early-access version 3905

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 3904.
+This is the source code for early-access 3905.
 
 ## Legal Notice
 

src/common/settings.h

@@ -353,6 +353,8 @@ struct Values {
                                           Category::RendererDebug};
     // TODO: remove this once AMDVLK supports VK_EXT_depth_bias_control
     bool renderer_amdvlk_depth_bias_workaround{};
+    Setting<bool> disable_buffer_reorder{linkage, false, "disable_buffer_reorder",
+                                         Category::RendererDebug};
 
     // System
     SwitchableSetting<Language, true> language_index{linkage,

src/core/CMakeLists.txt

@@ -698,6 +698,8 @@ add_library(core STATIC
     hle/service/nvnflinger/consumer_base.cpp
     hle/service/nvnflinger/consumer_base.h
     hle/service/nvnflinger/consumer_listener.h
+    hle/service/nvnflinger/fb_share_buffer_manager.cpp
+    hle/service/nvnflinger/fb_share_buffer_manager.h
     hle/service/nvnflinger/graphic_buffer_producer.cpp
     hle/service/nvnflinger/graphic_buffer_producer.h
     hle/service/nvnflinger/hos_binder_driver_server.cpp

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

@@ -19,6 +19,7 @@
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applet_ae.h"
 #include "core/hle/service/am/applet_oe.h"
+#include "core/hle/service/am/applets/applet_cabinet.h"
 #include "core/hle/service/am/applets/applet_mii_edit_types.h"
 #include "core/hle/service/am/applets/applet_profile_select.h"
 #include "core/hle/service/am/applets/applet_web_browser.h"
@@ -33,11 +34,13 @@
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/hle/service/ns/ns.h"
+#include "core/hle/service/nvnflinger/fb_share_buffer_manager.h"
 #include "core/hle/service/nvnflinger/nvnflinger.h"
 #include "core/hle/service/pm/pm.h"
 #include "core/hle/service/server_manager.h"
 #include "core/hle/service/sm/sm.h"
 #include "core/hle/service/vi/vi.h"
+#include "core/hle/service/vi/vi_results.h"
 #include "core/memory.h"
 
 namespace Service::AM {
@@ -190,7 +193,7 @@ IDisplayController::IDisplayController(Core::System& system_)
         {4, nullptr, "UpdateCallerAppletCaptureImage"},
         {5, nullptr, "GetLastForegroundCaptureImageEx"},
         {6, nullptr, "GetLastApplicationCaptureImageEx"},
-        {7, nullptr, "GetCallerAppletCaptureImageEx"},
+        {7, &IDisplayController::GetCallerAppletCaptureImageEx, "GetCallerAppletCaptureImageEx"},
         {8, &IDisplayController::TakeScreenShotOfOwnLayer, "TakeScreenShotOfOwnLayer"},
         {9, nullptr, "CopyBetweenCaptureBuffers"},
         {10, nullptr, "AcquireLastApplicationCaptureBuffer"},
@@ -208,8 +211,8 @@ IDisplayController::IDisplayController(Core::System& system_)
         {23, nullptr, "ReleaseLastApplicationCaptureSharedBuffer"},
         {24, nullptr, "AcquireLastForegroundCaptureSharedBuffer"},
         {25, nullptr, "ReleaseLastForegroundCaptureSharedBuffer"},
-        {26, nullptr, "AcquireCallerAppletCaptureSharedBuffer"},
+        {26, &IDisplayController::AcquireCallerAppletCaptureSharedBuffer, "AcquireCallerAppletCaptureSharedBuffer"},
-        {27, nullptr, "ReleaseCallerAppletCaptureSharedBuffer"},
+        {27, &IDisplayController::ReleaseCallerAppletCaptureSharedBuffer, "ReleaseCallerAppletCaptureSharedBuffer"},
         {28, nullptr, "TakeScreenShotOfOwnLayerEx"},
     };
     // clang-format on
@@ -219,6 +222,15 @@ IDisplayController::IDisplayController(Core::System& system_)
 
 IDisplayController::~IDisplayController() = default;
 
+void IDisplayController::GetCallerAppletCaptureImageEx(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push(1u);
+    rb.Push(0);
+}
+
 void IDisplayController::TakeScreenShotOfOwnLayer(HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
 
@@ -226,6 +238,22 @@ void IDisplayController::TakeScreenShotOfOwnLayer(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void IDisplayController::AcquireCallerAppletCaptureSharedBuffer(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push(1U);
+    rb.Push(0);
+}
+
+void IDisplayController::ReleaseCallerAppletCaptureSharedBuffer(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
 IDebugFunctions::IDebugFunctions(Core::System& system_)
     : ServiceFramework{system_, "IDebugFunctions"} {
     // clang-format off
@@ -285,14 +313,14 @@ ISelfController::ISelfController(Core::System& system_, Nvnflinger::Nvnflinger&
         {20, nullptr, "SetDesirableKeyboardLayout"},
         {21, nullptr, "GetScreenShotProgramId"},
         {40, &ISelfController::CreateManagedDisplayLayer, "CreateManagedDisplayLayer"},
-        {41, nullptr, "IsSystemBufferSharingEnabled"},
+        {41, &ISelfController::IsSystemBufferSharingEnabled, "IsSystemBufferSharingEnabled"},
-        {42, nullptr, "GetSystemSharedLayerHandle"},
+        {42, &ISelfController::GetSystemSharedLayerHandle, "GetSystemSharedLayerHandle"},
-        {43, nullptr, "GetSystemSharedBufferHandle"},
+        {43, &ISelfController::GetSystemSharedBufferHandle, "GetSystemSharedBufferHandle"},
         {44, &ISelfController::CreateManagedDisplaySeparableLayer, "CreateManagedDisplaySeparableLayer"},
         {45, nullptr, "SetManagedDisplayLayerSeparationMode"},
         {46, nullptr, "SetRecordingLayerCompositionEnabled"},
         {50, &ISelfController::SetHandlesRequestToDisplay, "SetHandlesRequestToDisplay"},
-        {51, nullptr, "ApproveToDisplay"},
+        {51, &ISelfController::ApproveToDisplay, "ApproveToDisplay"},
         {60, nullptr, "OverrideAutoSleepTimeAndDimmingTime"},
         {61, nullptr, "SetMediaPlaybackState"},
         {62, &ISelfController::SetIdleTimeDetectionExtension, "SetIdleTimeDetectionExtension"},
@@ -491,6 +519,44 @@ void ISelfController::CreateManagedDisplayLayer(HLERequestContext& ctx) {
     rb.Push(*layer_id);
 }
 
+void ISelfController::IsSystemBufferSharingEnabled(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(this->EnsureBufferSharingEnabled());
+}
+
+void ISelfController::GetSystemSharedLayerHandle(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 6};
+    rb.Push(this->EnsureBufferSharingEnabled());
+    rb.Push<s64>(system_shared_buffer_id);
+    rb.Push<s64>(system_shared_layer_id);
+}
+
+void ISelfController::GetSystemSharedBufferHandle(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(this->EnsureBufferSharingEnabled());
+    rb.Push<s64>(system_shared_buffer_id);
+}
+
+Result ISelfController::EnsureBufferSharingEnabled() {
+    R_SUCCEED_IF(buffer_sharing_enabled);
+    R_UNLESS(system.GetAppletManager().GetCurrentAppletId() > Applets::AppletId::Application,
+             VI::ResultOperationFailed);
+
+    ON_RESULT_SUCCESS {
+        buffer_sharing_enabled = true;
+    };
+
+    const auto display_id = nvnflinger.OpenDisplay("Default");
+    R_RETURN(nvnflinger.GetSystemBufferManager().Initialize(&system_shared_buffer_id,
+                                                            &system_shared_layer_id, *display_id));
+}
+
 void ISelfController::CreateManagedDisplaySeparableLayer(HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
 
@@ -516,6 +582,13 @@ void ISelfController::SetHandlesRequestToDisplay(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void ISelfController::ApproveToDisplay(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
 void ISelfController::SetIdleTimeDetectionExtension(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     idle_time_detection_extension = rp.Pop<u32>();
@@ -686,7 +759,8 @@ void AppletMessageQueue::OperationModeChanged() {
 
 ICommonStateGetter::ICommonStateGetter(Core::System& system_,
                                        std::shared_ptr<AppletMessageQueue> msg_queue_)
-    : ServiceFramework{system_, "ICommonStateGetter"}, msg_queue{std::move(msg_queue_)} {
+    : ServiceFramework{system_, "ICommonStateGetter"}, msg_queue{std::move(msg_queue_)},
+      service_context{system_, "ICommonStateGetter"} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {0, &ICommonStateGetter::GetEventHandle, "GetEventHandle"},
@@ -699,10 +773,10 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system_,
         {7, nullptr, "GetCradleStatus"},
         {8, &ICommonStateGetter::GetBootMode, "GetBootMode"},
         {9, &ICommonStateGetter::GetCurrentFocusState, "GetCurrentFocusState"},
-        {10, nullptr, "RequestToAcquireSleepLock"},
+        {10, &ICommonStateGetter::RequestToAcquireSleepLock, "RequestToAcquireSleepLock"},
         {11, nullptr, "ReleaseSleepLock"},
         {12, nullptr, "ReleaseSleepLockTransiently"},
-        {13, nullptr, "GetAcquiredSleepLockEvent"},
+        {13, &ICommonStateGetter::GetAcquiredSleepLockEvent, "GetAcquiredSleepLockEvent"},
         {14, nullptr, "GetWakeupCount"},
         {20, nullptr, "PushToGeneralChannel"},
         {30, nullptr, "GetHomeButtonReaderLockAccessor"},
@@ -745,6 +819,8 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system_,
 
     RegisterHandlers(functions);
 
+    sleep_lock_event = service_context.CreateEvent("ICommonStateGetter::SleepLockEvent");
+
     // Configure applets to be in foreground state
     msg_queue->PushMessage(AppletMessageQueue::AppletMessage::FocusStateChanged);
     msg_queue->PushMessage(AppletMessageQueue::AppletMessage::ChangeIntoForeground);
@@ -793,6 +869,24 @@ void ICommonStateGetter::GetCurrentFocusState(HLERequestContext& ctx) {
     rb.Push(static_cast<u8>(FocusState::InFocus));
 }
 
+void ICommonStateGetter::RequestToAcquireSleepLock(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    // Sleep lock is acquired immediately.
+    sleep_lock_event->Signal();
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
+void ICommonStateGetter::GetAcquiredSleepLockEvent(HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "called");
+
+    IPC::ResponseBuilder rb{ctx, 2, 1};
+    rb.Push(ResultSuccess);
+    rb.PushCopyObjects(sleep_lock_event->GetReadableEvent());
+}
+
 void ICommonStateGetter::IsVrModeEnabled(HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called");
 
@@ -1385,7 +1479,16 @@ ILibraryAppletSelfAccessor::ILibraryAppletSelfAccessor(Core::System& system_)
     // clang-format on
     RegisterHandlers(functions);
 
+    switch (system.GetAppletManager().GetCurrentAppletId()) {
+    case Applets::AppletId::Cabinet:
+        PushInShowCabinetData();
+        break;
+    case Applets::AppletId::MiiEdit:
         PushInShowMiiEditData();
+        break;
+    default:
+        break;
+    }
 }
 
 ILibraryAppletSelfAccessor::~ILibraryAppletSelfAccessor() = default;
@@ -1431,7 +1534,7 @@ void ILibraryAppletSelfAccessor::GetLibraryAppletInfo(HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
 
     const LibraryAppletInfo applet_info{
-        .applet_id = Applets::AppletId::MiiEdit,
+        .applet_id = system.GetAppletManager().GetCurrentAppletId(),
         .library_applet_mode = Applets::LibraryAppletMode::AllForeground,
     };
 
@@ -1459,6 +1562,35 @@ void ILibraryAppletSelfAccessor::GetCallerAppletIdentityInfo(HLERequestContext&
     rb.PushRaw(applet_info);
 }
 
+void ILibraryAppletSelfAccessor::PushInShowCabinetData() {
+    constexpr Applets::Applet::CommonArguments arguments{
+        .arguments_version = 3,
+        .size = 0x20,
+        .library_version = 1,
+        .theme_color = 3,
+        .play_startup_sound = true,
+        .system_tick = 0,
+    };
+
+    const Applets::StartParamForAmiiboSettings amiibo_settings{
+        .param_1 = 0,
+        .applet_mode = system.GetAppletManager().GetCabinetMode(),
+        .flags = 0,
+        .amiibo_settings_1 = 0,
+        .device_handle = 0,
+        .tag_info{},
+        .register_info{},
+        .amiibo_settings_3{},
+    };
+
+    std::vector<u8> argument_data(sizeof(arguments));
+    std::vector<u8> settings_data(sizeof(amiibo_settings));
+    std::memcpy(argument_data.data(), &arguments, sizeof(arguments));
+    std::memcpy(settings_data.data(), &amiibo_settings, sizeof(amiibo_settings));
+    queue_data.emplace_back(std::move(argument_data));
+    queue_data.emplace_back(std::move(settings_data));
+}
+
 void ILibraryAppletSelfAccessor::PushInShowMiiEditData() {
     struct MiiEditV3 {
         Applets::MiiEditAppletInputCommon common;
@@ -2235,7 +2367,7 @@ void IProcessWindingController::GetLaunchReason(HLERequestContext& ctx) {
 }
 
 void IProcessWindingController::OpenCallingLibraryApplet(HLERequestContext& ctx) {
-    const auto applet_id = Applets::AppletId::MiiEdit;
+    const auto applet_id = system.GetAppletManager().GetCurrentAppletId();
     const auto applet_mode = Applets::LibraryAppletMode::AllForeground;
 
     LOG_WARNING(Service_AM, "(STUBBED) called with applet_id={:08X}, applet_mode={:08X}", applet_id,
@@ -2256,4 +2388,5 @@ void IProcessWindingController::OpenCallingLibraryApplet(HLERequestContext& ctx)
     rb.Push(ResultSuccess);
     rb.PushIpcInterface<ILibraryAppletAccessor>(system, applet);
 }
+
 } // namespace Service::AM

src/core/hle/service/am/am.h

@@ -122,7 +122,10 @@ public:
     ~IDisplayController() override;
 
 private:
+    void GetCallerAppletCaptureImageEx(HLERequestContext& ctx);
     void TakeScreenShotOfOwnLayer(HLERequestContext& ctx);
+    void AcquireCallerAppletCaptureSharedBuffer(HLERequestContext& ctx);
+    void ReleaseCallerAppletCaptureSharedBuffer(HLERequestContext& ctx);
 };
 
 class IDebugFunctions final : public ServiceFramework<IDebugFunctions> {
@@ -150,9 +153,13 @@ private:
     void SetRestartMessageEnabled(HLERequestContext& ctx);
     void SetOutOfFocusSuspendingEnabled(HLERequestContext& ctx);
     void SetAlbumImageOrientation(HLERequestContext& ctx);
+    void IsSystemBufferSharingEnabled(HLERequestContext& ctx);
+    void GetSystemSharedBufferHandle(HLERequestContext& ctx);
+    void GetSystemSharedLayerHandle(HLERequestContext& ctx);
     void CreateManagedDisplayLayer(HLERequestContext& ctx);
     void CreateManagedDisplaySeparableLayer(HLERequestContext& ctx);
     void SetHandlesRequestToDisplay(HLERequestContext& ctx);
+    void ApproveToDisplay(HLERequestContext& ctx);
     void SetIdleTimeDetectionExtension(HLERequestContext& ctx);
     void GetIdleTimeDetectionExtension(HLERequestContext& ctx);
     void ReportUserIsActive(HLERequestContext& ctx);
@@ -164,6 +171,8 @@ private:
     void SaveCurrentScreenshot(HLERequestContext& ctx);
     void SetRecordVolumeMuted(HLERequestContext& ctx);
 
+    Result EnsureBufferSharingEnabled();
+
     enum class ScreenshotPermission : u32 {
         Inherit = 0,
         Enable = 1,
@@ -179,7 +188,10 @@ private:
 
     u32 idle_time_detection_extension = 0;
     u64 num_fatal_sections_entered = 0;
+    u64 system_shared_buffer_id = 0;
+    u64 system_shared_layer_id = 0;
     bool is_auto_sleep_disabled = false;
+    bool buffer_sharing_enabled = false;
     ScreenshotPermission screenshot_permission = ScreenshotPermission::Inherit;
 };
 
@@ -223,6 +235,8 @@ private:
     void GetEventHandle(HLERequestContext& ctx);
     void ReceiveMessage(HLERequestContext& ctx);
     void GetCurrentFocusState(HLERequestContext& ctx);
+    void RequestToAcquireSleepLock(HLERequestContext& ctx);
+    void GetAcquiredSleepLockEvent(HLERequestContext& ctx);
     void GetDefaultDisplayResolutionChangeEvent(HLERequestContext& ctx);
     void GetOperationMode(HLERequestContext& ctx);
     void GetPerformanceMode(HLERequestContext& ctx);
@@ -240,6 +254,8 @@ private:
 
     std::shared_ptr<AppletMessageQueue> msg_queue;
     bool vr_mode_state{};
+    Kernel::KEvent* sleep_lock_event;
+    KernelHelpers::ServiceContext service_context;
 };
 
 class IStorageImpl {
@@ -311,6 +327,7 @@ private:
     void ExitProcessAndReturn(HLERequestContext& ctx);
     void GetCallerAppletIdentityInfo(HLERequestContext& ctx);
 
+    void PushInShowCabinetData();
     void PushInShowMiiEditData();
 
     std::deque<std::vector<u8>> queue_data;

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

@@ -199,6 +199,14 @@ const AppletFrontendSet& AppletManager::GetAppletFrontendSet() const {
     return frontend;
 }
 
+NFP::CabinetMode AppletManager::GetCabinetMode() const {
+    return cabinet_mode;
+}
+
+AppletId AppletManager::GetCurrentAppletId() const {
+    return current_applet_id;
+}
+
 void AppletManager::SetAppletFrontendSet(AppletFrontendSet set) {
     if (set.cabinet != nullptr) {
         frontend.cabinet = std::move(set.cabinet);
@@ -237,6 +245,14 @@ void AppletManager::SetAppletFrontendSet(AppletFrontendSet set) {
     }
 }
 
+void AppletManager::SetCabinetMode(NFP::CabinetMode mode) {
+    cabinet_mode = mode;
+}
+
+void AppletManager::SetCurrentAppletId(AppletId applet_id) {
+    current_applet_id = applet_id;
+}
+
 void AppletManager::SetDefaultAppletFrontendSet() {
     ClearAll();
     SetDefaultAppletsIfMissing();

src/core/hle/service/am/applets/applets.h

@@ -34,6 +34,10 @@ class KEvent;
 class KReadableEvent;
 } // namespace Kernel
 
+namespace Service::NFP {
+enum class CabinetMode : u8;
+} // namespace Service::NFP
+
 namespace Service::AM {
 
 class IStorage;
@@ -41,6 +45,8 @@ class IStorage;
 namespace Applets {
 
 enum class AppletId : u32 {
+    None = 0x00,
+    Application = 0x01,
     OverlayDisplay = 0x02,
     QLaunch = 0x03,
     Starter = 0x04,
@@ -160,7 +166,6 @@ public:
         return initialized;
     }
 
-protected:
     struct CommonArguments {
         u32_le arguments_version;
         u32_le size;
@@ -171,6 +176,7 @@ protected:
     };
     static_assert(sizeof(CommonArguments) == 0x20, "CommonArguments has incorrect size.");
 
+protected:
     CommonArguments common_args{};
     AppletDataBroker broker;
     LibraryAppletMode applet_mode;
@@ -219,8 +225,12 @@ public:
     ~AppletManager();
 
     const AppletFrontendSet& GetAppletFrontendSet() const;
+    NFP::CabinetMode GetCabinetMode() const;
+    AppletId GetCurrentAppletId() const;
 
     void SetAppletFrontendSet(AppletFrontendSet set);
+    void SetCabinetMode(NFP::CabinetMode mode);
+    void SetCurrentAppletId(AppletId applet_id);
     void SetDefaultAppletFrontendSet();
     void SetDefaultAppletsIfMissing();
     void ClearAll();
@@ -228,6 +238,9 @@ public:
     std::shared_ptr<Applet> GetApplet(AppletId id, LibraryAppletMode mode) const;
 
 private:
+    AppletId current_applet_id{};
+    NFP::CabinetMode cabinet_mode{};
+
     AppletFrontendSet frontend;
     Core::System& system;
 };

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

@@ -23,19 +23,39 @@ public:
     explicit IMonitorService(Core::System& system_) : ServiceFramework{system_, "IMonitorService"} {
         // clang-format off
         static const FunctionInfo functions[] = {
-            {0, nullptr, "GetStateForMonitor"},
+            {0, &IMonitorService::GetStateForMonitor, "GetStateForMonitor"},
             {1, nullptr, "GetNetworkInfoForMonitor"},
             {2, nullptr, "GetIpv4AddressForMonitor"},
             {3, nullptr, "GetDisconnectReasonForMonitor"},
             {4, nullptr, "GetSecurityParameterForMonitor"},
             {5, nullptr, "GetNetworkConfigForMonitor"},
-            {100, nullptr, "InitializeMonitor"},
+            {100, &IMonitorService::InitializeMonitor, "InitializeMonitor"},
             {101, nullptr, "FinalizeMonitor"},
         };
         // clang-format on
 
         RegisterHandlers(functions);
     }
+
+private:
+    void GetStateForMonitor(HLERequestContext& ctx) {
+        LOG_INFO(Service_LDN, "called");
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(ResultSuccess);
+        rb.PushEnum(state);
+    }
+
+    void InitializeMonitor(HLERequestContext& ctx) {
+        LOG_INFO(Service_LDN, "called");
+
+        state = State::Initialized;
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+
+    State state{State::None};
 };
 
 class LDNM final : public ServiceFramework<LDNM> {
@@ -731,14 +751,81 @@ public:
     }
 };
 
+class ISfMonitorService final : public ServiceFramework<ISfMonitorService> {
+public:
+    explicit ISfMonitorService(Core::System& system_)
+        : ServiceFramework{system_, "ISfMonitorService"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, &ISfMonitorService::Initialize, "Initialize"},
+            {288, &ISfMonitorService::GetGroupInfo, "GetGroupInfo"},
+            {320, nullptr, "GetLinkLevel"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+private:
+    void Initialize(HLERequestContext& ctx) {
+        LOG_WARNING(Service_LDN, "(STUBBED) called");
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(ResultSuccess);
+        rb.Push(0);
+    }
+
+    void GetGroupInfo(HLERequestContext& ctx) {
+        LOG_WARNING(Service_LDN, "(STUBBED) called");
+
+        struct GroupInfo {
+            std::array<u8, 0x200> info;
+        };
+
+        GroupInfo group_info{};
+
+        ctx.WriteBuffer(group_info);
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+};
+
+class LP2PM final : public ServiceFramework<LP2PM> {
+public:
+    explicit LP2PM(Core::System& system_) : ServiceFramework{system_, "lp2p:m"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, &LP2PM::CreateMonitorService, "CreateMonitorService"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+private:
+    void CreateMonitorService(HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const u64 reserved_input = rp.Pop<u64>();
+
+        LOG_INFO(Service_LDN, "called, reserved_input={}", reserved_input);
+
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+        rb.Push(ResultSuccess);
+        rb.PushIpcInterface<ISfMonitorService>(system);
+    }
+};
+
 void LoopProcess(Core::System& system) {
     auto server_manager = std::make_unique<ServerManager>(system);
 
     server_manager->RegisterNamedService("ldn:m", std::make_shared<LDNM>(system));
     server_manager->RegisterNamedService("ldn:s", std::make_shared<LDNS>(system));
     server_manager->RegisterNamedService("ldn:u", std::make_shared<LDNU>(system));
+
     server_manager->RegisterNamedService("lp2p:app", std::make_shared<LP2PAPP>(system));
     server_manager->RegisterNamedService("lp2p:sys", std::make_shared<LP2PSYS>(system));
+    server_manager->RegisterNamedService("lp2p:m", std::make_shared<LP2PM>(system));
+
     ServerManager::RunServer(std::move(server_manager));
 }
 

src/core/hle/service/nfc/common/device.cpp

@@ -830,11 +830,6 @@ Result NfcDevice::SetRegisterInfoPrivate(const NFP::RegisterInfoPrivate& registe
         return ResultWrongDeviceState;
     }
 
-    Service::Mii::StoreData store_data{};
-    Service::Mii::NfpStoreDataExtension extension{};
-    store_data.BuildBase(Mii::Gender::Male);
-    extension.SetFromStoreData(store_data);
-
     auto& settings = tag_data.settings;
 
     if (tag_data.settings.settings.amiibo_initialized == 0) {
@@ -843,8 +838,8 @@ Result NfcDevice::SetRegisterInfoPrivate(const NFP::RegisterInfoPrivate& registe
     }
 
     SetAmiiboName(settings, register_info.amiibo_name);
-    tag_data.owner_mii.BuildFromStoreData(store_data);
+    tag_data.owner_mii.BuildFromStoreData(register_info.mii_store_data);
-    tag_data.mii_extension = extension;
+    tag_data.mii_extension.SetFromStoreData(register_info.mii_store_data);
     tag_data.unknown = 0;
     tag_data.unknown2 = {};
     settings.country_code_id = 0;

src/core/hle/service/nvdrv/devices/nvmap.h

@@ -45,13 +45,6 @@ public:
         IsSharedMemMapped = 6
     };
 
-private:
-    /// Id to use for the next handle that is created.
-    u32 next_handle = 0;
-
-    /// Id to use for the next object that is created.
-    u32 next_id = 0;
-
     struct IocCreateParams {
         // Input
         u32_le size{};
@@ -113,6 +106,13 @@ private:
     NvResult IocParam(std::span<const u8> input, std::span<u8> output);
     NvResult IocFree(std::span<const u8> input, std::span<u8> output);
 
+private:
+    /// Id to use for the next handle that is created.
+    u32 next_handle = 0;
+
+    /// Id to use for the next object that is created.
+    u32 next_id = 0;
+
     NvCore::Container& container;
     NvCore::NvMap& file;
 };

src/core/hle/service/nvnflinger/buffer_item.h

@@ -15,7 +15,7 @@
 
 namespace Service::android {
 
-class GraphicBuffer;
+struct GraphicBuffer;
 
 class BufferItem final {
 public:

src/core/hle/service/nvnflinger/buffer_slot.h

@@ -13,7 +13,7 @@
 
 namespace Service::android {
 
-class GraphicBuffer;
+struct GraphicBuffer;
 
 enum class BufferState : u32 {
     Free = 0,

src/core/hle/service/nvnflinger/fb_share_buffer_manager.cpp

@@ -0,0 +1,351 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <random>
+
+#include "core/core.h"
+#include "core/hle/kernel/k_process.h"
+#include "core/hle/kernel/k_system_resource.h"
+#include "core/hle/service/nvdrv/devices/nvmap.h"
+#include "core/hle/service/nvdrv/nvdrv.h"
+#include "core/hle/service/nvnflinger/buffer_queue_producer.h"
+#include "core/hle/service/nvnflinger/fb_share_buffer_manager.h"
+#include "core/hle/service/nvnflinger/pixel_format.h"
+#include "core/hle/service/nvnflinger/ui/graphic_buffer.h"
+#include "core/hle/service/vi/layer/vi_layer.h"
+#include "core/hle/service/vi/vi_results.h"
+
+namespace Service::Nvnflinger {
+
+namespace {
+
+Result AllocateIoForProcessAddressSpace(Common::ProcessAddress* out_map_address,
+                                        std::unique_ptr<Kernel::KPageGroup>* out_page_group,
+                                        Core::System& system, u32 size) {
+    using Core::Memory::YUZU_PAGESIZE;
+
+    // Allocate memory for the system shared buffer.
+    // FIXME: Because the gmmu can only point to cpu addresses, we need
+    //        to map this in the application space to allow it to be used.
+    // FIXME: Add proper smmu emulation.
+    // FIXME: This memory belongs to vi's .data section.
+    auto& kernel = system.Kernel();
+    auto* process = system.ApplicationProcess();
+    auto& page_table = process->GetPageTable();
+
+    // Hold a temporary page group reference while we try to map it.
+    auto pg = std::make_unique<Kernel::KPageGroup>(
+        kernel, std::addressof(kernel.GetSystemSystemResource().GetBlockInfoManager()));
+
+    // Allocate memory from secure pool.
+    R_TRY(kernel.MemoryManager().AllocateAndOpen(
+        pg.get(), size / YUZU_PAGESIZE,
+        Kernel::KMemoryManager::EncodeOption(Kernel::KMemoryManager::Pool::Secure,
+                                             Kernel::KMemoryManager::Direction::FromBack)));
+
+    // Get bounds of where mapping is possible.
+    const VAddr alias_code_begin = GetInteger(page_table.GetAliasCodeRegionStart());
+    const VAddr alias_code_size = page_table.GetAliasCodeRegionSize() / YUZU_PAGESIZE;
+    const auto state = Kernel::KMemoryState::Io;
+    const auto perm = Kernel::KMemoryPermission::UserReadWrite;
+    std::mt19937_64 rng{process->GetRandomEntropy(0)};
+
+    // Retry up to 64 times to map into alias code range.
+    Result res = ResultSuccess;
+    int i;
+    for (i = 0; i < 64; i++) {
+        *out_map_address = alias_code_begin + ((rng() % alias_code_size) * YUZU_PAGESIZE);
+        res = page_table.MapPageGroup(*out_map_address, *pg, state, perm);
+        if (R_SUCCEEDED(res)) {
+            break;
+        }
+    }
+
+    // Return failure, if necessary
+    R_UNLESS(i < 64, res);
+
+    // Return the mapped page group.
+    *out_page_group = std::move(pg);
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+template <typename T>
+std::span<u8> SerializeIoc(T& params) {
+    return std::span(reinterpret_cast<u8*>(std::addressof(params)), sizeof(T));
+}
+
+Result CreateNvMapHandle(u32* out_nv_map_handle, Nvidia::Devices::nvmap& nvmap, u32 size) {
+    // Create a handle.
+    Nvidia::Devices::nvmap::IocCreateParams create_in_params{
+        .size = size,
+        .handle = 0,
+    };
+    Nvidia::Devices::nvmap::IocCreateParams create_out_params{};
+    R_UNLESS(nvmap.IocCreate(SerializeIoc(create_in_params), SerializeIoc(create_out_params)) ==
+                 Nvidia::NvResult::Success,
+             VI::ResultOperationFailed);
+
+    // Assign the output handle.
+    *out_nv_map_handle = create_out_params.handle;
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+Result FreeNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle) {
+    // Free the handle.
+    Nvidia::Devices::nvmap::IocFreeParams free_in_params{
+        .handle = handle,
+    };
+    Nvidia::Devices::nvmap::IocFreeParams free_out_params{};
+    R_UNLESS(nvmap.IocFree(SerializeIoc(free_in_params), SerializeIoc(free_out_params)) ==
+                 Nvidia::NvResult::Success,
+             VI::ResultOperationFailed);
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+Result AllocNvMapHandle(Nvidia::Devices::nvmap& nvmap, u32 handle, Common::ProcessAddress buffer,
+                        u32 size) {
+    // Assign the allocated memory to the handle.
+    Nvidia::Devices::nvmap::IocAllocParams alloc_in_params{
+        .handle = handle,
+        .heap_mask = 0,
+        .flags = {},
+        .align = 0,
+        .kind = 0,
+        .address = GetInteger(buffer),
+    };
+    Nvidia::Devices::nvmap::IocAllocParams alloc_out_params{};
+    R_UNLESS(nvmap.IocAlloc(SerializeIoc(alloc_in_params), SerializeIoc(alloc_out_params)) ==
+                 Nvidia::NvResult::Success,
+             VI::ResultOperationFailed);
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+Result AllocateHandleForBuffer(u32* out_handle, Nvidia::Module& nvdrv,
+                               Common::ProcessAddress buffer, u32 size) {
+    // Get the nvmap device.
+    auto nvmap_fd = nvdrv.Open("/dev/nvmap");
+    auto nvmap = nvdrv.GetDevice<Nvidia::Devices::nvmap>(nvmap_fd);
+    ASSERT(nvmap != nullptr);
+
+    // Create a handle.
+    R_TRY(CreateNvMapHandle(out_handle, *nvmap, size));
+
+    // Ensure we maintain a clean state on failure.
+    ON_RESULT_FAILURE {
+        ASSERT(R_SUCCEEDED(FreeNvMapHandle(*nvmap, *out_handle)));
+    };
+
+    // Assign the allocated memory to the handle.
+    R_RETURN(AllocNvMapHandle(*nvmap, *out_handle, buffer, size));
+}
+
+constexpr auto SharedBufferBlockLinearFormat = android::PixelFormat::Rgba8888;
+constexpr u32 SharedBufferBlockLinearBpp = 4;
+
+constexpr u32 SharedBufferBlockLinearWidth = 1280;
+constexpr u32 SharedBufferBlockLinearHeight = 768;
+constexpr u32 SharedBufferBlockLinearStride =
+    SharedBufferBlockLinearWidth * SharedBufferBlockLinearBpp;
+constexpr u32 SharedBufferNumSlots = 7;
+
+constexpr u32 SharedBufferWidth = 1280;
+constexpr u32 SharedBufferHeight = 720;
+constexpr u32 SharedBufferAsync = false;
+
+constexpr u32 SharedBufferSlotSize =
+    SharedBufferBlockLinearWidth * SharedBufferBlockLinearHeight * SharedBufferBlockLinearBpp;
+constexpr u32 SharedBufferSize = SharedBufferSlotSize * SharedBufferNumSlots;
+
+constexpr SharedMemoryPoolLayout SharedBufferPoolLayout = [] {
+    SharedMemoryPoolLayout layout{};
+    layout.num_slots = SharedBufferNumSlots;
+
+    for (u32 i = 0; i < SharedBufferNumSlots; i++) {
+        layout.slots[i].buffer_offset = i * SharedBufferSlotSize;
+        layout.slots[i].size = SharedBufferSlotSize;
+        layout.slots[i].width = SharedBufferWidth;
+        layout.slots[i].height = SharedBufferHeight;
+    }
+
+    return layout;
+}();
+
+void MakeGraphicBuffer(android::BufferQueueProducer& producer, u32 slot, u32 handle) {
+    auto buffer = std::make_shared<android::GraphicBuffer>();
+    buffer->width = SharedBufferWidth;
+    buffer->height = SharedBufferHeight;
+    buffer->stride = SharedBufferBlockLinearStride;
+    buffer->format = SharedBufferBlockLinearFormat;
+    buffer->buffer_id = handle;
+    buffer->offset = slot * SharedBufferSlotSize;
+    ASSERT(producer.SetPreallocatedBuffer(slot, buffer) == android::Status::NoError);
+}
+
+} // namespace
+
+FbShareBufferManager::FbShareBufferManager(Core::System& system, Nvnflinger& flinger,
+                                           std::shared_ptr<Nvidia::Module> nvdrv)
+    : m_system(system), m_flinger(flinger), m_nvdrv(std::move(nvdrv)) {}
+
+FbShareBufferManager::~FbShareBufferManager() = default;
+
+Result FbShareBufferManager::Initialize(u64* out_buffer_id, u64* out_layer_id, u64 display_id) {
+    std::scoped_lock lk{m_guard};
+
+    // Ensure we have not already created a buffer.
+    R_UNLESS(m_buffer_id == 0, VI::ResultOperationFailed);
+
+    // Allocate memory and space for the shared buffer.
+    Common::ProcessAddress map_address;
+    R_TRY(AllocateIoForProcessAddressSpace(std::addressof(map_address),
+                                           std::addressof(m_buffer_page_group), m_system,
+                                           SharedBufferSize));
+
+    // Create an nvmap handle for the buffer and assign the memory to it.
+    R_TRY(AllocateHandleForBuffer(std::addressof(m_buffer_nvmap_handle), *m_nvdrv, map_address,
+                                  SharedBufferSize));
+
+    // Record the display id.
+    m_display_id = display_id;
+
+    // Create a layer for the display.
+    m_layer_id = m_flinger.CreateLayer(m_display_id).value();
+
+    // Set up the buffer.
+    m_buffer_id = m_next_buffer_id++;
+
+    // Get the layer.
+    VI::Layer* layer = m_flinger.FindLayer(m_display_id, m_layer_id);
+    ASSERT(layer != nullptr);
+
+    // Get the producer and set preallocated buffers.
+    auto& producer = layer->GetBufferQueue();
+    MakeGraphicBuffer(producer, 0, m_buffer_nvmap_handle);
+    MakeGraphicBuffer(producer, 1, m_buffer_nvmap_handle);
+
+    // Assign outputs.
+    *out_buffer_id = m_buffer_id;
+    *out_layer_id = m_layer_id;
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+Result FbShareBufferManager::GetSharedBufferMemoryHandleId(u64* out_buffer_size,
+                                                           s32* out_nvmap_handle,
+                                                           SharedMemoryPoolLayout* out_pool_layout,
+                                                           u64 buffer_id,
+                                                           u64 applet_resource_user_id) {
+    std::scoped_lock lk{m_guard};
+
+    R_UNLESS(m_buffer_id > 0, VI::ResultNotFound);
+    R_UNLESS(buffer_id == m_buffer_id, VI::ResultNotFound);
+
+    *out_pool_layout = SharedBufferPoolLayout;
+    *out_buffer_size = SharedBufferSize;
+    *out_nvmap_handle = m_buffer_nvmap_handle;
+
+    R_SUCCEED();
+}
+
+Result FbShareBufferManager::GetLayerFromId(VI::Layer** out_layer, u64 layer_id) {
+    // Ensure the layer id is valid.
+    R_UNLESS(m_layer_id > 0 && layer_id == m_layer_id, VI::ResultNotFound);
+
+    // Get the layer.
+    VI::Layer* layer = m_flinger.FindLayer(m_display_id, layer_id);
+    R_UNLESS(layer != nullptr, VI::ResultNotFound);
+
+    // We succeeded.
+    *out_layer = layer;
+    R_SUCCEED();
+}
+
+Result FbShareBufferManager::AcquireSharedFrameBuffer(android::Fence* out_fence,
+                                                      std::array<s32, 4>& out_slot_indexes,
+                                                      s64* out_target_slot, u64 layer_id) {
+    std::scoped_lock lk{m_guard};
+
+    // Get the layer.
+    VI::Layer* layer;
+    R_TRY(this->GetLayerFromId(std::addressof(layer), layer_id));
+
+    // Get the producer.
+    auto& producer = layer->GetBufferQueue();
+
+    // Get the next buffer from the producer.
+    s32 slot;
+    R_UNLESS(producer.DequeueBuffer(std::addressof(slot), out_fence, SharedBufferAsync != 0,
+                                    SharedBufferWidth, SharedBufferHeight,
+                                    SharedBufferBlockLinearFormat, 0) == android::Status::NoError,
+             VI::ResultOperationFailed);
+
+    // Assign remaining outputs.
+    *out_target_slot = slot;
+    out_slot_indexes = {0, 1, -1, -1};
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+Result FbShareBufferManager::PresentSharedFrameBuffer(android::Fence fence,
+                                                      Common::Rectangle<s32> crop_region,
+                                                      u32 transform, s32 swap_interval,
+                                                      u64 layer_id, s64 slot) {
+    std::scoped_lock lk{m_guard};
+
+    // Get the layer.
+    VI::Layer* layer;
+    R_TRY(this->GetLayerFromId(std::addressof(layer), layer_id));
+
+    // Get the producer.
+    auto& producer = layer->GetBufferQueue();
+
+    // Request to queue the buffer.
+    std::shared_ptr<android::GraphicBuffer> buffer;
+    R_UNLESS(producer.RequestBuffer(static_cast<s32>(slot), std::addressof(buffer)) ==
+                 android::Status::NoError,
+             VI::ResultOperationFailed);
+
+    // Queue the buffer to the producer.
+    android::QueueBufferInput input{};
+    android::QueueBufferOutput output{};
+    input.crop = crop_region;
+    input.fence = fence;
+    input.transform = static_cast<android::NativeWindowTransform>(transform);
+    input.swap_interval = swap_interval;
+    R_UNLESS(producer.QueueBuffer(static_cast<s32>(slot), input, std::addressof(output)) ==
+                 android::Status::NoError,
+             VI::ResultOperationFailed);
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+Result FbShareBufferManager::GetSharedFrameBufferAcquirableEvent(Kernel::KReadableEvent** out_event,
+                                                                 u64 layer_id) {
+    std::scoped_lock lk{m_guard};
+
+    // Get the layer.
+    VI::Layer* layer;
+    R_TRY(this->GetLayerFromId(std::addressof(layer), layer_id));
+
+    // Get the producer.
+    auto& producer = layer->GetBufferQueue();
+
+    // Set the event.
+    *out_event = std::addressof(producer.GetNativeHandle());
+
+    // We succeeded.
+    R_SUCCEED();
+}
+
+} // namespace Service::Nvnflinger

src/core/hle/service/nvnflinger/fb_share_buffer_manager.h

@@ -0,0 +1,65 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/math_util.h"
+#include "core/hle/service/nvnflinger/nvnflinger.h"
+#include "core/hle/service/nvnflinger/ui/fence.h"
+
+namespace Kernel {
+class KPageGroup;
+}
+
+namespace Service::Nvnflinger {
+
+struct SharedMemorySlot {
+    u64 buffer_offset;
+    u64 size;
+    s32 width;
+    s32 height;
+};
+static_assert(sizeof(SharedMemorySlot) == 0x18, "SharedMemorySlot has wrong size");
+
+struct SharedMemoryPoolLayout {
+    s32 num_slots;
+    SharedMemorySlot slots[0x10];
+};
+static_assert(sizeof(SharedMemoryPoolLayout) == 0x188, "SharedMemoryPoolLayout has wrong size");
+
+class FbShareBufferManager final {
+public:
+    explicit FbShareBufferManager(Core::System& system, Nvnflinger& flinger,
+                                  std::shared_ptr<Nvidia::Module> nvdrv);
+    ~FbShareBufferManager();
+
+    Result Initialize(u64* out_buffer_id, u64* out_layer_handle, u64 display_id);
+    Result GetSharedBufferMemoryHandleId(u64* out_buffer_size, s32* out_nvmap_handle,
+                                         SharedMemoryPoolLayout* out_pool_layout, u64 buffer_id,
+                                         u64 applet_resource_user_id);
+    Result AcquireSharedFrameBuffer(android::Fence* out_fence, std::array<s32, 4>& out_slots,
+                                    s64* out_target_slot, u64 layer_id);
+    Result PresentSharedFrameBuffer(android::Fence fence, Common::Rectangle<s32> crop_region,
+                                    u32 transform, s32 swap_interval, u64 layer_id, s64 slot);
+    Result GetSharedFrameBufferAcquirableEvent(Kernel::KReadableEvent** out_event, u64 layer_id);
+
+private:
+    Result GetLayerFromId(VI::Layer** out_layer, u64 layer_id);
+
+private:
+    u64 m_next_buffer_id = 1;
+    u64 m_display_id = 0;
+    u64 m_buffer_id = 0;
+    u64 m_layer_id = 0;
+    u32 m_buffer_nvmap_handle = 0;
+    SharedMemoryPoolLayout m_pool_layout = {};
+
+    std::unique_ptr<Kernel::KPageGroup> m_buffer_page_group;
+
+    std::mutex m_guard;
+    Core::System& m_system;
+    Nvnflinger& m_flinger;
+    std::shared_ptr<Nvidia::Module> m_nvdrv;
+};
+
+} // namespace Service::Nvnflinger

src/core/hle/service/nvnflinger/graphic_buffer_producer.h

@@ -19,6 +19,7 @@ class InputParcel;
 #pragma pack(push, 1)
 struct QueueBufferInput final {
     explicit QueueBufferInput(InputParcel& parcel);
+    explicit QueueBufferInput() = default;
 
     void Deflate(s64* timestamp_, bool* is_auto_timestamp_, Common::Rectangle<s32>* crop_,
                  NativeWindowScalingMode* scaling_mode_, NativeWindowTransform* transform_,
@@ -34,7 +35,6 @@ struct QueueBufferInput final {
         *fence_ = fence;
     }
 
-private:
     s64 timestamp{};
     s32 is_auto_timestamp{};
     Common::Rectangle<s32> crop{};

src/core/hle/service/nvnflinger/nvnflinger.cpp

@@ -17,6 +17,7 @@
 #include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/hle/service/nvnflinger/buffer_item_consumer.h"
 #include "core/hle/service/nvnflinger/buffer_queue_core.h"
+#include "core/hle/service/nvnflinger/fb_share_buffer_manager.h"
 #include "core/hle/service/nvnflinger/hos_binder_driver_server.h"
 #include "core/hle/service/nvnflinger/nvnflinger.h"
 #include "core/hle/service/nvnflinger/ui/graphic_buffer.h"
@@ -331,4 +332,14 @@ s64 Nvnflinger::GetNextTicks() const {
     return static_cast<s64>(speed_scale * (1000000000.f / effective_fps));
 }
 
+FbShareBufferManager& Nvnflinger::GetSystemBufferManager() {
+    const auto lock_guard = Lock();
+
+    if (!system_buffer_manager) {
+        system_buffer_manager = std::make_unique<FbShareBufferManager>(system, *this, nvdrv);
+    }
+
+    return *system_buffer_manager;
+}
+
 } // namespace Service::Nvnflinger

src/core/hle/service/nvnflinger/nvnflinger.h

@@ -45,6 +45,9 @@ class BufferQueueProducer;
 
 namespace Service::Nvnflinger {
 
+class FbShareBufferManager;
+class HosBinderDriverServer;
+
 class Nvnflinger final {
 public:
     explicit Nvnflinger(Core::System& system_, HosBinderDriverServer& hos_binder_driver_server_);
@@ -90,12 +93,16 @@ public:
 
     [[nodiscard]] s64 GetNextTicks() const;
 
+    FbShareBufferManager& GetSystemBufferManager();
+
 private:
     struct Layer {
         std::unique_ptr<android::BufferQueueCore> core;
         std::unique_ptr<android::BufferQueueProducer> producer;
     };
 
+    friend class FbShareBufferManager;
+
 private:
     [[nodiscard]] std::unique_lock<std::mutex> Lock() const {
         return std::unique_lock{*guard};
@@ -140,6 +147,8 @@ private:
     std::shared_ptr<Core::Timing::EventType> multi_composition_event;
     std::shared_ptr<Core::Timing::EventType> single_composition_event;
 
+    std::unique_ptr<FbShareBufferManager> system_buffer_manager;
+
     std::shared_ptr<std::mutex> guard;
 
     Core::System& system;

src/core/hle/service/nvnflinger/ui/fence.h

@@ -20,6 +20,9 @@ public:
     static constexpr Fence NoFence() {
         Fence fence;
         fence.fences[0].id = -1;
+        fence.fences[1].id = -1;
+        fence.fences[2].id = -1;
+        fence.fences[3].id = -1;
         return fence;
     }
 

src/core/hle/service/nvnflinger/ui/graphic_buffer.h

@@ -12,8 +12,7 @@
 
 namespace Service::android {
 
-class GraphicBuffer final {
+struct GraphicBuffer final {
-public:
     constexpr GraphicBuffer() = default;
 
     constexpr GraphicBuffer(u32 width_, u32 height_, PixelFormat format_, u32 usage_)
@@ -77,7 +76,6 @@ public:
         return false;
     }
 
-private:
     u32 magic{};
     s32 width{};
     s32 height{};

src/core/hle/service/vi/vi.cpp

@@ -20,9 +20,12 @@
 #include "core/hle/kernel/k_readable_event.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/service/ipc_helpers.h"
+#include "core/hle/service/nvdrv/devices/nvmap.h"
 #include "core/hle/service/nvdrv/nvdata.h"
+#include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/hle/service/nvnflinger/binder.h"
 #include "core/hle/service/nvnflinger/buffer_queue_producer.h"
+#include "core/hle/service/nvnflinger/fb_share_buffer_manager.h"
 #include "core/hle/service/nvnflinger/hos_binder_driver_server.h"
 #include "core/hle/service/nvnflinger/nvnflinger.h"
 #include "core/hle/service/nvnflinger/parcel.h"
@@ -131,8 +134,9 @@ private:
 
 class ISystemDisplayService final : public ServiceFramework<ISystemDisplayService> {
 public:
-    explicit ISystemDisplayService(Core::System& system_)
+    explicit ISystemDisplayService(Core::System& system_, Nvnflinger::Nvnflinger& nvnflinger_)
-        : ServiceFramework{system_, "ISystemDisplayService"} {
+        : ServiceFramework{system_, "ISystemDisplayService"}, nvnflinger{nvnflinger_} {
+        // clang-format off
         static const FunctionInfo functions[] = {
             {1200, nullptr, "GetZOrderCountMin"},
             {1202, nullptr, "GetZOrderCountMax"},
@@ -170,22 +174,126 @@ public:
             {3217, nullptr, "SetDisplayCmuLuma"},
             {3218, nullptr, "SetDisplayCrcMode"},
             {6013, nullptr, "GetLayerPresentationSubmissionTimestamps"},
-            {8225, nullptr, "GetSharedBufferMemoryHandleId"},
+            {8225, &ISystemDisplayService::GetSharedBufferMemoryHandleId, "GetSharedBufferMemoryHandleId"},
-            {8250, nullptr, "OpenSharedLayer"},
+            {8250, &ISystemDisplayService::OpenSharedLayer, "OpenSharedLayer"},
             {8251, nullptr, "CloseSharedLayer"},
-            {8252, nullptr, "ConnectSharedLayer"},
+            {8252, &ISystemDisplayService::ConnectSharedLayer, "ConnectSharedLayer"},
             {8253, nullptr, "DisconnectSharedLayer"},
-            {8254, nullptr, "AcquireSharedFrameBuffer"},
+            {8254, &ISystemDisplayService::AcquireSharedFrameBuffer, "AcquireSharedFrameBuffer"},
-            {8255, nullptr, "PresentSharedFrameBuffer"},
+            {8255, &ISystemDisplayService::PresentSharedFrameBuffer, "PresentSharedFrameBuffer"},
-            {8256, nullptr, "GetSharedFrameBufferAcquirableEvent"},
+            {8256, &ISystemDisplayService::GetSharedFrameBufferAcquirableEvent, "GetSharedFrameBufferAcquirableEvent"},
             {8257, nullptr, "FillSharedFrameBufferColor"},
             {8258, nullptr, "CancelSharedFrameBuffer"},
             {9000, nullptr, "GetDp2hdmiController"},
         };
+        // clang-format on
         RegisterHandlers(functions);
     }
 
 private:
+    void GetSharedBufferMemoryHandleId(HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const u64 buffer_id = rp.PopRaw<u64>();
+
+        LOG_INFO(Service_VI, "called. buffer_id={:#x}", buffer_id);
+
+        struct OutputParameters {
+            s32 nvmap_handle;
+            u64 size;
+        };
+
+        OutputParameters out{};
+        Nvnflinger::SharedMemoryPoolLayout layout{};
+        const auto result = nvnflinger.GetSystemBufferManager().GetSharedBufferMemoryHandleId(
+            &out.size, &out.nvmap_handle, &layout, buffer_id, 0);
+
+        ctx.WriteBuffer(&layout, sizeof(layout));
+
+        IPC::ResponseBuilder rb{ctx, 6};
+        rb.Push(result);
+        rb.PushRaw(out);
+    }
+
+    void OpenSharedLayer(HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const u64 layer_id = rp.PopRaw<u64>();
+
+        LOG_INFO(Service_VI, "(STUBBED) called. layer_id={:#x}", layer_id);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+
+    void ConnectSharedLayer(HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const u64 layer_id = rp.PopRaw<u64>();
+
+        LOG_INFO(Service_VI, "(STUBBED) called. layer_id={:#x}", layer_id);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+
+    void GetSharedFrameBufferAcquirableEvent(HLERequestContext& ctx) {
+        LOG_DEBUG(Service_VI, "called");
+
+        IPC::RequestParser rp{ctx};
+        const u64 layer_id = rp.PopRaw<u64>();
+
+        Kernel::KReadableEvent* event{};
+        const auto result = nvnflinger.GetSystemBufferManager().GetSharedFrameBufferAcquirableEvent(
+            &event, layer_id);
+
+        IPC::ResponseBuilder rb{ctx, 2, 1};
+        rb.Push(result);
+        rb.PushCopyObjects(event);
+    }
+
+    void AcquireSharedFrameBuffer(HLERequestContext& ctx) {
+        LOG_DEBUG(Service_VI, "called");
+
+        IPC::RequestParser rp{ctx};
+        const u64 layer_id = rp.PopRaw<u64>();
+
+        struct OutputParameters {
+            android::Fence fence;
+            std::array<s32, 4> slots;
+            s64 target_slot;
+        };
+        static_assert(sizeof(OutputParameters) == 0x40, "OutputParameters has wrong size");
+
+        OutputParameters out{};
+        const auto result = nvnflinger.GetSystemBufferManager().AcquireSharedFrameBuffer(
+            &out.fence, out.slots, &out.target_slot, layer_id);
+
+        IPC::ResponseBuilder rb{ctx, 18};
+        rb.Push(result);
+        rb.PushRaw(out);
+    }
+
+    void PresentSharedFrameBuffer(HLERequestContext& ctx) {
+        LOG_DEBUG(Service_VI, "called");
+
+        struct InputParameters {
+            android::Fence fence;
+            Common::Rectangle<s32> crop_region;
+            u32 window_transform;
+            s32 swap_interval;
+            u64 layer_id;
+            s64 surface_id;
+        };
+        static_assert(sizeof(InputParameters) == 0x50, "InputParameters has wrong size");
+
+        IPC::RequestParser rp{ctx};
+        auto input = rp.PopRaw<InputParameters>();
+
+        const auto result = nvnflinger.GetSystemBufferManager().PresentSharedFrameBuffer(
+            input.fence, input.crop_region, input.window_transform, input.swap_interval,
+            input.layer_id, input.surface_id);
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(result);
+    }
+
     void SetLayerZ(HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
         const u64 layer_id = rp.Pop<u64>();
@@ -228,6 +336,9 @@ private:
         rb.PushRaw<float>(60.0f); // This wouldn't seem to be correct for 30 fps games.
         rb.Push<u32>(0);
     }
+
+private:
+    Nvnflinger::Nvnflinger& nvnflinger;
 };
 
 class IManagerDisplayService final : public ServiceFramework<IManagerDisplayService> {
@@ -453,7 +564,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(ResultSuccess);
-        rb.PushIpcInterface<ISystemDisplayService>(system);
+        rb.PushIpcInterface<ISystemDisplayService>(system, nv_flinger);
     }
 
     void GetManagerDisplayService(HLERequestContext& ctx) {

src/video_core/CMakeLists.txt

@@ -15,6 +15,7 @@ add_library(video_core STATIC
     buffer_cache/buffer_cache.cpp
     buffer_cache/buffer_cache.h
     buffer_cache/memory_tracker_base.h
+    buffer_cache/usage_tracker.h
     buffer_cache/word_manager.h
     cache_types.h
     cdma_pusher.cpp

src/video_core/buffer_cache/buffer_cache.h

@@ -67,6 +67,7 @@ void BufferCache<P>::TickFrame() {
     if (!channel_state) {
         return;
     }
+    runtime.TickFrame(slot_buffers);
 
     // Calculate hits and shots and move hit bits to the right
     const u32 hits = std::reduce(channel_state->uniform_cache_hits.begin(),
@@ -230,7 +231,10 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
     for (const IntervalType& add_interval : tmp_intervals) {
         common_ranges.add(add_interval);
     }
-    runtime.CopyBuffer(dest_buffer, src_buffer, copies);
+    const auto& copy = copies[0];
+    src_buffer.MarkUsage(copy.src_offset, copy.size);
+    dest_buffer.MarkUsage(copy.dst_offset, copy.size);
+    runtime.CopyBuffer(dest_buffer, src_buffer, copies, true);
     if (has_new_downloads) {
         memory_tracker.MarkRegionAsGpuModified(*cpu_dest_address, amount);
     }
@@ -258,9 +262,10 @@ bool BufferCache<P>::DMAClear(GPUVAddr dst_address, u64 amount, u32 value) {
     common_ranges.subtract(subtract_interval);
 
     const BufferId buffer = FindBuffer(*cpu_dst_address, static_cast<u32>(size));
-    auto& dest_buffer = slot_buffers[buffer];
+    Buffer& dest_buffer = slot_buffers[buffer];
     const u32 offset = dest_buffer.Offset(*cpu_dst_address);
     runtime.ClearBuffer(dest_buffer, offset, size, value);
+    dest_buffer.MarkUsage(offset, size);
     return true;
 }
 
@@ -603,6 +608,7 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
             VAddr orig_cpu_addr = static_cast<VAddr>(second_copy.src_offset);
             const IntervalType base_interval{orig_cpu_addr, orig_cpu_addr + copy.size};
             async_downloads += std::make_pair(base_interval, 1);
+            buffer.MarkUsage(copy.src_offset, copy.size);
             runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
             normalized_copies.push_back(second_copy);
         }
@@ -621,8 +627,9 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
                     // Have in mind the staging buffer offset for the copy
                     copy.dst_offset += download_staging.offset;
                     const std::array copies{copy};
-                    runtime.CopyBuffer(download_staging.buffer, slot_buffers[buffer_id], copies,
+                    Buffer& buffer = slot_buffers[buffer_id];
-                                       false);
+                    buffer.MarkUsage(copy.src_offset, copy.size);
+                    runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
                 }
                 runtime.PostCopyBarrier();
                 runtime.Finish();
@@ -742,7 +749,7 @@ void BufferCache<P>::BindHostIndexBuffer() {
                 {BufferCopy{.src_offset = upload_staging.offset, .dst_offset = 0, .size = size}}};
             std::memcpy(upload_staging.mapped_span.data(),
                         draw_state.inline_index_draw_indexes.data(), size);
-            runtime.CopyBuffer(buffer, upload_staging.buffer, copies);
+            runtime.CopyBuffer(buffer, upload_staging.buffer, copies, true);
         } else {
             buffer.ImmediateUpload(0, draw_state.inline_index_draw_indexes);
         }
@@ -754,6 +761,7 @@ void BufferCache<P>::BindHostIndexBuffer() {
             offset + draw_state.index_buffer.first * draw_state.index_buffer.FormatSizeInBytes();
         runtime.BindIndexBuffer(buffer, new_offset, size);
     } else {
+        buffer.MarkUsage(offset, size);
         runtime.BindIndexBuffer(draw_state.topology, draw_state.index_buffer.format,
                                 draw_state.index_buffer.first, draw_state.index_buffer.count,
                                 buffer, offset, size);
@@ -790,6 +798,7 @@ void BufferCache<P>::BindHostVertexBuffers() {
 
             const u32 stride = maxwell3d->regs.vertex_streams[index].stride;
             const u32 offset = buffer.Offset(binding.cpu_addr);
+            buffer.MarkUsage(offset, binding.size);
 
             host_bindings.buffers.push_back(&buffer);
             host_bindings.offsets.push_back(offset);
@@ -895,6 +904,7 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
     if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
         channel_state->uniform_buffer_binding_sizes[stage][binding_index] = size;
     }
+    buffer.MarkUsage(offset, size);
     if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
         runtime.BindUniformBuffer(stage, binding_index, buffer, offset, size);
     } else {
@@ -913,6 +923,7 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
         const u32 offset = buffer.Offset(binding.cpu_addr);
+        buffer.MarkUsage(offset, size);
         const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
         if constexpr (NEEDS_BIND_STORAGE_INDEX) {
             runtime.BindStorageBuffer(stage, binding_index, buffer, offset, size, is_written);
@@ -933,6 +944,7 @@ void BufferCache<P>::BindHostGraphicsTextureBuffers(size_t stage) {
 
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const PixelFormat format = binding.format;
+        buffer.MarkUsage(offset, size);
         if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
             if (((channel_state->image_texture_buffers[stage] >> index) & 1) != 0) {
                 runtime.BindImageBuffer(buffer, offset, size, format);
@@ -963,9 +975,10 @@ void BufferCache<P>::BindHostTransformFeedbackBuffers() {
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
         const u32 offset = buffer.Offset(binding.cpu_addr);
+        buffer.MarkUsage(offset, size);
         host_bindings.buffers.push_back(&buffer);
         host_bindings.offsets.push_back(offset);
-        host_bindings.sizes.push_back(binding.size);
+        host_bindings.sizes.push_back(size);
     }
     if (host_bindings.buffers.size() > 0) {
         runtime.BindTransformFeedbackBuffers(host_bindings);
@@ -989,6 +1002,7 @@ void BufferCache<P>::BindHostComputeUniformBuffers() {
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
         const u32 offset = buffer.Offset(binding.cpu_addr);
+        buffer.MarkUsage(offset, size);
         if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
             runtime.BindComputeUniformBuffer(binding_index, buffer, offset, size);
             ++binding_index;
@@ -1009,6 +1023,7 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
         const u32 offset = buffer.Offset(binding.cpu_addr);
+        buffer.MarkUsage(offset, size);
         const bool is_written =
             ((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
         if constexpr (NEEDS_BIND_STORAGE_INDEX) {
@@ -1030,6 +1045,7 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
 
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const PixelFormat format = binding.format;
+        buffer.MarkUsage(offset, size);
         if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
             if (((channel_state->image_compute_texture_buffers >> index) & 1) != 0) {
                 runtime.BindImageBuffer(buffer, offset, size, format);
@@ -1148,10 +1164,11 @@ void BufferCache<P>::UpdateVertexBuffer(u32 index) {
     if (!gpu_memory->IsWithinGPUAddressRange(gpu_addr_end)) {
         size = static_cast<u32>(gpu_memory->MaxContinuousRange(gpu_addr_begin, size));
     }
+    const BufferId buffer_id = FindBuffer(*cpu_addr, size);
     channel_state->vertex_buffers[index] = Binding{
         .cpu_addr = *cpu_addr,
         .size = size,
-        .buffer_id = FindBuffer(*cpu_addr, size),
+        .buffer_id = buffer_id,
     };
 }
 
@@ -1403,7 +1420,8 @@ void BufferCache<P>::JoinOverlap(BufferId new_buffer_id, BufferId overlap_id,
         .dst_offset = dst_base_offset,
         .size = overlap.SizeBytes(),
     });
-    runtime.CopyBuffer(new_buffer, overlap, copies);
+    new_buffer.MarkUsage(copies[0].dst_offset, copies[0].size);
+    runtime.CopyBuffer(new_buffer, overlap, copies, true);
     DeleteBuffer(overlap_id, true);
 }
 
@@ -1416,7 +1434,9 @@ BufferId BufferCache<P>::CreateBuffer(VAddr cpu_addr, u32 wanted_size) {
     const u32 size = static_cast<u32>(overlap.end - overlap.begin);
     const BufferId new_buffer_id = slot_buffers.insert(runtime, rasterizer, overlap.begin, size);
     auto& new_buffer = slot_buffers[new_buffer_id];
-    runtime.ClearBuffer(new_buffer, 0, new_buffer.SizeBytes(), 0);
+    const size_t size_bytes = new_buffer.SizeBytes();
+    runtime.ClearBuffer(new_buffer, 0, size_bytes, 0);
+    new_buffer.MarkUsage(0, size_bytes);
     for (const BufferId overlap_id : overlap.ids) {
         JoinOverlap(new_buffer_id, overlap_id, !overlap.has_stream_leap);
     }
@@ -1469,11 +1489,6 @@ void BufferCache<P>::TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept {
 
 template <class P>
 bool BufferCache<P>::SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size) {
-    return SynchronizeBufferImpl(buffer, cpu_addr, size);
-}
-
-template <class P>
-bool BufferCache<P>::SynchronizeBufferImpl(Buffer& buffer, VAddr cpu_addr, u32 size) {
     boost::container::small_vector<BufferCopy, 4> copies;
     u64 total_size_bytes = 0;
     u64 largest_copy = 0;
@@ -1495,51 +1510,6 @@ bool BufferCache<P>::SynchronizeBufferImpl(Buffer& buffer, VAddr cpu_addr, u32 s
     return false;
 }
 
-template <class P>
-bool BufferCache<P>::SynchronizeBufferNoModified(Buffer& buffer, VAddr cpu_addr, u32 size) {
-    boost::container::small_vector<BufferCopy, 4> copies;
-    u64 total_size_bytes = 0;
-    u64 largest_copy = 0;
-    IntervalSet found_sets{};
-    auto make_copies = [&] {
-        for (auto& interval : found_sets) {
-            const std::size_t sub_size = interval.upper() - interval.lower();
-            const VAddr cpu_addr_ = interval.lower();
-            copies.push_back(BufferCopy{
-                .src_offset = total_size_bytes,
-                .dst_offset = cpu_addr_ - buffer.CpuAddr(),
-                .size = sub_size,
-            });
-            total_size_bytes += sub_size;
-            largest_copy = std::max<u64>(largest_copy, sub_size);
-        }
-        const std::span<BufferCopy> copies_span(copies.data(), copies.size());
-        UploadMemory(buffer, total_size_bytes, largest_copy, copies_span);
-    };
-    memory_tracker.ForEachUploadRange(cpu_addr, size, [&](u64 cpu_addr_out, u64 range_size) {
-        const VAddr base_adr = cpu_addr_out;
-        const VAddr end_adr = base_adr + range_size;
-        const IntervalType add_interval{base_adr, end_adr};
-        found_sets.add(add_interval);
-    });
-    if (found_sets.empty()) {
-        return true;
-    }
-    const IntervalType search_interval{cpu_addr, cpu_addr + size};
-    auto it = common_ranges.lower_bound(search_interval);
-    auto it_end = common_ranges.upper_bound(search_interval);
-    if (it == common_ranges.end()) {
-        make_copies();
-        return false;
-    }
-    while (it != it_end) {
-        found_sets.subtract(*it);
-        it++;
-    }
-    make_copies();
-    return false;
-}
-
 template <class P>
 void BufferCache<P>::UploadMemory(Buffer& buffer, u64 total_size_bytes, u64 largest_copy,
                                   std::span<BufferCopy> copies) {
@@ -1588,7 +1558,8 @@ void BufferCache<P>::MappedUploadMemory([[maybe_unused]] Buffer& buffer,
             // Apply the staging offset
             copy.src_offset += upload_staging.offset;
         }
-        runtime.CopyBuffer(buffer, upload_staging.buffer, copies);
+        const bool can_reorder = runtime.CanReorderUpload(buffer, copies);
+        runtime.CopyBuffer(buffer, upload_staging.buffer, copies, true, can_reorder);
     }
 }
 
@@ -1630,7 +1601,8 @@ void BufferCache<P>::InlineMemoryImplementation(VAddr dest_address, size_t copy_
         }};
         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);
+        const bool can_reorder = runtime.CanReorderUpload(buffer, copies);
+        runtime.CopyBuffer(buffer, upload_staging.buffer, copies, true, can_reorder);
     } else {
         buffer.ImmediateUpload(buffer.Offset(dest_address), inlined_buffer.first(copy_size));
     }
@@ -1683,8 +1655,9 @@ void BufferCache<P>::DownloadBufferMemory(Buffer& buffer, VAddr cpu_addr, u64 si
         for (BufferCopy& copy : copies) {
             // Modify copies to have the staging offset in mind
             copy.dst_offset += download_staging.offset;
+            buffer.MarkUsage(copy.src_offset, copy.size);
         }
-        runtime.CopyBuffer(download_staging.buffer, buffer, copies_span);
+        runtime.CopyBuffer(download_staging.buffer, buffer, copies_span, true);
         runtime.Finish();
         for (const BufferCopy& copy : copies) {
             const VAddr copy_cpu_addr = buffer.CpuAddr() + copy.src_offset;

src/video_core/buffer_cache/buffer_cache_base.h

@@ -525,10 +525,6 @@ private:
 
     bool SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size);
 
-    bool SynchronizeBufferImpl(Buffer& buffer, VAddr cpu_addr, u32 size);
-
-    bool SynchronizeBufferNoModified(Buffer& buffer, VAddr cpu_addr, u32 size);
-
     void UploadMemory(Buffer& buffer, u64 total_size_bytes, u64 largest_copy,
                       std::span<BufferCopy> copies);
 

src/video_core/buffer_cache/usage_tracker.h

@@ -0,0 +1,79 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include "common/alignment.h"
+#include "common/common_types.h"
+
+namespace VideoCommon {
+
+class UsageTracker {
+    static constexpr size_t BYTES_PER_BIT_SHIFT = 6;
+    static constexpr size_t PAGE_SHIFT = 6 + BYTES_PER_BIT_SHIFT;
+    static constexpr size_t PAGE_BYTES = 1 << PAGE_SHIFT;
+
+public:
+    explicit UsageTracker(size_t size) {
+        const size_t num_pages = (size >> PAGE_SHIFT) + 1;
+        pages.resize(num_pages, 0ULL);
+    }
+
+    void Reset() noexcept {
+        std::ranges::fill(pages, 0ULL);
+    }
+
+    void Track(u64 offset, u64 size) noexcept {
+        const size_t page = offset >> PAGE_SHIFT;
+        const size_t page_end = (offset + size) >> PAGE_SHIFT;
+        TrackPage(page, offset, size);
+        if (page == page_end) {
+            return;
+        }
+        for (size_t i = page + 1; i < page_end; i++) {
+            pages[i] = ~u64{0};
+        }
+        const size_t offset_end = offset + size;
+        const size_t offset_end_page_aligned = Common::AlignDown(offset_end, PAGE_BYTES);
+        TrackPage(page_end, offset_end_page_aligned, offset_end - offset_end_page_aligned);
+    }
+
+    [[nodiscard]] bool IsUsed(u64 offset, u64 size) const noexcept {
+        const size_t page = offset >> PAGE_SHIFT;
+        const size_t page_end = (offset + size) >> PAGE_SHIFT;
+        if (IsPageUsed(page, offset, size)) {
+            return true;
+        }
+        for (size_t i = page + 1; i < page_end; i++) {
+            if (pages[i] != 0) {
+                return true;
+            }
+        }
+        const size_t offset_end = offset + size;
+        const size_t offset_end_page_aligned = Common::AlignDown(offset_end, PAGE_BYTES);
+        return IsPageUsed(page_end, offset_end_page_aligned, offset_end - offset_end_page_aligned);
+    }
+
+private:
+    void TrackPage(u64 page, u64 offset, u64 size) noexcept {
+        const size_t offset_in_page = offset % PAGE_BYTES;
+        const size_t first_bit = offset_in_page >> BYTES_PER_BIT_SHIFT;
+        const size_t num_bits = std::min(size, PAGE_BYTES) >> BYTES_PER_BIT_SHIFT;
+        const size_t mask = ~u64{0} >> (64 - num_bits);
+        pages[page] |= (~u64{0} & mask) << first_bit;
+    }
+
+    bool IsPageUsed(u64 page, u64 offset, u64 size) const noexcept {
+        const size_t offset_in_page = offset % PAGE_BYTES;
+        const size_t first_bit = offset_in_page >> BYTES_PER_BIT_SHIFT;
+        const size_t num_bits = std::min(size, PAGE_BYTES) >> BYTES_PER_BIT_SHIFT;
+        const size_t mask = ~u64{0} >> (64 - num_bits);
+        const size_t mask2 = (~u64{0} & mask) << first_bit;
+        return (pages[page] & mask2) != 0;
+    }
+
+private:
+    std::vector<u64> pages;
+};
+
+} // namespace VideoCommon

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -178,13 +178,14 @@ void BufferCacheRuntime::CopyBuffer(GLuint dst_buffer, Buffer& src_buffer,
 }
 
 void BufferCacheRuntime::CopyBuffer(Buffer& dst_buffer, GLuint src_buffer,
-                                    std::span<const VideoCommon::BufferCopy> copies, bool barrier) {
+                                    std::span<const VideoCommon::BufferCopy> copies, bool barrier,
+                                    bool) {
     CopyBuffer(dst_buffer.Handle(), src_buffer, copies, barrier);
 }
 
 void BufferCacheRuntime::CopyBuffer(Buffer& dst_buffer, Buffer& src_buffer,
-                                    std::span<const VideoCommon::BufferCopy> copies) {
+                                    std::span<const VideoCommon::BufferCopy> copies, bool) {
-    CopyBuffer(dst_buffer.Handle(), src_buffer.Handle(), copies);
+    CopyBuffer(dst_buffer.Handle(), src_buffer.Handle(), copies, true);
 }
 
 void BufferCacheRuntime::PreCopyBarrier() {

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -30,6 +30,8 @@ public:
 
     void MakeResident(GLenum access) noexcept;
 
+    void MarkUsage(u64 offset, u64 size) {}
+
     [[nodiscard]] GLuint View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format);
 
     [[nodiscard]] GLuint64EXT HostGpuAddr() const noexcept {
@@ -66,22 +68,29 @@ public:
 
     [[nodiscard]] StagingBufferMap DownloadStagingBuffer(size_t size);
 
+    bool CanReorderUpload(const Buffer&, std::span<const VideoCommon::BufferCopy>) {
+        return false;
+    }
+
     void CopyBuffer(GLuint dst_buffer, GLuint src_buffer,
-                    std::span<const VideoCommon::BufferCopy> copies, bool barrier = true);
+                    std::span<const VideoCommon::BufferCopy> copies, bool barrier);
 
     void CopyBuffer(GLuint dst_buffer, Buffer& src_buffer,
-                    std::span<const VideoCommon::BufferCopy> copies, bool barrier = true);
+                    std::span<const VideoCommon::BufferCopy> copies, bool barrier);
 
     void CopyBuffer(Buffer& dst_buffer, GLuint src_buffer,
-                    std::span<const VideoCommon::BufferCopy> copies, bool barrier = true);
+                    std::span<const VideoCommon::BufferCopy> copies, bool barrier,
+                    bool can_reorder_upload = false);
 
     void CopyBuffer(Buffer& dst_buffer, Buffer& src_buffer,
-                    std::span<const VideoCommon::BufferCopy> copies);
+                    std::span<const VideoCommon::BufferCopy> copies, bool);
 
     void PreCopyBarrier();
     void PostCopyBarrier();
     void Finish();
 
+    void TickFrame(VideoCommon::SlotVector<Buffer>&) noexcept {}
+
     void ClearBuffer(Buffer& dest_buffer, u32 offset, size_t size, u32 value);
 
     void BindIndexBuffer(Buffer& buffer, u32 offset, u32 size);

src/video_core/renderer_vulkan/blit_image.cpp

@@ -454,7 +454,7 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView
     const VkPipeline pipeline = FindOrEmplaceColorPipeline(key);
     scheduler.RequestRenderpass(dst_framebuffer);
     scheduler.Record([this, dst_region, src_region, pipeline, layout, sampler,
-                      src_view](vk::CommandBuffer cmdbuf) {
+                      src_view](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         // TODO: Barriers
         const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
         UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
@@ -479,7 +479,8 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView
     const VkPipeline pipeline = FindOrEmplaceColorPipeline(key);
     scheduler.RequestOutsideRenderPassOperationContext();
     scheduler.Record([this, dst_framebuffer, src_image_view, src_image, src_sampler, dst_region,
-                      src_region, src_size, pipeline, layout](vk::CommandBuffer cmdbuf) {
+                      src_region, src_size, pipeline,
+                      layout](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         TransitionImageLayout(cmdbuf, src_image, VK_IMAGE_LAYOUT_READ_ONLY_OPTIMAL);
         BeginRenderPass(cmdbuf, dst_framebuffer);
         const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
@@ -512,7 +513,7 @@ void BlitImageHelper::BlitDepthStencil(const Framebuffer* dst_framebuffer,
     const VkPipeline pipeline = FindOrEmplaceDepthStencilPipeline(key);
     scheduler.RequestRenderpass(dst_framebuffer);
     scheduler.Record([dst_region, src_region, pipeline, layout, sampler, src_depth_view,
-                      src_stencil_view, this](vk::CommandBuffer cmdbuf) {
+                      src_stencil_view, this](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         // TODO: Barriers
         const VkDescriptorSet descriptor_set = two_textures_descriptor_allocator.Commit();
         UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_depth_view,
@@ -581,8 +582,8 @@ void BlitImageHelper::ClearColor(const Framebuffer* dst_framebuffer, u8 color_ma
     const VkPipeline pipeline = FindOrEmplaceClearColorPipeline(key);
     const VkPipelineLayout layout = *clear_color_pipeline_layout;
     scheduler.RequestRenderpass(dst_framebuffer);
-    scheduler.Record(
+    scheduler.Record([pipeline, layout, color_mask, clear_color,
-        [pipeline, layout, color_mask, clear_color, dst_region](vk::CommandBuffer cmdbuf) {
+                      dst_region](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
         const std::array blend_color = {
             (color_mask & 0x1) ? 1.0f : 0.0f, (color_mask & 0x2) ? 1.0f : 0.0f,
@@ -608,9 +609,8 @@ void BlitImageHelper::ClearDepthStencil(const Framebuffer* dst_framebuffer, bool
     const VkPipeline pipeline = FindOrEmplaceClearStencilPipeline(key);
     const VkPipelineLayout layout = *clear_color_pipeline_layout;
     scheduler.RequestRenderpass(dst_framebuffer);
-    scheduler.Record([pipeline, layout, clear_depth, dst_region](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
-        constexpr std::array blend_constants{0.0f, 0.0f, 0.0f, 0.0f};
+        [pipeline, layout, clear_depth, dst_region](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
-        cmdbuf.SetBlendConstants(blend_constants.data());
             cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
             BindBlitState(cmdbuf, dst_region);
             cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, clear_depth);
@@ -627,7 +627,8 @@ void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_frameb
     const VkExtent2D extent = GetConversionExtent(src_image_view);
 
     scheduler.RequestRenderpass(dst_framebuffer);
-    scheduler.Record([pipeline, layout, sampler, src_view, extent, this](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([pipeline, layout, sampler, src_view, extent, this](vk::CommandBuffer cmdbuf,
+                                                                         vk::CommandBuffer) {
         const VkOffset2D offset{
             .x = 0,
             .y = 0,
@@ -673,7 +674,7 @@ void BlitImageHelper::ConvertDepthStencil(VkPipeline pipeline, const Framebuffer
 
     scheduler.RequestRenderpass(dst_framebuffer);
     scheduler.Record([pipeline, layout, sampler, src_depth_view, src_stencil_view, extent,
-                      this](vk::CommandBuffer cmdbuf) {
+                      this](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const VkOffset2D offset{
             .x = 0,
             .y = 0,
@@ -867,7 +868,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceClearStencilPipeline(
         .sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
         .pNext = nullptr,
         .flags = 0,
-        .depthTestEnable = key.depth_clear,
+        .depthTestEnable = VK_FALSE,
         .depthWriteEnable = key.depth_clear,
         .depthCompareOp = VK_COMPARE_OP_ALWAYS,
         .depthBoundsTestEnable = VK_FALSE,

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -239,7 +239,7 @@ void Vulkan::RendererVulkan::RenderScreenshot(const Tegra::FramebufferConfig& fr
         memory_allocator.CreateBuffer(dst_buffer_info, MemoryUsage::Download);
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([&](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([&](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const VkImageMemoryBarrier read_barrier{
             .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
             .pNext = nullptr,

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -201,7 +201,8 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
                     .depth = 1,
                 },
         };
-        scheduler.Record([this, copy, index = image_index](vk::CommandBuffer cmdbuf) {
+        scheduler.Record(
+            [this, copy, index = image_index](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 const VkImage image = *raw_images[index];
                 const VkImageMemoryBarrier base_barrier{
                     .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
@@ -230,8 +231,8 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
                 write_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
                 write_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
 
-            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_HOST_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
+                cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_HOST_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
-                                   read_barrier);
+                                       0, read_barrier);
                 cmdbuf.CopyBufferToImage(*buffer, image, VK_IMAGE_LAYOUT_GENERAL, copy);
                 cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
                                        VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
@@ -250,7 +251,7 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
             .height = (up_scale * framebuffer.height) >> down_shift,
         };
         scheduler.Record([this, index = image_index, size,
-                          anti_alias_pass](vk::CommandBuffer cmdbuf) {
+                          anti_alias_pass](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             const VkImageMemoryBarrier base_barrier{
                 .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
                 .pNext = nullptr,
@@ -375,7 +376,7 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
     }
 
     scheduler.Record([this, host_framebuffer, index = image_index,
-                      size = render_area](vk::CommandBuffer cmdbuf) {
+                      size = render_area](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const f32 bg_red = Settings::values.bg_red.GetValue() / 255.0f;
         const f32 bg_green = Settings::values.bg_green.GetValue() / 255.0f;
         const f32 bg_blue = Settings::values.bg_blue.GetValue() / 255.0f;

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -79,13 +79,13 @@ vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allo
 } // Anonymous namespace
 
 Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params)
-    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params) {}
+    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params), tracker{4096} {}
 
 Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
                VAddr cpu_addr_, u64 size_bytes_)
     : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(rasterizer_, cpu_addr_, size_bytes_),
-      device{&runtime.device}, buffer{
+      device{&runtime.device}, buffer{CreateBuffer(*device, runtime.memory_allocator, SizeBytes())},
-                                   CreateBuffer(*device, runtime.memory_allocator, SizeBytes())} {
+      tracker{SizeBytes()} {
     if (runtime.device.HasDebuggingToolAttached()) {
         buffer.SetObjectNameEXT(fmt::format("Buffer 0x{:x}", CpuAddr()).c_str());
     }
@@ -179,7 +179,8 @@ public:
         if (!host_visible) {
             scheduler.RequestOutsideRenderPassOperationContext();
             scheduler.Record([src_buffer = staging.buffer, src_offset = staging.offset,
-                              dst_buffer = *buffer, size_bytes](vk::CommandBuffer cmdbuf) {
+                              dst_buffer = *buffer,
+                              size_bytes](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 const VkBufferCopy copy{
                     .srcOffset = src_offset,
                     .dstOffset = 0,
@@ -210,7 +211,8 @@ public:
         const size_t sub_first_offset = static_cast<size_t>(first % 4) * GetQuadsNum(num_indices);
         const size_t offset =
             (sub_first_offset + GetQuadsNum(first)) * 6ULL * BytesPerIndex(index_type);
-        scheduler.Record([buffer_ = *buffer, index_type_, offset](vk::CommandBuffer cmdbuf) {
+        scheduler.Record(
+            [buffer_ = *buffer, index_type_, offset](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 cmdbuf.BindIndexBuffer(buffer_, offset, index_type_);
             });
     }
@@ -355,12 +357,31 @@ bool BufferCacheRuntime::CanReportMemoryUsage() const {
     return device.CanReportMemoryUsage();
 }
 
+void BufferCacheRuntime::TickFrame(VideoCommon::SlotVector<Buffer>& slot_buffers) noexcept {
+    for (auto it = slot_buffers.begin(); it != slot_buffers.end(); it++) {
+        it->ResetUsageTracking();
+    }
+}
+
 void BufferCacheRuntime::Finish() {
     scheduler.Finish();
 }
 
+bool BufferCacheRuntime::CanReorderUpload(const Buffer& buffer,
+                                          std::span<const VideoCommon::BufferCopy> copies) {
+    if (Settings::values.disable_buffer_reorder) {
+        return false;
+    }
+    const bool can_use_upload_cmdbuf =
+        std::ranges::all_of(copies, [&](const VideoCommon::BufferCopy& copy) {
+            return !buffer.IsRegionUsed(copy.dst_offset, copy.size);
+        });
+    return can_use_upload_cmdbuf;
+}
+
 void BufferCacheRuntime::CopyBuffer(VkBuffer dst_buffer, VkBuffer src_buffer,
-                                    std::span<const VideoCommon::BufferCopy> copies, bool barrier) {
+                                    std::span<const VideoCommon::BufferCopy> copies, bool barrier,
+                                    bool can_reorder_upload) {
     if (dst_buffer == VK_NULL_HANDLE || src_buffer == VK_NULL_HANDLE) {
         return;
     }
@@ -376,11 +397,21 @@ void BufferCacheRuntime::CopyBuffer(VkBuffer dst_buffer, VkBuffer src_buffer,
         .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
         .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
     };
+
     // Measuring a popular game, this number never exceeds the specified size once data is warmed up
     boost::container::small_vector<VkBufferCopy, 8> vk_copies(copies.size());
     std::ranges::transform(copies, vk_copies.begin(), MakeBufferCopy);
+    if (src_buffer == staging_pool.StreamBuf() && can_reorder_upload) {
+        scheduler.Record([src_buffer, dst_buffer, vk_copies](vk::CommandBuffer,
+                                                             vk::CommandBuffer upload_cmdbuf) {
+            upload_cmdbuf.CopyBuffer(src_buffer, dst_buffer, vk_copies);
+        });
+        return;
+    }
+
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([src_buffer, dst_buffer, vk_copies, barrier](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
+        [src_buffer, dst_buffer, vk_copies, barrier](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             if (barrier) {
                 cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
                                        VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
@@ -401,7 +432,7 @@ void BufferCacheRuntime::PreCopyBarrier() {
         .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
     };
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
                                0, READ_BARRIER);
     });
@@ -415,7 +446,7 @@ void BufferCacheRuntime::PostCopyBarrier() {
         .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
     };
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
                                0, WRITE_BARRIER);
     });
@@ -439,12 +470,13 @@ void BufferCacheRuntime::ClearBuffer(VkBuffer dest_buffer, u32 offset, size_t si
     };
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([dest_buffer, offset, size, value](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
+        [dest_buffer, offset, size, value](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
-                               0, READ_BARRIER);
+            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
+                                   VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
             cmdbuf.FillBuffer(dest_buffer, offset, size, value);
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
+            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
-                               0, WRITE_BARRIER);
+                                   VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, WRITE_BARRIER);
         });
 }
 
@@ -470,7 +502,8 @@ void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat
         ReserveNullBuffer();
         vk_buffer = *null_buffer;
     }
-    scheduler.Record([vk_buffer, vk_offset, vk_index_type](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
+        [vk_buffer, vk_offset, vk_index_type](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.BindIndexBuffer(vk_buffer, vk_offset, vk_index_type);
         });
 }
@@ -478,7 +511,7 @@ void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat
 void BufferCacheRuntime::BindQuadIndexBuffer(PrimitiveTopology topology, u32 first, u32 count) {
     if (count == 0) {
         ReserveNullBuffer();
-        scheduler.Record([this](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([this](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.BindIndexBuffer(*null_buffer, 0, VK_INDEX_TYPE_UINT32);
         });
         return;
@@ -499,7 +532,8 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset
         return;
     }
     if (device.IsExtExtendedDynamicStateSupported()) {
-        scheduler.Record([index, buffer, offset, size, stride](vk::CommandBuffer cmdbuf) {
+        scheduler.Record(
+            [index, buffer, offset, size, stride](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 const VkDeviceSize vk_offset = buffer != VK_NULL_HANDLE ? offset : 0;
                 const VkDeviceSize vk_size = buffer != VK_NULL_HANDLE ? size : VK_WHOLE_SIZE;
                 const VkDeviceSize vk_stride = stride;
@@ -511,7 +545,7 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset
             buffer = *null_buffer;
             offset = 0;
         }
-        scheduler.Record([index, buffer, offset](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([index, buffer, offset](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.BindVertexBuffer(index, buffer, offset);
         });
     }
@@ -533,7 +567,8 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
     }
     if (device.IsExtExtendedDynamicStateSupported()) {
         scheduler.Record([this, bindings_ = std::move(bindings),
-                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
+                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf,
+                                                                       vk::CommandBuffer) {
             cmdbuf.BindVertexBuffers2EXT(bindings_.min_index,
                                          std::min(bindings_.max_index - bindings_.min_index,
                                                   device.GetMaxVertexInputBindings()),
@@ -542,7 +577,8 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
         });
     } else {
         scheduler.Record([this, bindings_ = std::move(bindings),
-                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
+                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf,
+                                                                       vk::CommandBuffer) {
             cmdbuf.BindVertexBuffers(bindings_.min_index,
                                      std::min(bindings_.max_index - bindings_.min_index,
                                               device.GetMaxVertexInputBindings()),
@@ -565,7 +601,7 @@ void BufferCacheRuntime::BindTransformFeedbackBuffer(u32 index, VkBuffer buffer,
         offset = 0;
         size = 0;
     }
-    scheduler.Record([index, buffer, offset, size](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([index, buffer, offset, size](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const VkDeviceSize vk_offset = offset;
         const VkDeviceSize vk_size = size;
         cmdbuf.BindTransformFeedbackBuffersEXT(index, 1, &buffer, &vk_offset, &vk_size);
@@ -581,8 +617,8 @@ void BufferCacheRuntime::BindTransformFeedbackBuffers(VideoCommon::HostBindings<
     for (u32 index = 0; index < bindings.buffers.size(); ++index) {
         buffer_handles.push_back(bindings.buffers[index]->Handle());
     }
-    scheduler.Record([bindings_ = std::move(bindings),
+    scheduler.Record([bindings_ = std::move(bindings), buffer_handles_ = std::move(buffer_handles)](
-                      buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
+                         vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.BindTransformFeedbackBuffersEXT(0, static_cast<u32>(buffer_handles_.size()),
                                                buffer_handles_.data(), bindings_.offsets.data(),
                                                bindings_.sizes.data());
@@ -613,7 +649,7 @@ void BufferCacheRuntime::ReserveNullBuffer() {
     }
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([buffer = *null_buffer](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([buffer = *null_buffer](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.FillBuffer(buffer, 0, VK_WHOLE_SIZE, 0);
     });
 }

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -5,6 +5,7 @@
 
 #include "video_core/buffer_cache/buffer_cache_base.h"
 #include "video_core/buffer_cache/memory_tracker_base.h"
+#include "video_core/buffer_cache/usage_tracker.h"
 #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"
@@ -34,6 +35,18 @@ public:
         return *buffer;
     }
 
+    [[nodiscard]] bool IsRegionUsed(u64 offset, u64 size) const noexcept {
+        return tracker.IsUsed(offset, size);
+    }
+
+    void MarkUsage(u64 offset, u64 size) noexcept {
+        tracker.Track(offset, size);
+    }
+
+    void ResetUsageTracking() noexcept {
+        tracker.Reset();
+    }
+
     operator VkBuffer() const noexcept {
         return *buffer;
     }
@@ -49,6 +62,7 @@ private:
     const Device* device{};
     vk::Buffer buffer;
     std::vector<BufferView> views;
+    VideoCommon::UsageTracker tracker;
 };
 
 class QuadArrayIndexBuffer;
@@ -67,6 +81,8 @@ public:
                                 ComputePassDescriptorQueue& compute_pass_descriptor_queue,
                                 DescriptorPool& descriptor_pool);
 
+    void TickFrame(VideoCommon::SlotVector<Buffer>& slot_buffers) noexcept;
+
     void Finish();
 
     u64 GetDeviceLocalMemory() const;
@@ -79,12 +95,15 @@ public:
 
     [[nodiscard]] StagingBufferRef DownloadStagingBuffer(size_t size, bool deferred = false);
 
+    bool CanReorderUpload(const Buffer& buffer, std::span<const VideoCommon::BufferCopy> copies);
+
     void FreeDeferredStagingBuffer(StagingBufferRef& ref);
 
     void PreCopyBarrier();
 
     void CopyBuffer(VkBuffer src_buffer, VkBuffer dst_buffer,
-                    std::span<const VideoCommon::BufferCopy> copies, bool barrier = true);
+                    std::span<const VideoCommon::BufferCopy> copies, bool barrier,
+                    bool can_reorder_upload = false);
 
     void PostCopyBarrier();
 

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -324,7 +324,8 @@ std::pair<VkBuffer, VkDeviceSize> Uint8Pass::Assemble(u32 num_vertices, VkBuffer
     const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([this, descriptor_data, num_vertices](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
+        [this, descriptor_data, num_vertices](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             static constexpr u32 DISPATCH_SIZE = 1024;
             static constexpr VkMemoryBarrier WRITE_BARRIER{
                 .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
@@ -383,7 +384,7 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
 
     scheduler.RequestOutsideRenderPassOperationContext();
     scheduler.Record([this, descriptor_data, num_tri_vertices, base_vertex, index_shift,
-                      is_strip](vk::CommandBuffer cmdbuf) {
+                      is_strip](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         static constexpr u32 DISPATCH_SIZE = 1024;
         static constexpr VkMemoryBarrier WRITE_BARRIER{
             .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
@@ -423,7 +424,7 @@ void ConditionalRenderingResolvePass::Resolve(VkBuffer dst_buffer, VkBuffer src_
     const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([this, descriptor_data](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([this, descriptor_data](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         static constexpr VkMemoryBarrier read_barrier{
             .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
             .pNext = nullptr,
@@ -483,7 +484,7 @@ void QueriesPrefixScanPass::Run(VkBuffer accumulation_buffer, VkBuffer dst_buffe
 
         scheduler.RequestOutsideRenderPassOperationContext();
         scheduler.Record([this, descriptor_data, min_accumulation_limit, max_accumulation_limit,
-                          runs_to_do, used_offset](vk::CommandBuffer cmdbuf) {
+                          runs_to_do, used_offset](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             static constexpr VkMemoryBarrier read_barrier{
                 .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
                 .pNext = nullptr,
@@ -548,8 +549,8 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
     const VkImageAspectFlags aspect_mask = image.AspectMask();
     const VkImage vk_image = image.Handle();
     const bool is_initialized = image.ExchangeInitialization();
-    scheduler.Record([vk_pipeline, vk_image, aspect_mask,
+    scheduler.Record([vk_pipeline, vk_image, aspect_mask, is_initialized](vk::CommandBuffer cmdbuf,
-                      is_initialized](vk::CommandBuffer cmdbuf) {
+                                                                          vk::CommandBuffer) {
         const VkImageMemoryBarrier image_barrier{
             .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
             .pNext = nullptr,
@@ -592,7 +593,7 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
         ASSERT(params.destination == (std::array<s32, 3>{0, 0, 0}));
         ASSERT(params.bytes_per_block_log2 == 4);
         scheduler.Record([this, num_dispatches_x, num_dispatches_y, num_dispatches_z, block_dims,
-                          params, descriptor_data](vk::CommandBuffer cmdbuf) {
+                          params, descriptor_data](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             const AstcPushConstants uniforms{
                 .blocks_dims = block_dims,
                 .layer_stride = params.layer_stride,
@@ -608,7 +609,7 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
             cmdbuf.Dispatch(num_dispatches_x, num_dispatches_y, num_dispatches_z);
         });
     }
-    scheduler.Record([vk_image, aspect_mask](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([vk_image, aspect_mask](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const VkImageMemoryBarrier image_barrier{
             .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
             .pNext = nullptr,
@@ -699,7 +700,7 @@ void MSAACopyPass::CopyImage(Image& dst_image, Image& src_image,
         };
 
         scheduler.Record([this, dst = dst_image.Handle(), msaa_pipeline, num_dispatches,
-                          descriptor_data](vk::CommandBuffer cmdbuf) {
+                          descriptor_data](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             const VkDescriptorSet set = descriptor_allocator.Commit();
             device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
             cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, msaa_pipeline);

src/video_core/renderer_vulkan/vk_compute_pipeline.cpp

@@ -199,15 +199,15 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
 
     if (!is_built.load(std::memory_order::relaxed)) {
         // Wait for the pipeline to be built
-        scheduler.Record([this](vk::CommandBuffer) {
+        scheduler.Record([this](vk::CommandBuffer, vk::CommandBuffer) {
             std::unique_lock lock{build_mutex};
             build_condvar.wait(lock, [this] { return is_built.load(std::memory_order::relaxed); });
         });
     }
     const void* const descriptor_data{guest_descriptor_queue.UpdateData()};
     const bool is_rescaling = !info.texture_descriptors.empty() || !info.image_descriptors.empty();
-    scheduler.Record([this, descriptor_data, is_rescaling,
+    scheduler.Record([this, descriptor_data, is_rescaling, rescaling_data = rescaling.Data()](
-                      rescaling_data = rescaling.Data()](vk::CommandBuffer cmdbuf) {
+                         vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
         if (!descriptor_set_layout) {
             return;

src/video_core/renderer_vulkan/vk_fsr.cpp

@@ -38,7 +38,8 @@ VkImageView FSR::Draw(Scheduler& scheduler, size_t image_index, VkImageView imag
 
     UpdateDescriptorSet(image_index, image_view);
 
-    scheduler.Record([this, image_index, input_image_extent, crop_rect](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([this, image_index, input_image_extent, crop_rect](vk::CommandBuffer cmdbuf,
+                                                                        vk::CommandBuffer) {
         const VkImageMemoryBarrier base_barrier{
             .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
             .pNext = nullptr,

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -493,7 +493,7 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
 
     if (!is_built.load(std::memory_order::relaxed)) {
         // Wait for the pipeline to be built
-        scheduler.Record([this](vk::CommandBuffer) {
+        scheduler.Record([this](vk::CommandBuffer, vk::CommandBuffer) {
             std::unique_lock lock{build_mutex};
             build_condvar.wait(lock, [this] { return is_built.load(std::memory_order::relaxed); });
         });
@@ -502,10 +502,10 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
     const bool update_rescaling{scheduler.UpdateRescaling(is_rescaling)};
     const bool bind_pipeline{scheduler.UpdateGraphicsPipeline(this)};
     const void* const descriptor_data{guest_descriptor_queue.UpdateData()};
-    scheduler.Record([this, descriptor_data, bind_pipeline, rescaling_data = rescaling.Data(),
+    scheduler.Record(
-                      is_rescaling, update_rescaling,
+        [this, descriptor_data, bind_pipeline, rescaling_data = rescaling.Data(), is_rescaling,
-                      uses_render_area = render_area.uses_render_area,
+         update_rescaling, uses_render_area = render_area.uses_render_area,
-                      render_area_data = render_area.words](vk::CommandBuffer cmdbuf) {
+         render_area_data = render_area.words](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             if (bind_pipeline) {
                 cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline);
             }
@@ -528,12 +528,13 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
                 return;
             }
             if (uses_push_descriptor) {
-            cmdbuf.PushDescriptorSetWithTemplateKHR(*descriptor_update_template, *pipeline_layout,
+                cmdbuf.PushDescriptorSetWithTemplateKHR(*descriptor_update_template,
-                                                    0, descriptor_data);
+                                                        *pipeline_layout, 0, descriptor_data);
             } else {
                 const VkDescriptorSet descriptor_set{descriptor_allocator.Commit()};
                 const vk::Device& dev{device.GetLogical()};
-            dev.UpdateDescriptorSet(descriptor_set, *descriptor_update_template, descriptor_data);
+                dev.UpdateDescriptorSet(descriptor_set, *descriptor_update_template,
+                                        descriptor_data);
                 cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline_layout, 0,
                                           descriptor_set, nullptr);
             }

src/video_core/renderer_vulkan/vk_master_semaphore.cpp

@@ -100,12 +100,14 @@ void MasterSemaphore::Wait(u64 tick) {
     Refresh();
 }
 
-VkResult MasterSemaphore::SubmitQueue(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
+VkResult MasterSemaphore::SubmitQueue(vk::CommandBuffer& cmdbuf, vk::CommandBuffer& upload_cmdbuf,
-                                      VkSemaphore wait_semaphore, u64 host_tick) {
+                                      VkSemaphore signal_semaphore, VkSemaphore wait_semaphore,
+                                      u64 host_tick) {
     if (semaphore) {
-        return SubmitQueueTimeline(cmdbuf, signal_semaphore, wait_semaphore, host_tick);
+        return SubmitQueueTimeline(cmdbuf, upload_cmdbuf, signal_semaphore, wait_semaphore,
+                                   host_tick);
     } else {
-        return SubmitQueueFence(cmdbuf, signal_semaphore, wait_semaphore, host_tick);
+        return SubmitQueueFence(cmdbuf, upload_cmdbuf, signal_semaphore, wait_semaphore, host_tick);
     }
 }
 
@@ -115,6 +117,7 @@ static constexpr std::array<VkPipelineStageFlags, 2> wait_stage_masks{
 };
 
 VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
+                                              vk::CommandBuffer& upload_cmdbuf,
                                               VkSemaphore signal_semaphore,
                                               VkSemaphore wait_semaphore, u64 host_tick) {
     const VkSemaphore timeline_semaphore = *semaphore;
@@ -123,6 +126,8 @@ VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
     const std::array signal_values{host_tick, u64(0)};
     const std::array signal_semaphores{timeline_semaphore, signal_semaphore};
 
+    const std::array cmdbuffers{*upload_cmdbuf, *cmdbuf};
+
     const u32 num_wait_semaphores = wait_semaphore ? 1 : 0;
     const VkTimelineSemaphoreSubmitInfo timeline_si{
         .sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO,
@@ -138,8 +143,8 @@ VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
         .waitSemaphoreCount = num_wait_semaphores,
         .pWaitSemaphores = &wait_semaphore,
         .pWaitDstStageMask = wait_stage_masks.data(),
-        .commandBufferCount = 1,
+        .commandBufferCount = static_cast<u32>(cmdbuffers.size()),
-        .pCommandBuffers = cmdbuf.address(),
+        .pCommandBuffers = cmdbuffers.data(),
         .signalSemaphoreCount = num_signal_semaphores,
         .pSignalSemaphores = signal_semaphores.data(),
     };
@@ -147,19 +152,23 @@ VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
     return device.GetGraphicsQueue().Submit(submit_info);
 }
 
-VkResult MasterSemaphore::SubmitQueueFence(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
+VkResult MasterSemaphore::SubmitQueueFence(vk::CommandBuffer& cmdbuf,
-                                           VkSemaphore wait_semaphore, u64 host_tick) {
+                                           vk::CommandBuffer& upload_cmdbuf,
+                                           VkSemaphore signal_semaphore, VkSemaphore wait_semaphore,
+                                           u64 host_tick) {
     const u32 num_signal_semaphores = signal_semaphore ? 1 : 0;
     const u32 num_wait_semaphores = wait_semaphore ? 1 : 0;
 
+    const std::array cmdbuffers{*upload_cmdbuf, *cmdbuf};
+
     const VkSubmitInfo submit_info{
         .sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
         .pNext = nullptr,
         .waitSemaphoreCount = num_wait_semaphores,
         .pWaitSemaphores = &wait_semaphore,
         .pWaitDstStageMask = wait_stage_masks.data(),
-        .commandBufferCount = 1,
+        .commandBufferCount = static_cast<u32>(cmdbuffers.size()),
-        .pCommandBuffers = cmdbuf.address(),
+        .pCommandBuffers = cmdbuffers.data(),
         .signalSemaphoreCount = num_signal_semaphores,
         .pSignalSemaphores = &signal_semaphore,
     };

src/video_core/renderer_vulkan/vk_master_semaphore.h

@@ -52,14 +52,16 @@ public:
     void Wait(u64 tick);
 
     /// Submits the device graphics queue, updating the tick as necessary
-    VkResult SubmitQueue(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
+    VkResult SubmitQueue(vk::CommandBuffer& cmdbuf, vk::CommandBuffer& upload_cmdbuf,
-                         VkSemaphore wait_semaphore, u64 host_tick);
+                         VkSemaphore signal_semaphore, VkSemaphore wait_semaphore, u64 host_tick);
 
 private:
-    VkResult SubmitQueueTimeline(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
+    VkResult SubmitQueueTimeline(vk::CommandBuffer& cmdbuf, vk::CommandBuffer& upload_cmdbuf,
-                                 VkSemaphore wait_semaphore, u64 host_tick);
+                                 VkSemaphore signal_semaphore, VkSemaphore wait_semaphore,
-    VkResult SubmitQueueFence(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
+                                 u64 host_tick);
-                              VkSemaphore wait_semaphore, u64 host_tick);
+    VkResult SubmitQueueFence(vk::CommandBuffer& cmdbuf, vk::CommandBuffer& upload_cmdbuf,
+                              VkSemaphore signal_semaphore, VkSemaphore wait_semaphore,
+                              u64 host_tick);
 
     void WaitThread(std::stop_token token);
 

src/video_core/renderer_vulkan/vk_present_manager.cpp

@@ -166,7 +166,7 @@ void PresentManager::Present(Frame* frame) {
         return;
     }
 
-    scheduler.Record([this, frame](vk::CommandBuffer) {
+    scheduler.Record([this, frame](vk::CommandBuffer, vk::CommandBuffer) {
         std::unique_lock lock{queue_mutex};
         present_queue.push(frame);
         frame_cv.notify_one();

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -138,7 +138,8 @@ public:
         };
         accumulation_buffer = memory_allocator.CreateBuffer(buffer_ci, MemoryUsage::DeviceLocal);
         scheduler.RequestOutsideRenderPassOperationContext();
-        scheduler.Record([buffer = *accumulation_buffer](vk::CommandBuffer cmdbuf) {
+        scheduler.Record(
+            [buffer = *accumulation_buffer](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 cmdbuf.FillBuffer(buffer, 0, 8, 0);
             });
     }
@@ -150,8 +151,8 @@ public:
             return;
         }
         ReserveHostQuery();
-        scheduler.Record([query_pool = current_query_pool,
+        scheduler.Record([query_pool = current_query_pool, query_index = current_bank_slot](
-                          query_index = current_bank_slot](vk::CommandBuffer cmdbuf) {
+                             vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             const bool use_precise = Settings::IsGPULevelHigh();
             cmdbuf.BeginQuery(query_pool, static_cast<u32>(query_index),
                               use_precise ? VK_QUERY_CONTROL_PRECISE_BIT : 0);
@@ -163,8 +164,8 @@ public:
         if (!has_started) {
             return;
         }
-        scheduler.Record([query_pool = current_query_pool,
+        scheduler.Record([query_pool = current_query_pool, query_index = current_bank_slot](
-                          query_index = current_bank_slot](vk::CommandBuffer cmdbuf) {
+                             vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.EndQuery(query_pool, static_cast<u32>(query_index));
         });
         has_started = false;
@@ -227,8 +228,8 @@ public:
             auto& resolve_buffer = buffers[resolve_buffer_index];
             VkQueryPool query_pool = bank->GetInnerPool();
             scheduler.RequestOutsideRenderPassOperationContext();
-            scheduler.Record([start, amount, base_offset, query_pool,
+            scheduler.Record([start, amount, base_offset, query_pool, buffer = *resolve_buffer](
-                              buffer = *resolve_buffer](vk::CommandBuffer cmdbuf) {
+                                 vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 const VkBufferMemoryBarrier copy_query_pool_barrier{
                     .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
                     .pNext = nullptr,
@@ -291,7 +292,8 @@ public:
 
         } else {
             scheduler.RequestOutsideRenderPassOperationContext();
-            scheduler.Record([buffer = *accumulation_buffer](vk::CommandBuffer cmdbuf) {
+            scheduler.Record(
+                [buffer = *accumulation_buffer](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                     cmdbuf.FillBuffer(buffer, 0, 8, 0);
                 });
         }
@@ -608,7 +610,7 @@ public:
     void Sync(StagingBufferRef& stagging_buffer, size_t extra_offset, size_t start, size_t size) {
         scheduler.RequestOutsideRenderPassOperationContext();
         scheduler.Record([this, dst_buffer = stagging_buffer.buffer, extra_offset, start,
-                          size](vk::CommandBuffer cmdbuf) {
+                          size](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             std::array<VkBufferCopy, 1> copy{VkBufferCopy{
                 .srcOffset = start * QUERY_SIZE,
                 .dstOffset = extra_offset,
@@ -793,7 +795,7 @@ public:
             .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
         };
         scheduler.RequestOutsideRenderPassOperationContext();
-        scheduler.Record([](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
                                    VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, WRITE_BARRIER);
         });
@@ -841,13 +843,14 @@ private:
         }
         has_flushed_end_pending = true;
         if (!has_started || buffers_count == 0) {
-            scheduler.Record([](vk::CommandBuffer cmdbuf) {
+            scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 cmdbuf.BeginTransformFeedbackEXT(0, 0, nullptr, nullptr);
             });
             UpdateBuffers();
             return;
         }
-        scheduler.Record([this, total = static_cast<u32>(buffers_count)](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([this, total = static_cast<u32>(buffers_count)](vk::CommandBuffer cmdbuf,
+                                                                         vk::CommandBuffer) {
             cmdbuf.BeginTransformFeedbackEXT(0, total, counter_buffers.data(), offsets.data());
         });
         UpdateBuffers();
@@ -861,12 +864,12 @@ private:
         has_flushed_end_pending = false;
 
         if (buffers_count == 0) {
-            scheduler.Record([](vk::CommandBuffer cmdbuf) {
+            scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 cmdbuf.EndTransformFeedbackEXT(0, 0, nullptr, nullptr);
             });
         } else {
-            scheduler.Record([this,
+            scheduler.Record([this, total = static_cast<u32>(buffers_count)](
-                              total = static_cast<u32>(buffers_count)](vk::CommandBuffer cmdbuf) {
+                                 vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 cmdbuf.EndTransformFeedbackEXT(0, total, counter_buffers.data(), offsets.data());
             });
         }
@@ -918,7 +921,7 @@ private:
         scheduler.RequestOutsideRenderPassOperationContext();
         scheduler.Record([dst_buffer = current_bank->GetBuffer(),
                           src_buffer = counter_buffers[stream], src_offset = offsets[stream],
-                          slot](vk::CommandBuffer cmdbuf) {
+                          slot](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT,
                                    VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
             std::array<VkBufferCopy, 1> copy{VkBufferCopy{
@@ -1253,8 +1256,9 @@ void QueryCacheRuntime::PauseHostConditionalRendering() {
         return;
     }
     if (impl->is_hcr_running) {
-        impl->scheduler.Record(
+        impl->scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
-            [](vk::CommandBuffer cmdbuf) { cmdbuf.EndConditionalRenderingEXT(); });
+            cmdbuf.EndConditionalRenderingEXT();
+        });
     }
     impl->is_hcr_running = false;
 }
@@ -1264,7 +1268,8 @@ void QueryCacheRuntime::ResumeHostConditionalRendering() {
         return;
     }
     if (!impl->is_hcr_running) {
-        impl->scheduler.Record([hcr_setup = impl->hcr_setup](vk::CommandBuffer cmdbuf) {
+        impl->scheduler.Record(
+            [hcr_setup = impl->hcr_setup](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 cmdbuf.BeginConditionalRenderingEXT(hcr_setup);
             });
     }
@@ -1436,12 +1441,12 @@ void QueryCacheRuntime::Barriers(bool is_prebarrier) {
         .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
     };
     if (is_prebarrier) {
-        impl->scheduler.Record([](vk::CommandBuffer cmdbuf) {
+        impl->scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
                                    VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
         });
     } else {
-        impl->scheduler.Record([](vk::CommandBuffer cmdbuf) {
+        impl->scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
                                    VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, WRITE_BARRIER);
         });
@@ -1536,8 +1541,9 @@ void QueryCacheRuntime::SyncValues(std::span<SyncValuesType> values, VkBuffer ba
     }
 
     impl->scheduler.RequestOutsideRenderPassOperationContext();
-    impl->scheduler.Record([src_buffer, dst_buffers = std::move(impl->buffers_to_upload_to),
+    impl->scheduler.Record(
-                            vk_copies = std::move(impl->copies_setup)](vk::CommandBuffer cmdbuf) {
+        [src_buffer, dst_buffers = std::move(impl->buffers_to_upload_to),
+         vk_copies = std::move(impl->copies_setup)](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             size_t size = dst_buffers.size();
             for (size_t i = 0; i < size; i++) {
                 cmdbuf.CopyBuffer(src_buffer, dst_buffers[i].first, vk_copies[i]);

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -216,7 +216,7 @@ void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
         const auto& draw_state = maxwell3d->draw_manager->GetDrawState();
         const u32 num_instances{instance_count};
         const DrawParams draw_params{MakeDrawParams(draw_state, num_instances, is_indexed)};
-        scheduler.Record([draw_params](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([draw_params](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             if (draw_params.is_indexed) {
                 cmdbuf.DrawIndexed(draw_params.num_vertices, draw_params.num_instances,
                                    draw_params.first_index, draw_params.base_vertex,
@@ -238,7 +238,7 @@ void RasterizerVulkan::DrawIndirect() {
         const auto& offset = indirect_buffer.second;
         if (params.is_byte_count) {
             scheduler.Record([buffer_obj = buffer->Handle(), offset,
-                              stride = params.stride](vk::CommandBuffer cmdbuf) {
+                              stride = params.stride](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 cmdbuf.DrawIndirectByteCountEXT(1, 0, buffer_obj, offset, 0,
                                                 static_cast<u32>(stride));
             });
@@ -250,7 +250,7 @@ void RasterizerVulkan::DrawIndirect() {
             const auto& offset_base = count.second;
             scheduler.Record([draw_buffer_obj = draw_buffer->Handle(),
                               buffer_obj = buffer->Handle(), offset_base, offset,
-                              params](vk::CommandBuffer cmdbuf) {
+                              params](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 if (params.is_indexed) {
                     cmdbuf.DrawIndexedIndirectCount(
                         buffer_obj, offset, draw_buffer_obj, offset_base,
@@ -263,7 +263,8 @@ void RasterizerVulkan::DrawIndirect() {
             });
             return;
         }
-        scheduler.Record([buffer_obj = buffer->Handle(), offset, params](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([buffer_obj = buffer->Handle(), offset, params](vk::CommandBuffer cmdbuf,
+                                                                         vk::CommandBuffer) {
             if (params.is_indexed) {
                 cmdbuf.DrawIndexedIndirect(buffer_obj, offset,
                                            static_cast<u32>(params.max_draw_counts),
@@ -387,7 +388,8 @@ void RasterizerVulkan::Clear(u32 layer_count) {
 
         if (regs.clear_surface.R && regs.clear_surface.G && regs.clear_surface.B &&
             regs.clear_surface.A) {
-            scheduler.Record([color_attachment, clear_value, clear_rect](vk::CommandBuffer cmdbuf) {
+            scheduler.Record([color_attachment, clear_value, clear_rect](vk::CommandBuffer cmdbuf,
+                                                                         vk::CommandBuffer) {
                 const VkClearAttachment attachment{
                     .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
                     .colorAttachment = color_attachment,
@@ -422,8 +424,7 @@ void RasterizerVulkan::Clear(u32 layer_count) {
         return;
     }
 
-    if (use_stencil && framebuffer->HasAspectStencilBit() && regs.stencil_front_mask != 0xFF &&
+    if (use_stencil && regs.stencil_front_mask != 0xFF && regs.stencil_front_mask != 0) {
-        regs.stencil_front_mask != 0) {
         Region2D dst_region = {
             Offset2D{.x = clear_rect.rect.offset.x, .y = clear_rect.rect.offset.y},
             Offset2D{.x = clear_rect.rect.offset.x + static_cast<s32>(clear_rect.rect.extent.width),
@@ -434,7 +435,7 @@ void RasterizerVulkan::Clear(u32 layer_count) {
                                      regs.stencil_front_func_mask, dst_region);
     } else {
         scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil,
-                          clear_rect, aspect_flags](vk::CommandBuffer cmdbuf) {
+                          clear_rect, aspect_flags](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             VkClearAttachment attachment;
             attachment.aspectMask = aspect_flags;
             attachment.colorAttachment = 0;
@@ -466,14 +467,16 @@ void RasterizerVulkan::DispatchCompute() {
             buffer_cache.ObtainBuffer(*indirect_address, 12, sync_info, post_op);
         scheduler.RequestOutsideRenderPassOperationContext();
         scheduler.Record([indirect_buffer = buffer->Handle(),
-                          indirect_offset = offset](vk::CommandBuffer cmdbuf) {
+                          indirect_offset = offset](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.DispatchIndirect(indirect_buffer, indirect_offset);
         });
         return;
     }
     const std::array<u32, 3> dim{qmd.grid_dim_x, qmd.grid_dim_y, qmd.grid_dim_z};
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([dim](vk::CommandBuffer cmdbuf) { cmdbuf.Dispatch(dim[0], dim[1], dim[2]); });
+    scheduler.Record([dim](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
+        cmdbuf.Dispatch(dim[0], dim[1], dim[2]);
+    });
 }
 
 void RasterizerVulkan::ResetCounter(VideoCommon::QueryType type) {
@@ -691,7 +694,7 @@ void RasterizerVulkan::WaitForIdle() {
     query_cache.NotifyWFI();
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([event = *wfi_event, flags](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([event = *wfi_event, flags](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.SetEvent(event, flags);
         cmdbuf.WaitEvents(event, flags, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, {}, {}, {});
     });
@@ -949,7 +952,9 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& reg
             .minDepth = 0.0f,
             .maxDepth = 1.0f,
         };
-        scheduler.Record([viewport](vk::CommandBuffer cmdbuf) { cmdbuf.SetViewport(0, viewport); });
+        scheduler.Record([viewport](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
+            cmdbuf.SetViewport(0, viewport);
+        });
         return;
     }
     const bool is_rescaling{texture_cache.IsRescaling()};
@@ -964,7 +969,7 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& reg
         GetViewportState(device, regs, 12, scale), GetViewportState(device, regs, 13, scale),
         GetViewportState(device, regs, 14, scale), GetViewportState(device, regs, 15, scale),
     };
-    scheduler.Record([this, viewport_list](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([this, viewport_list](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const u32 num_viewports = std::min<u32>(device.GetMaxViewports(), Maxwell::NumViewports);
         const vk::Span<VkViewport> viewports(viewport_list.data(), num_viewports);
         cmdbuf.SetViewport(0, viewports);
@@ -999,7 +1004,7 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs
         GetScissorState(regs, 14, up_scale, down_shift),
         GetScissorState(regs, 15, up_scale, down_shift),
     };
-    scheduler.Record([this, scissor_list](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([this, scissor_list](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const u32 num_scissors = std::min<u32>(device.GetMaxViewports(), Maxwell::NumViewports);
         const vk::Span<VkRect2D> scissors(scissor_list.data(), num_scissors);
         cmdbuf.SetScissor(0, scissors);
@@ -1032,9 +1037,10 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {
         units = static_cast<float>(static_cast<double>(units) * rescale_factor);
         return false;
     })();
-    scheduler.Record([constant = units, clamp = regs.depth_bias_clamp,
+    scheduler.Record(
-                      factor = regs.slope_scale_depth_bias, force_unorm,
+        [constant = units, clamp = regs.depth_bias_clamp, factor = regs.slope_scale_depth_bias,
-                      precise = device.HasExactDepthBiasControl()](vk::CommandBuffer cmdbuf) {
+         force_unorm,
+         precise = device.HasExactDepthBiasControl()](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             if (force_unorm) {
                 VkDepthBiasRepresentationInfoEXT info{
                     .sType = VK_STRUCTURE_TYPE_DEPTH_BIAS_REPRESENTATION_INFO_EXT,
@@ -1056,16 +1062,19 @@ void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& reg
     }
     const std::array blend_color = {regs.blend_color.r, regs.blend_color.g, regs.blend_color.b,
                                     regs.blend_color.a};
-    scheduler.Record(
+    scheduler.Record([blend_color](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
-        [blend_color](vk::CommandBuffer cmdbuf) { cmdbuf.SetBlendConstants(blend_color.data()); });
+        cmdbuf.SetBlendConstants(blend_color.data());
+    });
 }
 
 void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) {
     if (!state_tracker.TouchDepthBounds()) {
         return;
     }
-    scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]](
+    scheduler.Record([min = regs.depth_bounds[0],
-                         vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBounds(min, max); });
+                      max = regs.depth_bounds[1]](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
+        cmdbuf.SetDepthBounds(min, max);
+    });
 }
 
 void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1093,7 +1102,8 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs)
                 }
             }
             scheduler.Record([front_ref = regs.stencil_front_ref, back_ref = regs.stencil_back_ref,
-                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
+                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf,
+                                                                        vk::CommandBuffer) {
                 const bool set_back = two_sided && front_ref != back_ref;
                 // Front face
                 cmdbuf.SetStencilReference(set_back ? VK_STENCIL_FACE_FRONT_BIT
@@ -1119,7 +1129,8 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs)
             }
             scheduler.Record([front_write_mask = regs.stencil_front_mask,
                               back_write_mask = regs.stencil_back_mask,
-                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
+                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf,
+                                                                        vk::CommandBuffer) {
                 const bool set_back = two_sided && front_write_mask != back_write_mask;
                 // Front face
                 cmdbuf.SetStencilWriteMask(set_back ? VK_STENCIL_FACE_FRONT_BIT
@@ -1145,7 +1156,8 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs)
             }
             scheduler.Record([front_test_mask = regs.stencil_front_func_mask,
                               back_test_mask = regs.stencil_back_func_mask,
-                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
+                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf,
+                                                                        vk::CommandBuffer) {
                 const bool set_back = two_sided && front_test_mask != back_test_mask;
                 // Front face
                 cmdbuf.SetStencilCompareMask(set_back ? VK_STENCIL_FACE_FRONT_BIT
@@ -1166,7 +1178,8 @@ void RasterizerVulkan::UpdateLineWidth(Tegra::Engines::Maxwell3D::Regs& regs) {
     }
     const float width =
         regs.line_anti_alias_enable ? regs.line_width_smooth : regs.line_width_aliased;
-    scheduler.Record([width](vk::CommandBuffer cmdbuf) { cmdbuf.SetLineWidth(width); });
+    scheduler.Record(
+        [width](vk::CommandBuffer cmdbuf, vk::CommandBuffer) { cmdbuf.SetLineWidth(width); });
 }
 
 void RasterizerVulkan::UpdateCullMode(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1174,7 +1187,7 @@ void RasterizerVulkan::UpdateCullMode(Tegra::Engines::Maxwell3D::Regs& regs) {
         return;
     }
     scheduler.Record([enabled = regs.gl_cull_test_enabled,
-                      cull_face = regs.gl_cull_face](vk::CommandBuffer cmdbuf) {
+                      cull_face = regs.gl_cull_face](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.SetCullModeEXT(enabled ? MaxwellToVK::CullFace(cull_face) : VK_CULL_MODE_NONE);
     });
 }
@@ -1188,7 +1201,7 @@ void RasterizerVulkan::UpdateDepthBoundsTestEnable(Tegra::Engines::Maxwell3D::Re
         LOG_WARNING(Render_Vulkan, "Depth bounds is enabled but not supported");
         enabled = false;
     }
-    scheduler.Record([enable = enabled](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([enable = enabled](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.SetDepthBoundsTestEnableEXT(enable);
     });
 }
@@ -1197,7 +1210,8 @@ void RasterizerVulkan::UpdateDepthTestEnable(Tegra::Engines::Maxwell3D::Regs& re
     if (!state_tracker.TouchDepthTestEnable()) {
         return;
     }
-    scheduler.Record([enable = regs.depth_test_enable](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
+        [enable = regs.depth_test_enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.SetDepthTestEnableEXT(enable);
         });
 }
@@ -1206,7 +1220,8 @@ void RasterizerVulkan::UpdateDepthWriteEnable(Tegra::Engines::Maxwell3D::Regs& r
     if (!state_tracker.TouchDepthWriteEnable()) {
         return;
     }
-    scheduler.Record([enable = regs.depth_write_enabled](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
+        [enable = regs.depth_write_enabled](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.SetDepthWriteEnableEXT(enable);
         });
 }
@@ -1215,7 +1230,8 @@ void RasterizerVulkan::UpdatePrimitiveRestartEnable(Tegra::Engines::Maxwell3D::R
     if (!state_tracker.TouchPrimitiveRestartEnable()) {
         return;
     }
-    scheduler.Record([enable = regs.primitive_restart.enabled](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
+        [enable = regs.primitive_restart.enabled](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.SetPrimitiveRestartEnableEXT(enable);
         });
 }
@@ -1224,7 +1240,8 @@ void RasterizerVulkan::UpdateRasterizerDiscardEnable(Tegra::Engines::Maxwell3D::
     if (!state_tracker.TouchRasterizerDiscardEnable()) {
         return;
     }
-    scheduler.Record([disable = regs.rasterize_enable](vk::CommandBuffer cmdbuf) {
+    scheduler.Record(
+        [disable = regs.rasterize_enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.SetRasterizerDiscardEnableEXT(disable == 0);
         });
 }
@@ -1260,15 +1277,16 @@ void RasterizerVulkan::UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& re
     };
     const u32 topology_index = static_cast<u32>(maxwell3d->draw_manager->GetDrawState().topology);
     const u32 enable = enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]];
-    scheduler.Record(
+    scheduler.Record([enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
-        [enable](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBiasEnableEXT(enable != 0); });
+        cmdbuf.SetDepthBiasEnableEXT(enable != 0);
+    });
 }
 
 void RasterizerVulkan::UpdateLogicOpEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
     if (!state_tracker.TouchLogicOpEnable()) {
         return;
     }
-    scheduler.Record([enable = regs.logic_op.enable](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([enable = regs.logic_op.enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.SetLogicOpEnableEXT(enable != 0);
     });
 }
@@ -1283,15 +1301,16 @@ void RasterizerVulkan::UpdateDepthClampEnable(Tegra::Engines::Maxwell3D::Regs& r
                             Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
                         regs.viewport_clip_control.geometry_clip ==
                             Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
-    scheduler.Record(
+    scheduler.Record([is_enabled](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
-        [is_enabled](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthClampEnableEXT(is_enabled); });
+        cmdbuf.SetDepthClampEnableEXT(is_enabled);
+    });
 }
 
 void RasterizerVulkan::UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs) {
     if (!state_tracker.TouchDepthCompareOp()) {
         return;
     }
-    scheduler.Record([func = regs.depth_test_func](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([func = regs.depth_test_func](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.SetDepthCompareOpEXT(MaxwellToVK::ComparisonOp(func));
     });
 }
@@ -1306,8 +1325,9 @@ void RasterizerVulkan::UpdateFrontFace(Tegra::Engines::Maxwell3D::Regs& regs) {
         front_face = front_face == VK_FRONT_FACE_CLOCKWISE ? VK_FRONT_FACE_COUNTER_CLOCKWISE
                                                            : VK_FRONT_FACE_CLOCKWISE;
     }
-    scheduler.Record(
+    scheduler.Record([front_face](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
-        [front_face](vk::CommandBuffer cmdbuf) { cmdbuf.SetFrontFaceEXT(front_face); });
+        cmdbuf.SetFrontFaceEXT(front_face);
+    });
 }
 
 void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1325,7 +1345,7 @@ void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
         const Maxwell::StencilOp::Op back_zpass = regs.stencil_back_op.zpass;
         const Maxwell::ComparisonOp back_compare = regs.stencil_back_op.func;
         scheduler.Record([fail, zfail, zpass, compare, back_fail, back_zfail, back_zpass,
-                          back_compare](vk::CommandBuffer cmdbuf) {
+                          back_compare](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.SetStencilOpEXT(VK_STENCIL_FACE_FRONT_BIT, MaxwellToVK::StencilOp(fail),
                                    MaxwellToVK::StencilOp(zpass), MaxwellToVK::StencilOp(zfail),
                                    MaxwellToVK::ComparisonOp(compare));
@@ -1336,7 +1356,8 @@ void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
         });
     } else {
         // Front face defines the stencil op of both faces
-        scheduler.Record([fail, zfail, zpass, compare](vk::CommandBuffer cmdbuf) {
+        scheduler.Record(
+            [fail, zfail, zpass, compare](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
                 cmdbuf.SetStencilOpEXT(VK_STENCIL_FACE_FRONT_AND_BACK, MaxwellToVK::StencilOp(fail),
                                        MaxwellToVK::StencilOp(zpass), MaxwellToVK::StencilOp(zfail),
                                        MaxwellToVK::ComparisonOp(compare));
@@ -1351,7 +1372,8 @@ void RasterizerVulkan::UpdateLogicOp(Tegra::Engines::Maxwell3D::Regs& regs) {
     const auto op_value = static_cast<u32>(regs.logic_op.op);
     auto op = op_value >= 0x1500 && op_value < 0x1510 ? static_cast<VkLogicOp>(op_value - 0x1500)
                                                       : VK_LOGIC_OP_NO_OP;
-    scheduler.Record([op](vk::CommandBuffer cmdbuf) { cmdbuf.SetLogicOpEXT(op); });
+    scheduler.Record(
+        [op](vk::CommandBuffer cmdbuf, vk::CommandBuffer) { cmdbuf.SetLogicOpEXT(op); });
 }
 
 void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1377,7 +1399,7 @@ void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
                 current |= VK_COLOR_COMPONENT_A_BIT;
             }
         }
-        scheduler.Record([setup_masks](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([setup_masks](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.SetColorWriteMaskEXT(0, setup_masks);
         });
     }
@@ -1387,7 +1409,7 @@ void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
         std::ranges::transform(
             regs.blend.enable, setup_enables.begin(),
             [&](const auto& is_enabled) { return is_enabled != 0 ? VK_TRUE : VK_FALSE; });
-        scheduler.Record([setup_enables](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([setup_enables](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.SetColorBlendEnableEXT(0, setup_enables);
         });
     }
@@ -1410,7 +1432,7 @@ void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
             }
             blend_setup(regs.blend_per_target[index]);
         }
-        scheduler.Record([setup_blends](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([setup_blends](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             cmdbuf.SetColorBlendEquationEXT(0, setup_blends);
         });
     }
@@ -1420,7 +1442,7 @@ void RasterizerVulkan::UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs&
     if (!state_tracker.TouchStencilTestEnable()) {
         return;
     }
-    scheduler.Record([enable = regs.stencil_enable](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([enable = regs.stencil_enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.SetStencilTestEnableEXT(enable);
     });
 }
@@ -1478,7 +1500,7 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs)
             .divisor = is_instanced ? input_binding.frequency : 1,
         });
     }
-    scheduler.Record([bindings, attributes](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([bindings, attributes](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         cmdbuf.SetVertexInputEXT(bindings, attributes);
     });
 }

src/video_core/renderer_vulkan/vk_scheduler.cpp

@@ -22,11 +22,12 @@ namespace Vulkan {
 
 MICROPROFILE_DECLARE(Vulkan_WaitForWorker);
 
-void Scheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf) {
+void Scheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf,
+                                         vk::CommandBuffer upload_cmdbuf) {
     auto command = first;
     while (command != nullptr) {
         auto next = command->GetNext();
-        command->Execute(cmdbuf);
+        command->Execute(cmdbuf, upload_cmdbuf);
         command->~Command();
         command = next;
     }
@@ -102,7 +103,8 @@ void Scheduler::RequestRenderpass(const Framebuffer* framebuffer) {
     state.framebuffer = framebuffer_handle;
     state.render_area = render_area;
 
-    Record([renderpass, framebuffer_handle, render_area](vk::CommandBuffer cmdbuf) {
+    Record(
+        [renderpass, framebuffer_handle, render_area](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
             const VkRenderPassBeginInfo renderpass_bi{
                 .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
                 .pNext = nullptr,
@@ -180,7 +182,7 @@ void Scheduler::WorkerThread(std::stop_token stop_token) {
             // Perform the work, tracking whether the chunk was a submission
             // before executing.
             const bool has_submit = work->HasSubmit();
-            work->ExecuteAll(current_cmdbuf);
+            work->ExecuteAll(current_cmdbuf, current_upload_cmdbuf);
 
             // If the chunk was a submission, reallocate the command buffer.
             if (has_submit) {
@@ -205,6 +207,13 @@ void Scheduler::AllocateWorkerCommandBuffer() {
         .flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
         .pInheritanceInfo = nullptr,
     });
+    current_upload_cmdbuf = vk::CommandBuffer(command_pool->Commit(), device.GetDispatchLoader());
+    current_upload_cmdbuf.Begin({
+        .sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
+        .pNext = nullptr,
+        .flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
+        .pInheritanceInfo = nullptr,
+    });
 }
 
 u64 Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_semaphore) {
@@ -212,7 +221,9 @@ u64 Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_se
     InvalidateState();
 
     const u64 signal_value = master_semaphore->NextTick();
-    Record([signal_semaphore, wait_semaphore, signal_value, this](vk::CommandBuffer cmdbuf) {
+    Record([signal_semaphore, wait_semaphore, signal_value, this](vk::CommandBuffer cmdbuf,
+                                                                  vk::CommandBuffer upload_cmdbuf) {
+        upload_cmdbuf.End();
         cmdbuf.End();
 
         if (on_submit) {
@@ -221,7 +232,7 @@ u64 Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_se
 
         std::scoped_lock lock{submit_mutex};
         switch (const VkResult result = master_semaphore->SubmitQueue(
-                    cmdbuf, signal_semaphore, wait_semaphore, signal_value)) {
+                    cmdbuf, upload_cmdbuf, signal_semaphore, wait_semaphore, signal_value)) {
         case VK_SUCCESS:
             break;
         case VK_ERROR_DEVICE_LOST:
@@ -275,7 +286,7 @@ void Scheduler::EndRenderPass() {
         return;
     }
     Record([num_images = num_renderpass_images, images = renderpass_images,
-            ranges = renderpass_image_ranges](vk::CommandBuffer cmdbuf) {
+            ranges = renderpass_image_ranges](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         std::array<VkImageMemoryBarrier, 9> barriers;
         for (size_t i = 0; i < num_images; ++i) {
             barriers[i] = VkImageMemoryBarrier{

src/video_core/renderer_vulkan/vk_scheduler.h

@@ -119,7 +119,7 @@ private:
     public:
         virtual ~Command() = default;
 
-        virtual void Execute(vk::CommandBuffer cmdbuf) const = 0;
+        virtual void Execute(vk::CommandBuffer cmdbuf, vk::CommandBuffer upload_cmdbuf) const = 0;
 
         Command* GetNext() const {
             return next;
@@ -142,8 +142,8 @@ private:
         TypedCommand(TypedCommand&&) = delete;
         TypedCommand& operator=(TypedCommand&&) = delete;
 
-        void Execute(vk::CommandBuffer cmdbuf) const override {
+        void Execute(vk::CommandBuffer cmdbuf, vk::CommandBuffer upload_cmdbuf) const override {
-            command(cmdbuf);
+            command(cmdbuf, upload_cmdbuf);
         }
 
     private:
@@ -152,7 +152,7 @@ private:
 
     class CommandChunk final {
     public:
-        void ExecuteAll(vk::CommandBuffer cmdbuf);
+        void ExecuteAll(vk::CommandBuffer cmdbuf, vk::CommandBuffer upload_cmdbuf);
 
         template <typename T>
         bool Record(T& command) {
@@ -228,6 +228,7 @@ private:
     VideoCommon::QueryCacheBase<QueryCacheParams>* query_cache = nullptr;
 
     vk::CommandBuffer current_cmdbuf;
+    vk::CommandBuffer current_upload_cmdbuf;
 
     std::unique_ptr<CommandChunk> chunk;
     std::function<void()> on_submit;

src/video_core/renderer_vulkan/vk_smaa.cpp

@@ -103,7 +103,7 @@ void UploadImage(const Device& device, MemoryAllocator& allocator, Scheduler& sc
     }}};
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([&](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([&](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         TransitionImageLayout(cmdbuf, *image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
                               VK_IMAGE_LAYOUT_UNDEFINED);
         cmdbuf.CopyBufferToImage(*upload_buffer, *image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
@@ -672,7 +672,7 @@ void SMAA::UploadImages(Scheduler& scheduler) {
     UploadImage(m_device, m_allocator, scheduler, m_static_images[Search], search_extent,
                 VK_FORMAT_R8_UNORM, ARRAY_TO_SPAN(searchTexBytes));
 
-    scheduler.Record([&](vk::CommandBuffer& cmdbuf) {
+    scheduler.Record([&](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         for (auto& images : m_dynamic_images) {
             for (size_t i = 0; i < MaxDynamicImage; i++) {
                 ClearColorImage(cmdbuf, *images.images[i]);
@@ -707,7 +707,7 @@ VkImageView SMAA::Draw(Scheduler& scheduler, size_t image_index, VkImage source_
     UpdateDescriptorSets(source_image_view, image_index);
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([=, this](vk::CommandBuffer& cmdbuf) {
+    scheduler.Record([=, this](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         TransitionImageLayout(cmdbuf, source_image, VK_IMAGE_LAYOUT_GENERAL);
         TransitionImageLayout(cmdbuf, edges_image, VK_IMAGE_LAYOUT_GENERAL);
         BeginRenderPass(cmdbuf, m_renderpasses[EdgeDetection], edge_detection_framebuffer,

src/video_core/renderer_vulkan/vk_staging_buffer_pool.h

@@ -36,6 +36,10 @@ public:
     StagingBufferRef Request(size_t size, MemoryUsage usage, bool deferred = false);
     void FreeDeferred(StagingBufferRef& ref);
 
+    [[nodiscard]] VkBuffer StreamBuf() const noexcept {
+        return *stream_buffer;
+    }
+
     void TickFrame();
 
 private:

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -709,7 +709,7 @@ void BlitScale(Scheduler& scheduler, VkImage src_image, VkImage dst_image, const
 
     scheduler.RequestOutsideRenderPassOperationContext();
     scheduler.Record([dst_image, src_image, extent, resources, aspect_mask, resolution, is_2d,
-                      vk_filter, up_scaling](vk::CommandBuffer cmdbuf) {
+                      vk_filter, up_scaling](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const VkOffset2D src_size{
             .x = static_cast<s32>(up_scaling ? extent.width : resolution.ScaleUp(extent.width)),
             .y = static_cast<s32>(is_2d && up_scaling ? extent.height
@@ -955,7 +955,7 @@ void TextureCacheRuntime::ReinterpretImage(Image& dst, Image& src,
     const VkImage src_image = src.Handle();
     scheduler.RequestOutsideRenderPassOperationContext();
     scheduler.Record([dst_image, src_image, copy_buffer, src_aspect_mask, dst_aspect_mask,
-                      vk_in_copies, vk_out_copies](vk::CommandBuffer cmdbuf) {
+                      vk_in_copies, vk_out_copies](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         RangedBarrierRange dst_range;
         RangedBarrierRange src_range;
         for (const VkBufferImageCopy& copy : vk_in_copies) {
@@ -1104,7 +1104,7 @@ void TextureCacheRuntime::BlitImage(Framebuffer* dst_framebuffer, ImageView& dst
     const bool is_resolve = is_src_msaa && !is_dst_msaa;
     scheduler.RequestOutsideRenderPassOperationContext();
     scheduler.Record([filter, dst_region, src_region, dst_image, src_image, dst_layers, src_layers,
-                      aspect_mask, is_resolve](vk::CommandBuffer cmdbuf) {
+                      aspect_mask, is_resolve](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const std::array read_barriers{
             VkImageMemoryBarrier{
                 .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
@@ -1240,7 +1240,8 @@ void TextureCacheRuntime::CopyImage(Image& dst, Image& src,
     const VkImage dst_image = dst.Handle();
     const VkImage src_image = src.Handle();
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([dst_image, src_image, aspect_mask, vk_copies](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([dst_image, src_image, aspect_mask, vk_copies](vk::CommandBuffer cmdbuf,
+                                                                    vk::CommandBuffer) {
         RangedBarrierRange dst_range;
         RangedBarrierRange src_range;
         for (const VkImageCopy& copy : vk_copies) {
@@ -1402,7 +1403,7 @@ void Image::UploadMemory(VkBuffer buffer, VkDeviceSize offset,
     const VkImageAspectFlags vk_aspect_mask = aspect_mask;
     const bool is_initialized = std::exchange(initialized, true);
     scheduler->Record([src_buffer, vk_image, vk_aspect_mask, is_initialized,
-                       vk_copies](vk::CommandBuffer cmdbuf) {
+                       vk_copies](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         CopyBufferToImage(cmdbuf, src_buffer, vk_image, vk_aspect_mask, is_initialized, vk_copies);
     });
     if (is_rescaled) {
@@ -1441,7 +1442,8 @@ void Image::DownloadMemory(std::span<VkBuffer> buffers_span, std::span<VkDeviceS
     }
     scheduler->RequestOutsideRenderPassOperationContext();
     scheduler->Record([buffers = std::move(buffers_vector), image = *original_image,
-                       aspect_mask_ = aspect_mask, vk_copies](vk::CommandBuffer cmdbuf) {
+                       aspect_mask_ = aspect_mask,
+                       vk_copies](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
         const VkImageMemoryBarrier read_barrier{
             .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
             .pNext = nullptr,

src/video_core/texture_cache/slot_vector.h

@@ -138,6 +138,10 @@ public:
         return Iterator(this, SlotId{SlotId::INVALID_INDEX});
     }
 
+    [[nodiscard]] size_t size() const noexcept {
+        return values_capacity - free_list.size();
+    }
+
 private:
     struct NonTrivialDummy {
         NonTrivialDummy() noexcept {}

src/video_core/vulkan_common/vulkan_wrapper.h

@@ -1098,6 +1098,10 @@ public:
         return &handle;
     }
 
+    VkCommandBuffer operator*() const noexcept {
+        return handle;
+    }
+
     void Begin(const VkCommandBufferBeginInfo& begin_info) const {
         Check(dld->vkBeginCommandBuffer(handle, &begin_info));
     }

src/yuzu/configuration/configure_debug.cpp

@@ -61,6 +61,8 @@ void ConfigureDebug::SetConfiguration() {
     ui->enable_all_controllers->setChecked(Settings::values.enable_all_controllers.GetValue());
     ui->enable_renderdoc_hotkey->setEnabled(runtime_lock);
     ui->enable_renderdoc_hotkey->setChecked(Settings::values.enable_renderdoc_hotkey.GetValue());
+    ui->disable_buffer_reorder->setEnabled(runtime_lock);
+    ui->disable_buffer_reorder->setChecked(Settings::values.disable_buffer_reorder.GetValue());
     ui->enable_graphics_debugging->setEnabled(runtime_lock);
     ui->enable_graphics_debugging->setChecked(Settings::values.renderer_debug.GetValue());
     ui->enable_shader_feedback->setEnabled(runtime_lock);
@@ -114,6 +116,7 @@ void ConfigureDebug::ApplyConfiguration() {
     Settings::values.enable_all_controllers = ui->enable_all_controllers->isChecked();
     Settings::values.renderer_debug = ui->enable_graphics_debugging->isChecked();
     Settings::values.enable_renderdoc_hotkey = ui->enable_renderdoc_hotkey->isChecked();
+    Settings::values.disable_buffer_reorder = ui->disable_buffer_reorder->isChecked();
     Settings::values.renderer_shader_feedback = ui->enable_shader_feedback->isChecked();
     Settings::values.cpu_debug_mode = ui->enable_cpu_debugging->isChecked();
     Settings::values.enable_nsight_aftermath = ui->enable_nsight_aftermath->isChecked();

src/yuzu/configuration/configure_debug.ui

@@ -271,19 +271,6 @@
           </widget>
          </item>
          <item row="8" column="0">
-          <widget class="QCheckBox" name="disable_macro_hle">
-           <property name="enabled">
-            <bool>true</bool>
-           </property>
-           <property name="toolTip">
-            <string>When checked, it disables the macro HLE functions. Enabling this makes games run slower</string>
-           </property>
-           <property name="text">
-            <string>Disable Macro HLE</string>
-           </property>
-          </widget>
-         </item>
-         <item row="7" column="0">
           <widget class="QCheckBox" name="dump_macros">
            <property name="enabled">
             <bool>true</bool>
@@ -306,59 +293,7 @@
            </property>
           </widget>
          </item>
-         <item row="2" column="0">
-          <widget class="QCheckBox" name="enable_shader_feedback">
-           <property name="toolTip">
-            <string>When checked, yuzu will log statistics about the compiled pipeline cache</string>
-           </property>
-           <property name="text">
-            <string>Enable Shader Feedback</string>
-           </property>
-          </widget>
-         </item>
          <item row="6" column="0">
-          <widget class="QCheckBox" name="disable_macro_jit">
-           <property name="enabled">
-            <bool>true</bool>
-           </property>
-           <property name="toolTip">
-            <string>When checked, it disables the macro Just In Time compiler. Enabling this makes games run slower</string>
-           </property>
-           <property name="text">
-            <string>Disable Macro JIT</string>
-           </property>
-          </widget>
-         </item>
-         <item row="9" column="0">
-          <spacer name="verticalSpacer_5">
-           <property name="orientation">
-            <enum>Qt::Vertical</enum>
-           </property>
-           <property name="sizeType">
-            <enum>QSizePolicy::Preferred</enum>
-           </property>
-           <property name="sizeHint" stdset="0">
-            <size>
-             <width>20</width>
-             <height>0</height>
-            </size>
-           </property>
-          </spacer>
-         </item>
-         <item row="0" column="0">
-          <widget class="QCheckBox" name="enable_graphics_debugging">
-           <property name="enabled">
-            <bool>true</bool>
-           </property>
-           <property name="toolTip">
-            <string>When checked, the graphics API enters a slower debugging mode</string>
-           </property>
-           <property name="text">
-            <string>Enable Graphics Debugging</string>
-           </property>
-          </widget>
-         </item>
-         <item row="5" column="0">
           <widget class="QCheckBox" name="dump_shaders">
            <property name="enabled">
             <bool>true</bool>
@@ -378,6 +313,81 @@
            </property>
           </widget>
          </item>
+         <item row="7" column="0">
+          <widget class="QCheckBox" name="disable_macro_jit">
+           <property name="enabled">
+            <bool>true</bool>
+           </property>
+           <property name="toolTip">
+            <string>When checked, it disables the macro Just In Time compiler. Enabling this makes games run slower</string>
+           </property>
+           <property name="text">
+            <string>Disable Macro JIT</string>
+           </property>
+          </widget>
+         </item>
+         <item row="9" column="0">
+          <widget class="QCheckBox" name="disable_macro_hle">
+           <property name="enabled">
+            <bool>true</bool>
+           </property>
+           <property name="toolTip">
+            <string>When checked, it disables the macro HLE functions. Enabling this makes games run slower</string>
+           </property>
+           <property name="text">
+            <string>Disable Macro HLE</string>
+           </property>
+          </widget>
+         </item>
+         <item row="0" column="0">
+          <widget class="QCheckBox" name="enable_graphics_debugging">
+           <property name="enabled">
+            <bool>true</bool>
+           </property>
+           <property name="toolTip">
+            <string>When checked, the graphics API enters a slower debugging mode</string>
+           </property>
+           <property name="text">
+            <string>Enable Graphics Debugging</string>
+           </property>
+          </widget>
+         </item>
+         <item row="10" column="0">
+          <spacer name="verticalSpacer_5">
+           <property name="orientation">
+            <enum>Qt::Vertical</enum>
+           </property>
+           <property name="sizeType">
+            <enum>QSizePolicy::Preferred</enum>
+           </property>
+           <property name="sizeHint" stdset="0">
+            <size>
+             <width>20</width>
+             <height>0</height>
+            </size>
+           </property>
+          </spacer>
+         </item>
+         <item row="2" column="0">
+          <widget class="QCheckBox" name="enable_shader_feedback">
+           <property name="toolTip">
+            <string>When checked, yuzu will log statistics about the compiled pipeline cache</string>
+           </property>
+           <property name="text">
+            <string>Enable Shader Feedback</string>
+           </property>
+          </widget>
+         </item>
+         <item row="5" column="0">
+          <widget class="QCheckBox" name="disable_buffer_reorder">
+           <property name="toolTip">
+            <string>&lt;html&gt;&lt;head/&gt;&lt;body&gt;&lt;p&gt;When checked, disables reording of mapped memory uploads which allows to associate uploads with specific draws. May reduce performance in some cases.&lt;/p&gt;&lt;/body&gt;&lt;/html&gt;</string>
+           </property>
+           <property name="text">
+            <string>Disable Buffer Reorder</string>
+           </property>
+          </widget>
+         </item>
         </layout>
        </widget>
       </item>

src/yuzu/main.cpp

@@ -1556,6 +1556,10 @@ void GMainWindow::ConnectMenuEvents() {
     // Tools
     connect_menu(ui->action_Rederive, std::bind(&GMainWindow::OnReinitializeKeys, this,
                                                 ReinitializeKeyBehavior::Warning));
+    connect_menu(ui->action_Load_Cabinet_Nickname_Owner, &GMainWindow::OnCabinetNicknameAndOwner);
+    connect_menu(ui->action_Load_Cabinet_Eraser, &GMainWindow::OnCabinetEraser);
+    connect_menu(ui->action_Load_Cabinet_Restorer, &GMainWindow::OnCabinetRestorer);
+    connect_menu(ui->action_Load_Cabinet_Formatter, &GMainWindow::OnCabinetFormatter);
     connect_menu(ui->action_Load_Mii_Edit, &GMainWindow::OnMiiEdit);
     connect_menu(ui->action_Capture_Screenshot, &GMainWindow::OnCaptureScreenshot);
 
@@ -1573,6 +1577,7 @@ void GMainWindow::ConnectMenuEvents() {
 
 void GMainWindow::UpdateMenuState() {
     const bool is_paused = emu_thread == nullptr || !emu_thread->IsRunning();
+    const bool is_firmware_available = CheckFirmwarePresence();
 
     const std::array running_actions{
         ui->action_Stop,
@@ -1583,10 +1588,22 @@ void GMainWindow::UpdateMenuState() {
         ui->action_Pause,
     };
 
+    const std::array applet_actions{
+        ui->action_Load_Cabinet_Nickname_Owner,
+        ui->action_Load_Cabinet_Eraser,
+        ui->action_Load_Cabinet_Restorer,
+        ui->action_Load_Cabinet_Formatter,
+        ui->action_Load_Mii_Edit,
+    };
+
     for (QAction* action : running_actions) {
         action->setEnabled(emulation_running);
     }
 
+    for (QAction* action : applet_actions) {
+        action->setEnabled(is_firmware_available);
+    }
+
     ui->action_Capture_Screenshot->setEnabled(emulation_running && !is_paused);
 
     if (emulation_running && is_paused) {
@@ -1596,8 +1613,6 @@ void GMainWindow::UpdateMenuState() {
     }
 
     multiplayer_state->UpdateNotificationStatus();
-
-    ui->action_Load_Mii_Edit->setEnabled(CheckFirmwarePresence());
 }
 
 void GMainWindow::OnDisplayTitleBars(bool show) {
@@ -2108,6 +2123,8 @@ void GMainWindow::OnEmulationStopped() {
     OnTasStateChanged();
     render_window->FinalizeCamera();
 
+    system->GetAppletManager().SetCurrentAppletId(Service::AM::Applets::AppletId::None);
+
     // Enable all controllers
     system->HIDCore().SetSupportedStyleTag({Core::HID::NpadStyleSet::All});
 
@@ -4157,6 +4174,46 @@ void GMainWindow::OnToggleStatusBar() {
     statusBar()->setVisible(ui->action_Show_Status_Bar->isChecked());
 }
 
+void GMainWindow::OnCabinetNicknameAndOwner() {
+    OnCabinet(Service::NFP::CabinetMode::StartNicknameAndOwnerSettings);
+}
+
+void GMainWindow::OnCabinetEraser() {
+    OnCabinet(Service::NFP::CabinetMode::StartGameDataEraser);
+}
+
+void GMainWindow::OnCabinetRestorer() {
+    OnCabinet(Service::NFP::CabinetMode::StartRestorer);
+}
+
+void GMainWindow::OnCabinetFormatter() {
+    OnCabinet(Service::NFP::CabinetMode::StartFormatter);
+}
+
+void GMainWindow::OnCabinet(Service::NFP::CabinetMode mode) {
+    constexpr u64 CabinetId = 0x0100000000001002ull;
+    auto bis_system = system->GetFileSystemController().GetSystemNANDContents();
+    if (!bis_system) {
+        QMessageBox::warning(this, tr("No firmware available"),
+                             tr("Please install the firmware to use the Amiibo applet."));
+        return;
+    }
+
+    auto cabinet_nca = bis_system->GetEntry(CabinetId, FileSys::ContentRecordType::Program);
+    if (!cabinet_nca) {
+        QMessageBox::warning(this, tr("Cabinet Applet"),
+                             tr("Amiibo applet is not available. Please reinstall firmware."));
+        return;
+    }
+
+    system->GetAppletManager().SetCurrentAppletId(Service::AM::Applets::AppletId::Cabinet);
+    system->GetAppletManager().SetCabinetMode(mode);
+
+    QString filename = QString::fromStdString((cabinet_nca->GetFullPath()));
+    UISettings::values.roms_path = QFileInfo(filename).path();
+    BootGame(filename);
+}
+
 void GMainWindow::OnMiiEdit() {
     constexpr u64 MiiEditId = 0x0100000000001009ull;
     auto bis_system = system->GetFileSystemController().GetSystemNANDContents();
@@ -4173,6 +4230,8 @@ void GMainWindow::OnMiiEdit() {
         return;
     }
 
+    system->GetAppletManager().SetCurrentAppletId(Service::AM::Applets::AppletId::MiiEdit);
+
     const auto filename = QString::fromStdString((mii_applet_nca->GetFullPath()));
     UISettings::values.roms_path = QFileInfo(filename).path();
     BootGame(filename);

src/yuzu/main.h

@@ -106,6 +106,10 @@ namespace Service::NFC {
 class NfcDevice;
 } // namespace Service::NFC
 
+namespace Service::NFP {
+enum class CabinetMode : u8;
+} // namespace Service::NFP
+
 namespace Ui {
 class MainWindow;
 }
@@ -370,6 +374,11 @@ private slots:
     void ResetWindowSize720();
     void ResetWindowSize900();
     void ResetWindowSize1080();
+    void OnCabinetNicknameAndOwner();
+    void OnCabinetEraser();
+    void OnCabinetRestorer();
+    void OnCabinetFormatter();
+    void OnCabinet(Service::NFP::CabinetMode mode);
     void OnMiiEdit();
     void OnCaptureScreenshot();
     void OnReinitializeKeys(ReinitializeKeyBehavior behavior);

src/yuzu/main.ui

@@ -137,6 +137,15 @@
     <property name="title">
      <string>&amp;Tools</string>
     </property>
+    <widget class="QMenu" name="menu_cabinet_applet">
+     <property name="title">
+      <string>&amp;Amiibo</string>
+     </property>
+     <addaction name="action_Load_Cabinet_Nickname_Owner"/>
+     <addaction name="action_Load_Cabinet_Eraser"/>
+     <addaction name="action_Load_Cabinet_Restorer"/>
+     <addaction name="action_Load_Cabinet_Formatter"/>
+    </widget>
     <widget class="QMenu" name="menuTAS">
      <property name="title">
       <string>&amp;TAS</string>
@@ -150,6 +159,7 @@
     <addaction name="action_Rederive"/>
     <addaction name="action_Verify_installed_contents"/>
     <addaction name="separator"/>
+    <addaction name="menu_cabinet_applet"/>
     <addaction name="action_Load_Mii_Edit"/>
     <addaction name="separator"/>
     <addaction name="action_Capture_Screenshot"/>
@@ -370,6 +380,26 @@
     <string>&amp;Capture Screenshot</string>
    </property>
   </action>
+  <action name="action_Load_Cabinet_Nickname_Owner">
+   <property name="text">
+    <string>&amp;Set Nickname and Owner</string>
+   </property>
+  </action>
+  <action name="action_Load_Cabinet_Eraser">
+   <property name="text">
+    <string>&amp;Delete Game Data</string>
+   </property>
+  </action>
+  <action name="action_Load_Cabinet_Restorer">
+   <property name="text">
+    <string>&amp;Restore Amiibo</string>
+   </property>
+  </action>
+  <action name="action_Load_Cabinet_Formatter">
+   <property name="text">
+    <string>&amp;Format Amiibo</string>
+   </property>
+  </action>
   <action name="action_Load_Mii_Edit">
     <property name="text">
       <string>Open &amp;Mii Editor</string>