20 files changed, 265 insertions, 47 deletions

diff --git a/src/common/uint128.h b/src/common/uint128.h
index 83560a9ce..4780b2f9d 100644
--- a/src/common/uint128.h
+++ b/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
diff --git a/src/common/wall_clock.cpp b/src/common/wall_clock.cpp
index 1545993bd..49830b8ab 100644
--- a/src/common/wall_clock.cpp
+++ b/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
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 30f5e1128..de6305e2a 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -308,6 +308,9 @@ struct System::Impl {
         // Close all CPU/threading state
         cpu_manager.Shutdown();
 
+        // Release the Time Manager's resources
+        time_manager.Shutdown();
+
         // Shutdown kernel and core timing
         core_timing.Shutdown();
         kernel.Shutdown();
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 453695545..331cf3a60 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/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;
     }
diff --git a/src/core/hle/service/acc/acc.cpp b/src/core/hle/service/acc/acc.cpp
index 3ec0e1eca..615e20a54 100644
--- a/src/core/hle/service/acc/acc.cpp
+++ b/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};
diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp
index ffc3dfdc3..ba27bbb05 100644
--- a/src/core/hle/service/hid/hid.cpp
+++ b/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>()};
diff --git a/src/core/hle/service/hid/hid.h b/src/core/hle/service/hid/hid.h
index 06ddcf3e4..36ed228c8 100644
--- a/src/core/hle/service/hid/hid.h
+++ b/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{};
diff --git a/src/core/hle/service/time/time_manager.cpp b/src/core/hle/service/time/time_manager.cpp
index 858623e2b..1f7309f6b 100644
--- a/src/core/hle/service/time/time_manager.cpp
+++ b/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);
 }
diff --git a/src/core/hle/service/time/time_manager.h b/src/core/hle/service/time/time_manager.h
index 993c7c288..4db8cc0e1 100644
--- a/src/core/hle/service/time/time_manager.h
+++ b/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,
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index 48d5c4a5e..1ae5f1d62 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/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() {
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index ee053776d..f24bd0c7b 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/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
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 418644108..4610fd160 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/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;
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 529570ff0..5cf7cd151 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/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;
     }
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
index 848eedd66..668633e7b 100644
--- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
+++ b/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;
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h
index 041e6515c..982e92191 100644
--- a/src/video_core/renderer_vulkan/vk_buffer_cache.h
+++ b/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);
 
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 40e2e0d38..c6846d886 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/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);
     }
 
diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.cpp b/src/video_core/renderer_vulkan/vk_update_descriptor.cpp
index f99273c6a..dc45fdcb1 100644
--- a/src/video_core/renderer_vulkan/vk_update_descriptor.cpp
+++ b/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,
diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.h b/src/video_core/renderer_vulkan/vk_update_descriptor.h
index e214f7195..d35e77c44 100644
--- a/src/video_core/renderer_vulkan/vk_update_descriptor.h
+++ b/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 {
-    DescriptorUpdateEntry(VkDescriptorImageInfo image_) : image{image_} {}
+    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
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index 50f4e7d35..7728f600e 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/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);
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 833fa2a39..c69681e8d 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/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;