yuzu/game_list: Use non-deprecated version of QString's split() function

The previous overload of split() was deprecated in 5.14.

externals/FidelityFX-FSR/ffx-fsr/ffx_fsr1.h

@@ -747,12 +747,12 @@ AF1 sharpness){
   // Immediate constants for peak range.
   AF2 peakC=AF2(1.0,-1.0*4.0);
   // Limiters, these need to be high precision RCPs.
-  AF1 hitMinR=mn4R*ARcpF1(AF1_(4.0)*mx4R);
-  AF1 hitMinG=mn4G*ARcpF1(AF1_(4.0)*mx4G);
-  AF1 hitMinB=mn4B*ARcpF1(AF1_(4.0)*mx4B);
-  AF1 hitMaxR=(peakC.x-mx4R)*ARcpF1(AF1_(4.0)*mn4R+peakC.y);
-  AF1 hitMaxG=(peakC.x-mx4G)*ARcpF1(AF1_(4.0)*mn4G+peakC.y);
-  AF1 hitMaxB=(peakC.x-mx4B)*ARcpF1(AF1_(4.0)*mn4B+peakC.y);
+  AF1 hitMinR=min(mn4R,eR)*ARcpF1(AF1_(4.0)*mx4R);
+  AF1 hitMinG=min(mn4G,eG)*ARcpF1(AF1_(4.0)*mx4G);
+  AF1 hitMinB=min(mn4B,eB)*ARcpF1(AF1_(4.0)*mx4B);
+  AF1 hitMaxR=(peakC.x-max(mx4R,eR))*ARcpF1(AF1_(4.0)*mn4R+peakC.y);
+  AF1 hitMaxG=(peakC.x-max(mx4G,eG))*ARcpF1(AF1_(4.0)*mn4G+peakC.y);
+  AF1 hitMaxB=(peakC.x-max(mx4B,eB))*ARcpF1(AF1_(4.0)*mn4B+peakC.y);
   AF1 lobeR=max(-hitMinR,hitMaxR);
   AF1 lobeG=max(-hitMinG,hitMaxG);
   AF1 lobeB=max(-hitMinB,hitMaxB);
@@ -845,12 +845,12 @@ AF1 sharpness){
   // Immediate constants for peak range.
   AH2 peakC=AH2(1.0,-1.0*4.0);
   // Limiters, these need to be high precision RCPs.
-  AH1 hitMinR=mn4R*ARcpH1(AH1_(4.0)*mx4R);
-  AH1 hitMinG=mn4G*ARcpH1(AH1_(4.0)*mx4G);
-  AH1 hitMinB=mn4B*ARcpH1(AH1_(4.0)*mx4B);
-  AH1 hitMaxR=(peakC.x-mx4R)*ARcpH1(AH1_(4.0)*mn4R+peakC.y);
-  AH1 hitMaxG=(peakC.x-mx4G)*ARcpH1(AH1_(4.0)*mn4G+peakC.y);
-  AH1 hitMaxB=(peakC.x-mx4B)*ARcpH1(AH1_(4.0)*mn4B+peakC.y);
+  AH1 hitMinR=min(mn4R,eR)*ARcpH1(AH1_(4.0)*mx4R);
+  AH1 hitMinG=min(mn4G,eG)*ARcpH1(AH1_(4.0)*mx4G);
+  AH1 hitMinB=min(mn4B,eB)*ARcpH1(AH1_(4.0)*mx4B);
+  AH1 hitMaxR=(peakC.x-max(mx4R,eR))*ARcpH1(AH1_(4.0)*mn4R+peakC.y);
+  AH1 hitMaxG=(peakC.x-max(mx4G,eG))*ARcpH1(AH1_(4.0)*mn4G+peakC.y);
+  AH1 hitMaxB=(peakC.x-max(mx4B,eB))*ARcpH1(AH1_(4.0)*mn4B+peakC.y);
   AH1 lobeR=max(-hitMinR,hitMaxR);
   AH1 lobeG=max(-hitMinG,hitMaxG);
   AH1 lobeB=max(-hitMinB,hitMaxB);
@@ -963,12 +963,12 @@ AF1 sharpness){
   // Immediate constants for peak range.
   AH2 peakC=AH2(1.0,-1.0*4.0);
   // Limiters, these need to be high precision RCPs.
-  AH2 hitMinR=mn4R*ARcpH2(AH2_(4.0)*mx4R);
-  AH2 hitMinG=mn4G*ARcpH2(AH2_(4.0)*mx4G);
-  AH2 hitMinB=mn4B*ARcpH2(AH2_(4.0)*mx4B);
-  AH2 hitMaxR=(peakC.x-mx4R)*ARcpH2(AH2_(4.0)*mn4R+peakC.y);
-  AH2 hitMaxG=(peakC.x-mx4G)*ARcpH2(AH2_(4.0)*mn4G+peakC.y);
-  AH2 hitMaxB=(peakC.x-mx4B)*ARcpH2(AH2_(4.0)*mn4B+peakC.y);
+  AH2 hitMinR=min(mn4R,eR)*ARcpH2(AH2_(4.0)*mx4R);
+  AH2 hitMinG=min(mn4G,eG)*ARcpH2(AH2_(4.0)*mx4G);
+  AH2 hitMinB=min(mn4B,eB)*ARcpH2(AH2_(4.0)*mx4B);
+  AH2 hitMaxR=(peakC.x-max(mx4R,eR))*ARcpH2(AH2_(4.0)*mn4R+peakC.y);
+  AH2 hitMaxG=(peakC.x-max(mx4G,eG))*ARcpH2(AH2_(4.0)*mn4G+peakC.y);
+  AH2 hitMaxB=(peakC.x-max(mx4B,eB))*ARcpH2(AH2_(4.0)*mn4B+peakC.y);
   AH2 lobeR=max(-hitMinR,hitMaxR);
   AH2 lobeG=max(-hitMinG,hitMaxG);
   AH2 lobeB=max(-hitMinB,hitMaxB);

externals/ffmpeg/CMakeLists.txt

@@ -95,7 +95,8 @@ if (NOT WIN32)
         # ffnvenc could load CUDA libraries at the runtime using dlopen/dlsym or LoadLibrary/GetProcAddress
         # here we handle the hard-linking senario where CUDA is linked during compilation
         if (CUDA_FOUND)
-            list(APPEND FFmpeg_HWACCEL_FLAGS --extra-cflags=-I${CUDA_INCLUDE_DIRS})
+            # This line causes build error if CUDA_INCLUDE_DIRS is anything but a single non-empty value
+            #list(APPEND FFmpeg_HWACCEL_FLAGS --extra-cflags=-I${CUDA_INCLUDE_DIRS})
             list(APPEND FFmpeg_HWACCEL_LIBRARIES ${CUDA_LIBRARIES})
             list(APPEND FFmpeg_HWACCEL_INCLUDE_DIRS ${CUDA_INCLUDE_DIRS})
             list(APPEND FFmpeg_HWACCEL_LDFLAGS ${CUDA_LDFLAGS})
@@ -119,6 +120,8 @@ if (NOT WIN32)
 
     # `configure` parameters builds only exactly what yuzu needs from FFmpeg
     # `--disable-vdpau` is needed to avoid linking issues
+    set(FFmpeg_CC ${CMAKE_C_COMPILER_LAUNCHER}  ${CMAKE_C_COMPILER})
+    set(FFmpeg_CXX ${CMAKE_CXX_COMPILER_LAUNCHER}  ${CMAKE_CXX_COMPILER})
     add_custom_command(
         OUTPUT
             ${FFmpeg_MAKEFILE}
@@ -137,12 +140,14 @@ if (NOT WIN32)
                 --enable-decoder=h264
                 --enable-decoder=vp8
                 --enable-decoder=vp9
-                --cc="${CMAKE_C_COMPILER}"
-                --cxx="${CMAKE_CXX_COMPILER}"
+                --cc="${FFmpeg_CC}"
+                --cxx="${FFmpeg_CXX}"
                 ${FFmpeg_HWACCEL_FLAGS}
         WORKING_DIRECTORY
             ${FFmpeg_BUILD_DIR}
     )
+    unset(FFmpeg_CC)
+    unset(FFmpeg_CXX)
     unset(FFmpeg_HWACCEL_FLAGS)
 
     # Workaround for Ubuntu 18.04's older version of make not being able to call make as a child

src/audio_core/stream.cpp

@@ -79,8 +79,8 @@ static void VolumeAdjustSamples(std::vector<s16>& samples, float game_volume) {
         return;
     }
 
-    // Implementation of a volume slider with a dynamic range of 60 dB
-    const float volume_scale_factor = volume == 0 ? 0 : std::exp(6.90775f * volume) * 0.001f;
+    // Perceived volume is not the same as the volume level
+    const float volume_scale_factor = (0.85f * ((volume * volume) - volume)) + volume;
     for (auto& sample : samples) {
         sample = static_cast<s16>(sample * volume_scale_factor);
     }

src/common/bit_util.h

@@ -45,6 +45,12 @@ template <typename T>
     return static_cast<u32>(log2_f + static_cast<u64>((value ^ (1ULL << log2_f)) != 0ULL));
 }
 
+template <typename T>
+requires std::is_unsigned_v<T>
+[[nodiscard]] constexpr bool IsPow2(T value) {
+    return std::has_single_bit(value);
+}
+
 template <typename T>
 requires std::is_integral_v<T>
 [[nodiscard]] T NextPow2(T value) {

src/common/fiber.cpp

@@ -124,7 +124,10 @@ void Fiber::YieldTo(std::weak_ptr<Fiber> weak_from, Fiber& to) {
 
     // "from" might no longer be valid if the thread was killed
     if (auto from = weak_from.lock()) {
-        ASSERT(from->impl->previous_fiber != nullptr);
+        if (from->impl->previous_fiber == nullptr) {
+            ASSERT_MSG(false, "previous_fiber is nullptr!");
+            return;
+        }
         from->impl->previous_fiber->impl->context = transfer.fctx;
         from->impl->previous_fiber->impl->guard.unlock();
         from->impl->previous_fiber.reset();

src/common/fs/file.h

@@ -188,9 +188,8 @@ public:
 
 #ifdef _WIN32
     template <typename Path>
-    [[nodiscard]] void Open(const Path& path, FileAccessMode mode,
-                            FileType type = FileType::BinaryFile,
-                            FileShareFlag flag = FileShareFlag::ShareReadOnly) {
+    void Open(const Path& path, FileAccessMode mode, FileType type = FileType::BinaryFile,
+              FileShareFlag flag = FileShareFlag::ShareReadOnly) {
         using ValueType = typename Path::value_type;
         if constexpr (IsChar<ValueType>) {
             Open(ToU8String(path), mode, type, flag);

src/common/input.h

@@ -209,6 +209,13 @@ enum class ButtonNames {
     Triangle,
     Share,
     Options,
+
+    // Mouse buttons
+    ButtonMouseWheel,
+    ButtonBackward,
+    ButtonForward,
+    ButtonTask,
+    ButtonExtra,
 };
 
 // Callback data consisting of an input type and the equivalent data status

src/common/settings.h

@@ -554,6 +554,7 @@ struct Values {
     Setting<bool> use_docked_mode{true, "use_docked_mode"};
 
     BasicSetting<bool> enable_raw_input{false, "enable_raw_input"};
+    BasicSetting<bool> controller_navigation{true, "controller_navigation"};
 
     Setting<bool> vibration_enabled{true, "vibration_enabled"};
     Setting<bool> enable_accurate_vibrations{false, "enable_accurate_vibrations"};

src/common/wall_clock.cpp

@@ -72,7 +72,9 @@ std::unique_ptr<WallClock> CreateBestMatchingClock(u32 emulated_cpu_frequency,
     if (caps.invariant_tsc) {
         rtsc_frequency = EstimateRDTSCFrequency();
     }
-    if (rtsc_frequency == 0) {
+
+    // Fallback to StandardWallClock if rtsc period is higher than a nano second
+    if (rtsc_frequency <= 1000000000) {
         return std::make_unique<StandardWallClock>(emulated_cpu_frequency,
                                                    emulated_clock_frequency);
     } else {

src/common/x64/xbyak_abi.h

@@ -37,12 +37,12 @@ constexpr Xbyak::Reg IndexToReg(size_t reg_index) {
     }
 }
 
-inline std::bitset<32> BuildRegSet(std::initializer_list<Xbyak::Reg> regs) {
-    std::bitset<32> bits;
+constexpr std::bitset<32> BuildRegSet(std::initializer_list<Xbyak::Reg> regs) {
+    size_t bits = 0;
     for (const Xbyak::Reg& reg : regs) {
-        bits[RegToIndex(reg)] = true;
+        bits |= size_t{1} << RegToIndex(reg);
     }
-    return bits;
+    return {bits};
 }
 
 constexpr inline std::bitset<32> ABI_ALL_GPRS(0x0000FFFF);
@@ -57,7 +57,7 @@ constexpr inline Xbyak::Reg ABI_PARAM2 = Xbyak::util::rdx;
 constexpr inline Xbyak::Reg ABI_PARAM3 = Xbyak::util::r8;
 constexpr inline Xbyak::Reg ABI_PARAM4 = Xbyak::util::r9;
 
-const std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
+constexpr inline std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
     // GPRs
     Xbyak::util::rcx,
     Xbyak::util::rdx,
@@ -74,7 +74,7 @@ const std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
     Xbyak::util::xmm5,
 });
 
-const std::bitset<32> ABI_ALL_CALLEE_SAVED = BuildRegSet({
+constexpr inline std::bitset<32> ABI_ALL_CALLEE_SAVED = BuildRegSet({
     // GPRs
     Xbyak::util::rbx,
     Xbyak::util::rsi,
@@ -108,7 +108,7 @@ constexpr inline Xbyak::Reg ABI_PARAM2 = Xbyak::util::rsi;
 constexpr inline Xbyak::Reg ABI_PARAM3 = Xbyak::util::rdx;
 constexpr inline Xbyak::Reg ABI_PARAM4 = Xbyak::util::rcx;
 
-const std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
+constexpr inline std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
     // GPRs
     Xbyak::util::rcx,
     Xbyak::util::rdx,
@@ -137,7 +137,7 @@ const std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
     Xbyak::util::xmm15,
 });
 
-const std::bitset<32> ABI_ALL_CALLEE_SAVED = BuildRegSet({
+constexpr inline std::bitset<32> ABI_ALL_CALLEE_SAVED = BuildRegSet({
     // GPRs
     Xbyak::util::rbx,
     Xbyak::util::rbp,

src/core/CMakeLists.txt

@@ -247,6 +247,9 @@ add_library(core STATIC
     hle/kernel/k_trace.h
     hle/kernel/k_transfer_memory.cpp
     hle/kernel/k_transfer_memory.h
+    hle/kernel/k_worker_task.h
+    hle/kernel/k_worker_task_manager.cpp
+    hle/kernel/k_worker_task_manager.h
     hle/kernel/k_writable_event.cpp
     hle/kernel/k_writable_event.h
     hle/kernel/kernel.cpp

src/core/hid/emulated_console.cpp

@@ -158,7 +158,7 @@ void EmulatedConsole::SetMotion(const Common::Input::CallbackStatus& callback) {
     auto& motion = console.motion_state;
     motion.accel = emulated.GetAcceleration();
     motion.gyro = emulated.GetGyroscope();
-    motion.rotation = emulated.GetGyroscope();
+    motion.rotation = emulated.GetRotations();
     motion.orientation = emulated.GetOrientation();
     motion.quaternion = emulated.GetQuaternion();
     motion.gyro_bias = emulated.GetGyroBias();

src/core/hid/emulated_controller.cpp

@@ -145,7 +145,7 @@ void EmulatedController::LoadDevices() {
                    motion_devices.begin(), Common::Input::CreateDevice<Common::Input::InputDevice>);
     std::transform(trigger_params.begin(), trigger_params.end(), trigger_devices.begin(),
                    Common::Input::CreateDevice<Common::Input::InputDevice>);
-    std::transform(battery_params.begin(), battery_params.begin(), battery_devices.end(),
+    std::transform(battery_params.begin(), battery_params.end(), battery_devices.begin(),
                    Common::Input::CreateDevice<Common::Input::InputDevice>);
     std::transform(output_params.begin(), output_params.end(), output_devices.begin(),
                    Common::Input::CreateDevice<Common::Input::OutputDevice>);
@@ -749,6 +749,7 @@ void EmulatedController::SetMotion(const Common::Input::CallbackStatus& callback
         raw_status.gyro.y.value,
         raw_status.gyro.z.value,
     });
+    emulated.SetGyroThreshold(raw_status.gyro.x.properties.threshold);
     emulated.UpdateRotation(raw_status.delta_timestamp);
     emulated.UpdateOrientation(raw_status.delta_timestamp);
     force_update_motion = raw_status.force_update;

src/core/hid/input_converter.cpp

@@ -114,7 +114,7 @@ Common::Input::MotionStatus TransformToMotion(const Common::Input::CallbackStatu
         if (TransformToButton(callback).value) {
             std::random_device device;
             std::mt19937 gen(device());
-            std::uniform_int_distribution<s16> distribution(-1000, 1000);
+            std::uniform_int_distribution<s16> distribution(-5000, 5000);
             status.accel.x.raw_value = static_cast<f32>(distribution(gen)) * 0.001f;
             status.accel.y.raw_value = static_cast<f32>(distribution(gen)) * 0.001f;
             status.accel.z.raw_value = static_cast<f32>(distribution(gen)) * 0.001f;

src/core/hid/motion_input.cpp

@@ -10,7 +10,7 @@ namespace Core::HID {
 MotionInput::MotionInput() {
     // Initialize PID constants with default values
     SetPID(0.3f, 0.005f, 0.0f);
-    SetGyroThreshold(0.00005f);
+    SetGyroThreshold(0.007f);
 }
 
 void MotionInput::SetPID(f32 new_kp, f32 new_ki, f32 new_kd) {
@@ -31,7 +31,7 @@ void MotionInput::SetGyroscope(const Common::Vec3f& gyroscope) {
         gyro_bias = (gyro_bias * 0.9999f) + (gyroscope * 0.0001f);
     }
 
-    if (gyro.Length2() < gyro_threshold) {
+    if (gyro.Length() < gyro_threshold) {
         gyro = {};
     } else {
         only_accelerometer = false;

src/core/hle/kernel/hle_ipc.h

@@ -341,10 +341,6 @@ public:
         return *thread;
     }
 
-    bool IsThreadWaiting() const {
-        return is_thread_waiting;
-    }
-
 private:
     friend class IPC::ResponseBuilder;
 
@@ -379,7 +375,6 @@ private:
     u32 domain_offset{};
 
     std::shared_ptr<SessionRequestManager> manager;
-    bool is_thread_waiting{};
 
     KernelCore& kernel;
     Core::Memory::Memory& memory;

src/core/hle/kernel/k_affinity_mask.h

@@ -31,8 +31,6 @@ public:
     }
 
     constexpr void SetAffinity(s32 core, bool set) {
-        ASSERT(0 <= core && core < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
-
         if (set) {
             this->mask |= GetCoreBit(core);
         } else {

src/core/hle/kernel/k_page_table.cpp

@@ -276,22 +276,23 @@ ResultCode KPageTable::InitializeForProcess(FileSys::ProgramAddressSpaceType as_
 
 ResultCode KPageTable::MapProcessCode(VAddr addr, std::size_t num_pages, KMemoryState state,
                                       KMemoryPermission perm) {
-    std::lock_guard lock{page_table_lock};
-
     const u64 size{num_pages * PageSize};
 
-    if (!CanContain(addr, size, state)) {
-        return ResultInvalidCurrentMemory;
-    }
+    // Validate the mapping request.
+    R_UNLESS(this->CanContain(addr, size, state), ResultInvalidCurrentMemory);
 
-    if (IsRegionMapped(addr, size)) {
-        return ResultInvalidCurrentMemory;
-    }
+    // Lock the table.
+    std::lock_guard lock{page_table_lock};
+
+    // Verify that the destination memory is unmapped.
+    R_TRY(this->CheckMemoryState(addr, size, KMemoryState::All, KMemoryState::Free,
+                                 KMemoryPermission::None, KMemoryPermission::None,
+                                 KMemoryAttribute::None, KMemoryAttribute::None));
 
     KPageLinkedList page_linked_list;
-    CASCADE_CODE(system.Kernel().MemoryManager().Allocate(page_linked_list, num_pages, memory_pool,
-                                                          allocation_option));
-    CASCADE_CODE(Operate(addr, num_pages, page_linked_list, OperationType::MapGroup));
+    R_TRY(system.Kernel().MemoryManager().Allocate(page_linked_list, num_pages, memory_pool,
+                                                   allocation_option));
+    R_TRY(Operate(addr, num_pages, page_linked_list, OperationType::MapGroup));
 
     block_manager->Update(addr, num_pages, state, perm);
 
@@ -395,39 +396,12 @@ ResultCode KPageTable::UnmapProcessMemory(VAddr dst_addr, std::size_t size,
 
     return ResultSuccess;
 }
-void KPageTable::MapPhysicalMemory(KPageLinkedList& page_linked_list, VAddr start, VAddr end) {
-    auto node{page_linked_list.Nodes().begin()};
-    PAddr map_addr{node->GetAddress()};
-    std::size_t src_num_pages{node->GetNumPages()};
-
-    block_manager->IterateForRange(start, end, [&](const KMemoryInfo& info) {
-        if (info.state != KMemoryState::Free) {
-            return;
-        }
-
-        std::size_t dst_num_pages{GetSizeInRange(info, start, end) / PageSize};
-        VAddr dst_addr{GetAddressInRange(info, start)};
-
-        while (dst_num_pages) {
-            if (!src_num_pages) {
-                node = std::next(node);
-                map_addr = node->GetAddress();
-                src_num_pages = node->GetNumPages();
-            }
-
-            const std::size_t num_pages{std::min(src_num_pages, dst_num_pages)};
-            Operate(dst_addr, num_pages, KMemoryPermission::UserReadWrite, OperationType::Map,
-                    map_addr);
-
-            dst_addr += num_pages * PageSize;
-            map_addr += num_pages * PageSize;
-            src_num_pages -= num_pages;
-            dst_num_pages -= num_pages;
-        }
-    });
-}
 
 ResultCode KPageTable::MapPhysicalMemory(VAddr addr, std::size_t size) {
+    // Lock the physical memory lock.
+    std::lock_guard phys_lk(map_physical_memory_lock);
+
+    // Lock the table.
     std::lock_guard lock{page_table_lock};
 
     std::size_t mapped_size{};
@@ -463,7 +437,35 @@ ResultCode KPageTable::MapPhysicalMemory(VAddr addr, std::size_t size) {
     // We succeeded, so commit the memory reservation.
     memory_reservation.Commit();
 
-    MapPhysicalMemory(page_linked_list, addr, end_addr);
+    // Map the memory.
+    auto node{page_linked_list.Nodes().begin()};
+    PAddr map_addr{node->GetAddress()};
+    std::size_t src_num_pages{node->GetNumPages()};
+    block_manager->IterateForRange(addr, end_addr, [&](const KMemoryInfo& info) {
+        if (info.state != KMemoryState::Free) {
+            return;
+        }
+
+        std::size_t dst_num_pages{GetSizeInRange(info, addr, end_addr) / PageSize};
+        VAddr dst_addr{GetAddressInRange(info, addr)};
+
+        while (dst_num_pages) {
+            if (!src_num_pages) {
+                node = std::next(node);
+                map_addr = node->GetAddress();
+                src_num_pages = node->GetNumPages();
+            }
+
+            const std::size_t num_pages{std::min(src_num_pages, dst_num_pages)};
+            Operate(dst_addr, num_pages, KMemoryPermission::UserReadWrite, OperationType::Map,
+                    map_addr);
+
+            dst_addr += num_pages * PageSize;
+            map_addr += num_pages * PageSize;
+            src_num_pages -= num_pages;
+            dst_num_pages -= num_pages;
+        }
+    });
 
     mapped_physical_memory_size += remaining_size;
 
@@ -503,23 +505,8 @@ ResultCode KPageTable::UnmapPhysicalMemory(VAddr addr, std::size_t size) {
         return ResultSuccess;
     }
 
-    CASCADE_CODE(UnmapMemory(addr, size));
-
-    auto process{system.Kernel().CurrentProcess()};
-    process->GetResourceLimit()->Release(LimitableResource::PhysicalMemory, mapped_size);
-    mapped_physical_memory_size -= mapped_size;
-
-    return ResultSuccess;
-}
-
-ResultCode KPageTable::UnmapMemory(VAddr addr, std::size_t size) {
-    std::lock_guard lock{page_table_lock};
-
-    const VAddr end_addr{addr + size};
-    ResultCode result{ResultSuccess};
-    KPageLinkedList page_linked_list;
-
     // Unmap each region within the range
+    KPageLinkedList page_linked_list;
     block_manager->IterateForRange(addr, end_addr, [&](const KMemoryInfo& info) {
         if (info.state == KMemoryState::Normal) {
             const std::size_t block_size{GetSizeInRange(info, addr, end_addr)};
@@ -535,7 +522,6 @@ ResultCode KPageTable::UnmapMemory(VAddr addr, std::size_t size) {
             }
         }
     });
-
     if (result.IsError()) {
         return result;
     }
@@ -546,10 +532,14 @@ ResultCode KPageTable::UnmapMemory(VAddr addr, std::size_t size) {
 
     block_manager->Update(addr, num_pages, KMemoryState::Free);
 
+    auto process{system.Kernel().CurrentProcess()};
+    process->GetResourceLimit()->Release(LimitableResource::PhysicalMemory, mapped_size);
+    mapped_physical_memory_size -= mapped_size;
+
     return ResultSuccess;
 }
 
-ResultCode KPageTable::Map(VAddr dst_addr, VAddr src_addr, std::size_t size) {
+ResultCode KPageTable::MapMemory(VAddr dst_addr, VAddr src_addr, std::size_t size) {
     std::lock_guard lock{page_table_lock};
 
     KMemoryState src_state{};
@@ -588,7 +578,7 @@ ResultCode KPageTable::Map(VAddr dst_addr, VAddr src_addr, std::size_t size) {
     return ResultSuccess;
 }
 
-ResultCode KPageTable::Unmap(VAddr dst_addr, VAddr src_addr, std::size_t size) {
+ResultCode KPageTable::UnmapMemory(VAddr dst_addr, VAddr src_addr, std::size_t size) {
     std::lock_guard lock{page_table_lock};
 
     KMemoryState src_state{};
@@ -652,24 +642,26 @@ ResultCode KPageTable::MapPages(VAddr addr, const KPageLinkedList& page_linked_l
     return ResultSuccess;
 }
 
-ResultCode KPageTable::MapPages(VAddr addr, KPageLinkedList& page_linked_list, KMemoryState state,
-                                KMemoryPermission perm) {
-    std::lock_guard lock{page_table_lock};
-
+ResultCode KPageTable::MapPages(VAddr address, KPageLinkedList& page_linked_list,
+                                KMemoryState state, KMemoryPermission perm) {
+    // Check that the map is in range.
     const std::size_t num_pages{page_linked_list.GetNumPages()};
     const std::size_t size{num_pages * PageSize};
+    R_UNLESS(this->CanContain(address, size, state), ResultInvalidCurrentMemory);
 
-    if (!CanContain(addr, size, state)) {
-        return ResultInvalidCurrentMemory;
-    }
+    // Lock the table.
+    std::lock_guard lock{page_table_lock};
 
-    if (IsRegionMapped(addr, num_pages * PageSize)) {
-        return ResultInvalidCurrentMemory;
-    }
+    // Check the memory state.
+    R_TRY(this->CheckMemoryState(address, size, KMemoryState::All, KMemoryState::Free,
+                                 KMemoryPermission::None, KMemoryPermission::None,
+                                 KMemoryAttribute::None, KMemoryAttribute::None));
 
-    CASCADE_CODE(MapPages(addr, page_linked_list, perm));
+    // Map the pages.
+    R_TRY(MapPages(address, page_linked_list, perm));
 
-    block_manager->Update(addr, num_pages, state, perm);
+    // Update the blocks.
+    block_manager->Update(address, num_pages, state, perm);
 
     return ResultSuccess;
 }
@@ -693,21 +685,23 @@ ResultCode KPageTable::UnmapPages(VAddr addr, const KPageLinkedList& page_linked
 
 ResultCode KPageTable::UnmapPages(VAddr addr, KPageLinkedList& page_linked_list,
                                   KMemoryState state) {
-    std::lock_guard lock{page_table_lock};
-
+    // Check that the unmap is in range.
     const std::size_t num_pages{page_linked_list.GetNumPages()};
     const std::size_t size{num_pages * PageSize};
+    R_UNLESS(this->Contains(addr, size), ResultInvalidCurrentMemory);
 
-    if (!CanContain(addr, size, state)) {
-        return ResultInvalidCurrentMemory;
-    }
+    // Lock the table.
+    std::lock_guard lock{page_table_lock};
 
-    if (IsRegionMapped(addr, num_pages * PageSize)) {
-        return ResultInvalidCurrentMemory;
-    }
+    // Check the memory state.
+    R_TRY(this->CheckMemoryState(addr, size, KMemoryState::All, state, KMemoryPermission::None,
+                                 KMemoryPermission::None, KMemoryAttribute::All,
+                                 KMemoryAttribute::None));
 
-    CASCADE_CODE(UnmapPages(addr, page_linked_list));
+    // Perform the unmap.
+    R_TRY(UnmapPages(addr, page_linked_list));
 
+    // Update the blocks.
     block_manager->Update(addr, num_pages, state, KMemoryPermission::None);
 
     return ResultSuccess;
@@ -765,7 +759,6 @@ ResultCode KPageTable::SetProcessMemoryPermission(VAddr addr, std::size_t size,
 
     // Ensure cache coherency, if we're setting pages as executable.
     if (is_x) {
-        // Memory execution state is changing, invalidate CPU cache range
         system.InvalidateCpuInstructionCacheRange(addr, size);
     }
 
@@ -793,12 +786,12 @@ ResultCode KPageTable::ReserveTransferMemory(VAddr addr, std::size_t size, KMemo
     KMemoryState state{};
     KMemoryAttribute attribute{};
 
-    CASCADE_CODE(CheckMemoryState(
-        &state, nullptr, &attribute, nullptr, addr, size,
-        KMemoryState::FlagCanTransfer | KMemoryState::FlagReferenceCounted,
-        KMemoryState::FlagCanTransfer | KMemoryState::FlagReferenceCounted, KMemoryPermission::All,
-        KMemoryPermission::UserReadWrite, KMemoryAttribute::Mask, KMemoryAttribute::None,
-        KMemoryAttribute::IpcAndDeviceMapped));
+    R_TRY(CheckMemoryState(&state, nullptr, &attribute, nullptr, addr, size,
+                           KMemoryState::FlagCanTransfer | KMemoryState::FlagReferenceCounted,
+                           KMemoryState::FlagCanTransfer | KMemoryState::FlagReferenceCounted,
+                           KMemoryPermission::All, KMemoryPermission::UserReadWrite,
+                           KMemoryAttribute::Mask, KMemoryAttribute::None,
+                           KMemoryAttribute::IpcAndDeviceMapped));
 
     block_manager->Update(addr, size / PageSize, state, perm, attribute | KMemoryAttribute::Locked);
 
@@ -810,12 +803,11 @@ ResultCode KPageTable::ResetTransferMemory(VAddr addr, std::size_t size) {
 
     KMemoryState state{};
 
-    CASCADE_CODE(
-        CheckMemoryState(&state, nullptr, nullptr, nullptr, addr, size,
-                         KMemoryState::FlagCanTransfer | KMemoryState::FlagReferenceCounted,
-                         KMemoryState::FlagCanTransfer | KMemoryState::FlagReferenceCounted,
-                         KMemoryPermission::None, KMemoryPermission::None, KMemoryAttribute::Mask,
-                         KMemoryAttribute::Locked, KMemoryAttribute::IpcAndDeviceMapped));
+    R_TRY(CheckMemoryState(&state, nullptr, nullptr, nullptr, addr, size,
+                           KMemoryState::FlagCanTransfer | KMemoryState::FlagReferenceCounted,
+                           KMemoryState::FlagCanTransfer | KMemoryState::FlagReferenceCounted,
+                           KMemoryPermission::None, KMemoryPermission::None, KMemoryAttribute::Mask,
+                           KMemoryAttribute::Locked, KMemoryAttribute::IpcAndDeviceMapped));
 
     block_manager->Update(addr, size / PageSize, state, KMemoryPermission::UserReadWrite);
     return ResultSuccess;
@@ -871,8 +863,9 @@ ResultCode KPageTable::SetMemoryAttribute(VAddr addr, std::size_t size, u32 mask
         AttributeTestMask, KMemoryAttribute::None, ~AttributeTestMask));
 
     // Determine the new attribute.
-    const auto new_attr = ((old_attr & static_cast<KMemoryAttribute>(~mask)) |
-                           static_cast<KMemoryAttribute>(attr & mask));
+    const KMemoryAttribute new_attr =
+        static_cast<KMemoryAttribute>(((old_attr & static_cast<KMemoryAttribute>(~mask)) |
+                                       static_cast<KMemoryAttribute>(attr & mask)));
 
     // Perform operation.
     this->Operate(addr, num_pages, old_perm, OperationType::ChangePermissionsAndRefresh);
@@ -896,6 +889,9 @@ ResultCode KPageTable::SetMaxHeapSize(std::size_t size) {
 }
 
 ResultCode KPageTable::SetHeapSize(VAddr* out, std::size_t size) {
+    // Lock the physical memory lock.
+    std::lock_guard phys_lk(map_physical_memory_lock);
+
     // Try to perform a reduction in heap, instead of an extension.
     VAddr cur_address{};
     std::size_t allocation_size{};
@@ -1025,12 +1021,12 @@ ResultVal<VAddr> KPageTable::AllocateAndMapMemory(std::size_t needed_num_pages,
     }
 
     if (is_map_only) {
-        CASCADE_CODE(Operate(addr, needed_num_pages, perm, OperationType::Map, map_addr));
+        R_TRY(Operate(addr, needed_num_pages, perm, OperationType::Map, map_addr));
     } else {
         KPageLinkedList page_group;
-        CASCADE_CODE(system.Kernel().MemoryManager().Allocate(page_group, needed_num_pages,
-                                                              memory_pool, allocation_option));
-        CASCADE_CODE(Operate(addr, needed_num_pages, page_group, OperationType::MapGroup));
+        R_TRY(system.Kernel().MemoryManager().Allocate(page_group, needed_num_pages, memory_pool,
+                                                       allocation_option));
+        R_TRY(Operate(addr, needed_num_pages, page_group, OperationType::MapGroup));
     }
 
     block_manager->Update(addr, needed_num_pages, state, perm);
@@ -1186,7 +1182,7 @@ VAddr KPageTable::AllocateVirtualMemory(VAddr start, std::size_t region_num_page
 
 ResultCode KPageTable::Operate(VAddr addr, std::size_t num_pages, const KPageLinkedList& page_group,
                                OperationType operation) {
-    std::lock_guard lock{page_table_lock};
+    ASSERT(this->IsLockedByCurrentThread());
 
     ASSERT(Common::IsAligned(addr, PageSize));
     ASSERT(num_pages > 0);
@@ -1211,7 +1207,7 @@ ResultCode KPageTable::Operate(VAddr addr, std::size_t num_pages, const KPageLin
 
 ResultCode KPageTable::Operate(VAddr addr, std::size_t num_pages, KMemoryPermission perm,
                                OperationType operation, PAddr map_addr) {
-    std::lock_guard lock{page_table_lock};
+    ASSERT(this->IsLockedByCurrentThread());
 
     ASSERT(num_pages > 0);
     ASSERT(Common::IsAligned(addr, PageSize));

src/core/hle/kernel/k_page_table.h

@@ -37,9 +37,8 @@ public:
                                   VAddr src_addr);
     ResultCode MapPhysicalMemory(VAddr addr, std::size_t size);
     ResultCode UnmapPhysicalMemory(VAddr addr, std::size_t size);
-    ResultCode UnmapMemory(VAddr addr, std::size_t size);
-    ResultCode Map(VAddr dst_addr, VAddr src_addr, std::size_t size);
-    ResultCode Unmap(VAddr dst_addr, VAddr src_addr, std::size_t size);
+    ResultCode MapMemory(VAddr dst_addr, VAddr src_addr, std::size_t size);
+    ResultCode UnmapMemory(VAddr dst_addr, VAddr src_addr, std::size_t size);
     ResultCode MapPages(VAddr addr, KPageLinkedList& page_linked_list, KMemoryState state,
                         KMemoryPermission perm);
     ResultCode UnmapPages(VAddr addr, KPageLinkedList& page_linked_list, KMemoryState state);
@@ -88,7 +87,6 @@ private:
     ResultCode MapPages(VAddr addr, const KPageLinkedList& page_linked_list,
                         KMemoryPermission perm);
     ResultCode UnmapPages(VAddr addr, const KPageLinkedList& page_linked_list);
-    void MapPhysicalMemory(KPageLinkedList& page_linked_list, VAddr start, VAddr end);
     bool IsRegionMapped(VAddr address, u64 size);
     bool IsRegionContiguous(VAddr addr, u64 size) const;
     void AddRegionToPages(VAddr start, std::size_t num_pages, KPageLinkedList& page_linked_list);
@@ -148,6 +146,7 @@ private:
     }
 
     std::recursive_mutex page_table_lock;
+    std::mutex map_physical_memory_lock;
     std::unique_ptr<KMemoryBlockManager> block_manager;
 
 public:
@@ -249,7 +248,9 @@ public:
         return !IsOutsideASLRRegion(address, size);
     }
     constexpr PAddr GetPhysicalAddr(VAddr addr) {
-        return page_table_impl.backing_addr[addr >> PageBits] + addr;
+        const auto backing_addr = page_table_impl.backing_addr[addr >> PageBits];
+        ASSERT(backing_addr);
+        return backing_addr + addr;
     }
     constexpr bool Contains(VAddr addr) const {
         return address_space_start <= addr && addr <= address_space_end - 1;

src/core/hle/kernel/k_priority_queue.h

@@ -45,6 +45,7 @@ concept KPriorityQueueMember = !std::is_reference_v<T> && requires(T & t) {
 
     { t.GetActiveCore() } -> Common::ConvertibleTo<s32>;
     { t.GetPriority() } -> Common::ConvertibleTo<s32>;
+    { t.IsDummyThread() } -> Common::ConvertibleTo<bool>;
 };
 
 template <typename Member, size_t NumCores_, int LowestPriority, int HighestPriority>
@@ -257,7 +258,7 @@ private:
 
 private:
     constexpr void ClearAffinityBit(u64& affinity, s32 core) {
-        affinity &= ~(u64(1) << core);
+        affinity &= ~(UINT64_C(1) << core);
     }
 
     constexpr s32 GetNextCore(u64& affinity) {
@@ -349,24 +350,49 @@ public:
 
     // Mutators.
     constexpr void PushBack(Member* member) {
+        // This is for host (dummy) threads that we do not want to enter the priority queue.
+        if (member->IsDummyThread()) {
+            return;
+        }
+
         this->PushBack(member->GetPriority(), member);
     }
 
     constexpr void Remove(Member* member) {
+        // This is for host (dummy) threads that we do not want to enter the priority queue.
+        if (member->IsDummyThread()) {
+            return;
+        }
+
         this->Remove(member->GetPriority(), member);
     }
 
     constexpr void MoveToScheduledFront(Member* member) {
+        // This is for host (dummy) threads that we do not want to enter the priority queue.
+        if (member->IsDummyThread()) {
+            return;
+        }
+
         this->scheduled_queue.MoveToFront(member->GetPriority(), member->GetActiveCore(), member);
     }
 
     constexpr KThread* MoveToScheduledBack(Member* member) {
+        // This is for host (dummy) threads that we do not want to enter the priority queue.
+        if (member->IsDummyThread()) {
+            return {};
+        }
+
         return this->scheduled_queue.MoveToBack(member->GetPriority(), member->GetActiveCore(),
                                                 member);
     }
 
     // First class fancy operations.
     constexpr void ChangePriority(s32 prev_priority, bool is_running, Member* member) {
+        // This is for host (dummy) threads that we do not want to enter the priority queue.
+        if (member->IsDummyThread()) {
+            return;
+        }
+
         ASSERT(IsValidPriority(prev_priority));
 
         // Remove the member from the queues.
@@ -383,6 +409,11 @@ public:
 
     constexpr void ChangeAffinityMask(s32 prev_core, const AffinityMaskType& prev_affinity,
                                       Member* member) {
+        // This is for host (dummy) threads that we do not want to enter the priority queue.
+        if (member->IsDummyThread()) {
+            return;
+        }
+
         // Get the new information.
         const s32 priority = member->GetPriority();
         const AffinityMaskType& new_affinity = member->GetAffinityMask();
@@ -412,6 +443,11 @@ public:
     }
 
     constexpr void ChangeCore(s32 prev_core, Member* member, bool to_front = false) {
+        // This is for host (dummy) threads that we do not want to enter the priority queue.
+        if (member->IsDummyThread()) {
+            return;
+        }
+
         // Get the new information.
         const s32 new_core = member->GetActiveCore();
         const s32 priority = member->GetPriority();

src/core/hle/kernel/k_process.cpp

@@ -146,27 +146,35 @@ ResultCode KProcess::Initialize(KProcess* process, Core::System& system, std::st
     // Open a reference to the resource limit.
     process->resource_limit->Open();
 
+    // Clear remaining fields.
+    process->num_running_threads = 0;
+    process->is_signaled = false;
+    process->exception_thread = nullptr;
+    process->is_suspended = false;
+    process->schedule_count = 0;
+
     return ResultSuccess;
 }
 
+void KProcess::DoWorkerTaskImpl() {
+    UNIMPLEMENTED();
+}
+
 KResourceLimit* KProcess::GetResourceLimit() const {
     return resource_limit;
 }
 
-void KProcess::IncrementThreadCount() {
-    ASSERT(num_threads >= 0);
-    num_created_threads++;
-
-    if (const auto count = ++num_threads; count > peak_num_threads) {
-        peak_num_threads = count;
-    }
+void KProcess::IncrementRunningThreadCount() {
+    ASSERT(num_running_threads.load() >= 0);
+    ++num_running_threads;
 }
 
-void KProcess::DecrementThreadCount() {
-    ASSERT(num_threads > 0);
+void KProcess::DecrementRunningThreadCount() {
+    ASSERT(num_running_threads.load() > 0);
 
-    if (const auto count = --num_threads; count == 0) {
-        LOG_WARNING(Kernel, "Process termination is not fully implemented.");
+    if (const auto prev = num_running_threads--; prev == 1) {
+        // TODO(bunnei): Process termination to be implemented when multiprocess is supported.
+        UNIMPLEMENTED_MSG("KProcess termination is not implemennted!");
     }
 }
 
@@ -477,7 +485,7 @@ void KProcess::Finalize() {
     }
 
     // Perform inherited finalization.
-    KAutoObjectWithSlabHeapAndContainer<KProcess, KSynchronizationObject>::Finalize();
+    KAutoObjectWithSlabHeapAndContainer<KProcess, KWorkerTask>::Finalize();
 }
 
 /**

src/core/hle/kernel/k_process.h

@@ -15,6 +15,7 @@
 #include "core/hle/kernel/k_condition_variable.h"
 #include "core/hle/kernel/k_handle_table.h"
 #include "core/hle/kernel/k_synchronization_object.h"
+#include "core/hle/kernel/k_worker_task.h"
 #include "core/hle/kernel/process_capability.h"
 #include "core/hle/kernel/slab_helpers.h"
 #include "core/hle/result.h"
@@ -62,8 +63,7 @@ enum class ProcessStatus {
     DebugBreak,
 };
 
-class KProcess final
-    : public KAutoObjectWithSlabHeapAndContainer<KProcess, KSynchronizationObject> {
+class KProcess final : public KAutoObjectWithSlabHeapAndContainer<KProcess, KWorkerTask> {
     KERNEL_AUTOOBJECT_TRAITS(KProcess, KSynchronizationObject);
 
 public:
@@ -235,8 +235,8 @@ public:
         ++schedule_count;
     }
 
-    void IncrementThreadCount();
-    void DecrementThreadCount();
+    void IncrementRunningThreadCount();
+    void DecrementRunningThreadCount();
 
     void SetRunningThread(s32 core, KThread* thread, u64 idle_count) {
         running_threads[core] = thread;
@@ -345,6 +345,8 @@ public:
 
     bool IsSignaled() const override;
 
+    void DoWorkerTaskImpl();
+
     void PinCurrentThread(s32 core_id);
     void UnpinCurrentThread(s32 core_id);
     void UnpinThread(KThread* thread);
@@ -471,9 +473,7 @@ private:
     bool is_suspended{};
     bool is_initialized{};
 
-    std::atomic<s32> num_created_threads{};
-    std::atomic<u16> num_threads{};
-    u16 peak_num_threads{};
+    std::atomic<u16> num_running_threads{};
 
     std::array<KThread*, Core::Hardware::NUM_CPU_CORES> running_threads{};
     std::array<u64, Core::Hardware::NUM_CPU_CORES> running_thread_idle_counts{};

src/core/hle/kernel/k_scheduler.cpp

@@ -49,8 +49,6 @@ void KScheduler::RescheduleCores(KernelCore& kernel, u64 cores_pending_reschedul
         if (!must_context_switch || core != current_core) {
             auto& phys_core = kernel.PhysicalCore(core);
             phys_core.Interrupt();
-        } else {
-            must_context_switch = true;
         }
         cores_pending_reschedule &= ~(1ULL << core);
     }
@@ -408,6 +406,9 @@ void KScheduler::EnableScheduling(KernelCore& kernel, u64 cores_needing_scheduli
     } else {
         RescheduleCores(kernel, cores_needing_scheduling);
     }
+
+    // Special case to ensure dummy threads that are waiting block.
+    current_thread->IfDummyThreadTryWait();
 }
 
 u64 KScheduler::UpdateHighestPriorityThreads(KernelCore& kernel) {
@@ -709,23 +710,19 @@ void KScheduler::Unload(KThread* thread) {
 }
 
 void KScheduler::Reload(KThread* thread) {
-    LOG_TRACE(Kernel, "core {}, reload thread {}", core_id, thread ? thread->GetName() : "nullptr");
+    LOG_TRACE(Kernel, "core {}, reload thread {}", core_id, thread->GetName());
 
-    if (thread) {
-        ASSERT_MSG(thread->GetState() == ThreadState::Runnable, "Thread must be runnable.");
-
-        Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
-        cpu_core.LoadContext(thread->GetContext32());
-        cpu_core.LoadContext(thread->GetContext64());
-        cpu_core.SetTlsAddress(thread->GetTLSAddress());
-        cpu_core.SetTPIDR_EL0(thread->GetTPIDR_EL0());
-        cpu_core.ClearExclusiveState();
-    }
+    Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
+    cpu_core.LoadContext(thread->GetContext32());
+    cpu_core.LoadContext(thread->GetContext64());
+    cpu_core.SetTlsAddress(thread->GetTLSAddress());
+    cpu_core.SetTPIDR_EL0(thread->GetTPIDR_EL0());
+    cpu_core.ClearExclusiveState();
 }
 
 void KScheduler::SwitchContextStep2() {
     // Load context of new thread
-    Reload(current_thread.load());
+    Reload(GetCurrentThread());
 
     RescheduleCurrentCore();
 }
@@ -734,13 +731,23 @@ void KScheduler::ScheduleImpl() {
     KThread* previous_thread = GetCurrentThread();
     KThread* next_thread = state.highest_priority_thread;
 
-    state.needs_scheduling = false;
+    state.needs_scheduling.store(false);
 
     // We never want to schedule a null thread, so use the idle thread if we don't have a next.
     if (next_thread == nullptr) {
         next_thread = idle_thread;
     }
 
+    if (next_thread->GetCurrentCore() != core_id) {
+        next_thread->SetCurrentCore(core_id);
+    }
+
+    // We never want to schedule a dummy thread, as these are only used by host threads for locking.
+    if (next_thread->GetThreadType() == ThreadType::Dummy) {
+        ASSERT_MSG(false, "Dummy threads should never be scheduled!");
+        next_thread = idle_thread;
+    }
+
     // If we're not actually switching thread, there's nothing to do.
     if (next_thread == current_thread.load()) {
         previous_thread->EnableDispatch();
@@ -748,14 +755,8 @@ void KScheduler::ScheduleImpl() {
         return;
     }
 
-    if (next_thread->GetCurrentCore() != core_id) {
-        next_thread->SetCurrentCore(core_id);
-    }
-
-    current_thread.store(next_thread);
-
+    // Update the CPU time tracking variables.
     KProcess* const previous_process = system.Kernel().CurrentProcess();
-
     UpdateLastContextSwitchTime(previous_thread, previous_process);
 
     // Save context for previous thread
@@ -763,6 +764,10 @@ void KScheduler::ScheduleImpl() {
 
     std::shared_ptr<Common::Fiber>* old_context;
     old_context = &previous_thread->GetHostContext();
+
+    // Set the new thread.
+    current_thread.store(next_thread);
+
     guard.Unlock();
 
     Common::Fiber::YieldTo(*old_context, *switch_fiber);
@@ -790,8 +795,8 @@ void KScheduler::SwitchToCurrent() {
         do {
             auto next_thread = current_thread.load();
             if (next_thread != nullptr) {
-                next_thread->context_guard.Lock();
-                if (next_thread->GetRawState() != ThreadState::Runnable) {
+                const auto locked = next_thread->context_guard.TryLock();
+                if (state.needs_scheduling.load()) {
                     next_thread->context_guard.Unlock();
                     break;
                 }
@@ -799,6 +804,9 @@ void KScheduler::SwitchToCurrent() {
                     next_thread->context_guard.Unlock();
                     break;
                 }
+                if (!locked) {
+                    continue;
+                }
             }
             auto thread = next_thread ? next_thread : idle_thread;
             Common::Fiber::YieldTo(switch_fiber, *thread->GetHostContext());

src/core/hle/kernel/k_server_session.cpp

@@ -8,7 +8,6 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
-#include "common/scope_exit.h"
 #include "core/core_timing.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/hle_ipc.h"
@@ -123,20 +122,10 @@ ResultCode KServerSession::QueueSyncRequest(KThread* thread, Core::Memory::Memor
 
     context->PopulateFromIncomingCommandBuffer(kernel.CurrentProcess()->GetHandleTable(), cmd_buf);
 
-    // In the event that something fails here, stub a result to prevent the game from crashing.
-    // This is a work-around in the event that somehow we process a service request after the
-    // session has been closed by the game. This has been observed to happen rarely in Pokemon
-    // Sword/Shield and is likely a result of us using host threads/scheduling for services.
-    // TODO(bunnei): Find a better solution here.
-    auto error_guard = SCOPE_GUARD({ CompleteSyncRequest(*context); });
-
     // Ensure we have a session request handler
     if (manager->HasSessionRequestHandler(*context)) {
         if (auto strong_ptr = manager->GetServiceThread().lock()) {
             strong_ptr->QueueSyncRequest(*parent, std::move(context));
-
-            // We succeeded.
-            error_guard.Cancel();
         } else {
             ASSERT_MSG(false, "strong_ptr is nullptr!");
         }
@@ -171,13 +160,8 @@ ResultCode KServerSession::CompleteSyncRequest(HLERequestContext& context) {
         convert_to_domain = false;
     }
 
-    // Some service requests require the thread to block
-    {
-        KScopedSchedulerLock lock(kernel);
-        if (!context.IsThreadWaiting()) {
-            context.GetThread().EndWait(result);
-        }
-    }
+    // The calling thread is waiting for this request to complete, so wake it up.
+    context.GetThread().EndWait(result);
 
     return result;
 }

src/core/hle/kernel/k_thread.cpp

@@ -30,6 +30,7 @@
 #include "core/hle/kernel/k_system_control.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/k_thread_queue.h"
+#include "core/hle/kernel/k_worker_task_manager.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/svc_results.h"
 #include "core/hle/kernel/time_manager.h"
@@ -105,7 +106,7 @@ KThread::~KThread() = default;
 ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_stack_top, s32 prio,
                                s32 virt_core, KProcess* owner, ThreadType type) {
     // Assert parameters are valid.
-    ASSERT((type == ThreadType::Main) ||
+    ASSERT((type == ThreadType::Main) || (type == ThreadType::Dummy) ||
            (Svc::HighestThreadPriority <= prio && prio <= Svc::LowestThreadPriority));
     ASSERT((owner != nullptr) || (type != ThreadType::User));
     ASSERT(0 <= virt_core && virt_core < static_cast<s32>(Common::BitSize<u64>()));
@@ -139,7 +140,7 @@ ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_s
         UNREACHABLE_MSG("KThread::Initialize: Unknown ThreadType {}", static_cast<u32>(type));
         break;
     }
-    thread_type_for_debugging = type;
+    thread_type = type;
 
     // Set the ideal core ID and affinity mask.
     virtual_ideal_core_id = virt_core;
@@ -214,7 +215,6 @@ ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_s
 
         parent = owner;
         parent->Open();
-        parent->IncrementThreadCount();
     }
 
     // Initialize thread context.
@@ -261,7 +261,7 @@ ResultCode KThread::InitializeThread(KThread* thread, KThreadFunction func, uint
 }
 
 ResultCode KThread::InitializeDummyThread(KThread* thread) {
-    return thread->Initialize({}, {}, {}, DefaultThreadPriority, 3, {}, ThreadType::Dummy);
+    return thread->Initialize({}, {}, {}, DummyThreadPriority, 3, {}, ThreadType::Dummy);
 }
 
 ResultCode KThread::InitializeIdleThread(Core::System& system, KThread* thread, s32 virt_core) {
@@ -326,13 +326,8 @@ void KThread::Finalize() {
         }
     }
 
-    // Decrement the parent process's thread count.
-    if (parent != nullptr) {
-        parent->DecrementThreadCount();
-    }
-
     // Perform inherited finalization.
-    KAutoObjectWithSlabHeapAndContainer<KThread, KSynchronizationObject>::Finalize();
+    KSynchronizationObject::Finalize();
 }
 
 bool KThread::IsSignaled() const {
@@ -376,11 +371,28 @@ void KThread::StartTermination() {
 
     // Register terminated dpc flag.
     RegisterDpc(DpcFlag::Terminated);
+}
+
+void KThread::FinishTermination() {
+    // Ensure that the thread is not executing on any core.
+    if (parent != nullptr) {
+        for (std::size_t i = 0; i < static_cast<std::size_t>(Core::Hardware::NUM_CPU_CORES); ++i) {
+            KThread* core_thread{};
+            do {
+                core_thread = kernel.Scheduler(i).GetCurrentThread();
+            } while (core_thread == this);
+        }
+    }
 
     // Close the thread.
     this->Close();
 }
 
+void KThread::DoWorkerTaskImpl() {
+    // Finish the termination that was begun by Exit().
+    this->FinishTermination();
+}
+
 void KThread::Pin(s32 current_core) {
     ASSERT(kernel.GlobalSchedulerContext().IsLocked());
 
@@ -417,12 +429,7 @@ void KThread::Pin(s32 current_core) {
                                          static_cast<u32>(ThreadState::SuspendShift)));
 
         // Update our state.
-        const ThreadState old_state = thread_state;
-        thread_state = static_cast<ThreadState>(GetSuspendFlags() |
-                                                static_cast<u32>(old_state & ThreadState::Mask));
-        if (thread_state != old_state) {
-            KScheduler::OnThreadStateChanged(kernel, this, old_state);
-        }
+        UpdateState();
     }
 
     // TODO(bunnei): Update our SVC access permissions.
@@ -463,20 +470,13 @@ void KThread::Unpin() {
     }
 
     // Allow performing thread suspension (if termination hasn't been requested).
-    {
+    if (!IsTerminationRequested()) {
         // Update our allow flags.
-        if (!IsTerminationRequested()) {
-            suspend_allowed_flags |= (1 << (static_cast<u32>(SuspendType::Thread) +
-                                            static_cast<u32>(ThreadState::SuspendShift)));
-        }
+        suspend_allowed_flags |= (1 << (static_cast<u32>(SuspendType::Thread) +
+                                        static_cast<u32>(ThreadState::SuspendShift)));
 
         // Update our state.
-        const ThreadState old_state = thread_state;
-        thread_state = static_cast<ThreadState>(GetSuspendFlags() |
-                                                static_cast<u32>(old_state & ThreadState::Mask));
-        if (thread_state != old_state) {
-            KScheduler::OnThreadStateChanged(kernel, this, old_state);
-        }
+        UpdateState();
     }
 
     // TODO(bunnei): Update our SVC access permissions.
@@ -689,12 +689,7 @@ void KThread::Resume(SuspendType type) {
         ~(1u << (static_cast<u32>(ThreadState::SuspendShift) + static_cast<u32>(type)));
 
     // Update our state.
-    const ThreadState old_state = thread_state;
-    thread_state = static_cast<ThreadState>(GetSuspendFlags() |
-                                            static_cast<u32>(old_state & ThreadState::Mask));
-    if (thread_state != old_state) {
-        KScheduler::OnThreadStateChanged(kernel, this, old_state);
-    }
+    this->UpdateState();
 }
 
 void KThread::WaitCancel() {
@@ -721,19 +716,22 @@ void KThread::TrySuspend() {
     ASSERT(GetNumKernelWaiters() == 0);
 
     // Perform the suspend.
-    Suspend();
+    this->UpdateState();
 }
 
-void KThread::Suspend() {
+void KThread::UpdateState() {
     ASSERT(kernel.GlobalSchedulerContext().IsLocked());
-    ASSERT(IsSuspendRequested());
 
     // Set our suspend flags in state.
     const auto old_state = thread_state;
-    thread_state = static_cast<ThreadState>(GetSuspendFlags()) | (old_state & ThreadState::Mask);
+    const auto new_state =
+        static_cast<ThreadState>(this->GetSuspendFlags()) | (old_state & ThreadState::Mask);
+    thread_state = new_state;
 
     // Note the state change in scheduler.
-    KScheduler::OnThreadStateChanged(kernel, this, old_state);
+    if (new_state != old_state) {
+        KScheduler::OnThreadStateChanged(kernel, this, old_state);
+    }
 }
 
 void KThread::Continue() {
@@ -998,13 +996,16 @@ ResultCode KThread::Run() {
 
         // If the current thread has been asked to suspend, suspend it and retry.
         if (GetCurrentThread(kernel).IsSuspended()) {
-            GetCurrentThread(kernel).Suspend();
+            GetCurrentThread(kernel).UpdateState();
             continue;
         }
 
         // If we're not a kernel thread and we've been asked to suspend, suspend ourselves.
-        if (IsUserThread() && IsSuspended()) {
-            Suspend();
+        if (KProcess* owner = this->GetOwnerProcess(); owner != nullptr) {
+            if (IsUserThread() && IsSuspended()) {
+                this->UpdateState();
+            }
+            owner->IncrementRunningThreadCount();
         }
 
         // Set our state and finish.
@@ -1019,21 +1020,29 @@ ResultCode KThread::Run() {
 void KThread::Exit() {
     ASSERT(this == GetCurrentThreadPointer(kernel));
 
-    // Release the thread resource hint from parent.
+    // Release the thread resource hint, running thread count from parent.
     if (parent != nullptr) {
         parent->GetResourceLimit()->Release(Kernel::LimitableResource::Threads, 0, 1);
         resource_limit_release_hint = true;
+        parent->DecrementRunningThreadCount();
     }
 
     // Perform termination.
     {
         KScopedSchedulerLock sl{kernel};
 
+        // Disallow all suspension.
+        suspend_allowed_flags = 0;
+        this->UpdateState();
+
         // Disallow all suspension.
         suspend_allowed_flags = 0;
 
         // Start termination.
         StartTermination();
+
+        // Register the thread as a work task.
+        KWorkerTaskManager::AddTask(kernel, KWorkerTaskManager::WorkerType::Exit, this);
     }
 }
 
@@ -1061,12 +1070,46 @@ ResultCode KThread::Sleep(s64 timeout) {
     return ResultSuccess;
 }
 
+void KThread::IfDummyThreadTryWait() {
+    if (!IsDummyThread()) {
+        return;
+    }
+
+    if (GetState() != ThreadState::Waiting) {
+        return;
+    }
+
+    // Block until we can grab the lock.
+    KScopedSpinLock lk{dummy_wait_lock};
+}
+
+void KThread::IfDummyThreadBeginWait() {
+    if (!IsDummyThread()) {
+        return;
+    }
+
+    // Ensure the thread will block when IfDummyThreadTryWait is called.
+    dummy_wait_lock.Lock();
+}
+
+void KThread::IfDummyThreadEndWait() {
+    if (!IsDummyThread()) {
+        return;
+    }
+
+    // Ensure the thread will no longer block.
+    dummy_wait_lock.Unlock();
+}
+
 void KThread::BeginWait(KThreadQueue* queue) {
     // Set our state as waiting.
     SetState(ThreadState::Waiting);
 
     // Set our wait queue.
     wait_queue = queue;
+
+    // Special case for dummy threads to ensure they block.
+    IfDummyThreadBeginWait();
 }
 
 void KThread::NotifyAvailable(KSynchronizationObject* signaled_object, ResultCode wait_result_) {
@@ -1085,7 +1128,16 @@ void KThread::EndWait(ResultCode wait_result_) {
 
     // If we're waiting, notify our queue that we're available.
     if (GetState() == ThreadState::Waiting) {
+        if (wait_queue == nullptr) {
+            // This should never happen, but avoid a hard crash below to get this logged.
+            ASSERT_MSG(false, "wait_queue is nullptr!");
+            return;
+        }
+
         wait_queue->EndWait(this, wait_result_);
+
+        // Special case for dummy threads to wakeup if necessary.
+        IfDummyThreadEndWait();
     }
 }
 

src/core/hle/kernel/k_thread.h

@@ -19,6 +19,7 @@
 #include "core/hle/kernel/k_light_lock.h"
 #include "core/hle/kernel/k_spin_lock.h"
 #include "core/hle/kernel/k_synchronization_object.h"
+#include "core/hle/kernel/k_worker_task.h"
 #include "core/hle/kernel/slab_helpers.h"
 #include "core/hle/kernel/svc_common.h"
 #include "core/hle/kernel/svc_types.h"
@@ -100,7 +101,7 @@ enum class ThreadWaitReasonForDebugging : u32 {
 [[nodiscard]] KThread& GetCurrentThread(KernelCore& kernel);
 [[nodiscard]] s32 GetCurrentCoreId(KernelCore& kernel);
 
-class KThread final : public KAutoObjectWithSlabHeapAndContainer<KThread, KSynchronizationObject>,
+class KThread final : public KAutoObjectWithSlabHeapAndContainer<KThread, KWorkerTask>,
                       public boost::intrusive::list_base_hook<> {
     KERNEL_AUTOOBJECT_TRAITS(KThread, KSynchronizationObject);
 
@@ -111,6 +112,7 @@ private:
 public:
     static constexpr s32 DefaultThreadPriority = 44;
     static constexpr s32 IdleThreadPriority = Svc::LowestThreadPriority + 1;
+    static constexpr s32 DummyThreadPriority = Svc::LowestThreadPriority + 2;
 
     explicit KThread(KernelCore& kernel_);
     ~KThread() override;
@@ -192,9 +194,9 @@ public:
 
     void TrySuspend();
 
-    void Continue();
+    void UpdateState();
 
-    void Suspend();
+    void Continue();
 
     constexpr void SetSyncedIndex(s32 index) {
         synced_index = index;
@@ -385,6 +387,8 @@ public:
 
     void OnTimer();
 
+    void DoWorkerTaskImpl();
+
     static void PostDestroy(uintptr_t arg);
 
     [[nodiscard]] static ResultCode InitializeDummyThread(KThread* thread);
@@ -550,8 +554,12 @@ public:
         return wait_reason_for_debugging;
     }
 
-    [[nodiscard]] ThreadType GetThreadTypeForDebugging() const {
-        return thread_type_for_debugging;
+    [[nodiscard]] ThreadType GetThreadType() const {
+        return thread_type;
+    }
+
+    [[nodiscard]] bool IsDummyThread() const {
+        return GetThreadType() == ThreadType::Dummy;
     }
 
     void SetWaitObjectsForDebugging(const std::span<KSynchronizationObject*>& objects) {
@@ -628,6 +636,14 @@ public:
         return condvar_key;
     }
 
+    // Dummy threads (used for HLE host threads) cannot wait based on the guest scheduler, and
+    // therefore will not block on guest kernel synchronization primitives. These methods handle
+    // blocking as needed.
+
+    void IfDummyThreadTryWait();
+    void IfDummyThreadBeginWait();
+    void IfDummyThreadEndWait();
+
 private:
     static constexpr size_t PriorityInheritanceCountMax = 10;
     union SyncObjectBuffer {
@@ -679,6 +695,8 @@ private:
 
     void StartTermination();
 
+    void FinishTermination();
+
     [[nodiscard]] ResultCode Initialize(KThreadFunction func, uintptr_t arg, VAddr user_stack_top,
                                         s32 prio, s32 virt_core, KProcess* owner, ThreadType type);
 
@@ -744,16 +762,17 @@ private:
     bool resource_limit_release_hint{};
     StackParameters stack_parameters{};
     KSpinLock context_guard{};
+    KSpinLock dummy_wait_lock{};
 
     // For emulation
     std::shared_ptr<Common::Fiber> host_context{};
     bool is_single_core{};
+    ThreadType thread_type{};
 
     // For debugging
     std::vector<KSynchronizationObject*> wait_objects_for_debugging;
     VAddr mutex_wait_address_for_debugging{};
     ThreadWaitReasonForDebugging wait_reason_for_debugging{};
-    ThreadType thread_type_for_debugging{};
 
 public:
     using ConditionVariableThreadTreeType = ConditionVariableThreadTree;

src/core/hle/kernel/k_worker_task.h

@@ -0,0 +1,18 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/kernel/k_synchronization_object.h"
+
+namespace Kernel {
+
+class KWorkerTask : public KSynchronizationObject {
+public:
+    explicit KWorkerTask(KernelCore& kernel_);
+
+    void DoWorkerTask();
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_worker_task_manager.cpp

@@ -0,0 +1,42 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/hle/kernel/k_process.h"
+#include "core/hle/kernel/k_thread.h"
+#include "core/hle/kernel/k_worker_task.h"
+#include "core/hle/kernel/k_worker_task_manager.h"
+#include "core/hle/kernel/kernel.h"
+
+namespace Kernel {
+
+KWorkerTask::KWorkerTask(KernelCore& kernel_) : KSynchronizationObject{kernel_} {}
+
+void KWorkerTask::DoWorkerTask() {
+    if (auto* const thread = this->DynamicCast<KThread*>(); thread != nullptr) {
+        return thread->DoWorkerTaskImpl();
+    } else {
+        auto* const process = this->DynamicCast<KProcess*>();
+        ASSERT(process != nullptr);
+
+        return process->DoWorkerTaskImpl();
+    }
+}
+
+KWorkerTaskManager::KWorkerTaskManager() : m_waiting_thread(1, "yuzu:KWorkerTaskManager") {}
+
+void KWorkerTaskManager::AddTask(KernelCore& kernel, WorkerType type, KWorkerTask* task) {
+    ASSERT(type <= WorkerType::Count);
+    kernel.WorkerTaskManager().AddTask(kernel, task);
+}
+
+void KWorkerTaskManager::AddTask(KernelCore& kernel, KWorkerTask* task) {
+    KScopedSchedulerLock sl(kernel);
+    m_waiting_thread.QueueWork([task]() {
+        // Do the task.
+        task->DoWorkerTask();
+    });
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_worker_task_manager.h

@@ -0,0 +1,33 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "common/thread_worker.h"
+
+namespace Kernel {
+
+class KernelCore;
+class KWorkerTask;
+
+class KWorkerTaskManager final {
+public:
+    enum class WorkerType : u32 {
+        Exit,
+        Count,
+    };
+
+    KWorkerTaskManager();
+
+    static void AddTask(KernelCore& kernel_, WorkerType type, KWorkerTask* task);
+
+private:
+    void AddTask(KernelCore& kernel, KWorkerTask* task);
+
+private:
+    Common::ThreadWorker m_waiting_thread;
+};
+
+} // namespace Kernel

src/core/hle/kernel/kernel.cpp

@@ -37,6 +37,7 @@
 #include "core/hle/kernel/k_shared_memory.h"
 #include "core/hle/kernel/k_slab_heap.h"
 #include "core/hle/kernel/k_thread.h"
+#include "core/hle/kernel/k_worker_task_manager.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/physical_core.h"
 #include "core/hle/kernel/service_thread.h"
@@ -51,7 +52,8 @@ namespace Kernel {
 
 struct KernelCore::Impl {
     explicit Impl(Core::System& system_, KernelCore& kernel_)
-        : time_manager{system_}, object_list_container{kernel_}, system{system_} {}
+        : time_manager{system_}, object_list_container{kernel_},
+          service_threads_manager{1, "yuzu:ServiceThreadsManager"}, system{system_} {}
 
     void SetMulticore(bool is_multi) {
         is_multicore = is_multi;
@@ -299,12 +301,10 @@ struct KernelCore::Impl {
     // Gets the dummy KThread for the caller, allocating a new one if this is the first time
     KThread* GetHostDummyThread() {
         auto make_thread = [this]() {
-            std::lock_guard lk(dummy_thread_lock);
-            auto& thread = dummy_threads.emplace_back(std::make_unique<KThread>(system.Kernel()));
-            KAutoObject::Create(thread.get());
-            ASSERT(KThread::InitializeDummyThread(thread.get()).IsSuccess());
+            KThread* thread = KThread::Create(system.Kernel());
+            ASSERT(KThread::InitializeDummyThread(thread).IsSuccess());
             thread->SetName(fmt::format("DummyThread:{}", GetHostThreadId()));
-            return thread.get();
+            return thread;
         };
 
         thread_local KThread* saved_thread = make_thread();
@@ -707,32 +707,28 @@ struct KernelCore::Impl {
     std::weak_ptr<Kernel::ServiceThread> CreateServiceThread(KernelCore& kernel,
                                                              const std::string& name) {
         auto service_thread = std::make_shared<Kernel::ServiceThread>(kernel, 1, name);
-        {
-            std::lock_guard lk(service_threads_lock);
-            service_threads.emplace(service_thread);
-        }
+
+        service_threads_manager.QueueWork(
+            [this, service_thread]() { service_threads.emplace(service_thread); });
+
         return service_thread;
     }
 
     void ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread) {
-        auto strong_ptr = service_thread.lock();
-        {
-            std::lock_guard lk(service_threads_lock);
-            service_threads.erase(strong_ptr);
+        if (auto strong_ptr = service_thread.lock()) {
+            service_threads_manager.QueueWork(
+                [this, strong_ptr{std::move(strong_ptr)}]() { service_threads.erase(strong_ptr); });
         }
     }
 
     void ClearServiceThreads() {
-        std::lock_guard lk(service_threads_lock);
-        service_threads.clear();
+        service_threads_manager.QueueWork([this]() { service_threads.clear(); });
     }
 
     std::mutex server_ports_lock;
     std::mutex server_sessions_lock;
     std::mutex registered_objects_lock;
     std::mutex registered_in_use_objects_lock;
-    std::mutex dummy_thread_lock;
-    std::mutex service_threads_lock;
 
     std::atomic<u32> next_object_id{0};
     std::atomic<u64> next_kernel_process_id{KProcess::InitialKIPIDMin};
@@ -783,14 +779,12 @@ struct KernelCore::Impl {
 
     // Threads used for services
     std::unordered_set<std::shared_ptr<Kernel::ServiceThread>> service_threads;
+    Common::ThreadWorker service_threads_manager;
 
     std::array<KThread*, Core::Hardware::NUM_CPU_CORES> suspend_threads;
     std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES> interrupts{};
     std::array<std::unique_ptr<Kernel::KScheduler>, Core::Hardware::NUM_CPU_CORES> schedulers{};
 
-    // Specifically tracked to be automatically destroyed with kernel
-    std::vector<std::unique_ptr<KThread>> dummy_threads;
-
     bool is_multicore{};
     std::atomic_bool is_shutting_down{};
     bool is_phantom_mode_for_singlecore{};
@@ -798,6 +792,8 @@ struct KernelCore::Impl {
 
     std::array<u64, Core::Hardware::NUM_CPU_CORES> svc_ticks{};
 
+    KWorkerTaskManager worker_task_manager;
+
     // System context
     Core::System& system;
 };
@@ -1138,6 +1134,14 @@ const Init::KSlabResourceCounts& KernelCore::SlabResourceCounts() const {
     return impl->slab_resource_counts;
 }
 
+KWorkerTaskManager& KernelCore::WorkerTaskManager() {
+    return impl->worker_task_manager;
+}
+
+const KWorkerTaskManager& KernelCore::WorkerTaskManager() const {
+    return impl->worker_task_manager;
+}
+
 bool KernelCore::IsPhantomModeForSingleCore() const {
     return impl->IsPhantomModeForSingleCore();
 }

src/core/hle/kernel/kernel.h

@@ -52,6 +52,7 @@ class KSharedMemory;
 class KSharedMemoryInfo;
 class KThread;
 class KTransferMemory;
+class KWorkerTaskManager;
 class KWritableEvent;
 class KCodeMemory;
 class PhysicalCore;
@@ -343,6 +344,12 @@ public:
     /// Gets the current slab resource counts.
     const Init::KSlabResourceCounts& SlabResourceCounts() const;
 
+    /// Gets the current worker task manager, used for dispatching KThread/KProcess tasks.
+    KWorkerTaskManager& WorkerTaskManager();
+
+    /// Gets the current worker task manager, used for dispatching KThread/KProcess tasks.
+    const KWorkerTaskManager& WorkerTaskManager() const;
+
 private:
     friend class KProcess;
     friend class KThread;

src/core/hle/kernel/service_thread.cpp

@@ -12,6 +12,7 @@
 #include "common/scope_exit.h"
 #include "common/thread.h"
 #include "core/hle/kernel/k_session.h"
+#include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/service_thread.h"
 
@@ -50,6 +51,10 @@ ServiceThread::Impl::Impl(KernelCore& kernel, std::size_t num_threads, const std
 
             kernel.RegisterHostThread();
 
+            // Ensure the dummy thread allocated for this host thread is closed on exit.
+            auto* dummy_thread = kernel.GetCurrentEmuThread();
+            SCOPE_EXIT({ dummy_thread->Close(); });
+
             while (true) {
                 std::function<void()> task;
 

src/core/hle/kernel/svc.cpp

@@ -230,7 +230,7 @@ static ResultCode MapMemory(Core::System& system, VAddr dst_addr, VAddr src_addr
         return result;
     }
 
-    return page_table.Map(dst_addr, src_addr, size);
+    return page_table.MapMemory(dst_addr, src_addr, size);
 }
 
 static ResultCode MapMemory32(Core::System& system, u32 dst_addr, u32 src_addr, u32 size) {
@@ -249,7 +249,7 @@ static ResultCode UnmapMemory(Core::System& system, VAddr dst_addr, VAddr src_ad
         return result;
     }
 
-    return page_table.Unmap(dst_addr, src_addr, size);
+    return page_table.UnmapMemory(dst_addr, src_addr, size);
 }
 
 static ResultCode UnmapMemory32(Core::System& system, u32 dst_addr, u32 src_addr, u32 size) {
@@ -2613,7 +2613,7 @@ static const FunctionDef SVC_Table_32[] = {
     {0x33, SvcWrap32<GetThreadContext32>, "GetThreadContext32"},
     {0x34, SvcWrap32<WaitForAddress32>, "WaitForAddress32"},
     {0x35, SvcWrap32<SignalToAddress32>, "SignalToAddress32"},
-    {0x36, nullptr, "Unknown"},
+    {0x36, SvcWrap32<SynchronizePreemptionState>, "SynchronizePreemptionState32"},
     {0x37, nullptr, "Unknown"},
     {0x38, nullptr, "Unknown"},
     {0x39, nullptr, "Unknown"},

src/core/hle/service/acc/acc_su.cpp

@@ -37,8 +37,8 @@ ACC_SU::ACC_SU(std::shared_ptr<Module> module_, std::shared_ptr<ProfileManager>
         {130, nullptr, "ActivateOpenContextRetention"},
         {140, &ACC_SU::ListQualifiedUsers, "ListQualifiedUsers"},
         {150, nullptr, "AuthenticateApplicationAsync"},
-        {151, nullptr, "Unknown151"},
-        {152, nullptr, "Unknown152"},
+        {151, nullptr, "EnsureSignedDeviceIdentifierCacheForNintendoAccountAsync"},
+        {152, nullptr, "LoadSignedDeviceIdentifierCacheForNintendoAccount"},
         {190, nullptr, "GetUserLastOpenedApplication"},
         {191, nullptr, "ActivateOpenContextHolder"},
         {200, nullptr, "BeginUserRegistration"},

src/core/hle/service/acc/acc_u1.cpp

@@ -37,8 +37,8 @@ ACC_U1::ACC_U1(std::shared_ptr<Module> module_, std::shared_ptr<ProfileManager>
         {130, nullptr, "ActivateOpenContextRetention"},
         {140, &ACC_U1::ListQualifiedUsers, "ListQualifiedUsers"},
         {150, nullptr, "AuthenticateApplicationAsync"},
-        {151, nullptr, "Unknown151"},
-        {152, nullptr, "Unknown152"},
+        {151, nullptr, "EnsureSignedDeviceIdentifierCacheForNintendoAccountAsync"},
+        {152, nullptr, "LoadSignedDeviceIdentifierCacheForNintendoAccount"},
         {190, nullptr, "GetUserLastOpenedApplication"},
         {191, nullptr, "ActivateOpenContextHolder"},
         {997, nullptr, "DebugInvalidateTokenCacheForUser"},

src/core/hle/service/am/omm.cpp

@@ -37,6 +37,7 @@ OMM::OMM(Core::System& system_) : ServiceFramework{system_, "omm"} {
         {25, nullptr, "SetApplicationCecSettingsAndNotifyChanged"},
         {26, nullptr, "GetOperationModeSystemInfo"},
         {27, nullptr, "GetAppletFullAwakingSystemEvent"},
+        {28, nullptr, "CreateCradleFirmwareUpdater"},
     };
     // clang-format on
 

src/core/hle/service/apm/apm_interface.cpp

@@ -17,7 +17,7 @@ public:
         static const FunctionInfo functions[] = {
             {0, &ISession::SetPerformanceConfiguration, "SetPerformanceConfiguration"},
             {1, &ISession::GetPerformanceConfiguration, "GetPerformanceConfiguration"},
-            {2, nullptr, "SetCpuOverclockEnabled"},
+            {2, &ISession::SetCpuOverclockEnabled, "SetCpuOverclockEnabled"},
         };
         RegisterHandlers(functions);
     }
@@ -47,6 +47,18 @@ private:
         rb.PushEnum(controller.GetCurrentPerformanceConfiguration(mode));
     }
 
+    void SetCpuOverclockEnabled(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+
+        const auto cpu_overclock_enabled = rp.Pop<bool>();
+
+        LOG_WARNING(Service_APM, "(STUBBED) called, cpu_overclock_enabled={}",
+                    cpu_overclock_enabled);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+
     Controller& controller;
 };
 

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

@@ -41,14 +41,14 @@ AudCtl::AudCtl(Core::System& system_) : ServiceFramework{system_, "audctl"} {
         {27, nullptr, "SetVolumeMappingTableForDev"},
         {28, nullptr, "GetAudioOutputChannelCountForPlayReport"},
         {29, nullptr, "BindAudioOutputChannelCountUpdateEventForPlayReport"},
-        {30, nullptr, "Unknown30"},
-        {31, nullptr, "Unknown31"},
-        {32, nullptr, "Unknown32"},
-        {33, nullptr, "Unknown33"},
-        {34, nullptr, "Unknown34"},
-        {10000, nullptr, "Unknown10000"},
-        {10001, nullptr, "Unknown10001"},
-        {10002, nullptr, "Unknown10002"},
+        {30, nullptr, "SetSpeakerAutoMuteEnabled"},
+        {31, nullptr, "IsSpeakerAutoMuteEnabled"},
+        {32, nullptr, "GetActiveOutputTarget"},
+        {33, nullptr, "GetTargetDeviceInfo"},
+        {34, nullptr, "AcquireTargetNotification"},
+        {10000, nullptr, "NotifyAudioOutputTargetForPlayReport"},
+        {10001, nullptr, "NotifyAudioOutputChannelCountForPlayReport"},
+        {10002, nullptr, "NotifyUnsupportedUsbOutputDeviceAttachedForPlayReport"},
     };
     // clang-format on
 

src/core/hle/service/btm/btm.cpp

@@ -201,6 +201,22 @@ public:
             {62, nullptr, "Unknown62"},
             {63, nullptr, "Unknown63"},
             {64, nullptr, "Unknown64"},
+            {65, nullptr, "Unknown65"},
+            {66, nullptr, "Unknown66"},
+            {67, nullptr, "Unknown67"},
+            {68, nullptr, "Unknown68"},
+            {69, nullptr, "Unknown69"},
+            {70, nullptr, "Unknown70"},
+            {71, nullptr, "Unknown71"},
+            {72, nullptr, "Unknown72"},
+            {73, nullptr, "Unknown73"},
+            {74, nullptr, "Unknown74"},
+            {75, nullptr, "Unknown75"},
+            {76, nullptr, "Unknown76"},
+            {100, nullptr, "Unknown100"},
+            {101, nullptr, "Unknown101"},
+            {110, nullptr, "Unknown102"},
+            {111, nullptr, "Unknown103"},
         };
         // clang-format on
 
@@ -249,6 +265,20 @@ public:
             {7, nullptr, "AcquireRadioEvent"},
             {8, nullptr, "AcquireGamepadPairingEvent"},
             {9, nullptr, "IsGamepadPairingStarted"},
+            {10, nullptr, "StartAudioDeviceDiscovery"},
+            {11, nullptr, "StopAudioDeviceDiscovery"},
+            {12, nullptr, "IsDiscoveryingAudioDevice"},
+            {13, nullptr, "GetDiscoveredAudioDevice"},
+            {14, nullptr, "AcquireAudioDeviceConnectionEvent"},
+            {15, nullptr, "ConnectAudioDevice"},
+            {16, nullptr, "IsConnectingAudioDevice"},
+            {17, nullptr, "GetConnectedAudioDevices"},
+            {18, nullptr, "DisconnectAudioDevice"},
+            {19, nullptr, "AcquirePairedAudioDeviceInfoChangedEvent"},
+            {20, nullptr, "GetPairedAudioDevices"},
+            {21, nullptr, "RemoveAudioDevicePairing"},
+            {22, nullptr, "RequestAudioDeviceConnectionRejection"},
+            {23, nullptr, "CancelAudioDeviceConnectionRejection"}
         };
         // clang-format on
 

src/core/hle/service/filesystem/fsp_srv.cpp

@@ -744,6 +744,7 @@ FSP_SRV::FSP_SRV(Core::System& system_)
         {203, &FSP_SRV::OpenPatchDataStorageByCurrentProcess, "OpenPatchDataStorageByCurrentProcess"},
         {204, nullptr, "OpenDataFileSystemByProgramIndex"},
         {205, &FSP_SRV::OpenDataStorageWithProgramIndex, "OpenDataStorageWithProgramIndex"},
+        {206, nullptr, "OpenDataStorageByPath"},
         {400, nullptr, "OpenDeviceOperator"},
         {500, nullptr, "OpenSdCardDetectionEventNotifier"},
         {501, nullptr, "OpenGameCardDetectionEventNotifier"},
@@ -796,6 +797,8 @@ FSP_SRV::FSP_SRV(Core::System& system_)
         {1014, nullptr, "OutputMultiProgramTagAccessLog"},
         {1016, nullptr, "FlushAccessLogOnSdCard"},
         {1017, nullptr, "OutputApplicationInfoAccessLog"},
+        {1018, nullptr, "SetDebugOption"},
+        {1019, nullptr, "UnsetDebugOption"},
         {1100, nullptr, "OverrideSaveDataTransferTokenSignVerificationKey"},
         {1110, nullptr, "CorruptSaveDataFileSystemBySaveDataSpaceId2"},
         {1200, &FSP_SRV::OpenMultiCommitManager, "OpenMultiCommitManager"},

src/core/hle/service/friend/friend.cpp

@@ -27,13 +27,13 @@ public:
             {10101, &IFriendService::GetFriendList, "GetFriendList"},
             {10102, nullptr, "UpdateFriendInfo"},
             {10110, nullptr, "GetFriendProfileImage"},
-            {10120, nullptr, "Unknown10120"},
-            {10121, nullptr, "Unknown10121"},
+            {10120, nullptr, "IsFriendListCacheAvailable"},
+            {10121, nullptr, "EnsureFriendListAvailable"},
             {10200, nullptr, "SendFriendRequestForApplication"},
             {10211, nullptr, "AddFacedFriendRequestForApplication"},
             {10400, &IFriendService::GetBlockedUserListIds, "GetBlockedUserListIds"},
-            {10420, nullptr, "Unknown10420"},
-            {10421, nullptr, "Unknown10421"},
+            {10420, nullptr, "IsBlockedUserListCacheAvailable"},
+            {10421, nullptr, "EnsureBlockedUserListAvailable"},
             {10500, nullptr, "GetProfileList"},
             {10600, nullptr, "DeclareOpenOnlinePlaySession"},
             {10601, &IFriendService::DeclareCloseOnlinePlaySession, "DeclareCloseOnlinePlaySession"},
@@ -103,8 +103,8 @@ public:
             {30900, nullptr, "SendFriendInvitation"},
             {30910, nullptr, "ReadFriendInvitation"},
             {30911, nullptr, "ReadAllFriendInvitations"},
-            {40100, nullptr, "Unknown40100"},
-            {40400, nullptr, "Unknown40400"},
+            {40100, nullptr, "DeleteFriendListCache"},
+            {40400, nullptr, "DeleteBlockedUserListCache"},
             {49900, nullptr, "DeleteNetworkServiceAccountCache"},
         };
         // clang-format on

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

@@ -1176,7 +1176,8 @@ void Hid::SetNpadAnalogStickUseCenterClamp(Kernel::HLERequestContext& ctx) {
 
     const auto parameters{rp.PopRaw<Parameters>()};
 
-    applet_resource->GetController<Controller_NPad>(HidController::NPad)
+    GetAppletResource()
+        ->GetController<Controller_NPad>(HidController::NPad)
         .SetAnalogStickUseCenterClamp(parameters.analog_stick_use_center_clamp);
 
     LOG_WARNING(Service_HID,

src/core/hle/service/nim/nim.cpp

@@ -211,6 +211,11 @@ public:
             {127, nullptr, "Unknown127"},
             {128, nullptr, "Unknown128"},
             {129, nullptr, "Unknown129"},
+            {130, nullptr, "Unknown130"},
+            {131, nullptr, "Unknown131"},
+            {132, nullptr, "Unknown132"},
+            {133, nullptr, "Unknown133"},
+            {134, nullptr, "Unknown134"},
         };
         // clang-format on
 
@@ -287,6 +292,7 @@ public:
             {502, nullptr, "RequestDownloadTicketForPrepurchasedContents"},
             {503, nullptr, "RequestSyncTicket"},
             {504, nullptr, "RequestDownloadTicketForPrepurchasedContents2"},
+            {505, nullptr, "RequestDownloadTicketForPrepurchasedContentsForAccount"},
         };
         // clang-format on
 

src/core/hle/service/ns/ns.cpp

@@ -158,6 +158,7 @@ IApplicationManagerInterface::IApplicationManagerInterface(Core::System& system_
         {605, nullptr, "ListApplicationContentMetaStatusWithRightsCheck"},
         {606, nullptr, "GetContentMetaStorage"},
         {607, nullptr, "ListAvailableAddOnContent"},
+        {609, nullptr, "ListAvailabilityAssuredAddOnContent"},
         {700, nullptr, "PushDownloadTaskList"},
         {701, nullptr, "ClearTaskStatusList"},
         {702, nullptr, "RequestDownloadTaskList"},
@@ -289,6 +290,11 @@ IApplicationManagerInterface::IApplicationManagerInterface(Core::System& system_
         {2514, nullptr, "ClearTaskOfAsyncTaskManager"},
         {2515, nullptr, "CleanupAllPlaceHolderAndFragmentsIfNoTask"},
         {2516, nullptr, "EnsureApplicationCertificate"},
+        {2517, nullptr, "CreateApplicationInstance"},
+        {2518, nullptr, "UpdateQualificationForDebug"},
+        {2519, nullptr, "IsQualificationTransitionSupported"},
+        {2520, nullptr, "IsQualificationTransitionSupportedByProcessId"},
+        {2521, nullptr, "GetRightsUserChangedEvent"},
         {2800, nullptr, "GetApplicationIdOfPreomia"},
         {3000, nullptr, "RegisterDeviceLockKey"},
         {3001, nullptr, "UnregisterDeviceLockKey"},

src/core/hle/service/set/set_sys.cpp

@@ -307,6 +307,8 @@ SET_SYS::SET_SYS(Core::System& system_) : ServiceFramework{system_, "set:sys"} {
         {202, nullptr, "SetFieldTestingFlag"},
         {203, nullptr, "GetPanelCrcMode"},
         {204, nullptr, "SetPanelCrcMode"},
+        {205, nullptr, "GetNxControllerSettingsEx"},
+        {206, nullptr, "SetNxControllerSettingsEx"},
     };
     // clang-format on
 

src/core/hle/service/usb/usb.cpp

@@ -17,19 +17,19 @@ public:
     explicit IDsInterface(Core::System& system_) : ServiceFramework{system_, "IDsInterface"} {
         // clang-format off
         static const FunctionInfo functions[] = {
-            {0, nullptr, "GetDsEndpoint"},
-            {1, nullptr, "GetSetupEvent"},
-            {2, nullptr, "Unknown2"},
-            {3, nullptr, "EnableInterface"},
-            {4, nullptr, "DisableInterface"},
-            {5, nullptr, "CtrlInPostBufferAsync"},
-            {6, nullptr, "CtrlOutPostBufferAsync"},
-            {7, nullptr, "GetCtrlInCompletionEvent"},
-            {8, nullptr, "GetCtrlInReportData"},
-            {9, nullptr, "GetCtrlOutCompletionEvent"},
-            {10, nullptr, "GetCtrlOutReportData"},
-            {11, nullptr, "StallCtrl"},
-            {12, nullptr, "AppendConfigurationData"},
+            {0, nullptr, "BindDevice"},
+            {1, nullptr, "BindClientProcess"},
+            {2, nullptr, "AddInterface"},
+            {3, nullptr, "GetStateChangeEvent"},
+            {4, nullptr, "GetState"},
+            {5, nullptr, "ClearDeviceData"},
+            {6, nullptr, "AddUsbStringDescriptor"},
+            {7, nullptr, "DeleteUsbStringDescriptor"},
+            {8, nullptr, "SetUsbDeviceDescriptor"},
+            {9, nullptr, "SetBinaryObjectStore"},
+            {10, nullptr, "Enable"},
+            {11, nullptr, "Disable"},
+            {12, nullptr, "Unknown12"},
         };
         // clang-format on
 
@@ -42,19 +42,7 @@ public:
     explicit USB_DS(Core::System& system_) : ServiceFramework{system_, "usb:ds"} {
         // clang-format off
         static const FunctionInfo functions[] = {
-            {0, nullptr, "BindDevice"},
-            {1, nullptr, "BindClientProcess"},
-            {2, nullptr, "GetDsInterface"},
-            {3, nullptr, "GetStateChangeEvent"},
-            {4, nullptr, "GetState"},
-            {5, nullptr, "ClearDeviceData"},
-            {6, nullptr, "AddUsbStringDescriptor"},
-            {7, nullptr, "DeleteUsbStringDescriptor"},
-            {8, nullptr, "SetUsbDeviceDescriptor"},
-            {9, nullptr, "SetBinaryObjectStore"},
-            {10, nullptr, "Enable"},
-            {11, nullptr, "Disable"},
-            {12, nullptr, "Unknown12"},
+            {0, nullptr, "OpenDsService"},
         };
         // clang-format on
 
@@ -120,7 +108,7 @@ public:
             {5, nullptr, "DestroyInterfaceAvailableEvent"},
             {6, nullptr, "GetInterfaceStateChangeEvent"},
             {7, nullptr, "AcquireUsbIf"},
-            {8, nullptr, "Unknown8"},
+            {8, nullptr, "ResetDevice"},
         };
         // clang-format on
 

src/core/hle/service/wlan/wlan.cpp

@@ -53,6 +53,7 @@ public:
             {35, nullptr, "Unknown35"},
             {36, nullptr, "Unknown36"},
             {37, nullptr, "Unknown37"},
+            {38, nullptr, "Unknown38"},
         };
         // clang-format on
 
@@ -117,7 +118,6 @@ public:
             {49, nullptr, "Unknown49"},
             {50, nullptr, "Unknown50"},
             {51, nullptr, "Unknown51"},
-            {52, nullptr, "Unknown52"},
         };
         // clang-format on
 

src/input_common/drivers/mouse.cpp

@@ -16,6 +16,7 @@ constexpr int mouse_axis_x = 0;
 constexpr int mouse_axis_y = 1;
 constexpr int wheel_axis_x = 2;
 constexpr int wheel_axis_y = 3;
+constexpr int motion_wheel_y = 4;
 constexpr int touch_axis_x = 10;
 constexpr int touch_axis_y = 11;
 constexpr PadIdentifier identifier = {
@@ -30,8 +31,9 @@ Mouse::Mouse(std::string input_engine_) : InputEngine(std::move(input_engine_))
     PreSetAxis(identifier, mouse_axis_y);
     PreSetAxis(identifier, wheel_axis_x);
     PreSetAxis(identifier, wheel_axis_y);
+    PreSetAxis(identifier, motion_wheel_y);
     PreSetAxis(identifier, touch_axis_x);
-    PreSetAxis(identifier, touch_axis_x);
+    PreSetAxis(identifier, touch_axis_y);
     update_thread = std::jthread([this](std::stop_token stop_token) { UpdateThread(stop_token); });
 }
 
@@ -48,6 +50,8 @@ void Mouse::UpdateThread(std::stop_token stop_token) {
             SetAxis(identifier, mouse_axis_y, -last_mouse_change.y * sensitivity);
         }
 
+        SetAxis(identifier, motion_wheel_y, 0.0f);
+
         if (mouse_panning_timout++ > 20) {
             StopPanning();
         }
@@ -136,6 +140,7 @@ void Mouse::MouseWheelChange(int x, int y) {
     wheel_position.y += y;
     SetAxis(identifier, wheel_axis_x, static_cast<f32>(wheel_position.x));
     SetAxis(identifier, wheel_axis_y, static_cast<f32>(wheel_position.y));
+    SetAxis(identifier, motion_wheel_y, static_cast<f32>(y) / 100.0f);
 }
 
 void Mouse::ReleaseAllButtons() {
@@ -171,13 +176,39 @@ AnalogMapping Mouse::GetAnalogMappingForDevice(
     return mapping;
 }
 
+Common::Input::ButtonNames Mouse::GetUIButtonName(const Common::ParamPackage& params) const {
+    const auto button = static_cast<MouseButton>(params.Get("button", 0));
+    switch (button) {
+    case MouseButton::Left:
+        return Common::Input::ButtonNames::ButtonLeft;
+    case MouseButton::Right:
+        return Common::Input::ButtonNames::ButtonRight;
+    case MouseButton::Wheel:
+        return Common::Input::ButtonNames::ButtonMouseWheel;
+    case MouseButton::Backward:
+        return Common::Input::ButtonNames::ButtonBackward;
+    case MouseButton::Forward:
+        return Common::Input::ButtonNames::ButtonForward;
+    case MouseButton::Task:
+        return Common::Input::ButtonNames::ButtonTask;
+    case MouseButton::Extra:
+        return Common::Input::ButtonNames::ButtonExtra;
+    case MouseButton::Undefined:
+    default:
+        return Common::Input::ButtonNames::Undefined;
+    }
+}
+
 Common::Input::ButtonNames Mouse::GetUIName(const Common::ParamPackage& params) const {
     if (params.Has("button")) {
-        return Common::Input::ButtonNames::Value;
+        return GetUIButtonName(params);
     }
     if (params.Has("axis")) {
         return Common::Input::ButtonNames::Value;
     }
+    if (params.Has("axis_x") && params.Has("axis_y") && params.Has("axis_z")) {
+        return Common::Input::ButtonNames::Engine;
+    }
 
     return Common::Input::ButtonNames::Invalid;
 }

src/input_common/drivers/mouse.h

@@ -69,6 +69,8 @@ private:
     void UpdateThread(std::stop_token stop_token);
     void StopPanning();
 
+    Common::Input::ButtonNames GetUIButtonName(const Common::ParamPackage& params) const;
+
     Common::Vec2<int> mouse_origin;
     Common::Vec2<int> last_mouse_position;
     Common::Vec2<float> last_mouse_change;

src/input_common/drivers/sdl_driver.cpp

@@ -109,8 +109,9 @@ public:
 
     bool HasHDRumble() const {
         if (sdl_controller) {
-            return (SDL_GameControllerGetType(sdl_controller.get()) ==
-                    SDL_CONTROLLER_TYPE_NINTENDO_SWITCH_PRO);
+            const auto type = SDL_GameControllerGetType(sdl_controller.get());
+            return (type == SDL_CONTROLLER_TYPE_NINTENDO_SWITCH_PRO) ||
+                   (type == SDL_CONTROLLER_TYPE_PS5);
         }
         return false;
     }
@@ -198,9 +199,15 @@ public:
         if (sdl_controller) {
             switch (SDL_GameControllerGetType(sdl_controller.get())) {
             case SDL_CONTROLLER_TYPE_XBOX360:
-                return "XBox 360 Controller";
+                return "Xbox 360 Controller";
             case SDL_CONTROLLER_TYPE_XBOXONE:
-                return "XBox One Controller";
+                return "Xbox One Controller";
+            case SDL_CONTROLLER_TYPE_PS3:
+                return "DualShock 3 Controller";
+            case SDL_CONTROLLER_TYPE_PS4:
+                return "DualShock 4 Controller";
+            case SDL_CONTROLLER_TYPE_PS5:
+                return "DualSense Controller";
             default:
                 break;
             }

src/input_common/drivers/udp_client.cpp

@@ -192,6 +192,25 @@ std::size_t UDPClient::GetClientNumber(std::string_view host, u16 port) const {
     return MAX_UDP_CLIENTS;
 }
 
+BatteryLevel UDPClient::GetBatteryLevel(Response::Battery battery) const {
+    switch (battery) {
+    case Response::Battery::Dying:
+        return BatteryLevel::Empty;
+    case Response::Battery::Low:
+        return BatteryLevel::Critical;
+    case Response::Battery::Medium:
+        return BatteryLevel::Low;
+    case Response::Battery::High:
+        return BatteryLevel::Medium;
+    case Response::Battery::Full:
+    case Response::Battery::Charged:
+        return BatteryLevel::Full;
+    case Response::Battery::Charging:
+    default:
+        return BatteryLevel::Charging;
+    }
+}
+
 void UDPClient::OnVersion([[maybe_unused]] Response::Version data) {
     LOG_TRACE(Input, "Version packet received: {}", data.version);
 }
@@ -299,6 +318,8 @@ void UDPClient::OnPadData(Response::PadData data, std::size_t client) {
         const int button = static_cast<int>(buttons[i]);
         SetButton(identifier, button, button_status);
     }
+
+    SetBattery(identifier, GetBatteryLevel(data.info.battery));
 }
 
 void UDPClient::StartCommunication(std::size_t client, const std::string& host, u16 port) {
@@ -318,7 +339,7 @@ void UDPClient::StartCommunication(std::size_t client, const std::string& host,
     }
 }
 
-const PadIdentifier UDPClient::GetPadIdentifier(std::size_t pad_index) const {
+PadIdentifier UDPClient::GetPadIdentifier(std::size_t pad_index) const {
     const std::size_t client = pad_index / PADS_PER_CLIENT;
     return {
         .guid = clients[client].uuid,
@@ -327,9 +348,9 @@ const PadIdentifier UDPClient::GetPadIdentifier(std::size_t pad_index) const {
     };
 }
 
-const Common::UUID UDPClient::GetHostUUID(const std::string host) const {
-    const auto ip = boost::asio::ip::address_v4::from_string(host);
-    const auto hex_host = fmt::format("{:06x}", ip.to_ulong());
+Common::UUID UDPClient::GetHostUUID(const std::string& host) const {
+    const auto ip = boost::asio::ip::make_address_v4(host);
+    const auto hex_host = fmt::format("{:06x}", ip.to_uint());
     return Common::UUID{hex_host};
 }
 

src/input_common/drivers/udp_client.h

@@ -15,6 +15,7 @@ namespace InputCommon::CemuhookUDP {
 class Socket;
 
 namespace Response {
+enum class Battery : u8;
 struct PadData;
 struct PortInfo;
 struct TouchPad;
@@ -137,12 +138,15 @@ private:
     // Translates configuration to client number
     std::size_t GetClientNumber(std::string_view host, u16 port) const;
 
+    // Translates UDP battery level to input engine battery level
+    BatteryLevel GetBatteryLevel(Response::Battery battery) const;
+
     void OnVersion(Response::Version);
     void OnPortInfo(Response::PortInfo);
     void OnPadData(Response::PadData, std::size_t client);
     void StartCommunication(std::size_t client, const std::string& host, u16 port);
-    const PadIdentifier GetPadIdentifier(std::size_t pad_index) const;
-    const Common::UUID GetHostUUID(const std::string host) const;
+    PadIdentifier GetPadIdentifier(std::size_t pad_index) const;
+    Common::UUID GetHostUUID(const std::string& host) const;
 
     Common::Input::ButtonNames GetUIButtonName(const Common::ParamPackage& params) const;
 

src/input_common/input_engine.cpp

@@ -173,7 +173,7 @@ void InputEngine::ResetButtonState() {
             SetButton(controller.first, button.first, false);
         }
         for (const auto& button : controller.second.hat_buttons) {
-            SetHatButton(controller.first, button.first, false);
+            SetHatButton(controller.first, button.first, 0);
         }
     }
 }

src/input_common/input_engine.h

@@ -16,7 +16,7 @@
 
 // Pad Identifier of data source
 struct PadIdentifier {
-    Common::UUID guid{};
+    Common::UUID guid{Common::INVALID_UUID};
     std::size_t port{};
     std::size_t pad{};
 
@@ -89,7 +89,7 @@ struct UpdateCallback {
 
 // Triggered if data changed on the controller and the engine is on configuring mode
 struct MappingCallback {
-    std::function<void(MappingData)> on_data;
+    std::function<void(const MappingData&)> on_data;
 };
 
 // Input Identifier of data source

src/input_common/input_mapping.cpp

@@ -2,14 +2,13 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included
 
-#include "common/common_types.h"
 #include "common/settings.h"
 #include "input_common/input_engine.h"
 #include "input_common/input_mapping.h"
 
 namespace InputCommon {
 
-MappingFactory::MappingFactory() {}
+MappingFactory::MappingFactory() = default;
 
 void MappingFactory::BeginMapping(Polling::InputType type) {
     is_enabled = true;
@@ -19,7 +18,7 @@ void MappingFactory::BeginMapping(Polling::InputType type) {
     second_axis = -1;
 }
 
-[[nodiscard]] const Common::ParamPackage MappingFactory::GetNextInput() {
+Common::ParamPackage MappingFactory::GetNextInput() {
     Common::ParamPackage input;
     input_queue.Pop(input);
     return input;
@@ -57,7 +56,7 @@ void MappingFactory::StopMapping() {
 void MappingFactory::RegisterButton(const MappingData& data) {
     Common::ParamPackage new_input;
     new_input.Set("engine", data.engine);
-    if (data.pad.guid != Common::UUID{}) {
+    if (data.pad.guid.IsValid()) {
         new_input.Set("guid", data.pad.guid.Format());
     }
     new_input.Set("port", static_cast<int>(data.pad.port));
@@ -93,7 +92,7 @@ void MappingFactory::RegisterButton(const MappingData& data) {
 void MappingFactory::RegisterStick(const MappingData& data) {
     Common::ParamPackage new_input;
     new_input.Set("engine", data.engine);
-    if (data.pad.guid != Common::UUID{}) {
+    if (data.pad.guid.IsValid()) {
         new_input.Set("guid", data.pad.guid.Format());
     }
     new_input.Set("port", static_cast<int>(data.pad.port));
@@ -138,11 +137,24 @@ void MappingFactory::RegisterStick(const MappingData& data) {
 void MappingFactory::RegisterMotion(const MappingData& data) {
     Common::ParamPackage new_input;
     new_input.Set("engine", data.engine);
-    if (data.pad.guid != Common::UUID{}) {
+    if (data.pad.guid.IsValid()) {
         new_input.Set("guid", data.pad.guid.Format());
     }
     new_input.Set("port", static_cast<int>(data.pad.port));
     new_input.Set("pad", static_cast<int>(data.pad.pad));
+
+    // If engine is mouse map the mouse position as 3 axis motion
+    if (data.engine == "mouse") {
+        new_input.Set("axis_x", 1);
+        new_input.Set("invert_x", "-");
+        new_input.Set("axis_y", 0);
+        new_input.Set("axis_z", 4);
+        new_input.Set("range", 1.0f);
+        new_input.Set("deadzone", 0.0f);
+        input_queue.Push(new_input);
+        return;
+    }
+
     switch (data.type) {
     case EngineInputType::Button:
     case EngineInputType::HatButton:

src/input_common/input_mapping.h

@@ -3,8 +3,14 @@
 // Refer to the license.txt file included
 
 #pragma once
+
+#include "common/param_package.h"
 #include "common/threadsafe_queue.h"
 
+namespace InputCommon::Polling {
+enum class InputType;
+}
+
 namespace InputCommon {
 class InputEngine;
 struct MappingData;
@@ -20,7 +26,7 @@ public:
     void BeginMapping(Polling::InputType type);
 
     /// Returns an input event with mapping information from the input_queue
-    [[nodiscard]] const Common::ParamPackage GetNextInput();
+    [[nodiscard]] Common::ParamPackage GetNextInput();
 
     /**
      * Registers mapping input data from the driver

src/input_common/input_poller.cpp

@@ -504,9 +504,10 @@ private:
 
 class InputFromMotion final : public Common::Input::InputDevice {
 public:
-    explicit InputFromMotion(PadIdentifier identifier_, int motion_sensor_,
+    explicit InputFromMotion(PadIdentifier identifier_, int motion_sensor_, float gyro_threshold_,
                              InputEngine* input_engine_)
-        : identifier(identifier_), motion_sensor(motion_sensor_), input_engine(input_engine_) {
+        : identifier(identifier_), motion_sensor(motion_sensor_), gyro_threshold(gyro_threshold_),
+          input_engine(input_engine_) {
         UpdateCallback engine_callback{[this]() { OnChange(); }};
         const InputIdentifier input_identifier{
             .identifier = identifier,
@@ -525,8 +526,9 @@ public:
         const auto basic_motion = input_engine->GetMotion(identifier, motion_sensor);
         Common::Input::MotionStatus status{};
         const Common::Input::AnalogProperties properties = {
-            .deadzone = 0.001f,
+            .deadzone = 0.0f,
             .range = 1.0f,
+            .threshold = gyro_threshold,
             .offset = 0.0f,
         };
         status.accel.x = {.raw_value = basic_motion.accel_x, .properties = properties};
@@ -551,6 +553,7 @@ public:
 private:
     const PadIdentifier identifier;
     const int motion_sensor;
+    const float gyro_threshold;
     int callback_key;
     InputEngine* input_engine;
 };
@@ -873,9 +876,11 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateMotionDevice(
 
     if (params.Has("motion")) {
         const auto motion_sensor = params.Get("motion", 0);
+        const auto gyro_threshold = params.Get("threshold", 0.007f);
         input_engine->PreSetController(identifier);
         input_engine->PreSetMotion(identifier, motion_sensor);
-        return std::make_unique<InputFromMotion>(identifier, motion_sensor, input_engine.get());
+        return std::make_unique<InputFromMotion>(identifier, motion_sensor, gyro_threshold,
+                                                 input_engine.get());
     }
 
     const auto deadzone = std::clamp(params.Get("deadzone", 0.15f), 0.0f, 1.0f);

src/input_common/main.cpp

@@ -27,7 +27,7 @@ namespace InputCommon {
 struct InputSubsystem::Impl {
     void Initialize() {
         mapping_factory = std::make_shared<MappingFactory>();
-        MappingCallback mapping_callback{[this](MappingData data) { RegisterInput(data); }};
+        MappingCallback mapping_callback{[this](const MappingData& data) { RegisterInput(data); }};
 
         keyboard = std::make_shared<Keyboard>("keyboard");
         keyboard->SetMappingCallback(mapping_callback);
@@ -284,7 +284,7 @@ struct InputSubsystem::Impl {
 #endif
     }
 
-    void RegisterInput(MappingData data) {
+    void RegisterInput(const MappingData& data) {
         mapping_factory->RegisterInput(data);
     }
 
@@ -394,7 +394,7 @@ void InputSubsystem::BeginMapping(Polling::InputType type) {
     impl->mapping_factory->BeginMapping(type);
 }
 
-const Common::ParamPackage InputSubsystem::GetNextInput() const {
+Common::ParamPackage InputSubsystem::GetNextInput() const {
     return impl->mapping_factory->GetNextInput();
 }
 

src/input_common/main.h

@@ -126,7 +126,7 @@ public:
     void BeginMapping(Polling::InputType type);
 
     /// Returns an input event with mapping information.
-    [[nodiscard]] const Common::ParamPackage GetNextInput() const;
+    [[nodiscard]] Common::ParamPackage GetNextInput() const;
 
     /// Stop polling from all backends.
     void StopMapping() const;

src/shader_recompiler/backend/glsl/glsl_emit_context.cpp

@@ -458,9 +458,10 @@ void EmitContext::DefineGenericOutput(size_t index, u32 invocations) {
         std::string definition{fmt::format("layout(location={}", index)};
         const u32 remainder{4 - element};
         const TransformFeedbackVarying* xfb_varying{};
-        if (!runtime_info.xfb_varyings.empty()) {
-            xfb_varying = &runtime_info.xfb_varyings[base_index + element];
-            xfb_varying = xfb_varying && xfb_varying->components > 0 ? xfb_varying : nullptr;
+        const size_t xfb_varying_index{base_index + element};
+        if (xfb_varying_index < runtime_info.xfb_varyings.size()) {
+            xfb_varying = &runtime_info.xfb_varyings[xfb_varying_index];
+            xfb_varying = xfb_varying->components > 0 ? xfb_varying : nullptr;
         }
         const u32 num_components{xfb_varying ? xfb_varying->components : remainder};
         if (element > 0) {

src/shader_recompiler/backend/spirv/emit_spirv.cpp

@@ -387,6 +387,14 @@ void SetupSignedNanCapabilities(const Profile& profile, const IR::Program& progr
     }
 }
 
+void SetupTransformFeedbackCapabilities(EmitContext& ctx, Id main_func) {
+    if (ctx.runtime_info.xfb_varyings.empty()) {
+        return;
+    }
+    ctx.AddCapability(spv::Capability::TransformFeedback);
+    ctx.AddExecutionMode(main_func, spv::ExecutionMode::Xfb);
+}
+
 void SetupCapabilities(const Profile& profile, const Info& info, EmitContext& ctx) {
     if (info.uses_sampled_1d) {
         ctx.AddCapability(spv::Capability::Sampled1D);
@@ -442,9 +450,6 @@ void SetupCapabilities(const Profile& profile, const Info& info, EmitContext& ct
     if (info.uses_sample_id) {
         ctx.AddCapability(spv::Capability::SampleRateShading);
     }
-    if (!ctx.runtime_info.xfb_varyings.empty()) {
-        ctx.AddCapability(spv::Capability::TransformFeedback);
-    }
     if (info.uses_derivatives) {
         ctx.AddCapability(spv::Capability::DerivativeControl);
     }
@@ -484,6 +489,7 @@ std::vector<u32> EmitSPIRV(const Profile& profile, const RuntimeInfo& runtime_in
         SetupSignedNanCapabilities(profile, program, ctx, main);
     }
     SetupCapabilities(profile, program.info, ctx);
+    SetupTransformFeedbackCapabilities(ctx, main);
     PatchPhiNodes(program, ctx);
     return ctx.Assemble();
 }

src/shader_recompiler/backend/spirv/emit_spirv_atomic.cpp

@@ -74,7 +74,7 @@ Id StorageAtomicU64(EmitContext& ctx, const IR::Value& binding, const IR::Value&
         const auto [scope, semantics]{AtomicArgs(ctx)};
         return (ctx.*atomic_func)(ctx.U64, pointer, scope, semantics, value);
     }
-    LOG_ERROR(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
+    LOG_WARNING(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
     const Id pointer{StoragePointer(ctx, ctx.storage_types.U32x2, &StorageDefinitions::U32x2,
                                     binding, offset, sizeof(u32[2]))};
     const Id original_value{ctx.OpBitcast(ctx.U64, ctx.OpLoad(ctx.U32[2], pointer))};
@@ -267,7 +267,7 @@ Id EmitStorageAtomicExchange64(EmitContext& ctx, const IR::Value& binding, const
         const auto [scope, semantics]{AtomicArgs(ctx)};
         return ctx.OpAtomicExchange(ctx.U64, pointer, scope, semantics, value);
     }
-    LOG_ERROR(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
+    LOG_WARNING(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
     const Id pointer{StoragePointer(ctx, ctx.storage_types.U32x2, &StorageDefinitions::U32x2,
                                     binding, offset, sizeof(u32[2]))};
     const Id original{ctx.OpBitcast(ctx.U64, ctx.OpLoad(ctx.U32[2], pointer))};

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -164,9 +164,10 @@ void DefineGenericOutput(EmitContext& ctx, size_t index, std::optional<u32> invo
     while (element < 4) {
         const u32 remainder{4 - element};
         const TransformFeedbackVarying* xfb_varying{};
-        if (!ctx.runtime_info.xfb_varyings.empty()) {
-            xfb_varying = &ctx.runtime_info.xfb_varyings[base_attr_index + element];
-            xfb_varying = xfb_varying && xfb_varying->components > 0 ? xfb_varying : nullptr;
+        const size_t xfb_varying_index{base_attr_index + element};
+        if (xfb_varying_index < ctx.runtime_info.xfb_varyings.size()) {
+            xfb_varying = &ctx.runtime_info.xfb_varyings[xfb_varying_index];
+            xfb_varying = xfb_varying->components > 0 ? xfb_varying : nullptr;
         }
         const u32 num_components{xfb_varying ? xfb_varying->components : remainder};
 

src/shader_recompiler/frontend/maxwell/translate/impl/video_minimum_maximum.cpp

@@ -57,16 +57,6 @@ void TranslatorVisitor::VMNMX(u64 insn) {
     if (vmnmx.sat != 0) {
         throw NotImplementedException("VMNMX SAT");
     }
-    // Selectors were shown to default to 2 in unit tests
-    if (vmnmx.src_a_selector != 2) {
-        throw NotImplementedException("VMNMX Selector {}", vmnmx.src_a_selector.Value());
-    }
-    if (vmnmx.src_b_selector != 2) {
-        throw NotImplementedException("VMNMX Selector {}", vmnmx.src_b_selector.Value());
-    }
-    if (vmnmx.src_a_width != VideoWidth::Word) {
-        throw NotImplementedException("VMNMX Source Width {}", vmnmx.src_a_width.Value());
-    }
 
     const bool is_b_imm{vmnmx.is_src_b_reg == 0};
     const IR::U32 src_a{GetReg8(insn)};
@@ -76,10 +66,14 @@ void TranslatorVisitor::VMNMX(u64 insn) {
     const VideoWidth a_width{vmnmx.src_a_width};
     const VideoWidth b_width{GetVideoSourceWidth(vmnmx.src_b_width, is_b_imm)};
 
+    const u32 a_selector{static_cast<u32>(vmnmx.src_a_selector)};
+    // Immediate values can't have a selector
+    const u32 b_selector{is_b_imm ? 0U : static_cast<u32>(vmnmx.src_b_selector)};
+
     const bool src_a_signed{vmnmx.src_a_sign != 0};
     const bool src_b_signed{vmnmx.src_b_sign != 0};
-    const IR::U32 op_a{ExtractVideoOperandValue(ir, src_a, a_width, 0, src_a_signed)};
-    const IR::U32 op_b{ExtractVideoOperandValue(ir, src_b, b_width, 0, src_b_signed)};
+    const IR::U32 op_a{ExtractVideoOperandValue(ir, src_a, a_width, a_selector, src_a_signed)};
+    const IR::U32 op_b{ExtractVideoOperandValue(ir, src_b, b_width, b_selector, src_b_signed)};
 
     // First operation's sign is only dependent on operand b's sign
     const bool op_1_signed{src_b_signed};

src/shader_recompiler/frontend/maxwell/translate_program.h

@@ -21,7 +21,6 @@ namespace Shader::Maxwell {
 [[nodiscard]] IR::Program MergeDualVertexPrograms(IR::Program& vertex_a, IR::Program& vertex_b,
                                                   Environment& env_vertex_b);
 
-[[nodiscard]] void ConvertLegacyToGeneric(IR::Program& program,
-                                          const Shader::RuntimeInfo& runtime_info);
+void ConvertLegacyToGeneric(IR::Program& program, const RuntimeInfo& runtime_info);
 
 } // namespace Shader::Maxwell

src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp

@@ -688,7 +688,7 @@ void VisitUsages(Info& info, IR::Inst& inst) {
     case IR::Opcode::StorageAtomicAnd64:
     case IR::Opcode::StorageAtomicOr64:
     case IR::Opcode::StorageAtomicXor64:
-        info.used_storage_buffer_types |= IR::Type::U64;
+        info.used_storage_buffer_types |= IR::Type::U64 | IR::Type::U32x2;
         info.uses_int64_bit_atomics = true;
         break;
     case IR::Opcode::BindlessImageAtomicIAdd32:

src/video_core/buffer_cache/buffer_cache.h

@@ -131,6 +131,8 @@ public:
 
     void DownloadMemory(VAddr cpu_addr, u64 size);
 
+    bool InlineMemory(VAddr dest_address, size_t copy_size, std::span<u8> inlined_buffer);
+
     void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size);
 
     void DisableGraphicsUniformBuffer(size_t stage, u32 index);
@@ -808,6 +810,8 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
         return;
     }
     MICROPROFILE_SCOPE(GPU_DownloadMemory);
+    const bool is_accuracy_normal =
+        Settings::values.gpu_accuracy.GetValue() == Settings::GPUAccuracy::Normal;
 
     boost::container::small_vector<std::pair<BufferCopy, BufferId>, 1> downloads;
     u64 total_size_bytes = 0;
@@ -819,6 +823,9 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
             ForEachBufferInRange(cpu_addr, size, [&](BufferId buffer_id, Buffer& buffer) {
                 buffer.ForEachDownloadRangeAndClear(
                     cpu_addr, size, [&](u64 range_offset, u64 range_size) {
+                        if (is_accuracy_normal) {
+                            return;
+                        }
                         const VAddr buffer_addr = buffer.CpuAddr();
                         const auto add_download = [&](VAddr start, VAddr end) {
                             const u64 new_offset = start - buffer_addr;
@@ -1417,10 +1424,8 @@ void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, VAddr cpu_addr, u32 s
     const IntervalType base_interval{cpu_addr, cpu_addr + size};
     common_ranges.add(base_interval);
 
-    const bool is_accuracy_high =
-        Settings::values.gpu_accuracy.GetValue() == Settings::GPUAccuracy::High;
     const bool is_async = Settings::values.use_asynchronous_gpu_emulation.GetValue();
-    if (!is_async && !is_accuracy_high) {
+    if (!is_async) {
         return;
     }
     uncommitted_ranges.add(base_interval);
@@ -1474,6 +1479,8 @@ typename BufferCache<P>::OverlapResult BufferCache<P>::ResolveOverlaps(VAddr cpu
             // When this memory region has been joined a bunch of times, we assume it's being used
             // as a stream buffer. Increase the size to skip constantly recreating buffers.
             has_stream_leap = true;
+            begin -= PAGE_SIZE * 256;
+            cpu_addr = begin;
             end += PAGE_SIZE * 256;
         }
     }
@@ -1641,6 +1648,42 @@ void BufferCache<P>::MappedUploadMemory(Buffer& buffer, u64 total_size_bytes,
     runtime.CopyBuffer(buffer, upload_staging.buffer, copies);
 }
 
+template <class P>
+bool BufferCache<P>::InlineMemory(VAddr dest_address, size_t copy_size,
+                                  std::span<u8> inlined_buffer) {
+    const bool is_dirty = IsRegionRegistered(dest_address, copy_size);
+    if (!is_dirty) {
+        return false;
+    }
+    if (!IsRegionGpuModified(dest_address, copy_size)) {
+        return false;
+    }
+
+    const IntervalType subtract_interval{dest_address, dest_address + copy_size};
+    ClearDownload(subtract_interval);
+    common_ranges.subtract(subtract_interval);
+
+    BufferId buffer_id = FindBuffer(dest_address, static_cast<u32>(copy_size));
+    auto& buffer = slot_buffers[buffer_id];
+    SynchronizeBuffer(buffer, dest_address, static_cast<u32>(copy_size));
+
+    if constexpr (USE_MEMORY_MAPS) {
+        std::array copies{BufferCopy{
+            .src_offset = 0,
+            .dst_offset = buffer.Offset(dest_address),
+            .size = copy_size,
+        }};
+        auto upload_staging = runtime.UploadStagingBuffer(copy_size);
+        u8* const src_pointer = upload_staging.mapped_span.data();
+        std::memcpy(src_pointer, inlined_buffer.data(), copy_size);
+        runtime.CopyBuffer(buffer, upload_staging.buffer, copies);
+    } else {
+        buffer.ImmediateUpload(buffer.Offset(dest_address), inlined_buffer.first(copy_size));
+    }
+
+    return true;
+}
+
 template <class P>
 void BufferCache<P>::DownloadBufferMemory(Buffer& buffer) {
     DownloadBufferMemory(buffer, buffer.CpuAddr(), buffer.SizeBytes());

src/video_core/command_classes/codecs/codec.h

@@ -66,7 +66,7 @@ private:
     bool initialized{};
     NvdecCommon::VideoCodec current_codec{NvdecCommon::VideoCodec::None};
 
-    AVCodec* av_codec{nullptr};
+    const AVCodec* av_codec{nullptr};
     AVCodecContext* av_codec_ctx{nullptr};
     AVBufferRef* av_gpu_decoder{nullptr};
 

src/video_core/engines/engine_upload.cpp

@@ -7,6 +7,7 @@
 #include "common/assert.h"
 #include "video_core/engines/engine_upload.h"
 #include "video_core/memory_manager.h"
+#include "video_core/rasterizer_interface.h"
 #include "video_core/textures/decoders.h"
 
 namespace Tegra::Engines::Upload {
@@ -16,6 +17,10 @@ State::State(MemoryManager& memory_manager_, Registers& regs_)
 
 State::~State() = default;
 
+void State::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
+    rasterizer = rasterizer_;
+}
+
 void State::ProcessExec(const bool is_linear_) {
     write_offset = 0;
     copy_size = regs.line_length_in * regs.line_count;
@@ -32,7 +37,7 @@ void State::ProcessData(const u32 data, const bool is_last_call) {
     }
     const GPUVAddr address{regs.dest.Address()};
     if (is_linear) {
-        memory_manager.WriteBlock(address, inner_buffer.data(), copy_size);
+        rasterizer->AccelerateInlineToMemory(address, copy_size, inner_buffer);
     } else {
         UNIMPLEMENTED_IF(regs.dest.z != 0);
         UNIMPLEMENTED_IF(regs.dest.depth != 1);

src/video_core/engines/engine_upload.h

@@ -12,6 +12,10 @@ namespace Tegra {
 class MemoryManager;
 }
 
+namespace VideoCore {
+class RasterizerInterface;
+}
+
 namespace Tegra::Engines::Upload {
 
 struct Registers {
@@ -60,6 +64,9 @@ public:
     void ProcessExec(bool is_linear_);
     void ProcessData(u32 data, bool is_last_call);
 
+    /// Binds a rasterizer to this engine.
+    void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);
+
 private:
     u32 write_offset = 0;
     u32 copy_size = 0;
@@ -68,6 +75,7 @@ private:
     bool is_linear = false;
     Registers& regs;
     MemoryManager& memory_manager;
+    VideoCore::RasterizerInterface* rasterizer = nullptr;
 };
 
 } // namespace Tegra::Engines::Upload

src/video_core/engines/kepler_compute.cpp

@@ -22,6 +22,7 @@ KeplerCompute::~KeplerCompute() = default;
 
 void KeplerCompute::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
     rasterizer = rasterizer_;
+    upload_state.BindRasterizer(rasterizer);
 }
 
 void KeplerCompute::CallMethod(u32 method, u32 method_argument, bool is_last_call) {

src/video_core/engines/kepler_memory.cpp

@@ -19,6 +19,10 @@ KeplerMemory::KeplerMemory(Core::System& system_, MemoryManager& memory_manager)
 
 KeplerMemory::~KeplerMemory() = default;
 
+void KeplerMemory::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
+    upload_state.BindRasterizer(rasterizer_);
+}
+
 void KeplerMemory::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
     ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid KeplerMemory register, increase the size of the Regs structure");

src/video_core/engines/kepler_memory.h

@@ -22,6 +22,10 @@ namespace Tegra {
 class MemoryManager;
 }
 
+namespace VideoCore {
+class RasterizerInterface;
+}
+
 namespace Tegra::Engines {
 
 /**
@@ -38,6 +42,9 @@ public:
     explicit KeplerMemory(Core::System& system_, MemoryManager& memory_manager);
     ~KeplerMemory() override;
 
+    /// Binds a rasterizer to this engine.
+    void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);
+
     /// Write the value to the register identified by method.
     void CallMethod(u32 method, u32 method_argument, bool is_last_call) override;
 

src/video_core/engines/maxwell_3d.cpp

@@ -31,6 +31,7 @@ Maxwell3D::~Maxwell3D() = default;
 
 void Maxwell3D::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
     rasterizer = rasterizer_;
+    upload_state.BindRasterizer(rasterizer_);
 }
 
 void Maxwell3D::InitializeRegisterDefaults() {
@@ -240,7 +241,7 @@ void Maxwell3D::CallMacroMethod(u32 method, const std::vector<u32>& parameters)
         ((method - MacroRegistersStart) >> 1) % static_cast<u32>(macro_positions.size());
 
     // Execute the current macro.
-    macro_engine->Execute(*this, macro_positions[entry], parameters);
+    macro_engine->Execute(macro_positions[entry], parameters);
     if (mme_draw.current_mode != MMEDrawMode::Undefined) {
         FlushMMEInlineDraw();
     }

src/video_core/engines/maxwell_3d.h

@@ -1557,7 +1557,8 @@ private:
 
     static constexpr u32 null_cb_data = 0xFFFFFFFF;
     struct CBDataState {
-        std::array<std::array<u32, 0x4000>, 16> buffer;
+        static constexpr size_t inline_size = 0x4000;
+        std::array<std::array<u32, inline_size>, 16> buffer;
         u32 current{null_cb_data};
         u32 id{null_cb_data};
         u32 start_pos{};

src/video_core/gpu.cpp

@@ -59,6 +59,7 @@ struct GPU::Impl {
         maxwell_3d->BindRasterizer(rasterizer);
         fermi_2d->BindRasterizer(rasterizer);
         kepler_compute->BindRasterizer(rasterizer);
+        kepler_memory->BindRasterizer(rasterizer);
         maxwell_dma->BindRasterizer(rasterizer);
     }
 
@@ -502,8 +503,13 @@ struct GPU::Impl {
         case BufferMethods::SemaphoreAddressHigh:
         case BufferMethods::SemaphoreAddressLow:
         case BufferMethods::SemaphoreSequence:
+            break;
         case BufferMethods::UnkCacheFlush:
+            rasterizer->SyncGuestHost();
+            break;
         case BufferMethods::WrcacheFlush:
+            rasterizer->SignalReference();
+            break;
         case BufferMethods::FenceValue:
             break;
         case BufferMethods::RefCnt:
@@ -513,7 +519,7 @@ struct GPU::Impl {
             ProcessFenceActionMethod();
             break;
         case BufferMethods::WaitForInterrupt:
-            ProcessWaitForInterruptMethod();
+            rasterizer->WaitForIdle();
             break;
         case BufferMethods::SemaphoreTrigger: {
             ProcessSemaphoreTriggerMethod();

src/video_core/gpu.h

@@ -12,9 +12,6 @@
 #include "video_core/framebuffer_config.h"
 
 namespace Core {
-namespace Frontend {
-class EmuWindow;
-}
 class System;
 } // namespace Core
 
@@ -25,7 +22,6 @@ class ShaderNotify;
 
 namespace Tegra {
 class DmaPusher;
-class CDmaPusher;
 struct CommandList;
 
 enum class RenderTargetFormat : u32 {
@@ -88,15 +84,9 @@ enum class DepthFormat : u32 {
     D32_FLOAT_S8X24_UINT = 0x19,
 };
 
-struct CommandListHeader;
-class DebugContext;
-
 namespace Engines {
-class Fermi2D;
 class Maxwell3D;
-class MaxwellDMA;
 class KeplerCompute;
-class KeplerMemory;
 } // namespace Engines
 
 enum class EngineID {
@@ -190,12 +180,6 @@ public:
     /// Returns a const reference to the GPU DMA pusher.
     [[nodiscard]] const Tegra::DmaPusher& DmaPusher() const;
 
-    /// Returns a reference to the GPU CDMA pusher.
-    [[nodiscard]] Tegra::CDmaPusher& CDmaPusher();
-
-    /// Returns a const reference to the GPU CDMA pusher.
-    [[nodiscard]] const Tegra::CDmaPusher& CDmaPusher() const;
-
     /// Returns a reference to the underlying renderer.
     [[nodiscard]] VideoCore::RendererBase& Renderer();
 

src/video_core/host_shaders/astc_decoder.comp

@@ -155,9 +155,6 @@ uint SwizzleOffset(uvec2 pos) {
 // Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)]
 // is the same as [(num_bits - 1):0] and repeats all the way down.
 uint Replicate(uint val, uint num_bits, uint to_bit) {
-    if (num_bits == 0 || to_bit == 0) {
-        return 0;
-    }
     const uint v = val & uint((1 << num_bits) - 1);
     uint res = v;
     uint reslen = num_bits;
@@ -187,42 +184,57 @@ uint ReplicateBitTo9(uint value) {
     return REPLICATE_1_BIT_TO_9_TABLE[value];
 }
 
-uint FastReplicateTo8(uint value, uint num_bits) {
-    switch (num_bits) {
-    case 1:
-        return REPLICATE_1_BIT_TO_8_TABLE[value];
-    case 2:
-        return REPLICATE_2_BIT_TO_8_TABLE[value];
-    case 3:
-        return REPLICATE_3_BIT_TO_8_TABLE[value];
-    case 4:
-        return REPLICATE_4_BIT_TO_8_TABLE[value];
-    case 5:
-        return REPLICATE_5_BIT_TO_8_TABLE[value];
-    case 6:
-        return REPLICATE_6_BIT_TO_8_TABLE[value];
-    case 7:
-        return REPLICATE_7_BIT_TO_8_TABLE[value];
-    case 8:
+uint FastReplicate(uint value, uint num_bits, uint to_bit) {
+    if (num_bits == 0) {
+        return 0;
+    }
+    if (num_bits == to_bit) {
         return value;
     }
-    return Replicate(value, num_bits, 8);
+    if (to_bit == 6) {
+        switch (num_bits) {
+        case 1:
+            return REPLICATE_1_BIT_TO_6_TABLE[value];
+        case 2:
+            return REPLICATE_2_BIT_TO_6_TABLE[value];
+        case 3:
+            return REPLICATE_3_BIT_TO_6_TABLE[value];
+        case 4:
+            return REPLICATE_4_BIT_TO_6_TABLE[value];
+        case 5:
+            return REPLICATE_5_BIT_TO_6_TABLE[value];
+        default:
+            break;
+        }
+    } else { /* if (to_bit == 8) */
+        switch (num_bits) {
+        case 1:
+            return REPLICATE_1_BIT_TO_8_TABLE[value];
+        case 2:
+            return REPLICATE_2_BIT_TO_8_TABLE[value];
+        case 3:
+            return REPLICATE_3_BIT_TO_8_TABLE[value];
+        case 4:
+            return REPLICATE_4_BIT_TO_8_TABLE[value];
+        case 5:
+            return REPLICATE_5_BIT_TO_8_TABLE[value];
+        case 6:
+            return REPLICATE_6_BIT_TO_8_TABLE[value];
+        case 7:
+            return REPLICATE_7_BIT_TO_8_TABLE[value];
+        default:
+            break;
+        }
+    }
+    return Replicate(value, num_bits, to_bit);
+}
+
+uint FastReplicateTo8(uint value, uint num_bits) {
+    return FastReplicate(value, num_bits, 8);
 }
 
 uint FastReplicateTo6(uint value, uint num_bits) {
-    switch (num_bits) {
-    case 1:
-        return REPLICATE_1_BIT_TO_6_TABLE[value];
-    case 2:
-        return REPLICATE_2_BIT_TO_6_TABLE[value];
-    case 3:
-        return REPLICATE_3_BIT_TO_6_TABLE[value];
-    case 4:
-        return REPLICATE_4_BIT_TO_6_TABLE[value];
-    case 5:
-        return REPLICATE_5_BIT_TO_6_TABLE[value];
-    }
-    return Replicate(value, num_bits, 6);
+    return FastReplicate(value, num_bits, 6);
 }
 
 uint Div3Floor(uint v) {