core: Shutdown: Move kernel cleanup to later in shutdown.

- Fixes a shutdown crash due to a race condition with GPU still accessing memory.

.ci/scripts/clang/docker.sh

@@ -0,0 +1,18 @@
+#!/bin/bash -ex
+
+# Exit on error, rather than continuing with the rest of the script.
+set -e
+
+cd /yuzu
+
+ccache -s
+
+mkdir build || true && cd build
+cmake .. -DDISPLAY_VERSION=$1 -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=/usr/lib/ccache/clang -DCMAKE_CXX_COMPILER=/usr/lib/ccache/clang++ -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DUSE_DISCORD_PRESENCE=ON -DENABLE_QT_TRANSLATION=ON -DCMAKE_INSTALL_PREFIX="/usr"
+
+make -j$(nproc)
+
+ccache -s
+
+ctest -VV -C Release
+

.ci/scripts/clang/exec.sh

@@ -0,0 +1,8 @@
+#!/bin/bash -ex
+
+mkdir -p "ccache"  || true
+chmod a+x ./.ci/scripts/clang/docker.sh
+# the UID for the container yuzu user is 1027
+sudo chown -R 1027 ./
+docker run -e ENABLE_COMPATIBILITY_REPORTING -e CCACHE_DIR=/yuzu/ccache -v $(pwd):/yuzu yuzuemu/build-environments:linux-fresh /bin/bash /yuzu/.ci/scripts/clang/docker.sh $1
+sudo chown -R $UID ./

.ci/scripts/clang/upload.sh

@@ -0,0 +1,20 @@
+#!/bin/bash -ex
+
+. .ci/scripts/common/pre-upload.sh
+
+REV_NAME="yuzu-linux-${GITDATE}-${GITREV}"
+ARCHIVE_NAME="${REV_NAME}.tar.xz"
+COMPRESSION_FLAGS="-cJvf"
+
+if [ "${RELEASE_NAME}" = "mainline" ]; then
+    DIR_NAME="${REV_NAME}"
+else
+    DIR_NAME="${REV_NAME}_${RELEASE_NAME}"
+fi
+
+mkdir "$DIR_NAME"
+
+cp build/bin/yuzu-cmd "$DIR_NAME"
+cp build/bin/yuzu "$DIR_NAME"
+
+. .ci/scripts/common/post-upload.sh

.ci/templates/build-standard.yml

@@ -12,6 +12,9 @@ jobs:
       windows:
         BuildSuffix: 'windows-mingw'
         ScriptFolder: 'windows'
+      clang:
+        BuildSuffix: 'clang'
+        ScriptFolder: 'clang'
       linux:
         BuildSuffix: 'linux'
         ScriptFolder: 'linux'
@@ -24,4 +27,4 @@ jobs:
     parameters:
       artifactSource: 'false'
       cache: $(parameters.cache)
-      version: $(parameters.version)
+      version: $(parameters.version)

externals/glad/include/glad/glad.h

@@ -5156,6 +5156,9 @@ GLAPI PFNGLDEPTHRANGEARRAYVPROC glad_glDepthRangeArrayv;
 typedef void (APIENTRYP PFNGLDEPTHRANGEINDEXEDPROC)(GLuint index, GLdouble n, GLdouble f);
 GLAPI PFNGLDEPTHRANGEINDEXEDPROC glad_glDepthRangeIndexed;
 #define glDepthRangeIndexed glad_glDepthRangeIndexed
+typedef void (APIENTRYP PFNGLDEPTHRANGEINDEXEDDNVPROC)(GLuint index, GLdouble n, GLdouble f);
+GLAPI PFNGLDEPTHRANGEINDEXEDDNVPROC glad_glDepthRangeIndexeddNV;
+#define glDepthRangeIndexeddNV glad_glDepthRangeIndexeddNV
 typedef void (APIENTRYP PFNGLGETFLOATI_VPROC)(GLenum target, GLuint index, GLfloat *data);
 GLAPI PFNGLGETFLOATI_VPROC glad_glGetFloati_v;
 #define glGetFloati_v glad_glGetFloati_v

externals/glad/src/glad.c

@@ -1044,6 +1044,7 @@ PFNGLDEPTHMASKPROC glad_glDepthMask = NULL;
 PFNGLDEPTHRANGEPROC glad_glDepthRange = NULL;
 PFNGLDEPTHRANGEARRAYVPROC glad_glDepthRangeArrayv = NULL;
 PFNGLDEPTHRANGEINDEXEDPROC glad_glDepthRangeIndexed = NULL;
+PFNGLDEPTHRANGEINDEXEDDNVPROC glad_glDepthRangeIndexeddNV = NULL;
 PFNGLDEPTHRANGEFPROC glad_glDepthRangef = NULL;
 PFNGLDETACHSHADERPROC glad_glDetachShader = NULL;
 PFNGLDISABLEPROC glad_glDisable = NULL;
@@ -7971,6 +7972,7 @@ static void load_GL_NV_depth_buffer_float(GLADloadproc load) {
 	glad_glDepthRangedNV = (PFNGLDEPTHRANGEDNVPROC)load("glDepthRangedNV");
 	glad_glClearDepthdNV = (PFNGLCLEARDEPTHDNVPROC)load("glClearDepthdNV");
 	glad_glDepthBoundsdNV = (PFNGLDEPTHBOUNDSDNVPROC)load("glDepthBoundsdNV");
+	glad_glDepthRangeIndexeddNV = (PFNGLDEPTHRANGEINDEXEDDNVPROC)load("glDepthRangeIndexeddNV");
 }
 static void load_GL_NV_draw_texture(GLADloadproc load) {
 	if(!GLAD_GL_NV_draw_texture) return;

src/common/cityhash.h

@@ -61,6 +61,7 @@
 
 #pragma once
 
+#include <cstddef>
 #include "common/common_types.h"
 
 namespace Common {

src/common/uint128.h

@@ -98,4 +98,24 @@ namespace Common {
 #endif
 }
 
+// This function divides a u128 by a u32 value and produces two u64 values:
+// the result of division and the remainder
+[[nodiscard]] static inline std::pair<u64, u64> Divide128On32(u128 dividend, u32 divisor) {
+    u64 remainder = dividend[0] % divisor;
+    u64 accum = dividend[0] / divisor;
+    if (dividend[1] == 0)
+        return {accum, remainder};
+    // We ignore dividend[1] / divisor as that overflows
+    const u64 first_segment = (dividend[1] % divisor) << 32;
+    accum += (first_segment / divisor) << 32;
+    const u64 second_segment = (first_segment % divisor) << 32;
+    accum += (second_segment / divisor);
+    remainder += second_segment % divisor;
+    if (remainder >= divisor) {
+        accum++;
+        remainder -= divisor;
+    }
+    return {accum, remainder};
+}
+
 } // namespace Common

src/common/wall_clock.cpp

@@ -20,9 +20,7 @@ using base_time_point = std::chrono::time_point<base_timer>;
 class StandardWallClock final : public WallClock {
 public:
     explicit StandardWallClock(u64 emulated_cpu_frequency_, u64 emulated_clock_frequency_)
-        : WallClock(emulated_cpu_frequency_, emulated_clock_frequency_, false),
-          emulated_clock_factor{GetFixedPoint64Factor(emulated_clock_frequency, 1000000000)},
-          emulated_cpu_factor{GetFixedPoint64Factor(emulated_cpu_frequency, 1000000000)} {
+        : WallClock(emulated_cpu_frequency_, emulated_clock_frequency_, false) {
         start_time = base_timer::now();
     }
 
@@ -45,11 +43,16 @@ public:
     }
 
     u64 GetClockCycles() override {
-        return MultiplyHigh(GetTimeNS().count(), emulated_clock_factor);
+        std::chrono::nanoseconds time_now = GetTimeNS();
+        const u128 temporary =
+            Common::Multiply64Into128(time_now.count(), emulated_clock_frequency);
+        return Common::Divide128On32(temporary, 1000000000).first;
     }
 
     u64 GetCPUCycles() override {
-        return MultiplyHigh(GetTimeNS().count(), emulated_cpu_factor);
+        std::chrono::nanoseconds time_now = GetTimeNS();
+        const u128 temporary = Common::Multiply64Into128(time_now.count(), emulated_cpu_frequency);
+        return Common::Divide128On32(temporary, 1000000000).first;
     }
 
     void Pause([[maybe_unused]] bool is_paused) override {
@@ -58,8 +61,6 @@ public:
 
 private:
     base_time_point start_time;
-    const u64 emulated_clock_factor;
-    const u64 emulated_cpu_factor;
 };
 
 #ifdef ARCHITECTURE_x86_64

src/core/core.cpp

@@ -299,25 +299,17 @@ struct System::Impl {
             gpu_core->WaitIdle();
         }
 
-        // Shutdown emulation session
         services.reset();
         service_manager.reset();
         cheat_engine.reset();
         telemetry_session.reset();
-
-        // Close all CPU/threading state
         cpu_manager.Shutdown();
-
-        // Shutdown kernel and core timing
+        time_manager.Shutdown();
         core_timing.Shutdown();
-        kernel.Shutdown();
-
-        // Close app loader
         app_loader.reset();
         gpu_core.reset();
         perf_stats.reset();
-
-        // Clear all applets
+        kernel.Shutdown();
         applet_manager.ClearAll();
 
         LOG_DEBUG(Core, "Shutdown OK");

src/core/hle/kernel/kernel.cpp

@@ -101,8 +101,6 @@ struct KernelCore::Impl {
 
         current_process = nullptr;
 
-        system_resource_limit = nullptr;
-
         global_handle_table.Clear();
 
         preemption_event = nullptr;
@@ -111,6 +109,13 @@ struct KernelCore::Impl {
 
         exclusive_monitor.reset();
 
+        hid_shared_mem = nullptr;
+        font_shared_mem = nullptr;
+        irs_shared_mem = nullptr;
+        time_shared_mem = nullptr;
+
+        system_resource_limit = nullptr;
+
         // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
         next_host_thread_id = Core::Hardware::NUM_CPU_CORES;
     }

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

@@ -508,7 +508,7 @@ public:
             {1, &IManagerForApplication::GetAccountId, "GetAccountId"},
             {2, nullptr, "EnsureIdTokenCacheAsync"},
             {3, nullptr, "LoadIdTokenCache"},
-            {130, nullptr, "GetNintendoAccountUserResourceCacheForApplication"},
+            {130, &IManagerForApplication::GetNintendoAccountUserResourceCacheForApplication, "GetNintendoAccountUserResourceCacheForApplication"},
             {150, nullptr, "CreateAuthorizationRequest"},
             {160, &IManagerForApplication::StoreOpenContext, "StoreOpenContext"},
             {170, nullptr, "LoadNetworkServiceLicenseKindAsync"},
@@ -534,6 +534,22 @@ private:
         rb.PushRaw<u64>(user_id.GetNintendoID());
     }
 
+    void GetNintendoAccountUserResourceCacheForApplication(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_ACC, "(STUBBED) called");
+
+        std::vector<u8> nas_user_base_for_application(0x68);
+        ctx.WriteBuffer(nas_user_base_for_application, 0);
+
+        if (ctx.CanWriteBuffer(1)) {
+            std::vector<u8> unknown_out_buffer(ctx.GetWriteBufferSize(1));
+            ctx.WriteBuffer(unknown_out_buffer, 1);
+        }
+
+        IPC::ResponseBuilder rb{ctx, 4};
+        rb.Push(RESULT_SUCCESS);
+        rb.PushRaw<u64>(user_id.GetNintendoID());
+    }
+
     void StoreOpenContext(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_ACC, "(STUBBED) called");
         IPC::ResponseBuilder rb{ctx, 2};

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

@@ -5,15 +5,25 @@
 #include <cstring>
 #include "common/common_types.h"
 #include "core/core_timing.h"
+#include "core/frontend/emu_window.h"
 #include "core/hle/service/hid/controllers/gesture.h"
+#include "core/settings.h"
 
 namespace Service::HID {
 constexpr std::size_t SHARED_MEMORY_OFFSET = 0x3BA00;
+constexpr f32 angle_threshold = 0.08f;
+constexpr f32 pinch_threshold = 100.0f;
 
 Controller_Gesture::Controller_Gesture(Core::System& system) : ControllerBase(system) {}
 Controller_Gesture::~Controller_Gesture() = default;
 
-void Controller_Gesture::OnInit() {}
+void Controller_Gesture::OnInit() {
+    for (std::size_t id = 0; id < MAX_FINGERS; ++id) {
+        mouse_finger_id[id] = MAX_FINGERS;
+        keyboard_finger_id[id] = MAX_FINGERS;
+        udp_finger_id[id] = MAX_FINGERS;
+    }
+}
 
 void Controller_Gesture::OnRelease() {}
 
@@ -35,10 +45,153 @@ void Controller_Gesture::OnUpdate(const Core::Timing::CoreTiming& core_timing, u
 
     cur_entry.sampling_number = last_entry.sampling_number + 1;
     cur_entry.sampling_number2 = cur_entry.sampling_number;
-    // TODO(ogniK): Update gesture states
+
+    // TODO(german77): Implement all gesture types
+
+    const Input::TouchStatus& mouse_status = touch_mouse_device->GetStatus();
+    const Input::TouchStatus& udp_status = touch_udp_device->GetStatus();
+    for (std::size_t id = 0; id < mouse_status.size(); ++id) {
+        mouse_finger_id[id] = UpdateTouchInputEvent(mouse_status[id], mouse_finger_id[id]);
+        udp_finger_id[id] = UpdateTouchInputEvent(udp_status[id], udp_finger_id[id]);
+    }
+
+    if (Settings::values.use_touch_from_button) {
+        const Input::TouchStatus& keyboard_status = touch_btn_device->GetStatus();
+        for (std::size_t id = 0; id < mouse_status.size(); ++id) {
+            keyboard_finger_id[id] =
+                UpdateTouchInputEvent(keyboard_status[id], keyboard_finger_id[id]);
+        }
+    }
+
+    TouchType type = TouchType::Idle;
+    Attribute attributes{};
+    GestureProperties gesture = GetGestureProperties();
+    if (last_gesture.active_points != gesture.active_points) {
+        ++last_gesture.detection_count;
+    }
+    if (gesture.active_points > 0) {
+        if (last_gesture.active_points == 0) {
+            attributes.is_new_touch.Assign(true);
+            last_gesture.average_distance = gesture.average_distance;
+            last_gesture.angle = gesture.angle;
+        }
+
+        type = TouchType::Touch;
+        if (gesture.mid_point.x != last_entry.x || gesture.mid_point.y != last_entry.y) {
+            type = TouchType::Pan;
+        }
+        if (std::abs(gesture.average_distance - last_gesture.average_distance) > pinch_threshold) {
+            type = TouchType::Pinch;
+        }
+        if (std::abs(gesture.angle - last_gesture.angle) > angle_threshold) {
+            type = TouchType::Rotate;
+        }
+
+        cur_entry.delta_x = gesture.mid_point.x - last_entry.x;
+        cur_entry.delta_y = gesture.mid_point.y - last_entry.y;
+        // TODO: Find how velocities are calculated
+        cur_entry.vel_x = static_cast<float>(cur_entry.delta_x) * 150.1f;
+        cur_entry.vel_y = static_cast<float>(cur_entry.delta_y) * 150.1f;
+
+        // Slowdown the rate of change for less flapping
+        last_gesture.average_distance =
+            (last_gesture.average_distance * 0.9f) + (gesture.average_distance * 0.1f);
+        last_gesture.angle = (last_gesture.angle * 0.9f) + (gesture.angle * 0.1f);
+
+    } else {
+        cur_entry.delta_x = 0;
+        cur_entry.delta_y = 0;
+        cur_entry.vel_x = 0;
+        cur_entry.vel_y = 0;
+    }
+    last_gesture.active_points = gesture.active_points;
+    cur_entry.detection_count = last_gesture.detection_count;
+    cur_entry.type = type;
+    cur_entry.attributes = attributes;
+    cur_entry.x = gesture.mid_point.x;
+    cur_entry.y = gesture.mid_point.y;
+    cur_entry.point_count = static_cast<s32>(gesture.active_points);
+    for (size_t id = 0; id < MAX_POINTS; id++) {
+        cur_entry.points[id].x = gesture.points[id].x;
+        cur_entry.points[id].y = gesture.points[id].y;
+    }
+    cur_entry.rotation_angle = 0;
+    cur_entry.scale = 0;
 
     std::memcpy(data + SHARED_MEMORY_OFFSET, &shared_memory, sizeof(SharedMemory));
 }
 
-void Controller_Gesture::OnLoadInputDevices() {}
+void Controller_Gesture::OnLoadInputDevices() {
+    touch_mouse_device = Input::CreateDevice<Input::TouchDevice>("engine:emu_window");
+    touch_udp_device = Input::CreateDevice<Input::TouchDevice>("engine:cemuhookudp");
+    touch_btn_device = Input::CreateDevice<Input::TouchDevice>("engine:touch_from_button");
+}
+
+std::optional<std::size_t> Controller_Gesture::GetUnusedFingerID() const {
+    std::size_t first_free_id = 0;
+    while (first_free_id < MAX_POINTS) {
+        if (!fingers[first_free_id].pressed) {
+            return first_free_id;
+        } else {
+            first_free_id++;
+        }
+    }
+    return std::nullopt;
+}
+
+std::size_t Controller_Gesture::UpdateTouchInputEvent(
+    const std::tuple<float, float, bool>& touch_input, std::size_t finger_id) {
+    const auto& [x, y, pressed] = touch_input;
+    if (pressed) {
+        if (finger_id == MAX_POINTS) {
+            const auto first_free_id = GetUnusedFingerID();
+            if (!first_free_id) {
+                // Invalid finger id do nothing
+                return MAX_POINTS;
+            }
+            finger_id = first_free_id.value();
+            fingers[finger_id].pressed = true;
+        }
+        fingers[finger_id].x = x;
+        fingers[finger_id].y = y;
+        return finger_id;
+    }
+
+    if (finger_id != MAX_POINTS) {
+        fingers[finger_id].pressed = false;
+    }
+
+    return MAX_POINTS;
+}
+
+Controller_Gesture::GestureProperties Controller_Gesture::GetGestureProperties() {
+    GestureProperties gesture;
+    std::array<Finger, MAX_POINTS> active_fingers;
+    const auto end_iter = std::copy_if(fingers.begin(), fingers.end(), active_fingers.begin(),
+                                       [](const auto& finger) { return finger.pressed; });
+    gesture.active_points =
+        static_cast<std::size_t>(std::distance(active_fingers.begin(), end_iter));
+
+    for (size_t id = 0; id < gesture.active_points; ++id) {
+        gesture.points[id].x =
+            static_cast<int>(active_fingers[id].x * Layout::ScreenUndocked::Width);
+        gesture.points[id].y =
+            static_cast<int>(active_fingers[id].y * Layout::ScreenUndocked::Height);
+        gesture.mid_point.x += static_cast<int>(gesture.points[id].x / gesture.active_points);
+        gesture.mid_point.y += static_cast<int>(gesture.points[id].y / gesture.active_points);
+    }
+
+    for (size_t id = 0; id < gesture.active_points; ++id) {
+        const double distance =
+            std::pow(static_cast<float>(gesture.mid_point.x - gesture.points[id].x), 2) +
+            std::pow(static_cast<float>(gesture.mid_point.y - gesture.points[id].y), 2);
+        gesture.average_distance +=
+            static_cast<float>(distance) / static_cast<float>(gesture.active_points);
+    }
+
+    gesture.angle = std::atan2(static_cast<float>(gesture.mid_point.y - gesture.points[0].y),
+                               static_cast<float>(gesture.mid_point.x - gesture.points[0].x));
+    return gesture;
+}
+
 } // namespace Service::HID

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

@@ -5,8 +5,10 @@
 #pragma once
 
 #include <array>
+#include "common/bit_field.h"
 #include "common/common_types.h"
 #include "common/swap.h"
+#include "core/frontend/input.h"
 #include "core/hle/service/hid/controllers/controller_base.h"
 
 namespace Service::HID {
@@ -28,29 +30,64 @@ public:
     void OnLoadInputDevices() override;
 
 private:
-    struct Locations {
+    static constexpr size_t MAX_FINGERS = 16;
+    static constexpr size_t MAX_POINTS = 4;
+
+    enum class TouchType : u32 {
+        Idle,     // Nothing touching the screen
+        Complete, // Unknown. End of touch?
+        Cancel,   // Never triggered
+        Touch,    // Pressing without movement
+        Press,    // Never triggered
+        Tap,      // Fast press then release
+        Pan,      // All points moving together across the screen
+        Swipe,    // Fast press movement and release of a single point
+        Pinch,    // All points moving away/closer to the midpoint
+        Rotate,   // All points rotating from the midpoint
+    };
+
+    enum class Direction : u32 {
+        None,
+        Left,
+        Up,
+        Right,
+        Down,
+    };
+
+    struct Attribute {
+        union {
+            u32_le raw{};
+
+            BitField<0, 1, u32> is_new_touch;
+            BitField<1, 1, u32> is_double_tap;
+        };
+    };
+    static_assert(sizeof(Attribute) == 4, "Attribute is an invalid size");
+
+    struct Points {
         s32_le x;
         s32_le y;
     };
+    static_assert(sizeof(Points) == 8, "Points is an invalid size");
 
     struct GestureState {
         s64_le sampling_number;
         s64_le sampling_number2;
 
         s64_le detection_count;
-        s32_le type;
-        s32_le dir;
+        TouchType type;
+        Direction dir;
         s32_le x;
         s32_le y;
         s32_le delta_x;
         s32_le delta_y;
         f32 vel_x;
         f32 vel_y;
-        s32_le attributes;
-        f32 scale;
-        f32 rotation;
-        s32_le location_count;
-        std::array<Locations, 4> locations;
+        Attribute attributes;
+        u32 scale;
+        u32 rotation_angle;
+        s32_le point_count;
+        std::array<Points, 4> points;
     };
     static_assert(sizeof(GestureState) == 0x68, "GestureState is an invalid size");
 
@@ -58,6 +95,45 @@ private:
         CommonHeader header;
         std::array<GestureState, 17> gesture_states;
     };
+    static_assert(sizeof(SharedMemory) == 0x708, "SharedMemory is an invalid size");
+
+    struct Finger {
+        f32 x{};
+        f32 y{};
+        bool pressed{};
+    };
+
+    struct GestureProperties {
+        std::array<Points, MAX_POINTS> points{};
+        std::size_t active_points{};
+        Points mid_point{};
+        s64_le detection_count{};
+        u64_le delta_time{};
+        float average_distance{};
+        float angle{};
+    };
+
+    // Returns an unused finger id, if there is no fingers avaliable MAX_FINGERS will be returned
+    std::optional<size_t> GetUnusedFingerID() const;
+
+    /** If the touch is new it tries to assing a new finger id, if there is no fingers avaliable no
+     * changes will be made. Updates the coordinates if the finger id it's already set. If the touch
+     * ends delays the output by one frame to set the end_touch flag before finally freeing the
+     * finger id */
+    size_t UpdateTouchInputEvent(const std::tuple<float, float, bool>& touch_input,
+                                 size_t finger_id);
+
+    // Returns the average distance, angle and middle point of the active fingers
+    GestureProperties GetGestureProperties();
+
     SharedMemory shared_memory{};
+    std::unique_ptr<Input::TouchDevice> touch_mouse_device;
+    std::unique_ptr<Input::TouchDevice> touch_udp_device;
+    std::unique_ptr<Input::TouchDevice> touch_btn_device;
+    std::array<size_t, MAX_FINGERS> mouse_finger_id;
+    std::array<size_t, MAX_FINGERS> keyboard_finger_id;
+    std::array<size_t, MAX_FINGERS> udp_finger_id;
+    std::array<Finger, MAX_POINTS> fingers;
+    GestureProperties last_gesture;
 };
 } // namespace Service::HID

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

@@ -273,8 +273,8 @@ Hid::Hid(Core::System& system_) : ServiceFramework{system_, "hid"} {
         {204, &Hid::PermitVibration, "PermitVibration"},
         {205, &Hid::IsVibrationPermitted, "IsVibrationPermitted"},
         {206, &Hid::SendVibrationValues, "SendVibrationValues"},
-        {207, nullptr, "SendVibrationGcErmCommand"},
-        {208, nullptr, "GetActualVibrationGcErmCommand"},
+        {207, &Hid::SendVibrationGcErmCommand, "SendVibrationGcErmCommand"},
+        {208, &Hid::GetActualVibrationGcErmCommand, "GetActualVibrationGcErmCommand"},
         {209, &Hid::BeginPermitVibrationSession, "BeginPermitVibrationSession"},
         {210, &Hid::EndPermitVibrationSession, "EndPermitVibrationSession"},
         {211, &Hid::IsVibrationDeviceMounted, "IsVibrationDeviceMounted"},
@@ -1093,7 +1093,22 @@ void Hid::GetVibrationDeviceInfo(Kernel::HLERequestContext& ctx) {
 
     VibrationDeviceInfo vibration_device_info;
 
-    vibration_device_info.type = VibrationDeviceType::LinearResonantActuator;
+    switch (vibration_device_handle.npad_type) {
+    case Controller_NPad::NpadType::ProController:
+    case Controller_NPad::NpadType::Handheld:
+    case Controller_NPad::NpadType::JoyconDual:
+    case Controller_NPad::NpadType::JoyconLeft:
+    case Controller_NPad::NpadType::JoyconRight:
+    default:
+        vibration_device_info.type = VibrationDeviceType::LinearResonantActuator;
+        break;
+    case Controller_NPad::NpadType::GameCube:
+        vibration_device_info.type = VibrationDeviceType::GcErm;
+        break;
+    case Controller_NPad::NpadType::Pokeball:
+        vibration_device_info.type = VibrationDeviceType::Unknown;
+        break;
+    }
 
     switch (vibration_device_handle.device_index) {
     case Controller_NPad::DeviceIndex::Left:
@@ -1215,6 +1230,108 @@ void Hid::SendVibrationValues(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
+void Hid::SendVibrationGcErmCommand(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    struct Parameters {
+        Controller_NPad::DeviceHandle vibration_device_handle;
+        u64 applet_resource_user_id;
+        VibrationGcErmCommand gc_erm_command;
+    };
+    static_assert(sizeof(Parameters) == 0x18, "Parameters has incorrect size.");
+
+    const auto parameters{rp.PopRaw<Parameters>()};
+
+    /**
+     * Note: This uses yuzu-specific behavior such that the StopHard command produces
+     * vibrations where freq_low == 0.0f and freq_high == 0.0f, as defined below,
+     * in order to differentiate between Stop and StopHard commands.
+     * This is done to reuse the controller vibration functions made for regular controllers.
+     */
+    const auto vibration_value = [parameters] {
+        switch (parameters.gc_erm_command) {
+        case VibrationGcErmCommand::Stop:
+            return Controller_NPad::VibrationValue{
+                .amp_low = 0.0f,
+                .freq_low = 160.0f,
+                .amp_high = 0.0f,
+                .freq_high = 320.0f,
+            };
+        case VibrationGcErmCommand::Start:
+            return Controller_NPad::VibrationValue{
+                .amp_low = 1.0f,
+                .freq_low = 160.0f,
+                .amp_high = 1.0f,
+                .freq_high = 320.0f,
+            };
+        case VibrationGcErmCommand::StopHard:
+            return Controller_NPad::VibrationValue{
+                .amp_low = 0.0f,
+                .freq_low = 0.0f,
+                .amp_high = 0.0f,
+                .freq_high = 0.0f,
+            };
+        default:
+            return Controller_NPad::DEFAULT_VIBRATION_VALUE;
+        }
+    }();
+
+    applet_resource->GetController<Controller_NPad>(HidController::NPad)
+        .VibrateController(parameters.vibration_device_handle, vibration_value);
+
+    LOG_DEBUG(Service_HID,
+              "called, npad_type={}, npad_id={}, device_index={}, applet_resource_user_id={}, "
+              "gc_erm_command={}",
+              parameters.vibration_device_handle.npad_type,
+              parameters.vibration_device_handle.npad_id,
+              parameters.vibration_device_handle.device_index, parameters.applet_resource_user_id,
+              parameters.gc_erm_command);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void Hid::GetActualVibrationGcErmCommand(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    struct Parameters {
+        Controller_NPad::DeviceHandle vibration_device_handle;
+        INSERT_PADDING_WORDS_NOINIT(1);
+        u64 applet_resource_user_id;
+    };
+
+    const auto parameters{rp.PopRaw<Parameters>()};
+
+    const auto last_vibration = applet_resource->GetController<Controller_NPad>(HidController::NPad)
+                                    .GetLastVibration(parameters.vibration_device_handle);
+
+    const auto gc_erm_command = [last_vibration] {
+        if (last_vibration.amp_low != 0.0f || last_vibration.amp_high != 0.0f) {
+            return VibrationGcErmCommand::Start;
+        }
+
+        /**
+         * Note: This uses yuzu-specific behavior such that the StopHard command produces
+         * vibrations where freq_low == 0.0f and freq_high == 0.0f, as defined in the HID function
+         * SendVibrationGcErmCommand, in order to differentiate between Stop and StopHard commands.
+         * This is done to reuse the controller vibration functions made for regular controllers.
+         */
+        if (last_vibration.freq_low == 0.0f && last_vibration.freq_high == 0.0f) {
+            return VibrationGcErmCommand::StopHard;
+        }
+
+        return VibrationGcErmCommand::Stop;
+    }();
+
+    LOG_DEBUG(Service_HID,
+              "called, npad_type={}, npad_id={}, device_index={}, applet_resource_user_id={}",
+              parameters.vibration_device_handle.npad_type,
+              parameters.vibration_device_handle.npad_id,
+              parameters.vibration_device_handle.device_index, parameters.applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushEnum(gc_erm_command);
+}
+
 void Hid::BeginPermitVibrationSession(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto applet_resource_user_id{rp.Pop<u64>()};

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

@@ -136,6 +136,8 @@ private:
     void PermitVibration(Kernel::HLERequestContext& ctx);
     void IsVibrationPermitted(Kernel::HLERequestContext& ctx);
     void SendVibrationValues(Kernel::HLERequestContext& ctx);
+    void SendVibrationGcErmCommand(Kernel::HLERequestContext& ctx);
+    void GetActualVibrationGcErmCommand(Kernel::HLERequestContext& ctx);
     void BeginPermitVibrationSession(Kernel::HLERequestContext& ctx);
     void EndPermitVibrationSession(Kernel::HLERequestContext& ctx);
     void IsVibrationDeviceMounted(Kernel::HLERequestContext& ctx);
@@ -154,7 +156,9 @@ private:
     void GetNpadCommunicationMode(Kernel::HLERequestContext& ctx);
 
     enum class VibrationDeviceType : u32 {
+        Unknown = 0,
         LinearResonantActuator = 1,
+        GcErm = 2,
     };
 
     enum class VibrationDevicePosition : u32 {
@@ -163,6 +167,12 @@ private:
         Right = 2,
     };
 
+    enum class VibrationGcErmCommand : u64 {
+        Stop = 0,
+        Start = 1,
+        StopHard = 2,
+    };
+
     struct VibrationDeviceInfo {
         VibrationDeviceType type{};
         VibrationDevicePosition position{};

src/core/hle/service/ldn/ldn.cpp

@@ -156,7 +156,7 @@ public:
         is_initialized = true;
 
         IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(RESULT_SUCCESS);
+        rb.Push(ERROR_DISABLED);
     }
 
 private:

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

@@ -279,6 +279,10 @@ const SharedMemory& TimeManager::GetSharedMemory() const {
     return impl->shared_memory;
 }
 
+void TimeManager::Shutdown() {
+    impl.reset();
+}
+
 void TimeManager::UpdateLocalSystemClockTime(s64 posix_time) {
     impl->UpdateLocalSystemClockTime(system, posix_time);
 }

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

@@ -61,6 +61,8 @@ public:
 
     const SharedMemory& GetSharedMemory() const;
 
+    void Shutdown();
+
     void SetupTimeZoneManager(std::string location_name,
                               Clock::SteadyClockTimePoint time_zone_updated_time_point,
                               std::size_t total_location_name_count, u128 time_zone_rule_version,

src/video_core/gpu.cpp

@@ -459,7 +459,7 @@ void GPU::ProcessSemaphoreAcquire() {
 }
 
 void GPU::Start() {
-    gpu_thread.StartThread(*renderer, renderer->Context(), *dma_pusher, *cdma_pusher);
+    gpu_thread.StartThread(*renderer, renderer->Context(), *dma_pusher);
     cpu_context = renderer->GetRenderWindow().CreateSharedContext();
     cpu_context->MakeCurrent();
 }

src/video_core/gpu_thread.cpp

@@ -19,7 +19,7 @@ namespace VideoCommon::GPUThread {
 /// Runs the GPU thread
 static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
                       Core::Frontend::GraphicsContext& context, Tegra::DmaPusher& dma_pusher,
-                      SynchState& state, Tegra::CDmaPusher& cdma_pusher) {
+                      SynchState& state) {
     std::string name = "yuzu:GPU";
     MicroProfileOnThreadCreate(name.c_str());
     SCOPE_EXIT({ MicroProfileOnThreadExit(); });
@@ -46,9 +46,6 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
         if (auto* submit_list = std::get_if<SubmitListCommand>(&next.data)) {
             dma_pusher.Push(std::move(submit_list->entries));
             dma_pusher.DispatchCalls();
-        } else if (auto* command_list = std::get_if<SubmitChCommandEntries>(&next.data)) {
-            // NVDEC
-            cdma_pusher.ProcessEntries(std::move(command_list->entries));
         } else if (const auto* data = std::get_if<SwapBuffersCommand>(&next.data)) {
             renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr);
         } else if (std::holds_alternative<OnCommandListEndCommand>(next.data)) {
@@ -83,20 +80,16 @@ ThreadManager::~ThreadManager() {
 
 void ThreadManager::StartThread(VideoCore::RendererBase& renderer,
                                 Core::Frontend::GraphicsContext& context,
-                                Tegra::DmaPusher& dma_pusher, Tegra::CDmaPusher& cdma_pusher) {
+                                Tegra::DmaPusher& dma_pusher) {
     rasterizer = renderer.ReadRasterizer();
     thread = std::thread(RunThread, std::ref(system), std::ref(renderer), std::ref(context),
-                         std::ref(dma_pusher), std::ref(state), std::ref(cdma_pusher));
+                         std::ref(dma_pusher), std::ref(state));
 }
 
 void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
     PushCommand(SubmitListCommand(std::move(entries)));
 }
 
-void ThreadManager::SubmitCommandBuffer(Tegra::ChCommandHeaderList&& entries) {
-    PushCommand(SubmitChCommandEntries(std::move(entries)));
-}
-
 void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     PushCommand(SwapBuffersCommand(framebuffer ? std::make_optional(*framebuffer) : std::nullopt));
 }

src/video_core/gpu_thread.h

@@ -43,14 +43,6 @@ struct SubmitListCommand final {
     Tegra::CommandList entries;
 };
 
-/// Command to signal to the GPU thread that a cdma command list is ready for processing
-struct SubmitChCommandEntries final {
-    explicit SubmitChCommandEntries(Tegra::ChCommandHeaderList&& entries_)
-        : entries{std::move(entries_)} {}
-
-    Tegra::ChCommandHeaderList entries;
-};
-
 /// Command to signal to the GPU thread that a swap buffers is pending
 struct SwapBuffersCommand final {
     explicit SwapBuffersCommand(std::optional<const Tegra::FramebufferConfig> framebuffer_)
@@ -91,9 +83,9 @@ struct OnCommandListEndCommand final {};
 struct GPUTickCommand final {};
 
 using CommandData =
-    std::variant<EndProcessingCommand, SubmitListCommand, SubmitChCommandEntries,
-                 SwapBuffersCommand, FlushRegionCommand, InvalidateRegionCommand,
-                 FlushAndInvalidateRegionCommand, OnCommandListEndCommand, GPUTickCommand>;
+    std::variant<EndProcessingCommand, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
+                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand, OnCommandListEndCommand,
+                 GPUTickCommand>;
 
 struct CommandDataContainer {
     CommandDataContainer() = default;
@@ -123,14 +115,11 @@ public:
 
     /// Creates and starts the GPU thread.
     void StartThread(VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
-                     Tegra::DmaPusher& dma_pusher, Tegra::CDmaPusher& cdma_pusher);
+                     Tegra::DmaPusher& dma_pusher);
 
     /// Push GPU command entries to be processed
     void SubmitList(Tegra::CommandList&& entries);
 
-    /// Push GPU CDMA command buffer entries to be processed
-    void SubmitCommandBuffer(Tegra::ChCommandHeaderList&& entries);
-
     /// Swap buffers (render frame)
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
 

src/video_core/renderer_opengl/gl_device.cpp

@@ -239,6 +239,7 @@ Device::Device() {
     has_nv_viewport_array2 = GLAD_GL_NV_viewport_array2;
     has_vertex_buffer_unified_memory = GLAD_GL_NV_vertex_buffer_unified_memory;
     has_debugging_tool_attached = IsDebugToolAttached(extensions);
+    has_depth_buffer_float = HasExtension(extensions, "GL_NV_depth_buffer_float");
 
     // At the moment of writing this, only Nvidia's driver optimizes BufferSubData on exclusive
     // uniform buffers as "push constants"
@@ -275,6 +276,7 @@ Device::Device(std::nullptr_t) {
     has_image_load_formatted = true;
     has_texture_shadow_lod = true;
     has_variable_aoffi = true;
+    has_depth_buffer_float = true;
 }
 
 bool Device::TestVariableAoffi() {

src/video_core/renderer_opengl/gl_device.h

@@ -122,6 +122,10 @@ public:
         return use_driver_cache;
     }
 
+    bool HasDepthBufferFloat() const {
+        return has_depth_buffer_float;
+    }
+
 private:
     static bool TestVariableAoffi();
     static bool TestPreciseBug();
@@ -150,6 +154,7 @@ private:
     bool use_assembly_shaders{};
     bool use_asynchronous_shaders{};
     bool use_driver_cache{};
+    bool has_depth_buffer_float{};
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -889,7 +889,11 @@ void RasterizerOpenGL::SyncViewport() {
             const GLdouble reduce_z = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne;
             const GLdouble near_depth = src.translate_z - src.scale_z * reduce_z;
             const GLdouble far_depth = src.translate_z + src.scale_z;
-            glDepthRangeIndexed(static_cast<GLuint>(i), near_depth, far_depth);
+            if (device.HasDepthBufferFloat()) {
+                glDepthRangeIndexeddNV(static_cast<GLuint>(i), near_depth, far_depth);
+            } else {
+                glDepthRangeIndexed(static_cast<GLuint>(i), near_depth, far_depth);
+            }
 
             if (!GLAD_GL_NV_viewport_swizzle) {
                 continue;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -335,6 +335,10 @@ void ShaderCacheOpenGL::LoadDiskCache(u64 title_id, const std::atomic_bool& stop
                                       const VideoCore::DiskResourceLoadCallback& callback) {
     disk_cache.BindTitleID(title_id);
     const std::optional transferable = disk_cache.LoadTransferable();
+
+    LOG_INFO(Render_OpenGL, "Total Shader Count: {}",
+             transferable.has_value() ? transferable->size() : 0);
+
     if (!transferable) {
         return;
     }

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -201,10 +201,6 @@ void BufferCacheRuntime::BindTransformFeedbackBuffer(u32 index, VkBuffer buffer,
     });
 }
 
-void BufferCacheRuntime::BindBuffer(VkBuffer buffer, u32 offset, u32 size) {
-    update_descriptor_queue.AddBuffer(buffer, offset, size);
-}
-
 void BufferCacheRuntime::ReserveQuadArrayLUT(u32 num_indices, bool wait_for_idle) {
     if (num_indices <= current_num_indices) {
         return;

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -8,6 +8,7 @@
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
+#include "video_core/renderer_vulkan/vk_update_descriptor.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
@@ -16,7 +17,6 @@ namespace Vulkan {
 class Device;
 class VKDescriptorPool;
 class VKScheduler;
-class VKUpdateDescriptorQueue;
 
 class BufferCacheRuntime;
 
@@ -86,7 +86,9 @@ public:
     }
 
 private:
-    void BindBuffer(VkBuffer buffer, u32 offset, u32 size);
+    void BindBuffer(VkBuffer buffer, u32 offset, u32 size) {
+        update_descriptor_queue.AddBuffer(buffer, offset, size);
+    }
 
     void ReserveQuadArrayLUT(u32 num_indices, bool wait_for_idle);
 

src/video_core/renderer_vulkan/vk_shader_decompiler.cpp

@@ -1845,13 +1845,21 @@ private:
 
     Expression TextureGather(Operation operation) {
         const auto& meta = std::get<MetaTexture>(operation.GetMeta());
-        UNIMPLEMENTED_IF(!meta.aoffi.empty());
 
         const Id coords = GetCoordinates(operation, Type::Float);
+
+        spv::ImageOperandsMask mask = spv::ImageOperandsMask::MaskNone;
+        std::vector<Id> operands;
         Id texture{};
+
+        if (!meta.aoffi.empty()) {
+            mask = mask | spv::ImageOperandsMask::Offset;
+            operands.push_back(GetOffsetCoordinates(operation));
+        }
+
         if (meta.sampler.is_shadow) {
             texture = OpImageDrefGather(t_float4, GetTextureSampler(operation), coords,
-                                        AsFloat(Visit(meta.depth_compare)));
+                                        AsFloat(Visit(meta.depth_compare)), mask, operands);
         } else {
             u32 component_value = 0;
             if (meta.component) {
@@ -1860,7 +1868,7 @@ private:
                 component_value = component->GetValue();
             }
             texture = OpImageGather(t_float4, GetTextureSampler(operation), coords,
-                                    Constant(t_uint, component_value));
+                                    Constant(t_uint, component_value), mask, operands);
         }
         return GetTextureElement(operation, texture, Type::Float);
     }
@@ -1928,13 +1936,22 @@ private:
 
         const Id image = GetTextureImage(operation);
         const Id coords = GetCoordinates(operation, Type::Int);
+
+        spv::ImageOperandsMask mask = spv::ImageOperandsMask::MaskNone;
+        std::vector<Id> operands;
         Id fetch;
+
         if (meta.lod && !meta.sampler.is_buffer) {
-            fetch = OpImageFetch(t_float4, image, coords, spv::ImageOperandsMask::Lod,
-                                 AsInt(Visit(meta.lod)));
-        } else {
-            fetch = OpImageFetch(t_float4, image, coords);
+            mask = mask | spv::ImageOperandsMask::Lod;
+            operands.push_back(AsInt(Visit(meta.lod)));
         }
+
+        if (!meta.aoffi.empty()) {
+            mask = mask | spv::ImageOperandsMask::Offset;
+            operands.push_back(GetOffsetCoordinates(operation));
+        }
+
+        fetch = OpImageFetch(t_float4, image, coords, mask, operands);
         return GetTextureElement(operation, fetch, Type::Float);
     }
 

src/video_core/renderer_vulkan/vk_update_descriptor.cpp

@@ -20,20 +20,20 @@ VKUpdateDescriptorQueue::VKUpdateDescriptorQueue(const Device& device_, VKSchedu
 VKUpdateDescriptorQueue::~VKUpdateDescriptorQueue() = default;
 
 void VKUpdateDescriptorQueue::TickFrame() {
-    payload.clear();
+    payload_cursor = payload.data();
 }
 
 void VKUpdateDescriptorQueue::Acquire() {
     // Minimum number of entries required.
     // This is the maximum number of entries a single draw call migth use.
-    static constexpr std::size_t MIN_ENTRIES = 0x400;
+    static constexpr size_t MIN_ENTRIES = 0x400;
 
-    if (payload.size() + MIN_ENTRIES >= payload.max_size()) {
+    if (std::distance(payload.data(), payload_cursor) + MIN_ENTRIES >= payload.max_size()) {
         LOG_WARNING(Render_Vulkan, "Payload overflow, waiting for worker thread");
         scheduler.WaitWorker();
-        payload.clear();
+        payload_cursor = payload.data();
     }
-    upload_start = &*payload.end();
+    upload_start = payload_cursor;
 }
 
 void VKUpdateDescriptorQueue::Send(VkDescriptorUpdateTemplateKHR update_template,

src/video_core/renderer_vulkan/vk_update_descriptor.h

@@ -4,8 +4,7 @@
 
 #pragma once
 
-#include <variant>
-#include <boost/container/static_vector.hpp>
+#include <array>
 
 #include "common/common_types.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
@@ -16,13 +15,15 @@ class Device;
 class VKScheduler;
 
 struct DescriptorUpdateEntry {
+    struct Empty {};
+
+    DescriptorUpdateEntry() = default;
     DescriptorUpdateEntry(VkDescriptorImageInfo image_) : image{image_} {}
-
     DescriptorUpdateEntry(VkDescriptorBufferInfo buffer_) : buffer{buffer_} {}
-
     DescriptorUpdateEntry(VkBufferView texel_buffer_) : texel_buffer{texel_buffer_} {}
 
     union {
+        Empty empty{};
         VkDescriptorImageInfo image;
         VkDescriptorBufferInfo buffer;
         VkBufferView texel_buffer;
@@ -41,39 +42,40 @@ public:
     void Send(VkDescriptorUpdateTemplateKHR update_template, VkDescriptorSet set);
 
     void AddSampledImage(VkImageView image_view, VkSampler sampler) {
-        payload.emplace_back(VkDescriptorImageInfo{
+        *(payload_cursor++) = VkDescriptorImageInfo{
             .sampler = sampler,
             .imageView = image_view,
             .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
-        });
+        };
     }
 
     void AddImage(VkImageView image_view) {
-        payload.emplace_back(VkDescriptorImageInfo{
+        *(payload_cursor++) = VkDescriptorImageInfo{
             .sampler = VK_NULL_HANDLE,
             .imageView = image_view,
             .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
-        });
+        };
     }
 
-    void AddBuffer(VkBuffer buffer, u64 offset, size_t size) {
-        payload.emplace_back(VkDescriptorBufferInfo{
+    void AddBuffer(VkBuffer buffer, VkDeviceSize offset, VkDeviceSize size) {
+        *(payload_cursor++) = VkDescriptorBufferInfo{
             .buffer = buffer,
             .offset = offset,
             .range = size,
-        });
+        };
     }
 
     void AddTexelBuffer(VkBufferView texel_buffer) {
-        payload.emplace_back(texel_buffer);
+        *(payload_cursor++) = texel_buffer;
     }
 
 private:
     const Device& device;
     VKScheduler& scheduler;
 
+    DescriptorUpdateEntry* payload_cursor = nullptr;
     const DescriptorUpdateEntry* upload_start = nullptr;
-    boost::container::static_vector<DescriptorUpdateEntry, 0x10000> payload;
+    std::array<DescriptorUpdateEntry, 0x10000> payload;
 };
 
 } // namespace Vulkan

src/video_core/shader/decode/memory.cpp

@@ -330,6 +330,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         case StoreType::Bits32:
             (this->*set_memory)(bb, GetAddress(0), GetRegister(instr.gpr0));
             break;
+        case StoreType::Unsigned16:
         case StoreType::Signed16: {
             Node address = GetAddress(0);
             Node memory = (this->*get_memory)(address);

src/video_core/shader/decode/texture.cpp

@@ -806,6 +806,7 @@ Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is
 
     const std::size_t type_coord_count = GetCoordCount(texture_type);
     const bool lod_enabled = instr.tlds.GetTextureProcessMode() == TextureProcessMode::LL;
+    const bool aoffi_enabled = instr.tlds.UsesMiscMode(TextureMiscMode::AOFFI);
 
     // If enabled arrays index is always stored in the gpr8 field
     const u64 array_register = instr.gpr8.Value();
@@ -820,17 +821,23 @@ Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is
     std::vector<Node> coords;
     for (std::size_t i = 0; i < type_coord_count; ++i) {
         const bool last = (i == (type_coord_count - 1)) && (type_coord_count > 1);
-        coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
+        coords.push_back(
+            GetRegister(last && !aoffi_enabled ? last_coord_register : coord_register + i));
     }
 
     const Node array = is_array ? GetRegister(array_register) : nullptr;
     // When lod is used always is in gpr20
     const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0);
 
+    std::vector<Node> aoffi;
+    if (aoffi_enabled) {
+        aoffi = GetAoffiCoordinates(GetRegister(instr.gpr20), type_coord_count, false);
+    }
+
     Node4 values;
     for (u32 element = 0; element < values.size(); ++element) {
         auto coords_copy = coords;
-        MetaTexture meta{*sampler, array, {}, {}, {}, {}, {}, lod, {}, element, {}};
+        MetaTexture meta{*sampler, array, {}, aoffi, {}, {}, {}, lod, {}, element, {}};
         values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy));
     }
     return values;

src/video_core/shader_notify.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <mutex>
 #include "video_core/shader_notify.h"
 
 using namespace std::chrono_literals;

src/yuzu/configuration/config.cpp

@@ -648,7 +648,7 @@ void Config::ReadDebuggingValues() {
 void Config::ReadServiceValues() {
     qt_config->beginGroup(QStringLiteral("Services"));
     Settings::values.bcat_backend =
-        ReadSetting(QStringLiteral("bcat_backend"), QStringLiteral("null"))
+        ReadSetting(QStringLiteral("bcat_backend"), QStringLiteral("none"))
             .toString()
             .toStdString();
     Settings::values.bcat_boxcat_local =
@@ -1239,7 +1239,7 @@ void Config::SaveDebuggingValues() {
 void Config::SaveServiceValues() {
     qt_config->beginGroup(QStringLiteral("Services"));
     WriteSetting(QStringLiteral("bcat_backend"),
-                 QString::fromStdString(Settings::values.bcat_backend), QStringLiteral("null"));
+                 QString::fromStdString(Settings::values.bcat_backend), QStringLiteral("none"));
     WriteSetting(QStringLiteral("bcat_boxcat_local"), Settings::values.bcat_boxcat_local, false);
     qt_config->endGroup();
 }

src/yuzu_cmd/config.cpp

@@ -457,7 +457,7 @@ void Config::ReadValues() {
     Settings::values.yuzu_token = sdl2_config->Get("WebService", "yuzu_token", "");
 
     // Services
-    Settings::values.bcat_backend = sdl2_config->Get("Services", "bcat_backend", "null");
+    Settings::values.bcat_backend = sdl2_config->Get("Services", "bcat_backend", "none");
     Settings::values.bcat_boxcat_local =
         sdl2_config->GetBoolean("Services", "bcat_boxcat_local", false);
 }