the thread that depends on the priority of this one.
+    if (lock_owner)
+        lock_owner->UpdatePriority();
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
 /**

src/core/hle/kernel/thread.h

@@ -18,7 +18,7 @@
 enum ThreadPriority : u32 {
     THREADPRIO_HIGHEST = 0,       ///< Highest thread priority
     THREADPRIO_USERLAND_MAX = 24, ///< Highest thread priority for userland apps
-    THREADPRIO_DEFAULT = 48,      ///< Default thread priority for userland apps
+    THREADPRIO_DEFAULT = 44,      ///< Default thread priority for userland apps
     THREADPRIO_LOWEST = 63,       ///< Lowest thread priority
 };
 
@@ -43,6 +43,7 @@ enum ThreadStatus {
     THREADSTATUS_WAIT_IPC,       ///< Waiting for the reply from an IPC request
     THREADSTATUS_WAIT_SYNCH_ANY, ///< Waiting due to WaitSynch1 or WaitSynchN with wait_all = false
     THREADSTATUS_WAIT_SYNCH_ALL, ///< Waiting due to WaitSynchronizationN with wait_all = true
+    THREADSTATUS_WAIT_MUTEX,     ///< Waiting due to an ArbitrateLock/WaitProcessWideKey svc
     THREADSTATUS_DORMANT,        ///< Created but not yet made ready
     THREADSTATUS_DEAD            ///< Run to completion, or forcefully terminated
 };
@@ -54,7 +55,6 @@ enum class ThreadWakeupReason {
 
 namespace Kernel {
 
-class Mutex;
 class Process;
 
 class Thread final : public WaitObject {
@@ -103,18 +103,21 @@ public:
      */
     void SetPriority(u32 priority);
 
-    /**
-     * Boost's a thread's priority to the best priority among the thread's held mutexes.
-     * This prevents priority inversion via priority inheritance.
-     */
-    void UpdatePriority();
-
     /**
      * Temporarily boosts the thread's priority until the next time it is scheduled
      * @param priority The new priority
      */
     void BoostPriority(u32 priority);
 
+    /// Adds a thread to the list of threads that are waiting for a lock held by this thread.
+    void AddMutexWaiter(SharedPtr<Thread> thread);
+
+    /// Removes a thread from the list of threads that are waiting for a lock held by this thread.
+    void RemoveMutexWaiter(SharedPtr<Thread> thread);
+
+    /// Recalculates the current priority taking into account priority inheritance.
+    void UpdatePriority();
+
     /**
      * Gets the thread's thread ID
      * @return The thread's ID
@@ -205,19 +208,22 @@ public:
 
     VAddr tls_address; ///< Virtual address of the Thread Local Storage of the thread
 
-    /// Mutexes currently held by this thread, which will be released when it exits.
-    boost::container::flat_set<SharedPtr<Mutex>> held_mutexes;
-
-    /// Mutexes that this thread is currently waiting for.
-    boost::container::flat_set<SharedPtr<Mutex>> pending_mutexes;
-
     SharedPtr<Process> owner_process; ///< Process that owns this thread
 
     /// Objects that the thread is waiting on, in the same order as they were
     // passed to WaitSynchronization1/N.
     std::vector<SharedPtr<WaitObject>> wait_objects;
 
-    VAddr wait_address; ///< If waiting on an AddressArbiter, this is the arbitration address
+    /// List of threads that are waiting for a mutex that is held by this thread.
+    std::vector<SharedPtr<Thread>> wait_mutex_threads;
+
+    /// Thread that owns the lock that this thread is waiting for.
+    SharedPtr<Thread> lock_owner;
+
+    // If waiting on a ConditionVariable, this is the ConditionVariable  address
+    VAddr condvar_wait_address;
+    VAddr mutex_wait_address; ///< If waiting on a Mutex, this is the mutex address
+    Handle wait_handle;       ///< The handle used to wait for the mutex.
 
     std::string name;
 

src/core/hle/service/nvflinger/buffer_queue.cpp

@@ -9,7 +9,8 @@
 #include "core/core_timing.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
 
-namespace Service::NVFlinger {
+namespace Service {
+namespace NVFlinger {
 
 BufferQueue::BufferQueue(u32 id, u64 layer_id) : id(id), layer_id(layer_id) {
     native_handle = Kernel::Event::Create(Kernel::ResetType::OneShot, "BufferQueue NativeHandle");
@@ -110,4 +111,5 @@ void BufferQueue::SetBufferWaitEvent(Kernel::SharedPtr<Kernel::Event>&& wait_eve
     buffer_wait_event = std::move(wait_event);
 }
 
-} // namespace Service::NVFlinger
+} // namespace NVFlinger
+} // namespace Service

src/core/hle/service/nvflinger/buffer_queue.h

@@ -13,7 +13,8 @@ namespace CoreTiming {
 struct EventType;
 }
 
-namespace Service::NVFlinger {
+namespace Service {
+namespace NVFlinger {
 
 struct IGBPBuffer {
     u32_le magic;
@@ -97,4 +98,5 @@ private:
     Kernel::SharedPtr<Kernel::Event> buffer_wait_event;
 };
 
-} // namespace Service::NVFlinger
+} // namespace NVFlinger
+} // namespace Service

src/core/hle/service/service.cpp

@@ -145,7 +145,7 @@ ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& co
         return ResultCode(ErrorModule::HIPC, ErrorDescription::RemoteProcessDead);
     }
     case IPC::CommandType::Control: {
-        Core::System::GetInstance().ServiceManager().InvokeControlRequest(context);
+        SM::g_service_manager->InvokeControlRequest(context);
         break;
     }
     case IPC::CommandType::Request: {
@@ -170,40 +170,42 @@ void AddNamedPort(std::string name, SharedPtr<ClientPort> port) {
 }
 
 /// Initialize ServiceManager
-void Init(std::shared_ptr<SM::ServiceManager>& sm) {
+void Init() {
     // NVFlinger needs to be accessed by several services like Vi and AppletOE so we instantiate it
     // here and pass it into the respective InstallInterfaces functions.
     auto nv_flinger = std::make_shared<NVFlinger::NVFlinger>();
 
-    SM::ServiceManager::InstallInterfaces(sm);
+    SM::g_service_manager = std::make_shared<SM::ServiceManager>();
+    SM::ServiceManager::InstallInterfaces(SM::g_service_manager);
 
-    Account::InstallInterfaces(*sm);
-    AM::InstallInterfaces(*sm, nv_flinger);
-    AOC::InstallInterfaces(*sm);
-    APM::InstallInterfaces(*sm);
-    Audio::InstallInterfaces(*sm);
-    Fatal::InstallInterfaces(*sm);
-    FileSystem::InstallInterfaces(*sm);
-    Friend::InstallInterfaces(*sm);
-    HID::InstallInterfaces(*sm);
-    LM::InstallInterfaces(*sm);
-    NFP::InstallInterfaces(*sm);
-    NIFM::InstallInterfaces(*sm);
-    NS::InstallInterfaces(*sm);
-    Nvidia::InstallInterfaces(*sm);
-    PCTL::InstallInterfaces(*sm);
-    Sockets::InstallInterfaces(*sm);
-    SPL::InstallInterfaces(*sm);
-    SSL::InstallInterfaces(*sm);
-    Time::InstallInterfaces(*sm);
-    VI::InstallInterfaces(*sm, nv_flinger);
-    Set::InstallInterfaces(*sm);
+    Account::InstallInterfaces(*SM::g_service_manager);
+    AM::InstallInterfaces(*SM::g_service_manager, nv_flinger);
+    AOC::InstallInterfaces(*SM::g_service_manager);
+    APM::InstallInterfaces(*SM::g_service_manager);
+    Audio::InstallInterfaces(*SM::g_service_manager);
+    Fatal::InstallInterfaces(*SM::g_service_manager);
+    FileSystem::InstallInterfaces(*SM::g_service_manager);
+    Friend::InstallInterfaces(*SM::g_service_manager);
+    HID::InstallInterfaces(*SM::g_service_manager);
+    LM::InstallInterfaces(*SM::g_service_manager);
+    NFP::InstallInterfaces(*SM::g_service_manager);
+    NIFM::InstallInterfaces(*SM::g_service_manager);
+    NS::InstallInterfaces(*SM::g_service_manager);
+    Nvidia::InstallInterfaces(*SM::g_service_manager);
+    PCTL::InstallInterfaces(*SM::g_service_manager);
+    Sockets::InstallInterfaces(*SM::g_service_manager);
+    SPL::InstallInterfaces(*SM::g_service_manager);
+    SSL::InstallInterfaces(*SM::g_service_manager);
+    Time::InstallInterfaces(*SM::g_service_manager);
+    VI::InstallInterfaces(*SM::g_service_manager, nv_flinger);
+    Set::InstallInterfaces(*SM::g_service_manager);
 
     LOG_DEBUG(Service, "initialized OK");
 }
 
 /// Shutdown ServiceManager
 void Shutdown() {
+    SM::g_service_manager = nullptr;
     g_kernel_named_ports.clear();
     LOG_DEBUG(Service, "shutdown OK");
 }

src/core/hle/service/service.h

@@ -178,7 +178,7 @@ private:
 };
 
 /// Initialize ServiceManager
-void Init(std::shared_ptr<SM::ServiceManager>& sm);
+void Init();
 
 /// Shutdown ServiceManager
 void Shutdown();

src/core/hle/service/sm/sm.cpp

@@ -14,8 +14,6 @@
 
 namespace Service::SM {
 
-ServiceManager::~ServiceManager() = default;
-
 void ServiceManager::InvokeControlRequest(Kernel::HLERequestContext& context) {
     controller_interface->InvokeRequest(context);
 }
@@ -74,7 +72,7 @@ ResultVal<Kernel::SharedPtr<Kernel::ClientSession>> ServiceManager::ConnectToSer
     return client_port->Connect();
 }
 
-SM::~SM() = default;
+std::shared_ptr<ServiceManager> g_service_manager;
 
 /**
  * SM::Initialize service function

src/core/hle/service/sm/sm.h

@@ -23,7 +23,7 @@ namespace Service::SM {
 class SM final : public ServiceFramework<SM> {
 public:
     SM(std::shared_ptr<ServiceManager> service_manager);
-    ~SM() override;
+    ~SM() = default;
 
 private:
     void Initialize(Kernel::HLERequestContext& ctx);
@@ -44,8 +44,6 @@ class ServiceManager {
 public:
     static void InstallInterfaces(std::shared_ptr<ServiceManager> self);
 
-    ~ServiceManager();
-
     ResultVal<Kernel::SharedPtr<Kernel::ServerPort>> RegisterService(std::string name,
                                                                      unsigned int max_sessions);
     ResultVal<Kernel::SharedPtr<Kernel::ClientPort>> GetServicePort(const std::string& name);
@@ -61,4 +59,6 @@ private:
     std::unordered_map<std::string, Kernel::SharedPtr<Kernel::ClientPort>> registered_services;
 };
 
+extern std::shared_ptr<ServiceManager> g_service_manager;
+
 } // namespace Service::SM

src/core/hle/service/vi/vi.cpp

@@ -582,7 +582,7 @@ public:
             {2203, nullptr, "SetLayerSize"},
             {2204, nullptr, "GetLayerZ"},
             {2205, &ISystemDisplayService::SetLayerZ, "SetLayerZ"},
-            {2207, &ISystemDisplayService::SetLayerVisibility, "SetLayerVisibility"},
+            {2207, nullptr, "SetLayerVisibility"},
             {2209, nullptr, "SetLayerAlpha"},
             {2312, nullptr, "CreateStrayLayer"},
             {2400, nullptr, "OpenIndirectLayer"},
@@ -632,16 +632,6 @@ private:
         IPC::ResponseBuilder rb = rp.MakeBuilder(2, 0, 0);
         rb.Push(RESULT_SUCCESS);
     }
-
-    void SetLayerVisibility(Kernel::HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        u64 layer_id = rp.Pop<u64>();
-        bool visibility = rp.Pop<bool>();
-        IPC::ResponseBuilder rb = rp.MakeBuilder(2, 0, 0);
-        rb.Push(RESULT_SUCCESS);
-        LOG_WARNING(Service_VI, "(STUBBED) called, layer_id=0x%x, visibility=%u", layer_id,
-                    visibility);
-    }
 };
 
 class IManagerDisplayService final : public ServiceFramework<IManagerDisplayService> {
@@ -673,7 +663,7 @@ public:
             {4206, nullptr, "SetDefaultDisplay"},
             {6000, &IManagerDisplayService::AddToLayerStack, "AddToLayerStack"},
             {6001, nullptr, "RemoveFromLayerStack"},
-            {6002, &IManagerDisplayService::SetLayerVisibility, "SetLayerVisibility"},
+            {6002, nullptr, "SetLayerVisibility"},
             {6003, nullptr, "SetLayerConfig"},
             {6004, nullptr, "AttachLayerPresentationTracer"},
             {6005, nullptr, "DetachLayerPresentationTracer"},
@@ -755,16 +745,6 @@ private:
         rb.Push(RESULT_SUCCESS);
     }
 
-    void SetLayerVisibility(Kernel::HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        u64 layer_id = rp.Pop<u64>();
-        bool visibility = rp.Pop<bool>();
-        IPC::ResponseBuilder rb = rp.MakeBuilder(2, 0, 0);
-        rb.Push(RESULT_SUCCESS);
-        LOG_WARNING(Service_VI, "(STUBBED) called, layer_id=0x%x, visibility=%u", layer_id,
-                    visibility);
-    }
-
     std::shared_ptr<NVFlinger::NVFlinger> nv_flinger;
 };
 
@@ -835,15 +815,15 @@ private:
         IPC::RequestParser rp{ctx};
         u64 display_id = rp.Pop<u64>();
 
-        IPC::ResponseBuilder rb = rp.MakeBuilder(6, 0, 0);
+        IPC::ResponseBuilder rb = rp.MakeBuilder(4, 0, 0);
         rb.Push(RESULT_SUCCESS);
 
         if (Settings::values.use_docked_mode) {
-            rb.Push(static_cast<u64>(DisplayResolution::DockedWidth));
-            rb.Push(static_cast<u64>(DisplayResolution::DockedHeight));
+            rb.Push(static_cast<u32>(DisplayResolution::DockedWidth));
+            rb.Push(static_cast<u32>(DisplayResolution::DockedHeight));
         } else {
-            rb.Push(static_cast<u64>(DisplayResolution::UndockedWidth));
-            rb.Push(static_cast<u64>(DisplayResolution::UndockedHeight));
+            rb.Push(static_cast<u32>(DisplayResolution::UndockedWidth));
+            rb.Push(static_cast<u32>(DisplayResolution::UndockedHeight));
         }
     }
 

src/core/loader/nro.cpp

@@ -137,7 +137,7 @@ ResultStatus AppLoader_NRO::Load(Kernel::SharedPtr<Kernel::Process>& process) {
     process->address_mappings = default_address_mappings;
     process->resource_limit =
         Kernel::ResourceLimit::GetForCategory(Kernel::ResourceLimitCategory::APPLICATION);
-    process->Run(base_addr, 48, Memory::DEFAULT_STACK_SIZE);
+    process->Run(base_addr, THREADPRIO_DEFAULT, Memory::DEFAULT_STACK_SIZE);
 
     is_loaded = true;
     return ResultStatus::Success;

src/core/loader/nso.cpp

@@ -165,7 +165,7 @@ ResultStatus AppLoader_NSO::Load(Kernel::SharedPtr<Kernel::Process>& process) {
     process->address_mappings = default_address_mappings;
     process->resource_limit =
         Kernel::ResourceLimit::GetForCategory(Kernel::ResourceLimitCategory::APPLICATION);
-    process->Run(Memory::PROCESS_IMAGE_VADDR, 48, Memory::DEFAULT_STACK_SIZE);
+    process->Run(Memory::PROCESS_IMAGE_VADDR, THREADPRIO_DEFAULT, Memory::DEFAULT_STACK_SIZE);
 
     is_loaded = true;
     return ResultStatus::Success;

src/video_core/engines/maxwell_3d.cpp

@@ -147,36 +147,11 @@ void Maxwell3D::ProcessQueryGet() {
     // VAddr before writing.
     VAddr address = memory_manager.PhysicalToVirtualAddress(sequence_address);
 
-    // TODO(Subv): Support the other query units.
-    ASSERT_MSG(regs.query.query_get.unit == Regs::QueryUnit::Crop,
-               "Units other than CROP are unimplemented");
-    ASSERT_MSG(regs.query.query_get.short_query,
-               "Writing the entire query result structure is unimplemented");
-
-    u32 value = Memory::Read32(address);
-    u32 result = 0;
-
-    // TODO(Subv): Support the other query variables
-    switch (regs.query.query_get.select) {
-    case Regs::QuerySelect::Zero:
-        result = 0;
-        break;
-    default:
-        UNIMPLEMENTED_MSG("Unimplemented query select type %u",
-                          static_cast<u32>(regs.query.query_get.select.Value()));
-    }
-
-    // TODO(Subv): Research and implement how query sync conditions work.
-
     switch (regs.query.query_get.mode) {
-    case Regs::QueryMode::Write:
-    case Regs::QueryMode::Write2: {
+    case Regs::QueryMode::Write: {
         // Write the current query sequence to the sequence address.
         u32 sequence = regs.query.query_sequence;
         Memory::Write32(address, sequence);
-
-        // TODO(Subv): Write the proper query response structure to the address when not using short
-        // mode.
         break;
     }
     default:

src/video_core/engines/maxwell_3d.h

@@ -46,29 +46,6 @@ public:
         enum class QueryMode : u32 {
             Write = 0,
             Sync = 1,
-            // TODO(Subv): It is currently unknown what the difference between method 2 and method 0
-            // is.
-            Write2 = 2,
-        };
-
-        enum class QueryUnit : u32 {
-            VFetch = 1,
-            VP = 2,
-            Rast = 4,
-            StrmOut = 5,
-            GP = 6,
-            ZCull = 7,
-            Prop = 10,
-            Crop = 15,
-        };
-
-        enum class QuerySelect : u32 {
-            Zero = 0,
-        };
-
-        enum class QuerySyncCondition : u32 {
-            NotEqual = 0,
-            GreaterThan = 1,
         };
 
         enum class ShaderProgram : u32 {
@@ -499,10 +476,7 @@ public:
                         u32 raw;
                         BitField<0, 2, QueryMode> mode;
                         BitField<4, 1, u32> fence;
-                        BitField<12, 4, QueryUnit> unit;
-                        BitField<16, 1, QuerySyncCondition> sync_cond;
-                        BitField<23, 5, QuerySelect> select;
-                        BitField<28, 1, u32> short_query;
+                        BitField<12, 4, u32> unit;
                     } query_get;
 
                     GPUVAddr QueryAddress() const {

src/video_core/engines/shader_bytecode.h

@@ -4,24 +4,15 @@
 
 #pragma once
 
-#include <bitset>
 #include <cstring>
 #include <map>
 #include <string>
-#include <vector>
-
-#include <boost/optional.hpp>
-
 #include "common/bit_field.h"
-#include "common/common_types.h"
 
 namespace Tegra {
 namespace Shader {
 
 struct Register {
-    // Register 255 is special cased to always be 0
-    static constexpr size_t ZeroIndex = 255;
-
     constexpr Register() = default;
 
     constexpr Register(u64 value) : value(value) {}
@@ -95,12 +86,181 @@ union Uniform {
     BitField<34, 5, u64> index;
 };
 
+union OpCode {
+    enum class Id : u64 {
+        TEXS = 0x6C,
+        IPA = 0xE0,
+        FMUL32_IMM = 0x1E,
+        FFMA_IMM = 0x65,
+        FFMA_CR = 0x93,
+        FFMA_RC = 0xA3,
+        FFMA_RR = 0xB3,
+
+        FADD_C = 0x98B,
+        FMUL_C = 0x98D,
+        MUFU = 0xA10,
+        FADD_R = 0xB8B,
+        FMUL_R = 0xB8D,
+        LD_A = 0x1DFB,
+        ST_A = 0x1DFE,
+
+        FSETP_R = 0x5BB,
+        FSETP_C = 0x4BB,
+        EXIT = 0xE30,
+        KIL = 0xE33,
+
+        FMUL_IMM = 0x70D,
+        FMUL_IMM_x = 0x72D,
+        FADD_IMM = 0x70B,
+        FADD_IMM_x = 0x72B,
+    };
+
+    enum class Type {
+        Trivial,
+        Arithmetic,
+        Ffma,
+        Flow,
+        Memory,
+        Unknown,
+    };
+
+    struct Info {
+        Type type;
+        std::string name;
+    };
+
+    OpCode() = default;
+
+    constexpr OpCode(Id value) : value(static_cast<u64>(value)) {}
+
+    constexpr OpCode(u64 value) : value{value} {}
+
+    constexpr Id EffectiveOpCode() const {
+        switch (op1) {
+        case Id::TEXS:
+            return op1;
+        }
+
+        switch (op2) {
+        case Id::IPA:
+        case Id::FMUL32_IMM:
+            return op2;
+        }
+
+        switch (op3) {
+        case Id::FFMA_IMM:
+        case Id::FFMA_CR:
+        case Id::FFMA_RC:
+        case Id::FFMA_RR:
+            return op3;
+        }
+
+        switch (op4) {
+        case Id::EXIT:
+        case Id::FSETP_R:
+        case Id::FSETP_C:
+        case Id::KIL:
+            return op4;
+        }
+
+        switch (op5) {
+        case Id::MUFU:
+        case Id::LD_A:
+        case Id::ST_A:
+        case Id::FADD_R:
+        case Id::FADD_C:
+        case Id::FMUL_R:
+        case Id::FMUL_C:
+            return op5;
+
+        case Id::FMUL_IMM:
+        case Id::FMUL_IMM_x:
+            return Id::FMUL_IMM;
+
+        case Id::FADD_IMM:
+        case Id::FADD_IMM_x:
+            return Id::FADD_IMM;
+        }
+
+        return static_cast<Id>(value);
+    }
+
+    static const Info& GetInfo(const OpCode& opcode) {
+        static const std::map<Id, Info> info_table{BuildInfoTable()};
+        const auto& search{info_table.find(opcode.EffectiveOpCode())};
+        if (search != info_table.end()) {
+            return search->second;
+        }
+
+        static const Info unknown{Type::Unknown, "UNK"};
+        return unknown;
+    }
+
+    constexpr operator Id() const {
+        return static_cast<Id>(value);
+    }
+
+    constexpr OpCode operator<<(size_t bits) const {
+        return value << bits;
+    }
+
+    constexpr OpCode operator>>(size_t bits) const {
+        return value >> bits;
+    }
+
+    template <typename T>
+    constexpr u64 operator-(const T& oth) const {
+        return value - oth;
+    }
+
+    constexpr u64 operator&(const OpCode& oth) const {
+        return value & oth.value;
+    }
+
+    constexpr u64 operator~() const {
+        return ~value;
+    }
+
+    static std::map<Id, Info> BuildInfoTable() {
+        std::map<Id, Info> info_table;
+        info_table[Id::TEXS] = {Type::Memory, "texs"};
+        info_table[Id::LD_A] = {Type::Memory, "ld_a"};
+        info_table[Id::ST_A] = {Type::Memory, "st_a"};
+        info_table[Id::MUFU] = {Type::Arithmetic, "mufu"};
+        info_table[Id::FFMA_IMM] = {Type::Ffma, "ffma_imm"};
+        info_table[Id::FFMA_CR] = {Type::Ffma, "ffma_cr"};
+        info_table[Id::FFMA_RC] = {Type::Ffma, "ffma_rc"};
+        info_table[Id::FFMA_RR] = {Type::Ffma, "ffma_rr"};
+        info_table[Id::FADD_R] = {Type::Arithmetic, "fadd_r"};
+        info_table[Id::FADD_C] = {Type::Arithmetic, "fadd_c"};
+        info_table[Id::FADD_IMM] = {Type::Arithmetic, "fadd_imm"};
+        info_table[Id::FMUL_R] = {Type::Arithmetic, "fmul_r"};
+        info_table[Id::FMUL_C] = {Type::Arithmetic, "fmul_c"};
+        info_table[Id::FMUL_IMM] = {Type::Arithmetic, "fmul_imm"};
+        info_table[Id::FMUL32_IMM] = {Type::Arithmetic, "fmul32_imm"};
+        info_table[Id::FSETP_C] = {Type::Arithmetic, "fsetp_c"};
+        info_table[Id::FSETP_R] = {Type::Arithmetic, "fsetp_r"};
+        info_table[Id::EXIT] = {Type::Trivial, "exit"};
+        info_table[Id::IPA] = {Type::Trivial, "ipa"};
+        info_table[Id::KIL] = {Type::Flow, "kil"};
+        return info_table;
+    }
+
+    BitField<57, 7, Id> op1;
+    BitField<56, 8, Id> op2;
+    BitField<55, 9, Id> op3;
+    BitField<52, 12, Id> op4;
+    BitField<51, 13, Id> op5;
+    u64 value{};
+};
+static_assert(sizeof(OpCode) == 0x8, "Incorrect structure size");
+
 } // namespace Shader
 } // namespace Tegra
 
 namespace std {
 
-// TODO(bunnei): The below is forbidden by the C++ standard, but works fine. See #330.
+// TODO(bunne): The below is forbidden by the C++ standard, but works fine. See #330.
 template <>
 struct make_unsigned<Tegra::Shader::Attribute> {
     using type = Tegra::Shader::Attribute;
@@ -111,6 +271,11 @@ struct make_unsigned<Tegra::Shader::Register> {
     using type = Tegra::Shader::Register;
 };
 
+template <>
+struct make_unsigned<Tegra::Shader::OpCode> {
+    using type = Tegra::Shader::OpCode;
+};
+
 } // namespace std
 
 namespace Tegra {
@@ -118,23 +283,7 @@ namespace Shader {
 
 enum class Pred : u64 {
     UnusedIndex = 0x7,
-    NeverExecute = 0xF,
-};
-
-enum class PredCondition : u64 {
-    LessThan = 1,
-    Equal = 2,
-    LessEqual = 3,
-    GreaterThan = 4,
-    NotEqual = 5,
-    GreaterEqual = 6,
-    // TODO(Subv): Other condition types
-};
-
-enum class PredOperation : u64 {
-    And = 0,
-    Or = 1,
-    Xor = 2,
+    NeverExecute = 0xf,
 };
 
 enum class SubOp : u64 {
@@ -149,24 +298,18 @@ enum class SubOp : u64 {
 
 union Instruction {
     Instruction& operator=(const Instruction& instr) {
-        value = instr.value;
+        hex = instr.hex;
         return *this;
     }
 
-    constexpr Instruction(u64 value) : value{value} {}
-
+    OpCode opcode;
     BitField<0, 8, Register> gpr0;
     BitField<8, 8, Register> gpr8;
-    union {
-        BitField<16, 4, Pred> full_pred;
-        BitField<16, 3, u64> pred_index;
-    } pred;
-    BitField<19, 1, u64> negate_pred;
+    BitField<16, 4, Pred> pred;
     BitField<20, 8, Register> gpr20;
     BitField<20, 7, SubOp> sub_op;
     BitField<28, 8, Register> gpr28;
     BitField<39, 8, Register> gpr39;
-    BitField<48, 16, u64> opcode;
 
     union {
         BitField<20, 19, u64> imm20_19;
@@ -200,20 +343,6 @@ union Instruction {
         BitField<49, 1, u64> negate_c;
     } ffma;
 
-    union {
-        BitField<0, 3, u64> pred0;
-        BitField<3, 3, u64> pred3;
-        BitField<7, 1, u64> abs_a;
-        BitField<39, 3, u64> pred39;
-        BitField<42, 1, u64> neg_pred;
-        BitField<43, 1, u64> neg_a;
-        BitField<44, 1, u64> abs_b;
-        BitField<45, 2, PredOperation> op;
-        BitField<47, 1, u64> ftz;
-        BitField<48, 4, PredCondition> cond;
-        BitField<56, 1, u64> neg_b;
-    } fsetp;
-
     BitField<61, 1, u64> is_b_imm;
     BitField<60, 1, u64> is_b_gpr;
     BitField<59, 1, u64> is_c_gpr;
@@ -222,218 +351,11 @@ union Instruction {
     Uniform uniform;
     Sampler sampler;
 
-    u64 value;
+    u64 hex;
 };
 static_assert(sizeof(Instruction) == 0x8, "Incorrect structure size");
 static_assert(std::is_standard_layout<Instruction>::value,
               "Structure does not have standard layout");
 
-class OpCode {
-public:
-    enum class Id {
-        KIL,
-        LD_A,
-        ST_A,
-        TEXQ, // Texture Query
-        TEXS, // Texture Fetch with scalar/non-vec4 source/destinations
-        TLDS, // Texture Load with scalar/non-vec4 source/destinations
-        EXIT,
-        IPA,
-        FFMA_IMM, // Fused Multiply and Add
-        FFMA_CR,
-        FFMA_RC,
-        FFMA_RR,
-        FADD_C,
-        FADD_R,
-        FADD_IMM,
-        FMUL_C,
-        FMUL_R,
-        FMUL_IMM,
-        FMUL32_IMM,
-        MUFU, // Multi-Function Operator
-        RRO,  // Range Reduction Operator
-        F2F_C,
-        F2F_R,
-        F2F_IMM,
-        F2I_C,
-        F2I_R,
-        F2I_IMM,
-        I2F_C,
-        I2F_R,
-        I2F_IMM,
-        LOP32I,
-        MOV_C,
-        MOV_R,
-        MOV_IMM,
-        MOV32I,
-        SHR_C,
-        SHR_R,
-        SHR_IMM,
-        FSETP_C, // Set Predicate
-        FSETP_R,
-        FSETP_IMM,
-        ISETP_C,
-        ISETP_IMM,
-        ISETP_R,
-    };
-
-    enum class Type {
-        Trivial,
-        Arithmetic,
-        Ffma,
-        Flow,
-        Memory,
-        FloatPredicate,
-        IntegerPredicate,
-        Unknown,
-    };
-
-    class Matcher {
-    public:
-        Matcher(const char* const name, u16 mask, u16 expected, OpCode::Id id, OpCode::Type type)
-            : name{name}, mask{mask}, expected{expected}, id{id}, type{type} {}
-
-        const char* GetName() const {
-            return name;
-        }
-
-        u16 GetMask() const {
-            return mask;
-        }
-
-        Id GetId() const {
-            return id;
-        }
-
-        Type GetType() const {
-            return type;
-        }
-
-        /**
-         * Tests to see if the given instruction is the instruction this matcher represents.
-         * @param instruction The instruction to test
-         * @returns true if the given instruction matches.
-         */
-        bool Matches(u16 instruction) const {
-            return (instruction & mask) == expected;
-        }
-
-    private:
-        const char* name;
-        u16 mask;
-        u16 expected;
-        Id id;
-        Type type;
-    };
-
-    static boost::optional<const Matcher&> Decode(Instruction instr) {
-        static const auto table{GetDecodeTable()};
-
-        const auto matches_instruction = [instr](const auto& matcher) {
-            return matcher.Matches(static_cast<u16>(instr.opcode));
-        };
-
-        auto iter = std::find_if(table.begin(), table.end(), matches_instruction);
-        return iter != table.end() ? boost::optional<const Matcher&>(*iter) : boost::none;
-    }
-
-private:
-    struct Detail {
-    private:
-        static constexpr size_t opcode_bitsize = 16;
-
-        /**
-         * Generates the mask and the expected value after masking from a given bitstring.
-         * A '0' in a bitstring indicates that a zero must be present at that bit position.
-         * A '1' in a bitstring indicates that a one must be present at that bit position.
-         */
-        static auto GetMaskAndExpect(const char* const bitstring) {
-            u16 mask = 0, expect = 0;
-            for (size_t i = 0; i < opcode_bitsize; i++) {
-                const size_t bit_position = opcode_bitsize - i - 1;
-                switch (bitstring[i]) {
-                case '0':
-                    mask |= 1 << bit_position;
-                    break;
-                case '1':
-                    expect |= 1 << bit_position;
-                    mask |= 1 << bit_position;
-                    break;
-                default:
-                    // Ignore
-                    break;
-                }
-            }
-            return std::make_tuple(mask, expect);
-        }
-
-    public:
-        /// Creates a matcher that can match and parse instructions based on bitstring.
-        static auto GetMatcher(const char* const bitstring, OpCode::Id op, OpCode::Type type,
-                               const char* const name) {
-            const auto mask_expect = GetMaskAndExpect(bitstring);
-            return Matcher(name, std::get<0>(mask_expect), std::get<1>(mask_expect), op, type);
-        }
-    };
-
-    static std::vector<Matcher> GetDecodeTable() {
-        std::vector<Matcher> table = {
-#define INST(bitstring, op, type, name) Detail::GetMatcher(bitstring, op, type, name)
-            INST("111000110011----", Id::KIL, Type::Flow, "KIL"),
-            INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"),
-            INST("1110111111110---", Id::ST_A, Type::Memory, "ST_A"),
-            INST("1101111101001---", Id::TEXQ, Type::Memory, "TEXQ"),
-            INST("1101100---------", Id::TEXS, Type::Memory, "TEXS"),
-            INST("1101101---------", Id::TLDS, Type::Memory, "TLDS"),
-            INST("111000110000----", Id::EXIT, Type::Trivial, "EXIT"),
-            INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
-            INST("001100101-------", Id::FFMA_IMM, Type::Ffma, "FFMA_IMM"),
-            INST("010010011-------", Id::FFMA_CR, Type::Ffma, "FFMA_CR"),
-            INST("010100011-------", Id::FFMA_RC, Type::Ffma, "FFMA_RC"),
-            INST("010110011-------", Id::FFMA_RR, Type::Ffma, "FFMA_RR"),
-            INST("0100110001011---", Id::FADD_C, Type::Arithmetic, "FADD_C"),
-            INST("0101110001011---", Id::FADD_R, Type::Arithmetic, "FADD_R"),
-            INST("0011100-01011---", Id::FADD_IMM, Type::Arithmetic, "FADD_IMM"),
-            INST("0100110001101---", Id::FMUL_C, Type::Arithmetic, "FMUL_C"),
-            INST("0101110001101---", Id::FMUL_R, Type::Arithmetic, "FMUL_R"),
-            INST("0011100-01101---", Id::FMUL_IMM, Type::Arithmetic, "FMUL_IMM"),
-            INST("00011110--------", Id::FMUL32_IMM, Type::Arithmetic, "FMUL32_IMM"),
-            INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"),
-            INST("0101110010010---", Id::RRO, Type::Arithmetic, "RRO"),
-            INST("0100110010101---", Id::F2F_C, Type::Arithmetic, "F2F_C"),
-            INST("0101110010101---", Id::F2F_R, Type::Arithmetic, "F2F_R"),
-            INST("0011100-10101---", Id::F2F_IMM, Type::Arithmetic, "F2F_IMM"),
-            INST("0100110010110---", Id::F2I_C, Type::Arithmetic, "F2I_C"),
-            INST("0101110010110---", Id::F2I_R, Type::Arithmetic, "F2I_R"),
-            INST("0011100-10110---", Id::F2I_IMM, Type::Arithmetic, "F2I_IMM"),
-            INST("0100110010111---", Id::I2F_C, Type::Arithmetic, "I2F_C"),
-            INST("0101110010111---", Id::I2F_R, Type::Arithmetic, "I2F_R"),
-            INST("0011100-10111---", Id::I2F_IMM, Type::Arithmetic, "I2F_IMM"),
-            INST("000001----------", Id::LOP32I, Type::Arithmetic, "LOP32I"),
-            INST("0100110010011---", Id::MOV_C, Type::Arithmetic, "MOV_C"),
-            INST("0101110010011---", Id::MOV_R, Type::Arithmetic, "MOV_R"),
-            INST("0011100-10011---", Id::MOV_IMM, Type::Arithmetic, "MOV_IMM"),
-            INST("000000010000----", Id::MOV32I, Type::Arithmetic, "MOV32I"),
-            INST("0100110000101---", Id::SHR_C, Type::Arithmetic, "SHR_C"),
-            INST("0101110000101---", Id::SHR_R, Type::Arithmetic, "SHR_R"),
-            INST("0011100-00101---", Id::SHR_IMM, Type::Arithmetic, "SHR_IMM"),
-            INST("010010111011----", Id::FSETP_C, Type::FloatPredicate, "FSETP_C"),
-            INST("010110111011----", Id::FSETP_R, Type::FloatPredicate, "FSETP_R"),
-            INST("0011011-1011----", Id::FSETP_IMM, Type::FloatPredicate, "FSETP_IMM"),
-            INST("010010110110----", Id::ISETP_C, Type::IntegerPredicate, "ISETP_C"),
-            INST("010110110110----", Id::ISETP_R, Type::IntegerPredicate, "ISETP_R"),
-            INST("0011011-0110----", Id::ISETP_IMM, Type::IntegerPredicate, "ISETP_IMM"),
-        };
-#undef INST
-        std::stable_sort(table.begin(), table.end(), [](const auto& a, const auto& b) {
-            // If a matcher has more bits in its mask it is more specific, so it
-            // should come first.
-            return std::bitset<16>(a.GetMask()).count() > std::bitset<16>(b.GetMask()).count();
-        });
-
-        return table;
-    }
-};
-
 } // namespace Shader
 } // namespace Tegra

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -14,6 +14,7 @@
 #include "common/math_util.h"
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
+#include "common/vector_math.h"
 #include "core/core.h"
 #include "core/hle/kernel/process.h"
 #include "core/settings.h"

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -6,10 +6,15 @@
 
 #include <array>
 #include <cstddef>
+#include <cstring>
 #include <memory>
+#include <unordered_map>
 #include <vector>
 #include <glad/glad.h>
+#include "common/bit_field.h"
 #include "common/common_types.h"
+#include "common/hash.h"
+#include "common/vector_math.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -7,6 +7,7 @@
 #include <cstring>
 #include <iterator>
 #include <memory>
+#include <unordered_set>
 #include <utility>
 #include <vector>
 #include <boost/optional.hpp>
@@ -19,6 +20,7 @@
 #include "common/math_util.h"
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
+#include "common/vector_math.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
 #include "core/hle/kernel/process.h"
@@ -670,8 +672,7 @@ void CachedSurface::DownloadGLTexture(const MathUtil::Rectangle<u32>& rect, GLui
     glPixelStorei(GL_PACK_ROW_LENGTH, 0);
 }
 
-enum class MatchFlags {
-    None = 0,
+enum MatchFlags {
     Invalid = 1,      // Flag that can be applied to other match types, invalid matches require
                       // validation before they can be used
     Exact = 1 << 1,   // Surfaces perfectly match
@@ -685,10 +686,6 @@ constexpr MatchFlags operator|(MatchFlags lhs, MatchFlags rhs) {
     return static_cast<MatchFlags>(static_cast<int>(lhs) | static_cast<int>(rhs));
 }
 
-constexpr MatchFlags operator&(MatchFlags lhs, MatchFlags rhs) {
-    return static_cast<MatchFlags>(static_cast<int>(lhs) & static_cast<int>(rhs));
-}
-
 /// Get the best surface match (and its match type) for the given flags
 template <MatchFlags find_flags>
 Surface FindMatch(const SurfaceCache& surface_cache, const SurfaceParams& params,
@@ -706,15 +703,15 @@ Surface FindMatch(const SurfaceCache& surface_cache, const SurfaceParams& params
                                          : (params.res_scale <= surface->res_scale);
             // validity will be checked in GetCopyableInterval
             bool is_valid =
-                (find_flags & MatchFlags::Copy) != MatchFlags::None
+                find_flags & MatchFlags::Copy
                     ? true
                     : surface->IsRegionValid(validate_interval.value_or(params.GetInterval()));
 
-            if ((find_flags & MatchFlags::Invalid) == MatchFlags::None && !is_valid)
+            if (!(find_flags & MatchFlags::Invalid) && !is_valid)
                 continue;
 
             auto IsMatch_Helper = [&](auto check_type, auto match_fn) {
-                if ((find_flags & check_type) == MatchFlags::None)
+                if (!(find_flags & check_type))
                     return;
 
                 bool matched;

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -14,13 +14,13 @@ class OGLTexture : private NonCopyable {
 public:
     OGLTexture() = default;
 
-    OGLTexture(OGLTexture&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+    OGLTexture(OGLTexture&& o) : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLTexture() {
         Release();
     }
 
-    OGLTexture& operator=(OGLTexture&& o) noexcept {
+    OGLTexture& operator=(OGLTexture&& o) {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -49,13 +49,13 @@ class OGLSampler : private NonCopyable {
 public:
     OGLSampler() = default;
 
-    OGLSampler(OGLSampler&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+    OGLSampler(OGLSampler&& o) : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLSampler() {
         Release();
     }
 
-    OGLSampler& operator=(OGLSampler&& o) noexcept {
+    OGLSampler& operator=(OGLSampler&& o) {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -84,13 +84,13 @@ class OGLShader : private NonCopyable {
 public:
     OGLShader() = default;
 
-    OGLShader(OGLShader&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+    OGLShader(OGLShader&& o) : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLShader() {
         Release();
     }
 
-    OGLShader& operator=(OGLShader&& o) noexcept {
+    OGLShader& operator=(OGLShader&& o) {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -118,13 +118,13 @@ class OGLProgram : private NonCopyable {
 public:
     OGLProgram() = default;
 
-    OGLProgram(OGLProgram&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+    OGLProgram(OGLProgram&& o) : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLProgram() {
         Release();
     }
 
-    OGLProgram& operator=(OGLProgram&& o) noexcept {
+    OGLProgram& operator=(OGLProgram&& o) {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -165,12 +165,13 @@ public:
 class OGLPipeline : private NonCopyable {
 public:
     OGLPipeline() = default;
-    OGLPipeline(OGLPipeline&& o) noexcept : handle{std::exchange<GLuint>(o.handle, 0)} {}
-
+    OGLPipeline(OGLPipeline&& o) {
+        handle = std::exchange<GLuint>(o.handle, 0);
+    }
     ~OGLPipeline() {
         Release();
     }
-    OGLPipeline& operator=(OGLPipeline&& o) noexcept {
+    OGLPipeline& operator=(OGLPipeline&& o) {
         handle = std::exchange<GLuint>(o.handle, 0);
         return *this;
     }
@@ -198,13 +199,13 @@ class OGLBuffer : private NonCopyable {
 public:
     OGLBuffer() = default;
 
-    OGLBuffer(OGLBuffer&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+    OGLBuffer(OGLBuffer&& o) : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLBuffer() {
         Release();
     }
 
-    OGLBuffer& operator=(OGLBuffer&& o) noexcept {
+    OGLBuffer& operator=(OGLBuffer&& o) {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -233,12 +234,12 @@ class OGLSync : private NonCopyable {
 public:
     OGLSync() = default;
 
-    OGLSync(OGLSync&& o) noexcept : handle(std::exchange(o.handle, nullptr)) {}
+    OGLSync(OGLSync&& o) : handle(std::exchange(o.handle, nullptr)) {}
 
     ~OGLSync() {
         Release();
     }
-    OGLSync& operator=(OGLSync&& o) noexcept {
+    OGLSync& operator=(OGLSync&& o) {
         Release();
         handle = std::exchange(o.handle, nullptr);
         return *this;
@@ -266,13 +267,13 @@ class OGLVertexArray : private NonCopyable {
 public:
     OGLVertexArray() = default;
 
-    OGLVertexArray(OGLVertexArray&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+    OGLVertexArray(OGLVertexArray&& o) : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLVertexArray() {
         Release();
     }
 
-    OGLVertexArray& operator=(OGLVertexArray&& o) noexcept {
+    OGLVertexArray& operator=(OGLVertexArray&& o) {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -301,13 +302,13 @@ class OGLFramebuffer : private NonCopyable {
 public:
     OGLFramebuffer() = default;
 
-    OGLFramebuffer(OGLFramebuffer&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+    OGLFramebuffer(OGLFramebuffer&& o) : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLFramebuffer() {
         Release();
     }
 
-    OGLFramebuffer& operator=(OGLFramebuffer&& o) noexcept {
+    OGLFramebuffer& operator=(OGLFramebuffer&& o) {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -97,12 +97,11 @@ private:
             return exit_method;
 
         for (u32 offset = begin; offset != end && offset != PROGRAM_END; ++offset) {
-            if (const auto opcode = OpCode::Decode({program_code[offset]})) {
-                switch (opcode->GetId()) {
-                case OpCode::Id::EXIT: {
-                    return exit_method = ExitMethod::AlwaysEnd;
-                }
-                }
+            const Instruction instr = {program_code[offset]};
+            switch (instr.opcode.EffectiveOpCode()) {
+            case OpCode::Id::EXIT: {
+                return exit_method = ExitMethod::AlwaysEnd;
+            }
             }
         }
         return exit_method = ExitMethod::AlwaysReturn;
@@ -221,8 +220,6 @@ private:
 
     /// Generates code representing a temporary (GPR) register.
     std::string GetRegister(const Register& reg, unsigned elem = 0) {
-        if (reg == Register::ZeroIndex)
-            return "0";
         if (stage == Maxwell3D::Regs::ShaderStage::Fragment && reg < 4) {
             // GPRs 0-3 are output color for the fragment shader
             return std::string{"color."} + "rgba"[(reg + elem) & 3];
@@ -279,52 +276,6 @@ private:
         shader.AddLine(dest + " = " + src + ";");
     }
 
-    /*
-     * Writes code that assigns a predicate boolean variable.
-     * @param pred The id of the predicate to write to.
-     * @param value The expression value to assign to the predicate.
-     */
-    void SetPredicate(u64 pred, const std::string& value) {
-        using Tegra::Shader::Pred;
-        // Can't assign to the constant predicate.
-        ASSERT(pred != static_cast<u64>(Pred::UnusedIndex));
-
-        std::string variable = 'p' + std::to_string(pred);
-        shader.AddLine(variable + " = " + value + ';');
-        declr_predicates.insert(std::move(variable));
-    }
-
-    /*
-     * Returns the condition to use in the 'if' for a predicated instruction.
-     * @param instr Instruction to generate the if condition for.
-     * @returns string containing the predicate condition.
-     */
-    std::string GetPredicateCondition(Instruction instr) const {
-        using Tegra::Shader::Pred;
-        ASSERT(instr.pred.pred_index != static_cast<u64>(Pred::UnusedIndex));
-
-        std::string variable =
-            'p' + std::to_string(static_cast<u64>(instr.pred.pred_index.Value()));
-
-        if (instr.negate_pred) {
-            return "!(" + variable + ')';
-        }
-
-        return variable;
-    }
-
-    /*
-     * Returns whether the instruction at the specified offset is a 'sched' instruction.
-     * Sched instructions always appear before a sequence of 3 instructions.
-     */
-    bool IsSchedInstruction(u32 offset) const {
-        // sched instructions appear once every 4 instructions.
-        static constexpr size_t SchedPeriod = 4;
-        u32 absolute_offset = offset - main_offset;
-
-        return (absolute_offset % SchedPeriod) == 0;
-    }
-
     /**
      * Compiles a single instruction from Tegra to GLSL.
      * @param offset the offset of the Tegra shader instruction.
@@ -332,33 +283,11 @@ private:
      * + 1. If the current instruction always terminates the program, returns PROGRAM_END.
      */
     u32 CompileInstr(u32 offset) {
-        // Ignore sched instructions when generating code.
-        if (IsSchedInstruction(offset)) {
-            return offset + 1;
-        }
-
         const Instruction instr = {program_code[offset]};
-        const auto opcode = OpCode::Decode(instr);
 
-        // Decoding failure
-        if (!opcode) {
-            NGLOG_CRITICAL(HW_GPU, "Unhandled instruction: {0:x}", instr.value);
-            UNREACHABLE();
-        }
+        shader.AddLine("// " + std::to_string(offset) + ": " + OpCode::GetInfo(instr.opcode).name);
 
-        shader.AddLine("// " + std::to_string(offset) + ": " + opcode->GetName());
-
-        using Tegra::Shader::Pred;
-        ASSERT_MSG(instr.pred.full_pred != Pred::NeverExecute,
-                   "NeverExecute predicate not implemented");
-
-        if (instr.pred.pred_index != static_cast<u64>(Pred::UnusedIndex)) {
-            shader.AddLine("if (" + GetPredicateCondition(instr) + ')');
-            shader.AddLine('{');
-            ++shader.scope;
-        }
-
-        switch (opcode->GetType()) {
+        switch (OpCode::GetInfo(instr.opcode).type) {
         case OpCode::Type::Arithmetic: {
             std::string dest = GetRegister(instr.gpr0);
             std::string op_a = instr.alu.negate_a ? "-" : "";
@@ -383,7 +312,7 @@ private:
                 op_b = "abs(" + op_b + ")";
             }
 
-            switch (opcode->GetId()) {
+            switch (instr.opcode.EffectiveOpCode()) {
             case OpCode::Id::FMUL_C:
             case OpCode::Id::FMUL_R:
             case OpCode::Id::FMUL_IMM: {
@@ -425,18 +354,16 @@ private:
                     SetDest(0, dest, "min(" + op_a + "," + op_b + ")", 1, 1, instr.alu.abs_d);
                     break;
                 default:
-                    NGLOG_CRITICAL(HW_GPU, "Unhandled MUFU sub op: {0:x}",
+                    NGLOG_CRITICAL(HW_GPU, "Unhandled MUFU sub op: {}",
                                    static_cast<unsigned>(instr.sub_op.Value()));
                     UNREACHABLE();
                 }
                 break;
             }
-            case OpCode::Id::RRO: {
-                NGLOG_DEBUG(HW_GPU, "Skipping RRO instruction");
-                break;
-            }
             default: {
-                NGLOG_CRITICAL(HW_GPU, "Unhandled arithmetic instruction: {}", opcode->GetName());
+                NGLOG_CRITICAL(HW_GPU, "Unhandled arithmetic instruction: {} ({}): {}",
+                               static_cast<unsigned>(instr.opcode.EffectiveOpCode()),
+                               OpCode::GetInfo(instr.opcode).name, instr.hex);
                 UNREACHABLE();
             }
             }
@@ -448,7 +375,7 @@ private:
             std::string op_b = instr.ffma.negate_b ? "-" : "";
             std::string op_c = instr.ffma.negate_c ? "-" : "";
 
-            switch (opcode->GetId()) {
+            switch (instr.opcode.EffectiveOpCode()) {
             case OpCode::Id::FFMA_CR: {
                 op_b += GetUniform(instr.uniform);
                 op_c += GetRegister(instr.gpr39);
@@ -470,7 +397,9 @@ private:
                 break;
             }
             default: {
-                NGLOG_CRITICAL(HW_GPU, "Unhandled FFMA instruction: {}", opcode->GetName());
+                NGLOG_CRITICAL(HW_GPU, "Unhandled FFMA instruction: {} ({}): {}",
+                               static_cast<unsigned>(instr.opcode.EffectiveOpCode()),
+                               OpCode::GetInfo(instr.opcode).name, instr.hex);
                 UNREACHABLE();
             }
             }
@@ -482,7 +411,7 @@ private:
             std::string gpr0 = GetRegister(instr.gpr0);
             const Attribute::Index attribute = instr.attribute.fmt20.index;
 
-            switch (opcode->GetId()) {
+            switch (instr.opcode.EffectiveOpCode()) {
             case OpCode::Id::LD_A: {
                 ASSERT_MSG(instr.attribute.fmt20.size == 0, "untested");
                 SetDest(instr.attribute.fmt20.element, gpr0, GetInputAttribute(attribute), 1, 4);
@@ -513,76 +442,22 @@ private:
                 break;
             }
             default: {
-                NGLOG_CRITICAL(HW_GPU, "Unhandled memory instruction: {}", opcode->GetName());
+                NGLOG_CRITICAL(HW_GPU, "Unhandled memory instruction: {} ({}): {}",
+                               static_cast<unsigned>(instr.opcode.EffectiveOpCode()),
+                               OpCode::GetInfo(instr.opcode).name, instr.hex);
                 UNREACHABLE();
             }
             }
             break;
         }
-        case OpCode::Type::FloatPredicate: {
-            std::string op_a = instr.fsetp.neg_a ? "-" : "";
-            op_a += GetRegister(instr.gpr8);
 
-            if (instr.fsetp.abs_a) {
-                op_a = "abs(" + op_a + ')';
-            }
-
-            std::string op_b{};
-
-            if (instr.is_b_imm) {
-                if (instr.fsetp.neg_b) {
-                    // Only the immediate version of fsetp has a neg_b bit.
-                    op_b += '-';
-                }
-                op_b += '(' + GetImmediate19(instr) + ')';
-            } else {
-                if (instr.is_b_gpr) {
-                    op_b += GetRegister(instr.gpr20);
-                } else {
-                    op_b += GetUniform(instr.uniform);
-                }
-            }
-
-            if (instr.fsetp.abs_b) {
-                op_b = "abs(" + op_b + ')';
-            }
-
-            using Tegra::Shader::Pred;
-            ASSERT_MSG(instr.fsetp.pred0 == static_cast<u64>(Pred::UnusedIndex) &&
-                           instr.fsetp.pred39 == static_cast<u64>(Pred::UnusedIndex),
-                       "Compound predicates are not implemented");
-
-            // We can't use the constant predicate as destination.
-            ASSERT(instr.fsetp.pred3 != static_cast<u64>(Pred::UnusedIndex));
-
-            using Tegra::Shader::PredCondition;
-            switch (instr.fsetp.cond) {
-            case PredCondition::LessThan:
-                SetPredicate(instr.fsetp.pred3, '(' + op_a + ") < (" + op_b + ')');
-                break;
-            case PredCondition::Equal:
-                SetPredicate(instr.fsetp.pred3, '(' + op_a + ") == (" + op_b + ')');
-                break;
-            default:
-                NGLOG_CRITICAL(HW_GPU, "Unhandled predicate condition: {} (a: {}, b: {})",
-                               static_cast<unsigned>(instr.fsetp.cond.Value()), op_a, op_b);
-                UNREACHABLE();
-            }
-            break;
-        }
         default: {
-            switch (opcode->GetId()) {
+            switch (instr.opcode.EffectiveOpCode()) {
             case OpCode::Id::EXIT: {
-                ASSERT_MSG(instr.pred.pred_index == static_cast<u64>(Pred::UnusedIndex),
-                           "Predicated exits not implemented");
                 shader.AddLine("return true;");
                 offset = PROGRAM_END - 1;
                 break;
             }
-            case OpCode::Id::KIL: {
-                shader.AddLine("discard;");
-                break;
-            }
             case OpCode::Id::IPA: {
                 const auto& attribute = instr.attribute.fmt28;
                 std::string dest = GetRegister(instr.gpr0);
@@ -590,7 +465,9 @@ private:
                 break;
             }
             default: {
-                NGLOG_CRITICAL(HW_GPU, "Unhandled instruction: {}", opcode->GetName());
+                NGLOG_CRITICAL(HW_GPU, "Unhandled instruction: {} ({}): {}",
+                               static_cast<unsigned>(instr.opcode.EffectiveOpCode()),
+                               OpCode::GetInfo(instr.opcode).name, instr.hex);
                 UNREACHABLE();
             }
             }
@@ -599,12 +476,6 @@ private:
         }
         }
 
-        // Close the predicate condition scope.
-        if (instr.pred.pred_index != static_cast<u64>(Pred::UnusedIndex)) {
-            --shader.scope;
-            shader.AddLine('}');
-        }
-
         return offset + 1;
     }
 
@@ -734,12 +605,6 @@ private:
             declarations.AddNewLine();
             ++const_buffer_layout;
         }
-
-        declarations.AddNewLine();
-        for (const auto& pred : declr_predicates) {
-            declarations.AddLine("bool " + pred + " = false;");
-        }
-        declarations.AddNewLine();
     }
 
 private:
@@ -753,7 +618,6 @@ private:
 
     // Declarations
     std::set<std::string> declr_register;
-    std::set<std::string> declr_predicates;
     std::set<Attribute::Index> declr_input_attribute;
     std::set<Attribute::Index> declr_output_attribute;
     std::array<ConstBufferEntry, Maxwell3D::Regs::MaxConstBuffers> declr_const_buffers;

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -9,10 +9,13 @@
 #include <memory>
 #include <glad/glad.h>
 #include "common/assert.h"
+#include "common/bit_field.h"
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/frontend/emu_window.h"
+#include "core/hw/hw.h"
+#include "core/hw/lcd.h"
 #include "core/memory.h"
 #include "core/settings.h"
 #include "core/tracer/recorder.h"

src/yuzu/debugger/wait_tree.cpp

@@ -6,8 +6,8 @@
 #include "yuzu/util/util.h"
 
 #include "core/core.h"
-#include "core/hle/kernel/condition_variable.h"
 #include "core/hle/kernel/event.h"
+#include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/mutex.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/timer.h"
@@ -67,6 +67,29 @@ QString WaitTreeText::GetText() const {
     return text;
 }
 
+WaitTreeMutexInfo::WaitTreeMutexInfo(VAddr mutex_address) : mutex_address(mutex_address) {
+    mutex_value = Memory::Read32(mutex_address);
+    owner_handle = static_cast<Kernel::Handle>(mutex_value & Kernel::Mutex::MutexOwnerMask);
+    owner = Kernel::g_handle_table.Get<Kernel::Thread>(owner_handle);
+}
+
+QString WaitTreeMutexInfo::GetText() const {
+    return tr("waiting for mutex 0x%1").arg(mutex_address, 16, 16, QLatin1Char('0'));
+}
+
+std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeMutexInfo::GetChildren() const {
+    std::vector<std::unique_ptr<WaitTreeItem>> list;
+
+    bool has_waiters = (mutex_value & Kernel::Mutex::MutexHasWaitersFlag) != 0;
+
+    list.push_back(std::make_unique<WaitTreeText>(tr("has waiters: %1").arg(has_waiters)));
+    list.push_back(std::make_unique<WaitTreeText>(
+        tr("owner handle: 0x%1").arg(owner_handle, 8, 16, QLatin1Char('0'))));
+    if (owner != nullptr)
+        list.push_back(std::make_unique<WaitTreeThread>(*owner));
+    return list;
+}
+
 WaitTreeWaitObject::WaitTreeWaitObject(const Kernel::WaitObject& o) : object(o) {}
 
 bool WaitTreeExpandableItem::IsExpandable() const {
@@ -84,11 +107,6 @@ std::unique_ptr<WaitTreeWaitObject> WaitTreeWaitObject::make(const Kernel::WaitO
     switch (object.GetHandleType()) {
     case Kernel::HandleType::Event:
         return std::make_unique<WaitTreeEvent>(static_cast<const Kernel::Event&>(object));
-    case Kernel::HandleType::Mutex:
-        return std::make_unique<WaitTreeMutex>(static_cast<const Kernel::Mutex&>(object));
-    case Kernel::HandleType::ConditionVariable:
-        return std::make_unique<WaitTreeConditionVariable>(
-            static_cast<const Kernel::ConditionVariable&>(object));
     case Kernel::HandleType::Timer:
         return std::make_unique<WaitTreeTimer>(static_cast<const Kernel::Timer&>(object));
     case Kernel::HandleType::Thread:
@@ -160,6 +178,9 @@ QString WaitTreeThread::GetText() const {
     case THREADSTATUS_WAIT_SYNCH_ANY:
         status = tr("waiting for objects");
         break;
+    case THREADSTATUS_WAIT_MUTEX:
+        status = tr("waiting for mutex");
+        break;
     case THREADSTATUS_DORMANT:
         status = tr("dormant");
         break;
@@ -186,6 +207,7 @@ QColor WaitTreeThread::GetColor() const {
         return QColor(Qt::GlobalColor::darkYellow);
     case THREADSTATUS_WAIT_SYNCH_ALL:
     case THREADSTATUS_WAIT_SYNCH_ANY:
+    case THREADSTATUS_WAIT_MUTEX:
         return QColor(Qt::GlobalColor::red);
     case THREADSTATUS_DORMANT:
         return QColor(Qt::GlobalColor::darkCyan);
@@ -225,11 +247,11 @@ std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeThread::GetChildren() const {
     list.push_back(std::make_unique<WaitTreeText>(
         tr("last running ticks = %1").arg(thread.last_running_ticks)));
 
-    if (thread.held_mutexes.empty()) {
-        list.push_back(std::make_unique<WaitTreeText>(tr("not holding mutex")));
-    } else {
-        list.push_back(std::make_unique<WaitTreeMutexList>(thread.held_mutexes));
-    }
+    if (thread.mutex_wait_address != 0)
+        list.push_back(std::make_unique<WaitTreeMutexInfo>(thread.mutex_wait_address));
+    else
+        list.push_back(std::make_unique<WaitTreeText>(tr("not waiting for mutex")));
+
     if (thread.status == THREADSTATUS_WAIT_SYNCH_ANY ||
         thread.status == THREADSTATUS_WAIT_SYNCH_ALL) {
         list.push_back(std::make_unique<WaitTreeObjectList>(thread.wait_objects,
@@ -250,33 +272,6 @@ std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeEvent::GetChildren() const {
     return list;
 }
 
-WaitTreeMutex::WaitTreeMutex(const Kernel::Mutex& object) : WaitTreeWaitObject(object) {}
-
-std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeMutex::GetChildren() const {
-    std::vector<std::unique_ptr<WaitTreeItem>> list(WaitTreeWaitObject::GetChildren());
-
-    const auto& mutex = static_cast<const Kernel::Mutex&>(object);
-    if (mutex.GetHasWaiters()) {
-        list.push_back(std::make_unique<WaitTreeText>(tr("locked by thread:")));
-        list.push_back(std::make_unique<WaitTreeThread>(*mutex.GetHoldingThread()));
-    } else {
-        list.push_back(std::make_unique<WaitTreeText>(tr("free")));
-    }
-    return list;
-}
-
-WaitTreeConditionVariable::WaitTreeConditionVariable(const Kernel::ConditionVariable& object)
-    : WaitTreeWaitObject(object) {}
-
-std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeConditionVariable::GetChildren() const {
-    std::vector<std::unique_ptr<WaitTreeItem>> list(WaitTreeWaitObject::GetChildren());
-
-    const auto& condition_variable = static_cast<const Kernel::ConditionVariable&>(object);
-    list.push_back(std::make_unique<WaitTreeText>(
-        tr("available count = %1").arg(condition_variable.GetAvailableCount())));
-    return list;
-}
-
 WaitTreeTimer::WaitTreeTimer(const Kernel::Timer& object) : WaitTreeWaitObject(object) {}
 
 std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeTimer::GetChildren() const {
@@ -293,21 +288,6 @@ std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeTimer::GetChildren() const {
     return list;
 }
 
-WaitTreeMutexList::WaitTreeMutexList(
-    const boost::container::flat_set<Kernel::SharedPtr<Kernel::Mutex>>& list)
-    : mutex_list(list) {}
-
-QString WaitTreeMutexList::GetText() const {
-    return tr("holding mutexes");
-}
-
-std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeMutexList::GetChildren() const {
-    std::vector<std::unique_ptr<WaitTreeItem>> list(mutex_list.size());
-    std::transform(mutex_list.begin(), mutex_list.end(), list.begin(),
-                   [](const auto& t) { return std::make_unique<WaitTreeMutex>(*t); });
-    return list;
-}
-
 WaitTreeThreadList::WaitTreeThreadList(const std::vector<Kernel::SharedPtr<Kernel::Thread>>& list)
     : thread_list(list) {}
 

src/yuzu/debugger/wait_tree.h

@@ -16,8 +16,6 @@ class EmuThread;
 namespace Kernel {
 class WaitObject;
 class Event;
-class Mutex;
-class ConditionVariable;
 class Thread;
 class Timer;
 } // namespace Kernel
@@ -61,6 +59,20 @@ public:
     bool IsExpandable() const override;
 };
 
+class WaitTreeMutexInfo : public WaitTreeExpandableItem {
+    Q_OBJECT
+public:
+    explicit WaitTreeMutexInfo(VAddr mutex_address);
+    QString GetText() const override;
+    std::vector<std::unique_ptr<WaitTreeItem>> GetChildren() const override;
+
+private:
+    VAddr mutex_address;
+    u32 mutex_value;
+    Kernel::Handle owner_handle;
+    Kernel::SharedPtr<Kernel::Thread> owner;
+};
+
 class WaitTreeWaitObject : public WaitTreeExpandableItem {
     Q_OBJECT
 public:
@@ -104,20 +116,6 @@ public:
     std::vector<std::unique_ptr<WaitTreeItem>> GetChildren() const override;
 };
 
-class WaitTreeMutex : public WaitTreeWaitObject {
-    Q_OBJECT
-public:
-    explicit WaitTreeMutex(const Kernel::Mutex& object);
-    std::vector<std::unique_ptr<WaitTreeItem>> GetChildren() const override;
-};
-
-class WaitTreeConditionVariable : public WaitTreeWaitObject {
-    Q_OBJECT
-public:
-    explicit WaitTreeConditionVariable(const Kernel::ConditionVariable& object);
-    std::vector<std::unique_ptr<WaitTreeItem>> GetChildren() const override;
-};
-
 class WaitTreeTimer : public WaitTreeWaitObject {
     Q_OBJECT
 public:
@@ -125,19 +123,6 @@ public:
     std::vector<std::unique_ptr<WaitTreeItem>> GetChildren() const override;
 };
 
-class WaitTreeMutexList : public WaitTreeExpandableItem {
-    Q_OBJECT
-public:
-    explicit WaitTreeMutexList(
-        const boost::container::flat_set<Kernel::SharedPtr<Kernel::Mutex>>& list);
-
-    QString GetText() const override;
-    std::vector<std::unique_ptr<WaitTreeItem>> GetChildren() const override;
-
-private:
-    const boost::container::flat_set<Kernel::SharedPtr<Kernel::Mutex>>& mutex_list;
-};
-
 class WaitTreeThreadList : public WaitTreeExpandableItem {
     Q_OBJECT
 public:

src/yuzu_cmd/emu_window/emu_window_sdl2.cpp

@@ -56,28 +56,7 @@ void EmuWindow_SDL2::OnResize() {
     UpdateCurrentFramebufferLayout(width, height);
 }
 
-void EmuWindow_SDL2::Fullscreen() {
-    if (SDL_SetWindowFullscreen(render_window, SDL_WINDOW_FULLSCREEN) == 0) {
-        return;
-    }
-
-    NGLOG_ERROR(Frontend, "Fullscreening failed: {}", SDL_GetError());
-
-    // Try a different fullscreening method
-    NGLOG_INFO(Frontend, "Attempting to use borderless fullscreen...");
-    if (SDL_SetWindowFullscreen(render_window, SDL_WINDOW_FULLSCREEN_DESKTOP) == 0) {
-        return;
-    }
-
-    NGLOG_ERROR(Frontend, "Borderless fullscreening failed: {}", SDL_GetError());
-
-    // Fallback algorithm: Maximise window.
-    // Works on all systems (unless something is seriously wrong), so no fallback for this one.
-    NGLOG_INFO(Frontend, "Falling back on a maximised window...");
-    SDL_MaximizeWindow(render_window);
-}
-
-EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
+EmuWindow_SDL2::EmuWindow_SDL2() {
     InputCommon::Init();
 
     SDL_SetMainReady();
@@ -111,10 +90,6 @@ EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
         exit(1);
     }
 
-    if (fullscreen) {
-        Fullscreen();
-    }
-
     gl_context = SDL_GL_CreateContext(render_window);
 
     if (gl_context == nullptr) {

src/yuzu_cmd/emu_window/emu_window_sdl2.h

@@ -12,7 +12,7 @@ struct SDL_Window;
 
 class EmuWindow_SDL2 : public EmuWindow {
 public:
-    explicit EmuWindow_SDL2(bool fullscreen);
+    EmuWindow_SDL2();
     ~EmuWindow_SDL2();
 
     /// Swap buffers to display the next frame
@@ -43,9 +43,6 @@ private:
     /// Called by PollEvents when any event that may cause the window to be resized occurs
     void OnResize();
 
-    /// Called when user passes the fullscreen parameter flag
-    void Fullscreen();
-
     /// Called when a configuration change affects the minimal size of the window
     void OnMinimalClientAreaChangeRequest(
         const std::pair<unsigned, unsigned>& minimal_size) override;

src/yuzu_cmd/yuzu.cpp

@@ -50,7 +50,6 @@ static void PrintHelp(const char* argv0) {
     std::cout << "Usage: " << argv0
               << " [options] <filename>\n"
                  "-g, --gdbport=NUMBER  Enable gdb stub on port NUMBER\n"
-                 "-f, --fullscreen     Start in fullscreen mode\n"
                  "-h, --help            Display this help and exit\n"
                  "-v, --version         Output version information and exit\n";
 }
@@ -77,18 +76,15 @@ int main(int argc, char** argv) {
 #endif
     std::string filepath;
 
-    bool fullscreen = false;
-
     static struct option long_options[] = {
         {"gdbport", required_argument, 0, 'g'},
-        {"fullscreen", no_argument, 0, 'f'},
         {"help", no_argument, 0, 'h'},
         {"version", no_argument, 0, 'v'},
         {0, 0, 0, 0},
     };
 
     while (optind < argc) {
-        char arg = getopt_long(argc, argv, "g:fhv", long_options, &option_index);
+        char arg = getopt_long(argc, argv, "g:hv", long_options, &option_index);
         if (arg != -1) {
             switch (arg) {
             case 'g':
@@ -102,10 +98,6 @@ int main(int argc, char** argv) {
                     exit(1);
                 }
                 break;
-            case 'f':
-                fullscreen = true;
-                NGLOG_INFO(Frontend, "Starting in fullscreen mode...");
-                break;
             case 'h':
                 PrintHelp(argv[0]);
                 return 0;
@@ -145,7 +137,7 @@ int main(int argc, char** argv) {
     Settings::values.use_gdbstub = use_gdbstub;
     Settings::Apply();
 
-    std::unique_ptr<EmuWindow_SDL2> emu_window{std::make_unique<EmuWindow_SDL2>(fullscreen)};
+    std::unique_ptr<EmuWindow_SDL2> emu_window{std::make_unique<EmuWindow_SDL2>()};
 
     Core::System& system{Core::System::GetInstance()};