New upstream version 0-1284+ds

CMakeLists.txt

@@ -208,7 +208,7 @@ find_package(libusb 1.0.24)
 find_package(lz4 REQUIRED)
 find_package(nlohmann_json 3.8 REQUIRED)
 find_package(Opus 1.3)
-find_package(Vulkan 1.3.213)
+find_package(Vulkan 1.3.238)
 find_package(ZLIB 1.2 REQUIRED)
 find_package(zstd 1.5 REQUIRED)
 

src/common/CMakeLists.txt

@@ -78,6 +78,7 @@ add_library(common STATIC
     logging/types.h
     lz4_compression.cpp
     lz4_compression.h
+    make_unique_for_overwrite.h
     math_util.h
     memory_detect.cpp
     memory_detect.h
@@ -101,6 +102,7 @@ add_library(common STATIC
     ${CMAKE_CURRENT_BINARY_DIR}/scm_rev.cpp
     scm_rev.h
     scope_exit.h
+    scratch_buffer.h
     settings.cpp
     settings.h
     settings_input.cpp

src/common/make_unique_for_overwrite.h

@@ -0,0 +1,25 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <memory>
+#include <type_traits>
+
+namespace Common {
+
+template <class T>
+requires(!std::is_array_v<T>) std::unique_ptr<T> make_unique_for_overwrite() {
+    return std::unique_ptr<T>(new T);
+}
+
+template <class T>
+requires std::is_unbounded_array_v<T> std::unique_ptr<T> make_unique_for_overwrite(std::size_t n) {
+    return std::unique_ptr<T>(new std::remove_extent_t<T>[n]);
+}
+
+template <class T, class... Args>
+requires std::is_bounded_array_v<T>
+void make_unique_for_overwrite(Args&&...) = delete;
+
+} // namespace Common

src/common/scratch_buffer.h

@@ -0,0 +1,95 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/make_unique_for_overwrite.h"
+
+namespace Common {
+
+/**
+ * ScratchBuffer class
+ * This class creates a default initialized heap allocated buffer for cases such as intermediate
+ * buffers being copied into entirely, where value initializing members during allocation or resize
+ * is redundant.
+ */
+template <typename T>
+class ScratchBuffer {
+public:
+    ScratchBuffer() = default;
+
+    explicit ScratchBuffer(size_t initial_capacity)
+        : last_requested_size{initial_capacity}, buffer_capacity{initial_capacity},
+          buffer{Common::make_unique_for_overwrite<T[]>(initial_capacity)} {}
+
+    ~ScratchBuffer() = default;
+
+    /// This will only grow the buffer's capacity if size is greater than the current capacity.
+    /// The previously held data will remain intact.
+    void resize(size_t size) {
+        if (size > buffer_capacity) {
+            auto new_buffer = Common::make_unique_for_overwrite<T[]>(size);
+            std::move(buffer.get(), buffer.get() + buffer_capacity, new_buffer.get());
+            buffer = std::move(new_buffer);
+            buffer_capacity = size;
+        }
+        last_requested_size = size;
+    }
+
+    /// This will only grow the buffer's capacity if size is greater than the current capacity.
+    /// The previously held data will be destroyed if a reallocation occurs.
+    void resize_destructive(size_t size) {
+        if (size > buffer_capacity) {
+            buffer_capacity = size;
+            buffer = Common::make_unique_for_overwrite<T[]>(buffer_capacity);
+        }
+        last_requested_size = size;
+    }
+
+    [[nodiscard]] T* data() noexcept {
+        return buffer.get();
+    }
+
+    [[nodiscard]] const T* data() const noexcept {
+        return buffer.get();
+    }
+
+    [[nodiscard]] T* begin() noexcept {
+        return data();
+    }
+
+    [[nodiscard]] const T* begin() const noexcept {
+        return data();
+    }
+
+    [[nodiscard]] T* end() noexcept {
+        return data() + last_requested_size;
+    }
+
+    [[nodiscard]] const T* end() const noexcept {
+        return data() + last_requested_size;
+    }
+
+    [[nodiscard]] T& operator[](size_t i) {
+        return buffer[i];
+    }
+
+    [[nodiscard]] const T& operator[](size_t i) const {
+        return buffer[i];
+    }
+
+    [[nodiscard]] size_t size() const noexcept {
+        return last_requested_size;
+    }
+
+    [[nodiscard]] size_t capacity() const noexcept {
+        return buffer_capacity;
+    }
+
+private:
+    size_t last_requested_size{};
+    size_t buffer_capacity{};
+    std::unique_ptr<T[]> buffer{};
+};
+
+} // namespace Common

src/core/CMakeLists.txt

@@ -201,6 +201,9 @@ add_library(core STATIC
     hle/kernel/k_event_info.h
     hle/kernel/k_handle_table.cpp
     hle/kernel/k_handle_table.h
+    hle/kernel/k_hardware_timer_base.h
+    hle/kernel/k_hardware_timer.cpp
+    hle/kernel/k_hardware_timer.h
     hle/kernel/k_interrupt_manager.cpp
     hle/kernel/k_interrupt_manager.h
     hle/kernel/k_light_condition_variable.cpp
@@ -268,6 +271,7 @@ add_library(core STATIC
     hle/kernel/k_thread_local_page.h
     hle/kernel/k_thread_queue.cpp
     hle/kernel/k_thread_queue.h
+    hle/kernel/k_timer_task.h
     hle/kernel/k_trace.h
     hle/kernel/k_transfer_memory.cpp
     hle/kernel/k_transfer_memory.h
@@ -290,8 +294,6 @@ add_library(core STATIC
     hle/kernel/svc_common.h
     hle/kernel/svc_types.h
     hle/kernel/svc_wrap.h
-    hle/kernel/time_manager.cpp
-    hle/kernel/time_manager.h
     hle/result.h
     hle/service/acc/acc.cpp
     hle/service/acc/acc.h

src/core/core.cpp

@@ -183,26 +183,20 @@ struct System::Impl {
         Initialize(system);
     }
 
-    SystemResultStatus Run() {
+    void Run() {
         std::unique_lock<std::mutex> lk(suspend_guard);
-        status = SystemResultStatus::Success;
 
         kernel.Suspend(false);
         core_timing.SyncPause(false);
         is_paused.store(false, std::memory_order_relaxed);
-
-        return status;
     }
 
-    SystemResultStatus Pause() {
+    void Pause() {
         std::unique_lock<std::mutex> lk(suspend_guard);
-        status = SystemResultStatus::Success;
 
         core_timing.SyncPause(true);
         kernel.Suspend(true);
         is_paused.store(true, std::memory_order_relaxed);
-
-        return status;
     }
 
     bool IsPaused() const {
@@ -553,12 +547,12 @@ void System::Initialize() {
     impl->Initialize(*this);
 }
 
-SystemResultStatus System::Run() {
-    return impl->Run();
+void System::Run() {
+    impl->Run();
 }
 
-SystemResultStatus System::Pause() {
-    return impl->Pause();
+void System::Pause() {
+    impl->Pause();
 }
 
 bool System::IsPaused() const {

src/core/core.h

@@ -152,13 +152,13 @@ public:
      * Run the OS and Application
      * This function will start emulation and run the relevant devices
      */
-    [[nodiscard]] SystemResultStatus Run();
+    void Run();
 
     /**
      * Pause the OS and Application
      * This function will pause emulation and stop the relevant devices
      */
-    [[nodiscard]] SystemResultStatus Pause();
+    void Pause();
 
     /// Check if the core is currently paused.
     [[nodiscard]] bool IsPaused() const;

src/core/hid/emulated_controller.cpp

@@ -210,6 +210,13 @@ void EmulatedController::LoadTASParams() {
     tas_stick_params[Settings::NativeAnalog::LStick].Set("axis_y", 1);
     tas_stick_params[Settings::NativeAnalog::RStick].Set("axis_x", 2);
     tas_stick_params[Settings::NativeAnalog::RStick].Set("axis_y", 3);
+
+    // set to optimal stick to avoid sanitizing the stick and tweaking the coordinates
+    // making sure they play back in the game as originally written down in the script file
+    tas_stick_params[Settings::NativeAnalog::LStick].Set("deadzone", 0.0f);
+    tas_stick_params[Settings::NativeAnalog::LStick].Set("range", 1.0f);
+    tas_stick_params[Settings::NativeAnalog::RStick].Set("deadzone", 0.0f);
+    tas_stick_params[Settings::NativeAnalog::RStick].Set("range", 1.0f);
 }
 
 void EmulatedController::LoadVirtualGamepadParams() {

src/core/hle/kernel/k_address_arbiter.cpp

@@ -10,7 +10,6 @@
 #include "core/hle/kernel/k_thread_queue.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/svc_results.h"
-#include "core/hle/kernel/time_manager.h"
 #include "core/memory.h"
 
 namespace Kernel {

src/core/hle/kernel/k_hardware_timer.cpp

@@ -0,0 +1,74 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/hle/kernel/k_hardware_timer.h"
+#include "core/hle/kernel/k_scheduler.h"
+
+namespace Kernel {
+
+void KHardwareTimer::Initialize() {
+    // Create the timing callback to register with CoreTiming.
+    m_event_type = Core::Timing::CreateEvent(
+        "KHardwareTimer::Callback", [](std::uintptr_t timer_handle, s64, std::chrono::nanoseconds) {
+            reinterpret_cast<KHardwareTimer*>(timer_handle)->DoTask();
+            return std::nullopt;
+        });
+}
+
+void KHardwareTimer::Finalize() {
+    this->DisableInterrupt();
+    m_event_type.reset();
+}
+
+void KHardwareTimer::DoTask() {
+    // Handle the interrupt.
+    {
+        KScopedSchedulerLock slk{m_kernel};
+        KScopedSpinLock lk(this->GetLock());
+
+        //! Ignore this event if needed.
+        if (!this->GetInterruptEnabled()) {
+            return;
+        }
+
+        // Disable the timer interrupt while we handle this.
+        this->DisableInterrupt();
+
+        if (const s64 next_time = this->DoInterruptTaskImpl(GetTick());
+            0 < next_time && next_time <= m_wakeup_time) {
+            // We have a next time, so we should set the time to interrupt and turn the interrupt
+            // on.
+            this->EnableInterrupt(next_time);
+        }
+    }
+
+    // Clear the timer interrupt.
+    // Kernel::GetInterruptManager().ClearInterrupt(KInterruptName_NonSecurePhysicalTimer,
+    //                                              GetCurrentCoreId());
+}
+
+void KHardwareTimer::EnableInterrupt(s64 wakeup_time) {
+    this->DisableInterrupt();
+
+    m_wakeup_time = wakeup_time;
+    m_kernel.System().CoreTiming().ScheduleEvent(std::chrono::nanoseconds{m_wakeup_time},
+                                                 m_event_type, reinterpret_cast<uintptr_t>(this),
+                                                 true);
+}
+
+void KHardwareTimer::DisableInterrupt() {
+    m_kernel.System().CoreTiming().UnscheduleEvent(m_event_type, reinterpret_cast<uintptr_t>(this));
+    m_wakeup_time = std::numeric_limits<s64>::max();
+}
+
+s64 KHardwareTimer::GetTick() const {
+    return m_kernel.System().CoreTiming().GetGlobalTimeNs().count();
+}
+
+bool KHardwareTimer::GetInterruptEnabled() {
+    return m_wakeup_time != std::numeric_limits<s64>::max();
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_hardware_timer.h

@@ -0,0 +1,54 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "core/hle/kernel/k_hardware_timer_base.h"
+
+namespace Core::Timing {
+struct EventType;
+} // namespace Core::Timing
+
+namespace Kernel {
+
+class KHardwareTimer : /* public KInterruptTask, */ public KHardwareTimerBase {
+public:
+    explicit KHardwareTimer(KernelCore& kernel) : KHardwareTimerBase{kernel} {}
+
+    // Public API.
+    void Initialize();
+    void Finalize();
+
+    s64 GetCount() const {
+        return GetTick();
+    }
+
+    void RegisterTask(KTimerTask* task, s64 time_from_now) {
+        this->RegisterAbsoluteTask(task, GetTick() + time_from_now);
+    }
+
+    void RegisterAbsoluteTask(KTimerTask* task, s64 task_time) {
+        KScopedDisableDispatch dd{m_kernel};
+        KScopedSpinLock lk{this->GetLock()};
+
+        if (this->RegisterAbsoluteTaskImpl(task, task_time)) {
+            if (task_time <= m_wakeup_time) {
+                this->EnableInterrupt(task_time);
+            }
+        }
+    }
+
+private:
+    void EnableInterrupt(s64 wakeup_time);
+    void DisableInterrupt();
+    bool GetInterruptEnabled();
+    s64 GetTick() const;
+    void DoTask();
+
+private:
+    // Absolute time in nanoseconds
+    s64 m_wakeup_time{std::numeric_limits<s64>::max()};
+    std::shared_ptr<Core::Timing::EventType> m_event_type{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_hardware_timer_base.h

@@ -0,0 +1,92 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "core/hle/kernel/k_spin_lock.h"
+#include "core/hle/kernel/k_thread.h"
+#include "core/hle/kernel/k_timer_task.h"
+
+namespace Kernel {
+
+class KHardwareTimerBase {
+public:
+    explicit KHardwareTimerBase(KernelCore& kernel) : m_kernel{kernel} {}
+
+    void CancelTask(KTimerTask* task) {
+        KScopedDisableDispatch dd{m_kernel};
+        KScopedSpinLock lk{m_lock};
+
+        if (const s64 task_time = task->GetTime(); task_time > 0) {
+            this->RemoveTaskFromTree(task);
+        }
+    }
+
+protected:
+    KSpinLock& GetLock() {
+        return m_lock;
+    }
+
+    s64 DoInterruptTaskImpl(s64 cur_time) {
+        // We want to handle all tasks, returning the next time that a task is scheduled.
+        while (true) {
+            // Get the next task. If there isn't one, return 0.
+            KTimerTask* task = m_next_task;
+            if (task == nullptr) {
+                return 0;
+            }
+
+            // If the task needs to be done in the future, do it in the future and not now.
+            if (const s64 task_time = task->GetTime(); task_time > cur_time) {
+                return task_time;
+            }
+
+            // Remove the task from the tree of tasks, and update our next task.
+            this->RemoveTaskFromTree(task);
+
+            // Handle the task.
+            task->OnTimer();
+        }
+    }
+
+    bool RegisterAbsoluteTaskImpl(KTimerTask* task, s64 task_time) {
+        ASSERT(task_time > 0);
+
+        // Set the task's time, and insert it into our tree.
+        task->SetTime(task_time);
+        m_task_tree.insert(*task);
+
+        // Update our next task if relevant.
+        if (m_next_task != nullptr && m_next_task->GetTime() <= task_time) {
+            return false;
+        }
+        m_next_task = task;
+        return true;
+    }
+
+private:
+    void RemoveTaskFromTree(KTimerTask* task) {
+        // Erase from the tree.
+        auto it = m_task_tree.erase(m_task_tree.iterator_to(*task));
+
+        // Clear the task's scheduled time.
+        task->SetTime(0);
+
+        // Update our next task if relevant.
+        if (m_next_task == task) {
+            m_next_task = (it != m_task_tree.end()) ? std::addressof(*it) : nullptr;
+        }
+    }
+
+protected:
+    KernelCore& m_kernel;
+
+private:
+    using TimerTaskTree = Common::IntrusiveRedBlackTreeBaseTraits<KTimerTask>::TreeType<KTimerTask>;
+
+    KSpinLock m_lock{};
+    TimerTaskTree m_task_tree{};
+    KTimerTask* m_next_task{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h

@@ -5,9 +5,9 @@
 
 #include "common/common_types.h"
 #include "core/hle/kernel/global_scheduler_context.h"
+#include "core/hle/kernel/k_hardware_timer.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/time_manager.h"
 
 namespace Kernel {
 
@@ -22,7 +22,7 @@ public:
     ~KScopedSchedulerLockAndSleep() {
         // Register the sleep.
         if (timeout_tick > 0) {
-            kernel.TimeManager().ScheduleTimeEvent(thread, timeout_tick);
+            kernel.HardwareTimer().RegisterTask(thread, timeout_tick);
         }
 
         // Unlock the scheduler.

src/core/hle/kernel/k_shared_memory.cpp

@@ -6,6 +6,7 @@
 #include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_scoped_resource_reservation.h"
 #include "core/hle/kernel/k_shared_memory.h"
+#include "core/hle/kernel/k_system_resource.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/svc_results.h"
 
@@ -18,19 +19,19 @@ KSharedMemory::~KSharedMemory() {
 }
 
 Result KSharedMemory::Initialize(Core::DeviceMemory& device_memory_, KProcess* owner_process_,
-                                 KPageGroup&& page_list_, Svc::MemoryPermission owner_permission_,
-                                 Svc::MemoryPermission user_permission_, PAddr physical_address_,
-                                 std::size_t size_, std::string name_) {
+                                 Svc::MemoryPermission owner_permission_,
+                                 Svc::MemoryPermission user_permission_, std::size_t size_,
+                                 std::string name_) {
     // Set members.
     owner_process = owner_process_;
     device_memory = &device_memory_;
-    page_list = std::move(page_list_);
     owner_permission = owner_permission_;
     user_permission = user_permission_;
-    physical_address = physical_address_;
-    size = size_;
+    size = Common::AlignUp(size_, PageSize);
     name = std::move(name_);
 
+    const size_t num_pages = Common::DivideUp(size, PageSize);
+
     // Get the resource limit.
     KResourceLimit* reslimit = kernel.GetSystemResourceLimit();
 
@@ -39,6 +40,17 @@ Result KSharedMemory::Initialize(Core::DeviceMemory& device_memory_, KProcess* o
                                                   size_);
     R_UNLESS(memory_reservation.Succeeded(), ResultLimitReached);
 
+    // Allocate the memory.
+
+    //! HACK: Open continuous mapping from sysmodule pool.
+    auto option = KMemoryManager::EncodeOption(KMemoryManager::Pool::Secure,
+                                               KMemoryManager::Direction::FromBack);
+    physical_address = kernel.MemoryManager().AllocateAndOpenContinuous(num_pages, 1, option);
+    R_UNLESS(physical_address != 0, ResultOutOfMemory);
+
+    //! Insert the result into our page group.
+    page_group.emplace(physical_address, num_pages);
+
     // Commit our reservation.
     memory_reservation.Commit();
 
@@ -50,12 +62,23 @@ Result KSharedMemory::Initialize(Core::DeviceMemory& device_memory_, KProcess* o
     is_initialized = true;
 
     // Clear all pages in the memory.
-    std::memset(device_memory_.GetPointer<void>(physical_address_), 0, size_);
+    for (const auto& block : page_group->Nodes()) {
+        std::memset(device_memory_.GetPointer<void>(block.GetAddress()), 0, block.GetSize());
+    }
 
     return ResultSuccess;
 }
 
 void KSharedMemory::Finalize() {
+    // Close and finalize the page group.
+    // page_group->Close();
+    // page_group->Finalize();
+
+    //! HACK: Manually close.
+    for (const auto& block : page_group->Nodes()) {
+        kernel.MemoryManager().Close(block.GetAddress(), block.GetNumPages());
+    }
+
     // Release the memory reservation.
     resource_limit->Release(LimitableResource::PhysicalMemoryMax, size);
     resource_limit->Close();
@@ -65,32 +88,28 @@ void KSharedMemory::Finalize() {
 }
 
 Result KSharedMemory::Map(KProcess& target_process, VAddr address, std::size_t map_size,
-                          Svc::MemoryPermission permissions) {
-    const u64 page_count{(map_size + PageSize - 1) / PageSize};
+                          Svc::MemoryPermission map_perm) {
+    // Validate the size.
+    R_UNLESS(size == map_size, ResultInvalidSize);
 
-    if (page_list.GetNumPages() != page_count) {
-        UNIMPLEMENTED_MSG("Page count does not match");
-    }
-
-    const Svc::MemoryPermission expected =
+    // Validate the permission.
+    const Svc::MemoryPermission test_perm =
         &target_process == owner_process ? owner_permission : user_permission;
-
-    if (permissions != expected) {
-        UNIMPLEMENTED_MSG("Permission does not match");
+    if (test_perm == Svc::MemoryPermission::DontCare) {
+        ASSERT(map_perm == Svc::MemoryPermission::Read || map_perm == Svc::MemoryPermission::Write);
+    } else {
+        R_UNLESS(map_perm == test_perm, ResultInvalidNewMemoryPermission);
     }
 
-    return target_process.PageTable().MapPages(address, page_list, KMemoryState::Shared,
-                                               ConvertToKMemoryPermission(permissions));
+    return target_process.PageTable().MapPages(address, *page_group, KMemoryState::Shared,
+                                               ConvertToKMemoryPermission(map_perm));
 }
 
 Result KSharedMemory::Unmap(KProcess& target_process, VAddr address, std::size_t unmap_size) {
-    const u64 page_count{(unmap_size + PageSize - 1) / PageSize};
+    // Validate the size.
+    R_UNLESS(size == unmap_size, ResultInvalidSize);
 
-    if (page_list.GetNumPages() != page_count) {
-        UNIMPLEMENTED_MSG("Page count does not match");
-    }
-
-    return target_process.PageTable().UnmapPages(address, page_list, KMemoryState::Shared);
+    return target_process.PageTable().UnmapPages(address, *page_group, KMemoryState::Shared);
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_shared_memory.h

@@ -3,6 +3,7 @@
 
 #pragma once
 
+#include <optional>
 #include <string>
 
 #include "common/common_types.h"
@@ -26,9 +27,8 @@ public:
     ~KSharedMemory() override;
 
     Result Initialize(Core::DeviceMemory& device_memory_, KProcess* owner_process_,
-                      KPageGroup&& page_list_, Svc::MemoryPermission owner_permission_,
-                      Svc::MemoryPermission user_permission_, PAddr physical_address_,
-                      std::size_t size_, std::string name_);
+                      Svc::MemoryPermission owner_permission_,
+                      Svc::MemoryPermission user_permission_, std::size_t size_, std::string name_);
 
     /**
      * Maps a shared memory block to an address in the target process' address space
@@ -76,7 +76,7 @@ public:
 private:
     Core::DeviceMemory* device_memory{};
     KProcess* owner_process{};
-    KPageGroup page_list;
+    std::optional<KPageGroup> page_group{};
     Svc::MemoryPermission owner_permission{};
     Svc::MemoryPermission user_permission{};
     PAddr physical_address{};

src/core/hle/kernel/k_thread.h

@@ -22,6 +22,7 @@
 #include "core/hle/kernel/k_light_lock.h"
 #include "core/hle/kernel/k_spin_lock.h"
 #include "core/hle/kernel/k_synchronization_object.h"
+#include "core/hle/kernel/k_timer_task.h"
 #include "core/hle/kernel/k_worker_task.h"
 #include "core/hle/kernel/slab_helpers.h"
 #include "core/hle/kernel/svc_common.h"
@@ -112,7 +113,8 @@ void SetCurrentThread(KernelCore& kernel, KThread* thread);
 [[nodiscard]] s32 GetCurrentCoreId(KernelCore& kernel);
 
 class KThread final : public KAutoObjectWithSlabHeapAndContainer<KThread, KWorkerTask>,
-                      public boost::intrusive::list_base_hook<> {
+                      public boost::intrusive::list_base_hook<>,
+                      public KTimerTask {
     KERNEL_AUTOOBJECT_TRAITS(KThread, KSynchronizationObject);
 
 private:
@@ -840,4 +842,8 @@ private:
     KernelCore& kernel;
 };
 
+inline void KTimerTask::OnTimer() {
+    static_cast<KThread*>(this)->OnTimer();
+}
+
 } // namespace Kernel

src/core/hle/kernel/k_thread_queue.cpp

@@ -1,9 +1,9 @@
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "core/hle/kernel/k_hardware_timer.h"
 #include "core/hle/kernel/k_thread_queue.h"
 #include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/time_manager.h"
 
 namespace Kernel {
 
@@ -22,7 +22,7 @@ void KThreadQueue::EndWait(KThread* waiting_thread, Result wait_result) {
     waiting_thread->ClearWaitQueue();
 
     // Cancel the thread task.
-    kernel.TimeManager().UnscheduleTimeEvent(waiting_thread);
+    kernel.HardwareTimer().CancelTask(waiting_thread);
 }
 
 void KThreadQueue::CancelWait(KThread* waiting_thread, Result wait_result, bool cancel_timer_task) {
@@ -37,7 +37,7 @@ void KThreadQueue::CancelWait(KThread* waiting_thread, Result wait_result, bool
 
     // Cancel the thread task.
     if (cancel_timer_task) {
-        kernel.TimeManager().UnscheduleTimeEvent(waiting_thread);
+        kernel.HardwareTimer().CancelTask(waiting_thread);
     }
 }
 

src/core/hle/kernel/k_timer_task.h

@@ -0,0 +1,40 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/intrusive_red_black_tree.h"
+
+namespace Kernel {
+
+class KTimerTask : public Common::IntrusiveRedBlackTreeBaseNode<KTimerTask> {
+public:
+    static constexpr int Compare(const KTimerTask& lhs, const KTimerTask& rhs) {
+        if (lhs.GetTime() < rhs.GetTime()) {
+            return -1;
+        } else {
+            return 1;
+        }
+    }
+
+    constexpr explicit KTimerTask() = default;
+
+    constexpr void SetTime(s64 t) {
+        m_time = t;
+    }
+
+    constexpr s64 GetTime() const {
+        return m_time;
+    }
+
+    // NOTE: This is virtual in Nintendo's kernel. Prior to 13.0.0, KWaitObject was also a
+    // TimerTask; this is no longer the case. Since this is now KThread exclusive, we have
+    // devirtualized (see inline declaration for this inside k_thread.h).
+    void OnTimer();
+
+private:
+    // Absolute time in nanoseconds
+    s64 m_time{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/kernel.cpp

@@ -26,6 +26,7 @@
 #include "core/hle/kernel/k_client_port.h"
 #include "core/hle/kernel/k_dynamic_resource_manager.h"
 #include "core/hle/kernel/k_handle_table.h"
+#include "core/hle/kernel/k_hardware_timer.h"
 #include "core/hle/kernel/k_memory_layout.h"
 #include "core/hle/kernel/k_memory_manager.h"
 #include "core/hle/kernel/k_page_buffer.h"
@@ -39,7 +40,6 @@
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/physical_core.h"
 #include "core/hle/kernel/service_thread.h"
-#include "core/hle/kernel/time_manager.h"
 #include "core/hle/result.h"
 #include "core/hle/service/sm/sm.h"
 #include "core/memory.h"
@@ -55,7 +55,7 @@ struct KernelCore::Impl {
     static constexpr size_t ReservedDynamicPageCount = 64;
 
     explicit Impl(Core::System& system_, KernelCore& kernel_)
-        : time_manager{system_}, service_threads_manager{1, "ServiceThreadsManager"},
+        : service_threads_manager{1, "ServiceThreadsManager"},
           service_thread_barrier{2}, system{system_} {}
 
     void SetMulticore(bool is_multi) {
@@ -63,6 +63,9 @@ struct KernelCore::Impl {
     }
 
     void Initialize(KernelCore& kernel) {
+        hardware_timer = std::make_unique<Kernel::KHardwareTimer>(kernel);
+        hardware_timer->Initialize();
+
         global_object_list_container = std::make_unique<KAutoObjectWithListContainer>(kernel);
         global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
         global_handle_table = std::make_unique<Kernel::KHandleTable>(kernel);
@@ -91,6 +94,7 @@ struct KernelCore::Impl {
                                        pt_heap_region.GetSize());
         }
 
+        InitializeHackSharedMemory();
         RegisterHostThread(nullptr);
 
         default_service_thread = &CreateServiceThread(kernel, "DefaultServiceThread");
@@ -193,6 +197,9 @@ struct KernelCore::Impl {
         // Ensure that the object list container is finalized and properly shutdown.
         global_object_list_container->Finalize();
         global_object_list_container.reset();
+
+        hardware_timer->Finalize();
+        hardware_timer.reset();
     }
 
     void CloseServices() {
@@ -720,14 +727,14 @@ struct KernelCore::Impl {
     }
 
     void InitializeMemoryLayout() {
-        const auto system_pool = memory_layout->GetKernelSystemPoolRegionPhysicalExtents();
-
         // Initialize the memory manager.
         memory_manager = std::make_unique<KMemoryManager>(system);
         const auto& management_region = memory_layout->GetPoolManagementRegion();
         ASSERT(management_region.GetEndAddress() != 0);
         memory_manager->Initialize(management_region.GetAddress(), management_region.GetSize());
+    }
 
+    void InitializeHackSharedMemory() {
         // Setup memory regions for emulated processes
         // TODO(bunnei): These should not be hardcoded regions initialized within the kernel
         constexpr std::size_t hid_size{0x40000};
@@ -736,39 +743,23 @@ struct KernelCore::Impl {
         constexpr std::size_t time_size{0x1000};
         constexpr std::size_t hidbus_size{0x1000};
 
-        const PAddr hid_phys_addr{system_pool.GetAddress()};
-        const PAddr font_phys_addr{system_pool.GetAddress() + hid_size};
-        const PAddr irs_phys_addr{system_pool.GetAddress() + hid_size + font_size};
-        const PAddr time_phys_addr{system_pool.GetAddress() + hid_size + font_size + irs_size};
-        const PAddr hidbus_phys_addr{system_pool.GetAddress() + hid_size + font_size + irs_size +
-                                     time_size};
-
         hid_shared_mem = KSharedMemory::Create(system.Kernel());
         font_shared_mem = KSharedMemory::Create(system.Kernel());
         irs_shared_mem = KSharedMemory::Create(system.Kernel());
         time_shared_mem = KSharedMemory::Create(system.Kernel());
         hidbus_shared_mem = KSharedMemory::Create(system.Kernel());
 
-        hid_shared_mem->Initialize(system.DeviceMemory(), nullptr,
-                                   {hid_phys_addr, hid_size / PageSize},
-                                   Svc::MemoryPermission::None, Svc::MemoryPermission::Read,
-                                   hid_phys_addr, hid_size, "HID:SharedMemory");
-        font_shared_mem->Initialize(system.DeviceMemory(), nullptr,
-                                    {font_phys_addr, font_size / PageSize},
-                                    Svc::MemoryPermission::None, Svc::MemoryPermission::Read,
-                                    font_phys_addr, font_size, "Font:SharedMemory");
-        irs_shared_mem->Initialize(system.DeviceMemory(), nullptr,
-                                   {irs_phys_addr, irs_size / PageSize},
-                                   Svc::MemoryPermission::None, Svc::MemoryPermission::Read,
-                                   irs_phys_addr, irs_size, "IRS:SharedMemory");
-        time_shared_mem->Initialize(system.DeviceMemory(), nullptr,
-                                    {time_phys_addr, time_size / PageSize},
-                                    Svc::MemoryPermission::None, Svc::MemoryPermission::Read,
-                                    time_phys_addr, time_size, "Time:SharedMemory");
-        hidbus_shared_mem->Initialize(system.DeviceMemory(), nullptr,
-                                      {hidbus_phys_addr, hidbus_size / PageSize},
-                                      Svc::MemoryPermission::None, Svc::MemoryPermission::Read,
-                                      hidbus_phys_addr, hidbus_size, "HidBus:SharedMemory");
+        hid_shared_mem->Initialize(system.DeviceMemory(), nullptr, Svc::MemoryPermission::None,
+                                   Svc::MemoryPermission::Read, hid_size, "HID:SharedMemory");
+        font_shared_mem->Initialize(system.DeviceMemory(), nullptr, Svc::MemoryPermission::None,
+                                    Svc::MemoryPermission::Read, font_size, "Font:SharedMemory");
+        irs_shared_mem->Initialize(system.DeviceMemory(), nullptr, Svc::MemoryPermission::None,
+                                   Svc::MemoryPermission::Read, irs_size, "IRS:SharedMemory");
+        time_shared_mem->Initialize(system.DeviceMemory(), nullptr, Svc::MemoryPermission::None,
+                                    Svc::MemoryPermission::Read, time_size, "Time:SharedMemory");
+        hidbus_shared_mem->Initialize(system.DeviceMemory(), nullptr, Svc::MemoryPermission::None,
+                                      Svc::MemoryPermission::Read, hidbus_size,
+                                      "HidBus:SharedMemory");
     }
 
     KClientPort* CreateNamedServicePort(std::string name) {
@@ -832,7 +823,7 @@ struct KernelCore::Impl {
     std::vector<KProcess*> process_list;
     std::atomic<KProcess*> current_process{};
     std::unique_ptr<Kernel::GlobalSchedulerContext> global_scheduler_context;
-    Kernel::TimeManager time_manager;
+    std::unique_ptr<Kernel::KHardwareTimer> hardware_timer;
 
     Init::KSlabResourceCounts slab_resource_counts{};
     KResourceLimit* system_resource_limit{};
@@ -1019,12 +1010,8 @@ Kernel::KScheduler* KernelCore::CurrentScheduler() {
     return impl->schedulers[core_id].get();
 }
 
-Kernel::TimeManager& KernelCore::TimeManager() {
-    return impl->time_manager;
-}
-
-const Kernel::TimeManager& KernelCore::TimeManager() const {
-    return impl->time_manager;
+Kernel::KHardwareTimer& KernelCore::HardwareTimer() {
+    return *impl->hardware_timer;
 }
 
 Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() {

src/core/hle/kernel/kernel.h

@@ -39,6 +39,7 @@ class KDynamicPageManager;
 class KEvent;
 class KEventInfo;
 class KHandleTable;
+class KHardwareTimer;
 class KLinkedListNode;
 class KMemoryLayout;
 class KMemoryManager;
@@ -63,7 +64,6 @@ class KCodeMemory;
 class PhysicalCore;
 class ServiceThread;
 class Synchronization;
-class TimeManager;
 
 using ServiceInterfaceFactory =
     std::function<KClientPort&(Service::SM::ServiceManager&, Core::System&)>;
@@ -175,11 +175,8 @@ public:
     /// Gets the an instance of the current physical CPU core.
     const Kernel::PhysicalCore& CurrentPhysicalCore() const;
 
-    /// Gets the an instance of the TimeManager Interface.
-    Kernel::TimeManager& TimeManager();
-
-    /// Gets the an instance of the TimeManager Interface.
-    const Kernel::TimeManager& TimeManager() const;
+    /// Gets the an instance of the hardware timer.
+    Kernel::KHardwareTimer& HardwareTimer();
 
     /// Stops execution of 'id' core, in order to reschedule a new thread.
     void PrepareReschedule(std::size_t id);

src/core/hle/kernel/time_manager.cpp

@@ -1,44 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#include "common/assert.h"
-#include "core/core.h"
-#include "core/core_timing.h"
-#include "core/hle/kernel/k_scheduler.h"
-#include "core/hle/kernel/k_thread.h"
-#include "core/hle/kernel/time_manager.h"
-
-namespace Kernel {
-
-TimeManager::TimeManager(Core::System& system_) : system{system_} {
-    time_manager_event_type = Core::Timing::CreateEvent(
-        "Kernel::TimeManagerCallback",
-        [this](std::uintptr_t thread_handle, s64 time,
-               std::chrono::nanoseconds) -> std::optional<std::chrono::nanoseconds> {
-            KThread* thread = reinterpret_cast<KThread*>(thread_handle);
-            {
-                KScopedSchedulerLock sl(system.Kernel());
-                thread->OnTimer();
-            }
-            return std::nullopt;
-        });
-}
-
-void TimeManager::ScheduleTimeEvent(KThread* thread, s64 nanoseconds) {
-    std::scoped_lock lock{mutex};
-    if (nanoseconds > 0) {
-        ASSERT(thread);
-        ASSERT(thread->GetState() != ThreadState::Runnable);
-        system.CoreTiming().ScheduleEvent(std::chrono::nanoseconds{nanoseconds},
-                                          time_manager_event_type,
-                                          reinterpret_cast<uintptr_t>(thread));
-    }
-}
-
-void TimeManager::UnscheduleTimeEvent(KThread* thread) {
-    std::scoped_lock lock{mutex};
-    system.CoreTiming().UnscheduleEvent(time_manager_event_type,
-                                        reinterpret_cast<uintptr_t>(thread));
-}
-
-} // namespace Kernel

src/core/hle/kernel/time_manager.h

@@ -1,41 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#pragma once
-
-#include <memory>
-#include <mutex>
-
-namespace Core {
-class System;
-} // namespace Core
-
-namespace Core::Timing {
-struct EventType;
-} // namespace Core::Timing
-
-namespace Kernel {
-
-class KThread;
-
-/**
- * The `TimeManager` takes care of scheduling time events on threads and executes their TimeUp
- * method when the event is triggered.
- */
-class TimeManager {
-public:
-    explicit TimeManager(Core::System& system);
-
-    /// Schedule a time event on `timetask` thread that will expire in 'nanoseconds'
-    void ScheduleTimeEvent(KThread* time_task, s64 nanoseconds);
-
-    /// Unschedule an existing time event
-    void UnscheduleTimeEvent(KThread* thread);
-
-private:
-    Core::System& system;
-    std::shared_ptr<Core::Timing::EventType> time_manager_event_type;
-    std::mutex mutex;
-};
-
-} // namespace Kernel

src/core/hle/service/time/clock_types.h

@@ -49,6 +49,7 @@ struct SteadyClockContext {
 static_assert(sizeof(SteadyClockContext) == 0x18, "SteadyClockContext is incorrect size");
 static_assert(std::is_trivially_copyable_v<SteadyClockContext>,
               "SteadyClockContext must be trivially copyable");
+using StandardSteadyClockTimePointType = SteadyClockContext;
 
 struct SystemClockContext {
     s64 offset;

src/core/hle/service/time/time_sharedmemory.cpp

@@ -26,23 +26,24 @@ void SharedMemory::SetupStandardSteadyClock(const Common::UUID& clock_source_id,
     const Clock::SteadyClockContext context{
         static_cast<u64>(current_time_point.nanoseconds - ticks_time_span.nanoseconds),
         clock_source_id};
-    shared_memory_format.standard_steady_clock_timepoint.StoreData(
-        system.Kernel().GetTimeSharedMem().GetPointer(), context);
+    StoreToLockFreeAtomicType(&GetFormat()->standard_steady_clock_timepoint, context);
 }
 
 void SharedMemory::UpdateLocalSystemClockContext(const Clock::SystemClockContext& context) {
-    shared_memory_format.standard_local_system_clock_context.StoreData(
-        system.Kernel().GetTimeSharedMem().GetPointer(), context);
+    StoreToLockFreeAtomicType(&GetFormat()->standard_local_system_clock_context, context);
 }
 
 void SharedMemory::UpdateNetworkSystemClockContext(const Clock::SystemClockContext& context) {
-    shared_memory_format.standard_network_system_clock_context.StoreData(
-        system.Kernel().GetTimeSharedMem().GetPointer(), context);
+    StoreToLockFreeAtomicType(&GetFormat()->standard_network_system_clock_context, context);
 }
 
 void SharedMemory::SetAutomaticCorrectionEnabled(bool is_enabled) {
-    shared_memory_format.standard_user_system_clock_automatic_correction.StoreData(
-        system.Kernel().GetTimeSharedMem().GetPointer(), is_enabled);
+    StoreToLockFreeAtomicType(
+        &GetFormat()->is_standard_user_system_clock_automatic_correction_enabled, is_enabled);
+}
+
+SharedMemory::Format* SharedMemory::GetFormat() {
+    return reinterpret_cast<SharedMemory::Format*>(system.Kernel().GetTimeSharedMem().GetPointer());
 }
 
 } // namespace Service::Time

src/core/hle/service/time/time_sharedmemory.h

@@ -10,45 +10,68 @@
 
 namespace Service::Time {
 
+// Note: this type is not safe for concurrent writes.
+template <typename T>
+struct LockFreeAtomicType {
+    u32 counter_;
+    std::array<T, 2> value_;
+};
+
+template <typename T>
+static inline void StoreToLockFreeAtomicType(LockFreeAtomicType<T>* p, const T& value) {
+    // Get the current counter.
+    auto counter = p->counter_;
+
+    // Increment the counter.
+    ++counter;
+
+    // Store the updated value.
+    p->value_[counter % 2] = value;
+
+    // Fence memory.
+    std::atomic_thread_fence(std::memory_order_release);
+
+    // Set the updated counter.
+    p->counter_ = counter;
+}
+
+template <typename T>
+static inline T LoadFromLockFreeAtomicType(const LockFreeAtomicType<T>* p) {
+    while (true) {
+        // Get the counter.
+        auto counter = p->counter_;
+
+        // Get the value.
+        auto value = p->value_[counter % 2];
+
+        // Fence memory.
+        std::atomic_thread_fence(std::memory_order_acquire);
+
+        // Check that the counter matches.
+        if (counter == p->counter_) {
+            return value;
+        }
+    }
+}
+
 class SharedMemory final {
 public:
     explicit SharedMemory(Core::System& system_);
     ~SharedMemory();
 
-    // TODO(ogniK): We have to properly simulate memory barriers, how are we going to do this?
-    template <typename T, std::size_t Offset>
-    struct MemoryBarrier {
-        static_assert(std::is_trivially_copyable_v<T>, "T must be trivially copyable");
-        u32_le read_attempt{};
-        std::array<T, 2> data{};
-
-        // These are not actually memory barriers at the moment as we don't have multicore and all
-        // HLE is mutexed. This will need to properly be implemented when we start updating the time
-        // points on threads. As of right now, we'll be updated both values synchronously and just
-        // incrementing the read_attempt to indicate that we waited.
-        void StoreData(u8* shared_memory, T data_to_store) {
-            std::memcpy(this, shared_memory + Offset, sizeof(*this));
-            read_attempt++;
-            data[read_attempt & 1] = data_to_store;
-            std::memcpy(shared_memory + Offset, this, sizeof(*this));
-        }
-
-        // For reading we're just going to read the last stored value. If there was no value stored
-        // it will just end up reading an empty value as intended.
-        T ReadData(u8* shared_memory) {
-            std::memcpy(this, shared_memory + Offset, sizeof(*this));
-            return data[(read_attempt - 1) & 1];
-        }
-    };
-
     // Shared memory format
     struct Format {
-        MemoryBarrier<Clock::SteadyClockContext, 0x0> standard_steady_clock_timepoint;
-        MemoryBarrier<Clock::SystemClockContext, 0x38> standard_local_system_clock_context;
-        MemoryBarrier<Clock::SystemClockContext, 0x80> standard_network_system_clock_context;
-        MemoryBarrier<bool, 0xc8> standard_user_system_clock_automatic_correction;
-        u32_le format_version;
+        LockFreeAtomicType<Clock::StandardSteadyClockTimePointType> standard_steady_clock_timepoint;
+        LockFreeAtomicType<Clock::SystemClockContext> standard_local_system_clock_context;
+        LockFreeAtomicType<Clock::SystemClockContext> standard_network_system_clock_context;
+        LockFreeAtomicType<bool> is_standard_user_system_clock_automatic_correction_enabled;
+        u32 format_version;
     };
+    static_assert(offsetof(Format, standard_steady_clock_timepoint) == 0x0);
+    static_assert(offsetof(Format, standard_local_system_clock_context) == 0x38);
+    static_assert(offsetof(Format, standard_network_system_clock_context) == 0x80);
+    static_assert(offsetof(Format, is_standard_user_system_clock_automatic_correction_enabled) ==
+                  0xc8);
     static_assert(sizeof(Format) == 0xd8, "Format is an invalid size");
 
     void SetupStandardSteadyClock(const Common::UUID& clock_source_id,
@@ -56,10 +79,10 @@ public:
     void UpdateLocalSystemClockContext(const Clock::SystemClockContext& context);
     void UpdateNetworkSystemClockContext(const Clock::SystemClockContext& context);
     void SetAutomaticCorrectionEnabled(bool is_enabled);
+    Format* GetFormat();
 
 private:
     Core::System& system;
-    Format shared_memory_format{};
 };
 
 } // namespace Service::Time

src/input_common/input_mapping.cpp

@@ -200,12 +200,6 @@ bool MappingFactory::IsDriverValid(const MappingData& data) const {
         return false;
     }
     // The following drivers don't need to be mapped
-    if (data.engine == "tas") {
-        return false;
-    }
-    if (data.engine == "touch") {
-        return false;
-    }
     if (data.engine == "touch_from_button") {
         return false;
     }

src/input_common/main.cpp

@@ -26,79 +26,33 @@
 namespace InputCommon {
 
 struct InputSubsystem::Impl {
-    void Initialize() {
-        mapping_factory = std::make_shared<MappingFactory>();
+    template <typename Engine>
+    void RegisterEngine(std::string name, std::shared_ptr<Engine>& engine) {
         MappingCallback mapping_callback{[this](const MappingData& data) { RegisterInput(data); }};
 
-        keyboard = std::make_shared<Keyboard>("keyboard");
-        keyboard->SetMappingCallback(mapping_callback);
-        keyboard_factory = std::make_shared<InputFactory>(keyboard);
-        keyboard_output_factory = std::make_shared<OutputFactory>(keyboard);
-        Common::Input::RegisterInputFactory(keyboard->GetEngineName(), keyboard_factory);
-        Common::Input::RegisterOutputFactory(keyboard->GetEngineName(), keyboard_output_factory);
+        engine = std::make_shared<Engine>(name);
+        engine->SetMappingCallback(mapping_callback);
 
-        mouse = std::make_shared<Mouse>("mouse");
-        mouse->SetMappingCallback(mapping_callback);
-        mouse_factory = std::make_shared<InputFactory>(mouse);
-        mouse_output_factory = std::make_shared<OutputFactory>(mouse);
-        Common::Input::RegisterInputFactory(mouse->GetEngineName(), mouse_factory);
-        Common::Input::RegisterOutputFactory(mouse->GetEngineName(), mouse_output_factory);
+        std::shared_ptr<InputFactory> input_factory = std::make_shared<InputFactory>(engine);
+        std::shared_ptr<OutputFactory> output_factory = std::make_shared<OutputFactory>(engine);
+        Common::Input::RegisterInputFactory(engine->GetEngineName(), std::move(input_factory));
+        Common::Input::RegisterOutputFactory(engine->GetEngineName(), std::move(output_factory));
+    }
 
-        touch_screen = std::make_shared<TouchScreen>("touch");
-        touch_screen_factory = std::make_shared<InputFactory>(touch_screen);
-        Common::Input::RegisterInputFactory(touch_screen->GetEngineName(), touch_screen_factory);
-
-        gcadapter = std::make_shared<GCAdapter>("gcpad");
-        gcadapter->SetMappingCallback(mapping_callback);
-        gcadapter_input_factory = std::make_shared<InputFactory>(gcadapter);
-        gcadapter_output_factory = std::make_shared<OutputFactory>(gcadapter);
-        Common::Input::RegisterInputFactory(gcadapter->GetEngineName(), gcadapter_input_factory);
-        Common::Input::RegisterOutputFactory(gcadapter->GetEngineName(), gcadapter_output_factory);
-
-        udp_client = std::make_shared<CemuhookUDP::UDPClient>("cemuhookudp");
-        udp_client->SetMappingCallback(mapping_callback);
-        udp_client_input_factory = std::make_shared<InputFactory>(udp_client);
-        udp_client_output_factory = std::make_shared<OutputFactory>(udp_client);
-        Common::Input::RegisterInputFactory(udp_client->GetEngineName(), udp_client_input_factory);
-        Common::Input::RegisterOutputFactory(udp_client->GetEngineName(),
-                                             udp_client_output_factory);
-
-        tas_input = std::make_shared<TasInput::Tas>("tas");
-        tas_input->SetMappingCallback(mapping_callback);
-        tas_input_factory = std::make_shared<InputFactory>(tas_input);
-        tas_output_factory = std::make_shared<OutputFactory>(tas_input);
-        Common::Input::RegisterInputFactory(tas_input->GetEngineName(), tas_input_factory);
-        Common::Input::RegisterOutputFactory(tas_input->GetEngineName(), tas_output_factory);
-
-        camera = std::make_shared<Camera>("camera");
-        camera->SetMappingCallback(mapping_callback);
-        camera_input_factory = std::make_shared<InputFactory>(camera);
-        camera_output_factory = std::make_shared<OutputFactory>(camera);
-        Common::Input::RegisterInputFactory(camera->GetEngineName(), camera_input_factory);
-        Common::Input::RegisterOutputFactory(camera->GetEngineName(), camera_output_factory);
-
-        virtual_amiibo = std::make_shared<VirtualAmiibo>("virtual_amiibo");
-        virtual_amiibo->SetMappingCallback(mapping_callback);
-        virtual_amiibo_input_factory = std::make_shared<InputFactory>(virtual_amiibo);
-        virtual_amiibo_output_factory = std::make_shared<OutputFactory>(virtual_amiibo);
-        Common::Input::RegisterInputFactory(virtual_amiibo->GetEngineName(),
-                                            virtual_amiibo_input_factory);
-        Common::Input::RegisterOutputFactory(virtual_amiibo->GetEngineName(),
-                                             virtual_amiibo_output_factory);
-
-        virtual_gamepad = std::make_shared<VirtualGamepad>("virtual_gamepad");
-        virtual_gamepad->SetMappingCallback(mapping_callback);
-        virtual_gamepad_input_factory = std::make_shared<InputFactory>(virtual_gamepad);
-        Common::Input::RegisterInputFactory(virtual_gamepad->GetEngineName(),
-                                            virtual_gamepad_input_factory);
+    void Initialize() {
+        mapping_factory = std::make_shared<MappingFactory>();
 
+        RegisterEngine("keyboard", keyboard);
+        RegisterEngine("mouse", mouse);
+        RegisterEngine("touch", touch_screen);
+        RegisterEngine("gcpad", gcadapter);
+        RegisterEngine("cemuhookudp", udp_client);
+        RegisterEngine("tas", tas_input);
+        RegisterEngine("camera", camera);
+        RegisterEngine("virtual_amiibo", virtual_amiibo);
+        RegisterEngine("virtual_gamepad", virtual_gamepad);
 #ifdef HAVE_SDL2
-        sdl = std::make_shared<SDLDriver>("sdl");
-        sdl->SetMappingCallback(mapping_callback);
-        sdl_input_factory = std::make_shared<InputFactory>(sdl);
-        sdl_output_factory = std::make_shared<OutputFactory>(sdl);
-        Common::Input::RegisterInputFactory(sdl->GetEngineName(), sdl_input_factory);
-        Common::Input::RegisterOutputFactory(sdl->GetEngineName(), sdl_output_factory);
+        RegisterEngine("sdl", sdl);
 #endif
 
         Common::Input::RegisterInputFactory("touch_from_button",
@@ -107,45 +61,25 @@ struct InputSubsystem::Impl {
                                             std::make_shared<StickFromButton>());
     }
 
+    template <typename Engine>
+    void UnregisterEngine(std::shared_ptr<Engine>& engine) {
+        Common::Input::UnregisterInputFactory(engine->GetEngineName());
+        Common::Input::UnregisterOutputFactory(engine->GetEngineName());
+        engine.reset();
+    }
+
     void Shutdown() {
-        Common::Input::UnregisterInputFactory(keyboard->GetEngineName());
-        Common::Input::UnregisterOutputFactory(keyboard->GetEngineName());
-        keyboard.reset();
-
-        Common::Input::UnregisterInputFactory(mouse->GetEngineName());
-        Common::Input::UnregisterOutputFactory(mouse->GetEngineName());
-        mouse.reset();
-
-        Common::Input::UnregisterInputFactory(touch_screen->GetEngineName());
-        touch_screen.reset();
-
-        Common::Input::UnregisterInputFactory(gcadapter->GetEngineName());
-        Common::Input::UnregisterOutputFactory(gcadapter->GetEngineName());
-        gcadapter.reset();
-
-        Common::Input::UnregisterInputFactory(udp_client->GetEngineName());
-        Common::Input::UnregisterOutputFactory(udp_client->GetEngineName());
-        udp_client.reset();
-
-        Common::Input::UnregisterInputFactory(tas_input->GetEngineName());
-        Common::Input::UnregisterOutputFactory(tas_input->GetEngineName());
-        tas_input.reset();
-
-        Common::Input::UnregisterInputFactory(camera->GetEngineName());
-        Common::Input::UnregisterOutputFactory(camera->GetEngineName());
-        camera.reset();
-
-        Common::Input::UnregisterInputFactory(virtual_amiibo->GetEngineName());
-        Common::Input::UnregisterOutputFactory(virtual_amiibo->GetEngineName());
-        virtual_amiibo.reset();
-
-        Common::Input::UnregisterInputFactory(virtual_gamepad->GetEngineName());
-        virtual_gamepad.reset();
-
+        UnregisterEngine(keyboard);
+        UnregisterEngine(mouse);
+        UnregisterEngine(touch_screen);
+        UnregisterEngine(gcadapter);
+        UnregisterEngine(udp_client);
+        UnregisterEngine(tas_input);
+        UnregisterEngine(camera);
+        UnregisterEngine(virtual_amiibo);
+        UnregisterEngine(virtual_gamepad);
 #ifdef HAVE_SDL2
-        Common::Input::UnregisterInputFactory(sdl->GetEngineName());
-        Common::Input::UnregisterOutputFactory(sdl->GetEngineName());
-        sdl.reset();
+        UnregisterEngine(sdl);
 #endif
 
         Common::Input::UnregisterInputFactory("touch_from_button");
@@ -173,117 +107,86 @@ struct InputSubsystem::Impl {
         return devices;
     }
 
-    [[nodiscard]] AnalogMapping GetAnalogMappingForDevice(
+    [[nodiscard]] std::shared_ptr<InputEngine> GetInputEngine(
         const Common::ParamPackage& params) const {
         if (!params.Has("engine") || params.Get("engine", "") == "any") {
-            return {};
+            return nullptr;
         }
         const std::string engine = params.Get("engine", "");
+        if (engine == keyboard->GetEngineName()) {
+            return keyboard;
+        }
         if (engine == mouse->GetEngineName()) {
-            return mouse->GetAnalogMappingForDevice(params);
+            return mouse;
         }
         if (engine == gcadapter->GetEngineName()) {
-            return gcadapter->GetAnalogMappingForDevice(params);
+            return gcadapter;
         }
         if (engine == udp_client->GetEngineName()) {
-            return udp_client->GetAnalogMappingForDevice(params);
-        }
-        if (engine == tas_input->GetEngineName()) {
-            return tas_input->GetAnalogMappingForDevice(params);
+            return udp_client;
         }
 #ifdef HAVE_SDL2
         if (engine == sdl->GetEngineName()) {
-            return sdl->GetAnalogMappingForDevice(params);
+            return sdl;
         }
 #endif
-        return {};
+        return nullptr;
+    }
+
+    [[nodiscard]] AnalogMapping GetAnalogMappingForDevice(
+        const Common::ParamPackage& params) const {
+        const auto input_engine = GetInputEngine(params);
+
+        if (input_engine == nullptr) {
+            return {};
+        }
+
+        return input_engine->GetAnalogMappingForDevice(params);
     }
 
     [[nodiscard]] ButtonMapping GetButtonMappingForDevice(
         const Common::ParamPackage& params) const {
-        if (!params.Has("engine") || params.Get("engine", "") == "any") {
+        const auto input_engine = GetInputEngine(params);
+
+        if (input_engine == nullptr) {
             return {};
         }
-        const std::string engine = params.Get("engine", "");
-        if (engine == gcadapter->GetEngineName()) {
-            return gcadapter->GetButtonMappingForDevice(params);
-        }
-        if (engine == udp_client->GetEngineName()) {
-            return udp_client->GetButtonMappingForDevice(params);
-        }
-        if (engine == tas_input->GetEngineName()) {
-            return tas_input->GetButtonMappingForDevice(params);
-        }
-#ifdef HAVE_SDL2
-        if (engine == sdl->GetEngineName()) {
-            return sdl->GetButtonMappingForDevice(params);
-        }
-#endif
-        return {};
+
+        return input_engine->GetButtonMappingForDevice(params);
     }
 
     [[nodiscard]] MotionMapping GetMotionMappingForDevice(
         const Common::ParamPackage& params) const {
-        if (!params.Has("engine") || params.Get("engine", "") == "any") {
+        const auto input_engine = GetInputEngine(params);
+
+        if (input_engine == nullptr) {
             return {};
         }
-        const std::string engine = params.Get("engine", "");
-        if (engine == udp_client->GetEngineName()) {
-            return udp_client->GetMotionMappingForDevice(params);
-        }
-#ifdef HAVE_SDL2
-        if (engine == sdl->GetEngineName()) {
-            return sdl->GetMotionMappingForDevice(params);
-        }
-#endif
-        return {};
+
+        return input_engine->GetMotionMappingForDevice(params);
     }
 
     Common::Input::ButtonNames GetButtonName(const Common::ParamPackage& params) const {
         if (!params.Has("engine") || params.Get("engine", "") == "any") {
             return Common::Input::ButtonNames::Undefined;
         }
-        const std::string engine = params.Get("engine", "");
-        if (engine == mouse->GetEngineName()) {
-            return mouse->GetUIName(params);
+        const auto input_engine = GetInputEngine(params);
+
+        if (input_engine == nullptr) {
+            return Common::Input::ButtonNames::Invalid;
         }
-        if (engine == gcadapter->GetEngineName()) {
-            return gcadapter->GetUIName(params);
-        }
-        if (engine == udp_client->GetEngineName()) {
-            return udp_client->GetUIName(params);
-        }
-        if (engine == tas_input->GetEngineName()) {
-            return tas_input->GetUIName(params);
-        }
-#ifdef HAVE_SDL2
-        if (engine == sdl->GetEngineName()) {
-            return sdl->GetUIName(params);
-        }
-#endif
-        return Common::Input::ButtonNames::Invalid;
+
+        return input_engine->GetUIName(params);
     }
 
     bool IsStickInverted(const Common::ParamPackage& params) {
-        const std::string engine = params.Get("engine", "");
-        if (engine == mouse->GetEngineName()) {
-            return mouse->IsStickInverted(params);
+        const auto input_engine = GetInputEngine(params);
+
+        if (input_engine == nullptr) {
+            return false;
         }
-        if (engine == gcadapter->GetEngineName()) {
-            return gcadapter->IsStickInverted(params);
-        }
-        if (engine == udp_client->GetEngineName()) {
-            return udp_client->IsStickInverted(params);
-        }
-        if (engine == tas_input->GetEngineName()) {
-            return tas_input->IsStickInverted(params);
-        }
-#ifdef HAVE_SDL2
-        if (engine == sdl->GetEngineName()) {
-            return sdl->IsStickInverted(params);
-        }
-#endif
-        return false;
+
+        return input_engine->IsStickInverted(params);
     }
 
     bool IsController(const Common::ParamPackage& params) {
@@ -353,28 +256,8 @@ struct InputSubsystem::Impl {
     std::shared_ptr<VirtualAmiibo> virtual_amiibo;
     std::shared_ptr<VirtualGamepad> virtual_gamepad;
 
-    std::shared_ptr<InputFactory> keyboard_factory;
-    std::shared_ptr<InputFactory> mouse_factory;
-    std::shared_ptr<InputFactory> gcadapter_input_factory;
-    std::shared_ptr<InputFactory> touch_screen_factory;
-    std::shared_ptr<InputFactory> udp_client_input_factory;
-    std::shared_ptr<InputFactory> tas_input_factory;
-    std::shared_ptr<InputFactory> camera_input_factory;
-    std::shared_ptr<InputFactory> virtual_amiibo_input_factory;
-    std::shared_ptr<InputFactory> virtual_gamepad_input_factory;
-
-    std::shared_ptr<OutputFactory> keyboard_output_factory;
-    std::shared_ptr<OutputFactory> mouse_output_factory;
-    std::shared_ptr<OutputFactory> gcadapter_output_factory;
-    std::shared_ptr<OutputFactory> udp_client_output_factory;
-    std::shared_ptr<OutputFactory> tas_output_factory;
-    std::shared_ptr<OutputFactory> camera_output_factory;
-    std::shared_ptr<OutputFactory> virtual_amiibo_output_factory;
-
 #ifdef HAVE_SDL2
     std::shared_ptr<SDLDriver> sdl;
-    std::shared_ptr<InputFactory> sdl_input_factory;
-    std::shared_ptr<OutputFactory> sdl_output_factory;
 #endif
 };
 

src/tests/CMakeLists.txt

@@ -8,6 +8,7 @@ add_executable(tests
     common/host_memory.cpp
     common/param_package.cpp
     common/ring_buffer.cpp
+    common/scratch_buffer.cpp
     common/unique_function.cpp
     core/core_timing.cpp
     core/internal_network/network.cpp

src/tests/common/scratch_buffer.cpp

@@ -0,0 +1,200 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <algorithm>
+#include <array>
+#include <cstring>
+#include <span>
+#include <catch2/catch.hpp>
+#include "common/common_types.h"
+#include "common/scratch_buffer.h"
+
+namespace Common {
+
+TEST_CASE("ScratchBuffer: Basic Test", "[common]") {
+    ScratchBuffer<u8> buf;
+
+    REQUIRE(buf.size() == 0U);
+    REQUIRE(buf.capacity() == 0U);
+
+    std::array<u8, 10> payload;
+    payload.fill(66);
+
+    buf.resize(payload.size());
+    REQUIRE(buf.size() == payload.size());
+    REQUIRE(buf.capacity() == payload.size());
+
+    std::memcpy(buf.data(), payload.data(), payload.size());
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+}
+
+TEST_CASE("ScratchBuffer: resize_destructive Grow", "[common]") {
+    std::array<u8, 10> payload;
+    payload.fill(66);
+
+    ScratchBuffer<u8> buf(payload.size());
+    REQUIRE(buf.size() == payload.size());
+    REQUIRE(buf.capacity() == payload.size());
+
+    // Increasing the size should reallocate the buffer
+    buf.resize_destructive(payload.size() * 2);
+    REQUIRE(buf.size() == payload.size() * 2);
+    REQUIRE(buf.capacity() == payload.size() * 2);
+
+    // Since the buffer is not value initialized, reading its data will be garbage
+}
+
+TEST_CASE("ScratchBuffer: resize_destructive Shrink", "[common]") {
+    std::array<u8, 10> payload;
+    payload.fill(66);
+
+    ScratchBuffer<u8> buf(payload.size());
+    REQUIRE(buf.size() == payload.size());
+    REQUIRE(buf.capacity() == payload.size());
+
+    std::memcpy(buf.data(), payload.data(), payload.size());
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+
+    // Decreasing the size should not cause a buffer reallocation
+    // This can be tested by ensuring the buffer capacity and data has not changed,
+    buf.resize_destructive(1U);
+    REQUIRE(buf.size() == 1U);
+    REQUIRE(buf.capacity() == payload.size());
+
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+}
+
+TEST_CASE("ScratchBuffer: resize Grow u8", "[common]") {
+    std::array<u8, 10> payload;
+    payload.fill(66);
+
+    ScratchBuffer<u8> buf(payload.size());
+    REQUIRE(buf.size() == payload.size());
+    REQUIRE(buf.capacity() == payload.size());
+
+    std::memcpy(buf.data(), payload.data(), payload.size());
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+
+    // Increasing the size should reallocate the buffer
+    buf.resize(payload.size() * 2);
+    REQUIRE(buf.size() == payload.size() * 2);
+    REQUIRE(buf.capacity() == payload.size() * 2);
+
+    // resize() keeps the previous data intact
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+}
+
+TEST_CASE("ScratchBuffer: resize Grow u64", "[common]") {
+    std::array<u64, 10> payload;
+    payload.fill(6666);
+
+    ScratchBuffer<u64> buf(payload.size());
+    REQUIRE(buf.size() == payload.size());
+    REQUIRE(buf.capacity() == payload.size());
+
+    std::memcpy(buf.data(), payload.data(), payload.size() * sizeof(u64));
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+
+    // Increasing the size should reallocate the buffer
+    buf.resize(payload.size() * 2);
+    REQUIRE(buf.size() == payload.size() * 2);
+    REQUIRE(buf.capacity() == payload.size() * 2);
+
+    // resize() keeps the previous data intact
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+}
+
+TEST_CASE("ScratchBuffer: resize Shrink", "[common]") {
+    std::array<u8, 10> payload;
+    payload.fill(66);
+
+    ScratchBuffer<u8> buf(payload.size());
+    REQUIRE(buf.size() == payload.size());
+    REQUIRE(buf.capacity() == payload.size());
+
+    std::memcpy(buf.data(), payload.data(), payload.size());
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+
+    // Decreasing the size should not cause a buffer reallocation
+    // This can be tested by ensuring the buffer capacity and data has not changed,
+    buf.resize(1U);
+    REQUIRE(buf.size() == 1U);
+    REQUIRE(buf.capacity() == payload.size());
+
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+}
+
+TEST_CASE("ScratchBuffer: Span Size", "[common]") {
+    std::array<u8, 10> payload;
+    payload.fill(66);
+
+    ScratchBuffer<u8> buf(payload.size());
+    REQUIRE(buf.size() == payload.size());
+    REQUIRE(buf.capacity() == payload.size());
+
+    std::memcpy(buf.data(), payload.data(), payload.size());
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+
+    buf.resize(3U);
+    REQUIRE(buf.size() == 3U);
+    REQUIRE(buf.capacity() == payload.size());
+
+    const auto buf_span = std::span<u8>(buf);
+    // The span size is the last requested size of the buffer, not its capacity
+    REQUIRE(buf_span.size() == buf.size());
+
+    for (size_t i = 0; i < buf_span.size(); ++i) {
+        REQUIRE(buf_span[i] == buf[i]);
+        REQUIRE(buf_span[i] == payload[i]);
+    }
+}
+
+TEST_CASE("ScratchBuffer: Span Writes", "[common]") {
+    std::array<u8, 10> payload;
+    payload.fill(66);
+
+    ScratchBuffer<u8> buf(payload.size());
+    REQUIRE(buf.size() == payload.size());
+    REQUIRE(buf.capacity() == payload.size());
+
+    std::memcpy(buf.data(), payload.data(), payload.size());
+    for (size_t i = 0; i < payload.size(); ++i) {
+        REQUIRE(buf[i] == payload[i]);
+    }
+
+    buf.resize(3U);
+    REQUIRE(buf.size() == 3U);
+    REQUIRE(buf.capacity() == payload.size());
+
+    const auto buf_span = std::span<u8>(buf);
+    REQUIRE(buf_span.size() == buf.size());
+
+    for (size_t i = 0; i < buf_span.size(); ++i) {
+        const auto new_value = static_cast<u8>(i + 1U);
+        // Writes to a span of the scratch buffer will propogate to the buffer itself
+        buf_span[i] = new_value;
+        REQUIRE(buf[i] == new_value);
+    }
+}
+
+} // namespace Common

src/video_core/buffer_cache/buffer_cache.h

@@ -20,6 +20,7 @@
 #include "common/lru_cache.h"
 #include "common/microprofile.h"
 #include "common/polyfill_ranges.h"
+#include "common/scratch_buffer.h"
 #include "common/settings.h"
 #include "core/memory.h"
 #include "video_core/buffer_cache/buffer_base.h"
@@ -422,8 +423,7 @@ private:
     IntervalSet common_ranges;
     std::deque<IntervalSet> committed_ranges;
 
-    size_t immediate_buffer_capacity = 0;
-    std::unique_ptr<u8[]> immediate_buffer_alloc;
+    Common::ScratchBuffer<u8> immediate_buffer_alloc;
 
     struct LRUItemParams {
         using ObjectType = BufferId;
@@ -666,9 +666,10 @@ void BufferCache<P>::BindHostGeometryBuffers(bool is_indexed) {
         BindHostIndexBuffer();
     } else if constexpr (!HAS_FULL_INDEX_AND_PRIMITIVE_SUPPORT) {
         const auto& draw_state = maxwell3d->draw_manager->GetDrawState();
-        if (draw_state.topology == Maxwell::PrimitiveTopology::Quads) {
-            runtime.BindQuadArrayIndexBuffer(draw_state.vertex_buffer.first,
-                                             draw_state.vertex_buffer.count);
+        if (draw_state.topology == Maxwell::PrimitiveTopology::Quads ||
+            draw_state.topology == Maxwell::PrimitiveTopology::QuadStrip) {
+            runtime.BindQuadIndexBuffer(draw_state.topology, draw_state.vertex_buffer.first,
+                                        draw_state.vertex_buffer.count);
         }
     }
     BindHostVertexBuffers();
@@ -1926,11 +1927,8 @@ std::span<const u8> BufferCache<P>::ImmediateBufferWithData(VAddr cpu_addr, size
 
 template <class P>
 std::span<u8> BufferCache<P>::ImmediateBuffer(size_t wanted_capacity) {
-    if (wanted_capacity > immediate_buffer_capacity) {
-        immediate_buffer_capacity = wanted_capacity;
-        immediate_buffer_alloc = std::make_unique<u8[]>(wanted_capacity);
-    }
-    return std::span<u8>(immediate_buffer_alloc.get(), wanted_capacity);
+    immediate_buffer_alloc.resize_destructive(wanted_capacity);
+    return std::span<u8>(immediate_buffer_alloc.data(), wanted_capacity);
 }
 
 template <class P>

src/video_core/dma_pusher.cpp

@@ -56,7 +56,7 @@ bool DmaPusher::Step() {
 
     if (command_list.prefetch_command_list.size()) {
         // Prefetched command list from nvdrv, used for things like synchronization
-        command_headers = std::move(command_list.prefetch_command_list);
+        ProcessCommands(command_list.prefetch_command_list);
         dma_pushbuffer.pop();
     } else {
         const CommandListHeader command_list_header{
@@ -74,7 +74,7 @@ bool DmaPusher::Step() {
         }
 
         // Push buffer non-empty, read a word
-        command_headers.resize(command_list_header.size);
+        command_headers.resize_destructive(command_list_header.size);
         if (Settings::IsGPULevelHigh()) {
             memory_manager.ReadBlock(dma_get, command_headers.data(),
                                      command_list_header.size * sizeof(u32));
@@ -82,16 +82,21 @@ bool DmaPusher::Step() {
             memory_manager.ReadBlockUnsafe(dma_get, command_headers.data(),
                                            command_list_header.size * sizeof(u32));
         }
+        ProcessCommands(command_headers);
     }
-    for (std::size_t index = 0; index < command_headers.size();) {
-        const CommandHeader& command_header = command_headers[index];
+
+    return true;
+}
+
+void DmaPusher::ProcessCommands(std::span<const CommandHeader> commands) {
+    for (std::size_t index = 0; index < commands.size();) {
+        const CommandHeader& command_header = commands[index];
 
         if (dma_state.method_count) {
             // Data word of methods command
             if (dma_state.non_incrementing) {
                 const u32 max_write = static_cast<u32>(
-                    std::min<std::size_t>(index + dma_state.method_count, command_headers.size()) -
-                    index);
+                    std::min<std::size_t>(index + dma_state.method_count, commands.size()) - index);
                 CallMultiMethod(&command_header.argument, max_write);
                 dma_state.method_count -= max_write;
                 dma_state.is_last_call = true;
@@ -142,8 +147,6 @@ bool DmaPusher::Step() {
         }
         index++;
     }
-
-    return true;
 }
 
 void DmaPusher::SetState(const CommandHeader& command_header) {

src/video_core/dma_pusher.h

@@ -4,11 +4,13 @@
 #pragma once
 
 #include <array>
+#include <span>
 #include <vector>
 #include <queue>
 
 #include "common/bit_field.h"
 #include "common/common_types.h"
+#include "common/scratch_buffer.h"
 #include "video_core/engines/engine_interface.h"
 #include "video_core/engines/puller.h"
 
@@ -136,13 +138,15 @@ private:
     static constexpr u32 non_puller_methods = 0x40;
     static constexpr u32 max_subchannels = 8;
     bool Step();
+    void ProcessCommands(std::span<const CommandHeader> commands);
 
     void SetState(const CommandHeader& command_header);
 
     void CallMethod(u32 argument) const;
     void CallMultiMethod(const u32* base_start, u32 num_methods) const;
 
-    std::vector<CommandHeader> command_headers; ///< Buffer for list of commands fetched at once
+    Common::ScratchBuffer<CommandHeader>
+        command_headers; ///< Buffer for list of commands fetched at once
 
     std::queue<CommandList> dma_pushbuffer; ///< Queue of command lists to be processed
     std::size_t dma_pushbuffer_subindex{};  ///< Index within a command list within the pushbuffer
@@ -159,7 +163,7 @@ private:
     DmaState dma_state{};
     bool dma_increment_once{};
 
-    bool ib_enable{true}; ///< IB mode enabled
+    const bool ib_enable{true}; ///< IB mode enabled
 
     std::array<Engines::EngineInterface*, max_subchannels> subchannels{};
 

src/video_core/engines/engine_upload.cpp

@@ -24,7 +24,7 @@ void State::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
 void State::ProcessExec(const bool is_linear_) {
     write_offset = 0;
     copy_size = regs.line_length_in * regs.line_count;
-    inner_buffer.resize(copy_size);
+    inner_buffer.resize_destructive(copy_size);
     is_linear = is_linear_;
 }
 
@@ -70,7 +70,7 @@ void State::ProcessData(std::span<const u8> read_buffer) {
         const std::size_t dst_size = Tegra::Texture::CalculateSize(
             true, bytes_per_pixel, width, regs.dest.height, regs.dest.depth,
             regs.dest.BlockHeight(), regs.dest.BlockDepth());
-        tmp_buffer.resize(dst_size);
+        tmp_buffer.resize_destructive(dst_size);
         memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size);
         Tegra::Texture::SwizzleSubrect(tmp_buffer, read_buffer, bytes_per_pixel, width,
                                        regs.dest.height, regs.dest.depth, x_offset, regs.dest.y,

src/video_core/engines/engine_upload.h

@@ -4,9 +4,10 @@
 #pragma once
 
 #include <span>
-#include <vector>
+
 #include "common/bit_field.h"
 #include "common/common_types.h"
+#include "common/scratch_buffer.h"
 
 namespace Tegra {
 class MemoryManager;
@@ -73,8 +74,8 @@ private:
 
     u32 write_offset = 0;
     u32 copy_size = 0;
-    std::vector<u8> inner_buffer;
-    std::vector<u8> tmp_buffer;
+    Common::ScratchBuffer<u8> inner_buffer;
+    Common::ScratchBuffer<u8> tmp_buffer;
     bool is_linear = false;
     Registers& regs;
     MemoryManager& memory_manager;

src/video_core/engines/maxwell_dma.cpp

@@ -184,12 +184,8 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
     const size_t src_size =
         CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
 
-    if (read_buffer.size() < src_size) {
-        read_buffer.resize(src_size);
-    }
-    if (write_buffer.size() < dst_size) {
-        write_buffer.resize(dst_size);
-    }
+    read_buffer.resize_destructive(src_size);
+    write_buffer.resize_destructive(dst_size);
 
     memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
     memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
@@ -235,12 +231,8 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
         CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
     const size_t src_size = static_cast<size_t>(regs.pitch_in) * regs.line_count;
 
-    if (read_buffer.size() < src_size) {
-        read_buffer.resize(src_size);
-    }
-    if (write_buffer.size() < dst_size) {
-        write_buffer.resize(dst_size);
-    }
+    read_buffer.resize_destructive(src_size);
+    write_buffer.resize_destructive(dst_size);
 
     memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
     if (Settings::IsGPULevelExtreme()) {
@@ -269,12 +261,8 @@ void MaxwellDMA::FastCopyBlockLinearToPitch() {
     pos_x = pos_x % x_in_gob;
     pos_y = pos_y % 8;
 
-    if (read_buffer.size() < src_size) {
-        read_buffer.resize(src_size);
-    }
-    if (write_buffer.size() < dst_size) {
-        write_buffer.resize(dst_size);
-    }
+    read_buffer.resize_destructive(src_size);
+    write_buffer.resize_destructive(dst_size);
 
     if (Settings::IsGPULevelExtreme()) {
         memory_manager.ReadBlock(regs.offset_in + offset, read_buffer.data(), src_size);
@@ -333,14 +321,10 @@ void MaxwellDMA::CopyBlockLinearToBlockLinear() {
     const u32 pitch = x_elements * bytes_per_pixel;
     const size_t mid_buffer_size = pitch * regs.line_count;
 
-    if (read_buffer.size() < src_size) {
-        read_buffer.resize(src_size);
-    }
-    if (write_buffer.size() < dst_size) {
-        write_buffer.resize(dst_size);
-    }
+    read_buffer.resize_destructive(src_size);
+    write_buffer.resize_destructive(dst_size);
 
-    intermediate_buffer.resize(mid_buffer_size);
+    intermediate_buffer.resize_destructive(mid_buffer_size);
 
     memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
     memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);

src/video_core/engines/maxwell_dma.h

@@ -6,8 +6,10 @@
 #include <array>
 #include <cstddef>
 #include <vector>
+
 #include "common/bit_field.h"
 #include "common/common_types.h"
+#include "common/scratch_buffer.h"
 #include "video_core/engines/engine_interface.h"
 
 namespace Core {
@@ -234,9 +236,9 @@ private:
     MemoryManager& memory_manager;
     VideoCore::RasterizerInterface* rasterizer = nullptr;
 
-    std::vector<u8> read_buffer;
-    std::vector<u8> write_buffer;
-    std::vector<u8> intermediate_buffer;
+    Common::ScratchBuffer<u8> read_buffer;
+    Common::ScratchBuffer<u8> write_buffer;
+    Common::ScratchBuffer<u8> intermediate_buffer;
 
     static constexpr std::size_t NUM_REGS = 0x800;
     struct Regs {

src/video_core/host1x/vic.cpp

@@ -155,7 +155,7 @@ void Vic::WriteRGBFrame(const AVFrame* frame, const VicConfig& config) {
         // swizzle pitch linear to block linear
         const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
         const auto size = Texture::CalculateSize(true, 4, width, height, 1, block_height, 0);
-        luma_buffer.resize(size);
+        luma_buffer.resize_destructive(size);
         std::span<const u8> frame_buff(converted_frame_buf_addr, 4 * width * height);
         Texture::SwizzleSubrect(luma_buffer, frame_buff, 4, width, height, 1, 0, 0, width, height,
                                 block_height, 0, width * 4);
@@ -181,8 +181,8 @@ void Vic::WriteYUVFrame(const AVFrame* frame, const VicConfig& config) {
 
     const auto stride = static_cast<size_t>(frame->linesize[0]);
 
-    luma_buffer.resize(aligned_width * surface_height);
-    chroma_buffer.resize(aligned_width * surface_height / 2);
+    luma_buffer.resize_destructive(aligned_width * surface_height);
+    chroma_buffer.resize_destructive(aligned_width * surface_height / 2);
 
     // Populate luma buffer
     const u8* luma_src = frame->data[0];

src/video_core/host1x/vic.h

@@ -4,8 +4,9 @@
 #pragma once
 
 #include <memory>
-#include <vector>
+
 #include "common/common_types.h"
+#include "common/scratch_buffer.h"
 
 struct SwsContext;
 
@@ -49,8 +50,8 @@ private:
     /// size does not change during a stream
     using AVMallocPtr = std::unique_ptr<u8, decltype(&av_free)>;
     AVMallocPtr converted_frame_buffer;
-    std::vector<u8> luma_buffer;
-    std::vector<u8> chroma_buffer;
+    Common::ScratchBuffer<u8> luma_buffer;
+    Common::ScratchBuffer<u8> chroma_buffer;
 
     GPUVAddr config_struct_address{};
     GPUVAddr output_surface_luma_address{};

src/video_core/host_shaders/vulkan_quad_indexed.comp

@@ -16,6 +16,7 @@ layout (std430, set = 0, binding = 1) writeonly buffer OutputBuffer {
 layout (push_constant) uniform PushConstants {
     uint base_vertex;
     int index_shift; // 0: uint8, 1: uint16, 2: uint32
+    int is_strip; // 0: quads 1: quadstrip
 };
 
 void main() {
@@ -28,9 +29,10 @@ void main() {
     int flipped_shift = 2 - index_shift;
     int mask = (1 << flipped_shift) - 1;
 
-    const int quad_swizzle[6] = int[](0, 1, 2, 0, 2, 3);
+    const int quads_swizzle[6] = int[](0, 1, 2, 0, 2, 3);
+    const int quad_strip_swizzle[6] = int[](0, 3, 1, 0, 2, 3);
     for (uint vertex = 0; vertex < 6; ++vertex) {
-        int offset = primitive * 4 + quad_swizzle[vertex];
+        int offset = (is_strip == 0 ? primitive * 4 + quads_swizzle[vertex] : primitive * 2 + quad_strip_swizzle[vertex]);
         int int_offset = offset >> flipped_shift;
         int bit_offset = (offset & mask) * index_size;
         uint packed_input = input_indexes[int_offset];

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -301,6 +301,8 @@ VkPrimitiveTopology PrimitiveTopology([[maybe_unused]] const Device& device,
         return VK_PRIMITIVE_TOPOLOGY_POINT_LIST;
     case Maxwell::PrimitiveTopology::Lines:
         return VK_PRIMITIVE_TOPOLOGY_LINE_LIST;
+    case Maxwell::PrimitiveTopology::LineLoop:
+        return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST;
     case Maxwell::PrimitiveTopology::LineStrip:
         return VK_PRIMITIVE_TOPOLOGY_LINE_STRIP;
     case Maxwell::PrimitiveTopology::Triangles:
@@ -309,15 +311,28 @@ VkPrimitiveTopology PrimitiveTopology([[maybe_unused]] const Device& device,
         return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP;
     case Maxwell::PrimitiveTopology::TriangleFan:
         return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN;
+    case Maxwell::PrimitiveTopology::LinesAdjacency:
+        return VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY;
+    case Maxwell::PrimitiveTopology::LineStripAdjacency:
+        return VK_PRIMITIVE_TOPOLOGY_LINE_STRIP_WITH_ADJACENCY;
+    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
+        return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY;
+    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
+        return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP_WITH_ADJACENCY;
     case Maxwell::PrimitiveTopology::Quads:
-        // TODO(Rodrigo): Use VK_PRIMITIVE_TOPOLOGY_QUAD_LIST_EXT whenever it releases
+    case Maxwell::PrimitiveTopology::QuadStrip:
+        // TODO: Use VK_PRIMITIVE_TOPOLOGY_QUAD_LIST_EXT/VK_PRIMITIVE_TOPOLOGY_QUAD_STRIP_EXT
+        // whenever it releases
         return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST;
     case Maxwell::PrimitiveTopology::Patches:
         return VK_PRIMITIVE_TOPOLOGY_PATCH_LIST;
-    default:
-        UNIMPLEMENTED_MSG("Unimplemented topology={}", topology);
-        return {};
+    case Maxwell::PrimitiveTopology::Polygon:
+        LOG_WARNING(Render_Vulkan, "Draw mode is Polygon with a polygon mode of lines should be a "
+                                   "single body and not a bunch of triangles.");
+        return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN;
     }
+    UNIMPLEMENTED_MSG("Unimplemented topology={}", topology);
+    return {};
 }
 
 VkFormat VertexFormat(const Device& device, Maxwell::VertexAttribute::Type type,

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -51,15 +51,6 @@ size_t BytesPerIndex(VkIndexType index_type) {
     }
 }
 
-template <typename T>
-std::array<T, 6> MakeQuadIndices(u32 quad, u32 first) {
-    std::array<T, 6> indices{0, 1, 2, 0, 2, 3};
-    for (T& index : indices) {
-        index = static_cast<T>(first + index + quad * 4);
-    }
-    return indices;
-}
-
 vk::Buffer CreateBuffer(const Device& device, u64 size) {
     VkBufferUsageFlags flags =
         VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
@@ -123,6 +114,187 @@ VkBufferView Buffer::View(u32 offset, u32 size, VideoCore::Surface::PixelFormat
     return *views.back().handle;
 }
 
+class QuadIndexBuffer {
+public:
+    QuadIndexBuffer(const Device& device_, MemoryAllocator& memory_allocator_,
+                    Scheduler& scheduler_, StagingBufferPool& staging_pool_)
+        : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_},
+          staging_pool{staging_pool_} {}
+
+    virtual ~QuadIndexBuffer() = default;
+
+    void UpdateBuffer(u32 num_indices_) {
+        if (num_indices_ <= num_indices) {
+            return;
+        }
+
+        scheduler.Finish();
+
+        num_indices = num_indices_;
+        index_type = IndexTypeFromNumElements(device, num_indices);
+
+        const u32 num_quads = GetQuadsNum(num_indices);
+        const u32 num_triangle_indices = num_quads * 6;
+        const u32 num_first_offset_copies = 4;
+        const size_t bytes_per_index = BytesPerIndex(index_type);
+        const size_t size_bytes = num_triangle_indices * bytes_per_index * num_first_offset_copies;
+        buffer = device.GetLogical().CreateBuffer(VkBufferCreateInfo{
+            .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
+            .pNext = nullptr,
+            .flags = 0,
+            .size = size_bytes,
+            .usage = VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
+            .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
+            .queueFamilyIndexCount = 0,
+            .pQueueFamilyIndices = nullptr,
+        });
+        if (device.HasDebuggingToolAttached()) {
+            buffer.SetObjectNameEXT("Quad LUT");
+        }
+        memory_commit = memory_allocator.Commit(buffer, MemoryUsage::DeviceLocal);
+
+        const StagingBufferRef staging = staging_pool.Request(size_bytes, MemoryUsage::Upload);
+        u8* staging_data = staging.mapped_span.data();
+        const size_t quad_size = bytes_per_index * 6;
+
+        for (u32 first = 0; first < num_first_offset_copies; ++first) {
+            for (u32 quad = 0; quad < num_quads; ++quad) {
+                MakeAndUpdateIndices(staging_data, quad_size, quad, first);
+                staging_data += quad_size;
+            }
+        }
+
+        scheduler.RequestOutsideRenderPassOperationContext();
+        scheduler.Record([src_buffer = staging.buffer, src_offset = staging.offset,
+                          dst_buffer = *buffer, size_bytes](vk::CommandBuffer cmdbuf) {
+            const VkBufferCopy copy{
+                .srcOffset = src_offset,
+                .dstOffset = 0,
+                .size = size_bytes,
+            };
+            const VkBufferMemoryBarrier write_barrier{
+                .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
+                .pNext = nullptr,
+                .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+                .dstAccessMask = VK_ACCESS_INDEX_READ_BIT,
+                .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                .buffer = dst_buffer,
+                .offset = 0,
+                .size = size_bytes,
+            };
+            cmdbuf.CopyBuffer(src_buffer, dst_buffer, copy);
+            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
+                                   VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, write_barrier);
+        });
+    }
+
+    void BindBuffer(u32 first) {
+        const VkIndexType index_type_ = index_type;
+        const size_t sub_first_offset = static_cast<size_t>(first % 4) * GetQuadsNum(num_indices);
+        const size_t offset =
+            (sub_first_offset + GetQuadsNum(first)) * 6ULL * BytesPerIndex(index_type);
+        scheduler.Record([buffer = *buffer, index_type_, offset](vk::CommandBuffer cmdbuf) {
+            cmdbuf.BindIndexBuffer(buffer, offset, index_type_);
+        });
+    }
+
+protected:
+    virtual u32 GetQuadsNum(u32 num_indices) const = 0;
+
+    virtual void MakeAndUpdateIndices(u8* staging_data, size_t quad_size, u32 quad, u32 first) = 0;
+
+    const Device& device;
+    MemoryAllocator& memory_allocator;
+    Scheduler& scheduler;
+    StagingBufferPool& staging_pool;
+
+    vk::Buffer buffer{};
+    MemoryCommit memory_commit{};
+    VkIndexType index_type{};
+    u32 num_indices = 0;
+};
+
+class QuadArrayIndexBuffer : public QuadIndexBuffer {
+public:
+    QuadArrayIndexBuffer(const Device& device_, MemoryAllocator& memory_allocator_,
+                         Scheduler& scheduler_, StagingBufferPool& staging_pool_)
+        : QuadIndexBuffer(device_, memory_allocator_, scheduler_, staging_pool_) {}
+
+    ~QuadArrayIndexBuffer() = default;
+
+private:
+    u32 GetQuadsNum(u32 num_indices_) const override {
+        return num_indices_ / 4;
+    }
+
+    template <typename T>
+    static std::array<T, 6> MakeIndices(u32 quad, u32 first) {
+        std::array<T, 6> indices{0, 1, 2, 0, 2, 3};
+        for (T& index : indices) {
+            index = static_cast<T>(first + index + quad * 4);
+        }
+        return indices;
+    }
+
+    void MakeAndUpdateIndices(u8* staging_data, size_t quad_size, u32 quad, u32 first) {
+        switch (index_type) {
+        case VK_INDEX_TYPE_UINT8_EXT:
+            std::memcpy(staging_data, MakeIndices<u8>(quad, first).data(), quad_size);
+            break;
+        case VK_INDEX_TYPE_UINT16:
+            std::memcpy(staging_data, MakeIndices<u16>(quad, first).data(), quad_size);
+            break;
+        case VK_INDEX_TYPE_UINT32:
+            std::memcpy(staging_data, MakeIndices<u32>(quad, first).data(), quad_size);
+            break;
+        default:
+            ASSERT(false);
+            break;
+        }
+    }
+};
+
+class QuadStripIndexBuffer : public QuadIndexBuffer {
+public:
+    QuadStripIndexBuffer(const Device& device_, MemoryAllocator& memory_allocator_,
+                         Scheduler& scheduler_, StagingBufferPool& staging_pool_)
+        : QuadIndexBuffer(device_, memory_allocator_, scheduler_, staging_pool_) {}
+
+    ~QuadStripIndexBuffer() = default;
+
+private:
+    u32 GetQuadsNum(u32 num_indices_) const override {
+        return num_indices_ >= 4 ? (num_indices_ - 2) / 2 : 0;
+    }
+
+    template <typename T>
+    static std::array<T, 6> MakeIndices(u32 quad, u32 first) {
+        std::array<T, 6> indices{0, 3, 1, 0, 2, 3};
+        for (T& index : indices) {
+            index = static_cast<T>(first + index + quad * 2);
+        }
+        return indices;
+    }
+
+    void MakeAndUpdateIndices(u8* staging_data, size_t quad_size, u32 quad, u32 first) {
+        switch (index_type) {
+        case VK_INDEX_TYPE_UINT8_EXT:
+            std::memcpy(staging_data, MakeIndices<u8>(quad, first).data(), quad_size);
+            break;
+        case VK_INDEX_TYPE_UINT16:
+            std::memcpy(staging_data, MakeIndices<u16>(quad, first).data(), quad_size);
+            break;
+        case VK_INDEX_TYPE_UINT32:
+            std::memcpy(staging_data, MakeIndices<u32>(quad, first).data(), quad_size);
+            break;
+        default:
+            ASSERT(false);
+            break;
+        }
+    }
+};
+
 BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& memory_allocator_,
                                        Scheduler& scheduler_, StagingBufferPool& staging_pool_,
                                        UpdateDescriptorQueue& update_descriptor_queue_,
@@ -130,7 +302,12 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& m
     : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_},
       staging_pool{staging_pool_}, update_descriptor_queue{update_descriptor_queue_},
       uint8_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
-      quad_index_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue) {}
+      quad_index_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue) {
+    quad_array_index_buffer = std::make_shared<QuadArrayIndexBuffer>(device_, memory_allocator_,
+                                                                     scheduler_, staging_pool_);
+    quad_strip_index_buffer = std::make_shared<QuadStripIndexBuffer>(device_, memory_allocator_,
+                                                                     scheduler_, staging_pool_);
+}
 
 StagingBufferRef BufferCacheRuntime::UploadStagingBuffer(size_t size) {
     return staging_pool.Request(size, MemoryUsage::Upload);
@@ -245,10 +422,11 @@ void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat
     VkIndexType vk_index_type = MaxwellToVK::IndexFormat(index_format);
     VkDeviceSize vk_offset = offset;
     VkBuffer vk_buffer = buffer;
-    if (topology == PrimitiveTopology::Quads) {
+    if (topology == PrimitiveTopology::Quads || topology == PrimitiveTopology::QuadStrip) {
         vk_index_type = VK_INDEX_TYPE_UINT32;
         std::tie(vk_buffer, vk_offset) =
-            quad_index_pass.Assemble(index_format, num_indices, base_vertex, buffer, offset);
+            quad_index_pass.Assemble(index_format, num_indices, base_vertex, buffer, offset,
+                                     topology == PrimitiveTopology::QuadStrip);
     } else if (vk_index_type == VK_INDEX_TYPE_UINT8_EXT && !device.IsExtIndexTypeUint8Supported()) {
         vk_index_type = VK_INDEX_TYPE_UINT16;
         std::tie(vk_buffer, vk_offset) = uint8_pass.Assemble(num_indices, buffer, offset);
@@ -263,7 +441,7 @@ void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat
     });
 }
 
-void BufferCacheRuntime::BindQuadArrayIndexBuffer(u32 first, u32 count) {
+void BufferCacheRuntime::BindQuadIndexBuffer(PrimitiveTopology topology, u32 first, u32 count) {
     if (count == 0) {
         ReserveNullBuffer();
         scheduler.Record([this](vk::CommandBuffer cmdbuf) {
@@ -271,16 +449,14 @@ void BufferCacheRuntime::BindQuadArrayIndexBuffer(u32 first, u32 count) {
         });
         return;
     }
-    ReserveQuadArrayLUT(first + count, true);
 
-    // The LUT has the indices 0, 1, 2, and 3 copied as an array
-    // To apply these 'first' offsets we can apply an offset based on the modulus.
-    const VkIndexType index_type = quad_array_lut_index_type;
-    const size_t sub_first_offset = static_cast<size_t>(first % 4) * (current_num_indices / 4);
-    const size_t offset = (sub_first_offset + first / 4) * 6ULL * BytesPerIndex(index_type);
-    scheduler.Record([buffer = *quad_array_lut, index_type, offset](vk::CommandBuffer cmdbuf) {
-        cmdbuf.BindIndexBuffer(buffer, offset, index_type);
-    });
+    if (topology == PrimitiveTopology::Quads) {
+        quad_array_index_buffer->UpdateBuffer(first + count);
+        quad_array_index_buffer->BindBuffer(first);
+    } else if (topology == PrimitiveTopology::QuadStrip) {
+        quad_strip_index_buffer->UpdateBuffer(first + count);
+        quad_strip_index_buffer->BindBuffer(first);
+    }
 }
 
 void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset, u32 size,
@@ -323,83 +499,6 @@ void BufferCacheRuntime::BindTransformFeedbackBuffer(u32 index, VkBuffer buffer,
     });
 }
 
-void BufferCacheRuntime::ReserveQuadArrayLUT(u32 num_indices, bool wait_for_idle) {
-    if (num_indices <= current_num_indices) {
-        return;
-    }
-    if (wait_for_idle) {
-        scheduler.Finish();
-    }
-    current_num_indices = num_indices;
-    quad_array_lut_index_type = IndexTypeFromNumElements(device, num_indices);
-
-    const u32 num_quads = num_indices / 4;
-    const u32 num_triangle_indices = num_quads * 6;
-    const u32 num_first_offset_copies = 4;
-    const size_t bytes_per_index = BytesPerIndex(quad_array_lut_index_type);
-    const size_t size_bytes = num_triangle_indices * bytes_per_index * num_first_offset_copies;
-    quad_array_lut = device.GetLogical().CreateBuffer(VkBufferCreateInfo{
-        .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
-        .pNext = nullptr,
-        .flags = 0,
-        .size = size_bytes,
-        .usage = VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
-        .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
-        .queueFamilyIndexCount = 0,
-        .pQueueFamilyIndices = nullptr,
-    });
-    if (device.HasDebuggingToolAttached()) {
-        quad_array_lut.SetObjectNameEXT("Quad LUT");
-    }
-    quad_array_lut_commit = memory_allocator.Commit(quad_array_lut, MemoryUsage::DeviceLocal);
-
-    const StagingBufferRef staging = staging_pool.Request(size_bytes, MemoryUsage::Upload);
-    u8* staging_data = staging.mapped_span.data();
-    const size_t quad_size = bytes_per_index * 6;
-    for (u32 first = 0; first < num_first_offset_copies; ++first) {
-        for (u32 quad = 0; quad < num_quads; ++quad) {
-            switch (quad_array_lut_index_type) {
-            case VK_INDEX_TYPE_UINT8_EXT:
-                std::memcpy(staging_data, MakeQuadIndices<u8>(quad, first).data(), quad_size);
-                break;
-            case VK_INDEX_TYPE_UINT16:
-                std::memcpy(staging_data, MakeQuadIndices<u16>(quad, first).data(), quad_size);
-                break;
-            case VK_INDEX_TYPE_UINT32:
-                std::memcpy(staging_data, MakeQuadIndices<u32>(quad, first).data(), quad_size);
-                break;
-            default:
-                ASSERT(false);
-                break;
-            }
-            staging_data += quad_size;
-        }
-    }
-    scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([src_buffer = staging.buffer, src_offset = staging.offset,
-                      dst_buffer = *quad_array_lut, size_bytes](vk::CommandBuffer cmdbuf) {
-        const VkBufferCopy copy{
-            .srcOffset = src_offset,
-            .dstOffset = 0,
-            .size = size_bytes,
-        };
-        const VkBufferMemoryBarrier write_barrier{
-            .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_INDEX_READ_BIT,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .buffer = dst_buffer,
-            .offset = 0,
-            .size = size_bytes,
-        };
-        cmdbuf.CopyBuffer(src_buffer, dst_buffer, copy);
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT,
-                               0, write_barrier);
-    });
-}
-
 void BufferCacheRuntime::ReserveNullBuffer() {
     if (null_buffer) {
         return;

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -50,6 +50,9 @@ private:
     std::vector<BufferView> views;
 };
 
+class QuadArrayIndexBuffer;
+class QuadStripIndexBuffer;
+
 class BufferCacheRuntime {
     friend Buffer;
 
@@ -86,7 +89,7 @@ public:
     void BindIndexBuffer(PrimitiveTopology topology, IndexFormat index_format, u32 num_indices,
                          u32 base_vertex, VkBuffer buffer, u32 offset, u32 size);
 
-    void BindQuadArrayIndexBuffer(u32 first, u32 count);
+    void BindQuadIndexBuffer(PrimitiveTopology topology, u32 first, u32 count);
 
     void BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset, u32 size, u32 stride);
 
@@ -118,8 +121,6 @@ private:
         update_descriptor_queue.AddBuffer(buffer, offset, size);
     }
 
-    void ReserveQuadArrayLUT(u32 num_indices, bool wait_for_idle);
-
     void ReserveNullBuffer();
 
     const Device& device;
@@ -128,10 +129,8 @@ private:
     StagingBufferPool& staging_pool;
     UpdateDescriptorQueue& update_descriptor_queue;
 
-    vk::Buffer quad_array_lut;
-    MemoryCommit quad_array_lut_commit;
-    VkIndexType quad_array_lut_index_type{};
-    u32 current_num_indices = 0;
+    std::shared_ptr<QuadArrayIndexBuffer> quad_array_index_buffer;
+    std::shared_ptr<QuadStripIndexBuffer> quad_strip_index_buffer;
 
     vk::Buffer null_buffer;
     MemoryCommit null_buffer_commit;

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -245,7 +245,7 @@ QuadIndexedPass::QuadIndexedPass(const Device& device_, Scheduler& scheduler_,
                                  UpdateDescriptorQueue& update_descriptor_queue_)
     : ComputePass(device_, descriptor_pool_, INPUT_OUTPUT_DESCRIPTOR_SET_BINDINGS,
                   INPUT_OUTPUT_DESCRIPTOR_UPDATE_TEMPLATE, INPUT_OUTPUT_BANK_INFO,
-                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(u32) * 2>, VULKAN_QUAD_INDEXED_COMP_SPV),
+                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(u32) * 3>, VULKAN_QUAD_INDEXED_COMP_SPV),
       scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
       update_descriptor_queue{update_descriptor_queue_} {}
 
@@ -253,7 +253,7 @@ QuadIndexedPass::~QuadIndexedPass() = default;
 
 std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
     Tegra::Engines::Maxwell3D::Regs::IndexFormat index_format, u32 num_vertices, u32 base_vertex,
-    VkBuffer src_buffer, u32 src_offset) {
+    VkBuffer src_buffer, u32 src_offset, bool is_strip) {
     const u32 index_shift = [index_format] {
         switch (index_format) {
         case Tegra::Engines::Maxwell3D::Regs::IndexFormat::UnsignedByte:
@@ -267,7 +267,7 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
         return 2;
     }();
     const u32 input_size = num_vertices << index_shift;
-    const u32 num_tri_vertices = (num_vertices / 4) * 6;
+    const u32 num_tri_vertices = (is_strip ? (num_vertices - 2) / 2 : num_vertices / 4) * 6;
 
     const std::size_t staging_size = num_tri_vertices * sizeof(u32);
     const auto staging = staging_buffer_pool.Request(staging_size, MemoryUsage::DeviceLocal);
@@ -278,8 +278,8 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
     const void* const descriptor_data{update_descriptor_queue.UpdateData()};
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([this, descriptor_data, num_tri_vertices, base_vertex,
-                      index_shift](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([this, descriptor_data, num_tri_vertices, base_vertex, index_shift,
+                      is_strip](vk::CommandBuffer cmdbuf) {
         static constexpr u32 DISPATCH_SIZE = 1024;
         static constexpr VkMemoryBarrier WRITE_BARRIER{
             .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
@@ -287,7 +287,7 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
             .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
             .dstAccessMask = VK_ACCESS_INDEX_READ_BIT,
         };
-        const std::array<u32, 2> push_constants{base_vertex, index_shift};
+        const std::array<u32, 3> push_constants{base_vertex, index_shift, is_strip ? 1u : 0u};
         const VkDescriptorSet set = descriptor_allocator.Commit();
         device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
         cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);

src/video_core/renderer_vulkan/vk_compute_pass.h

@@ -74,7 +74,7 @@ public:
 
     std::pair<VkBuffer, VkDeviceSize> Assemble(
         Tegra::Engines::Maxwell3D::Regs::IndexFormat index_format, u32 num_vertices,
-        u32 base_vertex, VkBuffer src_buffer, u32 src_offset);
+        u32 base_vertex, VkBuffer src_buffer, u32 src_offset, bool is_strip);
 
 private:
     Scheduler& scheduler;

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -138,12 +138,16 @@ DrawParams MakeDrawParams(const MaxwellDrawState& draw_state, u32 num_instances,
         .first_index = is_indexed ? draw_state.index_buffer.first : 0,
         .is_indexed = is_indexed,
     };
+    // 6 triangle vertices per quad, base vertex is part of the index
+    // See BindQuadIndexBuffer for more details
     if (draw_state.topology == Maxwell::PrimitiveTopology::Quads) {
-        // 6 triangle vertices per quad, base vertex is part of the index
-        // See BindQuadArrayIndexBuffer for more details
         params.num_vertices = (params.num_vertices / 4) * 6;
         params.base_vertex = 0;
         params.is_indexed = true;
+    } else if (draw_state.topology == Maxwell::PrimitiveTopology::QuadStrip) {
+        params.num_vertices = (params.num_vertices - 2) / 2 * 6;
+        params.base_vertex = 0;
+        params.is_indexed = true;
     }
     return params;
 }

src/video_core/vulkan_common/vulkan_wrapper.cpp

@@ -314,6 +314,18 @@ const char* ToString(VkResult result) noexcept {
         return "VK_ERROR_VALIDATION_FAILED_EXT";
     case VkResult::VK_ERROR_INVALID_SHADER_NV:
         return "VK_ERROR_INVALID_SHADER_NV";
+    case VkResult::VK_ERROR_IMAGE_USAGE_NOT_SUPPORTED_KHR:
+        return "VK_ERROR_IMAGE_USAGE_NOT_SUPPORTED_KHR";
+    case VkResult::VK_ERROR_VIDEO_PICTURE_LAYOUT_NOT_SUPPORTED_KHR:
+        return "VK_ERROR_VIDEO_PICTURE_LAYOUT_NOT_SUPPORTED_KHR";
+    case VkResult::VK_ERROR_VIDEO_PROFILE_OPERATION_NOT_SUPPORTED_KHR:
+        return "VK_ERROR_VIDEO_PROFILE_OPERATION_NOT_SUPPORTED_KHR";
+    case VkResult::VK_ERROR_VIDEO_PROFILE_FORMAT_NOT_SUPPORTED_KHR:
+        return "VK_ERROR_VIDEO_PROFILE_FORMAT_NOT_SUPPORTED_KHR";
+    case VkResult::VK_ERROR_VIDEO_PROFILE_CODEC_NOT_SUPPORTED_KHR:
+        return "VK_ERROR_VIDEO_PROFILE_CODEC_NOT_SUPPORTED_KHR";
+    case VkResult::VK_ERROR_VIDEO_STD_VERSION_NOT_SUPPORTED_KHR:
+        return "VK_ERROR_VIDEO_STD_VERSION_NOT_SUPPORTED_KHR";
     case VkResult::VK_ERROR_INVALID_DRM_FORMAT_MODIFIER_PLANE_LAYOUT_EXT:
         return "VK_ERROR_INVALID_DRM_FORMAT_MODIFIER_PLANE_LAYOUT_EXT";
     case VkResult::VK_ERROR_FRAGMENTATION_EXT:

src/yuzu/bootmanager.cpp

@@ -46,30 +46,28 @@
 
 static Core::Frontend::WindowSystemType GetWindowSystemType();
 
-EmuThread::EmuThread(Core::System& system_) : system{system_} {}
+EmuThread::EmuThread(Core::System& system) : m_system{system} {}
 
 EmuThread::~EmuThread() = default;
 
 void EmuThread::run() {
-    std::string name = "EmuControlThread";
-    MicroProfileOnThreadCreate(name.c_str());
-    Common::SetCurrentThreadName(name.c_str());
+    const char* name = "EmuControlThread";
+    MicroProfileOnThreadCreate(name);
+    Common::SetCurrentThreadName(name);
 
-    auto& gpu = system.GPU();
-    auto stop_token = stop_source.get_token();
-    bool debugger_should_start = system.DebuggerEnabled();
+    auto& gpu = m_system.GPU();
+    auto stop_token = m_stop_source.get_token();
 
-    system.RegisterHostThread();
+    m_system.RegisterHostThread();
 
     // Main process has been loaded. Make the context current to this thread and begin GPU and CPU
     // execution.
     gpu.ObtainContext();
 
     emit LoadProgress(VideoCore::LoadCallbackStage::Prepare, 0, 0);
-
     if (Settings::values.use_disk_shader_cache.GetValue()) {
-        system.Renderer().ReadRasterizer()->LoadDiskResources(
-            system.GetCurrentProcessProgramID(), stop_token,
+        m_system.Renderer().ReadRasterizer()->LoadDiskResources(
+            m_system.GetCurrentProcessProgramID(), stop_token,
             [this](VideoCore::LoadCallbackStage stage, std::size_t value, std::size_t total) {
                 emit LoadProgress(stage, value, total);
             });
@@ -79,57 +77,34 @@ void EmuThread::run() {
     gpu.ReleaseContext();
     gpu.Start();
 
-    system.GetCpuManager().OnGpuReady();
+    m_system.GetCpuManager().OnGpuReady();
 
-    system.RegisterExitCallback([this]() {
-        stop_source.request_stop();
-        SetRunning(false);
-    });
+    if (m_system.DebuggerEnabled()) {
+        m_system.InitializeDebugger();
+    }
 
-    // Holds whether the cpu was running during the last iteration,
-    // so that the DebugModeLeft signal can be emitted before the
-    // next execution step
-    bool was_active = false;
     while (!stop_token.stop_requested()) {
-        if (running) {
-            if (was_active) {
-                emit DebugModeLeft();
-            }
+        std::unique_lock lk{m_should_run_mutex};
+        if (m_should_run) {
+            m_system.Run();
+            m_is_running.store(true);
+            m_is_running.notify_all();
 
-            running_guard = true;
-            Core::SystemResultStatus result = system.Run();
-            if (result != Core::SystemResultStatus::Success) {
-                running_guard = false;
-                this->SetRunning(false);
-                emit ErrorThrown(result, system.GetStatusDetails());
-            }
-
-            if (debugger_should_start) {
-                system.InitializeDebugger();
-                debugger_should_start = false;
-            }
-
-            running_wait.Wait();
-            result = system.Pause();
-            if (result != Core::SystemResultStatus::Success) {
-                running_guard = false;
-                this->SetRunning(false);
-                emit ErrorThrown(result, system.GetStatusDetails());
-            }
-            running_guard = false;
-
-            if (!stop_token.stop_requested()) {
-                was_active = true;
-                emit DebugModeEntered();
-            }
+            Common::CondvarWait(m_should_run_cv, lk, stop_token, [&] { return !m_should_run; });
         } else {
-            std::unique_lock lock{running_mutex};
-            Common::CondvarWait(running_cv, lock, stop_token, [&] { return IsRunning(); });
+            m_system.Pause();
+            m_is_running.store(false);
+            m_is_running.notify_all();
+
+            emit DebugModeEntered();
+            Common::CondvarWait(m_should_run_cv, lk, stop_token, [&] { return m_should_run; });
+            emit DebugModeLeft();
         }
     }
 
     // Shutdown the main emulated process
-    system.ShutdownMainProcess();
+    m_system.DetachDebugger();
+    m_system.ShutdownMainProcess();
 
 #if MICROPROFILE_ENABLED
     MicroProfileOnThreadExit();

src/yuzu/bootmanager.h

@@ -47,7 +47,7 @@ class EmuThread final : public QThread {
     Q_OBJECT
 
 public:
-    explicit EmuThread(Core::System& system_);
+    explicit EmuThread(Core::System& system);
     ~EmuThread() override;
 
     /**
@@ -57,30 +57,30 @@ public:
     void run() override;
 
     /**
-     * Sets whether the emulation thread is running or not
-     * @param running_ Boolean value, set the emulation thread to running if true
-     * @note This function is thread-safe
+     * Sets whether the emulation thread should run or not
+     * @param should_run Boolean value, set the emulation thread to running if true
      */
-    void SetRunning(bool running_) {
-        std::unique_lock lock{running_mutex};
-        running = running_;
-        lock.unlock();
-        running_cv.notify_all();
-        if (!running) {
-            running_wait.Set();
-            /// Wait until effectively paused
-            while (running_guard)
-                ;
+    void SetRunning(bool should_run) {
+        // TODO: Prevent other threads from modifying the state until we finish.
+        {
+            // Notify the running thread to change state.
+            std::unique_lock run_lk{m_should_run_mutex};
+            m_should_run = should_run;
+            m_should_run_cv.notify_one();
+        }
+
+        // Wait until paused, if pausing.
+        if (!should_run) {
+            m_is_running.wait(true);
         }
     }
 
     /**
      * Check if the emulation thread is running or not
      * @return True if the emulation thread is running, otherwise false
-     * @note This function is thread-safe
      */
     bool IsRunning() const {
-        return running;
+        return m_is_running.load() || m_should_run;
     }
 
     /**
@@ -88,18 +88,17 @@ public:
      */
     void ForceStop() {
         LOG_WARNING(Frontend, "Force stopping EmuThread");
-        stop_source.request_stop();
-        SetRunning(false);
+        m_stop_source.request_stop();
     }
 
 private:
-    bool running = false;
-    std::stop_source stop_source;
-    std::mutex running_mutex;
-    std::condition_variable_any running_cv;
-    Common::Event running_wait{};
-    std::atomic_bool running_guard{false};
-    Core::System& system;
+    Core::System& m_system;
+
+    std::stop_source m_stop_source;
+    std::mutex m_should_run_mutex;
+    std::condition_variable_any m_should_run_cv;
+    std::atomic<bool> m_is_running{false};
+    bool m_should_run{true};
 
 signals:
     /**
@@ -120,8 +119,6 @@ signals:
      */
     void DebugModeLeft();
 
-    void ErrorThrown(Core::SystemResultStatus, std::string);
-
     void LoadProgress(VideoCore::LoadCallbackStage stage, std::size_t value, std::size_t total);
 };
 

src/yuzu/main.cpp

@@ -1497,7 +1497,7 @@ void GMainWindow::SetupSigInterrupts() {
 
 void GMainWindow::HandleSigInterrupt(int sig) {
     if (sig == SIGINT) {
-        exit(1);
+        _exit(1);
     }
 
     // Calling into Qt directly from a signal handler is not safe,
@@ -1549,8 +1549,9 @@ void GMainWindow::AllowOSSleep() {
 
 bool GMainWindow::LoadROM(const QString& filename, u64 program_id, std::size_t program_index) {
     // Shutdown previous session if the emu thread is still active...
-    if (emu_thread != nullptr)
+    if (emu_thread != nullptr) {
         ShutdownGame();
+    }
 
     if (!render_window->InitRenderTarget()) {
         return false;
@@ -1709,6 +1710,11 @@ void GMainWindow::BootGame(const QString& filename, u64 program_id, std::size_t
     system->RegisterExecuteProgramCallback(
         [this](std::size_t program_index_) { render_window->ExecuteProgram(program_index_); });
 
+    system->RegisterExitCallback([this] {
+        emu_thread->ForceStop();
+        render_window->Exit();
+    });
+
     connect(render_window, &GRenderWindow::Closed, this, &GMainWindow::OnStopGame);
     connect(render_window, &GRenderWindow::MouseActivity, this, &GMainWindow::OnMouseActivity);
     // BlockingQueuedConnection is important here, it makes sure we've finished refreshing our views
@@ -1778,9 +1784,9 @@ void GMainWindow::BootGame(const QString& filename, u64 program_id, std::size_t
     OnStartGame();
 }
 
-void GMainWindow::ShutdownGame() {
+bool GMainWindow::OnShutdownBegin() {
     if (!emulation_running) {
-        return;
+        return false;
     }
 
     if (ui->action_Fullscreen->isChecked()) {
@@ -1792,21 +1798,55 @@ void GMainWindow::ShutdownGame() {
     // Disable unlimited frame rate
     Settings::values.use_speed_limit.SetValue(true);
 
+    if (system->IsShuttingDown()) {
+        return false;
+    }
+
     system->SetShuttingDown(true);
-    system->DetachDebugger();
     discord_rpc->Pause();
 
     RequestGameExit();
+    emu_thread->disconnect();
+    emu_thread->SetRunning(true);
 
     emit EmulationStopping();
 
-    // Wait for emulation thread to complete and delete it
-    if (!emu_thread->wait(5000)) {
+    shutdown_timer.setSingleShot(true);
+    shutdown_timer.start(system->DebuggerEnabled() ? 0 : 5000);
+    connect(&shutdown_timer, &QTimer::timeout, this, &GMainWindow::OnEmulationStopTimeExpired);
+    connect(emu_thread.get(), &QThread::finished, this, &GMainWindow::OnEmulationStopped);
+
+    // Disable everything to prevent anything from being triggered here
+    ui->action_Pause->setEnabled(false);
+    ui->action_Restart->setEnabled(false);
+    ui->action_Stop->setEnabled(false);
+
+    return true;
+}
+
+void GMainWindow::OnShutdownBeginDialog() {
+    shutdown_dialog = new OverlayDialog(this, *system, QString{}, tr("Closing software..."),
+                                        QString{}, QString{}, Qt::AlignHCenter | Qt::AlignVCenter);
+    shutdown_dialog->open();
+}
+
+void GMainWindow::OnEmulationStopTimeExpired() {
+    if (emu_thread) {
         emu_thread->ForceStop();
-        emu_thread->wait();
     }
+}
+
+void GMainWindow::OnEmulationStopped() {
+    shutdown_timer.stop();
+    emu_thread->disconnect();
+    emu_thread->wait();
     emu_thread = nullptr;
 
+    if (shutdown_dialog) {
+        shutdown_dialog->deleteLater();
+        shutdown_dialog = nullptr;
+    }
+
     emulation_running = false;
 
     discord_rpc->Update();
@@ -1852,6 +1892,20 @@ void GMainWindow::ShutdownGame() {
 
     // When closing the game, destroy the GLWindow to clear the context after the game is closed
     render_window->ReleaseRenderTarget();
+
+    Settings::RestoreGlobalState(system->IsPoweredOn());
+    system->HIDCore().ReloadInputDevices();
+    UpdateStatusButtons();
+}
+
+void GMainWindow::ShutdownGame() {
+    if (!emulation_running) {
+        return;
+    }
+
+    OnShutdownBegin();
+    OnEmulationStopTimeExpired();
+    OnEmulationStopped();
 }
 
 void GMainWindow::StoreRecentFile(const QString& filename) {
@@ -2918,8 +2972,6 @@ void GMainWindow::OnStartGame() {
 
     emu_thread->SetRunning(true);
 
-    connect(emu_thread.get(), &EmuThread::ErrorThrown, this, &GMainWindow::OnCoreError);
-
     UpdateMenuState();
     OnTasStateChanged();
 
@@ -2956,11 +3008,9 @@ void GMainWindow::OnStopGame() {
         return;
     }
 
-    ShutdownGame();
-
-    Settings::RestoreGlobalState(system->IsPoweredOn());
-    system->HIDCore().ReloadInputDevices();
-    UpdateStatusButtons();
+    if (OnShutdownBegin()) {
+        OnShutdownBeginDialog();
+    }
 }
 
 void GMainWindow::OnLoadComplete() {
@@ -3903,79 +3953,6 @@ void GMainWindow::OnMouseActivity() {
     mouse_center_timer.stop();
 }
 
-void GMainWindow::OnCoreError(Core::SystemResultStatus result, std::string details) {
-    QMessageBox::StandardButton answer;
-    QString status_message;
-    const QString common_message =
-        tr("The game you are trying to load requires additional files from your Switch to be "
-           "dumped "
-           "before playing.<br/><br/>For more information on dumping these files, please see the "
-           "following wiki page: <a "
-           "href='https://yuzu-emu.org/wiki/"
-           "dumping-system-archives-and-the-shared-fonts-from-a-switch-console/'>Dumping System "
-           "Archives and the Shared Fonts from a Switch Console</a>.<br/><br/>Would you like to "
-           "quit "
-           "back to the game list? Continuing emulation may result in crashes, corrupted save "
-           "data, or other bugs.");
-    switch (result) {
-    case Core::SystemResultStatus::ErrorSystemFiles: {
-        QString message;
-        if (details.empty()) {
-            message =
-                tr("yuzu was unable to locate a Switch system archive. %1").arg(common_message);
-        } else {
-            message = tr("yuzu was unable to locate a Switch system archive: %1. %2")
-                          .arg(QString::fromStdString(details), common_message);
-        }
-
-        answer = QMessageBox::question(this, tr("System Archive Not Found"), message,
-                                       QMessageBox::Yes | QMessageBox::No, QMessageBox::No);
-        status_message = tr("System Archive Missing");
-        break;
-    }
-
-    case Core::SystemResultStatus::ErrorSharedFont: {
-        const QString message =
-            tr("yuzu was unable to locate the Switch shared fonts. %1").arg(common_message);
-        answer = QMessageBox::question(this, tr("Shared Fonts Not Found"), message,
-                                       QMessageBox::Yes | QMessageBox::No, QMessageBox::No);
-        status_message = tr("Shared Font Missing");
-        break;
-    }
-
-    default:
-        answer = QMessageBox::question(
-            this, tr("Fatal Error"),
-            tr("yuzu has encountered a fatal error, please see the log for more details. "
-               "For more information on accessing the log, please see the following page: "
-               "<a href='https://community.citra-emu.org/t/how-to-upload-the-log-file/296'>How "
-               "to "
-               "Upload the Log File</a>.<br/><br/>Would you like to quit back to the game "
-               "list? "
-               "Continuing emulation may result in crashes, corrupted save data, or other "
-               "bugs."),
-            QMessageBox::Yes | QMessageBox::No, QMessageBox::No);
-        status_message = tr("Fatal Error encountered");
-        break;
-    }
-
-    if (answer == QMessageBox::Yes) {
-        if (emu_thread) {
-            ShutdownGame();
-
-            Settings::RestoreGlobalState(system->IsPoweredOn());
-            system->HIDCore().ReloadInputDevices();
-            UpdateStatusButtons();
-        }
-    } else {
-        // Only show the message if the game is still running.
-        if (emu_thread) {
-            emu_thread->SetRunning(true);
-            message_label->setText(status_message);
-        }
-    }
-}
-
 void GMainWindow::OnReinitializeKeys(ReinitializeKeyBehavior behavior) {
     if (behavior == ReinitializeKeyBehavior::Warning) {
         const auto res = QMessageBox::information(
@@ -4120,10 +4097,6 @@ void GMainWindow::closeEvent(QCloseEvent* event) {
     // Shutdown session if the emu thread is active...
     if (emu_thread != nullptr) {
         ShutdownGame();
-
-        Settings::RestoreGlobalState(system->IsPoweredOn());
-        system->HIDCore().ReloadInputDevices();
-        UpdateStatusButtons();
     }
 
     render_window->close();
@@ -4216,6 +4189,10 @@ bool GMainWindow::ConfirmForceLockedExit() {
 }
 
 void GMainWindow::RequestGameExit() {
+    if (!system->IsPoweredOn()) {
+        return;
+    }
+
     auto& sm{system->ServiceManager()};
     auto applet_oe = sm.GetService<Service::AM::AppletOE>("appletOE");
     auto applet_ae = sm.GetService<Service::AM::AppletAE>("appletAE");

src/yuzu/main.h

@@ -29,6 +29,7 @@ class GImageInfo;
 class GRenderWindow;
 class LoadingScreen;
 class MicroProfileDialog;
+class OverlayDialog;
 class ProfilerWidget;
 class ControllerDialog;
 class QLabel;
@@ -332,10 +333,13 @@ private slots:
     void ResetWindowSize900();
     void ResetWindowSize1080();
     void OnCaptureScreenshot();
-    void OnCoreError(Core::SystemResultStatus, std::string);
     void OnReinitializeKeys(ReinitializeKeyBehavior behavior);
     void OnLanguageChanged(const QString& locale);
     void OnMouseActivity();
+    bool OnShutdownBegin();
+    void OnShutdownBeginDialog();
+    void OnEmulationStopped();
+    void OnEmulationStopTimeExpired();
 
 private:
     QString GetGameListErrorRemoving(InstalledEntryType type) const;
@@ -385,6 +389,8 @@ private:
     GRenderWindow* render_window;
     GameList* game_list;
     LoadingScreen* loading_screen;
+    QTimer shutdown_timer;
+    OverlayDialog* shutdown_dialog{};
 
     GameListPlaceholder* game_list_placeholder;
 

src/yuzu/util/overlay_dialog.cpp

@@ -3,6 +3,7 @@
 
 #include <QKeyEvent>
 #include <QScreen>
+#include <QWindow>
 
 #include "core/core.h"
 #include "core/hid/hid_types.h"
@@ -42,7 +43,7 @@ OverlayDialog::OverlayDialog(QWidget* parent, Core::System& system, const QStrin
     MoveAndResizeWindow();
 
     // TODO (Morph): Remove this when InputInterpreter no longer relies on the HID backend
-    if (system.IsPoweredOn()) {
+    if (system.IsPoweredOn() && !ui->buttonsDialog->isHidden()) {
         input_interpreter = std::make_unique<InputInterpreter>(system);
 
         StartInputThread();
@@ -83,6 +84,11 @@ void OverlayDialog::InitializeRegularTextDialog(const QString& title_text, const
         ui->button_ok_label->setEnabled(false);
     }
 
+    if (ui->button_cancel->isHidden() && ui->button_ok_label->isHidden()) {
+        ui->buttonsDialog->hide();
+        return;
+    }
+
     connect(
         ui->button_cancel, &QPushButton::clicked, this,
         [this](bool) {
@@ -130,6 +136,11 @@ void OverlayDialog::InitializeRichTextDialog(const QString& title_text, const QS
         ui->button_ok_rich->setEnabled(false);
     }
 
+    if (ui->button_cancel_rich->isHidden() && ui->button_ok_rich->isHidden()) {
+        ui->buttonsRichDialog->hide();
+        return;
+    }
+
     connect(
         ui->button_cancel_rich, &QPushButton::clicked, this,
         [this](bool) {
@@ -152,7 +163,7 @@ void OverlayDialog::MoveAndResizeWindow() {
     const auto height = static_cast<float>(parentWidget()->height());
 
     // High DPI
-    const float dpi_scale = qApp->screenAt(pos)->logicalDotsPerInch() / 96.0f;
+    const float dpi_scale = parentWidget()->windowHandle()->screen()->logicalDotsPerInch() / 96.0f;
 
     const auto title_text_font_size = BASE_TITLE_FONT_SIZE * (height / BASE_HEIGHT) / dpi_scale;
     const auto body_text_font_size =
@@ -249,3 +260,9 @@ void OverlayDialog::InputThread() {
         std::this_thread::sleep_for(std::chrono::milliseconds(50));
     }
 }
+
+void OverlayDialog::keyPressEvent(QKeyEvent* e) {
+    if (!ui->buttonsDialog->isHidden() || e->key() != Qt::Key_Escape) {
+        QDialog::keyPressEvent(e);
+    }
+}

src/yuzu/util/overlay_dialog.h

@@ -94,6 +94,7 @@ private:
 
     /// The thread where input is being polled and processed.
     void InputThread();
+    void keyPressEvent(QKeyEvent* e) override;
 
     std::unique_ptr<Ui::OverlayDialog> ui;
 

src/yuzu_cmd/CMakeLists.txt

@@ -49,6 +49,15 @@ if(UNIX AND NOT APPLE)
     install(TARGETS yuzu-cmd)
 endif()
 
+if(WIN32)
+    # compile as a win32 gui application instead of a console application
+    if(MSVC)
+        set_target_properties(yuzu-cmd PROPERTIES LINK_FLAGS_RELEASE "/SUBSYSTEM:WINDOWS /ENTRY:mainCRTStartup")
+    elseif(MINGW)
+        set_target_properties(yuzu-cmd PROPERTIES LINK_FLAGS_RELEASE "-Wl,--subsystem,windows")
+    endif()
+endif()
+
 if (MSVC)
     include(CopyYuzuSDLDeps)
     copy_yuzu_SDL_deps(yuzu-cmd)

src/yuzu_cmd/yuzu.cpp

@@ -174,6 +174,13 @@ static void OnStatusMessageReceived(const Network::StatusMessageEntry& msg) {
 
 /// Application entry point
 int main(int argc, char** argv) {
+#ifdef _WIN32
+    if (AttachConsole(ATTACH_PARENT_PROCESS)) {
+        freopen("CONOUT$", "wb", stdout);
+        freopen("CONOUT$", "wb", stderr);
+    }
+#endif
+
     Common::Log::Initialize();
     Common::Log::SetColorConsoleBackendEnabled(true);
     Common::Log::Start();