speed up memory page mapping

vk_resource_manager: Implement fence and command buffer allocator

.gitmodules

@@ -37,3 +37,6 @@
 [submodule "discord-rpc"]
     path = externals/discord-rpc
     url = https://github.com/discordapp/discord-rpc.git
+[submodule "Vulkan-Headers"]
+    path = externals/Vulkan-Headers
+    url = https://github.com/KhronosGroup/Vulkan-Headers.git

CMakeLists.txt

@@ -23,6 +23,8 @@ option(YUZU_USE_QT_WEB_ENGINE "Use QtWebEngine for web applet implementation" OF
 
 option(ENABLE_CUBEB "Enables the cubeb audio backend" ON)
 
+option(ENABLE_VULKAN "Enables Vulkan backend" ON)
+
 option(USE_DISCORD_PRESENCE "Enables Discord Rich Presence" OFF)
 
 if(NOT EXISTS ${PROJECT_SOURCE_DIR}/.git/hooks/pre-commit)

src/audio_core/audio_out.cpp

@@ -26,14 +26,15 @@ static Stream::Format ChannelsToStreamFormat(u32 num_channels) {
     return {};
 }
 
-StreamPtr AudioOut::OpenStream(u32 sample_rate, u32 num_channels, std::string&& name,
+StreamPtr AudioOut::OpenStream(Core::Timing::CoreTiming& core_timing, u32 sample_rate,
+                               u32 num_channels, std::string&& name,
                                Stream::ReleaseCallback&& release_callback) {
     if (!sink) {
         sink = CreateSinkFromID(Settings::values.sink_id, Settings::values.audio_device_id);
     }
 
     return std::make_shared<Stream>(
-        sample_rate, ChannelsToStreamFormat(num_channels), std::move(release_callback),
+        core_timing, sample_rate, ChannelsToStreamFormat(num_channels), std::move(release_callback),
         sink->AcquireSinkStream(sample_rate, num_channels, name), std::move(name));
 }
 

src/audio_core/audio_out.h

@@ -13,6 +13,10 @@
 #include "audio_core/stream.h"
 #include "common/common_types.h"
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace AudioCore {
 
 /**
@@ -21,8 +25,8 @@ namespace AudioCore {
 class AudioOut {
 public:
     /// Opens a new audio stream
-    StreamPtr OpenStream(u32 sample_rate, u32 num_channels, std::string&& name,
-                         Stream::ReleaseCallback&& release_callback);
+    StreamPtr OpenStream(Core::Timing::CoreTiming& core_timing, u32 sample_rate, u32 num_channels,
+                         std::string&& name, Stream::ReleaseCallback&& release_callback);
 
     /// Returns a vector of recently released buffers specified by tag for the specified stream
     std::vector<Buffer::Tag> GetTagsAndReleaseBuffers(StreamPtr stream, std::size_t max_count);

src/audio_core/audio_renderer.cpp

@@ -8,6 +8,7 @@
 #include "audio_core/codec.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/hle/kernel/writable_event.h"
 #include "core/memory.h"
 
@@ -71,14 +72,14 @@ private:
     EffectOutStatus out_status{};
     EffectInStatus info{};
 };
-AudioRenderer::AudioRenderer(AudioRendererParameter params,
+AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, AudioRendererParameter params,
                              Kernel::SharedPtr<Kernel::WritableEvent> buffer_event)
     : worker_params{params}, buffer_event{buffer_event}, voices(params.voice_count),
       effects(params.effect_count) {
 
     audio_out = std::make_unique<AudioCore::AudioOut>();
-    stream = audio_out->OpenStream(STREAM_SAMPLE_RATE, STREAM_NUM_CHANNELS, "AudioRenderer",
-                                   [=]() { buffer_event->Signal(); });
+    stream = audio_out->OpenStream(core_timing, STREAM_SAMPLE_RATE, STREAM_NUM_CHANNELS,
+                                   "AudioRenderer", [=]() { buffer_event->Signal(); });
     audio_out->StartStream(stream);
 
     QueueMixedBuffer(0);

src/audio_core/audio_renderer.h

@@ -14,6 +14,10 @@
 #include "common/swap.h"
 #include "core/hle/kernel/object.h"
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Kernel {
 class WritableEvent;
 }
@@ -208,7 +212,7 @@ static_assert(sizeof(UpdateDataHeader) == 0x40, "UpdateDataHeader has wrong size
 
 class AudioRenderer {
 public:
-    AudioRenderer(AudioRendererParameter params,
+    AudioRenderer(Core::Timing::CoreTiming& core_timing, AudioRendererParameter params,
                   Kernel::SharedPtr<Kernel::WritableEvent> buffer_event);
     ~AudioRenderer();
 

src/audio_core/buffer.h

@@ -21,7 +21,7 @@ public:
     Buffer(Tag tag, std::vector<s16>&& samples) : tag{tag}, samples{std::move(samples)} {}
 
     /// Returns the raw audio data for the buffer
-    std::vector<s16>& Samples() {
+    std::vector<s16>& GetSamples() {
         return samples;
     }
 

src/audio_core/stream.cpp

@@ -32,12 +32,12 @@ u32 Stream::GetNumChannels() const {
     return {};
 }
 
-Stream::Stream(u32 sample_rate, Format format, ReleaseCallback&& release_callback,
-               SinkStream& sink_stream, std::string&& name_)
+Stream::Stream(Core::Timing::CoreTiming& core_timing, u32 sample_rate, Format format,
+               ReleaseCallback&& release_callback, SinkStream& sink_stream, std::string&& name_)
     : sample_rate{sample_rate}, format{format}, release_callback{std::move(release_callback)},
-      sink_stream{sink_stream}, name{std::move(name_)} {
+      sink_stream{sink_stream}, core_timing{core_timing}, name{std::move(name_)} {
 
-    release_event = CoreTiming::RegisterEvent(
+    release_event = core_timing.RegisterEvent(
         name, [this](u64 userdata, int cycles_late) { ReleaseActiveBuffer(); });
 }
 
@@ -57,7 +57,7 @@ Stream::State Stream::GetState() const {
 
 s64 Stream::GetBufferReleaseCycles(const Buffer& buffer) const {
     const std::size_t num_samples{buffer.GetSamples().size() / GetNumChannels()};
-    return CoreTiming::usToCycles((static_cast<u64>(num_samples) * 1000000) / sample_rate);
+    return Core::Timing::usToCycles((static_cast<u64>(num_samples) * 1000000) / sample_rate);
 }
 
 static void VolumeAdjustSamples(std::vector<s16>& samples) {
@@ -95,11 +95,11 @@ void Stream::PlayNextBuffer() {
     active_buffer = queued_buffers.front();
     queued_buffers.pop();
 
-    VolumeAdjustSamples(active_buffer->Samples());
+    VolumeAdjustSamples(active_buffer->GetSamples());
 
     sink_stream.EnqueueSamples(GetNumChannels(), active_buffer->GetSamples());
 
-    CoreTiming::ScheduleEventThreadsafe(GetBufferReleaseCycles(*active_buffer), release_event, {});
+    core_timing.ScheduleEventThreadsafe(GetBufferReleaseCycles(*active_buffer), release_event, {});
 }
 
 void Stream::ReleaseActiveBuffer() {

src/audio_core/stream.h

@@ -13,9 +13,10 @@
 #include "audio_core/buffer.h"
 #include "common/common_types.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
+class CoreTiming;
 struct EventType;
-}
+} // namespace Core::Timing
 
 namespace AudioCore {
 
@@ -42,8 +43,8 @@ public:
     /// Callback function type, used to change guest state on a buffer being released
     using ReleaseCallback = std::function<void()>;
 
-    Stream(u32 sample_rate, Format format, ReleaseCallback&& release_callback,
-           SinkStream& sink_stream, std::string&& name_);
+    Stream(Core::Timing::CoreTiming& core_timing, u32 sample_rate, Format format,
+           ReleaseCallback&& release_callback, SinkStream& sink_stream, std::string&& name_);
 
     /// Plays the audio stream
     void Play();
@@ -91,16 +92,17 @@ private:
     /// Gets the number of core cycles when the specified buffer will be released
     s64 GetBufferReleaseCycles(const Buffer& buffer) const;
 
-    u32 sample_rate;                        ///< Sample rate of the stream
-    Format format;                          ///< Format of the stream
-    ReleaseCallback release_callback;       ///< Buffer release callback for the stream
-    State state{State::Stopped};            ///< Playback state of the stream
-    CoreTiming::EventType* release_event{}; ///< Core timing release event for the stream
-    BufferPtr active_buffer;                ///< Actively playing buffer in the stream
-    std::queue<BufferPtr> queued_buffers;   ///< Buffers queued to be played in the stream
-    std::queue<BufferPtr> released_buffers; ///< Buffers recently released from the stream
-    SinkStream& sink_stream;                ///< Output sink for the stream
-    std::string name;                       ///< Name of the stream, must be unique
+    u32 sample_rate;                          ///< Sample rate of the stream
+    Format format;                            ///< Format of the stream
+    ReleaseCallback release_callback;         ///< Buffer release callback for the stream
+    State state{State::Stopped};              ///< Playback state of the stream
+    Core::Timing::EventType* release_event{}; ///< Core timing release event for the stream
+    BufferPtr active_buffer;                  ///< Actively playing buffer in the stream
+    std::queue<BufferPtr> queued_buffers;     ///< Buffers queued to be played in the stream
+    std::queue<BufferPtr> released_buffers;   ///< Buffers recently released from the stream
+    SinkStream& sink_stream;                  ///< Output sink for the stream
+    Core::Timing::CoreTiming& core_timing;    ///< Core timing instance.
+    std::string name;                         ///< Name of the stream, must be unique
 };
 
 using StreamPtr = std::shared_ptr<Stream>;

src/common/logging/backend.cpp

@@ -40,9 +40,7 @@ public:
     const Impl& operator=(Impl const&) = delete;
 
     void PushEntry(Entry e) {
-        std::lock_guard<std::mutex> lock(message_mutex);
         message_queue.Push(std::move(e));
-        message_cv.notify_one();
     }
 
     void AddBackend(std::unique_ptr<Backend> backend) {
@@ -86,15 +84,13 @@ private:
                 }
             };
             while (true) {
-                {
-                    std::unique_lock<std::mutex> lock(message_mutex);
-                    message_cv.wait(lock, [&] { return !running || message_queue.Pop(entry); });
-                }
-                if (!running) {
+                entry = message_queue.PopWait();
+                if (entry.final_entry) {
                     break;
                 }
                 write_logs(entry);
             }
+
             // Drain the logging queue. Only writes out up to MAX_LOGS_TO_WRITE to prevent a case
             // where a system is repeatedly spamming logs even on close.
             const int MAX_LOGS_TO_WRITE = filter.IsDebug() ? INT_MAX : 100;
@@ -106,14 +102,13 @@ private:
     }
 
     ~Impl() {
-        running = false;
-        message_cv.notify_one();
+        Entry entry;
+        entry.final_entry = true;
+        message_queue.Push(entry);
         backend_thread.join();
     }
 
-    std::atomic_bool running{true};
-    std::mutex message_mutex, writing_mutex;
-    std::condition_variable message_cv;
+    std::mutex writing_mutex;
     std::thread backend_thread;
     std::vector<std::unique_ptr<Backend>> backends;
     Common::MPSCQueue<Log::Entry> message_queue;
@@ -232,6 +227,7 @@ void DebuggerBackend::Write(const Entry& entry) {
     CLS(Render)                                                                                    \
     SUB(Render, Software)                                                                          \
     SUB(Render, OpenGL)                                                                            \
+    SUB(Render, Vulkan)                                                                            \
     CLS(Audio)                                                                                     \
     SUB(Audio, DSP)                                                                                \
     SUB(Audio, Sink)                                                                               \

src/common/logging/backend.h

@@ -27,6 +27,7 @@ struct Entry {
     unsigned int line_num;
     std::string function;
     std::string message;
+    bool final_entry = false;
 
     Entry() = default;
     Entry(Entry&& o) = default;

src/common/logging/log.h

@@ -112,6 +112,7 @@ enum class Class : ClassType {
     Render,            ///< Emulator video output and hardware acceleration
     Render_Software,   ///< Software renderer backend
     Render_OpenGL,     ///< OpenGL backend
+    Render_Vulkan,     ///< Vulkan backend
     Audio,             ///< Audio emulation
     Audio_DSP,         ///< The HLE implementation of the DSP
     Audio_Sink,        ///< Emulator audio output backend

src/common/threadsafe_queue.h

@@ -7,17 +7,17 @@
 // a simple lockless thread-safe,
 // single reader, single writer queue
 
-#include <algorithm>
 #include <atomic>
+#include <condition_variable>
 #include <cstddef>
 #include <mutex>
-#include "common/common_types.h"
+#include <utility>
 
 namespace Common {
-template <typename T, bool NeedSize = true>
+template <typename T>
 class SPSCQueue {
 public:
-    SPSCQueue() : size(0) {
+    SPSCQueue() {
         write_ptr = read_ptr = new ElementPtr();
     }
     ~SPSCQueue() {
@@ -25,13 +25,12 @@ public:
         delete read_ptr;
     }
 
-    u32 Size() const {
-        static_assert(NeedSize, "using Size() on FifoQueue without NeedSize");
+    std::size_t Size() const {
         return size.load();
     }
 
     bool Empty() const {
-        return !read_ptr->next.load();
+        return Size() == 0;
     }
 
     T& Front() const {
@@ -47,13 +46,14 @@ public:
         ElementPtr* new_ptr = new ElementPtr();
         write_ptr->next.store(new_ptr, std::memory_order_release);
         write_ptr = new_ptr;
-        if (NeedSize)
-            size++;
+        cv.notify_one();
+
+        ++size;
     }
 
     void Pop() {
-        if (NeedSize)
-            size--;
+        --size;
+
         ElementPtr* tmpptr = read_ptr;
         // advance the read pointer
         read_ptr = tmpptr->next.load();
@@ -66,8 +66,7 @@ public:
         if (Empty())
             return false;
 
-        if (NeedSize)
-            size--;
+        --size;
 
         ElementPtr* tmpptr = read_ptr;
         read_ptr = tmpptr->next.load(std::memory_order_acquire);
@@ -77,6 +76,16 @@ public:
         return true;
     }
 
+    T PopWait() {
+        if (Empty()) {
+            std::unique_lock<std::mutex> lock(cv_mutex);
+            cv.wait(lock, [this]() { return !Empty(); });
+        }
+        T t;
+        Pop(t);
+        return t;
+    }
+
     // not thread-safe
     void Clear() {
         size.store(0);
@@ -89,7 +98,7 @@ private:
     // and a pointer to the next ElementPtr
     class ElementPtr {
     public:
-        ElementPtr() : next(nullptr) {}
+        ElementPtr() {}
         ~ElementPtr() {
             ElementPtr* next_ptr = next.load();
 
@@ -98,21 +107,23 @@ private:
         }
 
         T current;
-        std::atomic<ElementPtr*> next;
+        std::atomic<ElementPtr*> next{nullptr};
     };
 
     ElementPtr* write_ptr;
     ElementPtr* read_ptr;
-    std::atomic<u32> size;
+    std::atomic_size_t size{0};
+    std::mutex cv_mutex;
+    std::condition_variable cv;
 };
 
 // a simple thread-safe,
 // single reader, multiple writer queue
 
-template <typename T, bool NeedSize = true>
+template <typename T>
 class MPSCQueue {
 public:
-    u32 Size() const {
+    std::size_t Size() const {
         return spsc_queue.Size();
     }
 
@@ -138,13 +149,17 @@ public:
         return spsc_queue.Pop(t);
     }
 
+    T PopWait() {
+        return spsc_queue.PopWait();
+    }
+
     // not thread-safe
     void Clear() {
         spsc_queue.Clear();
     }
 
 private:
-    SPSCQueue<T, NeedSize> spsc_queue;
+    SPSCQueue<T> spsc_queue;
     std::mutex write_lock;
 };
 } // namespace Common

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -112,14 +112,14 @@ public:
         // Always execute at least one tick.
         amortized_ticks = std::max<u64>(amortized_ticks, 1);
 
-        CoreTiming::AddTicks(amortized_ticks);
+        parent.core_timing.AddTicks(amortized_ticks);
         num_interpreted_instructions = 0;
     }
     u64 GetTicksRemaining() override {
-        return std::max(CoreTiming::GetDowncount(), 0);
+        return std::max(parent.core_timing.GetDowncount(), 0);
     }
     u64 GetCNTPCT() override {
-        return CoreTiming::GetTicks();
+        return parent.core_timing.GetTicks();
     }
 
     ARM_Dynarmic& parent;
@@ -172,8 +172,10 @@ void ARM_Dynarmic::Step() {
     cb->InterpreterFallback(jit->GetPC(), 1);
 }
 
-ARM_Dynarmic::ARM_Dynarmic(ExclusiveMonitor& exclusive_monitor, std::size_t core_index)
-    : cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), core_index{core_index},
+ARM_Dynarmic::ARM_Dynarmic(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
+                           std::size_t core_index)
+    : cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), inner_unicorn{core_timing},
+      core_index{core_index}, core_timing{core_timing},
       exclusive_monitor{dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {
     ThreadContext ctx{};
     inner_unicorn.SaveContext(ctx);

src/core/arm/dynarmic/arm_dynarmic.h

@@ -16,6 +16,10 @@ namespace Memory {
 struct PageTable;
 }
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Core {
 
 class ARM_Dynarmic_Callbacks;
@@ -23,7 +27,8 @@ class DynarmicExclusiveMonitor;
 
 class ARM_Dynarmic final : public ARM_Interface {
 public:
-    ARM_Dynarmic(ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ARM_Dynarmic(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
+                 std::size_t core_index);
     ~ARM_Dynarmic();
 
     void MapBackingMemory(VAddr address, std::size_t size, u8* memory,
@@ -62,6 +67,7 @@ private:
     ARM_Unicorn inner_unicorn;
 
     std::size_t core_index;
+    Timing::CoreTiming& core_timing;
     DynarmicExclusiveMonitor& exclusive_monitor;
 
     Memory::PageTable* current_page_table = nullptr;

src/core/arm/unicorn/arm_unicorn.cpp

@@ -72,7 +72,7 @@ static bool UnmappedMemoryHook(uc_engine* uc, uc_mem_type type, u64 addr, int si
     return {};
 }
 
-ARM_Unicorn::ARM_Unicorn() {
+ARM_Unicorn::ARM_Unicorn(Timing::CoreTiming& core_timing) : core_timing{core_timing} {
     CHECKED(uc_open(UC_ARCH_ARM64, UC_MODE_ARM, &uc));
 
     auto fpv = 3 << 20;
@@ -177,7 +177,7 @@ void ARM_Unicorn::Run() {
     if (GDBStub::IsServerEnabled()) {
         ExecuteInstructions(std::max(4000000, 0));
     } else {
-        ExecuteInstructions(std::max(CoreTiming::GetDowncount(), 0));
+        ExecuteInstructions(std::max(core_timing.GetDowncount(), 0));
     }
 }
 
@@ -190,7 +190,7 @@ MICROPROFILE_DEFINE(ARM_Jit_Unicorn, "ARM JIT", "Unicorn", MP_RGB(255, 64, 64));
 void ARM_Unicorn::ExecuteInstructions(int num_instructions) {
     MICROPROFILE_SCOPE(ARM_Jit_Unicorn);
     CHECKED(uc_emu_start(uc, GetPC(), 1ULL << 63, 0, num_instructions));
-    CoreTiming::AddTicks(num_instructions);
+    core_timing.AddTicks(num_instructions);
     if (GDBStub::IsServerEnabled()) {
         if (last_bkpt_hit) {
             uc_reg_write(uc, UC_ARM64_REG_PC, &last_bkpt.address);

src/core/arm/unicorn/arm_unicorn.h

@@ -9,12 +9,17 @@
 #include "core/arm/arm_interface.h"
 #include "core/gdbstub/gdbstub.h"
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Core {
 
 class ARM_Unicorn final : public ARM_Interface {
 public:
-    ARM_Unicorn();
+    explicit ARM_Unicorn(Timing::CoreTiming& core_timing);
     ~ARM_Unicorn();
+
     void MapBackingMemory(VAddr address, std::size_t size, u8* memory,
                           Kernel::VMAPermission perms) override;
     void UnmapMemory(VAddr address, std::size_t size) override;
@@ -43,6 +48,7 @@ public:
 
 private:
     uc_engine* uc{};
+    Timing::CoreTiming& core_timing;
     GDBStub::BreakpointAddress last_bkpt{};
     bool last_bkpt_hit;
 };

src/core/core.cpp

@@ -94,8 +94,8 @@ struct System::Impl {
     ResultStatus Init(System& system, Frontend::EmuWindow& emu_window) {
         LOG_DEBUG(HW_Memory, "initialized OK");
 
-        CoreTiming::Init();
-        kernel.Initialize();
+        core_timing.Initialize();
+        kernel.Initialize(core_timing);
 
         const auto current_time = std::chrono::duration_cast<std::chrono::seconds>(
             std::chrono::system_clock::now().time_since_epoch());
@@ -120,7 +120,7 @@ struct System::Impl {
         telemetry_session = std::make_unique<Core::TelemetrySession>();
         service_manager = std::make_shared<Service::SM::ServiceManager>();
 
-        Service::Init(service_manager, *virtual_filesystem);
+        Service::Init(service_manager, system, *virtual_filesystem);
         GDBStub::Init();
 
         renderer = VideoCore::CreateRenderer(emu_window, system);
@@ -205,7 +205,7 @@ struct System::Impl {
 
         // Shutdown kernel and core timing
         kernel.Shutdown();
-        CoreTiming::Shutdown();
+        core_timing.Shutdown();
 
         // Close app loader
         app_loader.reset();
@@ -232,9 +232,10 @@ struct System::Impl {
     }
 
     PerfStatsResults GetAndResetPerfStats() {
-        return perf_stats.GetAndResetStats(CoreTiming::GetGlobalTimeUs());
+        return perf_stats.GetAndResetStats(core_timing.GetGlobalTimeUs());
     }
 
+    Timing::CoreTiming core_timing;
     Kernel::KernelCore kernel;
     /// RealVfsFilesystem instance
     FileSys::VirtualFilesystem virtual_filesystem;
@@ -396,6 +397,14 @@ const Kernel::KernelCore& System::Kernel() const {
     return impl->kernel;
 }
 
+Timing::CoreTiming& System::CoreTiming() {
+    return impl->core_timing;
+}
+
+const Timing::CoreTiming& System::CoreTiming() const {
+    return impl->core_timing;
+}
+
 Core::PerfStats& System::GetPerfStats() {
     return impl->perf_stats;
 }

src/core/core.h

@@ -47,6 +47,10 @@ namespace VideoCore {
 class RendererBase;
 } // namespace VideoCore
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Core {
 
 class ARM_Interface;
@@ -205,6 +209,12 @@ public:
     /// Provides a constant pointer to the current process.
     const Kernel::Process* CurrentProcess() const;
 
+    /// Provides a reference to the core timing instance.
+    Timing::CoreTiming& CoreTiming();
+
+    /// Provides a constant reference to the core timing instance.
+    const Timing::CoreTiming& CoreTiming() const;
+
     /// Provides a reference to the kernel instance.
     Kernel::KernelCore& Kernel();
 

src/core/core_cpu.cpp

@@ -49,17 +49,18 @@ bool CpuBarrier::Rendezvous() {
     return false;
 }
 
-Cpu::Cpu(ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier, std::size_t core_index)
-    : cpu_barrier{cpu_barrier}, core_index{core_index} {
+Cpu::Cpu(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
+         CpuBarrier& cpu_barrier, std::size_t core_index)
+    : cpu_barrier{cpu_barrier}, core_timing{core_timing}, core_index{core_index} {
     if (Settings::values.use_cpu_jit) {
 #ifdef ARCHITECTURE_x86_64
-        arm_interface = std::make_unique<ARM_Dynarmic>(exclusive_monitor, core_index);
+        arm_interface = std::make_unique<ARM_Dynarmic>(core_timing, exclusive_monitor, core_index);
 #else
         arm_interface = std::make_unique<ARM_Unicorn>();
         LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
     } else {
-        arm_interface = std::make_unique<ARM_Unicorn>();
+        arm_interface = std::make_unique<ARM_Unicorn>(core_timing);
     }
 
     scheduler = std::make_unique<Kernel::Scheduler>(*arm_interface);
@@ -93,14 +94,14 @@ void Cpu::RunLoop(bool tight_loop) {
 
         if (IsMainCore()) {
             // TODO(Subv): Only let CoreTiming idle if all 4 cores are idling.
-            CoreTiming::Idle();
-            CoreTiming::Advance();
+            core_timing.Idle();
+            core_timing.Advance();
         }
 
         PrepareReschedule();
     } else {
         if (IsMainCore()) {
-            CoreTiming::Advance();
+            core_timing.Advance();
         }
 
         if (tight_loop) {

src/core/core_cpu.h

@@ -15,6 +15,10 @@ namespace Kernel {
 class Scheduler;
 }
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Core {
 
 class ARM_Interface;
@@ -41,7 +45,8 @@ private:
 
 class Cpu {
 public:
-    Cpu(ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier, std::size_t core_index);
+    Cpu(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
+        CpuBarrier& cpu_barrier, std::size_t core_index);
     ~Cpu();
 
     void RunLoop(bool tight_loop = true);
@@ -82,6 +87,7 @@ private:
     std::unique_ptr<ARM_Interface> arm_interface;
     CpuBarrier& cpu_barrier;
     std::unique_ptr<Kernel::Scheduler> scheduler;
+    Timing::CoreTiming& core_timing;
 
     std::atomic<bool> reschedule_pending = false;
     std::size_t core_index;

src/core/core_timing.cpp

@@ -8,71 +8,60 @@
 #include <mutex>
 #include <string>
 #include <tuple>
-#include <unordered_map>
-#include <vector>
+
 #include "common/assert.h"
 #include "common/thread.h"
-#include "common/threadsafe_queue.h"
 #include "core/core_timing_util.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
 
-static s64 global_timer;
-static int slice_length;
-static int downcount;
+constexpr int MAX_SLICE_LENGTH = 20000;
 
-struct EventType {
-    TimedCallback callback;
-    const std::string* name;
-};
-
-struct Event {
+struct CoreTiming::Event {
     s64 time;
     u64 fifo_order;
     u64 userdata;
     const EventType* type;
+
+    // Sort by time, unless the times are the same, in which case sort by
+    // the order added to the queue
+    friend bool operator>(const Event& left, const Event& right) {
+        return std::tie(left.time, left.fifo_order) > std::tie(right.time, right.fifo_order);
+    }
+
+    friend bool operator<(const Event& left, const Event& right) {
+        return std::tie(left.time, left.fifo_order) < std::tie(right.time, right.fifo_order);
+    }
 };
 
-// Sort by time, unless the times are the same, in which case sort by the order added to the queue
-static bool operator>(const Event& left, const Event& right) {
-    return std::tie(left.time, left.fifo_order) > std::tie(right.time, right.fifo_order);
+CoreTiming::CoreTiming() = default;
+CoreTiming::~CoreTiming() = default;
+
+void CoreTiming::Initialize() {
+    downcount = MAX_SLICE_LENGTH;
+    slice_length = MAX_SLICE_LENGTH;
+    global_timer = 0;
+    idled_cycles = 0;
+
+    // The time between CoreTiming being initialized and the first call to Advance() is considered
+    // the slice boundary between slice -1 and slice 0. Dispatcher loops must call Advance() before
+    // executing the first cycle of each slice to prepare the slice length and downcount for
+    // that slice.
+    is_global_timer_sane = true;
+
+    event_fifo_id = 0;
+
+    const auto empty_timed_callback = [](u64, s64) {};
+    ev_lost = RegisterEvent("_lost_event", empty_timed_callback);
 }
 
-static bool operator<(const Event& left, const Event& right) {
-    return std::tie(left.time, left.fifo_order) < std::tie(right.time, right.fifo_order);
+void CoreTiming::Shutdown() {
+    MoveEvents();
+    ClearPendingEvents();
+    UnregisterAllEvents();
 }
 
-// unordered_map stores each element separately as a linked list node so pointers to elements
-// remain stable regardless of rehashes/resizing.
-static std::unordered_map<std::string, EventType> event_types;
-
-// The queue is a min-heap using std::make_heap/push_heap/pop_heap.
-// We don't use std::priority_queue because we need to be able to serialize, unserialize and
-// erase arbitrary events (RemoveEvent()) regardless of the queue order. These aren't accomodated
-// by the standard adaptor class.
-static std::vector<Event> event_queue;
-static u64 event_fifo_id;
-// the queue for storing the events from other threads threadsafe until they will be added
-// to the event_queue by the emu thread
-static Common::MPSCQueue<Event, false> ts_queue;
-
-// the queue for unscheduling the events from other threads threadsafe
-static Common::MPSCQueue<std::pair<const EventType*, u64>, false> unschedule_queue;
-
-constexpr int MAX_SLICE_LENGTH = 20000;
-
-static s64 idled_cycles;
-
-// Are we in a function that has been called from Advance()
-// If events are sheduled from a function that gets called from Advance(),
-// don't change slice_length and downcount.
-static bool is_global_timer_sane;
-
-static EventType* ev_lost = nullptr;
-
-static void EmptyTimedCallback(u64 userdata, s64 cyclesLate) {}
-
-EventType* RegisterEvent(const std::string& name, TimedCallback callback) {
+EventType* CoreTiming::RegisterEvent(const std::string& name, TimedCallback callback) {
     // check for existing type with same name.
     // we want event type names to remain unique so that we can use them for serialization.
     ASSERT_MSG(event_types.find(name) == event_types.end(),
@@ -86,71 +75,31 @@ EventType* RegisterEvent(const std::string& name, TimedCallback callback) {
     return event_type;
 }
 
-void UnregisterAllEvents() {
+void CoreTiming::UnregisterAllEvents() {
     ASSERT_MSG(event_queue.empty(), "Cannot unregister events with events pending");
     event_types.clear();
 }
 
-void Init() {
-    downcount = MAX_SLICE_LENGTH;
-    slice_length = MAX_SLICE_LENGTH;
-    global_timer = 0;
-    idled_cycles = 0;
-
-    // The time between CoreTiming being intialized and the first call to Advance() is considered
-    // the slice boundary between slice -1 and slice 0. Dispatcher loops must call Advance() before
-    // executing the first cycle of each slice to prepare the slice length and downcount for
-    // that slice.
-    is_global_timer_sane = true;
-
-    event_fifo_id = 0;
-    ev_lost = RegisterEvent("_lost_event", &EmptyTimedCallback);
-}
-
-void Shutdown() {
-    MoveEvents();
-    ClearPendingEvents();
-    UnregisterAllEvents();
-}
-
-// This should only be called from the CPU thread. If you are calling
-// it from any other thread, you are doing something evil
-u64 GetTicks() {
-    u64 ticks = static_cast<u64>(global_timer);
-    if (!is_global_timer_sane) {
-        ticks += slice_length - downcount;
-    }
-    return ticks;
-}
-
-void AddTicks(u64 ticks) {
-    downcount -= static_cast<int>(ticks);
-}
-
-u64 GetIdleTicks() {
-    return static_cast<u64>(idled_cycles);
-}
-
-void ClearPendingEvents() {
-    event_queue.clear();
-}
-
-void ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata) {
+void CoreTiming::ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata) {
     ASSERT(event_type != nullptr);
-    s64 timeout = GetTicks() + cycles_into_future;
+    const s64 timeout = GetTicks() + cycles_into_future;
+
     // If this event needs to be scheduled before the next advance(), force one early
-    if (!is_global_timer_sane)
+    if (!is_global_timer_sane) {
         ForceExceptionCheck(cycles_into_future);
+    }
+
     event_queue.emplace_back(Event{timeout, event_fifo_id++, userdata, event_type});
     std::push_heap(event_queue.begin(), event_queue.end(), std::greater<>());
 }
 
-void ScheduleEventThreadsafe(s64 cycles_into_future, const EventType* event_type, u64 userdata) {
+void CoreTiming::ScheduleEventThreadsafe(s64 cycles_into_future, const EventType* event_type,
+                                         u64 userdata) {
     ts_queue.Push(Event{global_timer + cycles_into_future, 0, userdata, event_type});
 }
 
-void UnscheduleEvent(const EventType* event_type, u64 userdata) {
-    auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
+void CoreTiming::UnscheduleEvent(const EventType* event_type, u64 userdata) {
+    const auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
         return e.type == event_type && e.userdata == userdata;
     });
 
@@ -161,13 +110,33 @@ void UnscheduleEvent(const EventType* event_type, u64 userdata) {
     }
 }
 
-void UnscheduleEventThreadsafe(const EventType* event_type, u64 userdata) {
+void CoreTiming::UnscheduleEventThreadsafe(const EventType* event_type, u64 userdata) {
     unschedule_queue.Push(std::make_pair(event_type, userdata));
 }
 
-void RemoveEvent(const EventType* event_type) {
-    auto itr = std::remove_if(event_queue.begin(), event_queue.end(),
-                              [&](const Event& e) { return e.type == event_type; });
+u64 CoreTiming::GetTicks() const {
+    u64 ticks = static_cast<u64>(global_timer);
+    if (!is_global_timer_sane) {
+        ticks += slice_length - downcount;
+    }
+    return ticks;
+}
+
+u64 CoreTiming::GetIdleTicks() const {
+    return static_cast<u64>(idled_cycles);
+}
+
+void CoreTiming::AddTicks(u64 ticks) {
+    downcount -= static_cast<int>(ticks);
+}
+
+void CoreTiming::ClearPendingEvents() {
+    event_queue.clear();
+}
+
+void CoreTiming::RemoveEvent(const EventType* event_type) {
+    const auto itr = std::remove_if(event_queue.begin(), event_queue.end(),
+                                    [&](const Event& e) { return e.type == event_type; });
 
     // Removing random items breaks the invariant so we have to re-establish it.
     if (itr != event_queue.end()) {
@@ -176,22 +145,24 @@ void RemoveEvent(const EventType* event_type) {
     }
 }
 
-void RemoveNormalAndThreadsafeEvent(const EventType* event_type) {
+void CoreTiming::RemoveNormalAndThreadsafeEvent(const EventType* event_type) {
     MoveEvents();
     RemoveEvent(event_type);
 }
 
-void ForceExceptionCheck(s64 cycles) {
+void CoreTiming::ForceExceptionCheck(s64 cycles) {
     cycles = std::max<s64>(0, cycles);
-    if (downcount > cycles) {
-        // downcount is always (much) smaller than MAX_INT so we can safely cast cycles to an int
-        // here. Account for cycles already executed by adjusting the g.slice_length
-        slice_length -= downcount - static_cast<int>(cycles);
-        downcount = static_cast<int>(cycles);
+    if (downcount <= cycles) {
+        return;
     }
+
+    // downcount is always (much) smaller than MAX_INT so we can safely cast cycles to an int
+    // here. Account for cycles already executed by adjusting the g.slice_length
+    slice_length -= downcount - static_cast<int>(cycles);
+    downcount = static_cast<int>(cycles);
 }
 
-void MoveEvents() {
+void CoreTiming::MoveEvents() {
     for (Event ev; ts_queue.Pop(ev);) {
         ev.fifo_order = event_fifo_id++;
         event_queue.emplace_back(std::move(ev));
@@ -199,13 +170,13 @@ void MoveEvents() {
     }
 }
 
-void Advance() {
+void CoreTiming::Advance() {
     MoveEvents();
     for (std::pair<const EventType*, u64> ev; unschedule_queue.Pop(ev);) {
         UnscheduleEvent(ev.first, ev.second);
     }
 
-    int cycles_executed = slice_length - downcount;
+    const int cycles_executed = slice_length - downcount;
     global_timer += cycles_executed;
     slice_length = MAX_SLICE_LENGTH;
 
@@ -229,17 +200,17 @@ void Advance() {
     downcount = slice_length;
 }
 
-void Idle() {
+void CoreTiming::Idle() {
     idled_cycles += downcount;
     downcount = 0;
 }
 
-std::chrono::microseconds GetGlobalTimeUs() {
+std::chrono::microseconds CoreTiming::GetGlobalTimeUs() const {
     return std::chrono::microseconds{GetTicks() * 1000000 / BASE_CLOCK_RATE};
 }
 
-int GetDowncount() {
+int CoreTiming::GetDowncount() const {
     return downcount;
 }
 
-} // namespace CoreTiming
+} // namespace Core::Timing

src/core/core_timing.h

@@ -4,6 +4,27 @@
 
 #pragma once
 
+#include <chrono>
+#include <functional>
+#include <string>
+#include <unordered_map>
+#include <vector>
+#include "common/common_types.h"
+#include "common/threadsafe_queue.h"
+
+namespace Core::Timing {
+
+/// A callback that may be scheduled for a particular core timing event.
+using TimedCallback = std::function<void(u64 userdata, int cycles_late)>;
+
+/// Contains the characteristics of a particular event.
+struct EventType {
+    /// The event's callback function.
+    TimedCallback callback;
+    /// A pointer to the name of the event.
+    const std::string* name;
+};
+
 /**
  * This is a system to schedule events into the emulated machine's future. Time is measured
  * in main CPU clock cycles.
@@ -16,80 +37,120 @@
  * inside callback:
  *   ScheduleEvent(periodInCycles - cyclesLate, callback, "whatever")
  */
+class CoreTiming {
+public:
+    CoreTiming();
+    ~CoreTiming();
 
-#include <chrono>
-#include <functional>
-#include <string>
-#include "common/common_types.h"
+    CoreTiming(const CoreTiming&) = delete;
+    CoreTiming(CoreTiming&&) = delete;
 
-namespace CoreTiming {
+    CoreTiming& operator=(const CoreTiming&) = delete;
+    CoreTiming& operator=(CoreTiming&&) = delete;
 
-struct EventType;
+    /// CoreTiming begins at the boundary of timing slice -1. An initial call to Advance() is
+    /// required to end slice - 1 and start slice 0 before the first cycle of code is executed.
+    void Initialize();
 
-using TimedCallback = std::function<void(u64 userdata, int cycles_late)>;
+    /// Tears down all timing related functionality.
+    void Shutdown();
 
-/**
- * CoreTiming begins at the boundary of timing slice -1. An initial call to Advance() is
- * required to end slice -1 and start slice 0 before the first cycle of code is executed.
- */
-void Init();
-void Shutdown();
+    /// Registers a core timing event with the given name and callback.
+    ///
+    /// @param name     The name of the core timing event to register.
+    /// @param callback The callback to execute for the event.
+    ///
+    /// @returns An EventType instance representing the registered event.
+    ///
+    /// @pre The name of the event being registered must be unique among all
+    ///      registered events.
+    ///
+    EventType* RegisterEvent(const std::string& name, TimedCallback callback);
 
-/**
- * This should only be called from the emu thread, if you are calling it any other thread, you are
- * doing something evil
- */
-u64 GetTicks();
-u64 GetIdleTicks();
-void AddTicks(u64 ticks);
+    /// Unregisters all registered events thus far.
+    void UnregisterAllEvents();
 
-/**
- * Returns the event_type identifier. if name is not unique, it will assert.
- */
-EventType* RegisterEvent(const std::string& name, TimedCallback callback);
-void UnregisterAllEvents();
+    /// After the first Advance, the slice lengths and the downcount will be reduced whenever an
+    /// event is scheduled earlier than the current values.
+    ///
+    /// Scheduling from a callback will not update the downcount until the Advance() completes.
+    void ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata = 0);
 
-/**
- * After the first Advance, the slice lengths and the downcount will be reduced whenever an event
- * is scheduled earlier than the current values.
- * Scheduling from a callback will not update the downcount until the Advance() completes.
- */
-void ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata = 0);
+    /// This is to be called when outside of hle threads, such as the graphics thread, wants to
+    /// schedule things to be executed on the main thread.
+    ///
+    /// @note This doesn't change slice_length and thus events scheduled by this might be
+    /// called with a delay of up to MAX_SLICE_LENGTH
+    void ScheduleEventThreadsafe(s64 cycles_into_future, const EventType* event_type,
+                                 u64 userdata = 0);
 
-/**
- * This is to be called when outside of hle threads, such as the graphics thread, wants to
- * schedule things to be executed on the main thread.
- * Not that this doesn't change slice_length and thus events scheduled by this might be called
- * with a delay of up to MAX_SLICE_LENGTH
- */
-void ScheduleEventThreadsafe(s64 cycles_into_future, const EventType* event_type, u64 userdata);
+    void UnscheduleEvent(const EventType* event_type, u64 userdata);
+    void UnscheduleEventThreadsafe(const EventType* event_type, u64 userdata);
 
-void UnscheduleEvent(const EventType* event_type, u64 userdata);
-void UnscheduleEventThreadsafe(const EventType* event_type, u64 userdata);
+    /// We only permit one event of each type in the queue at a time.
+    void RemoveEvent(const EventType* event_type);
+    void RemoveNormalAndThreadsafeEvent(const EventType* event_type);
 
-/// We only permit one event of each type in the queue at a time.
-void RemoveEvent(const EventType* event_type);
-void RemoveNormalAndThreadsafeEvent(const EventType* event_type);
+    void ForceExceptionCheck(s64 cycles);
 
-/** Advance must be called at the beginning of dispatcher loops, not the end. Advance() ends
- * the previous timing slice and begins the next one, you must Advance from the previous
- * slice to the current one before executing any cycles. CoreTiming starts in slice -1 so an
- * Advance() is required to initialize the slice length before the first cycle of emulated
- * instructions is executed.
- */
-void Advance();
-void MoveEvents();
+    /// This should only be called from the emu thread, if you are calling it any other thread,
+    /// you are doing something evil
+    u64 GetTicks() const;
 
-/// Pretend that the main CPU has executed enough cycles to reach the next event.
-void Idle();
+    u64 GetIdleTicks() const;
 
-/// Clear all pending events. This should ONLY be done on exit.
-void ClearPendingEvents();
+    void AddTicks(u64 ticks);
 
-void ForceExceptionCheck(s64 cycles);
+    /// Advance must be called at the beginning of dispatcher loops, not the end. Advance() ends
+    /// the previous timing slice and begins the next one, you must Advance from the previous
+    /// slice to the current one before executing any cycles. CoreTiming starts in slice -1 so an
+    /// Advance() is required to initialize the slice length before the first cycle of emulated
+    /// instructions is executed.
+    void Advance();
 
-std::chrono::microseconds GetGlobalTimeUs();
+    /// Pretend that the main CPU has executed enough cycles to reach the next event.
+    void Idle();
 
-int GetDowncount();
+    std::chrono::microseconds GetGlobalTimeUs() const;
 
-} // namespace CoreTiming
+    int GetDowncount() const;
+
+private:
+    struct Event;
+
+    /// Clear all pending events. This should ONLY be done on exit.
+    void ClearPendingEvents();
+    void MoveEvents();
+
+    s64 global_timer = 0;
+    s64 idled_cycles = 0;
+    int slice_length = 0;
+    int downcount = 0;
+
+    // Are we in a function that has been called from Advance()
+    // If events are scheduled from a function that gets called from Advance(),
+    // don't change slice_length and downcount.
+    bool is_global_timer_sane = false;
+
+    // The queue is a min-heap using std::make_heap/push_heap/pop_heap.
+    // We don't use std::priority_queue because we need to be able to serialize, unserialize and
+    // erase arbitrary events (RemoveEvent()) regardless of the queue order. These aren't
+    // accomodated by the standard adaptor class.
+    std::vector<Event> event_queue;
+    u64 event_fifo_id = 0;
+
+    // Stores each element separately as a linked list node so pointers to elements
+    // remain stable regardless of rehashes/resizing.
+    std::unordered_map<std::string, EventType> event_types;
+
+    // The queue for storing the events from other threads threadsafe until they will be added
+    // to the event_queue by the emu thread
+    Common::MPSCQueue<Event> ts_queue;
+
+    // The queue for unscheduling the events from other threads threadsafe
+    Common::MPSCQueue<std::pair<const EventType*, u64>> unschedule_queue;
+
+    EventType* ev_lost = nullptr;
+};
+
+} // namespace Core::Timing

src/core/core_timing_util.cpp

@@ -8,7 +8,7 @@
 #include <limits>
 #include "common/logging/log.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
 
 constexpr u64 MAX_VALUE_TO_MULTIPLY = std::numeric_limits<s64>::max() / BASE_CLOCK_RATE;
 
@@ -60,4 +60,4 @@ s64 nsToCycles(u64 ns) {
     return (BASE_CLOCK_RATE * static_cast<s64>(ns)) / 1000000000;
 }
 
-} // namespace CoreTiming
+} // namespace Core::Timing

src/core/core_timing_util.h

@@ -6,7 +6,7 @@
 
 #include "common/common_types.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
 
 // The below clock rate is based on Switch's clockspeed being widely known as 1.020GHz
 // The exact value used is of course unverified.
@@ -61,4 +61,4 @@ inline u64 cyclesToMs(s64 cycles) {
     return cycles * 1000 / BASE_CLOCK_RATE;
 }
 
-} // namespace CoreTiming
+} // namespace Core::Timing

src/core/cpu_core_manager.cpp

@@ -27,7 +27,8 @@ void CpuCoreManager::Initialize(System& system) {
     exclusive_monitor = Cpu::MakeExclusiveMonitor(cores.size());
 
     for (std::size_t index = 0; index < cores.size(); ++index) {
-        cores[index] = std::make_unique<Cpu>(*exclusive_monitor, *barrier, index);
+        cores[index] =
+            std::make_unique<Cpu>(system.CoreTiming(), *exclusive_monitor, *barrier, index);
     }
 
     // Create threads for CPU cores 1-3, and build thread_to_cpu map

src/core/hle/kernel/address_arbiter.cpp

@@ -17,8 +17,7 @@
 #include "core/hle/result.h"
 #include "core/memory.h"
 
-namespace Kernel {
-namespace AddressArbiter {
+namespace Kernel::AddressArbiter {
 
 // Performs actual address waiting logic.
 static ResultCode WaitForAddress(VAddr address, s64 timeout) {
@@ -176,5 +175,4 @@ ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout) {
 
     return WaitForAddress(address, timeout);
 }
-} // namespace AddressArbiter
-} // namespace Kernel
+} // namespace Kernel::AddressArbiter

src/core/hle/kernel/address_arbiter.h

@@ -8,9 +8,8 @@
 
 union ResultCode;
 
-namespace Kernel {
+namespace Kernel::AddressArbiter {
 
-namespace AddressArbiter {
 enum class ArbitrationType {
     WaitIfLessThan = 0,
     DecrementAndWaitIfLessThan = 1,
@@ -29,6 +28,5 @@ ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 valu
 
 ResultCode WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout, bool should_decrement);
 ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout);
-} // namespace AddressArbiter
 
-} // namespace Kernel
+} // namespace Kernel::AddressArbiter

src/core/hle/kernel/kernel.cpp

@@ -86,11 +86,11 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] int cycles_
 }
 
 struct KernelCore::Impl {
-    void Initialize(KernelCore& kernel) {
+    void Initialize(KernelCore& kernel, Core::Timing::CoreTiming& core_timing) {
         Shutdown();
 
         InitializeSystemResourceLimit(kernel);
-        InitializeThreads();
+        InitializeThreads(core_timing);
     }
 
     void Shutdown() {
@@ -122,9 +122,9 @@ struct KernelCore::Impl {
         ASSERT(system_resource_limit->SetLimitValue(ResourceType::Sessions, 900).IsSuccess());
     }
 
-    void InitializeThreads() {
+    void InitializeThreads(Core::Timing::CoreTiming& core_timing) {
         thread_wakeup_event_type =
-            CoreTiming::RegisterEvent("ThreadWakeupCallback", ThreadWakeupCallback);
+            core_timing.RegisterEvent("ThreadWakeupCallback", ThreadWakeupCallback);
     }
 
     std::atomic<u32> next_object_id{0};
@@ -137,7 +137,7 @@ struct KernelCore::Impl {
 
     SharedPtr<ResourceLimit> system_resource_limit;
 
-    CoreTiming::EventType* thread_wakeup_event_type = nullptr;
+    Core::Timing::EventType* thread_wakeup_event_type = nullptr;
     // TODO(yuriks): This can be removed if Thread objects are explicitly pooled in the future,
     // allowing us to simply use a pool index or similar.
     Kernel::HandleTable thread_wakeup_callback_handle_table;
@@ -152,8 +152,8 @@ KernelCore::~KernelCore() {
     Shutdown();
 }
 
-void KernelCore::Initialize() {
-    impl->Initialize(*this);
+void KernelCore::Initialize(Core::Timing::CoreTiming& core_timing) {
+    impl->Initialize(*this, core_timing);
 }
 
 void KernelCore::Shutdown() {
@@ -213,7 +213,7 @@ u64 KernelCore::CreateNewProcessID() {
     return impl->next_process_id++;
 }
 
-CoreTiming::EventType* KernelCore::ThreadWakeupCallbackEventType() const {
+Core::Timing::EventType* KernelCore::ThreadWakeupCallbackEventType() const {
     return impl->thread_wakeup_event_type;
 }
 

src/core/hle/kernel/kernel.h

@@ -11,9 +11,10 @@
 template <typename T>
 class ResultVal;
 
-namespace CoreTiming {
+namespace Core::Timing {
+class CoreTiming;
 struct EventType;
-}
+} // namespace Core::Timing
 
 namespace Kernel {
 
@@ -39,7 +40,11 @@ public:
     KernelCore& operator=(KernelCore&&) = delete;
 
     /// Resets the kernel to a clean slate for use.
-    void Initialize();
+    ///
+    /// @param core_timing CoreTiming instance used to create any necessary
+    ///                    kernel-specific callback events.
+    ///
+    void Initialize(Core::Timing::CoreTiming& core_timing);
 
     /// Clears all resources in use by the kernel instance.
     void Shutdown();
@@ -89,7 +94,7 @@ private:
     u64 CreateNewThreadID();
 
     /// Retrieves the event type used for thread wakeup callbacks.
-    CoreTiming::EventType* ThreadWakeupCallbackEventType() const;
+    Core::Timing::EventType* ThreadWakeupCallbackEventType() const;
 
     /// Provides a reference to the thread wakeup callback handle table.
     Kernel::HandleTable& ThreadWakeupCallbackHandleTable();

src/core/hle/kernel/scheduler.cpp

@@ -111,7 +111,7 @@ void Scheduler::SwitchContext(Thread* new_thread) {
 
 void Scheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
     const u64 prev_switch_ticks = last_context_switch_time;
-    const u64 most_recent_switch_ticks = CoreTiming::GetTicks();
+    const u64 most_recent_switch_ticks = Core::System::GetInstance().CoreTiming().GetTicks();
     const u64 update_ticks = most_recent_switch_ticks - prev_switch_ticks;
 
     if (thread != nullptr) {

src/core/hle/kernel/svc.cpp

@@ -918,6 +918,7 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
         }
 
         const auto& system = Core::System::GetInstance();
+        const auto& core_timing = system.CoreTiming();
         const auto& scheduler = system.CurrentScheduler();
         const auto* const current_thread = scheduler.GetCurrentThread();
         const bool same_thread = current_thread == thread;
@@ -927,9 +928,9 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
         if (same_thread && info_sub_id == 0xFFFFFFFFFFFFFFFF) {
             const u64 thread_ticks = current_thread->GetTotalCPUTimeTicks();
 
-            out_ticks = thread_ticks + (CoreTiming::GetTicks() - prev_ctx_ticks);
+            out_ticks = thread_ticks + (core_timing.GetTicks() - prev_ctx_ticks);
         } else if (same_thread && info_sub_id == system.CurrentCoreIndex()) {
-            out_ticks = CoreTiming::GetTicks() - prev_ctx_ticks;
+            out_ticks = core_timing.GetTicks() - prev_ctx_ticks;
         }
 
         *result = out_ticks;
@@ -1546,10 +1547,11 @@ static ResultCode SignalToAddress(VAddr address, u32 type, s32 value, s32 num_to
 static u64 GetSystemTick() {
     LOG_TRACE(Kernel_SVC, "called");
 
-    const u64 result{CoreTiming::GetTicks()};
+    auto& core_timing = Core::System::GetInstance().CoreTiming();
+    const u64 result{core_timing.GetTicks()};
 
     // Advance time to defeat dumb games that busy-wait for the frame to end.
-    CoreTiming::AddTicks(400);
+    core_timing.AddTicks(400);
 
     return result;
 }

src/core/hle/kernel/thread.cpp

@@ -43,7 +43,8 @@ Thread::~Thread() = default;
 
 void Thread::Stop() {
     // Cancel any outstanding wakeup events for this thread
-    CoreTiming::UnscheduleEvent(kernel.ThreadWakeupCallbackEventType(), callback_handle);
+    Core::System::GetInstance().CoreTiming().UnscheduleEvent(kernel.ThreadWakeupCallbackEventType(),
+                                                             callback_handle);
     kernel.ThreadWakeupCallbackHandleTable().Close(callback_handle);
     callback_handle = 0;
 
@@ -85,12 +86,14 @@ void Thread::WakeAfterDelay(s64 nanoseconds) {
 
     // This function might be called from any thread so we have to be cautious and use the
     // thread-safe version of ScheduleEvent.
-    CoreTiming::ScheduleEventThreadsafe(CoreTiming::nsToCycles(nanoseconds),
-                                        kernel.ThreadWakeupCallbackEventType(), callback_handle);
+    Core::System::GetInstance().CoreTiming().ScheduleEventThreadsafe(
+        Core::Timing::nsToCycles(nanoseconds), kernel.ThreadWakeupCallbackEventType(),
+        callback_handle);
 }
 
 void Thread::CancelWakeupTimer() {
-    CoreTiming::UnscheduleEventThreadsafe(kernel.ThreadWakeupCallbackEventType(), callback_handle);
+    Core::System::GetInstance().CoreTiming().UnscheduleEventThreadsafe(
+        kernel.ThreadWakeupCallbackEventType(), callback_handle);
 }
 
 static std::optional<s32> GetNextProcessorId(u64 mask) {
@@ -189,6 +192,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
         return ResultCode(-1);
     }
 
+    auto& system = Core::System::GetInstance();
     SharedPtr<Thread> thread(new Thread(kernel));
 
     thread->thread_id = kernel.CreateNewThreadID();
@@ -197,7 +201,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     thread->stack_top = stack_top;
     thread->tpidr_el0 = 0;
     thread->nominal_priority = thread->current_priority = priority;
-    thread->last_running_ticks = CoreTiming::GetTicks();
+    thread->last_running_ticks = system.CoreTiming().GetTicks();
     thread->processor_id = processor_id;
     thread->ideal_core = processor_id;
     thread->affinity_mask = 1ULL << processor_id;
@@ -208,7 +212,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     thread->name = std::move(name);
     thread->callback_handle = kernel.ThreadWakeupCallbackHandleTable().Create(thread).Unwrap();
     thread->owner_process = &owner_process;
-    thread->scheduler = &Core::System::GetInstance().Scheduler(processor_id);
+    thread->scheduler = &system.Scheduler(processor_id);
     thread->scheduler->AddThread(thread, priority);
     thread->tls_address = thread->owner_process->MarkNextAvailableTLSSlotAsUsed(*thread);
 
@@ -257,7 +261,7 @@ void Thread::SetStatus(ThreadStatus new_status) {
     }
 
     if (status == ThreadStatus::Running) {
-        last_running_ticks = CoreTiming::GetTicks();
+        last_running_ticks = Core::System::GetInstance().CoreTiming().GetTicks();
     }
 
     status = new_status;

src/core/hle/service/audio/audout_u.cpp

@@ -68,12 +68,12 @@ public:
         RegisterHandlers(functions);
 
         // This is the event handle used to check if the audio buffer was released
-        auto& kernel = Core::System::GetInstance().Kernel();
-        buffer_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Sticky,
-                                                              "IAudioOutBufferReleased");
+        auto& system = Core::System::GetInstance();
+        buffer_event = Kernel::WritableEvent::CreateEventPair(
+            system.Kernel(), Kernel::ResetType::Sticky, "IAudioOutBufferReleased");
 
-        stream = audio_core.OpenStream(audio_params.sample_rate, audio_params.channel_count,
-                                       std::move(unique_name),
+        stream = audio_core.OpenStream(system.CoreTiming(), audio_params.sample_rate,
+                                       audio_params.channel_count, std::move(unique_name),
                                        [=]() { buffer_event.writable->Signal(); });
     }
 

src/core/hle/service/audio/audren_u.cpp

@@ -42,10 +42,11 @@ public:
         // clang-format on
         RegisterHandlers(functions);
 
-        auto& kernel = Core::System::GetInstance().Kernel();
-        system_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Sticky,
-                                                              "IAudioRenderer:SystemEvent");
-        renderer = std::make_unique<AudioCore::AudioRenderer>(audren_params, system_event.writable);
+        auto& system = Core::System::GetInstance();
+        system_event = Kernel::WritableEvent::CreateEventPair(
+            system.Kernel(), Kernel::ResetType::Sticky, "IAudioRenderer:SystemEvent");
+        renderer = std::make_unique<AudioCore::AudioRenderer>(system.CoreTiming(), audren_params,
+                                                              system_event.writable);
     }
 
 private:

src/core/hle/service/hid/controllers/controller_base.h

@@ -7,6 +7,10 @@
 #include "common/common_types.h"
 #include "common/swap.h"
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Service::HID {
 class ControllerBase {
 public:
@@ -20,7 +24,8 @@ public:
     virtual void OnRelease() = 0;
 
     // When the controller is requesting an update for the shared memory
-    virtual void OnUpdate(u8* data, std::size_t size) = 0;
+    virtual void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                          std::size_t size) = 0;
 
     // Called when input devices should be loaded
     virtual void OnLoadInputDevices() = 0;

src/core/hle/service/hid/controllers/debug_pad.cpp

@@ -21,8 +21,9 @@ void Controller_DebugPad::OnInit() {}
 
 void Controller_DebugPad::OnRelease() {}
 
-void Controller_DebugPad::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_DebugPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                   std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {

src/core/hle/service/hid/controllers/debug_pad.h

@@ -26,7 +26,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;

src/core/hle/service/hid/controllers/gesture.cpp

@@ -17,8 +17,9 @@ void Controller_Gesture::OnInit() {}
 
 void Controller_Gesture::OnRelease() {}
 
-void Controller_Gesture::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_Gesture::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                  std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {

src/core/hle/service/hid/controllers/gesture.h

@@ -22,7 +22,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;

src/core/hle/service/hid/controllers/keyboard.cpp

@@ -19,8 +19,9 @@ void Controller_Keyboard::OnInit() {}
 
 void Controller_Keyboard::OnRelease() {}
 
-void Controller_Keyboard::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_Keyboard::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                   std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {

src/core/hle/service/hid/controllers/keyboard.h

@@ -25,7 +25,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;

src/core/hle/service/hid/controllers/mouse.cpp

@@ -17,8 +17,9 @@ Controller_Mouse::~Controller_Mouse() = default;
 void Controller_Mouse::OnInit() {}
 void Controller_Mouse::OnRelease() {}
 
-void Controller_Mouse::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_Mouse::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {

src/core/hle/service/hid/controllers/mouse.h

@@ -24,7 +24,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;

src/core/hle/service/hid/controllers/npad.cpp

@@ -288,7 +288,8 @@ void Controller_NPad::RequestPadStateUpdate(u32 npad_id) {
     rstick_entry.y = static_cast<s32>(stick_r_y_f * HID_JOYSTICK_MAX);
 }
 
-void Controller_NPad::OnUpdate(u8* data, std::size_t data_len) {
+void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                               std::size_t data_len) {
     if (!IsControllerActivated())
         return;
     for (std::size_t i = 0; i < shared_memory_entries.size(); i++) {
@@ -308,7 +309,7 @@ void Controller_NPad::OnUpdate(u8* data, std::size_t data_len) {
             const auto& last_entry =
                 main_controller->npad[main_controller->common.last_entry_index];
 
-            main_controller->common.timestamp = CoreTiming::GetTicks();
+            main_controller->common.timestamp = core_timing.GetTicks();
             main_controller->common.last_entry_index =
                 (main_controller->common.last_entry_index + 1) % 17;
 

src/core/hle/service/hid/controllers/npad.h

@@ -30,7 +30,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;

src/core/hle/service/hid/controllers/stubbed.cpp

@@ -16,13 +16,14 @@ void Controller_Stubbed::OnInit() {}
 
 void Controller_Stubbed::OnRelease() {}
 
-void Controller_Stubbed::OnUpdate(u8* data, std::size_t size) {
+void Controller_Stubbed::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                  std::size_t size) {
     if (!smart_update) {
         return;
     }
 
     CommonHeader header{};
-    header.timestamp = CoreTiming::GetTicks();
+    header.timestamp = core_timing.GetTicks();
     header.total_entry_count = 17;
     header.entry_count = 0;
     header.last_entry_index = 0;

src/core/hle/service/hid/controllers/stubbed.h

@@ -20,7 +20,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;

src/core/hle/service/hid/controllers/touchscreen.cpp

@@ -20,8 +20,9 @@ void Controller_Touchscreen::OnInit() {}
 
 void Controller_Touchscreen::OnRelease() {}
 
-void Controller_Touchscreen::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_Touchscreen::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                      std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {
@@ -48,7 +49,7 @@ void Controller_Touchscreen::OnUpdate(u8* data, std::size_t size) {
         touch_entry.diameter_x = Settings::values.touchscreen.diameter_x;
         touch_entry.diameter_y = Settings::values.touchscreen.diameter_y;
         touch_entry.rotation_angle = Settings::values.touchscreen.rotation_angle;
-        const u64 tick = CoreTiming::GetTicks();
+        const u64 tick = core_timing.GetTicks();
         touch_entry.delta_time = tick - last_touch;
         last_touch = tick;
         touch_entry.finger = Settings::values.touchscreen.finger;

src/core/hle/service/hid/controllers/touchscreen.h

@@ -24,7 +24,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;

src/core/hle/service/hid/controllers/xpad.cpp

@@ -17,9 +17,10 @@ void Controller_XPad::OnInit() {}
 
 void Controller_XPad::OnRelease() {}
 
-void Controller_XPad::OnUpdate(u8* data, std::size_t size) {
+void Controller_XPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                               std::size_t size) {
     for (auto& xpad_entry : shared_memory.shared_memory_entries) {
-        xpad_entry.header.timestamp = CoreTiming::GetTicks();
+        xpad_entry.header.timestamp = core_timing.GetTicks();
         xpad_entry.header.total_entry_count = 17;
 
         if (!IsControllerActivated()) {

src/core/hle/service/hid/controllers/xpad.h

@@ -22,7 +22,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;

src/core/hle/service/hid/hid.cpp

@@ -36,9 +36,9 @@ namespace Service::HID {
 
 // Updating period for each HID device.
 // TODO(ogniK): Find actual polling rate of hid
-constexpr u64 pad_update_ticks = CoreTiming::BASE_CLOCK_RATE / 66;
-constexpr u64 accelerometer_update_ticks = CoreTiming::BASE_CLOCK_RATE / 100;
-constexpr u64 gyroscope_update_ticks = CoreTiming::BASE_CLOCK_RATE / 100;
+constexpr u64 pad_update_ticks = Core::Timing::BASE_CLOCK_RATE / 66;
+constexpr u64 accelerometer_update_ticks = Core::Timing::BASE_CLOCK_RATE / 100;
+constexpr u64 gyroscope_update_ticks = Core::Timing::BASE_CLOCK_RATE / 100;
 constexpr std::size_t SHARED_MEMORY_SIZE = 0x40000;
 
 IAppletResource::IAppletResource() : ServiceFramework("IAppletResource") {
@@ -73,14 +73,15 @@ IAppletResource::IAppletResource() : ServiceFramework("IAppletResource") {
     GetController<Controller_Stubbed>(HidController::Unknown3).SetCommonHeaderOffset(0x5000);
 
     // Register update callbacks
+    auto& core_timing = Core::System::GetInstance().CoreTiming();
     pad_update_event =
-        CoreTiming::RegisterEvent("HID::UpdatePadCallback", [this](u64 userdata, int cycles_late) {
+        core_timing.RegisterEvent("HID::UpdatePadCallback", [this](u64 userdata, int cycles_late) {
             UpdateControllers(userdata, cycles_late);
         });
 
     // TODO(shinyquagsire23): Other update callbacks? (accel, gyro?)
 
-    CoreTiming::ScheduleEvent(pad_update_ticks, pad_update_event);
+    core_timing.ScheduleEvent(pad_update_ticks, pad_update_event);
 
     ReloadInputDevices();
 }
@@ -94,7 +95,7 @@ void IAppletResource::DeactivateController(HidController controller) {
 }
 
 IAppletResource ::~IAppletResource() {
-    CoreTiming::UnscheduleEvent(pad_update_event, 0);
+    Core::System::GetInstance().CoreTiming().UnscheduleEvent(pad_update_event, 0);
 }
 
 void IAppletResource::GetSharedMemoryHandle(Kernel::HLERequestContext& ctx) {
@@ -106,15 +107,17 @@ void IAppletResource::GetSharedMemoryHandle(Kernel::HLERequestContext& ctx) {
 }
 
 void IAppletResource::UpdateControllers(u64 userdata, int cycles_late) {
+    auto& core_timing = Core::System::GetInstance().CoreTiming();
+
     const bool should_reload = Settings::values.is_device_reload_pending.exchange(false);
     for (const auto& controller : controllers) {
         if (should_reload) {
             controller->OnLoadInputDevices();
         }
-        controller->OnUpdate(shared_mem->GetPointer(), SHARED_MEMORY_SIZE);
+        controller->OnUpdate(core_timing, shared_mem->GetPointer(), SHARED_MEMORY_SIZE);
     }
 
-    CoreTiming::ScheduleEvent(pad_update_ticks - cycles_late, pad_update_event);
+    core_timing.ScheduleEvent(pad_update_ticks - cycles_late, pad_update_event);
 }
 
 class IActiveVibrationDeviceList final : public ServiceFramework<IActiveVibrationDeviceList> {

src/core/hle/service/hid/hid.h

@@ -7,7 +7,7 @@
 #include "controllers/controller_base.h"
 #include "core/hle/service/service.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
 struct EventType;
 }
 
@@ -66,7 +66,7 @@ private:
 
     Kernel::SharedPtr<Kernel::SharedMemory> shared_mem;
 
-    CoreTiming::EventType* pad_update_event;
+    Core::Timing::EventType* pad_update_event;
 
     std::array<std::unique_ptr<ControllerBase>, static_cast<size_t>(HidController::MaxControllers)>
         controllers{};

src/core/hle/service/hid/irs.cpp

@@ -98,7 +98,7 @@ void IRS::GetImageTransferProcessorState(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 5};
     rb.Push(RESULT_SUCCESS);
-    rb.PushRaw<u64>(CoreTiming::GetTicks());
+    rb.PushRaw<u64>(Core::System::GetInstance().CoreTiming().GetTicks());
     rb.PushRaw<u32>(0);
 }
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp

@@ -5,6 +5,7 @@
 #include <cstring>
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h"
@@ -184,7 +185,7 @@ u32 nvhost_ctrl_gpu::GetGpuTime(const std::vector<u8>& input, std::vector<u8>& o
 
     IoctlGetGpuTime params{};
     std::memcpy(&params, input.data(), input.size());
-    params.gpu_time = CoreTiming::cyclesToNs(CoreTiming::GetTicks());
+    params.gpu_time = Core::Timing::cyclesToNs(Core::System::GetInstance().CoreTiming().GetTicks());
     std::memcpy(output.data(), &params, output.size());
     return 0;
 }

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

@@ -13,10 +13,6 @@
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/writable_event.h"
 
-namespace CoreTiming {
-struct EventType;
-}
-
 namespace Service::NVFlinger {
 
 struct IGBPBuffer {

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

@@ -25,21 +25,21 @@
 namespace Service::NVFlinger {
 
 constexpr std::size_t SCREEN_REFRESH_RATE = 60;
-constexpr u64 frame_ticks = static_cast<u64>(CoreTiming::BASE_CLOCK_RATE / SCREEN_REFRESH_RATE);
+constexpr u64 frame_ticks = static_cast<u64>(Core::Timing::BASE_CLOCK_RATE / SCREEN_REFRESH_RATE);
 
-NVFlinger::NVFlinger() {
+NVFlinger::NVFlinger(Core::Timing::CoreTiming& core_timing) : core_timing{core_timing} {
     // Schedule the screen composition events
     composition_event =
-        CoreTiming::RegisterEvent("ScreenComposition", [this](u64 userdata, int cycles_late) {
+        core_timing.RegisterEvent("ScreenComposition", [this](u64 userdata, int cycles_late) {
             Compose();
-            CoreTiming::ScheduleEvent(frame_ticks - cycles_late, composition_event);
+            this->core_timing.ScheduleEvent(frame_ticks - cycles_late, composition_event);
         });
 
-    CoreTiming::ScheduleEvent(frame_ticks, composition_event);
+    core_timing.ScheduleEvent(frame_ticks, composition_event);
 }
 
 NVFlinger::~NVFlinger() {
-    CoreTiming::UnscheduleEvent(composition_event, 0);
+    core_timing.UnscheduleEvent(composition_event, 0);
 }
 
 void NVFlinger::SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance) {

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

@@ -14,9 +14,10 @@
 #include "common/common_types.h"
 #include "core/hle/kernel/object.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
+class CoreTiming;
 struct EventType;
-}
+} // namespace Core::Timing
 
 namespace Kernel {
 class ReadableEvent;
@@ -52,7 +53,7 @@ struct Display {
 
 class NVFlinger final {
 public:
-    NVFlinger();
+    explicit NVFlinger(Core::Timing::CoreTiming& core_timing);
     ~NVFlinger();
 
     /// Sets the NVDrv module instance to use to send buffers to the GPU.
@@ -115,8 +116,11 @@ private:
     /// layers.
     u32 next_buffer_queue_id = 1;
 
-    /// CoreTiming event that handles screen composition.
-    CoreTiming::EventType* composition_event;
+    /// Event that handles screen composition.
+    Core::Timing::EventType* composition_event;
+
+    /// Core timing instance for registering/unregistering the composition event.
+    Core::Timing::CoreTiming& core_timing;
 };
 
 } // namespace Service::NVFlinger

src/core/hle/service/service.cpp

@@ -194,10 +194,11 @@ ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& co
 // Module interface
 
 /// Initialize ServiceManager
-void Init(std::shared_ptr<SM::ServiceManager>& sm, FileSys::VfsFilesystem& vfs) {
+void Init(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system,
+          FileSys::VfsFilesystem& vfs) {
     // NVFlinger needs to be accessed by several services like Vi and AppletOE so we instantiate it
     // here and pass it into the respective InstallInterfaces functions.
-    auto nv_flinger = std::make_shared<NVFlinger::NVFlinger>();
+    auto nv_flinger = std::make_shared<NVFlinger::NVFlinger>(system.CoreTiming());
 
     SM::ServiceManager::InstallInterfaces(sm);
 

src/core/hle/service/service.h

@@ -14,6 +14,14 @@
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Namespace Service
 
+namespace Core {
+class System;
+}
+
+namespace FileSys {
+class VfsFilesystem;
+}
+
 namespace Kernel {
 class ClientPort;
 class ServerPort;
@@ -21,10 +29,6 @@ class ServerSession;
 class HLERequestContext;
 } // namespace Kernel
 
-namespace FileSys {
-class VfsFilesystem;
-}
-
 namespace Service {
 
 namespace SM {
@@ -178,7 +182,8 @@ private:
 };
 
 /// Initialize ServiceManager
-void Init(std::shared_ptr<SM::ServiceManager>& sm, FileSys::VfsFilesystem& vfs);
+void Init(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system,
+          FileSys::VfsFilesystem& vfs);
 
 /// Shutdown ServiceManager
 void Shutdown();

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

@@ -5,6 +5,7 @@
 #include <chrono>
 #include <ctime>
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hle/ipc_helpers.h"
@@ -106,8 +107,9 @@ private:
     void GetCurrentTimePoint(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_Time, "called");
 
-        SteadyClockTimePoint steady_clock_time_point{
-            CoreTiming::cyclesToMs(CoreTiming::GetTicks()) / 1000};
+        const auto& core_timing = Core::System::GetInstance().CoreTiming();
+        const SteadyClockTimePoint steady_clock_time_point{
+            Core::Timing::cyclesToMs(core_timing.GetTicks()) / 1000};
         IPC::ResponseBuilder rb{ctx, (sizeof(SteadyClockTimePoint) / 4) + 2};
         rb.Push(RESULT_SUCCESS);
         rb.PushRaw(steady_clock_time_point);
@@ -281,8 +283,9 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
         return;
     }
 
+    const auto& core_timing = Core::System::GetInstance().CoreTiming();
     const SteadyClockTimePoint steady_clock_time_point{
-        CoreTiming::cyclesToMs(CoreTiming::GetTicks()) / 1000, {}};
+        Core::Timing::cyclesToMs(core_timing.GetTicks()) / 1000, {}};
 
     CalendarTime calendar_time{};
     calendar_time.year = tm->tm_year + 1900;

src/core/memory.cpp

@@ -71,15 +71,20 @@ static void MapPages(PageTable& page_table, VAddr base, u64 size, u8* memory, Pa
                                  FlushMode::FlushAndInvalidate);
 
     VAddr end = base + size;
-    while (base != end) {
-        ASSERT_MSG(base < page_table.pointers.size(), "out of range mapping at {:016X}", base);
+    ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}",
+               base + page_table.pointers.size());
 
-        page_table.attributes[base] = type;
-        page_table.pointers[base] = memory;
+    std::fill(page_table.attributes.begin() + base, page_table.attributes.begin() + end, type);
 
-        base += 1;
-        if (memory != nullptr)
+    if (memory == nullptr) {
+        std::fill(page_table.pointers.begin() + base, page_table.pointers.begin() + end, memory);
+    } else {
+        while (base != end) {
+            page_table.pointers[base] = memory;
+
+            base += 1;
             memory += PAGE_SIZE;
+        }
     }
 }
 

src/tests/core/core_timing.cpp

@@ -28,100 +28,103 @@ void CallbackTemplate(u64 userdata, s64 cycles_late) {
     REQUIRE(lateness == cycles_late);
 }
 
-class ScopeInit final {
-public:
+struct ScopeInit final {
     ScopeInit() {
-        CoreTiming::Init();
+        core_timing.Initialize();
     }
     ~ScopeInit() {
-        CoreTiming::Shutdown();
+        core_timing.Shutdown();
     }
+
+    Core::Timing::CoreTiming core_timing;
 };
 
-static void AdvanceAndCheck(u32 idx, int downcount, int expected_lateness = 0,
-                            int cpu_downcount = 0) {
+static void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, int downcount,
+                            int expected_lateness = 0, int cpu_downcount = 0) {
     callbacks_ran_flags = 0;
     expected_callback = CB_IDS[idx];
     lateness = expected_lateness;
 
-    CoreTiming::AddTicks(CoreTiming::GetDowncount() -
-                         cpu_downcount); // Pretend we executed X cycles of instructions.
-    CoreTiming::Advance();
+    // Pretend we executed X cycles of instructions.
+    core_timing.AddTicks(core_timing.GetDowncount() - cpu_downcount);
+    core_timing.Advance();
 
     REQUIRE(decltype(callbacks_ran_flags)().set(idx) == callbacks_ran_flags);
-    REQUIRE(downcount == CoreTiming::GetDowncount());
+    REQUIRE(downcount == core_timing.GetDowncount());
 }
 
 TEST_CASE("CoreTiming[BasicOrder]", "[core]") {
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", CallbackTemplate<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", CallbackTemplate<1>);
-    CoreTiming::EventType* cb_c = CoreTiming::RegisterEvent("callbackC", CallbackTemplate<2>);
-    CoreTiming::EventType* cb_d = CoreTiming::RegisterEvent("callbackD", CallbackTemplate<3>);
-    CoreTiming::EventType* cb_e = CoreTiming::RegisterEvent("callbackE", CallbackTemplate<4>);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
+    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", CallbackTemplate<2>);
+    Core::Timing::EventType* cb_d = core_timing.RegisterEvent("callbackD", CallbackTemplate<3>);
+    Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", CallbackTemplate<4>);
 
     // Enter slice 0
-    CoreTiming::Advance();
+    core_timing.Advance();
 
     // D -> B -> C -> A -> E
-    CoreTiming::ScheduleEvent(1000, cb_a, CB_IDS[0]);
-    REQUIRE(1000 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEvent(500, cb_b, CB_IDS[1]);
-    REQUIRE(500 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEvent(800, cb_c, CB_IDS[2]);
-    REQUIRE(500 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEvent(100, cb_d, CB_IDS[3]);
-    REQUIRE(100 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEvent(1200, cb_e, CB_IDS[4]);
-    REQUIRE(100 == CoreTiming::GetDowncount());
+    core_timing.ScheduleEvent(1000, cb_a, CB_IDS[0]);
+    REQUIRE(1000 == core_timing.GetDowncount());
+    core_timing.ScheduleEvent(500, cb_b, CB_IDS[1]);
+    REQUIRE(500 == core_timing.GetDowncount());
+    core_timing.ScheduleEvent(800, cb_c, CB_IDS[2]);
+    REQUIRE(500 == core_timing.GetDowncount());
+    core_timing.ScheduleEvent(100, cb_d, CB_IDS[3]);
+    REQUIRE(100 == core_timing.GetDowncount());
+    core_timing.ScheduleEvent(1200, cb_e, CB_IDS[4]);
+    REQUIRE(100 == core_timing.GetDowncount());
 
-    AdvanceAndCheck(3, 400);
-    AdvanceAndCheck(1, 300);
-    AdvanceAndCheck(2, 200);
-    AdvanceAndCheck(0, 200);
-    AdvanceAndCheck(4, MAX_SLICE_LENGTH);
+    AdvanceAndCheck(core_timing, 3, 400);
+    AdvanceAndCheck(core_timing, 1, 300);
+    AdvanceAndCheck(core_timing, 2, 200);
+    AdvanceAndCheck(core_timing, 0, 200);
+    AdvanceAndCheck(core_timing, 4, MAX_SLICE_LENGTH);
 }
 
 TEST_CASE("CoreTiming[Threadsave]", "[core]") {
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", CallbackTemplate<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", CallbackTemplate<1>);
-    CoreTiming::EventType* cb_c = CoreTiming::RegisterEvent("callbackC", CallbackTemplate<2>);
-    CoreTiming::EventType* cb_d = CoreTiming::RegisterEvent("callbackD", CallbackTemplate<3>);
-    CoreTiming::EventType* cb_e = CoreTiming::RegisterEvent("callbackE", CallbackTemplate<4>);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
+    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", CallbackTemplate<2>);
+    Core::Timing::EventType* cb_d = core_timing.RegisterEvent("callbackD", CallbackTemplate<3>);
+    Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", CallbackTemplate<4>);
 
     // Enter slice 0
-    CoreTiming::Advance();
+    core_timing.Advance();
 
     // D -> B -> C -> A -> E
-    CoreTiming::ScheduleEventThreadsafe(1000, cb_a, CB_IDS[0]);
+    core_timing.ScheduleEventThreadsafe(1000, cb_a, CB_IDS[0]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(1000);
-    REQUIRE(1000 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEventThreadsafe(500, cb_b, CB_IDS[1]);
+    core_timing.ForceExceptionCheck(1000);
+    REQUIRE(1000 == core_timing.GetDowncount());
+    core_timing.ScheduleEventThreadsafe(500, cb_b, CB_IDS[1]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(500);
-    REQUIRE(500 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEventThreadsafe(800, cb_c, CB_IDS[2]);
+    core_timing.ForceExceptionCheck(500);
+    REQUIRE(500 == core_timing.GetDowncount());
+    core_timing.ScheduleEventThreadsafe(800, cb_c, CB_IDS[2]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(800);
-    REQUIRE(500 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEventThreadsafe(100, cb_d, CB_IDS[3]);
+    core_timing.ForceExceptionCheck(800);
+    REQUIRE(500 == core_timing.GetDowncount());
+    core_timing.ScheduleEventThreadsafe(100, cb_d, CB_IDS[3]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(100);
-    REQUIRE(100 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEventThreadsafe(1200, cb_e, CB_IDS[4]);
+    core_timing.ForceExceptionCheck(100);
+    REQUIRE(100 == core_timing.GetDowncount());
+    core_timing.ScheduleEventThreadsafe(1200, cb_e, CB_IDS[4]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(1200);
-    REQUIRE(100 == CoreTiming::GetDowncount());
+    core_timing.ForceExceptionCheck(1200);
+    REQUIRE(100 == core_timing.GetDowncount());
 
-    AdvanceAndCheck(3, 400);
-    AdvanceAndCheck(1, 300);
-    AdvanceAndCheck(2, 200);
-    AdvanceAndCheck(0, 200);
-    AdvanceAndCheck(4, MAX_SLICE_LENGTH);
+    AdvanceAndCheck(core_timing, 3, 400);
+    AdvanceAndCheck(core_timing, 1, 300);
+    AdvanceAndCheck(core_timing, 2, 200);
+    AdvanceAndCheck(core_timing, 0, 200);
+    AdvanceAndCheck(core_timing, 4, MAX_SLICE_LENGTH);
 }
 
 namespace SharedSlotTest {
@@ -142,59 +145,63 @@ TEST_CASE("CoreTiming[SharedSlot]", "[core]") {
     using namespace SharedSlotTest;
 
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", FifoCallback<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", FifoCallback<1>);
-    CoreTiming::EventType* cb_c = CoreTiming::RegisterEvent("callbackC", FifoCallback<2>);
-    CoreTiming::EventType* cb_d = CoreTiming::RegisterEvent("callbackD", FifoCallback<3>);
-    CoreTiming::EventType* cb_e = CoreTiming::RegisterEvent("callbackE", FifoCallback<4>);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", FifoCallback<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", FifoCallback<1>);
+    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", FifoCallback<2>);
+    Core::Timing::EventType* cb_d = core_timing.RegisterEvent("callbackD", FifoCallback<3>);
+    Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", FifoCallback<4>);
 
-    CoreTiming::ScheduleEvent(1000, cb_a, CB_IDS[0]);
-    CoreTiming::ScheduleEvent(1000, cb_b, CB_IDS[1]);
-    CoreTiming::ScheduleEvent(1000, cb_c, CB_IDS[2]);
-    CoreTiming::ScheduleEvent(1000, cb_d, CB_IDS[3]);
-    CoreTiming::ScheduleEvent(1000, cb_e, CB_IDS[4]);
+    core_timing.ScheduleEvent(1000, cb_a, CB_IDS[0]);
+    core_timing.ScheduleEvent(1000, cb_b, CB_IDS[1]);
+    core_timing.ScheduleEvent(1000, cb_c, CB_IDS[2]);
+    core_timing.ScheduleEvent(1000, cb_d, CB_IDS[3]);
+    core_timing.ScheduleEvent(1000, cb_e, CB_IDS[4]);
 
     // Enter slice 0
-    CoreTiming::Advance();
-    REQUIRE(1000 == CoreTiming::GetDowncount());
+    core_timing.Advance();
+    REQUIRE(1000 == core_timing.GetDowncount());
 
     callbacks_ran_flags = 0;
     counter = 0;
     lateness = 0;
-    CoreTiming::AddTicks(CoreTiming::GetDowncount());
-    CoreTiming::Advance();
-    REQUIRE(MAX_SLICE_LENGTH == CoreTiming::GetDowncount());
+    core_timing.AddTicks(core_timing.GetDowncount());
+    core_timing.Advance();
+    REQUIRE(MAX_SLICE_LENGTH == core_timing.GetDowncount());
     REQUIRE(0x1FULL == callbacks_ran_flags.to_ullong());
 }
 
-TEST_CASE("CoreTiming[PredictableLateness]", "[core]") {
+TEST_CASE("Core::Timing[PredictableLateness]", "[core]") {
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", CallbackTemplate<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", CallbackTemplate<1>);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
 
     // Enter slice 0
-    CoreTiming::Advance();
+    core_timing.Advance();
 
-    CoreTiming::ScheduleEvent(100, cb_a, CB_IDS[0]);
-    CoreTiming::ScheduleEvent(200, cb_b, CB_IDS[1]);
+    core_timing.ScheduleEvent(100, cb_a, CB_IDS[0]);
+    core_timing.ScheduleEvent(200, cb_b, CB_IDS[1]);
 
-    AdvanceAndCheck(0, 90, 10, -10); // (100 - 10)
-    AdvanceAndCheck(1, MAX_SLICE_LENGTH, 50, -50);
+    AdvanceAndCheck(core_timing, 0, 90, 10, -10); // (100 - 10)
+    AdvanceAndCheck(core_timing, 1, MAX_SLICE_LENGTH, 50, -50);
 }
 
 namespace ChainSchedulingTest {
 static int reschedules = 0;
 
-static void RescheduleCallback(u64 userdata, s64 cycles_late) {
+static void RescheduleCallback(Core::Timing::CoreTiming& core_timing, u64 userdata,
+                               s64 cycles_late) {
     --reschedules;
     REQUIRE(reschedules >= 0);
     REQUIRE(lateness == cycles_late);
 
-    if (reschedules > 0)
-        CoreTiming::ScheduleEvent(1000, reinterpret_cast<CoreTiming::EventType*>(userdata),
+    if (reschedules > 0) {
+        core_timing.ScheduleEvent(1000, reinterpret_cast<Core::Timing::EventType*>(userdata),
                                   userdata);
+    }
 }
 } // namespace ChainSchedulingTest
 
@@ -202,36 +209,39 @@ TEST_CASE("CoreTiming[ChainScheduling]", "[core]") {
     using namespace ChainSchedulingTest;
 
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", CallbackTemplate<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", CallbackTemplate<1>);
-    CoreTiming::EventType* cb_c = CoreTiming::RegisterEvent("callbackC", CallbackTemplate<2>);
-    CoreTiming::EventType* cb_rs =
-        CoreTiming::RegisterEvent("callbackReschedule", RescheduleCallback);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
+    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", CallbackTemplate<2>);
+    Core::Timing::EventType* cb_rs = core_timing.RegisterEvent(
+        "callbackReschedule", [&core_timing](u64 userdata, s64 cycles_late) {
+            RescheduleCallback(core_timing, userdata, cycles_late);
+        });
 
     // Enter slice 0
-    CoreTiming::Advance();
+    core_timing.Advance();
 
-    CoreTiming::ScheduleEvent(800, cb_a, CB_IDS[0]);
-    CoreTiming::ScheduleEvent(1000, cb_b, CB_IDS[1]);
-    CoreTiming::ScheduleEvent(2200, cb_c, CB_IDS[2]);
-    CoreTiming::ScheduleEvent(1000, cb_rs, reinterpret_cast<u64>(cb_rs));
-    REQUIRE(800 == CoreTiming::GetDowncount());
+    core_timing.ScheduleEvent(800, cb_a, CB_IDS[0]);
+    core_timing.ScheduleEvent(1000, cb_b, CB_IDS[1]);
+    core_timing.ScheduleEvent(2200, cb_c, CB_IDS[2]);
+    core_timing.ScheduleEvent(1000, cb_rs, reinterpret_cast<u64>(cb_rs));
+    REQUIRE(800 == core_timing.GetDowncount());
 
     reschedules = 3;
-    AdvanceAndCheck(0, 200);  // cb_a
-    AdvanceAndCheck(1, 1000); // cb_b, cb_rs
+    AdvanceAndCheck(core_timing, 0, 200);  // cb_a
+    AdvanceAndCheck(core_timing, 1, 1000); // cb_b, cb_rs
     REQUIRE(2 == reschedules);
 
-    CoreTiming::AddTicks(CoreTiming::GetDowncount());
-    CoreTiming::Advance(); // cb_rs
+    core_timing.AddTicks(core_timing.GetDowncount());
+    core_timing.Advance(); // cb_rs
     REQUIRE(1 == reschedules);
-    REQUIRE(200 == CoreTiming::GetDowncount());
+    REQUIRE(200 == core_timing.GetDowncount());
 
-    AdvanceAndCheck(2, 800); // cb_c
+    AdvanceAndCheck(core_timing, 2, 800); // cb_c
 
-    CoreTiming::AddTicks(CoreTiming::GetDowncount());
-    CoreTiming::Advance(); // cb_rs
+    core_timing.AddTicks(core_timing.GetDowncount());
+    core_timing.Advance(); // cb_rs
     REQUIRE(0 == reschedules);
-    REQUIRE(MAX_SLICE_LENGTH == CoreTiming::GetDowncount());
+    REQUIRE(MAX_SLICE_LENGTH == core_timing.GetDowncount());
 }

src/video_core/CMakeLists.txt

@@ -5,12 +5,12 @@ add_library(video_core STATIC
     debug_utils/debug_utils.h
     engines/fermi_2d.cpp
     engines/fermi_2d.h
+    engines/kepler_compute.cpp
+    engines/kepler_compute.h
     engines/kepler_memory.cpp
     engines/kepler_memory.h
     engines/maxwell_3d.cpp
     engines/maxwell_3d.h
-    engines/maxwell_compute.cpp
-    engines/maxwell_compute.h
     engines/maxwell_dma.cpp
     engines/maxwell_dma.h
     engines/shader_bytecode.h
@@ -101,6 +101,18 @@ add_library(video_core STATIC
     video_core.h
 )
 
+if (ENABLE_VULKAN)
+    target_sources(video_core PRIVATE
+        renderer_vulkan/declarations.h
+        renderer_vulkan/vk_device.cpp
+        renderer_vulkan/vk_device.h
+        renderer_vulkan/vk_resource_manager.cpp
+        renderer_vulkan/vk_resource_manager.h)
+
+    target_include_directories(video_core PRIVATE ../../externals/Vulkan-Headers/include)
+    target_compile_definitions(video_core PRIVATE HAS_VULKAN)
+endif()
+
 create_target_directory_groups(video_core)
 
 target_link_libraries(video_core PUBLIC common core)

src/video_core/engines/kepler_compute.cpp

@@ -0,0 +1,34 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/logging/log.h"
+#include "core/core.h"
+#include "core/memory.h"
+#include "video_core/engines/kepler_compute.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra::Engines {
+
+KeplerCompute::KeplerCompute(MemoryManager& memory_manager) : memory_manager{memory_manager} {}
+
+KeplerCompute::~KeplerCompute() = default;
+
+void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
+    ASSERT_MSG(method_call.method < Regs::NUM_REGS,
+               "Invalid KeplerCompute register, increase the size of the Regs structure");
+
+    regs.reg_array[method_call.method] = method_call.argument;
+
+    switch (method_call.method) {
+    case KEPLER_COMPUTE_REG_INDEX(launch):
+        // Abort execution since compute shaders can be used to alter game memory (e.g. CUDA
+        // kernels)
+        UNREACHABLE_MSG("Compute shaders are not implemented");
+        break;
+    default:
+        break;
+    }
+}
+
+} // namespace Tegra::Engines

src/video_core/engines/kepler_compute.h

@@ -10,47 +10,48 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
 
 namespace Tegra::Engines {
 
-#define MAXWELL_COMPUTE_REG_INDEX(field_name)                                                      \
-    (offsetof(Tegra::Engines::MaxwellCompute::Regs, field_name) / sizeof(u32))
+#define KEPLER_COMPUTE_REG_INDEX(field_name)                                                       \
+    (offsetof(Tegra::Engines::KeplerCompute::Regs, field_name) / sizeof(u32))
 
-class MaxwellCompute final {
+class KeplerCompute final {
 public:
-    MaxwellCompute() = default;
-    ~MaxwellCompute() = default;
+    explicit KeplerCompute(MemoryManager& memory_manager);
+    ~KeplerCompute();
+
+    static constexpr std::size_t NumConstBuffers = 8;
 
     struct Regs {
         static constexpr std::size_t NUM_REGS = 0xCF8;
 
         union {
             struct {
-                INSERT_PADDING_WORDS(0x281);
+                INSERT_PADDING_WORDS(0xAF);
 
-                union {
-                    u32 compute_end;
-                    BitField<0, 1, u32> unknown;
-                } compute;
+                u32 launch;
 
-                INSERT_PADDING_WORDS(0xA76);
+                INSERT_PADDING_WORDS(0xC48);
             };
             std::array<u32, NUM_REGS> reg_array;
         };
     } regs{};
-
     static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32),
-                  "MaxwellCompute Regs has wrong size");
+                  "KeplerCompute Regs has wrong size");
+
+    MemoryManager& memory_manager;
 
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
-    static_assert(offsetof(MaxwellCompute::Regs, field_name) == position * 4,                      \
+    static_assert(offsetof(KeplerCompute::Regs, field_name) == position * 4,                       \
                   "Field " #field_name " has invalid position")
 
-ASSERT_REG_POSITION(compute, 0x281);
+ASSERT_REG_POSITION(launch, 0xAF);
 
 #undef ASSERT_REG_POSITION
 

src/video_core/engines/maxwell_3d.cpp

@@ -317,7 +317,7 @@ void Maxwell3D::ProcessQueryGet() {
             LongQueryResult query_result{};
             query_result.value = result;
             // TODO(Subv): Generate a real GPU timestamp and write it here instead of CoreTiming
-            query_result.timestamp = CoreTiming::GetTicks();
+            query_result.timestamp = Core::System::GetInstance().CoreTiming().GetTicks();
             Memory::WriteBlock(*address, &query_result, sizeof(query_result));
         }
         dirty_flags.OnMemoryWrite();

src/video_core/engines/maxwell_compute.cpp

@@ -1,28 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "common/logging/log.h"
-#include "core/core.h"
-#include "video_core/engines/maxwell_compute.h"
-
-namespace Tegra::Engines {
-
-void MaxwellCompute::CallMethod(const GPU::MethodCall& method_call) {
-    ASSERT_MSG(method_call.method < Regs::NUM_REGS,
-               "Invalid MaxwellCompute register, increase the size of the Regs structure");
-
-    regs.reg_array[method_call.method] = method_call.argument;
-
-    switch (method_call.method) {
-    case MAXWELL_COMPUTE_REG_INDEX(compute): {
-        LOG_CRITICAL(HW_GPU, "Compute shaders are not implemented");
-        UNREACHABLE();
-        break;
-    }
-    default:
-        break;
-    }
-}
-
-} // namespace Tegra::Engines

src/video_core/engines/shader_bytecode.h

@@ -186,7 +186,7 @@ enum class SubOp : u64 {
 };
 
 enum class F2iRoundingOp : u64 {
-    None = 0,
+    RoundEven = 0,
     Floor = 1,
     Ceil = 2,
     Trunc = 3,

src/video_core/gpu.cpp

@@ -3,12 +3,13 @@
 // Refer to the license.txt file included.
 
 #include "common/assert.h"
+#include "core/core.h"
 #include "core/core_timing.h"
 #include "core/memory.h"
 #include "video_core/engines/fermi_2d.h"
+#include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/kepler_memory.h"
 #include "video_core/engines/maxwell_3d.h"
-#include "video_core/engines/maxwell_compute.h"
 #include "video_core/engines/maxwell_dma.h"
 #include "video_core/gpu.h"
 #include "video_core/rasterizer_interface.h"
@@ -18,6 +19,7 @@ namespace Tegra {
 u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
     switch (format) {
     case PixelFormat::ABGR8:
+    case PixelFormat::BGRA8:
         return 4;
     default:
         return 4;
@@ -31,7 +33,7 @@ GPU::GPU(VideoCore::RasterizerInterface& rasterizer) {
     dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
     maxwell_3d = std::make_unique<Engines::Maxwell3D>(rasterizer, *memory_manager);
     fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer, *memory_manager);
-    maxwell_compute = std::make_unique<Engines::MaxwellCompute>();
+    kepler_compute = std::make_unique<Engines::KeplerCompute>(*memory_manager);
     maxwell_dma = std::make_unique<Engines::MaxwellDMA>(rasterizer, *memory_manager);
     kepler_memory = std::make_unique<Engines::KeplerMemory>(rasterizer, *memory_manager);
 }
@@ -245,8 +247,8 @@ void GPU::CallEngineMethod(const MethodCall& method_call) {
     case EngineID::MAXWELL_B:
         maxwell_3d->CallMethod(method_call);
         break;
-    case EngineID::MAXWELL_COMPUTE_B:
-        maxwell_compute->CallMethod(method_call);
+    case EngineID::KEPLER_COMPUTE_B:
+        kepler_compute->CallMethod(method_call);
         break;
     case EngineID::MAXWELL_DMA_COPY_A:
         maxwell_dma->CallMethod(method_call);
@@ -282,7 +284,7 @@ void GPU::ProcessSemaphoreTriggerMethod() {
         block.sequence = regs.semaphore_sequence;
         // TODO(Kmather73): Generate a real GPU timestamp and write it here instead of
         // CoreTiming
-        block.timestamp = CoreTiming::GetTicks();
+        block.timestamp = Core::System::GetInstance().CoreTiming().GetTicks();
         Memory::WriteBlock(*address, &block, sizeof(block));
     } else {
         const auto address =

src/video_core/gpu.h

@@ -80,6 +80,7 @@ class DebugContext;
 struct FramebufferConfig {
     enum class PixelFormat : u32 {
         ABGR8 = 1,
+        BGRA8 = 5,
     };
 
     /**
@@ -102,15 +103,15 @@ struct FramebufferConfig {
 namespace Engines {
 class Fermi2D;
 class Maxwell3D;
-class MaxwellCompute;
 class MaxwellDMA;
+class KeplerCompute;
 class KeplerMemory;
 } // namespace Engines
 
 enum class EngineID {
     FERMI_TWOD_A = 0x902D, // 2D Engine
     MAXWELL_B = 0xB197,    // 3D Engine
-    MAXWELL_COMPUTE_B = 0xB1C0,
+    KEPLER_COMPUTE_B = 0xB1C0,
     KEPLER_INLINE_TO_MEMORY_B = 0xA140,
     MAXWELL_DMA_COPY_A = 0xB0B5,
 };
@@ -208,7 +209,7 @@ private:
     /// 2D engine
     std::unique_ptr<Engines::Fermi2D> fermi_2d;
     /// Compute engine
-    std::unique_ptr<Engines::MaxwellCompute> maxwell_compute;
+    std::unique_ptr<Engines::KeplerCompute> kepler_compute;
     /// DMA engine
     std::unique_ptr<Engines::MaxwellDMA> maxwell_dma;
     /// Inline memory engine

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -616,17 +616,8 @@ private:
 
     std::string VisitOperand(Operation operation, std::size_t operand_index, Type type) {
         std::string value = VisitOperand(operation, operand_index);
-
         switch (type) {
-        case Type::Bool:
-        case Type::Bool2:
-        case Type::Float:
-            return value;
-        case Type::Int:
-            return "ftoi(" + value + ')';
-        case Type::Uint:
-            return "ftou(" + value + ')';
-        case Type::HalfFloat:
+        case Type::HalfFloat: {
             const auto half_meta = std::get_if<MetaHalfArithmetic>(&operation.GetMeta());
             if (!half_meta) {
                 value = "toHalf2(" + value + ')';
@@ -643,6 +634,26 @@ private:
                 return "vec2(toHalf2(" + value + ")[1])";
             }
         }
+        default:
+            return CastOperand(value, type);
+        }
+    }
+
+    std::string CastOperand(const std::string& value, Type type) const {
+        switch (type) {
+        case Type::Bool:
+        case Type::Bool2:
+        case Type::Float:
+            return value;
+        case Type::Int:
+            return "ftoi(" + value + ')';
+        case Type::Uint:
+            return "ftou(" + value + ')';
+        case Type::HalfFloat:
+            // Can't be handled as a stand-alone value
+            UNREACHABLE();
+            return value;
+        }
         UNREACHABLE();
         return value;
     }
@@ -650,6 +661,7 @@ private:
     std::string BitwiseCastResult(std::string value, Type type, bool needs_parenthesis = false) {
         switch (type) {
         case Type::Bool:
+        case Type::Bool2:
         case Type::Float:
             if (needs_parenthesis) {
                 return '(' + value + ')';
@@ -719,45 +731,51 @@ private:
         constexpr std::array<const char*, 4> coord_constructors = {"float", "vec2", "vec3", "vec4"};
 
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
-        const auto count = static_cast<u32>(operation.GetOperandsCount());
         ASSERT(meta);
 
+        const std::size_t count = operation.GetOperandsCount();
+        const bool has_array = meta->sampler.IsArray();
+        const bool has_shadow = meta->sampler.IsShadow();
+
         std::string expr = func;
         expr += '(';
         expr += GetSampler(meta->sampler);
         expr += ", ";
 
-        expr += coord_constructors[meta->coords_count - 1];
+        expr += coord_constructors.at(count + (has_array ? 1 : 0) + (has_shadow ? 1 : 0) - 1);
         expr += '(';
-        for (u32 i = 0; i < count; ++i) {
-            const bool is_extra = i >= meta->coords_count;
-            const bool is_array = i == meta->array_index;
+        for (std::size_t i = 0; i < count; ++i) {
+            expr += Visit(operation[i]);
 
-            std::string operand = [&]() {
-                if (is_extra && is_extra_int) {
-                    if (const auto immediate = std::get_if<ImmediateNode>(operation[i])) {
-                        return std::to_string(static_cast<s32>(immediate->GetValue()));
-                    } else {
-                        return "ftoi(" + Visit(operation[i]) + ')';
-                    }
-                } else {
-                    return Visit(operation[i]);
-                }
-            }();
-            if (is_array) {
-                ASSERT(!is_extra);
-                operand = "float(ftoi(" + operand + "))";
-            }
-
-            expr += operand;
-
-            if (i + 1 == meta->coords_count) {
-                expr += ')';
-            }
-            if (i + 1 < count) {
+            const std::size_t next = i + 1;
+            if (next < count || has_array || has_shadow)
                 expr += ", ";
+        }
+        if (has_array) {
+            expr += "float(ftoi(" + Visit(meta->array) + "))";
+        }
+        if (has_shadow) {
+            if (has_array)
+                expr += ", ";
+            expr += Visit(meta->depth_compare);
+        }
+        expr += ')';
+
+        for (const Node extra : meta->extras) {
+            expr += ", ";
+            if (is_extra_int) {
+                if (const auto immediate = std::get_if<ImmediateNode>(extra)) {
+                    // Inline the string as an immediate integer in GLSL (some extra arguments are
+                    // required to be constant)
+                    expr += std::to_string(static_cast<s32>(immediate->GetValue()));
+                } else {
+                    expr += "ftoi(" + Visit(extra) + ')';
+                }
+            } else {
+                expr += Visit(extra);
             }
         }
+
         expr += ')';
         return expr;
     }
@@ -1134,7 +1152,7 @@ private:
                                   Type::HalfFloat);
     }
 
-    std::string F4Texture(Operation operation) {
+    std::string Texture(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
@@ -1145,7 +1163,7 @@ private:
         return expr + GetSwizzle(meta->element);
     }
 
-    std::string F4TextureLod(Operation operation) {
+    std::string TextureLod(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
@@ -1156,7 +1174,7 @@ private:
         return expr + GetSwizzle(meta->element);
     }
 
-    std::string F4TextureGather(Operation operation) {
+    std::string TextureGather(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
@@ -1164,7 +1182,7 @@ private:
                GetSwizzle(meta->element);
     }
 
-    std::string F4TextureQueryDimensions(Operation operation) {
+    std::string TextureQueryDimensions(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
@@ -1184,7 +1202,7 @@ private:
         return "0";
     }
 
-    std::string F4TextureQueryLod(Operation operation) {
+    std::string TextureQueryLod(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
@@ -1195,29 +1213,33 @@ private:
         return "0";
     }
 
-    std::string F4TexelFetch(Operation operation) {
+    std::string TexelFetch(Operation operation) {
         constexpr std::array<const char*, 4> constructors = {"int", "ivec2", "ivec3", "ivec4"};
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
-        const auto count = static_cast<u32>(operation.GetOperandsCount());
         ASSERT(meta);
+        UNIMPLEMENTED_IF(meta->sampler.IsArray());
+        const std::size_t count = operation.GetOperandsCount();
 
         std::string expr = "texelFetch(";
         expr += GetSampler(meta->sampler);
         expr += ", ";
 
-        expr += constructors[meta->coords_count - 1];
+        expr += constructors.at(operation.GetOperandsCount() - 1);
         expr += '(';
-        for (u32 i = 0; i < count; ++i) {
+        for (std::size_t i = 0; i < count; ++i) {
             expr += VisitOperand(operation, i, Type::Int);
-
-            if (i + 1 == meta->coords_count) {
+            const std::size_t next = i + 1;
+            if (next == count)
                 expr += ')';
-            }
-            if (i + 1 < count) {
+            else if (next < count)
                 expr += ", ";
-            }
+        }
+        for (std::size_t i = 0; i < meta->extras.size(); ++i) {
+            expr += ", ";
+            expr += CastOperand(Visit(meta->extras.at(i)), Type::Int);
         }
         expr += ')';
+
         return expr + GetSwizzle(meta->element);
     }
 
@@ -1454,12 +1476,12 @@ private:
         &GLSLDecompiler::Logical2HNotEqual,
         &GLSLDecompiler::Logical2HGreaterEqual,
 
-        &GLSLDecompiler::F4Texture,
-        &GLSLDecompiler::F4TextureLod,
-        &GLSLDecompiler::F4TextureGather,
-        &GLSLDecompiler::F4TextureQueryDimensions,
-        &GLSLDecompiler::F4TextureQueryLod,
-        &GLSLDecompiler::F4TexelFetch,
+        &GLSLDecompiler::Texture,
+        &GLSLDecompiler::TextureLod,
+        &GLSLDecompiler::TextureGather,
+        &GLSLDecompiler::TextureQueryDimensions,
+        &GLSLDecompiler::TextureQueryLod,
+        &GLSLDecompiler::TexelFetch,
 
         &GLSLDecompiler::Branch,
         &GLSLDecompiler::PushFlowStack,

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -107,7 +107,7 @@ RendererOpenGL::~RendererOpenGL() = default;
 void RendererOpenGL::SwapBuffers(
     std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
 
-    Core::System::GetInstance().GetPerfStats().EndSystemFrame();
+    system.GetPerfStats().EndSystemFrame();
 
     // Maintain the rasterizer's state as a priority
     OpenGLState prev_state = OpenGLState::GetCurState();
@@ -137,8 +137,8 @@ void RendererOpenGL::SwapBuffers(
 
     render_window.PollEvents();
 
-    Core::System::GetInstance().FrameLimiter().DoFrameLimiting(CoreTiming::GetGlobalTimeUs());
-    Core::System::GetInstance().GetPerfStats().BeginSystemFrame();
+    system.FrameLimiter().DoFrameLimiting(system.CoreTiming().GetGlobalTimeUs());
+    system.GetPerfStats().BeginSystemFrame();
 
     // Restore the rasterizer state
     prev_state.Apply();
@@ -380,7 +380,8 @@ void RendererOpenGL::CaptureScreenshot() {
     GLuint renderbuffer;
     glGenRenderbuffers(1, &renderbuffer);
     glBindRenderbuffer(GL_RENDERBUFFER, renderbuffer);
-    glRenderbufferStorage(GL_RENDERBUFFER, GL_RGB8, layout.width, layout.height);
+    glRenderbufferStorage(GL_RENDERBUFFER, state.GetsRGBUsed() ? GL_SRGB8 : GL_RGB8, layout.width,
+                          layout.height);
     glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, renderbuffer);
 
     DrawScreen(layout);

src/video_core/renderer_vulkan/declarations.h

@@ -0,0 +1,45 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <vulkan/vulkan.hpp>
+
+namespace Vulkan {
+
+// vulkan.hpp unique handlers use DispatchLoaderStatic
+template <typename T>
+using UniqueHandle = vk::UniqueHandle<T, vk::DispatchLoaderDynamic>;
+
+using UniqueAccelerationStructureNV = UniqueHandle<vk::AccelerationStructureNV>;
+using UniqueBuffer = UniqueHandle<vk::Buffer>;
+using UniqueBufferView = UniqueHandle<vk::BufferView>;
+using UniqueCommandBuffer = UniqueHandle<vk::CommandBuffer>;
+using UniqueCommandPool = UniqueHandle<vk::CommandPool>;
+using UniqueDescriptorPool = UniqueHandle<vk::DescriptorPool>;
+using UniqueDescriptorSet = UniqueHandle<vk::DescriptorSet>;
+using UniqueDescriptorSetLayout = UniqueHandle<vk::DescriptorSetLayout>;
+using UniqueDescriptorUpdateTemplate = UniqueHandle<vk::DescriptorUpdateTemplate>;
+using UniqueDevice = UniqueHandle<vk::Device>;
+using UniqueDeviceMemory = UniqueHandle<vk::DeviceMemory>;
+using UniqueEvent = UniqueHandle<vk::Event>;
+using UniqueFence = UniqueHandle<vk::Fence>;
+using UniqueFramebuffer = UniqueHandle<vk::Framebuffer>;
+using UniqueImage = UniqueHandle<vk::Image>;
+using UniqueImageView = UniqueHandle<vk::ImageView>;
+using UniqueIndirectCommandsLayoutNVX = UniqueHandle<vk::IndirectCommandsLayoutNVX>;
+using UniqueObjectTableNVX = UniqueHandle<vk::ObjectTableNVX>;
+using UniquePipeline = UniqueHandle<vk::Pipeline>;
+using UniquePipelineCache = UniqueHandle<vk::PipelineCache>;
+using UniquePipelineLayout = UniqueHandle<vk::PipelineLayout>;
+using UniqueQueryPool = UniqueHandle<vk::QueryPool>;
+using UniqueRenderPass = UniqueHandle<vk::RenderPass>;
+using UniqueSampler = UniqueHandle<vk::Sampler>;
+using UniqueSamplerYcbcrConversion = UniqueHandle<vk::SamplerYcbcrConversion>;
+using UniqueSemaphore = UniqueHandle<vk::Semaphore>;
+using UniqueShaderModule = UniqueHandle<vk::ShaderModule>;
+using UniqueSwapchainKHR = UniqueHandle<vk::SwapchainKHR>;
+using UniqueValidationCacheEXT = UniqueHandle<vk::ValidationCacheEXT>;
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_device.cpp

@@ -0,0 +1,231 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <map>
+#include <optional>
+#include <set>
+#include <vector>
+#include "common/assert.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+
+namespace Vulkan {
+
+namespace Alternatives {
+
+constexpr std::array<vk::Format, 3> Depth24UnormS8Uint = {
+    vk::Format::eD32SfloatS8Uint, vk::Format::eD16UnormS8Uint, {}};
+constexpr std::array<vk::Format, 3> Depth16UnormS8Uint = {
+    vk::Format::eD24UnormS8Uint, vk::Format::eD32SfloatS8Uint, {}};
+
+} // namespace Alternatives
+
+constexpr const vk::Format* GetFormatAlternatives(vk::Format format) {
+    switch (format) {
+    case vk::Format::eD24UnormS8Uint:
+        return Alternatives::Depth24UnormS8Uint.data();
+    case vk::Format::eD16UnormS8Uint:
+        return Alternatives::Depth16UnormS8Uint.data();
+    default:
+        return nullptr;
+    }
+}
+
+constexpr vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties,
+                                                   FormatType format_type) {
+    switch (format_type) {
+    case FormatType::Linear:
+        return properties.linearTilingFeatures;
+    case FormatType::Optimal:
+        return properties.optimalTilingFeatures;
+    case FormatType::Buffer:
+        return properties.bufferFeatures;
+    default:
+        return {};
+    }
+}
+
+VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+                   vk::SurfaceKHR surface)
+    : physical{physical}, format_properties{GetFormatProperties(dldi, physical)} {
+    SetupFamilies(dldi, surface);
+    SetupProperties(dldi);
+}
+
+VKDevice::~VKDevice() = default;
+
+bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance) {
+    const auto queue_cis = GetDeviceQueueCreateInfos();
+    vk::PhysicalDeviceFeatures device_features{};
+
+    const std::vector<const char*> extensions = {VK_KHR_SWAPCHAIN_EXTENSION_NAME};
+    const vk::DeviceCreateInfo device_ci({}, static_cast<u32>(queue_cis.size()), queue_cis.data(),
+                                         0, nullptr, static_cast<u32>(extensions.size()),
+                                         extensions.data(), &device_features);
+    vk::Device dummy_logical;
+    if (physical.createDevice(&device_ci, nullptr, &dummy_logical, dldi) != vk::Result::eSuccess) {
+        LOG_CRITICAL(Render_Vulkan, "Logical device failed to be created!");
+        return false;
+    }
+
+    dld.init(instance, dldi.vkGetInstanceProcAddr, dummy_logical, dldi.vkGetDeviceProcAddr);
+    logical = UniqueDevice(
+        dummy_logical, vk::ObjectDestroy<vk::NoParent, vk::DispatchLoaderDynamic>(nullptr, dld));
+
+    graphics_queue = logical->getQueue(graphics_family, 0, dld);
+    present_queue = logical->getQueue(present_family, 0, dld);
+    return true;
+}
+
+vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format,
+                                        vk::FormatFeatureFlags wanted_usage,
+                                        FormatType format_type) const {
+    if (IsFormatSupported(wanted_format, wanted_usage, format_type)) {
+        return wanted_format;
+    }
+    // The wanted format is not supported by hardware, search for alternatives
+    const vk::Format* alternatives = GetFormatAlternatives(wanted_format);
+    if (alternatives == nullptr) {
+        LOG_CRITICAL(Render_Vulkan,
+                     "Format={} with usage={} and type={} has no defined alternatives and host "
+                     "hardware does not support it",
+                     static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage),
+                     static_cast<u32>(format_type));
+        UNREACHABLE();
+        return wanted_format;
+    }
+
+    std::size_t i = 0;
+    for (vk::Format alternative = alternatives[0]; alternative != vk::Format{};
+         alternative = alternatives[++i]) {
+        if (!IsFormatSupported(alternative, wanted_usage, format_type))
+            continue;
+        LOG_WARNING(Render_Vulkan,
+                    "Emulating format={} with alternative format={} with usage={} and type={}",
+                    static_cast<u32>(wanted_format), static_cast<u32>(alternative),
+                    static_cast<u32>(wanted_usage), static_cast<u32>(format_type));
+        return alternative;
+    }
+
+    // No alternatives found, panic
+    LOG_CRITICAL(Render_Vulkan,
+                 "Format={} with usage={} and type={} is not supported by the host hardware and "
+                 "doesn't support any of the alternatives",
+                 static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage),
+                 static_cast<u32>(format_type));
+    UNREACHABLE();
+    return wanted_format;
+}
+
+bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
+                                 FormatType format_type) const {
+    const auto it = format_properties.find(wanted_format);
+    if (it == format_properties.end()) {
+        LOG_CRITICAL(Render_Vulkan, "Unimplemented format query={}",
+                     static_cast<u32>(wanted_format));
+        UNREACHABLE();
+        return true;
+    }
+    const vk::FormatFeatureFlags supported_usage = GetFormatFeatures(it->second, format_type);
+    return (supported_usage & wanted_usage) == wanted_usage;
+}
+
+bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+                          vk::SurfaceKHR surface) {
+    const std::string swapchain_extension = VK_KHR_SWAPCHAIN_EXTENSION_NAME;
+
+    bool has_swapchain{};
+    for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
+        has_swapchain |= prop.extensionName == swapchain_extension;
+    }
+    if (!has_swapchain) {
+        // The device doesn't support creating swapchains.
+        return false;
+    }
+
+    bool has_graphics{}, has_present{};
+    const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+    for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
+        const auto& family = queue_family_properties[i];
+        if (family.queueCount == 0)
+            continue;
+
+        has_graphics |=
+            (family.queueFlags & vk::QueueFlagBits::eGraphics) != static_cast<vk::QueueFlagBits>(0);
+        has_present |= physical.getSurfaceSupportKHR(i, surface, dldi) != 0;
+    }
+    if (!has_graphics || !has_present) {
+        // The device doesn't have a graphics and present queue.
+        return false;
+    }
+
+    // TODO(Rodrigo): Check if the device matches all requeriments.
+    const vk::PhysicalDeviceProperties props = physical.getProperties(dldi);
+    if (props.limits.maxUniformBufferRange < 65536) {
+        return false;
+    }
+
+    // Device is suitable.
+    return true;
+}
+
+void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface) {
+    std::optional<u32> graphics_family_, present_family_;
+
+    const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+    for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
+        if (graphics_family_ && present_family_)
+            break;
+
+        const auto& queue_family = queue_family_properties[i];
+        if (queue_family.queueCount == 0)
+            continue;
+
+        if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics)
+            graphics_family_ = i;
+        if (physical.getSurfaceSupportKHR(i, surface, dldi))
+            present_family_ = i;
+    }
+    ASSERT(graphics_family_ && present_family_);
+
+    graphics_family = *graphics_family_;
+    present_family = *present_family_;
+}
+
+void VKDevice::SetupProperties(const vk::DispatchLoaderDynamic& dldi) {
+    const vk::PhysicalDeviceProperties props = physical.getProperties(dldi);
+    device_type = props.deviceType;
+    uniform_buffer_alignment = static_cast<u64>(props.limits.minUniformBufferOffsetAlignment);
+}
+
+std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const {
+    static const float QUEUE_PRIORITY = 1.f;
+
+    std::set<u32> unique_queue_families = {graphics_family, present_family};
+    std::vector<vk::DeviceQueueCreateInfo> queue_cis;
+
+    for (u32 queue_family : unique_queue_families)
+        queue_cis.push_back({{}, queue_family, 1, &QUEUE_PRIORITY});
+
+    return queue_cis;
+}
+
+std::map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
+    const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) {
+    std::map<vk::Format, vk::FormatProperties> format_properties;
+
+    const auto AddFormatQuery = [&format_properties, &dldi, physical](vk::Format format) {
+        format_properties.emplace(format, physical.getFormatProperties(format, dldi));
+    };
+    AddFormatQuery(vk::Format::eA8B8G8R8UnormPack32);
+    AddFormatQuery(vk::Format::eR5G6B5UnormPack16);
+    AddFormatQuery(vk::Format::eD32Sfloat);
+    AddFormatQuery(vk::Format::eD16UnormS8Uint);
+    AddFormatQuery(vk::Format::eD24UnormS8Uint);
+    AddFormatQuery(vk::Format::eD32SfloatS8Uint);
+
+    return format_properties;
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_device.h

@@ -0,0 +1,116 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <map>
+#include <vector>
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+
+namespace Vulkan {
+
+/// Format usage descriptor
+enum class FormatType { Linear, Optimal, Buffer };
+
+/// Handles data specific to a physical device.
+class VKDevice final {
+public:
+    explicit VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+                      vk::SurfaceKHR surface);
+    ~VKDevice();
+
+    /// Initializes the device. Returns true on success.
+    bool Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance);
+
+    /**
+     * Returns a format supported by the device for the passed requeriments.
+     * @param wanted_format The ideal format to be returned. It may not be the returned format.
+     * @param wanted_usage The usage that must be fulfilled even if the format is not supported.
+     * @param format_type Format type usage.
+     * @returns A format supported by the device.
+     */
+    vk::Format GetSupportedFormat(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
+                                  FormatType format_type) const;
+
+    /// Returns the dispatch loader with direct function pointers of the device
+    const vk::DispatchLoaderDynamic& GetDispatchLoader() const {
+        return dld;
+    }
+
+    /// Returns the logical device
+    vk::Device GetLogical() const {
+        return logical.get();
+    }
+
+    /// Returns the physical device.
+    vk::PhysicalDevice GetPhysical() const {
+        return physical;
+    }
+
+    /// Returns the main graphics queue.
+    vk::Queue GetGraphicsQueue() const {
+        return graphics_queue;
+    }
+
+    /// Returns the main present queue.
+    vk::Queue GetPresentQueue() const {
+        return present_queue;
+    }
+
+    /// Returns main graphics queue family index.
+    u32 GetGraphicsFamily() const {
+        return graphics_family;
+    }
+
+    /// Returns main present queue family index.
+    u32 GetPresentFamily() const {
+        return present_family;
+    }
+
+    /// Returns if the device is integrated with the host CPU
+    bool IsIntegrated() const {
+        return device_type == vk::PhysicalDeviceType::eIntegratedGpu;
+    }
+
+    /// Returns uniform buffer alignment requeriment
+    u64 GetUniformBufferAlignment() const {
+        return uniform_buffer_alignment;
+    }
+
+    /// Checks if the physical device is suitable.
+    static bool IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+                           vk::SurfaceKHR surface);
+
+private:
+    /// Sets up queue families.
+    void SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface);
+
+    /// Sets up device properties.
+    void SetupProperties(const vk::DispatchLoaderDynamic& dldi);
+
+    /// Returns a list of queue initialization descriptors.
+    std::vector<vk::DeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
+
+    /// Returns true if a format is supported.
+    bool IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
+                           FormatType format_type) const;
+
+    /// Returns the device properties for Vulkan formats.
+    static std::map<vk::Format, vk::FormatProperties> GetFormatProperties(
+        const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical);
+
+    const vk::PhysicalDevice physical;  ///< Physical device
+    vk::DispatchLoaderDynamic dld;      ///< Device function pointers
+    UniqueDevice logical;               ///< Logical device
+    vk::Queue graphics_queue;           ///< Main graphics queue
+    vk::Queue present_queue;            ///< Main present queue
+    u32 graphics_family{};              ///< Main graphics queue family index
+    u32 present_family{};               ///< Main present queue family index
+    vk::PhysicalDeviceType device_type; ///< Physical device type
+    u64 uniform_buffer_alignment{};     ///< Uniform buffer alignment requeriment
+    std::map<vk::Format, vk::FormatProperties> format_properties; ///< Format properties dictionary
+};
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_resource_manager.cpp

@@ -0,0 +1,285 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <optional>
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+
+namespace Vulkan {
+
+// TODO(Rodrigo): Fine tune these numbers.
+constexpr std::size_t COMMAND_BUFFER_POOL_SIZE = 0x1000;
+constexpr std::size_t FENCES_GROW_STEP = 0x40;
+
+class CommandBufferPool final : public VKFencedPool {
+public:
+    CommandBufferPool(const VKDevice& device)
+        : VKFencedPool(COMMAND_BUFFER_POOL_SIZE), device{device} {}
+
+    void Allocate(std::size_t begin, std::size_t end) {
+        const auto dev = device.GetLogical();
+        const auto& dld = device.GetDispatchLoader();
+        const u32 graphics_family = device.GetGraphicsFamily();
+
+        auto pool = std::make_unique<Pool>();
+
+        // Command buffers are going to be commited, recorded, executed every single usage cycle.
+        // They are also going to be reseted when commited.
+        const auto pool_flags = vk::CommandPoolCreateFlagBits::eTransient |
+                                vk::CommandPoolCreateFlagBits::eResetCommandBuffer;
+        const vk::CommandPoolCreateInfo cmdbuf_pool_ci(pool_flags, graphics_family);
+        pool->handle = dev.createCommandPoolUnique(cmdbuf_pool_ci, nullptr, dld);
+
+        const vk::CommandBufferAllocateInfo cmdbuf_ai(*pool->handle,
+                                                      vk::CommandBufferLevel::ePrimary,
+                                                      static_cast<u32>(COMMAND_BUFFER_POOL_SIZE));
+        pool->cmdbufs =
+            dev.allocateCommandBuffersUnique<std::allocator<UniqueCommandBuffer>>(cmdbuf_ai, dld);
+
+        pools.push_back(std::move(pool));
+    }
+
+    vk::CommandBuffer Commit(VKFence& fence) {
+        const std::size_t index = CommitResource(fence);
+        const auto pool_index = index / COMMAND_BUFFER_POOL_SIZE;
+        const auto sub_index = index % COMMAND_BUFFER_POOL_SIZE;
+        return *pools[pool_index]->cmdbufs[sub_index];
+    }
+
+private:
+    struct Pool {
+        UniqueCommandPool handle;
+        std::vector<UniqueCommandBuffer> cmdbufs;
+    };
+
+    const VKDevice& device;
+
+    std::vector<std::unique_ptr<Pool>> pools;
+};
+
+VKResource::VKResource() = default;
+
+VKResource::~VKResource() = default;
+
+VKFence::VKFence(const VKDevice& device, UniqueFence handle)
+    : device{device}, handle{std::move(handle)} {}
+
+VKFence::~VKFence() = default;
+
+void VKFence::Wait() {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    dev.waitForFences({*handle}, true, std::numeric_limits<u64>::max(), dld);
+}
+
+void VKFence::Release() {
+    is_owned = false;
+}
+
+void VKFence::Commit() {
+    is_owned = true;
+    is_used = true;
+}
+
+bool VKFence::Tick(bool gpu_wait, bool owner_wait) {
+    if (!is_used) {
+        // If a fence is not used it's always free.
+        return true;
+    }
+    if (is_owned && !owner_wait) {
+        // The fence is still being owned (Release has not been called) and ownership wait has
+        // not been asked.
+        return false;
+    }
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    if (gpu_wait) {
+        // Wait for the fence if it has been requested.
+        dev.waitForFences({*handle}, true, std::numeric_limits<u64>::max(), dld);
+    } else {
+        if (dev.getFenceStatus(*handle, dld) != vk::Result::eSuccess) {
+            // Vulkan fence is not ready, not much it can do here
+            return false;
+        }
+    }
+
+    // Broadcast resources their free state.
+    for (auto* resource : protected_resources) {
+        resource->OnFenceRemoval(this);
+    }
+    protected_resources.clear();
+
+    // Prepare fence for reusage.
+    dev.resetFences({*handle}, dld);
+    is_used = false;
+    return true;
+}
+
+void VKFence::Protect(VKResource* resource) {
+    protected_resources.push_back(resource);
+}
+
+void VKFence::Unprotect(const VKResource* resource) {
+    const auto it = std::find(protected_resources.begin(), protected_resources.end(), resource);
+    if (it != protected_resources.end()) {
+        protected_resources.erase(it);
+    }
+}
+
+VKFenceWatch::VKFenceWatch() = default;
+
+VKFenceWatch::~VKFenceWatch() {
+    if (fence) {
+        fence->Unprotect(this);
+    }
+}
+
+void VKFenceWatch::Wait() {
+    if (!fence) {
+        return;
+    }
+    fence->Wait();
+    fence->Unprotect(this);
+    fence = nullptr;
+}
+
+void VKFenceWatch::Watch(VKFence& new_fence) {
+    Wait();
+    fence = &new_fence;
+    fence->Protect(this);
+}
+
+bool VKFenceWatch::TryWatch(VKFence& new_fence) {
+    if (fence) {
+        return false;
+    }
+    fence = &new_fence;
+    fence->Protect(this);
+    return true;
+}
+
+void VKFenceWatch::OnFenceRemoval(VKFence* signaling_fence) {
+    ASSERT_MSG(signaling_fence == fence, "Removing the wrong fence");
+    fence = nullptr;
+}
+
+VKFencedPool::VKFencedPool(std::size_t grow_step) : grow_step{grow_step} {}
+
+VKFencedPool::~VKFencedPool() = default;
+
+std::size_t VKFencedPool::CommitResource(VKFence& fence) {
+    const auto Search = [&](std::size_t begin, std::size_t end) -> std::optional<std::size_t> {
+        for (std::size_t iterator = begin; iterator < end; ++iterator) {
+            if (watches[iterator]->TryWatch(fence)) {
+                // The resource is now being watched, a free resource was successfully found.
+                return iterator;
+            }
+        }
+        return {};
+    };
+    // Try to find a free resource from the hinted position to the end.
+    auto found = Search(free_iterator, watches.size());
+    if (!found) {
+        // Search from beginning to the hinted position.
+        found = Search(0, free_iterator);
+        if (!found) {
+            // Both searches failed, the pool is full; handle it.
+            const std::size_t free_resource = ManageOverflow();
+
+            // Watch will wait for the resource to be free.
+            watches[free_resource]->Watch(fence);
+            found = free_resource;
+        }
+    }
+    // Free iterator is hinted to the resource after the one that's been commited.
+    free_iterator = (*found + 1) % watches.size();
+    return *found;
+}
+
+std::size_t VKFencedPool::ManageOverflow() {
+    const std::size_t old_capacity = watches.size();
+    Grow();
+
+    // The last entry is guaranted to be free, since it's the first element of the freshly
+    // allocated resources.
+    return old_capacity;
+}
+
+void VKFencedPool::Grow() {
+    const std::size_t old_capacity = watches.size();
+    watches.resize(old_capacity + grow_step);
+    std::generate(watches.begin() + old_capacity, watches.end(),
+                  []() { return std::make_unique<VKFenceWatch>(); });
+    Allocate(old_capacity, old_capacity + grow_step);
+}
+
+VKResourceManager::VKResourceManager(const VKDevice& device) : device{device} {
+    GrowFences(FENCES_GROW_STEP);
+    command_buffer_pool = std::make_unique<CommandBufferPool>(device);
+}
+
+VKResourceManager::~VKResourceManager() = default;
+
+VKFence& VKResourceManager::CommitFence() {
+    const auto StepFences = [&](bool gpu_wait, bool owner_wait) -> VKFence* {
+        const auto Tick = [=](auto& fence) { return fence->Tick(gpu_wait, owner_wait); };
+        const auto hinted = fences.begin() + fences_iterator;
+
+        auto it = std::find_if(hinted, fences.end(), Tick);
+        if (it == fences.end()) {
+            it = std::find_if(fences.begin(), hinted, Tick);
+            if (it == hinted) {
+                return nullptr;
+            }
+        }
+        fences_iterator = std::distance(fences.begin(), it) + 1;
+        if (fences_iterator >= fences.size())
+            fences_iterator = 0;
+
+        auto& fence = *it;
+        fence->Commit();
+        return fence.get();
+    };
+
+    VKFence* found_fence = StepFences(false, false);
+    if (!found_fence) {
+        // Try again, this time waiting.
+        found_fence = StepFences(true, false);
+
+        if (!found_fence) {
+            // Allocate new fences and try again.
+            LOG_INFO(Render_Vulkan, "Allocating new fences {} -> {}", fences.size(),
+                     fences.size() + FENCES_GROW_STEP);
+
+            GrowFences(FENCES_GROW_STEP);
+            found_fence = StepFences(true, false);
+            ASSERT(found_fence != nullptr);
+        }
+    }
+    return *found_fence;
+}
+
+vk::CommandBuffer VKResourceManager::CommitCommandBuffer(VKFence& fence) {
+    return command_buffer_pool->Commit(fence);
+}
+
+void VKResourceManager::GrowFences(std::size_t new_fences_count) {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    const vk::FenceCreateInfo fence_ci;
+
+    const std::size_t previous_size = fences.size();
+    fences.resize(previous_size + new_fences_count);
+
+    std::generate(fences.begin() + previous_size, fences.end(), [&]() {
+        return std::make_unique<VKFence>(device, dev.createFenceUnique(fence_ci, nullptr, dld));
+    });
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_resource_manager.h

@@ -0,0 +1,180 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <memory>
+#include <vector>
+#include "video_core/renderer_vulkan/declarations.h"
+
+namespace Vulkan {
+
+class VKDevice;
+class VKFence;
+class VKResourceManager;
+
+class CommandBufferPool;
+
+/// Interface for a Vulkan resource
+class VKResource {
+public:
+    explicit VKResource();
+    virtual ~VKResource();
+
+    /**
+     * Signals the object that an owning fence has been signaled.
+     * @param signaling_fence Fence that signals its usage end.
+     */
+    virtual void OnFenceRemoval(VKFence* signaling_fence) = 0;
+};
+
+/**
+ * Fences take ownership of objects, protecting them from GPU-side or driver-side concurrent access.
+ * They must be commited from the resource manager. Their usage flow is: commit the fence from the
+ * resource manager, protect resources with it and use them, send the fence to an execution queue
+ * and Wait for it if needed and then call Release. Used resources will automatically be signaled
+ * when they are free to be reused.
+ * @brief Protects resources for concurrent usage and signals its release.
+ */
+class VKFence {
+    friend class VKResourceManager;
+
+public:
+    explicit VKFence(const VKDevice& device, UniqueFence handle);
+    ~VKFence();
+
+    /**
+     * Waits for the fence to be signaled.
+     * @warning You must have ownership of the fence and it has to be previously sent to a queue to
+     * call this function.
+     */
+    void Wait();
+
+    /**
+     * Releases ownership of the fence. Pass after it has been sent to an execution queue.
+     * Unmanaged usage of the fence after the call will result in undefined behavior because it may
+     * be being used for something else.
+     */
+    void Release();
+
+    /// Protects a resource with this fence.
+    void Protect(VKResource* resource);
+
+    /// Removes protection for a resource.
+    void Unprotect(const VKResource* resource);
+
+    /// Retreives the fence.
+    operator vk::Fence() const {
+        return *handle;
+    }
+
+private:
+    /// Take ownership of the fence.
+    void Commit();
+
+    /**
+     * Updates the fence status.
+     * @warning Waiting for the owner might soft lock the execution.
+     * @param gpu_wait Wait for the fence to be signaled by the driver.
+     * @param owner_wait Wait for the owner to signal its freedom.
+     * @returns True if the fence is free. Waiting for gpu and owner will always return true.
+     */
+    bool Tick(bool gpu_wait, bool owner_wait);
+
+    const VKDevice& device;                       ///< Device handler
+    UniqueFence handle;                           ///< Vulkan fence
+    std::vector<VKResource*> protected_resources; ///< List of resources protected by this fence
+    bool is_owned = false; ///< The fence has been commited but not released yet.
+    bool is_used = false;  ///< The fence has been commited but it has not been checked to be free.
+};
+
+/**
+ * A fence watch is used to keep track of the usage of a fence and protect a resource or set of
+ * resources without having to inherit VKResource from their handlers.
+ */
+class VKFenceWatch final : public VKResource {
+public:
+    explicit VKFenceWatch();
+    ~VKFenceWatch();
+
+    /// Waits for the fence to be released.
+    void Wait();
+
+    /**
+     * Waits for a previous fence and watches a new one.
+     * @param new_fence New fence to wait to.
+     */
+    void Watch(VKFence& new_fence);
+
+    /**
+     * Checks if it's currently being watched and starts watching it if it's available.
+     * @returns True if a watch has started, false if it's being watched.
+     */
+    bool TryWatch(VKFence& new_fence);
+
+    void OnFenceRemoval(VKFence* signaling_fence) override;
+
+private:
+    VKFence* fence{}; ///< Fence watching this resource. nullptr when the watch is free.
+};
+
+/**
+ * Handles a pool of resources protected by fences. Manages resource overflow allocating more
+ * resources.
+ */
+class VKFencedPool {
+public:
+    explicit VKFencedPool(std::size_t grow_step);
+    virtual ~VKFencedPool();
+
+protected:
+    /**
+     * Commits a free resource and protects it with a fence. It may allocate new resources.
+     * @param fence Fence that protects the commited resource.
+     * @returns Index of the resource commited.
+     */
+    std::size_t CommitResource(VKFence& fence);
+
+    /// Called when a chunk of resources have to be allocated.
+    virtual void Allocate(std::size_t begin, std::size_t end) = 0;
+
+private:
+    /// Manages pool overflow allocating new resources.
+    std::size_t ManageOverflow();
+
+    /// Allocates a new page of resources.
+    void Grow();
+
+    std::size_t grow_step = 0;     ///< Number of new resources created after an overflow
+    std::size_t free_iterator = 0; ///< Hint to where the next free resources is likely to be found
+    std::vector<std::unique_ptr<VKFenceWatch>> watches; ///< Set of watched resources
+};
+
+/**
+ * The resource manager handles all resources that can be protected with a fence avoiding
+ * driver-side or GPU-side concurrent usage. Usage is documented in VKFence.
+ */
+class VKResourceManager final {
+public:
+    explicit VKResourceManager(const VKDevice& device);
+    ~VKResourceManager();
+
+    /// Commits a fence. It has to be sent to a queue and released.
+    VKFence& CommitFence();
+
+    /// Commits an unused command buffer and protects it with a fence.
+    vk::CommandBuffer CommitCommandBuffer(VKFence& fence);
+
+private:
+    /// Allocates new fences.
+    void GrowFences(std::size_t new_fences_count);
+
+    const VKDevice& device;          ///< Device handler.
+    std::size_t fences_iterator = 0; ///< Index where a free fence is likely to be found.
+    std::vector<std::unique_ptr<VKFence>> fences;           ///< Pool of fences.
+    std::unique_ptr<CommandBufferPool> command_buffer_pool; ///< Pool of command buffers.
+};
+
+} // namespace Vulkan

src/video_core/shader/decode/arithmetic_integer.cpp

@@ -41,7 +41,7 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
 
         const Node value = Operation(OperationCode::IAdd, PRECISE, op_a, op_b);
 
-        SetInternalFlagsFromInteger(bb, value, instr.op_32.generates_cc);
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
         SetRegister(bb, instr.gpr0, value);
         break;
     }
@@ -284,4 +284,4 @@ void ShaderIR::WriteLop3Instruction(NodeBlock& bb, Register dest, Node op_a, Nod
     SetRegister(bb, dest, value);
 }
 
-} // namespace VideoCommon::Shader
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/conversion.cpp

@@ -118,8 +118,8 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
 
         value = [&]() {
             switch (instr.conversion.f2i.rounding) {
-            case Tegra::Shader::F2iRoundingOp::None:
-                return value;
+            case Tegra::Shader::F2iRoundingOp::RoundEven:
+                return Operation(OperationCode::FRoundEven, PRECISE, value);
             case Tegra::Shader::F2iRoundingOp::Floor:
                 return Operation(OperationCode::FFloor, PRECISE, value);
             case Tegra::Shader::F2iRoundingOp::Ceil:
@@ -146,4 +146,4 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
     return pc;
 }
 
-} // namespace VideoCommon::Shader
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/memory.cpp

@@ -306,7 +306,6 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
     case OpCode::Id::TLD4S: {
         UNIMPLEMENTED_IF_MSG(instr.tld4s.UsesMiscMode(TextureMiscMode::AOFFI),
                              "AOFFI is not implemented");
-
         if (instr.tld4s.UsesMiscMode(TextureMiscMode::NODEP)) {
             LOG_WARNING(HW_GPU, "TLD4S.NODEP implementation is incomplete");
         }
@@ -315,9 +314,8 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         const Node op_a = GetRegister(instr.gpr8);
         const Node op_b = GetRegister(instr.gpr20);
 
-        std::vector<Node> coords;
-
         // TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
+        std::vector<Node> coords;
         if (depth_compare) {
             // Note: TLD4S coordinate encoding works just like TEXS's
             const Node op_y = GetRegister(instr.gpr8.Value() + 1);
@@ -328,18 +326,17 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
             coords.push_back(op_a);
             coords.push_back(op_b);
         }
-        const auto num_coords = static_cast<u32>(coords.size());
-        coords.push_back(Immediate(static_cast<u32>(instr.tld4s.component)));
+        std::vector<Node> extras;
+        extras.push_back(Immediate(static_cast<u32>(instr.tld4s.component)));
 
         const auto& sampler =
             GetSampler(instr.sampler, TextureType::Texture2D, false, depth_compare);
 
         Node4 values;
         for (u32 element = 0; element < values.size(); ++element) {
-            auto params = coords;
-            MetaTexture meta{sampler, element, num_coords};
-            values[element] =
-                Operation(OperationCode::F4TextureGather, std::move(meta), std::move(params));
+            auto coords_copy = coords;
+            MetaTexture meta{sampler, {}, {}, extras, element};
+            values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
         }
 
         WriteTexsInstructionFloat(bb, instr, values);
@@ -360,12 +357,13 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         switch (instr.txq.query_type) {
         case Tegra::Shader::TextureQueryType::Dimension: {
             for (u32 element = 0; element < 4; ++element) {
-                if (instr.txq.IsComponentEnabled(element)) {
-                    MetaTexture meta{sampler, element};
-                    const Node value = Operation(OperationCode::F4TextureQueryDimensions,
-                                                 std::move(meta), GetRegister(instr.gpr8));
-                    SetTemporal(bb, indexer++, value);
+                if (!instr.txq.IsComponentEnabled(element)) {
+                    continue;
                 }
+                MetaTexture meta{sampler, {}, {}, {}, element};
+                const Node value =
+                    Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8));
+                SetTemporal(bb, indexer++, value);
             }
             for (u32 i = 0; i < indexer; ++i) {
                 SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
@@ -412,9 +410,8 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
 
         for (u32 element = 0; element < 2; ++element) {
             auto params = coords;
-            MetaTexture meta_texture{sampler, element, static_cast<u32>(coords.size())};
-            const Node value =
-                Operation(OperationCode::F4TextureQueryLod, meta_texture, std::move(params));
+            MetaTexture meta{sampler, {}, {}, {}, element};
+            const Node value = Operation(OperationCode::TextureQueryLod, meta, std::move(params));
             SetTemporal(bb, element, value);
         }
         for (u32 element = 0; element < 2; ++element) {
@@ -432,7 +429,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::MZ), "MZ is not implemented");
 
         if (instr.tlds.UsesMiscMode(TextureMiscMode::NODEP)) {
-            LOG_WARNING(HW_GPU, "TMML.NODEP implementation is incomplete");
+            LOG_WARNING(HW_GPU, "TLDS.NODEP implementation is incomplete");
         }
 
         WriteTexsInstructionFloat(bb, instr, GetTldsCode(instr, texture_type, is_array));
@@ -535,15 +532,16 @@ void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
 }
 
 Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
-                               TextureProcessMode process_mode, bool depth_compare, bool is_array,
-                               std::size_t array_offset, std::size_t bias_offset,
-                               std::vector<Node>&& coords) {
-    UNIMPLEMENTED_IF_MSG(
-        (texture_type == TextureType::Texture3D && (is_array || depth_compare)) ||
-            (texture_type == TextureType::TextureCube && is_array && depth_compare),
-        "This method is not supported.");
+                               TextureProcessMode process_mode, std::vector<Node> coords,
+                               Node array, Node depth_compare, u32 bias_offset) {
+    const bool is_array = array;
+    const bool is_shadow = depth_compare;
 
-    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
+    UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) ||
+                             (texture_type == TextureType::TextureCube && is_array && is_shadow),
+                         "This method is not supported.");
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, is_shadow);
 
     const bool lod_needed = process_mode == TextureProcessMode::LZ ||
                             process_mode == TextureProcessMode::LL ||
@@ -552,35 +550,30 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
     // LOD selection (either via bias or explicit textureLod) not supported in GL for
     // sampler2DArrayShadow and samplerCubeArrayShadow.
     const bool gl_lod_supported =
-        !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && depth_compare) ||
-          (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && depth_compare));
+        !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && is_shadow) ||
+          (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && is_shadow));
 
     const OperationCode read_method =
-        lod_needed && gl_lod_supported ? OperationCode::F4TextureLod : OperationCode::F4Texture;
+        lod_needed && gl_lod_supported ? OperationCode::TextureLod : OperationCode::Texture;
 
     UNIMPLEMENTED_IF(process_mode != TextureProcessMode::None && !gl_lod_supported);
 
-    std::optional<u32> array_offset_value;
-    if (is_array)
-        array_offset_value = static_cast<u32>(array_offset);
-
-    const auto coords_count = static_cast<u32>(coords.size());
-
+    std::vector<Node> extras;
     if (process_mode != TextureProcessMode::None && gl_lod_supported) {
         if (process_mode == TextureProcessMode::LZ) {
-            coords.push_back(Immediate(0.0f));
+            extras.push_back(Immediate(0.0f));
         } else {
             // If present, lod or bias are always stored in the register indexed by the gpr20
             // field with an offset depending on the usage of the other registers
-            coords.push_back(GetRegister(instr.gpr20.Value() + bias_offset));
+            extras.push_back(GetRegister(instr.gpr20.Value() + bias_offset));
         }
     }
 
     Node4 values;
     for (u32 element = 0; element < values.size(); ++element) {
-        auto params = coords;
-        MetaTexture meta{sampler, element, coords_count, array_offset_value};
-        values[element] = Operation(read_method, std::move(meta), std::move(params));
+        auto copy_coords = coords;
+        MetaTexture meta{sampler, array, depth_compare, extras, element};
+        values[element] = Operation(read_method, meta, std::move(copy_coords));
     }
 
     return values;
@@ -602,28 +595,22 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
     for (std::size_t i = 0; i < coord_count; ++i) {
         coords.push_back(GetRegister(coord_register + i));
     }
-    // 1D.DC in opengl the 2nd component is ignored.
+    // 1D.DC in OpenGL the 2nd component is ignored.
     if (depth_compare && !is_array && texture_type == TextureType::Texture1D) {
         coords.push_back(Immediate(0.0f));
     }
-    std::size_t array_offset{};
-    if (is_array) {
-        array_offset = coords.size();
-        coords.push_back(GetRegister(array_register));
-    }
+
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+
+    Node dc{};
     if (depth_compare) {
-        // Depth is always stored in the register signaled by gpr20
-        // or in the next register if lod or bias are used
+        // Depth is always stored in the register signaled by gpr20 or in the next register if lod
+        // or bias are used
         const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
-        coords.push_back(GetRegister(depth_register));
-    }
-    // Fill ignored coordinates
-    while (coords.size() < total_coord_count) {
-        coords.push_back(Immediate(0));
+        dc = GetRegister(depth_register);
     }
 
-    return GetTextureCode(instr, texture_type, process_mode, depth_compare, is_array, array_offset,
-                          0, std::move(coords));
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0);
 }
 
 Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
@@ -641,6 +628,7 @@ Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
         (is_array || !(lod_bias_enabled || depth_compare) || (coord_count > 2))
             ? static_cast<u64>(instr.gpr20.Value())
             : coord_register + 1;
+    const u32 bias_offset = coord_count > 2 ? 1 : 0;
 
     std::vector<Node> coords;
     for (std::size_t i = 0; i < coord_count; ++i) {
@@ -648,24 +636,17 @@ Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
         coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
     }
 
-    std::size_t array_offset{};
-    if (is_array) {
-        array_offset = coords.size();
-        coords.push_back(GetRegister(array_register));
-    }
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+
+    Node dc{};
     if (depth_compare) {
-        // Depth is always stored in the register signaled by gpr20
-        // or in the next register if lod or bias are used
+        // Depth is always stored in the register signaled by gpr20 or in the next register if lod
+        // or bias are used
         const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
-        coords.push_back(GetRegister(depth_register));
-    }
-    // Fill ignored coordinates
-    while (coords.size() < total_coord_count) {
-        coords.push_back(Immediate(0));
+        dc = GetRegister(depth_register);
     }
 
-    return GetTextureCode(instr, texture_type, process_mode, depth_compare, is_array, array_offset,
-                          (coord_count > 2 ? 1 : 0), std::move(coords));
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset);
 }
 
 Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,
@@ -680,24 +661,16 @@ Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool de
     const u64 coord_register = array_register + (is_array ? 1 : 0);
 
     std::vector<Node> coords;
-
-    for (size_t i = 0; i < coord_count; ++i) {
+    for (size_t i = 0; i < coord_count; ++i)
         coords.push_back(GetRegister(coord_register + i));
-    }
-    std::optional<u32> array_offset;
-    if (is_array) {
-        array_offset = static_cast<u32>(coords.size());
-        coords.push_back(GetRegister(array_register));
-    }
 
     const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
 
     Node4 values;
     for (u32 element = 0; element < values.size(); ++element) {
-        auto params = coords;
-        MetaTexture meta{sampler, element, static_cast<u32>(coords.size()), array_offset};
-        values[element] =
-            Operation(OperationCode::F4TextureGather, std::move(meta), std::move(params));
+        auto coords_copy = coords;
+        MetaTexture meta{sampler, GetRegister(array_register), {}, {}, element};
+        values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
     }
 
     return values;
@@ -705,7 +678,6 @@ Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool de
 
 Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is_array) {
     const std::size_t type_coord_count = GetCoordCount(texture_type);
-    const std::size_t total_coord_count = type_coord_count + (is_array ? 1 : 0);
     const bool lod_enabled = instr.tlds.GetTextureProcessMode() == TextureProcessMode::LL;
 
     // If enabled arrays index is always stored in the gpr8 field
@@ -719,33 +691,22 @@ Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is
             : coord_register + 1;
 
     std::vector<Node> coords;
-
     for (std::size_t i = 0; i < type_coord_count; ++i) {
         const bool last = (i == (type_coord_count - 1)) && (type_coord_count > 1);
         coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
     }
-    std::optional<u32> array_offset;
-    if (is_array) {
-        array_offset = static_cast<u32>(coords.size());
-        coords.push_back(GetRegister(array_register));
-    }
-    const auto coords_count = static_cast<u32>(coords.size());
 
-    if (lod_enabled) {
-        // When lod is used always is in grp20
-        coords.push_back(GetRegister(instr.gpr20));
-    } else {
-        coords.push_back(Immediate(0));
-    }
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+    // When lod is used always is in gpr20
+    const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0);
 
     const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
 
     Node4 values;
     for (u32 element = 0; element < values.size(); ++element) {
-        auto params = coords;
-        MetaTexture meta{sampler, element, coords_count, array_offset};
-        values[element] =
-            Operation(OperationCode::F4TexelFetch, std::move(meta), std::move(params));
+        auto coords_copy = coords;
+        MetaTexture meta{sampler, array, {}, {lod}, element};
+        values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy));
     }
     return values;
 }

src/video_core/shader/shader_ir.h

@@ -156,12 +156,12 @@ enum class OperationCode {
     Logical2HNotEqual,     /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2
     Logical2HGreaterEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2
 
-    F4Texture,                /// (MetaTexture, float[N] coords, float[M] params) -> float4
-    F4TextureLod,             /// (MetaTexture, float[N] coords, float[M] params) -> float4
-    F4TextureGather,          /// (MetaTexture, float[N] coords, float[M] params) -> float4
-    F4TextureQueryDimensions, /// (MetaTexture, float a) -> float4
-    F4TextureQueryLod,        /// (MetaTexture, float[N] coords) -> float4
-    F4TexelFetch,             /// (MetaTexture, int[N], int) -> float4
+    Texture,                /// (MetaTexture, float[N] coords) -> float4
+    TextureLod,             /// (MetaTexture, float[N] coords) -> float4
+    TextureGather,          /// (MetaTexture, float[N] coords) -> float4
+    TextureQueryDimensions, /// (MetaTexture, float a) -> float4
+    TextureQueryLod,        /// (MetaTexture, float[N] coords) -> float4
+    TexelFetch,             /// (MetaTexture, int[N], int) -> float4
 
     Branch,        /// (uint branch_target) -> void
     PushFlowStack, /// (uint branch_target) -> void
@@ -288,9 +288,10 @@ struct MetaHalfArithmetic {
 
 struct MetaTexture {
     const Sampler& sampler;
+    Node array{};
+    Node depth_compare{};
+    std::vector<Node> extras;
     u32 element{};
-    u32 coords_count{};
-    std::optional<u32> array_index;
 };
 
 constexpr MetaArithmetic PRECISE = {true};
@@ -754,9 +755,8 @@ private:
         bool lod_bias_enabled, std::size_t max_coords, std::size_t max_inputs);
 
     Node4 GetTextureCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
-                         Tegra::Shader::TextureProcessMode process_mode, bool depth_compare,
-                         bool is_array, std::size_t array_offset, std::size_t bias_offset,
-                         std::vector<Node>&& coords);
+                         Tegra::Shader::TextureProcessMode process_mode, std::vector<Node> coords,
+                         Node array, Node depth_compare, u32 bias_offset);
 
     Node GetVideoOperand(Node op, bool is_chunk, bool is_signed, Tegra::Shader::VideoType type,
                          u64 byte_height);

src/video_core/surface.cpp

@@ -426,6 +426,8 @@ PixelFormat PixelFormatFromGPUPixelFormat(Tegra::FramebufferConfig::PixelFormat
     switch (format) {
     case Tegra::FramebufferConfig::PixelFormat::ABGR8:
         return PixelFormat::ABGR8U;
+    case Tegra::FramebufferConfig::PixelFormat::BGRA8:
+        return PixelFormat::BGRA8;
     default:
         LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
         UNREACHABLE();