Android #170

OpenGL: Add GL_PRIMITIVES_GENERATED and GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN queries

CMakeLists.txt

@@ -142,6 +142,9 @@ if (YUZU_USE_BUNDLED_VCPKG)
     if (ENABLE_WEB_SERVICE)
         list(APPEND VCPKG_MANIFEST_FEATURES "web-service")
     endif()
+    if (ANDROID)
+        list(APPEND VCPKG_MANIFEST_FEATURES "android")
+    endif()
 
     include(${CMAKE_SOURCE_DIR}/externals/vcpkg/scripts/buildsystems/vcpkg.cmake)
 elseif(NOT "$ENV{VCPKG_TOOLCHAIN_FILE}" STREQUAL "")
@@ -302,7 +305,7 @@ find_package(ZLIB 1.2 REQUIRED)
 find_package(zstd 1.5 REQUIRED)
 
 if (NOT YUZU_USE_EXTERNAL_VULKAN_HEADERS)
-    find_package(Vulkan 1.3.256 REQUIRED)
+    find_package(Vulkan 1.3.274 REQUIRED)
 endif()
 
 if (ENABLE_LIBUSB)

README.md

@@ -1,3 +1,12 @@
+| Pull Request | Commit | Title | Author | Merged? |
+|----|----|----|----|----|
+| [12394](https://github.com/yuzu-emu/yuzu//pull/12394) | [`c57ae803a`](https://github.com/yuzu-emu/yuzu//pull/12394/files) |  general: properly support multiple memory instances  | [liamwhite](https://github.com/liamwhite/) | Yes |
+
+
+End of merge log. You can find the original README.md below the break.
+
+-----
+
 <!--
 SPDX-FileCopyrightText: 2018 yuzu Emulator Project
 SPDX-License-Identifier: GPL-2.0-or-later

src/android/app/src/main/java/org/yuzu/yuzu_emu/ui/GamesFragment.kt

@@ -91,18 +91,20 @@ class GamesFragment : Fragment() {
         viewLifecycleOwner.lifecycleScope.apply {
             launch {
                 repeatOnLifecycle(Lifecycle.State.RESUMED) {
-                    gamesViewModel.isReloading.collect { binding.swipeRefresh.isRefreshing = it }
+                    gamesViewModel.isReloading.collect {
+                        binding.swipeRefresh.isRefreshing = it
+                        if (gamesViewModel.games.value.isEmpty() && !it) {
+                            binding.noticeText.visibility = View.VISIBLE
+                        } else {
+                            binding.noticeText.visibility = View.INVISIBLE
+                        }
+                    }
                 }
             }
             launch {
                 repeatOnLifecycle(Lifecycle.State.RESUMED) {
                     gamesViewModel.games.collectLatest {
                         (binding.gridGames.adapter as GameAdapter).submitList(it)
-                        if (it.isEmpty()) {
-                            binding.noticeText.visibility = View.VISIBLE
-                        } else {
-                            binding.noticeText.visibility = View.GONE
-                        }
                     }
                 }
             }

src/android/app/src/main/res/values/arrays.xml

@@ -256,11 +256,13 @@
 
     <string-array name="outputEngineEntries">
         <item>@string/auto</item>
+        <item>@string/oboe</item>
         <item>@string/cubeb</item>
         <item>@string/string_null</item>
     </string-array>
     <integer-array name="outputEngineValues">
         <item>0</item>
+        <item>4</item>
         <item>1</item>
         <item>3</item>
     </integer-array>

src/android/app/src/main/res/values/strings.xml

@@ -503,6 +503,7 @@
     <string name="theme_mode_dark">Dark</string>
 
     <!-- Audio output engines -->
+    <string name="oboe">oboe</string>
     <string name="cubeb">cubeb</string>
 
     <!-- Black backgrounds theme -->

src/audio_core/CMakeLists.txt

@@ -253,6 +253,17 @@ if (ENABLE_SDL2)
     target_compile_definitions(audio_core PRIVATE HAVE_SDL2)
 endif()
 
+if (ANDROID)
+    target_sources(audio_core PRIVATE
+        sink/oboe_sink.cpp
+        sink/oboe_sink.h
+    )
+
+    # FIXME: this port seems broken, it cannot be imported with find_package(oboe REQUIRED)
+    target_link_libraries(audio_core PRIVATE "${VCPKG_INSTALLED_DIR}/${VCPKG_TARGET_TRIPLET}/lib/liboboe.a")
+    target_compile_definitions(audio_core PRIVATE HAVE_OBOE)
+endif()
+
 if (YUZU_USE_PRECOMPILED_HEADERS)
     target_precompile_headers(audio_core PRIVATE precompiled_headers.h)
 endif()

src/audio_core/sink/cubeb_sink.cpp

@@ -253,8 +253,9 @@ CubebSink::~CubebSink() {
 #endif
 }
 
-SinkStream* CubebSink::AcquireSinkStream(Core::System& system, u32 system_channels,
+SinkStream* CubebSink::AcquireSinkStream(Core::System& system, u32 system_channels_,
                                          const std::string& name, StreamType type) {
+    system_channels = system_channels_;
     SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<CubebSinkStream>(
         ctx, device_channels, system_channels, output_device, input_device, name, type, system));
 

src/audio_core/sink/oboe_sink.cpp

@@ -0,0 +1,223 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <span>
+#include <vector>
+
+#include <oboe/Oboe.h>
+
+#include "audio_core/common/common.h"
+#include "audio_core/sink/oboe_sink.h"
+#include "audio_core/sink/sink_stream.h"
+#include "common/logging/log.h"
+#include "common/scope_exit.h"
+#include "core/core.h"
+
+namespace AudioCore::Sink {
+
+class OboeSinkStream final : public SinkStream,
+                             public oboe::AudioStreamDataCallback,
+                             public oboe::AudioStreamErrorCallback {
+public:
+    explicit OboeSinkStream(Core::System& system_, StreamType type_, const std::string& name_,
+                            u32 system_channels_)
+        : SinkStream(system_, type_) {
+        name = name_;
+        system_channels = system_channels_;
+
+        this->OpenStream();
+    }
+
+    ~OboeSinkStream() override {
+        LOG_INFO(Audio_Sink, "Destroyed Oboe stream");
+    }
+
+    void Finalize() override {
+        this->Stop();
+        m_stream.reset();
+    }
+
+    void Start(bool resume = false) override {
+        if (!m_stream || !paused) {
+            return;
+        }
+
+        paused = false;
+
+        if (m_stream->start() != oboe::Result::OK) {
+            LOG_CRITICAL(Audio_Sink, "Error starting Oboe stream");
+        }
+    }
+
+    void Stop() override {
+        if (!m_stream || paused) {
+            return;
+        }
+
+        this->SignalPause();
+
+        if (m_stream->stop() != oboe::Result::OK) {
+            LOG_CRITICAL(Audio_Sink, "Error stopping Oboe stream");
+        }
+    }
+
+public:
+    static s32 QueryChannelCount(oboe::Direction direction) {
+        std::shared_ptr<oboe::AudioStream> temp_stream;
+        oboe::AudioStreamBuilder builder;
+
+        const auto result = ConfigureBuilder(builder, direction)->openStream(temp_stream);
+        ASSERT(result == oboe::Result::OK);
+
+        return temp_stream->getChannelCount() >= 6 ? 6 : 2;
+    }
+
+protected:
+    oboe::DataCallbackResult onAudioReady(oboe::AudioStream*, void* audio_data,
+                                          s32 num_buffer_frames) override {
+        const size_t num_channels = this->GetDeviceChannels();
+        const size_t frame_size = num_channels;
+        const size_t num_frames = static_cast<size_t>(num_buffer_frames);
+
+        if (type == StreamType::In) {
+            std::span<const s16> input_buffer{reinterpret_cast<const s16*>(audio_data),
+                                              num_frames * frame_size};
+            this->ProcessAudioIn(input_buffer, num_frames);
+        } else {
+            std::span<s16> output_buffer{reinterpret_cast<s16*>(audio_data),
+                                         num_frames * frame_size};
+            this->ProcessAudioOutAndRender(output_buffer, num_frames);
+        }
+
+        return oboe::DataCallbackResult::Continue;
+    }
+
+    void onErrorAfterClose(oboe::AudioStream*, oboe::Result) override {
+        LOG_INFO(Audio_Sink, "Audio stream closed, reinitializing");
+
+        if (this->OpenStream()) {
+            m_stream->start();
+        }
+    }
+
+private:
+    static oboe::AudioStreamBuilder* ConfigureBuilder(oboe::AudioStreamBuilder& builder,
+                                                      oboe::Direction direction) {
+        // TODO: investigate callback delay issues when using AAudio
+        return builder.setPerformanceMode(oboe::PerformanceMode::LowLatency)
+            ->setAudioApi(oboe::AudioApi::OpenSLES)
+            ->setDirection(direction)
+            ->setSampleRate(TargetSampleRate)
+            ->setSampleRateConversionQuality(oboe::SampleRateConversionQuality::High)
+            ->setFormat(oboe::AudioFormat::I16)
+            ->setFormatConversionAllowed(true)
+            ->setUsage(oboe::Usage::Game)
+            ->setBufferCapacityInFrames(TargetSampleCount * 2);
+    }
+
+    bool OpenStream() {
+        const auto direction = [&]() {
+            switch (type) {
+            case StreamType::In:
+                return oboe::Direction::Input;
+            case StreamType::Out:
+            case StreamType::Render:
+                return oboe::Direction::Output;
+            default:
+                ASSERT(false);
+                return oboe::Direction::Output;
+            }
+        }();
+
+        const auto expected_channels = QueryChannelCount(direction);
+        const auto expected_mask = [&]() {
+            switch (expected_channels) {
+            case 1:
+                return oboe::ChannelMask::Mono;
+            case 2:
+                return oboe::ChannelMask::Stereo;
+            case 6:
+                return oboe::ChannelMask::CM5Point1;
+            default:
+                ASSERT(false);
+                return oboe::ChannelMask::Unspecified;
+            }
+        }();
+
+        oboe::AudioStreamBuilder builder;
+        const auto result = ConfigureBuilder(builder, direction)
+                                ->setChannelCount(expected_channels)
+                                ->setChannelMask(expected_mask)
+                                ->setChannelConversionAllowed(true)
+                                ->setDataCallback(this)
+                                ->setErrorCallback(this)
+                                ->openStream(m_stream);
+        ASSERT(result == oboe::Result::OK);
+        return result == oboe::Result::OK && this->SetStreamProperties();
+    }
+
+    bool SetStreamProperties() {
+        ASSERT(m_stream);
+
+        m_stream->setBufferSizeInFrames(TargetSampleCount * 2);
+        device_channels = m_stream->getChannelCount();
+
+        const auto sample_rate = m_stream->getSampleRate();
+        const auto buffer_capacity = m_stream->getBufferCapacityInFrames();
+        const auto stream_backend =
+            m_stream->getAudioApi() == oboe::AudioApi::AAudio ? "AAudio" : "OpenSLES";
+
+        LOG_INFO(Audio_Sink, "Opened Oboe {} stream with {} channels sample rate {} capacity {}",
+                 stream_backend, device_channels, sample_rate, buffer_capacity);
+
+        return true;
+    }
+
+    std::shared_ptr<oboe::AudioStream> m_stream{};
+};
+
+OboeSink::OboeSink() {
+    // TODO: This is not generally knowable
+    // The channel count is distinct based on direction and can change
+    device_channels = OboeSinkStream::QueryChannelCount(oboe::Direction::Output);
+}
+
+OboeSink::~OboeSink() = default;
+
+SinkStream* OboeSink::AcquireSinkStream(Core::System& system, u32 system_channels,
+                                        const std::string& name, StreamType type) {
+    SinkStreamPtr& stream = sink_streams.emplace_back(
+        std::make_unique<OboeSinkStream>(system, type, name, system_channels));
+
+    return stream.get();
+}
+
+void OboeSink::CloseStream(SinkStream* to_remove) {
+    sink_streams.remove_if([&](auto& stream) { return stream.get() == to_remove; });
+}
+
+void OboeSink::CloseStreams() {
+    sink_streams.clear();
+}
+
+f32 OboeSink::GetDeviceVolume() const {
+    if (sink_streams.empty()) {
+        return 1.0f;
+    }
+
+    return sink_streams.front()->GetDeviceVolume();
+}
+
+void OboeSink::SetDeviceVolume(f32 volume) {
+    for (auto& stream : sink_streams) {
+        stream->SetDeviceVolume(volume);
+    }
+}
+
+void OboeSink::SetSystemVolume(f32 volume) {
+    for (auto& stream : sink_streams) {
+        stream->SetSystemVolume(volume);
+    }
+}
+
+} // namespace AudioCore::Sink

src/audio_core/sink/oboe_sink.h

@@ -0,0 +1,75 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <list>
+#include <string>
+
+#include "audio_core/sink/sink.h"
+
+namespace Core {
+class System;
+}
+
+namespace AudioCore::Sink {
+class SinkStream;
+
+class OboeSink final : public Sink {
+public:
+    explicit OboeSink();
+    ~OboeSink() override;
+
+    /**
+     * Create a new sink stream.
+     *
+     * @param system          - Core system.
+     * @param system_channels - Number of channels the audio system expects.
+     *                          May differ from the device's channel count.
+     * @param name            - Name of this stream.
+     * @param type            - Type of this stream, render/in/out.
+     *
+     * @return A pointer to the created SinkStream
+     */
+    SinkStream* AcquireSinkStream(Core::System& system, u32 system_channels,
+                                  const std::string& name, StreamType type) override;
+
+    /**
+     * Close a given stream.
+     *
+     * @param stream - The stream to close.
+     */
+    void CloseStream(SinkStream* stream) override;
+
+    /**
+     * Close all streams.
+     */
+    void CloseStreams() override;
+
+    /**
+     * Get the device volume. Set from calls to the IAudioDevice service.
+     *
+     * @return Volume of the device.
+     */
+    f32 GetDeviceVolume() const override;
+
+    /**
+     * Set the device volume. Set from calls to the IAudioDevice service.
+     *
+     * @param volume - New volume of the device.
+     */
+    void SetDeviceVolume(f32 volume) override;
+
+    /**
+     * Set the system volume. Comes from the audio system using this stream.
+     *
+     * @param volume - New volume of the system.
+     */
+    void SetSystemVolume(f32 volume) override;
+
+private:
+    /// List of streams managed by this sink
+    std::list<SinkStreamPtr> sink_streams{};
+};
+
+} // namespace AudioCore::Sink

src/audio_core/sink/sdl2_sink.cpp

@@ -168,8 +168,9 @@ SDLSink::SDLSink(std::string_view target_device_name) {
 
 SDLSink::~SDLSink() = default;
 
-SinkStream* SDLSink::AcquireSinkStream(Core::System& system, u32 system_channels,
+SinkStream* SDLSink::AcquireSinkStream(Core::System& system, u32 system_channels_,
                                        const std::string&, StreamType type) {
+    system_channels = system_channels_;
     SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<SDLSinkStream>(
         device_channels, system_channels, output_device, input_device, type, system));
     return stream.get();

src/audio_core/sink/sink.h

@@ -85,9 +85,21 @@ public:
      */
     virtual void SetSystemVolume(f32 volume) = 0;
 
+    /**
+     * Get the number of channels the game has set, can be different to the host hardware's support.
+     * Either 2 or 6.
+     *
+     * @return Number of device channels.
+     */
+    u32 GetSystemChannels() const {
+        return system_channels;
+    }
+
 protected:
     /// Number of device channels supported by the hardware
     u32 device_channels{2};
+    /// Number of channels the game is sending
+    u32 system_channels{2};
 };
 
 using SinkPtr = std::unique_ptr<Sink>;

src/audio_core/sink/sink_details.cpp

@@ -7,6 +7,9 @@
 #include <vector>
 
 #include "audio_core/sink/sink_details.h"
+#ifdef HAVE_OBOE
+#include "audio_core/sink/oboe_sink.h"
+#endif
 #ifdef HAVE_CUBEB
 #include "audio_core/sink/cubeb_sink.h"
 #endif
@@ -36,6 +39,16 @@ struct SinkDetails {
 
 // sink_details is ordered in terms of desirability, with the best choice at the top.
 constexpr SinkDetails sink_details[] = {
+#ifdef HAVE_OBOE
+    SinkDetails{
+        Settings::AudioEngine::Oboe,
+        [](std::string_view device_id) -> std::unique_ptr<Sink> {
+            return std::make_unique<OboeSink>();
+        },
+        [](bool capture) { return std::vector<std::string>{"Default"}; },
+        []() { return true; },
+    },
+#endif
 #ifdef HAVE_CUBEB
     SinkDetails{
         Settings::AudioEngine::Cubeb,

src/audio_core/sink/sink_stream.cpp

@@ -40,29 +40,36 @@ void SinkStream::AppendBuffer(SinkBuffer& buffer, std::span<s16> samples) {
 
     if (system_channels == 6 && device_channels == 2) {
         // We're given 6 channels, but our device only outputs 2, so downmix.
-        static constexpr std::array<f32, 4> down_mix_coeff{1.0f, 0.707f, 0.251f, 0.707f};
+        // Front = 1.0
+        // Center = 0.596
+        // LFE = 0.354
+        // Back = 0.707
+        static constexpr std::array<f32, 4> down_mix_coeff{1.0, 0.596f, 0.354f, 0.707f};
 
         for (u32 read_index = 0, write_index = 0; read_index < samples.size();
              read_index += system_channels, write_index += device_channels) {
+            const auto fl =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::FrontLeft)]);
+            const auto fr =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::FrontRight)]);
+            const auto c =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::Center)]);
+            const auto lfe =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::LFE)]);
+            const auto bl =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::BackLeft)]);
+            const auto br =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::BackRight)]);
+
             const auto left_sample{
-                ((Common::FixedPoint<49, 15>(
-                      samples[read_index + static_cast<u32>(Channels::FrontLeft)]) *
-                      down_mix_coeff[0] +
-                  samples[read_index + static_cast<u32>(Channels::Center)] * down_mix_coeff[1] +
-                  samples[read_index + static_cast<u32>(Channels::LFE)] * down_mix_coeff[2] +
-                  samples[read_index + static_cast<u32>(Channels::BackLeft)] * down_mix_coeff[3]) *
-                 volume)
-                    .to_int()};
+                static_cast<s32>((fl * down_mix_coeff[0] + c * down_mix_coeff[1] +
+                                  lfe * down_mix_coeff[2] + bl * down_mix_coeff[3]) *
+                                 volume)};
 
             const auto right_sample{
-                ((Common::FixedPoint<49, 15>(
-                      samples[read_index + static_cast<u32>(Channels::FrontRight)]) *
-                      down_mix_coeff[0] +
-                  samples[read_index + static_cast<u32>(Channels::Center)] * down_mix_coeff[1] +
-                  samples[read_index + static_cast<u32>(Channels::LFE)] * down_mix_coeff[2] +
-                  samples[read_index + static_cast<u32>(Channels::BackRight)] * down_mix_coeff[3]) *
-                 volume)
-                    .to_int()};
+                static_cast<s32>((fr * down_mix_coeff[0] + c * down_mix_coeff[1] +
+                                  lfe * down_mix_coeff[2] + br * down_mix_coeff[3]) *
+                                 volume)};
 
             samples[write_index + static_cast<u32>(Channels::FrontLeft)] =
                 static_cast<s16>(std::clamp(left_sample, min, max));

src/common/settings_enums.h

@@ -82,16 +82,15 @@ enum class AudioEngine : u32 {
     Cubeb,
     Sdl2,
     Null,
+    Oboe,
 };
 
 template <>
 inline std::vector<std::pair<std::string, AudioEngine>>
 EnumMetadata<AudioEngine>::Canonicalizations() {
     return {
-        {"auto", AudioEngine::Auto},
-        {"cubeb", AudioEngine::Cubeb},
-        {"sdl2", AudioEngine::Sdl2},
-        {"null", AudioEngine::Null},
+        {"auto", AudioEngine::Auto}, {"cubeb", AudioEngine::Cubeb}, {"sdl2", AudioEngine::Sdl2},
+        {"null", AudioEngine::Null}, {"oboe", AudioEngine::Oboe},
     };
 }
 

src/core/arm/arm_interface.cpp

@@ -9,7 +9,7 @@
 
 namespace Core {
 
-void ArmInterface::LogBacktrace(const Kernel::KProcess* process) const {
+void ArmInterface::LogBacktrace(Kernel::KProcess* process) const {
     Kernel::Svc::ThreadContext ctx;
     this->GetContext(ctx);
 

src/core/arm/arm_interface.h

@@ -95,7 +95,7 @@ public:
     virtual void SignalInterrupt(Kernel::KThread* thread) = 0;
 
     // Stack trace generation.
-    void LogBacktrace(const Kernel::KProcess* process) const;
+    void LogBacktrace(Kernel::KProcess* process) const;
 
     // Debug functionality.
     virtual const Kernel::DebugWatchpoint* HaltedWatchpoint() const = 0;

src/core/arm/debug.cpp

@@ -79,7 +79,7 @@ constexpr std::array<u64, 2> SegmentBases{
     0x7100000000ULL,
 };
 
-void SymbolicateBacktrace(const Kernel::KProcess* process, std::vector<BacktraceEntry>& out) {
+void SymbolicateBacktrace(Kernel::KProcess* process, std::vector<BacktraceEntry>& out) {
     auto modules = FindModules(process);
 
     const bool is_64 = process->Is64Bit();
@@ -118,7 +118,7 @@ void SymbolicateBacktrace(const Kernel::KProcess* process, std::vector<Backtrace
     }
 }
 
-std::vector<BacktraceEntry> GetAArch64Backtrace(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetAArch64Backtrace(Kernel::KProcess* process,
                                                 const Kernel::Svc::ThreadContext& ctx) {
     std::vector<BacktraceEntry> out;
     auto& memory = process->GetMemory();
@@ -144,7 +144,7 @@ std::vector<BacktraceEntry> GetAArch64Backtrace(const Kernel::KProcess* process,
     return out;
 }
 
-std::vector<BacktraceEntry> GetAArch32Backtrace(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetAArch32Backtrace(Kernel::KProcess* process,
                                                 const Kernel::Svc::ThreadContext& ctx) {
     std::vector<BacktraceEntry> out;
     auto& memory = process->GetMemory();
@@ -173,7 +173,7 @@ std::vector<BacktraceEntry> GetAArch32Backtrace(const Kernel::KProcess* process,
 } // namespace
 
 std::optional<std::string> GetThreadName(const Kernel::KThread* thread) {
-    const auto* process = thread->GetOwnerProcess();
+    auto* process = thread->GetOwnerProcess();
     if (process->Is64Bit()) {
         return GetNameFromThreadType64(process->GetMemory(), *thread);
     } else {
@@ -248,7 +248,7 @@ Kernel::KProcessAddress GetModuleEnd(const Kernel::KProcess* process,
     return cur_addr - 1;
 }
 
-Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process) {
+Loader::AppLoader::Modules FindModules(Kernel::KProcess* process) {
     Loader::AppLoader::Modules modules;
 
     auto& page_table = process->GetPageTable();
@@ -312,7 +312,7 @@ Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process) {
     return modules;
 }
 
-Kernel::KProcessAddress FindMainModuleEntrypoint(const Kernel::KProcess* process) {
+Kernel::KProcessAddress FindMainModuleEntrypoint(Kernel::KProcess* process) {
     // Do we have any loaded executable sections?
     auto modules = FindModules(process);
 
@@ -337,7 +337,7 @@ void InvalidateInstructionCacheRange(const Kernel::KProcess* process, u64 addres
     }
 }
 
-std::vector<BacktraceEntry> GetBacktraceFromContext(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetBacktraceFromContext(Kernel::KProcess* process,
                                                     const Kernel::Svc::ThreadContext& ctx) {
     if (process->Is64Bit()) {
         return GetAArch64Backtrace(process, ctx);

src/core/arm/debug.h

@@ -14,9 +14,9 @@ std::optional<std::string> GetThreadName(const Kernel::KThread* thread);
 std::string_view GetThreadWaitReason(const Kernel::KThread* thread);
 std::string GetThreadState(const Kernel::KThread* thread);
 
-Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process);
+Loader::AppLoader::Modules FindModules(Kernel::KProcess* process);
 Kernel::KProcessAddress GetModuleEnd(const Kernel::KProcess* process, Kernel::KProcessAddress base);
-Kernel::KProcessAddress FindMainModuleEntrypoint(const Kernel::KProcess* process);
+Kernel::KProcessAddress FindMainModuleEntrypoint(Kernel::KProcess* process);
 
 void InvalidateInstructionCacheRange(const Kernel::KProcess* process, u64 address, u64 size);
 
@@ -28,7 +28,7 @@ struct BacktraceEntry {
     std::string name;
 };
 
-std::vector<BacktraceEntry> GetBacktraceFromContext(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetBacktraceFromContext(Kernel::KProcess* process,
                                                     const Kernel::Svc::ThreadContext& ctx);
 std::vector<BacktraceEntry> GetBacktrace(const Kernel::KThread* thread);
 

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -15,7 +15,7 @@ using namespace Common::Literals;
 
 class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
 public:
-    explicit DynarmicCallbacks32(ArmDynarmic32& parent, const Kernel::KProcess* process)
+    explicit DynarmicCallbacks32(ArmDynarmic32& parent, Kernel::KProcess* process)
         : m_parent{parent}, m_memory(process->GetMemory()),
           m_process(process), m_debugger_enabled{parent.m_system.DebuggerEnabled()},
           m_check_memory_access{m_debugger_enabled ||
@@ -169,7 +169,7 @@ public:
 
     ArmDynarmic32& m_parent;
     Core::Memory::Memory& m_memory;
-    const Kernel::KProcess* m_process{};
+    Kernel::KProcess* m_process{};
     const bool m_debugger_enabled{};
     const bool m_check_memory_access{};
     static constexpr u64 MinimumRunCycles = 10000U;
@@ -370,7 +370,7 @@ void ArmDynarmic32::RewindBreakpointInstruction() {
     this->SetContext(m_breakpoint_context);
 }
 
-ArmDynarmic32::ArmDynarmic32(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+ArmDynarmic32::ArmDynarmic32(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                              DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index)
     : ArmInterface{uses_wall_clock}, m_system{system}, m_exclusive_monitor{exclusive_monitor},
       m_cb(std::make_unique<DynarmicCallbacks32>(*this, process)),

src/core/arm/dynarmic/arm_dynarmic_32.h

@@ -20,7 +20,7 @@ class System;
 
 class ArmDynarmic32 final : public ArmInterface {
 public:
-    ArmDynarmic32(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+    ArmDynarmic32(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                   DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index);
     ~ArmDynarmic32() override;
 

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -15,7 +15,7 @@ using namespace Common::Literals;
 
 class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
-    explicit DynarmicCallbacks64(ArmDynarmic64& parent, const Kernel::KProcess* process)
+    explicit DynarmicCallbacks64(ArmDynarmic64& parent, Kernel::KProcess* process)
         : m_parent{parent}, m_memory(process->GetMemory()),
           m_process(process), m_debugger_enabled{parent.m_system.DebuggerEnabled()},
           m_check_memory_access{m_debugger_enabled ||
@@ -216,7 +216,7 @@ public:
     Core::Memory::Memory& m_memory;
     u64 m_tpidrro_el0{};
     u64 m_tpidr_el0{};
-    const Kernel::KProcess* m_process{};
+    Kernel::KProcess* m_process{};
     const bool m_debugger_enabled{};
     const bool m_check_memory_access{};
     static constexpr u64 MinimumRunCycles = 10000U;
@@ -399,7 +399,7 @@ void ArmDynarmic64::RewindBreakpointInstruction() {
     this->SetContext(m_breakpoint_context);
 }
 
-ArmDynarmic64::ArmDynarmic64(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+ArmDynarmic64::ArmDynarmic64(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                              DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index)
     : ArmInterface{uses_wall_clock}, m_system{system}, m_exclusive_monitor{exclusive_monitor},
       m_cb(std::make_unique<DynarmicCallbacks64>(*this, process)), m_core_index{core_index} {

src/core/arm/dynarmic/arm_dynarmic_64.h

@@ -25,7 +25,7 @@ class System;
 
 class ArmDynarmic64 final : public ArmInterface {
 public:
-    ArmDynarmic64(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+    ArmDynarmic64(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                   DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index);
     ~ArmDynarmic64() override;
 

src/core/arm/nce/arm_nce.cpp

@@ -39,7 +39,7 @@ fpsimd_context* GetFloatingPointState(mcontext_t& host_ctx) {
 }
 
 using namespace Common::Literals;
-constexpr u32 StackSize = 32_KiB;
+constexpr u32 StackSize = 128_KiB;
 
 } // namespace
 

src/core/arm/nce/interpreter_visitor.cpp

@@ -5,8 +5,6 @@
 #include "common/bit_cast.h"
 #include "core/arm/nce/interpreter_visitor.h"
 
-#include <dynarmic/frontend/A64/decoder/a64.h>
-
 namespace Core {
 
 template <u32 BitSize>
@@ -249,6 +247,7 @@ bool InterpreterVisitor::LDR_lit_fpsimd(Imm<2> opc, Imm<19> imm19, Vec Vt) {
         return false;
     }
 
+    // Size in bytes
     const u64 size = 4 << opc.ZeroExtend();
     const u64 offset = imm19.SignExtend<u64>() << 2;
     const u64 address = this->GetPc() + offset;
@@ -530,7 +529,7 @@ bool InterpreterVisitor::SIMDImmediate(bool wback, bool postindex, size_t scale,
     }
     case MemOp::Load: {
         u128 data{};
-        m_memory.ReadBlock(address, &data, datasize);
+        m_memory.ReadBlock(address, &data, datasize / 8);
         this->SetVec(Vt, data);
         break;
     }

src/core/arm/nce/visitor_base.h

@@ -4,9 +4,15 @@
 
 #pragma once
 
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wshadow"
+
 #include <dynarmic/frontend/A64/a64_types.h>
+#include <dynarmic/frontend/A64/decoder/a64.h>
 #include <dynarmic/frontend/imm.h>
 
+#pragma GCC diagnostic pop
+
 namespace Core {
 
 class VisitorBase {

src/core/core.cpp

@@ -28,7 +28,6 @@
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
-#include "core/gpu_dirty_memory_manager.h"
 #include "core/hid/hid_core.h"
 #include "core/hle/kernel/k_memory_manager.h"
 #include "core/hle/kernel/k_process.h"
@@ -130,11 +129,8 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
 
 struct System::Impl {
     explicit Impl(System& system)
-        : kernel{system}, fs_controller{system}, memory{system}, hid_core{}, room_network{},
-          cpu_manager{system}, reporter{system}, applet_manager{system}, profile_manager{},
-          time_manager{system}, gpu_dirty_memory_write_manager{} {
-        memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager);
-    }
+        : kernel{system}, fs_controller{system}, hid_core{}, room_network{}, cpu_manager{system},
+          reporter{system}, applet_manager{system}, profile_manager{}, time_manager{system} {}
 
     void Initialize(System& system) {
         device_memory = std::make_unique<Core::DeviceMemory>();
@@ -241,17 +237,17 @@ struct System::Impl {
         debugger = std::make_unique<Debugger>(system, port);
     }
 
-    SystemResultStatus SetupForApplicationProcess(System& system, Frontend::EmuWindow& emu_window) {
+    void InitializeKernel(System& system) {
         LOG_DEBUG(Core, "initialized OK");
 
         // Setting changes may require a full system reinitialization (e.g., disabling multicore).
         ReinitializeIfNecessary(system);
 
-        memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager);
-
         kernel.Initialize();
         cpu_manager.Initialize();
+    }
 
+    SystemResultStatus SetupForApplicationProcess(System& system, Frontend::EmuWindow& emu_window) {
         /// Reset all glue registrations
         arp_manager.ResetAll();
 
@@ -300,17 +296,9 @@ struct System::Impl {
             return SystemResultStatus::ErrorGetLoader;
         }
 
-        SystemResultStatus init_result{SetupForApplicationProcess(system, emu_window)};
-        if (init_result != SystemResultStatus::Success) {
-            LOG_CRITICAL(Core, "Failed to initialize system (Error {})!",
-                         static_cast<int>(init_result));
-            ShutdownMainProcess();
-            return init_result;
-        }
+        InitializeKernel(system);
 
-        telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider);
-
-        // Create the process.
+        // Create the application process.
         auto main_process = Kernel::KProcess::Create(system.Kernel());
         Kernel::KProcess::Register(system.Kernel(), main_process);
         kernel.AppendNewProcess(main_process);
@@ -323,7 +311,18 @@ struct System::Impl {
             return static_cast<SystemResultStatus>(
                 static_cast<u32>(SystemResultStatus::ErrorLoader) + static_cast<u32>(load_result));
         }
+
+        // Set up the rest of the system.
+        SystemResultStatus init_result{SetupForApplicationProcess(system, emu_window)};
+        if (init_result != SystemResultStatus::Success) {
+            LOG_CRITICAL(Core, "Failed to initialize system (Error {})!",
+                         static_cast<int>(init_result));
+            ShutdownMainProcess();
+            return init_result;
+        }
+
         AddGlueRegistrationForProcess(*app_loader, *main_process);
+        telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider);
 
         // Initialize cheat engine
         if (cheat_engine) {
@@ -426,7 +425,6 @@ struct System::Impl {
         cpu_manager.Shutdown();
         debugger.reset();
         kernel.Shutdown();
-        memory.Reset();
         Network::RestartSocketOperations();
 
         if (auto room_member = room_network.GetRoomMember().lock()) {
@@ -507,7 +505,6 @@ struct System::Impl {
     std::unique_ptr<Tegra::Host1x::Host1x> host1x_core;
     std::unique_ptr<Core::DeviceMemory> device_memory;
     std::unique_ptr<AudioCore::AudioCore> audio_core;
-    Core::Memory::Memory memory;
     Core::HID::HIDCore hid_core;
     Network::RoomNetwork room_network;
 
@@ -567,9 +564,6 @@ struct System::Impl {
     std::array<u64, Core::Hardware::NUM_CPU_CORES> dynarmic_ticks{};
     std::array<MicroProfileToken, Core::Hardware::NUM_CPU_CORES> microprofile_cpu{};
 
-    std::array<Core::GPUDirtyMemoryManager, Core::Hardware::NUM_CPU_CORES>
-        gpu_dirty_memory_write_manager{};
-
     std::deque<std::vector<u8>> user_channel;
 };
 
@@ -652,29 +646,12 @@ void System::PrepareReschedule(const u32 core_index) {
     impl->kernel.PrepareReschedule(core_index);
 }
 
-Core::GPUDirtyMemoryManager& System::CurrentGPUDirtyMemoryManager() {
-    const std::size_t core = impl->kernel.GetCurrentHostThreadID();
-    return impl->gpu_dirty_memory_write_manager[core < Core::Hardware::NUM_CPU_CORES
-                                                    ? core
-                                                    : Core::Hardware::NUM_CPU_CORES - 1];
-}
-
-/// Provides a constant reference to the current gou dirty memory manager.
-const Core::GPUDirtyMemoryManager& System::CurrentGPUDirtyMemoryManager() const {
-    const std::size_t core = impl->kernel.GetCurrentHostThreadID();
-    return impl->gpu_dirty_memory_write_manager[core < Core::Hardware::NUM_CPU_CORES
-                                                    ? core
-                                                    : Core::Hardware::NUM_CPU_CORES - 1];
-}
-
 size_t System::GetCurrentHostThreadID() const {
     return impl->kernel.GetCurrentHostThreadID();
 }
 
 void System::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
-    for (auto& manager : impl->gpu_dirty_memory_write_manager) {
-        manager.Gather(callback);
-    }
+    return this->ApplicationProcess()->GatherGPUDirtyMemory(callback);
 }
 
 PerfStatsResults System::GetAndResetPerfStats() {
@@ -723,20 +700,12 @@ const Kernel::KProcess* System::ApplicationProcess() const {
     return impl->kernel.ApplicationProcess();
 }
 
-ExclusiveMonitor& System::Monitor() {
-    return impl->kernel.GetExclusiveMonitor();
-}
-
-const ExclusiveMonitor& System::Monitor() const {
-    return impl->kernel.GetExclusiveMonitor();
-}
-
 Memory::Memory& System::ApplicationMemory() {
-    return impl->memory;
+    return impl->kernel.ApplicationProcess()->GetMemory();
 }
 
 const Core::Memory::Memory& System::ApplicationMemory() const {
-    return impl->memory;
+    return impl->kernel.ApplicationProcess()->GetMemory();
 }
 
 Tegra::GPU& System::GPU() {

src/core/core.h

@@ -116,7 +116,6 @@ class CpuManager;
 class Debugger;
 class DeviceMemory;
 class ExclusiveMonitor;
-class GPUDirtyMemoryManager;
 class PerfStats;
 class Reporter;
 class SpeedLimiter;
@@ -225,12 +224,6 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule(u32 core_index);
 
-    /// Provides a reference to the gou dirty memory manager.
-    [[nodiscard]] Core::GPUDirtyMemoryManager& CurrentGPUDirtyMemoryManager();
-
-    /// Provides a constant reference to the current gou dirty memory manager.
-    [[nodiscard]] const Core::GPUDirtyMemoryManager& CurrentGPUDirtyMemoryManager() const;
-
     void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
 
     [[nodiscard]] size_t GetCurrentHostThreadID() const;
@@ -250,12 +243,6 @@ public:
     /// Gets a const reference to the underlying CPU manager
     [[nodiscard]] const CpuManager& GetCpuManager() const;
 
-    /// Gets a reference to the exclusive monitor
-    [[nodiscard]] ExclusiveMonitor& Monitor();
-
-    /// Gets a constant reference to the exclusive monitor
-    [[nodiscard]] const ExclusiveMonitor& Monitor() const;
-
     /// Gets a mutable reference to the system memory instance.
     [[nodiscard]] Core::Memory::Memory& ApplicationMemory();
 

src/core/file_sys/program_metadata.cpp

@@ -166,6 +166,10 @@ u32 ProgramMetadata::GetSystemResourceSize() const {
     return npdm_header.system_resource_size;
 }
 
+PoolPartition ProgramMetadata::GetPoolPartition() const {
+    return acid_header.pool_partition;
+}
+
 const ProgramMetadata::KernelCapabilityDescriptors& ProgramMetadata::GetKernelCapabilities() const {
     return aci_kernel_capabilities;
 }
@@ -201,7 +205,7 @@ void ProgramMetadata::Print() const {
     // Begin ACID printing (potential perms, signed)
     LOG_DEBUG(Service_FS, "Magic:                  {:.4}", acid_header.magic.data());
     LOG_DEBUG(Service_FS, "Flags:                  0x{:02X}", acid_header.flags);
-    LOG_DEBUG(Service_FS, " > Is Retail:           {}", acid_header.is_retail ? "YES" : "NO");
+    LOG_DEBUG(Service_FS, " > Is Retail:           {}", acid_header.production_flag ? "YES" : "NO");
     LOG_DEBUG(Service_FS, "Title ID Min:           0x{:016X}", acid_header.title_id_min);
     LOG_DEBUG(Service_FS, "Title ID Max:           0x{:016X}", acid_header.title_id_max);
     LOG_DEBUG(Service_FS, "Filesystem Access:      0x{:016X}\n", acid_file_access.permissions);

src/core/file_sys/program_metadata.h

@@ -34,6 +34,13 @@ enum class ProgramFilePermission : u64 {
     Everything = 1ULL << 63,
 };
 
+enum class PoolPartition : u32 {
+    Application = 0,
+    Applet = 1,
+    System = 2,
+    SystemNonSecure = 3,
+};
+
 /**
  * Helper which implements an interface to parse Program Description Metadata (NPDM)
  * Data can either be loaded from a file path or with data and an offset into it.
@@ -72,6 +79,7 @@ public:
     u64 GetTitleID() const;
     u64 GetFilesystemPermissions() const;
     u32 GetSystemResourceSize() const;
+    PoolPartition GetPoolPartition() const;
     const KernelCapabilityDescriptors& GetKernelCapabilities() const;
     const std::array<u8, 0x10>& GetName() const {
         return npdm_header.application_name;
@@ -116,8 +124,9 @@ private:
         union {
             u32 flags;
 
-            BitField<0, 1, u32> is_retail;
-            BitField<1, 31, u32> flags_unk;
+            BitField<0, 1, u32> production_flag;
+            BitField<1, 1, u32> unqualified_approval;
+            BitField<2, 4, PoolPartition> pool_partition;
         };
         u64_le title_id_min;
         u64_le title_id_max;

src/core/hle/kernel/k_address_arbiter.cpp

@@ -4,6 +4,7 @@
 #include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
 #include "core/hle/kernel/k_address_arbiter.h"
+#include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/k_thread.h"
@@ -26,9 +27,9 @@ bool ReadFromUser(KernelCore& kernel, s32* out, KProcessAddress address) {
     return true;
 }
 
-bool DecrementIfLessThan(Core::System& system, s32* out, KProcessAddress address, s32 value) {
-    auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+bool DecrementIfLessThan(KernelCore& kernel, s32* out, KProcessAddress address, s32 value) {
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     // NOTE: If scheduler lock is not held here, interrupt disable is required.
     // KScopedInterruptDisable di;
@@ -66,10 +67,10 @@ bool DecrementIfLessThan(Core::System& system, s32* out, KProcessAddress address
     return true;
 }
 
-bool UpdateIfEqual(Core::System& system, s32* out, KProcessAddress address, s32 value,
+bool UpdateIfEqual(KernelCore& kernel, s32* out, KProcessAddress address, s32 value,
                    s32 new_value) {
-    auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     // NOTE: If scheduler lock is not held here, interrupt disable is required.
     // KScopedInterruptDisable di;
@@ -159,7 +160,7 @@ Result KAddressArbiter::SignalAndIncrementIfEqual(uint64_t addr, s32 value, s32
 
         // Check the userspace value.
         s32 user_value{};
-        R_UNLESS(UpdateIfEqual(m_system, std::addressof(user_value), addr, value, value + 1),
+        R_UNLESS(UpdateIfEqual(m_kernel, std::addressof(user_value), addr, value, value + 1),
                  ResultInvalidCurrentMemory);
         R_UNLESS(user_value == value, ResultInvalidState);
 
@@ -219,7 +220,7 @@ Result KAddressArbiter::SignalAndModifyByWaitingCountIfEqual(uint64_t addr, s32
         s32 user_value{};
         bool succeeded{};
         if (value != new_value) {
-            succeeded = UpdateIfEqual(m_system, std::addressof(user_value), addr, value, new_value);
+            succeeded = UpdateIfEqual(m_kernel, std::addressof(user_value), addr, value, new_value);
         } else {
             succeeded = ReadFromUser(m_kernel, std::addressof(user_value), addr);
         }
@@ -262,7 +263,7 @@ Result KAddressArbiter::WaitIfLessThan(uint64_t addr, s32 value, bool decrement,
         s32 user_value{};
         bool succeeded{};
         if (decrement) {
-            succeeded = DecrementIfLessThan(m_system, std::addressof(user_value), addr, value);
+            succeeded = DecrementIfLessThan(m_kernel, std::addressof(user_value), addr, value);
         } else {
             succeeded = ReadFromUser(m_kernel, std::addressof(user_value), addr);
         }

src/core/hle/kernel/k_client_port.cpp

@@ -58,9 +58,8 @@ Result KClientPort::CreateSession(KClientSession** out) {
     KSession* session{};
 
     // Reserve a new session from the resource limit.
-    //! FIXME: we are reserving this from the wrong resource limit!
-    KScopedResourceReservation session_reservation(
-        m_kernel.ApplicationProcess()->GetResourceLimit(), LimitableResource::SessionCountMax);
+    KScopedResourceReservation session_reservation(GetCurrentProcessPointer(m_kernel),
+                                                   LimitableResource::SessionCountMax);
     R_UNLESS(session_reservation.Succeeded(), ResultLimitReached);
 
     // Allocate a session normally.

src/core/hle/kernel/k_condition_variable.cpp

@@ -28,10 +28,10 @@ bool WriteToUser(KernelCore& kernel, KProcessAddress address, const u32* p) {
     return true;
 }
 
-bool UpdateLockAtomic(Core::System& system, u32* out, KProcessAddress address, u32 if_zero,
+bool UpdateLockAtomic(KernelCore& kernel, u32* out, KProcessAddress address, u32 if_zero,
                       u32 new_orr_mask) {
-    auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     u32 expected{};
 
@@ -208,7 +208,7 @@ void KConditionVariable::SignalImpl(KThread* thread) {
         // TODO(bunnei): We should call CanAccessAtomic(..) here.
         can_access = true;
         if (can_access) [[likely]] {
-            UpdateLockAtomic(m_system, std::addressof(prev_tag), address, own_tag,
+            UpdateLockAtomic(m_kernel, std::addressof(prev_tag), address, own_tag,
                              Svc::HandleWaitMask);
         }
     }

src/core/hle/kernel/k_handle_table.h

@@ -30,7 +30,7 @@ public:
 public:
     explicit KHandleTable(KernelCore& kernel) : m_kernel(kernel) {}
 
-    Result Initialize(s32 size) {
+    Result Initialize(KProcess* owner, s32 size) {
         // Check that the table size is valid.
         R_UNLESS(size <= static_cast<s32>(MaxTableSize), ResultOutOfMemory);
 
@@ -44,6 +44,7 @@ public:
         m_next_linear_id = MinLinearId;
         m_count = 0;
         m_free_head_index = -1;
+        m_owner = owner;
 
         // Free all entries.
         for (s32 i = 0; i < static_cast<s32>(m_table_size); ++i) {
@@ -90,8 +91,8 @@ public:
         // Handle pseudo-handles.
         if constexpr (std::derived_from<KProcess, T>) {
             if (handle == Svc::PseudoHandle::CurrentProcess) {
-                //! FIXME: this is the wrong process!
-                auto* const cur_process = m_kernel.ApplicationProcess();
+                // TODO: this should be the current process
+                auto* const cur_process = m_owner;
                 ASSERT(cur_process != nullptr);
                 return cur_process;
             }
@@ -301,6 +302,7 @@ private:
 
 private:
     KernelCore& m_kernel;
+    KProcess* m_owner{};
     std::array<EntryInfo, MaxTableSize> m_entry_infos{};
     std::array<KAutoObject*, MaxTableSize> m_objects{};
     mutable KSpinLock m_lock;

src/core/hle/kernel/k_process.cpp

@@ -306,12 +306,16 @@ Result KProcess::Initialize(const Svc::CreateProcessParameter& params, const KPa
             False(params.flags & Svc::CreateProcessFlag::DisableDeviceAddressSpaceMerge);
         R_TRY(m_page_table.Initialize(as_type, enable_aslr, enable_das_merge, !enable_aslr, pool,
                                       params.code_address, params.code_num_pages * PageSize,
-                                      m_system_resource, res_limit, this->GetMemory(), 0));
+                                      m_system_resource, res_limit, m_memory, 0));
     }
     ON_RESULT_FAILURE_2 {
         m_page_table.Finalize();
     };
 
+    // Ensure our memory is initialized.
+    m_memory.SetCurrentPageTable(*this);
+    m_memory.SetGPUDirtyManagers(m_dirty_memory_managers);
+
     // Ensure we can insert the code region.
     R_UNLESS(m_page_table.CanContain(params.code_address, params.code_num_pages * PageSize,
                                      KMemoryState::Code),
@@ -399,12 +403,16 @@ Result KProcess::Initialize(const Svc::CreateProcessParameter& params,
             False(params.flags & Svc::CreateProcessFlag::DisableDeviceAddressSpaceMerge);
         R_TRY(m_page_table.Initialize(as_type, enable_aslr, enable_das_merge, !enable_aslr, pool,
                                       params.code_address, code_size, m_system_resource, res_limit,
-                                      this->GetMemory(), aslr_space_start));
+                                      m_memory, aslr_space_start));
     }
     ON_RESULT_FAILURE_2 {
         m_page_table.Finalize();
     };
 
+    // Ensure our memory is initialized.
+    m_memory.SetCurrentPageTable(*this);
+    m_memory.SetGPUDirtyManagers(m_dirty_memory_managers);
+
     // Ensure we can insert the code region.
     R_UNLESS(m_page_table.CanContain(params.code_address, code_size, KMemoryState::Code),
              ResultInvalidMemoryRegion);
@@ -1094,8 +1102,7 @@ void KProcess::UnpinThread(KThread* thread) {
 
 Result KProcess::GetThreadList(s32* out_num_threads, KProcessAddress out_thread_ids,
                                s32 max_out_count) {
-    // TODO: use current memory reference
-    auto& memory = m_kernel.System().ApplicationMemory();
+    auto& memory = this->GetMemory();
 
     // Lock the list.
     KScopedLightLock lk(m_list_lock);
@@ -1128,14 +1135,15 @@ void KProcess::Switch(KProcess* cur_process, KProcess* next_process) {}
 KProcess::KProcess(KernelCore& kernel)
     : KAutoObjectWithSlabHeapAndContainer(kernel), m_page_table{kernel}, m_state_lock{kernel},
       m_list_lock{kernel}, m_cond_var{kernel.System()}, m_address_arbiter{kernel.System()},
-      m_handle_table{kernel} {}
+      m_handle_table{kernel}, m_dirty_memory_managers{},
+      m_exclusive_monitor{}, m_memory{kernel.System()} {}
 KProcess::~KProcess() = default;
 
 Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size,
                                   KProcessAddress aslr_space_start, bool is_hbl) {
     // Create a resource limit for the process.
-    const auto physical_memory_size =
-        m_kernel.MemoryManager().GetSize(Kernel::KMemoryManager::Pool::Application);
+    const auto pool = static_cast<KMemoryManager::Pool>(metadata.GetPoolPartition());
+    const auto physical_memory_size = m_kernel.MemoryManager().GetSize(pool);
     auto* res_limit =
         Kernel::CreateResourceLimitForProcess(m_kernel.System(), physical_memory_size);
 
@@ -1146,8 +1154,10 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     Svc::CreateProcessFlag flag{};
     u64 code_address{};
 
-    // We are an application.
-    flag |= Svc::CreateProcessFlag::IsApplication;
+    // Determine if we are an application.
+    if (pool == KMemoryManager::Pool::Application) {
+        flag |= Svc::CreateProcessFlag::IsApplication;
+    }
 
     // If we are 64-bit, create as such.
     if (metadata.Is64BitProgram()) {
@@ -1196,8 +1206,8 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     std::memcpy(params.name.data(), name.data(), sizeof(params.name));
 
     // Initialize for application process.
-    R_TRY(this->Initialize(params, metadata.GetKernelCapabilities(), res_limit,
-                           KMemoryManager::Pool::Application, aslr_space_start));
+    R_TRY(this->Initialize(params, metadata.GetKernelCapabilities(), res_limit, pool,
+                           aslr_space_start));
 
     // Assign remaining properties.
     m_is_hbl = is_hbl;
@@ -1223,7 +1233,7 @@ void KProcess::LoadModule(CodeSet code_set, KProcessAddress base_addr) {
     ReprotectSegment(code_set.DataSegment(), Svc::MemoryPermission::ReadWrite);
 
 #ifdef HAS_NCE
-    if (Settings::IsNceEnabled()) {
+    if (this->IsApplication() && Settings::IsNceEnabled()) {
         auto& buffer = m_kernel.System().DeviceMemory().buffer;
         const auto& code = code_set.CodeSegment();
         const auto& patch = code_set.PatchSegment();
@@ -1235,10 +1245,11 @@ void KProcess::LoadModule(CodeSet code_set, KProcessAddress base_addr) {
 }
 
 void KProcess::InitializeInterfaces() {
-    this->GetMemory().SetCurrentPageTable(*this);
+    m_exclusive_monitor =
+        Core::MakeExclusiveMonitor(this->GetMemory(), Core::Hardware::NUM_CPU_CORES);
 
 #ifdef HAS_NCE
-    if (this->Is64Bit() && Settings::IsNceEnabled()) {
+    if (this->IsApplication() && Settings::IsNceEnabled()) {
         for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             m_arm_interfaces[i] = std::make_unique<Core::ArmNce>(m_kernel.System(), true, i);
         }
@@ -1248,13 +1259,13 @@ void KProcess::InitializeInterfaces() {
         for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             m_arm_interfaces[i] = std::make_unique<Core::ArmDynarmic64>(
                 m_kernel.System(), m_kernel.IsMulticore(), this,
-                static_cast<Core::DynarmicExclusiveMonitor&>(m_kernel.GetExclusiveMonitor()), i);
+                static_cast<Core::DynarmicExclusiveMonitor&>(*m_exclusive_monitor), i);
         }
     } else {
         for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             m_arm_interfaces[i] = std::make_unique<Core::ArmDynarmic32>(
                 m_kernel.System(), m_kernel.IsMulticore(), this,
-                static_cast<Core::DynarmicExclusiveMonitor&>(m_kernel.GetExclusiveMonitor()), i);
+                static_cast<Core::DynarmicExclusiveMonitor&>(*m_exclusive_monitor), i);
         }
     }
 }
@@ -1305,9 +1316,10 @@ bool KProcess::RemoveWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointT
     return true;
 }
 
-Core::Memory::Memory& KProcess::GetMemory() const {
-    // TODO: per-process memory
-    return m_kernel.System().ApplicationMemory();
+void KProcess::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
+    for (auto& manager : m_dirty_memory_managers) {
+        manager.Gather(callback);
+    }
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_process.h

@@ -7,6 +7,7 @@
 
 #include "core/arm/arm_interface.h"
 #include "core/file_sys/program_metadata.h"
+#include "core/gpu_dirty_memory_manager.h"
 #include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/k_address_arbiter.h"
 #include "core/hle/kernel/k_capabilities.h"
@@ -17,6 +18,7 @@
 #include "core/hle/kernel/k_system_resource.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/k_thread_local_page.h"
+#include "core/memory.h"
 
 namespace Kernel {
 
@@ -126,6 +128,9 @@ private:
 #ifdef HAS_NCE
     std::unordered_map<u64, u64> m_post_handlers{};
 #endif
+    std::array<Core::GPUDirtyMemoryManager, Core::Hardware::NUM_CPU_CORES> m_dirty_memory_managers;
+    std::unique_ptr<Core::ExclusiveMonitor> m_exclusive_monitor;
+    Core::Memory::Memory m_memory;
 
 private:
     Result StartTermination();
@@ -502,7 +507,15 @@ public:
 
     void InitializeInterfaces();
 
-    Core::Memory::Memory& GetMemory() const;
+    Core::Memory::Memory& GetMemory() {
+        return m_memory;
+    }
+
+    void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
+
+    Core::ExclusiveMonitor& GetExclusiveMonitor() const {
+        return *m_exclusive_monitor;
+    }
 
 public:
     // Overridden parent functions.
@@ -539,7 +552,7 @@ private:
 
     Result InitializeHandleTable(s32 size) {
         // Try to initialize the handle table.
-        R_TRY(m_handle_table.Initialize(size));
+        R_TRY(m_handle_table.Initialize(this, size));
 
         // We succeeded, so note that we did.
         m_is_handle_table_initialized = true;

src/core/hle/kernel/k_server_session.h

@@ -49,14 +49,21 @@ public:
     bool IsSignaled() const override;
     void OnClientClosed();
 
-    /// TODO: flesh these out to match the real kernel
     Result OnRequest(KSessionRequest* request);
-    Result SendReply(bool is_hle = false);
-    Result ReceiveRequest(std::shared_ptr<Service::HLERequestContext>* out_context = nullptr,
+    Result SendReply(uintptr_t server_message, uintptr_t server_buffer_size,
+                     KPhysicalAddress server_message_paddr, bool is_hle = false);
+    Result ReceiveRequest(uintptr_t server_message, uintptr_t server_buffer_size,
+                          KPhysicalAddress server_message_paddr,
+                          std::shared_ptr<Service::HLERequestContext>* out_context = nullptr,
                           std::weak_ptr<Service::SessionRequestManager> manager = {});
 
     Result SendReplyHLE() {
-        return SendReply(true);
+        R_RETURN(this->SendReply(0, 0, 0, true));
+    }
+
+    Result ReceiveRequestHLE(std::shared_ptr<Service::HLERequestContext>* out_context,
+                             std::weak_ptr<Service::SessionRequestManager> manager) {
+        R_RETURN(this->ReceiveRequest(0, 0, 0, out_context, manager));
     }
 
 private:

src/core/hle/kernel/k_session.cpp

@@ -33,8 +33,7 @@ void KSession::Initialize(KClientPort* client_port, uintptr_t name) {
     m_name = name;
 
     // Set our owner process.
-    //! FIXME: this is the wrong process!
-    m_process = m_kernel.ApplicationProcess();
+    m_process = GetCurrentProcessPointer(m_kernel);
     m_process->Open();
 
     // Set our port.

src/core/hle/kernel/k_thread.cpp

@@ -1422,8 +1422,7 @@ s32 GetCurrentCoreId(KernelCore& kernel) {
 }
 
 Core::Memory::Memory& GetCurrentMemory(KernelCore& kernel) {
-    // TODO: per-process memory
-    return kernel.System().ApplicationMemory();
+    return GetCurrentProcess(kernel).GetMemory();
 }
 
 KScopedDisableDispatch::~KScopedDisableDispatch() {

src/core/hle/kernel/k_thread.h

@@ -314,11 +314,7 @@ public:
         m_current_core_id = core;
     }
 
-    KProcess* GetOwnerProcess() {
-        return m_parent;
-    }
-
-    const KProcess* GetOwnerProcess() const {
+    KProcess* GetOwnerProcess() const {
         return m_parent;
     }
 

src/core/hle/kernel/kernel.cpp

@@ -68,8 +68,6 @@ struct KernelCore::Impl {
 
         global_object_list_container = std::make_unique<KAutoObjectWithListContainer>(kernel);
         global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
-        global_handle_table = std::make_unique<Kernel::KHandleTable>(kernel);
-        global_handle_table->Initialize(KHandleTable::MaxTableSize);
 
         is_phantom_mode_for_singlecore = false;
 
@@ -121,13 +119,8 @@ struct KernelCore::Impl {
         next_user_process_id = KProcess::ProcessIdMin;
         next_thread_id = 1;
 
-        global_handle_table->Finalize();
-        global_handle_table.reset();
-
         preemption_event = nullptr;
 
-        exclusive_monitor.reset();
-
         // Cleanup persistent kernel objects
         auto CleanupObject = [](KAutoObject* obj) {
             if (obj) {
@@ -191,8 +184,6 @@ struct KernelCore::Impl {
     }
 
     void InitializePhysicalCores() {
-        exclusive_monitor =
-            Core::MakeExclusiveMonitor(system.ApplicationMemory(), Core::Hardware::NUM_CPU_CORES);
         for (u32 i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             const s32 core{static_cast<s32>(i)};
 
@@ -791,10 +782,6 @@ struct KernelCore::Impl {
 
     std::shared_ptr<Core::Timing::EventType> preemption_event;
 
-    // This is the kernel's handle table or supervisor handle table which
-    // stores all the objects in place.
-    std::unique_ptr<KHandleTable> global_handle_table;
-
     std::unique_ptr<KAutoObjectWithListContainer> global_object_list_container;
 
     std::unique_ptr<KObjectNameGlobalData> object_name_global_data;
@@ -805,7 +792,6 @@ struct KernelCore::Impl {
     std::mutex server_lock;
     std::vector<std::unique_ptr<Service::ServerManager>> server_managers;
 
-    std::unique_ptr<Core::ExclusiveMonitor> exclusive_monitor;
     std::array<std::unique_ptr<Kernel::PhysicalCore>, Core::Hardware::NUM_CPU_CORES> cores;
 
     // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
@@ -882,10 +868,6 @@ KResourceLimit* KernelCore::GetSystemResourceLimit() {
     return impl->system_resource_limit;
 }
 
-KScopedAutoObject<KThread> KernelCore::RetrieveThreadFromGlobalHandleTable(Handle handle) const {
-    return impl->global_handle_table->GetObject<KThread>(handle);
-}
-
 void KernelCore::AppendNewProcess(KProcess* process) {
     impl->process_list.push_back(process);
 }
@@ -959,14 +941,6 @@ Kernel::KHardwareTimer& KernelCore::HardwareTimer() {
     return *impl->hardware_timer;
 }
 
-Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() {
-    return *impl->exclusive_monitor;
-}
-
-const Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() const {
-    return *impl->exclusive_monitor;
-}
-
 KAutoObjectWithListContainer& KernelCore::ObjectListContainer() {
     return *impl->global_object_list_container;
 }
@@ -1030,14 +1004,6 @@ u64 KernelCore::CreateNewUserProcessID() {
     return impl->next_user_process_id++;
 }
 
-KHandleTable& KernelCore::GlobalHandleTable() {
-    return *impl->global_handle_table;
-}
-
-const KHandleTable& KernelCore::GlobalHandleTable() const {
-    return *impl->global_handle_table;
-}
-
 void KernelCore::RegisterCoreThread(std::size_t core_id) {
     impl->RegisterCoreThread(core_id);
 }

src/core/hle/kernel/kernel.h

@@ -116,9 +116,6 @@ public:
     /// Retrieves a shared pointer to the system resource limit instance.
     KResourceLimit* GetSystemResourceLimit();
 
-    /// Retrieves a shared pointer to a Thread instance within the thread wakeup handle table.
-    KScopedAutoObject<KThread> RetrieveThreadFromGlobalHandleTable(Handle handle) const;
-
     /// Adds the given shared pointer to an internal list of active processes.
     void AppendNewProcess(KProcess* process);
 
@@ -170,10 +167,6 @@ public:
     /// Stops execution of 'id' core, in order to reschedule a new thread.
     void PrepareReschedule(std::size_t id);
 
-    Core::ExclusiveMonitor& GetExclusiveMonitor();
-
-    const Core::ExclusiveMonitor& GetExclusiveMonitor() const;
-
     KAutoObjectWithListContainer& ObjectListContainer();
 
     const KAutoObjectWithListContainer& ObjectListContainer() const;

src/core/hle/kernel/message_buffer.h

@@ -18,13 +18,13 @@ public:
         static constexpr inline u64 NullTag = 0;
 
     public:
-        enum class ReceiveListCountType : u32 {
-            None = 0,
-            ToMessageBuffer = 1,
-            ToSingleBuffer = 2,
+        enum ReceiveListCountType : u32 {
+            ReceiveListCountType_None = 0,
+            ReceiveListCountType_ToMessageBuffer = 1,
+            ReceiveListCountType_ToSingleBuffer = 2,
 
-            CountOffset = 2,
-            CountMax = 13,
+            ReceiveListCountType_CountOffset = 2,
+            ReceiveListCountType_CountMax = 13,
         };
 
     private:
@@ -591,16 +591,16 @@ public:
         // Add the size of the receive list.
         const auto count = hdr.GetReceiveListCount();
         switch (count) {
-        case MessageHeader::ReceiveListCountType::None:
+        case MessageHeader::ReceiveListCountType_None:
             break;
-        case MessageHeader::ReceiveListCountType::ToMessageBuffer:
+        case MessageHeader::ReceiveListCountType_ToMessageBuffer:
             break;
-        case MessageHeader::ReceiveListCountType::ToSingleBuffer:
+        case MessageHeader::ReceiveListCountType_ToSingleBuffer:
             msg_size += ReceiveListEntry::GetDataSize();
             break;
         default:
             msg_size += (static_cast<s32>(count) -
-                         static_cast<s32>(MessageHeader::ReceiveListCountType::CountOffset)) *
+                         static_cast<s32>(MessageHeader::ReceiveListCountType_CountOffset)) *
                         ReceiveListEntry::GetDataSize();
             break;
         }

src/core/hle/kernel/svc/svc_info.cpp

@@ -118,7 +118,6 @@ Result GetInfo(Core::System& system, u64* result, InfoType info_id_type, Handle
             R_SUCCEED();
 
         case InfoType::IsApplication:
-            LOG_WARNING(Kernel_SVC, "(STUBBED) Assuming process is application");
             *result = process->IsApplication();
             R_SUCCEED();
 

src/core/hle/kernel/svc/svc_ipc.cpp

@@ -48,8 +48,7 @@ Result ReplyAndReceiveImpl(KernelCore& kernel, int32_t* out_index, uintptr_t mes
         };
 
         // Send the reply.
-        R_TRY(session->SendReply());
-        // R_TRY(session->SendReply(message, buffer_size, message_paddr));
+        R_TRY(session->SendReply(message, buffer_size, message_paddr));
     }
 
     // Receive a message.
@@ -85,8 +84,7 @@ Result ReplyAndReceiveImpl(KernelCore& kernel, int32_t* out_index, uintptr_t mes
             if (R_SUCCEEDED(result)) {
                 KServerSession* session = objs[index]->DynamicCast<KServerSession*>();
                 if (session != nullptr) {
-                    // result = session->ReceiveRequest(message, buffer_size, message_paddr);
-                    result = session->ReceiveRequest();
+                    result = session->ReceiveRequest(message, buffer_size, message_paddr);
                     if (ResultNotFound == result) {
                         continue;
                     }

src/core/hle/kernel/svc_results.h

@@ -38,7 +38,9 @@ constexpr Result ResultInvalidState{ErrorModule::Kernel, 125};
 constexpr Result ResultReservedUsed{ErrorModule::Kernel, 126};
 constexpr Result ResultPortClosed{ErrorModule::Kernel, 131};
 constexpr Result ResultLimitReached{ErrorModule::Kernel, 132};
+constexpr Result ResultReceiveListBroken{ErrorModule::Kernel, 258};
 constexpr Result ResultOutOfAddressSpace{ErrorModule::Kernel, 259};
+constexpr Result ResultMessageTooLarge{ErrorModule::Kernel, 260};
 constexpr Result ResultInvalidId{ErrorModule::Kernel, 519};
 
 } // namespace Kernel

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

@@ -359,7 +359,7 @@ private:
 
     void GetActiveChannelCount(HLERequestContext& ctx) {
         const auto& sink{system.AudioCore().GetOutputSink()};
-        u32 channel_count{sink.GetDeviceChannels()};
+        u32 channel_count{sink.GetSystemChannels()};
 
         LOG_DEBUG(Service_Audio, "(STUBBED) called. Channels={}", channel_count);
 

src/core/hle/service/fatal/fatal.cpp

@@ -89,7 +89,7 @@ static void GenerateErrorReport(Core::System& system, Result error_code, const F
         crash_report += fmt::format("    ESR:                         {:016x}\n", info.esr);
         crash_report += fmt::format("    FAR:                         {:016x}\n", info.far);
         crash_report += "\nBacktrace:\n";
-        for (size_t i = 0; i < info.backtrace_size; i++) {
+        for (u32 i = 0; i < std::min<u32>(info.backtrace_size, 32); i++) {
             crash_report +=
                 fmt::format("    Backtrace[{:02d}]:               {:016x}\n", i, info.backtrace[i]);
         }

src/core/hle/service/filesystem/filesystem.h

@@ -54,6 +54,13 @@ enum class ImageDirectoryId : u32 {
     SdCard,
 };
 
+enum class OpenDirectoryMode : u64 {
+    Directory = (1 << 0),
+    File = (1 << 1),
+    All = Directory | File
+};
+DECLARE_ENUM_FLAG_OPERATORS(OpenDirectoryMode);
+
 class FileSystemController {
 public:
     explicit FileSystemController(Core::System& system_);

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

@@ -259,7 +259,7 @@ static void BuildEntryIndex(std::vector<FileSys::Entry>& entries, const std::vec
 
 class IDirectory final : public ServiceFramework<IDirectory> {
 public:
-    explicit IDirectory(Core::System& system_, FileSys::VirtualDir backend_)
+    explicit IDirectory(Core::System& system_, FileSys::VirtualDir backend_, OpenDirectoryMode mode)
         : ServiceFramework{system_, "IDirectory"}, backend(std::move(backend_)) {
         static const FunctionInfo functions[] = {
             {0, &IDirectory::Read, "Read"},
@@ -269,8 +269,12 @@ public:
 
         // TODO(DarkLordZach): Verify that this is the correct behavior.
         // Build entry index now to save time later.
-        BuildEntryIndex(entries, backend->GetFiles(), FileSys::EntryType::File);
-        BuildEntryIndex(entries, backend->GetSubdirectories(), FileSys::EntryType::Directory);
+        if (True(mode & OpenDirectoryMode::Directory)) {
+            BuildEntryIndex(entries, backend->GetSubdirectories(), FileSys::EntryType::Directory);
+        }
+        if (True(mode & OpenDirectoryMode::File)) {
+            BuildEntryIndex(entries, backend->GetFiles(), FileSys::EntryType::File);
+        }
     }
 
 private:
@@ -446,11 +450,9 @@ public:
 
         const auto file_buffer = ctx.ReadBuffer();
         const std::string name = Common::StringFromBuffer(file_buffer);
+        const auto mode = rp.PopRaw<OpenDirectoryMode>();
 
-        // TODO(Subv): Implement this filter.
-        const u32 filter_flags = rp.Pop<u32>();
-
-        LOG_DEBUG(Service_FS, "called. directory={}, filter={}", name, filter_flags);
+        LOG_DEBUG(Service_FS, "called. directory={}, mode={}", name, mode);
 
         FileSys::VirtualDir vfs_dir{};
         auto result = backend.OpenDirectory(&vfs_dir, name);
@@ -460,7 +462,7 @@ public:
             return;
         }
 
-        auto directory = std::make_shared<IDirectory>(system, vfs_dir);
+        auto directory = std::make_shared<IDirectory>(system, vfs_dir, mode);
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(ResultSuccess);

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

@@ -51,7 +51,7 @@ private:
         IPC::RequestParser rp{ctx};
         const auto vibration_device_handle{rp.PopRaw<Core::HID::VibrationDeviceHandle>()};
 
-        if (resource_manager != nullptr) {
+        if (resource_manager != nullptr && resource_manager->GetNpad()) {
             resource_manager->GetNpad()->InitializeVibrationDevice(vibration_device_handle);
         }
 

src/core/hle/service/ipc_helpers.h

@@ -151,8 +151,8 @@ public:
         if (manager->IsDomain()) {
             context->AddDomainObject(std::move(iface));
         } else {
-            kernel.ApplicationProcess()->GetResourceLimit()->Reserve(
-                Kernel::LimitableResource::SessionCountMax, 1);
+            ASSERT(Kernel::GetCurrentProcess(kernel).GetResourceLimit()->Reserve(
+                Kernel::LimitableResource::SessionCountMax, 1));
 
             auto* session = Kernel::KSession::Create(kernel);
             session->Initialize(nullptr, 0);

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

@@ -90,6 +90,18 @@ Status BufferQueueConsumer::AcquireBuffer(BufferItem* out_buffer,
 
     LOG_DEBUG(Service_Nvnflinger, "acquiring slot={}", slot);
 
+    // If the front buffer is still being tracked, update its slot state
+    if (core->StillTracking(*front)) {
+        slots[slot].acquire_called = true;
+        slots[slot].needs_cleanup_on_release = false;
+        slots[slot].buffer_state = BufferState::Acquired;
+
+        // TODO: for now, avoid resetting the fence, so that when we next return this
+        // slot to the producer, it will wait for the fence to pass. We should fix this
+        // by properly waiting for the fence in the BufferItemConsumer.
+        // slots[slot].fence = Fence::NoFence();
+    }
+
     // If the buffer has previously been acquired by the consumer, set graphic_buffer to nullptr to
     // avoid unnecessarily remapping this buffer on the consumer side.
     if (out_buffer->acquire_called) {
@@ -132,11 +144,28 @@ Status BufferQueueConsumer::ReleaseBuffer(s32 slot, u64 frame_number, const Fenc
             ++current;
         }
 
-        slots[slot].buffer_state = BufferState::Free;
+        if (slots[slot].buffer_state == BufferState::Acquired) {
+            // TODO: for now, avoid resetting the fence, so that when we next return this
+            // slot to the producer, it can wait for its own fence to pass. We should fix this
+            // by properly waiting for the fence in the BufferItemConsumer.
+            // slots[slot].fence = release_fence;
+            slots[slot].buffer_state = BufferState::Free;
 
-        listener = core->connected_producer_listener;
+            listener = core->connected_producer_listener;
 
-        LOG_DEBUG(Service_Nvnflinger, "releasing slot {}", slot);
+            LOG_DEBUG(Service_Nvnflinger, "releasing slot {}", slot);
+        } else if (slots[slot].needs_cleanup_on_release) {
+            LOG_DEBUG(Service_Nvnflinger, "releasing a stale buffer slot {} (state = {})", slot,
+                      slots[slot].buffer_state);
+            slots[slot].needs_cleanup_on_release = false;
+            return Status::StaleBufferSlot;
+        } else {
+            LOG_ERROR(Service_Nvnflinger,
+                      "attempted to release buffer slot {} but its state was {}", slot,
+                      slots[slot].buffer_state);
+
+            return Status::BadValue;
+        }
 
         core->SignalDequeueCondition();
     }

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

@@ -74,6 +74,10 @@ void BufferQueueCore::FreeBufferLocked(s32 slot) {
 
     slots[slot].graphic_buffer.reset();
 
+    if (slots[slot].buffer_state == BufferState::Acquired) {
+        slots[slot].needs_cleanup_on_release = true;
+    }
+
     slots[slot].buffer_state = BufferState::Free;
     slots[slot].frame_number = UINT32_MAX;
     slots[slot].acquire_called = false;

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

@@ -31,6 +31,7 @@ struct BufferSlot final {
     u64 frame_number{};
     Fence fence;
     bool acquire_called{};
+    bool needs_cleanup_on_release{};
     bool attached_by_consumer{};
     bool is_preallocated{};
 };

src/core/hle/service/server_manager.cpp

@@ -47,7 +47,7 @@ ServerManager::~ServerManager() {
     m_stopped.Wait();
     m_threads.clear();
 
-    // Clean up ports.
+    // Clean up server ports.
     for (const auto& [port, handler] : m_ports) {
         port->Close();
     }
@@ -97,22 +97,15 @@ Result ServerManager::RegisterNamedService(const std::string& service_name,
                                            u32 max_sessions) {
     ASSERT(m_sessions.size() + m_ports.size() < MaximumWaitObjects);
 
-    // Add the new server to sm:.
-    ASSERT(R_SUCCEEDED(
-        m_system.ServiceManager().RegisterService(service_name, max_sessions, handler_factory)));
-
-    // Get the registered port.
-    Kernel::KPort* port{};
-    ASSERT(
-        R_SUCCEEDED(m_system.ServiceManager().GetServicePort(std::addressof(port), service_name)));
-
-    // Open a new reference to the server port.
-    port->GetServerPort().Open();
+    // Add the new server to sm: and get the moved server port.
+    Kernel::KServerPort* server_port{};
+    R_ASSERT(m_system.ServiceManager().RegisterService(std::addressof(server_port), service_name,
+                                                       max_sessions, handler_factory));
 
     // Begin tracking the server port.
     {
         std::scoped_lock ll{m_list_mutex};
-        m_ports.emplace(std::addressof(port->GetServerPort()), std::move(handler_factory));
+        m_ports.emplace(server_port, std::move(handler_factory));
     }
 
     // Signal the wakeup event.
@@ -372,7 +365,7 @@ Result ServerManager::OnSessionEvent(Kernel::KServerSession* session,
 
     // Try to receive a message.
     std::shared_ptr<HLERequestContext> context;
-    rc = session->ReceiveRequest(&context, manager);
+    rc = session->ReceiveRequestHLE(&context, manager);
 
     // If the session has been closed, we're done.
     if (rc == Kernel::ResultSessionClosed) {

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

@@ -507,6 +507,14 @@ void SET_SYS::SetTvSettings(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void SET_SYS::GetDebugModeFlag(HLERequestContext& ctx) {
+    LOG_DEBUG(Service_SET, "called");
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(ResultSuccess);
+    rb.Push<u32>(0);
+}
+
 void SET_SYS::GetQuestFlag(HLERequestContext& ctx) {
     LOG_WARNING(Service_SET, "(STUBBED) called");
 
@@ -926,7 +934,7 @@ SET_SYS::SET_SYS(Core::System& system_) : ServiceFramework{system_, "set:sys"},
         {59, &SET_SYS::SetNetworkSystemClockContext, "SetNetworkSystemClockContext"},
         {60, &SET_SYS::IsUserSystemClockAutomaticCorrectionEnabled, "IsUserSystemClockAutomaticCorrectionEnabled"},
         {61, &SET_SYS::SetUserSystemClockAutomaticCorrectionEnabled, "SetUserSystemClockAutomaticCorrectionEnabled"},
-        {62, nullptr, "GetDebugModeFlag"},
+        {62, &SET_SYS::GetDebugModeFlag, "GetDebugModeFlag"},
         {63, &SET_SYS::GetPrimaryAlbumStorage, "GetPrimaryAlbumStorage"},
         {64, nullptr, "SetPrimaryAlbumStorage"},
         {65, nullptr, "GetUsb30EnableFlag"},
@@ -1143,6 +1151,8 @@ void SET_SYS::StoreSettings() {
 }
 
 void SET_SYS::StoreSettingsThreadFunc(std::stop_token stop_token) {
+    Common::SetCurrentThreadName("SettingsStore");
+
     while (Common::StoppableTimedWait(stop_token, std::chrono::minutes(1))) {
         std::scoped_lock l{m_save_needed_mutex};
         if (!std::exchange(m_save_needed, false)) {

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

@@ -98,6 +98,7 @@ private:
     void GetSettingsItemValue(HLERequestContext& ctx);
     void GetTvSettings(HLERequestContext& ctx);
     void SetTvSettings(HLERequestContext& ctx);
+    void GetDebugModeFlag(HLERequestContext& ctx);
     void GetQuestFlag(HLERequestContext& ctx);
     void GetDeviceTimeZoneLocationName(HLERequestContext& ctx);
     void SetDeviceTimeZoneLocationName(HLERequestContext& ctx);

src/core/hle/service/sm/sm.cpp

@@ -29,8 +29,7 @@ ServiceManager::ServiceManager(Kernel::KernelCore& kernel_) : kernel{kernel_} {
 
 ServiceManager::~ServiceManager() {
     for (auto& [name, port] : service_ports) {
-        port->GetClientPort().Close();
-        port->GetServerPort().Close();
+        port->Close();
     }
 
     if (deferral_event) {
@@ -50,8 +49,8 @@ static Result ValidateServiceName(const std::string& name) {
     return ResultSuccess;
 }
 
-Result ServiceManager::RegisterService(std::string name, u32 max_sessions,
-                                       SessionRequestHandlerFactory handler) {
+Result ServiceManager::RegisterService(Kernel::KServerPort** out_server_port, std::string name,
+                                       u32 max_sessions, SessionRequestHandlerFactory handler) {
     R_TRY(ValidateServiceName(name));
 
     std::scoped_lock lk{lock};
@@ -66,13 +65,17 @@ Result ServiceManager::RegisterService(std::string name, u32 max_sessions,
     // Register the port.
     Kernel::KPort::Register(kernel, port);
 
-    service_ports.emplace(name, port);
+    service_ports.emplace(name, std::addressof(port->GetClientPort()));
     registered_services.emplace(name, handler);
     if (deferral_event) {
         deferral_event->Signal();
     }
 
-    return ResultSuccess;
+    // Set our output.
+    *out_server_port = std::addressof(port->GetServerPort());
+
+    // We succeeded.
+    R_SUCCEED();
 }
 
 Result ServiceManager::UnregisterService(const std::string& name) {
@@ -91,7 +94,8 @@ Result ServiceManager::UnregisterService(const std::string& name) {
     return ResultSuccess;
 }
 
-Result ServiceManager::GetServicePort(Kernel::KPort** out_port, const std::string& name) {
+Result ServiceManager::GetServicePort(Kernel::KClientPort** out_client_port,
+                                      const std::string& name) {
     R_TRY(ValidateServiceName(name));
 
     std::scoped_lock lk{lock};
@@ -101,7 +105,7 @@ Result ServiceManager::GetServicePort(Kernel::KPort** out_port, const std::strin
         return Service::SM::ResultNotRegistered;
     }
 
-    *out_port = it->second;
+    *out_client_port = it->second;
     return ResultSuccess;
 }
 
@@ -172,8 +176,8 @@ Result SM::GetServiceImpl(Kernel::KClientSession** out_client_session, HLEReques
     std::string name(PopServiceName(rp));
 
     // Find the named port.
-    Kernel::KPort* port{};
-    auto port_result = service_manager.GetServicePort(&port, name);
+    Kernel::KClientPort* client_port{};
+    auto port_result = service_manager.GetServicePort(&client_port, name);
     if (port_result == Service::SM::ResultInvalidServiceName) {
         LOG_ERROR(Service_SM, "Invalid service name '{}'", name);
         return Service::SM::ResultInvalidServiceName;
@@ -187,7 +191,7 @@ Result SM::GetServiceImpl(Kernel::KClientSession** out_client_session, HLEReques
 
     // Create a new session.
     Kernel::KClientSession* session{};
-    if (const auto result = port->GetClientPort().CreateSession(&session); result.IsError()) {
+    if (const auto result = client_port->CreateSession(&session); result.IsError()) {
         LOG_ERROR(Service_SM, "called service={} -> error 0x{:08X}", name, result.raw);
         return result;
     }
@@ -221,7 +225,9 @@ void SM::RegisterServiceImpl(HLERequestContext& ctx, std::string name, u32 max_s
     LOG_DEBUG(Service_SM, "called with name={}, max_session_count={}, is_light={}", name,
               max_session_count, is_light);
 
-    if (const auto result = service_manager.RegisterService(name, max_session_count, nullptr);
+    Kernel::KServerPort* server_port{};
+    if (const auto result = service_manager.RegisterService(std::addressof(server_port), name,
+                                                            max_session_count, nullptr);
         result.IsError()) {
         LOG_ERROR(Service_SM, "failed to register service with error_code={:08X}", result.raw);
         IPC::ResponseBuilder rb{ctx, 2};
@@ -229,13 +235,9 @@ void SM::RegisterServiceImpl(HLERequestContext& ctx, std::string name, u32 max_s
         return;
     }
 
-    auto* port = Kernel::KPort::Create(kernel);
-    port->Initialize(ServerSessionCountMax, is_light, 0);
-    SCOPE_EXIT({ port->GetClientPort().Close(); });
-
     IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
     rb.Push(ResultSuccess);
-    rb.PushMoveObjects(port->GetServerPort());
+    rb.PushMoveObjects(server_port);
 }
 
 void SM::UnregisterService(HLERequestContext& ctx) {

src/core/hle/service/sm/sm.h

@@ -56,10 +56,10 @@ public:
     explicit ServiceManager(Kernel::KernelCore& kernel_);
     ~ServiceManager();
 
-    Result RegisterService(std::string name, u32 max_sessions,
-                           SessionRequestHandlerFactory handler_factory);
+    Result RegisterService(Kernel::KServerPort** out_server_port, std::string name,
+                           u32 max_sessions, SessionRequestHandlerFactory handler_factory);
     Result UnregisterService(const std::string& name);
-    Result GetServicePort(Kernel::KPort** out_port, const std::string& name);
+    Result GetServicePort(Kernel::KClientPort** out_client_port, const std::string& name);
 
     template <Common::DerivedFrom<SessionRequestHandler> T>
     std::shared_ptr<T> GetService(const std::string& service_name) const {
@@ -84,7 +84,7 @@ private:
     /// Map of registered services, retrieved using GetServicePort.
     std::mutex lock;
     std::unordered_map<std::string, SessionRequestHandlerFactory> registered_services;
-    std::unordered_map<std::string, Kernel::KPort*> service_ports;
+    std::unordered_map<std::string, Kernel::KClientPort*> service_ports;
 
     /// Kernel context
     Kernel::KernelCore& kernel;

src/core/hle/service/sm/sm_controller.cpp

@@ -28,7 +28,6 @@ void Controller::ConvertCurrentObjectToDomain(HLERequestContext& ctx) {
 void Controller::CloneCurrentObject(HLERequestContext& ctx) {
     LOG_DEBUG(Service, "called");
 
-    auto& process = *ctx.GetThread().GetOwnerProcess();
     auto session_manager = ctx.GetManager();
 
     // FIXME: this is duplicated from the SVC, it should just call it instead
@@ -36,11 +35,11 @@ void Controller::CloneCurrentObject(HLERequestContext& ctx) {
 
     // Reserve a new session from the process resource limit.
     Kernel::KScopedResourceReservation session_reservation(
-        &process, Kernel::LimitableResource::SessionCountMax);
+        Kernel::GetCurrentProcessPointer(kernel), Kernel::LimitableResource::SessionCountMax);
     ASSERT(session_reservation.Succeeded());
 
     // Create the session.
-    Kernel::KSession* session = Kernel::KSession::Create(system.Kernel());
+    Kernel::KSession* session = Kernel::KSession::Create(kernel);
     ASSERT(session != nullptr);
 
     // Initialize the session.
@@ -50,7 +49,7 @@ void Controller::CloneCurrentObject(HLERequestContext& ctx) {
     session_reservation.Commit();
 
     // Register the session.
-    Kernel::KSession::Register(system.Kernel(), session);
+    Kernel::KSession::Register(kernel, session);
 
     // Register with server manager.
     session_manager->GetServerManager().RegisterSession(&session->GetServerSession(),

src/core/loader/deconstructed_rom_directory.cpp

@@ -129,9 +129,10 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
     }
     metadata.Print();
 
-    // Enable NCE only for programs with 39-bit address space.
+    // Enable NCE only for applications with 39-bit address space.
     const bool is_39bit =
         metadata.GetAddressSpaceType() == FileSys::ProgramAddressSpaceType::Is39Bit;
+    const bool is_application = metadata.GetPoolPartition() == FileSys::PoolPartition::Application;
     Settings::SetNceEnabled(is_39bit);
 
     const std::array static_modules = {"rtld",    "main",    "subsdk0", "subsdk1", "subsdk2",
@@ -147,7 +148,7 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
 
     const auto GetPatcher = [&](size_t i) -> Core::NCE::Patcher* {
 #ifdef HAS_NCE
-        if (Settings::IsNceEnabled()) {
+        if (is_application && Settings::IsNceEnabled()) {
             return &module_patchers[i];
         }
 #endif
@@ -175,7 +176,7 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
 
     // Enable direct memory mapping in case of NCE.
     const u64 fastmem_base = [&]() -> size_t {
-        if (Settings::IsNceEnabled()) {
+        if (is_application && Settings::IsNceEnabled()) {
             auto& buffer = system.DeviceMemory().buffer;
             buffer.EnableDirectMappedAddress();
             return reinterpret_cast<u64>(buffer.VirtualBasePointer());

src/core/memory.cpp

@@ -45,7 +45,13 @@ struct Memory::Impl {
 
     void SetCurrentPageTable(Kernel::KProcess& process) {
         current_page_table = &process.GetPageTable().GetImpl();
-        current_page_table->fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer();
+
+        if (std::addressof(process) == system.ApplicationProcess() &&
+            Settings::IsFastmemEnabled()) {
+            current_page_table->fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer();
+        } else {
+            current_page_table->fastmem_arena = nullptr;
+        }
     }
 
     void MapMemoryRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size,
@@ -57,7 +63,7 @@ struct Memory::Impl {
         MapPages(page_table, base / YUZU_PAGESIZE, size / YUZU_PAGESIZE, target,
                  Common::PageType::Memory);
 
-        if (Settings::IsFastmemEnabled()) {
+        if (current_page_table->fastmem_arena) {
             system.DeviceMemory().buffer.Map(GetInteger(base),
                                              GetInteger(target) - DramMemoryMap::Base, size, perms);
         }
@@ -69,7 +75,7 @@ struct Memory::Impl {
         MapPages(page_table, base / YUZU_PAGESIZE, size / YUZU_PAGESIZE, 0,
                  Common::PageType::Unmapped);
 
-        if (Settings::IsFastmemEnabled()) {
+        if (current_page_table->fastmem_arena) {
             system.DeviceMemory().buffer.Unmap(GetInteger(base), size);
         }
     }
@@ -79,7 +85,7 @@ struct Memory::Impl {
         ASSERT_MSG((size & YUZU_PAGEMASK) == 0, "non-page aligned size: {:016X}", size);
         ASSERT_MSG((vaddr & YUZU_PAGEMASK) == 0, "non-page aligned base: {:016X}", vaddr);
 
-        if (!Settings::IsFastmemEnabled()) {
+        if (!current_page_table->fastmem_arena) {
             return;
         }
 
@@ -88,11 +94,6 @@ struct Memory::Impl {
         const bool is_x =
             True(perms & Common::MemoryPermission::Execute) && Settings::IsNceEnabled();
 
-        if (!current_page_table) {
-            system.DeviceMemory().buffer.Protect(vaddr, size, is_r, is_w, is_x);
-            return;
-        }
-
         u64 protect_bytes{};
         u64 protect_begin{};
         for (u64 addr = vaddr; addr < vaddr + size; addr += YUZU_PAGESIZE) {
@@ -239,7 +240,7 @@ struct Memory::Impl {
 
     bool WalkBlock(const Common::ProcessAddress addr, const std::size_t size, auto on_unmapped,
                    auto on_memory, auto on_rasterizer, auto increment) {
-        const auto& page_table = system.ApplicationProcess()->GetPageTable().GetImpl();
+        const auto& page_table = *current_page_table;
         std::size_t remaining_size = size;
         std::size_t page_index = addr >> YUZU_PAGEBITS;
         std::size_t page_offset = addr & YUZU_PAGEMASK;
@@ -484,7 +485,7 @@ struct Memory::Impl {
             return;
         }
 
-        if (Settings::IsFastmemEnabled()) {
+        if (current_page_table->fastmem_arena) {
             system.DeviceMemory().buffer.Protect(vaddr, size, !debug, !debug);
         }
 
@@ -541,7 +542,7 @@ struct Memory::Impl {
             return;
         }
 
-        if (Settings::IsFastmemEnabled()) {
+        if (current_page_table->fastmem_arena) {
             const bool is_read_enable =
                 !Settings::values.use_reactive_flushing.GetValue() || !cached;
             system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
@@ -886,8 +887,7 @@ void Memory::ProtectRegion(Common::PageTable& page_table, Common::ProcessAddress
 }
 
 bool Memory::IsValidVirtualAddress(const Common::ProcessAddress vaddr) const {
-    const Kernel::KProcess& process = *system.ApplicationProcess();
-    const auto& page_table = process.GetPageTable().GetImpl();
+    const auto& page_table = *impl->current_page_table;
     const size_t page = vaddr >> YUZU_PAGEBITS;
     if (page >= page_table.pointers.size()) {
         return false;

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -74,6 +74,11 @@ std::optional<OutAttr> OutputAttrPointer(EmitContext& ctx, IR::Attribute attr) {
     case IR::Attribute::ClipDistance7: {
         const u32 base{static_cast<u32>(IR::Attribute::ClipDistance0)};
         const u32 index{static_cast<u32>(attr) - base};
+        if (index >= ctx.profile.max_user_clip_distances) {
+            LOG_WARNING(Shader, "Ignoring clip distance store {} >= {} supported", index,
+                        ctx.profile.max_user_clip_distances);
+            return std::nullopt;
+        }
         const Id clip_num{ctx.Const(index)};
         return OutputAccessChain(ctx, ctx.output_f32, ctx.clip_distances, clip_num);
     }

src/shader_recompiler/backend/spirv/emit_spirv_image.cpp

@@ -214,16 +214,16 @@ Id TextureImage(EmitContext& ctx, IR::TextureInstInfo info, const IR::Value& ind
     }
 }
 
-Id Image(EmitContext& ctx, const IR::Value& index, IR::TextureInstInfo info) {
+std::pair<Id, bool> Image(EmitContext& ctx, const IR::Value& index, IR::TextureInstInfo info) {
     if (!index.IsImmediate() || index.U32() != 0) {
         throw NotImplementedException("Indirect image indexing");
     }
     if (info.type == TextureType::Buffer) {
         const ImageBufferDefinition def{ctx.image_buffers.at(info.descriptor_index)};
-        return ctx.OpLoad(def.image_type, def.id);
+        return {ctx.OpLoad(def.image_type, def.id), def.is_integer};
     } else {
         const ImageDefinition def{ctx.images.at(info.descriptor_index)};
-        return ctx.OpLoad(def.image_type, def.id);
+        return {ctx.OpLoad(def.image_type, def.id), def.is_integer};
     }
 }
 
@@ -566,13 +566,23 @@ Id EmitImageRead(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id co
         LOG_WARNING(Shader_SPIRV, "Typeless image read not supported by host");
         return ctx.ConstantNull(ctx.U32[4]);
     }
-    return Emit(&EmitContext::OpImageSparseRead, &EmitContext::OpImageRead, ctx, inst, ctx.U32[4],
-                Image(ctx, index, info), coords, std::nullopt, std::span<const Id>{});
+    const auto [image, is_integer] = Image(ctx, index, info);
+    const Id result_type{is_integer ? ctx.U32[4] : ctx.F32[4]};
+    Id color{Emit(&EmitContext::OpImageSparseRead, &EmitContext::OpImageRead, ctx, inst,
+                  result_type, image, coords, std::nullopt, std::span<const Id>{})};
+    if (!is_integer) {
+        color = ctx.OpBitcast(ctx.U32[4], color);
+    }
+    return color;
 }
 
 void EmitImageWrite(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id color) {
     const auto info{inst->Flags<IR::TextureInstInfo>()};
-    ctx.OpImageWrite(Image(ctx, index, info), coords, color);
+    const auto [image, is_integer] = Image(ctx, index, info);
+    if (!is_integer) {
+        color = ctx.OpBitcast(ctx.F32[4], color);
+    }
+    ctx.OpImageWrite(image, coords, color);
 }
 
 Id EmitIsTextureScaled(EmitContext& ctx, const IR::Value& index) {

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -74,20 +74,19 @@ spv::ImageFormat GetImageFormat(ImageFormat format) {
     throw InvalidArgument("Invalid image format {}", format);
 }
 
-Id ImageType(EmitContext& ctx, const ImageDescriptor& desc) {
+Id ImageType(EmitContext& ctx, const ImageDescriptor& desc, Id sampled_type) {
     const spv::ImageFormat format{GetImageFormat(desc.format)};
-    const Id type{ctx.U32[1]};
     switch (desc.type) {
     case TextureType::Color1D:
-        return ctx.TypeImage(type, spv::Dim::Dim1D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim1D, false, false, false, 2, format);
     case TextureType::ColorArray1D:
-        return ctx.TypeImage(type, spv::Dim::Dim1D, false, true, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim1D, false, true, false, 2, format);
     case TextureType::Color2D:
-        return ctx.TypeImage(type, spv::Dim::Dim2D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim2D, false, false, false, 2, format);
     case TextureType::ColorArray2D:
-        return ctx.TypeImage(type, spv::Dim::Dim2D, false, true, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim2D, false, true, false, 2, format);
     case TextureType::Color3D:
-        return ctx.TypeImage(type, spv::Dim::Dim3D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim3D, false, false, false, 2, format);
     case TextureType::Buffer:
         throw NotImplementedException("Image buffer");
     default:
@@ -1273,7 +1272,9 @@ void EmitContext::DefineImageBuffers(const Info& info, u32& binding) {
             throw NotImplementedException("Array of image buffers");
         }
         const spv::ImageFormat format{GetImageFormat(desc.format)};
-        const Id image_type{TypeImage(U32[1], spv::Dim::Buffer, false, false, false, 2, format)};
+        const Id sampled_type{desc.is_integer ? U32[1] : F32[1]};
+        const Id image_type{
+            TypeImage(sampled_type, spv::Dim::Buffer, false, false, false, 2, format)};
         const Id pointer_type{TypePointer(spv::StorageClass::UniformConstant, image_type)};
         const Id id{AddGlobalVariable(pointer_type, spv::StorageClass::UniformConstant)};
         Decorate(id, spv::Decoration::Binding, binding);
@@ -1283,6 +1284,7 @@ void EmitContext::DefineImageBuffers(const Info& info, u32& binding) {
             .id = id,
             .image_type = image_type,
             .count = desc.count,
+            .is_integer = desc.is_integer,
         });
         if (profile.supported_spirv >= 0x00010400) {
             interfaces.push_back(id);
@@ -1327,7 +1329,8 @@ void EmitContext::DefineImages(const Info& info, u32& binding, u32& scaling_inde
         if (desc.count != 1) {
             throw NotImplementedException("Array of images");
         }
-        const Id image_type{ImageType(*this, desc)};
+        const Id sampled_type{desc.is_integer ? U32[1] : F32[1]};
+        const Id image_type{ImageType(*this, desc, sampled_type)};
         const Id pointer_type{TypePointer(spv::StorageClass::UniformConstant, image_type)};
         const Id id{AddGlobalVariable(pointer_type, spv::StorageClass::UniformConstant)};
         Decorate(id, spv::Decoration::Binding, binding);
@@ -1337,6 +1340,7 @@ void EmitContext::DefineImages(const Info& info, u32& binding, u32& scaling_inde
             .id = id,
             .image_type = image_type,
             .count = desc.count,
+            .is_integer = desc.is_integer,
         });
         if (profile.supported_spirv >= 0x00010400) {
             interfaces.push_back(id);
@@ -1528,7 +1532,8 @@ void EmitContext::DefineOutputs(const IR::Program& program) {
         if (stage == Stage::Fragment) {
             throw NotImplementedException("Storing ClipDistance in fragment stage");
         }
-        const Id type{TypeArray(F32[1], Const(8U))};
+        const Id type{TypeArray(
+            F32[1], Const(std::min(info.used_clip_distances, profile.max_user_clip_distances)))};
         clip_distances = DefineOutput(*this, type, invocations, spv::BuiltIn::ClipDistance);
     }
     if (info.stores[IR::Attribute::Layer] &&

src/shader_recompiler/backend/spirv/spirv_emit_context.h

@@ -47,12 +47,14 @@ struct ImageBufferDefinition {
     Id id;
     Id image_type;
     u32 count;
+    bool is_integer;
 };
 
 struct ImageDefinition {
     Id id;
     Id image_type;
     u32 count;
+    bool is_integer;
 };
 
 struct UniformDefinitions {

src/shader_recompiler/environment.h

@@ -24,6 +24,8 @@ public:
 
     [[nodiscard]] virtual TexturePixelFormat ReadTexturePixelFormat(u32 raw_handle) = 0;
 
+    [[nodiscard]] virtual bool IsTexturePixelFormatInteger(u32 raw_handle) = 0;
+
     [[nodiscard]] virtual u32 ReadViewportTransformState() = 0;
 
     [[nodiscard]] virtual u32 TextureBoundBuffer() const = 0;

src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp

@@ -913,7 +913,11 @@ void GatherInfoFromHeader(Environment& env, Info& info) {
         }
         for (size_t index = 0; index < 8; ++index) {
             const u16 mask{header.vtg.omap_systemc.clip_distances};
-            info.stores.Set(IR::Attribute::ClipDistance0 + index, ((mask >> index) & 1) != 0);
+            const bool used{((mask >> index) & 1) != 0};
+            info.stores.Set(IR::Attribute::ClipDistance0 + index, used);
+            if (used) {
+                info.used_clip_distances = static_cast<u32>(index) + 1;
+            }
         }
         info.stores.Set(IR::Attribute::PrimitiveId,
                         header.vtg.omap_systemb.primitive_array_id != 0);

src/shader_recompiler/ir_opt/constant_propagation_pass.cpp

@@ -815,6 +815,15 @@ bool FindGradient3DDerivatives(std::array<IR::Value, 3>& results, IR::Value coor
     return true;
 }
 
+void ConvertDerivatives(std::array<IR::Value, 3>& results, IR::IREmitter& ir) {
+    for (size_t i = 0; i < 3; i++) {
+        if (results[i].Type() == IR::Type::U32) {
+            results[i] = results[i].IsImmediate() ? ir.Imm32(Common::BitCast<f32>(results[i].U32()))
+                                                  : ir.BitCast<IR::F32>(IR::U32(results[i]));
+        }
+    }
+}
+
 void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
     IR::TextureInstInfo info = inst.Flags<IR::TextureInstInfo>();
     auto orig_opcode = inst.GetOpcode();
@@ -831,12 +840,14 @@ void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
     if (!offset.IsImmediate()) {
         return;
     }
+    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::Inst* const inst2 = coords.InstRecursive();
     std::array<std::array<IR::Value, 3>, 3> results_matrix;
     for (size_t i = 0; i < 3; i++) {
         if (!FindGradient3DDerivatives(results_matrix[i], inst2->Arg(i).Resolve())) {
             return;
         }
+        ConvertDerivatives(results_matrix[i], ir);
     }
     IR::F32 lod_clamp{};
     if (info.has_lod_clamp != 0) {
@@ -846,7 +857,6 @@ void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
             lod_clamp = IR::F32{bias_lc};
         }
     }
-    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::Value new_coords =
         ir.CompositeConstruct(results_matrix[0][0], results_matrix[1][0], results_matrix[2][0]);
     IR::Value derivatives_1 = ir.CompositeConstruct(results_matrix[0][1], results_matrix[0][2],

src/shader_recompiler/ir_opt/texture_pass.cpp

@@ -372,6 +372,10 @@ TexturePixelFormat ReadTexturePixelFormat(Environment& env, const ConstBufferAdd
     return env.ReadTexturePixelFormat(GetTextureHandle(env, cbuf));
 }
 
+bool IsTexturePixelFormatInteger(Environment& env, const ConstBufferAddr& cbuf) {
+    return env.IsTexturePixelFormatInteger(GetTextureHandle(env, cbuf));
+}
+
 class Descriptors {
 public:
     explicit Descriptors(TextureBufferDescriptors& texture_buffer_descriptors_,
@@ -403,6 +407,7 @@ public:
         })};
         image_buffer_descriptors[index].is_written |= desc.is_written;
         image_buffer_descriptors[index].is_read |= desc.is_read;
+        image_buffer_descriptors[index].is_integer |= desc.is_integer;
         return index;
     }
 
@@ -432,6 +437,7 @@ public:
         })};
         image_descriptors[index].is_written |= desc.is_written;
         image_descriptors[index].is_read |= desc.is_read;
+        image_descriptors[index].is_integer |= desc.is_integer;
         return index;
     }
 
@@ -469,6 +475,20 @@ void PatchImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
                         ir.FPRecip(ir.ConvertUToF(32, 32, ir.CompositeExtract(texture_size, 1))))));
 }
 
+bool IsPixelFormatSNorm(TexturePixelFormat pixel_format) {
+    switch (pixel_format) {
+    case TexturePixelFormat::A8B8G8R8_SNORM:
+    case TexturePixelFormat::R8G8_SNORM:
+    case TexturePixelFormat::R8_SNORM:
+    case TexturePixelFormat::R16G16B16A16_SNORM:
+    case TexturePixelFormat::R16G16_SNORM:
+    case TexturePixelFormat::R16_SNORM:
+        return true;
+    default:
+        return false;
+    }
+}
+
 void PatchTexelFetch(IR::Block& block, IR::Inst& inst, TexturePixelFormat pixel_format) {
     const auto it{IR::Block::InstructionList::s_iterator_to(inst)};
     IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
@@ -587,11 +607,13 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
             }
             const bool is_written{inst->GetOpcode() != IR::Opcode::ImageRead};
             const bool is_read{inst->GetOpcode() != IR::Opcode::ImageWrite};
+            const bool is_integer{IsTexturePixelFormatInteger(env, cbuf)};
             if (flags.type == TextureType::Buffer) {
                 index = descriptors.Add(ImageBufferDescriptor{
                     .format = flags.image_format,
                     .is_written = is_written,
                     .is_read = is_read,
+                    .is_integer = is_integer,
                     .cbuf_index = cbuf.index,
                     .cbuf_offset = cbuf.offset,
                     .count = cbuf.count,
@@ -603,6 +625,7 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
                     .format = flags.image_format,
                     .is_written = is_written,
                     .is_read = is_read,
+                    .is_integer = is_integer,
                     .cbuf_index = cbuf.index,
                     .cbuf_offset = cbuf.offset,
                     .count = cbuf.count,
@@ -658,7 +681,7 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
         if (!host_info.support_snorm_render_buffer && inst->GetOpcode() == IR::Opcode::ImageFetch &&
             flags.type == TextureType::Buffer) {
             const auto pixel_format = ReadTexturePixelFormat(env, cbuf);
-            if (pixel_format != TexturePixelFormat::OTHER) {
+            if (IsPixelFormatSNorm(pixel_format)) {
                 PatchTexelFetch(*texture_inst.block, *texture_inst.inst, pixel_format);
             }
         }

src/shader_recompiler/profile.h

@@ -87,6 +87,8 @@ struct Profile {
     bool has_broken_robust{};
 
     u64 min_ssbo_alignment{};
+
+    u32 max_user_clip_distances{};
 };
 
 } // namespace Shader

src/shader_recompiler/shader_info.h

@@ -35,14 +35,109 @@ enum class TextureType : u32 {
 };
 constexpr u32 NUM_TEXTURE_TYPES = 9;
 
-enum class TexturePixelFormat : u32 {
+enum class TexturePixelFormat {
+    A8B8G8R8_UNORM,
     A8B8G8R8_SNORM,
+    A8B8G8R8_SINT,
+    A8B8G8R8_UINT,
+    R5G6B5_UNORM,
+    B5G6R5_UNORM,
+    A1R5G5B5_UNORM,
+    A2B10G10R10_UNORM,
+    A2B10G10R10_UINT,
+    A2R10G10B10_UNORM,
+    A1B5G5R5_UNORM,
+    A5B5G5R1_UNORM,
+    R8_UNORM,
     R8_SNORM,
-    R8G8_SNORM,
+    R8_SINT,
+    R8_UINT,
+    R16G16B16A16_FLOAT,
+    R16G16B16A16_UNORM,
     R16G16B16A16_SNORM,
-    R16G16_SNORM,
+    R16G16B16A16_SINT,
+    R16G16B16A16_UINT,
+    B10G11R11_FLOAT,
+    R32G32B32A32_UINT,
+    BC1_RGBA_UNORM,
+    BC2_UNORM,
+    BC3_UNORM,
+    BC4_UNORM,
+    BC4_SNORM,
+    BC5_UNORM,
+    BC5_SNORM,
+    BC7_UNORM,
+    BC6H_UFLOAT,
+    BC6H_SFLOAT,
+    ASTC_2D_4X4_UNORM,
+    B8G8R8A8_UNORM,
+    R32G32B32A32_FLOAT,
+    R32G32B32A32_SINT,
+    R32G32_FLOAT,
+    R32G32_SINT,
+    R32_FLOAT,
+    R16_FLOAT,
+    R16_UNORM,
     R16_SNORM,
-    OTHER
+    R16_UINT,
+    R16_SINT,
+    R16G16_UNORM,
+    R16G16_FLOAT,
+    R16G16_UINT,
+    R16G16_SINT,
+    R16G16_SNORM,
+    R32G32B32_FLOAT,
+    A8B8G8R8_SRGB,
+    R8G8_UNORM,
+    R8G8_SNORM,
+    R8G8_SINT,
+    R8G8_UINT,
+    R32G32_UINT,
+    R16G16B16X16_FLOAT,
+    R32_UINT,
+    R32_SINT,
+    ASTC_2D_8X8_UNORM,
+    ASTC_2D_8X5_UNORM,
+    ASTC_2D_5X4_UNORM,
+    B8G8R8A8_SRGB,
+    BC1_RGBA_SRGB,
+    BC2_SRGB,
+    BC3_SRGB,
+    BC7_SRGB,
+    A4B4G4R4_UNORM,
+    G4R4_UNORM,
+    ASTC_2D_4X4_SRGB,
+    ASTC_2D_8X8_SRGB,
+    ASTC_2D_8X5_SRGB,
+    ASTC_2D_5X4_SRGB,
+    ASTC_2D_5X5_UNORM,
+    ASTC_2D_5X5_SRGB,
+    ASTC_2D_10X8_UNORM,
+    ASTC_2D_10X8_SRGB,
+    ASTC_2D_6X6_UNORM,
+    ASTC_2D_6X6_SRGB,
+    ASTC_2D_10X6_UNORM,
+    ASTC_2D_10X6_SRGB,
+    ASTC_2D_10X5_UNORM,
+    ASTC_2D_10X5_SRGB,
+    ASTC_2D_10X10_UNORM,
+    ASTC_2D_10X10_SRGB,
+    ASTC_2D_12X10_UNORM,
+    ASTC_2D_12X10_SRGB,
+    ASTC_2D_12X12_UNORM,
+    ASTC_2D_12X12_SRGB,
+    ASTC_2D_8X6_UNORM,
+    ASTC_2D_8X6_SRGB,
+    ASTC_2D_6X5_UNORM,
+    ASTC_2D_6X5_SRGB,
+    E5B9G9R9_FLOAT,
+    D32_FLOAT,
+    D16_UNORM,
+    X8_D24_UNORM,
+    S8_UINT,
+    D24_UNORM_S8_UINT,
+    S8_UINT_D24_UNORM,
+    D32_FLOAT_S8_UINT,
 };
 
 enum class ImageFormat : u32 {
@@ -97,6 +192,7 @@ struct ImageBufferDescriptor {
     ImageFormat format;
     bool is_written;
     bool is_read;
+    bool is_integer;
     u32 cbuf_index;
     u32 cbuf_offset;
     u32 count;
@@ -129,6 +225,7 @@ struct ImageDescriptor {
     ImageFormat format;
     bool is_written;
     bool is_read;
+    bool is_integer;
     u32 cbuf_index;
     u32 cbuf_offset;
     u32 count;
@@ -227,6 +324,8 @@ struct Info {
     bool requires_layer_emulation{};
     IR::Attribute emulated_layer{};
 
+    u32 used_clip_distances{};
+
     boost::container::static_vector<ConstantBufferDescriptor, MAX_CBUFS>
         constant_buffer_descriptors;
     boost::container::static_vector<StorageBufferDescriptor, MAX_SSBOS> storage_buffers_descriptors;

src/tests/video_core/memory_tracker.cpp

@@ -23,13 +23,13 @@ constexpr VAddr c = 16 * HIGH_PAGE_SIZE;
 
 class RasterizerInterface {
 public:
-    void UpdatePagesCachedCount(VAddr addr, u64 size, bool cache) {
+    void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {
         const u64 page_start{addr >> Core::Memory::YUZU_PAGEBITS};
         const u64 page_end{(addr + size + Core::Memory::YUZU_PAGESIZE - 1) >>
                            Core::Memory::YUZU_PAGEBITS};
         for (u64 page = page_start; page < page_end; ++page) {
             int& value = page_table[page];
-            value += (cache ? 1 : -1);
+            value += delta;
             if (value < 0) {
                 throw std::logic_error{"negative page"};
             }
@@ -546,4 +546,4 @@ TEST_CASE("MemoryTracker: Cached write downloads") {
     REQUIRE(!memory_track->IsRegionGpuModified(c + PAGE, PAGE));
     memory_track->MarkRegionAsCpuModified(c, WORD);
     REQUIRE(rasterizer.Count() == 0);
-}
+}

src/video_core/buffer_cache/word_manager.h

@@ -473,7 +473,7 @@ private:
         VAddr addr = cpu_addr + word_index * BYTES_PER_WORD;
         IteratePages(changed_bits, [&](size_t offset, size_t size) {
             rasterizer->UpdatePagesCachedCount(addr + offset * BYTES_PER_PAGE,
-                                               size * BYTES_PER_PAGE, add_to_rasterizer);
+                                               size * BYTES_PER_PAGE, add_to_rasterizer ? 1 : -1);
         });
     }
 

src/video_core/engines/maxwell_3d.cpp

@@ -586,14 +586,22 @@ void Maxwell3D::ProcessQueryCondition() {
 }
 
 void Maxwell3D::ProcessCounterReset() {
-    switch (regs.clear_report_value) {
-    case Regs::ClearReport::ZPassPixelCount:
-        rasterizer->ResetCounter(VideoCommon::QueryType::ZPassPixelCount64);
-        break;
-    default:
-        LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}", regs.clear_report_value);
-        break;
-    }
+    const auto query_type = [clear_report = regs.clear_report_value]() {
+        switch (clear_report) {
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::ZPassPixelCount:
+            return VideoCommon::QueryType::ZPassPixelCount64;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::StreamingPrimitivesSucceeded:
+            return VideoCommon::QueryType::StreamingPrimitivesSucceeded;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::PrimitivesGenerated:
+            return VideoCommon::QueryType::PrimitivesGenerated;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::VtgPrimitivesOut:
+            return VideoCommon::QueryType::VtgPrimitivesOut;
+        default:
+            LOG_DEBUG(HW_GPU, "Unimplemented counter reset={}", clear_report);
+            return VideoCommon::QueryType::Payload;
+        }
+    }();
+    rasterizer->ResetCounter(query_type);
 }
 
 void Maxwell3D::ProcessSyncPoint() {

src/video_core/fence_manager.h

@@ -270,7 +270,7 @@ private:
 
     std::jthread fence_thread;
 
-    DelayedDestructionRing<TFence, 6> delayed_destruction_ring;
+    DelayedDestructionRing<TFence, 8> delayed_destruction_ring;
 };
 
 } // namespace VideoCommon

src/video_core/query_cache.h

@@ -28,8 +28,11 @@
 namespace VideoCore {
 enum class QueryType {
     SamplesPassed,
+    PrimitivesGenerated,
+    TfbPrimitivesWritten,
+    Count,
 };
-constexpr std::size_t NumQueryTypes = 1;
+constexpr std::size_t NumQueryTypes = static_cast<size_t>(QueryType::Count);
 } // namespace VideoCore
 
 namespace VideoCommon {
@@ -44,15 +47,6 @@ public:
     explicit CounterStreamBase(QueryCache& cache_, VideoCore::QueryType type_)
         : cache{cache_}, type{type_} {}
 
-    /// Updates the state of the stream, enabling or disabling as needed.
-    void Update(bool enabled) {
-        if (enabled) {
-            Enable();
-        } else {
-            Disable();
-        }
-    }
-
     /// Resets the stream to zero. It doesn't disable the query after resetting.
     void Reset() {
         if (current) {
@@ -80,7 +74,6 @@ public:
         return current != nullptr;
     }
 
-private:
     /// Enables the stream.
     void Enable() {
         if (current) {
@@ -97,6 +90,7 @@ private:
         last = std::exchange(current, nullptr);
     }
 
+private:
     QueryCache& cache;
     const VideoCore::QueryType type;
 
@@ -112,8 +106,14 @@ public:
         : rasterizer{rasterizer_},
           // Use reinterpret_cast instead of static_cast as workaround for
           // UBSan bug (https://github.com/llvm/llvm-project/issues/59060)
-          cpu_memory{cpu_memory_}, streams{{CounterStream{reinterpret_cast<QueryCache&>(*this),
-                                                          VideoCore::QueryType::SamplesPassed}}} {
+          cpu_memory{cpu_memory_}, streams{{
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::SamplesPassed}},
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::PrimitivesGenerated}},
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::TfbPrimitivesWritten}},
+                                   }} {
         (void)slot_async_jobs.insert(); // Null value
     }
 
@@ -157,12 +157,11 @@ public:
         AsyncFlushQuery(query, timestamp, lock);
     }
 
-    /// Updates counters from GPU state. Expected to be called once per draw, clear or dispatch.
-    void UpdateCounters() {
+    /// Enables all available GPU counters
+    void EnableCounters() {
         std::unique_lock lock{mutex};
-        if (maxwell3d) {
-            const auto& regs = maxwell3d->regs;
-            Stream(VideoCore::QueryType::SamplesPassed).Update(regs.zpass_pixel_count_enable);
+        for (auto& stream : streams) {
+            stream.Enable();
         }
     }
 
@@ -176,7 +175,7 @@ public:
     void DisableStreams() {
         std::unique_lock lock{mutex};
         for (auto& stream : streams) {
-            stream.Update(false);
+            stream.Disable();
         }
     }
 
@@ -353,7 +352,7 @@ private:
 
     std::shared_ptr<std::vector<AsyncJobId>> uncommitted_flushes{};
     std::list<std::shared_ptr<std::vector<AsyncJobId>>> committed_flushes;
-};
+}; // namespace VideoCommon
 
 template <class QueryCache, class HostCounter>
 class HostCounterBase {

src/video_core/rasterizer_accelerated.cpp

@@ -3,7 +3,6 @@
 
 #include <atomic>
 
-#include "common/alignment.h"
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/div_ceil.h"
@@ -12,65 +11,61 @@
 
 namespace VideoCore {
 
-static constexpr u16 IdentityValue = 1;
-
 using namespace Core::Memory;
 
-RasterizerAccelerated::RasterizerAccelerated(Memory& cpu_memory_) : map{}, cpu_memory{cpu_memory_} {
-    // We are tracking CPU memory, which cannot map more than 39 bits.
-    const VAddr start_address = 0;
-    const VAddr end_address = (1ULL << 39);
-    const IntervalType address_space_interval(start_address, end_address);
-    const auto value = std::make_pair(address_space_interval, IdentityValue);
-
-    map.add(value);
-}
+RasterizerAccelerated::RasterizerAccelerated(Memory& cpu_memory_)
+    : cached_pages(std::make_unique<CachedPages>()), cpu_memory{cpu_memory_} {}
 
 RasterizerAccelerated::~RasterizerAccelerated() = default;
 
-void RasterizerAccelerated::UpdatePagesCachedCount(VAddr addr, u64 size, bool cache) {
-    std::scoped_lock lk{map_lock};
+void RasterizerAccelerated::UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {
+    u64 uncache_begin = 0;
+    u64 cache_begin = 0;
+    u64 uncache_bytes = 0;
+    u64 cache_bytes = 0;
 
-    // Align sizes.
-    addr = Common::AlignDown(addr, YUZU_PAGESIZE);
-    size = Common::AlignUp(size, YUZU_PAGESIZE);
+    std::atomic_thread_fence(std::memory_order_acquire);
+    const u64 page_end = Common::DivCeil(addr + size, YUZU_PAGESIZE);
+    for (u64 page = addr >> YUZU_PAGEBITS; page != page_end; ++page) {
+        std::atomic_uint16_t& count = cached_pages->at(page >> 2).Count(page);
 
-    // Declare the overall interval we are going to operate on.
-    const VAddr start_address = addr;
-    const VAddr end_address = addr + size;
-    const IntervalType modification_range(start_address, end_address);
+        if (delta > 0) {
+            ASSERT_MSG(count.load(std::memory_order::relaxed) < UINT16_MAX, "Count may overflow!");
+        } else if (delta < 0) {
+            ASSERT_MSG(count.load(std::memory_order::relaxed) > 0, "Count may underflow!");
+        } else {
+            ASSERT_MSG(false, "Delta must be non-zero!");
+        }
 
-    // Find the boundaries of where to iterate.
-    const auto lower = map.lower_bound(modification_range);
-    const auto upper = map.upper_bound(modification_range);
+        // Adds or subtracts 1, as count is a unsigned 8-bit value
+        count.fetch_add(static_cast<u16>(delta), std::memory_order_release);
 
-    // Iterate over the contained intervals.
-    for (auto it = lower; it != upper; it++) {
-        // Intersect interval range with modification range.
-        const auto current_range = modification_range & it->first;
-
-        // Calculate the address and size to operate over.
-        const auto current_addr = current_range.lower();
-        const auto current_size = current_range.upper() - current_addr;
-
-        // Get the current value of the range.
-        const auto value = it->second;
-
-        if (cache && value == IdentityValue) {
-            // If we are going to cache, and the value is not yet referenced, then cache this range.
-            cpu_memory.RasterizerMarkRegionCached(current_addr, current_size, true);
-        } else if (!cache && value == IdentityValue + 1) {
-            // If we are going to uncache, and this is the last reference, then uncache this range.
-            cpu_memory.RasterizerMarkRegionCached(current_addr, current_size, false);
+        // Assume delta is either -1 or 1
+        if (count.load(std::memory_order::relaxed) == 0) {
+            if (uncache_bytes == 0) {
+                uncache_begin = page;
+            }
+            uncache_bytes += YUZU_PAGESIZE;
+        } else if (uncache_bytes > 0) {
+            cpu_memory.RasterizerMarkRegionCached(uncache_begin << YUZU_PAGEBITS, uncache_bytes,
+                                                  false);
+            uncache_bytes = 0;
+        }
+        if (count.load(std::memory_order::relaxed) == 1 && delta > 0) {
+            if (cache_bytes == 0) {
+                cache_begin = page;
+            }
+            cache_bytes += YUZU_PAGESIZE;
+        } else if (cache_bytes > 0) {
+            cpu_memory.RasterizerMarkRegionCached(cache_begin << YUZU_PAGEBITS, cache_bytes, true);
+            cache_bytes = 0;
         }
     }
-
-    // Update the set.
-    const auto value = std::make_pair(modification_range, IdentityValue);
-    if (cache) {
-        map.add(value);
-    } else {
-        map.subtract(value);
+    if (uncache_bytes > 0) {
+        cpu_memory.RasterizerMarkRegionCached(uncache_begin << YUZU_PAGEBITS, uncache_bytes, false);
+    }
+    if (cache_bytes > 0) {
+        cpu_memory.RasterizerMarkRegionCached(cache_begin << YUZU_PAGEBITS, cache_bytes, true);
     }
 }
 

src/video_core/rasterizer_accelerated.h

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <mutex>
-#include <boost/icl/interval_map.hpp>
+#include <array>
+#include <atomic>
 
 #include "common/common_types.h"
 #include "video_core/rasterizer_interface.h"
@@ -21,17 +21,28 @@ public:
     explicit RasterizerAccelerated(Core::Memory::Memory& cpu_memory_);
     ~RasterizerAccelerated() override;
 
-    void UpdatePagesCachedCount(VAddr addr, u64 size, bool cache) override;
+    void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) override;
 
 private:
-    using PageIndex = VAddr;
-    using PageReferenceCount = u16;
+    class CacheEntry final {
+    public:
+        CacheEntry() = default;
 
-    using IntervalMap = boost::icl::interval_map<PageIndex, PageReferenceCount>;
-    using IntervalType = IntervalMap::interval_type;
+        std::atomic_uint16_t& Count(std::size_t page) {
+            return values[page & 3];
+        }
 
-    IntervalMap map;
-    std::mutex map_lock;
+        const std::atomic_uint16_t& Count(std::size_t page) const {
+            return values[page & 3];
+        }
+
+    private:
+        std::array<std::atomic_uint16_t, 4> values{};
+    };
+    static_assert(sizeof(CacheEntry) == 8, "CacheEntry should be 8 bytes!");
+
+    using CachedPages = std::array<CacheEntry, 0x2000000>;
+    std::unique_ptr<CachedPages> cached_pages;
     Core::Memory::Memory& cpu_memory;
 };
 

src/video_core/rasterizer_interface.h

@@ -162,7 +162,7 @@ public:
     }
 
     /// Increase/decrease the number of object in pages touching the specified region
-    virtual void UpdatePagesCachedCount(VAddr addr, u64 size, bool cache) {}
+    virtual void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {}
 
     /// Initialize disk cached resources for the game being emulated
     virtual void LoadDiskResources(u64 title_id, std::stop_token stop_loading,

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -58,6 +58,9 @@ Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rast
         glObjectLabel(GL_BUFFER, buffer.handle, static_cast<GLsizei>(name.size()), name.data());
     }
     glNamedBufferData(buffer.handle, SizeBytes(), nullptr, GL_DYNAMIC_DRAW);
+    if (runtime.has_unified_vertex_buffers) {
+        glGetNamedBufferParameterui64vNV(buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &address);
+    }
 }
 
 void Buffer::ImmediateUpload(size_t offset, std::span<const u8> data) noexcept {
@@ -109,6 +112,7 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_,
     : device{device_}, staging_buffer_pool{staging_buffer_pool_},
       has_fast_buffer_sub_data{device.HasFastBufferSubData()},
       use_assembly_shaders{device.UseAssemblyShaders()},
+      has_unified_vertex_buffers{device.HasVertexBufferUnifiedMemory()},
       stream_buffer{has_fast_buffer_sub_data ? std::nullopt : std::make_optional<StreamBuffer>()} {
     GLint gl_max_attributes;
     glGetIntegerv(GL_MAX_VERTEX_ATTRIBS, &gl_max_attributes);
@@ -210,8 +214,14 @@ void BufferCacheRuntime::ClearBuffer(Buffer& dest_buffer, u32 offset, size_t siz
 }
 
 void BufferCacheRuntime::BindIndexBuffer(Buffer& buffer, u32 offset, u32 size) {
-    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer.Handle());
-    index_buffer_offset = offset;
+    if (has_unified_vertex_buffers) {
+        buffer.MakeResident(GL_READ_ONLY);
+        glBufferAddressRangeNV(GL_ELEMENT_ARRAY_ADDRESS_NV, 0, buffer.HostGpuAddr() + offset,
+                               static_cast<GLsizeiptr>(Common::AlignUp(size, 4)));
+    } else {
+        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer.Handle());
+        index_buffer_offset = offset;
+    }
 }
 
 void BufferCacheRuntime::BindVertexBuffer(u32 index, Buffer& buffer, u32 offset, u32 size,
@@ -219,8 +229,15 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, Buffer& buffer, u32 offset,
     if (index >= max_attributes) {
         return;
     }
-    glBindVertexBuffer(index, buffer.Handle(), static_cast<GLintptr>(offset),
-                       static_cast<GLsizei>(stride));
+    if (has_unified_vertex_buffers) {
+        buffer.MakeResident(GL_READ_ONLY);
+        glBindVertexBuffer(index, 0, 0, static_cast<GLsizei>(stride));
+        glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, index,
+                               buffer.HostGpuAddr() + offset, static_cast<GLsizeiptr>(size));
+    } else {
+        glBindVertexBuffer(index, buffer.Handle(), static_cast<GLintptr>(offset),
+                           static_cast<GLsizei>(stride));
+    }
 }
 
 void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bindings) {
@@ -233,9 +250,23 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
                            [](u64 stride) { return static_cast<GLsizei>(stride); });
     const u32 count =
         std::min(static_cast<u32>(bindings.buffers.size()), max_attributes - bindings.min_index);
-    glBindVertexBuffers(bindings.min_index, static_cast<GLsizei>(count), buffer_handles.data(),
-                        reinterpret_cast<const GLintptr*>(bindings.offsets.data()),
-                        buffer_strides.data());
+    if (has_unified_vertex_buffers) {
+        for (u32 index = 0; index < count; ++index) {
+            Buffer& buffer = *bindings.buffers[index];
+            buffer.MakeResident(GL_READ_ONLY);
+            glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, bindings.min_index + index,
+                                   buffer.HostGpuAddr() + bindings.offsets[index],
+                                   static_cast<GLsizeiptr>(bindings.sizes[index]));
+        }
+        static constexpr std::array<size_t, 32> ZEROS{};
+        glBindVertexBuffers(bindings.min_index, static_cast<GLsizei>(count),
+                            reinterpret_cast<const GLuint*>(ZEROS.data()),
+                            reinterpret_cast<const GLintptr*>(ZEROS.data()), buffer_strides.data());
+    } else {
+        glBindVertexBuffers(bindings.min_index, static_cast<GLsizei>(count), buffer_handles.data(),
+                            reinterpret_cast<const GLintptr*>(bindings.offsets.data()),
+                            buffer_strides.data());
+    }
 }
 
 void BufferCacheRuntime::BindUniformBuffer(size_t stage, u32 binding_index, Buffer& buffer,

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -209,6 +209,7 @@ private:
 
     bool has_fast_buffer_sub_data = false;
     bool use_assembly_shaders = false;
+    bool has_unified_vertex_buffers = false;
 
     bool use_storage_buffers = false;
 

src/video_core/renderer_opengl/gl_device.cpp

@@ -200,6 +200,7 @@ Device::Device(Core::Frontend::EmuWindow& emu_window) {
     has_broken_texture_view_formats = is_amd || (!is_linux && is_intel);
     has_nv_viewport_array2 = GLAD_GL_NV_viewport_array2;
     has_derivative_control = GLAD_GL_ARB_derivative_control;
+    has_vertex_buffer_unified_memory = GLAD_GL_NV_vertex_buffer_unified_memory;
     has_debugging_tool_attached = IsDebugToolAttached(extensions);
     has_depth_buffer_float = HasExtension(extensions, "GL_NV_depth_buffer_float");
     has_geometry_shader_passthrough = GLAD_GL_NV_geometry_shader_passthrough;

src/video_core/renderer_opengl/gl_device.h

@@ -72,6 +72,10 @@ public:
         return has_texture_shadow_lod;
     }
 
+    bool HasVertexBufferUnifiedMemory() const {
+        return has_vertex_buffer_unified_memory;
+    }
+
     bool HasASTC() const {
         return has_astc;
     }
@@ -211,6 +215,7 @@ private:
     bool has_vertex_viewport_layer{};
     bool has_image_load_formatted{};
     bool has_texture_shadow_lod{};
+    bool has_vertex_buffer_unified_memory{};
     bool has_astc{};
     bool has_variable_aoffi{};
     bool has_component_indexing_bug{};

src/video_core/renderer_opengl/gl_query_cache.cpp

@@ -18,16 +18,27 @@ namespace OpenGL {
 
 namespace {
 
-constexpr std::array<GLenum, VideoCore::NumQueryTypes> QueryTargets = {GL_SAMPLES_PASSED};
-
 constexpr GLenum GetTarget(VideoCore::QueryType type) {
-    return QueryTargets[static_cast<std::size_t>(type)];
+    switch (type) {
+    case VideoCore::QueryType::SamplesPassed:
+        return GL_SAMPLES_PASSED;
+    case VideoCore::QueryType::PrimitivesGenerated:
+        return GL_PRIMITIVES_GENERATED;
+    case VideoCore::QueryType::TfbPrimitivesWritten:
+        return GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN;
+    default:
+        break;
+    }
+    UNIMPLEMENTED_MSG("Query type {}", type);
+    return 0;
 }
 
 } // Anonymous namespace
 
 QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_)
-    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {}
+    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {
+    EnableCounters();
+}
 
 QueryCache::~QueryCache() = default;
 
@@ -103,13 +114,13 @@ u64 CachedQuery::Flush([[maybe_unused]] bool async) {
     auto& stream = cache->Stream(type);
     const bool slice_counter = WaitPending() && stream.IsEnabled();
     if (slice_counter) {
-        stream.Update(false);
+        stream.Disable();
     }
 
     auto result = VideoCommon::CachedQueryBase<HostCounter>::Flush();
 
     if (slice_counter) {
-        stream.Update(true);
+        stream.Enable();
     }
 
     return result;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -51,6 +51,22 @@ constexpr size_t NUM_SUPPORTED_VERTEX_ATTRIBUTES = 16;
 void oglEnable(GLenum cap, bool state) {
     (state ? glEnable : glDisable)(cap);
 }
+
+std::optional<VideoCore::QueryType> MaxwellToVideoCoreQuery(VideoCommon::QueryType type) {
+    switch (type) {
+    case VideoCommon::QueryType::PrimitivesGenerated:
+    case VideoCommon::QueryType::VtgPrimitivesOut:
+        return VideoCore::QueryType::PrimitivesGenerated;
+    case VideoCommon::QueryType::ZPassPixelCount64:
+        return VideoCore::QueryType::SamplesPassed;
+    case VideoCommon::QueryType::StreamingPrimitivesSucceeded:
+        // case VideoCommon::QueryType::StreamingByteCount:
+        // TODO: StreamingByteCount = StreamingPrimitivesSucceeded * num_verts * vert_stride
+        return VideoCore::QueryType::TfbPrimitivesWritten;
+    default:
+        return std::nullopt;
+    }
+}
 } // Anonymous namespace
 
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
@@ -162,14 +178,18 @@ void RasterizerOpenGL::Clear(u32 layer_count) {
         SyncFramebufferSRGB();
     }
     if (regs.clear_surface.Z) {
-        ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear Z but buffer is not enabled!");
+        if (regs.zeta_enable != 0) {
+            LOG_DEBUG(Render_OpenGL, "Tried to clear Z but buffer is not enabled!");
+        }
         use_depth = true;
 
         state_tracker.NotifyDepthMask();
         glDepthMask(GL_TRUE);
     }
     if (regs.clear_surface.S) {
-        ASSERT_MSG(regs.zeta_enable, "Tried to clear stencil but buffer is not enabled!");
+        if (regs.zeta_enable) {
+            LOG_DEBUG(Render_OpenGL, "Tried to clear stencil but buffer is not enabled!");
+        }
         use_stencil = true;
     }
 
@@ -212,7 +232,6 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) {
 
     SCOPE_EXIT({ gpu.TickWork(); });
     gpu_memory->FlushCaching();
-    query_cache.UpdateCounters();
 
     GraphicsPipeline* const pipeline{shader_cache.CurrentGraphicsPipeline()};
     if (!pipeline) {
@@ -330,7 +349,6 @@ void RasterizerOpenGL::DrawTexture() {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
-    query_cache.UpdateCounters();
 
     texture_cache.SynchronizeGraphicsDescriptors();
     texture_cache.UpdateRenderTargets(false);
@@ -397,21 +415,28 @@ void RasterizerOpenGL::DispatchCompute() {
 }
 
 void RasterizerOpenGL::ResetCounter(VideoCommon::QueryType type) {
-    if (type == VideoCommon::QueryType::ZPassPixelCount64) {
-        query_cache.ResetCounter(VideoCore::QueryType::SamplesPassed);
+    const auto query_cache_type = MaxwellToVideoCoreQuery(type);
+    if (!query_cache_type.has_value()) {
+        UNIMPLEMENTED_IF_MSG(type != VideoCommon::QueryType::Payload, "Reset query type: {}", type);
+        return;
     }
+    query_cache.ResetCounter(*query_cache_type);
 }
 
 void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCommon::QueryType type,
                              VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) {
-    if (type == VideoCommon::QueryType::ZPassPixelCount64) {
-        if (True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)) {
-            query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, {gpu.GetTicks()});
-        } else {
-            query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, std::nullopt);
-        }
-        return;
+    const auto query_cache_type = MaxwellToVideoCoreQuery(type);
+    if (!query_cache_type.has_value()) {
+        return QueryFallback(gpu_addr, type, flags, payload, subreport);
     }
+    const bool has_timeout = True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout);
+    const auto timestamp = has_timeout ? std::optional<u64>{gpu.GetTicks()} : std::nullopt;
+    query_cache.Query(gpu_addr, *query_cache_type, timestamp);
+}
+
+void RasterizerOpenGL::QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
+                                     VideoCommon::QueryPropertiesFlags flags, u32 payload,
+                                     u32 subreport) {
     if (type != VideoCommon::QueryType::Payload) {
         payload = 1u;
     }
@@ -1294,15 +1319,13 @@ void RasterizerOpenGL::BeginTransformFeedback(GraphicsPipeline* program, GLenum
     program->ConfigureTransformFeedback();
 
     UNIMPLEMENTED_IF(regs.IsShaderConfigEnabled(Maxwell::ShaderType::TessellationInit) ||
-                     regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation) ||
-                     regs.IsShaderConfigEnabled(Maxwell::ShaderType::Geometry));
-    UNIMPLEMENTED_IF(primitive_mode != GL_POINTS);
+                     regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation));
 
     // We may have to call BeginTransformFeedbackNV here since they seem to call different
     // implementations on Nvidia's driver (the pointer is different) but we are using
     // ARB_transform_feedback3 features with NV_transform_feedback interactions and the ARB
     // extension doesn't define BeginTransformFeedback (without NV) interactions. It just works.
-    glBeginTransformFeedback(GL_POINTS);
+    glBeginTransformFeedback(primitive_mode);
 }
 
 void RasterizerOpenGL::EndTransformFeedback() {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -225,6 +225,9 @@ private:
     /// End a transform feedback
     void EndTransformFeedback();
 
+    void QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
+                       VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport);
+
     Tegra::GPU& gpu;
 
     const Device& device;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -51,7 +51,7 @@ using VideoCommon::LoadPipelines;
 using VideoCommon::SerializePipeline;
 using Context = ShaderContext::Context;
 
-constexpr u32 CACHE_VERSION = 9;
+constexpr u32 CACHE_VERSION = 10;
 
 template <typename Container>
 auto MakeSpan(Container& container) {
@@ -233,6 +233,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .ignore_nan_fp_comparisons = true,
           .gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(),
           .min_ssbo_alignment = device.GetShaderStorageBufferAlignment(),
+          .max_user_clip_distances = 8,
       },
       host_info{
           .support_float64 = true,

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -168,6 +168,14 @@ RendererOpenGL::RendererOpenGL(Core::TelemetrySession& telemetry_session_,
     if (!GLAD_GL_ARB_seamless_cubemap_per_texture && !GLAD_GL_AMD_seamless_cubemap_per_texture) {
         glEnable(GL_TEXTURE_CUBE_MAP_SEAMLESS);
     }
+    // Enable unified vertex attributes and query vertex buffer address when the driver supports it
+    if (device.HasVertexBufferUnifiedMemory()) {
+        glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV);
+        glEnableClientState(GL_ELEMENT_ARRAY_UNIFIED_NV);
+        glMakeNamedBufferResidentNV(vertex_buffer.handle, GL_READ_ONLY);
+        glGetNamedBufferParameterui64vNV(vertex_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV,
+                                         &vertex_buffer_address);
+    }
 }
 
 RendererOpenGL::~RendererOpenGL() = default;
@@ -667,7 +675,13 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
                          offsetof(ScreenRectVertex, tex_coord));
     glVertexAttribBinding(PositionLocation, 0);
     glVertexAttribBinding(TexCoordLocation, 0);
-    glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+    if (device.HasVertexBufferUnifiedMemory()) {
+        glBindVertexBuffer(0, 0, 0, sizeof(ScreenRectVertex));
+        glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, 0, vertex_buffer_address,
+                               sizeof(vertices));
+    } else {
+        glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+    }
 
     if (Settings::values.scaling_filter.GetValue() != Settings::ScalingFilter::NearestNeighbor) {
         glBindSampler(0, present_sampler.handle);

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -78,8 +78,15 @@ 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), tracker{4096} {}
+Buffer::Buffer(BufferCacheRuntime& runtime, VideoCommon::NullBufferParams null_params)
+    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params), tracker{4096} {
+    if (runtime.device.HasNullDescriptor()) {
+        return;
+    }
+    device = &runtime.device;
+    buffer = runtime.CreateNullBuffer();
+    is_null = true;
+}
 
 Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
                VAddr cpu_addr_, u64 size_bytes_)
@@ -93,8 +100,12 @@ Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rast
 
 VkBufferView Buffer::View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format) {
     if (!device) {
-        // Null buffer, return a null descriptor
+        // Null buffer supported, return a null descriptor
         return VK_NULL_HANDLE;
+    } else if (is_null) {
+        // Null buffer not supported, adjust offset and size
+        offset = 0;
+        size = 0;
     }
     const auto it{std::ranges::find_if(views, [offset, size, format](const BufferView& view) {
         return offset == view.offset && size == view.size && format == view.format;
@@ -563,22 +574,27 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
         }
         buffer_handles.push_back(handle);
     }
+    const u32 device_max = device.GetMaxVertexInputBindings();
+    const u32 min_binding = std::min(bindings.min_index, device_max);
+    const u32 max_binding = std::min(bindings.max_index, device_max);
+    const u32 binding_count = max_binding - min_binding;
+    if (binding_count == 0) {
+        return;
+    }
     if (device.IsExtExtendedDynamicStateSupported()) {
-        scheduler.Record([this, bindings_ = std::move(bindings),
-                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
-            cmdbuf.BindVertexBuffers2EXT(bindings_.min_index,
-                                         std::min(bindings_.max_index - bindings_.min_index,
-                                                  device.GetMaxVertexInputBindings()),
-                                         buffer_handles_.data(), bindings_.offsets.data(),
-                                         bindings_.sizes.data(), bindings_.strides.data());
+        scheduler.Record([bindings_ = std::move(bindings),
+                          buffer_handles_ = std::move(buffer_handles),
+                          binding_count](vk::CommandBuffer cmdbuf) {
+            cmdbuf.BindVertexBuffers2EXT(bindings_.min_index, binding_count, buffer_handles_.data(),
+                                         bindings_.offsets.data(), bindings_.sizes.data(),
+                                         bindings_.strides.data());
         });
     } else {
-        scheduler.Record([this, bindings_ = std::move(bindings),
-                          buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
-            cmdbuf.BindVertexBuffers(bindings_.min_index,
-                                     std::min(bindings_.max_index - bindings_.min_index,
-                                              device.GetMaxVertexInputBindings()),
-                                     buffer_handles_.data(), bindings_.offsets.data());
+        scheduler.Record([bindings_ = std::move(bindings),
+                          buffer_handles_ = std::move(buffer_handles),
+                          binding_count](vk::CommandBuffer cmdbuf) {
+            cmdbuf.BindVertexBuffers(bindings_.min_index, binding_count, buffer_handles_.data(),
+                                     bindings_.offsets.data());
         });
     }
 }
@@ -622,9 +638,12 @@ void BufferCacheRuntime::BindTransformFeedbackBuffers(VideoCommon::HostBindings<
 }
 
 void BufferCacheRuntime::ReserveNullBuffer() {
-    if (null_buffer) {
-        return;
+    if (!null_buffer) {
+        null_buffer = CreateNullBuffer();
     }
+}
+
+vk::Buffer BufferCacheRuntime::CreateNullBuffer() {
     VkBufferCreateInfo create_info{
         .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
         .pNext = nullptr,
@@ -639,15 +658,17 @@ void BufferCacheRuntime::ReserveNullBuffer() {
     if (device.IsExtTransformFeedbackSupported()) {
         create_info.usage |= VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT;
     }
-    null_buffer = memory_allocator.CreateBuffer(create_info, MemoryUsage::DeviceLocal);
+    vk::Buffer ret = memory_allocator.CreateBuffer(create_info, MemoryUsage::DeviceLocal);
     if (device.HasDebuggingToolAttached()) {
-        null_buffer.SetObjectNameEXT("Null buffer");
+        ret.SetObjectNameEXT("Null buffer");
     }
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([buffer = *null_buffer](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([buffer = *ret](vk::CommandBuffer cmdbuf) {
         cmdbuf.FillBuffer(buffer, 0, VK_WHOLE_SIZE, 0);
     });
+
+    return ret;
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -63,6 +63,7 @@ private:
     vk::Buffer buffer;
     std::vector<BufferView> views;
     VideoCommon::UsageTracker tracker;
+    bool is_null{};
 };
 
 class QuadArrayIndexBuffer;
@@ -151,6 +152,7 @@ private:
     }
 
     void ReserveNullBuffer();
+    vk::Buffer CreateNullBuffer();
 
     const Device& device;
     MemoryAllocator& memory_allocator;

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -54,7 +54,7 @@ using VideoCommon::FileEnvironment;
 using VideoCommon::GenericEnvironment;
 using VideoCommon::GraphicsEnvironment;
 
-constexpr u32 CACHE_VERSION = 10;
+constexpr u32 CACHE_VERSION = 11;
 constexpr std::array<char, 8> VULKAN_CACHE_MAGIC_NUMBER{'y', 'u', 'z', 'u', 'v', 'k', 'c', 'h'};
 
 template <typename Container>
@@ -374,6 +374,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .has_broken_robust =
             device.IsNvidia() && device.GetNvidiaArch() <= NvidiaArchitecture::Arch_Pascal,
         .min_ssbo_alignment = device.GetStorageBufferAlignment(),
+        .max_user_clip_distances = device.GetMaxUserClipDistances(),
     };
 
     host_info = Shader::HostTranslateInfo{

src/video_core/renderer_vulkan/vk_present_manager.cpp

@@ -293,10 +293,10 @@ void PresentManager::RecreateSwapchain(Frame* frame) {
 }
 
 void PresentManager::SetImageCount() {
-    // We cannot have more than 5 images in flight at any given time.
-    // FRAMES_IN_FLIGHT is 7, and the cache TICKS_TO_DESTROY is 6.
+    // We cannot have more than 7 images in flight at any given time.
+    // FRAMES_IN_FLIGHT is 8, and the cache TICKS_TO_DESTROY is 8.
     // Mali drivers will give us 6.
-    image_count = std::min<size_t>(swapchain.GetImageCount(), 5);
+    image_count = std::min<size_t>(swapchain.GetImageCount(), 7);
 }
 
 void PresentManager::CopyToSwapchain(Frame* frame) {

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -289,12 +289,15 @@ public:
         }
 
         if (has_multi_queries) {
-            size_t intermediary_buffer_index = ObtainBuffer<false>(num_slots_used);
+            const size_t min_accumulation_limit =
+                std::min(first_accumulation_checkpoint, num_slots_used);
+            const size_t max_accumulation_limit =
+                std::max(last_accumulation_checkpoint, num_slots_used);
+            const size_t intermediary_buffer_index = ObtainBuffer<false>(num_slots_used);
             resolve_buffers.push_back(intermediary_buffer_index);
             queries_prefix_scan_pass->Run(*accumulation_buffer, *buffers[intermediary_buffer_index],
                                           *buffers[resolve_buffer_index], num_slots_used,
-                                          std::min(first_accumulation_checkpoint, num_slots_used),
-                                          last_accumulation_checkpoint);
+                                          min_accumulation_limit, max_accumulation_limit);
 
         } else {
             scheduler.RequestOutsideRenderPassOperationContext();

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -485,6 +485,10 @@ void RasterizerVulkan::DispatchCompute() {
 }
 
 void RasterizerVulkan::ResetCounter(VideoCommon::QueryType type) {
+    if (type != VideoCommon::QueryType::ZPassPixelCount64) {
+        LOG_DEBUG(Render_Vulkan, "Unimplemented counter reset={}", type);
+        return;
+    }
     query_cache.CounterReset(type);
 }
 

src/video_core/renderer_vulkan/vk_update_descriptor.h

@@ -31,7 +31,7 @@ struct DescriptorUpdateEntry {
 class UpdateDescriptorQueue final {
     // This should be plenty for the vast majority of cases. Most desktop platforms only
     // provide up to 3 swapchain images.
-    static constexpr size_t FRAMES_IN_FLIGHT = 7;
+    static constexpr size_t FRAMES_IN_FLIGHT = 8;
     static constexpr size_t FRAME_PAYLOAD_SIZE = 0x20000;
     static constexpr size_t PAYLOAD_SIZE = FRAME_PAYLOAD_SIZE * FRAMES_IN_FLIGHT;
 

src/video_core/shader_cache.cpp

@@ -132,7 +132,7 @@ void ShaderCache::Register(std::unique_ptr<ShaderInfo> data, VAddr addr, size_t
 
     storage.push_back(std::move(data));
 
-    rasterizer.UpdatePagesCachedCount(addr, size, true);
+    rasterizer.UpdatePagesCachedCount(addr, size, 1);
 }
 
 void ShaderCache::InvalidatePagesInRegion(VAddr addr, size_t size) {
@@ -209,7 +209,7 @@ void ShaderCache::UnmarkMemory(Entry* entry) {
 
     const VAddr addr = entry->addr_start;
     const size_t size = entry->addr_end - addr;
-    rasterizer.UpdatePagesCachedCount(addr, size, false);
+    rasterizer.UpdatePagesCachedCount(addr, size, -1);
 }
 
 void ShaderCache::RemoveShadersFromStorage(std::span<ShaderInfo*> removed_shaders) {

src/video_core/shader_environment.cpp

@@ -62,23 +62,9 @@ static Shader::TextureType ConvertTextureType(const Tegra::Texture::TICEntry& en
 }
 
 static Shader::TexturePixelFormat ConvertTexturePixelFormat(const Tegra::Texture::TICEntry& entry) {
-    switch (PixelFormatFromTextureInfo(entry.format, entry.r_type, entry.g_type, entry.b_type,
-                                       entry.a_type, entry.srgb_conversion)) {
-    case VideoCore::Surface::PixelFormat::A8B8G8R8_SNORM:
-        return Shader::TexturePixelFormat::A8B8G8R8_SNORM;
-    case VideoCore::Surface::PixelFormat::R8_SNORM:
-        return Shader::TexturePixelFormat::R8_SNORM;
-    case VideoCore::Surface::PixelFormat::R8G8_SNORM:
-        return Shader::TexturePixelFormat::R8G8_SNORM;
-    case VideoCore::Surface::PixelFormat::R16G16B16A16_SNORM:
-        return Shader::TexturePixelFormat::R16G16B16A16_SNORM;
-    case VideoCore::Surface::PixelFormat::R16G16_SNORM:
-        return Shader::TexturePixelFormat::R16G16_SNORM;
-    case VideoCore::Surface::PixelFormat::R16_SNORM:
-        return Shader::TexturePixelFormat::R16_SNORM;
-    default:
-        return Shader::TexturePixelFormat::OTHER;
-    }
+    return static_cast<Shader::TexturePixelFormat>(
+        PixelFormatFromTextureInfo(entry.format, entry.r_type, entry.g_type, entry.b_type,
+                                   entry.a_type, entry.srgb_conversion));
 }
 
 static std::string_view StageToPrefix(Shader::Stage stage) {
@@ -398,6 +384,11 @@ Shader::TexturePixelFormat GraphicsEnvironment::ReadTexturePixelFormat(u32 handl
     return result;
 }
 
+bool GraphicsEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 GraphicsEnvironment::ReadViewportTransformState() {
     const auto& regs{maxwell3d->regs};
     viewport_transform_state = regs.viewport_scale_offset_enabled;
@@ -448,6 +439,11 @@ Shader::TexturePixelFormat ComputeEnvironment::ReadTexturePixelFormat(u32 handle
     return result;
 }
 
+bool ComputeEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 ComputeEnvironment::ReadViewportTransformState() {
     return viewport_transform_state;
 }
@@ -551,6 +547,11 @@ Shader::TexturePixelFormat FileEnvironment::ReadTexturePixelFormat(u32 handle) {
     return it->second;
 }
 
+bool FileEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 FileEnvironment::ReadViewportTransformState() {
     return viewport_transform_state;
 }