Use the rendertarget format of the correct RT rather than the first valid

vfs_vector: avoid n^2 lookup in layeredfs building

src/common/settings.cpp

@@ -62,6 +62,7 @@ void LogSettings() {
     log_setting("Renderer_AccelerateASTC", values.accelerate_astc.GetValue());
     log_setting("Renderer_AsyncASTC", values.async_astc.GetValue());
     log_setting("Renderer_UseVsync", values.vsync_mode.GetValue());
+    log_setting("Renderer_UseReactiveFlushing", values.use_reactive_flushing.GetValue());
     log_setting("Renderer_ShaderBackend", values.shader_backend.GetValue());
     log_setting("Renderer_UseAsynchronousShaders", values.use_asynchronous_shaders.GetValue());
     log_setting("Renderer_AnisotropicFilteringLevel", values.max_anisotropy.GetValue());
@@ -223,6 +224,7 @@ void RestoreGlobalState(bool is_powered_on) {
     values.nvdec_emulation.SetGlobal(true);
     values.accelerate_astc.SetGlobal(true);
     values.async_astc.SetGlobal(true);
+    values.use_reactive_flushing.SetGlobal(true);
     values.shader_backend.SetGlobal(true);
     values.use_asynchronous_shaders.SetGlobal(true);
     values.use_fast_gpu_time.SetGlobal(true);

src/common/settings.h

@@ -465,6 +465,7 @@ struct Values {
     SwitchableSetting<bool> async_astc{false, "async_astc"};
     Setting<VSyncMode, true> vsync_mode{VSyncMode::FIFO, VSyncMode::Immediate,
                                         VSyncMode::FIFORelaxed, "use_vsync"};
+    SwitchableSetting<bool> use_reactive_flushing{true, "use_reactive_flushing"};
     SwitchableSetting<ShaderBackend, true> shader_backend{ShaderBackend::GLSL, ShaderBackend::GLSL,
                                                           ShaderBackend::SPIRV, "shader_backend"};
     SwitchableSetting<bool> use_asynchronous_shaders{false, "use_asynchronous_shaders"};

src/core/core.cpp

@@ -612,6 +612,10 @@ void System::PrepareReschedule(const u32 core_index) {
     impl->kernel.PrepareReschedule(core_index);
 }
 
+size_t System::GetCurrentHostThreadID() const {
+    return impl->kernel.GetCurrentHostThreadID();
+}
+
 PerfStatsResults System::GetAndResetPerfStats() {
     return impl->GetAndResetPerfStats();
 }

src/core/core.h

@@ -222,6 +222,8 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule(u32 core_index);
 
+    [[nodiscard]] size_t GetCurrentHostThreadID() const;
+
     /// Gets and resets core performance statistics
     [[nodiscard]] PerfStatsResults GetAndResetPerfStats();
 

src/core/file_sys/vfs_layered.cpp

@@ -2,6 +2,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <algorithm>
+#include <set>
 #include <utility>
 #include "core/file_sys/vfs_layered.h"
 
@@ -58,11 +59,13 @@ std::string LayeredVfsDirectory::GetFullPath() const {
 
 std::vector<VirtualFile> LayeredVfsDirectory::GetFiles() const {
     std::vector<VirtualFile> out;
+    std::set<std::string, std::less<>> out_names;
+
     for (const auto& layer : dirs) {
         for (const auto& file : layer->GetFiles()) {
-            if (std::find_if(out.begin(), out.end(), [&file](const VirtualFile& comp) {
-                    return comp->GetName() == file->GetName();
-                }) == out.end()) {
+            auto file_name = file->GetName();
+            if (!out_names.contains(file_name)) {
+                out_names.emplace(std::move(file_name));
                 out.push_back(file);
             }
         }

src/core/file_sys/vfs_vector.cpp

@@ -67,6 +67,23 @@ VectorVfsDirectory::VectorVfsDirectory(std::vector<VirtualFile> files_,
 
 VectorVfsDirectory::~VectorVfsDirectory() = default;
 
+VirtualFile VectorVfsDirectory::GetFile(std::string_view file_name) const {
+    if (!optimized_file_index_built) {
+        optimized_file_index.clear();
+        for (size_t i = 0; i < files.size(); i++) {
+            optimized_file_index.emplace(files[i]->GetName(), i);
+        }
+        optimized_file_index_built = true;
+    }
+
+    const auto it = optimized_file_index.find(file_name);
+    if (it != optimized_file_index.end()) {
+        return files[it->second];
+    }
+
+    return nullptr;
+}
+
 std::vector<VirtualFile> VectorVfsDirectory::GetFiles() const {
     return files;
 }
@@ -107,6 +124,7 @@ bool VectorVfsDirectory::DeleteSubdirectory(std::string_view subdir_name) {
 }
 
 bool VectorVfsDirectory::DeleteFile(std::string_view file_name) {
+    optimized_file_index_built = false;
     return FindAndRemoveVectorElement(files, file_name);
 }
 
@@ -124,6 +142,7 @@ VirtualFile VectorVfsDirectory::CreateFile(std::string_view file_name) {
 }
 
 void VectorVfsDirectory::AddFile(VirtualFile file) {
+    optimized_file_index_built = false;
     files.push_back(std::move(file));
 }
 

src/core/file_sys/vfs_vector.h

@@ -105,6 +105,7 @@ public:
                                 VirtualDir parent = nullptr);
     ~VectorVfsDirectory() override;
 
+    VirtualFile GetFile(std::string_view file_name) const override;
     std::vector<VirtualFile> GetFiles() const override;
     std::vector<VirtualDir> GetSubdirectories() const override;
     bool IsWritable() const override;
@@ -126,6 +127,9 @@ private:
 
     VirtualDir parent;
     std::string name;
+
+    mutable std::map<std::string, size_t, std::less<>> optimized_file_index;
+    mutable bool optimized_file_index_built{};
 };
 
 } // namespace FileSys

src/core/hid/emulated_controller.cpp

@@ -688,6 +688,12 @@ void EmulatedController::SetMotionParam(std::size_t index, Common::ParamPackage
     ReloadInput();
 }
 
+void EmulatedController::StartMotionCalibration() {
+    for (ControllerMotionInfo& motion : controller.motion_values) {
+        motion.emulated.Calibrate();
+    }
+}
+
 void EmulatedController::SetButton(const Common::Input::CallbackStatus& callback, std::size_t index,
                                    Common::UUID uuid) {
     if (index >= controller.button_values.size()) {
@@ -979,7 +985,6 @@ void EmulatedController::SetMotion(const Common::Input::CallbackStatus& callback
     emulated.SetUserGyroThreshold(raw_status.gyro.x.properties.threshold);
     emulated.UpdateRotation(raw_status.delta_timestamp);
     emulated.UpdateOrientation(raw_status.delta_timestamp);
-    force_update_motion = raw_status.force_update;
 
     auto& motion = controller.motion_state[index];
     motion.accel = emulated.GetAcceleration();
@@ -1618,19 +1623,6 @@ NpadGcTriggerState EmulatedController::GetTriggers() const {
 
 MotionState EmulatedController::GetMotions() const {
     std::unique_lock lock{mutex};
-
-    // Some drivers like mouse motion need constant refreshing
-    if (force_update_motion) {
-        for (auto& device : motion_devices) {
-            if (!device) {
-                continue;
-            }
-            lock.unlock();
-            device->ForceUpdate();
-            lock.lock();
-        }
-    }
-
     return controller.motion_state;
 }
 
@@ -1696,8 +1688,21 @@ void EmulatedController::DeleteCallback(int key) {
     callback_list.erase(iterator);
 }
 
-void EmulatedController::TurboButtonUpdate() {
+void EmulatedController::StatusUpdate() {
     turbo_button_state = (turbo_button_state + 1) % (TURBO_BUTTON_DELAY * 2);
+
+    // Some drivers like key motion need constant refreshing
+    for (std::size_t index = 0; index < motion_devices.size(); ++index) {
+        const auto& raw_status = controller.motion_values[index].raw_status;
+        auto& device = motion_devices[index];
+        if (!raw_status.force_update) {
+            continue;
+        }
+        if (!device) {
+            continue;
+        }
+        device->ForceUpdate();
+    }
 }
 
 NpadButton EmulatedController::GetTurboButtonMask() const {

src/core/hid/emulated_controller.h

@@ -290,6 +290,9 @@ public:
      */
     void SetMotionParam(std::size_t index, Common::ParamPackage param);
 
+    /// Auto calibrates the current motion devices
+    void StartMotionCalibration();
+
     /// Returns the latest button status from the controller with parameters
     ButtonValues GetButtonsValues() const;
 
@@ -415,8 +418,8 @@ public:
      */
     void DeleteCallback(int key);
 
-    /// Swaps the state of the turbo buttons
-    void TurboButtonUpdate();
+    /// Swaps the state of the turbo buttons and updates motion input
+    void StatusUpdate();
 
 private:
     /// creates input devices from params
@@ -528,7 +531,6 @@ private:
     bool is_configuring{false};
     bool system_buttons_enabled{true};
     f32 motion_sensitivity{Core::HID::MotionInput::IsAtRestStandard};
-    bool force_update_motion{false};
     u32 turbo_button_state{0};
 
     // Temporary values to avoid doing changes while the controller is in configuring mode

src/core/hid/input_converter.cpp

@@ -86,7 +86,7 @@ Common::Input::MotionStatus TransformToMotion(const Common::Input::CallbackStatu
             .range = 1.0f,
             .offset = 0.0f,
         };
-        status.delta_timestamp = 5000;
+        status.delta_timestamp = 1000;
         status.force_update = true;
         status.accel.x = {
             .value = 0.0f,

src/core/hid/motion_input.cpp

@@ -37,11 +37,17 @@ void MotionInput::SetGyroscope(const Common::Vec3f& gyroscope) {
     gyro.y = std::clamp(gyro.y, -GyroMaxValue, GyroMaxValue);
     gyro.z = std::clamp(gyro.z, -GyroMaxValue, GyroMaxValue);
 
-    // Auto adjust drift to minimize drift
+    // Auto adjust gyro_bias to minimize drift
     if (!IsMoving(IsAtRestRelaxed)) {
         gyro_bias = (gyro_bias * 0.9999f) + (gyroscope * 0.0001f);
     }
 
+    // Adjust drift when calibration mode is enabled
+    if (calibration_mode) {
+        gyro_bias = (gyro_bias * 0.99f) + (gyroscope * 0.01f);
+        StopCalibration();
+    }
+
     if (gyro.Length() < gyro_threshold * user_gyro_threshold) {
         gyro = {};
     } else {
@@ -107,6 +113,19 @@ void MotionInput::UpdateRotation(u64 elapsed_time) {
     rotations += gyro * sample_period;
 }
 
+void MotionInput::Calibrate() {
+    calibration_mode = true;
+    calibration_counter = 0;
+}
+
+void MotionInput::StopCalibration() {
+    if (calibration_counter++ > CalibrationSamples) {
+        calibration_mode = false;
+        ResetQuaternion();
+        ResetRotations();
+    }
+}
+
 // Based on Madgwick's implementation of Mayhony's AHRS algorithm.
 // https://github.com/xioTechnologies/Open-Source-AHRS-With-x-IMU/blob/master/x-IMU%20IMU%20and%20AHRS%20Algorithms/x-IMU%20IMU%20and%20AHRS%20Algorithms/AHRS/MahonyAHRS.cs
 void MotionInput::UpdateOrientation(u64 elapsed_time) {

src/core/hid/motion_input.h

@@ -23,6 +23,8 @@ public:
     static constexpr float GyroMaxValue = 5.0f;
     static constexpr float AccelMaxValue = 7.0f;
 
+    static constexpr std::size_t CalibrationSamples = 300;
+
     explicit MotionInput();
 
     MotionInput(const MotionInput&) = default;
@@ -49,6 +51,8 @@ public:
     void UpdateRotation(u64 elapsed_time);
     void UpdateOrientation(u64 elapsed_time);
 
+    void Calibrate();
+
     [[nodiscard]] std::array<Common::Vec3f, 3> GetOrientation() const;
     [[nodiscard]] Common::Vec3f GetAcceleration() const;
     [[nodiscard]] Common::Vec3f GetGyroscope() const;
@@ -61,6 +65,7 @@ public:
     [[nodiscard]] bool IsCalibrated(f32 sensitivity) const;
 
 private:
+    void StopCalibration();
     void ResetOrientation();
     void SetOrientationFromAccelerometer();
 
@@ -103,6 +108,12 @@ private:
 
     // Use accelerometer values to calculate position
     bool only_accelerometer = true;
+
+    // When enabled it will aggressively adjust for gyro drift
+    bool calibration_mode = false;
+
+    // Used to auto disable calibration mode
+    std::size_t calibration_counter = 0;
 };
 
 } // namespace Core::HID

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

@@ -423,8 +423,8 @@ void Controller_NPad::RequestPadStateUpdate(Core::HID::NpadIdType npad_id) {
         return;
     }
 
-    // This function is unique to yuzu for the turbo buttons to work properly
-    controller.device->TurboButtonUpdate();
+    // This function is unique to yuzu for the turbo buttons and motion to work properly
+    controller.device->StatusUpdate();
 
     auto& pad_entry = controller.npad_pad_state;
     auto& trigger_entry = controller.npad_trigger_state;

src/core/hle/service/nifm/nifm.cpp

@@ -218,7 +218,7 @@ public:
 
 private:
     void Submit(HLERequestContext& ctx) {
-        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        LOG_DEBUG(Service_NIFM, "(STUBBED) called");
 
         if (state == RequestState::NotSubmitted) {
             UpdateState(RequestState::OnHold);
@@ -229,7 +229,7 @@ private:
     }
 
     void GetRequestState(HLERequestContext& ctx) {
-        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        LOG_DEBUG(Service_NIFM, "(STUBBED) called");
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(ResultSuccess);
@@ -237,7 +237,7 @@ private:
     }
 
     void GetResult(HLERequestContext& ctx) {
-        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        LOG_DEBUG(Service_NIFM, "(STUBBED) called");
 
         const auto result = [this] {
             const auto has_connection = Network::GetHostIPv4Address().has_value();

src/core/internal_network/network.cpp

@@ -356,7 +356,7 @@ NetworkInstance::~NetworkInstance() {
 std::optional<IPv4Address> GetHostIPv4Address() {
     const auto network_interface = Network::GetSelectedNetworkInterface();
     if (!network_interface.has_value()) {
-        LOG_ERROR(Network, "GetSelectedNetworkInterface returned no interface");
+        LOG_DEBUG(Network, "GetSelectedNetworkInterface returned no interface");
         return {};
     }
 

src/core/internal_network/network_interface.cpp

@@ -200,7 +200,7 @@ std::optional<NetworkInterface> GetSelectedNetworkInterface() {
         });
 
     if (res == network_interfaces.end()) {
-        LOG_ERROR(Network, "Couldn't find selected interface \"{}\"", selected_network_interface);
+        LOG_DEBUG(Network, "Couldn't find selected interface \"{}\"", selected_network_interface);
         return std::nullopt;
     }
 

src/core/memory.cpp

@@ -13,10 +13,12 @@
 #include "common/swap.h"
 #include "core/core.h"
 #include "core/device_memory.h"
+#include "core/hardware_properties.h"
 #include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/memory.h"
 #include "video_core/gpu.h"
+#include "video_core/rasterizer_download_area.h"
 
 namespace Core::Memory {
 
@@ -243,7 +245,7 @@ struct Memory::Impl {
             [&](const Common::ProcessAddress current_vaddr, const std::size_t copy_amount,
                 const u8* const host_ptr) {
                 if constexpr (!UNSAFE) {
-                    system.GPU().FlushRegion(GetInteger(current_vaddr), copy_amount);
+                    HandleRasterizerDownload(GetInteger(current_vaddr), copy_amount);
                 }
                 std::memcpy(dest_buffer, host_ptr, copy_amount);
             },
@@ -334,7 +336,7 @@ struct Memory::Impl {
             },
             [&](const Common::ProcessAddress current_vaddr, const std::size_t copy_amount,
                 u8* const host_ptr) {
-                system.GPU().FlushRegion(GetInteger(current_vaddr), copy_amount);
+                HandleRasterizerDownload(GetInteger(current_vaddr), copy_amount);
                 WriteBlockImpl<false>(process, dest_addr, host_ptr, copy_amount);
             },
             [&](const std::size_t copy_amount) {
@@ -373,7 +375,7 @@ struct Memory::Impl {
                                  const std::size_t block_size) {
             // dc ivac: Invalidate to point of coherency
             // GPU flush -> CPU invalidate
-            system.GPU().FlushRegion(GetInteger(current_vaddr), block_size);
+            HandleRasterizerDownload(GetInteger(current_vaddr), block_size);
         };
         return PerformCacheOperation(process, dest_addr, size, on_rasterizer);
     }
@@ -462,7 +464,8 @@ struct Memory::Impl {
         }
 
         if (Settings::IsFastmemEnabled()) {
-            const bool is_read_enable = !Settings::IsGPULevelExtreme() || !cached;
+            const bool is_read_enable =
+                !Settings::values.use_reactive_flushing.GetValue() || !cached;
             system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
         }
 
@@ -651,7 +654,7 @@ struct Memory::Impl {
                 LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:016X}", sizeof(T) * 8,
                           GetInteger(vaddr));
             },
-            [&]() { system.GPU().FlushRegion(GetInteger(vaddr), sizeof(T)); });
+            [&]() { HandleRasterizerDownload(GetInteger(vaddr), sizeof(T)); });
         if (ptr) {
             std::memcpy(&result, ptr, sizeof(T));
         }
@@ -712,7 +715,19 @@ struct Memory::Impl {
         return true;
     }
 
+    void HandleRasterizerDownload(VAddr address, size_t size) {
+        const size_t core = system.GetCurrentHostThreadID();
+        auto& current_area = rasterizer_areas[core];
+        const VAddr end_address = address + size;
+        if (current_area.start_address <= address && end_address <= current_area.end_address)
+            [[likely]] {
+            return;
+        }
+        current_area = system.GPU().OnCPURead(address, size);
+    }
+
     Common::PageTable* current_page_table = nullptr;
+    std::array<VideoCore::RasterizerDownloadArea, Core::Hardware::NUM_CPU_CORES> rasterizer_areas{};
     Core::System& system;
 };
 

src/input_common/helpers/joycon_protocol/common_protocol.cpp

@@ -236,13 +236,13 @@ DriverResult JoyconCommonProtocol::GetMCUDataResponse(ReportMode report_mode,
     return DriverResult::Success;
 }
 
-DriverResult JoyconCommonProtocol::SendMCUData(ReportMode report_mode, SubCommand sc,
+DriverResult JoyconCommonProtocol::SendMCUData(ReportMode report_mode, MCUSubCommand sc,
                                                std::span<const u8> buffer,
                                                MCUCommandResponse& output) {
     SubCommandPacket packet{
         .output_report = OutputReport::MCU_DATA,
         .packet_counter = GetCounter(),
-        .sub_command = sc,
+        .mcu_sub_command = sc,
         .command_data = {},
     };
 
@@ -269,8 +269,7 @@ DriverResult JoyconCommonProtocol::WaitSetMCUMode(ReportMode report_mode, MCUMod
     std::size_t tries{};
 
     do {
-        const std::vector<u8> mcu_data{static_cast<u8>(MCUMode::Standby)};
-        const auto result = SendMCUData(report_mode, SubCommand::STATE, mcu_data, output);
+        const auto result = SendMCUData(report_mode, MCUSubCommand::SetDeviceMode, {}, output);
 
         if (result != DriverResult::Success) {
             return result;

src/input_common/helpers/joycon_protocol/common_protocol.h

@@ -156,7 +156,7 @@ public:
      * @param buffer data to be send
      * @returns output buffer containing the response
      */
-    DriverResult SendMCUData(ReportMode report_mode, SubCommand sc, std::span<const u8> buffer,
+    DriverResult SendMCUData(ReportMode report_mode, MCUSubCommand sc, std::span<const u8> buffer,
                              MCUCommandResponse& output);
 
     /**

src/input_common/helpers/joycon_protocol/joycon_types.h

@@ -575,7 +575,6 @@ struct NFCPollingCommandData {
 static_assert(sizeof(NFCPollingCommandData) == 0x05, "NFCPollingCommandData is an invalid size");
 
 struct NFCRequestState {
-    MCUSubCommand sub_command;
     NFCReadCommand command_argument;
     u8 packet_id;
     INSERT_PADDING_BYTES(0x1);
@@ -587,6 +586,7 @@ struct NFCRequestState {
         NFCPollingCommandData nfc_polling;
     };
     u8 crc;
+    INSERT_PADDING_BYTES(0x1);
 };
 static_assert(sizeof(NFCRequestState) == 0x26, "NFCRequestState is an invalid size");
 
@@ -659,7 +659,10 @@ struct SubCommandPacket {
     OutputReport output_report;
     u8 packet_counter;
     INSERT_PADDING_BYTES(0x8); // This contains vibration data
-    SubCommand sub_command;
+    union {
+        SubCommand sub_command;
+        MCUSubCommand mcu_sub_command;
+    };
     std::array<u8, 0x26> command_data;
 };
 static_assert(sizeof(SubCommandPacket) == 0x31, "SubCommandPacket is an invalid size");

src/input_common/helpers/joycon_protocol/nfc.cpp

@@ -278,7 +278,6 @@ DriverResult NfcProtocol::GetAmiiboData(std::vector<u8>& ntag_data) {
 
 DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output) {
     NFCRequestState request{
-        .sub_command = MCUSubCommand::ReadDeviceMode,
         .command_argument = NFCReadCommand::StartPolling,
         .packet_id = 0x0,
         .packet_flag = MCUPacketFlag::LastCommandPacket,
@@ -296,13 +295,13 @@ DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output) {
 
     std::array<u8, sizeof(NFCRequestState)> request_data{};
     memcpy(request_data.data(), &request, sizeof(NFCRequestState));
-    request_data[37] = CalculateMCU_CRC8(request_data.data() + 1, 36);
-    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, SubCommand::STATE, request_data, output);
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
 }
 
 DriverResult NfcProtocol::SendStopPollingRequest(MCUCommandResponse& output) {
     NFCRequestState request{
-        .sub_command = MCUSubCommand::ReadDeviceMode,
         .command_argument = NFCReadCommand::StopPolling,
         .packet_id = 0x0,
         .packet_flag = MCUPacketFlag::LastCommandPacket,
@@ -313,13 +312,13 @@ DriverResult NfcProtocol::SendStopPollingRequest(MCUCommandResponse& output) {
 
     std::array<u8, sizeof(NFCRequestState)> request_data{};
     memcpy(request_data.data(), &request, sizeof(NFCRequestState));
-    request_data[37] = CalculateMCU_CRC8(request_data.data() + 1, 36);
-    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, SubCommand::STATE, request_data, output);
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
 }
 
 DriverResult NfcProtocol::SendStartWaitingRecieveRequest(MCUCommandResponse& output) {
     NFCRequestState request{
-        .sub_command = MCUSubCommand::ReadDeviceMode,
         .command_argument = NFCReadCommand::StartWaitingRecieve,
         .packet_id = 0x0,
         .packet_flag = MCUPacketFlag::LastCommandPacket,
@@ -330,13 +329,13 @@ DriverResult NfcProtocol::SendStartWaitingRecieveRequest(MCUCommandResponse& out
 
     std::vector<u8> request_data(sizeof(NFCRequestState));
     memcpy(request_data.data(), &request, sizeof(NFCRequestState));
-    request_data[37] = CalculateMCU_CRC8(request_data.data() + 1, 36);
-    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, SubCommand::STATE, request_data, output);
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
 }
 
 DriverResult NfcProtocol::SendReadAmiiboRequest(MCUCommandResponse& output, NFCPages ntag_pages) {
     NFCRequestState request{
-        .sub_command = MCUSubCommand::ReadDeviceMode,
         .command_argument = NFCReadCommand::Ntag,
         .packet_id = 0x0,
         .packet_flag = MCUPacketFlag::LastCommandPacket,
@@ -355,8 +354,9 @@ DriverResult NfcProtocol::SendReadAmiiboRequest(MCUCommandResponse& output, NFCP
 
     std::array<u8, sizeof(NFCRequestState)> request_data{};
     memcpy(request_data.data(), &request, sizeof(NFCRequestState));
-    request_data[37] = CalculateMCU_CRC8(request_data.data() + 1, 36);
-    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, SubCommand::STATE, request_data, output);
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
 }
 
 NFCReadBlockCommand NfcProtocol::GetReadBlockCommand(NFCPages pages) const {

src/input_common/input_poller.cpp

@@ -667,7 +667,7 @@ public:
             .raw_value = input_engine->GetAxis(identifier, axis_z),
             .properties = properties_z,
         };
-        status.delta_timestamp = 5000;
+        status.delta_timestamp = 1000;
         status.force_update = true;
         return status;
     }

src/tests/video_core/memory_tracker.cpp

@@ -535,12 +535,12 @@ TEST_CASE("MemoryTracker: Cached write downloads") {
     memory_track->MarkRegionAsGpuModified(c + PAGE, PAGE);
     int num = 0;
     memory_track->ForEachDownloadRangeAndClear(c, WORD, [&](u64 offset, u64 size) { ++num; });
-    REQUIRE(num == 1);
+    REQUIRE(num == 0);
     num = 0;
     memory_track->ForEachUploadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
     REQUIRE(num == 0);
     REQUIRE(!memory_track->IsRegionCpuModified(c + PAGE, PAGE));
-    REQUIRE(!memory_track->IsRegionGpuModified(c + PAGE, PAGE));
+    REQUIRE(memory_track->IsRegionGpuModified(c + PAGE, PAGE));
     memory_track->FlushCachedWrites();
     REQUIRE(memory_track->IsRegionCpuModified(c + PAGE, PAGE));
     REQUIRE(!memory_track->IsRegionGpuModified(c + PAGE, PAGE));

src/video_core/buffer_cache/buffer_base.h

@@ -18,6 +18,7 @@ namespace VideoCommon {
 enum class BufferFlagBits {
     Picked = 1 << 0,
     CachedWrites = 1 << 1,
+    PreemtiveDownload = 1 << 2,
 };
 DECLARE_ENUM_FLAG_OPERATORS(BufferFlagBits)
 
@@ -54,6 +55,10 @@ public:
         flags |= BufferFlagBits::Picked;
     }
 
+    void MarkPreemtiveDownload() noexcept {
+        flags |= BufferFlagBits::PreemtiveDownload;
+    }
+
     /// Unmark buffer as picked
     void Unpick() noexcept {
         flags &= ~BufferFlagBits::Picked;
@@ -84,6 +89,10 @@ public:
         return True(flags & BufferFlagBits::CachedWrites);
     }
 
+    bool IsPreemtiveDownload() const noexcept {
+        return True(flags & BufferFlagBits::PreemtiveDownload);
+    }
+
     /// Returns the base CPU address of the buffer
     [[nodiscard]] VAddr CpuAddr() const noexcept {
         return cpu_addr;

src/video_core/buffer_cache/buffer_cache.h

@@ -23,8 +23,6 @@ BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
     common_ranges.clear();
     inline_buffer_id = NULL_BUFFER_ID;
 
-    active_async_buffers = !Settings::IsGPULevelHigh();
-
     if (!runtime.CanReportMemoryUsage()) {
         minimum_memory = DEFAULT_EXPECTED_MEMORY;
         critical_memory = DEFAULT_CRITICAL_MEMORY;
@@ -75,8 +73,6 @@ void BufferCache<P>::TickFrame() {
     uniform_cache_hits[0] = 0;
     uniform_cache_shots[0] = 0;
 
-    active_async_buffers = !Settings::IsGPULevelHigh();
-
     const bool skip_preferred = hits * 256 < shots * 251;
     uniform_buffer_skip_cache_size = skip_preferred ? DEFAULT_SKIP_CACHE_SIZE : 0;
 
@@ -111,9 +107,25 @@ void BufferCache<P>::WriteMemory(VAddr cpu_addr, u64 size) {
 template <class P>
 void BufferCache<P>::CachedWriteMemory(VAddr cpu_addr, u64 size) {
     memory_tracker.CachedCpuWrite(cpu_addr, size);
-    const IntervalType add_interval{Common::AlignDown(cpu_addr, YUZU_PAGESIZE),
-                                    Common::AlignUp(cpu_addr + size, YUZU_PAGESIZE)};
-    cached_ranges.add(add_interval);
+}
+
+template <class P>
+std::optional<VideoCore::RasterizerDownloadArea> BufferCache<P>::GetFlushArea(VAddr cpu_addr,
+                                                                              u64 size) {
+    std::optional<VideoCore::RasterizerDownloadArea> area{};
+    area.emplace();
+    VAddr cpu_addr_start_aligned = Common::AlignDown(cpu_addr, Core::Memory::YUZU_PAGESIZE);
+    VAddr cpu_addr_end_aligned = Common::AlignUp(cpu_addr + size, Core::Memory::YUZU_PAGESIZE);
+    area->start_address = cpu_addr_start_aligned;
+    area->end_address = cpu_addr_end_aligned;
+    if (memory_tracker.IsRegionPreflushable(cpu_addr, size)) {
+        area->preemtive = true;
+        return area;
+    };
+    memory_tracker.MarkRegionAsPreflushable(cpu_addr_start_aligned,
+                                            cpu_addr_end_aligned - cpu_addr_start_aligned);
+    area->preemtive = !IsRegionGpuModified(cpu_addr, size);
+    return area;
 }
 
 template <class P>
@@ -205,7 +217,7 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
     if (has_new_downloads) {
         memory_tracker.MarkRegionAsGpuModified(*cpu_dest_address, amount);
     }
-    std::vector<u8> tmp_buffer(amount);
+    tmp_buffer.resize(amount);
     cpu_memory.ReadBlockUnsafe(*cpu_src_address, tmp_buffer.data(), amount);
     cpu_memory.WriteBlockUnsafe(*cpu_dest_address, tmp_buffer.data(), amount);
     return true;
@@ -441,9 +453,7 @@ void BufferCache<P>::BindComputeTextureBuffer(size_t tbo_index, GPUVAddr gpu_add
 
 template <class P>
 void BufferCache<P>::FlushCachedWrites() {
-    cached_write_buffer_ids.clear();
     memory_tracker.FlushCachedWrites();
-    cached_ranges.clear();
 }
 
 template <class P>
@@ -474,9 +484,8 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
 
     if (committed_ranges.empty()) {
         if constexpr (IMPLEMENTS_ASYNC_DOWNLOADS) {
-            if (active_async_buffers) {
-                async_buffers.emplace_back(std::optional<Async_Buffer>{});
-            }
+
+            async_buffers.emplace_back(std::optional<Async_Buffer>{});
         }
         return;
     }
@@ -537,64 +546,65 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
     committed_ranges.clear();
     if (downloads.empty()) {
         if constexpr (IMPLEMENTS_ASYNC_DOWNLOADS) {
-            if (active_async_buffers) {
-                async_buffers.emplace_back(std::optional<Async_Buffer>{});
-            }
+
+            async_buffers.emplace_back(std::optional<Async_Buffer>{});
         }
         return;
     }
-    if (active_async_buffers) {
-        if constexpr (IMPLEMENTS_ASYNC_DOWNLOADS) {
-            auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes, true);
-            boost::container::small_vector<BufferCopy, 4> normalized_copies;
-            IntervalSet new_async_range{};
-            runtime.PreCopyBarrier();
-            for (auto& [copy, buffer_id] : downloads) {
-                copy.dst_offset += download_staging.offset;
-                const std::array copies{copy};
-                BufferCopy second_copy{copy};
-                Buffer& buffer = slot_buffers[buffer_id];
-                second_copy.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
-                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);
-                runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
-                normalized_copies.push_back(second_copy);
-            }
-            runtime.PostCopyBarrier();
-            pending_downloads.emplace_back(std::move(normalized_copies));
-            async_buffers.emplace_back(download_staging);
-        } else {
+    if constexpr (IMPLEMENTS_ASYNC_DOWNLOADS) {
+        auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes, true);
+        boost::container::small_vector<BufferCopy, 4> normalized_copies;
+        IntervalSet new_async_range{};
+        runtime.PreCopyBarrier();
+        for (auto& [copy, buffer_id] : downloads) {
+            copy.dst_offset += download_staging.offset;
+            const std::array copies{copy};
+            BufferCopy second_copy{copy};
+            Buffer& buffer = slot_buffers[buffer_id];
+            second_copy.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
+            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);
+            runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
+            normalized_copies.push_back(second_copy);
+        }
+        runtime.PostCopyBarrier();
+        pending_downloads.emplace_back(std::move(normalized_copies));
+        async_buffers.emplace_back(download_staging);
+    } else {
+        if (!Settings::IsGPULevelHigh()) {
             committed_ranges.clear();
             uncommitted_ranges.clear();
-        }
-    } else {
-        if constexpr (USE_MEMORY_MAPS) {
-            auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
-            runtime.PreCopyBarrier();
-            for (auto& [copy, buffer_id] : downloads) {
-                // 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, false);
-            }
-            runtime.PostCopyBarrier();
-            runtime.Finish();
-            for (const auto& [copy, buffer_id] : downloads) {
-                const Buffer& buffer = slot_buffers[buffer_id];
-                const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
-                // Undo the modified offset
-                const u64 dst_offset = copy.dst_offset - download_staging.offset;
-                const u8* read_mapped_memory = download_staging.mapped_span.data() + dst_offset;
-                cpu_memory.WriteBlockUnsafe(cpu_addr, read_mapped_memory, copy.size);
-            }
         } else {
-            const std::span<u8> immediate_buffer = ImmediateBuffer(largest_copy);
-            for (const auto& [copy, buffer_id] : downloads) {
-                Buffer& buffer = slot_buffers[buffer_id];
-                buffer.ImmediateDownload(copy.src_offset, immediate_buffer.subspan(0, copy.size));
-                const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
-                cpu_memory.WriteBlockUnsafe(cpu_addr, immediate_buffer.data(), copy.size);
+            if constexpr (USE_MEMORY_MAPS) {
+                auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
+                runtime.PreCopyBarrier();
+                for (auto& [copy, buffer_id] : downloads) {
+                    // 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,
+                                       false);
+                }
+                runtime.PostCopyBarrier();
+                runtime.Finish();
+                for (const auto& [copy, buffer_id] : downloads) {
+                    const Buffer& buffer = slot_buffers[buffer_id];
+                    const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
+                    // Undo the modified offset
+                    const u64 dst_offset = copy.dst_offset - download_staging.offset;
+                    const u8* read_mapped_memory = download_staging.mapped_span.data() + dst_offset;
+                    cpu_memory.WriteBlockUnsafe(cpu_addr, read_mapped_memory, copy.size);
+                }
+            } else {
+                const std::span<u8> immediate_buffer = ImmediateBuffer(largest_copy);
+                for (const auto& [copy, buffer_id] : downloads) {
+                    Buffer& buffer = slot_buffers[buffer_id];
+                    buffer.ImmediateDownload(copy.src_offset,
+                                             immediate_buffer.subspan(0, copy.size));
+                    const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
+                    cpu_memory.WriteBlockUnsafe(cpu_addr, immediate_buffer.data(), copy.size);
+                }
             }
         }
     }
@@ -1629,7 +1639,6 @@ void BufferCache<P>::DeleteBuffer(BufferId buffer_id, bool do_not_mark) {
     replace(transform_feedback_buffers);
     replace(compute_uniform_buffers);
     replace(compute_storage_buffers);
-    std::erase(cached_write_buffer_ids, buffer_id);
 
     // Mark the whole buffer as CPU written to stop tracking CPU writes
     if (!do_not_mark) {

src/video_core/buffer_cache/buffer_cache_base.h

@@ -188,6 +188,8 @@ public:
 
     void DownloadMemory(VAddr cpu_addr, u64 size);
 
+    std::optional<VideoCore::RasterizerDownloadArea> GetFlushArea(VAddr cpu_addr, u64 size);
+
     bool InlineMemory(VAddr dest_address, size_t copy_size, std::span<const u8> inlined_buffer);
 
     void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size);
@@ -541,8 +543,6 @@ private:
                        std::array<std::array<u32, NUM_GRAPHICS_UNIFORM_BUFFERS>, NUM_STAGES>, Empty>
         uniform_buffer_binding_sizes{};
 
-    std::vector<BufferId> cached_write_buffer_ids;
-
     MemoryTracker memory_tracker;
     IntervalSet uncommitted_ranges;
     IntervalSet common_ranges;
@@ -572,9 +572,8 @@ private:
     u64 critical_memory = 0;
     BufferId inline_buffer_id;
 
-    bool active_async_buffers = false;
-
     std::array<BufferId, ((1ULL << 39) >> CACHING_PAGEBITS)> page_table;
+    std::vector<u8> tmp_buffer;
 };
 
 } // namespace VideoCommon

src/video_core/buffer_cache/memory_tracker_base.h

@@ -66,6 +66,14 @@ public:
             });
     }
 
+    /// Returns true if a region has been marked as Preflushable
+    [[nodiscard]] bool IsRegionPreflushable(VAddr query_cpu_addr, u64 query_size) noexcept {
+        return IteratePages<false>(
+            query_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) {
+                return manager->template IsRegionModified<Type::Preflushable>(offset, size);
+            });
+    }
+
     /// Mark region as CPU modified, notifying the rasterizer about this change
     void MarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 query_size) {
         IteratePages<true>(dirty_cpu_addr, query_size,
@@ -93,6 +101,15 @@ public:
                            });
     }
 
+    /// Mark region as modified from the host GPU
+    void MarkRegionAsPreflushable(VAddr dirty_cpu_addr, u64 query_size) noexcept {
+        IteratePages<true>(dirty_cpu_addr, query_size,
+                           [](Manager* manager, u64 offset, size_t size) {
+                               manager->template ChangeRegionState<Type::Preflushable, true>(
+                                   manager->GetCpuAddr() + offset, size);
+                           });
+    }
+
     /// Unmark region as modified from the host GPU
     void UnmarkRegionAsGpuModified(VAddr dirty_cpu_addr, u64 query_size) noexcept {
         IteratePages<true>(dirty_cpu_addr, query_size,
@@ -102,6 +119,15 @@ public:
                            });
     }
 
+    /// Unmark region as modified from the host GPU
+    void UnmarkRegionAsPreflushable(VAddr dirty_cpu_addr, u64 query_size) noexcept {
+        IteratePages<true>(dirty_cpu_addr, query_size,
+                           [](Manager* manager, u64 offset, size_t size) {
+                               manager->template ChangeRegionState<Type::Preflushable, false>(
+                                   manager->GetCpuAddr() + offset, size);
+                           });
+    }
+
     /// Mark region as modified from the CPU
     /// but don't mark it as modified until FlusHCachedWrites is called.
     void CachedCpuWrite(VAddr dirty_cpu_addr, u64 query_size) {

src/video_core/buffer_cache/word_manager.h

@@ -26,6 +26,7 @@ enum class Type {
     GPU,
     CachedCPU,
     Untracked,
+    Preflushable,
 };
 
 /// Vector tracking modified pages tightly packed with small vector optimization
@@ -55,17 +56,20 @@ struct Words {
             gpu.stack.fill(0);
             cached_cpu.stack.fill(0);
             untracked.stack.fill(~u64{0});
+            preflushable.stack.fill(0);
         } else {
             // Share allocation between CPU and GPU pages and set their default values
-            u64* const alloc = new u64[num_words * 4];
+            u64* const alloc = new u64[num_words * 5];
             cpu.heap = alloc;
             gpu.heap = alloc + num_words;
             cached_cpu.heap = alloc + num_words * 2;
             untracked.heap = alloc + num_words * 3;
+            preflushable.heap = alloc + num_words * 4;
             std::fill_n(cpu.heap, num_words, ~u64{0});
             std::fill_n(gpu.heap, num_words, 0);
             std::fill_n(cached_cpu.heap, num_words, 0);
             std::fill_n(untracked.heap, num_words, ~u64{0});
+            std::fill_n(preflushable.heap, num_words, 0);
         }
         // Clean up tailing bits
         const u64 last_word_size = size_bytes % BYTES_PER_WORD;
@@ -88,13 +92,14 @@ struct Words {
         gpu = rhs.gpu;
         cached_cpu = rhs.cached_cpu;
         untracked = rhs.untracked;
+        preflushable = rhs.preflushable;
         rhs.cpu.heap = nullptr;
         return *this;
     }
 
     Words(Words&& rhs) noexcept
         : size_bytes{rhs.size_bytes}, num_words{rhs.num_words}, cpu{rhs.cpu}, gpu{rhs.gpu},
-          cached_cpu{rhs.cached_cpu}, untracked{rhs.untracked} {
+          cached_cpu{rhs.cached_cpu}, untracked{rhs.untracked}, preflushable{rhs.preflushable} {
         rhs.cpu.heap = nullptr;
     }
 
@@ -129,6 +134,8 @@ struct Words {
             return std::span<u64>(cached_cpu.Pointer(IsShort()), num_words);
         } else if constexpr (type == Type::Untracked) {
             return std::span<u64>(untracked.Pointer(IsShort()), num_words);
+        } else if constexpr (type == Type::Preflushable) {
+            return std::span<u64>(preflushable.Pointer(IsShort()), num_words);
         }
     }
 
@@ -142,6 +149,8 @@ struct Words {
             return std::span<const u64>(cached_cpu.Pointer(IsShort()), num_words);
         } else if constexpr (type == Type::Untracked) {
             return std::span<const u64>(untracked.Pointer(IsShort()), num_words);
+        } else if constexpr (type == Type::Preflushable) {
+            return std::span<const u64>(preflushable.Pointer(IsShort()), num_words);
         }
     }
 
@@ -151,6 +160,7 @@ struct Words {
     WordsArray<stack_words> gpu;
     WordsArray<stack_words> cached_cpu;
     WordsArray<stack_words> untracked;
+    WordsArray<stack_words> preflushable;
 };
 
 template <class RasterizerInterface, size_t stack_words = 1>
@@ -292,6 +302,9 @@ public:
                  (pending_pointer - pending_offset) * BYTES_PER_PAGE);
         };
         IterateWords(offset, size, [&](size_t index, u64 mask) {
+            if constexpr (type == Type::GPU) {
+                mask &= ~untracked_words[index];
+            }
             const u64 word = state_words[index] & mask;
             if constexpr (clear) {
                 if constexpr (type == Type::CPU || type == Type::CachedCPU) {
@@ -340,8 +353,13 @@ public:
         static_assert(type != Type::Untracked);
 
         const std::span<const u64> state_words = words.template Span<type>();
+        [[maybe_unused]] const std::span<const u64> untracked_words =
+            words.template Span<Type::Untracked>();
         bool result = false;
         IterateWords(offset, size, [&](size_t index, u64 mask) {
+            if constexpr (type == Type::GPU) {
+                mask &= ~untracked_words[index];
+            }
             const u64 word = state_words[index] & mask;
             if (word != 0) {
                 result = true;
@@ -362,9 +380,14 @@ public:
     [[nodiscard]] std::pair<u64, u64> ModifiedRegion(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
         const std::span<const u64> state_words = words.template Span<type>();
+        [[maybe_unused]] const std::span<const u64> untracked_words =
+            words.template Span<Type::Untracked>();
         u64 begin = std::numeric_limits<u64>::max();
         u64 end = 0;
         IterateWords(offset, size, [&](size_t index, u64 mask) {
+            if constexpr (type == Type::GPU) {
+                mask &= ~untracked_words[index];
+            }
             const u64 word = state_words[index] & mask;
             if (word == 0) {
                 return;

src/video_core/engines/maxwell_dma.cpp

@@ -223,7 +223,7 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
     write_buffer.resize_destructive(dst_size);
 
     memory_manager.ReadBlock(src_operand.address, read_buffer.data(), src_size);
-    memory_manager.ReadBlockUnsafe(dst_operand.address, write_buffer.data(), dst_size);
+    memory_manager.ReadBlock(dst_operand.address, write_buffer.data(), dst_size);
 
     UnswizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, width, height, depth, x_offset,
                      src_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
@@ -288,11 +288,7 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
     write_buffer.resize_destructive(dst_size);
 
     memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
-    if (Settings::IsGPULevelExtreme()) {
-        memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
-    } else {
-        memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size);
-    }
+    memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size);
 
     // If the input is linear and the output is tiled, swizzle the input and copy it over.
     SwizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, width, height, depth, x_offset,

src/video_core/fence_manager.h

@@ -59,6 +59,11 @@ public:
         buffer_cache.AccumulateFlushes();
     }
 
+    void SignalReference() {
+        std::function<void()> do_nothing([] {});
+        SignalFence(std::move(do_nothing));
+    }
+
     void SyncOperation(std::function<void()>&& func) {
         uncommitted_operations.emplace_back(std::move(func));
     }

src/video_core/gpu.cpp

@@ -283,6 +283,21 @@ struct GPU::Impl {
         gpu_thread.FlushRegion(addr, size);
     }
 
+    VideoCore::RasterizerDownloadArea OnCPURead(VAddr addr, u64 size) {
+        auto raster_area = rasterizer->GetFlushArea(addr, size);
+        if (raster_area.preemtive) {
+            return raster_area;
+        }
+        raster_area.preemtive = true;
+        const u64 fence = RequestSyncOperation([this, &raster_area]() {
+            rasterizer->FlushRegion(raster_area.start_address,
+                                    raster_area.end_address - raster_area.start_address);
+        });
+        gpu_thread.TickGPU();
+        WaitForSyncOperation(fence);
+        return raster_area;
+    }
+
     /// Notify rasterizer that any caches of the specified region should be invalidated
     void InvalidateRegion(VAddr addr, u64 size) {
         gpu_thread.InvalidateRegion(addr, size);
@@ -538,6 +553,10 @@ void GPU::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     impl->SwapBuffers(framebuffer);
 }
 
+VideoCore::RasterizerDownloadArea GPU::OnCPURead(VAddr addr, u64 size) {
+    return impl->OnCPURead(addr, size);
+}
+
 void GPU::FlushRegion(VAddr addr, u64 size) {
     impl->FlushRegion(addr, size);
 }

src/video_core/gpu.h

@@ -10,6 +10,7 @@
 #include "core/hle/service/nvdrv/nvdata.h"
 #include "video_core/cdma_pusher.h"
 #include "video_core/framebuffer_config.h"
+#include "video_core/rasterizer_download_area.h"
 
 namespace Core {
 class System;
@@ -240,6 +241,9 @@ public:
     /// Swap buffers (render frame)
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
 
+    /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
+    [[nodiscard]] VideoCore::RasterizerDownloadArea OnCPURead(VAddr addr, u64 size);
+
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     void FlushRegion(VAddr addr, u64 size);
 

src/video_core/query_cache.h

@@ -255,7 +255,6 @@ private:
                 if (!in_range(query)) {
                     continue;
                 }
-                rasterizer.UpdatePagesCachedCount(query.GetCpuAddr(), query.SizeInBytes(), -1);
                 AsyncJobId async_job_id = query.GetAsyncJob();
                 auto flush_result = query.Flush(async);
                 if (async_job_id == NULL_ASYNC_JOB_ID) {
@@ -273,7 +272,6 @@ private:
 
     /// Registers the passed parameters as cached and returns a pointer to the stored cached query.
     CachedQuery* Register(VideoCore::QueryType type, VAddr cpu_addr, u8* host_ptr, bool timestamp) {
-        rasterizer.UpdatePagesCachedCount(cpu_addr, CachedQuery::SizeInBytes(timestamp), 1);
         const u64 page = static_cast<u64>(cpu_addr) >> YUZU_PAGEBITS;
         return &cached_queries[page].emplace_back(static_cast<QueryCache&>(*this), type, cpu_addr,
                                                   host_ptr);

src/video_core/rasterizer_download_area.h

@@ -0,0 +1,16 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace VideoCore {
+
+struct RasterizerDownloadArea {
+    VAddr start_address;
+    VAddr end_address;
+    bool preemtive;
+};
+
+} // namespace VideoCore

src/video_core/rasterizer_interface.h

@@ -12,6 +12,7 @@
 #include "video_core/cache_types.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/gpu.h"
+#include "video_core/rasterizer_download_area.h"
 
 namespace Tegra {
 class MemoryManager;
@@ -95,6 +96,8 @@ public:
     virtual bool MustFlushRegion(VAddr addr, u64 size,
                                  VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
 
+    virtual RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) = 0;
+
     /// Notify rasterizer that any caches of the specified region should be invalidated
     virtual void InvalidateRegion(VAddr addr, u64 size,
                                   VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;

src/video_core/renderer_null/null_rasterizer.cpp

@@ -1,6 +1,8 @@
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/alignment.h"
+#include "core/memory.h"
 #include "video_core/host1x/host1x.h"
 #include "video_core/memory_manager.h"
 #include "video_core/renderer_null/null_rasterizer.h"
@@ -46,6 +48,14 @@ bool RasterizerNull::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheTyp
 }
 void RasterizerNull::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType) {}
 void RasterizerNull::OnCPUWrite(VAddr addr, u64 size) {}
+VideoCore::RasterizerDownloadArea RasterizerNull::GetFlushArea(VAddr addr, u64 size) {
+    VideoCore::RasterizerDownloadArea new_area{
+        .start_address = Common::AlignDown(addr, Core::Memory::YUZU_PAGESIZE),
+        .end_address = Common::AlignUp(addr + size, Core::Memory::YUZU_PAGESIZE),
+        .preemtive = true,
+    };
+    return new_area;
+}
 void RasterizerNull::InvalidateGPUCache() {}
 void RasterizerNull::UnmapMemory(VAddr addr, u64 size) {}
 void RasterizerNull::ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) {}

src/video_core/renderer_null/null_rasterizer.h

@@ -54,6 +54,7 @@ public:
     void InvalidateRegion(VAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void OnCPUWrite(VAddr addr, u64 size) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
     void InvalidateGPUCache() override;
     void UnmapMemory(VAddr addr, u64 size) override;
     void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) override;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -433,6 +433,29 @@ bool RasterizerOpenGL::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheT
     return false;
 }
 
+VideoCore::RasterizerDownloadArea RasterizerOpenGL::GetFlushArea(VAddr addr, u64 size) {
+    {
+        std::scoped_lock lock{texture_cache.mutex};
+        auto area = texture_cache.GetFlushArea(addr, size);
+        if (area) {
+            return *area;
+        }
+    }
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        auto area = buffer_cache.GetFlushArea(addr, size);
+        if (area) {
+            return *area;
+        }
+    }
+    VideoCore::RasterizerDownloadArea new_area{
+        .start_address = Common::AlignDown(addr, Core::Memory::YUZU_PAGESIZE),
+        .end_address = Common::AlignUp(addr + size, Core::Memory::YUZU_PAGESIZE),
+        .preemtive = true,
+    };
+    return new_area;
+}
+
 void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     if (addr == 0 || size == 0) {
@@ -1281,7 +1304,7 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
                                        const Tegra::DMA::BufferOperand& buffer_operand,
                                        const Tegra::DMA::ImageOperand& image_operand) {
     std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
-    const auto image_id = texture_cache.DmaImageId(image_operand);
+    const auto image_id = texture_cache.DmaImageId(image_operand, IS_IMAGE_UPLOAD);
     if (image_id == VideoCommon::NULL_IMAGE_ID) {
         return false;
     }

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -95,6 +95,7 @@ public:
                      VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     bool MustFlushRegion(VAddr addr, u64 size,
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void OnCPUWrite(VAddr addr, u64 size) override;

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -231,7 +231,7 @@ void ApplySwizzle(GLuint handle, PixelFormat format, std::array<SwizzleSource, 4
 
 [[nodiscard]] bool CanBeAccelerated(const TextureCacheRuntime& runtime,
                                     const VideoCommon::ImageInfo& info) {
-    if (IsPixelFormatASTC(info.format) && !runtime.HasNativeASTC()) {
+    if (IsPixelFormatASTC(info.format) && info.size.depth == 1 && !runtime.HasNativeASTC()) {
         return Settings::values.accelerate_astc.GetValue() &&
                !Settings::values.async_astc.GetValue();
     }

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -349,25 +349,12 @@ void RasterizerVulkan::Clear(u32 layer_count) {
 
     const u32 color_attachment = regs.clear_surface.RT;
     if (use_color && framebuffer->HasAspectColorBit(color_attachment)) {
-        VkClearValue clear_value;
-        bool is_integer = false;
-        bool is_signed = false;
-        size_t int_size = 8;
-        for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; ++i) {
-            const auto& this_rt = regs.rt[i];
-            if (this_rt.Address() == 0) {
-                continue;
-            }
-            if (this_rt.format == Tegra::RenderTargetFormat::NONE) {
-                continue;
-            }
-            const auto format =
-                VideoCore::Surface::PixelFormatFromRenderTargetFormat(this_rt.format);
-            is_integer = IsPixelFormatInteger(format);
-            is_signed = IsPixelFormatSignedInteger(format);
-            int_size = PixelComponentSizeBitsInteger(format);
-            break;
-        }
+        const auto format =
+            VideoCore::Surface::PixelFormatFromRenderTargetFormat(regs.rt[color_attachment].format);
+        bool is_integer = IsPixelFormatInteger(format);
+        bool is_signed = IsPixelFormatSignedInteger(format);
+        size_t int_size = PixelComponentSizeBitsInteger(format);
+        VkClearValue clear_value{};
         if (!is_integer) {
             std::memcpy(clear_value.color.float32, regs.clear_color.data(),
                         regs.clear_color.size() * sizeof(f32));
@@ -502,6 +489,22 @@ bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheT
     return false;
 }
 
+VideoCore::RasterizerDownloadArea RasterizerVulkan::GetFlushArea(VAddr addr, u64 size) {
+    {
+        std::scoped_lock lock{texture_cache.mutex};
+        auto area = texture_cache.GetFlushArea(addr, size);
+        if (area) {
+            return *area;
+        }
+    }
+    VideoCore::RasterizerDownloadArea new_area{
+        .start_address = Common::AlignDown(addr, Core::Memory::YUZU_PAGESIZE),
+        .end_address = Common::AlignUp(addr + size, Core::Memory::YUZU_PAGESIZE),
+        .preemtive = true,
+    };
+    return new_area;
+}
+
 void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
     if (addr == 0 || size == 0) {
         return;
@@ -598,7 +601,7 @@ void RasterizerVulkan::SignalSyncPoint(u32 value) {
 }
 
 void RasterizerVulkan::SignalReference() {
-    fence_manager.SignalOrdering();
+    fence_manager.SignalReference();
 }
 
 void RasterizerVulkan::ReleaseFences() {
@@ -631,7 +634,7 @@ void RasterizerVulkan::WaitForIdle() {
         cmdbuf.SetEvent(event, flags);
         cmdbuf.WaitEvents(event, flags, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, {}, {}, {});
     });
-    SignalReference();
+    fence_manager.SignalOrdering();
 }
 
 void RasterizerVulkan::FragmentBarrier() {
@@ -777,7 +780,7 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
                                        const Tegra::DMA::BufferOperand& buffer_operand,
                                        const Tegra::DMA::ImageOperand& image_operand) {
     std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
-    const auto image_id = texture_cache.DmaImageId(image_operand);
+    const auto image_id = texture_cache.DmaImageId(image_operand, IS_IMAGE_UPLOAD);
     if (image_id == VideoCommon::NULL_IMAGE_ID) {
         return false;
     }

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -92,6 +92,7 @@ public:
                      VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     bool MustFlushRegion(VAddr addr, u64 size,
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) override;

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -1268,7 +1268,7 @@ Image::Image(TextureCacheRuntime& runtime_, const ImageInfo& info_, GPUVAddr gpu
     if (IsPixelFormatASTC(info.format) && !runtime->device.IsOptimalAstcSupported()) {
         if (Settings::values.async_astc.GetValue()) {
             flags |= VideoCommon::ImageFlagBits::AsynchronousDecode;
-        } else if (Settings::values.accelerate_astc.GetValue()) {
+        } else if (Settings::values.accelerate_astc.GetValue() && info.size.depth == 1) {
             flags |= VideoCommon::ImageFlagBits::AcceleratedUpload;
         }
         flags |= VideoCommon::ImageFlagBits::Converted;

src/video_core/texture_cache/image_info.cpp

@@ -4,6 +4,7 @@
 #include <fmt/format.h>
 
 #include "common/assert.h"
+#include "common/settings.h"
 #include "video_core/surface.h"
 #include "video_core/texture_cache/format_lookup_table.h"
 #include "video_core/texture_cache/image_info.h"
@@ -22,6 +23,8 @@ using VideoCore::Surface::PixelFormat;
 using VideoCore::Surface::SurfaceType;
 
 ImageInfo::ImageInfo(const TICEntry& config) noexcept {
+    forced_flushed = config.IsPitchLinear() && !Settings::values.use_reactive_flushing.GetValue();
+    dma_downloaded = forced_flushed;
     format = PixelFormatFromTextureInfo(config.format, config.r_type, config.g_type, config.b_type,
                                         config.a_type, config.srgb_conversion);
     num_samples = NumSamples(config.msaa_mode);
@@ -117,6 +120,9 @@ ImageInfo::ImageInfo(const TICEntry& config) noexcept {
 
 ImageInfo::ImageInfo(const Maxwell3D::Regs::RenderTargetConfig& ct,
                      Tegra::Texture::MsaaMode msaa_mode) noexcept {
+    forced_flushed =
+        ct.tile_mode.is_pitch_linear && !Settings::values.use_reactive_flushing.GetValue();
+    dma_downloaded = forced_flushed;
     format = VideoCore::Surface::PixelFormatFromRenderTargetFormat(ct.format);
     rescaleable = false;
     if (ct.tile_mode.is_pitch_linear) {
@@ -155,6 +161,9 @@ ImageInfo::ImageInfo(const Maxwell3D::Regs::RenderTargetConfig& ct,
 
 ImageInfo::ImageInfo(const Maxwell3D::Regs::Zeta& zt, const Maxwell3D::Regs::ZetaSize& zt_size,
                      Tegra::Texture::MsaaMode msaa_mode) noexcept {
+    forced_flushed =
+        zt.tile_mode.is_pitch_linear && !Settings::values.use_reactive_flushing.GetValue();
+    dma_downloaded = forced_flushed;
     format = VideoCore::Surface::PixelFormatFromDepthFormat(zt.format);
     size.width = zt_size.width;
     size.height = zt_size.height;
@@ -195,6 +204,9 @@ ImageInfo::ImageInfo(const Maxwell3D::Regs::Zeta& zt, const Maxwell3D::Regs::Zet
 
 ImageInfo::ImageInfo(const Fermi2D::Surface& config) noexcept {
     UNIMPLEMENTED_IF_MSG(config.layer != 0, "Surface layer is not zero");
+    forced_flushed = config.linear == Fermi2D::MemoryLayout::Pitch &&
+                     !Settings::values.use_reactive_flushing.GetValue();
+    dma_downloaded = forced_flushed;
     format = VideoCore::Surface::PixelFormatFromRenderTargetFormat(config.format);
     rescaleable = false;
     if (config.linear == Fermi2D::MemoryLayout::Pitch) {

src/video_core/texture_cache/image_info.h

@@ -39,6 +39,8 @@ struct ImageInfo {
     u32 tile_width_spacing = 0;
     bool rescaleable = false;
     bool downscaleable = false;
+    bool forced_flushed = false;
+    bool dma_downloaded = false;
 };
 
 } // namespace VideoCommon

src/video_core/texture_cache/image_view_base.cpp

@@ -4,7 +4,6 @@
 #include <algorithm>
 
 #include "common/assert.h"
-#include "common/settings.h"
 #include "video_core/compatible_formats.h"
 #include "video_core/surface.h"
 #include "video_core/texture_cache/formatter.h"
@@ -26,8 +25,7 @@ ImageViewBase::ImageViewBase(const ImageViewInfo& info, const ImageInfo& image_i
     ASSERT_MSG(VideoCore::Surface::IsViewCompatible(image_info.format, info.format, false, true),
                "Image view format {} is incompatible with image format {}", info.format,
                image_info.format);
-    const bool is_async = Settings::values.use_asynchronous_gpu_emulation.GetValue();
-    if (image_info.type == ImageType::Linear && is_async) {
+    if (image_info.forced_flushed) {
         flags |= ImageViewFlagBits::PreemtiveDownload;
     }
     if (image_info.type == ImageType::e3D && info.type != ImageViewType::e3D) {

src/video_core/texture_cache/texture_cache.h

@@ -490,6 +490,32 @@ void TextureCache<P>::DownloadMemory(VAddr cpu_addr, size_t size) {
     }
 }
 
+template <class P>
+std::optional<VideoCore::RasterizerDownloadArea> TextureCache<P>::GetFlushArea(VAddr cpu_addr,
+                                                                               u64 size) {
+    std::optional<VideoCore::RasterizerDownloadArea> area{};
+    ForEachImageInRegion(cpu_addr, size, [&](ImageId, ImageBase& image) {
+        if (False(image.flags & ImageFlagBits::GpuModified)) {
+            return;
+        }
+        if (!area) {
+            area.emplace();
+            area->start_address = cpu_addr;
+            area->end_address = cpu_addr + size;
+            area->preemtive = true;
+        }
+        area->start_address = std::min(area->start_address, image.cpu_addr);
+        area->end_address = std::max(area->end_address, image.cpu_addr_end);
+        for (auto image_view_id : image.image_view_ids) {
+            auto& image_view = slot_image_views[image_view_id];
+            image_view.flags |= ImageViewFlagBits::PreemtiveDownload;
+        }
+        area->preemtive &= image.info.forced_flushed;
+        image.info.forced_flushed = true;
+    });
+    return area;
+}
+
 template <class P>
 void TextureCache<P>::UnmapMemory(VAddr cpu_addr, size_t size) {
     std::vector<ImageId> deleted_images;
@@ -683,6 +709,7 @@ void TextureCache<P>::CommitAsyncFlushes() {
                 download_info.async_buffer_id = last_async_buffer_id;
             }
         }
+
         if (any_none_dma) {
             auto download_map = runtime.DownloadStagingBuffer(total_size_bytes, true);
             for (const PendingDownload& download_info : download_ids) {
@@ -695,6 +722,7 @@ void TextureCache<P>::CommitAsyncFlushes() {
             }
             uncommitted_async_buffers.emplace_back(download_map);
         }
+
         async_buffers.emplace_back(std::move(uncommitted_async_buffers));
         uncommitted_async_buffers.clear();
     }
@@ -783,17 +811,22 @@ void TextureCache<P>::PopAsyncFlushes() {
 }
 
 template <class P>
-ImageId TextureCache<P>::DmaImageId(const Tegra::DMA::ImageOperand& operand) {
+ImageId TextureCache<P>::DmaImageId(const Tegra::DMA::ImageOperand& operand, bool is_upload) {
     const ImageInfo dst_info(operand);
     const ImageId dst_id = FindDMAImage(dst_info, operand.address);
     if (!dst_id) {
         return NULL_IMAGE_ID;
     }
-    const auto& image = slot_images[dst_id];
+    auto& image = slot_images[dst_id];
     if (False(image.flags & ImageFlagBits::GpuModified)) {
         // No need to waste time on an image that's synced with guest
         return NULL_IMAGE_ID;
     }
+    if (!is_upload && !image.info.dma_downloaded) {
+        // Force a full sync.
+        image.info.dma_downloaded = true;
+        return NULL_IMAGE_ID;
+    }
     const auto base = image.TryFindBase(operand.address);
     if (!base) {
         return NULL_IMAGE_ID;
@@ -1290,7 +1323,6 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VA
             all_siblings.push_back(overlap_id);
         } else {
             bad_overlap_ids.push_back(overlap_id);
-            overlap.flags |= ImageFlagBits::BadOverlap;
         }
     };
     ForEachImageInRegion(cpu_addr, size_bytes, region_check);
@@ -1401,7 +1433,12 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VA
         ImageBase& aliased = slot_images[aliased_id];
         aliased.overlapping_images.push_back(new_image_id);
         new_image.overlapping_images.push_back(aliased_id);
-        new_image.flags |= ImageFlagBits::BadOverlap;
+        if (aliased.info.resources.levels == 1 && aliased.overlapping_images.size() > 1) {
+            aliased.flags |= ImageFlagBits::BadOverlap;
+        }
+        if (new_image.info.resources.levels == 1 && new_image.overlapping_images.size() > 1) {
+            new_image.flags |= ImageFlagBits::BadOverlap;
+        }
     }
     RegisterImage(new_image_id);
     return new_image_id;

src/video_core/texture_cache/texture_cache_base.h

@@ -179,6 +179,8 @@ public:
     /// Download contents of host images to guest memory in a region
     void DownloadMemory(VAddr cpu_addr, size_t size);
 
+    std::optional<VideoCore::RasterizerDownloadArea> GetFlushArea(VAddr cpu_addr, u64 size);
+
     /// Remove images in a region
     void UnmapMemory(VAddr cpu_addr, size_t size);
 
@@ -205,7 +207,7 @@ public:
     /// Pop asynchronous downloads
     void PopAsyncFlushes();
 
-    [[nodiscard]] ImageId DmaImageId(const Tegra::DMA::ImageOperand& operand);
+    [[nodiscard]] ImageId DmaImageId(const Tegra::DMA::ImageOperand& operand, bool is_upload);
 
     [[nodiscard]] std::pair<Image*, BufferImageCopy> DmaBufferImageCopy(
         const Tegra::DMA::ImageCopy& copy_info, const Tegra::DMA::BufferOperand& buffer_operand,

src/video_core/texture_cache/util.cpp

@@ -896,11 +896,11 @@ void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8
         ASSERT(copy.buffer_row_length == Common::AlignUp(mip_size.width, tile_size.width));
         ASSERT(copy.buffer_image_height == Common::AlignUp(mip_size.height, tile_size.height));
         if (IsPixelFormatASTC(info.format)) {
-            ASSERT(copy.image_extent.depth == 1);
-            Tegra::Texture::ASTC::Decompress(input.subspan(copy.buffer_offset),
-                                             copy.image_extent.width, copy.image_extent.height,
-                                             copy.image_subresource.num_layers, tile_size.width,
-                                             tile_size.height, output.subspan(output_offset));
+            Tegra::Texture::ASTC::Decompress(
+                input.subspan(copy.buffer_offset), copy.image_extent.width,
+                copy.image_extent.height,
+                copy.image_subresource.num_layers * copy.image_extent.depth, tile_size.width,
+                tile_size.height, output.subspan(output_offset));
         } else {
             DecompressBC4(input.subspan(copy.buffer_offset), copy.image_extent,
                           output.subspan(output_offset));

src/yuzu/bootmanager.h

@@ -8,7 +8,6 @@
 #include <cstddef>
 #include <memory>
 #include <mutex>
-#include <stop_token>
 #include <utility>
 #include <vector>
 

src/yuzu/configuration/config.cpp

@@ -710,6 +710,7 @@ void Config::ReadRendererValues() {
     ReadGlobalSetting(Settings::values.nvdec_emulation);
     ReadGlobalSetting(Settings::values.accelerate_astc);
     ReadGlobalSetting(Settings::values.async_astc);
+    ReadGlobalSetting(Settings::values.use_reactive_flushing);
     ReadGlobalSetting(Settings::values.shader_backend);
     ReadGlobalSetting(Settings::values.use_asynchronous_shaders);
     ReadGlobalSetting(Settings::values.use_fast_gpu_time);
@@ -1355,6 +1356,7 @@ void Config::SaveRendererValues() {
                  Settings::values.nvdec_emulation.UsingGlobal());
     WriteGlobalSetting(Settings::values.accelerate_astc);
     WriteGlobalSetting(Settings::values.async_astc);
+    WriteGlobalSetting(Settings::values.use_reactive_flushing);
     WriteSetting(QString::fromStdString(Settings::values.shader_backend.GetLabel()),
                  static_cast<u32>(Settings::values.shader_backend.GetValue(global)),
                  static_cast<u32>(Settings::values.shader_backend.GetDefault()),

src/yuzu/configuration/configure_graphics_advanced.cpp

@@ -21,6 +21,7 @@ ConfigureGraphicsAdvanced::~ConfigureGraphicsAdvanced() = default;
 
 void ConfigureGraphicsAdvanced::SetConfiguration() {
     const bool runtime_lock = !system.IsPoweredOn();
+    ui->use_reactive_flushing->setEnabled(runtime_lock);
     ui->async_present->setEnabled(runtime_lock);
     ui->renderer_force_max_clock->setEnabled(runtime_lock);
     ui->async_astc->setEnabled(runtime_lock);
@@ -29,6 +30,7 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
 
     ui->async_present->setChecked(Settings::values.async_presentation.GetValue());
     ui->renderer_force_max_clock->setChecked(Settings::values.renderer_force_max_clock.GetValue());
+    ui->use_reactive_flushing->setChecked(Settings::values.use_reactive_flushing.GetValue());
     ui->async_astc->setChecked(Settings::values.async_astc.GetValue());
     ui->use_asynchronous_shaders->setChecked(Settings::values.use_asynchronous_shaders.GetValue());
     ui->use_fast_gpu_time->setChecked(Settings::values.use_fast_gpu_time.GetValue());
@@ -60,6 +62,8 @@ void ConfigureGraphicsAdvanced::ApplyConfiguration() {
                                              renderer_force_max_clock);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.max_anisotropy,
                                              ui->anisotropic_filtering_combobox);
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_reactive_flushing,
+                                             ui->use_reactive_flushing, use_reactive_flushing);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.async_astc, ui->async_astc,
                                              async_astc);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_asynchronous_shaders,
@@ -91,6 +95,7 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
         ui->async_present->setEnabled(Settings::values.async_presentation.UsingGlobal());
         ui->renderer_force_max_clock->setEnabled(
             Settings::values.renderer_force_max_clock.UsingGlobal());
+        ui->use_reactive_flushing->setEnabled(Settings::values.use_reactive_flushing.UsingGlobal());
         ui->async_astc->setEnabled(Settings::values.async_astc.UsingGlobal());
         ui->use_asynchronous_shaders->setEnabled(
             Settings::values.use_asynchronous_shaders.UsingGlobal());
@@ -108,6 +113,8 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
     ConfigurationShared::SetColoredTristate(ui->renderer_force_max_clock,
                                             Settings::values.renderer_force_max_clock,
                                             renderer_force_max_clock);
+    ConfigurationShared::SetColoredTristate(
+        ui->use_reactive_flushing, Settings::values.use_reactive_flushing, use_reactive_flushing);
     ConfigurationShared::SetColoredTristate(ui->async_astc, Settings::values.async_astc,
                                             async_astc);
     ConfigurationShared::SetColoredTristate(ui->use_asynchronous_shaders,

src/yuzu/configuration/configure_graphics_advanced.h

@@ -40,6 +40,7 @@ private:
     ConfigurationShared::CheckState renderer_force_max_clock;
     ConfigurationShared::CheckState use_vsync;
     ConfigurationShared::CheckState async_astc;
+    ConfigurationShared::CheckState use_reactive_flushing;
     ConfigurationShared::CheckState use_asynchronous_shaders;
     ConfigurationShared::CheckState use_fast_gpu_time;
     ConfigurationShared::CheckState use_vulkan_driver_pipeline_cache;

src/yuzu/configuration/configure_graphics_advanced.ui

@@ -96,6 +96,16 @@
           </property>
          </widget>
         </item>
+        <item>
+         <widget class="QCheckBox" name="use_reactive_flushing">
+          <property name="toolTip">
+           <string>Uses reactive flushing instead of predictive flushing. Allowing a more accurate syncing of memory.</string>
+          </property>
+          <property name="text">
+           <string>Enable Reactive Flushing</string>
+          </property>
+         </widget>
+        </item>
         <item>
          <widget class="QCheckBox" name="use_asynchronous_shaders">
           <property name="toolTip">

src/yuzu/configuration/configure_input_player.cpp

@@ -479,6 +479,9 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i
                             param.Set("threshold", new_threshold / 1000.0f);
                             emulated_controller->SetMotionParam(motion_id, param);
                         });
+                        context_menu.addAction(tr("Calibrate sensor"), [&] {
+                            emulated_controller->StartMotionCalibration();
+                        });
                     }
                     context_menu.exec(motion_map[motion_id]->mapToGlobal(menu_location));
                 });

src/yuzu/configuration/configure_input_player_widget.cpp

@@ -582,9 +582,9 @@ void PlayerControlPreview::DrawDualController(QPainter& p, const QPointF center)
         using namespace Settings::NativeMotion;
         p.setPen(colors.outline);
         p.setBrush(colors.transparent);
-        Draw3dCube(p, center + QPointF(-180, -5),
+        Draw3dCube(p, center + QPointF(-180, 90),
                    motion_values[Settings::NativeMotion::MotionLeft].euler, 20.0f);
-        Draw3dCube(p, center + QPointF(180, -5),
+        Draw3dCube(p, center + QPointF(180, 90),
                    motion_values[Settings::NativeMotion::MotionRight].euler, 20.0f);
     }
 
@@ -2926,14 +2926,14 @@ void PlayerControlPreview::DrawArrow(QPainter& p, const QPointF center, const Di
 void PlayerControlPreview::Draw3dCube(QPainter& p, QPointF center, const Common::Vec3f& euler,
                                       float size) {
     std::array<Common::Vec3f, 8> cube{
-        Common::Vec3f{-1, -1, -1},
-        {-1, 1, -1},
-        {1, 1, -1},
-        {1, -1, -1},
-        {-1, -1, 1},
-        {-1, 1, 1},
-        {1, 1, 1},
-        {1, -1, 1},
+        Common::Vec3f{-0.7f, -1, -0.5f},
+        {-0.7f, 1, -0.5f},
+        {0.7f, 1, -0.5f},
+        {0.7f, -1, -0.5f},
+        {-0.7f, -1, 0.5f},
+        {-0.7f, 1, 0.5f},
+        {0.7f, 1, 0.5f},
+        {0.7f, -1, 0.5f},
     };
 
     for (Common::Vec3f& point : cube) {

src/yuzu/qt_common.cpp

@@ -8,7 +8,7 @@
 #include "core/frontend/emu_window.h"
 #include "yuzu/qt_common.h"
 
-#ifdef __linux__
+#if !defined(WIN32) && !defined(__APPLE__)
 #include <qpa/qplatformnativeinterface.h>
 #endif
 

src/yuzu_cmd/config.cpp

@@ -312,6 +312,7 @@ void Config::ReadValues() {
     ReadSetting("Renderer", Settings::values.use_asynchronous_gpu_emulation);
     ReadSetting("Renderer", Settings::values.vsync_mode);
     ReadSetting("Renderer", Settings::values.shader_backend);
+    ReadSetting("Renderer", Settings::values.use_reactive_flushing);
     ReadSetting("Renderer", Settings::values.use_asynchronous_shaders);
     ReadSetting("Renderer", Settings::values.nvdec_emulation);
     ReadSetting("Renderer", Settings::values.accelerate_astc);

src/yuzu_cmd/default_ini.h

@@ -340,6 +340,10 @@ use_vsync =
 # 0: GLSL, 1 (default): GLASM, 2: SPIR-V
 shader_backend =
 
+# Uses reactive flushing instead of predictive flushing. Allowing a more accurate syncing of memory.
+# 0: Off, 1 (default): On
+use_reactive_flushing =
+
 # Whether to allow asynchronous shader building.
 # 0 (default): Off, 1: On
 use_asynchronous_shaders =