kernel: Rewrite virtual memory management

.ci/ios.sh

@@ -1,10 +1,5 @@
 #!/bin/bash -ex
 
-# TODO: Work around pip install issues with Python 3.12 in the GitHub runner image.
-# See: https://github.com/actions/runner-images/issues/8709
-PYTHON_PATH=$(brew --prefix python@3.11)
-export PATH=$PYTHON_PATH/bin:$PYTHON_PATH/libexec/bin:$PATH
-
 mkdir build && cd build
 cmake .. -GNinja \
     -DCMAKE_BUILD_TYPE=Release \

.ci/macos.sh

@@ -1,10 +1,5 @@
 #!/bin/bash -ex
 
-# TODO: Work around pip install issues with Python 3.12 in the GitHub runner image.
-# See: https://github.com/actions/runner-images/issues/8709
-PYTHON_PATH=$(brew --prefix python@3.11)
-export PATH=$PYTHON_PATH/bin:$PYTHON_PATH/libexec/bin:$PATH
-
 mkdir build && cd build
 cmake .. -GNinja \
     -DCMAKE_BUILD_TYPE=Release \

.github/workflows/build.yml

@@ -59,7 +59,7 @@ jobs:
           name: ${{ env.OS }}-${{ env.TARGET }}
           path: artifacts/
   macos:
-    runs-on: macos-13
+    runs-on: macos-latest
     strategy:
       matrix:
         target: ["x86_64", "arm64"]
@@ -92,7 +92,7 @@ jobs:
           path: ${{ env.OS }}-${{ env.TARGET }}
           key: ${{ runner.os }}-${{ matrix.target }}-${{ github.sha }}-${{ github.run_id }}-${{ github.run_attempt }}
   macos-universal:
-    runs-on: macos-13
+    runs-on: macos-latest
     needs: macos
     env:
       OS: macos
@@ -234,7 +234,7 @@ jobs:
           name: ${{ env.OS }}-${{ env.TARGET }}
           path: src/android/app/artifacts/
   ios:
-    runs-on: macos-13
+    runs-on: macos-latest
     if: ${{ !startsWith(github.ref, 'refs/tags/') }}
     env:
       CCACHE_DIR: ${{ github.workspace }}/.ccache

CMakeLists.txt

@@ -236,7 +236,7 @@ find_package(Threads REQUIRED)
 
 if (ENABLE_QT)
     if (NOT USE_SYSTEM_QT)
-        download_qt(6.6.0)
+        download_qt(6.5.1)
     endif()
 
     find_package(Qt6 REQUIRED COMPONENTS Widgets Multimedia Concurrent)

dist/qt_themes/colorful_dark/icons/index.theme

@@ -2,13 +2,7 @@
 Name=colorful_dark
 Comment=Colorful theme (Dark style)
 Inherits=default
-Directories=16x16,48x48,256x256
+Directories=16x16
  
 [16x16]
 Size=16
-
-[48x48]
-Size=48
-
-[256x256]
-Size=256

dist/qt_themes/colorful_midnight_blue/icons/index.theme

@@ -2,13 +2,7 @@
 Name=colorful_midnight_blue
 Comment=Colorful theme (Midnight Blue style)
 Inherits=default
-Directories=16x16,48x48,256x256
+Directories=16x16
 
 [16x16]
 Size=16
-
-[48x48]
-Size=48
-
-[256x256]
-Size=256

externals/CMakeLists.txt

@@ -46,17 +46,11 @@ set(CATCH_INSTALL_EXTRAS OFF CACHE BOOL "")
 add_subdirectory(catch2)
 
 # Crypto++
-if(USE_SYSTEM_CRYPTOPP)
+set(CRYPTOPP_BUILD_DOCUMENTATION OFF CACHE BOOL "")
-    find_package(cryptopp REQUIRED)
+set(CRYPTOPP_BUILD_TESTING OFF CACHE BOOL "")
-    add_library(cryptopp INTERFACE)
+set(CRYPTOPP_INSTALL OFF CACHE BOOL "")
-    target_link_libraries(cryptopp INTERFACE cryptopp::cryptopp)
+set(CRYPTOPP_SOURCES "${CMAKE_SOURCE_DIR}/externals/cryptopp" CACHE STRING "")
-else()
+add_subdirectory(cryptopp-cmake)
-    set(CRYPTOPP_BUILD_DOCUMENTATION OFF CACHE BOOL "")
-    set(CRYPTOPP_BUILD_TESTING OFF CACHE BOOL "")
-    set(CRYPTOPP_INSTALL OFF CACHE BOOL "")
-    set(CRYPTOPP_SOURCES "${CMAKE_SOURCE_DIR}/externals/cryptopp" CACHE STRING "")
-    add_subdirectory(cryptopp-cmake)
-endif()
 
 # dds-ktx
 add_library(dds-ktx INTERFACE)
@@ -234,30 +228,15 @@ else()
 endif()
 
 # ENet
-if(USE_SYSTEM_ENET)
+add_subdirectory(enet)
-    find_package(libenet REQUIRED)
+target_include_directories(enet INTERFACE ./enet/include)
-    add_library(enet INTERFACE)
-    target_link_libraries(enet INTERFACE libenet::libenet)
-else()
-    add_subdirectory(enet)
-    target_include_directories(enet INTERFACE ./enet/include)
-endif()
 
 # Cubeb
 if (ENABLE_CUBEB)
-    if(USE_SYSTEM_CUBEB)
-        find_package(cubeb REQUIRED)
-        add_library(cubeb INTERFACE)
-        target_link_libraries(cubeb INTERFACE cubeb::cubeb)
-        if(TARGET cubeb::cubeb)
-            message(STATUS "Found system cubeb")
-        endif()
-    else()
     set(BUILD_TESTS OFF CACHE BOOL "")
     set(BUILD_TOOLS OFF CACHE BOOL "")
     set(BUNDLE_SPEEX ON CACHE BOOL "")
     add_subdirectory(cubeb EXCLUDE_FROM_ALL)
-    endif()
 endif()
 
 # DiscordRPC

externals/cmake-modules/CitraHandleSystemLibs.cmake

@@ -19,9 +19,6 @@ option(USE_SYSTEM_FDK_AAC_HEADERS "Use the system fdk-aac headers (instead of th
 option(USE_SYSTEM_FFMPEG_HEADERS "Use the system FFmpeg headers (instead of the bundled one)" OFF)
 option(USE_SYSTEM_GLSLANG "Use the system glslang and SPIR-V libraries (instead of the bundled ones)" OFF)
 option(USE_SYSTEM_ZSTD "Use the system Zstandard library (instead of the bundled one)" OFF)
-option(USE_SYSTEM_ENET "Use the system libenet (instead of the bundled one)" OFF)
-option(USE_SYSTEM_CRYPTOPP "Use the system cryptopp (instead of the bundled one)" OFF)
-option(USE_SYSTEM_CUBEB "Use the system cubeb (instead of the bundled one)" OFF)
 
 # Qt and MoltenVK are handled separately
 CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_SDL2 "Disable system SDL2" OFF "USE_SYSTEM_LIBS" OFF)
@@ -40,9 +37,6 @@ CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_FDK_AAC_HEADERS "Disable system fdk_aac" O
 CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_FFMPEG_HEADERS "Disable system ffmpeg" OFF "USE_SYSTEM_LIBS" OFF)
 CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_GLSLANG "Disable system glslang" OFF "USE_SYSTEM_LIBS" OFF)
 CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_ZSTD "Disable system Zstandard" OFF "USE_SYSTEM_LIBS" OFF)
-CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_ENET "Disable system libenet" OFF "USE_SYSTEM_LIBS" OFF)
-CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_CRYPTOPP "Disable system cryptopp" OFF "USE_SYSTEM_LIBS" OFF)
-CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_CUBEB "Disable system cubeb" OFF "USE_SYSTEM_LIBS" OFF)
 
 set(LIB_VAR_LIST
     SDL2
@@ -61,9 +55,6 @@ set(LIB_VAR_LIST
     FFMPEG_HEADERS
     GLSLANG
     ZSTD
-    ENET
-    CRYPTOPP
-    CUBEB
     )
 
 # First, check that USE_SYSTEM_XXX is not used with USE_SYSTEM_LIBS

externals/cmake-modules/Findcryptopp.cmake

@@ -1,36 +0,0 @@
-if(NOT CRYPTOPP_FOUND)
-    pkg_check_modules(CRYPTOPP_TMP libcrypto++)
-
-    find_path(CRYPTOPP_INCLUDE_DIRS NAMES cryptlib.h
-            PATHS
-            ${CRYPTOPP_TMP_INCLUDE_DIRS}
-            /usr/include
-            /usr/include/crypto++
-            /usr/local/include
-            /usr/local/include/crypto++
-            )
-
-    find_library(CRYPTOPP_LIBRARY_DIRS NAMES crypto++
-            PATHS
-            ${CRYPTOPP_TMP_LIBRARY_DIRS}
-            /usr/lib
-            /usr/locallib
-            )
-
-    if(CRYPTOPP_INCLUDE_DIRS AND CRYPTOPP_LIBRARY_DIRS)
-        set(CRYPTOPP_FOUND TRUE CACHE INTERNAL "Found cryptopp")
-        message(STATUS "Found cryptopp: ${CRYPTOPP_LIBRARY_DIRS}, ${CRYPTOPP_INCLUDE_DIRS}")
-    else()
-        set(CRYPTOPP_FOUND FALSE CACHE INTERNAL "Found cryptopp")
-        message(STATUS "Cryptopp not found.")
-    endif()
-endif()
-
-if(CRYPTOPP_FOUND AND NOT TARGET cryptopp::cryptopp)
-    add_library(cryptopp::cryptopp UNKNOWN IMPORTED)
-    set_target_properties(cryptopp::cryptopp PROPERTIES
-        INCLUDE_DIRECTORIES ${CRYPTOPP_INCLUDE_DIRS}
-        INTERFACE_LINK_LIBRARIES ${CRYPTOPP_LIBRARY_DIRS}
-        IMPORTED_LOCATION ${CRYPTOPP_LIBRARY_DIRS}
-        )
-endif()

externals/cmake-modules/Findlibenet.cmake

@@ -1,34 +0,0 @@
-if(NOT libenet_FOUND)
-    pkg_check_modules(ENET_TMP libenet)
-
-    find_path(libenet_INCLUDE_DIRS NAMES enet.h PATH_SUFFIXES enet
-            PATHS
-            ${ENET_TMP_INCLUDE_DIRS}
-            /usr/include
-            /usr/local/include
-            )
-
-    find_library(libenet_LIBRARY_DIRS NAMES enet
-            PATHS
-            ${ENET_TMP_LIBRARY_DIRS}
-            /usr/lib
-            /usr/local/lib
-            )
-
-    if(libenet_INCLUDE_DIRS AND libenet_LIBRARY_DIRS)
-        set(libenet_FOUND TRUE CACHE INTERNAL "Found libenet")
-        message(STATUS "Found libenet ${libenet_LIBRARY_DIRS}, ${libenet_INCLUDE_DIRS}")
-    else()
-        set(libenet_FOUND FALSE CACHE INTERNAL "Found libenet")
-        message(STATUS "Libenet not found.")
-    endif()
-endif()
-
-if(libenet_FOUND AND NOT TARGET libenet::libenet)
-    add_library(libenet::libenet UNKNOWN IMPORTED)
-    set_target_properties(libenet::libenet PROPERTIES
-            INCLUDE_DIRECTORIES ${libenet_INCLUDE_DIRS}
-            INTERFACE_LINK_LIBRARIES ${libenet_LIBRARY_DIRS}
-            IMPORTED_LOCATION ${libenet_LIBRARY_DIRS}
-            )
-endif()

src/citra_qt/game_list.cpp

@@ -258,15 +258,13 @@ void GameList::OnUpdateThemedIcons() {
     for (int i = 0; i < item_model->invisibleRootItem()->rowCount(); i++) {
         QStandardItem* child = item_model->invisibleRootItem()->child(i);
 
-        const int icon_size = IconSizes.at(UISettings::values.game_list_icon_size.GetValue());
         switch (child->data(GameListItem::TypeRole).value<GameListItemType>()) {
         case GameListItemType::InstalledDir:
-            child->setData(QIcon::fromTheme(QStringLiteral("sd_card")).pixmap(icon_size),
+            child->setData(QIcon::fromTheme(QStringLiteral("sd_card")).pixmap(48),
                            Qt::DecorationRole);
             break;
         case GameListItemType::SystemDir:
-            child->setData(QIcon::fromTheme(QStringLiteral("chip")).pixmap(icon_size),
+            child->setData(QIcon::fromTheme(QStringLiteral("chip")).pixmap(48), Qt::DecorationRole);
-                           Qt::DecorationRole);
             break;
         case GameListItemType::CustomDir: {
             const UISettings::GameDir& game_dir =
@@ -274,12 +272,11 @@ void GameList::OnUpdateThemedIcons() {
             const QString icon_name = QFileInfo::exists(game_dir.path)
                                           ? QStringLiteral("folder")
                                           : QStringLiteral("bad_folder");
-            child->setData(QIcon::fromTheme(icon_name).pixmap(icon_size), Qt::DecorationRole);
+            child->setData(QIcon::fromTheme(icon_name).pixmap(48), Qt::DecorationRole);
             break;
         }
         case GameListItemType::AddDir:
-            child->setData(QIcon::fromTheme(QStringLiteral("plus")).pixmap(icon_size),
+            child->setData(QIcon::fromTheme(QStringLiteral("plus")).pixmap(48), Qt::DecorationRole);
-                           Qt::DecorationRole);
             break;
         default:
             break;

src/citra_qt/main.cpp

@@ -2741,10 +2741,7 @@ void GMainWindow::filterBarSetChecked(bool state) {
 }
 
 void GMainWindow::UpdateUITheme() {
-    const QString icons_base_path = QStringLiteral(":/icons/");
+    const QString default_icons = QStringLiteral(":/icons/default");
-    const QString default_theme = QStringLiteral("default");
-    const QString default_theme_path = icons_base_path + default_theme;
-
     const QString& current_theme = UISettings::values.theme;
     const bool is_default_theme = current_theme == QString::fromUtf8(UISettings::themes[0].second);
     QStringList theme_paths(default_theme_paths);
@@ -2762,8 +2759,8 @@ void GMainWindow::UpdateUITheme() {
             qApp->setStyleSheet({});
             setStyleSheet({});
         }
-        theme_paths.append(default_theme_path);
+        theme_paths.append(default_icons);
-        QIcon::setThemeName(default_theme);
+        QIcon::setThemeName(default_icons);
     } else {
         const QString theme_uri(QLatin1Char{':'} + current_theme + QStringLiteral("/style.qss"));
         QFile f(theme_uri);
@@ -2775,9 +2772,9 @@ void GMainWindow::UpdateUITheme() {
             LOG_ERROR(Frontend, "Unable to set style, stylesheet file not found");
         }
 
-        const QString current_theme_path = icons_base_path + current_theme;
+        const QString theme_name = QStringLiteral(":/icons/") + current_theme;
-        theme_paths.append({default_theme_path, current_theme_path});
+        theme_paths.append({default_icons, theme_name});
-        QIcon::setThemeName(current_theme);
+        QIcon::setThemeName(theme_name);
     }
 
     QIcon::setThemeSearchPaths(theme_paths);

src/citra_qt/updater/updater.cpp

@@ -17,7 +17,7 @@
 #include "common/file_util.h"
 #include "common/logging/log.h"
 
-#ifdef Q_OS_MACOS
+#ifdef Q_OS_OSX
 #define DEFAULT_TOOL_PATH QStringLiteral("../../../../maintenancetool")
 #else
 #define DEFAULT_TOOL_PATH QStringLiteral("../maintenancetool")
@@ -102,7 +102,7 @@ QString UpdaterPrivate::ToSystemExe(QString base_path) {
         return base_path + QStringLiteral(".exe");
     else
         return base_path;
-#elif defined(Q_OS_MACOS)
+#elif defined(Q_OS_OSX)
     if (base_path.endsWith(QStringLiteral(".app")))
         base_path.truncate(base_path.lastIndexOf(QStringLiteral(".")));
     return base_path + QStringLiteral(".app/Contents/MacOS/") + QFileInfo(base_path).fileName();
@@ -112,7 +112,7 @@ QString UpdaterPrivate::ToSystemExe(QString base_path) {
 }
 
 QFileInfo UpdaterPrivate::GetMaintenanceTool() const {
-#if defined(Q_OS_UNIX) && !defined(Q_OS_MACOS)
+#if defined(Q_OS_UNIX) && !defined(Q_OS_OSX)
     const auto appimage_path = QProcessEnvironment::systemEnvironment()
                                    .value(QStringLiteral("APPIMAGE"), {})
                                    .toStdString();

src/common/CMakeLists.txt

@@ -86,6 +86,7 @@ add_library(citra_common STATIC
     file_util.cpp
     file_util.h
     hash.h
+    intrusive_list.h
     linear_disk_cache.h
     literals.h
     logging/backend.cpp
@@ -107,8 +108,11 @@ add_library(citra_common STATIC
     microprofile.h
     microprofileui.h
     misc.cpp
+    page_table.cpp
+    page_table.h
     param_package.cpp
     param_package.h
+    parent_of_member.h
     polyfill_thread.h
     precompiled_headers.h
     quaternion.h
@@ -124,7 +128,6 @@ add_library(citra_common STATIC
     serialization/boost_flat_set.h
     serialization/boost_small_vector.hpp
     serialization/boost_vector.hpp
-    static_lru_cache.h
     string_literal.h
     string_util.cpp
     string_util.h

src/common/common_funcs.h

@@ -110,6 +110,14 @@ __declspec(dllimport) void __stdcall DebugBreak(void);
         return static_cast<T>(key) == 0;                                                           \
     }
 
+#define CITRA_NON_COPYABLE(cls)                                                                     \
+    cls(const cls&) = delete;                                                                       \
+    cls& operator=(const cls&) = delete
+
+#define CITRA_NON_MOVEABLE(cls)                                                                     \
+    cls(cls&&) = delete;                                                                            \
+    cls& operator=(cls&&) = delete
+
 // Generic function to get last error message.
 // Call directly after the command or use the error num.
 // This function might change the error code.

src/common/file_util.cpp

@@ -1155,43 +1155,6 @@ std::size_t IOFile::ReadImpl(void* data, std::size_t length, std::size_t data_si
     return std::fread(data, data_size, length, m_file);
 }
 
-#ifdef _WIN32
-static std::size_t pread(int fd, void* buf, size_t count, uint64_t offset) {
-    long unsigned int read_bytes = 0;
-    OVERLAPPED overlapped = {0};
-    HANDLE file = reinterpret_cast<HANDLE>(_get_osfhandle(fd));
-
-    overlapped.OffsetHigh = static_cast<uint32_t>(offset >> 32);
-    overlapped.Offset = static_cast<uint32_t>(offset & 0xFFFF'FFFFLL);
-    SetLastError(0);
-    bool ret = ReadFile(file, buf, static_cast<uint32_t>(count), &read_bytes, &overlapped);
-
-    if (!ret && GetLastError() != ERROR_HANDLE_EOF) {
-        errno = GetLastError();
-        return std::numeric_limits<std::size_t>::max();
-    }
-    return read_bytes;
-}
-#else
-#define pread ::pread
-#endif
-
-std::size_t IOFile::ReadAtImpl(void* data, std::size_t length, std::size_t data_size,
-                               std::size_t offset) {
-    if (!IsOpen()) {
-        m_good = false;
-        return std::numeric_limits<std::size_t>::max();
-    }
-
-    if (length == 0) {
-        return 0;
-    }
-
-    DEBUG_ASSERT(data != nullptr);
-
-    return pread(fileno(m_file), data, data_size * length, offset);
-}
-
 std::size_t IOFile::WriteImpl(const void* data, std::size_t length, std::size_t data_size) {
     if (!IsOpen()) {
         m_good = false;

src/common/file_util.h

@@ -294,18 +294,6 @@ public:
         return items_read;
     }
 
-    template <typename T>
-    std::size_t ReadAtArray(T* data, std::size_t length, std::size_t offset) {
-        static_assert(std::is_trivially_copyable_v<T>,
-                      "Given array does not consist of trivially copyable objects");
-
-        std::size_t items_read = ReadAtImpl(data, length, sizeof(T), offset);
-        if (items_read != length)
-            m_good = false;
-
-        return items_read;
-    }
-
     template <typename T>
     std::size_t WriteArray(const T* data, std::size_t length) {
         static_assert(std::is_trivially_copyable_v<T>,
@@ -324,12 +312,6 @@ public:
         return ReadArray(reinterpret_cast<char*>(data), length);
     }
 
-    template <typename T>
-    std::size_t ReadAtBytes(T* data, std::size_t length, std::size_t offset) {
-        static_assert(std::is_trivially_copyable_v<T>, "T must be trivially copyable");
-        return ReadAtArray(reinterpret_cast<char*>(data), length, offset);
-    }
-
     template <typename T>
     std::size_t WriteBytes(const T* data, std::size_t length) {
         static_assert(std::is_trivially_copyable_v<T>, "T must be trivially copyable");
@@ -381,8 +363,6 @@ public:
 
 private:
     std::size_t ReadImpl(void* data, std::size_t length, std::size_t data_size);
-    std::size_t ReadAtImpl(void* data, std::size_t length, std::size_t data_size,
-                           std::size_t offset);
     std::size_t WriteImpl(const void* data, std::size_t length, std::size_t data_size);
 
     bool Open();

src/common/intrusive_list.h

@@ -0,0 +1,631 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "common/parent_of_member.h"
+
+namespace Common {
+
+// Forward declare implementation class for Node.
+namespace impl {
+
+class IntrusiveListImpl;
+
+}
+
+class IntrusiveListNode {
+    CITRA_NON_COPYABLE(IntrusiveListNode);
+
+private:
+    friend class impl::IntrusiveListImpl;
+
+    IntrusiveListNode* m_prev;
+    IntrusiveListNode* m_next;
+
+public:
+    constexpr IntrusiveListNode() : m_prev(this), m_next(this) {}
+
+    constexpr bool IsLinked() const {
+        return m_next != this;
+    }
+
+private:
+    constexpr void LinkPrev(IntrusiveListNode* node) {
+        // We can't link an already linked node.
+        ASSERT(!node->IsLinked());
+        this->SplicePrev(node, node);
+    }
+
+    constexpr void SplicePrev(IntrusiveListNode* first, IntrusiveListNode* last) {
+        // Splice a range into the list.
+        auto last_prev = last->m_prev;
+        first->m_prev = m_prev;
+        last_prev->m_next = this;
+        m_prev->m_next = first;
+        m_prev = last_prev;
+    }
+
+    constexpr void LinkNext(IntrusiveListNode* node) {
+        // We can't link an already linked node.
+        ASSERT(!node->IsLinked());
+        return this->SpliceNext(node, node);
+    }
+
+    constexpr void SpliceNext(IntrusiveListNode* first, IntrusiveListNode* last) {
+        // Splice a range into the list.
+        auto last_prev = last->m_prev;
+        first->m_prev = this;
+        last_prev->m_next = m_next;
+        m_next->m_prev = last_prev;
+        m_next = first;
+    }
+
+    constexpr void Unlink() {
+        this->Unlink(m_next);
+    }
+
+    constexpr void Unlink(IntrusiveListNode* last) {
+        // Unlink a node from a next node.
+        auto last_prev = last->m_prev;
+        m_prev->m_next = last;
+        last->m_prev = m_prev;
+        last_prev->m_next = this;
+        m_prev = last_prev;
+    }
+
+    constexpr IntrusiveListNode* GetPrev() {
+        return m_prev;
+    }
+
+    constexpr const IntrusiveListNode* GetPrev() const {
+        return m_prev;
+    }
+
+    constexpr IntrusiveListNode* GetNext() {
+        return m_next;
+    }
+
+    constexpr const IntrusiveListNode* GetNext() const {
+        return m_next;
+    }
+};
+// DEPRECATED: static_assert(std::is_literal_type<IntrusiveListNode>::value);
+
+namespace impl {
+
+class IntrusiveListImpl {
+    CITRA_NON_COPYABLE(IntrusiveListImpl);
+
+private:
+    IntrusiveListNode m_root_node;
+
+public:
+    template <bool Const>
+    class Iterator;
+
+    using value_type = IntrusiveListNode;
+    using size_type = size_t;
+    using difference_type = ptrdiff_t;
+    using pointer = value_type*;
+    using const_pointer = const value_type*;
+    using reference = value_type&;
+    using const_reference = const value_type&;
+    using iterator = Iterator<false>;
+    using const_iterator = Iterator<true>;
+    using reverse_iterator = std::reverse_iterator<iterator>;
+    using const_reverse_iterator = std::reverse_iterator<const_iterator>;
+
+    template <bool Const>
+    class Iterator {
+    public:
+        using iterator_category = std::bidirectional_iterator_tag;
+        using value_type = typename IntrusiveListImpl::value_type;
+        using difference_type = typename IntrusiveListImpl::difference_type;
+        using pointer =
+            std::conditional_t<Const, IntrusiveListImpl::const_pointer, IntrusiveListImpl::pointer>;
+        using reference = std::conditional_t<Const, IntrusiveListImpl::const_reference,
+                                             IntrusiveListImpl::reference>;
+
+    private:
+        pointer m_node;
+
+    public:
+        constexpr explicit Iterator(pointer n) : m_node(n) {}
+
+        constexpr bool operator==(const Iterator& rhs) const {
+            return m_node == rhs.m_node;
+        }
+
+        constexpr pointer operator->() const {
+            return m_node;
+        }
+
+        constexpr reference operator*() const {
+            return *m_node;
+        }
+
+        constexpr Iterator& operator++() {
+            m_node = m_node->m_next;
+            return *this;
+        }
+
+        constexpr Iterator& operator--() {
+            m_node = m_node->m_prev;
+            return *this;
+        }
+
+        constexpr Iterator operator++(int) {
+            const Iterator it{*this};
+            ++(*this);
+            return it;
+        }
+
+        constexpr Iterator operator--(int) {
+            const Iterator it{*this};
+            --(*this);
+            return it;
+        }
+
+        constexpr operator Iterator<true>() const {
+            return Iterator<true>(m_node);
+        }
+
+        constexpr Iterator<false> GetNonConstIterator() const {
+            return Iterator<false>(const_cast<IntrusiveListImpl::pointer>(m_node));
+        }
+    };
+
+public:
+    constexpr IntrusiveListImpl() : m_root_node() {}
+
+    // Iterator accessors.
+    constexpr iterator begin() {
+        return iterator(m_root_node.GetNext());
+    }
+
+    constexpr const_iterator begin() const {
+        return const_iterator(m_root_node.GetNext());
+    }
+
+    constexpr iterator end() {
+        return iterator(std::addressof(m_root_node));
+    }
+
+    constexpr const_iterator end() const {
+        return const_iterator(std::addressof(m_root_node));
+    }
+
+    constexpr iterator iterator_to(reference v) {
+        // Only allow iterator_to for values in lists.
+        ASSERT(v.IsLinked());
+        return iterator(std::addressof(v));
+    }
+
+    constexpr const_iterator iterator_to(const_reference v) const {
+        // Only allow iterator_to for values in lists.
+        ASSERT(v.IsLinked());
+        return const_iterator(std::addressof(v));
+    }
+
+    // Content management.
+    constexpr bool empty() const {
+        return !m_root_node.IsLinked();
+    }
+
+    constexpr size_type size() const {
+        return static_cast<size_type>(std::distance(this->begin(), this->end()));
+    }
+
+    constexpr reference back() {
+        return *m_root_node.GetPrev();
+    }
+
+    constexpr const_reference back() const {
+        return *m_root_node.GetPrev();
+    }
+
+    constexpr reference front() {
+        return *m_root_node.GetNext();
+    }
+
+    constexpr const_reference front() const {
+        return *m_root_node.GetNext();
+    }
+
+    constexpr void push_back(reference node) {
+        m_root_node.LinkPrev(std::addressof(node));
+    }
+
+    constexpr void push_front(reference node) {
+        m_root_node.LinkNext(std::addressof(node));
+    }
+
+    constexpr void pop_back() {
+        m_root_node.GetPrev()->Unlink();
+    }
+
+    constexpr void pop_front() {
+        m_root_node.GetNext()->Unlink();
+    }
+
+    constexpr iterator insert(const_iterator pos, reference node) {
+        pos.GetNonConstIterator()->LinkPrev(std::addressof(node));
+        return iterator(std::addressof(node));
+    }
+
+    constexpr void splice(const_iterator pos, IntrusiveListImpl& o) {
+        splice_impl(pos, o.begin(), o.end());
+    }
+
+    constexpr void splice(const_iterator pos, IntrusiveListImpl&, const_iterator first) {
+        const_iterator last(first);
+        std::advance(last, 1);
+        splice_impl(pos, first, last);
+    }
+
+    constexpr void splice(const_iterator pos, IntrusiveListImpl&, const_iterator first,
+                          const_iterator last) {
+        splice_impl(pos, first, last);
+    }
+
+    constexpr iterator erase(const_iterator pos) {
+        if (pos == this->end()) {
+            return this->end();
+        }
+        iterator it(pos.GetNonConstIterator());
+        (it++)->Unlink();
+        return it;
+    }
+
+    constexpr void clear() {
+        while (!this->empty()) {
+            this->pop_front();
+        }
+    }
+
+private:
+    constexpr void splice_impl(const_iterator _pos, const_iterator _first, const_iterator _last) {
+        if (_first == _last) {
+            return;
+        }
+        iterator pos(_pos.GetNonConstIterator());
+        iterator first(_first.GetNonConstIterator());
+        iterator last(_last.GetNonConstIterator());
+        first->Unlink(std::addressof(*last));
+        pos->SplicePrev(std::addressof(*first), std::addressof(*first));
+    }
+};
+
+} // namespace impl
+
+template <class T, class Traits>
+class IntrusiveList {
+    CITRA_NON_COPYABLE(IntrusiveList);
+
+private:
+    impl::IntrusiveListImpl m_impl;
+
+public:
+    template <bool Const>
+    class Iterator;
+
+    using value_type = T;
+    using size_type = size_t;
+    using difference_type = ptrdiff_t;
+    using pointer = value_type*;
+    using const_pointer = const value_type*;
+    using reference = value_type&;
+    using const_reference = const value_type&;
+    using iterator = Iterator<false>;
+    using const_iterator = Iterator<true>;
+    using reverse_iterator = std::reverse_iterator<iterator>;
+    using const_reverse_iterator = std::reverse_iterator<const_iterator>;
+
+    template <bool Const>
+    class Iterator {
+    public:
+        friend class Common::IntrusiveList<T, Traits>;
+
+        using ImplIterator =
+            std::conditional_t<Const, Common::impl::IntrusiveListImpl::const_iterator,
+                               Common::impl::IntrusiveListImpl::iterator>;
+
+        using iterator_category = std::bidirectional_iterator_tag;
+        using value_type = typename IntrusiveList::value_type;
+        using difference_type = typename IntrusiveList::difference_type;
+        using pointer =
+            std::conditional_t<Const, IntrusiveList::const_pointer, IntrusiveList::pointer>;
+        using reference =
+            std::conditional_t<Const, IntrusiveList::const_reference, IntrusiveList::reference>;
+
+    private:
+        ImplIterator m_iterator;
+
+    private:
+        constexpr explicit Iterator(ImplIterator it) : m_iterator(it) {}
+
+        constexpr ImplIterator GetImplIterator() const {
+            return m_iterator;
+        }
+
+    public:
+        constexpr bool operator==(const Iterator& rhs) const {
+            return m_iterator == rhs.m_iterator;
+        }
+
+        constexpr pointer operator->() const {
+            return std::addressof(Traits::GetParent(*m_iterator));
+        }
+
+        constexpr reference operator*() const {
+            return Traits::GetParent(*m_iterator);
+        }
+
+        constexpr Iterator& operator++() {
+            ++m_iterator;
+            return *this;
+        }
+
+        constexpr Iterator& operator--() {
+            --m_iterator;
+            return *this;
+        }
+
+        constexpr Iterator operator++(int) {
+            const Iterator it{*this};
+            ++m_iterator;
+            return it;
+        }
+
+        constexpr Iterator operator--(int) {
+            const Iterator it{*this};
+            --m_iterator;
+            return it;
+        }
+
+        constexpr operator Iterator<true>() const {
+            return Iterator<true>(m_iterator);
+        }
+    };
+
+private:
+    static constexpr IntrusiveListNode& GetNode(reference ref) {
+        return Traits::GetNode(ref);
+    }
+
+    static constexpr IntrusiveListNode const& GetNode(const_reference ref) {
+        return Traits::GetNode(ref);
+    }
+
+    static constexpr reference GetParent(IntrusiveListNode& node) {
+        return Traits::GetParent(node);
+    }
+
+    static constexpr const_reference GetParent(IntrusiveListNode const& node) {
+        return Traits::GetParent(node);
+    }
+
+public:
+    constexpr IntrusiveList() : m_impl() {}
+
+    // Iterator accessors.
+    constexpr iterator begin() {
+        return iterator(m_impl.begin());
+    }
+
+    constexpr const_iterator begin() const {
+        return const_iterator(m_impl.begin());
+    }
+
+    constexpr iterator end() {
+        return iterator(m_impl.end());
+    }
+
+    constexpr const_iterator end() const {
+        return const_iterator(m_impl.end());
+    }
+
+    constexpr const_iterator cbegin() const {
+        return this->begin();
+    }
+
+    constexpr const_iterator cend() const {
+        return this->end();
+    }
+
+    constexpr reverse_iterator rbegin() {
+        return reverse_iterator(this->end());
+    }
+
+    constexpr const_reverse_iterator rbegin() const {
+        return const_reverse_iterator(this->end());
+    }
+
+    constexpr reverse_iterator rend() {
+        return reverse_iterator(this->begin());
+    }
+
+    constexpr const_reverse_iterator rend() const {
+        return const_reverse_iterator(this->begin());
+    }
+
+    constexpr const_reverse_iterator crbegin() const {
+        return this->rbegin();
+    }
+
+    constexpr const_reverse_iterator crend() const {
+        return this->rend();
+    }
+
+    constexpr iterator iterator_to(reference v) {
+        return iterator(m_impl.iterator_to(GetNode(v)));
+    }
+
+    constexpr const_iterator iterator_to(const_reference v) const {
+        return const_iterator(m_impl.iterator_to(GetNode(v)));
+    }
+
+    // Content management.
+    constexpr bool empty() const {
+        return m_impl.empty();
+    }
+
+    constexpr size_type size() const {
+        return m_impl.size();
+    }
+
+    constexpr reference back() {
+        return GetParent(m_impl.back());
+    }
+
+    constexpr const_reference back() const {
+        return GetParent(m_impl.back());
+    }
+
+    constexpr reference front() {
+        return GetParent(m_impl.front());
+    }
+
+    constexpr const_reference front() const {
+        return GetParent(m_impl.front());
+    }
+
+    constexpr void push_back(reference ref) {
+        m_impl.push_back(GetNode(ref));
+    }
+
+    constexpr void push_front(reference ref) {
+        m_impl.push_front(GetNode(ref));
+    }
+
+    constexpr void pop_back() {
+        m_impl.pop_back();
+    }
+
+    constexpr void pop_front() {
+        m_impl.pop_front();
+    }
+
+    constexpr iterator insert(const_iterator pos, reference ref) {
+        return iterator(m_impl.insert(pos.GetImplIterator(), GetNode(ref)));
+    }
+
+    constexpr void splice(const_iterator pos, IntrusiveList& o) {
+        m_impl.splice(pos.GetImplIterator(), o.m_impl);
+    }
+
+    constexpr void splice(const_iterator pos, IntrusiveList& o, const_iterator first) {
+        m_impl.splice(pos.GetImplIterator(), o.m_impl, first.GetImplIterator());
+    }
+
+    constexpr void splice(const_iterator pos, IntrusiveList& o, const_iterator first,
+                          const_iterator last) {
+        m_impl.splice(pos.GetImplIterator(), o.m_impl, first.GetImplIterator(),
+                      last.GetImplIterator());
+    }
+
+    constexpr iterator erase(const_iterator pos) {
+        return iterator(m_impl.erase(pos.GetImplIterator()));
+    }
+
+    constexpr void clear() {
+        m_impl.clear();
+    }
+};
+
+template <auto T, class Derived = Common::impl::GetParentType<T>>
+class IntrusiveListMemberTraits;
+
+template <class Parent, IntrusiveListNode Parent::*Member, class Derived>
+class IntrusiveListMemberTraits<Member, Derived> {
+public:
+    using ListType = IntrusiveList<Derived, IntrusiveListMemberTraits>;
+
+private:
+    friend class IntrusiveList<Derived, IntrusiveListMemberTraits>;
+
+    static constexpr IntrusiveListNode& GetNode(Derived& parent) {
+        return parent.*Member;
+    }
+
+    static constexpr IntrusiveListNode const& GetNode(Derived const& parent) {
+        return parent.*Member;
+    }
+
+    static Derived& GetParent(IntrusiveListNode& node) {
+        return Common::GetParentReference<Member, Derived>(std::addressof(node));
+    }
+
+    static Derived const& GetParent(IntrusiveListNode const& node) {
+        return Common::GetParentReference<Member, Derived>(std::addressof(node));
+    }
+};
+
+template <auto T, class Derived = Common::impl::GetParentType<T>>
+class IntrusiveListMemberTraitsByNonConstexprOffsetOf;
+
+template <class Parent, IntrusiveListNode Parent::*Member, class Derived>
+class IntrusiveListMemberTraitsByNonConstexprOffsetOf<Member, Derived> {
+public:
+    using ListType = IntrusiveList<Derived, IntrusiveListMemberTraitsByNonConstexprOffsetOf>;
+
+private:
+    friend class IntrusiveList<Derived, IntrusiveListMemberTraitsByNonConstexprOffsetOf>;
+
+    static constexpr IntrusiveListNode& GetNode(Derived& parent) {
+        return parent.*Member;
+    }
+
+    static constexpr IntrusiveListNode const& GetNode(Derived const& parent) {
+        return parent.*Member;
+    }
+
+    static Derived& GetParent(IntrusiveListNode& node) {
+        return *reinterpret_cast<Derived*>(reinterpret_cast<char*>(std::addressof(node)) -
+                                           GetOffset());
+    }
+
+    static Derived const& GetParent(IntrusiveListNode const& node) {
+        return *reinterpret_cast<const Derived*>(
+            reinterpret_cast<const char*>(std::addressof(node)) - GetOffset());
+    }
+
+    static uintptr_t GetOffset() {
+        return reinterpret_cast<uintptr_t>(std::addressof(reinterpret_cast<Derived*>(0)->*Member));
+    }
+};
+
+template <class Derived>
+class IntrusiveListBaseNode : public IntrusiveListNode {};
+
+template <class Derived>
+class IntrusiveListBaseTraits {
+public:
+    using ListType = IntrusiveList<Derived, IntrusiveListBaseTraits>;
+
+private:
+    friend class IntrusiveList<Derived, IntrusiveListBaseTraits>;
+
+    static constexpr IntrusiveListNode& GetNode(Derived& parent) {
+        return static_cast<IntrusiveListNode&>(
+            static_cast<IntrusiveListBaseNode<Derived>&>(parent));
+    }
+
+    static constexpr IntrusiveListNode const& GetNode(Derived const& parent) {
+        return static_cast<const IntrusiveListNode&>(
+            static_cast<const IntrusiveListBaseNode<Derived>&>(parent));
+    }
+
+    static constexpr Derived& GetParent(IntrusiveListNode& node) {
+        return static_cast<Derived&>(static_cast<IntrusiveListBaseNode<Derived>&>(node));
+    }
+
+    static constexpr Derived const& GetParent(IntrusiveListNode const& node) {
+        return static_cast<const Derived&>(
+            static_cast<const IntrusiveListBaseNode<Derived>&>(node));
+    }
+};
+
+} // namespace Common

src/common/page_table.cpp

@@ -0,0 +1,64 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/page_table.h"
+
+namespace Common {
+
+PageTable::PageTable() = default;
+
+PageTable::~PageTable() noexcept = default;
+
+bool PageTable::BeginTraversal(TraversalEntry* out_entry, TraversalContext* out_context, VAddr address) const {
+    // Setup invalid defaults.
+    out_entry->phys_addr = 0;
+    out_entry->block_size = page_size;
+    out_context->next_page = 0;
+
+    // Validate that we can read the actual entry.
+    const auto page = address / page_size;
+    if (page >= backing_addr.size()) {
+        return false;
+    }
+
+    // Validate that the entry is mapped.
+    const auto phys_addr = backing_addr[page];
+    if (phys_addr == 0) {
+        return false;
+    }
+
+    // Populate the results.
+    out_entry->phys_addr = phys_addr + address;
+    out_context->next_page = page + 1;
+    out_context->next_offset = address + page_size;
+
+    return true;
+}
+
+bool PageTable::ContinueTraversal(TraversalEntry* out_entry, TraversalContext* context) const {
+    // Setup invalid defaults.
+    out_entry->phys_addr = 0;
+    out_entry->block_size = page_size;
+
+    // Validate that we can read the actual entry.
+    const auto page = context->next_page;
+    if (page >= backing_addr.size()) {
+        return false;
+    }
+
+    // Validate that the entry is mapped.
+    const auto phys_addr = backing_addr[page];
+    if (phys_addr == 0) {
+        return false;
+    }
+
+    // Populate the results.
+    out_entry->phys_addr = phys_addr + context->next_offset;
+    context->next_page = page + 1;
+    context->next_offset += page_size;
+
+    return true;
+}
+
+} // namespace Common

src/common/page_table.h

@@ -0,0 +1,116 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Common {
+
+enum class PageType : u8 {
+    /// Page is unmapped and should cause an access error.
+    Unmapped,
+    /// Page is mapped to regular memory. This is the only type you can get pointers to.
+    Memory,
+    /// Page is mapped to regular memory, but also needs to check for rasterizer cache flushing and
+    /// invalidation
+    RasterizerCachedMemory,
+};
+
+/**
+ * A (reasonably) fast way of allowing switchable and remappable process address spaces. It loosely
+ * mimics the way a real CPU page table works.
+ */
+struct PageTable {
+    struct TraversalEntry {
+        u64 phys_addr{};
+        std::size_t block_size{};
+    };
+
+    struct TraversalContext {
+        u64 next_page{};
+        u64 next_offset{};
+    };
+
+    /// Number of bits reserved for attribute tagging.
+    /// This can be at most the guaranteed alignment of the pointers in the page table.
+    static constexpr int ATTRIBUTE_BITS = 2;
+    static constexpr size_t PAGE_BITS = 12;
+    static constexpr size_t NUM_ENTRIES = 1 << (32 - PAGE_BITS);
+
+    /**
+     * Pair of host pointer and page type attribute.
+     * This uses the lower bits of a given pointer to store the attribute tag.
+     * Writing and reading the pointer attribute pair is guaranteed to be atomic for the same method
+     * call. In other words, they are guaranteed to be synchronized at all times.
+     */
+    class PageInfo {
+    public:
+        /// Returns the page pointer
+        [[nodiscard]] uintptr_t Pointer() const noexcept {
+            return ExtractPointer(raw);
+        }
+
+        /// Returns the page type attribute
+        [[nodiscard]] PageType Type() const noexcept {
+            return ExtractType(raw);
+        }
+
+        /// Returns the page pointer and attribute pair, extracted from the same atomic read
+        [[nodiscard]] std::pair<uintptr_t, PageType> PointerType() const noexcept {
+            return {ExtractPointer(raw), ExtractType(raw)};
+        }
+
+        /// Returns the raw representation of the page information.
+        /// Use ExtractPointer and ExtractType to unpack the value.
+        [[nodiscard]] uintptr_t Raw() const noexcept {
+            return raw;
+        }
+
+        /// Write a page pointer and type pair atomically
+        void Store(uintptr_t pointer, PageType type) noexcept {
+            raw = pointer | static_cast<uintptr_t>(type);
+        }
+
+        /// Unpack a pointer from a page info raw representation
+        [[nodiscard]] static uintptr_t ExtractPointer(uintptr_t raw) noexcept {
+            return raw & (~uintptr_t{0} << ATTRIBUTE_BITS);
+        }
+
+        /// Unpack a page type from a page info raw representation
+        [[nodiscard]] static PageType ExtractType(uintptr_t raw) noexcept {
+            return static_cast<PageType>(raw & ((uintptr_t{1} << ATTRIBUTE_BITS) - 1));
+        }
+
+    private:
+        uintptr_t raw;
+    };
+
+    PageTable();
+    ~PageTable() noexcept;
+
+    PageTable(const PageTable&) = delete;
+    PageTable& operator=(const PageTable&) = delete;
+
+    PageTable(PageTable&&) noexcept = default;
+    PageTable& operator=(PageTable&&) noexcept = default;
+
+    bool BeginTraversal(TraversalEntry* out_entry, TraversalContext* out_context, VAddr address) const;
+    bool ContinueTraversal(TraversalEntry* out_entry, TraversalContext* context) const;
+
+    PAddr GetPhysicalAddress(VAddr virt_addr) const {
+        return backing_addr[virt_addr / page_size] + virt_addr;
+    }
+
+    /**
+     * Vector of memory pointers backing each page. An entry can only be non-null if the
+     * corresponding attribute element is of type `Memory`.
+     */
+    std::array<PageInfo, NUM_ENTRIES> pointers;
+    std::array<u64, NUM_ENTRIES> blocks;
+    std::array<u64, NUM_ENTRIES> backing_addr;
+    std::size_t page_size{};
+};
+
+} // namespace Common

src/common/parent_of_member.h

@@ -0,0 +1,190 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <type_traits>
+
+#include "common/assert.h"
+
+namespace Common {
+namespace detail {
+template <typename T, size_t Size, size_t Align>
+struct TypedStorageImpl {
+    alignas(Align) u8 storage_[Size];
+};
+} // namespace detail
+
+template <typename T>
+using TypedStorage = detail::TypedStorageImpl<T, sizeof(T), alignof(T)>;
+
+template <typename T>
+static constexpr T* GetPointer(TypedStorage<T>& ts) {
+    return static_cast<T*>(static_cast<void*>(std::addressof(ts.storage_)));
+}
+
+template <typename T>
+static constexpr const T* GetPointer(const TypedStorage<T>& ts) {
+    return static_cast<const T*>(static_cast<const void*>(std::addressof(ts.storage_)));
+}
+
+namespace impl {
+
+template <size_t MaxDepth>
+struct OffsetOfUnionHolder {
+    template <typename ParentType, typename MemberType, size_t Offset>
+    union UnionImpl {
+        using PaddingMember = char;
+        static constexpr size_t GetOffset() {
+            return Offset;
+        }
+
+#pragma pack(push, 1)
+        struct {
+            PaddingMember padding[Offset];
+            MemberType members[(sizeof(ParentType) / sizeof(MemberType)) + 1];
+        } data;
+#pragma pack(pop)
+        UnionImpl<ParentType, MemberType, Offset + 1> next_union;
+    };
+
+    template <typename ParentType, typename MemberType>
+    union UnionImpl<ParentType, MemberType, 0> {
+        static constexpr size_t GetOffset() {
+            return 0;
+        }
+
+        struct {
+            MemberType members[(sizeof(ParentType) / sizeof(MemberType)) + 1];
+        } data;
+        UnionImpl<ParentType, MemberType, 1> next_union;
+    };
+
+    template <typename ParentType, typename MemberType>
+    union UnionImpl<ParentType, MemberType, MaxDepth> {};
+};
+
+template <typename ParentType, typename MemberType>
+struct OffsetOfCalculator {
+    using UnionHolder =
+        typename OffsetOfUnionHolder<sizeof(MemberType)>::template UnionImpl<ParentType, MemberType,
+                                                                             0>;
+    union Union {
+        char c{};
+        UnionHolder first_union;
+        TypedStorage<ParentType> parent;
+
+        constexpr Union() : c() {}
+    };
+    static constexpr Union U = {};
+
+    static constexpr const MemberType* GetNextAddress(const MemberType* start,
+                                                      const MemberType* target) {
+        while (start < target) {
+            start++;
+        }
+        return start;
+    }
+
+    static constexpr std::ptrdiff_t GetDifference(const MemberType* start,
+                                                  const MemberType* target) {
+        return (target - start) * sizeof(MemberType);
+    }
+
+    template <typename CurUnion>
+    static constexpr std::ptrdiff_t OffsetOfImpl(MemberType ParentType::*member,
+                                                 CurUnion& cur_union) {
+        constexpr size_t Offset = CurUnion::GetOffset();
+        const auto target = std::addressof(GetPointer(U.parent)->*member);
+        const auto start = std::addressof(cur_union.data.members[0]);
+        const auto next = GetNextAddress(start, target);
+
+        if (next != target) {
+            if constexpr (Offset < sizeof(MemberType) - 1) {
+                return OffsetOfImpl(member, cur_union.next_union);
+            } else {
+                UNREACHABLE();
+            }
+        }
+
+        return static_cast<ptrdiff_t>(static_cast<size_t>(next - start) * sizeof(MemberType) +
+                                      Offset);
+    }
+
+    static constexpr std::ptrdiff_t OffsetOf(MemberType ParentType::*member) {
+        return OffsetOfImpl(member, U.first_union);
+    }
+};
+
+template <typename T>
+struct GetMemberPointerTraits;
+
+template <typename P, typename M>
+struct GetMemberPointerTraits<M P::*> {
+    using Parent = P;
+    using Member = M;
+};
+
+template <auto MemberPtr>
+using GetParentType = typename GetMemberPointerTraits<decltype(MemberPtr)>::Parent;
+
+template <auto MemberPtr>
+using GetMemberType = typename GetMemberPointerTraits<decltype(MemberPtr)>::Member;
+
+template <auto MemberPtr, typename RealParentType = GetParentType<MemberPtr>>
+constexpr std::ptrdiff_t OffsetOf() {
+    using DeducedParentType = GetParentType<MemberPtr>;
+    using MemberType = GetMemberType<MemberPtr>;
+    static_assert(std::is_base_of<DeducedParentType, RealParentType>::value ||
+                  std::is_same<RealParentType, DeducedParentType>::value);
+
+    return OffsetOfCalculator<RealParentType, MemberType>::OffsetOf(MemberPtr);
+};
+
+} // namespace impl
+
+template <auto MemberPtr, typename RealParentType = impl::GetParentType<MemberPtr>>
+constexpr RealParentType& GetParentReference(impl::GetMemberType<MemberPtr>* member) {
+    std::ptrdiff_t Offset = impl::OffsetOf<MemberPtr, RealParentType>();
+    return *static_cast<RealParentType*>(
+        static_cast<void*>(static_cast<uint8_t*>(static_cast<void*>(member)) - Offset));
+}
+
+template <auto MemberPtr, typename RealParentType = impl::GetParentType<MemberPtr>>
+constexpr RealParentType const& GetParentReference(impl::GetMemberType<MemberPtr> const* member) {
+    std::ptrdiff_t Offset = impl::OffsetOf<MemberPtr, RealParentType>();
+    return *static_cast<const RealParentType*>(static_cast<const void*>(
+        static_cast<const uint8_t*>(static_cast<const void*>(member)) - Offset));
+}
+
+template <auto MemberPtr, typename RealParentType = impl::GetParentType<MemberPtr>>
+constexpr RealParentType* GetParentPointer(impl::GetMemberType<MemberPtr>* member) {
+    return std::addressof(GetParentReference<MemberPtr, RealParentType>(member));
+}
+
+template <auto MemberPtr, typename RealParentType = impl::GetParentType<MemberPtr>>
+constexpr RealParentType const* GetParentPointer(impl::GetMemberType<MemberPtr> const* member) {
+    return std::addressof(GetParentReference<MemberPtr, RealParentType>(member));
+}
+
+template <auto MemberPtr, typename RealParentType = impl::GetParentType<MemberPtr>>
+constexpr RealParentType& GetParentReference(impl::GetMemberType<MemberPtr>& member) {
+    return GetParentReference<MemberPtr, RealParentType>(std::addressof(member));
+}
+
+template <auto MemberPtr, typename RealParentType = impl::GetParentType<MemberPtr>>
+constexpr RealParentType const& GetParentReference(impl::GetMemberType<MemberPtr> const& member) {
+    return GetParentReference<MemberPtr, RealParentType>(std::addressof(member));
+}
+
+template <auto MemberPtr, typename RealParentType = impl::GetParentType<MemberPtr>>
+constexpr RealParentType* GetParentPointer(impl::GetMemberType<MemberPtr>& member) {
+    return std::addressof(GetParentReference<MemberPtr, RealParentType>(member));
+}
+
+template <auto MemberPtr, typename RealParentType = impl::GetParentType<MemberPtr>>
+constexpr RealParentType const* GetParentPointer(impl::GetMemberType<MemberPtr> const& member) {
+    return std::addressof(GetParentReference<MemberPtr, RealParentType>(member));
+}
+
+} // namespace Common

src/common/static_lru_cache.h

@@ -1,113 +0,0 @@
-// Modified version of: https://www.boost.org/doc/libs/1_79_0/boost/compute/detail/lru_cache.hpp
-// Most important change is the use of an array instead of a map, so that elements are
-// statically allocated. The insert and get methods have been merged into the request method.
-// Original license:
-//
-//---------------------------------------------------------------------------//
-// Copyright (c) 2013 Kyle Lutz <kyle.r.lutz@gmail.com>
-//
-// Distributed under the Boost Software License, Version 1.0
-// See accompanying file LICENSE_1_0.txt or copy at
-// http://www.boost.org/LICENSE_1_0.txt
-//
-// See http://boostorg.github.com/compute for more information.
-//---------------------------------------------------------------------------//
-#pragma once
-
-#include <array>
-#include <list>
-#include <tuple>
-#include <utility>
-
-namespace Common {
-
-// a cache which evicts the least recently used item when it is full
-// the cache elements are statically allocated.
-template <class Key, class Value, size_t Size>
-class StaticLRUCache {
-public:
-    using key_type = Key;
-    using value_type = Value;
-    using list_type = std::list<std::pair<Key, size_t>>;
-    using array_type = std::array<Value, Size>;
-
-    StaticLRUCache() = default;
-
-    ~StaticLRUCache() = default;
-
-    size_t size() const {
-        return m_list.size();
-    }
-
-    constexpr size_t capacity() const {
-        return m_array.size();
-    }
-
-    bool empty() const {
-        return m_list.empty();
-    }
-
-    bool contains(const key_type& key) const {
-        return find(key) != m_list.end();
-    }
-
-    // Requests an element from the cache. If it is not found,
-    // the element is inserted using its key.
-    // Returns whether the element was present in the cache
-    // and a reference to the element itself.
-    std::pair<bool, value_type&> request(const key_type& key) {
-        // lookup value in the cache
-        auto i = find(key);
-        if (i == m_list.cend()) {
-            size_t next_index = size();
-            // insert item into the cache, but first check if it is full
-            if (next_index >= capacity()) {
-                // cache is full, evict the least recently used item
-                next_index = evict();
-            }
-
-            // insert the new item
-            m_list.push_front(std::make_pair(key, next_index));
-            return std::pair<bool, value_type&>(false, m_array[next_index]);
-        }
-        // return the value, but first update its place in the most
-        // recently used list
-        if (i != m_list.cbegin()) {
-            // move item to the front of the most recently used list
-            auto backup = *i;
-            m_list.erase(i);
-            m_list.push_front(backup);
-
-            // return the value
-            return std::pair<bool, value_type&>(true, m_array[backup.second]);
-        } else {
-            // the item is already at the front of the most recently
-            // used list so just return it
-            return std::pair<bool, value_type&>(true, m_array[i->second]);
-        }
-    }
-
-    void clear() {
-        m_list.clear();
-    }
-
-private:
-    typename list_type::const_iterator find(const key_type& key) const {
-        return std::find_if(m_list.cbegin(), m_list.cend(),
-                            [&key](const auto& el) { return el.first == key; });
-    }
-
-    size_t evict() {
-        // evict item from the end of most recently used list
-        typename list_type::iterator i = --m_list.end();
-        size_t evicted_index = i->second;
-        m_list.erase(i);
-        return evicted_index;
-    }
-
-private:
-    array_type m_array;
-    list_type m_list;
-};
-
-} // namespace Common

src/core/CMakeLists.txt

@@ -149,6 +149,27 @@ add_library(citra_core STATIC
     hle/kernel/ipc_debugger/recorder.h
     hle/kernel/kernel.cpp
     hle/kernel/kernel.h
+    hle/kernel/k_auto_object.cpp
+    hle/kernel/k_auto_object.h
+    hle/kernel/k_class_token.cpp
+    hle/kernel/k_class_token.h
+    hle/kernel/k_linked_list.cpp
+    hle/kernel/k_linked_list.h
+    hle/kernel/k_memory_block.cpp
+    hle/kernel/k_memory_block.h
+    hle/kernel/k_memory_block_manager.cpp
+    hle/kernel/k_memory_block_manager.h
+    hle/kernel/k_memory_manager.cpp
+    hle/kernel/k_memory_manager.h
+    hle/kernel/k_page_group.cpp
+    hle/kernel/k_page_group.h
+    hle/kernel/k_page_heap.cpp
+    hle/kernel/k_page_heap.h
+    hle/kernel/k_page_manager.cpp
+    hle/kernel/k_page_manager.h
+    hle/kernel/k_page_table.cpp
+    hle/kernel/k_page_table.h
+    hle/kernel/k_slab_heap.h
     hle/kernel/memory.cpp
     hle/kernel/memory.h
     hle/kernel/mutex.cpp
@@ -473,6 +494,7 @@ add_library(citra_core STATIC
     tracer/citrace.h
     tracer/recorder.cpp
     tracer/recorder.h
+    hle/kernel/slab_helpers.h
 )
 
 create_target_directory_groups(citra_core)

src/core/file_sys/file_backend.h

@@ -86,20 +86,6 @@ public:
      */
     virtual void Flush() const = 0;
 
-    /**
-     * Whether the backend supports cached reads.
-     */
-    virtual bool AllowsCachedReads() const {
-        return false;
-    }
-
-    /**
-     * Whether the cache is ready for a specified offset and length.
-     */
-    virtual bool CacheReady(std::size_t file_offset, std::size_t length) {
-        return false;
-    }
-
 protected:
     std::unique_ptr<DelayGenerator> delay_generator;
 

src/core/file_sys/ivfc_archive.h

@@ -131,14 +131,6 @@ public:
     }
     void Flush() const override {}
 
-    bool AllowsCachedReads() const override {
-        return romfs_file->AllowsCachedReads();
-    }
-
-    bool CacheReady(std::size_t file_offset, std::size_t length) override {
-        return romfs_file->CacheReady(file_offset, length);
-    }
-
 private:
     std::shared_ptr<RomFSReader> romfs_file;
 

src/core/file_sys/layered_fs.h

@@ -53,14 +53,6 @@ public:
 
     bool DumpRomFS(const std::string& target_path);
 
-    bool AllowsCachedReads() const override {
-        return false;
-    }
-
-    bool CacheReady(std::size_t file_offset, std::size_t length) override {
-        return false;
-    }
-
 private:
     struct File;
     struct Directory {

src/core/file_sys/romfs_reader.cpp

@@ -1,5 +1,4 @@
 #include <algorithm>
-#include <vector>
 #include <cryptopp/aes.h>
 #include <cryptopp/modes.h>
 #include "common/archives.h"
@@ -10,102 +9,17 @@ SERIALIZE_EXPORT_IMPL(FileSys::DirectRomFSReader)
 namespace FileSys {
 
 std::size_t DirectRomFSReader::ReadFile(std::size_t offset, std::size_t length, u8* buffer) {
-    length = std::min(length, static_cast<std::size_t>(data_size) - offset);
     if (length == 0)
         return 0; // Crypto++ does not like zero size buffer
-
+    file.Seek(file_offset + offset, SEEK_SET);
-    const auto segments = BreakupRead(offset, length);
+    std::size_t read_length = std::min(length, static_cast<std::size_t>(data_size) - offset);
-    size_t read_progress = 0;
+    read_length = file.ReadBytes(buffer, read_length);
-
-    // Skip cache if the read is too big
-    if (segments.size() == 1 && segments[0].second > cache_line_size) {
-        length = file.ReadAtBytes(buffer, length, file_offset + offset);
     if (is_encrypted) {
         CryptoPP::CTR_Mode<CryptoPP::AES>::Decryption d(key.data(), key.size(), ctr.data());
         d.Seek(crypto_offset + offset);
-            d.ProcessData(buffer, buffer, length);
+        d.ProcessData(buffer, buffer, read_length);
     }
-        // LOG_INFO(Service_FS, "Cache SKIP: offset={}, length={}", offset, length);
+    return read_length;
-        return length;
-    }
-
-    // TODO(PabloMK7): Make cache thread safe, read the comment in CacheReady function.
-    // std::unique_lock<std::shared_mutex> read_guard(cache_mutex);
-    for (const auto& seg : segments) {
-        size_t read_size = cache_line_size;
-        size_t page = OffsetToPage(seg.first);
-        // Check if segment is in cache
-        auto cache_entry = cache.request(page);
-        if (!cache_entry.first) {
-            // If not found, read from disk and cache the data
-            read_size = file.ReadAtBytes(cache_entry.second.data(), read_size, file_offset + page);
-            if (is_encrypted && read_size) {
-                CryptoPP::CTR_Mode<CryptoPP::AES>::Decryption d(key.data(), key.size(), ctr.data());
-                d.Seek(crypto_offset + page);
-                d.ProcessData(cache_entry.second.data(), cache_entry.second.data(), read_size);
-            }
-            // LOG_INFO(Service_FS, "Cache MISS: page={}, length={}, into={}", page, seg.second,
-            //          (seg.first - page));
-        } else {
-            // LOG_INFO(Service_FS, "Cache HIT: page={}, length={}, into={}", page, seg.second,
-            //          (seg.first - page));
-        }
-        size_t copy_amount =
-            (read_size > (seg.first - page))
-                ? std::min((seg.first - page) + seg.second, read_size) - (seg.first - page)
-                : 0;
-        std::memcpy(buffer + read_progress, cache_entry.second.data() + (seg.first - page),
-                    copy_amount);
-        read_progress += copy_amount;
-    }
-    return read_progress;
-}
-
-bool DirectRomFSReader::AllowsCachedReads() const {
-    return true;
-}
-
-bool DirectRomFSReader::CacheReady(std::size_t file_offset, std::size_t length) {
-    auto segments = BreakupRead(file_offset, length);
-    if (segments.size() == 1 && segments[0].second > cache_line_size) {
-        return false;
-    } else {
-        // TODO(PabloMK7): Since the LRU cache is not thread safe, a lock must be used.
-        // However, this completely breaks the point of using a cache, because
-        // smaller reads may be blocked by bigger reads. For now, always return
-        // data being in cache to prevent the need of a lock, and only read data
-        // asynchronously if it is too big to use the cache.
-        /*
-        std::shared_lock<std::shared_mutex> read_guard(cache_mutex);
-        for (auto it = segments.begin(); it != segments.end(); it++) {
-            if (!cache.contains(OffsetToPage(it->first)))
-                return false;
-        }
-        */
-        return true;
-    }
-}
-
-std::vector<std::pair<std::size_t, std::size_t>> DirectRomFSReader::BreakupRead(
-    std::size_t offset, std::size_t length) {
-
-    std::vector<std::pair<std::size_t, std::size_t>> ret;
-
-    // Reads bigger than the cache line size will probably never hit again
-    if (length > cache_line_size) {
-        ret.push_back(std::make_pair(offset, length));
-        return ret;
-    }
-
-    size_t curr_offset = offset;
-    while (length) {
-        size_t next_page = OffsetToPage(curr_offset + cache_line_size);
-        size_t curr_page_len = std::min(length, next_page - curr_offset);
-        ret.push_back(std::make_pair(curr_offset, curr_page_len));
-        curr_offset = next_page;
-        length -= curr_page_len;
-    }
-    return ret;
 }
 
 } // namespace FileSys

src/core/file_sys/romfs_reader.h

@@ -1,14 +1,11 @@
 #pragma once
 
 #include <array>
-#include <shared_mutex>
 #include <boost/serialization/array.hpp>
 #include <boost/serialization/base_object.hpp>
 #include <boost/serialization/export.hpp>
-#include "common/alignment.h"
 #include "common/common_types.h"
 #include "common/file_util.h"
-#include "common/static_lru_cache.h"
 
 namespace FileSys {
 
@@ -21,8 +18,6 @@ public:
 
     virtual std::size_t GetSize() const = 0;
     virtual std::size_t ReadFile(std::size_t offset, std::size_t length, u8* buffer) = 0;
-    virtual bool AllowsCachedReads() const = 0;
-    virtual bool CacheReady(std::size_t file_offset, std::size_t length) = 0;
 
 private:
     template <class Archive>
@@ -53,10 +48,6 @@ public:
 
     std::size_t ReadFile(std::size_t offset, std::size_t length, u8* buffer) override;
 
-    bool AllowsCachedReads() const override;
-
-    bool CacheReady(std::size_t file_offset, std::size_t length) override;
-
 private:
     bool is_encrypted;
     FileUtil::IOFile file;
@@ -66,23 +57,8 @@ private:
     u64 crypto_offset;
     u64 data_size;
 
-    // Total cache size: 128KB
-    static constexpr size_t cache_line_size = (1 << 13); // About 8KB
-    static constexpr size_t cache_line_count = 16;
-
-    Common::StaticLRUCache<std::size_t, std::array<u8, cache_line_size>, cache_line_count> cache;
-    // TODO(PabloMK7): Make cache thread safe, read the comment in CacheReady function.
-    // std::shared_mutex cache_mutex;
-
     DirectRomFSReader() = default;
 
-    std::size_t OffsetToPage(std::size_t offset) {
-        return Common::AlignDown<std::size_t>(offset, cache_line_size);
-    }
-
-    std::vector<std::pair<std::size_t, std::size_t>> BreakupRead(std::size_t offset,
-                                                                 std::size_t length);
-
     template <class Archive>
     void serialize(Archive& ar, const unsigned int) {
         ar& boost::serialization::base_object<RomFSReader>(*this);

src/core/hle/kernel/hle_ipc.h

@@ -221,13 +221,6 @@ public:
         return session;
     }
 
-    /**
-     * Returns the client thread that made the service request.
-     */
-    std::shared_ptr<Thread> ClientThread() const {
-        return thread;
-    }
-
     class WakeupCallback {
     public:
         virtual ~WakeupCallback() = default;

src/core/hle/kernel/k_auto_object.cpp

@@ -0,0 +1,22 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/hle/kernel/k_auto_object.h"
+#include "core/hle/kernel/kernel.h"
+
+namespace Kernel {
+
+KAutoObject* KAutoObject::Create(KAutoObject* obj) {
+    obj->m_ref_count = 1;
+    return obj;
+}
+
+void KAutoObject::RegisterWithKernel() {
+    m_kernel.RegisterKernelObject(this);
+}
+
+void KAutoObject::UnregisterWithKernel(KernelCore& kernel, KAutoObject* self) {
+    kernel.UnregisterKernelObject(self);
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_auto_object.h

@@ -0,0 +1,268 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <atomic>
+#include <string>
+
+#include "common/assert.h"
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "core/hle/kernel/k_class_token.h"
+
+namespace Kernel {
+
+class KernelSystem;
+class KProcess;
+
+#define KERNEL_AUTOOBJECT_TRAITS_IMPL(CLASS, BASE_CLASS, ATTRIBUTE)                                \
+private:                                                                                           \
+    friend class ::Kernel::KClassTokenGenerator;                                                   \
+    static constexpr inline auto ObjectType = ::Kernel::KClassTokenGenerator::ObjectType::CLASS;   \
+    static constexpr inline const char* const TypeName = #CLASS;                                   \
+    static constexpr inline ClassTokenType ClassToken() { return ::Kernel::ClassToken<CLASS>; }    \
+                                                                                                   \
+public:                                                                                            \
+    CITRA_NON_COPYABLE(CLASS);                                                                     \
+    CITRA_NON_MOVEABLE(CLASS);                                                                     \
+                                                                                                   \
+    using BaseClass = BASE_CLASS;                                                                  \
+    static constexpr TypeObj GetStaticTypeObj() {                                                  \
+        constexpr ClassTokenType Token = ClassToken();                                             \
+        return TypeObj(TypeName, Token);                                                           \
+    }                                                                                              \
+    static constexpr const char* GetStaticTypeName() { return TypeName; }                          \
+    virtual TypeObj GetTypeObj() ATTRIBUTE { return GetStaticTypeObj(); }                          \
+    virtual const char* GetTypeName() ATTRIBUTE { return GetStaticTypeName(); }                    \
+                                                                                                   \
+private:                                                                                           \
+    constexpr bool operator!=(const TypeObj& rhs)
+
+#define KERNEL_AUTOOBJECT_TRAITS(CLASS, BASE_CLASS)                                                \
+    KERNEL_AUTOOBJECT_TRAITS_IMPL(CLASS, BASE_CLASS, const override)
+
+class KAutoObject {
+protected:
+    class TypeObj {
+    public:
+        constexpr explicit TypeObj(const char* n, ClassTokenType tok)
+            : m_name(n), m_class_token(tok) {}
+
+        constexpr const char* GetName() const {
+            return m_name;
+        }
+        constexpr ClassTokenType GetClassToken() const {
+            return m_class_token;
+        }
+
+        constexpr bool operator==(const TypeObj& rhs) const {
+            return this->GetClassToken() == rhs.GetClassToken();
+        }
+
+        constexpr bool operator!=(const TypeObj& rhs) const {
+            return this->GetClassToken() != rhs.GetClassToken();
+        }
+
+        constexpr bool IsDerivedFrom(const TypeObj& rhs) const {
+            return (this->GetClassToken() | rhs.GetClassToken()) == this->GetClassToken();
+        }
+
+    private:
+        const char* m_name;
+        ClassTokenType m_class_token;
+    };
+
+private:
+    KERNEL_AUTOOBJECT_TRAITS_IMPL(KAutoObject, KAutoObject, const);
+
+public:
+    explicit KAutoObject(KernelSystem& kernel) : m_kernel(kernel) {
+        RegisterWithKernel();
+    }
+    virtual ~KAutoObject() = default;
+
+    static KAutoObject* Create(KAutoObject* ptr);
+
+    // Destroy is responsible for destroying the auto object's resources when ref_count hits zero.
+    virtual void Destroy() {
+        UNIMPLEMENTED();
+    }
+
+    // Finalize is responsible for cleaning up resource, but does not destroy the object.
+    virtual void Finalize() {}
+
+    virtual KProcess* GetOwner() const {
+        return nullptr;
+    }
+
+    u32 GetReferenceCount() const {
+        return m_ref_count.load();
+    }
+
+    bool IsDerivedFrom(const TypeObj& rhs) const {
+        return this->GetTypeObj().IsDerivedFrom(rhs);
+    }
+
+    bool IsDerivedFrom(const KAutoObject& rhs) const {
+        return this->IsDerivedFrom(rhs.GetTypeObj());
+    }
+
+    template <typename Derived>
+    Derived DynamicCast() {
+        static_assert(std::is_pointer_v<Derived>);
+        using DerivedType = std::remove_pointer_t<Derived>;
+
+        if (this->IsDerivedFrom(DerivedType::GetStaticTypeObj())) {
+            return static_cast<Derived>(this);
+        } else {
+            return nullptr;
+        }
+    }
+
+    template <typename Derived>
+    const Derived DynamicCast() const {
+        static_assert(std::is_pointer_v<Derived>);
+        using DerivedType = std::remove_pointer_t<Derived>;
+
+        if (this->IsDerivedFrom(DerivedType::GetStaticTypeObj())) {
+            return static_cast<Derived>(this);
+        } else {
+            return nullptr;
+        }
+    }
+
+    bool Open() {
+        // Atomically increment the reference count, only if it's positive.
+        u32 cur_ref_count = m_ref_count.load(std::memory_order_acquire);
+        do {
+            if (cur_ref_count == 0) {
+                return false;
+            }
+            ASSERT(cur_ref_count < cur_ref_count + 1);
+        } while (!m_ref_count.compare_exchange_weak(cur_ref_count, cur_ref_count + 1,
+                                                    std::memory_order_relaxed));
+
+        return true;
+    }
+
+    void Close() {
+        // Atomically decrement the reference count, not allowing it to become negative.
+        u32 cur_ref_count = m_ref_count.load(std::memory_order_acquire);
+        do {
+            ASSERT(cur_ref_count > 0);
+        } while (!m_ref_count.compare_exchange_weak(cur_ref_count, cur_ref_count - 1,
+                                                    std::memory_order_acq_rel));
+
+        // If ref count hits zero, destroy the object.
+        if (cur_ref_count - 1 == 0) {
+            KernelSystem& kernel = m_kernel;
+            this->Destroy();
+            KAutoObject::UnregisterWithKernel(kernel, this);
+        }
+    }
+
+private:
+    void RegisterWithKernel();
+    static void UnregisterWithKernel(KernelSystem& kernel, KAutoObject* self);
+
+protected:
+    KernelSystem& m_kernel;
+
+private:
+    std::atomic<u32> m_ref_count{};
+};
+
+template <typename T>
+class KScopedAutoObject {
+public:
+    CITRA_NON_COPYABLE(KScopedAutoObject);
+
+    constexpr KScopedAutoObject() = default;
+
+    constexpr KScopedAutoObject(T* o) : m_obj(o) {
+        if (m_obj != nullptr) {
+            m_obj->Open();
+        }
+    }
+
+    ~KScopedAutoObject() {
+        if (m_obj != nullptr) {
+            m_obj->Close();
+        }
+        m_obj = nullptr;
+    }
+
+    template <typename U>
+        requires(std::derived_from<T, U> || std::derived_from<U, T>)
+    constexpr KScopedAutoObject(KScopedAutoObject<U>&& rhs) {
+        if constexpr (std::derived_from<U, T>) {
+            // Upcast.
+            m_obj = rhs.m_obj;
+            rhs.m_obj = nullptr;
+        } else {
+            // Downcast.
+            T* derived = nullptr;
+            if (rhs.m_obj != nullptr) {
+                derived = rhs.m_obj->template DynamicCast<T*>();
+                if (derived == nullptr) {
+                    rhs.m_obj->Close();
+                }
+            }
+
+            m_obj = derived;
+            rhs.m_obj = nullptr;
+        }
+    }
+
+    constexpr KScopedAutoObject<T>& operator=(KScopedAutoObject<T>&& rhs) {
+        rhs.Swap(*this);
+        return *this;
+    }
+
+    constexpr T* operator->() {
+        return m_obj;
+    }
+    constexpr T& operator*() {
+        return *m_obj;
+    }
+
+    constexpr void Reset(T* o) {
+        KScopedAutoObject(o).Swap(*this);
+    }
+
+    constexpr T* GetPointerUnsafe() {
+        return m_obj;
+    }
+
+    constexpr T* GetPointerUnsafe() const {
+        return m_obj;
+    }
+
+    constexpr T* ReleasePointerUnsafe() {
+        T* ret = m_obj;
+        m_obj = nullptr;
+        return ret;
+    }
+
+    constexpr bool IsNull() const {
+        return m_obj == nullptr;
+    }
+    constexpr bool IsNotNull() const {
+        return m_obj != nullptr;
+    }
+
+private:
+    template <typename U>
+    friend class KScopedAutoObject;
+
+private:
+    T* m_obj{};
+
+private:
+    constexpr void Swap(KScopedAutoObject& rhs) noexcept {
+        std::swap(m_obj, rhs.m_obj);
+    }
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_class_token.cpp

@@ -0,0 +1,125 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/hle/kernel/k_auto_object.h"
+#include "core/hle/kernel/k_class_token.h"
+#include "core/hle/kernel/k_client_port.h"
+#include "core/hle/kernel/k_client_session.h"
+#include "core/hle/kernel/k_code_memory.h"
+#include "core/hle/kernel/k_event.h"
+#include "core/hle/kernel/k_port.h"
+#include "core/hle/kernel/k_process.h"
+#include "core/hle/kernel/k_readable_event.h"
+#include "core/hle/kernel/k_resource_limit.h"
+#include "core/hle/kernel/k_server_port.h"
+#include "core/hle/kernel/k_server_session.h"
+#include "core/hle/kernel/k_session.h"
+#include "core/hle/kernel/k_shared_memory.h"
+#include "core/hle/kernel/k_synchronization_object.h"
+#include "core/hle/kernel/k_system_resource.h"
+#include "core/hle/kernel/k_thread.h"
+#include "core/hle/kernel/k_transfer_memory.h"
+
+namespace Kernel {
+
+// Ensure that we generate correct class tokens for all types.
+
+// Ensure that the absolute token values are correct.
+static_assert(ClassToken<KAutoObject> == 0b00000000'00000000);
+static_assert(ClassToken<KSynchronizationObject> == 0b00000000'00000001);
+static_assert(ClassToken<KReadableEvent> == 0b00000000'00000011);
+// static_assert(ClassToken<KInterruptEvent> == 0b00000111'00000011);
+// static_assert(ClassToken<KDebug> == 0b00001011'00000001);
+static_assert(ClassToken<KThread> == 0b00010011'00000001);
+static_assert(ClassToken<KServerPort> == 0b00100011'00000001);
+static_assert(ClassToken<KServerSession> == 0b01000011'00000001);
+static_assert(ClassToken<KClientPort> == 0b10000011'00000001);
+static_assert(ClassToken<KClientSession> == 0b00001101'00000000);
+static_assert(ClassToken<KProcess> == 0b00010101'00000001);
+static_assert(ClassToken<KResourceLimit> == 0b00100101'00000000);
+// static_assert(ClassToken<KLightSession> == 0b01000101'00000000);
+static_assert(ClassToken<KPort> == 0b10000101'00000000);
+static_assert(ClassToken<KSession> == 0b00011001'00000000);
+static_assert(ClassToken<KSharedMemory> == 0b00101001'00000000);
+static_assert(ClassToken<KEvent> == 0b01001001'00000000);
+// static_assert(ClassToken<KLightClientSession> == 0b00110001'00000000);
+// static_assert(ClassToken<KLightServerSession> == 0b01010001'00000000);
+static_assert(ClassToken<KTransferMemory> == 0b01010001'00000000);
+// static_assert(ClassToken<KDeviceAddressSpace> == 0b01100001'00000000);
+// static_assert(ClassToken<KSessionRequest> == 0b10100001'00000000);
+static_assert(ClassToken<KCodeMemory> == 0b10100001'00000000);
+
+// Ensure that the token hierarchy is correct.
+
+// Base classes
+static_assert(ClassToken<KAutoObject> == (0b00000000));
+static_assert(ClassToken<KSynchronizationObject> == (0b00000001 | ClassToken<KAutoObject>));
+static_assert(ClassToken<KReadableEvent> == (0b00000010 | ClassToken<KSynchronizationObject>));
+
+// Final classes
+// static_assert(ClassToken<KInterruptEvent> == ((0b00000111 << 8) | ClassToken<KReadableEvent>));
+// static_assert(ClassToken<KDebug> == ((0b00001011 << 8) | ClassToken<KSynchronizationObject>));
+static_assert(ClassToken<KThread> == ((0b00010011 << 8) | ClassToken<KSynchronizationObject>));
+static_assert(ClassToken<KServerPort> == ((0b00100011 << 8) | ClassToken<KSynchronizationObject>));
+static_assert(ClassToken<KServerSession> ==
+              ((0b01000011 << 8) | ClassToken<KSynchronizationObject>));
+static_assert(ClassToken<KClientPort> == ((0b10000011 << 8) | ClassToken<KSynchronizationObject>));
+static_assert(ClassToken<KClientSession> == ((0b00001101 << 8) | ClassToken<KAutoObject>));
+static_assert(ClassToken<KProcess> == ((0b00010101 << 8) | ClassToken<KSynchronizationObject>));
+static_assert(ClassToken<KResourceLimit> == ((0b00100101 << 8) | ClassToken<KAutoObject>));
+// static_assert(ClassToken<KLightSession> == ((0b01000101 << 8) | ClassToken<KAutoObject>));
+static_assert(ClassToken<KPort> == ((0b10000101 << 8) | ClassToken<KAutoObject>));
+static_assert(ClassToken<KSession> == ((0b00011001 << 8) | ClassToken<KAutoObject>));
+static_assert(ClassToken<KSharedMemory> == ((0b00101001 << 8) | ClassToken<KAutoObject>));
+static_assert(ClassToken<KEvent> == ((0b01001001 << 8) | ClassToken<KAutoObject>));
+// static_assert(ClassToken<KLightClientSession> == ((0b00110001 << 8) | ClassToken<KAutoObject>));
+// static_assert(ClassToken<KLightServerSession> == ((0b01010001 << 8) | ClassToken<KAutoObject>));
+static_assert(ClassToken<KTransferMemory> == ((0b01010001 << 8) | ClassToken<KAutoObject>));
+// static_assert(ClassToken<KDeviceAddressSpace> == ((0b01100001 << 8) | ClassToken<KAutoObject>));
+// static_assert(ClassToken<KSessionRequest> == ((0b10100001 << 8) | ClassToken<KAutoObject>));
+static_assert(ClassToken<KCodeMemory> == ((0b10100001 << 8) | ClassToken<KAutoObject>));
+
+// Ensure that the token hierarchy reflects the class hierarchy.
+
+// Base classes.
+static_assert(!std::is_final_v<KSynchronizationObject> &&
+              std::is_base_of_v<KAutoObject, KSynchronizationObject>);
+static_assert(!std::is_final_v<KReadableEvent> &&
+              std::is_base_of_v<KSynchronizationObject, KReadableEvent>);
+
+// Final classes
+// static_assert(std::is_final_v<KInterruptEvent> &&
+//              std::is_base_of_v<KReadableEvent, KInterruptEvent>);
+// static_assert(std::is_final_v<KDebug> &&
+//              std::is_base_of_v<KSynchronizationObject, KDebug>);
+static_assert(std::is_final_v<KThread> && std::is_base_of_v<KSynchronizationObject, KThread>);
+static_assert(std::is_final_v<KServerPort> &&
+              std::is_base_of_v<KSynchronizationObject, KServerPort>);
+static_assert(std::is_final_v<KServerSession> &&
+              std::is_base_of_v<KSynchronizationObject, KServerSession>);
+static_assert(std::is_final_v<KClientPort> &&
+              std::is_base_of_v<KSynchronizationObject, KClientPort>);
+static_assert(std::is_final_v<KClientSession> && std::is_base_of_v<KAutoObject, KClientSession>);
+static_assert(std::is_final_v<KProcess> && std::is_base_of_v<KSynchronizationObject, KProcess>);
+static_assert(std::is_final_v<KResourceLimit> && std::is_base_of_v<KAutoObject, KResourceLimit>);
+// static_assert(std::is_final_v<KLightSession> &&
+//              std::is_base_of_v<KAutoObject, KLightSession>);
+static_assert(std::is_final_v<KPort> && std::is_base_of_v<KAutoObject, KPort>);
+static_assert(std::is_final_v<KSession> && std::is_base_of_v<KAutoObject, KSession>);
+static_assert(std::is_final_v<KSharedMemory> && std::is_base_of_v<KAutoObject, KSharedMemory>);
+static_assert(std::is_final_v<KEvent> && std::is_base_of_v<KAutoObject, KEvent>);
+// static_assert(std::is_final_v<KLightClientSession> &&
+//              std::is_base_of_v<KAutoObject, KLightClientSession>);
+// static_assert(std::is_final_v<KLightServerSession> &&
+//              std::is_base_of_v<KAutoObject, KLightServerSession>);
+static_assert(std::is_final_v<KTransferMemory> && std::is_base_of_v<KAutoObject, KTransferMemory>);
+// static_assert(std::is_final_v<KDeviceAddressSpace> &&
+//              std::is_base_of_v<KAutoObject, KDeviceAddressSpace>);
+// static_assert(std::is_final_v<KSessionRequest> &&
+//              std::is_base_of_v<KAutoObject, KSessionRequest>);
+// static_assert(std::is_final_v<KCodeMemory> &&
+//              std::is_base_of_v<KAutoObject, KCodeMemory>);
+
+static_assert(std::is_base_of_v<KAutoObject, KSystemResource>);
+
+} // namespace Kernel

src/core/hle/kernel/k_class_token.h

@@ -0,0 +1,111 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Kernel {
+
+class KAutoObject;
+class KSynchronizationObject;
+
+class KClassTokenGenerator {
+public:
+    using TokenBaseType = u8;
+
+public:
+    static constexpr size_t BaseClassBits = 1;
+    static constexpr size_t FinalClassBits = (sizeof(TokenBaseType) * CHAR_BIT) - BaseClassBits - 1;
+    // One bit per base class.
+    static constexpr size_t NumBaseClasses = BaseClassBits;
+    // Final classes are permutations of three bits.
+    static constexpr size_t NumFinalClasses = [] {
+        TokenBaseType index = 0;
+        for (size_t i = 0; i < FinalClassBits; i++) {
+            for (size_t j = i + 1; j < FinalClassBits; j++) {
+                for (size_t k = j + 1; k < FinalClassBits; k++) {
+                    index++;
+                }
+            }
+        }
+        return index;
+    }();
+
+private:
+    template <TokenBaseType Index>
+    static constexpr inline TokenBaseType BaseClassToken = 1U << Index;
+
+    template <TokenBaseType Index>
+    static constexpr inline TokenBaseType FinalClassToken = [] {
+        TokenBaseType index = 0;
+        for (size_t i = 0; i < FinalClassBits; i++) {
+            for (size_t j = i + 1; j < FinalClassBits; j++) {
+                for (size_t k = j + 1; k < FinalClassBits; k++) {
+                    if ((index++) == Index) {
+                        return static_cast<TokenBaseType>(((1ULL << i) | (1ULL << j) | (1ULL << k))
+                                                          << BaseClassBits);
+                    }
+                }
+            }
+        }
+        UNREACHABLE();
+    }();
+
+    template <typename T>
+    static constexpr inline TokenBaseType GetClassToken() {
+        static_assert(std::is_base_of<KAutoObject, T>::value);
+        if constexpr (std::is_same<T, KAutoObject>::value) {
+            static_assert(T::ObjectType == ObjectType::KAutoObject);
+            return 0;
+        } else if constexpr (std::is_same<T, KSynchronizationObject>::value) {
+            static_assert(T::ObjectType == ObjectType::KSynchronizationObject);
+            return 1;
+        } else if constexpr (ObjectType::FinalClassesStart <= T::ObjectType &&
+                             T::ObjectType < ObjectType::FinalClassesEnd) {
+            constexpr auto ClassIndex = static_cast<TokenBaseType>(T::ObjectType) -
+                                        static_cast<TokenBaseType>(ObjectType::FinalClassesStart);
+            return FinalClassToken<ClassIndex> | GetClassToken<typename T::BaseClass>();
+        } else {
+            static_assert(!std::is_same<T, T>::value, "GetClassToken: Invalid Type");
+        }
+    };
+
+public:
+    enum class ObjectType {
+        KAutoObject,
+        KSynchronizationObject,
+
+        FinalClassesStart,
+        KSemaphore,
+        KEvent,
+        KTimer,
+        KMutex,
+        KDebug,
+        KServerPort,
+        KDmaObject,
+        KClientPort,
+        KCodeSet,
+        KSession,
+        KThread,
+        KServerSession,
+        KAddressArbiter,
+        KClientSession,
+        KPort,
+        KSharedMemory,
+        KProcess,
+        KResourceLimit,
+
+        FinalClassesEnd = FinalClassesStart + NumFinalClasses,
+    };
+
+    template <typename T>
+    static constexpr inline TokenBaseType ClassToken = GetClassToken<T>();
+};
+
+using ClassTokenType = KClassTokenGenerator::TokenBaseType;
+
+template <typename T>
+static constexpr inline ClassTokenType ClassToken = KClassTokenGenerator::ClassToken<T>;
+
+} // namespace Kernel

src/core/hle/kernel/k_linked_list.h

@@ -0,0 +1,237 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/intrusive_list.h"
+#include "core/hle/kernel/slab_helpers.h"
+
+namespace Kernel {
+
+class KernelSystem;
+
+class KLinkedListNode : public Common::IntrusiveListBaseNode<KLinkedListNode>,
+                        public KSlabAllocated<KLinkedListNode> {
+
+public:
+    explicit KLinkedListNode(KernelSystem&) {}
+    KLinkedListNode() = default;
+
+    void Initialize(void* it) {
+        m_item = it;
+    }
+
+    void* GetItem() const {
+        return m_item;
+    }
+
+private:
+    void* m_item = nullptr;
+};
+
+template <typename T>
+class KLinkedList : private Common::IntrusiveListBaseTraits<KLinkedListNode>::ListType {
+private:
+    using BaseList = Common::IntrusiveListBaseTraits<KLinkedListNode>::ListType;
+
+public:
+    template <bool Const>
+    class Iterator;
+
+    using value_type = T;
+    using size_type = size_t;
+    using difference_type = ptrdiff_t;
+    using pointer = value_type*;
+    using const_pointer = const value_type*;
+    using reference = value_type&;
+    using const_reference = const value_type&;
+    using iterator = Iterator<false>;
+    using const_iterator = Iterator<true>;
+    using reverse_iterator = std::reverse_iterator<iterator>;
+    using const_reverse_iterator = std::reverse_iterator<const_iterator>;
+
+    template <bool Const>
+    class Iterator {
+    private:
+        using BaseIterator = BaseList::iterator;
+        friend class KLinkedList;
+
+    public:
+        using iterator_category = std::bidirectional_iterator_tag;
+        using value_type = typename KLinkedList::value_type;
+        using difference_type = typename KLinkedList::difference_type;
+        using pointer = std::conditional_t<Const, KLinkedList::const_pointer, KLinkedList::pointer>;
+        using reference =
+            std::conditional_t<Const, KLinkedList::const_reference, KLinkedList::reference>;
+
+    public:
+        explicit Iterator(BaseIterator it) : m_base_it(it) {}
+
+        pointer GetItem() const {
+            return static_cast<pointer>(m_base_it->GetItem());
+        }
+
+        bool operator==(const Iterator& rhs) const {
+            return m_base_it == rhs.m_base_it;
+        }
+
+        bool operator!=(const Iterator& rhs) const {
+            return !(*this == rhs);
+        }
+
+        pointer operator->() const {
+            return this->GetItem();
+        }
+
+        reference operator*() const {
+            return *this->GetItem();
+        }
+
+        Iterator& operator++() {
+            ++m_base_it;
+            return *this;
+        }
+
+        Iterator& operator--() {
+            --m_base_it;
+            return *this;
+        }
+
+        Iterator operator++(int) {
+            const Iterator it{*this};
+            ++(*this);
+            return it;
+        }
+
+        Iterator operator--(int) {
+            const Iterator it{*this};
+            --(*this);
+            return it;
+        }
+
+        operator Iterator<true>() const {
+            return Iterator<true>(m_base_it);
+        }
+
+    private:
+        BaseIterator m_base_it;
+    };
+
+public:
+    constexpr KLinkedList(KernelSystem& kernel_) : BaseList(), kernel{kernel_} {}
+
+    ~KLinkedList() {
+        // Erase all elements.
+        for (auto it = begin(); it != end(); it = erase(it)) {
+        }
+
+        // Ensure we succeeded.
+        ASSERT(this->empty());
+    }
+
+    // Iterator accessors.
+    iterator begin() {
+        return iterator(BaseList::begin());
+    }
+
+    const_iterator begin() const {
+        return const_iterator(BaseList::begin());
+    }
+
+    iterator end() {
+        return iterator(BaseList::end());
+    }
+
+    const_iterator end() const {
+        return const_iterator(BaseList::end());
+    }
+
+    const_iterator cbegin() const {
+        return this->begin();
+    }
+
+    const_iterator cend() const {
+        return this->end();
+    }
+
+    reverse_iterator rbegin() {
+        return reverse_iterator(this->end());
+    }
+
+    const_reverse_iterator rbegin() const {
+        return const_reverse_iterator(this->end());
+    }
+
+    reverse_iterator rend() {
+        return reverse_iterator(this->begin());
+    }
+
+    const_reverse_iterator rend() const {
+        return const_reverse_iterator(this->begin());
+    }
+
+    const_reverse_iterator crbegin() const {
+        return this->rbegin();
+    }
+
+    const_reverse_iterator crend() const {
+        return this->rend();
+    }
+
+    // Content management.
+    using BaseList::empty;
+    using BaseList::size;
+
+    reference back() {
+        return *(--this->end());
+    }
+
+    const_reference back() const {
+        return *(--this->end());
+    }
+
+    reference front() {
+        return *this->begin();
+    }
+
+    const_reference front() const {
+        return *this->begin();
+    }
+
+    iterator insert(const_iterator pos, reference ref) {
+        KLinkedListNode* new_node = KLinkedListNode::Allocate(kernel);
+        ASSERT(new_node != nullptr);
+        new_node->Initialize(std::addressof(ref));
+        return iterator(BaseList::insert(pos.m_base_it, *new_node));
+    }
+
+    void push_back(reference ref) {
+        this->insert(this->end(), ref);
+    }
+
+    void push_front(reference ref) {
+        this->insert(this->begin(), ref);
+    }
+
+    void pop_back() {
+        this->erase(--this->end());
+    }
+
+    void pop_front() {
+        this->erase(this->begin());
+    }
+
+    iterator erase(const iterator pos) {
+        KLinkedListNode* freed_node = std::addressof(*pos.m_base_it);
+        iterator ret = iterator(BaseList::erase(pos.m_base_it));
+        KLinkedListNode::Free(kernel, freed_node);
+
+        return ret;
+    }
+
+private:
+    KernelSystem& kernel;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_memory_block.cpp

@@ -0,0 +1,41 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/k_memory_block.h"
+
+namespace Kernel {
+
+void KMemoryBlock::ShrinkBlock(VAddr addr, u32 num_pages) {
+    const VAddr end_addr = addr + (num_pages << Memory::CITRA_PAGE_BITS) - 1;
+    const VAddr last_addr = this->GetLastAddress();
+    if (m_base_addr < end_addr && end_addr < last_addr) {
+        m_base_addr = end_addr + 1;
+        m_num_pages = (last_addr - end_addr) >> Memory::CITRA_PAGE_BITS;
+        return;
+    }
+    if (m_base_addr < addr && addr < last_addr) {
+        m_num_pages = (addr - m_base_addr) >> Memory::CITRA_PAGE_BITS;
+        return;
+    }
+}
+
+void KMemoryBlock::GrowBlock(VAddr addr, u32 num_pages) {
+    const u32 end_addr = addr + (num_pages << Memory::CITRA_PAGE_BITS) - 1;
+    const u32 last_addr = this->GetLastAddress();
+    if (addr < m_base_addr) {
+        m_base_addr = addr;
+        m_num_pages = (last_addr - addr + 1) >> Memory::CITRA_PAGE_BITS;
+    }
+    if (last_addr < end_addr) {
+        m_num_pages = (end_addr - m_base_addr + 1) >> Memory::CITRA_PAGE_BITS;
+    }
+}
+
+bool KMemoryBlock::IncludesRange(VAddr addr, u32 num_pages) {
+    const u32 end_addr = addr + (num_pages << Memory::CITRA_PAGE_BITS) - 1;
+    const u32 last_addr = this->GetLastAddress();
+    return m_base_addr >= addr && last_addr <= end_addr;
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_memory_block.h

@@ -0,0 +1,187 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "core/hle/kernel/slab_helpers.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+enum class KMemoryPermission : u32 {
+    None = 0x0,
+    UserRead = 0x1,
+    UserWrite = 0x2,
+    UserReadWrite = UserRead | UserWrite,
+    UserExecute = 0x4,
+    UserReadExecute = UserRead | UserExecute,
+    KernelRead = 0x8,
+    KernelWrite = 0x10,
+    KernelExecute = 0x20,
+    KernelReadWrite = KernelRead | KernelWrite,
+    DontCare = 0x10000000,
+};
+DECLARE_ENUM_FLAG_OPERATORS(KMemoryPermission)
+
+enum class KMemoryState : u32 {
+    Free = 0x0,
+    Reserved = 0x1,
+    Io = 0x2,
+    Static = 0x3,
+    Code = 0x4,
+    Private = 0x5,
+    Shared = 0x6,
+    Continuous = 0x7,
+    Aliased = 0x8,
+    Alias = 0x9,
+    Aliascode = 0xA,
+    Locked = 0xB,
+    KernelMask = 0xFF,
+
+    FlagDeallocatable = 0x100,
+    FlagProtectible = 0x200,
+    FlagDebuggable = 0x400,
+    FlagIpcAllowed = 0x800,
+    FlagMapped = 0x1000,
+    FlagPrivate = 0x2000,
+    FlagShared = 0x4000,
+    FlagsPrivateOrShared = 0x6000,
+    FlagCodeAllowed = 0x8000,
+    FlagsIpc = 0x1800,
+    FlagsPrivateData = 0x3800,
+    FlagsPrivateCodeAllowed = 0xB800,
+    FlagsPrivateCode = 0xBC00,
+    FlagsCode = 0x9C00,
+
+    KernelIo = 0x1002,
+    KernelStatic = 0x1003,
+    KernelShared = 0x5806,
+    KernelLinear = 0x3907,
+    KernelAliased = 0x3A08,
+    KernelAlias = 0x1A09,
+    KernelAliasCode = 0x9C0A,
+    PrivateAliasCode = 0xBC0A,
+    PrivateCode = 0xBC04,
+    PrivateData = 0xBB05,
+    KernelLocked = 0x380B,
+    FlagsAny = 0xFFFFFFFF,
+};
+DECLARE_ENUM_FLAG_OPERATORS(KMemoryState)
+
+struct KMemoryInfo {
+    VAddr m_base_address;
+    u32 m_size;
+    KMemoryPermission m_perms;
+    KMemoryState m_state;
+
+    constexpr VAddr GetAddress() const {
+        return m_base_address;
+    }
+
+    constexpr u32 GetSize() const {
+        return m_size;
+    }
+
+    constexpr u32 GetNumPages() const {
+        return this->GetSize() >> Memory::CITRA_PAGE_BITS;
+    }
+
+    constexpr VAddr GetEndAddress() const {
+        return this->GetAddress() + this->GetSize();
+    }
+
+    constexpr VAddr GetLastAddress() const {
+        return this->GetEndAddress() - 1;
+    }
+
+    constexpr KMemoryPermission GetPerms() const {
+        return m_perms;
+    }
+
+    constexpr KMemoryState GetState() const {
+        return m_state;
+    }
+};
+
+struct KMemoryBlock : public KSlabAllocated<KMemoryBlock> {
+public:
+    explicit KMemoryBlock() = default;
+
+    constexpr void Initialize(VAddr base_addr, u32 num_pages, u32 tag, KMemoryState state,
+                              KMemoryPermission perms) {
+        m_base_addr = base_addr;
+        m_num_pages = num_pages;
+        m_permission = perms;
+        m_memory_state = state;
+        m_tag = tag;
+    }
+
+    constexpr bool Contains(VAddr addr) const {
+        return this->GetAddress() <= addr && addr <= this->GetLastAddress();
+    }
+
+    constexpr KMemoryInfo GetInfo() const {
+        return {
+            .m_base_address = m_base_addr,
+            .m_size = this->GetSize(),
+            .m_perms = m_permission,
+            .m_state = m_memory_state,
+        };
+    }
+
+    constexpr bool HasProperties(KMemoryState s, KMemoryPermission p, u32 t) const {
+        return m_memory_state == s && m_permission == p && m_tag == t;
+    }
+
+    constexpr bool HasSameProperties(const KMemoryBlock& rhs) const {
+        return m_memory_state == rhs.m_memory_state && m_permission == rhs.m_permission &&
+               m_tag == rhs.m_tag;
+    }
+
+    constexpr u32 GetSize() const {
+        return m_num_pages << Memory::CITRA_PAGE_BITS;
+    }
+
+    constexpr u32 GetEndAddress() const {
+        return this->GetAddress() + this->GetSize();
+    }
+
+    constexpr u32 GetLastAddress() const {
+        return this->GetEndAddress() - 1;
+    }
+
+    constexpr u32 GetAddress() const {
+        return m_base_addr;
+    }
+
+    constexpr u32 GetNumPages() const {
+        return m_num_pages;
+    }
+
+    constexpr KMemoryPermission GetPermission() const {
+        return m_permission;
+    }
+
+    constexpr KMemoryState GetState() const {
+        return m_memory_state;
+    }
+
+    constexpr u32 GetTag() const {
+        return m_tag;
+    }
+
+    void ShrinkBlock(VAddr addr, u32 num_pages);
+    void GrowBlock(VAddr addr, u32 num_pages);
+    bool IncludesRange(VAddr addr, u32 num_pages);
+
+private:
+    u32 m_base_addr{};
+    u32 m_num_pages{};
+    KMemoryPermission m_permission{};
+    KMemoryState m_memory_state{};
+    u32 m_tag{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_memory_block_manager.cpp

@@ -0,0 +1,199 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/scope_exit.h"
+#include "core/hle/kernel/k_memory_block_manager.h"
+
+namespace Kernel {
+
+void KMemoryBlockManager::Initialize(u32 addr_space_start, u32 addr_space_end) {
+    const u32 num_pages = (addr_space_end - addr_space_start) >> Memory::CITRA_PAGE_BITS;
+    KMemoryBlock* block = KMemoryBlock::Allocate(m_kernel);
+    block->Initialize(addr_space_start, num_pages, 0, KMemoryState::Free, KMemoryPermission::None);
+    m_blocks.push_back(*block);
+}
+
+s64 KMemoryBlockManager::GetTotalCommittedMemory() {
+    u32 total_commited_memory{};
+    for (const auto& block : m_blocks) {
+        const KMemoryInfo info = block.GetInfo();
+        if (info.GetAddress() - 0x1C000000 >= 0x4000000 &&
+            True(info.GetState() & KMemoryState::Private)) {
+            total_commited_memory += info.GetSize();
+        }
+    }
+    return total_commited_memory;
+}
+
+KMemoryBlock* KMemoryBlockManager::FindFreeBlockInRegion(VAddr start, u32 num_pages,
+                                                         u32 block_num_pages) {
+    const VAddr end = start + (num_pages << Memory::CITRA_PAGE_BITS);
+    const u32 block_size = block_num_pages << Memory::CITRA_PAGE_BITS;
+    for (auto& block : m_blocks) {
+        const KMemoryInfo info = block.GetInfo();
+        if (info.GetState() != KMemoryState::Free) {
+            continue;
+        }
+        const VAddr block_start = std::max(info.GetAddress(), start);
+        const VAddr block_end = block_start + block_size;
+        if (block_end <= end && block_end <= info.GetEndAddress()) {
+            return std::addressof(block);
+        }
+    }
+    return nullptr;
+}
+
+void KMemoryBlockManager::CoalesceBlocks() {
+    auto it = m_blocks.begin();
+    while (true) {
+        iterator prev = it++;
+        if (it == m_blocks.end()) {
+            break;
+        }
+
+        // Merge adjacent blocks with the same properties.
+        if (prev->HasSameProperties(*it)) {
+            KMemoryBlock* block = std::addressof(*it);
+            const KMemoryInfo info = block->GetInfo();
+            prev->GrowBlock(info.GetAddress(), info.GetNumPages());
+            KMemoryBlock::Free(m_kernel, block);
+            m_blocks.erase(it);
+            it = prev;
+        }
+    }
+}
+
+ResultCode KMemoryBlockManager::MutateRange(VAddr addr, u32 num_pages, KMemoryState state,
+                                            KMemoryPermission perms, u32 tag) {
+    // Initialize iterators.
+    const VAddr last_addr = addr + (num_pages << Memory::CITRA_PAGE_BITS) - 1;
+    iterator begin = FindIterator(addr);
+    iterator end = FindIterator(last_addr);
+
+    // Before returning we have to coalesce.
+    SCOPE_EXIT({ this->CoalesceBlocks(); });
+
+    // Begin and end addresses are in different blocks. We need to shrink/remove
+    // any blocks in that range and insert a new one with the new attributes.
+    if (begin != end) {
+        // Any blocks in-between begin and end can be completely erased.
+        for (auto it = std::next(begin); it != end;) {
+            KMemoryBlock::Free(m_kernel, std::addressof(*it));
+            it = m_blocks.erase(it);
+        }
+
+        // If begin block has same properties, grow it to accomodate the range.
+        if (begin->HasProperties(state, perms, tag)) {
+            begin->GrowBlock(addr, num_pages);
+            // If the end block is fully overwritten, remove it.
+            if (end->GetLastAddress() == last_addr) {
+                KMemoryBlock::Free(m_kernel, std::addressof(*end));
+                m_blocks.erase(end);
+                R_SUCCEED();
+            }
+        } else if (end->HasProperties(state, perms, tag)) {
+            // If end block has same properties, grow it to accomodate the range.
+            end->GrowBlock(addr, num_pages);
+
+            // Remove start block if fully overwritten
+            if (begin->GetAddress() == addr) {
+                KMemoryBlock::Free(m_kernel, std::addressof(*begin));
+                m_blocks.erase(begin);
+                R_SUCCEED();
+            }
+        } else {
+            // Neither begin and end blocks have required properties.
+            // Shrink them both and create a new block in-between.
+            if (begin->IncludesRange(addr, num_pages)) {
+                KMemoryBlock::Free(m_kernel, std::addressof(*begin));
+                begin = m_blocks.erase(begin);
+            } else {
+                // Otherwise cut off the part that inside our range
+                begin->ShrinkBlock(addr, num_pages);
+            }
+
+            // If the end block is fully inside the range, remove it
+            if (end->IncludesRange(addr, num_pages)) {
+                KMemoryBlock::Free(m_kernel, std::addressof(*end));
+                end = m_blocks.erase(end);
+            } else {
+                // Otherwise cut off the part that inside our range
+                end->ShrinkBlock(addr, num_pages);
+            }
+
+            // The range [va, endVa] is now void, create new block in its place.
+            KMemoryBlock* block = KMemoryBlock::Allocate(m_kernel);
+            block->Initialize(addr, num_pages, 0, state, perms);
+
+            // Insert it to the block list
+            m_blocks.insert(end, *block);
+            R_SUCCEED();
+        }
+
+        // Shrink the block containing the start va
+        begin->ShrinkBlock(addr, num_pages);
+        R_SUCCEED();
+    }
+
+    // Start and end address are in same block, we have to split that.
+    if (!begin->HasProperties(state, perms, tag)) {
+        const KMemoryInfo info = begin->GetInfo();
+        const u32 pages_in_block = (addr - info.GetAddress()) >> Memory::CITRA_PAGE_BITS;
+
+        // Block has same starting address, we can just adjust the size.
+        if (info.GetAddress() == addr) {
+            // Block size matches, simply change attributes.
+            if (info.GetSize() == num_pages << Memory::CITRA_PAGE_BITS) {
+                begin->Initialize(addr, num_pages, tag, state, perms);
+                R_SUCCEED();
+            }
+            // Block size is bigger, split, insert new block after and update
+            begin->ShrinkBlock(addr, num_pages);
+            KMemoryBlock* block = KMemoryBlock::Allocate(m_kernel);
+            block->Initialize(addr, num_pages, tag, state, perms);
+
+            // Insert it to the block list.
+            m_blocks.insert(begin, *block);
+            R_SUCCEED();
+        }
+
+        // Same end address, but different base addr.
+        if (info.GetLastAddress() == last_addr) {
+            begin->ShrinkBlock(addr, num_pages);
+            KMemoryBlock* block = KMemoryBlock::Allocate(m_kernel);
+            block->Initialize(addr, num_pages, tag, state, perms);
+
+            // Insert it to the block list
+            m_blocks.insert(++begin, *block);
+            R_SUCCEED();
+        }
+
+        // Block fully contains start and end addresses. Shrink it to [last_addr, block_end] range.
+        begin->ShrinkBlock(0, num_pages + (addr >> Memory::CITRA_PAGE_BITS));
+
+        // Create a new block for [addr, last_addr] with the provided attributes.
+        KMemoryBlock* middle_block = KMemoryBlock::Allocate(m_kernel);
+        middle_block->Initialize(addr, num_pages, tag, state, perms);
+        begin = m_blocks.insert(begin, *middle_block);
+
+        // Create another block for the third range [block_addr, addr].
+        KMemoryBlock* start_block = KMemoryBlock::Allocate(m_kernel);
+        start_block->Initialize(info.GetAddress(), pages_in_block, 0, info.GetState(),
+                                info.GetPerms());
+        m_blocks.insert(begin, *start_block);
+    }
+
+    // We are done :)
+    R_SUCCEED();
+}
+
+void KMemoryBlockManager::Finalize() {
+    auto it = m_blocks.begin();
+    while (it != m_blocks.end()) {
+        KMemoryBlock::Free(m_kernel, std::addressof(*it));
+        it = m_blocks.erase(it);
+    }
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_memory_block_manager.h

@@ -0,0 +1,41 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/kernel/k_linked_list.h"
+#include "core/hle/kernel/k_memory_block.h"
+
+namespace Kernel {
+
+class KMemoryBlockManager final {
+    using BlockList = KLinkedList<KMemoryBlock>;
+    using iterator = BlockList::iterator;
+
+public:
+    explicit KMemoryBlockManager(KernelSystem& kernel) : m_kernel{kernel}, m_blocks{kernel} {}
+    ~KMemoryBlockManager() = default;
+
+    void Initialize(u32 addr_space_start, u32 addr_sce_end);
+    void Finalize();
+
+    void CoalesceBlocks();
+    s64 GetTotalCommittedMemory();
+    ResultCode MutateRange(VAddr addr, u32 num_pages, KMemoryState state, KMemoryPermission perms,
+                           u32 tag);
+
+    KMemoryBlock* GetMemoryBlockContainingAddr(u32 addr);
+    KMemoryBlock* FindFreeBlockInRegion(VAddr start, u32 num_pages, u32 block_num_pages);
+
+    iterator FindIterator(VAddr address) {
+        return std::find_if(m_blocks.begin(), m_blocks.end(),
+                            [address](auto& block) { return block.Contains(address); });
+    }
+
+private:
+    KernelSystem& m_kernel;
+    BlockList m_blocks;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_memory_manager.cpp

@@ -0,0 +1,73 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/k_memory_manager.h"
+
+namespace Kernel {
+
+void KMemoryManager::Initialize(FcramLayout* layout, u32 fcram_addr, u32 fcram_size) {
+    m_application_heap.Initialize(layout->application_addr, layout->application_size);
+    m_system_heap.Initialize(layout->system_addr, layout->system_size);
+    m_base_heap.Initialize(layout->base_addr, layout->base_size);
+    m_page_manager.Initialize(fcram_addr, fcram_size >> Memory::CITRA_PAGE_BITS);
+}
+
+u32 KMemoryManager::ConvertSharedMemPaLinearWithAppMemType(PAddr addr) {
+    int v2; // r1
+
+    const u32 fcram_offset = addr - Memory::FCRAM_PADDR;
+    if ((unsigned __int8)g_kernelSharedConfigPagePtr->appMemType == 7) {
+        v2 = 0x30000000;
+    } else {
+        v2 = 0x14000000;
+    }
+    return fcram_offset + v2;
+}
+
+VAddr KMemoryManager::AllocateContiguous(u32 num_pages, u32 page_alignment, MemoryOperation op) {
+    // KLightScopedMutex m{m_page_manager.GetMutex()};
+
+    if (True(op & MemoryOperation::Kernel)) {
+        m_page_manager.GetKernelMemoryUsage() += num_pages << Memory::CITRA_PAGE_BITS;
+    }
+
+    switch (op & MemoryOperation::RegionMask) {
+    case MemoryOperation::RegionApplication:
+        return m_application_heap.AllocateContiguous(num_pages, page_alignment);
+    case MemoryOperation::RegionSystem:
+        return m_system_heap.AllocateContiguous(num_pages, page_alignment);
+    case MemoryOperation::RegionBase:
+        return m_base_heap.AllocateContiguous(num_pages, page_alignment);
+    default:
+        UNREACHABLE();
+        return 0;
+    }
+}
+
+VAddr KMemoryManager::AllocateContiguousBackwards(u32 num_pages, MemoryOperation op) {
+    // KLightScopedMutex m{m_page_manager.GetMutex()};
+
+    if (True(op & MemoryOperation::Kernel)) {
+        m_page_manager.GetKernelMemoryUsage() += num_pages << Memory::CITRA_PAGE_BITS;
+    }
+
+    switch (op & MemoryOperation::RegionMask) {
+    case MemoryOperation::RegionApplication:
+        return m_application_heap.AllocateBackwards(num_pages);
+    case MemoryOperation::RegionSystem:
+        return m_system_heap.AllocateBackwards(num_pages);
+    case MemoryOperation::RegionBase:
+        return m_base_heap.AllocateBackwards(num_pages);
+    default:
+        UNREACHABLE();
+        return 0;
+    }
+}
+
+void KMemoryManager::FreeContiguousLocked(u32 addr, u32 num_pages) {
+    // KLightScopedMutex m{m_page_manager.GetMutex()};
+    m_page_manager.FreeContiguous(addr, num_pages, MemoryOperation::None);
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_memory_manager.h

@@ -0,0 +1,60 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/kernel/k_page_heap.h"
+#include "core/hle/kernel/k_page_manager.h"
+
+namespace Memory {
+class MemorySystem;
+}
+
+namespace Kernel {
+
+struct FcramLayout {
+    u32 application_addr;
+    u32 application_size;
+    u32 system_addr;
+    u32 system_size;
+    u32 base_addr;
+    u32 base_size;
+};
+
+class KMemoryManager {
+public:
+    explicit KMemoryManager(Memory::MemorySystem& memory)
+        : m_application_heap{memory}, m_system_heap{memory}, m_base_heap{memory}, m_page_manager{
+                                                                                      memory,
+                                                                                      this} {}
+    ~KMemoryManager() = default;
+
+    void Initialize(FcramLayout* layout, u32 fcram_addr, u32 fcram_size);
+
+    u32 ConvertSharedMemPaLinearWithAppMemType(PAddr addr);
+
+    KPageHeap& GetApplicationHeap() noexcept {
+        return m_application_heap;
+    }
+
+    KPageHeap& GetSystemHeap() noexcept {
+        return m_system_heap;
+    }
+
+    KPageHeap& GetBaseHeap() noexcept {
+        return m_base_heap;
+    }
+
+    VAddr AllocateContiguous(u32 num_pages, u32 page_alignment, MemoryOperation op);
+    VAddr AllocateContiguousBackwards(u32 num_pages, MemoryOperation op);
+    void FreeContiguousLocked(u32 addr, u32 num_pages);
+
+private:
+    KPageHeap m_application_heap;
+    KPageHeap m_system_heap;
+    KPageHeap m_base_heap;
+    KPageManager m_page_manager;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_page_group.cpp

@@ -0,0 +1,82 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/k_page_group.h"
+#include "core/hle/kernel/k_page_manager.h"
+
+namespace Kernel {
+
+KPageGroup::~KPageGroup() {
+    EraseAll();
+}
+
+void KPageGroup::AddRange(u32 addr, u32 num_pages) {
+    // If the provided range is empty there is nothing to do.
+    if (num_pages == 0 || addr + (num_pages << Memory::CITRA_PAGE_BITS) == 0) {
+        return;
+    }
+
+    // KScopedSchedulerLock lk{m_kernel};
+
+    // Attempt to coaelse with last block if possible.
+    if (!m_blocks.empty()) {
+        KBlockInfo& last = m_blocks.back();
+        if (addr != 0 && addr == last.GetEndAddress()) {
+            last.m_num_pages += num_pages;
+            return;
+        }
+    }
+
+    // Allocate and initialize the new block.
+    KBlockInfo* new_block = KBlockInfo::Allocate(m_kernel);
+    new_block->Initialize(addr, num_pages);
+
+    // Push the block to the list.
+    m_blocks.push_back(*new_block);
+}
+
+void KPageGroup::IncrefPages() {
+    // Iterate over block list and increment page reference counts.
+    for (const auto& block : m_blocks) {
+        m_page_manager->IncrefPages(block.GetAddress(), block.GetNumPages());
+    }
+}
+
+u32 KPageGroup::GetTotalNumPages() {
+    // Iterate over block list and count number of pages.
+    u32 total_num_pages{};
+    for (const auto& block : m_blocks) {
+        total_num_pages = block.GetNumPages();
+    }
+    return total_num_pages;
+}
+
+void KPageGroup::EraseAll() {
+    // Free all blocks referenced in the linked list.
+    auto it = m_blocks.begin();
+    while (it != m_blocks.end()) {
+        KBlockInfo::Free(m_kernel, std::addressof(*it));
+        it = m_blocks.erase(it);
+    }
+}
+
+bool KPageGroup::IsEquivalentTo(const KPageGroup& rhs) const {
+    auto lit = m_blocks.begin();
+    auto rit = rhs.m_blocks.begin();
+    auto lend = m_blocks.end();
+    auto rend = rhs.m_blocks.end();
+
+    while (lit != lend && rit != rend) {
+        if (*lit != *rit) {
+            return false;
+        }
+
+        ++lit;
+        ++rit;
+    }
+
+    return lit == lend && rit == rend;
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_page_group.h

@@ -0,0 +1,102 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/kernel/k_linked_list.h"
+#include "core/hle/kernel/slab_helpers.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+struct KBlockInfo final : public KSlabAllocated<KBlockInfo> {
+public:
+    explicit KBlockInfo() = default;
+    ~KBlockInfo() = default;
+
+    void Initialize(u32 address, u32 num_pages) {
+        m_base_address = address;
+        m_num_pages = num_pages;
+    }
+
+    constexpr u32 GetAddress() const {
+        return m_base_address;
+    }
+
+    constexpr u32 GetEndAddress() const {
+        return this->GetAddress() + this->GetSize();
+    }
+
+    constexpr u32 GetSize() const {
+        return m_num_pages << Memory::CITRA_PAGE_BITS;
+    }
+
+    constexpr u32 GetNumPages() const {
+        return m_num_pages;
+    }
+
+    constexpr bool IsEquivalentTo(const KBlockInfo& rhs) const {
+        return m_base_address == rhs.m_base_address && m_num_pages == rhs.m_num_pages;
+    }
+
+    constexpr bool operator==(const KBlockInfo& rhs) const {
+        return this->IsEquivalentTo(rhs);
+    }
+
+    constexpr bool operator!=(const KBlockInfo& rhs) const {
+        return !(*this == rhs);
+    }
+
+public:
+    u32 m_base_address;
+    u32 m_num_pages;
+};
+
+class KPageManager;
+class KernelSystem;
+
+class KPageGroup {
+    using BlockInfoList = KLinkedList<KBlockInfo>;
+    using iterator = BlockInfoList::const_iterator;
+
+public:
+    explicit KPageGroup(KernelSystem& kernel, KPageManager* page_manager)
+        : m_kernel{kernel}, m_page_manager{page_manager}, m_blocks{kernel} {}
+    ~KPageGroup();
+
+    iterator begin() const {
+        return this->m_blocks.begin();
+    }
+    iterator end() const {
+        return this->m_blocks.end();
+    }
+    bool empty() const {
+        return this->m_blocks.empty();
+    }
+
+    void AddRange(u32 addr, u32 num_pages);
+    void IncrefPages();
+
+    void EraseAll();
+    void FreeMemory();
+
+    u32 GetTotalNumPages();
+
+    bool IsEquivalentTo(const KPageGroup& rhs) const;
+
+    bool operator==(const KPageGroup& rhs) const {
+        return this->IsEquivalentTo(rhs);
+    }
+
+    bool operator!=(const KPageGroup& rhs) const {
+        return !(*this == rhs);
+    }
+
+private:
+    KernelSystem& m_kernel;
+    KPageManager* m_page_manager{};
+    BlockInfoList m_blocks;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_page_heap.cpp

@@ -0,0 +1,408 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/k_page_heap.h"
+
+namespace Kernel {
+
+void KPageHeap::Initialize(VAddr region_start, u32 region_size) {
+    m_region_start = region_start;
+    m_region_size = region_size;
+
+    // Retrieve the first block in the provided region.
+    Block* first_block = m_memory.GetPointer<Block>(m_region_start);
+    ASSERT(first_block);
+
+    // Initialize the block.
+    first_block->num_pages = this->GetNumPages();
+    first_block->current = first_block;
+
+    // Insert the block to our block list.
+    m_blocks.push_front(*first_block);
+}
+
+u32 KPageHeap::GetTotalNumPages() {
+    // Iterate over the blocks.
+    u32 total_num_pages{};
+    for (const auto& block : m_blocks) {
+        total_num_pages = block.num_pages;
+    }
+    return total_num_pages;
+}
+
+void KPageHeap::FreeBlock(u32 addr, u32 num_pages) {
+    // Return if there are no pages to free.
+    if (num_pages == 0) {
+        return;
+    }
+
+    // Return if unable to insert block at the beginning.
+    auto start_block = std::addressof(m_blocks.front());
+    if (this->TryInsert(addr, num_pages, nullptr, start_block)) {
+        return;
+    }
+
+    // Iterate over the blocks.
+    for (auto it = m_blocks.begin(); it != m_blocks.end();) {
+        // Attempt to insert.
+        Block* block = std::addressof(*it++);
+        Block* next_block = std::addressof(*it++);
+        if (this->TryInsert(addr, num_pages, block, next_block)) {
+            break;
+        }
+    }
+}
+
+void* KPageHeap::AllocateBackwards(u32 size) {
+    // Ensure allocation is possible.
+    if (size == 0) [[unlikely]] {
+        return nullptr;
+    }
+
+    // Iterate over block list backwards.
+    u32 remaining = size;
+    for (auto it = m_blocks.rbegin(); it != m_blocks.rend(); it++) {
+        // If block does not cover remaining pages continue.
+        auto block = std::addressof(*it);
+        const u32 num_pages = block->num_pages;
+        if (remaining > num_pages) {
+            remaining -= num_pages;
+            continue;
+        }
+
+        // Split last block at our boundary.
+        const u32 new_block_pages = num_pages - remaining;
+        auto new_block = this->SplitBlock(block, new_block_pages);
+        ASSERT(new_block && new_block->num_pages == new_block_pages);
+
+        // new_block.prev = 0;
+        this->SetLastBlock(block);
+
+        // Return final block which points to our allocated memory.
+        return new_block;
+    }
+
+    return nullptr;
+}
+
+void* KPageHeap::AllocateContiguous(u32 size, u32 page_alignment) {
+    KPageHeapBlock* next;  // r6
+    KPageHeapBlock* v13;   // r1
+    KPageHeapBlock* prev;  // [sp+0h] [bp-30h]
+    KPageHeapBlock* block; // [sp+4h] [bp-2Ch]
+
+    // Ensure allocation is possible.
+    if (m_blocks.empty() || size == 0) [[unlikely]] {
+        return nullptr;
+    }
+
+    for (auto it = m_blocks.begin(); it != m_blocks.end(); it++) {
+        // Ensure block is valid.
+        auto block = std::addressof(*it);
+        this->ValidateBlock(block);
+    }
+
+    KPageHeapBlock* current_node = m_link.next;
+    while (current_node) {
+        u32 misalignment = 0;
+        KPageHeap::ValidateBlock(current_node);
+        const u32 num_pages = current_node->num_pages;
+        // if (current_node->num_pages > this->GetNumPages() || this->GetRegionEnd() < (unsigned
+        // int)current_node + 4096 * num_pages) {
+        //     UNREACHABLE();
+        // }
+        if (page_alignment > 1) {
+            const u32 v11 = ((unsigned int)current_node >> 12) % page_alignment;
+            if (v11) {
+                misalignment = page_alignment - v11;
+            }
+        }
+
+        if (size + misalignment <= num_pages) {
+            block = current_node;
+            if (misalignment) {
+                block = KPageHeap::SplitBlock(current_node, misalignment);
+            }
+
+            KPageHeap::SplitBlock(block, size);
+            KPageHeap::ValidateBlock(block);
+            prev = block->link.prev;
+            next = block->link.next;
+            KPageHeap::ValidateBlock(prev);
+            KPageHeap::ValidateBlock(next);
+
+            if (prev) {
+                prev->link.next = next;
+                v13 = prev;
+            } else {
+                m_link.next = next;
+                if (!next) {
+                    m_link.prev = 0;
+                    goto LABEL_28;
+                }
+                m_link.next->link.prev = 0;
+                v13 = m_link.next;
+            }
+            KPageHeap::UpdateBlockMac(v13);
+            if (next) {
+                next->link.prev = prev;
+                KPageHeap::UpdateBlockMac(next);
+            LABEL_29:
+                if (block->num_pages != size) {
+                    UNREACHABLE();
+                }
+                return block;
+            }
+        LABEL_28:
+            KPageHeap::SetLastBlock(prev);
+            goto LABEL_29;
+        }
+        current_node = current_node->link.next;
+    }
+
+    for (KPageHeapBlock* j = m_link.next; j; j = j->link.next) {
+        KPageHeap::ValidateBlock(j);
+    }
+
+    return 0;
+}
+
+void KPageHeap::SetLastBlock(KPageHeapBlock* block) {
+    m_link.prev = block;
+    if (!block) [[unlikely]] {
+        m_link.next = nullptr;
+        return;
+    }
+
+    /*u32 v2 = m_key[0];
+    u32 v3 = m_key[1];
+    u32 v4 = m_key[2];
+    u32 v5 = m_key[3];
+    for (int i = 0; i < 2; i++) {
+        int v7 = 0;
+        do {
+            v2 -= *(u32 *)((char *)&block->num_pages + v7) - __ROR4__(v3, 3);
+            v7 += 4;
+            v3 -= __ROR4__(v4, (v5 & 0xF) + 3) ^ __ROR4__(v5, (v2 & 0xF) + 13);
+            v4 -= __ROR4__(v5, v2) * v3;
+            v5 -= __ROR4__(v2, v3) * v4;
+        } while ( v7 < 20 );
+    }
+
+    if ((v2 ^ v3) != block->mac) {
+        UNREACHABLE();
+    }*/
+
+    m_link.prev->link.next = nullptr;
+}
+
+KPageHeap::Block* KPageHeap::SplitBlock(Block* block, u32 new_block_size) {
+    const u32 num_pages = block->num_pages;
+    ASSERT(block->num_pages <= this->GetNumPages());
+    // if (block->num_pages > this->GetNumPages() || this->GetRegionEnd() < (unsigned int)block +
+    // 4096 * num_pages) {
+    //     UNREACHABLE();
+    // }
+
+    if (!new_block_size || num_pages == new_block_size) [[unlikely]] {
+        return nullptr;
+    }
+
+    Block* new_block = (Block*)((char*)block + Memory::CITRA_PAGE_SIZE * new_block_size);
+    Block* next = block->link.next;
+    const u32 v12 = num_pages - new_block_size;
+    new_block->nonce = 0;
+    new_block->num_pages = v12;
+    new_block->mac = 0;
+    new_block->link.next = next;
+    new_block->link.prev = block;
+    new_block->current = new_block;
+
+    if (new_block->num_pages != v12) {
+        UNREACHABLE();
+    }
+
+    block->link.next = new_block;
+    block->num_pages = new_block_size;
+
+    if (block->num_pages != new_block_size) {
+        UNREACHABLE();
+    }
+
+    KPageHeapBlock* v13 = new_block->link.next;
+    KPageHeap::ValidateBlock(v13);
+    if (v13) {
+        v13->link.prev = new_block;
+        KPageHeap::UpdateBlockMac(v13);
+    } else {
+        KPageHeap::SetLastBlock(new_block);
+    }
+
+    return new_block;
+}
+
+bool KPageHeap::TryInsert(u32 freedAddr, u32 freedNumPages, Block* prev, Block* next) {
+    KPageHeapBlock* v6;    // r5
+    u32 numPages;          // r8
+    u32 freedAddrEnd;      // r11
+    u32 regionSize;        // r0
+    u32 prevRightNeighour; // r10
+    KPageHeapBlock* nxt;   // r9
+    u32 regionStart;       // r0
+    bool v14;              // cc
+    BOOL result;           // r0
+    bool v16;              // zf
+    u32 v17;               // r8
+    u32 v18;               // r1
+    u32 v19;               // r9
+    KPageHeapBlock* next;  // r5
+    u32 v21;               // r3
+    u32 v22;               // r0
+    u32 v23;               // r1
+    u32 v24;               // r2
+    int i;                 // r11
+    int v26;               // r12
+    int v27;               // r10
+    bool v28;              // zf
+    u32 v29;               // r4
+    u32 v30;               // r1
+    u32 v31;               // r8
+    KPageHeapBlock* v32;   // r4
+    u32 v33;               // [sp+4h] [bp-3Ch]
+
+    v6 = (KPageHeapBlock*)freedAddr;
+    numPages = 0;
+    freedAddrEnd = freedAddr + (freedNumPages << 12);
+    v33 = 0;
+    regionSize = this->regionSize;
+    if (freedNumPages > regionSize >> 12 || regionSize + this->regionStart < freedAddrEnd)
+        kernelpanic();
+    if (prev) {
+        KPageHeap::ValidateBlock(this, prev);
+        numPages = prev->numPages;
+    }
+    if (next) {
+        KPageHeap::ValidateBlock(this, next);
+        v33 = next->numPages;
+    }
+    if (prev) {
+        if ((KPageHeapBlock*)((char*)prev + 4096 * prev->numPages - 1) < prev)
+            kernelpanic();
+        prevRightNeighour = (u32)prev + 4096 * prev->numPages;
+    } else {
+        prevRightNeighour = this->regionStart;
+    }
+    if (next)
+        nxt = next;
+    else
+        nxt = (KPageHeapBlock*)(this->regionStart + this->regionSize);
+    regionStart = this->regionStart;
+    if (regionStart > prevRightNeighour || regionStart + this->regionSize < (unsigned int)nxt)
+        kernelpanic();
+    v14 = prevRightNeighour > (unsigned int)v6;
+    if (prevRightNeighour <= (unsigned int)v6)
+        v14 = freedAddrEnd > (unsigned int)nxt;
+    result = 0;
+    if (!v14) {
+        v6->nonce = 0;
+        v6->link.prev = prev;
+        v6->mac = 0;
+        v6->numPages = freedNumPages;
+        v6->link.next = next;
+        v6->current = v6;
+        KPageHeap::UpdateBlockMac(this, v6);
+        if (v6->numPages != freedNumPages)
+            kernelpanic();
+        if (prev) {
+            prev->link.next = v6;
+            KPageHeap::UpdateBlockMac(this, prev);
+            if (prev->numPages != numPages)
+                kernelpanic();
+        } else {
+            this->link.next = v6;
+            if (v6) {
+                v21 = this->key[2];
+                v22 = this->key[0];
+                v23 = this->key[1];
+                v24 = this->key[3];
+                for (i = 0; i < 2; ++i) {
+                    v26 = 0;
+                    do {
+                        v27 = *(u32*)((char*)&v6->numPages + v26);
+                        v26 += 4;
+                        v22 -= v27 - __ROR4__(v23, 3);
+                        v23 -= __ROR4__(v21, (v24 & 0xF) + 3) ^ __ROR4__(v24, (v22 & 0xF) + 13);
+                        v21 -= __ROR4__(v24, v22) * v23;
+                        v24 -= __ROR4__(v22, v23) * v21;
+                    } while (v26 < 20);
+                }
+                if ((v22 ^ v23) != v6->mac)
+                    kernelpanic();
+                this->link.next->link.prev = 0;
+                KPageHeap::UpdateBlockMac(this, this->link.next);
+            } else {
+                this->link.prev = 0;
+            }
+        }
+        if (next) {
+            next->link.prev = v6;
+            KPageHeap::UpdateBlockMac(this, next);
+            if (next->numPages != v33)
+                kernelpanic();
+        } else {
+            KPageHeap::SetLastBlock(this, v6);
+        }
+        v16 = prev == 0;
+        if (prev)
+            v16 = v6 == 0;
+        if (!v16 && (KPageHeapBlock*)((char*)prev + 4096 * prev->numPages) == v6) {
+            KPageHeap::ValidateBlock(this, prev);
+            v17 = prev->numPages;
+            KPageHeap::ValidateBlock(this, v6);
+            v18 = prev->numPages;
+            v19 = v6->numPages;
+            prev->link.next = v6->link.next;
+            prev->numPages = v18 + v19;
+            KPageHeap::UpdateBlockMac(this, prev);
+            if (prev->numPages != v17 + v19)
+                kernelpanic();
+            next = v6->link.next;
+            KPageHeap::ValidateBlock(this, next);
+            if (next) {
+                next->link.prev = prev;
+                KPageHeap::UpdateBlockMac(this, next);
+            } else {
+                KPageHeap::SetLastBlock(this, prev);
+            }
+            v6 = prev;
+        }
+        v28 = v6 == 0;
+        if (v6)
+            v28 = next == 0;
+        if (!v28 && (KPageHeapBlock*)((char*)v6 + 4096 * v6->numPages) == next) {
+            KPageHeap::ValidateBlock(this, v6);
+            v29 = v6->numPages;
+            KPageHeap::ValidateBlock(this, next);
+            v30 = v6->numPages;
+            v31 = next->numPages;
+            v6->link.next = next->link.next;
+            v6->numPages = v30 + v31;
+            KPageHeap::UpdateBlockMac(this, v6);
+            if (v6->numPages != v29 + v31)
+                kernelpanic();
+            v32 = next->link.next;
+            KPageHeap::ValidateBlock(this, v32);
+            if (v32) {
+                v32->link.prev = v6;
+                KPageHeap::UpdateBlockMac(this, v32);
+            } else {
+                KPageHeap::SetLastBlock(this, v6);
+            }
+        }
+        return 1;
+    }
+    return result;
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_page_heap.h

@@ -0,0 +1,64 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/intrusive_list.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+class KPageHeap final {
+public:
+    explicit KPageHeap(Memory::MemorySystem& memory) : m_memory{memory} {}
+    ~KPageHeap() = default;
+
+    constexpr u32 GetNumPages() const {
+        return m_region_size >> Memory::CITRA_PAGE_BITS;
+    }
+
+    constexpr u32 GetRegionStart() const {
+        return m_region_start;
+    }
+
+    constexpr u32 GetRegionEnd() const {
+        return m_region_start + m_region_size;
+    }
+
+    constexpr bool Contains(u32 addr) const {
+        return this->GetRegionStart() <= addr && addr < this->GetRegionEnd();
+    }
+
+public:
+    void Initialize(VAddr region_start, u32 region_size);
+    u32 GetTotalNumPages();
+
+    VAddr AllocateBackwards(u32 size);
+    VAddr AllocateContiguous(u32 size, u32 page_alignment);
+    void FreeBlock(u32 addr, u32 num_pages);
+
+private:
+    struct Block final : public Common::IntrusiveListBaseNode<Block> {
+        u32 num_pages;
+        Block* current;
+        u32 nonce;
+        u32 mac;
+    };
+
+    using BlockList = Common::IntrusiveListBaseTraits<Block>::ListType;
+    using iterator = BlockList::iterator;
+
+    Block* SplitBlock(Block* block, u32 new_block_size);
+    bool TryInsert(u32 freed_addr, u32 num_freed_pages, Block* prev_block, Block* next_block);
+    void SetLastBlock(Block* block);
+
+private:
+    BlockList m_blocks{};
+    Memory::MemorySystem& m_memory;
+    u32 m_region_start{};
+    u32 m_region_size{};
+    std::array<u32, 4> m_key{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_page_manager.cpp

@@ -0,0 +1,107 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/k_memory_manager.h"
+#include "core/hle/kernel/k_page_manager.h"
+#include "core/memory.h"
+
+namespace Kernel {
+
+void KPageManager::Initialize(u32 start_addr, u32 num_pages) {
+    // Initialize page manager address range.
+    m_start_addr = start_addr;
+    m_num_pages = num_pages;
+
+    // Compute the number of pages to allocate from the base heap.
+    const u32 num_ref_counts_pages = ((sizeof(u32) * num_pages - 1) >> Memory::CITRA_PAGE_BITS) + 1;
+    auto& base_heap = m_memory_manager->GetBaseHeap();
+
+    // Allocate page refcounting memory.
+    u32 ref_counts_addr{};
+    {
+        // KLightScopedMutex m{m_mutex};
+        m_kernel_memory_usage += num_ref_counts_pages << Memory::CITRA_PAGE_BITS;
+        ref_counts_addr = base_heap.AllocateContiguous(num_ref_counts_pages, 0);
+        m_page_ref_counts = m_memory.GetPointer<u32>(ref_counts_addr);
+        ASSERT(m_page_ref_counts);
+    }
+
+    // Zero-initialize reference counts.
+    if (num_pages) {
+        std::memset(m_page_ref_counts, 0, num_ref_counts_pages << Memory::CITRA_PAGE_BITS);
+    }
+
+    // Track allocated pages.
+    this->IncrefPages(ref_counts_addr, num_ref_counts_pages);
+}
+
+void KPageManager::IncrefPages(u32 addr, u32 num_pages) {
+    // KLightScopedMutex m{m_mutex};
+
+    // Increment page reference counts.
+    const u32 page_start = (addr - m_start_addr) >> Memory::CITRA_PAGE_BITS;
+    const u32 page_end = num_pages + page_start;
+    for (u32 page = page_start; page < page_end; page++) {
+        m_page_ref_counts[page_start]++;
+    }
+}
+
+void KPageManager::FreeContiguous(u32 addr, u32 num_pages, MemoryOperation op) {
+    // Ensure the provided address is in range.
+    const u32 page_start = (addr - m_start_addr) >> Memory::CITRA_PAGE_BITS;
+    const u32 page_end = page_start + num_pages;
+    if (page_start >= page_end) [[unlikely]] {
+        return;
+    }
+
+    // Retrieve page heaps from the memory manager.
+    auto& application_heap = m_memory_manager->GetApplicationHeap();
+    auto& base_heap = m_memory_manager->GetBaseHeap();
+    auto& system_heap = m_memory_manager->GetSystemHeap();
+
+    // Frees the range of pages provided from the appropriate heap.
+    const auto FreePages = [&](u32 start_page, u32 num_pages) {
+        const u32 current_addr = m_start_addr + (start_page << Memory::CITRA_PAGE_BITS);
+        if (base_heap.Contains(current_addr)) {
+            base_heap.FreeBlock(current_addr, num_pages);
+        } else if (system_heap.Contains(current_addr)) {
+            system_heap.FreeBlock(current_addr, num_pages);
+        } else {
+            application_heap.FreeBlock(current_addr, num_pages);
+        }
+        // Update kernel memory usage if requested.
+        if (True(op & MemoryOperation::Kernel)) {
+            m_kernel_memory_usage -= num_pages << Memory::CITRA_PAGE_BITS;
+        }
+    };
+
+    // Iterate over the range of pages to free.
+    u32 start_free_page = 0;
+    u32 num_pages_to_free = 0;
+    for (u32 page = page_start; page < page_end; page++) {
+        const u32 new_count = --m_page_ref_counts[page];
+        if (new_count) {
+            // Nothing to free, continue to next page.
+            if (num_pages_to_free <= 0) {
+                continue;
+            }
+            // Free accumulated pages and reset.
+            FreePages(start_free_page, num_pages_to_free);
+            num_pages_to_free = 0;
+        } else if (num_pages_to_free <= 0) {
+            start_free_page = page;
+            num_pages_to_free = 1;
+        } else {
+            // Advance number of pages to free.
+            num_pages_to_free++;
+        }
+    }
+
+    // Free any remaining pages.
+    if (num_pages_to_free > 0) {
+        FreePages(start_free_page, num_pages_to_free);
+    }
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_page_manager.h

@@ -0,0 +1,63 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Memory {
+class MemorySystem;
+}
+
+namespace Kernel {
+
+enum class MemoryOperation : u32 {
+    None = 0x0,
+    RegionApplication = 0x100,
+    RegionSystem = 0x200,
+    RegionBase = 0x300,
+    Kernel = 0x80000000,
+    RegionBaseKernel = Kernel | RegionBase,
+    Free = 0x1,
+    Reserve = 0x2,
+    Alloc = 0x3,
+    Map = 0x4,
+    Unmap = 0x5,
+    Prot = 0x6,
+    OpMask = 0xFF,
+    RegionMask = 0xF00,
+    LinearFlag = 0x10000,
+};
+DECLARE_ENUM_FLAG_OPERATORS(MemoryOperation)
+
+class KMemoryManager;
+
+class KPageManager {
+public:
+    explicit KPageManager(Memory::MemorySystem& memory, KMemoryManager* memory_manager)
+        : m_memory{memory}, m_memory_manager{memory_manager} {}
+    ~KPageManager() = default;
+
+    std::atomic<u32>& GetKernelMemoryUsage() noexcept {
+        return m_kernel_memory_usage;
+    }
+
+    void Initialize(u32 start_addr, u32 num_pages);
+    void IncrefPages(u32 addr, u32 num_pages);
+    void FreeContiguous(u32 data, u32 num_pages, MemoryOperation op);
+
+private:
+    Memory::MemorySystem& m_memory;
+    KMemoryManager* m_memory_manager{};
+    u32 m_start_addr{};
+    u32 m_num_pages{};
+    u32* m_page_ref_counts{};
+    std::atomic<u32> m_kernel_memory_usage{};
+    // KLightMutex m_mutex;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_page_table.h

@@ -0,0 +1,134 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "core/hle/kernel/k_memory_block_manager.h"
+#include "core/hle/result.h"
+
+namespace Common {
+class PageTable;
+}
+
+namespace Kernel {
+
+enum class KMemoryUpdateFlags {
+    None = 0x0,
+    State = 0x1,
+    Perms = 0x100,
+    StateAndPerms = State | Perms,
+};
+DECLARE_ENUM_FLAG_OPERATORS(KMemoryUpdateFlags)
+
+enum class MemoryOperation : u32;
+
+class KPageGroup;
+class KPageManager;
+
+class KPageTable {
+public:
+    explicit KPageTable(KernelSystem& kernel, KPageManager* page_manager)
+        : m_kernel{kernel}, m_page_manager{page_manager}, m_memory_block_manager{kernel} {}
+    ~KPageTable() = default;
+
+    Common::PageTable& GetImpl() {
+        return *m_impl;
+    }
+
+    void InitizalizeL1Table(u32** outL1TablePtr, u32* L1Table);
+
+    ResultCode CheckAndUpdateAddrRangeMaskedStateAndPerms(
+        u32 addr, u32 num_pages, KMemoryState state_mask, KMemoryState expected_state,
+        KMemoryPermission min_perms, KMemoryState new_state, KMemoryPermission new_perms);
+    ResultCode CheckAddressRangeSizeAndState(u32 addr, u32 size, KMemoryState state);
+    ResultCode CheckAddressRangeSizeAndStateFlags(u32 addr, u32 size, KMemoryState stateMask,
+                                                  KMemoryState expectedStateFlags);
+    ResultCode CheckMemoryBlockAttributes(u32 addr, u32 size, KMemoryState state,
+                                          KMemoryPermission perms);
+    ResultCode CheckAddrRangeMaskedStateAndPerms(u32 addr, u32 size, KMemoryState stateMask,
+                                                 KMemoryState expectedState,
+                                                 KMemoryPermission minPerms);
+    ResultCode CheckAndChangeGroupStateAndPerms(u32 addr, KPageGroup* pgGroup,
+                                                KMemoryState stateMask, KMemoryState expectedState,
+                                                KMemoryPermission minPerms, KMemoryState newState,
+                                                KMemoryPermission newPerms);
+
+    ResultCode MapL2Entries(u32 va, u32 pa, u32 numPages_reused, u32* attribsPtr, bool isLarge);
+    ResultCode MapL1Entries(u32 va, u32 pa, u32 numPages, u32* attribsPtr, bool isLarge);
+    ResultCode MapContiguousPhysicalAddressRange(u32 va, u32 pa, u32 numPages, u32* mmuAttribs);
+    ResultCode MergeContiguousEntries(u32 va);
+    ResultCode MapNewlyAllocatedPhysicalAddressRange(u32 va, u32 pa, u32 numPages, u32* mmuAttrbis);
+
+    ResultCode RemapMemoryInterprocess(KPageTable* dstPgTbl, KPageTable* srcPgTbl, u32 dstAddr,
+                                       u32 srcAddr, u32 numPages, KMemoryState dstMemState,
+                                       KMemoryPermission dstMemPerms);
+
+    ResultCode ChangePageAttributes(u32 addr, u32 size, u32* mmuAttribs);
+    ResultCode CheckAndUnmapPageGroup(u32 addr, KPageGroup* pgGroup);
+
+    ResultCode CreateAlias(u32 srcAddr, u32 dstAddr, u32 numPages, KMemoryState expectedStateSrc,
+                           KMemoryPermission expectedMinPermsSrc, KMemoryState newStateSrc,
+                           KMemoryPermission newPermsSrc, KMemoryState newStateDst,
+                           KMemoryPermission newPermsDst);
+    ResultCode DestroyAlias(u32 srcAddr, u32 dstAddr, u32 numPages, KMemoryState expectedStateSrc,
+                            KMemoryPermission expectedMinPermsSrc, KMemoryState expectedStateDst,
+                            KMemoryPermission expectedMinPermsDst, KMemoryState newStateSrc,
+                            KMemoryPermission newPermsSrc);
+
+    void Unmap(u32 addr, u32 numPages);
+    void UnmapEntries(u32 currentVa, u32 numPages, KPageGroup* outPgGroupUnmapped);
+
+    ResultCode OperateOnGroup(u32 addr, KPageGroup* pgGroup, KMemoryState state,
+                              KMemoryPermission perms, KMemoryUpdateFlags updateFlags);
+    ResultCode OperateOnAnyFreeBlockInRegionWithGuardPage(u32* outAddr, u32 blockNumPages,
+                                                          u32 regionStart, u32 regionNumPages,
+                                                          u32 pa, KMemoryState state,
+                                                          KMemoryPermission perms,
+                                                          KMemoryUpdateFlags updateFlags,
+                                                          MemoryOperation region);
+    ResultCode Operate(u32 va, u32 numPages, u32 pa, KMemoryState state, KMemoryPermission perms,
+                       KMemoryUpdateFlags updateFlags, MemoryOperation region);
+
+    ResultCode MakePageGroup(KPageGroup& pg, VAddr addr, u32 num_pages);
+    ResultCode QueryInfo(KMemoryInfo* outMemoryInfo, u32* pageInfo, u32 addr);
+    ResultCode CopyMemoryInterprocessForIpc(u32 dstAddr, KPageTable* srcPgTbl, u32 srcAddr,
+                                            u32 size);
+    ResultCode SplitContiguousEntries(u32 va, u32 size);
+
+    u32 ConvertVaToPa(u32** L1TablePtr, u32 va);
+
+    void InvalidateAllTlbEntries();
+    void InvalidateEntireInstructionCache();
+    void InvalidateEntireInstructionCacheLocal();
+    void InvalidateTlbEntryByMva(u32 addr);
+    void InvalidateDataCacheRange(u32 addr, u32 size);
+    void InvalidateDataCacheRangeLocal(u32 addr, u32 size);
+
+    void CleanInvalidateEntireDataCacheLocal();
+    void CleanInvalidateDataCacheRangeLocal(u32 addr, u32 size);
+    void CleanInvalidateDataCacheRange(u32 addr, u32 size);
+    void CleanInvalidateInstructionCacheRange(u32 addr, u32 size);
+    void CleanInvalidateEntireDataCache();
+    void CleanDataCacheRange(u32 addr, u32 size);
+
+private:
+    KernelSystem& m_kernel;
+    KPageManager* m_page_manager;
+    // KLightMutex mutex;
+    std::unique_ptr<Common::PageTable> m_impl{};
+    std::array<bool, 4> m_tlb_needs_invalidating{};
+    KMemoryBlockManager m_memory_block_manager;
+    u32 m_translation_table_base{};
+    u8 m_asid{};
+    bool m_is_kernel{};
+    bool m_use_small_pages{};
+    u32 m_address_space_start{};
+    u32 m_address_space_end{};
+    u32 m_linear_address_range_start{};
+    u32 m_translation_table_size{};
+    u32* m_l1_table{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_slab_heap.h

@@ -0,0 +1,191 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+
+#include "common/assert.h"
+#include "common/atomic_ops.h"
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Kernel {
+
+class KernelSystem;
+
+namespace impl {
+
+class KSlabHeapImpl {
+    CITRA_NON_COPYABLE(KSlabHeapImpl);
+    CITRA_NON_MOVEABLE(KSlabHeapImpl);
+
+public:
+    struct Node {
+        Node* next{};
+    };
+
+public:
+    constexpr KSlabHeapImpl() = default;
+
+    void Initialize() {
+        ASSERT(m_head == nullptr);
+    }
+
+    Node* GetHead() const {
+        return m_head;
+    }
+
+    void* Allocate() {
+        Node* ret = m_head;
+        if (ret != nullptr) [[likely]] {
+            m_head = ret->next;
+        }
+        return ret;
+    }
+
+    void Free(void* obj) {
+        Node* node = static_cast<Node*>(obj);
+        node->next = m_head;
+        m_head = node;
+    }
+
+private:
+    std::atomic<Node*> m_head{};
+};
+
+} // namespace impl
+
+class KSlabHeapBase : protected impl::KSlabHeapImpl {
+    CITRA_NON_COPYABLE(KSlabHeapBase);
+    CITRA_NON_MOVEABLE(KSlabHeapBase);
+
+private:
+    size_t m_obj_size{};
+    uintptr_t m_peak{};
+    uintptr_t m_start{};
+    uintptr_t m_end{};
+
+private:
+    void UpdatePeakImpl(uintptr_t obj) {
+        const uintptr_t alloc_peak = obj + this->GetObjectSize();
+        uintptr_t cur_peak = m_peak;
+        do {
+            if (alloc_peak <= cur_peak) {
+                break;
+            }
+        } while (
+            !Common::AtomicCompareAndSwap(std::addressof(m_peak), alloc_peak, cur_peak, cur_peak));
+    }
+
+public:
+    constexpr KSlabHeapBase() = default;
+
+    bool Contains(uintptr_t address) const {
+        return m_start <= address && address < m_end;
+    }
+
+    void Initialize(size_t obj_size, void* memory, size_t memory_size) {
+        // Ensure we don't initialize a slab using null memory.
+        ASSERT(memory != nullptr);
+
+        // Set our object size.
+        m_obj_size = obj_size;
+
+        // Initialize the base allocator.
+        KSlabHeapImpl::Initialize();
+
+        // Set our tracking variables.
+        const size_t num_obj = (memory_size / obj_size);
+        m_start = reinterpret_cast<uintptr_t>(memory);
+        m_end = m_start + num_obj * obj_size;
+        m_peak = m_start;
+
+        // Free the objects.
+        u8* cur = reinterpret_cast<u8*>(m_end);
+
+        for (size_t i = 0; i < num_obj; i++) {
+            cur -= obj_size;
+            KSlabHeapImpl::Free(cur);
+        }
+    }
+
+    size_t GetSlabHeapSize() const {
+        return (m_end - m_start) / this->GetObjectSize();
+    }
+
+    size_t GetObjectSize() const {
+        return m_obj_size;
+    }
+
+    void* Allocate() {
+        void* obj = KSlabHeapImpl::Allocate();
+        return obj;
+    }
+
+    void Free(void* obj) {
+        // Don't allow freeing an object that wasn't allocated from this heap.
+        const bool contained = this->Contains(reinterpret_cast<uintptr_t>(obj));
+        ASSERT(contained);
+        KSlabHeapImpl::Free(obj);
+    }
+
+    size_t GetObjectIndex(const void* obj) const {
+        return (reinterpret_cast<uintptr_t>(obj) - m_start) / this->GetObjectSize();
+    }
+
+    size_t GetPeakIndex() const {
+        return this->GetObjectIndex(reinterpret_cast<const void*>(m_peak));
+    }
+
+    uintptr_t GetSlabHeapAddress() const {
+        return m_start;
+    }
+
+    size_t GetNumRemaining() const {
+        // Only calculate the number of remaining objects under debug configuration.
+        return 0;
+    }
+};
+
+template <typename T>
+class KSlabHeap final : public KSlabHeapBase {
+private:
+    using BaseHeap = KSlabHeapBase;
+
+public:
+    constexpr KSlabHeap() = default;
+
+    void Initialize(void* memory, size_t memory_size) {
+        BaseHeap::Initialize(sizeof(T), memory, memory_size);
+    }
+
+    T* Allocate() {
+        T* obj = static_cast<T*>(BaseHeap::Allocate());
+
+        if (obj != nullptr) [[likely]] {
+            std::construct_at(obj);
+        }
+        return obj;
+    }
+
+    T* Allocate(KernelSystem& kernel) {
+        T* obj = static_cast<T*>(BaseHeap::Allocate());
+
+        if (obj != nullptr) [[likely]] {
+            std::construct_at(obj, kernel);
+        }
+        return obj;
+    }
+
+    void Free(T* obj) {
+        BaseHeap::Free(obj);
+    }
+
+    size_t GetObjectIndex(const T* obj) const {
+        return BaseHeap::GetObjectIndex(obj);
+    }
+};
+
+} // namespace Kernel

src/core/hle/kernel/kernel.cpp

@@ -8,9 +8,11 @@
 #include "common/archives.h"
 #include "common/serialization/atomic.h"
 #include "core/hle/kernel/client_port.h"
-#include "core/hle/kernel/config_mem.h"
-#include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/ipc_debugger/recorder.h"
+#include "core/hle/kernel/k_linked_list.h"
+#include "core/hle/kernel/k_memory_block.h"
+#include "core/hle/kernel/k_page_group.h"
+#include "core/hle/kernel/k_slab_heap.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/memory.h"
 #include "core/hle/kernel/process.h"
@@ -29,6 +31,7 @@ KernelSystem::KernelSystem(Memory::MemorySystem& memory, Core::Timing& timing,
     : memory(memory), timing(timing),
       prepare_reschedule_callback(std::move(prepare_reschedule_callback)), memory_mode(memory_mode),
       n3ds_hw_caps(n3ds_hw_caps) {
+    slab_heap_container = std::make_unique<SlabHeapContainer>();
     std::generate(memory_regions.begin(), memory_regions.end(),
                   [] { return std::make_shared<MemoryRegionInfo>(); });
     MemoryInit(memory_mode, n3ds_hw_caps.memory_mode, override_init_time);
@@ -192,6 +195,27 @@ void KernelSystem::serialize(Archive& ar, const unsigned int file_version) {
     }
 }
 
+struct KernelSystem::SlabHeapContainer {
+    KSlabHeap<KLinkedListNode> linked_list_node;
+    KSlabHeap<KBlockInfo> block_info;
+    KSlabHeap<KMemoryBlock> memory_block;
+};
+
+template <typename T>
+KSlabHeap<T>& KernelSystem::SlabHeap() {
+    if constexpr (std::is_same_v<T, KLinkedListNode>) {
+        return slab_heap_container->linked_list_node;
+    } else if constexpr (std::is_same_v<T, KBlockInfo>) {
+        return slab_heap_container->block_info;
+    } else if constexpr (std::is_same_v<T, KMemoryBlock>) {
+        return slab_heap_container->memory_block;
+    }
+}
+
+template KSlabHeap<KLinkedListNode>& KernelSystem::SlabHeap();
+template KSlabHeap<KBlockInfo>& KernelSystem::SlabHeap();
+template KSlabHeap<KMemoryBlock>& KernelSystem::SlabHeap();
+
 SERIALIZE_IMPL(KernelSystem)
 
 } // namespace Kernel

src/core/hle/kernel/kernel.h

@@ -130,6 +130,9 @@ private:
     friend class boost::serialization::access;
 };
 
+template <typename T>
+class KSlabHeap;
+
 class KernelSystem {
 public:
     explicit KernelSystem(Memory::MemorySystem& memory, Core::Timing& timing,
@@ -260,6 +263,10 @@ public:
                                                               MemoryPermission other_permissions,
                                                               std::string name = "Unknown Applet");
 
+    /// Gets the slab heap for the specified kernel object type.
+    template <typename T>
+    KSlabHeap<T>& SlabHeap();
+
     u32 GenerateObjectID();
 
     /// Retrieves a process from the current list of processes.
@@ -369,6 +376,10 @@ private:
     MemoryMode memory_mode;
     New3dsHwCapabilities n3ds_hw_caps;
 
+    /// Helper to encapsulate all slab heaps in a single heap allocated container
+    struct SlabHeapContainer;
+    std::unique_ptr<SlabHeapContainer> slab_heap_container;
+
     friend class boost::serialization::access;
     template <class Archive>
     void serialize(Archive& ar, const unsigned int file_version);

src/core/hle/kernel/slab_helpers.h

@@ -0,0 +1,130 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/kernel/k_auto_object.h"
+#include "core/hle/kernel/kernel.h"
+
+namespace Kernel {
+
+template <class Derived>
+class KSlabAllocated {
+public:
+    constexpr KSlabAllocated() = default;
+
+    size_t GetSlabIndex(KernelSystem& kernel) const {
+        return kernel.SlabHeap<Derived>().GetIndex(static_cast<const Derived*>(this));
+    }
+
+public:
+    static void InitializeSlabHeap(KernelSystem& kernel, void* memory, size_t memory_size) {
+        kernel.SlabHeap<Derived>().Initialize(memory, memory_size);
+    }
+
+    static Derived* Allocate(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().Allocate(kernel);
+    }
+
+    static void Free(KernelSystem& kernel, Derived* obj) {
+        kernel.SlabHeap<Derived>().Free(obj);
+    }
+
+    static size_t GetObjectSize(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetObjectSize();
+    }
+
+    static size_t GetSlabHeapSize(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetSlabHeapSize();
+    }
+
+    static size_t GetPeakIndex(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetPeakIndex();
+    }
+
+    static uintptr_t GetSlabHeapAddress(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetSlabHeapAddress();
+    }
+
+    static size_t GetNumRemaining(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetNumRemaining();
+    }
+};
+
+template <typename Derived, typename Base>
+class KAutoObjectWithSlabHeap : public Base {
+    static_assert(std::is_base_of<KAutoObject, Base>::value);
+
+private:
+    static Derived* Allocate(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().Allocate(kernel);
+    }
+
+    static void Free(KernelSystem& kernel, Derived* obj) {
+        kernel.SlabHeap<Derived>().Free(obj);
+    }
+
+public:
+    explicit KAutoObjectWithSlabHeap(KernelSystem& kernel) : Base(kernel) {}
+    virtual ~KAutoObjectWithSlabHeap() = default;
+
+    virtual void Destroy() override {
+        const bool is_initialized = this->IsInitialized();
+        uintptr_t arg = 0;
+        if (is_initialized) {
+            arg = this->GetPostDestroyArgument();
+            this->Finalize();
+        }
+        Free(Base::m_kernel, static_cast<Derived*>(this));
+        if (is_initialized) {
+            Derived::PostDestroy(arg);
+        }
+    }
+
+    virtual bool IsInitialized() const {
+        return true;
+    }
+    virtual uintptr_t GetPostDestroyArgument() const {
+        return 0;
+    }
+
+    size_t GetSlabIndex() const {
+        return SlabHeap<Derived>(Base::m_kernel).GetObjectIndex(static_cast<const Derived*>(this));
+    }
+
+public:
+    static void InitializeSlabHeap(KernelSystem& kernel, void* memory, size_t memory_size) {
+        kernel.SlabHeap<Derived>().Initialize(memory, memory_size);
+    }
+
+    static Derived* Create(KernelSystem& kernel) {
+        Derived* obj = Allocate(kernel);
+        if (obj != nullptr) {
+            KAutoObject::Create(obj);
+        }
+        return obj;
+    }
+
+    static size_t GetObjectSize(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetObjectSize();
+    }
+
+    static size_t GetSlabHeapSize(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetSlabHeapSize();
+    }
+
+    static size_t GetPeakIndex(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetPeakIndex();
+    }
+
+    static uintptr_t GetSlabHeapAddress(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetSlabHeapAddress();
+    }
+
+    static size_t GetNumRemaining(KernelSystem& kernel) {
+        return kernel.SlabHeap<Derived>().GetNumRemaining();
+    }
+};
+
+} // namespace Kernel

src/core/hle/result.h

@@ -408,3 +408,130 @@ private:
     auto CONCAT2(check_result_L, __LINE__) = source;                                               \
     if (CONCAT2(check_result_L, __LINE__).IsError())                                               \
         return CONCAT2(check_result_L, __LINE__);
+
+#define R_SUCCEEDED(res) (static_cast<ResultCode>(res).IsSuccess())
+#define R_FAILED(res) (!static_cast<ResultCode>(res).IsSuccess())
+
+namespace ResultImpl {
+template <auto EvaluateResult, class F>
+class ScopedResultGuard {
+private:
+    ResultCode& m_ref;
+    F m_f;
+
+public:
+    constexpr ScopedResultGuard(ResultCode& ref, F f) : m_ref(ref), m_f(std::move(f)) {}
+    constexpr ~ScopedResultGuard() {
+        if (EvaluateResult(m_ref)) {
+            m_f();
+        }
+    }
+};
+
+template <auto EvaluateResult>
+class ResultReferenceForScopedResultGuard {
+private:
+    ResultCode& m_ref;
+
+public:
+    constexpr ResultReferenceForScopedResultGuard(ResultCode& r) : m_ref(r) {}
+    constexpr operator ResultCode&() const {
+        return m_ref;
+    }
+};
+
+template <auto EvaluateResult, typename F>
+constexpr ScopedResultGuard<EvaluateResult, F> operator+(
+    ResultReferenceForScopedResultGuard<EvaluateResult> ref, F&& f) {
+    return ScopedResultGuard<EvaluateResult, F>(static_cast<ResultCode&>(ref), std::forward<F>(f));
+}
+
+constexpr bool EvaluateResultSuccess(const ResultCode& r) {
+    return R_SUCCEEDED(r);
+}
+constexpr bool EvaluateResultFailure(const ResultCode& r) {
+    return R_FAILED(r);
+}
+
+template <typename T>
+constexpr void UpdateCurrentResultReference(T result_reference, ResultCode result) = delete;
+// Intentionally not defined
+
+template <>
+constexpr void UpdateCurrentResultReference<ResultCode&>(ResultCode& result_reference,
+                                                         ResultCode result) {
+    result_reference = result;
+}
+
+template <>
+constexpr void UpdateCurrentResultReference<const ResultCode>(ResultCode result_reference,
+                                                              ResultCode result) {}
+} // namespace ResultImpl
+
+#define DECLARE_CURRENT_RESULT_REFERENCE_AND_STORAGE(COUNTER_VALUE)                                \
+    [[maybe_unused]] constexpr bool CONCAT2(HasPrevRef_, COUNTER_VALUE) =                          \
+        std::same_as<decltype(__TmpCurrentResultReference), ResultCode&>;                          \
+    [[maybe_unused]] Result CONCAT2(PrevRef_, COUNTER_VALUE) = __TmpCurrentResultReference;        \
+    [[maybe_unused]] Result CONCAT2(__tmp_result_, COUNTER_VALUE) = ResultSuccess;                 \
+    Result& __TmpCurrentResultReference = CONCAT2(HasPrevRef_, COUNTER_VALUE)                      \
+                                              ? CONCAT2(PrevRef_, COUNTER_VALUE)                   \
+                                              : CONCAT2(__tmp_result_, COUNTER_VALUE)
+
+#define ON_RESULT_RETURN_IMPL(...)                                                                 \
+    static_assert(std::same_as<decltype(__TmpCurrentResultReference), ResultCode&>);               \
+    auto CONCAT2(RESULT_GUARD_STATE_, __COUNTER__) =                                               \
+        ResultImpl::ResultReferenceForScopedResultGuard<__VA_ARGS__>(                              \
+            __TmpCurrentResultReference) +                                                         \
+        [&]()
+
+#define ON_RESULT_FAILURE_2 ON_RESULT_RETURN_IMPL(ResultImpl::EvaluateResultFailure)
+
+#define ON_RESULT_FAILURE                                                                          \
+    DECLARE_CURRENT_RESULT_REFERENCE_AND_STORAGE(__COUNTER__);                                     \
+    ON_RESULT_FAILURE_2
+
+#define ON_RESULT_SUCCESS_2 ON_RESULT_RETURN_IMPL(ResultImpl::EvaluateResultSuccess)
+
+#define ON_RESULT_SUCCESS                                                                          \
+    DECLARE_CURRENT_RESULT_REFERENCE_AND_STORAGE(__COUNTER__);                                     \
+    ON_RESULT_SUCCESS_2
+
+constexpr inline ResultCode __TmpCurrentResultReference = RESULT_SUCCESS;
+
+/// Returns a result.
+#define R_RETURN(res_expr)                                                                         \
+    {                                                                                              \
+        const ResultCode _tmp_r_throw_rc = (res_expr);                                             \
+        ResultImpl::UpdateCurrentResultReference<decltype(__TmpCurrentResultReference)>(           \
+            __TmpCurrentResultReference, _tmp_r_throw_rc);                                         \
+        return _tmp_r_throw_rc;                                                                    \
+    }
+
+/// Returns ResultSuccess()
+#define R_SUCCEED() R_RETURN(RESULT_SUCCESS)
+
+/// Throws a result.
+#define R_THROW(res_expr) R_RETURN(res_expr)
+
+/// Evaluates a boolean expression, and returns a result unless that expression is true.
+#define R_UNLESS(expr, res)                                                                        \
+    {                                                                                              \
+        if (!(expr)) {                                                                             \
+            R_THROW(res);                                                                          \
+        }                                                                                          \
+    }
+
+/// Evaluates an expression that returns a result, and returns the result if it would fail.
+#define R_TRY(res_expr)                                                                            \
+    {                                                                                              \
+        const auto _tmp_r_try_rc = (res_expr);                                                     \
+        if (R_FAILED(_tmp_r_try_rc)) {                                                             \
+            R_THROW(_tmp_r_try_rc);                                                                \
+        }                                                                                          \
+    }
+
+/// Evaluates a boolean expression, and succeeds if that expression is true.
+#define R_SUCCEED_IF(expr) R_UNLESS(!(expr), RESULT_SUCCESS)
+
+/// Evaluates a boolean expression, and asserts if that expression is false.
+#define R_ASSERT(expr) ASSERT(R_SUCCEEDED(expr))

src/core/hle/service/fs/file.cpp

@@ -57,6 +57,7 @@ void File::Read(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp(ctx);
     u64 offset = rp.Pop<u64>();
     u32 length = rp.Pop<u32>();
+    auto& buffer = rp.PopMappedBuffer();
     LOG_TRACE(Service_FS, "Read {}: offset=0x{:x} length=0x{:08X}", GetName(), offset, length);
 
     const FileSessionSlot* file = GetSessionData(ctx.Session());
@@ -75,17 +76,15 @@ void File::Read(Kernel::HLERequestContext& ctx) {
                   offset, length, backend->GetSize());
     }
 
-    // Conventional reading if the backend does not support cache.
-    if (!backend->AllowsCachedReads()) {
-        auto& buffer = rp.PopMappedBuffer();
     IPC::RequestBuilder rb = rp.MakeBuilder(2, 2);
-        std::unique_ptr<u8*> data = std::make_unique<u8*>(static_cast<u8*>(operator new(length)));
+
-        const auto read = backend->Read(offset, length, *data);
+    std::vector<u8> data(length);
+    ResultVal<std::size_t> read = backend->Read(offset, data.size(), data.data());
     if (read.Failed()) {
         rb.Push(read.Code());
         rb.Push<u32>(0);
     } else {
-            buffer.Write(*data, 0, *read);
+        buffer.Write(data.data(), 0, *read);
         rb.Push(RESULT_SUCCESS);
         rb.Push<u32>(static_cast<u32>(*read));
     }
@@ -93,76 +92,6 @@ void File::Read(Kernel::HLERequestContext& ctx) {
 
     std::chrono::nanoseconds read_timeout_ns{backend->GetReadDelayNs(length)};
     ctx.SleepClientThread("file::read", read_timeout_ns, nullptr);
-        return;
-    }
-
-    struct AsyncData {
-        // Input
-        u32 length;
-        u64 offset;
-        std::chrono::steady_clock::time_point pre_timer;
-        bool cache_ready;
-
-        // Output
-        ResultCode ret{0};
-        Kernel::MappedBuffer* buffer;
-        std::unique_ptr<u8*> data;
-        size_t read_size;
-    };
-
-    auto async_data = std::make_shared<AsyncData>();
-    async_data->buffer = &rp.PopMappedBuffer();
-    async_data->length = length;
-    async_data->offset = offset;
-    async_data->cache_ready = backend->CacheReady(offset, length);
-    if (!async_data->cache_ready) {
-        async_data->pre_timer = std::chrono::steady_clock::now();
-    }
-
-    // LOG_DEBUG(Service_FS, "cache={}, offset={}, length={}", cache_ready, offset, length);
-    ctx.RunAsync(
-        [this, async_data](Kernel::HLERequestContext& ctx) {
-            async_data->data =
-                std::make_unique<u8*>(static_cast<u8*>(operator new(async_data->length)));
-            const auto read =
-                backend->Read(async_data->offset, async_data->length, *async_data->data);
-            if (read.Failed()) {
-                async_data->ret = read.Code();
-                async_data->read_size = 0;
-            } else {
-                async_data->ret = RESULT_SUCCESS;
-                async_data->read_size = *read;
-            }
-
-            const auto read_delay = static_cast<s64>(backend->GetReadDelayNs(async_data->length));
-            if (!async_data->cache_ready) {
-                const auto time_took = std::chrono::duration_cast<std::chrono::nanoseconds>(
-                                           std::chrono::steady_clock::now() - async_data->pre_timer)
-                                           .count();
-                /*
-                if (time_took > read_delay) {
-                    LOG_DEBUG(Service_FS, "Took longer! length={}, time_took={}, read_delay={}",
-                              async_data->length, time_took, read_delay);
-                }
-                */
-                return static_cast<s64>((read_delay > time_took) ? (read_delay - time_took) : 0);
-            } else {
-                return static_cast<s64>(read_delay);
-            }
-        },
-        [async_data](Kernel::HLERequestContext& ctx) {
-            IPC::RequestBuilder rb(ctx, 0x0802, 2, 2);
-            if (async_data->ret.IsError()) {
-                rb.Push(async_data->ret);
-                rb.Push<u32>(0);
-            } else {
-                async_data->buffer->Write(*async_data->data, 0, async_data->read_size);
-                rb.Push(RESULT_SUCCESS);
-                rb.Push<u32>(static_cast<u32>(async_data->read_size));
-            }
-            rb.PushMappedBuffer(*async_data->buffer);
-        },
-        !async_data->cache_ready);
 }
 
 void File::Write(Kernel::HLERequestContext& ctx) {

src/core/hle/service/fs/fs_user.cpp

@@ -670,26 +670,6 @@ void FS_USER::GetFormatInfo(Kernel::HLERequestContext& ctx) {
     rb.Push<bool>(format_info->duplicate_data != 0);
 }
 
-void FS_USER::GetProductInfo(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp(ctx);
-
-    u32 process_id = rp.Pop<u32>();
-
-    LOG_DEBUG(Service_FS, "called, process_id={}", process_id);
-
-    IPC::RequestBuilder rb = rp.MakeBuilder(6, 0);
-
-    const auto product_info = GetProductInfo(process_id);
-    if (!product_info.has_value()) {
-        rb.Push(ResultCode(FileSys::ErrCodes::ArchiveNotMounted, ErrorModule::FS,
-                           ErrorSummary::NotFound, ErrorLevel::Status));
-        return;
-    }
-
-    rb.Push(RESULT_SUCCESS);
-    rb.PushRaw<ProductInfo>(product_info.value());
-}
-
 void FS_USER::GetProgramLaunchInfo(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp(ctx);
     const auto process_id = rp.Pop<u32>();
@@ -707,20 +687,8 @@ void FS_USER::GetProgramLaunchInfo(Kernel::HLERequestContext& ctx) {
         return;
     }
 
-    ProgramInfo program_info = program_info_result.Unwrap();
-
-    // Always report the launched program mediatype is SD if the friends module is requesting this
-    // information and the media type is game card. Otherwise, friends will append a "romid" field
-    // to the NASC request with a cartridge unique identifier. Using a dump of a game card and the
-    // game card itself at the same time online is known to have caused issues in the past.
-    auto process = ctx.ClientThread()->owner_process.lock();
-    if (process && process->codeset->name == "friends" &&
-        program_info.media_type == MediaType::GameCard) {
-        program_info.media_type = MediaType::SDMC;
-    }
-
     rb.Push(RESULT_SUCCESS);
-    rb.PushRaw<ProgramInfo>(program_info);
+    rb.PushRaw(program_info_result.Unwrap());
 }
 
 void FS_USER::ObsoletedCreateExtSaveData(Kernel::HLERequestContext& ctx) {
@@ -807,12 +775,12 @@ void FS_USER::AddSeed(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
-void FS_USER::ObsoletedSetSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
+void FS_USER::SetSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp(ctx);
-    const u64 value = rp.Pop<u64>();
+    u64 value = rp.Pop<u64>();
-    const u32 secure_value_slot = rp.Pop<u32>();
+    u32 secure_value_slot = rp.Pop<u32>();
-    const u32 unique_id = rp.Pop<u32>();
+    u32 unique_id = rp.Pop<u32>();
-    const u8 title_variation = rp.Pop<u8>();
+    u8 title_variation = rp.Pop<u8>();
 
     // TODO: Generate and Save the Secure Value
 
@@ -826,11 +794,12 @@ void FS_USER::ObsoletedSetSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
-void FS_USER::ObsoletedGetSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
+void FS_USER::GetSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp(ctx);
-    const u32 secure_value_slot = rp.Pop<u32>();
+
-    const u32 unique_id = rp.Pop<u32>();
+    u32 secure_value_slot = rp.Pop<u32>();
-    const u8 title_variation = rp.Pop<u8>();
+    u32 unique_id = rp.Pop<u32>();
+    u8 title_variation = rp.Pop<u8>();
 
     LOG_WARNING(
         Service_FS,
@@ -847,77 +816,7 @@ void FS_USER::ObsoletedGetSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
     rb.Push<u64>(0);      // the secure value
 }
 
-void FS_USER::SetThisSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
+void FS_USER::Register(u32 process_id, u64 program_id, const std::string& filepath) {
-    IPC::RequestParser rp(ctx);
-    const u32 secure_value_slot = rp.Pop<u32>();
-    const u64 value = rp.Pop<u64>();
-
-    // TODO: Generate and Save the Secure Value
-
-    LOG_WARNING(Service_FS, "(STUBBED) called, value=0x{:016x} secure_value_slot=0x{:08X}", value,
-                secure_value_slot);
-
-    IPC::RequestBuilder rb = rp.MakeBuilder(1, 0);
-
-    rb.Push(RESULT_SUCCESS);
-}
-
-void FS_USER::GetThisSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp(ctx);
-    const u32 secure_value_slot = rp.Pop<u32>();
-
-    LOG_WARNING(Service_FS, "(STUBBED) called secure_value_slot=0x{:08X}", secure_value_slot);
-
-    IPC::RequestBuilder rb = rp.MakeBuilder(5, 0);
-
-    rb.Push(RESULT_SUCCESS);
-
-    // TODO: Implement Secure Value Lookup & Generation
-
-    rb.Push<bool>(false); // indicates that the secure value doesn't exist
-    rb.Push<bool>(false); // looks like a boolean value, purpose unknown
-    rb.Push<u64>(0);      // the secure value
-}
-
-void FS_USER::SetSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp(ctx);
-    const auto archive_handle = rp.PopRaw<ArchiveHandle>();
-    const u32 secure_value_slot = rp.Pop<u32>();
-    const u64 value = rp.Pop<u64>();
-    const bool flush = rp.Pop<bool>();
-
-    // TODO: Generate and Save the Secure Value
-
-    LOG_WARNING(Service_FS,
-                "(STUBBED) called, value=0x{:016x} secure_value_slot=0x{:04X} "
-                "archive_handle=0x{:08X} flush={}",
-                value, secure_value_slot, archive_handle, flush);
-
-    IPC::RequestBuilder rb = rp.MakeBuilder(1, 0);
-
-    rb.Push(RESULT_SUCCESS);
-}
-
-void FS_USER::GetSaveDataSecureValue(Kernel::HLERequestContext& ctx) {
-    IPC::RequestParser rp(ctx);
-    const auto archive_handle = rp.PopRaw<ArchiveHandle>();
-    const u32 secure_value_slot = rp.Pop<u32>();
-
-    LOG_WARNING(Service_FS, "(STUBBED) called secure_value_slot=0x{:08X} archive_handle=0x{:08X}",
-                secure_value_slot, archive_handle);
-
-    IPC::RequestBuilder rb = rp.MakeBuilder(5, 0);
-
-    rb.Push(RESULT_SUCCESS);
-
-    // TODO: Implement Secure Value Lookup & Generation
-
-    rb.Push<bool>(false); // indicates that the secure value doesn't exist
-    rb.Push<bool>(false); // looks like a boolean value, purpose unknown
-    rb.Push<u64>(0);      // the secure value
-}
-
-void FS_USER::RegisterProgramInfo(u32 process_id, u64 program_id, const std::string& filepath) {
     const MediaType media_type = GetMediaTypeFromPath(filepath);
     program_info_map.insert_or_assign(process_id, ProgramInfo{program_id, media_type});
     if (media_type == MediaType::GameCard) {
@@ -929,19 +828,6 @@ std::string FS_USER::GetCurrentGamecardPath() const {
     return current_gamecard_path;
 }
 
-void FS_USER::RegisterProductInfo(u32 process_id, const ProductInfo& product_info) {
-    product_info_map.insert_or_assign(process_id, product_info);
-}
-
-std::optional<FS_USER::ProductInfo> FS_USER::GetProductInfo(u32 process_id) {
-    auto it = product_info_map.find(process_id);
-    if (it != product_info_map.end()) {
-        return it->second;
-    } else {
-        return {};
-    }
-}
-
 ResultVal<u16> FS_USER::GetSpecialContentIndexFromGameCard(u64 title_id, SpecialContentType type) {
     // TODO(B3N30) check if on real 3DS NCSD is checked if partition exists
 
@@ -1043,7 +929,7 @@ FS_USER::FS_USER(Core::System& system)
         {0x082B, nullptr, "CardNorDirectRead_4xIO"},
         {0x082C, nullptr, "CardNorDirectCpuWriteWithoutVerify"},
         {0x082D, nullptr, "CardNorDirectSectorEraseWithoutVerify"},
-        {0x082E, &FS_USER::GetProductInfo, "GetProductInfo"},
+        {0x082E, nullptr, "GetProductInfo"},
         {0x082F, &FS_USER::GetProgramLaunchInfo, "GetProgramLaunchInfo"},
         {0x0830, &FS_USER::ObsoletedCreateExtSaveData, "Obsoleted_3_0_CreateExtSaveData"},
         {0x0831, nullptr, "CreateSharedExtSaveData"},
@@ -1098,16 +984,12 @@ FS_USER::FS_USER(Core::System& system)
         {0x0862, &FS_USER::SetPriority, "SetPriority"},
         {0x0863, &FS_USER::GetPriority, "GetPriority"},
         {0x0864, nullptr, "GetNandInfo"},
-        {0x0865, &FS_USER::ObsoletedSetSaveDataSecureValue, "SetSaveDataSecureValue"},
+        {0x0865, &FS_USER::SetSaveDataSecureValue, "SetSaveDataSecureValue"},
-        {0x0866, &FS_USER::ObsoletedGetSaveDataSecureValue, "GetSaveDataSecureValue"},
+        {0x0866, &FS_USER::GetSaveDataSecureValue, "GetSaveDataSecureValue"},
         {0x0867, nullptr, "ControlSecureSave"},
         {0x0868, nullptr, "GetMediaType"},
         {0x0869, nullptr, "GetNandEraseCount"},
         {0x086A, nullptr, "ReadNandReport"},
-        {0x086E, &FS_USER::SetThisSaveDataSecureValue, "SetThisSaveDataSecureValue" },
-        {0x086F, &FS_USER::GetThisSaveDataSecureValue, "GetThisSaveDataSecureValue" },
-        {0x0875, &FS_USER::SetSaveDataSecureValue, "SetSaveDataSecureValue" },
-        {0x0876, &FS_USER::GetSaveDataSecureValue, "GetSaveDataSecureValue" },
         {0x087A, &FS_USER::AddSeed, "AddSeed"},
         {0x087D, &FS_USER::GetNumSeeds, "GetNumSeeds"},
         {0x0886, nullptr, "CheckUpdatedDat"},

src/core/hle/service/fs/fs_user.h

@@ -4,12 +4,10 @@
 
 #pragma once
 
-#include <optional>
 #include <unordered_map>
 #include <boost/serialization/base_object.hpp>
 #include "common/common_types.h"
 #include "core/file_sys/errors.h"
-#include "core/hle/service/fs/archive.h"
 #include "core/hle/service/service.h"
 
 namespace Core {
@@ -49,23 +47,12 @@ class FS_USER final : public ServiceFramework<FS_USER, ClientSlot> {
 public:
     explicit FS_USER(Core::System& system);
 
-    // On real HW this is part of FSReg (FSReg:Register). But since that module is only used by
+    // On real HW this is part of FS:Reg. But since that module is only used by loader and pm, which
-    // loader and pm, which we HLEed, we can just directly use it here
+    // we HLEed, we can just directly use it here
-    void RegisterProgramInfo(u32 process_id, u64 program_id, const std::string& filepath);
+    void Register(u32 process_id, u64 program_id, const std::string& filepath);
 
     std::string GetCurrentGamecardPath() const;
 
-    struct ProductInfo {
-        std::array<u8, 0x10> product_code;
-        u16_le maker_code;
-        u16_le remaster_version;
-    };
-    static_assert(sizeof(ProductInfo) == 0x14);
-
-    void RegisterProductInfo(u32 process_id, const ProductInfo& product_info);
-
-    std::optional<ProductInfo> GetProductInfo(u32 process_id);
-
     /// Gets the registered program info of a process.
     ResultVal<ProgramInfo> GetProgramLaunchInfo(u32 process_id) const {
         auto info = program_info_map.find(process_id);
@@ -522,17 +509,6 @@ private:
      */
     void GetFormatInfo(Kernel::HLERequestContext& ctx);
 
-    /**
-     * FS_User::GetProductInfo service function.
-     *  Inputs:
-     *      0 : 0x082E0040
-     *      1 : Process ID
-     *  Outputs:
-     *      1 : Result of function, 0 on success, otherwise error code
-     *      2-6 : Product info
-     */
-    void GetProductInfo(Kernel::HLERequestContext& ctx);
-
     /**
      * FS_User::GetProgramLaunchInfo service function.
      *  Inputs:
@@ -624,7 +600,7 @@ private:
      *      0 : 0x08650140
      *      1 : Result of function, 0 on success, otherwise error code
      */
-    void ObsoletedSetSaveDataSecureValue(Kernel::HLERequestContext& ctx);
+    void SetSaveDataSecureValue(Kernel::HLERequestContext& ctx);
 
     /**
      * FS_User::GetSaveDataSecureValue service function.
@@ -639,57 +615,6 @@ private:
      *      2 : If Secure Value doesn't exist, 0, if it exists, 1
      *      3-4 : Secure Value
      */
-    void ObsoletedGetSaveDataSecureValue(Kernel::HLERequestContext& ctx);
-
-    /**
-     * FS_User::SetThisSaveDataSecureValue service function.
-     *  Inputs:
-     *      1 : Secure Value Slot
-     *      2-3 : Secure Value
-     *  Outputs:
-     *      1 : Result of function, 0 on success, otherwise error code
-     */
-    void SetThisSaveDataSecureValue(Kernel::HLERequestContext& ctx);
-
-    /**
-     * FS_User::GetSaveDataSecureValue service function.
-     *  Inputs:
-     *      1 : Secure Value Slot
-     *  Outputs:
-     *      1 : Result of function, 0 on success, otherwise error code
-     *      2 : If Secure Value doesn't exist, 0, if it exists, 1
-     *      3 : Unknown
-     *      4-5 : Secure Value
-     */
-    void GetThisSaveDataSecureValue(Kernel::HLERequestContext& ctx);
-
-    /**
-     * FS_User::SetSaveDataSecureValue service function.
-     *  Inputs:
-     *      0 : 0x08750180
-     *      1-2 : Archive
-     *      3 : Secure Value Slot
-     *      4 : value
-     *      5 : flush
-     *  Outputs:
-     *      0 : header
-     *      1 : Result of function, 0 on success, otherwise error code
-     */
-    void SetSaveDataSecureValue(Kernel::HLERequestContext& ctx);
-
-    /**
-     * FS_User::GetSaveDataSecureValue service function.
-     *  Inputs:
-     *      0 : 0x087600C0
-     *      1-2 : Archive
-     *      2 : Secure Value slot
-     *  Outputs:
-     *      0 : Header
-     *      1 : Result of function, 0 on success, otherwise error code
-     *      2 : If Secure Value doesn't exist, 0, if it exists, 1
-     *      3 : unknown
-     *      4-5 : Secure Value
-     */
     void GetSaveDataSecureValue(Kernel::HLERequestContext& ctx);
 
     static ResultVal<u16> GetSpecialContentIndexFromGameCard(u64 title_id, SpecialContentType type);
@@ -699,8 +624,6 @@ private:
     std::unordered_map<u32, ProgramInfo> program_info_map;
     std::string current_gamecard_path;
 
-    std::unordered_map<u32, ProductInfo> product_info_map;
-
     u32 priority = -1; ///< For SetPriority and GetPriority service functions
 
     Core::System& system;

src/core/loader/3dsx.cpp

@@ -282,7 +282,7 @@ ResultStatus AppLoader_THREEDSX::Load(std::shared_ptr<Kernel::Process>& process)
     // On real HW this is done with FS:Reg, but we can be lazy
     auto fs_user =
         Core::System::GetInstance().ServiceManager().GetService<Service::FS::FS_USER>("fs:USER");
-    fs_user->RegisterProgramInfo(process->GetObjectId(), process->codeset->program_id, filepath);
+    fs_user->Register(process->GetObjectId(), process->codeset->program_id, filepath);
 
     process->Run(48, Kernel::DEFAULT_STACK_SIZE);
 

src/core/loader/ncch.cpp

@@ -174,16 +174,7 @@ ResultStatus AppLoader_NCCH::LoadExec(std::shared_ptr<Kernel::Process>& process)
         auto fs_user =
             Core::System::GetInstance().ServiceManager().GetService<Service::FS::FS_USER>(
                 "fs:USER");
-        fs_user->RegisterProgramInfo(process->process_id, process->codeset->program_id, filepath);
+        fs_user->Register(process->process_id, process->codeset->program_id, filepath);
-
-        Service::FS::FS_USER::ProductInfo product_info{};
-        std::memcpy(product_info.product_code.data(), overlay_ncch->ncch_header.product_code,
-                    product_info.product_code.size());
-        std::memcpy(&product_info.remaster_version,
-                    overlay_ncch->exheader_header.codeset_info.flags.remaster_version,
-                    sizeof(product_info.remaster_version));
-        product_info.maker_code = overlay_ncch->ncch_header.maker_code;
-        fs_user->RegisterProductInfo(process->process_id, product_info);
 
         process->Run(priority, stack_size);
         return ResultStatus::Success;

src/core/memory.h

@@ -329,6 +329,11 @@ public:
      */
     u8* GetPointer(VAddr vaddr);
 
+    template <typename T>
+    T* GetPointer(VAddr vaddr) {
+        return reinterpret_cast<T*>(GetPointer(vaddr));
+    }
+
     /**
      * Gets a pointer to the given address.
      *

src/video_core/command_processor.cpp

@@ -78,11 +78,12 @@ static void WriteUniformFloatReg(ShaderRegs& config, Shader::ShaderSetup& setup,
         (float_regs_counter >= 3 && !uniform_setup.IsFloat32())) {
         float_regs_counter = 0;
 
-        if (uniform_setup.index >= setup.uniforms.f.size()) {
+        auto& uniform = setup.uniforms.f[uniform_setup.index];
+
+        if (uniform_setup.index >= 96) {
             LOG_ERROR(HW_GPU, "Invalid {} float uniform index {}", GetShaderSetupTypeName(setup),
                       (int)uniform_setup.index);
         } else {
-            auto& uniform = setup.uniforms.f[uniform_setup.index];
 
             // NOTE: The destination component order indeed is "backwards"
             if (uniform_setup.IsFloat32()) {

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -78,8 +78,8 @@ RasterizerOpenGL::RasterizerOpenGL(Memory::MemorySystem& memory,
                                    VideoCore::CustomTexManager& custom_tex_manager,
                                    VideoCore::RendererBase& renderer, Driver& driver_)
     : VideoCore::RasterizerAccelerated{memory}, driver{driver_},
-      shader_manager{renderer.GetRenderWindow(), driver}, runtime{driver, renderer},
+      shader_manager{renderer.GetRenderWindow(), driver, !driver.IsOpenGLES()},
-      res_cache{memory, custom_tex_manager, runtime, regs, renderer},
+      runtime{driver, renderer}, res_cache{memory, custom_tex_manager, runtime, regs, renderer},
       texture_buffer_size{TextureBufferSize()}, vertex_buffer{driver, GL_ARRAY_BUFFER,
                                                               VERTEX_BUFFER_SIZE},
       uniform_buffer{driver, GL_UNIFORM_BUFFER, UNIFORM_BUFFER_SIZE},

src/video_core/renderer_opengl/gl_resource_manager.cpp

@@ -2,10 +2,14 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/microprofile.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
 #include "video_core/renderer_opengl/gl_state.h"
 
+MICROPROFILE_DEFINE(OpenGL_ResourceCreation, "OpenGL", "Resource Creation", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_RGB(128, 128, 192));
+
 namespace OpenGL {
 
 void OGLRenderbuffer::Create() {
@@ -13,6 +17,7 @@ void OGLRenderbuffer::Create() {
         return;
     }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenRenderbuffers(1, &handle);
 }
 
@@ -21,6 +26,7 @@ void OGLRenderbuffer::Release() {
         return;
     }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteRenderbuffers(1, &handle);
     OpenGLState::GetCurState().ResetRenderbuffer(handle).Apply();
     handle = 0;
@@ -31,6 +37,7 @@ void OGLTexture::Create() {
         return;
     }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenTextures(1, &handle);
 }
 
@@ -39,6 +46,7 @@ void OGLTexture::Release() {
         return;
     }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteTextures(1, &handle);
     OpenGLState::GetCurState().ResetTexture(handle).Apply();
     handle = 0;
@@ -80,6 +88,7 @@ void OGLSampler::Create() {
         return;
     }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenSamplers(1, &handle);
 }
 
@@ -88,41 +97,37 @@ void OGLSampler::Release() {
         return;
     }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteSamplers(1, &handle);
     OpenGLState::GetCurState().ResetSampler(handle).Apply();
     handle = 0;
 }
 
 void OGLShader::Create(std::string_view source, GLenum type) {
-    if (handle != 0) {
+    if (handle != 0)
         return;
-    }
+    if (source.empty())
-    if (source.empty()) {
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     handle = LoadShader(source, type);
 }
 
 void OGLShader::Release() {
-    if (handle == 0) {
+    if (handle == 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteShader(handle);
     handle = 0;
 }
 
-void OGLProgram::Create(std::string_view source, GLenum type) {
+void OGLProgram::Create(bool separable_program, std::span<const GLuint> shaders) {
-    if (handle != 0) {
+    if (handle != 0)
         return;
-    }
-    if (source.empty()) {
-        return;
-    }
 
-    const std::array sources{GetPreamble().data(), source.data()};
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
-    handle = glCreateShaderProgramv(type, 2, sources.data());
+    handle = LoadProgram(separable_program, shaders);
 }
 
 void OGLProgram::Create(std::string_view vert_shader, std::string_view frag_shader) {
@@ -130,87 +135,88 @@ void OGLProgram::Create(std::string_view vert_shader, std::string_view frag_shad
     vert.Create(vert_shader, GL_VERTEX_SHADER);
     frag.Create(frag_shader, GL_FRAGMENT_SHADER);
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     const std::array shaders{vert.handle, frag.handle};
-    handle = LoadProgram(shaders);
+    Create(false, shaders);
 }
 
 void OGLProgram::Release() {
-    if (handle == 0) {
+    if (handle == 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteProgram(handle);
     OpenGLState::GetCurState().ResetProgram(handle).Apply();
     handle = 0;
 }
 
 void OGLPipeline::Create() {
-    if (handle != 0) {
+    if (handle != 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenProgramPipelines(1, &handle);
 }
 
 void OGLPipeline::Release() {
-    if (handle == 0) {
+    if (handle == 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteProgramPipelines(1, &handle);
     OpenGLState::GetCurState().ResetPipeline(handle).Apply();
     handle = 0;
 }
 
 void OGLBuffer::Create() {
-    if (handle != 0) {
+    if (handle != 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenBuffers(1, &handle);
 }
 
 void OGLBuffer::Release() {
-    if (handle == 0) {
+    if (handle == 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteBuffers(1, &handle);
     OpenGLState::GetCurState().ResetBuffer(handle).Apply();
     handle = 0;
 }
 
 void OGLVertexArray::Create() {
-    if (handle != 0) {
+    if (handle != 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenVertexArrays(1, &handle);
 }
 
 void OGLVertexArray::Release() {
-    if (handle == 0) {
+    if (handle == 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteVertexArrays(1, &handle);
     OpenGLState::GetCurState().ResetVertexArray(handle).Apply();
     handle = 0;
 }
 
 void OGLFramebuffer::Create() {
-    if (handle != 0) {
+    if (handle != 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenFramebuffers(1, &handle);
 }
 
 void OGLFramebuffer::Release() {
-    if (handle == 0) {
+    if (handle == 0)
         return;
-    }
 
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteFramebuffers(1, &handle);
     OpenGLState::GetCurState().ResetFramebuffer(handle).Apply();
     handle = 0;

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -130,7 +130,7 @@ public:
     }
 
     /// Creates a new program from given shader objects
-    void Create(std::string_view source, GLenum type);
+    void Create(bool separable_program, std::span<const GLuint> shaders);
 
     /// Creates a new program from given shader soruce code
     void Create(std::string_view vert_shader, std::string_view frag_shader);

src/video_core/renderer_opengl/gl_shader_disk_cache.cpp

@@ -103,8 +103,10 @@ bool ShaderDiskCacheRaw::Save(FileUtil::IOFile& file) const {
     return true;
 }
 
-ShaderDiskCache::ShaderDiskCache()
+ShaderDiskCache::ShaderDiskCache(bool separable)
-    : transferable_file(AppendTransferableFile()), precompiled_file(AppendPrecompiledFile()) {}
+    : separable{separable}, transferable_file(AppendTransferableFile()),
+      // seperable shaders use the virtual precompile file, that already has a header.
+      precompiled_file(AppendPrecompiledFile(!separable)) {}
 
 std::optional<std::vector<ShaderDiskCacheRaw>> ShaderDiskCache::LoadTransferable() {
     const bool has_title_id = GetProgramID() != 0;
@@ -175,7 +177,7 @@ std::optional<std::vector<ShaderDiskCacheRaw>> ShaderDiskCache::LoadTransferable
 }
 
 std::pair<std::unordered_map<u64, ShaderDiskCacheDecompiled>, ShaderDumpsMap>
-ShaderDiskCache::LoadPrecompiled() {
+ShaderDiskCache::LoadPrecompiled(bool compressed) {
     if (!IsUsable())
         return {};
 
@@ -185,7 +187,7 @@ ShaderDiskCache::LoadPrecompiled() {
         return {};
     }
 
-    const auto result = LoadPrecompiledFile(precompiled_file);
+    const auto result = LoadPrecompiledFile(precompiled_file, compressed);
     if (!result) {
         LOG_INFO(Render_OpenGL,
                  "Failed to load precompiled cache for game with title id={} - removing",
@@ -197,16 +199,22 @@ ShaderDiskCache::LoadPrecompiled() {
 }
 
 std::optional<std::pair<std::unordered_map<u64, ShaderDiskCacheDecompiled>, ShaderDumpsMap>>
-ShaderDiskCache::LoadPrecompiledFile(FileUtil::IOFile& file) {
+ShaderDiskCache::LoadPrecompiledFile(FileUtil::IOFile& file, bool compressed) {
     // Read compressed file from disk and decompress to virtual precompiled cache file
     std::vector<u8> precompiled_file(file.GetSize());
     file.ReadBytes(precompiled_file.data(), precompiled_file.size());
-    const std::vector<u8> decompressed = Common::Compression::DecompressDataZSTD(precompiled_file);
+    if (compressed) {
+        const std::vector<u8> decompressed =
+            Common::Compression::DecompressDataZSTD(precompiled_file);
         if (decompressed.empty()) {
             LOG_ERROR(Render_OpenGL, "Could not decompress precompiled shader cache.");
             return std::nullopt;
         }
         SaveArrayToPrecompiled(decompressed.data(), decompressed.size());
+    } else {
+        SaveArrayToPrecompiled(precompiled_file.data(), precompiled_file.size());
+    }
+
     decompressed_precompiled_cache_offset = 0;
 
     ShaderCacheVersionHash file_hash{};
@@ -345,7 +353,7 @@ void ShaderDiskCache::InvalidatePrecompiled() {
     if (!FileUtil::Delete(GetPrecompiledPath())) {
         LOG_ERROR(Render_OpenGL, "Failed to invalidate precompiled file={}", GetPrecompiledPath());
     }
-    precompiled_file = AppendPrecompiledFile();
+    precompiled_file = AppendPrecompiledFile(!separable);
 }
 
 void ShaderDiskCache::SaveRaw(const ShaderDiskCacheRaw& entry) {
@@ -463,11 +471,12 @@ FileUtil::IOFile ShaderDiskCache::AppendTransferableFile() {
     return file;
 }
 
-FileUtil::IOFile ShaderDiskCache::AppendPrecompiledFile() {
+FileUtil::IOFile ShaderDiskCache::AppendPrecompiledFile(bool write_header) {
     if (!EnsureDirectories())
         return {};
 
     const auto precompiled_path{GetPrecompiledPath()};
+    const bool existed = FileUtil::Exists(precompiled_path);
 
     FileUtil::IOFile file(precompiled_path, "ab+");
     if (!file.IsOpen()) {
@@ -475,6 +484,15 @@ FileUtil::IOFile ShaderDiskCache::AppendPrecompiledFile() {
         return {};
     }
 
+    // If the file didn't exist, write its version
+    if (write_header && (!existed || file.GetSize() == 0)) {
+        const auto hash{GetShaderCacheVersionHash()};
+        if (file.WriteArray(hash.data(), hash.size()) != hash.size()) {
+            LOG_ERROR(Render_OpenGL, "Failed to write precompiled cache version in path={}",
+                      precompiled_path);
+            return {};
+        }
+    }
     return file;
 }
 
@@ -498,7 +516,7 @@ void ShaderDiskCache::SaveVirtualPrecompiledFile() {
     if (!FileUtil::Delete(GetPrecompiledPath())) {
         LOG_ERROR(Render_OpenGL, "Failed to invalidate precompiled file={}", GetPrecompiledPath());
     }
-    precompiled_file = AppendPrecompiledFile();
+    precompiled_file = AppendPrecompiledFile(!separable);
 
     if (precompiled_file.WriteBytes(compressed.data(), compressed.size()) != compressed.size()) {
         LOG_ERROR(Render_OpenGL, "Failed to write precompiled cache version in path={}",
@@ -540,7 +558,10 @@ std::string ShaderDiskCache::GetPrecompiledDir() const {
 }
 
 std::string ShaderDiskCache::GetPrecompiledShaderDir() const {
+    if (separable) {
         return GetPrecompiledDir() + DIR_SEP "separable";
+    }
+    return GetPrecompiledDir() + DIR_SEP "conventional";
 }
 
 std::string ShaderDiskCache::GetBaseDir() const {

src/video_core/renderer_opengl/gl_shader_disk_cache.h

@@ -16,8 +16,11 @@
 
 #include <glad/glad.h>
 
+#include "common/assert.h"
+#include "common/common_types.h"
 #include "common/file_util.h"
-#include "video_core/shader/generator/shader_gen.h"
+#include "video_core/regs.h"
+#include "video_core/shader/generator/glsl_shader_gen.h"
 
 namespace Core {
 class System;
@@ -87,14 +90,14 @@ struct ShaderDiskCacheDump {
 
 class ShaderDiskCache {
 public:
-    explicit ShaderDiskCache();
+    explicit ShaderDiskCache(bool separable);
     ~ShaderDiskCache() = default;
 
     /// Loads transferable cache. If file has a old version or on failure, it deletes the file.
     std::optional<std::vector<ShaderDiskCacheRaw>> LoadTransferable();
 
     /// Loads current game's precompiled cache. Invalidates on failure.
-    std::pair<ShaderDecompiledMap, ShaderDumpsMap> LoadPrecompiled();
+    std::pair<ShaderDecompiledMap, ShaderDumpsMap> LoadPrecompiled(bool compressed);
 
     /// Removes the transferable (and precompiled) cache file.
     void InvalidateAll();
@@ -120,7 +123,7 @@ public:
 private:
     /// Loads the transferable cache. Returns empty on failure.
     std::optional<std::pair<ShaderDecompiledMap, ShaderDumpsMap>> LoadPrecompiledFile(
-        FileUtil::IOFile& file);
+        FileUtil::IOFile& file, bool compressed);
 
     /// Loads a decompiled cache entry from m_precompiled_cache_virtual_file. Returns empty on
     /// failure.
@@ -140,7 +143,7 @@ private:
     FileUtil::IOFile AppendTransferableFile();
 
     /// Opens current game's precompiled file and write it's header if it doesn't exist
-    FileUtil::IOFile AppendPrecompiledFile();
+    FileUtil::IOFile AppendPrecompiledFile(bool write_header);
 
     /// Save precompiled header to precompiled_cache_in_memory
     void SavePrecompiledHeaderToVirtualPrecompiledCache();
@@ -216,6 +219,8 @@ private:
     // The cache has been loaded at boot
     bool tried_to_load{};
 
+    bool separable{};
+
     u64 program_id{};
     std::string title_id;
 

src/video_core/renderer_opengl/gl_shader_manager.cpp

@@ -14,7 +14,7 @@
 #include "video_core/renderer_opengl/gl_shader_disk_cache.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
-#include "video_core/shader/generator/glsl_shader_gen.h"
+#include "video_core/shader/generator/shader_uniforms.h"
 #include "video_core/video_core.h"
 
 using namespace Pica::Shader::Generator;
@@ -35,16 +35,19 @@ static u64 GetUniqueIdentifier(const Pica::Regs& regs, const ProgramCode& code)
 }
 
 static OGLProgram GeneratePrecompiledProgram(const ShaderDiskCacheDump& dump,
-                                             const std::set<GLenum>& supported_formats) {
+                                             const std::set<GLenum>& supported_formats,
+                                             bool separable) {
 
     if (supported_formats.find(dump.binary_format) == supported_formats.end()) {
         LOG_INFO(Render_OpenGL, "Precompiled cache entry with unsupported format - removing");
         return {};
     }
 
-    OGLProgram shader{};
+    auto shader = OGLProgram();
     shader.handle = glCreateProgram();
+    if (separable) {
         glProgramParameteri(shader.handle, GL_PROGRAM_SEPARABLE, GL_TRUE);
+    }
     glProgramBinary(shader.handle, dump.binary_format, dump.binary.data(),
                     static_cast<GLsizei>(dump.binary.size()));
 
@@ -87,42 +90,91 @@ static std::tuple<PicaVSConfig, Pica::Shader::ShaderSetup> BuildVSConfigFromRaw(
             setup};
 }
 
-class TrivialVertexShader {
+/**
+ * An object representing a shader program staging. It can be either a shader object or a program
+ * object, depending on whether separable program is used.
+ */
+class OGLShaderStage {
 public:
-    explicit TrivialVertexShader(const Driver& driver) {
+    explicit OGLShaderStage(bool separable) {
-        const auto code = GLSL::GenerateTrivialVertexShader(driver.HasClipCullDistance(), true);
+        if (separable) {
-        program.Create(code, GL_VERTEX_SHADER);
+            shader_or_program = OGLProgram();
+        } else {
+            shader_or_program = OGLShader();
         }
-    GLuint Get() const {
+    }
-        return program.handle;
+
+    void Create(const char* source, GLenum type) {
+        if (shader_or_program.index() == 0) {
+            std::get<OGLShader>(shader_or_program).Create(source, type);
+        } else {
+            OGLShader shader;
+            shader.Create(source, type);
+            OGLProgram& program = std::get<OGLProgram>(shader_or_program);
+            program.Create(true, std::array{shader.handle});
+        }
+    }
+
+    GLuint GetHandle() const {
+        if (shader_or_program.index() == 0) {
+            return std::get<OGLShader>(shader_or_program).handle;
+        } else {
+            return std::get<OGLProgram>(shader_or_program).handle;
+        }
+    }
+
+    void Inject(OGLProgram&& program) {
+        shader_or_program = std::move(program);
     }
 
 private:
-    OGLProgram program;
+    std::variant<OGLShader, OGLProgram> shader_or_program;
 };
 
-template <typename KeyConfigType, auto CodeGenerator, GLenum ShaderType>
+class TrivialVertexShader {
+public:
+    explicit TrivialVertexShader(const Driver& driver, bool separable) : program(separable) {
+        const auto code =
+            GLSL::GenerateTrivialVertexShader(driver.HasClipCullDistance(), separable);
+        program.Create(code.c_str(), GL_VERTEX_SHADER);
+    }
+    GLuint Get() const {
+        return program.GetHandle();
+    }
+
+private:
+    OGLShaderStage program;
+};
+
+template <typename KeyConfigType, std::string (*CodeGenerator)(const KeyConfigType&, bool),
+          GLenum ShaderType>
 class ShaderCache {
 public:
-    explicit ShaderCache() = default;
+    explicit ShaderCache(bool separable) : separable(separable) {}
-
     std::tuple<GLuint, std::optional<std::string>> Get(const KeyConfigType& config) {
-        auto [iter, new_shader] = shaders.try_emplace(config);
+        auto [iter, new_shader] = shaders.emplace(config, OGLShaderStage{separable});
-        OGLProgram& cached_shader = iter->second;
+        OGLShaderStage& cached_shader = iter->second;
         std::optional<std::string> result{};
         if (new_shader) {
-            result = CodeGenerator(config, true);
+            result = CodeGenerator(config, separable);
-            cached_shader.Create(result.value(), ShaderType);
+            cached_shader.Create(result->c_str(), ShaderType);
         }
-        return {cached_shader.handle, std::move(result)};
+        return {cached_shader.GetHandle(), std::move(result)};
     }
 
-    void Inject(const KeyConfigType& key, OGLProgram&& stage) {
+    void Inject(const KeyConfigType& key, OGLProgram&& program) {
+        OGLShaderStage stage{separable};
+        stage.Inject(std::move(program));
+        shaders.emplace(key, std::move(stage));
+    }
+
+    void Inject(const KeyConfigType& key, OGLShaderStage&& stage) {
         shaders.emplace(key, std::move(stage));
     }
 
 private:
-    std::unordered_map<KeyConfigType, OGLProgram> shaders;
+    bool separable;
+    std::unordered_map<KeyConfigType, OGLShaderStage> shaders;
 };
 
 // This is a cache designed for shaders translated from PICA shaders. The first cache matches the
@@ -130,48 +182,59 @@ private:
 // GLSL code. The configuration is like this because there might be leftover code in the PICA shader
 // program buffer from the previous shader, which is hashed into the config, resulting several
 // different config values from the same shader program.
-template <typename KeyConfigType, auto CodeGenerator, GLenum ShaderType>
+template <typename KeyConfigType,
+          std::string (*CodeGenerator)(const Pica::Shader::ShaderSetup&, const KeyConfigType&,
+                                       bool),
+          GLenum ShaderType>
 class ShaderDoubleCache {
 public:
-    explicit ShaderDoubleCache() = default;
+    explicit ShaderDoubleCache(bool separable) : separable(separable) {}
-
     std::tuple<GLuint, std::optional<std::string>> Get(const KeyConfigType& key,
                                                        const Pica::Shader::ShaderSetup& setup) {
         std::optional<std::string> result{};
         auto map_it = shader_map.find(key);
         if (map_it == shader_map.end()) {
-            const auto program = CodeGenerator(setup, key, separable);
+            auto program = CodeGenerator(setup, key, separable);
             if (program.empty()) {
                 shader_map[key] = nullptr;
                 return {0, std::nullopt};
             }
 
-            const auto [iter, new_shader] = shader_cache.try_emplace(program);
+            auto [iter, new_shader] = shader_cache.emplace(program, OGLShaderStage{separable});
-            OGLProgram& cached_shader = iter->second;
+            OGLShaderStage& cached_shader = iter->second;
             if (new_shader) {
                 result = program;
-                cached_shader.Create(program, ShaderType);
+                cached_shader.Create(program.c_str(), ShaderType);
             }
             shader_map[key] = &cached_shader;
-            return {cached_shader.handle, std::move(result)};
+            return {cached_shader.GetHandle(), std::move(result)};
         }
 
         if (map_it->second == nullptr) {
             return {0, std::nullopt};
         }
 
-        return {map_it->second->handle, std::nullopt};
+        return {map_it->second->GetHandle(), std::nullopt};
     }
 
     void Inject(const KeyConfigType& key, std::string decomp, OGLProgram&& program) {
-        const auto iter = shader_cache.emplace(std::move(decomp), std::move(program)).first;
+        OGLShaderStage stage{separable};
-        shader_map.insert_or_assign(key, &iter->second);
+        stage.Inject(std::move(program));
+        const auto iter = shader_cache.emplace(std::move(decomp), std::move(stage)).first;
+        OGLShaderStage& cached_shader = iter->second;
+        shader_map.insert_or_assign(key, &cached_shader);
+    }
+
+    void Inject(const KeyConfigType& key, std::string decomp, OGLShaderStage&& stage) {
+        const auto iter = shader_cache.emplace(std::move(decomp), std::move(stage)).first;
+        OGLShaderStage& cached_shader = iter->second;
+        shader_map.insert_or_assign(key, &cached_shader);
     }
 
 private:
     bool separable;
-    std::unordered_map<KeyConfigType, OGLProgram*> shader_map;
+    std::unordered_map<KeyConfigType, OGLShaderStage*> shader_map;
-    std::unordered_map<std::string, OGLProgram> shader_cache;
+    std::unordered_map<std::string, OGLShaderStage> shader_cache;
 };
 
 using ProgrammableVertexShaders =
@@ -185,7 +248,11 @@ using FragmentShaders =
 
 class ShaderProgramManager::Impl {
 public:
-    explicit Impl(const Driver& driver) : trivial_vertex_shader(driver) {
+    explicit Impl(const Driver& driver, bool separable)
+        : separable(separable), programmable_vertex_shaders(separable),
+          trivial_vertex_shader(driver, separable), fixed_geometry_shaders(separable),
+          fragment_shaders(separable), disk_cache(separable) {
+        if (separable)
             pipeline.Create();
     }
 
@@ -215,6 +282,8 @@ public:
     static_assert(offsetof(ShaderTuple, fs_hash) == sizeof(std::size_t) * 2,
                   "ShaderTuple layout changed!");
 
+    bool separable;
+
     ShaderTuple current;
 
     ProgrammableVertexShaders programmable_vertex_shaders;
@@ -228,10 +297,11 @@ public:
     ShaderDiskCache disk_cache;
 };
 
-ShaderProgramManager::ShaderProgramManager(Frontend::EmuWindow& emu_window_, const Driver& driver_)
+ShaderProgramManager::ShaderProgramManager(Frontend::EmuWindow& emu_window_, const Driver& driver_,
+                                           bool separable)
     : emu_window{emu_window_}, driver{driver_},
       strict_context_required{emu_window.StrictContextRequired()}, impl{std::make_unique<Impl>(
-                                                                       driver_)} {}
+                                                                       driver_, separable)} {}
 
 ShaderProgramManager::~ShaderProgramManager() = default;
 
@@ -293,18 +363,30 @@ void ShaderProgramManager::UseFragmentShader(const Pica::Regs& regs, bool use_no
 }
 
 void ShaderProgramManager::ApplyTo(OpenGLState& state) {
+    if (impl->separable) {
         if (driver.HasBug(DriverBug::ShaderStageChangeFreeze)) {
-        glUseProgramStages(impl->pipeline.handle,
+            glUseProgramStages(
-                           GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT | GL_FRAGMENT_SHADER_BIT,
+                impl->pipeline.handle,
-                           0);
+                GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT | GL_FRAGMENT_SHADER_BIT, 0);
         }
 
         glUseProgramStages(impl->pipeline.handle, GL_VERTEX_SHADER_BIT, impl->current.vs);
         glUseProgramStages(impl->pipeline.handle, GL_GEOMETRY_SHADER_BIT, impl->current.gs);
         glUseProgramStages(impl->pipeline.handle, GL_FRAGMENT_SHADER_BIT, impl->current.fs);
-
         state.draw.shader_program = 0;
         state.draw.program_pipeline = impl->pipeline.handle;
+    } else {
+        const u64 unique_identifier = impl->current.GetConfigHash();
+        OGLProgram& cached_program = impl->program_cache[unique_identifier];
+        if (cached_program.handle == 0) {
+            cached_program.Create(false,
+                                  std::array{impl->current.vs, impl->current.gs, impl->current.fs});
+            auto& disk_cache = impl->disk_cache;
+            disk_cache.SaveDumpToFile(unique_identifier, cached_program.handle,
+                                      VideoCore::g_hw_shader_accurate_mul);
+        }
+        state.draw.shader_program = cached_program.handle;
+    }
 }
 
 void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
@@ -318,7 +400,7 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
 
     // Load uncompressed precompiled file for non-separable shaders.
     // Precompiled file for separable shaders is compressed.
-    auto [decompiled, dumps] = disk_cache.LoadPrecompiled();
+    auto [decompiled, dumps] = disk_cache.LoadPrecompiled(impl->separable);
 
     if (stop_loading) {
         return;
@@ -336,11 +418,11 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
         callback(VideoCore::LoadCallbackStage::Decompile, 0, raws.size());
     }
     std::vector<std::size_t> load_raws_index;
-
     // Loads both decompiled and precompiled shaders from the cache. If either one is missing for
-    const auto LoadPrecompiledShader =
+    const auto LoadPrecompiledShader = [&](std::size_t begin, std::size_t end,
-        [&](std::size_t begin, std::size_t end, std::span<const ShaderDiskCacheRaw> raw_cache,
+                                           std::span<const ShaderDiskCacheRaw> raw_cache,
-            const ShaderDecompiledMap& decompiled_map, const ShaderDumpsMap& dump_map) {
+                                           const ShaderDecompiledMap& decompiled_map,
+                                           const ShaderDumpsMap& dump_map) {
         for (std::size_t i = begin; i < end; ++i) {
             if (stop_loading || compilation_failed) {
                 return;
@@ -372,23 +454,24 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
                 }
 
                 // If the shader is dumped, attempt to load it
-                    shader = GeneratePrecompiledProgram(dump->second, supported_formats);
+                shader =
+                    GeneratePrecompiledProgram(dump->second, supported_formats, impl->separable);
                 if (shader.handle == 0) {
                     // If any shader failed, stop trying to compile, delete the cache, and start
-                        // loading from raws.
+                    // loading from raws
                     compilation_failed = true;
                     return;
                 }
-                    // We have both the binary shader and the decompiled, so inject it into the
+                // we have both the binary shader and the decompiled, so inject it into the
-                    // cache.
+                // cache
                 if (raw.GetProgramType() == ProgramType::VS) {
                     auto [conf, setup] = BuildVSConfigFromRaw(raw, driver);
                     std::scoped_lock lock(mutex);
                     impl->programmable_vertex_shaders.Inject(conf, decomp->second.code,
                                                              std::move(shader));
                 } else if (raw.GetProgramType() == ProgramType::FS) {
-                        PicaFSConfig conf(raw.GetRawShaderConfig(), false, driver.IsOpenGLES(),
+                    PicaFSConfig conf(raw.GetRawShaderConfig(), false, driver.IsOpenGLES(), false,
-                                          false, driver.HasBlendMinMaxFactor());
+                                      driver.HasBlendMinMaxFactor());
                     std::scoped_lock lock(mutex);
                     impl->fragment_shaders.Inject(conf, std::move(shader));
                 } else {
@@ -410,7 +493,42 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
         }
     };
 
+    const auto LoadPrecompiledProgram = [&](const ShaderDecompiledMap& decompiled_map,
+                                            const ShaderDumpsMap& dump_map) {
+        std::size_t i{0};
+        for (const auto& dump : dump_map) {
+            if (stop_loading) {
+                break;
+            }
+            const u64 unique_identifier{dump.first};
+            const auto decomp{decompiled_map.find(unique_identifier)};
+
+            // Only load the program if its sanitize_mul setting matches
+            if (decomp->second.sanitize_mul != VideoCore::g_hw_shader_accurate_mul) {
+                continue;
+            }
+
+            // If the shader program is dumped, attempt to load it
+            OGLProgram shader =
+                GeneratePrecompiledProgram(dump.second, supported_formats, impl->separable);
+            if (shader.handle != 0) {
+                impl->program_cache.emplace(unique_identifier, std::move(shader));
+            } else {
+                LOG_ERROR(Frontend, "Failed to link Precompiled program!");
+                compilation_failed = true;
+                break;
+            }
+            if (callback) {
+                callback(VideoCore::LoadCallbackStage::Decompile, ++i, dump_map.size());
+            }
+        }
+    };
+
+    if (impl->separable) {
         LoadPrecompiledShader(0, raws.size(), raws, decompiled, dumps);
+    } else {
+        LoadPrecompiledProgram(decompiled, dumps);
+    }
 
     bool load_all_raws = false;
     if (compilation_failed) {
@@ -421,6 +539,11 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
         precompiled_cache_altered = true;
         load_all_raws = true;
     }
+    // TODO(SachinV): Skip loading raws until we implement a proper way to link non-seperable
+    // shaders.
+    if (!impl->separable) {
+        return;
+    }
 
     const std::size_t load_raws_size = load_all_raws ? raws.size() : load_raws_index.size();
 
@@ -447,25 +570,25 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
             GLuint handle{0};
             std::string code;
             // Otherwise decompile and build the shader at boot and save the result to the
-            // precompiled file.
+            // precompiled file
             if (raw.GetProgramType() == ProgramType::VS) {
-                const auto [conf, setup] = BuildVSConfigFromRaw(raw, driver);
+                auto [conf, setup] = BuildVSConfigFromRaw(raw, driver);
-                code = GLSL::GenerateVertexShader(setup, conf, true);
+                code = GLSL::GenerateVertexShader(setup, conf, impl->separable);
-                OGLProgram program{};
+                OGLShaderStage stage{impl->separable};
-                program.Create(code, GL_VERTEX_SHADER);
+                stage.Create(code.c_str(), GL_VERTEX_SHADER);
-                handle = program.handle;
+                handle = stage.GetHandle();
                 sanitize_mul = conf.state.sanitize_mul;
                 std::scoped_lock lock(mutex);
-                impl->programmable_vertex_shaders.Inject(conf, code, std::move(program));
+                impl->programmable_vertex_shaders.Inject(conf, code, std::move(stage));
             } else if (raw.GetProgramType() == ProgramType::FS) {
                 PicaFSConfig conf(raw.GetRawShaderConfig(), false, driver.IsOpenGLES(), false,
                                   driver.HasBlendMinMaxFactor());
-                code = GLSL::GenerateFragmentShader(conf, true);
+                code = GLSL::GenerateFragmentShader(conf, impl->separable);
-                OGLProgram program{};
+                OGLShaderStage stage{impl->separable};
-                program.Create(code, GL_FRAGMENT_SHADER);
+                stage.Create(code.c_str(), GL_FRAGMENT_SHADER);
-                handle = program.handle;
+                handle = stage.GetHandle();
                 std::scoped_lock lock(mutex);
-                impl->fragment_shaders.Inject(conf, std::move(program));
+                impl->fragment_shaders.Inject(conf, std::move(stage));
             } else {
                 // Unsupported shader type got stored somehow so nuke the cache
                 LOG_ERROR(Frontend, "failed to load raw ProgramType {}", raw.GetProgramType());

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -33,29 +33,22 @@ enum UniformBindings {
 /// A class that manage different shader stages and configures them with given config data.
 class ShaderProgramManager {
 public:
-    explicit ShaderProgramManager(Frontend::EmuWindow& emu_window, const Driver& driver);
+    ShaderProgramManager(Frontend::EmuWindow& emu_window, const Driver& driver, bool separable);
     ~ShaderProgramManager();
 
-    /// Loads the pipeline cache stored to disk.
     void LoadDiskCache(const std::atomic_bool& stop_loading,
                        const VideoCore::DiskResourceLoadCallback& callback);
 
-    /// Binds a PICA decompiled vertex shader.
     bool UseProgrammableVertexShader(const Pica::Regs& config, Pica::Shader::ShaderSetup& setup);
 
-    /// Binds a passthrough vertex shader.
     void UseTrivialVertexShader();
 
-    /// Binds a passthrough geometry shader.
     void UseFixedGeometryShader(const Pica::Regs& regs);
 
-    /// Binds no geometry shader.
     void UseTrivialGeometryShader();
 
-    /// Binds a fragment shader generated from PICA state.
     void UseFragmentShader(const Pica::Regs& config, bool use_normal);
 
-    /// Binds current shader state to provided OpenGLState.
     void ApplyTo(OpenGLState& state);
 
 private:

src/video_core/renderer_opengl/gl_shader_util.cpp

@@ -13,9 +13,10 @@
 
 namespace OpenGL {
 
-std::string_view GetPreamble() {
+GLuint LoadShader(std::string_view source, GLenum type) {
+    std::string preamble;
     if (GLES) {
-        return R"(#version 320 es
+        preamble = R"(#version 320 es
 
 #if defined(GL_ANDROID_extension_pack_es31a)
 #extension GL_ANDROID_extension_pack_es31a : enable
@@ -25,12 +26,9 @@ std::string_view GetPreamble() {
 #extension GL_EXT_clip_cull_distance : enable
 #endif // defined(GL_EXT_clip_cull_distance)
 )";
+    } else {
+        preamble = "#version 430 core\n";
     }
-    return "#version 430 core\n";
-}
-
-GLuint LoadShader(std::string_view source, GLenum type) {
-    const auto preamble = GetPreamble();
 
     std::string_view debug_type;
     switch (type) {
@@ -74,9 +72,11 @@ GLuint LoadShader(std::string_view source, GLenum type) {
     return shader_id;
 }
 
-GLuint LoadProgram(std::span<const GLuint> shaders) {
+GLuint LoadProgram(bool separable_program, std::span<const GLuint> shaders) {
+    // Link the program
     LOG_DEBUG(Render_OpenGL, "Linking program...");
-    const GLuint program_id = glCreateProgram();
+
+    GLuint program_id = glCreateProgram();
 
     for (GLuint shader : shaders) {
         if (shader != 0) {
@@ -84,6 +84,10 @@ GLuint LoadProgram(std::span<const GLuint> shaders) {
         }
     }
 
+    if (separable_program) {
+        glProgramParameteri(program_id, GL_PROGRAM_SEPARABLE, GL_TRUE);
+    }
+
     glProgramParameteri(program_id, GL_PROGRAM_BINARY_RETRIEVABLE_HINT, GL_TRUE);
     glLinkProgram(program_id);
 

src/video_core/renderer_opengl/gl_shader_util.h

@@ -9,11 +9,6 @@
 
 namespace OpenGL {
 
-/**
- * Utility function to retrieve the preamble to compile an OpenGL GLSL shader
- */
-std::string_view GetPreamble();
-
 /**
  * Utility function to create and compile an OpenGL GLSL shader
  * @param source String of the GLSL shader program
@@ -27,6 +22,6 @@ GLuint LoadShader(std::string_view source, GLenum type);
  * @param shaders ID of shaders to attach to the program
  * @returns Handle of the newly created OpenGL program object
  */
-GLuint LoadProgram(std::span<const GLuint> shaders);
+GLuint LoadProgram(bool separable_program, std::span<const GLuint> shaders);
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_state.cpp

@@ -59,7 +59,6 @@ OpenGLState::OpenGLState() {
     texture_buffer_lut_lf.texture_buffer = 0;
     texture_buffer_lut_rg.texture_buffer = 0;
     texture_buffer_lut_rgba.texture_buffer = 0;
-    color_buffer.texture_2d = 0;
 
     image_shadow_buffer = 0;
     image_shadow_texture_px = 0;

src/video_core/renderer_vulkan/vk_texture_runtime.cpp

@@ -1562,7 +1562,7 @@ DebugScope::DebugScope(TextureRuntime& runtime, Common::Vec4f color, std::string
     if (!has_debug_tool) {
         return;
     }
-    scheduler.Record([color, label = std::string(label)](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([color, label](vk::CommandBuffer cmdbuf) {
         const vk::DebugUtilsLabelEXT debug_label = {
             .pLabelName = label.data(),
             .color = std::array{color[0], color[1], color[2], color[3]},