53 files changed, 1187 insertions, 673 deletions

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 71efbb40d..dbda528ce 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -32,7 +32,6 @@ if (MSVC)
     # /Zc:inline          - Let codegen omit inline functions in object files
     # /Zc:throwingNew     - Let codegen assume `operator new` (without std::nothrow) will never return null
     add_compile_options(
-        /W3
         /MP
         /Zi
         /Zo
@@ -43,6 +42,13 @@ if (MSVC)
         /Zc:externConstexpr
         /Zc:inline
         /Zc:throwingNew
+
+        # Warnings
+        /W3
+        /we4547 # 'operator' : operator before comma has no effect; expected operator with side-effect
+        /we4549 # 'operator1': operator before comma has no effect; did you intend 'operator2'?
+        /we4555 # Expression has no effect; expected expression with side-effect
+        /we4834 # Discarding return value of function with 'nodiscard' attribute
     )
 
     # /GS- - No stack buffer overflow checks
@@ -56,6 +62,7 @@ else()
         -Werror=implicit-fallthrough
         -Werror=missing-declarations
         -Werror=reorder
+        -Werror=unused-result
         -Wextra
         -Wmissing-declarations
         -Wno-attributes
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index e50ab2922..207c7a0a6 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -190,6 +190,22 @@ if(ARCHITECTURE_x86_64)
     )
 endif()
 
+if (MSVC)
+  target_compile_definitions(common PRIVATE
+    # The standard library doesn't provide any replacement for codecvt yet
+    # so we can disable this deprecation warning for the time being.
+    _SILENCE_CXX17_CODECVT_HEADER_DEPRECATION_WARNING
+  )
+  target_compile_options(common PRIVATE
+    /W4
+    /WX
+  )
+else()
+  target_compile_options(common PRIVATE
+    -Werror
+  )
+endif()
+
 create_target_directory_groups(common)
 find_package(Boost 1.71 COMPONENTS context headers REQUIRED)
 
diff --git a/src/common/fiber.cpp b/src/common/fiber.cpp
index e186ed880..b209f52fc 100644
--- a/src/common/fiber.cpp
+++ b/src/common/fiber.cpp
@@ -79,9 +79,9 @@ void Fiber::Exit() {
     released = true;
 }
 
-void Fiber::SetRewindPoint(std::function<void(void*)>&& rewind_func, void* start_parameter) {
+void Fiber::SetRewindPoint(std::function<void(void*)>&& rewind_func, void* rewind_param) {
     rewind_point = std::move(rewind_func);
-    rewind_parameter = start_parameter;
+    rewind_parameter = rewind_param;
 }
 
 void Fiber::Rewind() {
@@ -161,9 +161,9 @@ Fiber::Fiber(std::function<void(void*)>&& entry_point_func, void* start_paramete
         boost::context::detail::make_fcontext(stack_base, impl->stack.size(), FiberStartFunc);
 }
 
-void Fiber::SetRewindPoint(std::function<void(void*)>&& rewind_func, void* start_parameter) {
+void Fiber::SetRewindPoint(std::function<void(void*)>&& rewind_func, void* rewind_param) {
     rewind_point = std::move(rewind_func);
-    rewind_parameter = start_parameter;
+    rewind_parameter = rewind_param;
 }
 
 Fiber::Fiber() : impl{std::make_unique<FiberImpl>()} {}
diff --git a/src/common/fiber.h b/src/common/fiber.h
index cefd61df9..699286ee2 100644
--- a/src/common/fiber.h
+++ b/src/common/fiber.h
@@ -49,7 +49,7 @@ public:
     static void YieldTo(std::shared_ptr<Fiber> from, std::shared_ptr<Fiber> to);
     [[nodiscard]] static std::shared_ptr<Fiber> ThreadToFiber();
 
-    void SetRewindPoint(std::function<void(void*)>&& rewind_func, void* start_parameter);
+    void SetRewindPoint(std::function<void(void*)>&& rewind_func, void* rewind_param);
 
     void Rewind();
 
diff --git a/src/common/file_util.cpp b/src/common/file_util.cpp
index 16c3713e0..18fbfa25b 100644
--- a/src/common/file_util.cpp
+++ b/src/common/file_util.cpp
@@ -472,13 +472,14 @@ u64 ScanDirectoryTree(const std::string& directory, FSTEntry& parent_entry,
 }
 
 bool DeleteDirRecursively(const std::string& directory, unsigned int recursion) {
-    const auto callback = [recursion](u64* num_entries_out, const std::string& directory,
-                                      const std::string& virtual_name) -> bool {
-        std::string new_path = directory + DIR_SEP_CHR + virtual_name;
+    const auto callback = [recursion](u64*, const std::string& directory,
+                                      const std::string& virtual_name) {
+        const std::string new_path = directory + DIR_SEP_CHR + virtual_name;
 
         if (IsDirectory(new_path)) {
-            if (recursion == 0)
+            if (recursion == 0) {
                 return false;
+            }
             return DeleteDirRecursively(new_path, recursion - 1);
         }
         return Delete(new_path);
@@ -492,7 +493,8 @@ bool DeleteDirRecursively(const std::string& directory, unsigned int recursion)
     return true;
 }
 
-void CopyDir(const std::string& source_path, const std::string& dest_path) {
+void CopyDir([[maybe_unused]] const std::string& source_path,
+             [[maybe_unused]] const std::string& dest_path) {
 #ifndef _WIN32
     if (source_path == dest_path) {
         return;
@@ -553,7 +555,7 @@ std::optional<std::string> GetCurrentDir() {
     std::string strDir = dir;
 #endif
     free(dir);
-    return std::move(strDir);
+    return strDir;
 }
 
 bool SetCurrentDir(const std::string& directory) {
@@ -772,21 +774,23 @@ std::size_t ReadFileToString(bool text_file, const std::string& filename, std::s
 
 void SplitFilename83(const std::string& filename, std::array<char, 9>& short_name,
                      std::array<char, 4>& extension) {
-    const std::string forbidden_characters = ".\"/\\[]:;=, ";
+    static constexpr std::string_view forbidden_characters = ".\"/\\[]:;=, ";
 
     // On a FAT32 partition, 8.3 names are stored as a 11 bytes array, filled with spaces.
     short_name = {{' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', '\0'}};
     extension = {{' ', ' ', ' ', '\0'}};
 
-    std::string::size_type point = filename.rfind('.');
-    if (point == filename.size() - 1)
+    auto point = filename.rfind('.');
+    if (point == filename.size() - 1) {
         point = filename.rfind('.', point);
+    }
 
     // Get short name.
     int j = 0;
     for (char letter : filename.substr(0, point)) {
-        if (forbidden_characters.find(letter, 0) != std::string::npos)
+        if (forbidden_characters.find(letter, 0) != std::string::npos) {
             continue;
+        }
         if (j == 8) {
             // TODO(Link Mauve): also do that for filenames containing a space.
             // TODO(Link Mauve): handle multiple files having the same short name.
@@ -794,14 +798,15 @@ void SplitFilename83(const std::string& filename, std::array<char, 9>& short_nam
             short_name[7] = '1';
             break;
         }
-        short_name[j++] = toupper(letter);
+        short_name[j++] = static_cast<char>(std::toupper(letter));
     }
 
     // Get extension.
     if (point != std::string::npos) {
         j = 0;
-        for (char letter : filename.substr(point + 1, 3))
-            extension[j++] = toupper(letter);
+        for (char letter : filename.substr(point + 1, 3)) {
+            extension[j++] = static_cast<char>(std::toupper(letter));
+        }
     }
 }
 
diff --git a/src/common/file_util.h b/src/common/file_util.h
index 8b587320f..840cde2a6 100644
--- a/src/common/file_util.h
+++ b/src/common/file_util.h
@@ -232,7 +232,7 @@ public:
 
     void Swap(IOFile& other) noexcept;
 
-    [[nodiscard]] bool Open(const std::string& filename, const char openmode[], int flags = 0);
+    bool Open(const std::string& filename, const char openmode[], int flags = 0);
     bool Close();
 
     template <typename T>
diff --git a/src/common/logging/backend.cpp b/src/common/logging/backend.cpp
index 62cfde397..90dfa22ca 100644
--- a/src/common/logging/backend.cpp
+++ b/src/common/logging/backend.cpp
@@ -274,7 +274,6 @@ const char* GetLogClassName(Class log_class) {
     case Class::Count:
         break;
     }
-    UNREACHABLE();
     return "Invalid";
 }
 
@@ -293,7 +292,6 @@ const char* GetLevelName(Level log_level) {
         break;
     }
 #undef LVL
-    UNREACHABLE();
     return "Invalid";
 }
 
diff --git a/src/common/misc.cpp b/src/common/misc.cpp
index 68cb86cd1..1d5393597 100644
--- a/src/common/misc.cpp
+++ b/src/common/misc.cpp
@@ -16,16 +16,23 @@
 // Call directly after the command or use the error num.
 // This function might change the error code.
 std::string GetLastErrorMsg() {
-    static const std::size_t buff_size = 255;
+    static constexpr std::size_t buff_size = 255;
     char err_str[buff_size];
 
 #ifdef _WIN32
     FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM, nullptr, GetLastError(),
                    MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), err_str, buff_size, nullptr);
+    return std::string(err_str, buff_size);
+#elif defined(__GLIBC__) && (_GNU_SOURCE || (_POSIX_C_SOURCE < 200112L && _XOPEN_SOURCE < 600))
+    // Thread safe (GNU-specific)
+    const char* str = strerror_r(errno, err_str, buff_size);
+    return std::string(str);
 #else
     // Thread safe (XSI-compliant)
-    strerror_r(errno, err_str, buff_size);
+    const int success = strerror_r(errno, err_str, buff_size);
+    if (success != 0) {
+        return {};
+    }
+    return std::string(err_str);
 #endif
-
-    return std::string(err_str, buff_size);
 }
diff --git a/src/common/stream.h b/src/common/stream.h
index 2585c16af..0e40692de 100644
--- a/src/common/stream.h
+++ b/src/common/stream.h
@@ -21,6 +21,12 @@ public:
     explicit Stream();
     ~Stream();
 
+    Stream(const Stream&) = delete;
+    Stream& operator=(const Stream&) = delete;
+
+    Stream(Stream&&) = default;
+    Stream& operator=(Stream&&) = default;
+
     /// Reposition bitstream "cursor" to the specified offset from origin
     void Seek(s32 offset, SeekOrigin origin);
 
@@ -30,15 +36,15 @@ public:
     /// Writes byte at current position
     void WriteByte(u8 byte);
 
-    std::size_t GetPosition() const {
+    [[nodiscard]] std::size_t GetPosition() const {
         return position;
     }
 
-    std::vector<u8>& GetBuffer() {
+    [[nodiscard]] std::vector<u8>& GetBuffer() {
         return buffer;
     }
 
-    const std::vector<u8>& GetBuffer() const {
+    [[nodiscard]] const std::vector<u8>& GetBuffer() const {
         return buffer;
     }
 
diff --git a/src/common/string_util.cpp b/src/common/string_util.cpp
index 84883a1d3..4cba2aaa4 100644
--- a/src/common/string_util.cpp
+++ b/src/common/string_util.cpp
@@ -8,6 +8,7 @@
 #include <cstdlib>
 #include <locale>
 #include <sstream>
+
 #include "common/common_paths.h"
 #include "common/logging/log.h"
 #include "common/string_util.h"
@@ -21,14 +22,14 @@ namespace Common {
 /// Make a string lowercase
 std::string ToLower(std::string str) {
     std::transform(str.begin(), str.end(), str.begin(),
-                   [](unsigned char c) { return std::tolower(c); });
+                   [](unsigned char c) { return static_cast<char>(std::tolower(c)); });
     return str;
 }
 
 /// Make a string uppercase
 std::string ToUpper(std::string str) {
     std::transform(str.begin(), str.end(), str.begin(),
-                   [](unsigned char c) { return std::toupper(c); });
+                   [](unsigned char c) { return static_cast<char>(std::toupper(c)); });
     return str;
 }
 
diff --git a/src/common/timer.cpp b/src/common/timer.cpp
index 2dc15e434..d17dc2a50 100644
--- a/src/common/timer.cpp
+++ b/src/common/timer.cpp
@@ -142,20 +142,18 @@ std::string Timer::GetTimeFormatted() {
 // ----------------
 double Timer::GetDoubleTime() {
     // Get continuous timestamp
-    u64 TmpSeconds = static_cast<u64>(Common::Timer::GetTimeSinceJan1970().count());
-    double ms = static_cast<u64>(GetTimeMs().count()) % 1000;
+    auto tmp_seconds = static_cast<u64>(GetTimeSinceJan1970().count());
+    const auto ms = static_cast<double>(static_cast<u64>(GetTimeMs().count()) % 1000);
 
     // Remove a few years. We only really want enough seconds to make
     // sure that we are detecting actual actions, perhaps 60 seconds is
     // enough really, but I leave a year of seconds anyway, in case the
     // user's clock is incorrect or something like that.
-    TmpSeconds = TmpSeconds - (38 * 365 * 24 * 60 * 60);
+    tmp_seconds = tmp_seconds - (38 * 365 * 24 * 60 * 60);
 
     // Make a smaller integer that fits in the double
-    u32 Seconds = static_cast<u32>(TmpSeconds);
-    double TmpTime = Seconds + ms;
-
-    return TmpTime;
+    const auto seconds = static_cast<u32>(tmp_seconds);
+    return seconds + ms;
 }
 
 } // Namespace Common
diff --git a/src/common/wall_clock.cpp b/src/common/wall_clock.cpp
index 7a20e95b7..452a2837e 100644
--- a/src/common/wall_clock.cpp
+++ b/src/common/wall_clock.cpp
@@ -53,7 +53,7 @@ public:
         return Common::Divide128On32(temporary, 1000000000).first;
     }
 
-    void Pause(bool is_paused) override {
+    void Pause([[maybe_unused]] bool is_paused) override {
         // Do nothing in this clock type.
     }
 
diff --git a/src/common/x64/native_clock.h b/src/common/x64/native_clock.h
index 7c503df26..97aab6ac9 100644
--- a/src/common/x64/native_clock.h
+++ b/src/common/x64/native_clock.h
@@ -34,7 +34,7 @@ private:
     /// value used to reduce the native clocks accuracy as some apss rely on
     /// undefined behavior where the level of accuracy in the clock shouldn't
     /// be higher.
-    static constexpr u64 inaccuracy_mask = ~(0x400 - 1);
+    static constexpr u64 inaccuracy_mask = ~(UINT64_C(0x400) - 1);
 
     SpinLock rtsc_serialize{};
     u64 last_measure{};
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index e0f207f3e..9a983e81d 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -454,6 +454,8 @@ add_library(core STATIC
     hle/service/nvdrv/nvdrv.h
     hle/service/nvdrv/nvmemp.cpp
     hle/service/nvdrv/nvmemp.h
+    hle/service/nvdrv/syncpoint_manager.cpp
+    hle/service/nvdrv/syncpoint_manager.h
     hle/service/nvflinger/buffer_queue.cpp
     hle/service/nvflinger/buffer_queue.h
     hle/service/nvflinger/nvflinger.cpp
diff --git a/src/core/core.cpp b/src/core/core.cpp
index fde2ccc09..242796008 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -179,16 +179,18 @@ struct System::Impl {
         arp_manager.ResetAll();
 
         telemetry_session = std::make_unique<Core::TelemetrySession>();
+
+        gpu_core = VideoCore::CreateGPU(emu_window, system);
+        if (!gpu_core) {
+            return ResultStatus::ErrorVideoCore;
+        }
+
         service_manager = std::make_shared<Service::SM::ServiceManager>(kernel);
 
         Service::Init(service_manager, system);
         GDBStub::DeferStart();
 
         interrupt_manager = std::make_unique<Core::Hardware::InterruptManager>(system);
-        gpu_core = VideoCore::CreateGPU(emu_window, system);
-        if (!gpu_core) {
-            return ResultStatus::ErrorVideoCore;
-        }
 
         // Initialize time manager, which must happen after kernel is created
         time_manager.Initialize();
diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index ff9d9248b..b17529dee 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -4,6 +4,7 @@
 
 #include <algorithm>
 #include <bitset>
+#include <ctime>
 #include <memory>
 #include <random>
 #include "common/alignment.h"
@@ -123,7 +124,7 @@ std::shared_ptr<Process> Process::Create(Core::System& system, std::string name,
                                                               : kernel.CreateNewUserProcessID();
     process->capabilities.InitializeForMetadatalessProcess();
 
-    std::mt19937 rng(Settings::values.rng_seed.GetValue().value_or(0));
+    std::mt19937 rng(Settings::values.rng_seed.GetValue().value_or(std::time(nullptr)));
     std::uniform_int_distribution<u64> distribution;
     std::generate(process->random_entropy.begin(), process->random_entropy.end(),
                   [&] { return distribution(rng); });
diff --git a/src/core/hle/service/ldr/ldr.cpp b/src/core/hle/service/ldr/ldr.cpp
index 9ad5bbf0d..eeaca44b6 100644
--- a/src/core/hle/service/ldr/ldr.cpp
+++ b/src/core/hle/service/ldr/ldr.cpp
@@ -166,7 +166,7 @@ public:
             {0, &RelocatableObject::LoadNro, "LoadNro"},
             {1, &RelocatableObject::UnloadNro, "UnloadNro"},
             {2, &RelocatableObject::LoadNrr, "LoadNrr"},
-            {3, nullptr, "UnloadNrr"},
+            {3, &RelocatableObject::UnloadNrr, "UnloadNrr"},
             {4, &RelocatableObject::Initialize, "Initialize"},
             {10, nullptr, "LoadNrrEx"},
         };
@@ -272,6 +272,20 @@ public:
         rb.Push(RESULT_SUCCESS);
     }
 
+    void UnloadNrr(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto pid = rp.Pop<u64>();
+        const auto nrr_address = rp.Pop<VAddr>();
+
+        LOG_DEBUG(Service_LDR, "called with pid={}, nrr_address={:016X}", pid, nrr_address);
+
+        nrr.erase(nrr_address);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+
+        rb.Push(RESULT_SUCCESS);
+    }
+
     bool ValidateRegionForMap(Kernel::Memory::PageTable& page_table, VAddr start,
                               std::size_t size) const {
         constexpr std::size_t padding_size{4 * Kernel::Memory::PageSize};
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
index b27ee0502..8356a8139 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
@@ -15,8 +15,9 @@
 
 namespace Service::Nvidia::Devices {
 
-nvhost_ctrl::nvhost_ctrl(Core::System& system, EventInterface& events_interface)
-    : nvdevice(system), events_interface{events_interface} {}
+nvhost_ctrl::nvhost_ctrl(Core::System& system, EventInterface& events_interface,
+                         SyncpointManager& syncpoint_manager)
+    : nvdevice(system), events_interface{events_interface}, syncpoint_manager{syncpoint_manager} {}
 nvhost_ctrl::~nvhost_ctrl() = default;
 
 u32 nvhost_ctrl::ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@@ -36,8 +37,8 @@ u32 nvhost_ctrl::ioctl(Ioctl command, const std::vector<u8>& input, const std::v
         return IocCtrlEventRegister(input, output);
     case IoctlCommand::IocCtrlEventUnregisterCommand:
         return IocCtrlEventUnregister(input, output);
-    case IoctlCommand::IocCtrlEventSignalCommand:
-        return IocCtrlEventSignal(input, output);
+    case IoctlCommand::IocCtrlClearEventWaitCommand:
+        return IocCtrlClearEventWait(input, output);
     default:
         UNIMPLEMENTED_MSG("Unimplemented ioctl");
         return 0;
@@ -70,19 +71,33 @@ u32 nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>&
         return NvResult::BadParameter;
     }
 
+    if (syncpoint_manager.IsSyncpointExpired(params.syncpt_id, params.threshold)) {
+        params.value = syncpoint_manager.GetSyncpointMin(params.syncpt_id);
+        std::memcpy(output.data(), &params, sizeof(params));
+        return NvResult::Success;
+    }
+
+    if (const auto new_value = syncpoint_manager.RefreshSyncpoint(params.syncpt_id);
+        syncpoint_manager.IsSyncpointExpired(params.syncpt_id, params.threshold)) {
+        params.value = new_value;
+        std::memcpy(output.data(), &params, sizeof(params));
+        return NvResult::Success;
+    }
+
     auto event = events_interface.events[event_id];
     auto& gpu = system.GPU();
+
     // This is mostly to take into account unimplemented features. As synced
     // gpu is always synced.
     if (!gpu.IsAsync()) {
-        event.writable->Signal();
+        event.event.writable->Signal();
         return NvResult::Success;
     }
     auto lock = gpu.LockSync();
-    const u32 current_syncpoint_value = gpu.GetSyncpointValue(params.syncpt_id);
+    const u32 current_syncpoint_value = event.fence.value;
     const s32 diff = current_syncpoint_value - params.threshold;
     if (diff >= 0) {
-        event.writable->Signal();
+        event.event.writable->Signal();
         params.value = current_syncpoint_value;
         std::memcpy(output.data(), &params, sizeof(params));
         return NvResult::Success;
@@ -109,7 +124,7 @@ u32 nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>&
             params.value = ((params.syncpt_id & 0xfff) << 16) | 0x10000000;
         }
         params.value |= event_id;
-        event.writable->Clear();
+        event.event.writable->Clear();
         gpu.RegisterSyncptInterrupt(params.syncpt_id, target_value);
         if (!is_async && ctrl.fresh_call) {
             ctrl.must_delay = true;
@@ -154,24 +169,22 @@ u32 nvhost_ctrl::IocCtrlEventUnregister(const std::vector<u8>& input, std::vecto
     return NvResult::Success;
 }
 
-u32 nvhost_ctrl::IocCtrlEventSignal(const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_ctrl::IocCtrlClearEventWait(const std::vector<u8>& input, std::vector<u8>& output) {
     IocCtrlEventSignalParams params{};
     std::memcpy(&params, input.data(), sizeof(params));
-    // TODO(Blinkhawk): This is normally called when an NvEvents timeout on WaitSynchronization
-    // It is believed from RE to cancel the GPU Event. However, better research is required
-    u32 event_id = params.user_event_id & 0x00FF;
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called, user_event_id: {:X}", event_id);
+
+    u32 event_id = params.event_id & 0x00FF;
+    LOG_WARNING(Service_NVDRV, "cleared event wait on, event_id: {:X}", event_id);
+
     if (event_id >= MaxNvEvents) {
         return NvResult::BadParameter;
     }
     if (events_interface.status[event_id] == EventState::Waiting) {
-        auto& gpu = system.GPU();
-        if (gpu.CancelSyncptInterrupt(events_interface.assigned_syncpt[event_id],
-                                      events_interface.assigned_value[event_id])) {
-            events_interface.LiberateEvent(event_id);
-            events_interface.events[event_id].writable->Signal();
-        }
+        events_interface.LiberateEvent(event_id);
     }
+
+    syncpoint_manager.RefreshSyncpoint(events_interface.events[event_id].fence.id);
+
     return NvResult::Success;
 }
 
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
index 9898623de..24ad96cb9 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
@@ -14,7 +14,8 @@ namespace Service::Nvidia::Devices {
 
 class nvhost_ctrl final : public nvdevice {
 public:
-    explicit nvhost_ctrl(Core::System& system, EventInterface& events_interface);
+    explicit nvhost_ctrl(Core::System& system, EventInterface& events_interface,
+                         SyncpointManager& syncpoint_manager);
     ~nvhost_ctrl() override;
 
     u32 ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@@ -31,7 +32,7 @@ private:
         IocSyncptWaitexCommand = 0xC0100019,
         IocSyncptReadMaxCommand = 0xC008001A,
         IocGetConfigCommand = 0xC183001B,
-        IocCtrlEventSignalCommand = 0xC004001C,
+        IocCtrlClearEventWaitCommand = 0xC004001C,
         IocCtrlEventWaitCommand = 0xC010001D,
         IocCtrlEventWaitAsyncCommand = 0xC010001E,
         IocCtrlEventRegisterCommand = 0xC004001F,
@@ -94,7 +95,7 @@ private:
     static_assert(sizeof(IocGetConfigParams) == 387, "IocGetConfigParams is incorrect size");
 
     struct IocCtrlEventSignalParams {
-        u32_le user_event_id;
+        u32_le event_id;
     };
     static_assert(sizeof(IocCtrlEventSignalParams) == 4,
                   "IocCtrlEventSignalParams is incorrect size");
@@ -142,9 +143,10 @@ private:
 
     u32 IocCtrlEventUnregister(const std::vector<u8>& input, std::vector<u8>& output);
 
-    u32 IocCtrlEventSignal(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 IocCtrlClearEventWait(const std::vector<u8>& input, std::vector<u8>& output);
 
     EventInterface& events_interface;
+    SyncpointManager& syncpoint_manager;
 };
 
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
index f1966ac0e..152019548 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
@@ -7,14 +7,20 @@
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/hle/service/nvdrv/devices/nvhost_gpu.h"
+#include "core/hle/service/nvdrv/syncpoint_manager.h"
 #include "core/memory.h"
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 
 namespace Service::Nvidia::Devices {
 
-nvhost_gpu::nvhost_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
-    : nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {}
+nvhost_gpu::nvhost_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev,
+                       SyncpointManager& syncpoint_manager)
+    : nvdevice(system), nvmap_dev(std::move(nvmap_dev)), syncpoint_manager{syncpoint_manager} {
+    channel_fence.id = syncpoint_manager.AllocateSyncpoint();
+    channel_fence.value = system.GPU().GetSyncpointValue(channel_fence.id);
+}
+
 nvhost_gpu::~nvhost_gpu() = default;
 
 u32 nvhost_gpu::ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@@ -126,10 +132,10 @@ u32 nvhost_gpu::AllocGPFIFOEx2(const std::vector<u8>& input, std::vector<u8>& ou
                 params.num_entries, params.flags, params.unk0, params.unk1, params.unk2,
                 params.unk3);
 
-    auto& gpu = system.GPU();
-    params.fence_out.id = assigned_syncpoints;
-    params.fence_out.value = gpu.GetSyncpointValue(assigned_syncpoints);
-    assigned_syncpoints++;
+    channel_fence.value = system.GPU().GetSyncpointValue(channel_fence.id);
+
+    params.fence_out = channel_fence;
+
     std::memcpy(output.data(), &params, output.size());
     return 0;
 }
@@ -145,39 +151,100 @@ u32 nvhost_gpu::AllocateObjectContext(const std::vector<u8>& input, std::vector<
     return 0;
 }
 
-u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output) {
-    if (input.size() < sizeof(IoctlSubmitGpfifo)) {
-        UNIMPLEMENTED();
+static std::vector<Tegra::CommandHeader> BuildWaitCommandList(Fence fence) {
+    return {
+        Tegra::BuildCommandHeader(Tegra::BufferMethods::FenceValue, 1,
+                                  Tegra::SubmissionMode::Increasing),
+        {fence.value},
+        Tegra::BuildCommandHeader(Tegra::BufferMethods::FenceAction, 1,
+                                  Tegra::SubmissionMode::Increasing),
+        Tegra::GPU::FenceAction::Build(Tegra::GPU::FenceOperation::Acquire, fence.id),
+    };
+}
+
+static std::vector<Tegra::CommandHeader> BuildIncrementCommandList(Fence fence, u32 add_increment) {
+    std::vector<Tegra::CommandHeader> result{
+        Tegra::BuildCommandHeader(Tegra::BufferMethods::FenceValue, 1,
+                                  Tegra::SubmissionMode::Increasing),
+        {}};
+
+    for (u32 count = 0; count < add_increment; ++count) {
+        result.emplace_back(Tegra::BuildCommandHeader(Tegra::BufferMethods::FenceAction, 1,
+                                                      Tegra::SubmissionMode::Increasing));
+        result.emplace_back(
+            Tegra::GPU::FenceAction::Build(Tegra::GPU::FenceOperation::Increment, fence.id));
     }
-    IoctlSubmitGpfifo params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
+
+    return result;
+}
+
+static std::vector<Tegra::CommandHeader> BuildIncrementWithWfiCommandList(Fence fence,
+                                                                          u32 add_increment) {
+    std::vector<Tegra::CommandHeader> result{
+        Tegra::BuildCommandHeader(Tegra::BufferMethods::WaitForInterrupt, 1,
+                                  Tegra::SubmissionMode::Increasing),
+        {}};
+    const std::vector<Tegra::CommandHeader> increment{
+        BuildIncrementCommandList(fence, add_increment)};
+
+    result.insert(result.end(), increment.begin(), increment.end());
+
+    return result;
+}
+
+u32 nvhost_gpu::SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, std::vector<u8>& output,
+                                 Tegra::CommandList&& entries) {
     LOG_TRACE(Service_NVDRV, "called, gpfifo={:X}, num_entries={:X}, flags={:X}", params.address,
               params.num_entries, params.flags.raw);
 
-    ASSERT_MSG(input.size() == sizeof(IoctlSubmitGpfifo) +
-                                   params.num_entries * sizeof(Tegra::CommandListHeader),
-               "Incorrect input size");
+    auto& gpu = system.GPU();
 
-    Tegra::CommandList entries(params.num_entries);
-    std::memcpy(entries.data(), &input[sizeof(IoctlSubmitGpfifo)],
-                params.num_entries * sizeof(Tegra::CommandListHeader));
+    params.fence_out.id = channel_fence.id;
 
-    UNIMPLEMENTED_IF(params.flags.add_wait.Value() != 0);
-    UNIMPLEMENTED_IF(params.flags.add_increment.Value() != 0);
+    if (params.flags.add_wait.Value() &&
+        !syncpoint_manager.IsSyncpointExpired(params.fence_out.id, params.fence_out.value)) {
+        gpu.PushGPUEntries(Tegra::CommandList{BuildWaitCommandList(params.fence_out)});
+    }
 
-    auto& gpu = system.GPU();
-    u32 current_syncpoint_value = gpu.GetSyncpointValue(params.fence_out.id);
-    if (params.flags.increment.Value()) {
-        params.fence_out.value += current_syncpoint_value;
+    if (params.flags.add_increment.Value() || params.flags.increment.Value()) {
+        const u32 increment_value = params.flags.increment.Value() ? params.fence_out.value : 0;
+        params.fence_out.value = syncpoint_manager.IncreaseSyncpoint(
+            params.fence_out.id, params.AddIncrementValue() + increment_value);
     } else {
-        params.fence_out.value = current_syncpoint_value;
+        params.fence_out.value = syncpoint_manager.GetSyncpointMax(params.fence_out.id);
     }
+
+    entries.RefreshIntegrityChecks(gpu);
     gpu.PushGPUEntries(std::move(entries));
 
+    if (params.flags.add_increment.Value()) {
+        if (params.flags.suppress_wfi) {
+            gpu.PushGPUEntries(Tegra::CommandList{
+                BuildIncrementCommandList(params.fence_out, params.AddIncrementValue())});
+        } else {
+            gpu.PushGPUEntries(Tegra::CommandList{
+                BuildIncrementWithWfiCommandList(params.fence_out, params.AddIncrementValue())});
+        }
+    }
+
     std::memcpy(output.data(), &params, sizeof(IoctlSubmitGpfifo));
     return 0;
 }
 
+u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output) {
+    if (input.size() < sizeof(IoctlSubmitGpfifo)) {
+        UNIMPLEMENTED();
+    }
+    IoctlSubmitGpfifo params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
+
+    Tegra::CommandList entries(params.num_entries);
+    std::memcpy(entries.command_lists.data(), &input[sizeof(IoctlSubmitGpfifo)],
+                params.num_entries * sizeof(Tegra::CommandListHeader));
+
+    return SubmitGPFIFOImpl(params, output, std::move(entries));
+}
+
 u32 nvhost_gpu::KickoffPB(const std::vector<u8>& input, std::vector<u8>& output,
                           const std::vector<u8>& input2, IoctlVersion version) {
     if (input.size() < sizeof(IoctlSubmitGpfifo)) {
@@ -185,31 +252,17 @@ u32 nvhost_gpu::KickoffPB(const std::vector<u8>& input, std::vector<u8>& output,
     }
     IoctlSubmitGpfifo params{};
     std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
-    LOG_TRACE(Service_NVDRV, "called, gpfifo={:X}, num_entries={:X}, flags={:X}", params.address,
-              params.num_entries, params.flags.raw);
 
     Tegra::CommandList entries(params.num_entries);
     if (version == IoctlVersion::Version2) {
-        std::memcpy(entries.data(), input2.data(),
+        std::memcpy(entries.command_lists.data(), input2.data(),
                     params.num_entries * sizeof(Tegra::CommandListHeader));
     } else {
-        system.Memory().ReadBlock(params.address, entries.data(),
+        system.Memory().ReadBlock(params.address, entries.command_lists.data(),
                                   params.num_entries * sizeof(Tegra::CommandListHeader));
     }
-    UNIMPLEMENTED_IF(params.flags.add_wait.Value() != 0);
-    UNIMPLEMENTED_IF(params.flags.add_increment.Value() != 0);
-
-    auto& gpu = system.GPU();
-    u32 current_syncpoint_value = gpu.GetSyncpointValue(params.fence_out.id);
-    if (params.flags.increment.Value()) {
-        params.fence_out.value += current_syncpoint_value;
-    } else {
-        params.fence_out.value = current_syncpoint_value;
-    }
-    gpu.PushGPUEntries(std::move(entries));
 
-    std::memcpy(output.data(), &params, output.size());
-    return 0;
+    return SubmitGPFIFOImpl(params, output, std::move(entries));
 }
 
 u32 nvhost_gpu::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
index 2ac74743f..a252fc06d 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
@@ -11,6 +11,11 @@
 #include "common/swap.h"
 #include "core/hle/service/nvdrv/devices/nvdevice.h"
 #include "core/hle/service/nvdrv/nvdata.h"
+#include "video_core/dma_pusher.h"
+
+namespace Service::Nvidia {
+class SyncpointManager;
+}
 
 namespace Service::Nvidia::Devices {
 
@@ -21,7 +26,8 @@ constexpr u32 NVGPU_IOCTL_CHANNEL_KICKOFF_PB(0x1b);
 
 class nvhost_gpu final : public nvdevice {
 public:
-    explicit nvhost_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
+    explicit nvhost_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev,
+                        SyncpointManager& syncpoint_manager);
     ~nvhost_gpu() override;
 
     u32 ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@@ -162,10 +168,15 @@ private:
             u32_le raw;
             BitField<0, 1, u32_le> add_wait;      // append a wait sync_point to the list
             BitField<1, 1, u32_le> add_increment; // append an increment to the list
-            BitField<2, 1, u32_le> new_hw_format; // Mostly ignored
+            BitField<2, 1, u32_le> new_hw_format; // mostly ignored
+            BitField<4, 1, u32_le> suppress_wfi;  // suppress wait for interrupt
             BitField<8, 1, u32_le> increment;     // increment the returned fence
         } flags;
         Fence fence_out; // returned new fence object for others to wait on
+
+        u32 AddIncrementValue() const {
+            return flags.add_increment.Value() << 1;
+        }
     };
     static_assert(sizeof(IoctlSubmitGpfifo) == 16 + sizeof(Fence),
                   "IoctlSubmitGpfifo is incorrect size");
@@ -190,6 +201,8 @@ private:
     u32 SetChannelPriority(const std::vector<u8>& input, std::vector<u8>& output);
     u32 AllocGPFIFOEx2(const std::vector<u8>& input, std::vector<u8>& output);
     u32 AllocateObjectContext(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, std::vector<u8>& output,
+                         Tegra::CommandList&& entries);
     u32 SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output);
     u32 KickoffPB(const std::vector<u8>& input, std::vector<u8>& output,
                   const std::vector<u8>& input2, IoctlVersion version);
@@ -198,7 +211,8 @@ private:
     u32 ChannelSetTimeslice(const std::vector<u8>& input, std::vector<u8>& output);
 
     std::shared_ptr<nvmap> nvmap_dev;
-    u32 assigned_syncpoints{};
+    SyncpointManager& syncpoint_manager;
+    Fence channel_fence;
 };
 
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
index 85792495f..30f03f845 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
@@ -38,7 +38,7 @@ std::size_t WriteVectors(std::vector<u8>& dst, const std::vector<T>& src, std::s
 
 namespace NvErrCodes {
 constexpr u32 Success{};
-constexpr u32 OutOfMemory{static_cast<u32>(-12)};
+[[maybe_unused]] constexpr u32 OutOfMemory{static_cast<u32>(-12)};
 constexpr u32 InvalidInput{static_cast<u32>(-22)};
 } // namespace NvErrCodes
 
diff --git a/src/core/hle/service/nvdrv/nvdrv.cpp b/src/core/hle/service/nvdrv/nvdrv.cpp
index 803c1a984..a46755cdc 100644
--- a/src/core/hle/service/nvdrv/nvdrv.cpp
+++ b/src/core/hle/service/nvdrv/nvdrv.cpp
@@ -21,6 +21,7 @@
 #include "core/hle/service/nvdrv/interface.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/hle/service/nvdrv/nvmemp.h"
+#include "core/hle/service/nvdrv/syncpoint_manager.h"
 #include "core/hle/service/nvflinger/nvflinger.h"
 
 namespace Service::Nvidia {
@@ -36,21 +37,23 @@ void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger
     nvflinger.SetNVDrvInstance(module_);
 }
 
-Module::Module(Core::System& system) {
+Module::Module(Core::System& system) : syncpoint_manager{system.GPU()} {
     auto& kernel = system.Kernel();
     for (u32 i = 0; i < MaxNvEvents; i++) {
         std::string event_label = fmt::format("NVDRV::NvEvent_{}", i);
-        events_interface.events[i] = Kernel::WritableEvent::CreateEventPair(kernel, event_label);
+        events_interface.events[i] = {Kernel::WritableEvent::CreateEventPair(kernel, event_label)};
         events_interface.status[i] = EventState::Free;
         events_interface.registered[i] = false;
     }
     auto nvmap_dev = std::make_shared<Devices::nvmap>(system);
     devices["/dev/nvhost-as-gpu"] = std::make_shared<Devices::nvhost_as_gpu>(system, nvmap_dev);
-    devices["/dev/nvhost-gpu"] = std::make_shared<Devices::nvhost_gpu>(system, nvmap_dev);
+    devices["/dev/nvhost-gpu"] =
+        std::make_shared<Devices::nvhost_gpu>(system, nvmap_dev, syncpoint_manager);
     devices["/dev/nvhost-ctrl-gpu"] = std::make_shared<Devices::nvhost_ctrl_gpu>(system);
     devices["/dev/nvmap"] = nvmap_dev;
     devices["/dev/nvdisp_disp0"] = std::make_shared<Devices::nvdisp_disp0>(system, nvmap_dev);
-    devices["/dev/nvhost-ctrl"] = std::make_shared<Devices::nvhost_ctrl>(system, events_interface);
+    devices["/dev/nvhost-ctrl"] =
+        std::make_shared<Devices::nvhost_ctrl>(system, events_interface, syncpoint_manager);
     devices["/dev/nvhost-nvdec"] = std::make_shared<Devices::nvhost_nvdec>(system, nvmap_dev);
     devices["/dev/nvhost-nvjpg"] = std::make_shared<Devices::nvhost_nvjpg>(system);
     devices["/dev/nvhost-vic"] = std::make_shared<Devices::nvhost_vic>(system, nvmap_dev);
@@ -95,17 +98,17 @@ void Module::SignalSyncpt(const u32 syncpoint_id, const u32 value) {
         if (events_interface.assigned_syncpt[i] == syncpoint_id &&
             events_interface.assigned_value[i] == value) {
             events_interface.LiberateEvent(i);
-            events_interface.events[i].writable->Signal();
+            events_interface.events[i].event.writable->Signal();
         }
     }
 }
 
 std::shared_ptr<Kernel::ReadableEvent> Module::GetEvent(const u32 event_id) const {
-    return events_interface.events[event_id].readable;
+    return events_interface.events[event_id].event.readable;
 }
 
 std::shared_ptr<Kernel::WritableEvent> Module::GetEventWriteable(const u32 event_id) const {
-    return events_interface.events[event_id].writable;
+    return events_interface.events[event_id].event.writable;
 }
 
 } // namespace Service::Nvidia
diff --git a/src/core/hle/service/nvdrv/nvdrv.h b/src/core/hle/service/nvdrv/nvdrv.h
index 7706a5590..f3d863dac 100644
--- a/src/core/hle/service/nvdrv/nvdrv.h
+++ b/src/core/hle/service/nvdrv/nvdrv.h
@@ -10,6 +10,7 @@
 #include "common/common_types.h"
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/nvdrv/nvdata.h"
+#include "core/hle/service/nvdrv/syncpoint_manager.h"
 #include "core/hle/service/service.h"
 
 namespace Core {
@@ -22,15 +23,23 @@ class NVFlinger;
 
 namespace Service::Nvidia {
 
+class SyncpointManager;
+
 namespace Devices {
 class nvdevice;
 }
 
+/// Represents an Nvidia event
+struct NvEvent {
+    Kernel::EventPair event;
+    Fence fence{};
+};
+
 struct EventInterface {
     // Mask representing currently busy events
     u64 events_mask{};
     // Each kernel event associated to an NV event
-    std::array<Kernel::EventPair, MaxNvEvents> events;
+    std::array<NvEvent, MaxNvEvents> events;
     // The status of the current NVEvent
     std::array<EventState, MaxNvEvents> status{};
     // Tells if an NVEvent is registered or not
@@ -119,6 +128,9 @@ public:
     std::shared_ptr<Kernel::WritableEvent> GetEventWriteable(u32 event_id) const;
 
 private:
+    /// Manages syncpoints on the host
+    SyncpointManager syncpoint_manager;
+
     /// Id to use for the next open file descriptor.
     u32 next_fd = 1;
 
diff --git a/src/core/hle/service/nvdrv/syncpoint_manager.cpp b/src/core/hle/service/nvdrv/syncpoint_manager.cpp
new file mode 100644
index 000000000..0151a03b7
--- /dev/null
+++ b/src/core/hle/service/nvdrv/syncpoint_manager.cpp
@@ -0,0 +1,39 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/hle/service/nvdrv/syncpoint_manager.h"
+#include "video_core/gpu.h"
+
+namespace Service::Nvidia {
+
+SyncpointManager::SyncpointManager(Tegra::GPU& gpu) : gpu{gpu} {}
+
+SyncpointManager::~SyncpointManager() = default;
+
+u32 SyncpointManager::RefreshSyncpoint(u32 syncpoint_id) {
+    syncpoints[syncpoint_id].min = gpu.GetSyncpointValue(syncpoint_id);
+    return GetSyncpointMin(syncpoint_id);
+}
+
+u32 SyncpointManager::AllocateSyncpoint() {
+    for (u32 syncpoint_id = 1; syncpoint_id < MaxSyncPoints; syncpoint_id++) {
+        if (!syncpoints[syncpoint_id].is_allocated) {
+            syncpoints[syncpoint_id].is_allocated = true;
+            return syncpoint_id;
+        }
+    }
+    UNREACHABLE_MSG("No more available syncpoints!");
+    return {};
+}
+
+u32 SyncpointManager::IncreaseSyncpoint(u32 syncpoint_id, u32 value) {
+    for (u32 index = 0; index < value; ++index) {
+        syncpoints[syncpoint_id].max.fetch_add(1, std::memory_order_relaxed);
+    }
+
+    return GetSyncpointMax(syncpoint_id);
+}
+
+} // namespace Service::Nvidia
diff --git a/src/core/hle/service/nvdrv/syncpoint_manager.h b/src/core/hle/service/nvdrv/syncpoint_manager.h
new file mode 100644
index 000000000..4168b6c7e
--- /dev/null
+++ b/src/core/hle/service/nvdrv/syncpoint_manager.h
@@ -0,0 +1,85 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <atomic>
+
+#include "common/common_types.h"
+#include "core/hle/service/nvdrv/nvdata.h"
+
+namespace Tegra {
+class GPU;
+}
+
+namespace Service::Nvidia {
+
+class SyncpointManager final {
+public:
+    explicit SyncpointManager(Tegra::GPU& gpu);
+    ~SyncpointManager();
+
+    /**
+     * Returns true if the specified syncpoint is expired for the given value.
+     * @param syncpoint_id Syncpoint ID to check.
+     * @param value Value to check against the specified syncpoint.
+     * @returns True if the specified syncpoint is expired for the given value, otherwise False.
+     */
+    bool IsSyncpointExpired(u32 syncpoint_id, u32 value) const {
+        return (GetSyncpointMax(syncpoint_id) - value) >= (GetSyncpointMin(syncpoint_id) - value);
+    }
+
+    /**
+     * Gets the lower bound for the specified syncpoint.
+     * @param syncpoint_id Syncpoint ID to get the lower bound for.
+     * @returns The lower bound for the specified syncpoint.
+     */
+    u32 GetSyncpointMin(u32 syncpoint_id) const {
+        return syncpoints[syncpoint_id].min.load(std::memory_order_relaxed);
+    }
+
+    /**
+     * Gets the uper bound for the specified syncpoint.
+     * @param syncpoint_id Syncpoint ID to get the upper bound for.
+     * @returns The upper bound for the specified syncpoint.
+     */
+    u32 GetSyncpointMax(u32 syncpoint_id) const {
+        return syncpoints[syncpoint_id].max.load(std::memory_order_relaxed);
+    }
+
+    /**
+     * Refreshes the minimum value for the specified syncpoint.
+     * @param syncpoint_id Syncpoint ID to be refreshed.
+     * @returns The new syncpoint minimum value.
+     */
+    u32 RefreshSyncpoint(u32 syncpoint_id);
+
+    /**
+     * Allocates a new syncoint.
+     * @returns The syncpoint ID for the newly allocated syncpoint.
+     */
+    u32 AllocateSyncpoint();
+
+    /**
+     * Increases the maximum value for the specified syncpoint.
+     * @param syncpoint_id Syncpoint ID to be increased.
+     * @param value Value to increase the specified syncpoint by.
+     * @returns The new syncpoint maximum value.
+     */
+    u32 IncreaseSyncpoint(u32 syncpoint_id, u32 value);
+
+private:
+    struct Syncpoint {
+        std::atomic<u32> min;
+        std::atomic<u32> max;
+        std::atomic<bool> is_allocated;
+    };
+
+    std::array<Syncpoint, MaxSyncPoints> syncpoints{};
+
+    Tegra::GPU& gpu;
+};
+
+} // namespace Service::Nvidia
diff --git a/src/core/hle/service/nvflinger/nvflinger.cpp b/src/core/hle/service/nvflinger/nvflinger.cpp
index c64673dba..44aa2bdae 100644
--- a/src/core/hle/service/nvflinger/nvflinger.cpp
+++ b/src/core/hle/service/nvflinger/nvflinger.cpp
@@ -242,6 +242,10 @@ void NVFlinger::Compose() {
 
         const auto& igbp_buffer = buffer->get().igbp_buffer;
 
+        if (!system.IsPoweredOn()) {
+            return; // We are likely shutting down
+        }
+
         auto& gpu = system.GPU();
         const auto& multi_fence = buffer->get().multi_fence;
         guard->unlock();
diff --git a/src/input_common/gcadapter/gc_adapter.cpp b/src/input_common/gcadapter/gc_adapter.cpp
index c95feb0d7..b912188b6 100644
--- a/src/input_common/gcadapter/gc_adapter.cpp
+++ b/src/input_common/gcadapter/gc_adapter.cpp
@@ -21,26 +21,6 @@
 
 namespace GCAdapter {
 
-// Used to loop through and assign button in poller
-constexpr std::array<PadButton, 12> PadButtonArray{
-    PadButton::PAD_BUTTON_LEFT, PadButton::PAD_BUTTON_RIGHT, PadButton::PAD_BUTTON_DOWN,
-    PadButton::PAD_BUTTON_UP,   PadButton::PAD_TRIGGER_Z,    PadButton::PAD_TRIGGER_R,
-    PadButton::PAD_TRIGGER_L,   PadButton::PAD_BUTTON_A,     PadButton::PAD_BUTTON_B,
-    PadButton::PAD_BUTTON_X,    PadButton::PAD_BUTTON_Y,     PadButton::PAD_BUTTON_START,
-};
-
-static void PadToState(const GCPadStatus& pad, GCState& out_state) {
-    for (const auto& button : PadButtonArray) {
-        const auto button_key = static_cast<u16>(button);
-        const auto button_value = (pad.button & button_key) != 0;
-        out_state.buttons.insert_or_assign(static_cast<s32>(button_key), button_value);
-    }
-
-    for (std::size_t i = 0; i < pad.axis_values.size(); ++i) {
-        out_state.axes.insert_or_assign(static_cast<u32>(i), pad.axis_values[i]);
-    }
-}
-
 Adapter::Adapter() {
     if (usb_adapter_handle != nullptr) {
         return;
@@ -49,168 +29,263 @@ Adapter::Adapter() {
 
     const int init_res = libusb_init(&libusb_ctx);
     if (init_res == LIBUSB_SUCCESS) {
-        Setup();
+        adapter_scan_thread = std::thread(&Adapter::AdapterScanThread, this);
     } else {
         LOG_ERROR(Input, "libusb could not be initialized. failed with error = {}", init_res);
     }
 }
 
-GCPadStatus Adapter::GetPadStatus(std::size_t port, const std::array<u8, 37>& adapter_payload) {
-    GCPadStatus pad = {};
-    const std::size_t offset = 1 + (9 * port);
+Adapter::~Adapter() {
+    Reset();
+}
+
+void Adapter::AdapterInputThread() {
+    LOG_DEBUG(Input, "GC Adapter input thread started");
+    s32 payload_size{};
+    AdapterPayload adapter_payload{};
+
+    if (adapter_scan_thread.joinable()) {
+        adapter_scan_thread.join();
+    }
+
+    while (adapter_input_thread_running) {
+        libusb_interrupt_transfer(usb_adapter_handle, input_endpoint, adapter_payload.data(),
+                                  static_cast<s32>(adapter_payload.size()), &payload_size, 16);
+        if (IsPayloadCorrect(adapter_payload, payload_size)) {
+            UpdateControllers(adapter_payload);
+            UpdateVibrations();
+        }
+        std::this_thread::yield();
+    }
 
-    adapter_controllers_status[port] = static_cast<ControllerTypes>(adapter_payload[offset] >> 4);
+    if (restart_scan_thread) {
+        adapter_scan_thread = std::thread(&Adapter::AdapterScanThread, this);
+        restart_scan_thread = false;
+    }
+}
+
+bool Adapter::IsPayloadCorrect(const AdapterPayload& adapter_payload, s32 payload_size) {
+    if (payload_size != static_cast<s32>(adapter_payload.size()) ||
+        adapter_payload[0] != LIBUSB_DT_HID) {
+        LOG_DEBUG(Input, "Error reading payload (size: {}, type: {:02x})", payload_size,
+                  adapter_payload[0]);
+        if (input_error_counter++ > 20) {
+            LOG_ERROR(Input, "GC adapter timeout, Is the adapter connected?");
+            adapter_input_thread_running = false;
+            restart_scan_thread = true;
+        }
+        return false;
+    }
+
+    input_error_counter = 0;
+    return true;
+}
+
+void Adapter::UpdateControllers(const AdapterPayload& adapter_payload) {
+    for (std::size_t port = 0; port < pads.size(); ++port) {
+        const std::size_t offset = 1 + (9 * port);
+        const auto type = static_cast<ControllerTypes>(adapter_payload[offset] >> 4);
+        UpdatePadType(port, type);
+        if (DeviceConnected(port)) {
+            const u8 b1 = adapter_payload[offset + 1];
+            const u8 b2 = adapter_payload[offset + 2];
+            UpdateStateButtons(port, b1, b2);
+            UpdateStateAxes(port, adapter_payload);
+            if (configuring) {
+                UpdateYuzuSettings(port);
+            }
+        }
+    }
+}
+
+void Adapter::UpdatePadType(std::size_t port, ControllerTypes pad_type) {
+    if (pads[port].type == pad_type) {
+        return;
+    }
+    // Device changed reset device and set new type
+    ResetDevice(port);
+    pads[port].type = pad_type;
+}
+
+void Adapter::UpdateStateButtons(std::size_t port, u8 b1, u8 b2) {
+    if (port >= pads.size()) {
+        return;
+    }
 
     static constexpr std::array<PadButton, 8> b1_buttons{
-        PadButton::PAD_BUTTON_A,    PadButton::PAD_BUTTON_B,    PadButton::PAD_BUTTON_X,
-        PadButton::PAD_BUTTON_Y,    PadButton::PAD_BUTTON_LEFT, PadButton::PAD_BUTTON_RIGHT,
-        PadButton::PAD_BUTTON_DOWN, PadButton::PAD_BUTTON_UP,
+        PadButton::ButtonA,    PadButton::ButtonB,     PadButton::ButtonX,    PadButton::ButtonY,
+        PadButton::ButtonLeft, PadButton::ButtonRight, PadButton::ButtonDown, PadButton::ButtonUp,
     };
 
     static constexpr std::array<PadButton, 4> b2_buttons{
-        PadButton::PAD_BUTTON_START,
-        PadButton::PAD_TRIGGER_Z,
-        PadButton::PAD_TRIGGER_R,
-        PadButton::PAD_TRIGGER_L,
+        PadButton::ButtonStart,
+        PadButton::TriggerZ,
+        PadButton::TriggerR,
+        PadButton::TriggerL,
     };
+    pads[port].buttons = 0;
+    for (std::size_t i = 0; i < b1_buttons.size(); ++i) {
+        if ((b1 & (1U << i)) != 0) {
+            pads[port].buttons =
+                static_cast<u16>(pads[port].buttons | static_cast<u16>(b1_buttons[i]));
+            pads[port].last_button = b1_buttons[i];
+        }
+    }
 
+    for (std::size_t j = 0; j < b2_buttons.size(); ++j) {
+        if ((b2 & (1U << j)) != 0) {
+            pads[port].buttons =
+                static_cast<u16>(pads[port].buttons | static_cast<u16>(b2_buttons[j]));
+            pads[port].last_button = b2_buttons[j];
+        }
+    }
+}
+
+void Adapter::UpdateStateAxes(std::size_t port, const AdapterPayload& adapter_payload) {
+    if (port >= pads.size()) {
+        return;
+    }
+
+    const std::size_t offset = 1 + (9 * port);
     static constexpr std::array<PadAxes, 6> axes{
         PadAxes::StickX,    PadAxes::StickY,      PadAxes::SubstickX,
         PadAxes::SubstickY, PadAxes::TriggerLeft, PadAxes::TriggerRight,
     };
 
-    if (adapter_controllers_status[port] == ControllerTypes::None && !get_origin[port]) {
-        // Controller may have been disconnected, recalibrate if reconnected.
-        get_origin[port] = true;
+    for (const PadAxes axis : axes) {
+        const auto index = static_cast<std::size_t>(axis);
+        const u8 axis_value = adapter_payload[offset + 3 + index];
+        if (pads[port].axis_origin[index] == 255) {
+            pads[port].axis_origin[index] = axis_value;
+        }
+        pads[port].axis_values[index] =
+            static_cast<s16>(axis_value - pads[port].axis_origin[index]);
     }
+}
 
-    if (adapter_controllers_status[port] != ControllerTypes::None) {
-        const u8 b1 = adapter_payload[offset + 1];
-        const u8 b2 = adapter_payload[offset + 2];
+void Adapter::UpdateYuzuSettings(std::size_t port) {
+    if (port >= pads.size()) {
+        return;
+    }
 
-        for (std::size_t i = 0; i < b1_buttons.size(); ++i) {
-            if ((b1 & (1U << i)) != 0) {
-                pad.button = static_cast<u16>(pad.button | static_cast<u16>(b1_buttons[i]));
-            }
-        }
+    constexpr u8 axis_threshold = 50;
+    GCPadStatus pad_status = {.port = port};
 
-        for (std::size_t j = 0; j < b2_buttons.size(); ++j) {
-            if ((b2 & (1U << j)) != 0) {
-                pad.button = static_cast<u16>(pad.button | static_cast<u16>(b2_buttons[j]));
-            }
-        }
-        for (PadAxes axis : axes) {
-            const auto index = static_cast<std::size_t>(axis);
-            pad.axis_values[index] = adapter_payload[offset + 3 + index];
-        }
+    if (pads[port].buttons != 0) {
+        pad_status.button = pads[port].last_button;
+        pad_queue.Push(pad_status);
+    }
+
+    // Accounting for a threshold here to ensure an intentional press
+    for (std::size_t i = 0; i < pads[port].axis_values.size(); ++i) {
+        const s16 value = pads[port].axis_values[i];
 
-        if (get_origin[port]) {
-            origin_status[port].axis_values = pad.axis_values;
-            get_origin[port] = false;
+        if (value > axis_threshold || value < -axis_threshold) {
+            pad_status.axis = static_cast<PadAxes>(i);
+            pad_status.axis_value = value;
+            pad_status.axis_threshold = axis_threshold;
+            pad_queue.Push(pad_status);
         }
     }
-    return pad;
 }
 
-void Adapter::Read() {
-    LOG_DEBUG(Input, "GC Adapter Read() thread started");
+void Adapter::UpdateVibrations() {
+    // Use 8 states to keep the switching between on/off fast enough for
+    // a human to not notice the difference between switching from on/off
+    // More states = more rumble strengths = slower update time
+    constexpr u8 vibration_states = 8;
 
-    int payload_size;
-    std::array<u8, 37> adapter_payload;
-    std::array<GCPadStatus, 4> pads;
-
-    while (adapter_thread_running) {
-        libusb_interrupt_transfer(usb_adapter_handle, input_endpoint, adapter_payload.data(),
-                                  sizeof(adapter_payload), &payload_size, 16);
-
-        if (payload_size != sizeof(adapter_payload) || adapter_payload[0] != LIBUSB_DT_HID) {
-            LOG_ERROR(Input,
-                      "Error reading payload (size: {}, type: {:02x}) Is the adapter connected?",
-                      payload_size, adapter_payload[0]);
-            adapter_thread_running = false; // error reading from adapter, stop reading.
-            break;
-        }
-        for (std::size_t port = 0; port < pads.size(); ++port) {
-            pads[port] = GetPadStatus(port, adapter_payload);
-            if (DeviceConnected(port) && configuring) {
-                if (pads[port].button != 0) {
-                    pad_queue[port].Push(pads[port]);
-                }
+    vibration_counter = (vibration_counter + 1) % vibration_states;
 
-                // Accounting for a threshold here to ensure an intentional press
-                for (size_t i = 0; i < pads[port].axis_values.size(); ++i) {
-                    const u8 value = pads[port].axis_values[i];
-                    const u8 origin = origin_status[port].axis_values[i];
-
-                    if (value > origin + pads[port].THRESHOLD ||
-                        value < origin - pads[port].THRESHOLD) {
-                        pads[port].axis = static_cast<PadAxes>(i);
-                        pads[port].axis_value = pads[port].axis_values[i];
-                        pad_queue[port].Push(pads[port]);
-                    }
-                }
-            }
-            PadToState(pads[port], state[port]);
-        }
-        std::this_thread::yield();
+    for (GCController& pad : pads) {
+        const bool vibrate = pad.rumble_amplitude > vibration_counter;
+        vibration_changed |= vibrate != pad.enable_vibration;
+        pad.enable_vibration = vibrate;
     }
+    SendVibrations();
 }
 
-void Adapter::Setup() {
-    // Initialize all controllers as unplugged
-    adapter_controllers_status.fill(ControllerTypes::None);
-    // Initialize all ports to store axis origin values
-    get_origin.fill(true);
-
-    // pointer to list of connected usb devices
-    libusb_device** devices{};
-
-    // populate the list of devices, get the count
-    const ssize_t device_count = libusb_get_device_list(libusb_ctx, &devices);
-    if (device_count < 0) {
-        LOG_ERROR(Input, "libusb_get_device_list failed with error: {}", device_count);
+void Adapter::SendVibrations() {
+    if (!rumble_enabled || !vibration_changed) {
         return;
     }
-
-    if (devices != nullptr) {
-        for (std::size_t index = 0; index < static_cast<std::size_t>(device_count); ++index) {
-            if (CheckDeviceAccess(devices[index])) {
-                // GC Adapter found and accessible, registering it
-                GetGCEndpoint(devices[index]);
-                break;
-            }
+    s32 size{};
+    constexpr u8 rumble_command = 0x11;
+    const u8 p1 = pads[0].enable_vibration;
+    const u8 p2 = pads[1].enable_vibration;
+    const u8 p3 = pads[2].enable_vibration;
+    const u8 p4 = pads[3].enable_vibration;
+    std::array<u8, 5> payload = {rumble_command, p1, p2, p3, p4};
+    const int err = libusb_interrupt_transfer(usb_adapter_handle, output_endpoint, payload.data(),
+                                              static_cast<s32>(payload.size()), &size, 16);
+    if (err) {
+        LOG_DEBUG(Input, "Adapter libusb write failed: {}", libusb_error_name(err));
+        if (output_error_counter++ > 5) {
+            LOG_ERROR(Input, "GC adapter output timeout, Rumble disabled");
+            rumble_enabled = false;
         }
-        libusb_free_device_list(devices, 1);
+        return;
     }
+    output_error_counter = 0;
+    vibration_changed = false;
 }
 
-bool Adapter::CheckDeviceAccess(libusb_device* device) {
-    libusb_device_descriptor desc;
-    const int get_descriptor_error = libusb_get_device_descriptor(device, &desc);
-    if (get_descriptor_error) {
-        // could not acquire the descriptor, no point in trying to use it.
-        LOG_ERROR(Input, "libusb_get_device_descriptor failed with error: {}",
-                  get_descriptor_error);
-        return false;
+bool Adapter::RumblePlay(std::size_t port, f32 amplitude) {
+    amplitude = std::clamp(amplitude, 0.0f, 1.0f);
+    const auto raw_amp = static_cast<u8>(amplitude * 0x8);
+    pads[port].rumble_amplitude = raw_amp;
+
+    return rumble_enabled;
+}
+
+void Adapter::AdapterScanThread() {
+    adapter_scan_thread_running = true;
+    adapter_input_thread_running = false;
+    if (adapter_input_thread.joinable()) {
+        adapter_input_thread.join();
+    }
+    ClearLibusbHandle();
+    ResetDevices();
+    while (adapter_scan_thread_running && !adapter_input_thread_running) {
+        Setup();
+        std::this_thread::sleep_for(std::chrono::seconds(1));
     }
+}
 
-    if (desc.idVendor != 0x057e || desc.idProduct != 0x0337) {
-        // This isn't the device we are looking for.
-        return false;
+void Adapter::Setup() {
+    usb_adapter_handle = libusb_open_device_with_vid_pid(libusb_ctx, 0x057e, 0x0337);
+
+    if (usb_adapter_handle == NULL) {
+        return;
+    }
+    if (!CheckDeviceAccess()) {
+        ClearLibusbHandle();
+        return;
     }
-    const int open_error = libusb_open(device, &usb_adapter_handle);
 
-    if (open_error == LIBUSB_ERROR_ACCESS) {
-        LOG_ERROR(Input, "Yuzu can not gain access to this device: ID {:04X}:{:04X}.",
-                  desc.idVendor, desc.idProduct);
-        return false;
+    libusb_device* device = libusb_get_device(usb_adapter_handle);
+
+    LOG_INFO(Input, "GC adapter is now connected");
+    // GC Adapter found and accessible, registering it
+    if (GetGCEndpoint(device)) {
+        adapter_scan_thread_running = false;
+        adapter_input_thread_running = true;
+        rumble_enabled = true;
+        input_error_counter = 0;
+        output_error_counter = 0;
+        adapter_input_thread = std::thread(&Adapter::AdapterInputThread, this);
     }
-    if (open_error) {
-        LOG_ERROR(Input, "libusb_open failed to open device with error = {}", open_error);
-        return false;
+}
+
+bool Adapter::CheckDeviceAccess() {
+    // This fixes payload problems from offbrand GCAdapters
+    const s32 control_transfer_error =
+        libusb_control_transfer(usb_adapter_handle, 0x21, 11, 0x0001, 0, nullptr, 0, 1000);
+    if (control_transfer_error < 0) {
+        LOG_ERROR(Input, "libusb_control_transfer failed with error= {}", control_transfer_error);
     }
 
-    int kernel_driver_error = libusb_kernel_driver_active(usb_adapter_handle, 0);
+    s32 kernel_driver_error = libusb_kernel_driver_active(usb_adapter_handle, 0);
     if (kernel_driver_error == 1) {
         kernel_driver_error = libusb_detach_kernel_driver(usb_adapter_handle, 0);
         if (kernel_driver_error != 0 && kernel_driver_error != LIBUSB_ERROR_NOT_SUPPORTED) {
@@ -236,13 +311,13 @@ bool Adapter::CheckDeviceAccess(libusb_device* device) {
     return true;
 }
 
-void Adapter::GetGCEndpoint(libusb_device* device) {
+bool Adapter::GetGCEndpoint(libusb_device* device) {
     libusb_config_descriptor* config = nullptr;
     const int config_descriptor_return = libusb_get_config_descriptor(device, 0, &config);
     if (config_descriptor_return != LIBUSB_SUCCESS) {
         LOG_ERROR(Input, "libusb_get_config_descriptor failed with error = {}",
                   config_descriptor_return);
-        return;
+        return false;
     }
 
     for (u8 ic = 0; ic < config->bNumInterfaces; ic++) {
@@ -264,31 +339,51 @@ void Adapter::GetGCEndpoint(libusb_device* device) {
     unsigned char clear_payload = 0x13;
     libusb_interrupt_transfer(usb_adapter_handle, output_endpoint, &clear_payload,
                               sizeof(clear_payload), nullptr, 16);
-
-    adapter_thread_running = true;
-    adapter_input_thread = std::thread(&Adapter::Read, this);
+    return true;
 }
 
-Adapter::~Adapter() {
-    Reset();
-}
+void Adapter::JoinThreads() {
+    restart_scan_thread = false;
+    adapter_input_thread_running = false;
+    adapter_scan_thread_running = false;
 
-void Adapter::Reset() {
-    if (adapter_thread_running) {
-        adapter_thread_running = false;
+    if (adapter_scan_thread.joinable()) {
+        adapter_scan_thread.join();
     }
+
     if (adapter_input_thread.joinable()) {
         adapter_input_thread.join();
     }
+}
 
-    adapter_controllers_status.fill(ControllerTypes::None);
-    get_origin.fill(true);
-
+void Adapter::ClearLibusbHandle() {
     if (usb_adapter_handle) {
         libusb_release_interface(usb_adapter_handle, 1);
         libusb_close(usb_adapter_handle);
         usb_adapter_handle = nullptr;
     }
+}
+
+void Adapter::ResetDevices() {
+    for (std::size_t i = 0; i < pads.size(); ++i) {
+        ResetDevice(i);
+    }
+}
+
+void Adapter::ResetDevice(std::size_t port) {
+    pads[port].type = ControllerTypes::None;
+    pads[port].enable_vibration = false;
+    pads[port].rumble_amplitude = 0;
+    pads[port].buttons = 0;
+    pads[port].last_button = PadButton::Undefined;
+    pads[port].axis_values.fill(0);
+    pads[port].axis_origin.fill(255);
+}
+
+void Adapter::Reset() {
+    JoinThreads();
+    ClearLibusbHandle();
+    ResetDevices();
 
     if (libusb_ctx) {
         libusb_exit(libusb_ctx);
@@ -297,11 +392,11 @@ void Adapter::Reset() {
 
 std::vector<Common::ParamPackage> Adapter::GetInputDevices() const {
     std::vector<Common::ParamPackage> devices;
-    for (std::size_t port = 0; port < state.size(); ++port) {
+    for (std::size_t port = 0; port < pads.size(); ++port) {
         if (!DeviceConnected(port)) {
             continue;
         }
-        std::string name = fmt::format("Gamecube Controller {}", port);
+        std::string name = fmt::format("Gamecube Controller {}", port + 1);
         devices.emplace_back(Common::ParamPackage{
             {"class", "gcpad"},
             {"display", std::move(name)},
@@ -318,18 +413,18 @@ InputCommon::ButtonMapping Adapter::GetButtonMappingForDevice(
     // This list also excludes any button that can't be really mapped
     static constexpr std::array<std::pair<Settings::NativeButton::Values, PadButton>, 12>
         switch_to_gcadapter_button = {
-            std::pair{Settings::NativeButton::A, PadButton::PAD_BUTTON_A},
-            {Settings::NativeButton::B, PadButton::PAD_BUTTON_B},
-            {Settings::NativeButton::X, PadButton::PAD_BUTTON_X},
-            {Settings::NativeButton::Y, PadButton::PAD_BUTTON_Y},
-            {Settings::NativeButton::Plus, PadButton::PAD_BUTTON_START},
-            {Settings::NativeButton::DLeft, PadButton::PAD_BUTTON_LEFT},
-            {Settings::NativeButton::DUp, PadButton::PAD_BUTTON_UP},
-            {Settings::NativeButton::DRight, PadButton::PAD_BUTTON_RIGHT},
-            {Settings::NativeButton::DDown, PadButton::PAD_BUTTON_DOWN},
-            {Settings::NativeButton::SL, PadButton::PAD_TRIGGER_L},
-            {Settings::NativeButton::SR, PadButton::PAD_TRIGGER_R},
-            {Settings::NativeButton::R, PadButton::PAD_TRIGGER_Z},
+            std::pair{Settings::NativeButton::A, PadButton::ButtonA},
+            {Settings::NativeButton::B, PadButton::ButtonB},
+            {Settings::NativeButton::X, PadButton::ButtonX},
+            {Settings::NativeButton::Y, PadButton::ButtonY},
+            {Settings::NativeButton::Plus, PadButton::ButtonStart},
+            {Settings::NativeButton::DLeft, PadButton::ButtonLeft},
+            {Settings::NativeButton::DUp, PadButton::ButtonUp},
+            {Settings::NativeButton::DRight, PadButton::ButtonRight},
+            {Settings::NativeButton::DDown, PadButton::ButtonDown},
+            {Settings::NativeButton::SL, PadButton::TriggerL},
+            {Settings::NativeButton::SR, PadButton::TriggerR},
+            {Settings::NativeButton::R, PadButton::TriggerZ},
         };
     if (!params.Has("port")) {
         return {};
@@ -352,8 +447,10 @@ InputCommon::ButtonMapping Adapter::GetButtonMappingForDevice(
     for (const auto& [switch_button, gcadapter_axis] : switch_to_gcadapter_axis) {
         Common::ParamPackage button_params({{"engine", "gcpad"}});
         button_params.Set("port", params.Get("port", 0));
-        button_params.Set("button", static_cast<int>(PadButton::PAD_STICK));
-        button_params.Set("axis", static_cast<int>(gcadapter_axis));
+        button_params.Set("button", static_cast<s32>(PadButton::Stick));
+        button_params.Set("axis", static_cast<s32>(gcadapter_axis));
+        button_params.Set("threshold", 0.5f);
+        button_params.Set("direction", "+");
         mapping.insert_or_assign(switch_button, std::move(button_params));
     }
     return mapping;
@@ -382,46 +479,33 @@ InputCommon::AnalogMapping Adapter::GetAnalogMappingForDevice(
 }
 
 bool Adapter::DeviceConnected(std::size_t port) const {
-    return adapter_controllers_status[port] != ControllerTypes::None;
-}
-
-void Adapter::ResetDeviceType(std::size_t port) {
-    adapter_controllers_status[port] = ControllerTypes::None;
+    return pads[port].type != ControllerTypes::None;
 }
 
 void Adapter::BeginConfiguration() {
-    get_origin.fill(true);
-    for (auto& pq : pad_queue) {
-        pq.Clear();
-    }
+    pad_queue.Clear();
     configuring = true;
 }
 
 void Adapter::EndConfiguration() {
-    for (auto& pq : pad_queue) {
-        pq.Clear();
-    }
+    pad_queue.Clear();
     configuring = false;
 }
 
-std::array<Common::SPSCQueue<GCPadStatus>, 4>& Adapter::GetPadQueue() {
+Common::SPSCQueue<GCPadStatus>& Adapter::GetPadQueue() {
     return pad_queue;
 }
 
-const std::array<Common::SPSCQueue<GCPadStatus>, 4>& Adapter::GetPadQueue() const {
+const Common::SPSCQueue<GCPadStatus>& Adapter::GetPadQueue() const {
     return pad_queue;
 }
 
-std::array<GCState, 4>& Adapter::GetPadState() {
-    return state;
-}
-
-const std::array<GCState, 4>& Adapter::GetPadState() const {
-    return state;
+GCController& Adapter::GetPadState(std::size_t port) {
+    return pads.at(port);
 }
 
-int Adapter::GetOriginValue(u32 port, u32 axis) const {
-    return origin_status[port].axis_values[axis];
+const GCController& Adapter::GetPadState(std::size_t port) const {
+    return pads.at(port);
 }
 
 } // namespace GCAdapter
diff --git a/src/input_common/gcadapter/gc_adapter.h b/src/input_common/gcadapter/gc_adapter.h
index 4f5f3de8e..d28dcfad3 100644
--- a/src/input_common/gcadapter/gc_adapter.h
+++ b/src/input_common/gcadapter/gc_adapter.h
@@ -19,24 +19,23 @@ struct libusb_device_handle;
 namespace GCAdapter {
 
 enum class PadButton {
-    PAD_BUTTON_LEFT = 0x0001,
-    PAD_BUTTON_RIGHT = 0x0002,
-    PAD_BUTTON_DOWN = 0x0004,
-    PAD_BUTTON_UP = 0x0008,
-    PAD_TRIGGER_Z = 0x0010,
-    PAD_TRIGGER_R = 0x0020,
-    PAD_TRIGGER_L = 0x0040,
-    PAD_BUTTON_A = 0x0100,
-    PAD_BUTTON_B = 0x0200,
-    PAD_BUTTON_X = 0x0400,
-    PAD_BUTTON_Y = 0x0800,
-    PAD_BUTTON_START = 0x1000,
+    Undefined = 0x0000,
+    ButtonLeft = 0x0001,
+    ButtonRight = 0x0002,
+    ButtonDown = 0x0004,
+    ButtonUp = 0x0008,
+    TriggerZ = 0x0010,
+    TriggerR = 0x0020,
+    TriggerL = 0x0040,
+    ButtonA = 0x0100,
+    ButtonB = 0x0200,
+    ButtonX = 0x0400,
+    ButtonY = 0x0800,
+    ButtonStart = 0x1000,
     // Below is for compatibility with "AxisButton" type
-    PAD_STICK = 0x2000,
+    Stick = 0x2000,
 };
 
-extern const std::array<PadButton, 12> PadButtonArray;
-
 enum class PadAxes : u8 {
     StickX,
     StickY,
@@ -47,87 +46,122 @@ enum class PadAxes : u8 {
     Undefined,
 };
 
+enum class ControllerTypes {
+    None,
+    Wired,
+    Wireless,
+};
+
 struct GCPadStatus {
-    u16 button{}; // Or-ed PAD_BUTTON_* and PAD_TRIGGER_* bits
+    std::size_t port{};
 
-    std::array<u8, 6> axis_values{};    // Triggers and sticks, following indices defined in PadAxes
-    static constexpr u8 THRESHOLD = 50; // Threshold for axis press for polling
+    PadButton button{PadButton::Undefined}; // Or-ed PAD_BUTTON_* and PAD_TRIGGER_* bits
 
-    u8 port{};
     PadAxes axis{PadAxes::Undefined};
-    u8 axis_value{255};
+    s16 axis_value{};
+    u8 axis_threshold{50};
 };
 
-struct GCState {
-    std::unordered_map<int, bool> buttons;
-    std::unordered_map<u32, u16> axes;
+struct GCController {
+    ControllerTypes type{};
+    bool enable_vibration{};
+    u8 rumble_amplitude{};
+    u16 buttons{};
+    PadButton last_button{};
+    std::array<s16, 6> axis_values{};
+    std::array<u8, 6> axis_origin{};
 };
 
-enum class ControllerTypes { None, Wired, Wireless };
-
 class Adapter {
 public:
-    /// Initialize the GC Adapter capture and read sequence
     Adapter();
-
-    /// Close the adapter read thread and release the adapter
     ~Adapter();
+
+    /// Request a vibration for a controlelr
+    bool RumblePlay(std::size_t port, f32 amplitude);
+
     /// Used for polling
     void BeginConfiguration();
     void EndConfiguration();
 
+    Common::SPSCQueue<GCPadStatus>& GetPadQueue();
+    const Common::SPSCQueue<GCPadStatus>& GetPadQueue() const;
+
+    GCController& GetPadState(std::size_t port);
+    const GCController& GetPadState(std::size_t port) const;
+
+    /// Returns true if there is a device connected to port
+    bool DeviceConnected(std::size_t port) const;
+
+    /// Used for automapping features
     std::vector<Common::ParamPackage> GetInputDevices() const;
     InputCommon::ButtonMapping GetButtonMappingForDevice(const Common::ParamPackage& params) const;
     InputCommon::AnalogMapping GetAnalogMappingForDevice(const Common::ParamPackage& params) const;
 
-    /// Returns true if there is a device connected to port
-    bool DeviceConnected(std::size_t port) const;
+private:
+    using AdapterPayload = std::array<u8, 37>;
 
-    std::array<Common::SPSCQueue<GCPadStatus>, 4>& GetPadQueue();
-    const std::array<Common::SPSCQueue<GCPadStatus>, 4>& GetPadQueue() const;
+    void UpdatePadType(std::size_t port, ControllerTypes pad_type);
+    void UpdateControllers(const AdapterPayload& adapter_payload);
+    void UpdateYuzuSettings(std::size_t port);
+    void UpdateStateButtons(std::size_t port, u8 b1, u8 b2);
+    void UpdateStateAxes(std::size_t port, const AdapterPayload& adapter_payload);
+    void UpdateVibrations();
 
-    std::array<GCState, 4>& GetPadState();
-    const std::array<GCState, 4>& GetPadState() const;
+    void AdapterInputThread();
 
-    int GetOriginValue(u32 port, u32 axis) const;
+    void AdapterScanThread();
 
-private:
-    GCPadStatus GetPadStatus(std::size_t port, const std::array<u8, 37>& adapter_payload);
+    bool IsPayloadCorrect(const AdapterPayload& adapter_payload, s32 payload_size);
+
+    // Updates vibration state of all controllers
+    void SendVibrations();
+
+    /// For use in initialization, querying devices to find the adapter
+    void Setup();
 
-    void Read();
+    /// Resets status of all GC controller devices to a disconected state
+    void ResetDevices();
 
-    /// Resets status of device connected to port
-    void ResetDeviceType(std::size_t port);
+    /// Resets status of device connected to a disconected state
+    void ResetDevice(std::size_t port);
 
     /// Returns true if we successfully gain access to GC Adapter
-    bool CheckDeviceAccess(libusb_device* device);
+    bool CheckDeviceAccess();
 
-    /// Captures GC Adapter endpoint address,
-    void GetGCEndpoint(libusb_device* device);
+    /// Captures GC Adapter endpoint address
+    /// Returns true if the endpoind was set correctly
+    bool GetGCEndpoint(libusb_device* device);
 
     /// For shutting down, clear all data, join all threads, release usb
     void Reset();
 
-    /// For use in initialization, querying devices to find the adapter
-    void Setup();
+    // Join all threads
+    void JoinThreads();
+
+    // Release usb handles
+    void ClearLibusbHandle();
 
     libusb_device_handle* usb_adapter_handle = nullptr;
+    std::array<GCController, 4> pads;
+    Common::SPSCQueue<GCPadStatus> pad_queue;
 
     std::thread adapter_input_thread;
-    bool adapter_thread_running;
+    std::thread adapter_scan_thread;
+    bool adapter_input_thread_running;
+    bool adapter_scan_thread_running;
+    bool restart_scan_thread;
 
     libusb_context* libusb_ctx;
 
-    u8 input_endpoint = 0;
-    u8 output_endpoint = 0;
-
-    bool configuring = false;
+    u8 input_endpoint{0};
+    u8 output_endpoint{0};
+    u8 input_error_counter{0};
+    u8 output_error_counter{0};
+    int vibration_counter{0};
 
-    std::array<GCState, 4> state;
-    std::array<bool, 4> get_origin;
-    std::array<GCPadStatus, 4> origin_status;
-    std::array<Common::SPSCQueue<GCPadStatus>, 4> pad_queue;
-    std::array<ControllerTypes, 4> adapter_controllers_status{};
+    bool configuring{false};
+    bool rumble_enabled{true};
+    bool vibration_changed{true};
 };
-
 } // namespace GCAdapter
diff --git a/src/input_common/gcadapter/gc_poller.cpp b/src/input_common/gcadapter/gc_poller.cpp
index 893556916..6bd6f57fc 100644
--- a/src/input_common/gcadapter/gc_poller.cpp
+++ b/src/input_common/gcadapter/gc_poller.cpp
@@ -15,22 +15,30 @@ namespace InputCommon {
 
 class GCButton final : public Input::ButtonDevice {
 public:
-    explicit GCButton(u32 port_, int button_, const GCAdapter::Adapter* adapter)
+    explicit GCButton(u32 port_, s32 button_, GCAdapter::Adapter* adapter)
         : port(port_), button(button_), gcadapter(adapter) {}
 
     ~GCButton() override;
 
     bool GetStatus() const override {
         if (gcadapter->DeviceConnected(port)) {
-            return gcadapter->GetPadState()[port].buttons.at(button);
+            return (gcadapter->GetPadState(port).buttons & button) != 0;
         }
         return false;
     }
 
+    bool SetRumblePlay(f32 amp_high, f32 amp_low, f32 freq_high, f32 freq_low) const override {
+        const float amplitude = amp_high + amp_low > 2.0f ? 1.0f : (amp_high + amp_low) * 0.5f;
+        const auto new_amp =
+            static_cast<f32>(pow(amplitude, 0.5f) * (3.0f - 2.0f * pow(amplitude, 0.15f)));
+
+        return gcadapter->RumblePlay(port, new_amp);
+    }
+
 private:
     const u32 port;
-    const int button;
-    const GCAdapter::Adapter* gcadapter;
+    const s32 button;
+    GCAdapter::Adapter* gcadapter;
 };
 
 class GCAxisButton final : public Input::ButtonDevice {
@@ -38,13 +46,12 @@ public:
     explicit GCAxisButton(u32 port_, u32 axis_, float threshold_, bool trigger_if_greater_,
                           const GCAdapter::Adapter* adapter)
         : port(port_), axis(axis_), threshold(threshold_), trigger_if_greater(trigger_if_greater_),
-          gcadapter(adapter),
-          origin_value(static_cast<float>(adapter->GetOriginValue(port_, axis_))) {}
+          gcadapter(adapter) {}
 
     bool GetStatus() const override {
         if (gcadapter->DeviceConnected(port)) {
-            const float current_axis_value = gcadapter->GetPadState()[port].axes.at(axis);
-            const float axis_value = (current_axis_value - origin_value) / 128.0f;
+            const float current_axis_value = gcadapter->GetPadState(port).axis_values.at(axis);
+            const float axis_value = current_axis_value / 128.0f;
             if (trigger_if_greater) {
                 // TODO: Might be worthwile to set a slider for the trigger threshold. It is
                 // currently always set to 0.5 in configure_input_player.cpp ZL/ZR HandleClick
@@ -61,7 +68,6 @@ private:
     float threshold;
     bool trigger_if_greater;
     const GCAdapter::Adapter* gcadapter;
-    const float origin_value;
 };
 
 GCButtonFactory::GCButtonFactory(std::shared_ptr<GCAdapter::Adapter> adapter_)
@@ -73,7 +79,7 @@ std::unique_ptr<Input::ButtonDevice> GCButtonFactory::Create(const Common::Param
     const auto button_id = params.Get("button", 0);
     const auto port = static_cast<u32>(params.Get("port", 0));
 
-    constexpr int PAD_STICK_ID = static_cast<u16>(GCAdapter::PadButton::PAD_STICK);
+    constexpr s32 PAD_STICK_ID = static_cast<s32>(GCAdapter::PadButton::Stick);
 
     // button is not an axis/stick button
     if (button_id != PAD_STICK_ID) {
@@ -106,32 +112,25 @@ Common::ParamPackage GCButtonFactory::GetNextInput() const {
     Common::ParamPackage params;
     GCAdapter::GCPadStatus pad;
     auto& queue = adapter->GetPadQueue();
-    for (std::size_t port = 0; port < queue.size(); ++port) {
-        while (queue[port].Pop(pad)) {
-            // This while loop will break on the earliest detected button
-            params.Set("engine", "gcpad");
-            params.Set("port", static_cast<int>(port));
-            for (const auto& button : GCAdapter::PadButtonArray) {
-                const u16 button_value = static_cast<u16>(button);
-                if (pad.button & button_value) {
-                    params.Set("button", button_value);
-                    break;
-                }
-            }
+    while (queue.Pop(pad)) {
+        // This while loop will break on the earliest detected button
+        params.Set("engine", "gcpad");
+        params.Set("port", static_cast<s32>(pad.port));
+        if (pad.button != GCAdapter::PadButton::Undefined) {
+            params.Set("button", static_cast<u16>(pad.button));
+        }
 
-            // For Axis button implementation
-            if (pad.axis != GCAdapter::PadAxes::Undefined) {
-                params.Set("axis", static_cast<u8>(pad.axis));
-                params.Set("button", static_cast<u16>(GCAdapter::PadButton::PAD_STICK));
-                if (pad.axis_value > 128) {
-                    params.Set("direction", "+");
-                    params.Set("threshold", "0.25");
-                } else {
-                    params.Set("direction", "-");
-                    params.Set("threshold", "-0.25");
-                }
-                break;
+        // For Axis button implementation
+        if (pad.axis != GCAdapter::PadAxes::Undefined) {
+            params.Set("axis", static_cast<u8>(pad.axis));
+            params.Set("button", static_cast<u16>(GCAdapter::PadButton::Stick));
+            params.Set("threshold", "0.25");
+            if (pad.axis_value > 0) {
+                params.Set("direction", "+");
+            } else {
+                params.Set("direction", "-");
             }
+            break;
         }
     }
     return params;
@@ -152,17 +151,14 @@ public:
     explicit GCAnalog(u32 port_, u32 axis_x_, u32 axis_y_, float deadzone_,
                       const GCAdapter::Adapter* adapter, float range_)
         : port(port_), axis_x(axis_x_), axis_y(axis_y_), deadzone(deadzone_), gcadapter(adapter),
-          origin_value_x(static_cast<float>(adapter->GetOriginValue(port_, axis_x_))),
-          origin_value_y(static_cast<float>(adapter->GetOriginValue(port_, axis_y_))),
           range(range_) {}
 
     float GetAxis(u32 axis) const {
         if (gcadapter->DeviceConnected(port)) {
             std::lock_guard lock{mutex};
-            const auto origin_value = axis % 2 == 0 ? origin_value_x : origin_value_y;
             const auto axis_value =
-                static_cast<float>(gcadapter->GetPadState()[port].axes.at(axis));
-            return (axis_value - origin_value) / (100.0f * range);
+                static_cast<float>(gcadapter->GetPadState(port).axis_values.at(axis));
+            return (axis_value) / (100.0f * range);
         }
         return 0.0f;
     }
@@ -215,8 +211,6 @@ private:
     const u32 axis_y;
     const float deadzone;
     const GCAdapter::Adapter* gcadapter;
-    const float origin_value_x;
-    const float origin_value_y;
     const float range;
     mutable std::mutex mutex;
 };
@@ -254,26 +248,44 @@ void GCAnalogFactory::EndConfiguration() {
 
 Common::ParamPackage GCAnalogFactory::GetNextInput() {
     GCAdapter::GCPadStatus pad;
+    Common::ParamPackage params;
     auto& queue = adapter->GetPadQueue();
-    for (std::size_t port = 0; port < queue.size(); ++port) {
-        while (queue[port].Pop(pad)) {
-            if (pad.axis == GCAdapter::PadAxes::Undefined ||
-                std::abs((static_cast<float>(pad.axis_value) - 128.0f) / 128.0f) < 0.1f) {
-                continue;
-            }
-            // An analog device needs two axes, so we need to store the axis for later and wait for
-            // a second input event. The axes also must be from the same joystick.
-            const u8 axis = static_cast<u8>(pad.axis);
-            if (analog_x_axis == -1) {
-                analog_x_axis = axis;
-                controller_number = static_cast<int>(port);
-            } else if (analog_y_axis == -1 && analog_x_axis != axis &&
-                       controller_number == static_cast<int>(port)) {
-                analog_y_axis = axis;
-            }
+    while (queue.Pop(pad)) {
+        if (pad.button != GCAdapter::PadButton::Undefined) {
+            params.Set("engine", "gcpad");
+            params.Set("port", static_cast<s32>(pad.port));
+            params.Set("button", static_cast<u16>(pad.button));
+            return params;
+        }
+        if (pad.axis == GCAdapter::PadAxes::Undefined ||
+            std::abs(static_cast<float>(pad.axis_value) / 128.0f) < 0.1f) {
+            continue;
+        }
+        // An analog device needs two axes, so we need to store the axis for later and wait for
+        // a second input event. The axes also must be from the same joystick.
+        const u8 axis = static_cast<u8>(pad.axis);
+        if (axis == 0 || axis == 1) {
+            analog_x_axis = 0;
+            analog_y_axis = 1;
+            controller_number = static_cast<s32>(pad.port);
+            break;
+        }
+        if (axis == 2 || axis == 3) {
+            analog_x_axis = 2;
+            analog_y_axis = 3;
+            controller_number = static_cast<s32>(pad.port);
+            break;
+        }
+
+        if (analog_x_axis == -1) {
+            analog_x_axis = axis;
+            controller_number = static_cast<s32>(pad.port);
+        } else if (analog_y_axis == -1 && analog_x_axis != axis &&
+                   controller_number == static_cast<s32>(pad.port)) {
+            analog_y_axis = axis;
+            break;
         }
     }
-    Common::ParamPackage params;
     if (analog_x_axis != -1 && analog_y_axis != -1) {
         params.Set("engine", "gcpad");
         params.Set("port", controller_number);
diff --git a/src/video_core/command_classes/codecs/codec.cpp b/src/video_core/command_classes/codecs/codec.cpp
index 2df410be8..1adf3cd13 100644
--- a/src/video_core/command_classes/codecs/codec.cpp
+++ b/src/video_core/command_classes/codecs/codec.cpp
@@ -4,6 +4,7 @@
 
 #include <cstring>
 #include <fstream>
+#include <vector>
 #include "common/assert.h"
 #include "video_core/command_classes/codecs/codec.h"
 #include "video_core/command_classes/codecs/h264.h"
diff --git a/src/video_core/command_classes/codecs/codec.h b/src/video_core/command_classes/codecs/codec.h
index 2e56daf29..5bbe6a332 100644
--- a/src/video_core/command_classes/codecs/codec.h
+++ b/src/video_core/command_classes/codecs/codec.h
@@ -5,8 +5,6 @@
 #pragma once
 
 #include <memory>
-#include <vector>
-#include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/command_classes/nvdec_common.h"
 
@@ -44,11 +42,11 @@ public:
     void Decode();
 
     /// Returns most recently decoded frame
-    AVFrame* GetCurrentFrame();
-    const AVFrame* GetCurrentFrame() const;
+    [[nodiscard]] AVFrame* GetCurrentFrame();
+    [[nodiscard]] const AVFrame* GetCurrentFrame() const;
 
     /// Returns the value of current_codec
-    NvdecCommon::VideoCodec GetCurrentCodec() const;
+    [[nodiscard]] NvdecCommon::VideoCodec GetCurrentCodec() const;
 
 private:
     bool initialized{};
diff --git a/src/video_core/command_classes/codecs/h264.cpp b/src/video_core/command_classes/codecs/h264.cpp
index 1a39f7b23..33e063e20 100644
--- a/src/video_core/command_classes/codecs/h264.cpp
+++ b/src/video_core/command_classes/codecs/h264.cpp
@@ -18,17 +18,33 @@
 // OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 //
 
+#include <array>
 #include "common/bit_util.h"
 #include "video_core/command_classes/codecs/h264.h"
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 
 namespace Tegra::Decoder {
+namespace {
+// ZigZag LUTs from libavcodec.
+constexpr std::array<u8, 64> zig_zag_direct{
+    0,  1,  8,  16, 9,  2,  3,  10, 17, 24, 32, 25, 18, 11, 4,  5,  12, 19, 26, 33, 40, 48,
+    41, 34, 27, 20, 13, 6,  7,  14, 21, 28, 35, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23,
+    30, 37, 44, 51, 58, 59, 52, 45, 38, 31, 39, 46, 53, 60, 61, 54, 47, 55, 62, 63,
+};
+
+constexpr std::array<u8, 16> zig_zag_scan{
+    0 + 0 * 4, 1 + 0 * 4, 0 + 1 * 4, 0 + 2 * 4, 1 + 1 * 4, 2 + 0 * 4, 3 + 0 * 4, 2 + 1 * 4,
+    1 + 2 * 4, 0 + 3 * 4, 1 + 3 * 4, 2 + 2 * 4, 3 + 1 * 4, 3 + 2 * 4, 2 + 3 * 4, 3 + 3 * 4,
+};
+} // Anonymous namespace
+
 H264::H264(GPU& gpu_) : gpu(gpu_) {}
 
 H264::~H264() = default;
 
-std::vector<u8>& H264::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state, bool is_first_frame) {
+const std::vector<u8>& H264::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state,
+                                                bool is_first_frame) {
     H264DecoderContext context{};
     gpu.MemoryManager().ReadBlock(state.picture_info_offset, &context, sizeof(H264DecoderContext));
 
@@ -48,7 +64,8 @@ std::vector<u8>& H264::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state, bo
         writer.WriteU(0, 8);
         writer.WriteU(31, 8);
         writer.WriteUe(0);
-        const s32 chroma_format_idc = (context.h264_parameter_set.flags >> 12) & 0x3;
+        const auto chroma_format_idc =
+            static_cast<u32>((context.h264_parameter_set.flags >> 12) & 3);
         writer.WriteUe(chroma_format_idc);
         if (chroma_format_idc == 3) {
             writer.WriteBit(false);
@@ -59,8 +76,8 @@ std::vector<u8>& H264::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state, bo
         writer.WriteBit(false); // QpprimeYZeroTransformBypassFlag
         writer.WriteBit(false); // Scaling matrix present flag
 
-        const s32 order_cnt_type = static_cast<s32>((context.h264_parameter_set.flags >> 14) & 3);
-        writer.WriteUe(static_cast<s32>((context.h264_parameter_set.flags >> 8) & 0xf));
+        const auto order_cnt_type = static_cast<u32>((context.h264_parameter_set.flags >> 14) & 3);
+        writer.WriteUe(static_cast<u32>((context.h264_parameter_set.flags >> 8) & 0xf));
         writer.WriteUe(order_cnt_type);
         if (order_cnt_type == 0) {
             writer.WriteUe(context.h264_parameter_set.log2_max_pic_order_cnt);
@@ -100,7 +117,7 @@ std::vector<u8>& H264::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state, bo
         writer.WriteUe(0);
         writer.WriteUe(0);
 
-        writer.WriteBit(context.h264_parameter_set.entropy_coding_mode_flag);
+        writer.WriteBit(context.h264_parameter_set.entropy_coding_mode_flag != 0);
         writer.WriteBit(false);
         writer.WriteUe(0);
         writer.WriteUe(context.h264_parameter_set.num_refidx_l0_default_active);
@@ -172,8 +189,8 @@ void H264BitWriter::WriteSe(s32 value) {
     WriteExpGolombCodedInt(value);
 }
 
-void H264BitWriter::WriteUe(s32 value) {
-    WriteExpGolombCodedUInt((u32)value);
+void H264BitWriter::WriteUe(u32 value) {
+    WriteExpGolombCodedUInt(value);
 }
 
 void H264BitWriter::End() {
diff --git a/src/video_core/command_classes/codecs/h264.h b/src/video_core/command_classes/codecs/h264.h
index 21752dd90..273449495 100644
--- a/src/video_core/command_classes/codecs/h264.h
+++ b/src/video_core/command_classes/codecs/h264.h
@@ -38,7 +38,7 @@ public:
     /// WriteSe and WriteUe write in the Exp-Golomb-coded syntax
     void WriteU(s32 value, s32 value_sz);
     void WriteSe(s32 value);
-    void WriteUe(s32 value);
+    void WriteUe(u32 value);
 
     /// Finalize the bitstream
     void End();
@@ -51,26 +51,14 @@ public:
     void WriteScalingList(const std::vector<u8>& list, s32 start, s32 count);
 
     /// Return the bitstream as a vector.
-    std::vector<u8>& GetByteArray();
-    const std::vector<u8>& GetByteArray() const;
+    [[nodiscard]] std::vector<u8>& GetByteArray();
+    [[nodiscard]] const std::vector<u8>& GetByteArray() const;
 
 private:
-    // ZigZag LUTs from libavcodec.
-    static constexpr std::array<u8, 64> zig_zag_direct{
-        0,  1,  8,  16, 9,  2,  3,  10, 17, 24, 32, 25, 18, 11, 4,  5,  12, 19, 26, 33, 40, 48,
-        41, 34, 27, 20, 13, 6,  7,  14, 21, 28, 35, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23,
-        30, 37, 44, 51, 58, 59, 52, 45, 38, 31, 39, 46, 53, 60, 61, 54, 47, 55, 62, 63,
-    };
-
-    static constexpr std::array<u8, 16> zig_zag_scan{
-        0 + 0 * 4, 1 + 0 * 4, 0 + 1 * 4, 0 + 2 * 4, 1 + 1 * 4, 2 + 0 * 4, 3 + 0 * 4, 2 + 1 * 4,
-        1 + 2 * 4, 0 + 3 * 4, 1 + 3 * 4, 2 + 2 * 4, 3 + 1 * 4, 3 + 2 * 4, 2 + 3 * 4, 3 + 3 * 4,
-    };
-
     void WriteBits(s32 value, s32 bit_count);
     void WriteExpGolombCodedInt(s32 value);
     void WriteExpGolombCodedUInt(u32 value);
-    s32 GetFreeBufferBits();
+    [[nodiscard]] s32 GetFreeBufferBits();
     void Flush();
 
     s32 buffer_size{8};
@@ -86,8 +74,8 @@ public:
     ~H264();
 
     /// Compose the H264 header of the frame for FFmpeg decoding
-    std::vector<u8>& ComposeFrameHeader(NvdecCommon::NvdecRegisters& state,
-                                        bool is_first_frame = false);
+    [[nodiscard]] const std::vector<u8>& ComposeFrameHeader(NvdecCommon::NvdecRegisters& state,
+                                                            bool is_first_frame = false);
 
 private:
     struct H264ParameterSet {
diff --git a/src/video_core/command_classes/codecs/vp9.cpp b/src/video_core/command_classes/codecs/vp9.cpp
index d205a8f5d..ab44fdc9e 100644
--- a/src/video_core/command_classes/codecs/vp9.cpp
+++ b/src/video_core/command_classes/codecs/vp9.cpp
@@ -9,7 +9,7 @@
 #include "video_core/memory_manager.h"
 
 namespace Tegra::Decoder {
-
+namespace {
 // Default compressed header probabilities once frame context resets
 constexpr Vp9EntropyProbs default_probs{
     .y_mode_prob{
@@ -170,6 +170,89 @@ constexpr Vp9EntropyProbs default_probs{
     .high_precision{128, 128},
 };
 
+constexpr std::array<s32, 256> norm_lut{
+    0, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+};
+
+constexpr std::array<s32, 254> map_lut{
+    20,  21,  22,  23,  24,  25,  0,   26,  27,  28,  29,  30,  31,  32,  33,  34,  35,  36,  37,
+    1,   38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49,  2,   50,  51,  52,  53,  54,
+    55,  56,  57,  58,  59,  60,  61,  3,   62,  63,  64,  65,  66,  67,  68,  69,  70,  71,  72,
+    73,  4,   74,  75,  76,  77,  78,  79,  80,  81,  82,  83,  84,  85,  5,   86,  87,  88,  89,
+    90,  91,  92,  93,  94,  95,  96,  97,  6,   98,  99,  100, 101, 102, 103, 104, 105, 106, 107,
+    108, 109, 7,   110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 8,   122, 123, 124,
+    125, 126, 127, 128, 129, 130, 131, 132, 133, 9,   134, 135, 136, 137, 138, 139, 140, 141, 142,
+    143, 144, 145, 10,  146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 11,  158, 159,
+    160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 12,  170, 171, 172, 173, 174, 175, 176, 177,
+    178, 179, 180, 181, 13,  182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 14,  194,
+    195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 15,  206, 207, 208, 209, 210, 211, 212,
+    213, 214, 215, 216, 217, 16,  218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 17,
+    230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 18,  242, 243, 244, 245, 246, 247,
+    248, 249, 250, 251, 252, 253, 19,
+};
+
+// 6.2.14 Tile size calculation
+
+[[nodiscard]] s32 CalcMinLog2TileCols(s32 frame_width) {
+    const s32 sb64_cols = (frame_width + 63) / 64;
+    s32 min_log2 = 0;
+
+    while ((64 << min_log2) < sb64_cols) {
+        min_log2++;
+    }
+
+    return min_log2;
+}
+
+[[nodiscard]] s32 CalcMaxLog2TileCols(s32 frame_width) {
+    const s32 sb64_cols = (frame_width + 63) / 64;
+    s32 max_log2 = 1;
+
+    while ((sb64_cols >> max_log2) >= 4) {
+        max_log2++;
+    }
+
+    return max_log2 - 1;
+}
+
+// Recenters probability. Based on section 6.3.6 of VP9 Specification
+[[nodiscard]] s32 RecenterNonNeg(s32 new_prob, s32 old_prob) {
+    if (new_prob > old_prob * 2) {
+        return new_prob;
+    }
+
+    if (new_prob >= old_prob) {
+        return (new_prob - old_prob) * 2;
+    }
+
+    return (old_prob - new_prob) * 2 - 1;
+}
+
+// Adjusts old_prob depending on new_prob. Based on section 6.3.5 of VP9 Specification
+[[nodiscard]] s32 RemapProbability(s32 new_prob, s32 old_prob) {
+    new_prob--;
+    old_prob--;
+
+    std::size_t index{};
+
+    if (old_prob * 2 <= 0xff) {
+        index = static_cast<std::size_t>(std::max(0, RecenterNonNeg(new_prob, old_prob) - 1));
+    } else {
+        index = static_cast<std::size_t>(
+            std::max(0, RecenterNonNeg(0xff - 1 - new_prob, 0xff - 1 - old_prob) - 1));
+    }
+
+    return map_lut[index];
+}
+} // Anonymous namespace
+
 VP9::VP9(GPU& gpu) : gpu(gpu) {}
 
 VP9::~VP9() = default;
@@ -207,32 +290,6 @@ void VP9::WriteProbabilityDelta(VpxRangeEncoder& writer, u8 new_prob, u8 old_pro
     EncodeTermSubExp(writer, delta);
 }
 
-s32 VP9::RemapProbability(s32 new_prob, s32 old_prob) {
-    new_prob--;
-    old_prob--;
-
-    std::size_t index{};
-
-    if (old_prob * 2 <= 0xff) {
-        index = static_cast<std::size_t>(std::max(0, RecenterNonNeg(new_prob, old_prob) - 1));
-    } else {
-        index = static_cast<std::size_t>(
-            std::max(0, RecenterNonNeg(0xff - 1 - new_prob, 0xff - 1 - old_prob) - 1));
-    }
-
-    return map_lut[index];
-}
-
-s32 VP9::RecenterNonNeg(s32 new_prob, s32 old_prob) {
-    if (new_prob > old_prob * 2) {
-        return new_prob;
-    } else if (new_prob >= old_prob) {
-        return (new_prob - old_prob) * 2;
-    } else {
-        return (old_prob - new_prob) * 2 - 1;
-    }
-}
-
 void VP9::EncodeTermSubExp(VpxRangeEncoder& writer, s32 value) {
     if (WriteLessThan(writer, value, 16)) {
         writer.Write(value, 4);
@@ -332,28 +389,6 @@ void VP9::WriteMvProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_
     }
 }
 
-s32 VP9::CalcMinLog2TileCols(s32 frame_width) {
-    const s32 sb64_cols = (frame_width + 63) / 64;
-    s32 min_log2 = 0;
-
-    while ((64 << min_log2) < sb64_cols) {
-        min_log2++;
-    }
-
-    return min_log2;
-}
-
-s32 VP9::CalcMaxLog2TileCols(s32 frameWidth) {
-    const s32 sb64_cols = (frameWidth + 63) / 64;
-    s32 max_log2 = 1;
-
-    while ((sb64_cols >> max_log2) >= 4) {
-        max_log2++;
-    }
-
-    return max_log2 - 1;
-}
-
 Vp9PictureInfo VP9::GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state) {
     PictureInfo picture_info{};
     gpu.MemoryManager().ReadBlock(state.picture_info_offset, &picture_info, sizeof(PictureInfo));
@@ -379,14 +414,14 @@ Vp9FrameContainer VP9::GetCurrentFrame(const NvdecCommon::NvdecRegisters& state)
     Vp9FrameContainer frame{};
     {
         gpu.SyncGuestHost();
-        frame.info = std::move(GetVp9PictureInfo(state));
+        frame.info = GetVp9PictureInfo(state);
 
         frame.bit_stream.resize(frame.info.bitstream_size);
         gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, frame.bit_stream.data(),
                                       frame.info.bitstream_size);
     }
     // Buffer two frames, saving the last show frame info
-    if (next_next_frame.bit_stream.size() != 0) {
+    if (!next_next_frame.bit_stream.empty()) {
         Vp9FrameContainer temp{
             .info = frame.info,
             .bit_stream = frame.bit_stream,
@@ -396,15 +431,15 @@ Vp9FrameContainer VP9::GetCurrentFrame(const NvdecCommon::NvdecRegisters& state)
         frame.bit_stream = next_next_frame.bit_stream;
         next_next_frame = std::move(temp);
 
-        if (next_frame.bit_stream.size() != 0) {
-            Vp9FrameContainer temp{
+        if (!next_frame.bit_stream.empty()) {
+            Vp9FrameContainer temp2{
                 .info = frame.info,
                 .bit_stream = frame.bit_stream,
             };
             next_frame.info.show_frame = frame.info.last_frame_shown;
             frame.info = next_frame.info;
             frame.bit_stream = next_frame.bit_stream;
-            next_frame = std::move(temp);
+            next_frame = std::move(temp2);
         } else {
             next_frame.info = frame.info;
             next_frame.bit_stream = frame.bit_stream;
@@ -605,12 +640,6 @@ std::vector<u8> VP9::ComposeCompressedHeader() {
 
     writer.End();
     return writer.GetBuffer();
-
-    const auto writer_bytearray = writer.GetBuffer();
-
-    std::vector<u8> compressed_header(writer_bytearray.size());
-    std::memcpy(compressed_header.data(), writer_bytearray.data(), writer_bytearray.size());
-    return compressed_header;
 }
 
 VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
@@ -648,7 +677,6 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
         current_frame_info.intra_only = true;
 
     } else {
-        std::array<s32, 3> ref_frame_index;
 
         if (!current_frame_info.show_frame) {
             uncomp_writer.WriteBit(current_frame_info.intra_only);
@@ -663,9 +691,9 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
         }
 
         // Last, Golden, Altref frames
-        ref_frame_index = std::array<s32, 3>{0, 1, 2};
+        std::array<s32, 3> ref_frame_index{0, 1, 2};
 
-        // set when next frame is hidden
+        // Set when next frame is hidden
         // altref and golden references are swapped
         if (swap_next_golden) {
             ref_frame_index = std::array<s32, 3>{0, 2, 1};
@@ -754,17 +782,19 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
         for (std::size_t index = 0; index < current_frame_info.ref_deltas.size(); index++) {
             const s8 old_deltas = loop_filter_ref_deltas[index];
             const s8 new_deltas = current_frame_info.ref_deltas[index];
+            const bool differing_delta = old_deltas != new_deltas;
 
-            loop_filter_delta_update |=
-                (update_loop_filter_ref_deltas[index] = old_deltas != new_deltas);
+            update_loop_filter_ref_deltas[index] = differing_delta;
+            loop_filter_delta_update |= differing_delta;
         }
 
         for (std::size_t index = 0; index < current_frame_info.mode_deltas.size(); index++) {
             const s8 old_deltas = loop_filter_mode_deltas[index];
             const s8 new_deltas = current_frame_info.mode_deltas[index];
+            const bool differing_delta = old_deltas != new_deltas;
 
-            loop_filter_delta_update |=
-                (update_loop_filter_mode_deltas[index] = old_deltas != new_deltas);
+            update_loop_filter_mode_deltas[index] = differing_delta;
+            loop_filter_delta_update |= differing_delta;
         }
 
         uncomp_writer.WriteBit(loop_filter_delta_update);
@@ -824,12 +854,12 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
     return uncomp_writer;
 }
 
-std::vector<u8>& VP9::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state) {
+const std::vector<u8>& VP9::ComposeFrameHeader(NvdecCommon::NvdecRegisters& state) {
     std::vector<u8> bitstream;
     {
         Vp9FrameContainer curr_frame = GetCurrentFrame(state);
         current_frame_info = curr_frame.info;
-        bitstream = curr_frame.bit_stream;
+        bitstream = std::move(curr_frame.bit_stream);
     }
 
     // The uncompressed header routine sets PrevProb parameters needed for the compressed header
diff --git a/src/video_core/command_classes/codecs/vp9.h b/src/video_core/command_classes/codecs/vp9.h
index 748e11bae..e2504512c 100644
--- a/src/video_core/command_classes/codecs/vp9.h
+++ b/src/video_core/command_classes/codecs/vp9.h
@@ -4,9 +4,9 @@
 
 #pragma once
 
-#include <unordered_map>
+#include <array>
 #include <vector>
-#include "common/common_funcs.h"
+
 #include "common/common_types.h"
 #include "common/stream.h"
 #include "video_core/command_classes/codecs/vp9_types.h"
@@ -25,6 +25,12 @@ public:
     VpxRangeEncoder();
     ~VpxRangeEncoder();
 
+    VpxRangeEncoder(const VpxRangeEncoder&) = delete;
+    VpxRangeEncoder& operator=(const VpxRangeEncoder&) = delete;
+
+    VpxRangeEncoder(VpxRangeEncoder&&) = default;
+    VpxRangeEncoder& operator=(VpxRangeEncoder&&) = default;
+
     /// Writes the rightmost value_size bits from value into the stream
     void Write(s32 value, s32 value_size);
 
@@ -37,11 +43,11 @@ public:
     /// Signal the end of the bitstream
     void End();
 
-    std::vector<u8>& GetBuffer() {
+    [[nodiscard]] std::vector<u8>& GetBuffer() {
         return base_stream.GetBuffer();
     }
 
-    const std::vector<u8>& GetBuffer() const {
+    [[nodiscard]] const std::vector<u8>& GetBuffer() const {
         return base_stream.GetBuffer();
     }
 
@@ -52,17 +58,6 @@ private:
     u32 range{0xff};
     s32 count{-24};
     s32 half_probability{128};
-    static constexpr std::array<s32, 256> norm_lut{
-        0, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
-        3, 3, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
-        2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
-        1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
-        1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-    };
 };
 
 class VpxBitStreamWriter {
@@ -70,6 +65,12 @@ public:
     VpxBitStreamWriter();
     ~VpxBitStreamWriter();
 
+    VpxBitStreamWriter(const VpxBitStreamWriter&) = delete;
+    VpxBitStreamWriter& operator=(const VpxBitStreamWriter&) = delete;
+
+    VpxBitStreamWriter(VpxBitStreamWriter&&) = default;
+    VpxBitStreamWriter& operator=(VpxBitStreamWriter&&) = default;
+
     /// Write an unsigned integer value
     void WriteU(u32 value, u32 value_size);
 
@@ -86,10 +87,10 @@ public:
     void Flush();
 
     /// Returns byte_array
-    std::vector<u8>& GetByteArray();
+    [[nodiscard]] std::vector<u8>& GetByteArray();
 
     /// Returns const byte_array
-    const std::vector<u8>& GetByteArray() const;
+    [[nodiscard]] const std::vector<u8>& GetByteArray() const;
 
 private:
     /// Write bit_count bits from value into buffer
@@ -110,12 +111,18 @@ public:
     explicit VP9(GPU& gpu);
     ~VP9();
 
+    VP9(const VP9&) = delete;
+    VP9& operator=(const VP9&) = delete;
+
+    VP9(VP9&&) = default;
+    VP9& operator=(VP9&&) = delete;
+
     /// Composes the VP9 frame from the GPU state information. Based on the official VP9 spec
     /// documentation
-    std::vector<u8>& ComposeFrameHeader(NvdecCommon::NvdecRegisters& state);
+    [[nodiscard]] const std::vector<u8>& ComposeFrameHeader(NvdecCommon::NvdecRegisters& state);
 
     /// Returns true if the most recent frame was a hidden frame.
-    bool WasFrameHidden() const {
+    [[nodiscard]] bool WasFrameHidden() const {
         return hidden;
     }
 
@@ -132,12 +139,6 @@ private:
     /// Generates compressed header probability deltas in the bitstream writer
     void WriteProbabilityDelta(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob);
 
-    /// Adjusts old_prob depending on new_prob. Based on section 6.3.5 of VP9 Specification
-    s32 RemapProbability(s32 new_prob, s32 old_prob);
-
-    /// Recenters probability. Based on section 6.3.6 of VP9 Specification
-    s32 RecenterNonNeg(s32 new_prob, s32 old_prob);
-
     /// Inverse of 6.3.4 Decode term subexp
     void EncodeTermSubExp(VpxRangeEncoder& writer, s32 value);
 
@@ -157,22 +158,18 @@ private:
     /// Write motion vector probability updates. 6.3.17 in the spec
     void WriteMvProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob);
 
-    /// 6.2.14 Tile size calculation
-    s32 CalcMinLog2TileCols(s32 frame_width);
-    s32 CalcMaxLog2TileCols(s32 frame_width);
-
     /// Returns VP9 information from NVDEC provided offset and size
-    Vp9PictureInfo GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state);
+    [[nodiscard]] Vp9PictureInfo GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state);
 
     /// Read and convert NVDEC provided entropy probs to Vp9EntropyProbs struct
     void InsertEntropy(u64 offset, Vp9EntropyProbs& dst);
 
     /// Returns frame to be decoded after buffering
-    Vp9FrameContainer GetCurrentFrame(const NvdecCommon::NvdecRegisters& state);
+    [[nodiscard]] Vp9FrameContainer GetCurrentFrame(const NvdecCommon::NvdecRegisters& state);
 
     /// Use NVDEC providied information to compose the headers for the current frame
-    std::vector<u8> ComposeCompressedHeader();
-    VpxBitStreamWriter ComposeUncompressedHeader();
+    [[nodiscard]] std::vector<u8> ComposeCompressedHeader();
+    [[nodiscard]] VpxBitStreamWriter ComposeUncompressedHeader();
 
     GPU& gpu;
     std::vector<u8> frame;
@@ -180,7 +177,7 @@ private:
     std::array<s8, 4> loop_filter_ref_deltas{};
     std::array<s8, 2> loop_filter_mode_deltas{};
 
-    bool hidden;
+    bool hidden = false;
     s64 current_frame_number = -2; // since we buffer 2 frames
     s32 grace_period = 6;          // frame offsets need to stabilize
     std::array<FrameContexts, 4> frame_ctxs{};
@@ -193,23 +190,6 @@ private:
 
     s32 diff_update_probability = 252;
     s32 frame_sync_code = 0x498342;
-    static constexpr std::array<s32, 254> map_lut = {
-        20,  21,  22,  23,  24,  25,  0,   26,  27,  28,  29,  30,  31,  32,  33,  34,  35,
-        36,  37,  1,   38,  39,  40,  41,  42,  43,  44,  45,  46,  47,  48,  49,  2,   50,
-        51,  52,  53,  54,  55,  56,  57,  58,  59,  60,  61,  3,   62,  63,  64,  65,  66,
-        67,  68,  69,  70,  71,  72,  73,  4,   74,  75,  76,  77,  78,  79,  80,  81,  82,
-        83,  84,  85,  5,   86,  87,  88,  89,  90,  91,  92,  93,  94,  95,  96,  97,  6,
-        98,  99,  100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 7,   110, 111, 112, 113,
-        114, 115, 116, 117, 118, 119, 120, 121, 8,   122, 123, 124, 125, 126, 127, 128, 129,
-        130, 131, 132, 133, 9,   134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145,
-        10,  146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 11,  158, 159, 160,
-        161, 162, 163, 164, 165, 166, 167, 168, 169, 12,  170, 171, 172, 173, 174, 175, 176,
-        177, 178, 179, 180, 181, 13,  182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192,
-        193, 14,  194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 15,  206, 207,
-        208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 16,  218, 219, 220, 221, 222, 223,
-        224, 225, 226, 227, 228, 229, 17,  230, 231, 232, 233, 234, 235, 236, 237, 238, 239,
-        240, 241, 18,  242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 19,
-    };
 };
 
 } // namespace Decoder
diff --git a/src/video_core/command_classes/codecs/vp9_types.h b/src/video_core/command_classes/codecs/vp9_types.h
index 8688fdac0..4f0b05d22 100644
--- a/src/video_core/command_classes/codecs/vp9_types.h
+++ b/src/video_core/command_classes/codecs/vp9_types.h
@@ -4,13 +4,11 @@
 
 #pragma once
 
-#include <algorithm>
-#include <list>
+#include <array>
+#include <cstring>
 #include <vector>
-#include "common/cityhash.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
-#include "video_core/command_classes/nvdec_common.h"
 
 namespace Tegra {
 class GPU;
@@ -233,9 +231,8 @@ struct PictureInfo {
     u32 surface_params{};
     INSERT_PADDING_WORDS(3);
 
-    Vp9PictureInfo Convert() const {
-
-        return Vp9PictureInfo{
+    [[nodiscard]] Vp9PictureInfo Convert() const {
+        return {
             .is_key_frame = (vp9_flags & FrameFlags::IsKeyFrame) != 0,
             .intra_only = (vp9_flags & FrameFlags::IntraOnly) != 0,
             .last_frame_was_key = (vp9_flags & FrameFlags::LastFrameIsKeyFrame) != 0,
diff --git a/src/video_core/command_classes/host1x.cpp b/src/video_core/command_classes/host1x.cpp
index a5234ee47..c4dd4881a 100644
--- a/src/video_core/command_classes/host1x.cpp
+++ b/src/video_core/command_classes/host1x.cpp
@@ -15,7 +15,7 @@ void Tegra::Host1x::StateWrite(u32 offset, u32 arguments) {
     std::memcpy(state_offset, &arguments, sizeof(u32));
 }
 
-void Tegra::Host1x::ProcessMethod(Host1x::Method method, const std::vector<u32>& arguments) {
+void Tegra::Host1x::ProcessMethod(Method method, const std::vector<u32>& arguments) {
     StateWrite(static_cast<u32>(method), arguments[0]);
     switch (method) {
     case Method::WaitSyncpt:
diff --git a/src/video_core/command_classes/host1x.h b/src/video_core/command_classes/host1x.h
index 501a5ed2e..013eaa0c1 100644
--- a/src/video_core/command_classes/host1x.h
+++ b/src/video_core/command_classes/host1x.h
@@ -61,7 +61,7 @@ public:
     ~Host1x();
 
     /// Writes the method into the state, Invoke Execute() if encountered
-    void ProcessMethod(Host1x::Method method, const std::vector<u32>& arguments);
+    void ProcessMethod(Method method, const std::vector<u32>& arguments);
 
 private:
     /// For Host1x, execute is waiting on a syncpoint previously written into the state
diff --git a/src/video_core/command_classes/nvdec.cpp b/src/video_core/command_classes/nvdec.cpp
index ede9466eb..8ca7a7b06 100644
--- a/src/video_core/command_classes/nvdec.cpp
+++ b/src/video_core/command_classes/nvdec.cpp
@@ -2,13 +2,9 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <bitset>
 #include "common/assert.h"
-#include "common/bit_util.h"
-#include "core/memory.h"
 #include "video_core/command_classes/nvdec.h"
 #include "video_core/gpu.h"
-#include "video_core/memory_manager.h"
 
 namespace Tegra {
 
@@ -16,7 +12,7 @@ Nvdec::Nvdec(GPU& gpu_) : gpu(gpu_), codec(std::make_unique<Codec>(gpu)) {}
 
 Nvdec::~Nvdec() = default;
 
-void Nvdec::ProcessMethod(Nvdec::Method method, const std::vector<u32>& arguments) {
+void Nvdec::ProcessMethod(Method method, const std::vector<u32>& arguments) {
     if (method == Method::SetVideoCodec) {
         codec->StateWrite(static_cast<u32>(method), arguments[0]);
     } else {
diff --git a/src/video_core/command_classes/nvdec.h b/src/video_core/command_classes/nvdec.h
index c1a9d843e..eec4443f9 100644
--- a/src/video_core/command_classes/nvdec.h
+++ b/src/video_core/command_classes/nvdec.h
@@ -4,8 +4,8 @@
 
 #pragma once
 
+#include <memory>
 #include <vector>
-#include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/command_classes/codecs/codec.h"
 
@@ -23,17 +23,17 @@ public:
     ~Nvdec();
 
     /// Writes the method into the state, Invoke Execute() if encountered
-    void ProcessMethod(Nvdec::Method method, const std::vector<u32>& arguments);
+    void ProcessMethod(Method method, const std::vector<u32>& arguments);
 
     /// Return most recently decoded frame
-    AVFrame* GetFrame();
-    const AVFrame* GetFrame() const;
+    [[nodiscard]] AVFrame* GetFrame();
+    [[nodiscard]] const AVFrame* GetFrame() const;
 
 private:
     /// Invoke codec to decode a frame
     void Execute();
 
     GPU& gpu;
-    std::unique_ptr<Tegra::Codec> codec;
+    std::unique_ptr<Codec> codec;
 };
 } // namespace Tegra
diff --git a/src/video_core/command_classes/sync_manager.cpp b/src/video_core/command_classes/sync_manager.cpp
index a0ab44855..19dc9e0ab 100644
--- a/src/video_core/command_classes/sync_manager.cpp
+++ b/src/video_core/command_classes/sync_manager.cpp
@@ -27,22 +27,22 @@ SyncptIncrManager::SyncptIncrManager(GPU& gpu_) : gpu(gpu_) {}
 SyncptIncrManager::~SyncptIncrManager() = default;
 
 void SyncptIncrManager::Increment(u32 id) {
-    increments.push_back(SyncptIncr{0, id, true});
+    increments.emplace_back(0, 0, id, true);
     IncrementAllDone();
 }
 
 u32 SyncptIncrManager::IncrementWhenDone(u32 class_id, u32 id) {
     const u32 handle = current_id++;
-    increments.push_back(SyncptIncr{handle, class_id, id});
+    increments.emplace_back(handle, class_id, id);
     return handle;
 }
 
 void SyncptIncrManager::SignalDone(u32 handle) {
-    auto done_incr = std::find_if(increments.begin(), increments.end(),
-                                  [handle](SyncptIncr incr) { return incr.id == handle; });
-    if (done_incr != increments.end()) {
-        const SyncptIncr incr = *done_incr;
-        *done_incr = SyncptIncr{incr.id, incr.class_id, incr.syncpt_id, true};
+    const auto done_incr =
+        std::find_if(increments.begin(), increments.end(),
+                     [handle](const SyncptIncr& incr) { return incr.id == handle; });
+    if (done_incr != increments.cend()) {
+        done_incr->complete = true;
     }
     IncrementAllDone();
 }
diff --git a/src/video_core/command_classes/sync_manager.h b/src/video_core/command_classes/sync_manager.h
index 353b67573..2c321ec58 100644
--- a/src/video_core/command_classes/sync_manager.h
+++ b/src/video_core/command_classes/sync_manager.h
@@ -32,8 +32,8 @@ struct SyncptIncr {
     u32 syncpt_id;
     bool complete;
 
-    SyncptIncr(u32 id, u32 syncpt_id_, u32 class_id_, bool done = false)
-        : id(id), class_id(class_id_), syncpt_id(syncpt_id_), complete(done) {}
+    SyncptIncr(u32 id_, u32 class_id_, u32 syncpt_id_, bool done = false)
+        : id(id_), class_id(class_id_), syncpt_id(syncpt_id_), complete(done) {}
 };
 
 class SyncptIncrManager {
diff --git a/src/video_core/command_classes/vic.cpp b/src/video_core/command_classes/vic.cpp
index 66e15a1a8..5b52da277 100644
--- a/src/video_core/command_classes/vic.cpp
+++ b/src/video_core/command_classes/vic.cpp
@@ -26,7 +26,7 @@ void Vic::VicStateWrite(u32 offset, u32 arguments) {
     std::memcpy(state_offset, &arguments, sizeof(u32));
 }
 
-void Vic::ProcessMethod(Vic::Method method, const std::vector<u32>& arguments) {
+void Vic::ProcessMethod(Method method, const std::vector<u32>& arguments) {
     LOG_DEBUG(HW_GPU, "Vic method 0x{:X}", static_cast<u32>(method));
     VicStateWrite(static_cast<u32>(method), arguments[0]);
     const u64 arg = static_cast<u64>(arguments[0]) << 8;
diff --git a/src/video_core/command_classes/vic.h b/src/video_core/command_classes/vic.h
index dd0a2aed8..8c4e284a1 100644
--- a/src/video_core/command_classes/vic.h
+++ b/src/video_core/command_classes/vic.h
@@ -63,11 +63,11 @@ public:
         SetOutputSurfaceChromaVOffset = 0x1ca
     };
 
-    explicit Vic(GPU& gpu, std::shared_ptr<Tegra::Nvdec> nvdec_processor);
+    explicit Vic(GPU& gpu, std::shared_ptr<Nvdec> nvdec_processor);
     ~Vic();
 
     /// Write to the device state.
-    void ProcessMethod(Vic::Method method, const std::vector<u32>& arguments);
+    void ProcessMethod(Method method, const std::vector<u32>& arguments);
 
 private:
     void Execute();
diff --git a/src/video_core/dma_pusher.cpp b/src/video_core/dma_pusher.cpp
index f2f96ac33..105b85a92 100644
--- a/src/video_core/dma_pusher.cpp
+++ b/src/video_core/dma_pusher.cpp
@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/cityhash.h"
 #include "common/microprofile.h"
 #include "core/core.h"
 #include "core/memory.h"
@@ -12,6 +13,20 @@
 
 namespace Tegra {
 
+void CommandList::RefreshIntegrityChecks(GPU& gpu) {
+    command_list_hashes.resize(command_lists.size());
+
+    for (std::size_t index = 0; index < command_lists.size(); ++index) {
+        const CommandListHeader command_list_header = command_lists[index];
+        std::vector<CommandHeader> command_headers(command_list_header.size);
+        gpu.MemoryManager().ReadBlockUnsafe(command_list_header.addr, command_headers.data(),
+                                            command_list_header.size * sizeof(u32));
+        command_list_hashes[index] =
+            Common::CityHash64(reinterpret_cast<char*>(command_headers.data()),
+                               command_list_header.size * sizeof(u32));
+    }
+}
+
 DmaPusher::DmaPusher(Core::System& system, GPU& gpu) : gpu{gpu}, system{system} {}
 
 DmaPusher::~DmaPusher() = default;
@@ -45,32 +60,51 @@ bool DmaPusher::Step() {
         return false;
     }
 
-    const CommandList& command_list{dma_pushbuffer.front()};
-    ASSERT_OR_EXECUTE(!command_list.empty(), {
-        // Somehow the command_list is empty, in order to avoid a crash
-        // We ignore it and assume its size is 0.
-        dma_pushbuffer.pop();
-        dma_pushbuffer_subindex = 0;
-        return true;
-    });
-    const CommandListHeader command_list_header{command_list[dma_pushbuffer_subindex++]};
-    const GPUVAddr dma_get = command_list_header.addr;
-
-    if (dma_pushbuffer_subindex >= command_list.size()) {
-        // We've gone through the current list, remove it from the queue
-        dma_pushbuffer.pop();
-        dma_pushbuffer_subindex = 0;
-    }
+    CommandList& command_list{dma_pushbuffer.front()};
 
-    if (command_list_header.size == 0) {
-        return true;
-    }
+    ASSERT_OR_EXECUTE(
+        command_list.command_lists.size() || command_list.prefetch_command_list.size(), {
+            // Somehow the command_list is empty, in order to avoid a crash
+            // We ignore it and assume its size is 0.
+            dma_pushbuffer.pop();
+            dma_pushbuffer_subindex = 0;
+            return true;
+        });
 
-    // Push buffer non-empty, read a word
-    command_headers.resize(command_list_header.size);
-    gpu.MemoryManager().ReadBlockUnsafe(dma_get, command_headers.data(),
-                                        command_list_header.size * sizeof(u32));
+    if (command_list.prefetch_command_list.size()) {
+        // Prefetched command list from nvdrv, used for things like synchronization
+        command_headers = std::move(command_list.prefetch_command_list);
+        dma_pushbuffer.pop();
+    } else {
+        const CommandListHeader command_list_header{
+            command_list.command_lists[dma_pushbuffer_subindex]};
+        const u64 next_hash = command_list.command_list_hashes[dma_pushbuffer_subindex++];
+        const GPUVAddr dma_get = command_list_header.addr;
+
+        if (dma_pushbuffer_subindex >= command_list.command_lists.size()) {
+            // We've gone through the current list, remove it from the queue
+            dma_pushbuffer.pop();
+            dma_pushbuffer_subindex = 0;
+        }
 
+        if (command_list_header.size == 0) {
+            return true;
+        }
+
+        // Push buffer non-empty, read a word
+        command_headers.resize(command_list_header.size);
+        gpu.MemoryManager().ReadBlockUnsafe(dma_get, command_headers.data(),
+                                            command_list_header.size * sizeof(u32));
+
+        // Integrity check
+        const u64 new_hash = Common::CityHash64(reinterpret_cast<char*>(command_headers.data()),
+                                                command_list_header.size * sizeof(u32));
+        if (new_hash != next_hash) {
+            LOG_CRITICAL(HW_GPU, "CommandList at addr=0x{:X} is corrupt, skipping!", dma_get);
+            dma_pushbuffer.pop();
+            return true;
+        }
+    }
     for (std::size_t index = 0; index < command_headers.size();) {
         const CommandHeader& command_header = command_headers[index];
 
diff --git a/src/video_core/dma_pusher.h b/src/video_core/dma_pusher.h
index efa90d170..8496ba2da 100644
--- a/src/video_core/dma_pusher.h
+++ b/src/video_core/dma_pusher.h
@@ -27,6 +27,31 @@ enum class SubmissionMode : u32 {
     IncreaseOnce = 5
 };
 
+// Note that, traditionally, methods are treated as 4-byte addressable locations, and hence
+// their numbers are written down multiplied by 4 in Docs. Here we are not multiply by 4.
+// So the values you see in docs might be multiplied by 4.
+enum class BufferMethods : u32 {
+    BindObject = 0x0,
+    Nop = 0x2,
+    SemaphoreAddressHigh = 0x4,
+    SemaphoreAddressLow = 0x5,
+    SemaphoreSequence = 0x6,
+    SemaphoreTrigger = 0x7,
+    NotifyIntr = 0x8,
+    WrcacheFlush = 0x9,
+    Unk28 = 0xA,
+    UnkCacheFlush = 0xB,
+    RefCnt = 0x14,
+    SemaphoreAcquire = 0x1A,
+    SemaphoreRelease = 0x1B,
+    FenceValue = 0x1C,
+    FenceAction = 0x1D,
+    WaitForInterrupt = 0x1E,
+    Unk7c = 0x1F,
+    Yield = 0x20,
+    NonPullerMethods = 0x40,
+};
+
 struct CommandListHeader {
     union {
         u64 raw;
@@ -49,9 +74,29 @@ union CommandHeader {
 static_assert(std::is_standard_layout_v<CommandHeader>, "CommandHeader is not standard layout");
 static_assert(sizeof(CommandHeader) == sizeof(u32), "CommandHeader has incorrect size!");
 
+static constexpr CommandHeader BuildCommandHeader(BufferMethods method, u32 arg_count,
+                                                  SubmissionMode mode) {
+    CommandHeader result{};
+    result.method.Assign(static_cast<u32>(method));
+    result.arg_count.Assign(arg_count);
+    result.mode.Assign(mode);
+    return result;
+}
+
 class GPU;
 
-using CommandList = std::vector<Tegra::CommandListHeader>;
+struct CommandList final {
+    CommandList() = default;
+    explicit CommandList(std::size_t size) : command_lists(size) {}
+    explicit CommandList(std::vector<Tegra::CommandHeader>&& prefetch_command_list)
+        : prefetch_command_list{std::move(prefetch_command_list)} {}
+
+    void RefreshIntegrityChecks(GPU& gpu);
+
+    std::vector<Tegra::CommandListHeader> command_lists;
+    std::vector<u64> command_list_hashes;
+    std::vector<Tegra::CommandHeader> prefetch_command_list;
+};
 
 /**
  * The DmaPusher class implements DMA submission to FIFOs, providing an area of memory that the
@@ -60,7 +105,7 @@ using CommandList = std::vector<Tegra::CommandListHeader>;
  * See https://envytools.readthedocs.io/en/latest/hw/fifo/dma-pusher.html#fifo-dma-pusher for
  * details on this implementation.
  */
-class DmaPusher {
+class DmaPusher final {
 public:
     explicit DmaPusher(Core::System& system, GPU& gpu);
     ~DmaPusher();
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index d374b73cf..a3c05d1b0 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -1893,6 +1893,7 @@ public:
         ICMP_IMM,
         FCMP_RR,
         FCMP_RC,
+        FCMP_IMMR,
         MUFU,  // Multi-Function Operator
         RRO_C, // Range Reduction Operator
         RRO_R,
@@ -2205,6 +2206,7 @@ private:
             INST("0111110-0-------", Id::HSET2_IMM, Type::HalfSet, "HSET2_IMM"),
             INST("010110111010----", Id::FCMP_RR, Type::Arithmetic, "FCMP_RR"),
             INST("010010111010----", Id::FCMP_RC, Type::Arithmetic, "FCMP_RC"),
+            INST("0011011-1010----", Id::FCMP_IMMR, Type::Arithmetic, "FCMP_IMMR"),
             INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"),
             INST("0100110010010---", Id::RRO_C, Type::Arithmetic, "RRO_C"),
             INST("0101110010010---", Id::RRO_R, Type::Arithmetic, "RRO_R"),
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 171f78183..ebd149c3a 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -194,30 +194,6 @@ void GPU::SyncGuestHost() {
 void GPU::OnCommandListEnd() {
     renderer->Rasterizer().ReleaseFences();
 }
-// Note that, traditionally, methods are treated as 4-byte addressable locations, and hence
-// their numbers are written down multiplied by 4 in Docs. Here we are not multiply by 4.
-// So the values you see in docs might be multiplied by 4.
-enum class BufferMethods {
-    BindObject = 0x0,
-    Nop = 0x2,
-    SemaphoreAddressHigh = 0x4,
-    SemaphoreAddressLow = 0x5,
-    SemaphoreSequence = 0x6,
-    SemaphoreTrigger = 0x7,
-    NotifyIntr = 0x8,
-    WrcacheFlush = 0x9,
-    Unk28 = 0xA,
-    UnkCacheFlush = 0xB,
-    RefCnt = 0x14,
-    SemaphoreAcquire = 0x1A,
-    SemaphoreRelease = 0x1B,
-    FenceValue = 0x1C,
-    FenceAction = 0x1D,
-    Unk78 = 0x1E,
-    Unk7c = 0x1F,
-    Yield = 0x20,
-    NonPullerMethods = 0x40,
-};
 
 enum class GpuSemaphoreOperation {
     AcquireEqual = 0x1,
@@ -277,7 +253,12 @@ void GPU::CallPullerMethod(const MethodCall& method_call) {
     case BufferMethods::UnkCacheFlush:
     case BufferMethods::WrcacheFlush:
     case BufferMethods::FenceValue:
+        break;
     case BufferMethods::FenceAction:
+        ProcessFenceActionMethod();
+        break;
+    case BufferMethods::WaitForInterrupt:
+        ProcessWaitForInterruptMethod();
         break;
     case BufferMethods::SemaphoreTrigger: {
         ProcessSemaphoreTriggerMethod();
@@ -391,6 +372,25 @@ void GPU::ProcessBindMethod(const MethodCall& method_call) {
     }
 }
 
+void GPU::ProcessFenceActionMethod() {
+    switch (regs.fence_action.op) {
+    case FenceOperation::Acquire:
+        WaitFence(regs.fence_action.syncpoint_id, regs.fence_value);
+        break;
+    case FenceOperation::Increment:
+        IncrementSyncPoint(regs.fence_action.syncpoint_id);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented operation {}",
+                          static_cast<u32>(regs.fence_action.op.Value()));
+    }
+}
+
+void GPU::ProcessWaitForInterruptMethod() {
+    // TODO(bunnei) ImplementMe
+    LOG_WARNING(HW_GPU, "(STUBBED) called");
+}
+
 void GPU::ProcessSemaphoreTriggerMethod() {
     const auto semaphoreOperationMask = 0xF;
     const auto op =
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index b8c613b11..5444b49f3 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -263,6 +263,24 @@ public:
         return use_nvdec;
     }
 
+    enum class FenceOperation : u32 {
+        Acquire = 0,
+        Increment = 1,
+    };
+
+    union FenceAction {
+        u32 raw;
+        BitField<0, 1, FenceOperation> op;
+        BitField<8, 24, u32> syncpoint_id;
+
+        static constexpr CommandHeader Build(FenceOperation op, u32 syncpoint_id) {
+            FenceAction result{};
+            result.op.Assign(op);
+            result.syncpoint_id.Assign(syncpoint_id);
+            return {result.raw};
+        }
+    };
+
     struct Regs {
         static constexpr size_t NUM_REGS = 0x40;
 
@@ -291,10 +309,7 @@ public:
                 u32 semaphore_acquire;
                 u32 semaphore_release;
                 u32 fence_value;
-                union {
-                    BitField<4, 4, u32> operation;
-                    BitField<8, 8, u32> id;
-                } fence_action;
+                FenceAction fence_action;
                 INSERT_UNION_PADDING_WORDS(0xE2);
 
                 // Puller state
@@ -342,6 +357,8 @@ protected:
 
 private:
     void ProcessBindMethod(const MethodCall& method_call);
+    void ProcessFenceActionMethod();
+    void ProcessWaitForInterruptMethod();
     void ProcessSemaphoreTriggerMethod();
     void ProcessSemaphoreRelease();
     void ProcessSemaphoreAcquire();
diff --git a/src/video_core/shader/async_shaders.cpp b/src/video_core/shader/async_shaders.cpp
index aabd62c5c..39cc3b869 100644
--- a/src/video_core/shader/async_shaders.cpp
+++ b/src/video_core/shader/async_shaders.cpp
@@ -20,14 +20,15 @@ AsyncShaders::~AsyncShaders() {
 }
 
 void AsyncShaders::AllocateWorkers() {
-    // Max worker threads we should allow
-    constexpr u32 MAX_THREADS = 4;
-    // Deduce how many threads we can use
-    const u32 threads_used = std::thread::hardware_concurrency() / 4;
-    // Always allow at least 1 thread regardless of our settings
-    const auto max_worker_count = std::max(1U, threads_used);
-    // Don't use more than MAX_THREADS
-    const auto num_workers = std::min(max_worker_count, MAX_THREADS);
+    // Use at least one thread
+    u32 num_workers = 1;
+
+    // Deduce how many more threads we can use
+    const u32 thread_count = std::thread::hardware_concurrency();
+    if (thread_count >= 8) {
+        // Increase async workers by 1 for every 2 threads >= 8
+        num_workers += 1 + (thread_count - 8) / 2;
+    }
 
     // If we already have workers queued, ignore
     if (num_workers == worker_threads.size()) {
diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp
index 4db329fa5..afef5948d 100644
--- a/src/video_core/shader/decode/arithmetic.cpp
+++ b/src/video_core/shader/decode/arithmetic.cpp
@@ -137,7 +137,8 @@ u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) {
         break;
     }
     case OpCode::Id::FCMP_RR:
-    case OpCode::Id::FCMP_RC: {
+    case OpCode::Id::FCMP_RC:
+    case OpCode::Id::FCMP_IMMR: {
         UNIMPLEMENTED_IF(instr.fcmp.ftz == 0);
         Node op_c = GetRegister(instr.gpr39);
         Node comp = GetPredicateComparisonFloat(instr.fcmp.cond, std::move(op_c), Immediate(0.0f));
diff --git a/src/web_service/web_backend.cpp b/src/web_service/web_backend.cpp
index f264b98a0..67183e64c 100644
--- a/src/web_service/web_backend.cpp
+++ b/src/web_service/web_backend.cpp
@@ -71,11 +71,6 @@ struct Client::Impl {
             return {};
         }
 
-        if (!cli->is_socket_open()) {
-            LOG_ERROR(WebService, "Failed to open socket, skipping request!");
-            return {};
-        }
-
         cli->set_connection_timeout(TIMEOUT_SECONDS);
         cli->set_read_timeout(TIMEOUT_SECONDS);
         cli->set_write_timeout(TIMEOUT_SECONDS);