vulkan_device: Enable VK_KHR_swapchain_mutable_format if available

Silences validation errors when creating sRGB image views of linear swapchain images

src/common/logging/backend.cpp

@@ -18,7 +18,6 @@
 #include "common/fs/fs_paths.h"
 #include "common/fs/path_util.h"
 #include "common/literals.h"
-#include "common/thread.h"
 
 #include "common/logging/backend.h"
 #include "common/logging/log.h"

src/common/logging/filter.cpp

@@ -111,7 +111,6 @@ bool ParseFilterRule(Filter& instance, Iterator begin, Iterator end) {
     SUB(Service, NCM)                                                                              \
     SUB(Service, NFC)                                                                              \
     SUB(Service, NFP)                                                                              \
-    SUB(Service, NGCT)                                                                             \
     SUB(Service, NIFM)                                                                             \
     SUB(Service, NIM)                                                                              \
     SUB(Service, NPNS)                                                                             \

src/common/logging/types.h

@@ -81,7 +81,6 @@ enum class Class : u8 {
     Service_NCM,       ///< The NCM service
     Service_NFC,       ///< The NFC (Near-field communication) service
     Service_NFP,       ///< The NFP service
-    Service_NGCT,      ///< The NGCT (No Good Content for Terra) service
     Service_NIFM,      ///< The NIFM (Network interface) service
     Service_NIM,       ///< The NIM service
     Service_NPNS,      ///< The NPNS service

src/common/lru_cache.h

@@ -1,140 +0,0 @@
-// Copyright 2021 yuzu Emulator Project
-// Licensed under GPLv2+ or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <deque>
-#include <memory>
-#include <type_traits>
-
-#include "common/common_types.h"
-
-namespace Common {
-
-template <class Traits>
-class LeastRecentlyUsedCache {
-    using ObjectType = typename Traits::ObjectType;
-    using TickType = typename Traits::TickType;
-
-    struct Item {
-        ObjectType obj;
-        TickType tick;
-        Item* next{};
-        Item* prev{};
-    };
-
-public:
-    LeastRecentlyUsedCache() : first_item{}, last_item{} {}
-    ~LeastRecentlyUsedCache() = default;
-
-    size_t Insert(ObjectType obj, TickType tick) {
-        const auto new_id = Build();
-        auto& item = item_pool[new_id];
-        item.obj = obj;
-        item.tick = tick;
-        Attach(item);
-        return new_id;
-    }
-
-    void Touch(size_t id, TickType tick) {
-        auto& item = item_pool[id];
-        if (item.tick >= tick) {
-            return;
-        }
-        item.tick = tick;
-        if (&item == last_item) {
-            return;
-        }
-        Detach(item);
-        Attach(item);
-    }
-
-    void Free(size_t id) {
-        auto& item = item_pool[id];
-        Detach(item);
-        item.prev = nullptr;
-        item.next = nullptr;
-        free_items.push_back(id);
-    }
-
-    template <typename Func>
-    void ForEachItemBelow(TickType tick, Func&& func) {
-        static constexpr bool RETURNS_BOOL =
-            std::is_same_v<std::invoke_result<Func, ObjectType>, bool>;
-        Item* iterator = first_item;
-        while (iterator) {
-            if (static_cast<s64>(tick) - static_cast<s64>(iterator->tick) < 0) {
-                return;
-            }
-            Item* next = iterator->next;
-            if constexpr (RETURNS_BOOL) {
-                if (func(iterator->obj)) {
-                    return;
-                }
-            } else {
-                func(iterator->obj);
-            }
-            iterator = next;
-        }
-    }
-
-private:
-    size_t Build() {
-        if (free_items.empty()) {
-            const size_t item_id = item_pool.size();
-            auto& item = item_pool.emplace_back();
-            item.next = nullptr;
-            item.prev = nullptr;
-            return item_id;
-        }
-        const size_t item_id = free_items.front();
-        free_items.pop_front();
-        auto& item = item_pool[item_id];
-        item.next = nullptr;
-        item.prev = nullptr;
-        return item_id;
-    }
-
-    void Attach(Item& item) {
-        if (!first_item) {
-            first_item = &item;
-        }
-        if (!last_item) {
-            last_item = &item;
-        } else {
-            item.prev = last_item;
-            last_item->next = &item;
-            item.next = nullptr;
-            last_item = &item;
-        }
-    }
-
-    void Detach(Item& item) {
-        if (item.prev) {
-            item.prev->next = item.next;
-        }
-        if (item.next) {
-            item.next->prev = item.prev;
-        }
-        if (&item == first_item) {
-            first_item = item.next;
-            if (first_item) {
-                first_item->prev = nullptr;
-            }
-        }
-        if (&item == last_item) {
-            last_item = item.prev;
-            if (last_item) {
-                last_item->next = nullptr;
-            }
-        }
-    }
-
-    std::deque<Item> item_pool;
-    std::deque<size_t> free_items;
-    Item* first_item{};
-    Item* last_item{};
-};
-
-} // namespace Common

src/common/settings.cpp

@@ -59,6 +59,7 @@ void LogSettings() {
     log_setting("Renderer_UseVsync", values.use_vsync.GetValue());
     log_setting("Renderer_ShaderBackend", values.shader_backend.GetValue());
     log_setting("Renderer_UseAsynchronousShaders", values.use_asynchronous_shaders.GetValue());
+    log_setting("Renderer_UseGarbageCollection", values.use_caches_gc.GetValue());
     log_setting("Renderer_AnisotropicFilteringLevel", values.max_anisotropy.GetValue());
     log_setting("Audio_OutputEngine", values.sink_id.GetValue());
     log_setting("Audio_EnableAudioStretching", values.enable_audio_stretching.GetValue());
@@ -142,6 +143,7 @@ void RestoreGlobalState(bool is_powered_on) {
     values.shader_backend.SetGlobal(true);
     values.use_asynchronous_shaders.SetGlobal(true);
     values.use_fast_gpu_time.SetGlobal(true);
+    values.use_caches_gc.SetGlobal(true);
     values.bg_red.SetGlobal(true);
     values.bg_green.SetGlobal(true);
     values.bg_blue.SetGlobal(true);

src/common/settings.h

@@ -475,6 +475,7 @@ struct Values {
                                                 ShaderBackend::SPIRV, "shader_backend"};
     Setting<bool> use_asynchronous_shaders{false, "use_asynchronous_shaders"};
     Setting<bool> use_fast_gpu_time{true, "use_fast_gpu_time"};
+    Setting<bool> use_caches_gc{false, "use_caches_gc"};
 
     Setting<u8> bg_red{0, "bg_red"};
     Setting<u8> bg_green{0, "bg_green"};

src/core/CMakeLists.txt

@@ -452,8 +452,6 @@ add_library(core STATIC
     hle/service/nfp/nfp.h
     hle/service/nfp/nfp_user.cpp
     hle/service/nfp/nfp_user.h
-    hle/service/ngct/ngct.cpp
-    hle/service/ngct/ngct.h
     hle/service/nifm/nifm.cpp
     hle/service/nifm/nifm.h
     hle/service/nim/nim.cpp

src/core/core.cpp

@@ -507,12 +507,6 @@ const ARM_Interface& System::CurrentArmInterface() const {
     return impl->kernel.CurrentPhysicalCore().ArmInterface();
 }
 
-std::size_t System::CurrentCoreIndex() const {
-    std::size_t core = impl->kernel.GetCurrentHostThreadID();
-    ASSERT(core < Core::Hardware::NUM_CPU_CORES);
-    return core;
-}
-
 Kernel::PhysicalCore& System::CurrentPhysicalCore() {
     return impl->kernel.CurrentPhysicalCore();
 }

src/core/core.h

@@ -205,9 +205,6 @@ public:
     /// Gets an ARM interface to the CPU core that is currently running
     [[nodiscard]] const ARM_Interface& CurrentArmInterface() const;
 
-    /// Gets the index of the currently running CPU core
-    [[nodiscard]] std::size_t CurrentCoreIndex() const;
-
     /// Gets the physical core for the CPU core that is currently running
     [[nodiscard]] Kernel::PhysicalCore& CurrentPhysicalCore();
 

src/core/cpu_manager.cpp

@@ -21,34 +21,25 @@ namespace Core {
 CpuManager::CpuManager(System& system_) : system{system_} {}
 CpuManager::~CpuManager() = default;
 
-void CpuManager::ThreadStart(CpuManager& cpu_manager, std::size_t core) {
-    cpu_manager.RunThread(core);
+void CpuManager::ThreadStart(std::stop_token stop_token, CpuManager& cpu_manager,
+                             std::size_t core) {
+    cpu_manager.RunThread(stop_token, core);
 }
 
 void CpuManager::Initialize() {
     running_mode = true;
     if (is_multicore) {
         for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-            core_data[core].host_thread =
-                std::make_unique<std::thread>(ThreadStart, std::ref(*this), core);
+            core_data[core].host_thread = std::jthread(ThreadStart, std::ref(*this), core);
         }
     } else {
-        core_data[0].host_thread = std::make_unique<std::thread>(ThreadStart, std::ref(*this), 0);
+        core_data[0].host_thread = std::jthread(ThreadStart, std::ref(*this), 0);
     }
 }
 
 void CpuManager::Shutdown() {
     running_mode = false;
     Pause(false);
-    if (is_multicore) {
-        for (auto& data : core_data) {
-            data.host_thread->join();
-            data.host_thread.reset();
-        }
-    } else {
-        core_data[0].host_thread->join();
-        core_data[0].host_thread.reset();
-    }
 }
 
 std::function<void(void*)> CpuManager::GetGuestThreadStartFunc() {
@@ -127,17 +118,18 @@ void CpuManager::MultiCoreRunGuestLoop() {
             physical_core = &kernel.CurrentPhysicalCore();
         }
         system.ExitDynarmicProfile();
-        physical_core->ArmInterface().ClearExclusiveState();
-        kernel.CurrentScheduler()->RescheduleCurrentCore();
+        {
+            Kernel::KScopedDisableDispatch dd(kernel);
+            physical_core->ArmInterface().ClearExclusiveState();
+        }
     }
 }
 
 void CpuManager::MultiCoreRunIdleThread() {
     auto& kernel = system.Kernel();
     while (true) {
-        auto& physical_core = kernel.CurrentPhysicalCore();
-        physical_core.Idle();
-        kernel.CurrentScheduler()->RescheduleCurrentCore();
+        Kernel::KScopedDisableDispatch dd(kernel);
+        kernel.CurrentPhysicalCore().Idle();
     }
 }
 
@@ -145,12 +137,12 @@ void CpuManager::MultiCoreRunSuspendThread() {
     auto& kernel = system.Kernel();
     kernel.CurrentScheduler()->OnThreadStart();
     while (true) {
-        auto core = kernel.GetCurrentHostThreadID();
+        auto core = kernel.CurrentPhysicalCoreIndex();
         auto& scheduler = *kernel.CurrentScheduler();
         Kernel::KThread* current_thread = scheduler.GetCurrentThread();
         Common::Fiber::YieldTo(current_thread->GetHostContext(), *core_data[core].host_context);
         ASSERT(scheduler.ContextSwitchPending());
-        ASSERT(core == kernel.GetCurrentHostThreadID());
+        ASSERT(core == kernel.CurrentPhysicalCoreIndex());
         scheduler.RescheduleCurrentCore();
     }
 }
@@ -317,7 +309,7 @@ void CpuManager::Pause(bool paused) {
     }
 }
 
-void CpuManager::RunThread(std::size_t core) {
+void CpuManager::RunThread(std::stop_token stop_token, std::size_t core) {
     /// Initialization
     system.RegisterCoreThread(core);
     std::string name;
@@ -356,8 +348,8 @@ void CpuManager::RunThread(std::size_t core) {
             sc_sync_first_use = false;
         }
 
-        // Abort if emulation was killed before the session really starts
-        if (!system.IsPoweredOn()) {
+        // Emulation was stopped
+        if (stop_token.stop_requested()) {
             return;
         }
 

src/core/cpu_manager.h

@@ -78,9 +78,9 @@ private:
     void SingleCoreRunSuspendThread();
     void SingleCorePause(bool paused);
 
-    static void ThreadStart(CpuManager& cpu_manager, std::size_t core);
+    static void ThreadStart(std::stop_token stop_token, CpuManager& cpu_manager, std::size_t core);
 
-    void RunThread(std::size_t core);
+    void RunThread(std::stop_token stop_token, std::size_t core);
 
     struct CoreData {
         std::shared_ptr<Common::Fiber> host_context;
@@ -89,7 +89,7 @@ private:
         std::atomic<bool> is_running;
         std::atomic<bool> is_paused;
         std::atomic<bool> initialized;
-        std::unique_ptr<std::thread> host_thread;
+        std::jthread host_thread;
     };
 
     std::atomic<bool> running_mode{};

src/core/hle/kernel/k_address_arbiter.cpp

@@ -28,7 +28,7 @@ bool ReadFromUser(Core::System& system, s32* out, VAddr address) {
 
 bool DecrementIfLessThan(Core::System& system, s32* out, VAddr address, s32 value) {
     auto& monitor = system.Monitor();
-    const auto current_core = system.CurrentCoreIndex();
+    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
 
     // TODO(bunnei): We should disable interrupts here via KScopedInterruptDisable.
     // TODO(bunnei): We should call CanAccessAtomic(..) here.
@@ -58,7 +58,7 @@ bool DecrementIfLessThan(Core::System& system, s32* out, VAddr address, s32 valu
 
 bool UpdateIfEqual(Core::System& system, s32* out, VAddr address, s32 value, s32 new_value) {
     auto& monitor = system.Monitor();
-    const auto current_core = system.CurrentCoreIndex();
+    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
 
     // TODO(bunnei): We should disable interrupts here via KScopedInterruptDisable.
     // TODO(bunnei): We should call CanAccessAtomic(..) here.

src/core/hle/kernel/k_auto_object.h

@@ -170,6 +170,10 @@ public:
         }
     }
 
+    const std::string& GetName() const {
+        return name;
+    }
+
 private:
     void RegisterWithKernel();
     void UnregisterWithKernel();

src/core/hle/kernel/k_condition_variable.cpp

@@ -35,7 +35,7 @@ bool WriteToUser(Core::System& system, VAddr address, const u32* p) {
 bool UpdateLockAtomic(Core::System& system, u32* out, VAddr address, u32 if_zero,
                       u32 new_orr_mask) {
     auto& monitor = system.Monitor();
-    const auto current_core = system.CurrentCoreIndex();
+    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
 
     // Load the value from the address.
     const auto expected = monitor.ExclusiveRead32(current_core, address);

src/core/hle/kernel/k_handle_table.cpp

@@ -13,6 +13,7 @@ ResultCode KHandleTable::Finalize() {
     // Get the table and clear our record of it.
     u16 saved_table_size = 0;
     {
+        KScopedDisableDispatch dd(kernel);
         KScopedSpinLock lk(m_lock);
 
         std::swap(m_table_size, saved_table_size);
@@ -43,6 +44,7 @@ bool KHandleTable::Remove(Handle handle) {
     // Find the object and free the entry.
     KAutoObject* obj = nullptr;
     {
+        KScopedDisableDispatch dd(kernel);
         KScopedSpinLock lk(m_lock);
 
         if (this->IsValidHandle(handle)) {
@@ -61,6 +63,7 @@ bool KHandleTable::Remove(Handle handle) {
 }
 
 ResultCode KHandleTable::Add(Handle* out_handle, KAutoObject* obj, u16 type) {
+    KScopedDisableDispatch dd(kernel);
     KScopedSpinLock lk(m_lock);
 
     // Never exceed our capacity.
@@ -83,6 +86,7 @@ ResultCode KHandleTable::Add(Handle* out_handle, KAutoObject* obj, u16 type) {
 }
 
 ResultCode KHandleTable::Reserve(Handle* out_handle) {
+    KScopedDisableDispatch dd(kernel);
     KScopedSpinLock lk(m_lock);
 
     // Never exceed our capacity.
@@ -93,6 +97,7 @@ ResultCode KHandleTable::Reserve(Handle* out_handle) {
 }
 
 void KHandleTable::Unreserve(Handle handle) {
+    KScopedDisableDispatch dd(kernel);
     KScopedSpinLock lk(m_lock);
 
     // Unpack the handle.
@@ -111,6 +116,7 @@ void KHandleTable::Unreserve(Handle handle) {
 }
 
 void KHandleTable::Register(Handle handle, KAutoObject* obj, u16 type) {
+    KScopedDisableDispatch dd(kernel);
     KScopedSpinLock lk(m_lock);
 
     // Unpack the handle.

src/core/hle/kernel/k_handle_table.h

@@ -69,6 +69,7 @@ public:
     template <typename T = KAutoObject>
     KScopedAutoObject<T> GetObjectWithoutPseudoHandle(Handle handle) const {
         // Lock and look up in table.
+        KScopedDisableDispatch dd(kernel);
         KScopedSpinLock lk(m_lock);
 
         if constexpr (std::is_same_v<T, KAutoObject>) {
@@ -123,6 +124,7 @@ public:
         size_t num_opened;
         {
             // Lock the table.
+            KScopedDisableDispatch dd(kernel);
             KScopedSpinLock lk(m_lock);
             for (num_opened = 0; num_opened < num_handles; num_opened++) {
                 // Get the current handle.

src/core/hle/kernel/k_process.cpp

@@ -59,6 +59,7 @@ void SetupMainThread(Core::System& system, KProcess& owner_process, u32 priority
     thread->GetContext64().cpu_registers[0] = 0;
     thread->GetContext32().cpu_registers[1] = thread_handle;
     thread->GetContext64().cpu_registers[1] = thread_handle;
+    thread->DisableDispatch();
 
     auto& kernel = system.Kernel();
     // Threads by default are dormant, wake up the main thread so it runs when the scheduler fires

src/core/hle/kernel/k_scheduler.cpp

@@ -376,20 +376,18 @@ void KScheduler::ClearSchedulerUpdateNeeded(KernelCore& kernel) {
 }
 
 void KScheduler::DisableScheduling(KernelCore& kernel) {
-    if (auto* scheduler = kernel.CurrentScheduler(); scheduler) {
-        ASSERT(scheduler->GetCurrentThread()->GetDisableDispatchCount() >= 0);
-        scheduler->GetCurrentThread()->DisableDispatch();
-    }
+    ASSERT(GetCurrentThreadPointer(kernel)->GetDisableDispatchCount() >= 0);
+    GetCurrentThreadPointer(kernel)->DisableDispatch();
 }
 
 void KScheduler::EnableScheduling(KernelCore& kernel, u64 cores_needing_scheduling) {
-    if (auto* scheduler = kernel.CurrentScheduler(); scheduler) {
-        ASSERT(scheduler->GetCurrentThread()->GetDisableDispatchCount() >= 1);
-        if (scheduler->GetCurrentThread()->GetDisableDispatchCount() >= 1) {
-            scheduler->GetCurrentThread()->EnableDispatch();
-        }
+    ASSERT(GetCurrentThreadPointer(kernel)->GetDisableDispatchCount() >= 1);
+
+    if (GetCurrentThreadPointer(kernel)->GetDisableDispatchCount() > 1) {
+        GetCurrentThreadPointer(kernel)->EnableDispatch();
+    } else {
+        RescheduleCores(kernel, cores_needing_scheduling);
     }
-    RescheduleCores(kernel, cores_needing_scheduling);
 }
 
 u64 KScheduler::UpdateHighestPriorityThreads(KernelCore& kernel) {
@@ -617,13 +615,17 @@ KScheduler::KScheduler(Core::System& system_, s32 core_id_) : system{system_}, c
     state.highest_priority_thread = nullptr;
 }
 
-KScheduler::~KScheduler() {
+void KScheduler::Finalize() {
     if (idle_thread) {
         idle_thread->Close();
         idle_thread = nullptr;
     }
 }
 
+KScheduler::~KScheduler() {
+    ASSERT(!idle_thread);
+}
+
 KThread* KScheduler::GetCurrentThread() const {
     if (auto result = current_thread.load(); result) {
         return result;
@@ -642,10 +644,12 @@ void KScheduler::RescheduleCurrentCore() {
     if (phys_core.IsInterrupted()) {
         phys_core.ClearInterrupt();
     }
+
     guard.Lock();
     if (state.needs_scheduling.load()) {
         Schedule();
     } else {
+        GetCurrentThread()->EnableDispatch();
         guard.Unlock();
     }
 }
@@ -655,26 +659,33 @@ void KScheduler::OnThreadStart() {
 }
 
 void KScheduler::Unload(KThread* thread) {
+    ASSERT(thread);
+
     LOG_TRACE(Kernel, "core {}, unload thread {}", core_id, thread ? thread->GetName() : "nullptr");
 
-    if (thread) {
-        if (thread->IsCallingSvc()) {
-            thread->ClearIsCallingSvc();
-        }
-        if (!thread->IsTerminationRequested()) {
-            prev_thread = thread;
-
-            Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
-            cpu_core.SaveContext(thread->GetContext32());
-            cpu_core.SaveContext(thread->GetContext64());
-            // Save the TPIDR_EL0 system register in case it was modified.
-            thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
-            cpu_core.ClearExclusiveState();
-        } else {
-            prev_thread = nullptr;
-        }
-        thread->context_guard.Unlock();
+    if (thread->IsCallingSvc()) {
+        thread->ClearIsCallingSvc();
     }
+
+    auto& physical_core = system.Kernel().PhysicalCore(core_id);
+    if (!physical_core.IsInitialized()) {
+        return;
+    }
+
+    Core::ARM_Interface& cpu_core = physical_core.ArmInterface();
+    cpu_core.SaveContext(thread->GetContext32());
+    cpu_core.SaveContext(thread->GetContext64());
+    // Save the TPIDR_EL0 system register in case it was modified.
+    thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+    cpu_core.ClearExclusiveState();
+
+    if (!thread->IsTerminationRequested() && thread->GetActiveCore() == core_id) {
+        prev_thread = thread;
+    } else {
+        prev_thread = nullptr;
+    }
+
+    thread->context_guard.Unlock();
 }
 
 void KScheduler::Reload(KThread* thread) {
@@ -683,11 +694,6 @@ void KScheduler::Reload(KThread* thread) {
     if (thread) {
         ASSERT_MSG(thread->GetState() == ThreadState::Runnable, "Thread must be runnable.");
 
-        auto* const thread_owner_process = thread->GetOwnerProcess();
-        if (thread_owner_process != nullptr) {
-            system.Kernel().MakeCurrentProcess(thread_owner_process);
-        }
-
         Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
         cpu_core.LoadContext(thread->GetContext32());
         cpu_core.LoadContext(thread->GetContext64());
@@ -705,7 +711,7 @@ void KScheduler::SwitchContextStep2() {
 }
 
 void KScheduler::ScheduleImpl() {
-    KThread* previous_thread = current_thread.load();
+    KThread* previous_thread = GetCurrentThread();
     KThread* next_thread = state.highest_priority_thread;
 
     state.needs_scheduling = false;
@@ -717,10 +723,15 @@ void KScheduler::ScheduleImpl() {
 
     // If we're not actually switching thread, there's nothing to do.
     if (next_thread == current_thread.load()) {
+        previous_thread->EnableDispatch();
         guard.Unlock();
         return;
     }
 
+    if (next_thread->GetCurrentCore() != core_id) {
+        next_thread->SetCurrentCore(core_id);
+    }
+
     current_thread.store(next_thread);
 
     KProcess* const previous_process = system.Kernel().CurrentProcess();
@@ -731,11 +742,7 @@ void KScheduler::ScheduleImpl() {
     Unload(previous_thread);
 
     std::shared_ptr<Common::Fiber>* old_context;
-    if (previous_thread != nullptr) {
-        old_context = &previous_thread->GetHostContext();
-    } else {
-        old_context = &idle_thread->GetHostContext();
-    }
+    old_context = &previous_thread->GetHostContext();
     guard.Unlock();
 
     Common::Fiber::YieldTo(*old_context, *switch_fiber);

src/core/hle/kernel/k_scheduler.h

@@ -33,6 +33,8 @@ public:
     explicit KScheduler(Core::System& system_, s32 core_id_);
     ~KScheduler();
 
+    void Finalize();
+
     /// Reschedules to the next available thread (call after current thread is suspended)
     void RescheduleCurrentCore();
 

src/core/hle/kernel/k_thread.cpp

@@ -14,6 +14,7 @@
 #include "common/fiber.h"
 #include "common/logging/log.h"
 #include "common/scope_exit.h"
+#include "common/settings.h"
 #include "common/thread_queue_list.h"
 #include "core/core.h"
 #include "core/cpu_manager.h"
@@ -188,7 +189,7 @@ ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_s
     // Setup the stack parameters.
     StackParameters& sp = GetStackParameters();
     sp.cur_thread = this;
-    sp.disable_count = 1;
+    sp.disable_count = 0;
     SetInExceptionHandler();
 
     // Set thread ID.
@@ -215,9 +216,10 @@ ResultCode KThread::InitializeThread(KThread* thread, KThreadFunction func, uint
     // Initialize the thread.
     R_TRY(thread->Initialize(func, arg, user_stack_top, prio, core, owner, type));
 
-    // Initialize host context.
+    // Initialize emulation parameters.
     thread->host_context =
         std::make_shared<Common::Fiber>(std::move(init_func), init_func_parameter);
+    thread->is_single_core = !Settings::values.use_multi_core.GetValue();
 
     return ResultSuccess;
 }
@@ -970,6 +972,9 @@ ResultCode KThread::Run() {
 
         // Set our state and finish.
         SetState(ThreadState::Runnable);
+
+        DisableDispatch();
+
         return ResultSuccess;
     }
 }
@@ -1054,4 +1059,16 @@ s32 GetCurrentCoreId(KernelCore& kernel) {
     return GetCurrentThread(kernel).GetCurrentCore();
 }
 
+KScopedDisableDispatch::~KScopedDisableDispatch() {
+    if (GetCurrentThread(kernel).GetDisableDispatchCount() <= 1) {
+        auto scheduler = kernel.CurrentScheduler();
+
+        if (scheduler) {
+            scheduler->RescheduleCurrentCore();
+        }
+    } else {
+        GetCurrentThread(kernel).EnableDispatch();
+    }
+}
+
 } // namespace Kernel

src/core/hle/kernel/k_thread.h

@@ -450,16 +450,39 @@ public:
         sleeping_queue = q;
     }
 
+    [[nodiscard]] bool IsKernelThread() const {
+        return GetActiveCore() == 3;
+    }
+
+    [[nodiscard]] bool IsDispatchTrackingDisabled() const {
+        return is_single_core || IsKernelThread();
+    }
+
     [[nodiscard]] s32 GetDisableDispatchCount() const {
+        if (IsDispatchTrackingDisabled()) {
+            // TODO(bunnei): Until kernel threads are emulated, we cannot enable/disable dispatch.
+            return 1;
+        }
+
         return this->GetStackParameters().disable_count;
     }
 
     void DisableDispatch() {
+        if (IsDispatchTrackingDisabled()) {
+            // TODO(bunnei): Until kernel threads are emulated, we cannot enable/disable dispatch.
+            return;
+        }
+
         ASSERT(GetCurrentThread(kernel).GetDisableDispatchCount() >= 0);
         this->GetStackParameters().disable_count++;
     }
 
     void EnableDispatch() {
+        if (IsDispatchTrackingDisabled()) {
+            // TODO(bunnei): Until kernel threads are emulated, we cannot enable/disable dispatch.
+            return;
+        }
+
         ASSERT(GetCurrentThread(kernel).GetDisableDispatchCount() > 0);
         this->GetStackParameters().disable_count--;
     }
@@ -708,6 +731,7 @@ private:
 
     // For emulation
     std::shared_ptr<Common::Fiber> host_context{};
+    bool is_single_core{};
 
     // For debugging
     std::vector<KSynchronizationObject*> wait_objects_for_debugging;
@@ -752,4 +776,16 @@ public:
     }
 };
 
+class KScopedDisableDispatch {
+public:
+    [[nodiscard]] explicit KScopedDisableDispatch(KernelCore& kernel_) : kernel{kernel_} {
+        GetCurrentThread(kernel).DisableDispatch();
+    }
+
+    ~KScopedDisableDispatch();
+
+private:
+    KernelCore& kernel;
+};
+
 } // namespace Kernel

src/core/hle/kernel/kernel.cpp

@@ -85,8 +85,9 @@ struct KernelCore::Impl {
     }
 
     void InitializeCores() {
-        for (auto& core : cores) {
-            core.Initialize(current_process->Is64BitProcess());
+        for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
+            cores[core_id].Initialize(current_process->Is64BitProcess());
+            system.Memory().SetCurrentPageTable(*current_process, core_id);
         }
     }
 
@@ -131,15 +132,6 @@ struct KernelCore::Impl {
         next_user_process_id = KProcess::ProcessIDMin;
         next_thread_id = 1;
 
-        for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            if (suspend_threads[core_id]) {
-                suspend_threads[core_id]->Close();
-                suspend_threads[core_id] = nullptr;
-            }
-
-            schedulers[core_id].reset();
-        }
-
         cores.clear();
 
         global_handle_table->Finalize();
@@ -167,6 +159,16 @@ struct KernelCore::Impl {
         CleanupObject(time_shared_mem);
         CleanupObject(system_resource_limit);
 
+        for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
+            if (suspend_threads[core_id]) {
+                suspend_threads[core_id]->Close();
+                suspend_threads[core_id] = nullptr;
+            }
+
+            schedulers[core_id]->Finalize();
+            schedulers[core_id].reset();
+        }
+
         // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
         next_host_thread_id = Core::Hardware::NUM_CPU_CORES;
 
@@ -257,14 +259,6 @@ struct KernelCore::Impl {
 
     void MakeCurrentProcess(KProcess* process) {
         current_process = process;
-        if (process == nullptr) {
-            return;
-        }
-
-        const u32 core_id = GetCurrentHostThreadID();
-        if (core_id < Core::Hardware::NUM_CPU_CORES) {
-            system.Memory().SetCurrentPageTable(*process, core_id);
-        }
     }
 
     static inline thread_local u32 host_thread_id = UINT32_MAX;
@@ -827,16 +821,20 @@ const Kernel::PhysicalCore& KernelCore::PhysicalCore(std::size_t id) const {
     return impl->cores[id];
 }
 
+size_t KernelCore::CurrentPhysicalCoreIndex() const {
+    const u32 core_id = impl->GetCurrentHostThreadID();
+    if (core_id >= Core::Hardware::NUM_CPU_CORES) {
+        return Core::Hardware::NUM_CPU_CORES - 1;
+    }
+    return core_id;
+}
+
 Kernel::PhysicalCore& KernelCore::CurrentPhysicalCore() {
-    u32 core_id = impl->GetCurrentHostThreadID();
-    ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
-    return impl->cores[core_id];
+    return impl->cores[CurrentPhysicalCoreIndex()];
 }
 
 const Kernel::PhysicalCore& KernelCore::CurrentPhysicalCore() const {
-    u32 core_id = impl->GetCurrentHostThreadID();
-    ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
-    return impl->cores[core_id];
+    return impl->cores[CurrentPhysicalCoreIndex()];
 }
 
 Kernel::KScheduler* KernelCore::CurrentScheduler() {
@@ -1029,6 +1027,9 @@ void KernelCore::Suspend(bool in_suspention) {
             impl->suspend_threads[core_id]->SetState(state);
             impl->suspend_threads[core_id]->SetWaitReasonForDebugging(
                 ThreadWaitReasonForDebugging::Suspended);
+            if (!should_suspend) {
+                impl->suspend_threads[core_id]->DisableDispatch();
+            }
         }
     }
 }
@@ -1043,13 +1044,11 @@ void KernelCore::ExceptionalExit() {
 }
 
 void KernelCore::EnterSVCProfile() {
-    std::size_t core = impl->GetCurrentHostThreadID();
-    impl->svc_ticks[core] = MicroProfileEnter(MICROPROFILE_TOKEN(Kernel_SVC));
+    impl->svc_ticks[CurrentPhysicalCoreIndex()] = MicroProfileEnter(MICROPROFILE_TOKEN(Kernel_SVC));
 }
 
 void KernelCore::ExitSVCProfile() {
-    std::size_t core = impl->GetCurrentHostThreadID();
-    MicroProfileLeave(MICROPROFILE_TOKEN(Kernel_SVC), impl->svc_ticks[core]);
+    MicroProfileLeave(MICROPROFILE_TOKEN(Kernel_SVC), impl->svc_ticks[CurrentPhysicalCoreIndex()]);
 }
 
 std::weak_ptr<Kernel::ServiceThread> KernelCore::CreateServiceThread(const std::string& name) {

src/core/hle/kernel/kernel.h

@@ -146,6 +146,9 @@ public:
     /// Gets the an instance of the respective physical CPU core.
     const Kernel::PhysicalCore& PhysicalCore(std::size_t id) const;
 
+    /// Gets the current physical core index for the running host thread.
+    std::size_t CurrentPhysicalCoreIndex() const;
+
     /// Gets the sole instance of the Scheduler at the current running core.
     Kernel::KScheduler* CurrentScheduler();
 

src/core/hle/kernel/svc.cpp

@@ -877,7 +877,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, Handle
             const u64 thread_ticks = current_thread->GetCpuTime();
 
             out_ticks = thread_ticks + (core_timing.GetCPUTicks() - prev_ctx_ticks);
-        } else if (same_thread && info_sub_id == system.CurrentCoreIndex()) {
+        } else if (same_thread && info_sub_id == system.Kernel().CurrentPhysicalCoreIndex()) {
             out_ticks = core_timing.GetCPUTicks() - prev_ctx_ticks;
         }
 

src/core/hle/service/ngct/ngct.cpp

@@ -1,46 +0,0 @@
-// Copyright 2021 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included
-
-#include "common/string_util.h"
-#include "core/core.h"
-#include "core/hle/ipc_helpers.h"
-#include "core/hle/service/ngct/ngct.h"
-#include "core/hle/service/service.h"
-
-namespace Service::NGCT {
-
-class IService final : public ServiceFramework<IService> {
-public:
-    explicit IService(Core::System& system_) : ServiceFramework{system_, "ngct:u"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {0, nullptr, "Match"},
-            {1, &IService::Filter, "Filter"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-
-private:
-    void Filter(Kernel::HLERequestContext& ctx) {
-        const auto buffer = ctx.ReadBuffer();
-        const auto text = Common::StringFromFixedZeroTerminatedBuffer(
-            reinterpret_cast<const char*>(buffer.data()), buffer.size());
-
-        LOG_WARNING(Service_NGCT, "(STUBBED) called, text={}", text);
-
-        // Return the same string since we don't censor anything
-        ctx.WriteBuffer(buffer);
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultSuccess);
-    }
-};
-
-void InstallInterfaces(SM::ServiceManager& service_manager, Core::System& system) {
-    std::make_shared<IService>(system)->InstallAsService(system.ServiceManager());
-}
-
-} // namespace Service::NGCT

src/core/hle/service/ngct/ngct.h

@@ -1,20 +0,0 @@
-// Copyright 2021 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included
-
-#pragma once
-
-namespace Core {
-class System;
-}
-
-namespace Service::SM {
-class ServiceManager;
-}
-
-namespace Service::NGCT {
-
-/// Registers all NGCT services with the specified service manager.
-void InstallInterfaces(SM::ServiceManager& service_manager, Core::System& system);
-
-} // namespace Service::NGCT

src/core/hle/service/nifm/nifm.cpp

@@ -277,45 +277,37 @@ private:
     void GetCurrentNetworkProfile(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
-        const auto net_iface = Network::GetSelectedNetworkInterface();
-
-        const SfNetworkProfileData network_profile_data = [&net_iface] {
-            if (!net_iface) {
-                return SfNetworkProfileData{};
-            }
-
-            return SfNetworkProfileData{
-                .ip_setting_data{
-                    .ip_address_setting{
-                        .is_automatic{true},
-                        .current_address{Network::TranslateIPv4(net_iface->ip_address)},
-                        .subnet_mask{Network::TranslateIPv4(net_iface->subnet_mask)},
-                        .gateway{Network::TranslateIPv4(net_iface->gateway)},
-                    },
-                    .dns_setting{
-                        .is_automatic{true},
-                        .primary_dns{1, 1, 1, 1},
-                        .secondary_dns{1, 0, 0, 1},
-                    },
-                    .proxy_setting{
-                        .enabled{false},
-                        .port{},
-                        .proxy_server{},
-                        .automatic_auth_enabled{},
-                        .user{},
-                        .password{},
-                    },
-                    .mtu{1500},
+        const SfNetworkProfileData network_profile_data{
+            .ip_setting_data{
+                .ip_address_setting{
+                    .is_automatic{true},
+                    .current_address{192, 168, 1, 100},
+                    .subnet_mask{255, 255, 255, 0},
+                    .gateway{192, 168, 1, 1},
                 },
-                .uuid{0xdeadbeef, 0xdeadbeef},
-                .network_name{"yuzu Network"},
-                .wireless_setting_data{
-                    .ssid_length{12},
-                    .ssid{"yuzu Network"},
-                    .passphrase{"yuzupassword"},
+                .dns_setting{
+                    .is_automatic{true},
+                    .primary_dns{1, 1, 1, 1},
+                    .secondary_dns{1, 0, 0, 1},
                 },
-            };
-        }();
+                .proxy_setting{
+                    .enabled{false},
+                    .port{},
+                    .proxy_server{},
+                    .automatic_auth_enabled{},
+                    .user{},
+                    .password{},
+                },
+                .mtu{1500},
+            },
+            .uuid{0xdeadbeef, 0xdeadbeef},
+            .network_name{"yuzu Network"},
+            .wireless_setting_data{
+                .ssid_length{12},
+                .ssid{"yuzu Network"},
+                .passphrase{"yuzupassword"},
+            },
+        };
 
         ctx.WriteBuffer(network_profile_data);
 
@@ -360,33 +352,38 @@ private:
         LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
         struct IpConfigInfo {
-            IpAddressSetting ip_address_setting{};
-            DnsSetting dns_setting{};
+            IpAddressSetting ip_address_setting;
+            DnsSetting dns_setting;
         };
         static_assert(sizeof(IpConfigInfo) == sizeof(IpAddressSetting) + sizeof(DnsSetting),
                       "IpConfigInfo has incorrect size.");
 
-        const auto net_iface = Network::GetSelectedNetworkInterface();
+        IpConfigInfo ip_config_info{
+            .ip_address_setting{
+                .is_automatic{true},
+                .current_address{0, 0, 0, 0},
+                .subnet_mask{255, 255, 255, 0},
+                .gateway{192, 168, 1, 1},
+            },
+            .dns_setting{
+                .is_automatic{true},
+                .primary_dns{1, 1, 1, 1},
+                .secondary_dns{1, 0, 0, 1},
+            },
+        };
 
-        const IpConfigInfo ip_config_info = [&net_iface] {
-            if (!net_iface) {
-                return IpConfigInfo{};
-            }
+        const auto iface = Network::GetSelectedNetworkInterface();
+        if (iface) {
+            ip_config_info.ip_address_setting =
+                IpAddressSetting{.is_automatic{true},
+                                 .current_address{Network::TranslateIPv4(iface->ip_address)},
+                                 .subnet_mask{Network::TranslateIPv4(iface->subnet_mask)},
+                                 .gateway{Network::TranslateIPv4(iface->gateway)}};
 
-            return IpConfigInfo{
-                .ip_address_setting{
-                    .is_automatic{true},
-                    .current_address{Network::TranslateIPv4(net_iface->ip_address)},
-                    .subnet_mask{Network::TranslateIPv4(net_iface->subnet_mask)},
-                    .gateway{Network::TranslateIPv4(net_iface->gateway)},
-                },
-                .dns_setting{
-                    .is_automatic{true},
-                    .primary_dns{1, 1, 1, 1},
-                    .secondary_dns{1, 0, 0, 1},
-                },
-            };
-        }();
+        } else {
+            LOG_ERROR(Service_NIFM,
+                      "Couldn't get host network configuration info, using default values");
+        }
 
         IPC::ResponseBuilder rb{ctx, 2 + (sizeof(IpConfigInfo) + 3) / sizeof(u32)};
         rb.Push(ResultSuccess);

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

@@ -9,17 +9,20 @@
 #include "core/core.h"
 #include "core/hle/kernel/k_writable_event.h"
 #include "core/hle/kernel/kernel.h"
+#include "core/hle/service/kernel_helpers.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
 
 namespace Service::NVFlinger {
 
-BufferQueue::BufferQueue(Kernel::KernelCore& kernel, u32 id_, u64 layer_id_)
-    : id(id_), layer_id(layer_id_), buffer_wait_event{kernel} {
-    Kernel::KAutoObject::Create(std::addressof(buffer_wait_event));
-    buffer_wait_event.Initialize("BufferQueue:WaitEvent");
+BufferQueue::BufferQueue(Kernel::KernelCore& kernel, u32 id_, u64 layer_id_,
+                         KernelHelpers::ServiceContext& service_context_)
+    : id(id_), layer_id(layer_id_), service_context{service_context_} {
+    buffer_wait_event = service_context.CreateEvent("BufferQueue:WaitEvent");
 }
 
-BufferQueue::~BufferQueue() = default;
+BufferQueue::~BufferQueue() {
+    service_context.CloseEvent(buffer_wait_event);
+}
 
 void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer) {
     ASSERT(slot < buffer_slots);
@@ -41,7 +44,7 @@ void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer)
         .multi_fence = {},
     };
 
-    buffer_wait_event.GetWritableEvent().Signal();
+    buffer_wait_event->GetWritableEvent().Signal();
 }
 
 std::optional<std::pair<u32, Service::Nvidia::MultiFence*>> BufferQueue::DequeueBuffer(u32 width,
@@ -119,7 +122,7 @@ void BufferQueue::CancelBuffer(u32 slot, const Service::Nvidia::MultiFence& mult
     }
     free_buffers_condition.notify_one();
 
-    buffer_wait_event.GetWritableEvent().Signal();
+    buffer_wait_event->GetWritableEvent().Signal();
 }
 
 std::optional<std::reference_wrapper<const BufferQueue::Buffer>> BufferQueue::AcquireBuffer() {
@@ -154,7 +157,7 @@ void BufferQueue::ReleaseBuffer(u32 slot) {
     }
     free_buffers_condition.notify_one();
 
-    buffer_wait_event.GetWritableEvent().Signal();
+    buffer_wait_event->GetWritableEvent().Signal();
 }
 
 void BufferQueue::Connect() {
@@ -169,7 +172,7 @@ void BufferQueue::Disconnect() {
         std::unique_lock lock{queue_sequence_mutex};
         queue_sequence.clear();
     }
-    buffer_wait_event.GetWritableEvent().Signal();
+    buffer_wait_event->GetWritableEvent().Signal();
     is_connect = false;
     free_buffers_condition.notify_one();
 }
@@ -189,11 +192,11 @@ u32 BufferQueue::Query(QueryType type) {
 }
 
 Kernel::KWritableEvent& BufferQueue::GetWritableBufferWaitEvent() {
-    return buffer_wait_event.GetWritableEvent();
+    return buffer_wait_event->GetWritableEvent();
 }
 
 Kernel::KReadableEvent& BufferQueue::GetBufferWaitEvent() {
-    return buffer_wait_event.GetReadableEvent();
+    return buffer_wait_event->GetReadableEvent();
 }
 
 } // namespace Service::NVFlinger

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

@@ -24,6 +24,10 @@ class KReadableEvent;
 class KWritableEvent;
 } // namespace Kernel
 
+namespace Service::KernelHelpers {
+class ServiceContext;
+} // namespace Service::KernelHelpers
+
 namespace Service::NVFlinger {
 
 constexpr u32 buffer_slots = 0x40;
@@ -54,7 +58,8 @@ public:
         NativeWindowFormat = 2,
     };
 
-    explicit BufferQueue(Kernel::KernelCore& kernel, u32 id_, u64 layer_id_);
+    explicit BufferQueue(Kernel::KernelCore& kernel, u32 id_, u64 layer_id_,
+                         KernelHelpers::ServiceContext& service_context_);
     ~BufferQueue();
 
     enum class BufferTransformFlags : u32 {
@@ -130,12 +135,14 @@ private:
     std::list<u32> free_buffers;
     std::array<Buffer, buffer_slots> buffers;
     std::list<u32> queue_sequence;
-    Kernel::KEvent buffer_wait_event;
+    Kernel::KEvent* buffer_wait_event{};
 
     std::mutex free_buffers_mutex;
     std::condition_variable free_buffers_condition;
 
     std::mutex queue_sequence_mutex;
+
+    KernelHelpers::ServiceContext& service_context;
 };
 
 } // namespace Service::NVFlinger

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

@@ -147,7 +147,7 @@ std::optional<u64> NVFlinger::CreateLayer(u64 display_id) {
 void NVFlinger::CreateLayerAtId(VI::Display& display, u64 layer_id) {
     const u32 buffer_queue_id = next_buffer_queue_id++;
     buffer_queues.emplace_back(
-        std::make_unique<BufferQueue>(system.Kernel(), buffer_queue_id, layer_id));
+        std::make_unique<BufferQueue>(system.Kernel(), buffer_queue_id, layer_id, service_context));
     display.CreateLayer(layer_id, *buffer_queues.back());
 }
 

src/core/hle/service/service.cpp

@@ -46,7 +46,6 @@
 #include "core/hle/service/ncm/ncm.h"
 #include "core/hle/service/nfc/nfc.h"
 #include "core/hle/service/nfp/nfp.h"
-#include "core/hle/service/ngct/ngct.h"
 #include "core/hle/service/nifm/nifm.h"
 #include "core/hle/service/nim/nim.h"
 #include "core/hle/service/npns/npns.h"
@@ -272,7 +271,6 @@ Services::Services(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system
     NCM::InstallInterfaces(*sm, system);
     NFC::InstallInterfaces(*sm, system);
     NFP::InstallInterfaces(*sm, system);
-    NGCT::InstallInterfaces(*sm, system);
     NIFM::InstallInterfaces(*sm, system);
     NIM::InstallInterfaces(*sm, system);
     NPNS::InstallInterfaces(*sm, system);

src/core/network/network_interface.cpp

@@ -37,73 +37,73 @@ std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
             AF_INET, GAA_FLAG_SKIP_MULTICAST | GAA_FLAG_SKIP_DNS_SERVER | GAA_FLAG_INCLUDE_GATEWAYS,
             nullptr, adapter_addresses.data(), &buf_size);
 
-        if (ret != ERROR_BUFFER_OVERFLOW) {
+        if (ret == ERROR_BUFFER_OVERFLOW) {
+            adapter_addresses.resize((buf_size / sizeof(IP_ADAPTER_ADDRESSES)) + 1);
+        } else {
             break;
         }
-
-        adapter_addresses.resize((buf_size / sizeof(IP_ADAPTER_ADDRESSES)) + 1);
     }
 
-    if (ret != NO_ERROR) {
+    if (ret == NO_ERROR) {
+        std::vector<NetworkInterface> result;
+
+        for (auto current_address = adapter_addresses.data(); current_address != nullptr;
+             current_address = current_address->Next) {
+            if (current_address->FirstUnicastAddress == nullptr ||
+                current_address->FirstUnicastAddress->Address.lpSockaddr == nullptr) {
+                continue;
+            }
+
+            if (current_address->OperStatus != IfOperStatusUp) {
+                continue;
+            }
+
+            const auto ip_addr = Common::BitCast<struct sockaddr_in>(
+                                     *current_address->FirstUnicastAddress->Address.lpSockaddr)
+                                     .sin_addr;
+
+            ULONG mask = 0;
+            if (ConvertLengthToIpv4Mask(current_address->FirstUnicastAddress->OnLinkPrefixLength,
+                                        &mask) != NO_ERROR) {
+                LOG_ERROR(Network, "Failed to convert IPv4 prefix length to subnet mask");
+                continue;
+            }
+
+            struct in_addr gateway = {.S_un{.S_addr{0}}};
+            if (current_address->FirstGatewayAddress != nullptr &&
+                current_address->FirstGatewayAddress->Address.lpSockaddr != nullptr) {
+                gateway = Common::BitCast<struct sockaddr_in>(
+                              *current_address->FirstGatewayAddress->Address.lpSockaddr)
+                              .sin_addr;
+            }
+
+            result.push_back(NetworkInterface{
+                .name{Common::UTF16ToUTF8(std::wstring{current_address->FriendlyName})},
+                .ip_address{ip_addr},
+                .subnet_mask = in_addr{.S_un{.S_addr{mask}}},
+                .gateway = gateway});
+        }
+
+        return result;
+    } else {
         LOG_ERROR(Network, "Failed to get network interfaces with GetAdaptersAddresses");
         return {};
     }
-
-    std::vector<NetworkInterface> result;
-
-    for (auto current_address = adapter_addresses.data(); current_address != nullptr;
-         current_address = current_address->Next) {
-        if (current_address->FirstUnicastAddress == nullptr ||
-            current_address->FirstUnicastAddress->Address.lpSockaddr == nullptr) {
-            continue;
-        }
-
-        if (current_address->OperStatus != IfOperStatusUp) {
-            continue;
-        }
-
-        const auto ip_addr = Common::BitCast<struct sockaddr_in>(
-                                 *current_address->FirstUnicastAddress->Address.lpSockaddr)
-                                 .sin_addr;
-
-        ULONG mask = 0;
-        if (ConvertLengthToIpv4Mask(current_address->FirstUnicastAddress->OnLinkPrefixLength,
-                                    &mask) != NO_ERROR) {
-            LOG_ERROR(Network, "Failed to convert IPv4 prefix length to subnet mask");
-            continue;
-        }
-
-        struct in_addr gateway = {.S_un{.S_addr{0}}};
-        if (current_address->FirstGatewayAddress != nullptr &&
-            current_address->FirstGatewayAddress->Address.lpSockaddr != nullptr) {
-            gateway = Common::BitCast<struct sockaddr_in>(
-                          *current_address->FirstGatewayAddress->Address.lpSockaddr)
-                          .sin_addr;
-        }
-
-        result.emplace_back(NetworkInterface{
-            .name{Common::UTF16ToUTF8(std::wstring{current_address->FriendlyName})},
-            .ip_address{ip_addr},
-            .subnet_mask = in_addr{.S_un{.S_addr{mask}}},
-            .gateway = gateway});
-    }
-
-    return result;
 }
 
 #else
 
 std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
+    std::vector<NetworkInterface> result;
+
     struct ifaddrs* ifaddr = nullptr;
 
     if (getifaddrs(&ifaddr) != 0) {
         LOG_ERROR(Network, "Failed to get network interfaces with getifaddrs: {}",
                   std::strerror(errno));
-        return {};
+        return result;
     }
 
-    std::vector<NetworkInterface> result;
-
     for (auto ifa = ifaddr; ifa != nullptr; ifa = ifa->ifa_next) {
         if (ifa->ifa_addr == nullptr || ifa->ifa_netmask == nullptr) {
             continue;
@@ -117,62 +117,55 @@ std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
             continue;
         }
 
-        u32 gateway{};
-
+        std::uint32_t gateway{0};
         std::ifstream file{"/proc/net/route"};
-        if (!file.is_open()) {
+        if (file.is_open()) {
+
+            // ignore header
+            file.ignore(std::numeric_limits<std::streamsize>::max(), '\n');
+
+            bool gateway_found = false;
+
+            for (std::string line; std::getline(file, line);) {
+                std::istringstream iss{line};
+
+                std::string iface_name{};
+                iss >> iface_name;
+                if (iface_name != ifa->ifa_name) {
+                    continue;
+                }
+
+                iss >> std::hex;
+
+                std::uint32_t dest{0};
+                iss >> dest;
+                if (dest != 0) {
+                    // not the default route
+                    continue;
+                }
+
+                iss >> gateway;
+
+                std::uint16_t flags{0};
+                iss >> flags;
+
+                // flag RTF_GATEWAY (defined in <linux/route.h>)
+                if ((flags & 0x2) == 0) {
+                    continue;
+                }
+
+                gateway_found = true;
+                break;
+            }
+
+            if (!gateway_found) {
+                gateway = 0;
+            }
+        } else {
             LOG_ERROR(Network, "Failed to open \"/proc/net/route\"");
-
-            result.emplace_back(NetworkInterface{
-                .name{ifa->ifa_name},
-                .ip_address{Common::BitCast<struct sockaddr_in>(*ifa->ifa_addr).sin_addr},
-                .subnet_mask{Common::BitCast<struct sockaddr_in>(*ifa->ifa_netmask).sin_addr},
-                .gateway{in_addr{.s_addr = gateway}}});
-            continue;
         }
 
-        // ignore header
-        file.ignore(std::numeric_limits<std::streamsize>::max(), '\n');
-
-        bool gateway_found = false;
-
-        for (std::string line; std::getline(file, line);) {
-            std::istringstream iss{line};
-
-            std::string iface_name;
-            iss >> iface_name;
-            if (iface_name != ifa->ifa_name) {
-                continue;
-            }
-
-            iss >> std::hex;
-
-            u32 dest{};
-            iss >> dest;
-            if (dest != 0) {
-                // not the default route
-                continue;
-            }
-
-            iss >> gateway;
-
-            u16 flags{};
-            iss >> flags;
-
-            // flag RTF_GATEWAY (defined in <linux/route.h>)
-            if ((flags & 0x2) == 0) {
-                continue;
-            }
-
-            gateway_found = true;
-            break;
-        }
-
-        if (!gateway_found) {
-            gateway = 0;
-        }
-
-        result.emplace_back(NetworkInterface{
+        result.push_back(NetworkInterface{
             .name{ifa->ifa_name},
             .ip_address{Common::BitCast<struct sockaddr_in>(*ifa->ifa_addr).sin_addr},
             .subnet_mask{Common::BitCast<struct sockaddr_in>(*ifa->ifa_netmask).sin_addr},
@@ -187,11 +180,11 @@ std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
 #endif
 
 std::optional<NetworkInterface> GetSelectedNetworkInterface() {
-    const auto& selected_network_interface = Settings::values.network_interface.GetValue();
+    const std::string& selected_network_interface = Settings::values.network_interface.GetValue();
     const auto network_interfaces = Network::GetAvailableNetworkInterfaces();
     if (network_interfaces.size() == 0) {
         LOG_ERROR(Network, "GetAvailableNetworkInterfaces returned no interfaces");
-        return std::nullopt;
+        return {};
     }
 
     const auto res =
@@ -199,12 +192,12 @@ std::optional<NetworkInterface> GetSelectedNetworkInterface() {
             return iface.name == selected_network_interface;
         });
 
-    if (res == network_interfaces.end()) {
+    if (res != network_interfaces.end()) {
+        return *res;
+    } else {
         LOG_ERROR(Network, "Couldn't find selected interface \"{}\"", selected_network_interface);
-        return std::nullopt;
+        return {};
     }
-
-    return *res;
 }
 
 } // namespace Network

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -333,9 +333,8 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
             return ctx.OpBitcast(ctx.F32[1], ctx.OpISub(ctx.U32[1], index, base));
         }
     case IR::Attribute::FrontFace:
-        return ctx.OpSelect(ctx.F32[1], ctx.OpLoad(ctx.U1, ctx.front_face),
-                            ctx.OpBitcast(ctx.F32[1], ctx.Const(std::numeric_limits<u32>::max())),
-                            ctx.f32_zero_value);
+        return ctx.OpSelect(ctx.U32[1], ctx.OpLoad(ctx.U1, ctx.front_face),
+                            ctx.Const(std::numeric_limits<u32>::max()), ctx.u32_zero_value);
     case IR::Attribute::PointSpriteS:
         return ctx.OpLoad(ctx.F32[1],
                           ctx.OpAccessChain(ctx.input_f32, ctx.point_coord, ctx.u32_zero_value));

src/shader_recompiler/frontend/maxwell/structured_control_flow.cpp

@@ -20,7 +20,6 @@
 #include "shader_recompiler/frontend/maxwell/decode.h"
 #include "shader_recompiler/frontend/maxwell/structured_control_flow.h"
 #include "shader_recompiler/frontend/maxwell/translate/translate.h"
-#include "shader_recompiler/host_translate_info.h"
 #include "shader_recompiler/object_pool.h"
 
 namespace Shader::Maxwell {
@@ -653,7 +652,7 @@ class TranslatePass {
 public:
     TranslatePass(ObjectPool<IR::Inst>& inst_pool_, ObjectPool<IR::Block>& block_pool_,
                   ObjectPool<Statement>& stmt_pool_, Environment& env_, Statement& root_stmt,
-                  IR::AbstractSyntaxList& syntax_list_, const HostTranslateInfo& host_info)
+                  IR::AbstractSyntaxList& syntax_list_)
         : stmt_pool{stmt_pool_}, inst_pool{inst_pool_}, block_pool{block_pool_}, env{env_},
           syntax_list{syntax_list_} {
         Visit(root_stmt, nullptr, nullptr);
@@ -661,9 +660,6 @@ public:
         IR::Block& first_block{*syntax_list.front().data.block};
         IR::IREmitter ir(first_block, first_block.begin());
         ir.Prologue();
-        if (uses_demote_to_helper && host_info.needs_demote_reorder) {
-            DemoteCombinationPass();
-        }
     }
 
 private:
@@ -813,14 +809,7 @@ private:
             }
             case StatementType::Return: {
                 ensure_block();
-                IR::Block* return_block{block_pool.Create(inst_pool)};
-                IR::IREmitter{*return_block}.Epilogue();
-                current_block->AddBranch(return_block);
-
-                auto& merge{syntax_list.emplace_back()};
-                merge.type = IR::AbstractSyntaxNode::Type::Block;
-                merge.data.block = return_block;
-
+                IR::IREmitter{*current_block}.Epilogue();
                 current_block = nullptr;
                 syntax_list.emplace_back().type = IR::AbstractSyntaxNode::Type::Return;
                 break;
@@ -835,7 +824,6 @@ private:
                 auto& merge{syntax_list.emplace_back()};
                 merge.type = IR::AbstractSyntaxNode::Type::Block;
                 merge.data.block = demote_block;
-                uses_demote_to_helper = true;
                 break;
             }
             case StatementType::Unreachable: {
@@ -867,117 +855,11 @@ private:
         return block_pool.Create(inst_pool);
     }
 
-    void DemoteCombinationPass() {
-        using Type = IR::AbstractSyntaxNode::Type;
-        std::vector<IR::Block*> demote_blocks;
-        std::vector<IR::U1> demote_conds;
-        u32 num_epilogues{};
-        u32 branch_depth{};
-        for (const IR::AbstractSyntaxNode& node : syntax_list) {
-            if (node.type == Type::If) {
-                ++branch_depth;
-            }
-            if (node.type == Type::EndIf) {
-                --branch_depth;
-            }
-            if (node.type != Type::Block) {
-                continue;
-            }
-            if (branch_depth > 1) {
-                // Skip reordering nested demote branches.
-                continue;
-            }
-            for (const IR::Inst& inst : node.data.block->Instructions()) {
-                const IR::Opcode op{inst.GetOpcode()};
-                if (op == IR::Opcode::DemoteToHelperInvocation) {
-                    demote_blocks.push_back(node.data.block);
-                    break;
-                }
-                if (op == IR::Opcode::Epilogue) {
-                    ++num_epilogues;
-                }
-            }
-        }
-        if (demote_blocks.size() == 0) {
-            return;
-        }
-        if (num_epilogues > 1) {
-            LOG_DEBUG(Shader, "Combining demotes with more than one return is not implemented.");
-            return;
-        }
-        s64 last_iterator_offset{};
-        auto& asl{syntax_list};
-        for (const IR::Block* demote_block : demote_blocks) {
-            const auto start_it{asl.begin() + last_iterator_offset};
-            auto asl_it{std::find_if(start_it, asl.end(), [&](const IR::AbstractSyntaxNode& asn) {
-                return asn.type == Type::If && asn.data.if_node.body == demote_block;
-            })};
-            if (asl_it == asl.end()) {
-                // Demote without a conditional branch.
-                // No need to proceed since all fragment instances will be demoted regardless.
-                return;
-            }
-            const IR::Block* const end_if = asl_it->data.if_node.merge;
-            demote_conds.push_back(asl_it->data.if_node.cond);
-            last_iterator_offset = std::distance(asl.begin(), asl_it);
-
-            asl_it = asl.erase(asl_it);
-            asl_it = std::find_if(asl_it, asl.end(), [&](const IR::AbstractSyntaxNode& asn) {
-                return asn.type == Type::Block && asn.data.block == demote_block;
-            });
-
-            asl_it = asl.erase(asl_it);
-            asl_it = std::find_if(asl_it, asl.end(), [&](const IR::AbstractSyntaxNode& asn) {
-                return asn.type == Type::EndIf && asn.data.end_if.merge == end_if;
-            });
-            asl_it = asl.erase(asl_it);
-        }
-        const auto epilogue_func{[](const IR::AbstractSyntaxNode& asn) {
-            if (asn.type != Type::Block) {
-                return false;
-            }
-            for (const auto& inst : asn.data.block->Instructions()) {
-                if (inst.GetOpcode() == IR::Opcode::Epilogue) {
-                    return true;
-                }
-            }
-            return false;
-        }};
-        const auto reverse_it{std::find_if(asl.rbegin(), asl.rend(), epilogue_func)};
-        const auto return_block_it{(reverse_it + 1).base()};
-
-        IR::IREmitter ir{*(return_block_it - 1)->data.block};
-        IR::U1 cond(IR::Value(false));
-        for (const auto& demote_cond : demote_conds) {
-            cond = ir.LogicalOr(cond, demote_cond);
-        }
-        cond.Inst()->DestructiveAddUsage(1);
-
-        IR::AbstractSyntaxNode demote_if_node{};
-        demote_if_node.type = Type::If;
-        demote_if_node.data.if_node.cond = cond;
-        demote_if_node.data.if_node.body = demote_blocks[0];
-        demote_if_node.data.if_node.merge = return_block_it->data.block;
-
-        IR::AbstractSyntaxNode demote_node{};
-        demote_node.type = Type::Block;
-        demote_node.data.block = demote_blocks[0];
-
-        IR::AbstractSyntaxNode demote_endif_node{};
-        demote_endif_node.type = Type::EndIf;
-        demote_endif_node.data.end_if.merge = return_block_it->data.block;
-
-        asl.insert(return_block_it, demote_endif_node);
-        asl.insert(return_block_it, demote_node);
-        asl.insert(return_block_it, demote_if_node);
-    }
-
     ObjectPool<Statement>& stmt_pool;
     ObjectPool<IR::Inst>& inst_pool;
     ObjectPool<IR::Block>& block_pool;
     Environment& env;
     IR::AbstractSyntaxList& syntax_list;
-    bool uses_demote_to_helper{};
 
 // TODO: C++20 Remove this when all compilers support constexpr std::vector
 #if __cpp_lib_constexpr_vector >= 201907
@@ -989,13 +871,12 @@ private:
 } // Anonymous namespace
 
 IR::AbstractSyntaxList BuildASL(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
-                                Environment& env, Flow::CFG& cfg,
-                                const HostTranslateInfo& host_info) {
+                                Environment& env, Flow::CFG& cfg) {
     ObjectPool<Statement> stmt_pool{64};
     GotoPass goto_pass{cfg, stmt_pool};
     Statement& root{goto_pass.RootStatement()};
     IR::AbstractSyntaxList syntax_list;
-    TranslatePass{inst_pool, block_pool, stmt_pool, env, root, syntax_list, host_info};
+    TranslatePass{inst_pool, block_pool, stmt_pool, env, root, syntax_list};
     return syntax_list;
 }
 

src/shader_recompiler/frontend/maxwell/structured_control_flow.h

@@ -11,13 +11,10 @@
 #include "shader_recompiler/frontend/maxwell/control_flow.h"
 #include "shader_recompiler/object_pool.h"
 
-namespace Shader {
-struct HostTranslateInfo;
-namespace Maxwell {
+namespace Shader::Maxwell {
 
 [[nodiscard]] IR::AbstractSyntaxList BuildASL(ObjectPool<IR::Inst>& inst_pool,
                                               ObjectPool<IR::Block>& block_pool, Environment& env,
-                                              Flow::CFG& cfg, const HostTranslateInfo& host_info);
+                                              Flow::CFG& cfg);
 
-} // namespace Maxwell
-} // namespace Shader
+} // namespace Shader::Maxwell

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -130,7 +130,7 @@ void AddNVNStorageBuffers(IR::Program& program) {
 IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
                              Environment& env, Flow::CFG& cfg, const HostTranslateInfo& host_info) {
     IR::Program program;
-    program.syntax_list = BuildASL(inst_pool, block_pool, env, cfg, host_info);
+    program.syntax_list = BuildASL(inst_pool, block_pool, env, cfg);
     program.blocks = GenerateBlocks(program.syntax_list);
     program.post_order_blocks = PostOrder(program.syntax_list.front());
     program.stage = env.ShaderStage();

src/shader_recompiler/host_translate_info.h

@@ -11,9 +11,8 @@ namespace Shader {
 
 /// Misc information about the host
 struct HostTranslateInfo {
-    bool support_float16{};      ///< True when the device supports 16-bit floats
-    bool support_int64{};        ///< True when the device supports 64-bit integers
-    bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
+    bool support_float16{}; ///< True when the device supports 16-bit floats
+    bool support_int64{};   ///< True when the device supports 64-bit integers
 };
 
 } // namespace Shader

src/video_core/buffer_cache/buffer_base.h

@@ -261,6 +261,16 @@ public:
         stream_score += score;
     }
 
+    /// Sets the new frame tick
+    void SetFrameTick(u64 new_frame_tick) noexcept {
+        frame_tick = new_frame_tick;
+    }
+
+    /// Returns the new frame tick
+    [[nodiscard]] u64 FrameTick() const noexcept {
+        return frame_tick;
+    }
+
     /// Returns the likeliness of this being a stream buffer
     [[nodiscard]] int StreamScore() const noexcept {
         return stream_score;
@@ -297,14 +307,6 @@ public:
         return words.size_bytes;
     }
 
-    size_t getLRUID() const noexcept {
-        return lru_id;
-    }
-
-    void setLRUID(size_t lru_id_) {
-        lru_id = lru_id_;
-    }
-
 private:
     template <Type type>
     u64* Array() noexcept {
@@ -601,9 +603,9 @@ private:
     RasterizerInterface* rasterizer = nullptr;
     VAddr cpu_addr = 0;
     Words words;
+    u64 frame_tick = 0;
     BufferFlagBits flags{};
     int stream_score = 0;
-    size_t lru_id = SIZE_MAX;
 };
 
 } // namespace VideoCommon

src/video_core/buffer_cache/buffer_cache.h

@@ -20,7 +20,6 @@
 #include "common/common_types.h"
 #include "common/div_ceil.h"
 #include "common/literals.h"
-#include "common/lru_cache.h"
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
 #include "common/settings.h"
@@ -331,7 +330,7 @@ private:
     template <bool insert>
     void ChangeRegister(BufferId buffer_id);
 
-    void TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept;
+    void TouchBuffer(Buffer& buffer) const noexcept;
 
     bool SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size);
 
@@ -429,11 +428,7 @@ private:
     size_t immediate_buffer_capacity = 0;
     std::unique_ptr<u8[]> immediate_buffer_alloc;
 
-    struct LRUItemParams {
-        using ObjectType = BufferId;
-        using TickType = u64;
-    };
-    Common::LeastRecentlyUsedCache<LRUItemParams> lru_cache;
+    typename SlotVector<Buffer>::Iterator deletion_iterator;
     u64 frame_tick = 0;
     u64 total_used_memory = 0;
 
@@ -450,6 +445,7 @@ BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
       kepler_compute{kepler_compute_}, gpu_memory{gpu_memory_}, cpu_memory{cpu_memory_} {
     // Ensure the first slot is used for the null buffer
     void(slot_buffers.insert(runtime, NullBufferParams{}));
+    deletion_iterator = slot_buffers.end();
     common_ranges.clear();
 }
 
@@ -458,17 +454,20 @@ void BufferCache<P>::RunGarbageCollector() {
     const bool aggressive_gc = total_used_memory >= CRITICAL_MEMORY;
     const u64 ticks_to_destroy = aggressive_gc ? 60 : 120;
     int num_iterations = aggressive_gc ? 64 : 32;
-    const auto clean_up = [this, &num_iterations](BufferId buffer_id) {
-        if (num_iterations == 0) {
-            return true;
+    for (; num_iterations > 0; --num_iterations) {
+        if (deletion_iterator == slot_buffers.end()) {
+            deletion_iterator = slot_buffers.begin();
         }
-        --num_iterations;
-        auto& buffer = slot_buffers[buffer_id];
-        DownloadBufferMemory(buffer);
-        DeleteBuffer(buffer_id);
-        return false;
-    };
-    lru_cache.ForEachItemBelow(frame_tick - ticks_to_destroy, clean_up);
+        ++deletion_iterator;
+        if (deletion_iterator == slot_buffers.end()) {
+            break;
+        }
+        const auto [buffer_id, buffer] = *deletion_iterator;
+        if (buffer->FrameTick() + ticks_to_destroy < frame_tick) {
+            DownloadBufferMemory(*buffer);
+            DeleteBuffer(buffer_id);
+        }
+    }
 }
 
 template <class P>
@@ -486,7 +485,7 @@ void BufferCache<P>::TickFrame() {
     const bool skip_preferred = hits * 256 < shots * 251;
     uniform_buffer_skip_cache_size = skip_preferred ? DEFAULT_SKIP_CACHE_SIZE : 0;
 
-    if (total_used_memory >= EXPECTED_MEMORY) {
+    if (Settings::values.use_caches_gc.GetValue() && total_used_memory >= EXPECTED_MEMORY) {
         RunGarbageCollector();
     }
     ++frame_tick;
@@ -955,7 +954,7 @@ bool BufferCache<P>::IsRegionCpuModified(VAddr addr, size_t size) {
 template <class P>
 void BufferCache<P>::BindHostIndexBuffer() {
     Buffer& buffer = slot_buffers[index_buffer.buffer_id];
-    TouchBuffer(buffer, index_buffer.buffer_id);
+    TouchBuffer(buffer);
     const u32 offset = buffer.Offset(index_buffer.cpu_addr);
     const u32 size = index_buffer.size;
     SynchronizeBuffer(buffer, index_buffer.cpu_addr, size);
@@ -976,7 +975,7 @@ void BufferCache<P>::BindHostVertexBuffers() {
     for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
         const Binding& binding = vertex_buffers[index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer, binding.buffer_id);
+        TouchBuffer(buffer);
         SynchronizeBuffer(buffer, binding.cpu_addr, binding.size);
         if (!flags[Dirty::VertexBuffer0 + index]) {
             continue;
@@ -1012,7 +1011,7 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
     const VAddr cpu_addr = binding.cpu_addr;
     const u32 size = std::min(binding.size, (*uniform_buffer_sizes)[stage][index]);
     Buffer& buffer = slot_buffers[binding.buffer_id];
-    TouchBuffer(buffer, binding.buffer_id);
+    TouchBuffer(buffer);
     const bool use_fast_buffer = binding.buffer_id != NULL_BUFFER_ID &&
                                  size <= uniform_buffer_skip_cache_size &&
                                  !buffer.IsRegionGpuModified(cpu_addr, size);
@@ -1084,7 +1083,7 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
     ForEachEnabledBit(enabled_storage_buffers[stage], [&](u32 index) {
         const Binding& binding = storage_buffers[stage][index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer, binding.buffer_id);
+        TouchBuffer(buffer);
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
@@ -1129,7 +1128,7 @@ void BufferCache<P>::BindHostTransformFeedbackBuffers() {
     for (u32 index = 0; index < NUM_TRANSFORM_FEEDBACK_BUFFERS; ++index) {
         const Binding& binding = transform_feedback_buffers[index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer, binding.buffer_id);
+        TouchBuffer(buffer);
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
@@ -1149,7 +1148,7 @@ void BufferCache<P>::BindHostComputeUniformBuffers() {
     ForEachEnabledBit(enabled_compute_uniform_buffer_mask, [&](u32 index) {
         const Binding& binding = compute_uniform_buffers[index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer, binding.buffer_id);
+        TouchBuffer(buffer);
         const u32 size = std::min(binding.size, (*compute_uniform_buffer_sizes)[index]);
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
@@ -1169,7 +1168,7 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
     ForEachEnabledBit(enabled_compute_storage_buffers, [&](u32 index) {
         const Binding& binding = compute_storage_buffers[index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer, binding.buffer_id);
+        TouchBuffer(buffer);
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
@@ -1514,11 +1513,11 @@ BufferId BufferCache<P>::CreateBuffer(VAddr cpu_addr, u32 wanted_size) {
     const OverlapResult overlap = ResolveOverlaps(cpu_addr, wanted_size);
     const u32 size = static_cast<u32>(overlap.end - overlap.begin);
     const BufferId new_buffer_id = slot_buffers.insert(runtime, rasterizer, overlap.begin, size);
+    TouchBuffer(slot_buffers[new_buffer_id]);
     for (const BufferId overlap_id : overlap.ids) {
         JoinOverlap(new_buffer_id, overlap_id, !overlap.has_stream_leap);
     }
     Register(new_buffer_id);
-    TouchBuffer(slot_buffers[new_buffer_id], new_buffer_id);
     return new_buffer_id;
 }
 
@@ -1535,14 +1534,12 @@ void BufferCache<P>::Unregister(BufferId buffer_id) {
 template <class P>
 template <bool insert>
 void BufferCache<P>::ChangeRegister(BufferId buffer_id) {
-    Buffer& buffer = slot_buffers[buffer_id];
+    const Buffer& buffer = slot_buffers[buffer_id];
     const auto size = buffer.SizeBytes();
     if (insert) {
         total_used_memory += Common::AlignUp(size, 1024);
-        buffer.setLRUID(lru_cache.Insert(buffer_id, frame_tick));
     } else {
         total_used_memory -= Common::AlignUp(size, 1024);
-        lru_cache.Free(buffer.getLRUID());
     }
     const VAddr cpu_addr_begin = buffer.CpuAddr();
     const VAddr cpu_addr_end = cpu_addr_begin + size;
@@ -1558,10 +1555,8 @@ void BufferCache<P>::ChangeRegister(BufferId buffer_id) {
 }
 
 template <class P>
-void BufferCache<P>::TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept {
-    if (buffer_id != NULL_BUFFER_ID) {
-        lru_cache.Touch(buffer.getLRUID(), frame_tick);
-    }
+void BufferCache<P>::TouchBuffer(Buffer& buffer) const noexcept {
+    buffer.SetFrameTick(frame_tick);
 }
 
 template <class P>

src/video_core/command_classes/codecs/vp9.cpp

@@ -742,7 +742,6 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
     uncomp_writer.WriteDeltaQ(current_frame_info.uv_dc_delta_q);
     uncomp_writer.WriteDeltaQ(current_frame_info.uv_ac_delta_q);
 
-    ASSERT(!current_frame_info.segment_enabled);
     uncomp_writer.WriteBit(false); // Segmentation enabled (TODO).
 
     const s32 min_tile_cols_log2 = CalcMinLog2TileCols(current_frame_info.frame_size.width);

src/video_core/command_classes/codecs/vp9_types.h

@@ -22,7 +22,7 @@ struct Vp9FrameDimensions {
 };
 static_assert(sizeof(Vp9FrameDimensions) == 0x8, "Vp9 Vp9FrameDimensions is an invalid size");
 
-enum class FrameFlags : u32 {
+enum FrameFlags : u32 {
     IsKeyFrame = 1 << 0,
     LastFrameIsKeyFrame = 1 << 1,
     FrameSizeChanged = 1 << 2,
@@ -30,7 +30,6 @@ enum class FrameFlags : u32 {
     LastShowFrame = 1 << 4,
     IntraOnly = 1 << 5,
 };
-DECLARE_ENUM_FLAG_OPERATORS(FrameFlags)
 
 enum class TxSize {
     Tx4x4 = 0,   // 4x4 transform
@@ -93,34 +92,44 @@ struct Vp9EntropyProbs {
 static_assert(sizeof(Vp9EntropyProbs) == 0x7B4, "Vp9EntropyProbs is an invalid size");
 
 struct Vp9PictureInfo {
-    u32 bitstream_size;
-    std::array<u64, 4> frame_offsets;
+    bool is_key_frame;
+    bool intra_only;
+    bool last_frame_was_key;
+    bool frame_size_changed;
+    bool error_resilient_mode;
+    bool last_frame_shown;
+    bool show_frame;
     std::array<s8, 4> ref_frame_sign_bias;
     s32 base_q_index;
     s32 y_dc_delta_q;
     s32 uv_dc_delta_q;
     s32 uv_ac_delta_q;
+    bool lossless;
     s32 transform_mode;
+    bool allow_high_precision_mv;
     s32 interp_filter;
     s32 reference_mode;
+    s8 comp_fixed_ref;
+    std::array<s8, 2> comp_var_ref;
     s32 log2_tile_cols;
     s32 log2_tile_rows;
+    bool segment_enabled;
+    bool segment_map_update;
+    bool segment_map_temporal_update;
+    s32 segment_abs_delta;
+    std::array<u32, 8> segment_feature_enable;
+    std::array<std::array<s16, 4>, 8> segment_feature_data;
+    bool mode_ref_delta_enabled;
+    bool use_prev_in_find_mv_refs;
     std::array<s8, 4> ref_deltas;
     std::array<s8, 2> mode_deltas;
     Vp9EntropyProbs entropy;
     Vp9FrameDimensions frame_size;
     u8 first_level;
     u8 sharpness_level;
-    bool is_key_frame;
-    bool intra_only;
-    bool last_frame_was_key;
-    bool error_resilient_mode;
-    bool last_frame_shown;
-    bool show_frame;
-    bool lossless;
-    bool allow_high_precision_mv;
-    bool segment_enabled;
-    bool mode_ref_delta_enabled;
+    u32 bitstream_size;
+    std::array<u64, 4> frame_offsets;
+    std::array<bool, 4> refresh_frame;
 };
 
 struct Vp9FrameContainer {
@@ -136,7 +145,7 @@ struct PictureInfo {
     Vp9FrameDimensions golden_frame_size;  ///< 0x50
     Vp9FrameDimensions alt_frame_size;     ///< 0x58
     Vp9FrameDimensions current_frame_size; ///< 0x60
-    FrameFlags vp9_flags;                  ///< 0x68
+    u32 vp9_flags;                         ///< 0x68
     std::array<s8, 4> ref_frame_sign_bias; ///< 0x6C
     u8 first_level;                        ///< 0x70
     u8 sharpness_level;                    ///< 0x71
@@ -149,43 +158,60 @@ struct PictureInfo {
     u8 allow_high_precision_mv;            ///< 0x78
     u8 interp_filter;                      ///< 0x79
     u8 reference_mode;                     ///< 0x7A
-    INSERT_PADDING_BYTES_NOINIT(3);        ///< 0x7B
+    s8 comp_fixed_ref;                     ///< 0x7B
+    std::array<s8, 2> comp_var_ref;        ///< 0x7C
     u8 log2_tile_cols;                     ///< 0x7E
     u8 log2_tile_rows;                     ///< 0x7F
     Segmentation segmentation;             ///< 0x80
     LoopFilter loop_filter;                ///< 0xE4
-    INSERT_PADDING_BYTES_NOINIT(21);       ///< 0xEB
+    INSERT_PADDING_BYTES_NOINIT(5);        ///< 0xEB
+    u32 surface_params;                    ///< 0xF0
+    INSERT_PADDING_WORDS_NOINIT(3);        ///< 0xF4
 
     [[nodiscard]] Vp9PictureInfo Convert() const {
         return {
-            .bitstream_size = bitstream_size,
-            .frame_offsets{},
+            .is_key_frame = (vp9_flags & FrameFlags::IsKeyFrame) != 0,
+            .intra_only = (vp9_flags & FrameFlags::IntraOnly) != 0,
+            .last_frame_was_key = (vp9_flags & FrameFlags::LastFrameIsKeyFrame) != 0,
+            .frame_size_changed = (vp9_flags & FrameFlags::FrameSizeChanged) != 0,
+            .error_resilient_mode = (vp9_flags & FrameFlags::ErrorResilientMode) != 0,
+            .last_frame_shown = (vp9_flags & FrameFlags::LastShowFrame) != 0,
+            .show_frame = true,
             .ref_frame_sign_bias = ref_frame_sign_bias,
             .base_q_index = base_q_index,
             .y_dc_delta_q = y_dc_delta_q,
             .uv_dc_delta_q = uv_dc_delta_q,
             .uv_ac_delta_q = uv_ac_delta_q,
+            .lossless = lossless != 0,
             .transform_mode = tx_mode,
+            .allow_high_precision_mv = allow_high_precision_mv != 0,
             .interp_filter = interp_filter,
             .reference_mode = reference_mode,
+            .comp_fixed_ref = comp_fixed_ref,
+            .comp_var_ref = comp_var_ref,
             .log2_tile_cols = log2_tile_cols,
             .log2_tile_rows = log2_tile_rows,
+            .segment_enabled = segmentation.enabled != 0,
+            .segment_map_update = segmentation.update_map != 0,
+            .segment_map_temporal_update = segmentation.temporal_update != 0,
+            .segment_abs_delta = segmentation.abs_delta,
+            .segment_feature_enable = segmentation.feature_mask,
+            .segment_feature_data = segmentation.feature_data,
+            .mode_ref_delta_enabled = loop_filter.mode_ref_delta_enabled != 0,
+            .use_prev_in_find_mv_refs = !(vp9_flags == (FrameFlags::ErrorResilientMode)) &&
+                                        !(vp9_flags == (FrameFlags::FrameSizeChanged)) &&
+                                        !(vp9_flags == (FrameFlags::IntraOnly)) &&
+                                        (vp9_flags == (FrameFlags::LastShowFrame)) &&
+                                        !(vp9_flags == (FrameFlags::LastFrameIsKeyFrame)),
             .ref_deltas = loop_filter.ref_deltas,
             .mode_deltas = loop_filter.mode_deltas,
             .entropy{},
             .frame_size = current_frame_size,
             .first_level = first_level,
             .sharpness_level = sharpness_level,
-            .is_key_frame = True(vp9_flags & FrameFlags::IsKeyFrame),
-            .intra_only = True(vp9_flags & FrameFlags::IntraOnly),
-            .last_frame_was_key = True(vp9_flags & FrameFlags::LastFrameIsKeyFrame),
-            .error_resilient_mode = True(vp9_flags & FrameFlags::ErrorResilientMode),
-            .last_frame_shown = True(vp9_flags & FrameFlags::LastShowFrame),
-            .show_frame = true,
-            .lossless = lossless != 0,
-            .allow_high_precision_mv = allow_high_precision_mv != 0,
-            .segment_enabled = segmentation.enabled != 0,
-            .mode_ref_delta_enabled = loop_filter.mode_ref_delta_enabled != 0,
+            .bitstream_size = bitstream_size,
+            .frame_offsets{},
+            .refresh_frame{},
         };
     }
 };
@@ -290,6 +316,7 @@ ASSERT_POSITION(last_frame_size, 0x48);
 ASSERT_POSITION(first_level, 0x70);
 ASSERT_POSITION(segmentation, 0x80);
 ASSERT_POSITION(loop_filter, 0xE4);
+ASSERT_POSITION(surface_params, 0xF0);
 #undef ASSERT_POSITION
 
 #define ASSERT_POSITION(field_name, position)                                                      \

src/video_core/renderer_opengl/gl_device.h

@@ -156,10 +156,6 @@ public:
         return shader_backend;
     }
 
-    bool IsAmd() const {
-        return vendor_name == "ATI Technologies Inc.";
-    }
-
 private:
     static bool TestVariableAoffi();
     static bool TestPreciseBug();

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -219,7 +219,6 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
       host_info{
           .support_float16 = false,
           .support_int64 = device.HasShaderInt64(),
-          .needs_demote_reorder = device.IsAmd(),
       } {
     if (use_asynchronous_shaders) {
         workers = CreateWorkers();

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -159,13 +159,11 @@ VkSemaphore VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
 
         const VAddr framebuffer_addr = framebuffer.address + framebuffer.offset;
         const u8* const host_ptr = cpu_memory.GetPointer(framebuffer_addr);
+        const size_t size_bytes = GetSizeInBytes(framebuffer);
 
         // TODO(Rodrigo): Read this from HLE
         constexpr u32 block_height_log2 = 4;
         const u32 bytes_per_pixel = GetBytesPerPixel(framebuffer);
-        const u64 size_bytes{Tegra::Texture::CalculateSize(true, bytes_per_pixel,
-                                                           framebuffer.stride, framebuffer.height,
-                                                           1, block_height_log2, 0)};
         Tegra::Texture::UnswizzleTexture(
             mapped_span.subspan(image_offset, size_bytes), std::span(host_ptr, size_bytes),
             bytes_per_pixel, framebuffer.width, framebuffer.height, 1, block_height_log2, 0);
@@ -358,7 +356,7 @@ void VKBlitScreen::CreateDescriptorPool() {
 void VKBlitScreen::CreateRenderPass() {
     const VkAttachmentDescription color_attachment{
         .flags = 0,
-        .format = swapchain.GetImageFormat(),
+        .format = swapchain.GetImageViewFormat(),
         .samples = VK_SAMPLE_COUNT_1_BIT,
         .loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR,
         .storeOp = VK_ATTACHMENT_STORE_OP_STORE,

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -325,8 +325,6 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, Tegra::Engines::Maxw
     host_info = Shader::HostTranslateInfo{
         .support_float16 = device.IsFloat16Supported(),
         .support_int64 = device.IsShaderInt64Supported(),
-        .needs_demote_reorder = driver_id == VK_DRIVER_ID_AMD_PROPRIETARY_KHR ||
-                                driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE_KHR,
     };
 }
 

src/video_core/renderer_vulkan/vk_swapchain.cpp

@@ -20,16 +20,15 @@ namespace Vulkan {
 
 namespace {
 
-VkSurfaceFormatKHR ChooseSwapSurfaceFormat(vk::Span<VkSurfaceFormatKHR> formats, bool srgb) {
+VkSurfaceFormatKHR ChooseSwapSurfaceFormat(vk::Span<VkSurfaceFormatKHR> formats) {
     if (formats.size() == 1 && formats[0].format == VK_FORMAT_UNDEFINED) {
         VkSurfaceFormatKHR format;
         format.format = VK_FORMAT_B8G8R8A8_UNORM;
         format.colorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR;
         return format;
     }
-    const auto& found = std::find_if(formats.begin(), formats.end(), [srgb](const auto& format) {
-        const auto request_format = srgb ? VK_FORMAT_B8G8R8A8_SRGB : VK_FORMAT_B8G8R8A8_UNORM;
-        return format.format == request_format &&
+    const auto& found = std::find_if(formats.begin(), formats.end(), [](const auto& format) {
+        return format.format == VK_FORMAT_B8G8R8A8_UNORM &&
                format.colorSpace == VK_COLOR_SPACE_SRGB_NONLINEAR_KHR;
     });
     return found != formats.end() ? *found : formats[0];
@@ -145,7 +144,7 @@ void VKSwapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities,
     const auto formats{physical_device.GetSurfaceFormatsKHR(surface)};
     const auto present_modes{physical_device.GetSurfacePresentModesKHR(surface)};
 
-    const VkSurfaceFormatKHR surface_format{ChooseSwapSurfaceFormat(formats, srgb)};
+    const VkSurfaceFormatKHR surface_format{ChooseSwapSurfaceFormat(formats)};
     const VkPresentModeKHR present_mode{ChooseSwapPresentMode(present_modes)};
 
     u32 requested_image_count{capabilities.minImageCount + 1};
@@ -180,6 +179,17 @@ void VKSwapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities,
         swapchain_ci.queueFamilyIndexCount = static_cast<u32>(queue_indices.size());
         swapchain_ci.pQueueFamilyIndices = queue_indices.data();
     }
+    static constexpr std::array view_formats{VK_FORMAT_B8G8R8A8_UNORM, VK_FORMAT_B8G8R8A8_SRGB};
+    VkImageFormatListCreateInfo format_list{
+        .sType = VK_STRUCTURE_TYPE_IMAGE_FORMAT_LIST_CREATE_INFO_KHR,
+        .pNext = nullptr,
+        .viewFormatCount = static_cast<u32>(view_formats.size()),
+        .pViewFormats = view_formats.data(),
+    };
+    if (device.IsKhrSwapchainMutableFormatEnabled()) {
+        format_list.pNext = std::exchange(swapchain_ci.pNext, &format_list);
+        swapchain_ci.flags |= VK_SWAPCHAIN_CREATE_MUTABLE_FORMAT_BIT_KHR;
+    }
     // Request the size again to reduce the possibility of a TOCTOU race condition.
     const auto updated_capabilities = physical_device.GetSurfaceCapabilitiesKHR(surface);
     swapchain_ci.imageExtent = ChooseSwapExtent(updated_capabilities, width, height);
@@ -191,7 +201,7 @@ void VKSwapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities,
 
     images = swapchain.GetImages();
     image_count = static_cast<u32>(images.size());
-    image_format = surface_format.format;
+    image_view_format = srgb ? VK_FORMAT_B8G8R8A8_SRGB : VK_FORMAT_B8G8R8A8_UNORM;
 }
 
 void VKSwapchain::CreateSemaphores() {
@@ -207,7 +217,7 @@ void VKSwapchain::CreateImageViews() {
         .flags = 0,
         .image = {},
         .viewType = VK_IMAGE_VIEW_TYPE_2D,
-        .format = image_format,
+        .format = image_view_format,
         .components =
             {
                 .r = VK_COMPONENT_SWIZZLE_IDENTITY,

src/video_core/renderer_vulkan/vk_swapchain.h

@@ -68,8 +68,8 @@ public:
         return *image_views[index];
     }
 
-    VkFormat GetImageFormat() const {
-        return image_format;
+    VkFormat GetImageViewFormat() const {
+        return image_view_format;
     }
 
 private:
@@ -96,7 +96,7 @@ private:
     u32 image_index{};
     u32 frame_index{};
 
-    VkFormat image_format{};
+    VkFormat image_view_format{};
     VkExtent2D extent{};
 
     bool current_srgb{};

src/video_core/texture_cache/image_base.h

@@ -80,7 +80,7 @@ struct ImageBase {
     VAddr cpu_addr_end = 0;
 
     u64 modification_tick = 0;
-    size_t lru_index = SIZE_MAX;
+    u64 frame_tick = 0;
 
     std::array<u32, MAX_MIP_LEVELS> mip_level_offsets{};
 

src/video_core/texture_cache/texture_cache.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include "common/alignment.h"
+#include "common/settings.h"
 #include "video_core/dirty_flags.h"
 #include "video_core/texture_cache/samples_helper.h"
 #include "video_core/texture_cache/texture_cache_base.h"
@@ -42,6 +43,8 @@ TextureCache<P>::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface&
     void(slot_image_views.insert(runtime, NullImageParams{}));
     void(slot_samplers.insert(runtime, sampler_descriptor));
 
+    deletion_iterator = slot_images.begin();
+
     if constexpr (HAS_DEVICE_MEMORY_INFO) {
         const auto device_memory = runtime.GetDeviceLocalMemory();
         const u64 possible_expected_memory = (device_memory * 3) / 10;
@@ -61,38 +64,70 @@ template <class P>
 void TextureCache<P>::RunGarbageCollector() {
     const bool high_priority_mode = total_used_memory >= expected_memory;
     const bool aggressive_mode = total_used_memory >= critical_memory;
-    const u64 ticks_to_destroy = aggressive_mode ? 10ULL : high_priority_mode ? 25ULL : 100ULL;
-    size_t num_iterations = aggressive_mode ? 10000 : (high_priority_mode ? 100 : 5);
-    const auto clean_up = [this, &num_iterations, high_priority_mode](ImageId image_id) {
-        if (num_iterations == 0) {
-            return true;
+    const u64 ticks_to_destroy = high_priority_mode ? 60 : 100;
+    int num_iterations = aggressive_mode ? 256 : (high_priority_mode ? 128 : 64);
+    for (; num_iterations > 0; --num_iterations) {
+        if (deletion_iterator == slot_images.end()) {
+            deletion_iterator = slot_images.begin();
+            if (deletion_iterator == slot_images.end()) {
+                break;
+            }
         }
-        --num_iterations;
-        auto& image = slot_images[image_id];
-        const bool must_download = image.IsSafeDownload();
-        if (!high_priority_mode && must_download) {
-            return false;
+        auto [image_id, image_tmp] = *deletion_iterator;
+        Image* image = image_tmp; // fix clang error.
+        const bool is_alias = True(image->flags & ImageFlagBits::Alias);
+        const bool is_bad_overlap = True(image->flags & ImageFlagBits::BadOverlap);
+        const bool must_download = image->IsSafeDownload();
+        bool should_care = is_bad_overlap || is_alias || (high_priority_mode && !must_download);
+        const u64 ticks_needed =
+            is_bad_overlap
+                ? ticks_to_destroy >> 4
+                : ((should_care && aggressive_mode) ? ticks_to_destroy >> 1 : ticks_to_destroy);
+        should_care |= aggressive_mode;
+        if (should_care && image->frame_tick + ticks_needed < frame_tick) {
+            if (is_bad_overlap) {
+                const bool overlap_check = std::ranges::all_of(
+                    image->overlapping_images, [&, image](const ImageId& overlap_id) {
+                        auto& overlap = slot_images[overlap_id];
+                        return overlap.frame_tick >= image->frame_tick;
+                    });
+                if (!overlap_check) {
+                    ++deletion_iterator;
+                    continue;
+                }
+            }
+            if (!is_bad_overlap && must_download) {
+                const bool alias_check = std::ranges::none_of(
+                    image->aliased_images, [&, image](const AliasedImage& alias) {
+                        auto& alias_image = slot_images[alias.id];
+                        return (alias_image.frame_tick < image->frame_tick) ||
+                               (alias_image.modification_tick < image->modification_tick);
+                    });
+
+                if (alias_check) {
+                    auto map = runtime.DownloadStagingBuffer(image->unswizzled_size_bytes);
+                    const auto copies = FullDownloadCopies(image->info);
+                    image->DownloadMemory(map, copies);
+                    runtime.Finish();
+                    SwizzleImage(gpu_memory, image->gpu_addr, image->info, copies, map.mapped_span);
+                }
+            }
+            if (True(image->flags & ImageFlagBits::Tracked)) {
+                UntrackImage(*image, image_id);
+            }
+            UnregisterImage(image_id);
+            DeleteImage(image_id);
+            if (is_bad_overlap) {
+                ++num_iterations;
+            }
         }
-        if (must_download) {
-            auto map = runtime.DownloadStagingBuffer(image.unswizzled_size_bytes);
-            const auto copies = FullDownloadCopies(image.info);
-            image.DownloadMemory(map, copies);
-            runtime.Finish();
-            SwizzleImage(gpu_memory, image.gpu_addr, image.info, copies, map.mapped_span);
-        }
-        if (True(image.flags & ImageFlagBits::Tracked)) {
-            UntrackImage(image, image_id);
-        }
-        UnregisterImage(image_id);
-        DeleteImage(image_id);
-        return false;
-    };
-    lru_cache.ForEachItemBelow(frame_tick - ticks_to_destroy, clean_up);
+        ++deletion_iterator;
+    }
 }
 
 template <class P>
 void TextureCache<P>::TickFrame() {
-    if (total_used_memory > minimum_memory) {
+    if (Settings::values.use_caches_gc.GetValue() && total_used_memory > minimum_memory) {
         RunGarbageCollector();
     }
     sentenced_images.Tick();
@@ -1043,8 +1078,6 @@ void TextureCache<P>::RegisterImage(ImageId image_id) {
         tentative_size = EstimatedDecompressedSize(tentative_size, image.info.format);
     }
     total_used_memory += Common::AlignUp(tentative_size, 1024);
-    image.lru_index = lru_cache.Insert(image_id, frame_tick);
-
     ForEachGPUPage(image.gpu_addr, image.guest_size_bytes,
                    [this, image_id](u64 page) { gpu_page_table[page].push_back(image_id); });
     if (False(image.flags & ImageFlagBits::Sparse)) {
@@ -1082,7 +1115,6 @@ void TextureCache<P>::UnregisterImage(ImageId image_id) {
         tentative_size = EstimatedDecompressedSize(tentative_size, image.info.format);
     }
     total_used_memory -= Common::AlignUp(tentative_size, 1024);
-    lru_cache.Free(image.lru_index);
     const auto& clear_page_table =
         [this, image_id](
             u64 page,
@@ -1352,7 +1384,7 @@ void TextureCache<P>::PrepareImage(ImageId image_id, bool is_modification, bool
     if (is_modification) {
         MarkModification(image);
     }
-    lru_cache.Touch(image.lru_index, frame_tick);
+    image.frame_tick = frame_tick;
 }
 
 template <class P>

src/video_core/texture_cache/texture_cache_base.h

@@ -14,7 +14,6 @@
 
 #include "common/common_types.h"
 #include "common/literals.h"
-#include "common/lru_cache.h"
 #include "video_core/compatible_formats.h"
 #include "video_core/delayed_destruction_ring.h"
 #include "video_core/engines/fermi_2d.h"
@@ -371,12 +370,6 @@ private:
     std::vector<ImageId> uncommitted_downloads;
     std::queue<std::vector<ImageId>> committed_downloads;
 
-    struct LRUItemParams {
-        using ObjectType = ImageId;
-        using TickType = u64;
-    };
-    Common::LeastRecentlyUsedCache<LRUItemParams> lru_cache;
-
     static constexpr size_t TICKS_TO_DESTROY = 6;
     DelayedDestructionRing<Image, TICKS_TO_DESTROY> sentenced_images;
     DelayedDestructionRing<ImageView, TICKS_TO_DESTROY> sentenced_image_view;
@@ -386,6 +379,7 @@ private:
 
     u64 modification_tick = 0;
     u64 frame_tick = 0;
+    typename SlotVector<Image>::Iterator deletion_iterator;
 };
 
 } // namespace VideoCommon

src/video_core/textures/decoders.cpp

@@ -63,6 +63,14 @@ void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32
                 const u32 unswizzled_offset =
                     slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL;
 
+                if (const auto offset = (TO_LINEAR ? unswizzled_offset : swizzled_offset);
+                    offset >= input.size()) {
+                    // TODO(Rodrigo): This is an out of bounds access that should never happen. To
+                    // avoid crashing the emulator, break.
+                    ASSERT_MSG(false, "offset {} exceeds input size {}!", offset, input.size());
+                    break;
+                }
+
                 u8* const dst = &output[TO_LINEAR ? swizzled_offset : unswizzled_offset];
                 const u8* const src = &input[TO_LINEAR ? unswizzled_offset : swizzled_offset];
 

src/video_core/vulkan_common/vulkan_device.cpp

@@ -839,6 +839,8 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
     bool has_khr_shader_float16_int8{};
     bool has_khr_workgroup_memory_explicit_layout{};
     bool has_khr_pipeline_executable_properties{};
+    bool has_khr_image_format_list{};
+    bool has_khr_swapchain_mutable_format{};
     bool has_ext_subgroup_size_control{};
     bool has_ext_transform_feedback{};
     bool has_ext_custom_border_color{};
@@ -888,6 +890,9 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
         test(has_ext_shader_atomic_int64, VK_KHR_SHADER_ATOMIC_INT64_EXTENSION_NAME, false);
         test(has_khr_workgroup_memory_explicit_layout,
              VK_KHR_WORKGROUP_MEMORY_EXPLICIT_LAYOUT_EXTENSION_NAME, false);
+        test(has_khr_image_format_list, VK_KHR_IMAGE_FORMAT_LIST_EXTENSION_NAME, false);
+        test(has_khr_swapchain_mutable_format, VK_KHR_SWAPCHAIN_MUTABLE_FORMAT_EXTENSION_NAME,
+             false);
         test(has_ext_line_rasterization, VK_EXT_LINE_RASTERIZATION_EXTENSION_NAME, false);
         if (Settings::values.enable_nsight_aftermath) {
             test(nv_device_diagnostics_config, VK_NV_DEVICE_DIAGNOSTICS_CONFIG_EXTENSION_NAME,
@@ -1066,6 +1071,11 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
             khr_pipeline_executable_properties = true;
         }
     }
+    if (has_khr_image_format_list && has_khr_swapchain_mutable_format) {
+        extensions.push_back(VK_KHR_IMAGE_FORMAT_LIST_EXTENSION_NAME);
+        extensions.push_back(VK_KHR_SWAPCHAIN_MUTABLE_FORMAT_EXTENSION_NAME);
+        khr_swapchain_mutable_format = true;
+    }
     if (khr_push_descriptor) {
         VkPhysicalDevicePushDescriptorPropertiesKHR push_descriptor;
         push_descriptor.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PUSH_DESCRIPTOR_PROPERTIES_KHR;

src/video_core/vulkan_common/vulkan_device.h

@@ -224,6 +224,11 @@ public:
         return khr_pipeline_executable_properties;
     }
 
+    /// Returns true if VK_KHR_swapchain_mutable_format is enabled.
+    bool IsKhrSwapchainMutableFormatEnabled() const {
+        return khr_swapchain_mutable_format;
+    }
+
     /// Returns true if the device supports VK_KHR_workgroup_memory_explicit_layout.
     bool IsKhrWorkgroupMemoryExplicitLayoutSupported() const {
         return khr_workgroup_memory_explicit_layout;
@@ -390,6 +395,7 @@ private:
     bool khr_workgroup_memory_explicit_layout{}; ///< Support for explicit workgroup layouts.
     bool khr_push_descriptor{};                  ///< Support for VK_KHR_push_descritor.
     bool khr_pipeline_executable_properties{};   ///< Support for executable properties.
+    bool khr_swapchain_mutable_format{};         ///< Support for VK_KHR_swapchain_mutable_format.
     bool ext_index_type_uint8{};                 ///< Support for VK_EXT_index_type_uint8.
     bool ext_sampler_filter_minmax{};            ///< Support for VK_EXT_sampler_filter_minmax.
     bool ext_depth_range_unrestricted{};         ///< Support for VK_EXT_depth_range_unrestricted.

src/yuzu/configuration/config.cpp

@@ -818,6 +818,7 @@ void Config::ReadRendererValues() {
     ReadGlobalSetting(Settings::values.shader_backend);
     ReadGlobalSetting(Settings::values.use_asynchronous_shaders);
     ReadGlobalSetting(Settings::values.use_fast_gpu_time);
+    ReadGlobalSetting(Settings::values.use_caches_gc);
     ReadGlobalSetting(Settings::values.bg_red);
     ReadGlobalSetting(Settings::values.bg_green);
     ReadGlobalSetting(Settings::values.bg_blue);
@@ -1358,6 +1359,7 @@ void Config::SaveRendererValues() {
                  Settings::values.shader_backend.UsingGlobal());
     WriteGlobalSetting(Settings::values.use_asynchronous_shaders);
     WriteGlobalSetting(Settings::values.use_fast_gpu_time);
+    WriteGlobalSetting(Settings::values.use_caches_gc);
     WriteGlobalSetting(Settings::values.bg_red);
     WriteGlobalSetting(Settings::values.bg_green);
     WriteGlobalSetting(Settings::values.bg_blue);

src/yuzu/configuration/configure_graphics.ui

@@ -156,7 +156,7 @@
         <item>
          <widget class="QCheckBox" name="use_disk_shader_cache">
           <property name="text">
-           <string>Use disk pipeline cache</string>
+           <string>Use disk shader cache</string>
           </property>
          </widget>
         </item>

src/yuzu/configuration/configure_graphics_advanced.cpp

@@ -28,6 +28,7 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
 
     ui->use_vsync->setChecked(Settings::values.use_vsync.GetValue());
     ui->use_asynchronous_shaders->setChecked(Settings::values.use_asynchronous_shaders.GetValue());
+    ui->use_caches_gc->setChecked(Settings::values.use_caches_gc.GetValue());
     ui->use_fast_gpu_time->setChecked(Settings::values.use_fast_gpu_time.GetValue());
 
     if (Settings::IsConfiguringGlobal()) {
@@ -54,6 +55,8 @@ void ConfigureGraphicsAdvanced::ApplyConfiguration() {
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_asynchronous_shaders,
                                              ui->use_asynchronous_shaders,
                                              use_asynchronous_shaders);
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_caches_gc, ui->use_caches_gc,
+                                             use_caches_gc);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_fast_gpu_time,
                                              ui->use_fast_gpu_time, use_fast_gpu_time);
 }
@@ -78,6 +81,7 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
         ui->use_asynchronous_shaders->setEnabled(
             Settings::values.use_asynchronous_shaders.UsingGlobal());
         ui->use_fast_gpu_time->setEnabled(Settings::values.use_fast_gpu_time.UsingGlobal());
+        ui->use_caches_gc->setEnabled(Settings::values.use_caches_gc.UsingGlobal());
         ui->anisotropic_filtering_combobox->setEnabled(
             Settings::values.max_anisotropy.UsingGlobal());
 
@@ -90,6 +94,8 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
                                             use_asynchronous_shaders);
     ConfigurationShared::SetColoredTristate(ui->use_fast_gpu_time,
                                             Settings::values.use_fast_gpu_time, use_fast_gpu_time);
+    ConfigurationShared::SetColoredTristate(ui->use_caches_gc, Settings::values.use_caches_gc,
+                                            use_caches_gc);
     ConfigurationShared::SetColoredComboBox(
         ui->gpu_accuracy, ui->label_gpu_accuracy,
         static_cast<int>(Settings::values.gpu_accuracy.GetValue(true)));

src/yuzu/configuration/configure_graphics_advanced.h

@@ -37,4 +37,5 @@ private:
     ConfigurationShared::CheckState use_vsync;
     ConfigurationShared::CheckState use_asynchronous_shaders;
     ConfigurationShared::CheckState use_fast_gpu_time;
+    ConfigurationShared::CheckState use_caches_gc;
 };

src/yuzu/configuration/configure_graphics_advanced.ui

@@ -82,7 +82,7 @@
            <string>Enables asynchronous shader compilation, which may reduce shader stutter. This feature is experimental.</string>
           </property>
           <property name="text">
-           <string>Use asynchronous shader building (Hack)</string>
+           <string>Use asynchronous shader building (hack)</string>
           </property>
          </widget>
         </item>
@@ -92,7 +92,17 @@
             <string>Enables Fast GPU Time. This option will force most games to run at their highest native resolution.</string>
           </property>
           <property name="text">
-           <string>Use Fast GPU Time (Hack)</string>
+           <string>Use Fast GPU Time (hack)</string>
+          </property>
+         </widget>
+        </item>
+        <item>
+         <widget class="QCheckBox" name="use_caches_gc">
+          <property name="toolTip">
+           <string>Enables garbage collection for the GPU caches, this will try to keep VRAM within 3-4 GB by flushing the least used textures/buffers. May cause issues in a few games.</string>
+          </property>
+          <property name="text">
+           <string>Enable GPU cache garbage collection (experimental)</string>
           </property>
          </widget>
         </item>

src/yuzu_cmd/config.cpp

@@ -468,6 +468,7 @@ void Config::ReadValues() {
     ReadSetting("Renderer", Settings::values.use_nvdec_emulation);
     ReadSetting("Renderer", Settings::values.accelerate_astc);
     ReadSetting("Renderer", Settings::values.use_fast_gpu_time);
+    ReadSetting("Renderer", Settings::values.use_caches_gc);
 
     ReadSetting("Renderer", Settings::values.bg_red);
     ReadSetting("Renderer", Settings::values.bg_green);