vk_scheduler: Use std::jthread

Useful for jthreads which make use of the threadsafe queues.

.ci/scripts/linux/docker.sh

@@ -18,7 +18,8 @@ cmake .. \
       -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON \
       -DENABLE_QT_TRANSLATION=ON \
       -DUSE_DISCORD_PRESENCE=ON \
-      -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"}
+      -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} \
+      -DYUZU_USE_BUNDLED_FFMPEG=ON
 
 make -j$(nproc)
 

CMakeLists.txt

@@ -25,7 +25,7 @@ option(YUZU_USE_BUNDLED_BOOST "Download bundled Boost" OFF)
 
 option(YUZU_USE_BUNDLED_LIBUSB "Compile bundled libusb" OFF)
 
-CMAKE_DEPENDENT_OPTION(YUZU_USE_BUNDLED_FFMPEG "Download/Build bundled FFmpeg" ON "WIN32" OFF)
+option(YUZU_USE_BUNDLED_FFMPEG "Download/Build bundled FFmpeg" "${WIN32}")
 
 option(YUZU_USE_QT_WEB_ENGINE "Use QtWebEngine for web applet implementation" OFF)
 
@@ -376,7 +376,7 @@ if (ENABLE_SDL2)
     if (YUZU_USE_BUNDLED_SDL2)
         # Detect toolchain and platform
         if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1930) AND ARCHITECTURE_x86_64)
-            set(SDL2_VER "SDL2-2.0.15-prerelease")
+            set(SDL2_VER "SDL2-2.0.16")
         else()
             message(FATAL_ERROR "No bundled SDL2 binaries for your toolchain. Disable YUZU_USE_BUNDLED_SDL2 and provide your own.")
         endif()
@@ -396,7 +396,7 @@ if (ENABLE_SDL2)
     elseif (YUZU_USE_EXTERNAL_SDL2)
         message(STATUS "Using SDL2 from externals.")
     else()
-        find_package(SDL2 2.0.15 REQUIRED)
+        find_package(SDL2 2.0.16 REQUIRED)
 
         # Some installations don't set SDL2_LIBRARIES
         if("${SDL2_LIBRARIES}" STREQUAL "")
@@ -496,7 +496,7 @@ endif()
 # Ensure libusb is properly configured (based on dolphin libusb include)
 if(NOT APPLE AND NOT YUZU_USE_BUNDLED_LIBUSB)
     include(FindPkgConfig)
-    if (PKG_CONFIG_FOUND)
+    if (PKG_CONFIG_FOUND AND NOT CMAKE_SYSTEM_NAME MATCHES "DragonFly|FreeBSD")
         pkg_check_modules(LIBUSB QUIET libusb-1.0>=1.0.24)
     else()
         find_package(LibUSB)
@@ -518,6 +518,10 @@ set(FFmpeg_COMPONENTS
     avutil
     swscale)
 
+if (${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
+    Include(FindPkgConfig REQUIRED)
+    pkg_check_modules(LIBVA libva)
+endif()
 if (NOT YUZU_USE_BUNDLED_FFMPEG)
     # Use system installed FFmpeg
     find_package(FFmpeg QUIET COMPONENTS ${FFmpeg_COMPONENTS})
@@ -540,6 +544,9 @@ endif()
 
 if (YUZU_USE_BUNDLED_FFMPEG)
     if (NOT WIN32)
+        # TODO(lat9nq): Move this to externals/ffmpeg/CMakeLists.txt (and move externals/ffmpeg to
+        # externals/ffmpeg/ffmpeg)
+
         # Build FFmpeg from externals
         message(STATUS "Using FFmpeg from externals")
 
@@ -579,12 +586,84 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                 CACHE PATH "Paths to FFmpeg libraries" FORCE)
         endforeach()
 
-        set(FFmpeg_INCLUDE_DIR
-            "${FFmpeg_PREFIX};${FFmpeg_BUILD_DIR}"
-            CACHE PATH "Path to FFmpeg headers" FORCE)
+        Include(FindPkgConfig REQUIRED)
+        pkg_check_modules(LIBVA libva)
+        pkg_check_modules(CUDA cuda)
+        pkg_check_modules(FFNVCODEC ffnvcodec)
+        pkg_check_modules(VDPAU vdpau)
+
+        set(FFmpeg_HWACCEL_LIBRARIES)
+        set(FFmpeg_HWACCEL_FLAGS)
+        set(FFmpeg_HWACCEL_INCLUDE_DIRS)
+        set(FFmpeg_HWACCEL_LDFLAGS)
+
+        if(LIBVA_FOUND)
+            pkg_check_modules(LIBDRM libdrm REQUIRED)
+            find_package(X11 REQUIRED)
+            pkg_check_modules(LIBVA-DRM libva-drm REQUIRED)
+            pkg_check_modules(LIBVA-X11 libva-x11 REQUIRED)
+            list(APPEND FFmpeg_HWACCEL_LIBRARIES
+                ${LIBDRM_LIBRARIES}
+                ${X11_LIBRARIES}
+                ${LIBVA-DRM_LIBRARIES}
+                ${LIBVA-X11_LIBRARIES}
+                ${LIBVA_LIBRARIES})
+            set(FFmpeg_HWACCEL_FLAGS
+                --enable-hwaccel=h264_vaapi
+                --enable-hwaccel=vp9_vaapi
+                --enable-libdrm)
+            list(APPEND FFmpeg_HWACCEL_INCLUDE_DIRS
+                ${LIBDRM_INCLUDE_DIRS}
+                ${X11_INCLUDE_DIRS}
+                ${LIBVA-DRM_INCLUDE_DIRS}
+                ${LIBVA-X11_INCLUDE_DIRS}
+                ${LIBVA_INCLUDE_DIRS}
+            )
+            message(STATUS "VA-API found")
+        else()
+            set(FFmpeg_HWACCEL_FLAGS --disable-vaapi)
+        endif()
+
+        if (FFNVCODEC_FOUND AND CUDA_FOUND)
+            list(APPEND FFmpeg_HWACCEL_FLAGS
+                --enable-cuvid
+                --enable-ffnvcodec
+                --enable-nvdec
+                --enable-hwaccel=h264_nvdec
+                --enable-hwaccel=vp9_nvdec
+                --extra-cflags=-I${CUDA_INCLUDE_DIRS}
+            )
+            list(APPEND FFmpeg_HWACCEL_LIBRARIES
+                ${FFNVCODEC_LIBRARIES}
+                ${CUDA_LIBRARIES}
+            )
+            list(APPEND FFmpeg_HWACCEL_INCLUDE_DIRS
+                ${FFNVCODEC_INCLUDE_DIRS}
+                ${CUDA_INCLUDE_DIRS}
+            )
+            list(APPEND FFmpeg_HWACCEL_LDFLAGS
+                ${FFNVCODEC_LDFLAGS}
+                ${CUDA_LDFLAGS}
+            )
+            message(STATUS "ffnvcodec libraries version ${FFNVCODEC_VERSION} found")
+        endif()
+
+        if (VDPAU_FOUND)
+            list(APPEND FFmpeg_HWACCEL_FLAGS
+                --enable-vdpau
+                --enable-hwaccel=h264_vdpau
+                --enable-hwaccel=vp9_vdpau
+            )
+            list(APPEND FFmpeg_HWACCEL_LIBRARIES ${VDPAU_LIBRARIES})
+            list(APPEND FFmpeg_HWACCEL_INCLUDE_DIRS ${VDPAU_INCLUDE_DIRS})
+            list(APPEND FFmpeg_HWACCEL_LDFLAGS ${VDPAU_LDFLAGS})
+            message(STATUS "vdpau libraries version ${VDPAU_VERSION} found")
+        else()
+            list(APPEND FFmpeg_HWACCEL_FLAGS --disable-vdpau)
+        endif()
 
         # `configure` parameters builds only exactly what yuzu needs from FFmpeg
-        # `--disable-{vaapi,vdpau}` is needed to avoid linking issues
+        # `--disable-vdpau` is needed to avoid linking issues
         add_custom_command(
             OUTPUT
                 ${FFmpeg_MAKEFILE}
@@ -600,13 +679,15 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                     --disable-network
                     --disable-postproc
                     --disable-swresample
-                    --disable-vaapi
-                    --disable-vdpau
                     --enable-decoder=h264
                     --enable-decoder=vp9
+                    --cc="${CMAKE_C_COMPILER}"
+                    --cxx="${CMAKE_CXX_COMPILER}"
+                    ${FFmpeg_HWACCEL_FLAGS}
             WORKING_DIRECTORY
                 ${FFmpeg_BUILD_DIR}
         )
+        unset(FFmpeg_HWACCEL_FLAGS)
 
         # Workaround for Ubuntu 18.04's older version of make not being able to call make as a child
         # with context of the jobserver. Also helps ninja users.
@@ -616,19 +697,36 @@ if (YUZU_USE_BUNDLED_FFMPEG)
             OUTPUT_VARIABLE
                 SYSTEM_THREADS)
 
+        set(FFmpeg_BUILD_LIBRARIES ${FFmpeg_LIBRARIES})
         add_custom_command(
             OUTPUT
-                ${FFmpeg_LIBRARIES}
+                ${FFmpeg_BUILD_LIBRARIES}
             COMMAND
                 make -j${SYSTEM_THREADS}
             WORKING_DIRECTORY
                 ${FFmpeg_BUILD_DIR}
         )
 
+        set(FFmpeg_INCLUDE_DIR
+            "${FFmpeg_PREFIX};${FFmpeg_BUILD_DIR};${FFmpeg_HWACCEL_INCLUDE_DIRS}"
+            CACHE PATH "Path to FFmpeg headers" FORCE)
+
+        set(FFmpeg_LDFLAGS
+            "${FFmpeg_HWACCEL_LDFLAGS}"
+            CACHE STRING "FFmpeg linker flags" FORCE)
+
         # ALL makes this custom target build every time
         # but it won't actually build if the DEPENDS parameter is up to date
         add_custom_target(ffmpeg-configure ALL DEPENDS ${FFmpeg_MAKEFILE})
-        add_custom_target(ffmpeg-build ALL DEPENDS ${FFmpeg_LIBRARIES} ffmpeg-configure)
+        add_custom_target(ffmpeg-build ALL DEPENDS ${FFmpeg_BUILD_LIBRARIES} ffmpeg-configure)
+        link_libraries(${FFmpeg_LIBVA_LIBRARIES})
+        set(FFmpeg_LIBRARIES ${FFmpeg_BUILD_LIBRARIES} ${FFmpeg_HWACCEL_LIBRARIES}
+            CACHE PATH "Paths to FFmpeg libraries" FORCE)
+        unset(FFmpeg_BUILD_LIBRARIES)
+        unset(FFmpeg_HWACCEL_FLAGS)
+        unset(FFmpeg_HWACCEL_INCLUDE_DIRS)
+        unset(FFmpeg_HWACCEL_LDFLAGS)
+        unset(FFmpeg_HWACCEL_LIBRARIES)
 
         if (FFmpeg_FOUND)
             message(STATUS "Found FFmpeg version ${FFmpeg_VERSION}")
@@ -637,12 +735,13 @@ if (YUZU_USE_BUNDLED_FFMPEG)
         endif()
     else() # WIN32
         # Use yuzu FFmpeg binaries
-        set(FFmpeg_EXT_NAME "ffmpeg-4.3.1")
+        set(FFmpeg_EXT_NAME "ffmpeg-4.4")
         set(FFmpeg_PATH "${CMAKE_BINARY_DIR}/externals/${FFmpeg_EXT_NAME}")
         download_bundled_external("ffmpeg/" ${FFmpeg_EXT_NAME} "")
         set(FFmpeg_FOUND YES)
         set(FFmpeg_INCLUDE_DIR "${FFmpeg_PATH}/include" CACHE PATH "Path to FFmpeg headers" FORCE)
         set(FFmpeg_LIBRARY_DIR "${FFmpeg_PATH}/bin" CACHE PATH "Path to FFmpeg library directory" FORCE)
+        set(FFmpeg_LDFLAGS "" CACHE STRING "FFmpeg linker flags" FORCE)
         set(FFmpeg_DLL_DIR "${FFmpeg_PATH}/bin" CACHE PATH "Path to FFmpeg dll's" FORCE)
         set(FFmpeg_LIBRARIES
             ${FFmpeg_LIBRARY_DIR}/swscale.lib
@@ -668,7 +767,7 @@ if (APPLE)
 elseif (WIN32)
     # WSAPoll and SHGetKnownFolderPath (AppData/Roaming) didn't exist before WinNT 6.x (Vista)
     add_definitions(-D_WIN32_WINNT=0x0600 -DWINVER=0x0600)
-    set(PLATFORM_LIBRARIES winmm ws2_32)
+    set(PLATFORM_LIBRARIES winmm ws2_32 iphlpapi)
     if (MINGW)
         # PSAPI is the Process Status API
         set(PLATFORM_LIBRARIES ${PLATFORM_LIBRARIES} psapi imm32 version)

dist/qt_themes/default/style.qss

@@ -51,11 +51,11 @@ QPushButton#GPUStatusBarButton:hover {
 }
 
 QPushButton#GPUStatusBarButton:checked {
-    color: #ff8040;
+    color: #b06020;
 }
 
 QPushButton#GPUStatusBarButton:!checked {
-    color: #40dd40;
+    color: #109010;
 }
 
 QPushButton#buttonRefreshDevices {

externals/CMakeLists.txt

@@ -7,7 +7,9 @@ include(DownloadExternals)
 # xbyak
 if (ARCHITECTURE_x86 OR ARCHITECTURE_x86_64)
     add_library(xbyak INTERFACE)
-    target_include_directories(xbyak SYSTEM INTERFACE ./xbyak/xbyak)
+    file(MAKE_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/xbyak/include)
+    file(COPY ${CMAKE_CURRENT_SOURCE_DIR}/xbyak/xbyak DESTINATION ${CMAKE_CURRENT_BINARY_DIR}/xbyak/include)
+    target_include_directories(xbyak SYSTEM INTERFACE ${CMAKE_CURRENT_BINARY_DIR}/xbyak/include)
     target_compile_definitions(xbyak INTERFACE XBYAK_NO_OP_NAMES)
 endif()
 
@@ -19,6 +21,7 @@ target_include_directories(catch-single-include INTERFACE catch/single_include)
 if (ARCHITECTURE_x86_64)
     set(DYNARMIC_TESTS OFF)
     set(DYNARMIC_NO_BUNDLED_FMT ON)
+    set(DYNARMIC_IGNORE_ASSERTS ON CACHE BOOL "" FORCE)
     add_subdirectory(dynarmic)
 endif()
 

externals/libusb/CMakeLists.txt

@@ -67,6 +67,8 @@ if (MINGW OR (${CMAKE_SYSTEM_NAME} MATCHES "Linux") OR APPLE)
             "${LIBUSB_MAKEFILE}"
         COMMAND
             env
+                CC="${CMAKE_C_COMPILER}"
+                CXX="${CMAKE_CXX_COMPILER}"
                 CFLAGS="${LIBUSB_CFLAGS}"
             sh "${LIBUSB_CONFIGURE}"
                 ${LIBUSB_CONFIGURE_ARGS}

src/common/CMakeLists.txt

@@ -53,6 +53,8 @@ add_library(common STATIC
     div_ceil.h
     dynamic_library.cpp
     dynamic_library.h
+    error.cpp
+    error.h
     fiber.cpp
     fiber.h
     fs/file.cpp
@@ -88,7 +90,6 @@ add_library(common STATIC
     microprofile.cpp
     microprofile.h
     microprofileui.h
-    misc.cpp
     nvidia_flags.cpp
     nvidia_flags.h
     page_table.cpp

src/common/assert.h

@@ -52,8 +52,12 @@ assert_noinline_call(const Fn& fn) {
 #define DEBUG_ASSERT(_a_) ASSERT(_a_)
 #define DEBUG_ASSERT_MSG(_a_, ...) ASSERT_MSG(_a_, __VA_ARGS__)
 #else // not debug
-#define DEBUG_ASSERT(_a_)
-#define DEBUG_ASSERT_MSG(_a_, _desc_, ...)
+#define DEBUG_ASSERT(_a_)                                                                          \
+    do {                                                                                           \
+    } while (0)
+#define DEBUG_ASSERT_MSG(_a_, _desc_, ...)                                                         \
+    do {                                                                                           \
+    } while (0)
 #endif
 
 #define UNIMPLEMENTED() ASSERT_MSG(false, "Unimplemented code!")

src/common/common_funcs.h

@@ -4,9 +4,8 @@
 
 #pragma once
 
-#include <algorithm>
 #include <array>
-#include <string>
+#include <iterator>
 
 #if !defined(ARCHITECTURE_x86_64)
 #include <cstdlib> // for exit
@@ -49,16 +48,6 @@ __declspec(dllimport) void __stdcall DebugBreak(void);
 
 #endif // _MSC_VER ndef
 
-// Generic function to get last error message.
-// Call directly after the command or use the error num.
-// This function might change the error code.
-// Defined in misc.cpp.
-[[nodiscard]] std::string GetLastErrorMsg();
-
-// Like GetLastErrorMsg(), but passing an explicit error code.
-// Defined in misc.cpp.
-[[nodiscard]] std::string NativeErrorToString(int e);
-
 #define DECLARE_ENUM_FLAG_OPERATORS(type)                                                          \
     [[nodiscard]] constexpr type operator|(type a, type b) noexcept {                              \
         using T = std::underlying_type_t<type>;                                                    \
@@ -72,6 +61,14 @@ __declspec(dllimport) void __stdcall DebugBreak(void);
         using T = std::underlying_type_t<type>;                                                    \
         return static_cast<type>(static_cast<T>(a) ^ static_cast<T>(b));                           \
     }                                                                                              \
+    [[nodiscard]] constexpr type operator<<(type a, type b) noexcept {                             \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<type>(static_cast<T>(a) << static_cast<T>(b));                          \
+    }                                                                                              \
+    [[nodiscard]] constexpr type operator>>(type a, type b) noexcept {                             \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<type>(static_cast<T>(a) >> static_cast<T>(b));                          \
+    }                                                                                              \
     constexpr type& operator|=(type& a, type b) noexcept {                                         \
         a = a | b;                                                                                 \
         return a;                                                                                  \
@@ -84,6 +81,14 @@ __declspec(dllimport) void __stdcall DebugBreak(void);
         a = a ^ b;                                                                                 \
         return a;                                                                                  \
     }                                                                                              \
+    constexpr type& operator<<=(type& a, type b) noexcept {                                        \
+        a = a << b;                                                                                \
+        return a;                                                                                  \
+    }                                                                                              \
+    constexpr type& operator>>=(type& a, type b) noexcept {                                        \
+        a = a >> b;                                                                                \
+        return a;                                                                                  \
+    }                                                                                              \
     [[nodiscard]] constexpr type operator~(type key) noexcept {                                    \
         using T = std::underlying_type_t<type>;                                                    \
         return static_cast<type>(~static_cast<T>(key));                                            \

src/common/error.cpp

@@ -10,7 +10,9 @@
 #include <cstring>
 #endif
 
-#include "common/common_funcs.h"
+#include "common/error.h"
+
+namespace Common {
 
 std::string NativeErrorToString(int e) {
 #ifdef _WIN32
@@ -50,3 +52,5 @@ std::string GetLastErrorMsg() {
     return NativeErrorToString(errno);
 #endif
 }
+
+} // namespace Common

src/common/error.h

@@ -0,0 +1,21 @@
+// Copyright 2013 Dolphin Emulator Project / 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <string>
+
+namespace Common {
+
+// Generic function to get last error message.
+// Call directly after the command or use the error num.
+// This function might change the error code.
+// Defined in error.cpp.
+[[nodiscard]] std::string GetLastErrorMsg();
+
+// Like GetLastErrorMsg(), but passing an explicit error code.
+// Defined in error.cpp.
+[[nodiscard]] std::string NativeErrorToString(int e);
+
+} // namespace Common

src/common/hex_util.h

@@ -61,7 +61,7 @@ template <typename ContiguousContainer>
     return out;
 }
 
-[[nodiscard]] constexpr std::array<u8, 16> AsArray(const char (&data)[17]) {
+[[nodiscard]] constexpr std::array<u8, 16> AsArray(const char (&data)[33]) {
     return HexStringToArray<16>(data);
 }
 

src/common/host_memory.cpp

@@ -6,7 +6,7 @@
 #include <windows.h>
 #include "common/dynamic_library.h"
 
-#elif defined(__linux__) // ^^^ Windows ^^^ vvv Linux vvv
+#elif defined(__linux__) || defined(__FreeBSD__) // ^^^ Windows ^^^ vvv Linux vvv
 
 #ifndef _GNU_SOURCE
 #define _GNU_SOURCE
@@ -343,7 +343,7 @@ private:
     std::unordered_map<size_t, size_t> placeholder_host_pointers; ///< Placeholder backing offset
 };
 
-#elif defined(__linux__) // ^^^ Windows ^^^ vvv Linux vvv
+#elif defined(__linux__) || defined(__FreeBSD__) // ^^^ Windows ^^^ vvv Linux vvv
 
 class HostMemory::Impl {
 public:
@@ -357,7 +357,12 @@ public:
         });
 
         // Backing memory initialization
+#if defined(__FreeBSD__) && __FreeBSD__ < 13
+        // XXX Drop after FreeBSD 12.* reaches EOL on 2024-06-30
+        fd = shm_open(SHM_ANON, O_RDWR, 0600);
+#else
         fd = memfd_create("HostMemory", 0);
+#endif
         if (fd == -1) {
             LOG_CRITICAL(HW_Memory, "memfd_create failed: {}", strerror(errno));
             throw std::bad_alloc{};

src/common/logging/backend.cpp

@@ -2,13 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <algorithm>
 #include <atomic>
 #include <chrono>
 #include <climits>
-#include <condition_variable>
-#include <memory>
-#include <mutex>
+#include <exception>
 #include <thread>
 #include <vector>
 
@@ -16,28 +13,174 @@
 #include <windows.h> // For OutputDebugStringW
 #endif
 
-#include "common/assert.h"
 #include "common/fs/file.h"
 #include "common/fs/fs.h"
+#include "common/fs/fs_paths.h"
+#include "common/fs/path_util.h"
 #include "common/literals.h"
+#include "common/thread.h"
 
 #include "common/logging/backend.h"
 #include "common/logging/log.h"
 #include "common/logging/text_formatter.h"
 #include "common/settings.h"
+#ifdef _WIN32
 #include "common/string_util.h"
+#endif
 #include "common/threadsafe_queue.h"
 
 namespace Common::Log {
 
+namespace {
+
+/**
+ * Interface for logging backends.
+ */
+class Backend {
+public:
+    virtual ~Backend() = default;
+
+    virtual void Write(const Entry& entry) = 0;
+
+    virtual void EnableForStacktrace() = 0;
+
+    virtual void Flush() = 0;
+};
+
+/**
+ * Backend that writes to stderr and with color
+ */
+class ColorConsoleBackend final : public Backend {
+public:
+    explicit ColorConsoleBackend() = default;
+
+    ~ColorConsoleBackend() override = default;
+
+    void Write(const Entry& entry) override {
+        if (enabled.load(std::memory_order_relaxed)) {
+            PrintColoredMessage(entry);
+        }
+    }
+
+    void Flush() override {
+        // stderr shouldn't be buffered
+    }
+
+    void EnableForStacktrace() override {
+        enabled = true;
+    }
+
+    void SetEnabled(bool enabled_) {
+        enabled = enabled_;
+    }
+
+private:
+    std::atomic_bool enabled{false};
+};
+
+/**
+ * Backend that writes to a file passed into the constructor
+ */
+class FileBackend final : public Backend {
+public:
+    explicit FileBackend(const std::filesystem::path& filename) {
+        auto old_filename = filename;
+        old_filename += ".old.txt";
+
+        // Existence checks are done within the functions themselves.
+        // We don't particularly care if these succeed or not.
+        static_cast<void>(FS::RemoveFile(old_filename));
+        static_cast<void>(FS::RenameFile(filename, old_filename));
+
+        file = std::make_unique<FS::IOFile>(filename, FS::FileAccessMode::Write,
+                                            FS::FileType::TextFile);
+    }
+
+    ~FileBackend() override = default;
+
+    void Write(const Entry& entry) override {
+        if (!enabled) {
+            return;
+        }
+
+        bytes_written += file->WriteString(FormatLogMessage(entry).append(1, '\n'));
+
+        using namespace Common::Literals;
+        // Prevent logs from exceeding a set maximum size in the event that log entries are spammed.
+        const auto write_limit = Settings::values.extended_logging ? 1_GiB : 100_MiB;
+        const bool write_limit_exceeded = bytes_written > write_limit;
+        if (entry.log_level >= Level::Error || write_limit_exceeded) {
+            if (write_limit_exceeded) {
+                // Stop writing after the write limit is exceeded.
+                // Don't close the file so we can print a stacktrace if necessary
+                enabled = false;
+            }
+            file->Flush();
+        }
+    }
+
+    void Flush() override {
+        file->Flush();
+    }
+
+    void EnableForStacktrace() override {
+        enabled = true;
+        bytes_written = 0;
+    }
+
+private:
+    std::unique_ptr<FS::IOFile> file;
+    bool enabled = true;
+    std::size_t bytes_written = 0;
+};
+
+/**
+ * Backend that writes to Visual Studio's output window
+ */
+class DebuggerBackend final : public Backend {
+public:
+    explicit DebuggerBackend() = default;
+
+    ~DebuggerBackend() override = default;
+
+    void Write(const Entry& entry) override {
+#ifdef _WIN32
+        ::OutputDebugStringW(UTF8ToUTF16W(FormatLogMessage(entry).append(1, '\n')).c_str());
+#endif
+    }
+
+    void Flush() override {}
+
+    void EnableForStacktrace() override {}
+};
+
+bool initialization_in_progress_suppress_logging = true;
+
 /**
  * Static state as a singleton.
  */
 class Impl {
 public:
     static Impl& Instance() {
-        static Impl backend;
-        return backend;
+        if (!instance) {
+            throw std::runtime_error("Using Logging instance before its initialization");
+        }
+        return *instance;
+    }
+
+    static void Initialize() {
+        if (instance) {
+            LOG_WARNING(Log, "Reinitializing logging backend");
+            return;
+        }
+        using namespace Common::FS;
+        const auto& log_dir = GetYuzuPath(YuzuPath::LogDir);
+        void(CreateDir(log_dir));
+        Filter filter;
+        filter.ParseFilterString(Settings::values.log_filter.GetValue());
+        instance = std::unique_ptr<Impl, decltype(&Deleter)>(new Impl(log_dir / LOG_FILE, filter),
+                                                             Deleter);
+        initialization_in_progress_suppress_logging = false;
     }
 
     Impl(const Impl&) = delete;
@@ -46,74 +189,54 @@ public:
     Impl(Impl&&) = delete;
     Impl& operator=(Impl&&) = delete;
 
-    void PushEntry(Class log_class, Level log_level, const char* filename, unsigned int line_num,
-                   const char* function, std::string message) {
-        message_queue.Push(
-            CreateEntry(log_class, log_level, filename, line_num, function, std::move(message)));
-    }
-
-    void AddBackend(std::unique_ptr<Backend> backend) {
-        std::lock_guard lock{writing_mutex};
-        backends.push_back(std::move(backend));
-    }
-
-    void RemoveBackend(std::string_view backend_name) {
-        std::lock_guard lock{writing_mutex};
-
-        std::erase_if(backends, [&backend_name](const auto& backend) {
-            return backend_name == backend->GetName();
-        });
-    }
-
-    const Filter& GetGlobalFilter() const {
-        return filter;
-    }
-
     void SetGlobalFilter(const Filter& f) {
         filter = f;
     }
 
-    Backend* GetBackend(std::string_view backend_name) {
-        const auto it =
-            std::find_if(backends.begin(), backends.end(),
-                         [&backend_name](const auto& i) { return backend_name == i->GetName(); });
-        if (it == backends.end())
-            return nullptr;
-        return it->get();
+    void SetColorConsoleBackendEnabled(bool enabled) {
+        color_console_backend.SetEnabled(enabled);
+    }
+
+    void PushEntry(Class log_class, Level log_level, const char* filename, unsigned int line_num,
+                   const char* function, std::string message) {
+        if (!filter.CheckMessage(log_class, log_level))
+            return;
+        const Entry& entry =
+            CreateEntry(log_class, log_level, filename, line_num, function, std::move(message));
+        message_queue.Push(entry);
     }
 
 private:
-    Impl() {
-        backend_thread = std::thread([&] {
-            Entry entry;
-            auto write_logs = [&](Entry& e) {
-                std::lock_guard lock{writing_mutex};
-                for (const auto& backend : backends) {
-                    backend->Write(e);
-                }
-            };
-            while (true) {
-                entry = message_queue.PopWait();
-                if (entry.final_entry) {
-                    break;
-                }
-                write_logs(entry);
-            }
-
-            // Drain the logging queue. Only writes out up to MAX_LOGS_TO_WRITE to prevent a
-            // case where a system is repeatedly spamming logs even on close.
-            const int MAX_LOGS_TO_WRITE = filter.IsDebug() ? INT_MAX : 100;
-            int logs_written = 0;
-            while (logs_written++ < MAX_LOGS_TO_WRITE && message_queue.Pop(entry)) {
-                write_logs(entry);
-            }
-        });
-    }
+    Impl(const std::filesystem::path& file_backend_filename, const Filter& filter_)
+        : filter{filter_}, file_backend{file_backend_filename}, backend_thread{std::thread([this] {
+              Common::SetCurrentThreadName("yuzu:Log");
+              Entry entry;
+              const auto write_logs = [this, &entry]() {
+                  ForEachBackend([&entry](Backend& backend) { backend.Write(entry); });
+              };
+              while (true) {
+                  entry = message_queue.PopWait();
+                  if (entry.final_entry) {
+                      break;
+                  }
+                  write_logs();
+              }
+              // Drain the logging queue. Only writes out up to MAX_LOGS_TO_WRITE to prevent a
+              // case where a system is repeatedly spamming logs even on close.
+              int max_logs_to_write = filter.IsDebug() ? INT_MAX : 100;
+              while (max_logs_to_write-- && message_queue.Pop(entry)) {
+                  write_logs();
+              }
+          })} {}
 
     ~Impl() {
-        Entry entry;
-        entry.final_entry = true;
-        message_queue.Push(entry);
+        StopBackendThread();
+    }
+
+    void StopBackendThread() {
+        Entry stop_entry{};
+        stop_entry.final_entry = true;
+        message_queue.Push(stop_entry);
         backend_thread.join();
     }
 
@@ -135,100 +258,51 @@ private:
         };
     }
 
-    std::mutex writing_mutex;
-    std::thread backend_thread;
-    std::vector<std::unique_ptr<Backend>> backends;
-    MPSCQueue<Entry> message_queue;
+    void ForEachBackend(auto lambda) {
+        lambda(static_cast<Backend&>(debugger_backend));
+        lambda(static_cast<Backend&>(color_console_backend));
+        lambda(static_cast<Backend&>(file_backend));
+    }
+
+    static void Deleter(Impl* ptr) {
+        delete ptr;
+    }
+
+    static inline std::unique_ptr<Impl, decltype(&Deleter)> instance{nullptr, Deleter};
+
     Filter filter;
+    DebuggerBackend debugger_backend{};
+    ColorConsoleBackend color_console_backend{};
+    FileBackend file_backend;
+
+    std::thread backend_thread;
+    MPSCQueue<Entry> message_queue{};
     std::chrono::steady_clock::time_point time_origin{std::chrono::steady_clock::now()};
 };
+} // namespace
 
-ConsoleBackend::~ConsoleBackend() = default;
-
-void ConsoleBackend::Write(const Entry& entry) {
-    PrintMessage(entry);
+void Initialize() {
+    Impl::Initialize();
 }
 
-ColorConsoleBackend::~ColorConsoleBackend() = default;
-
-void ColorConsoleBackend::Write(const Entry& entry) {
-    PrintColoredMessage(entry);
-}
-
-FileBackend::FileBackend(const std::filesystem::path& filename) {
-    auto old_filename = filename;
-    old_filename += ".old.txt";
-
-    // Existence checks are done within the functions themselves.
-    // We don't particularly care if these succeed or not.
-    FS::RemoveFile(old_filename);
-    void(FS::RenameFile(filename, old_filename));
-
-    file =
-        std::make_unique<FS::IOFile>(filename, FS::FileAccessMode::Write, FS::FileType::TextFile);
-}
-
-FileBackend::~FileBackend() = default;
-
-void FileBackend::Write(const Entry& entry) {
-    if (!file->IsOpen()) {
-        return;
-    }
-
-    using namespace Common::Literals;
-    // Prevent logs from exceeding a set maximum size in the event that log entries are spammed.
-    constexpr std::size_t MAX_BYTES_WRITTEN = 100_MiB;
-    constexpr std::size_t MAX_BYTES_WRITTEN_EXTENDED = 1_GiB;
-
-    const bool write_limit_exceeded =
-        bytes_written > MAX_BYTES_WRITTEN_EXTENDED ||
-        (bytes_written > MAX_BYTES_WRITTEN && !Settings::values.extended_logging);
-
-    // Close the file after the write limit is exceeded.
-    if (write_limit_exceeded) {
-        file->Close();
-        return;
-    }
-
-    bytes_written += file->WriteString(FormatLogMessage(entry).append(1, '\n'));
-    if (entry.log_level >= Level::Error) {
-        file->Flush();
-    }
-}
-
-DebuggerBackend::~DebuggerBackend() = default;
-
-void DebuggerBackend::Write(const Entry& entry) {
-#ifdef _WIN32
-    ::OutputDebugStringW(UTF8ToUTF16W(FormatLogMessage(entry).append(1, '\n')).c_str());
-#endif
+void DisableLoggingInTests() {
+    initialization_in_progress_suppress_logging = true;
 }
 
 void SetGlobalFilter(const Filter& filter) {
     Impl::Instance().SetGlobalFilter(filter);
 }
 
-void AddBackend(std::unique_ptr<Backend> backend) {
-    Impl::Instance().AddBackend(std::move(backend));
-}
-
-void RemoveBackend(std::string_view backend_name) {
-    Impl::Instance().RemoveBackend(backend_name);
-}
-
-Backend* GetBackend(std::string_view backend_name) {
-    return Impl::Instance().GetBackend(backend_name);
+void SetColorConsoleBackendEnabled(bool enabled) {
+    Impl::Instance().SetColorConsoleBackendEnabled(enabled);
 }
 
 void FmtLogMessageImpl(Class log_class, Level log_level, const char* filename,
                        unsigned int line_num, const char* function, const char* format,
                        const fmt::format_args& args) {
-    auto& instance = Impl::Instance();
-    const auto& filter = instance.GetGlobalFilter();
-    if (!filter.CheckMessage(log_class, log_level))
-        return;
-
-    instance.PushEntry(log_class, log_level, filename, line_num, function,
-                       fmt::vformat(format, args));
+    if (!initialization_in_progress_suppress_logging) {
+        Impl::Instance().PushEntry(log_class, log_level, filename, line_num, function,
+                                   fmt::vformat(format, args));
+    }
 }
 } // namespace Common::Log

src/common/logging/backend.h

@@ -5,120 +5,21 @@
 #pragma once
 
 #include <filesystem>
-#include <memory>
-#include <string>
-#include <string_view>
 #include "common/logging/filter.h"
-#include "common/logging/log.h"
-
-namespace Common::FS {
-class IOFile;
-}
 
 namespace Common::Log {
 
 class Filter;
 
-/**
- * Interface for logging backends. As loggers can be created and removed at runtime, this can be
- * used by a frontend for adding a custom logging backend as needed
- */
-class Backend {
-public:
-    virtual ~Backend() = default;
+/// Initializes the logging system. This should be the first thing called in main.
+void Initialize();
 
-    virtual void SetFilter(const Filter& new_filter) {
-        filter = new_filter;
-    }
-    virtual const char* GetName() const = 0;
-    virtual void Write(const Entry& entry) = 0;
-
-private:
-    Filter filter;
-};
+void DisableLoggingInTests();
 
 /**
- * Backend that writes to stderr without any color commands
- */
-class ConsoleBackend : public Backend {
-public:
-    ~ConsoleBackend() override;
-
-    static const char* Name() {
-        return "console";
-    }
-    const char* GetName() const override {
-        return Name();
-    }
-    void Write(const Entry& entry) override;
-};
-
-/**
- * Backend that writes to stderr and with color
- */
-class ColorConsoleBackend : public Backend {
-public:
-    ~ColorConsoleBackend() override;
-
-    static const char* Name() {
-        return "color_console";
-    }
-
-    const char* GetName() const override {
-        return Name();
-    }
-    void Write(const Entry& entry) override;
-};
-
-/**
- * Backend that writes to a file passed into the constructor
- */
-class FileBackend : public Backend {
-public:
-    explicit FileBackend(const std::filesystem::path& filename);
-    ~FileBackend() override;
-
-    static const char* Name() {
-        return "file";
-    }
-
-    const char* GetName() const override {
-        return Name();
-    }
-
-    void Write(const Entry& entry) override;
-
-private:
-    std::unique_ptr<FS::IOFile> file;
-    std::size_t bytes_written = 0;
-};
-
-/**
- * Backend that writes to Visual Studio's output window
- */
-class DebuggerBackend : public Backend {
-public:
-    ~DebuggerBackend() override;
-
-    static const char* Name() {
-        return "debugger";
-    }
-    const char* GetName() const override {
-        return Name();
-    }
-    void Write(const Entry& entry) override;
-};
-
-void AddBackend(std::unique_ptr<Backend> backend);
-
-void RemoveBackend(std::string_view backend_name);
-
-Backend* GetBackend(std::string_view backend_name);
-
-/**
- * The global filter will prevent any messages from even being processed if they are filtered. Each
- * backend can have a filter, but if the level is lower than the global filter, the backend will
- * never get the message
+ * The global filter will prevent any messages from even being processed if they are filtered.
  */
 void SetGlobalFilter(const Filter& filter);
-} // namespace Common::Log
+
+void SetColorConsoleBackendEnabled(bool enabled);
+} // namespace Common::Log

src/common/logging/filter.cpp

@@ -111,6 +111,7 @@ bool ParseFilterRule(Filter& instance, Iterator begin, Iterator end) {
     SUB(Service, NCM)                                                                              \
     SUB(Service, NFC)                                                                              \
     SUB(Service, NFP)                                                                              \
+    SUB(Service, NGCT)                                                                             \
     SUB(Service, NIFM)                                                                             \
     SUB(Service, NIM)                                                                              \
     SUB(Service, NPNS)                                                                             \

src/common/logging/types.h

@@ -81,6 +81,7 @@ enum class Class : u8 {
     Service_NCM,       ///< The NCM service
     Service_NFC,       ///< The NFC (Near-field communication) service
     Service_NFP,       ///< The NFP service
+    Service_NGCT,      ///< The NGCT (No Good Content for Terra) service
     Service_NIFM,      ///< The NIFM (Network interface) service
     Service_NIM,       ///< The NIM service
     Service_NPNS,      ///< The NPNS service

src/common/lru_cache.h

@@ -0,0 +1,140 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2+ or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <deque>
+#include <memory>
+#include <type_traits>
+
+#include "common/common_types.h"
+
+namespace Common {
+
+template <class Traits>
+class LeastRecentlyUsedCache {
+    using ObjectType = typename Traits::ObjectType;
+    using TickType = typename Traits::TickType;
+
+    struct Item {
+        ObjectType obj;
+        TickType tick;
+        Item* next{};
+        Item* prev{};
+    };
+
+public:
+    LeastRecentlyUsedCache() : first_item{}, last_item{} {}
+    ~LeastRecentlyUsedCache() = default;
+
+    size_t Insert(ObjectType obj, TickType tick) {
+        const auto new_id = Build();
+        auto& item = item_pool[new_id];
+        item.obj = obj;
+        item.tick = tick;
+        Attach(item);
+        return new_id;
+    }
+
+    void Touch(size_t id, TickType tick) {
+        auto& item = item_pool[id];
+        if (item.tick >= tick) {
+            return;
+        }
+        item.tick = tick;
+        if (&item == last_item) {
+            return;
+        }
+        Detach(item);
+        Attach(item);
+    }
+
+    void Free(size_t id) {
+        auto& item = item_pool[id];
+        Detach(item);
+        item.prev = nullptr;
+        item.next = nullptr;
+        free_items.push_back(id);
+    }
+
+    template <typename Func>
+    void ForEachItemBelow(TickType tick, Func&& func) {
+        static constexpr bool RETURNS_BOOL =
+            std::is_same_v<std::invoke_result<Func, ObjectType>, bool>;
+        Item* iterator = first_item;
+        while (iterator) {
+            if (static_cast<s64>(tick) - static_cast<s64>(iterator->tick) < 0) {
+                return;
+            }
+            Item* next = iterator->next;
+            if constexpr (RETURNS_BOOL) {
+                if (func(iterator->obj)) {
+                    return;
+                }
+            } else {
+                func(iterator->obj);
+            }
+            iterator = next;
+        }
+    }
+
+private:
+    size_t Build() {
+        if (free_items.empty()) {
+            const size_t item_id = item_pool.size();
+            auto& item = item_pool.emplace_back();
+            item.next = nullptr;
+            item.prev = nullptr;
+            return item_id;
+        }
+        const size_t item_id = free_items.front();
+        free_items.pop_front();
+        auto& item = item_pool[item_id];
+        item.next = nullptr;
+        item.prev = nullptr;
+        return item_id;
+    }
+
+    void Attach(Item& item) {
+        if (!first_item) {
+            first_item = &item;
+        }
+        if (!last_item) {
+            last_item = &item;
+        } else {
+            item.prev = last_item;
+            last_item->next = &item;
+            item.next = nullptr;
+            last_item = &item;
+        }
+    }
+
+    void Detach(Item& item) {
+        if (item.prev) {
+            item.prev->next = item.next;
+        }
+        if (item.next) {
+            item.next->prev = item.prev;
+        }
+        if (&item == first_item) {
+            first_item = item.next;
+            if (first_item) {
+                first_item->prev = nullptr;
+            }
+        }
+        if (&item == last_item) {
+            last_item = item.prev;
+            if (last_item) {
+                last_item->next = nullptr;
+            }
+        }
+    }
+
+    std::deque<Item> item_pool;
+    std::deque<size_t> free_items;
+    Item* first_item{};
+    Item* last_item{};
+};
+
+} // namespace Common

src/common/settings.cpp

@@ -54,12 +54,11 @@ void LogSettings() {
     log_setting("Renderer_GPUAccuracyLevel", values.gpu_accuracy.GetValue());
     log_setting("Renderer_UseAsynchronousGpuEmulation",
                 values.use_asynchronous_gpu_emulation.GetValue());
-    log_setting("Renderer_UseNvdecEmulation", values.use_nvdec_emulation.GetValue());
+    log_setting("Renderer_NvdecEmulation", values.nvdec_emulation.GetValue());
     log_setting("Renderer_AccelerateASTC", values.accelerate_astc.GetValue());
     log_setting("Renderer_UseVsync", values.use_vsync.GetValue());
     log_setting("Renderer_ShaderBackend", values.shader_backend.GetValue());
     log_setting("Renderer_UseAsynchronousShaders", values.use_asynchronous_shaders.GetValue());
-    log_setting("Renderer_UseGarbageCollection", values.use_caches_gc.GetValue());
     log_setting("Renderer_AnisotropicFilteringLevel", values.max_anisotropy.GetValue());
     log_setting("Audio_OutputEngine", values.sink_id.GetValue());
     log_setting("Audio_EnableAudioStretching", values.enable_audio_stretching.GetValue());
@@ -137,13 +136,12 @@ void RestoreGlobalState(bool is_powered_on) {
     values.use_disk_shader_cache.SetGlobal(true);
     values.gpu_accuracy.SetGlobal(true);
     values.use_asynchronous_gpu_emulation.SetGlobal(true);
-    values.use_nvdec_emulation.SetGlobal(true);
+    values.nvdec_emulation.SetGlobal(true);
     values.accelerate_astc.SetGlobal(true);
     values.use_vsync.SetGlobal(true);
     values.shader_backend.SetGlobal(true);
     values.use_asynchronous_shaders.SetGlobal(true);
     values.use_fast_gpu_time.SetGlobal(true);
-    values.use_caches_gc.SetGlobal(true);
     values.bg_red.SetGlobal(true);
     values.bg_green.SetGlobal(true);
     values.bg_blue.SetGlobal(true);

src/common/settings.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <algorithm>
 #include <array>
 #include <atomic>
 #include <chrono>
@@ -42,6 +43,17 @@ enum class CPUAccuracy : u32 {
     Unsafe = 2,
 };
 
+enum class FullscreenMode : u32 {
+    Borderless = 0,
+    Exclusive = 1,
+};
+
+enum class NvdecEmulation : u32 {
+    Off = 0,
+    CPU = 1,
+    GPU = 2,
+};
+
 /** The BasicSetting class is a simple resource manager. It defines a label and default value
  * alongside the actual value of the setting for simpler and less-error prone use with frontend
  * configurations. Setting a default value and label is required, though subclasses may deviate from
@@ -69,14 +81,14 @@ public:
      */
     explicit BasicSetting(const Type& default_val, const std::string& name)
         : default_value{default_val}, global{default_val}, label{name} {}
-    ~BasicSetting() = default;
+    virtual ~BasicSetting() = default;
 
     /**
      *  Returns a reference to the setting's value.
      *
      * @returns A reference to the setting
      */
-    [[nodiscard]] const Type& GetValue() const {
+    [[nodiscard]] virtual const Type& GetValue() const {
         return global;
     }
 
@@ -85,7 +97,7 @@ public:
      *
      * @param value The desired value
      */
-    void SetValue(const Type& value) {
+    virtual void SetValue(const Type& value) {
         Type temp{value};
         std::swap(global, temp);
     }
@@ -115,7 +127,7 @@ public:
      *
      * @returns A reference to the setting
      */
-    const Type& operator=(const Type& value) {
+    virtual const Type& operator=(const Type& value) {
         Type temp{value};
         std::swap(global, temp);
         return global;
@@ -126,7 +138,7 @@ public:
      *
      * @returns A reference to the setting
      */
-    explicit operator const Type&() const {
+    explicit virtual operator const Type&() const {
         return global;
     }
 
@@ -136,6 +148,51 @@ protected:
     const std::string label{};  ///< The setting's label
 };
 
+/**
+ * BasicRangedSetting class is intended for use with quantifiable settings that need a more
+ * restrictive range than implicitly defined by its type. Implements a minimum and maximum that is
+ * simply used to sanitize SetValue and the assignment overload.
+ */
+template <typename Type>
+class BasicRangedSetting : virtual public BasicSetting<Type> {
+public:
+    /**
+     * Sets a default value, minimum value, maximum value, and label.
+     *
+     * @param default_val Intial value of the setting, and default value of the setting
+     * @param min_val Sets the minimum allowed value of the setting
+     * @param max_val Sets the maximum allowed value of the setting
+     * @param name Label for the setting
+     */
+    explicit BasicRangedSetting(const Type& default_val, const Type& min_val, const Type& max_val,
+                                const std::string& name)
+        : BasicSetting<Type>{default_val, name}, minimum{min_val}, maximum{max_val} {}
+    virtual ~BasicRangedSetting() = default;
+
+    /**
+     * Like BasicSetting's SetValue, except value is clamped to the range of the setting.
+     *
+     * @param value The desired value
+     */
+    void SetValue(const Type& value) override {
+        this->global = std::clamp(value, minimum, maximum);
+    }
+
+    /**
+     * Like BasicSetting's assignment overload, except value is clamped to the range of the setting.
+     *
+     * @param value The desired value
+     * @returns A reference to the setting's value
+     */
+    const Type& operator=(const Type& value) override {
+        this->global = std::clamp(value, minimum, maximum);
+        return this->global;
+    }
+
+    const Type minimum; ///< Minimum allowed value of the setting
+    const Type maximum; ///< Maximum allowed value of the setting
+};
+
 /**
  * The Setting class is a slightly more complex version of the BasicSetting class. This adds a
  * custom setting to switch to when a guest application specifically requires it. The effect is that
@@ -147,7 +204,7 @@ protected:
  * Like the BasicSetting, this requires setting a default value and label to use.
  */
 template <typename Type>
-class Setting final : public BasicSetting<Type> {
+class Setting : virtual public BasicSetting<Type> {
 public:
     /**
      * Sets a default value, label, and setting value.
@@ -157,7 +214,7 @@ public:
      */
     explicit Setting(const Type& default_val, const std::string& name)
         : BasicSetting<Type>(default_val, name) {}
-    ~Setting() = default;
+    virtual ~Setting() = default;
 
     /**
      * Tells this setting to represent either the global or custom setting when other member
@@ -186,7 +243,13 @@ public:
      *
      * @returns The required value of the setting
      */
-    [[nodiscard]] const Type& GetValue(bool need_global = false) const {
+    [[nodiscard]] virtual const Type& GetValue() const override {
+        if (use_global) {
+            return this->global;
+        }
+        return custom;
+    }
+    [[nodiscard]] virtual const Type& GetValue(bool need_global) const {
         if (use_global || need_global) {
             return this->global;
         }
@@ -198,7 +261,7 @@ public:
      *
      * @param value The new value
      */
-    void SetValue(const Type& value) {
+    void SetValue(const Type& value) override {
         Type temp{value};
         if (use_global) {
             std::swap(this->global, temp);
@@ -214,7 +277,7 @@ public:
      *
      * @returns A reference to the current setting value
      */
-    const Type& operator=(const Type& value) {
+    const Type& operator=(const Type& value) override {
         Type temp{value};
         if (use_global) {
             std::swap(this->global, temp);
@@ -229,18 +292,87 @@ public:
      *
      * @returns A reference to the current setting value
      */
-    explicit operator const Type&() const {
+    virtual explicit operator const Type&() const override {
         if (use_global) {
             return this->global;
         }
         return custom;
     }
 
-private:
+protected:
     bool use_global{true}; ///< The setting's global state
     Type custom{};         ///< The custom value of the setting
 };
 
+/**
+ * RangedSetting is a Setting that implements a maximum and minimum value for its setting. Intended
+ * for use with quantifiable settings.
+ */
+template <typename Type>
+class RangedSetting final : public BasicRangedSetting<Type>, public Setting<Type> {
+public:
+    /**
+     * Sets a default value, minimum value, maximum value, and label.
+     *
+     * @param default_val Intial value of the setting, and default value of the setting
+     * @param min_val Sets the minimum allowed value of the setting
+     * @param max_val Sets the maximum allowed value of the setting
+     * @param name Label for the setting
+     */
+    explicit RangedSetting(const Type& default_val, const Type& min_val, const Type& max_val,
+                           const std::string& name)
+        : BasicSetting<Type>{default_val, name},
+          BasicRangedSetting<Type>{default_val, min_val, max_val, name}, Setting<Type>{default_val,
+                                                                                       name} {}
+    virtual ~RangedSetting() = default;
+
+    // The following are needed to avoid a MSVC bug
+    // (source: https://stackoverflow.com/questions/469508)
+    [[nodiscard]] const Type& GetValue() const override {
+        return Setting<Type>::GetValue();
+    }
+    [[nodiscard]] const Type& GetValue(bool need_global) const override {
+        return Setting<Type>::GetValue(need_global);
+    }
+    explicit operator const Type&() const override {
+        if (this->use_global) {
+            return this->global;
+        }
+        return this->custom;
+    }
+
+    /**
+     * Like BasicSetting's SetValue, except value is clamped to the range of the setting. Sets the
+     * appropriate value depending on the global state.
+     *
+     * @param value The desired value
+     */
+    void SetValue(const Type& value) override {
+        const Type temp = std::clamp(value, this->minimum, this->maximum);
+        if (this->use_global) {
+            this->global = temp;
+        }
+        this->custom = temp;
+    }
+
+    /**
+     * Like BasicSetting's assignment overload, except value is clamped to the range of the setting.
+     * Uses the appropriate value depending on the global state.
+     *
+     * @param value The desired value
+     * @returns A reference to the setting's value
+     */
+    const Type& operator=(const Type& value) override {
+        const Type temp = std::clamp(value, this->minimum, this->maximum);
+        if (this->use_global) {
+            this->global = temp;
+            return this->global;
+        }
+        this->custom = temp;
+        return this->custom;
+    }
+};
+
 /**
  * The InputSetting class allows for getting a reference to either the global or custom members.
  * This is required as we cannot easily modify the values of user-defined types within containers
@@ -284,13 +416,14 @@ struct Values {
     BasicSetting<std::string> sink_id{"auto", "output_engine"};
     BasicSetting<bool> audio_muted{false, "audio_muted"};
     Setting<bool> enable_audio_stretching{true, "enable_audio_stretching"};
-    Setting<u8> volume{100, "volume"};
+    RangedSetting<u8> volume{100, 0, 100, "volume"};
 
     // Core
     Setting<bool> use_multi_core{true, "use_multi_core"};
 
     // Cpu
-    Setting<CPUAccuracy> cpu_accuracy{CPUAccuracy::Auto, "cpu_accuracy"};
+    RangedSetting<CPUAccuracy> cpu_accuracy{CPUAccuracy::Auto, CPUAccuracy::Auto,
+                                            CPUAccuracy::Unsafe, "cpu_accuracy"};
     // TODO: remove cpu_accuracy_first_time, migration setting added 8 July 2021
     BasicSetting<bool> cpu_accuracy_first_time{true, "cpu_accuracy_first_time"};
     BasicSetting<bool> cpu_debug_mode{false, "cpu_debug_mode"};
@@ -312,8 +445,10 @@ struct Values {
     Setting<bool> cpuopt_unsafe_fastmem_check{true, "cpuopt_unsafe_fastmem_check"};
 
     // Renderer
-    Setting<RendererBackend> renderer_backend{RendererBackend::OpenGL, "backend"};
+    RangedSetting<RendererBackend> renderer_backend{
+        RendererBackend::OpenGL, RendererBackend::OpenGL, RendererBackend::Vulkan, "backend"};
     BasicSetting<bool> renderer_debug{false, "debug"};
+    BasicSetting<bool> renderer_shader_feedback{false, "shader_feedback"};
     BasicSetting<bool> enable_nsight_aftermath{false, "nsight_aftermath"};
     BasicSetting<bool> disable_shader_loop_safety_checks{false,
                                                          "disable_shader_loop_safety_checks"};
@@ -322,29 +457,30 @@ struct Values {
     Setting<u16> resolution_factor{1, "resolution_factor"};
     // *nix platforms may have issues with the borderless windowed fullscreen mode.
     // Default to exclusive fullscreen on these platforms for now.
-    Setting<int> fullscreen_mode{
+    RangedSetting<FullscreenMode> fullscreen_mode{
 #ifdef _WIN32
-        0,
+        FullscreenMode::Borderless,
 #else
-        1,
+        FullscreenMode::Exclusive,
 #endif
-        "fullscreen_mode"};
-    Setting<int> aspect_ratio{0, "aspect_ratio"};
-    Setting<int> max_anisotropy{0, "max_anisotropy"};
+        FullscreenMode::Borderless, FullscreenMode::Exclusive, "fullscreen_mode"};
+    RangedSetting<int> aspect_ratio{0, 0, 3, "aspect_ratio"};
+    RangedSetting<int> max_anisotropy{0, 0, 4, "max_anisotropy"};
     Setting<bool> use_speed_limit{true, "use_speed_limit"};
-    Setting<u16> speed_limit{100, "speed_limit"};
+    RangedSetting<u16> speed_limit{100, 0, 9999, "speed_limit"};
     Setting<bool> use_disk_shader_cache{true, "use_disk_shader_cache"};
-    Setting<GPUAccuracy> gpu_accuracy{GPUAccuracy::High, "gpu_accuracy"};
+    RangedSetting<GPUAccuracy> gpu_accuracy{GPUAccuracy::High, GPUAccuracy::Normal,
+                                            GPUAccuracy::Extreme, "gpu_accuracy"};
     Setting<bool> use_asynchronous_gpu_emulation{true, "use_asynchronous_gpu_emulation"};
-    Setting<bool> use_nvdec_emulation{true, "use_nvdec_emulation"};
+    Setting<NvdecEmulation> nvdec_emulation{NvdecEmulation::GPU, "nvdec_emulation"};
     Setting<bool> accelerate_astc{true, "accelerate_astc"};
     Setting<bool> use_vsync{true, "use_vsync"};
-    BasicSetting<u16> fps_cap{1000, "fps_cap"};
+    BasicRangedSetting<u16> fps_cap{1000, 1, 1000, "fps_cap"};
     BasicSetting<bool> disable_fps_limit{false, "disable_fps_limit"};
-    Setting<ShaderBackend> shader_backend{ShaderBackend::GLASM, "shader_backend"};
+    RangedSetting<ShaderBackend> shader_backend{ShaderBackend::GLASM, ShaderBackend::GLSL,
+                                                ShaderBackend::SPIRV, "shader_backend"};
     Setting<bool> use_asynchronous_shaders{false, "use_asynchronous_shaders"};
     Setting<bool> use_fast_gpu_time{true, "use_fast_gpu_time"};
-    Setting<bool> use_caches_gc{false, "use_caches_gc"};
 
     Setting<u8> bg_red{0, "bg_red"};
     Setting<u8> bg_green{0, "bg_green"};
@@ -358,10 +494,10 @@ struct Values {
     std::chrono::seconds custom_rtc_differential;
 
     BasicSetting<s32> current_user{0, "current_user"};
-    Setting<s32> language_index{1, "language_index"};
-    Setting<s32> region_index{1, "region_index"};
-    Setting<s32> time_zone_index{0, "time_zone_index"};
-    Setting<s32> sound_index{1, "sound_index"};
+    RangedSetting<s32> language_index{1, 0, 17, "language_index"};
+    RangedSetting<s32> region_index{1, 0, 6, "region_index"};
+    RangedSetting<s32> time_zone_index{0, 0, 45, "time_zone_index"};
+    RangedSetting<s32> sound_index{1, 0, 2, "sound_index"};
 
     // Controls
     InputSetting<std::array<PlayerInput, 10>> players;
@@ -378,7 +514,7 @@ struct Values {
                                                 "udp_input_servers"};
 
     BasicSetting<bool> mouse_panning{false, "mouse_panning"};
-    BasicSetting<u8> mouse_panning_sensitivity{10, "mouse_panning_sensitivity"};
+    BasicRangedSetting<u8> mouse_panning_sensitivity{10, 1, 100, "mouse_panning_sensitivity"};
     BasicSetting<bool> mouse_enabled{false, "mouse_enabled"};
     std::string mouse_device;
     MouseButtonsRaw mouse_buttons;
@@ -427,9 +563,10 @@ struct Values {
     BasicSetting<std::string> log_filter{"*:Info", "log_filter"};
     BasicSetting<bool> use_dev_keys{false, "use_dev_keys"};
 
-    // Services
+    // Network
     BasicSetting<std::string> bcat_backend{"none", "bcat_backend"};
     BasicSetting<bool> bcat_boxcat_local{false, "bcat_boxcat_local"};
+    BasicSetting<std::string> network_interface{std::string(), "network_interface"};
 
     // WebService
     BasicSetting<bool> enable_telemetry{true, "enable_telemetry"};

src/common/thread.cpp

@@ -2,7 +2,9 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "common/common_funcs.h"
+#include <string>
+
+#include "common/error.h"
 #include "common/logging/log.h"
 #include "common/thread.h"
 #ifdef __APPLE__
@@ -21,8 +23,6 @@
 #include <unistd.h>
 #endif
 
-#include <string>
-
 #ifdef __FreeBSD__
 #define cpu_set_t cpuset_t
 #endif

src/common/threadsafe_queue.h

@@ -14,7 +14,7 @@
 #include <utility>
 
 namespace Common {
-template <typename T>
+template <typename T, bool with_stop_token = false>
 class SPSCQueue {
 public:
     SPSCQueue() {
@@ -46,15 +46,13 @@ public:
         ElementPtr* new_ptr = new ElementPtr();
         write_ptr->next.store(new_ptr, std::memory_order_release);
         write_ptr = new_ptr;
+        ++size;
 
-        const size_t previous_size{size++};
-
-        // Acquire the mutex and then immediately release it as a fence.
+        // cv_mutex must be held or else there will be a missed wakeup if the other thread is in the
+        // line before cv.wait
         // TODO(bunnei): This can be replaced with C++20 waitable atomics when properly supported.
         // See discussion on https://github.com/yuzu-emu/yuzu/pull/3173 for details.
-        if (previous_size == 0) {
-            std::lock_guard lock{cv_mutex};
-        }
+        std::lock_guard lock{cv_mutex};
         cv.notify_one();
     }
 
@@ -86,7 +84,7 @@ public:
     void Wait() {
         if (Empty()) {
             std::unique_lock lock{cv_mutex};
-            cv.wait(lock, [this]() { return !Empty(); });
+            cv.wait(lock, [this] { return !Empty(); });
         }
     }
 
@@ -97,6 +95,19 @@ public:
         return t;
     }
 
+    T PopWait(std::stop_token stop_token) {
+        if (Empty()) {
+            std::unique_lock lock{cv_mutex};
+            cv.wait(lock, stop_token, [this] { return !Empty(); });
+        }
+        if (stop_token.stop_requested()) {
+            return T{};
+        }
+        T t;
+        Pop(t);
+        return t;
+    }
+
     // not thread-safe
     void Clear() {
         size.store(0);
@@ -125,13 +136,13 @@ private:
     ElementPtr* read_ptr;
     std::atomic_size_t size{0};
     std::mutex cv_mutex;
-    std::condition_variable cv;
+    std::conditional_t<with_stop_token, std::condition_variable_any, std::condition_variable> cv;
 };
 
 // a simple thread-safe,
 // single reader, multiple writer queue
 
-template <typename T>
+template <typename T, bool with_stop_token = false>
 class MPSCQueue {
 public:
     [[nodiscard]] std::size_t Size() const {
@@ -168,13 +179,17 @@ public:
         return spsc_queue.PopWait();
     }
 
+    T PopWait(std::stop_token stop_token) {
+        return spsc_queue.PopWait(stop_token);
+    }
+
     // not thread-safe
     void Clear() {
         spsc_queue.Clear();
     }
 
 private:
-    SPSCQueue<T> spsc_queue;
+    SPSCQueue<T, with_stop_token> spsc_queue;
     std::mutex write_lock;
 };
 } // namespace Common

src/common/uuid.cpp

@@ -6,10 +6,64 @@
 
 #include <fmt/format.h>
 
+#include "common/assert.h"
 #include "common/uuid.h"
 
 namespace Common {
 
+namespace {
+
+bool IsHexDigit(char c) {
+    return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F');
+}
+
+u8 HexCharToByte(char c) {
+    if (c >= '0' && c <= '9') {
+        return static_cast<u8>(c - '0');
+    }
+    if (c >= 'a' && c <= 'f') {
+        return static_cast<u8>(c - 'a' + 10);
+    }
+    if (c >= 'A' && c <= 'F') {
+        return static_cast<u8>(c - 'A' + 10);
+    }
+    ASSERT_MSG(false, "{} is not a hexadecimal digit!", c);
+    return u8{0};
+}
+
+} // Anonymous namespace
+
+u128 HexStringToU128(std::string_view hex_string) {
+    const size_t length = hex_string.length();
+
+    // Detect "0x" prefix.
+    const bool has_0x_prefix = length > 2 && hex_string[0] == '0' && hex_string[1] == 'x';
+    const size_t offset = has_0x_prefix ? 2 : 0;
+
+    // Check length.
+    if (length > 32 + offset) {
+        ASSERT_MSG(false, "hex_string has more than 32 hexadecimal characters!");
+        return INVALID_UUID;
+    }
+
+    u64 lo = 0;
+    u64 hi = 0;
+    for (size_t i = 0; i < length - offset; ++i) {
+        const char c = hex_string[length - 1 - i];
+        if (!IsHexDigit(c)) {
+            ASSERT_MSG(false, "{} is not a hexadecimal digit!", c);
+            return INVALID_UUID;
+        }
+        if (i < 16) {
+            lo |= u64{HexCharToByte(c)} << (i * 4);
+        }
+        if (i >= 16) {
+            hi |= u64{HexCharToByte(c)} << ((i - 16) * 4);
+        }
+    }
+    return u128{lo, hi};
+}
+
 UUID UUID::Generate() {
     std::random_device device;
     std::mt19937 gen(device());
@@ -18,7 +72,7 @@ UUID UUID::Generate() {
 }
 
 std::string UUID::Format() const {
-    return fmt::format("0x{:016X}{:016X}", uuid[1], uuid[0]);
+    return fmt::format("{:016x}{:016x}", uuid[1], uuid[0]);
 }
 
 std::string UUID::FormatSwitch() const {

src/common/uuid.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <string>
+#include <string_view>
 
 #include "common/common_types.h"
 
@@ -12,12 +13,30 @@ namespace Common {
 
 constexpr u128 INVALID_UUID{{0, 0}};
 
+/**
+ * Converts a hex string to a 128-bit unsigned integer.
+ *
+ * The hex string can be formatted in lowercase or uppercase, with or without the "0x" prefix.
+ *
+ * This function will assert and return INVALID_UUID under the following conditions:
+ * - If the hex string is more than 32 characters long
+ * - If the hex string contains non-hexadecimal characters
+ *
+ * @param hex_string Hexadecimal string
+ *
+ * @returns A 128-bit unsigned integer if successfully converted, INVALID_UUID otherwise.
+ */
+[[nodiscard]] u128 HexStringToU128(std::string_view hex_string);
+
 struct UUID {
     // UUIDs which are 0 are considered invalid!
     u128 uuid;
     UUID() = default;
     constexpr explicit UUID(const u128& id) : uuid{id} {}
     constexpr explicit UUID(const u64 lo, const u64 hi) : uuid{{lo, hi}} {}
+    explicit UUID(std::string_view hex_string) {
+        uuid = HexStringToU128(hex_string);
+    }
 
     [[nodiscard]] constexpr explicit operator bool() const {
         return uuid != INVALID_UUID;
@@ -50,3 +69,14 @@ struct UUID {
 static_assert(sizeof(UUID) == 16, "UUID is an invalid size!");
 
 } // namespace Common
+
+namespace std {
+
+template <>
+struct hash<Common::UUID> {
+    size_t operator()(const Common::UUID& uuid) const noexcept {
+        return uuid.uuid[1] ^ uuid.uuid[0];
+    }
+};
+
+} // namespace std

src/common/x64/xbyak_abi.h

@@ -6,7 +6,7 @@
 
 #include <bitset>
 #include <initializer_list>
-#include <xbyak.h>
+#include <xbyak/xbyak.h>
 #include "common/assert.h"
 
 namespace Common::X64 {

src/common/x64/xbyak_util.h

@@ -5,7 +5,7 @@
 #pragma once
 
 #include <type_traits>
-#include <xbyak.h>
+#include <xbyak/xbyak.h>
 #include "common/x64/xbyak_abi.h"
 
 namespace Common::X64 {

src/core/CMakeLists.txt

@@ -263,6 +263,8 @@ add_library(core STATIC
     hle/service/acc/acc_u0.h
     hle/service/acc/acc_u1.cpp
     hle/service/acc/acc_u1.h
+    hle/service/acc/async_context.cpp
+    hle/service/acc/async_context.h
     hle/service/acc/errors.h
     hle/service/acc/profile_manager.cpp
     hle/service/acc/profile_manager.h
@@ -452,6 +454,8 @@ add_library(core STATIC
     hle/service/nfp/nfp.h
     hle/service/nfp/nfp_user.cpp
     hle/service/nfp/nfp_user.h
+    hle/service/ngct/ngct.cpp
+    hle/service/ngct/ngct.h
     hle/service/nifm/nifm.cpp
     hle/service/nifm/nifm.h
     hle/service/nim/nim.cpp
@@ -636,6 +640,8 @@ add_library(core STATIC
     memory.h
     network/network.cpp
     network/network.h
+    network/network_interface.cpp
+    network/network_interface.h
     network/sockets.h
     perf_stats.cpp
     perf_stats.h

src/core/core.cpp

@@ -4,6 +4,7 @@
 
 #include <array>
 #include <atomic>
+#include <exception>
 #include <memory>
 #include <utility>
 
@@ -82,9 +83,13 @@ FileSys::StorageId GetStorageIdForFrontendSlot(
     }
 }
 
-} // Anonymous namespace
+void KProcessDeleter(Kernel::KProcess* process) {
+    process->Destroy();
+}
 
-/*static*/ System System::s_instance;
+using KProcessPtr = std::unique_ptr<Kernel::KProcess, decltype(&KProcessDeleter)>;
+
+} // Anonymous namespace
 
 FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
                                          const std::string& path) {
@@ -234,8 +239,8 @@ struct System::Impl {
         }
 
         telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider);
-        auto main_process = Kernel::KProcess::Create(system.Kernel());
-        ASSERT(Kernel::KProcess::Initialize(main_process, system, "main",
+        main_process = KProcessPtr{Kernel::KProcess::Create(system.Kernel()), KProcessDeleter};
+        ASSERT(Kernel::KProcess::Initialize(main_process.get(), system, "main",
                                             Kernel::KProcess::ProcessType::Userland)
                    .IsSuccess());
         main_process->Open();
@@ -248,7 +253,7 @@ struct System::Impl {
                                              static_cast<u32>(load_result));
         }
         AddGlueRegistrationForProcess(*app_loader, *main_process);
-        kernel.MakeCurrentProcess(main_process);
+        kernel.MakeCurrentProcess(main_process.get());
         kernel.InitializeCores();
 
         // Initialize cheat engine
@@ -300,10 +305,7 @@ struct System::Impl {
         is_powered_on = false;
         exit_lock = false;
 
-        if (gpu_core) {
-            gpu_core->ShutDown();
-        }
-
+        gpu_core.reset();
         services.reset();
         service_manager.reset();
         cheat_engine.reset();
@@ -312,11 +314,12 @@ struct System::Impl {
         time_manager.Shutdown();
         core_timing.Shutdown();
         app_loader.reset();
-        gpu_core.reset();
         perf_stats.reset();
         kernel.Shutdown();
         memory.Reset();
         applet_manager.ClearAll();
+        // TODO: The main process should be freed based on KAutoObject ref counting.
+        main_process.reset();
 
         LOG_DEBUG(Core, "Shutdown OK");
     }
@@ -375,6 +378,7 @@ struct System::Impl {
     std::unique_ptr<Tegra::GPU> gpu_core;
     std::unique_ptr<Hardware::InterruptManager> interrupt_manager;
     std::unique_ptr<Core::DeviceMemory> device_memory;
+    KProcessPtr main_process{nullptr, KProcessDeleter};
     Core::Memory::Memory memory;
     CpuManager cpu_manager;
     std::atomic_bool is_powered_on{};
@@ -425,6 +429,20 @@ struct System::Impl {
 System::System() : impl{std::make_unique<Impl>(*this)} {}
 System::~System() = default;
 
+System& System::GetInstance() {
+    if (!s_instance) {
+        throw std::runtime_error("Using System instance before its initialization");
+    }
+    return *s_instance;
+}
+
+void System::InitializeGlobalInstance() {
+    if (s_instance) {
+        throw std::runtime_error("Reinitializing Global System instance.");
+    }
+    s_instance = std::unique_ptr<System>(new System);
+}
+
 CpuManager& System::GetCpuManager() {
     return impl->cpu_manager;
 }

src/core/core.h

@@ -120,9 +120,9 @@ public:
      * Gets the instance of the System singleton class.
      * @returns Reference to the instance of the System singleton class.
      */
-    [[deprecated("Use of the global system instance is deprecated")]] static System& GetInstance() {
-        return s_instance;
-    }
+    [[deprecated("Use of the global system instance is deprecated")]] static System& GetInstance();
+
+    static void InitializeGlobalInstance();
 
     /// Enumeration representing the return values of the System Initialize and Load process.
     enum class ResultStatus : u32 {
@@ -396,7 +396,7 @@ private:
     struct Impl;
     std::unique_ptr<Impl> impl;
 
-    static System s_instance;
+    inline static std::unique_ptr<System> s_instance{};
 };
 
 } // namespace Core

src/core/cpu_manager.cpp

@@ -21,34 +21,25 @@ namespace Core {
 CpuManager::CpuManager(System& system_) : system{system_} {}
 CpuManager::~CpuManager() = default;
 
-void CpuManager::ThreadStart(CpuManager& cpu_manager, std::size_t core) {
-    cpu_manager.RunThread(core);
+void CpuManager::ThreadStart(std::stop_token stop_token, CpuManager& cpu_manager,
+                             std::size_t core) {
+    cpu_manager.RunThread(stop_token, core);
 }
 
 void CpuManager::Initialize() {
     running_mode = true;
     if (is_multicore) {
         for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-            core_data[core].host_thread =
-                std::make_unique<std::thread>(ThreadStart, std::ref(*this), core);
+            core_data[core].host_thread = std::jthread(ThreadStart, std::ref(*this), core);
         }
     } else {
-        core_data[0].host_thread = std::make_unique<std::thread>(ThreadStart, std::ref(*this), 0);
+        core_data[0].host_thread = std::jthread(ThreadStart, std::ref(*this), 0);
     }
 }
 
 void CpuManager::Shutdown() {
     running_mode = false;
     Pause(false);
-    if (is_multicore) {
-        for (auto& data : core_data) {
-            data.host_thread->join();
-            data.host_thread.reset();
-        }
-    } else {
-        core_data[0].host_thread->join();
-        core_data[0].host_thread.reset();
-    }
 }
 
 std::function<void(void*)> CpuManager::GetGuestThreadStartFunc() {
@@ -317,7 +308,7 @@ void CpuManager::Pause(bool paused) {
     }
 }
 
-void CpuManager::RunThread(std::size_t core) {
+void CpuManager::RunThread(std::stop_token stop_token, std::size_t core) {
     /// Initialization
     system.RegisterCoreThread(core);
     std::string name;
@@ -361,6 +352,10 @@ void CpuManager::RunThread(std::size_t core) {
             return;
         }
 
+        if (stop_token.stop_requested()) {
+            break;
+        }
+
         auto current_thread = system.Kernel().CurrentScheduler()->GetCurrentThread();
         data.is_running = true;
         Common::Fiber::YieldTo(data.host_context, *current_thread->GetHostContext());

src/core/cpu_manager.h

@@ -78,9 +78,9 @@ private:
     void SingleCoreRunSuspendThread();
     void SingleCorePause(bool paused);
 
-    static void ThreadStart(CpuManager& cpu_manager, std::size_t core);
+    static void ThreadStart(std::stop_token stop_token, CpuManager& cpu_manager, std::size_t core);
 
-    void RunThread(std::size_t core);
+    void RunThread(std::stop_token stop_token, std::size_t core);
 
     struct CoreData {
         std::shared_ptr<Common::Fiber> host_context;
@@ -89,7 +89,7 @@ private:
         std::atomic<bool> is_running;
         std::atomic<bool> is_paused;
         std::atomic<bool> initialized;
-        std::unique_ptr<std::thread> host_thread;
+        std::jthread host_thread;
     };
 
     std::atomic<bool> running_mode{};

src/core/file_sys/kernel_executable.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <string>
 #include <vector>
 
 #include "common/common_funcs.h"

src/core/hle/api_version.h

@@ -12,9 +12,9 @@ namespace HLE::ApiVersion {
 
 // Horizon OS version constants.
 
-constexpr u8 HOS_VERSION_MAJOR = 11;
-constexpr u8 HOS_VERSION_MINOR = 0;
-constexpr u8 HOS_VERSION_MICRO = 1;
+constexpr u8 HOS_VERSION_MAJOR = 12;
+constexpr u8 HOS_VERSION_MINOR = 1;
+constexpr u8 HOS_VERSION_MICRO = 0;
 
 // NintendoSDK version constants.
 
@@ -22,19 +22,26 @@ constexpr u8 SDK_REVISION_MAJOR = 1;
 constexpr u8 SDK_REVISION_MINOR = 0;
 
 constexpr char PLATFORM_STRING[] = "NX";
-constexpr char VERSION_HASH[] = "69103fcb2004dace877094c2f8c29e6113be5dbf";
-constexpr char DISPLAY_VERSION[] = "11.0.1";
-constexpr char DISPLAY_TITLE[] = "NintendoSDK Firmware for NX 11.0.1-1.0";
+constexpr char VERSION_HASH[] = "76b10c2dab7d3aa73fc162f8dff1655e6a21caf4";
+constexpr char DISPLAY_VERSION[] = "12.1.0";
+constexpr char DISPLAY_TITLE[] = "NintendoSDK Firmware for NX 12.1.0-1.0";
 
 // Atmosphere version constants.
 
-constexpr u8 ATMOSPHERE_RELEASE_VERSION_MAJOR = 0;
-constexpr u8 ATMOSPHERE_RELEASE_VERSION_MINOR = 19;
-constexpr u8 ATMOSPHERE_RELEASE_VERSION_MICRO = 4;
+constexpr u8 ATMOSPHERE_RELEASE_VERSION_MAJOR = 1;
+constexpr u8 ATMOSPHERE_RELEASE_VERSION_MINOR = 0;
+constexpr u8 ATMOSPHERE_RELEASE_VERSION_MICRO = 0;
+
+constexpr u32 AtmosphereTargetFirmwareWithRevision(u8 major, u8 minor, u8 micro, u8 rev) {
+    return u32{major} << 24 | u32{minor} << 16 | u32{micro} << 8 | u32{rev};
+}
+
+constexpr u32 AtmosphereTargetFirmware(u8 major, u8 minor, u8 micro) {
+    return AtmosphereTargetFirmwareWithRevision(major, minor, micro, 0);
+}
 
 constexpr u32 GetTargetFirmware() {
-    return u32{HOS_VERSION_MAJOR} << 24 | u32{HOS_VERSION_MINOR} << 16 |
-           u32{HOS_VERSION_MICRO} << 8 | 0U;
+    return AtmosphereTargetFirmware(HOS_VERSION_MAJOR, HOS_VERSION_MINOR, HOS_VERSION_MICRO);
 }
 
 } // namespace HLE::ApiVersion

src/core/hle/kernel/kernel.cpp

@@ -267,20 +267,23 @@ struct KernelCore::Impl {
         }
     }
 
-    /// Creates a new host thread ID, should only be called by GetHostThreadId
-    u32 AllocateHostThreadId(std::optional<std::size_t> core_id) {
-        if (core_id) {
-            // The first for slots are reserved for CPU core threads
-            ASSERT(*core_id < Core::Hardware::NUM_CPU_CORES);
-            return static_cast<u32>(*core_id);
-        } else {
-            return next_host_thread_id++;
+    static inline thread_local u32 host_thread_id = UINT32_MAX;
+
+    /// Gets the host thread ID for the caller, allocating a new one if this is the first time
+    u32 GetHostThreadId(std::size_t core_id) {
+        if (host_thread_id == UINT32_MAX) {
+            // The first four slots are reserved for CPU core threads
+            ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+            host_thread_id = static_cast<u32>(core_id);
         }
+        return host_thread_id;
     }
 
     /// Gets the host thread ID for the caller, allocating a new one if this is the first time
-    u32 GetHostThreadId(std::optional<std::size_t> core_id = std::nullopt) {
-        const thread_local auto host_thread_id{AllocateHostThreadId(core_id)};
+    u32 GetHostThreadId() {
+        if (host_thread_id == UINT32_MAX) {
+            host_thread_id = next_host_thread_id++;
+        }
         return host_thread_id;
     }
 

src/core/hle/kernel/kernel.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <functional>
 #include <memory>
 #include <string>
 #include <unordered_map>

src/core/hle/kernel/svc.cpp

@@ -1078,8 +1078,8 @@ static ResultCode GetThreadContext(Core::System& system, VAddr out_context, Hand
             for (auto i = 0; i < static_cast<s32>(Core::Hardware::NUM_CPU_CORES); ++i) {
                 if (thread.GetPointerUnsafe() == kernel.Scheduler(i).GetCurrentThread()) {
                     current = true;
+                    break;
                 }
-                break;
             }
 
             // If the thread is current, retry until it isn't.

src/core/hle/service/acc/acc.cpp

@@ -23,6 +23,7 @@
 #include "core/hle/service/acc/acc_su.h"
 #include "core/hle/service/acc/acc_u0.h"
 #include "core/hle/service/acc/acc_u1.h"
+#include "core/hle/service/acc/async_context.h"
 #include "core/hle/service/acc/errors.h"
 #include "core/hle/service/acc/profile_manager.h"
 #include "core/hle/service/glue/arp.h"
@@ -292,7 +293,7 @@ public:
 
 protected:
     void Get(Kernel::HLERequestContext& ctx) {
-        LOG_DEBUG(Service_ACC, "called user_id={}", user_id.Format());
+        LOG_DEBUG(Service_ACC, "called user_id=0x{}", user_id.Format());
         ProfileBase profile_base{};
         ProfileData data{};
         if (profile_manager.GetProfileBaseAndData(user_id, profile_base, data)) {
@@ -301,7 +302,7 @@ protected:
             rb.Push(ResultSuccess);
             rb.PushRaw(profile_base);
         } else {
-            LOG_ERROR(Service_ACC, "Failed to get profile base and data for user={}",
+            LOG_ERROR(Service_ACC, "Failed to get profile base and data for user=0x{}",
                       user_id.Format());
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ResultUnknown); // TODO(ogniK): Get actual error code
@@ -309,14 +310,14 @@ protected:
     }
 
     void GetBase(Kernel::HLERequestContext& ctx) {
-        LOG_DEBUG(Service_ACC, "called user_id={}", user_id.Format());
+        LOG_DEBUG(Service_ACC, "called user_id=0x{}", user_id.Format());
         ProfileBase profile_base{};
         if (profile_manager.GetProfileBase(user_id, profile_base)) {
             IPC::ResponseBuilder rb{ctx, 16};
             rb.Push(ResultSuccess);
             rb.PushRaw(profile_base);
         } else {
-            LOG_ERROR(Service_ACC, "Failed to get profile base for user={}", user_id.Format());
+            LOG_ERROR(Service_ACC, "Failed to get profile base for user=0x{}", user_id.Format());
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ResultUnknown); // TODO(ogniK): Get actual error code
         }
@@ -372,7 +373,7 @@ protected:
 
         const auto user_data = ctx.ReadBuffer();
 
-        LOG_DEBUG(Service_ACC, "called, username='{}', timestamp={:016X}, uuid={}",
+        LOG_DEBUG(Service_ACC, "called, username='{}', timestamp={:016X}, uuid=0x{}",
                   Common::StringFromFixedZeroTerminatedBuffer(
                       reinterpret_cast<const char*>(base.username.data()), base.username.size()),
                   base.timestamp, base.user_uuid.Format());
@@ -405,7 +406,7 @@ protected:
         const auto user_data = ctx.ReadBuffer();
         const auto image_data = ctx.ReadBuffer(1);
 
-        LOG_DEBUG(Service_ACC, "called, username='{}', timestamp={:016X}, uuid={}",
+        LOG_DEBUG(Service_ACC, "called, username='{}', timestamp={:016X}, uuid=0x{}",
                   Common::StringFromFixedZeroTerminatedBuffer(
                       reinterpret_cast<const char*>(base.username.data()), base.username.size()),
                   base.timestamp, base.user_uuid.Format());
@@ -454,22 +455,6 @@ public:
         : IProfileCommon{system_, "IProfileEditor", true, user_id_, profile_manager_} {}
 };
 
-class IAsyncContext final : public ServiceFramework<IAsyncContext> {
-public:
-    explicit IAsyncContext(Core::System& system_) : ServiceFramework{system_, "IAsyncContext"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {0, nullptr, "GetSystemEvent"},
-            {1, nullptr, "Cancel"},
-            {2, nullptr, "HasDone"},
-            {3, nullptr, "GetResult"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
 class ISessionObject final : public ServiceFramework<ISessionObject> {
 public:
     explicit ISessionObject(Core::System& system_, Common::UUID)
@@ -504,16 +489,44 @@ public:
     }
 };
 
+class EnsureTokenIdCacheAsyncInterface final : public IAsyncContext {
+public:
+    explicit EnsureTokenIdCacheAsyncInterface(Core::System& system_) : IAsyncContext{system_} {
+        MarkComplete();
+    }
+    ~EnsureTokenIdCacheAsyncInterface() = default;
+
+    void LoadIdTokenCache(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_ACC, "(STUBBED) called");
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+
+protected:
+    bool IsComplete() const override {
+        return true;
+    }
+
+    void Cancel() override {}
+
+    ResultCode GetResult() const override {
+        return ResultSuccess;
+    }
+};
+
 class IManagerForApplication final : public ServiceFramework<IManagerForApplication> {
 public:
     explicit IManagerForApplication(Core::System& system_, Common::UUID user_id_)
-        : ServiceFramework{system_, "IManagerForApplication"}, user_id{user_id_} {
+        : ServiceFramework{system_, "IManagerForApplication"},
+          ensure_token_id{std::make_shared<EnsureTokenIdCacheAsyncInterface>(system)},
+          user_id{user_id_} {
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, &IManagerForApplication::CheckAvailability, "CheckAvailability"},
             {1, &IManagerForApplication::GetAccountId, "GetAccountId"},
-            {2, nullptr, "EnsureIdTokenCacheAsync"},
-            {3, nullptr, "LoadIdTokenCache"},
+            {2, &IManagerForApplication::EnsureIdTokenCacheAsync, "EnsureIdTokenCacheAsync"},
+            {3, &IManagerForApplication::LoadIdTokenCache, "LoadIdTokenCache"},
             {130, &IManagerForApplication::GetNintendoAccountUserResourceCacheForApplication, "GetNintendoAccountUserResourceCacheForApplication"},
             {150, nullptr, "CreateAuthorizationRequest"},
             {160, &IManagerForApplication::StoreOpenContext, "StoreOpenContext"},
@@ -540,6 +553,20 @@ private:
         rb.PushRaw<u64>(user_id.GetNintendoID());
     }
 
+    void EnsureIdTokenCacheAsync(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_ACC, "(STUBBED) called");
+
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+        rb.Push(ResultSuccess);
+        rb.PushIpcInterface(ensure_token_id);
+    }
+
+    void LoadIdTokenCache(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_ACC, "(STUBBED) called");
+
+        ensure_token_id->LoadIdTokenCache(ctx);
+    }
+
     void GetNintendoAccountUserResourceCacheForApplication(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_ACC, "(STUBBED) called");
 
@@ -562,6 +589,7 @@ private:
         rb.Push(ResultSuccess);
     }
 
+    std::shared_ptr<EnsureTokenIdCacheAsyncInterface> ensure_token_id{};
     Common::UUID user_id{Common::INVALID_UUID};
 };
 
@@ -662,7 +690,7 @@ void Module::Interface::GetUserCount(Kernel::HLERequestContext& ctx) {
 void Module::Interface::GetUserExistence(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     Common::UUID user_id = rp.PopRaw<Common::UUID>();
-    LOG_DEBUG(Service_ACC, "called user_id={}", user_id.Format());
+    LOG_DEBUG(Service_ACC, "called user_id=0x{}", user_id.Format());
 
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(ResultSuccess);
@@ -693,7 +721,7 @@ void Module::Interface::GetLastOpenedUser(Kernel::HLERequestContext& ctx) {
 void Module::Interface::GetProfile(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     Common::UUID user_id = rp.PopRaw<Common::UUID>();
-    LOG_DEBUG(Service_ACC, "called user_id={}", user_id.Format());
+    LOG_DEBUG(Service_ACC, "called user_id=0x{}", user_id.Format());
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(ResultSuccess);
@@ -802,7 +830,7 @@ void Module::Interface::GetProfileEditor(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     Common::UUID user_id = rp.PopRaw<Common::UUID>();
 
-    LOG_DEBUG(Service_ACC, "called, user_id={}", user_id.Format());
+    LOG_DEBUG(Service_ACC, "called, user_id=0x{}", user_id.Format());
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(ResultSuccess);
@@ -844,7 +872,7 @@ void Module::Interface::StoreSaveDataThumbnailApplication(Kernel::HLERequestCont
     IPC::RequestParser rp{ctx};
     const auto uuid = rp.PopRaw<Common::UUID>();
 
-    LOG_WARNING(Service_ACC, "(STUBBED) called, uuid={}", uuid.Format());
+    LOG_WARNING(Service_ACC, "(STUBBED) called, uuid=0x{}", uuid.Format());
 
     // TODO(ogniK): Check if application ID is zero on acc initialize. As we don't have a reliable
     // way of confirming things like the TID, we're going to assume a non zero value for the time
@@ -858,7 +886,7 @@ void Module::Interface::StoreSaveDataThumbnailSystem(Kernel::HLERequestContext&
     const auto uuid = rp.PopRaw<Common::UUID>();
     const auto tid = rp.Pop<u64_le>();
 
-    LOG_WARNING(Service_ACC, "(STUBBED) called, uuid={}, tid={:016X}", uuid.Format(), tid);
+    LOG_WARNING(Service_ACC, "(STUBBED) called, uuid=0x{}, tid={:016X}", uuid.Format(), tid);
     StoreSaveDataThumbnail(ctx, uuid, tid);
 }
 

src/core/hle/service/acc/async_context.cpp

@@ -0,0 +1,68 @@
+// Copyright 2021 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/core.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/service/acc/async_context.h"
+
+namespace Service::Account {
+IAsyncContext::IAsyncContext(Core::System& system_)
+    : ServiceFramework{system_, "IAsyncContext"}, compeletion_event{system_.Kernel()} {
+
+    Kernel::KAutoObject::Create(std::addressof(compeletion_event));
+    compeletion_event.Initialize("IAsyncContext:CompletionEvent");
+
+    // clang-format off
+    static const FunctionInfo functions[] = {
+        {0, &IAsyncContext::GetSystemEvent, "GetSystemEvent"},
+        {1, &IAsyncContext::Cancel, "Cancel"},
+        {2, &IAsyncContext::HasDone, "HasDone"},
+        {3, &IAsyncContext::GetResult, "GetResult"},
+    };
+    // clang-format on
+
+    RegisterHandlers(functions);
+}
+
+void IAsyncContext::GetSystemEvent(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_ACC, "called");
+
+    IPC::ResponseBuilder rb{ctx, 2, 1};
+    rb.Push(ResultSuccess);
+    rb.PushCopyObjects(compeletion_event.GetReadableEvent());
+}
+
+void IAsyncContext::Cancel(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_ACC, "called");
+
+    Cancel();
+    MarkComplete();
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
+void IAsyncContext::HasDone(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_ACC, "called");
+
+    is_complete.store(IsComplete());
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(ResultSuccess);
+    rb.Push(is_complete.load());
+}
+
+void IAsyncContext::GetResult(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_ACC, "called");
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(GetResult());
+}
+
+void IAsyncContext::MarkComplete() {
+    is_complete.store(true);
+    compeletion_event.GetWritableEvent().Signal();
+}
+
+} // namespace Service::Account

src/core/hle/service/acc/async_context.h

@@ -0,0 +1,37 @@
+// Copyright 2021 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+#include "core/hle/kernel/k_event.h"
+#include "core/hle/service/service.h"
+
+namespace Core {
+class System;
+}
+
+namespace Service::Account {
+
+class IAsyncContext : public ServiceFramework<IAsyncContext> {
+public:
+    explicit IAsyncContext(Core::System& system_);
+
+    void GetSystemEvent(Kernel::HLERequestContext& ctx);
+    void Cancel(Kernel::HLERequestContext& ctx);
+    void HasDone(Kernel::HLERequestContext& ctx);
+    void GetResult(Kernel::HLERequestContext& ctx);
+
+protected:
+    virtual bool IsComplete() const = 0;
+    virtual void Cancel() = 0;
+    virtual ResultCode GetResult() const = 0;
+
+    void MarkComplete();
+
+    std::atomic<bool> is_complete{false};
+    Kernel::KEvent compeletion_event;
+};
+
+} // namespace Service::Account

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

@@ -1270,7 +1270,8 @@ void ILibraryAppletCreator::CreateHandleStorage(Kernel::HLERequestContext& ctx)
 IApplicationFunctions::IApplicationFunctions(Core::System& system_)
     : ServiceFramework{system_, "IApplicationFunctions"}, gpu_error_detected_event{system.Kernel()},
       friend_invitation_storage_channel_event{system.Kernel()},
-      health_warning_disappeared_system_event{system.Kernel()} {
+      notification_storage_channel_event{system.Kernel()}, health_warning_disappeared_system_event{
+                                                               system.Kernel()} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {1, &IApplicationFunctions::PopLaunchParameter, "PopLaunchParameter"},
@@ -1322,7 +1323,7 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
         {131, nullptr, "SetDelayTimeToAbortOnGpuError"},
         {140, &IApplicationFunctions::GetFriendInvitationStorageChannelEvent, "GetFriendInvitationStorageChannelEvent"},
         {141, &IApplicationFunctions::TryPopFromFriendInvitationStorageChannel, "TryPopFromFriendInvitationStorageChannel"},
-        {150, nullptr, "GetNotificationStorageChannelEvent"},
+        {150, &IApplicationFunctions::GetNotificationStorageChannelEvent, "GetNotificationStorageChannelEvent"},
         {151, nullptr, "TryPopFromNotificationStorageChannel"},
         {160, &IApplicationFunctions::GetHealthWarningDisappearedSystemEvent, "GetHealthWarningDisappearedSystemEvent"},
         {170, nullptr, "SetHdcpAuthenticationActivated"},
@@ -1340,11 +1341,14 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
 
     Kernel::KAutoObject::Create(std::addressof(gpu_error_detected_event));
     Kernel::KAutoObject::Create(std::addressof(friend_invitation_storage_channel_event));
+    Kernel::KAutoObject::Create(std::addressof(notification_storage_channel_event));
     Kernel::KAutoObject::Create(std::addressof(health_warning_disappeared_system_event));
 
     gpu_error_detected_event.Initialize("IApplicationFunctions:GpuErrorDetectedSystemEvent");
     friend_invitation_storage_channel_event.Initialize(
         "IApplicationFunctions:FriendInvitationStorageChannelEvent");
+    notification_storage_channel_event.Initialize(
+        "IApplicationFunctions:NotificationStorageChannelEvent");
     health_warning_disappeared_system_event.Initialize(
         "IApplicationFunctions:HealthWarningDisappearedSystemEvent");
 }
@@ -1762,6 +1766,14 @@ void IApplicationFunctions::TryPopFromFriendInvitationStorageChannel(
     rb.Push(ERR_NO_DATA_IN_CHANNEL);
 }
 
+void IApplicationFunctions::GetNotificationStorageChannelEvent(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_AM, "called");
+
+    IPC::ResponseBuilder rb{ctx, 2, 1};
+    rb.Push(ResultSuccess);
+    rb.PushCopyObjects(notification_storage_channel_event.GetReadableEvent());
+}
+
 void IApplicationFunctions::GetHealthWarningDisappearedSystemEvent(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called");
 

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

@@ -295,6 +295,7 @@ private:
     void GetGpuErrorDetectedSystemEvent(Kernel::HLERequestContext& ctx);
     void GetFriendInvitationStorageChannelEvent(Kernel::HLERequestContext& ctx);
     void TryPopFromFriendInvitationStorageChannel(Kernel::HLERequestContext& ctx);
+    void GetNotificationStorageChannelEvent(Kernel::HLERequestContext& ctx);
     void GetHealthWarningDisappearedSystemEvent(Kernel::HLERequestContext& ctx);
 
     bool launch_popped_application_specific = false;
@@ -302,6 +303,7 @@ private:
     s32 previous_program_index{-1};
     Kernel::KEvent gpu_error_detected_event;
     Kernel::KEvent friend_invitation_storage_channel_event;
+    Kernel::KEvent notification_storage_channel_event;
     Kernel::KEvent health_warning_disappeared_system_event;
 };
 

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

@@ -16,6 +16,30 @@
 
 namespace Service::AM::Applets {
 
+struct ErrorCode {
+    u32 error_category{};
+    u32 error_number{};
+
+    static constexpr ErrorCode FromU64(u64 error_code) {
+        return {
+            .error_category{static_cast<u32>(error_code >> 32)},
+            .error_number{static_cast<u32>(error_code & 0xFFFFFFFF)},
+        };
+    }
+
+    static constexpr ErrorCode FromResultCode(ResultCode result) {
+        return {
+            .error_category{2000 + static_cast<u32>(result.module.Value())},
+            .error_number{result.description.Value()},
+        };
+    }
+
+    constexpr ResultCode ToResultCode() const {
+        return ResultCode{static_cast<ErrorModule>(error_category - 2000), error_number};
+    }
+};
+static_assert(sizeof(ErrorCode) == 0x8, "ErrorCode has incorrect size.");
+
 #pragma pack(push, 4)
 struct ShowError {
     u8 mode;
@@ -76,12 +100,7 @@ void CopyArgumentData(const std::vector<u8>& data, T& variable) {
 }
 
 ResultCode Decode64BitError(u64 error) {
-    const auto description = (error >> 32) & 0x1FFF;
-    auto module = error & 0x3FF;
-    if (module >= 2000)
-        module -= 2000;
-    module &= 0x1FF;
-    return {static_cast<ErrorModule>(module), static_cast<u32>(description)};
+    return ErrorCode::FromU64(error).ToResultCode();
 }
 
 } // Anonymous namespace

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

@@ -377,7 +377,8 @@ void SoftwareKeyboard::SubmitForTextCheck(std::u16string submitted_text) {
 
     if (swkbd_config_common.use_utf8) {
         std::string utf8_submitted_text = Common::UTF16ToUTF8(current_text);
-        const u64 buffer_size = sizeof(u64) + utf8_submitted_text.size();
+        // Include the null terminator in the buffer size.
+        const u64 buffer_size = utf8_submitted_text.size() + 1;
 
         LOG_DEBUG(Service_AM, "\nBuffer Size: {}\nUTF-8 Submitted Text: {}", buffer_size,
                   utf8_submitted_text);
@@ -386,7 +387,8 @@ void SoftwareKeyboard::SubmitForTextCheck(std::u16string submitted_text) {
         std::memcpy(out_data.data() + sizeof(u64), utf8_submitted_text.data(),
                     utf8_submitted_text.size());
     } else {
-        const u64 buffer_size = sizeof(u64) + current_text.size() * sizeof(char16_t);
+        // Include the null terminator in the buffer size.
+        const u64 buffer_size = (current_text.size() + 1) * sizeof(char16_t);
 
         LOG_DEBUG(Service_AM, "\nBuffer Size: {}\nUTF-16 Submitted Text: {}", buffer_size,
                   Common::UTF16ToUTF8(current_text));

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

@@ -97,14 +97,24 @@ ResultCode VfsDirectoryServiceWrapper::DeleteFile(const std::string& path_) cons
 
 ResultCode VfsDirectoryServiceWrapper::CreateDirectory(const std::string& path_) const {
     std::string path(Common::FS::SanitizePath(path_));
-    auto dir = GetDirectoryRelativeWrapped(backing, Common::FS::GetParentPath(path));
-    if (dir == nullptr || Common::FS::GetFilename(Common::FS::GetParentPath(path)).empty()) {
-        dir = backing;
-    }
-    auto new_dir = dir->CreateSubdirectory(Common::FS::GetFilename(path));
-    if (new_dir == nullptr) {
-        // TODO(DarkLordZach): Find a better error code for this
-        return ResultUnknown;
+
+    // NOTE: This is inaccurate behavior. CreateDirectory is not recursive.
+    // CreateDirectory should return PathNotFound if the parent directory does not exist.
+    // This is here temporarily in order to have UMM "work" in the meantime.
+    // TODO (Morph): Remove this when a hardware test verifies the correct behavior.
+    const auto components = Common::FS::SplitPathComponents(path);
+    std::string relative_path;
+    for (const auto& component : components) {
+        // Skip empty path components
+        if (component.empty()) {
+            continue;
+        }
+        relative_path = Common::FS::SanitizePath(relative_path + '/' + component);
+        auto new_dir = backing->CreateSubdirectory(relative_path);
+        if (new_dir == nullptr) {
+            // TODO(DarkLordZach): Find a better error code for this
+            return ResultUnknown;
+        }
     }
     return ResultSuccess;
 }

src/core/hle/service/friend/friend.cpp

@@ -158,7 +158,7 @@ private:
         const auto local_play = rp.Pop<bool>();
         const auto uuid = rp.PopRaw<Common::UUID>();
 
-        LOG_WARNING(Service_Friend, "(STUBBED) called local_play={} uuid={}", local_play,
+        LOG_WARNING(Service_Friend, "(STUBBED) called, local_play={}, uuid=0x{}", local_play,
                     uuid.Format());
 
         IPC::ResponseBuilder rb{ctx, 2};
@@ -171,7 +171,7 @@ private:
         const auto uuid = rp.PopRaw<Common::UUID>();
         [[maybe_unused]] const auto filter = rp.PopRaw<SizedFriendFilter>();
         const auto pid = rp.Pop<u64>();
-        LOG_WARNING(Service_Friend, "(STUBBED) called, offset={}, uuid={}, pid={}", friend_offset,
+        LOG_WARNING(Service_Friend, "(STUBBED) called, offset={}, uuid=0x{}, pid={}", friend_offset,
                     uuid.Format(), pid);
 
         IPC::ResponseBuilder rb{ctx, 3};
@@ -289,7 +289,7 @@ void Module::Interface::CreateNotificationService(Kernel::HLERequestContext& ctx
     IPC::RequestParser rp{ctx};
     auto uuid = rp.PopRaw<Common::UUID>();
 
-    LOG_DEBUG(Service_Friend, "called, uuid={}", uuid.Format());
+    LOG_DEBUG(Service_Friend, "called, uuid=0x{}", uuid.Format());
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(ResultSuccess);

src/core/hle/service/hid/controllers/touchscreen.h

@@ -15,6 +15,20 @@
 namespace Service::HID {
 class Controller_Touchscreen final : public ControllerBase {
 public:
+    enum class TouchScreenModeForNx : u8 {
+        UseSystemSetting,
+        Finger,
+        Heat2,
+    };
+
+    struct TouchScreenConfigurationForNx {
+        TouchScreenModeForNx mode;
+        INSERT_PADDING_BYTES_NOINIT(0x7);
+        INSERT_PADDING_BYTES_NOINIT(0xF); // Reserved
+    };
+    static_assert(sizeof(TouchScreenConfigurationForNx) == 0x17,
+                  "TouchScreenConfigurationForNx is an invalid size");
+
     explicit Controller_Touchscreen(Core::System& system_);
     ~Controller_Touchscreen() override;
 

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

@@ -331,7 +331,7 @@ Hid::Hid(Core::System& system_)
         {529, nullptr, "SetDisallowedPalmaConnection"},
         {1000, &Hid::SetNpadCommunicationMode, "SetNpadCommunicationMode"},
         {1001, &Hid::GetNpadCommunicationMode, "GetNpadCommunicationMode"},
-        {1002, nullptr, "SetTouchScreenConfiguration"},
+        {1002, &Hid::SetTouchScreenConfiguration, "SetTouchScreenConfiguration"},
         {1003, nullptr, "IsFirmwareUpdateNeededForNotification"},
         {2000, nullptr, "ActivateDigitizer"},
     };
@@ -1631,6 +1631,18 @@ void Hid::GetNpadCommunicationMode(Kernel::HLERequestContext& ctx) {
                     .GetNpadCommunicationMode());
 }
 
+void Hid::SetTouchScreenConfiguration(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto touchscreen_mode{rp.PopRaw<Controller_Touchscreen::TouchScreenConfigurationForNx>()};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_WARNING(Service_HID, "(STUBBED) called, touchscreen_mode={}, applet_resource_user_id={}",
+                touchscreen_mode.mode, applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
 class HidDbg final : public ServiceFramework<HidDbg> {
 public:
     explicit HidDbg(Core::System& system_) : ServiceFramework{system_, "hid:dbg"} {

src/core/hle/service/hid/hid.h

@@ -159,6 +159,7 @@ private:
     void SetPalmaBoostMode(Kernel::HLERequestContext& ctx);
     void SetNpadCommunicationMode(Kernel::HLERequestContext& ctx);
     void GetNpadCommunicationMode(Kernel::HLERequestContext& ctx);
+    void SetTouchScreenConfiguration(Kernel::HLERequestContext& ctx);
 
     enum class VibrationDeviceType : u32 {
         Unknown = 0,

src/core/hle/service/ngct/ngct.cpp

@@ -0,0 +1,46 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included
+
+#include "common/string_util.h"
+#include "core/core.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/service/ngct/ngct.h"
+#include "core/hle/service/service.h"
+
+namespace Service::NGCT {
+
+class IService final : public ServiceFramework<IService> {
+public:
+    explicit IService(Core::System& system_) : ServiceFramework{system_, "ngct:u"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, nullptr, "Match"},
+            {1, &IService::Filter, "Filter"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+private:
+    void Filter(Kernel::HLERequestContext& ctx) {
+        const auto buffer = ctx.ReadBuffer();
+        const auto text = Common::StringFromFixedZeroTerminatedBuffer(
+            reinterpret_cast<const char*>(buffer.data()), buffer.size());
+
+        LOG_WARNING(Service_NGCT, "(STUBBED) called, text={}", text);
+
+        // Return the same string since we don't censor anything
+        ctx.WriteBuffer(buffer);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+};
+
+void InstallInterfaces(SM::ServiceManager& service_manager, Core::System& system) {
+    std::make_shared<IService>(system)->InstallAsService(system.ServiceManager());
+}
+
+} // namespace Service::NGCT

src/core/hle/service/ngct/ngct.h

@@ -0,0 +1,20 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included
+
+#pragma once
+
+namespace Core {
+class System;
+}
+
+namespace Service::SM {
+class ServiceManager;
+}
+
+namespace Service::NGCT {
+
+/// Registers all NGCT services with the specified service manager.
+void InstallInterfaces(SM::ServiceManager& service_manager, Core::System& system);
+
+} // namespace Service::NGCT

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

@@ -11,6 +11,7 @@
 #include "core/hle/service/nifm/nifm.h"
 #include "core/hle/service/service.h"
 #include "core/network/network.h"
+#include "core/network/network_interface.h"
 
 namespace Service::NIFM {
 
@@ -179,10 +180,10 @@ private:
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(ResultSuccess);
 
-        if (Settings::values.bcat_backend.GetValue() == "none") {
-            rb.PushEnum(RequestState::NotSubmitted);
-        } else {
+        if (Network::GetHostIPv4Address().has_value()) {
             rb.PushEnum(RequestState::Connected);
+        } else {
+            rb.PushEnum(RequestState::NotSubmitted);
         }
     }
 
@@ -276,37 +277,45 @@ private:
     void GetCurrentNetworkProfile(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
-        const SfNetworkProfileData network_profile_data{
-            .ip_setting_data{
-                .ip_address_setting{
-                    .is_automatic{true},
-                    .current_address{192, 168, 1, 100},
-                    .subnet_mask{255, 255, 255, 0},
-                    .gateway{192, 168, 1, 1},
+        const auto net_iface = Network::GetSelectedNetworkInterface();
+
+        const SfNetworkProfileData network_profile_data = [&net_iface] {
+            if (!net_iface) {
+                return SfNetworkProfileData{};
+            }
+
+            return SfNetworkProfileData{
+                .ip_setting_data{
+                    .ip_address_setting{
+                        .is_automatic{true},
+                        .current_address{Network::TranslateIPv4(net_iface->ip_address)},
+                        .subnet_mask{Network::TranslateIPv4(net_iface->subnet_mask)},
+                        .gateway{Network::TranslateIPv4(net_iface->gateway)},
+                    },
+                    .dns_setting{
+                        .is_automatic{true},
+                        .primary_dns{1, 1, 1, 1},
+                        .secondary_dns{1, 0, 0, 1},
+                    },
+                    .proxy_setting{
+                        .enabled{false},
+                        .port{},
+                        .proxy_server{},
+                        .automatic_auth_enabled{},
+                        .user{},
+                        .password{},
+                    },
+                    .mtu{1500},
                 },
-                .dns_setting{
-                    .is_automatic{true},
-                    .primary_dns{1, 1, 1, 1},
-                    .secondary_dns{1, 0, 0, 1},
+                .uuid{0xdeadbeef, 0xdeadbeef},
+                .network_name{"yuzu Network"},
+                .wireless_setting_data{
+                    .ssid_length{12},
+                    .ssid{"yuzu Network"},
+                    .passphrase{"yuzupassword"},
                 },
-                .proxy_setting{
-                    .enabled{false},
-                    .port{},
-                    .proxy_server{},
-                    .automatic_auth_enabled{},
-                    .user{},
-                    .password{},
-                },
-                .mtu{1500},
-            },
-            .uuid{0xdeadbeef, 0xdeadbeef},
-            .network_name{"yuzu Network"},
-            .wireless_setting_data{
-                .ssid_length{12},
-                .ssid{"yuzu Network"},
-                .passphrase{"yuzupassword"},
-            },
-        };
+            };
+        }();
 
         ctx.WriteBuffer(network_profile_data);
 
@@ -322,12 +331,15 @@ private:
     void GetCurrentIpAddress(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
-        const auto [ipv4, error] = Network::GetHostIPv4Address();
-        UNIMPLEMENTED_IF(error != Network::Errno::SUCCESS);
+        auto ipv4 = Network::GetHostIPv4Address();
+        if (!ipv4) {
+            LOG_ERROR(Service_NIFM, "Couldn't get host IPv4 address, defaulting to 0.0.0.0");
+            ipv4.emplace(Network::IPv4Address{0, 0, 0, 0});
+        }
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(ResultSuccess);
-        rb.PushRaw(ipv4);
+        rb.PushRaw(*ipv4);
     }
     void CreateTemporaryNetworkProfile(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_NIFM, "called");
@@ -348,25 +360,33 @@ private:
         LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
         struct IpConfigInfo {
-            IpAddressSetting ip_address_setting;
-            DnsSetting dns_setting;
+            IpAddressSetting ip_address_setting{};
+            DnsSetting dns_setting{};
         };
         static_assert(sizeof(IpConfigInfo) == sizeof(IpAddressSetting) + sizeof(DnsSetting),
                       "IpConfigInfo has incorrect size.");
 
-        const IpConfigInfo ip_config_info{
-            .ip_address_setting{
-                .is_automatic{true},
-                .current_address{192, 168, 1, 100},
-                .subnet_mask{255, 255, 255, 0},
-                .gateway{192, 168, 1, 1},
-            },
-            .dns_setting{
-                .is_automatic{true},
-                .primary_dns{1, 1, 1, 1},
-                .secondary_dns{1, 0, 0, 1},
-            },
-        };
+        const auto net_iface = Network::GetSelectedNetworkInterface();
+
+        const IpConfigInfo ip_config_info = [&net_iface] {
+            if (!net_iface) {
+                return IpConfigInfo{};
+            }
+
+            return IpConfigInfo{
+                .ip_address_setting{
+                    .is_automatic{true},
+                    .current_address{Network::TranslateIPv4(net_iface->ip_address)},
+                    .subnet_mask{Network::TranslateIPv4(net_iface->subnet_mask)},
+                    .gateway{Network::TranslateIPv4(net_iface->gateway)},
+                },
+                .dns_setting{
+                    .is_automatic{true},
+                    .primary_dns{1, 1, 1, 1},
+                    .secondary_dns{1, 0, 0, 1},
+                },
+            };
+        }();
 
         IPC::ResponseBuilder rb{ctx, 2 + (sizeof(IpConfigInfo) + 3) / sizeof(u32)};
         rb.Push(ResultSuccess);
@@ -384,10 +404,10 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(ResultSuccess);
-        if (Settings::values.bcat_backend.GetValue() == "none") {
-            rb.Push<u8>(0);
-        } else {
+        if (Network::GetHostIPv4Address().has_value()) {
             rb.Push<u8>(1);
+        } else {
+            rb.Push<u8>(0);
         }
     }
     void IsAnyInternetRequestAccepted(Kernel::HLERequestContext& ctx) {
@@ -395,10 +415,10 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(ResultSuccess);
-        if (Settings::values.bcat_backend.GetValue() == "none") {
-            rb.Push<u8>(0);
-        } else {
+        if (Network::GetHostIPv4Address().has_value()) {
             rb.Push<u8>(1);
+        } else {
+            rb.Push<u8>(0);
         }
     }
 };

src/core/hle/service/ns/language.cpp

@@ -339,13 +339,16 @@ std::optional<ApplicationLanguage> ConvertToApplicationLanguage(
     case Set::LanguageCode::FR_CA:
         return ApplicationLanguage::CanadianFrench;
     case Set::LanguageCode::PT:
+    case Set::LanguageCode::PT_BR:
         return ApplicationLanguage::Portuguese;
     case Set::LanguageCode::RU:
         return ApplicationLanguage::Russian;
     case Set::LanguageCode::KO:
         return ApplicationLanguage::Korean;
+    case Set::LanguageCode::ZH_TW:
     case Set::LanguageCode::ZH_HANT:
         return ApplicationLanguage::TraditionalChinese;
+    case Set::LanguageCode::ZH_CN:
     case Set::LanguageCode::ZH_HANS:
         return ApplicationLanguage::SimplifiedChinese;
     default:

src/core/hle/service/ns/ns_language.h

@@ -1,42 +0,0 @@
-// Copyright 2019 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-#include <optional>
-#include <string>
-#include "common/common_types.h"
-#include "core/hle/service/set/set.h"
-
-namespace Service::NS {
-/// This is nn::ns::detail::ApplicationLanguage
-enum class ApplicationLanguage : u8 {
-    AmericanEnglish = 0,
-    BritishEnglish,
-    Japanese,
-    French,
-    German,
-    LatinAmericanSpanish,
-    Spanish,
-    Italian,
-    Dutch,
-    CanadianFrench,
-    Portuguese,
-    Russian,
-    Korean,
-    TraditionalChinese,
-    SimplifiedChinese,
-    Count
-};
-using ApplicationLanguagePriorityList =
-    const std::array<ApplicationLanguage, static_cast<std::size_t>(ApplicationLanguage::Count)>;
-
-constexpr u32 GetSupportedLanguageFlag(const ApplicationLanguage lang) {
-    return 1U << static_cast<u32>(lang);
-}
-
-const ApplicationLanguagePriorityList* GetApplicationLanguagePriorityList(ApplicationLanguage lang);
-std::optional<ApplicationLanguage> ConvertToApplicationLanguage(
-    Service::Set::LanguageCode language_code);
-std::optional<Service::Set::LanguageCode> ConvertToLanguageCode(ApplicationLanguage lang);
-} // namespace Service::NS

src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp

@@ -42,15 +42,14 @@ void nvdisp_disp0::OnClose(DeviceFD fd) {}
 void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height,
                         u32 stride, NVFlinger::BufferQueue::BufferTransformFlags transform,
                         const Common::Rectangle<int>& crop_rect) {
-    VAddr addr = nvmap_dev->GetObjectAddress(buffer_handle);
+    const VAddr addr = nvmap_dev->GetObjectAddress(buffer_handle);
     LOG_TRACE(Service,
               "Drawing from address {:X} offset {:08X} Width {} Height {} Stride {} Format {}",
               addr, offset, width, height, stride, format);
 
-    using PixelFormat = Tegra::FramebufferConfig::PixelFormat;
-    const Tegra::FramebufferConfig framebuffer{
-        addr,      offset,   width, height, stride, static_cast<PixelFormat>(format),
-        transform, crop_rect};
+    const auto pixel_format = static_cast<Tegra::FramebufferConfig::PixelFormat>(format);
+    const Tegra::FramebufferConfig framebuffer{addr,   offset,       width,     height,
+                                               stride, pixel_format, transform, crop_rect};
 
     system.GetPerfStats().EndSystemFrame();
     system.GPU().SwapBuffers(&framebuffer);

src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp

@@ -166,8 +166,6 @@ NvResult nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vecto
             LOG_ERROR(Service_NVDRV, "failed to map size={}", object->size);
         } else {
             cmd_buffer.map_address = object->dma_map_addr;
-            AddBufferMap(object->dma_map_addr, object->size, object->addr,
-                         object->status == nvmap::Object::Status::Allocated);
         }
     }
     std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
@@ -178,30 +176,11 @@ NvResult nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vecto
 }
 
 NvResult nvhost_nvdec_common::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlMapBuffer params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
-    std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
-    SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
-
-    auto& gpu = system.GPU();
-
-    for (auto& cmd_buffer : cmd_buffer_handles) {
-        const auto object{nvmap_dev->GetObject(cmd_buffer.map_handle)};
-        if (!object) {
-            LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmd_buffer.map_handle);
-            std::memcpy(output.data(), &params, output.size());
-            return NvResult::InvalidState;
-        }
-        if (const auto size{RemoveBufferMap(object->dma_map_addr)}; size) {
-            gpu.MemoryManager().Unmap(object->dma_map_addr, *size);
-        } else {
-            // This occurs quite frequently, however does not seem to impact functionality
-            LOG_DEBUG(Service_NVDRV, "invalid offset=0x{:X} dma=0x{:X}", object->addr,
-                      object->dma_map_addr);
-        }
-        object->dma_map_addr = 0;
-    }
+    // This is intntionally stubbed.
+    // Skip unmapping buffers here, as to not break the continuity of the VP9 reference frame
+    // addresses, and risk invalidating data before the async GPU thread is done with it
     std::memset(output.data(), 0, output.size());
+    LOG_DEBUG(Service_NVDRV, "(STUBBED) called");
     return NvResult::Success;
 }
 
@@ -212,33 +191,4 @@ NvResult nvhost_nvdec_common::SetSubmitTimeout(const std::vector<u8>& input,
     return NvResult::Success;
 }
 
-std::optional<nvhost_nvdec_common::BufferMap> nvhost_nvdec_common::FindBufferMap(
-    GPUVAddr gpu_addr) const {
-    const auto it = std::find_if(
-        buffer_mappings.begin(), buffer_mappings.upper_bound(gpu_addr), [&](const auto& entry) {
-            return (gpu_addr >= entry.second.StartAddr() && gpu_addr < entry.second.EndAddr());
-        });
-
-    ASSERT(it != buffer_mappings.end());
-    return it->second;
-}
-
-void nvhost_nvdec_common::AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr,
-                                       bool is_allocated) {
-    buffer_mappings.insert_or_assign(gpu_addr, BufferMap{gpu_addr, size, cpu_addr, is_allocated});
-}
-
-std::optional<std::size_t> nvhost_nvdec_common::RemoveBufferMap(GPUVAddr gpu_addr) {
-    const auto iter{buffer_mappings.find(gpu_addr)};
-    if (iter == buffer_mappings.end()) {
-        return std::nullopt;
-    }
-    std::size_t size = 0;
-    if (iter->second.IsAllocated()) {
-        size = iter->second.Size();
-    }
-    buffer_mappings.erase(iter);
-    return size;
-}
-
 } // namespace Service::Nvidia::Devices

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

@@ -23,45 +23,6 @@ public:
     ~nvhost_nvdec_common() override;
 
 protected:
-    class BufferMap final {
-    public:
-        constexpr BufferMap() = default;
-
-        constexpr BufferMap(GPUVAddr start_addr_, std::size_t size_)
-            : start_addr{start_addr_}, end_addr{start_addr_ + size_} {}
-
-        constexpr BufferMap(GPUVAddr start_addr_, std::size_t size_, VAddr cpu_addr_,
-                            bool is_allocated_)
-            : start_addr{start_addr_}, end_addr{start_addr_ + size_}, cpu_addr{cpu_addr_},
-              is_allocated{is_allocated_} {}
-
-        constexpr VAddr StartAddr() const {
-            return start_addr;
-        }
-
-        constexpr VAddr EndAddr() const {
-            return end_addr;
-        }
-
-        constexpr std::size_t Size() const {
-            return end_addr - start_addr;
-        }
-
-        constexpr VAddr CpuAddr() const {
-            return cpu_addr;
-        }
-
-        constexpr bool IsAllocated() const {
-            return is_allocated;
-        }
-
-    private:
-        GPUVAddr start_addr{};
-        GPUVAddr end_addr{};
-        VAddr cpu_addr{};
-        bool is_allocated{};
-    };
-
     struct IoctlSetNvmapFD {
         s32_le nvmap_fd{};
     };
@@ -154,17 +115,11 @@ protected:
     NvResult UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
     NvResult SetSubmitTimeout(const std::vector<u8>& input, std::vector<u8>& output);
 
-    std::optional<BufferMap> FindBufferMap(GPUVAddr gpu_addr) const;
-    void AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr, bool is_allocated);
-    std::optional<std::size_t> RemoveBufferMap(GPUVAddr gpu_addr);
-
     s32_le nvmap_fd{};
     u32_le submit_timeout{};
     std::shared_ptr<nvmap> nvmap_dev;
     SyncpointManager& syncpoint_manager;
     std::array<u32, MaxSyncPoints> device_syncpoints{};
-    // This is expected to be ordered, therefore we must use a map, not unordered_map
-    std::map<GPUVAddr, BufferMap> buffer_mappings;
 };
 }; // namespace Devices
 } // namespace Service::Nvidia

src/core/hle/service/nvflinger/buffer_queue.cpp

@@ -9,17 +9,20 @@
 #include "core/core.h"
 #include "core/hle/kernel/k_writable_event.h"
 #include "core/hle/kernel/kernel.h"
+#include "core/hle/service/kernel_helpers.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
 
 namespace Service::NVFlinger {
 
-BufferQueue::BufferQueue(Kernel::KernelCore& kernel, u32 id_, u64 layer_id_)
-    : id(id_), layer_id(layer_id_), buffer_wait_event{kernel} {
-    Kernel::KAutoObject::Create(std::addressof(buffer_wait_event));
-    buffer_wait_event.Initialize("BufferQueue:WaitEvent");
+BufferQueue::BufferQueue(Kernel::KernelCore& kernel, u32 id_, u64 layer_id_,
+                         KernelHelpers::ServiceContext& service_context_)
+    : id(id_), layer_id(layer_id_), service_context{service_context_} {
+    buffer_wait_event = service_context.CreateEvent("BufferQueue:WaitEvent");
 }
 
-BufferQueue::~BufferQueue() = default;
+BufferQueue::~BufferQueue() {
+    service_context.CloseEvent(buffer_wait_event);
+}
 
 void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer) {
     ASSERT(slot < buffer_slots);
@@ -41,7 +44,7 @@ void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer)
         .multi_fence = {},
     };
 
-    buffer_wait_event.GetWritableEvent().Signal();
+    buffer_wait_event->GetWritableEvent().Signal();
 }
 
 std::optional<std::pair<u32, Service::Nvidia::MultiFence*>> BufferQueue::DequeueBuffer(u32 width,
@@ -119,7 +122,7 @@ void BufferQueue::CancelBuffer(u32 slot, const Service::Nvidia::MultiFence& mult
     }
     free_buffers_condition.notify_one();
 
-    buffer_wait_event.GetWritableEvent().Signal();
+    buffer_wait_event->GetWritableEvent().Signal();
 }
 
 std::optional<std::reference_wrapper<const BufferQueue::Buffer>> BufferQueue::AcquireBuffer() {
@@ -154,7 +157,7 @@ void BufferQueue::ReleaseBuffer(u32 slot) {
     }
     free_buffers_condition.notify_one();
 
-    buffer_wait_event.GetWritableEvent().Signal();
+    buffer_wait_event->GetWritableEvent().Signal();
 }
 
 void BufferQueue::Connect() {
@@ -169,7 +172,7 @@ void BufferQueue::Disconnect() {
         std::unique_lock lock{queue_sequence_mutex};
         queue_sequence.clear();
     }
-    buffer_wait_event.GetWritableEvent().Signal();
+    buffer_wait_event->GetWritableEvent().Signal();
     is_connect = false;
     free_buffers_condition.notify_one();
 }
@@ -189,11 +192,11 @@ u32 BufferQueue::Query(QueryType type) {
 }
 
 Kernel::KWritableEvent& BufferQueue::GetWritableBufferWaitEvent() {
-    return buffer_wait_event.GetWritableEvent();
+    return buffer_wait_event->GetWritableEvent();
 }
 
 Kernel::KReadableEvent& BufferQueue::GetBufferWaitEvent() {
-    return buffer_wait_event.GetReadableEvent();
+    return buffer_wait_event->GetReadableEvent();
 }
 
 } // namespace Service::NVFlinger

src/core/hle/service/nvflinger/buffer_queue.h

@@ -24,6 +24,10 @@ class KReadableEvent;
 class KWritableEvent;
 } // namespace Kernel
 
+namespace Service::KernelHelpers {
+class ServiceContext;
+} // namespace Service::KernelHelpers
+
 namespace Service::NVFlinger {
 
 constexpr u32 buffer_slots = 0x40;
@@ -38,7 +42,9 @@ struct IGBPBuffer {
     u32_le index;
     INSERT_PADDING_WORDS(3);
     u32_le gpu_buffer_id;
-    INSERT_PADDING_WORDS(17);
+    INSERT_PADDING_WORDS(6);
+    u32_le external_format;
+    INSERT_PADDING_WORDS(10);
     u32_le nvmap_handle;
     u32_le offset;
     INSERT_PADDING_WORDS(60);
@@ -54,7 +60,8 @@ public:
         NativeWindowFormat = 2,
     };
 
-    explicit BufferQueue(Kernel::KernelCore& kernel, u32 id_, u64 layer_id_);
+    explicit BufferQueue(Kernel::KernelCore& kernel, u32 id_, u64 layer_id_,
+                         KernelHelpers::ServiceContext& service_context_);
     ~BufferQueue();
 
     enum class BufferTransformFlags : u32 {
@@ -130,12 +137,14 @@ private:
     std::list<u32> free_buffers;
     std::array<Buffer, buffer_slots> buffers;
     std::list<u32> queue_sequence;
-    Kernel::KEvent buffer_wait_event;
+    Kernel::KEvent* buffer_wait_event{};
 
     std::mutex free_buffers_mutex;
     std::condition_variable free_buffers_condition;
 
     std::mutex queue_sequence_mutex;
+
+    KernelHelpers::ServiceContext& service_context;
 };
 
 } // namespace Service::NVFlinger

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -61,12 +61,13 @@ void NVFlinger::SplitVSync() {
     }
 }
 
-NVFlinger::NVFlinger(Core::System& system_) : system(system_) {
-    displays.emplace_back(0, "Default", system);
-    displays.emplace_back(1, "External", system);
-    displays.emplace_back(2, "Edid", system);
-    displays.emplace_back(3, "Internal", system);
-    displays.emplace_back(4, "Null", system);
+NVFlinger::NVFlinger(Core::System& system_)
+    : system(system_), service_context(system_, "nvflinger") {
+    displays.emplace_back(0, "Default", service_context, system);
+    displays.emplace_back(1, "External", service_context, system);
+    displays.emplace_back(2, "Edid", service_context, system);
+    displays.emplace_back(3, "Internal", service_context, system);
+    displays.emplace_back(4, "Null", service_context, system);
     guard = std::make_shared<std::mutex>();
 
     // Schedule the screen composition events
@@ -146,7 +147,7 @@ std::optional<u64> NVFlinger::CreateLayer(u64 display_id) {
 void NVFlinger::CreateLayerAtId(VI::Display& display, u64 layer_id) {
     const u32 buffer_queue_id = next_buffer_queue_id++;
     buffer_queues.emplace_back(
-        std::make_unique<BufferQueue>(system.Kernel(), buffer_queue_id, layer_id));
+        std::make_unique<BufferQueue>(system.Kernel(), buffer_queue_id, layer_id, service_context));
     display.CreateLayer(layer_id, *buffer_queues.back());
 }
 
@@ -297,7 +298,7 @@ void NVFlinger::Compose() {
         auto nvdisp = nvdrv->GetDevice<Nvidia::Devices::nvdisp_disp0>("/dev/nvdisp_disp0");
         ASSERT(nvdisp);
 
-        nvdisp->flip(igbp_buffer.gpu_buffer_id, igbp_buffer.offset, igbp_buffer.format,
+        nvdisp->flip(igbp_buffer.gpu_buffer_id, igbp_buffer.offset, igbp_buffer.external_format,
                      igbp_buffer.width, igbp_buffer.height, igbp_buffer.stride,
                      buffer->get().transform, buffer->get().crop_rect);
 

src/core/hle/service/nvflinger/nvflinger.h

@@ -15,6 +15,7 @@
 #include <vector>
 
 #include "common/common_types.h"
+#include "core/hle/service/kernel_helpers.h"
 
 namespace Common {
 class Event;
@@ -135,6 +136,8 @@ private:
     std::unique_ptr<std::thread> vsync_thread;
     std::unique_ptr<Common::Event> wait_event;
     std::atomic<bool> is_running{};
+
+    KernelHelpers::ServiceContext service_context;
 };
 
 } // namespace Service::NVFlinger

src/core/hle/service/service.cpp

@@ -46,6 +46,7 @@
 #include "core/hle/service/ncm/ncm.h"
 #include "core/hle/service/nfc/nfc.h"
 #include "core/hle/service/nfp/nfp.h"
+#include "core/hle/service/ngct/ngct.h"
 #include "core/hle/service/nifm/nifm.h"
 #include "core/hle/service/nim/nim.h"
 #include "core/hle/service/npns/npns.h"
@@ -271,6 +272,7 @@ Services::Services(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system
     NCM::InstallInterfaces(*sm, system);
     NFC::InstallInterfaces(*sm, system);
     NFP::InstallInterfaces(*sm, system);
+    NGCT::InstallInterfaces(*sm, system);
     NIFM::InstallInterfaces(*sm, system);
     NIM::InstallInterfaces(*sm, system);
     NPNS::InstallInterfaces(*sm, system);

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

@@ -12,7 +12,7 @@
 
 namespace Service::Set {
 namespace {
-constexpr std::array<LanguageCode, 17> available_language_codes = {{
+constexpr std::array<LanguageCode, 18> available_language_codes = {{
     LanguageCode::JA,
     LanguageCode::EN_US,
     LanguageCode::FR,
@@ -30,6 +30,7 @@ constexpr std::array<LanguageCode, 17> available_language_codes = {{
     LanguageCode::ES_419,
     LanguageCode::ZH_HANS,
     LanguageCode::ZH_HANT,
+    LanguageCode::PT_BR,
 }};
 
 enum class KeyboardLayout : u64 {
@@ -50,7 +51,7 @@ enum class KeyboardLayout : u64 {
     ChineseTraditional = 14,
 };
 
-constexpr std::array<std::pair<LanguageCode, KeyboardLayout>, 17> language_to_layout{{
+constexpr std::array<std::pair<LanguageCode, KeyboardLayout>, 18> language_to_layout{{
     {LanguageCode::JA, KeyboardLayout::Japanese},
     {LanguageCode::EN_US, KeyboardLayout::EnglishUs},
     {LanguageCode::FR, KeyboardLayout::French},
@@ -68,10 +69,11 @@ constexpr std::array<std::pair<LanguageCode, KeyboardLayout>, 17> language_to_la
     {LanguageCode::ES_419, KeyboardLayout::SpanishLatin},
     {LanguageCode::ZH_HANS, KeyboardLayout::ChineseSimplified},
     {LanguageCode::ZH_HANT, KeyboardLayout::ChineseTraditional},
+    {LanguageCode::PT_BR, KeyboardLayout::Portuguese},
 }};
 
-constexpr std::size_t pre4_0_0_max_entries = 15;
-constexpr std::size_t post4_0_0_max_entries = 17;
+constexpr std::size_t PRE_4_0_0_MAX_ENTRIES = 0xF;
+constexpr std::size_t POST_4_0_0_MAX_ENTRIES = 0x40;
 
 constexpr ResultCode ERR_INVALID_LANGUAGE{ErrorModule::Settings, 625};
 
@@ -81,9 +83,10 @@ void PushResponseLanguageCode(Kernel::HLERequestContext& ctx, std::size_t num_la
     rb.Push(static_cast<u32>(num_language_codes));
 }
 
-void GetAvailableLanguageCodesImpl(Kernel::HLERequestContext& ctx, std::size_t max_size) {
+void GetAvailableLanguageCodesImpl(Kernel::HLERequestContext& ctx, std::size_t max_entries) {
     const std::size_t requested_amount = ctx.GetWriteBufferSize() / sizeof(LanguageCode);
-    const std::size_t copy_amount = std::min(requested_amount, max_size);
+    const std::size_t max_amount = std::min(requested_amount, max_entries);
+    const std::size_t copy_amount = std::min(available_language_codes.size(), max_amount);
     const std::size_t copy_size = copy_amount * sizeof(LanguageCode);
 
     ctx.WriteBuffer(available_language_codes.data(), copy_size);
@@ -118,7 +121,7 @@ LanguageCode GetLanguageCodeFromIndex(std::size_t index) {
 void SET::GetAvailableLanguageCodes(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
-    GetAvailableLanguageCodesImpl(ctx, pre4_0_0_max_entries);
+    GetAvailableLanguageCodesImpl(ctx, PRE_4_0_0_MAX_ENTRIES);
 }
 
 void SET::MakeLanguageCode(Kernel::HLERequestContext& ctx) {
@@ -140,19 +143,19 @@ void SET::MakeLanguageCode(Kernel::HLERequestContext& ctx) {
 void SET::GetAvailableLanguageCodes2(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
-    GetAvailableLanguageCodesImpl(ctx, post4_0_0_max_entries);
+    GetAvailableLanguageCodesImpl(ctx, POST_4_0_0_MAX_ENTRIES);
 }
 
 void SET::GetAvailableLanguageCodeCount(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
-    PushResponseLanguageCode(ctx, pre4_0_0_max_entries);
+    PushResponseLanguageCode(ctx, PRE_4_0_0_MAX_ENTRIES);
 }
 
 void SET::GetAvailableLanguageCodeCount2(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_SET, "called");
 
-    PushResponseLanguageCode(ctx, post4_0_0_max_entries);
+    PushResponseLanguageCode(ctx, POST_4_0_0_MAX_ENTRIES);
 }
 
 void SET::GetQuestFlag(Kernel::HLERequestContext& ctx) {

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

@@ -31,6 +31,7 @@ enum class LanguageCode : u64 {
     ES_419 = 0x00003931342D7365,
     ZH_HANS = 0x00736E61482D687A,
     ZH_HANT = 0x00746E61482D687A,
+    PT_BR = 0x00000052422D7470,
 };
 LanguageCode GetLanguageCodeFromIndex(std::size_t idx);
 

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

@@ -12,18 +12,21 @@
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_readable_event.h"
 #include "core/hle/kernel/k_writable_event.h"
+#include "core/hle/service/kernel_helpers.h"
 #include "core/hle/service/vi/display/vi_display.h"
 #include "core/hle/service/vi/layer/vi_layer.h"
 
 namespace Service::VI {
 
-Display::Display(u64 id, std::string name_, Core::System& system)
-    : display_id{id}, name{std::move(name_)}, vsync_event{system.Kernel()} {
-    Kernel::KAutoObject::Create(std::addressof(vsync_event));
-    vsync_event.Initialize(fmt::format("Display VSync Event {}", id));
+Display::Display(u64 id, std::string name_, KernelHelpers::ServiceContext& service_context_,
+                 Core::System& system_)
+    : display_id{id}, name{std::move(name_)}, service_context{service_context_} {
+    vsync_event = service_context.CreateEvent(fmt::format("Display VSync Event {}", id));
 }
 
-Display::~Display() = default;
+Display::~Display() {
+    service_context.CloseEvent(vsync_event);
+}
 
 Layer& Display::GetLayer(std::size_t index) {
     return *layers.at(index);
@@ -34,11 +37,11 @@ const Layer& Display::GetLayer(std::size_t index) const {
 }
 
 Kernel::KReadableEvent& Display::GetVSyncEvent() {
-    return vsync_event.GetReadableEvent();
+    return vsync_event->GetReadableEvent();
 }
 
 void Display::SignalVSyncEvent() {
-    vsync_event.GetWritableEvent().Signal();
+    vsync_event->GetWritableEvent().Signal();
 }
 
 void Display::CreateLayer(u64 layer_id, NVFlinger::BufferQueue& buffer_queue) {

src/core/hle/service/vi/display/vi_display.h

@@ -18,6 +18,9 @@ class KEvent;
 namespace Service::NVFlinger {
 class BufferQueue;
 }
+namespace Service::KernelHelpers {
+class ServiceContext;
+} // namespace Service::KernelHelpers
 
 namespace Service::VI {
 
@@ -31,10 +34,13 @@ class Display {
 public:
     /// Constructs a display with a given unique ID and name.
     ///
-    /// @param id   The unique ID for this display.
+    /// @param id The unique ID for this display.
+    /// @param service_context_ The ServiceContext for the owning service.
     /// @param name_ The name for this display.
+    /// @param system_ The global system instance.
     ///
-    Display(u64 id, std::string name_, Core::System& system);
+    Display(u64 id, std::string name_, KernelHelpers::ServiceContext& service_context_,
+            Core::System& system_);
     ~Display();
 
     /// Gets the unique ID assigned to this display.
@@ -98,9 +104,10 @@ public:
 private:
     u64 display_id;
     std::string name;
+    KernelHelpers::ServiceContext& service_context;
 
     std::vector<std::shared_ptr<Layer>> layers;
-    Kernel::KEvent vsync_event;
+    Kernel::KEvent* vsync_event{};
 };
 
 } // namespace Service::VI

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

@@ -1158,7 +1158,7 @@ private:
 
         const auto layer_id = nv_flinger.CreateLayer(display_id);
         if (!layer_id) {
-            LOG_ERROR(Service_VI, "Layer not found! layer_id={}", *layer_id);
+            LOG_ERROR(Service_VI, "Layer not found! display_id={}", display_id);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERR_NOT_FOUND);
             return;

src/core/memory.cpp

@@ -4,8 +4,6 @@
 
 #include <algorithm>
 #include <cstring>
-#include <optional>
-#include <utility>
 
 #include "common/assert.h"
 #include "common/atomic_ops.h"
@@ -14,12 +12,10 @@
 #include "common/page_table.h"
 #include "common/settings.h"
 #include "common/swap.h"
-#include "core/arm/arm_interface.h"
 #include "core/core.h"
 #include "core/device_memory.h"
 #include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_process.h"
-#include "core/hle/kernel/physical_memory.h"
 #include "core/memory.h"
 #include "video_core/gpu.h"
 
@@ -62,17 +58,7 @@ struct Memory::Impl {
         }
     }
 
-    bool IsValidVirtualAddress(const Kernel::KProcess& process, const VAddr vaddr) const {
-        const auto& page_table = process.PageTable().PageTableImpl();
-        const auto [pointer, type] = page_table.pointers[vaddr >> PAGE_BITS].PointerType();
-        return pointer != nullptr || type == Common::PageType::RasterizerCachedMemory;
-    }
-
-    bool IsValidVirtualAddress(VAddr vaddr) const {
-        return IsValidVirtualAddress(*system.CurrentProcess(), vaddr);
-    }
-
-    u8* GetPointerFromRasterizerCachedMemory(VAddr vaddr) const {
+    [[nodiscard]] u8* GetPointerFromRasterizerCachedMemory(VAddr vaddr) const {
         const PAddr paddr{current_page_table->backing_addr[vaddr >> PAGE_BITS]};
 
         if (!paddr) {
@@ -82,18 +68,6 @@ struct Memory::Impl {
         return system.DeviceMemory().GetPointer(paddr) + vaddr;
     }
 
-    u8* GetPointer(const VAddr vaddr) const {
-        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
-        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
-            return pointer + vaddr;
-        }
-        const auto type = Common::PageTable::PageInfo::ExtractType(raw_pointer);
-        if (type == Common::PageType::RasterizerCachedMemory) {
-            return GetPointerFromRasterizerCachedMemory(vaddr);
-        }
-        return nullptr;
-    }
-
     u8 Read8(const VAddr addr) {
         return Read<u8>(addr);
     }
@@ -179,7 +153,7 @@ struct Memory::Impl {
         std::string string;
         string.reserve(max_length);
         for (std::size_t i = 0; i < max_length; ++i) {
-            const char c = Read8(vaddr);
+            const char c = Read<s8>(vaddr);
             if (c == '\0') {
                 break;
             }
@@ -190,15 +164,14 @@ struct Memory::Impl {
         return string;
     }
 
-    void ReadBlock(const Kernel::KProcess& process, const VAddr src_addr, void* dest_buffer,
-                   const std::size_t size) {
+    void WalkBlock(const Kernel::KProcess& process, const VAddr addr, const std::size_t size,
+                   auto on_unmapped, auto on_memory, auto on_rasterizer, auto increment) {
         const auto& page_table = process.PageTable().PageTableImpl();
-
         std::size_t remaining_size = size;
-        std::size_t page_index = src_addr >> PAGE_BITS;
-        std::size_t page_offset = src_addr & PAGE_MASK;
+        std::size_t page_index = addr >> PAGE_BITS;
+        std::size_t page_offset = addr & PAGE_MASK;
 
-        while (remaining_size > 0) {
+        while (remaining_size) {
             const std::size_t copy_amount =
                 std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
             const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
@@ -206,22 +179,18 @@ struct Memory::Impl {
             const auto [pointer, type] = page_table.pointers[page_index].PointerType();
             switch (type) {
             case Common::PageType::Unmapped: {
-                LOG_ERROR(HW_Memory,
-                          "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
-                          current_vaddr, src_addr, size);
-                std::memset(dest_buffer, 0, copy_amount);
+                on_unmapped(copy_amount, current_vaddr);
                 break;
             }
             case Common::PageType::Memory: {
                 DEBUG_ASSERT(pointer);
-                const u8* const src_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                std::memcpy(dest_buffer, src_ptr, copy_amount);
+                u8* mem_ptr = pointer + page_offset + (page_index << PAGE_BITS);
+                on_memory(copy_amount, mem_ptr);
                 break;
             }
             case Common::PageType::RasterizerCachedMemory: {
-                const u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)};
-                system.GPU().FlushRegion(current_vaddr, copy_amount);
-                std::memcpy(dest_buffer, host_ptr, copy_amount);
+                u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)};
+                on_rasterizer(current_vaddr, copy_amount, host_ptr);
                 break;
             }
             default:
@@ -230,248 +199,122 @@ struct Memory::Impl {
 
             page_index++;
             page_offset = 0;
-            dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
+            increment(copy_amount);
             remaining_size -= copy_amount;
         }
     }
 
-    void ReadBlockUnsafe(const Kernel::KProcess& process, const VAddr src_addr, void* dest_buffer,
-                         const std::size_t size) {
-        const auto& page_table = process.PageTable().PageTableImpl();
-
-        std::size_t remaining_size = size;
-        std::size_t page_index = src_addr >> PAGE_BITS;
-        std::size_t page_offset = src_addr & PAGE_MASK;
-
-        while (remaining_size > 0) {
-            const std::size_t copy_amount =
-                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
-            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-
-            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
-            switch (type) {
-            case Common::PageType::Unmapped: {
+    template <bool UNSAFE>
+    void ReadBlockImpl(const Kernel::KProcess& process, const VAddr src_addr, void* dest_buffer,
+                       const std::size_t size) {
+        WalkBlock(
+            process, src_addr, size,
+            [src_addr, size, &dest_buffer](const std::size_t copy_amount,
+                                           const VAddr current_vaddr) {
                 LOG_ERROR(HW_Memory,
                           "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                           current_vaddr, src_addr, size);
                 std::memset(dest_buffer, 0, copy_amount);
-                break;
-            }
-            case Common::PageType::Memory: {
-                DEBUG_ASSERT(pointer);
-                const u8* const src_ptr = pointer + page_offset + (page_index << PAGE_BITS);
+            },
+            [&dest_buffer](const std::size_t copy_amount, const u8* const src_ptr) {
                 std::memcpy(dest_buffer, src_ptr, copy_amount);
-                break;
-            }
-            case Common::PageType::RasterizerCachedMemory: {
-                const u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)};
+            },
+            [&system = system, &dest_buffer](const VAddr current_vaddr,
+                                             const std::size_t copy_amount,
+                                             const u8* const host_ptr) {
+                if constexpr (!UNSAFE) {
+                    system.GPU().FlushRegion(current_vaddr, copy_amount);
+                }
                 std::memcpy(dest_buffer, host_ptr, copy_amount);
-                break;
-            }
-            default:
-                UNREACHABLE();
-            }
-
-            page_index++;
-            page_offset = 0;
-            dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
-            remaining_size -= copy_amount;
-        }
+            },
+            [&dest_buffer](const std::size_t copy_amount) {
+                dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
+            });
     }
 
     void ReadBlock(const VAddr src_addr, void* dest_buffer, const std::size_t size) {
-        ReadBlock(*system.CurrentProcess(), src_addr, dest_buffer, size);
+        ReadBlockImpl<false>(*system.CurrentProcess(), src_addr, dest_buffer, size);
     }
 
     void ReadBlockUnsafe(const VAddr src_addr, void* dest_buffer, const std::size_t size) {
-        ReadBlockUnsafe(*system.CurrentProcess(), src_addr, dest_buffer, size);
+        ReadBlockImpl<true>(*system.CurrentProcess(), src_addr, dest_buffer, size);
     }
 
-    void WriteBlock(const Kernel::KProcess& process, const VAddr dest_addr, const void* src_buffer,
-                    const std::size_t size) {
-        const auto& page_table = process.PageTable().PageTableImpl();
-        std::size_t remaining_size = size;
-        std::size_t page_index = dest_addr >> PAGE_BITS;
-        std::size_t page_offset = dest_addr & PAGE_MASK;
-
-        while (remaining_size > 0) {
-            const std::size_t copy_amount =
-                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
-            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-
-            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
-            switch (type) {
-            case Common::PageType::Unmapped: {
+    template <bool UNSAFE>
+    void WriteBlockImpl(const Kernel::KProcess& process, const VAddr dest_addr,
+                        const void* src_buffer, const std::size_t size) {
+        WalkBlock(
+            process, dest_addr, size,
+            [dest_addr, size](const std::size_t copy_amount, const VAddr current_vaddr) {
                 LOG_ERROR(HW_Memory,
                           "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                           current_vaddr, dest_addr, size);
-                break;
-            }
-            case Common::PageType::Memory: {
-                DEBUG_ASSERT(pointer);
-                u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS);
+            },
+            [&src_buffer](const std::size_t copy_amount, u8* const dest_ptr) {
                 std::memcpy(dest_ptr, src_buffer, copy_amount);
-                break;
-            }
-            case Common::PageType::RasterizerCachedMemory: {
-                u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)};
-                system.GPU().InvalidateRegion(current_vaddr, copy_amount);
+            },
+            [&system = system, &src_buffer](const VAddr current_vaddr,
+                                            const std::size_t copy_amount, u8* const host_ptr) {
+                if constexpr (!UNSAFE) {
+                    system.GPU().InvalidateRegion(current_vaddr, copy_amount);
+                }
                 std::memcpy(host_ptr, src_buffer, copy_amount);
-                break;
-            }
-            default:
-                UNREACHABLE();
-            }
-
-            page_index++;
-            page_offset = 0;
-            src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
-            remaining_size -= copy_amount;
-        }
-    }
-
-    void WriteBlockUnsafe(const Kernel::KProcess& process, const VAddr dest_addr,
-                          const void* src_buffer, const std::size_t size) {
-        const auto& page_table = process.PageTable().PageTableImpl();
-        std::size_t remaining_size = size;
-        std::size_t page_index = dest_addr >> PAGE_BITS;
-        std::size_t page_offset = dest_addr & PAGE_MASK;
-
-        while (remaining_size > 0) {
-            const std::size_t copy_amount =
-                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
-            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-
-            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
-            switch (type) {
-            case Common::PageType::Unmapped: {
-                LOG_ERROR(HW_Memory,
-                          "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
-                          current_vaddr, dest_addr, size);
-                break;
-            }
-            case Common::PageType::Memory: {
-                DEBUG_ASSERT(pointer);
-                u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                std::memcpy(dest_ptr, src_buffer, copy_amount);
-                break;
-            }
-            case Common::PageType::RasterizerCachedMemory: {
-                u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)};
-                std::memcpy(host_ptr, src_buffer, copy_amount);
-                break;
-            }
-            default:
-                UNREACHABLE();
-            }
-
-            page_index++;
-            page_offset = 0;
-            src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
-            remaining_size -= copy_amount;
-        }
+            },
+            [&src_buffer](const std::size_t copy_amount) {
+                src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
+            });
     }
 
     void WriteBlock(const VAddr dest_addr, const void* src_buffer, const std::size_t size) {
-        WriteBlock(*system.CurrentProcess(), dest_addr, src_buffer, size);
+        WriteBlockImpl<false>(*system.CurrentProcess(), dest_addr, src_buffer, size);
     }
 
     void WriteBlockUnsafe(const VAddr dest_addr, const void* src_buffer, const std::size_t size) {
-        WriteBlockUnsafe(*system.CurrentProcess(), dest_addr, src_buffer, size);
+        WriteBlockImpl<true>(*system.CurrentProcess(), dest_addr, src_buffer, size);
     }
 
     void ZeroBlock(const Kernel::KProcess& process, const VAddr dest_addr, const std::size_t size) {
-        const auto& page_table = process.PageTable().PageTableImpl();
-        std::size_t remaining_size = size;
-        std::size_t page_index = dest_addr >> PAGE_BITS;
-        std::size_t page_offset = dest_addr & PAGE_MASK;
-
-        while (remaining_size > 0) {
-            const std::size_t copy_amount =
-                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
-            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-
-            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
-            switch (type) {
-            case Common::PageType::Unmapped: {
+        WalkBlock(
+            process, dest_addr, size,
+            [dest_addr, size](const std::size_t copy_amount, const VAddr current_vaddr) {
                 LOG_ERROR(HW_Memory,
                           "Unmapped ZeroBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                           current_vaddr, dest_addr, size);
-                break;
-            }
-            case Common::PageType::Memory: {
-                DEBUG_ASSERT(pointer);
-                u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS);
+            },
+            [](const std::size_t copy_amount, u8* const dest_ptr) {
                 std::memset(dest_ptr, 0, copy_amount);
-                break;
-            }
-            case Common::PageType::RasterizerCachedMemory: {
-                u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)};
+            },
+            [&system = system](const VAddr current_vaddr, const std::size_t copy_amount,
+                               u8* const host_ptr) {
                 system.GPU().InvalidateRegion(current_vaddr, copy_amount);
                 std::memset(host_ptr, 0, copy_amount);
-                break;
-            }
-            default:
-                UNREACHABLE();
-            }
-
-            page_index++;
-            page_offset = 0;
-            remaining_size -= copy_amount;
-        }
-    }
-
-    void ZeroBlock(const VAddr dest_addr, const std::size_t size) {
-        ZeroBlock(*system.CurrentProcess(), dest_addr, size);
+            },
+            [](const std::size_t copy_amount) {});
     }
 
     void CopyBlock(const Kernel::KProcess& process, VAddr dest_addr, VAddr src_addr,
                    const std::size_t size) {
-        const auto& page_table = process.PageTable().PageTableImpl();
-        std::size_t remaining_size = size;
-        std::size_t page_index = src_addr >> PAGE_BITS;
-        std::size_t page_offset = src_addr & PAGE_MASK;
-
-        while (remaining_size > 0) {
-            const std::size_t copy_amount =
-                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
-            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-
-            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
-            switch (type) {
-            case Common::PageType::Unmapped: {
+        WalkBlock(
+            process, dest_addr, size,
+            [this, &process, &dest_addr, &src_addr, size](const std::size_t copy_amount,
+                                                          const VAddr current_vaddr) {
                 LOG_ERROR(HW_Memory,
                           "Unmapped CopyBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                           current_vaddr, src_addr, size);
                 ZeroBlock(process, dest_addr, copy_amount);
-                break;
-            }
-            case Common::PageType::Memory: {
-                DEBUG_ASSERT(pointer);
-                const u8* src_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                WriteBlock(process, dest_addr, src_ptr, copy_amount);
-                break;
-            }
-            case Common::PageType::RasterizerCachedMemory: {
-                const u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)};
+            },
+            [this, &process, &dest_addr](const std::size_t copy_amount, const u8* const src_ptr) {
+                WriteBlockImpl<false>(process, dest_addr, src_ptr, copy_amount);
+            },
+            [this, &system = system, &process, &dest_addr](
+                const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
                 system.GPU().FlushRegion(current_vaddr, copy_amount);
-                WriteBlock(process, dest_addr, host_ptr, copy_amount);
-                break;
-            }
-            default:
-                UNREACHABLE();
-            }
-
-            page_index++;
-            page_offset = 0;
-            dest_addr += static_cast<VAddr>(copy_amount);
-            src_addr += static_cast<VAddr>(copy_amount);
-            remaining_size -= copy_amount;
-        }
-    }
-
-    void CopyBlock(VAddr dest_addr, VAddr src_addr, std::size_t size) {
-        return CopyBlock(*system.CurrentProcess(), dest_addr, src_addr, size);
+                WriteBlockImpl<false>(process, dest_addr, host_ptr, copy_amount);
+            },
+            [&dest_addr, &src_addr](const std::size_t copy_amount) {
+                dest_addr += static_cast<VAddr>(copy_amount);
+                src_addr += static_cast<VAddr>(copy_amount);
+            });
     }
 
     void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
@@ -514,7 +357,7 @@ struct Memory::Impl {
             } else {
                 // Switch page type to uncached if now uncached
                 switch (page_type) {
-                case Common::PageType::Unmapped:
+                case Common::PageType::Unmapped: // NOLINT(bugprone-branch-clone)
                     // It is not necessary for a process to have this region mapped into its address
                     // space, for example, a system module need not have a VRAM mapping.
                     break;
@@ -597,6 +440,44 @@ struct Memory::Impl {
         }
     }
 
+    [[nodiscard]] u8* GetPointerImpl(VAddr vaddr, auto on_unmapped, auto on_rasterizer) const {
+        // AARCH64 masks the upper 16 bit of all memory accesses
+        vaddr &= 0xffffffffffffLL;
+
+        if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) {
+            on_unmapped();
+            return nullptr;
+        }
+
+        // Avoid adding any extra logic to this fast-path block
+        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
+        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
+            return &pointer[vaddr];
+        }
+        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
+        case Common::PageType::Unmapped:
+            on_unmapped();
+            return nullptr;
+        case Common::PageType::Memory:
+            ASSERT_MSG(false, "Mapped memory page without a pointer @ 0x{:016X}", vaddr);
+            return nullptr;
+        case Common::PageType::RasterizerCachedMemory: {
+            u8* const host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)};
+            on_rasterizer();
+            return host_ptr;
+        }
+        default:
+            UNREACHABLE();
+        }
+        return nullptr;
+    }
+
+    [[nodiscard]] u8* GetPointer(const VAddr vaddr) const {
+        return GetPointerImpl(
+            vaddr, [vaddr]() { LOG_ERROR(HW_Memory, "Unmapped GetPointer @ 0x{:016X}", vaddr); },
+            []() {});
+    }
+
     /**
      * Reads a particular data type out of memory at the given virtual address.
      *
@@ -610,39 +491,17 @@ struct Memory::Impl {
      */
     template <typename T>
     T Read(VAddr vaddr) {
-        // AARCH64 masks the upper 16 bit of all memory accesses
-        vaddr &= 0xffffffffffffLL;
-
-        if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) {
-            LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:08X}", sizeof(T) * 8, vaddr);
-            return 0;
+        T result = 0;
+        const u8* const ptr = GetPointerImpl(
+            vaddr,
+            [vaddr]() {
+                LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:016X}", sizeof(T) * 8, vaddr);
+            },
+            [&system = system, vaddr]() { system.GPU().FlushRegion(vaddr, sizeof(T)); });
+        if (ptr) {
+            std::memcpy(&result, ptr, sizeof(T));
         }
-
-        // Avoid adding any extra logic to this fast-path block
-        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
-        if (const u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
-            T value;
-            std::memcpy(&value, &pointer[vaddr], sizeof(T));
-            return value;
-        }
-        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
-        case Common::PageType::Unmapped:
-            LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:08X}", sizeof(T) * 8, vaddr);
-            return 0;
-        case Common::PageType::Memory:
-            ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr);
-            break;
-        case Common::PageType::RasterizerCachedMemory: {
-            const u8* const host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)};
-            system.GPU().FlushRegion(vaddr, sizeof(T));
-            T value;
-            std::memcpy(&value, host_ptr, sizeof(T));
-            return value;
-        }
-        default:
-            UNREACHABLE();
-        }
-        return {};
+        return result;
     }
 
     /**
@@ -656,110 +515,46 @@ struct Memory::Impl {
      */
     template <typename T>
     void Write(VAddr vaddr, const T data) {
-        // AARCH64 masks the upper 16 bit of all memory accesses
-        vaddr &= 0xffffffffffffLL;
-
-        if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) {
-            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8,
-                      static_cast<u32>(data), vaddr);
-            return;
-        }
-
-        // Avoid adding any extra logic to this fast-path block
-        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
-        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
-            std::memcpy(&pointer[vaddr], &data, sizeof(T));
-            return;
-        }
-        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
-        case Common::PageType::Unmapped:
-            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8,
-                      static_cast<u32>(data), vaddr);
-            return;
-        case Common::PageType::Memory:
-            ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr);
-            break;
-        case Common::PageType::RasterizerCachedMemory: {
-            u8* const host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)};
-            system.GPU().InvalidateRegion(vaddr, sizeof(T));
-            std::memcpy(host_ptr, &data, sizeof(T));
-            break;
-        }
-        default:
-            UNREACHABLE();
+        u8* const ptr = GetPointerImpl(
+            vaddr,
+            [vaddr, data]() {
+                LOG_ERROR(HW_Memory, "Unmapped Write{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8,
+                          vaddr, static_cast<u64>(data));
+            },
+            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
+        if (ptr) {
+            std::memcpy(ptr, &data, sizeof(T));
         }
     }
 
     template <typename T>
     bool WriteExclusive(VAddr vaddr, const T data, const T expected) {
-        // AARCH64 masks the upper 16 bit of all memory accesses
-        vaddr &= 0xffffffffffffLL;
-
-        if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) {
-            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8,
-                      static_cast<u32>(data), vaddr);
-            return true;
-        }
-
-        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
-        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
-            // NOTE: Avoid adding any extra logic to this fast-path block
-            const auto volatile_pointer = reinterpret_cast<volatile T*>(&pointer[vaddr]);
+        u8* const ptr = GetPointerImpl(
+            vaddr,
+            [vaddr, data]() {
+                LOG_ERROR(HW_Memory, "Unmapped WriteExclusive{} @ 0x{:016X} = 0x{:016X}",
+                          sizeof(T) * 8, vaddr, static_cast<u64>(data));
+            },
+            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
+        if (ptr) {
+            const auto volatile_pointer = reinterpret_cast<volatile T*>(ptr);
             return Common::AtomicCompareAndSwap(volatile_pointer, data, expected);
         }
-        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
-        case Common::PageType::Unmapped:
-            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8,
-                      static_cast<u32>(data), vaddr);
-            return true;
-        case Common::PageType::Memory:
-            ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr);
-            break;
-        case Common::PageType::RasterizerCachedMemory: {
-            u8* host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)};
-            system.GPU().InvalidateRegion(vaddr, sizeof(T));
-            auto* pointer = reinterpret_cast<volatile T*>(&host_ptr);
-            return Common::AtomicCompareAndSwap(pointer, data, expected);
-        }
-        default:
-            UNREACHABLE();
-        }
         return true;
     }
 
     bool WriteExclusive128(VAddr vaddr, const u128 data, const u128 expected) {
-        // AARCH64 masks the upper 16 bit of all memory accesses
-        vaddr &= 0xffffffffffffLL;
-
-        if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) {
-            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8,
-                      static_cast<u32>(data[0]), vaddr);
-            return true;
-        }
-
-        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
-        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
-            // NOTE: Avoid adding any extra logic to this fast-path block
-            const auto volatile_pointer = reinterpret_cast<volatile u64*>(&pointer[vaddr]);
+        u8* const ptr = GetPointerImpl(
+            vaddr,
+            [vaddr, data]() {
+                LOG_ERROR(HW_Memory, "Unmapped WriteExclusive128 @ 0x{:016X} = 0x{:016X}{:016X}",
+                          vaddr, static_cast<u64>(data[1]), static_cast<u64>(data[0]));
+            },
+            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(u128)); });
+        if (ptr) {
+            const auto volatile_pointer = reinterpret_cast<volatile u64*>(ptr);
             return Common::AtomicCompareAndSwap(volatile_pointer, data, expected);
         }
-        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
-        case Common::PageType::Unmapped:
-            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}{:016X}", sizeof(data) * 8,
-                      static_cast<u64>(data[1]), static_cast<u64>(data[0]), vaddr);
-            return true;
-        case Common::PageType::Memory:
-            ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr);
-            break;
-        case Common::PageType::RasterizerCachedMemory: {
-            u8* host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)};
-            system.GPU().InvalidateRegion(vaddr, sizeof(u128));
-            auto* pointer = reinterpret_cast<volatile u64*>(&host_ptr);
-            return Common::AtomicCompareAndSwap(pointer, data, expected);
-        }
-        default:
-            UNREACHABLE();
-        }
         return true;
     }
 
@@ -789,12 +584,11 @@ void Memory::UnmapRegion(Common::PageTable& page_table, VAddr base, u64 size) {
     impl->UnmapRegion(page_table, base, size);
 }
 
-bool Memory::IsValidVirtualAddress(const Kernel::KProcess& process, const VAddr vaddr) const {
-    return impl->IsValidVirtualAddress(process, vaddr);
-}
-
 bool Memory::IsValidVirtualAddress(const VAddr vaddr) const {
-    return impl->IsValidVirtualAddress(vaddr);
+    const Kernel::KProcess& process = *system.CurrentProcess();
+    const auto& page_table = process.PageTable().PageTableImpl();
+    const auto [pointer, type] = page_table.pointers[vaddr >> PAGE_BITS].PointerType();
+    return pointer != nullptr || type == Common::PageType::RasterizerCachedMemory;
 }
 
 u8* Memory::GetPointer(VAddr vaddr) {
@@ -863,64 +657,38 @@ std::string Memory::ReadCString(VAddr vaddr, std::size_t max_length) {
 
 void Memory::ReadBlock(const Kernel::KProcess& process, const VAddr src_addr, void* dest_buffer,
                        const std::size_t size) {
-    impl->ReadBlock(process, src_addr, dest_buffer, size);
+    impl->ReadBlockImpl<false>(process, src_addr, dest_buffer, size);
 }
 
 void Memory::ReadBlock(const VAddr src_addr, void* dest_buffer, const std::size_t size) {
     impl->ReadBlock(src_addr, dest_buffer, size);
 }
 
-void Memory::ReadBlockUnsafe(const Kernel::KProcess& process, const VAddr src_addr,
-                             void* dest_buffer, const std::size_t size) {
-    impl->ReadBlockUnsafe(process, src_addr, dest_buffer, size);
-}
-
 void Memory::ReadBlockUnsafe(const VAddr src_addr, void* dest_buffer, const std::size_t size) {
     impl->ReadBlockUnsafe(src_addr, dest_buffer, size);
 }
 
 void Memory::WriteBlock(const Kernel::KProcess& process, VAddr dest_addr, const void* src_buffer,
                         std::size_t size) {
-    impl->WriteBlock(process, dest_addr, src_buffer, size);
+    impl->WriteBlockImpl<false>(process, dest_addr, src_buffer, size);
 }
 
 void Memory::WriteBlock(const VAddr dest_addr, const void* src_buffer, const std::size_t size) {
     impl->WriteBlock(dest_addr, src_buffer, size);
 }
 
-void Memory::WriteBlockUnsafe(const Kernel::KProcess& process, VAddr dest_addr,
-                              const void* src_buffer, std::size_t size) {
-    impl->WriteBlockUnsafe(process, dest_addr, src_buffer, size);
-}
-
 void Memory::WriteBlockUnsafe(const VAddr dest_addr, const void* src_buffer,
                               const std::size_t size) {
     impl->WriteBlockUnsafe(dest_addr, src_buffer, size);
 }
 
-void Memory::ZeroBlock(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size) {
-    impl->ZeroBlock(process, dest_addr, size);
-}
-
-void Memory::ZeroBlock(VAddr dest_addr, std::size_t size) {
-    impl->ZeroBlock(dest_addr, size);
-}
-
 void Memory::CopyBlock(const Kernel::KProcess& process, VAddr dest_addr, VAddr src_addr,
                        const std::size_t size) {
     impl->CopyBlock(process, dest_addr, src_addr, size);
 }
 
-void Memory::CopyBlock(VAddr dest_addr, VAddr src_addr, std::size_t size) {
-    impl->CopyBlock(dest_addr, src_addr, size);
-}
-
 void Memory::RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
     impl->RasterizerMarkRegionCached(vaddr, size, cached);
 }
 
-bool IsKernelVirtualAddress(const VAddr vaddr) {
-    return KERNEL_REGION_VADDR <= vaddr && vaddr < KERNEL_REGION_END;
-}
-
 } // namespace Core::Memory

src/core/memory.h

@@ -39,11 +39,6 @@ enum : VAddr {
 
     /// Application stack
     DEFAULT_STACK_SIZE = 0x100000,
-
-    /// Kernel Virtual Address Range
-    KERNEL_REGION_VADDR = 0xFFFFFF8000000000,
-    KERNEL_REGION_SIZE = 0x7FFFE00000,
-    KERNEL_REGION_END = KERNEL_REGION_VADDR + KERNEL_REGION_SIZE,
 };
 
 /// Central class that handles all memory operations and state.
@@ -56,7 +51,7 @@ public:
     Memory& operator=(const Memory&) = delete;
 
     Memory(Memory&&) = default;
-    Memory& operator=(Memory&&) = default;
+    Memory& operator=(Memory&&) = delete;
 
     /**
      * Resets the state of the Memory system.
@@ -90,17 +85,6 @@ public:
      */
     void UnmapRegion(Common::PageTable& page_table, VAddr base, u64 size);
 
-    /**
-     * Checks whether or not the supplied address is a valid virtual
-     * address for the given process.
-     *
-     * @param process The emulated process to check the address against.
-     * @param vaddr   The virtual address to check the validity of.
-     *
-     * @returns True if the given virtual address is valid, false otherwise.
-     */
-    bool IsValidVirtualAddress(const Kernel::KProcess& process, VAddr vaddr) const;
-
     /**
      * Checks whether or not the supplied address is a valid virtual
      * address for the current process.
@@ -109,7 +93,7 @@ public:
      *
      * @returns True if the given virtual address is valid, false otherwise.
      */
-    bool IsValidVirtualAddress(VAddr vaddr) const;
+    [[nodiscard]] bool IsValidVirtualAddress(VAddr vaddr) const;
 
     /**
      * Gets a pointer to the given address.
@@ -134,7 +118,7 @@ public:
      * @returns The pointer to the given address, if the address is valid.
      *          If the address is not valid, nullptr will be returned.
      */
-    const u8* GetPointer(VAddr vaddr) const;
+    [[nodiscard]] const u8* GetPointer(VAddr vaddr) const;
 
     template <typename T>
     const T* GetPointer(VAddr vaddr) const {
@@ -327,27 +311,6 @@ public:
     void ReadBlock(const Kernel::KProcess& process, VAddr src_addr, void* dest_buffer,
                    std::size_t size);
 
-    /**
-     * Reads a contiguous block of bytes from a specified process' address space.
-     * This unsafe version does not trigger GPU flushing.
-     *
-     * @param process     The process to read the data from.
-     * @param src_addr    The virtual address to begin reading from.
-     * @param dest_buffer The buffer to place the read bytes into.
-     * @param size        The amount of data to read, in bytes.
-     *
-     * @note If a size of 0 is specified, then this function reads nothing and
-     *       no attempts to access memory are made at all.
-     *
-     * @pre dest_buffer must be at least size bytes in length, otherwise a
-     *      buffer overrun will occur.
-     *
-     * @post The range [dest_buffer, size) contains the read bytes from the
-     *       process' address space.
-     */
-    void ReadBlockUnsafe(const Kernel::KProcess& process, VAddr src_addr, void* dest_buffer,
-                         std::size_t size);
-
     /**
      * Reads a contiguous block of bytes from the current process' address space.
      *
@@ -408,26 +371,6 @@ public:
     void WriteBlock(const Kernel::KProcess& process, VAddr dest_addr, const void* src_buffer,
                     std::size_t size);
 
-    /**
-     * Writes a range of bytes into a given process' address space at the specified
-     * virtual address.
-     * This unsafe version does not invalidate GPU Memory.
-     *
-     * @param process    The process to write data into the address space of.
-     * @param dest_addr  The destination virtual address to begin writing the data at.
-     * @param src_buffer The data to write into the process' address space.
-     * @param size       The size of the data to write, in bytes.
-     *
-     * @post The address range [dest_addr, size) in the process' address space
-     *       contains the data that was within src_buffer.
-     *
-     * @post If an attempt is made to write into an unmapped region of memory, the writes
-     *       will be ignored and an error will be logged.
-     *
-     */
-    void WriteBlockUnsafe(const Kernel::KProcess& process, VAddr dest_addr, const void* src_buffer,
-                          std::size_t size);
-
     /**
      * Writes a range of bytes into the current process' address space at the specified
      * virtual address.
@@ -467,29 +410,6 @@ public:
      */
     void WriteBlockUnsafe(VAddr dest_addr, const void* src_buffer, std::size_t size);
 
-    /**
-     * Fills the specified address range within a process' address space with zeroes.
-     *
-     * @param process   The process that will have a portion of its memory zeroed out.
-     * @param dest_addr The starting virtual address of the range to zero out.
-     * @param size      The size of the address range to zero out, in bytes.
-     *
-     * @post The range [dest_addr, size) within the process' address space is
-     *       filled with zeroes.
-     */
-    void ZeroBlock(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size);
-
-    /**
-     * Fills the specified address range within the current process' address space with zeroes.
-     *
-     * @param dest_addr The starting virtual address of the range to zero out.
-     * @param size      The size of the address range to zero out, in bytes.
-     *
-     * @post The range [dest_addr, size) within the current process' address space is
-     *       filled with zeroes.
-     */
-    void ZeroBlock(VAddr dest_addr, std::size_t size);
-
     /**
      * Copies data within a process' address space to another location within the
      * same address space.
@@ -505,19 +425,6 @@ public:
     void CopyBlock(const Kernel::KProcess& process, VAddr dest_addr, VAddr src_addr,
                    std::size_t size);
 
-    /**
-     * Copies data within the current process' address space to another location within the
-     * same address space.
-     *
-     * @param dest_addr The destination virtual address to begin copying the data into.
-     * @param src_addr  The source virtual address to begin copying the data from.
-     * @param size      The size of the data to copy, in bytes.
-     *
-     * @post The range [dest_addr, size) within the current process' address space
-     *       contains the same data within the range [src_addr, size).
-     */
-    void CopyBlock(VAddr dest_addr, VAddr src_addr, std::size_t size);
-
     /**
      * Marks each page within the specified address range as cached or uncached.
      *
@@ -535,7 +442,4 @@ private:
     std::unique_ptr<Impl> impl;
 };
 
-/// Determines if the given VAddr is a kernel address
-bool IsKernelVirtualAddress(VAddr vaddr);
-
 } // namespace Core::Memory

src/core/network/network.cpp

@@ -7,12 +7,14 @@
 #include <limits>
 #include <utility>
 #include <vector>
-#include "common/common_funcs.h"
+
+#include "common/error.h"
 
 #ifdef _WIN32
-#define _WINSOCK_DEPRECATED_NO_WARNINGS // gethostname
 #include <winsock2.h>
+#include <ws2tcpip.h>
 #elif YUZU_UNIX
+#include <arpa/inet.h>
 #include <errno.h>
 #include <fcntl.h>
 #include <netdb.h>
@@ -27,7 +29,9 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
+#include "common/settings.h"
 #include "core/network/network.h"
+#include "core/network/network_interface.h"
 #include "core/network/sockets.h"
 
 namespace Network {
@@ -47,11 +51,6 @@ void Finalize() {
     WSACleanup();
 }
 
-constexpr IPv4Address TranslateIPv4(in_addr addr) {
-    auto& bytes = addr.S_un.S_un_b;
-    return IPv4Address{bytes.s_b1, bytes.s_b2, bytes.s_b3, bytes.s_b4};
-}
-
 sockaddr TranslateFromSockAddrIn(SockAddrIn input) {
     sockaddr_in result;
 
@@ -138,12 +137,6 @@ void Initialize() {}
 
 void Finalize() {}
 
-constexpr IPv4Address TranslateIPv4(in_addr addr) {
-    const u32 bytes = addr.s_addr;
-    return IPv4Address{static_cast<u8>(bytes), static_cast<u8>(bytes >> 8),
-                       static_cast<u8>(bytes >> 16), static_cast<u8>(bytes >> 24)};
-}
-
 sockaddr TranslateFromSockAddrIn(SockAddrIn input) {
     sockaddr_in result;
 
@@ -182,7 +175,7 @@ linger MakeLinger(bool enable, u32 linger_value) {
 }
 
 bool EnableNonBlock(int fd, bool enable) {
-    int flags = fcntl(fd, F_GETFD);
+    int flags = fcntl(fd, F_GETFL);
     if (flags == -1) {
         return false;
     }
@@ -191,7 +184,7 @@ bool EnableNonBlock(int fd, bool enable) {
     } else {
         flags &= ~O_NONBLOCK;
     }
-    return fcntl(fd, F_SETFD, flags) == 0;
+    return fcntl(fd, F_SETFL, flags) == 0;
 }
 
 Errno TranslateNativeError(int e) {
@@ -227,8 +220,12 @@ Errno GetAndLogLastError() {
 #else
     int e = errno;
 #endif
-    LOG_ERROR(Network, "Socket operation error: {}", NativeErrorToString(e));
-    return TranslateNativeError(e);
+    const Errno err = TranslateNativeError(e);
+    if (err == Errno::AGAIN) {
+        return err;
+    }
+    LOG_ERROR(Network, "Socket operation error: {}", Common::NativeErrorToString(e));
+    return err;
 }
 
 int TranslateDomain(Domain domain) {
@@ -353,27 +350,29 @@ NetworkInstance::~NetworkInstance() {
     Finalize();
 }
 
-std::pair<IPv4Address, Errno> GetHostIPv4Address() {
-    std::array<char, 256> name{};
-    if (gethostname(name.data(), static_cast<int>(name.size()) - 1) == SOCKET_ERROR) {
-        return {IPv4Address{}, GetAndLogLastError()};
+std::optional<IPv4Address> GetHostIPv4Address() {
+    const std::string& selected_network_interface = Settings::values.network_interface.GetValue();
+    const auto network_interfaces = Network::GetAvailableNetworkInterfaces();
+    if (network_interfaces.size() == 0) {
+        LOG_ERROR(Network, "GetAvailableNetworkInterfaces returned no interfaces");
+        return {};
     }
 
-    hostent* const ent = gethostbyname(name.data());
-    if (!ent) {
-        return {IPv4Address{}, GetAndLogLastError()};
-    }
-    if (ent->h_addr_list == nullptr) {
-        UNIMPLEMENTED_MSG("No addr provided in hostent->h_addr_list");
-        return {IPv4Address{}, Errno::SUCCESS};
-    }
-    if (ent->h_length != sizeof(in_addr)) {
-        UNIMPLEMENTED_MSG("Unexpected size={} in hostent->h_length", ent->h_length);
-    }
+    const auto res =
+        std::ranges::find_if(network_interfaces, [&selected_network_interface](const auto& iface) {
+            return iface.name == selected_network_interface;
+        });
 
-    in_addr addr;
-    std::memcpy(&addr, ent->h_addr_list[0], sizeof(addr));
-    return {TranslateIPv4(addr), Errno::SUCCESS};
+    if (res != network_interfaces.end()) {
+        char ip_addr[16] = {};
+        ASSERT(inet_ntop(AF_INET, &res->ip_address, ip_addr, sizeof(ip_addr)) != nullptr);
+        LOG_INFO(Network, "IP address: {}", ip_addr);
+
+        return TranslateIPv4(res->ip_address);
+    } else {
+        LOG_ERROR(Network, "Couldn't find selected interface \"{}\"", selected_network_interface);
+        return {};
+    }
 }
 
 std::pair<s32, Errno> Poll(std::vector<PollFD>& pollfds, s32 timeout) {
@@ -570,7 +569,7 @@ std::pair<s32, Errno> Socket::SendTo(u32 flags, const std::vector<u8>& message,
     ASSERT(flags == 0);
 
     const sockaddr* to = nullptr;
-    const int tolen = addr ? 0 : sizeof(sockaddr);
+    const int tolen = addr ? sizeof(sockaddr) : 0;
     sockaddr host_addr_in;
 
     if (addr) {

src/core/network/network.h

@@ -5,11 +5,18 @@
 #pragma once
 
 #include <array>
+#include <optional>
 #include <utility>
 
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 
+#ifdef _WIN32
+#include <winsock2.h>
+#elif YUZU_UNIX
+#include <netinet/in.h>
+#endif
+
 namespace Network {
 
 class Socket;
@@ -92,8 +99,21 @@ public:
     ~NetworkInstance();
 };
 
+#ifdef _WIN32
+constexpr IPv4Address TranslateIPv4(in_addr addr) {
+    auto& bytes = addr.S_un.S_un_b;
+    return IPv4Address{bytes.s_b1, bytes.s_b2, bytes.s_b3, bytes.s_b4};
+}
+#elif YUZU_UNIX
+constexpr IPv4Address TranslateIPv4(in_addr addr) {
+    const u32 bytes = addr.s_addr;
+    return IPv4Address{static_cast<u8>(bytes), static_cast<u8>(bytes >> 8),
+                       static_cast<u8>(bytes >> 16), static_cast<u8>(bytes >> 24)};
+}
+#endif
+
 /// @brief Returns host's IPv4 address
-/// @return Pair of an array of human ordered IPv4 address (e.g. 192.168.0.1) and an error code
-std::pair<IPv4Address, Errno> GetHostIPv4Address();
+/// @return human ordered IPv4 address (e.g. 192.168.0.1) as an array
+std::optional<IPv4Address> GetHostIPv4Address();
 
 } // namespace Network

src/core/network/network_interface.cpp

@@ -0,0 +1,210 @@
+// Copyright 2021 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <fstream>
+#include <sstream>
+#include <vector>
+
+#include "common/bit_cast.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "common/settings.h"
+#include "common/string_util.h"
+#include "core/network/network_interface.h"
+
+#ifdef _WIN32
+#include <iphlpapi.h>
+#else
+#include <cerrno>
+#include <ifaddrs.h>
+#include <net/if.h>
+#endif
+
+namespace Network {
+
+#ifdef _WIN32
+
+std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
+    std::vector<IP_ADAPTER_ADDRESSES> adapter_addresses;
+    DWORD ret = ERROR_BUFFER_OVERFLOW;
+    DWORD buf_size = 0;
+
+    // retry up to 5 times
+    for (int i = 0; i < 5 && ret == ERROR_BUFFER_OVERFLOW; i++) {
+        ret = GetAdaptersAddresses(
+            AF_INET, GAA_FLAG_SKIP_MULTICAST | GAA_FLAG_SKIP_DNS_SERVER | GAA_FLAG_INCLUDE_GATEWAYS,
+            nullptr, adapter_addresses.data(), &buf_size);
+
+        if (ret != ERROR_BUFFER_OVERFLOW) {
+            break;
+        }
+
+        adapter_addresses.resize((buf_size / sizeof(IP_ADAPTER_ADDRESSES)) + 1);
+    }
+
+    if (ret != NO_ERROR) {
+        LOG_ERROR(Network, "Failed to get network interfaces with GetAdaptersAddresses");
+        return {};
+    }
+
+    std::vector<NetworkInterface> result;
+
+    for (auto current_address = adapter_addresses.data(); current_address != nullptr;
+         current_address = current_address->Next) {
+        if (current_address->FirstUnicastAddress == nullptr ||
+            current_address->FirstUnicastAddress->Address.lpSockaddr == nullptr) {
+            continue;
+        }
+
+        if (current_address->OperStatus != IfOperStatusUp) {
+            continue;
+        }
+
+        const auto ip_addr = Common::BitCast<struct sockaddr_in>(
+                                 *current_address->FirstUnicastAddress->Address.lpSockaddr)
+                                 .sin_addr;
+
+        ULONG mask = 0;
+        if (ConvertLengthToIpv4Mask(current_address->FirstUnicastAddress->OnLinkPrefixLength,
+                                    &mask) != NO_ERROR) {
+            LOG_ERROR(Network, "Failed to convert IPv4 prefix length to subnet mask");
+            continue;
+        }
+
+        struct in_addr gateway = {.S_un{.S_addr{0}}};
+        if (current_address->FirstGatewayAddress != nullptr &&
+            current_address->FirstGatewayAddress->Address.lpSockaddr != nullptr) {
+            gateway = Common::BitCast<struct sockaddr_in>(
+                          *current_address->FirstGatewayAddress->Address.lpSockaddr)
+                          .sin_addr;
+        }
+
+        result.emplace_back(NetworkInterface{
+            .name{Common::UTF16ToUTF8(std::wstring{current_address->FriendlyName})},
+            .ip_address{ip_addr},
+            .subnet_mask = in_addr{.S_un{.S_addr{mask}}},
+            .gateway = gateway});
+    }
+
+    return result;
+}
+
+#else
+
+std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
+    struct ifaddrs* ifaddr = nullptr;
+
+    if (getifaddrs(&ifaddr) != 0) {
+        LOG_ERROR(Network, "Failed to get network interfaces with getifaddrs: {}",
+                  std::strerror(errno));
+        return {};
+    }
+
+    std::vector<NetworkInterface> result;
+
+    for (auto ifa = ifaddr; ifa != nullptr; ifa = ifa->ifa_next) {
+        if (ifa->ifa_addr == nullptr || ifa->ifa_netmask == nullptr) {
+            continue;
+        }
+
+        if (ifa->ifa_addr->sa_family != AF_INET) {
+            continue;
+        }
+
+        if ((ifa->ifa_flags & IFF_UP) == 0 || (ifa->ifa_flags & IFF_LOOPBACK) != 0) {
+            continue;
+        }
+
+        u32 gateway{};
+
+        std::ifstream file{"/proc/net/route"};
+        if (!file.is_open()) {
+            LOG_ERROR(Network, "Failed to open \"/proc/net/route\"");
+
+            result.emplace_back(NetworkInterface{
+                .name{ifa->ifa_name},
+                .ip_address{Common::BitCast<struct sockaddr_in>(*ifa->ifa_addr).sin_addr},
+                .subnet_mask{Common::BitCast<struct sockaddr_in>(*ifa->ifa_netmask).sin_addr},
+                .gateway{in_addr{.s_addr = gateway}}});
+            continue;
+        }
+
+        // ignore header
+        file.ignore(std::numeric_limits<std::streamsize>::max(), '\n');
+
+        bool gateway_found = false;
+
+        for (std::string line; std::getline(file, line);) {
+            std::istringstream iss{line};
+
+            std::string iface_name;
+            iss >> iface_name;
+            if (iface_name != ifa->ifa_name) {
+                continue;
+            }
+
+            iss >> std::hex;
+
+            u32 dest{};
+            iss >> dest;
+            if (dest != 0) {
+                // not the default route
+                continue;
+            }
+
+            iss >> gateway;
+
+            u16 flags{};
+            iss >> flags;
+
+            // flag RTF_GATEWAY (defined in <linux/route.h>)
+            if ((flags & 0x2) == 0) {
+                continue;
+            }
+
+            gateway_found = true;
+            break;
+        }
+
+        if (!gateway_found) {
+            gateway = 0;
+        }
+
+        result.emplace_back(NetworkInterface{
+            .name{ifa->ifa_name},
+            .ip_address{Common::BitCast<struct sockaddr_in>(*ifa->ifa_addr).sin_addr},
+            .subnet_mask{Common::BitCast<struct sockaddr_in>(*ifa->ifa_netmask).sin_addr},
+            .gateway{in_addr{.s_addr = gateway}}});
+    }
+
+    freeifaddrs(ifaddr);
+
+    return result;
+}
+
+#endif
+
+std::optional<NetworkInterface> GetSelectedNetworkInterface() {
+    const auto& selected_network_interface = Settings::values.network_interface.GetValue();
+    const auto network_interfaces = Network::GetAvailableNetworkInterfaces();
+    if (network_interfaces.size() == 0) {
+        LOG_ERROR(Network, "GetAvailableNetworkInterfaces returned no interfaces");
+        return std::nullopt;
+    }
+
+    const auto res =
+        std::ranges::find_if(network_interfaces, [&selected_network_interface](const auto& iface) {
+            return iface.name == selected_network_interface;
+        });
+
+    if (res == network_interfaces.end()) {
+        LOG_ERROR(Network, "Couldn't find selected interface \"{}\"", selected_network_interface);
+        return std::nullopt;
+    }
+
+    return *res;
+}
+
+} // namespace Network

src/core/network/network_interface.h

@@ -0,0 +1,29 @@
+// Copyright 2021 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <optional>
+#include <string>
+#include <vector>
+
+#ifdef _WIN32
+#include <winsock2.h>
+#else
+#include <netinet/in.h>
+#endif
+
+namespace Network {
+
+struct NetworkInterface {
+    std::string name;
+    struct in_addr ip_address;
+    struct in_addr subnet_mask;
+    struct in_addr gateway;
+};
+
+std::vector<NetworkInterface> GetAvailableNetworkInterfaces();
+std::optional<NetworkInterface> GetSelectedNetworkInterface();
+
+} // namespace Network

src/core/telemetry_session.cpp

@@ -72,6 +72,18 @@ static const char* TranslateGPUAccuracyLevel(Settings::GPUAccuracy backend) {
     return "Unknown";
 }
 
+static const char* TranslateNvdecEmulation(Settings::NvdecEmulation backend) {
+    switch (backend) {
+    case Settings::NvdecEmulation::Off:
+        return "Off";
+    case Settings::NvdecEmulation::CPU:
+        return "CPU";
+    case Settings::NvdecEmulation::GPU:
+        return "GPU";
+    }
+    return "Unknown";
+}
+
 u64 GetTelemetryId() {
     u64 telemetry_id{};
     const auto filename = Common::FS::GetYuzuPath(Common::FS::YuzuPath::ConfigDir) / "telemetry_id";
@@ -229,8 +241,8 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader,
              TranslateGPUAccuracyLevel(Settings::values.gpu_accuracy.GetValue()));
     AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
              Settings::values.use_asynchronous_gpu_emulation.GetValue());
-    AddField(field_type, "Renderer_UseNvdecEmulation",
-             Settings::values.use_nvdec_emulation.GetValue());
+    AddField(field_type, "Renderer_NvdecEmulation",
+             TranslateNvdecEmulation(Settings::values.nvdec_emulation.GetValue()));
     AddField(field_type, "Renderer_AccelerateASTC", Settings::values.accelerate_astc.GetValue());
     AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync.GetValue());
     AddField(field_type, "Renderer_ShaderBackend",

src/input_common/main.cpp

@@ -304,10 +304,10 @@ std::vector<std::unique_ptr<Polling::DevicePoller>> InputSubsystem::GetPollers([
 }
 
 std::string GenerateKeyboardParam(int key_code) {
-    Common::ParamPackage param{
-        {"engine", "keyboard"},
-        {"code", std::to_string(key_code)},
-    };
+    Common::ParamPackage param;
+    param.Set("engine", "keyboard");
+    param.Set("code", key_code);
+    param.Set("toggle", false);
     return param.Serialize();
 }
 

src/input_common/mouse/mouse_poller.cpp

@@ -57,6 +57,7 @@ Common::ParamPackage MouseButtonFactory::GetNextInput() const {
         if (pad.button != MouseInput::MouseButton::Undefined) {
             params.Set("engine", "mouse");
             params.Set("button", static_cast<u16>(pad.button));
+            params.Set("toggle", false);
             return params;
         }
     }

src/input_common/sdl/sdl_impl.cpp

@@ -82,6 +82,12 @@ public:
         state.buttons.insert_or_assign(button, value);
     }
 
+    void PreSetButton(int button) {
+        if (!state.buttons.contains(button)) {
+            SetButton(button, false);
+        }
+    }
+
     void SetMotion(SDL_ControllerSensorEvent event) {
         constexpr float gravity_constant = 9.80665f;
         std::lock_guard lock{mutex};
@@ -155,9 +161,16 @@ public:
         state.axes.insert_or_assign(axis, value);
     }
 
-    float GetAxis(int axis, float range) const {
+    void PreSetAxis(int axis) {
+        if (!state.axes.contains(axis)) {
+            SetAxis(axis, 0);
+        }
+    }
+
+    float GetAxis(int axis, float range, float offset) const {
         std::lock_guard lock{mutex};
-        return static_cast<float>(state.axes.at(axis)) / (32767.0f * range);
+        const float value = static_cast<float>(state.axes.at(axis)) / 32767.0f;
+        return (value + offset) / range;
     }
 
     bool RumblePlay(u16 amp_low, u16 amp_high) {
@@ -174,9 +187,10 @@ public:
         return false;
     }
 
-    std::tuple<float, float> GetAnalog(int axis_x, int axis_y, float range) const {
-        float x = GetAxis(axis_x, range);
-        float y = GetAxis(axis_y, range);
+    std::tuple<float, float> GetAnalog(int axis_x, int axis_y, float range, float offset_x,
+                                       float offset_y) const {
+        float x = GetAxis(axis_x, range, offset_x);
+        float y = GetAxis(axis_y, range, offset_y);
         y = -y; // 3DS uses an y-axis inverse from SDL
 
         // Make sure the coordinates are in the unit circle,
@@ -483,7 +497,7 @@ public:
           trigger_if_greater(trigger_if_greater_) {}
 
     bool GetStatus() const override {
-        const float axis_value = joystick->GetAxis(axis, 1.0f);
+        const float axis_value = joystick->GetAxis(axis, 1.0f, 0.0f);
         if (trigger_if_greater) {
             return axis_value > threshold;
         }
@@ -500,12 +514,14 @@ private:
 class SDLAnalog final : public Input::AnalogDevice {
 public:
     explicit SDLAnalog(std::shared_ptr<SDLJoystick> joystick_, int axis_x_, int axis_y_,
-                       bool invert_x_, bool invert_y_, float deadzone_, float range_)
+                       bool invert_x_, bool invert_y_, float deadzone_, float range_,
+                       float offset_x_, float offset_y_)
         : joystick(std::move(joystick_)), axis_x(axis_x_), axis_y(axis_y_), invert_x(invert_x_),
-          invert_y(invert_y_), deadzone(deadzone_), range(range_) {}
+          invert_y(invert_y_), deadzone(deadzone_), range(range_), offset_x(offset_x_),
+          offset_y(offset_y_) {}
 
     std::tuple<float, float> GetStatus() const override {
-        auto [x, y] = joystick->GetAnalog(axis_x, axis_y, range);
+        auto [x, y] = joystick->GetAnalog(axis_x, axis_y, range, offset_x, offset_y);
         const float r = std::sqrt((x * x) + (y * y));
         if (invert_x) {
             x = -x;
@@ -522,8 +538,8 @@ public:
     }
 
     std::tuple<float, float> GetRawStatus() const override {
-        const float x = joystick->GetAxis(axis_x, range);
-        const float y = joystick->GetAxis(axis_y, range);
+        const float x = joystick->GetAxis(axis_x, range, offset_x);
+        const float y = joystick->GetAxis(axis_y, range, offset_y);
         return {x, -y};
     }
 
@@ -555,6 +571,8 @@ private:
     const bool invert_y;
     const float deadzone;
     const float range;
+    const float offset_x;
+    const float offset_y;
 };
 
 class SDLVibration final : public Input::VibrationDevice {
@@ -621,7 +639,7 @@ public:
           trigger_if_greater(trigger_if_greater_) {}
 
     Input::MotionStatus GetStatus() const override {
-        const float axis_value = joystick->GetAxis(axis, 1.0f);
+        const float axis_value = joystick->GetAxis(axis, 1.0f, 0.0f);
         bool trigger = axis_value < threshold;
         if (trigger_if_greater) {
             trigger = axis_value > threshold;
@@ -720,13 +738,13 @@ public:
                 LOG_ERROR(Input, "Unknown direction {}", direction_name);
             }
             // This is necessary so accessing GetAxis with axis won't crash
-            joystick->SetAxis(axis, 0);
+            joystick->PreSetAxis(axis);
             return std::make_unique<SDLAxisButton>(joystick, axis, threshold, trigger_if_greater);
         }
 
         const int button = params.Get("button", 0);
         // This is necessary so accessing GetButton with button won't crash
-        joystick->SetButton(button, false);
+        joystick->PreSetButton(button);
         return std::make_unique<SDLButton>(joystick, button, toggle);
     }
 
@@ -757,13 +775,15 @@ public:
         const std::string invert_y_value = params.Get("invert_y", "+");
         const bool invert_x = invert_x_value == "-";
         const bool invert_y = invert_y_value == "-";
+        const float offset_x = params.Get("offset_x", 0.0f);
+        const float offset_y = params.Get("offset_y", 0.0f);
         auto joystick = state.GetSDLJoystickByGUID(guid, port);
 
         // This is necessary so accessing GetAxis with axis_x and axis_y won't crash
-        joystick->SetAxis(axis_x, 0);
-        joystick->SetAxis(axis_y, 0);
+        joystick->PreSetAxis(axis_x);
+        joystick->PreSetAxis(axis_y);
         return std::make_unique<SDLAnalog>(joystick, axis_x, axis_y, invert_x, invert_y, deadzone,
-                                           range);
+                                           range, offset_x, offset_y);
     }
 
 private:
@@ -844,13 +864,13 @@ public:
                 LOG_ERROR(Input, "Unknown direction {}", direction_name);
             }
             // This is necessary so accessing GetAxis with axis won't crash
-            joystick->SetAxis(axis, 0);
+            joystick->PreSetAxis(axis);
             return std::make_unique<SDLAxisMotion>(joystick, axis, threshold, trigger_if_greater);
         }
 
         const int button = params.Get("button", 0);
         // This is necessary so accessing GetButton with button won't crash
-        joystick->SetButton(button, false);
+        joystick->PreSetButton(button);
         return std::make_unique<SDLButtonMotion>(joystick, button);
     }
 
@@ -869,6 +889,9 @@ SDLState::SDLState() {
     RegisterFactory<VibrationDevice>("sdl", vibration_factory);
     RegisterFactory<MotionDevice>("sdl", motion_factory);
 
+    // Disable raw input. When enabled this setting causes SDL to die when a web applet opens
+    SDL_SetHint(SDL_HINT_JOYSTICK_RAWINPUT, "0");
+
     // Enable HIDAPI rumble. This prevents SDL from disabling motion on PS4 and PS5 controllers
     SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_PS4_RUMBLE, "1");
     SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_PS5_RUMBLE, "1");
@@ -995,6 +1018,7 @@ Common::ParamPackage BuildButtonParamPackageForButton(int port, std::string guid
     params.Set("port", port);
     params.Set("guid", std::move(guid));
     params.Set("button", button);
+    params.Set("toggle", false);
     return params;
 }
 
@@ -1134,13 +1158,15 @@ Common::ParamPackage BuildParamPackageForBinding(int port, const std::string& gu
 }
 
 Common::ParamPackage BuildParamPackageForAnalog(int port, const std::string& guid, int axis_x,
-                                                int axis_y) {
+                                                int axis_y, float offset_x, float offset_y) {
     Common::ParamPackage params;
     params.Set("engine", "sdl");
     params.Set("port", port);
     params.Set("guid", guid);
     params.Set("axis_x", axis_x);
     params.Set("axis_y", axis_y);
+    params.Set("offset_x", offset_x);
+    params.Set("offset_y", offset_y);
     params.Set("invert_x", "+");
     params.Set("invert_y", "+");
     return params;
@@ -1342,24 +1368,39 @@ AnalogMapping SDLState::GetAnalogMappingForDevice(const Common::ParamPackage& pa
     const auto& binding_left_y =
         SDL_GameControllerGetBindForAxis(controller, SDL_CONTROLLER_AXIS_LEFTY);
     if (params.Has("guid2")) {
+        joystick2->PreSetAxis(binding_left_x.value.axis);
+        joystick2->PreSetAxis(binding_left_y.value.axis);
+        const auto left_offset_x = -joystick2->GetAxis(binding_left_x.value.axis, 1.0f, 0);
+        const auto left_offset_y = -joystick2->GetAxis(binding_left_y.value.axis, 1.0f, 0);
         mapping.insert_or_assign(
             Settings::NativeAnalog::LStick,
             BuildParamPackageForAnalog(joystick2->GetPort(), joystick2->GetGUID(),
-                                       binding_left_x.value.axis, binding_left_y.value.axis));
+                                       binding_left_x.value.axis, binding_left_y.value.axis,
+                                       left_offset_x, left_offset_y));
     } else {
+        joystick->PreSetAxis(binding_left_x.value.axis);
+        joystick->PreSetAxis(binding_left_y.value.axis);
+        const auto left_offset_x = -joystick->GetAxis(binding_left_x.value.axis, 1.0f, 0);
+        const auto left_offset_y = -joystick->GetAxis(binding_left_y.value.axis, 1.0f, 0);
         mapping.insert_or_assign(
             Settings::NativeAnalog::LStick,
             BuildParamPackageForAnalog(joystick->GetPort(), joystick->GetGUID(),
-                                       binding_left_x.value.axis, binding_left_y.value.axis));
+                                       binding_left_x.value.axis, binding_left_y.value.axis,
+                                       left_offset_x, left_offset_y));
     }
     const auto& binding_right_x =
         SDL_GameControllerGetBindForAxis(controller, SDL_CONTROLLER_AXIS_RIGHTX);
     const auto& binding_right_y =
         SDL_GameControllerGetBindForAxis(controller, SDL_CONTROLLER_AXIS_RIGHTY);
+    joystick->PreSetAxis(binding_right_x.value.axis);
+    joystick->PreSetAxis(binding_right_y.value.axis);
+    const auto right_offset_x = -joystick->GetAxis(binding_right_x.value.axis, 1.0f, 0);
+    const auto right_offset_y = -joystick->GetAxis(binding_right_y.value.axis, 1.0f, 0);
     mapping.insert_or_assign(Settings::NativeAnalog::RStick,
                              BuildParamPackageForAnalog(joystick->GetPort(), joystick->GetGUID(),
                                                         binding_right_x.value.axis,
-                                                        binding_right_y.value.axis));
+                                                        binding_right_y.value.axis, right_offset_x,
+                                                        right_offset_y));
     return mapping;
 }
 
@@ -1563,8 +1604,9 @@ public:
             }
 
             if (const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which)) {
+                // Set offset to zero since the joystick is not on center
                 auto params = BuildParamPackageForAnalog(joystick->GetPort(), joystick->GetGUID(),
-                                                         first_axis, axis);
+                                                         first_axis, axis, 0, 0);
                 first_axis = -1;
                 return params;
             }

src/shader_recompiler/backend/glsl/emit_glsl_warp.cpp

@@ -11,6 +11,8 @@
 
 namespace Shader::Backend::GLSL {
 namespace {
+constexpr char THREAD_ID[]{"gl_SubGroupInvocationARB"};
+
 void SetInBoundsFlag(EmitContext& ctx, IR::Inst& inst) {
     IR::Inst* const in_bounds{inst.GetAssociatedPseudoOperation(IR::Opcode::GetInBoundsFromOp)};
     if (!in_bounds) {
@@ -43,84 +45,100 @@ void UseShuffleNv(EmitContext& ctx, IR::Inst& inst, std::string_view shfl_op,
     ctx.AddU32("{}={}({},{},{},shfl_in_bounds);", inst, shfl_op, value, index, width);
     SetInBoundsFlag(ctx, inst);
 }
+
+std::string_view BallotIndex(EmitContext& ctx) {
+    if (!ctx.profile.warp_size_potentially_larger_than_guest) {
+        return ".x";
+    }
+    return "[gl_SubGroupInvocationARB>>5]";
+}
+
+std::string GetMask(EmitContext& ctx, std::string_view mask) {
+    const auto ballot_index{BallotIndex(ctx)};
+    return fmt::format("uint(uvec2({}){})", mask, ballot_index);
+}
 } // Anonymous namespace
 
 void EmitLaneId(EmitContext& ctx, IR::Inst& inst) {
-    ctx.AddU32("{}=gl_SubGroupInvocationARB&31u;", inst);
+    ctx.AddU32("{}={}&31u;", inst, THREAD_ID);
 }
 
 void EmitVoteAll(EmitContext& ctx, IR::Inst& inst, std::string_view pred) {
     if (!ctx.profile.warp_size_potentially_larger_than_guest) {
         ctx.AddU1("{}=allInvocationsEqualARB({});", inst, pred);
-    } else {
-        const auto active_mask{fmt::format("uvec2(ballotARB(true))[gl_SubGroupInvocationARB]")};
-        const auto ballot{fmt::format("uvec2(ballotARB({}))[gl_SubGroupInvocationARB]", pred)};
-        ctx.AddU1("{}=({}&{})=={};", inst, ballot, active_mask, active_mask);
+        return;
     }
+    const auto ballot_index{BallotIndex(ctx)};
+    const auto active_mask{fmt::format("uvec2(ballotARB(true)){}", ballot_index)};
+    const auto ballot{fmt::format("uvec2(ballotARB({})){}", pred, ballot_index)};
+    ctx.AddU1("{}=({}&{})=={};", inst, ballot, active_mask, active_mask);
 }
 
 void EmitVoteAny(EmitContext& ctx, IR::Inst& inst, std::string_view pred) {
     if (!ctx.profile.warp_size_potentially_larger_than_guest) {
         ctx.AddU1("{}=anyInvocationARB({});", inst, pred);
-    } else {
-        const auto active_mask{fmt::format("uvec2(ballotARB(true))[gl_SubGroupInvocationARB]")};
-        const auto ballot{fmt::format("uvec2(ballotARB({}))[gl_SubGroupInvocationARB]", pred)};
-        ctx.AddU1("{}=({}&{})!=0u;", inst, ballot, active_mask, active_mask);
+        return;
     }
+    const auto ballot_index{BallotIndex(ctx)};
+    const auto active_mask{fmt::format("uvec2(ballotARB(true)){}", ballot_index)};
+    const auto ballot{fmt::format("uvec2(ballotARB({})){}", pred, ballot_index)};
+    ctx.AddU1("{}=({}&{})!=0u;", inst, ballot, active_mask, active_mask);
 }
 
 void EmitVoteEqual(EmitContext& ctx, IR::Inst& inst, std::string_view pred) {
     if (!ctx.profile.warp_size_potentially_larger_than_guest) {
         ctx.AddU1("{}=allInvocationsEqualARB({});", inst, pred);
-    } else {
-        const auto active_mask{fmt::format("uvec2(ballotARB(true))[gl_SubGroupInvocationARB]")};
-        const auto ballot{fmt::format("uvec2(ballotARB({}))[gl_SubGroupInvocationARB]", pred)};
-        const auto value{fmt::format("({}^{})", ballot, active_mask)};
-        ctx.AddU1("{}=({}==0)||({}=={});", inst, value, value, active_mask);
+        return;
     }
+    const auto ballot_index{BallotIndex(ctx)};
+    const auto active_mask{fmt::format("uvec2(ballotARB(true)){}", ballot_index)};
+    const auto ballot{fmt::format("uvec2(ballotARB({})){}", pred, ballot_index)};
+    const auto value{fmt::format("({}^{})", ballot, active_mask)};
+    ctx.AddU1("{}=({}==0)||({}=={});", inst, value, value, active_mask);
 }
 
 void EmitSubgroupBallot(EmitContext& ctx, IR::Inst& inst, std::string_view pred) {
-    if (!ctx.profile.warp_size_potentially_larger_than_guest) {
-        ctx.AddU32("{}=uvec2(ballotARB({})).x;", inst, pred);
-    } else {
-        ctx.AddU32("{}=uvec2(ballotARB({}))[gl_SubGroupInvocationARB];", inst, pred);
-    }
+    const auto ballot_index{BallotIndex(ctx)};
+    ctx.AddU32("{}=uvec2(ballotARB({})){};", inst, pred, ballot_index);
 }
 
 void EmitSubgroupEqMask(EmitContext& ctx, IR::Inst& inst) {
-    ctx.AddU32("{}=uint(gl_SubGroupEqMaskARB.x);", inst);
+    ctx.AddU32("{}={};", inst, GetMask(ctx, "gl_SubGroupEqMaskARB"));
 }
 
 void EmitSubgroupLtMask(EmitContext& ctx, IR::Inst& inst) {
-    ctx.AddU32("{}=uint(gl_SubGroupLtMaskARB.x);", inst);
+    ctx.AddU32("{}={};", inst, GetMask(ctx, "gl_SubGroupLtMaskARB"));
 }
 
 void EmitSubgroupLeMask(EmitContext& ctx, IR::Inst& inst) {
-    ctx.AddU32("{}=uint(gl_SubGroupLeMaskARB.x);", inst);
+    ctx.AddU32("{}={};", inst, GetMask(ctx, "gl_SubGroupLeMaskARB"));
 }
 
 void EmitSubgroupGtMask(EmitContext& ctx, IR::Inst& inst) {
-    ctx.AddU32("{}=uint(gl_SubGroupGtMaskARB.x);", inst);
+    ctx.AddU32("{}={};", inst, GetMask(ctx, "gl_SubGroupGtMaskARB"));
 }
 
 void EmitSubgroupGeMask(EmitContext& ctx, IR::Inst& inst) {
-    ctx.AddU32("{}=uint(gl_SubGroupGeMaskARB.x);", inst);
+    ctx.AddU32("{}={};", inst, GetMask(ctx, "gl_SubGroupGeMaskARB"));
 }
 
 void EmitShuffleIndex(EmitContext& ctx, IR::Inst& inst, std::string_view value,
-                      std::string_view index, std::string_view clamp,
-                      std::string_view segmentation_mask) {
+                      std::string_view index, std::string_view clamp, std::string_view seg_mask) {
     if (ctx.profile.support_gl_warp_intrinsics) {
-        UseShuffleNv(ctx, inst, "shuffleNV", value, index, clamp, segmentation_mask);
+        UseShuffleNv(ctx, inst, "shuffleNV", value, index, clamp, seg_mask);
         return;
     }
-    const auto not_seg_mask{fmt::format("(~{})", segmentation_mask)};
-    const auto thread_id{"gl_SubGroupInvocationARB"};
-    const auto min_thread_id{ComputeMinThreadId(thread_id, segmentation_mask)};
-    const auto max_thread_id{ComputeMaxThreadId(min_thread_id, clamp, not_seg_mask)};
+    const bool big_warp{ctx.profile.warp_size_potentially_larger_than_guest};
+    const auto is_upper_partition{"int(gl_SubGroupInvocationARB)>=32"};
+    const auto upper_index{fmt::format("{}?{}+32:{}", is_upper_partition, index, index)};
+    const auto upper_clamp{fmt::format("{}?{}+32:{}", is_upper_partition, clamp, clamp)};
 
-    const auto lhs{fmt::format("({}&{})", index, not_seg_mask)};
+    const auto not_seg_mask{fmt::format("(~{})", seg_mask)};
+    const auto min_thread_id{ComputeMinThreadId(THREAD_ID, seg_mask)};
+    const auto max_thread_id{
+        ComputeMaxThreadId(min_thread_id, big_warp ? upper_clamp : clamp, not_seg_mask)};
+
+    const auto lhs{fmt::format("({}&{})", big_warp ? upper_index : index, not_seg_mask)};
     const auto src_thread_id{fmt::format("({})|({})", lhs, min_thread_id)};
     ctx.Add("shfl_in_bounds=int({})<=int({});", src_thread_id, max_thread_id);
     SetInBoundsFlag(ctx, inst);
@@ -128,29 +146,34 @@ void EmitShuffleIndex(EmitContext& ctx, IR::Inst& inst, std::string_view value,
 }
 
 void EmitShuffleUp(EmitContext& ctx, IR::Inst& inst, std::string_view value, std::string_view index,
-                   std::string_view clamp, std::string_view segmentation_mask) {
+                   std::string_view clamp, std::string_view seg_mask) {
     if (ctx.profile.support_gl_warp_intrinsics) {
-        UseShuffleNv(ctx, inst, "shuffleUpNV", value, index, clamp, segmentation_mask);
+        UseShuffleNv(ctx, inst, "shuffleUpNV", value, index, clamp, seg_mask);
         return;
     }
-    const auto thread_id{"gl_SubGroupInvocationARB"};
-    const auto max_thread_id{GetMaxThreadId(thread_id, clamp, segmentation_mask)};
-    const auto src_thread_id{fmt::format("({}-{})", thread_id, index)};
+    const bool big_warp{ctx.profile.warp_size_potentially_larger_than_guest};
+    const auto is_upper_partition{"int(gl_SubGroupInvocationARB)>=32"};
+    const auto upper_clamp{fmt::format("{}?{}+32:{}", is_upper_partition, clamp, clamp)};
+
+    const auto max_thread_id{GetMaxThreadId(THREAD_ID, big_warp ? upper_clamp : clamp, seg_mask)};
+    const auto src_thread_id{fmt::format("({}-{})", THREAD_ID, index)};
     ctx.Add("shfl_in_bounds=int({})>=int({});", src_thread_id, max_thread_id);
     SetInBoundsFlag(ctx, inst);
     ctx.AddU32("{}=shfl_in_bounds?readInvocationARB({},{}):{};", inst, value, src_thread_id, value);
 }
 
 void EmitShuffleDown(EmitContext& ctx, IR::Inst& inst, std::string_view value,
-                     std::string_view index, std::string_view clamp,
-                     std::string_view segmentation_mask) {
+                     std::string_view index, std::string_view clamp, std::string_view seg_mask) {
     if (ctx.profile.support_gl_warp_intrinsics) {
-        UseShuffleNv(ctx, inst, "shuffleDownNV", value, index, clamp, segmentation_mask);
+        UseShuffleNv(ctx, inst, "shuffleDownNV", value, index, clamp, seg_mask);
         return;
     }
-    const auto thread_id{"gl_SubGroupInvocationARB"};
-    const auto max_thread_id{GetMaxThreadId(thread_id, clamp, segmentation_mask)};
-    const auto src_thread_id{fmt::format("({}+{})", thread_id, index)};
+    const bool big_warp{ctx.profile.warp_size_potentially_larger_than_guest};
+    const auto is_upper_partition{"int(gl_SubGroupInvocationARB)>=32"};
+    const auto upper_clamp{fmt::format("{}?{}+32:{}", is_upper_partition, clamp, clamp)};
+
+    const auto max_thread_id{GetMaxThreadId(THREAD_ID, big_warp ? upper_clamp : clamp, seg_mask)};
+    const auto src_thread_id{fmt::format("({}+{})", THREAD_ID, index)};
     ctx.Add("shfl_in_bounds=int({})<=int({});", src_thread_id, max_thread_id);
     SetInBoundsFlag(ctx, inst);
     ctx.AddU32("{}=shfl_in_bounds?readInvocationARB({},{}):{};", inst, value, src_thread_id, value);
@@ -158,14 +181,17 @@ void EmitShuffleDown(EmitContext& ctx, IR::Inst& inst, std::string_view value,
 
 void EmitShuffleButterfly(EmitContext& ctx, IR::Inst& inst, std::string_view value,
                           std::string_view index, std::string_view clamp,
-                          std::string_view segmentation_mask) {
+                          std::string_view seg_mask) {
     if (ctx.profile.support_gl_warp_intrinsics) {
-        UseShuffleNv(ctx, inst, "shuffleXorNV", value, index, clamp, segmentation_mask);
+        UseShuffleNv(ctx, inst, "shuffleXorNV", value, index, clamp, seg_mask);
         return;
     }
-    const auto thread_id{"gl_SubGroupInvocationARB"};
-    const auto max_thread_id{GetMaxThreadId(thread_id, clamp, segmentation_mask)};
-    const auto src_thread_id{fmt::format("({}^{})", thread_id, index)};
+    const bool big_warp{ctx.profile.warp_size_potentially_larger_than_guest};
+    const auto is_upper_partition{"int(gl_SubGroupInvocationARB)>=32"};
+    const auto upper_clamp{fmt::format("{}?{}+32:{}", is_upper_partition, clamp, clamp)};
+
+    const auto max_thread_id{GetMaxThreadId(THREAD_ID, big_warp ? upper_clamp : clamp, seg_mask)};
+    const auto src_thread_id{fmt::format("({}^{})", THREAD_ID, index)};
     ctx.Add("shfl_in_bounds=int({})<=int({});", src_thread_id, max_thread_id);
     SetInBoundsFlag(ctx, inst);
     ctx.AddU32("{}=shfl_in_bounds?readInvocationARB({},{}):{};", inst, value, src_thread_id, value);

src/shader_recompiler/backend/spirv/emit_context.cpp

@@ -15,6 +15,8 @@
 
 namespace Shader::Backend::SPIRV {
 namespace {
+constexpr size_t NUM_FIXEDFNCTEXTURE = 10;
+
 enum class Operation {
     Increment,
     Decrement,
@@ -427,6 +429,16 @@ Id DescType(EmitContext& ctx, Id sampled_type, Id pointer_type, u32 count) {
         return pointer_type;
     }
 }
+
+size_t FindNextUnusedLocation(const std::bitset<IR::NUM_GENERICS>& used_locations,
+                              size_t start_offset) {
+    for (size_t location = start_offset; location < used_locations.size(); ++location) {
+        if (!used_locations.test(location)) {
+            return location;
+        }
+    }
+    throw RuntimeError("Unable to get an unused location for legacy attribute");
+}
 } // Anonymous namespace
 
 void VectorTypes::Define(Sirit::Module& sirit_ctx, Id base_type, std::string_view name) {
@@ -1227,6 +1239,7 @@ void EmitContext::DefineInputs(const IR::Program& program) {
         loads[IR::Attribute::TessellationEvaluationPointV]) {
         tess_coord = DefineInput(*this, F32[3], false, spv::BuiltIn::TessCoord);
     }
+    std::bitset<IR::NUM_GENERICS> used_locations{};
     for (size_t index = 0; index < IR::NUM_GENERICS; ++index) {
         const AttributeType input_type{runtime_info.generic_input_types[index]};
         if (!runtime_info.previous_stage_stores.Generic(index)) {
@@ -1238,6 +1251,7 @@ void EmitContext::DefineInputs(const IR::Program& program) {
         if (input_type == AttributeType::Disabled) {
             continue;
         }
+        used_locations.set(index);
         const Id type{GetAttributeType(*this, input_type)};
         const Id id{DefineInput(*this, type, true)};
         Decorate(id, spv::Decoration::Location, static_cast<u32>(index));
@@ -1263,6 +1277,26 @@ void EmitContext::DefineInputs(const IR::Program& program) {
             break;
         }
     }
+    size_t previous_unused_location = 0;
+    if (loads.AnyComponent(IR::Attribute::ColorFrontDiffuseR)) {
+        const size_t location = FindNextUnusedLocation(used_locations, previous_unused_location);
+        previous_unused_location = location;
+        used_locations.set(location);
+        const Id id{DefineInput(*this, F32[4], true)};
+        Decorate(id, spv::Decoration::Location, location);
+        input_front_color = id;
+    }
+    for (size_t index = 0; index < NUM_FIXEDFNCTEXTURE; ++index) {
+        if (loads.AnyComponent(IR::Attribute::FixedFncTexture0S + index * 4)) {
+            const size_t location =
+                FindNextUnusedLocation(used_locations, previous_unused_location);
+            previous_unused_location = location;
+            used_locations.set(location);
+            const Id id{DefineInput(*this, F32[4], true)};
+            Decorate(id, spv::Decoration::Location, location);
+            input_fixed_fnc_textures[index] = id;
+        }
+    }
     if (stage == Stage::TessellationEval) {
         for (size_t index = 0; index < info.uses_patches.size(); ++index) {
             if (!info.uses_patches[index]) {
@@ -1313,9 +1347,31 @@ void EmitContext::DefineOutputs(const IR::Program& program) {
         viewport_mask = DefineOutput(*this, TypeArray(U32[1], Const(1u)), std::nullopt,
                                      spv::BuiltIn::ViewportMaskNV);
     }
+    std::bitset<IR::NUM_GENERICS> used_locations{};
     for (size_t index = 0; index < IR::NUM_GENERICS; ++index) {
         if (info.stores.Generic(index)) {
             DefineGenericOutput(*this, index, invocations);
+            used_locations.set(index);
+        }
+    }
+    size_t previous_unused_location = 0;
+    if (info.stores.AnyComponent(IR::Attribute::ColorFrontDiffuseR)) {
+        const size_t location = FindNextUnusedLocation(used_locations, previous_unused_location);
+        previous_unused_location = location;
+        used_locations.set(location);
+        const Id id{DefineOutput(*this, F32[4], invocations)};
+        Decorate(id, spv::Decoration::Location, static_cast<u32>(location));
+        output_front_color = id;
+    }
+    for (size_t index = 0; index < NUM_FIXEDFNCTEXTURE; ++index) {
+        if (info.stores.AnyComponent(IR::Attribute::FixedFncTexture0S + index * 4)) {
+            const size_t location =
+                FindNextUnusedLocation(used_locations, previous_unused_location);
+            previous_unused_location = location;
+            used_locations.set(location);
+            const Id id{DefineOutput(*this, F32[4], invocations)};
+            Decorate(id, spv::Decoration::Location, location);
+            output_fixed_fnc_textures[index] = id;
         }
     }
     switch (stage) {

src/shader_recompiler/backend/spirv/emit_context.h

@@ -268,10 +268,14 @@ public:
     Id write_global_func_u32x4{};
 
     Id input_position{};
+    Id input_front_color{};
+    std::array<Id, 10> input_fixed_fnc_textures{};
     std::array<Id, 32> input_generics{};
 
     Id output_point_size{};
     Id output_position{};
+    Id output_front_color{};
+    std::array<Id, 10> output_fixed_fnc_textures{};
     std::array<std::array<GenericElementInfo, 4>, 32> output_generics{};
 
     Id output_tess_level_outer{};

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -43,6 +43,25 @@ Id AttrPointer(EmitContext& ctx, Id pointer_type, Id vertex, Id base, Args&&...
     }
 }
 
+bool IsFixedFncTexture(IR::Attribute attribute) {
+    return attribute >= IR::Attribute::FixedFncTexture0S &&
+           attribute <= IR::Attribute::FixedFncTexture9Q;
+}
+
+u32 FixedFncTextureAttributeIndex(IR::Attribute attribute) {
+    if (!IsFixedFncTexture(attribute)) {
+        throw InvalidArgument("Attribute {} is not a FixedFncTexture", attribute);
+    }
+    return (static_cast<u32>(attribute) - static_cast<u32>(IR::Attribute::FixedFncTexture0S)) / 4u;
+}
+
+u32 FixedFncTextureAttributeElement(IR::Attribute attribute) {
+    if (!IsFixedFncTexture(attribute)) {
+        throw InvalidArgument("Attribute {} is not a FixedFncTexture", attribute);
+    }
+    return static_cast<u32>(attribute) % 4u;
+}
+
 template <typename... Args>
 Id OutputAccessChain(EmitContext& ctx, Id result_type, Id base, Args&&... args) {
     if (ctx.stage == Stage::TessellationControl) {
@@ -74,6 +93,13 @@ std::optional<OutAttr> OutputAttrPointer(EmitContext& ctx, IR::Attribute attr) {
             return OutputAccessChain(ctx, ctx.output_f32, info.id, index_id);
         }
     }
+    if (IsFixedFncTexture(attr)) {
+        const u32 index{FixedFncTextureAttributeIndex(attr)};
+        const u32 element{FixedFncTextureAttributeElement(attr)};
+        const Id element_id{ctx.Const(element)};
+        return OutputAccessChain(ctx, ctx.output_f32, ctx.output_fixed_fnc_textures[index],
+                                 element_id);
+    }
     switch (attr) {
     case IR::Attribute::PointSize:
         return ctx.output_point_size;
@@ -85,6 +111,14 @@ std::optional<OutAttr> OutputAttrPointer(EmitContext& ctx, IR::Attribute attr) {
         const Id element_id{ctx.Const(element)};
         return OutputAccessChain(ctx, ctx.output_f32, ctx.output_position, element_id);
     }
+    case IR::Attribute::ColorFrontDiffuseR:
+    case IR::Attribute::ColorFrontDiffuseG:
+    case IR::Attribute::ColorFrontDiffuseB:
+    case IR::Attribute::ColorFrontDiffuseA: {
+        const u32 element{static_cast<u32>(attr) % 4};
+        const Id element_id{ctx.Const(element)};
+        return OutputAccessChain(ctx, ctx.output_f32, ctx.output_front_color, element_id);
+    }
     case IR::Attribute::ClipDistance0:
     case IR::Attribute::ClipDistance1:
     case IR::Attribute::ClipDistance2:
@@ -298,19 +332,21 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
     if (IR::IsGeneric(attr)) {
         const u32 index{IR::GenericAttributeIndex(attr)};
         const std::optional<AttrInfo> type{AttrTypes(ctx, index)};
-        if (!type) {
-            // Attribute is disabled
+        if (!type || !ctx.runtime_info.previous_stage_stores.Generic(index, element)) {
+            // Attribute is disabled or varying component is not written
             return ctx.Const(element == 3 ? 1.0f : 0.0f);
         }
-        if (!ctx.runtime_info.previous_stage_stores.Generic(index, element)) {
-            // Varying component is not written
-            return ctx.Const(type && element == 3 ? 1.0f : 0.0f);
-        }
         const Id generic_id{ctx.input_generics.at(index)};
         const Id pointer{AttrPointer(ctx, type->pointer, vertex, generic_id, ctx.Const(element))};
         const Id value{ctx.OpLoad(type->id, pointer)};
         return type->needs_cast ? ctx.OpBitcast(ctx.F32[1], value) : value;
     }
+    if (IsFixedFncTexture(attr)) {
+        const u32 index{FixedFncTextureAttributeIndex(attr)};
+        const Id attr_id{ctx.input_fixed_fnc_textures[index]};
+        const Id attr_ptr{AttrPointer(ctx, ctx.input_f32, vertex, attr_id, ctx.Const(element))};
+        return ctx.OpLoad(ctx.F32[1], attr_ptr);
+    }
     switch (attr) {
     case IR::Attribute::PrimitiveId:
         return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.primitive_id));
@@ -320,6 +356,13 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
     case IR::Attribute::PositionW:
         return ctx.OpLoad(ctx.F32[1], AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position,
                                                   ctx.Const(element)));
+    case IR::Attribute::ColorFrontDiffuseR:
+    case IR::Attribute::ColorFrontDiffuseG:
+    case IR::Attribute::ColorFrontDiffuseB:
+    case IR::Attribute::ColorFrontDiffuseA: {
+        return ctx.OpLoad(ctx.F32[1], AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_front_color,
+                                                  ctx.Const(element)));
+    }
     case IR::Attribute::InstanceId:
         if (ctx.profile.support_vertex_instance_id) {
             return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.instance_id));
@@ -337,8 +380,9 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
             return ctx.OpBitcast(ctx.F32[1], ctx.OpISub(ctx.U32[1], index, base));
         }
     case IR::Attribute::FrontFace:
-        return ctx.OpSelect(ctx.U32[1], ctx.OpLoad(ctx.U1, ctx.front_face),
-                            ctx.Const(std::numeric_limits<u32>::max()), ctx.u32_zero_value);
+        return ctx.OpSelect(ctx.F32[1], ctx.OpLoad(ctx.U1, ctx.front_face),
+                            ctx.OpBitcast(ctx.F32[1], ctx.Const(std::numeric_limits<u32>::max())),
+                            ctx.f32_zero_value);
     case IR::Attribute::PointSpriteS:
         return ctx.OpLoad(ctx.F32[1],
                           ctx.OpAccessChain(ctx.input_f32, ctx.point_coord, ctx.u32_zero_value));

src/shader_recompiler/backend/spirv/emit_spirv_warp.cpp

@@ -7,8 +7,13 @@
 
 namespace Shader::Backend::SPIRV {
 namespace {
+Id GetThreadId(EmitContext& ctx) {
+    return ctx.OpLoad(ctx.U32[1], ctx.subgroup_local_invocation_id);
+}
+
 Id WarpExtract(EmitContext& ctx, Id value) {
-    const Id local_index{ctx.OpLoad(ctx.U32[1], ctx.subgroup_local_invocation_id)};
+    const Id thread_id{GetThreadId(ctx)};
+    const Id local_index{ctx.OpShiftRightArithmetic(ctx.U32[1], thread_id, ctx.Const(5U))};
     return ctx.OpVectorExtractDynamic(ctx.U32[1], value, local_index);
 }
 
@@ -48,10 +53,17 @@ Id SelectValue(EmitContext& ctx, Id in_range, Id value, Id src_thread_id) {
     return ctx.OpSelect(ctx.U32[1], in_range,
                         ctx.OpSubgroupReadInvocationKHR(ctx.U32[1], value, src_thread_id), value);
 }
+
+Id GetUpperClamp(EmitContext& ctx, Id invocation_id, Id clamp) {
+    const Id thirty_two{ctx.Const(32u)};
+    const Id is_upper_partition{ctx.OpSGreaterThanEqual(ctx.U1, invocation_id, thirty_two)};
+    const Id upper_clamp{ctx.OpIAdd(ctx.U32[1], thirty_two, clamp)};
+    return ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_clamp, clamp);
+}
 } // Anonymous namespace
 
 Id EmitLaneId(EmitContext& ctx) {
-    const Id id{ctx.OpLoad(ctx.U32[1], ctx.subgroup_local_invocation_id)};
+    const Id id{GetThreadId(ctx)};
     if (!ctx.profile.warp_size_potentially_larger_than_guest) {
         return id;
     }
@@ -123,7 +135,15 @@ Id EmitSubgroupGeMask(EmitContext& ctx) {
 Id EmitShuffleIndex(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                     Id segmentation_mask) {
     const Id not_seg_mask{ctx.OpNot(ctx.U32[1], segmentation_mask)};
-    const Id thread_id{ctx.OpLoad(ctx.U32[1], ctx.subgroup_local_invocation_id)};
+    const Id thread_id{GetThreadId(ctx)};
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        const Id thirty_two{ctx.Const(32u)};
+        const Id is_upper_partition{ctx.OpSGreaterThanEqual(ctx.U1, thread_id, thirty_two)};
+        const Id upper_index{ctx.OpIAdd(ctx.U32[1], thirty_two, index)};
+        const Id upper_clamp{ctx.OpIAdd(ctx.U32[1], thirty_two, clamp)};
+        index = ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_index, index);
+        clamp = ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_clamp, clamp);
+    }
     const Id min_thread_id{ComputeMinThreadId(ctx, thread_id, segmentation_mask)};
     const Id max_thread_id{ComputeMaxThreadId(ctx, min_thread_id, clamp, not_seg_mask)};
 
@@ -137,7 +157,10 @@ Id EmitShuffleIndex(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id cla
 
 Id EmitShuffleUp(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                  Id segmentation_mask) {
-    const Id thread_id{ctx.OpLoad(ctx.U32[1], ctx.subgroup_local_invocation_id)};
+    const Id thread_id{GetThreadId(ctx)};
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        clamp = GetUpperClamp(ctx, thread_id, clamp);
+    }
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
     const Id src_thread_id{ctx.OpISub(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSGreaterThanEqual(ctx.U1, src_thread_id, max_thread_id)};
@@ -148,7 +171,10 @@ Id EmitShuffleUp(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
 
 Id EmitShuffleDown(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                    Id segmentation_mask) {
-    const Id thread_id{ctx.OpLoad(ctx.U32[1], ctx.subgroup_local_invocation_id)};
+    const Id thread_id{GetThreadId(ctx)};
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        clamp = GetUpperClamp(ctx, thread_id, clamp);
+    }
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
     const Id src_thread_id{ctx.OpIAdd(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
@@ -159,7 +185,10 @@ Id EmitShuffleDown(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clam
 
 Id EmitShuffleButterfly(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                         Id segmentation_mask) {
-    const Id thread_id{ctx.OpLoad(ctx.U32[1], ctx.subgroup_local_invocation_id)};
+    const Id thread_id{GetThreadId(ctx)};
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        clamp = GetUpperClamp(ctx, thread_id, clamp);
+    }
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
     const Id src_thread_id{ctx.OpBitwiseXor(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};

src/shader_recompiler/frontend/ir/value.h

@@ -57,6 +57,7 @@ public:
 
     [[nodiscard]] IR::Inst* Inst() const;
     [[nodiscard]] IR::Inst* InstRecursive() const;
+    [[nodiscard]] IR::Inst* TryInstRecursive() const;
     [[nodiscard]] IR::Value Resolve() const;
     [[nodiscard]] IR::Reg Reg() const;
     [[nodiscard]] IR::Pred Pred() const;
@@ -308,6 +309,13 @@ inline IR::Inst* Value::InstRecursive() const {
     return inst;
 }
 
+inline IR::Inst* Value::TryInstRecursive() const {
+    if (IsIdentity()) {
+        return inst->Arg(0).TryInstRecursive();
+    }
+    return type == Type::Opaque ? inst : nullptr;
+}
+
 inline IR::Value Value::Resolve() const {
     if (IsIdentity()) {
         return inst->Arg(0).Resolve();

src/shader_recompiler/frontend/maxwell/structured_control_flow.cpp

@@ -20,6 +20,7 @@
 #include "shader_recompiler/frontend/maxwell/decode.h"
 #include "shader_recompiler/frontend/maxwell/structured_control_flow.h"
 #include "shader_recompiler/frontend/maxwell/translate/translate.h"
+#include "shader_recompiler/host_translate_info.h"
 #include "shader_recompiler/object_pool.h"
 
 namespace Shader::Maxwell {
@@ -652,7 +653,7 @@ class TranslatePass {
 public:
     TranslatePass(ObjectPool<IR::Inst>& inst_pool_, ObjectPool<IR::Block>& block_pool_,
                   ObjectPool<Statement>& stmt_pool_, Environment& env_, Statement& root_stmt,
-                  IR::AbstractSyntaxList& syntax_list_)
+                  IR::AbstractSyntaxList& syntax_list_, const HostTranslateInfo& host_info)
         : stmt_pool{stmt_pool_}, inst_pool{inst_pool_}, block_pool{block_pool_}, env{env_},
           syntax_list{syntax_list_} {
         Visit(root_stmt, nullptr, nullptr);
@@ -660,6 +661,9 @@ public:
         IR::Block& first_block{*syntax_list.front().data.block};
         IR::IREmitter ir(first_block, first_block.begin());
         ir.Prologue();
+        if (uses_demote_to_helper && host_info.needs_demote_reorder) {
+            DemoteCombinationPass();
+        }
     }
 
 private:
@@ -809,7 +813,14 @@ private:
             }
             case StatementType::Return: {
                 ensure_block();
-                IR::IREmitter{*current_block}.Epilogue();
+                IR::Block* return_block{block_pool.Create(inst_pool)};
+                IR::IREmitter{*return_block}.Epilogue();
+                current_block->AddBranch(return_block);
+
+                auto& merge{syntax_list.emplace_back()};
+                merge.type = IR::AbstractSyntaxNode::Type::Block;
+                merge.data.block = return_block;
+
                 current_block = nullptr;
                 syntax_list.emplace_back().type = IR::AbstractSyntaxNode::Type::Return;
                 break;
@@ -824,6 +835,7 @@ private:
                 auto& merge{syntax_list.emplace_back()};
                 merge.type = IR::AbstractSyntaxNode::Type::Block;
                 merge.data.block = demote_block;
+                uses_demote_to_helper = true;
                 break;
             }
             case StatementType::Unreachable: {
@@ -855,11 +867,117 @@ private:
         return block_pool.Create(inst_pool);
     }
 
+    void DemoteCombinationPass() {
+        using Type = IR::AbstractSyntaxNode::Type;
+        std::vector<IR::Block*> demote_blocks;
+        std::vector<IR::U1> demote_conds;
+        u32 num_epilogues{};
+        u32 branch_depth{};
+        for (const IR::AbstractSyntaxNode& node : syntax_list) {
+            if (node.type == Type::If) {
+                ++branch_depth;
+            }
+            if (node.type == Type::EndIf) {
+                --branch_depth;
+            }
+            if (node.type != Type::Block) {
+                continue;
+            }
+            if (branch_depth > 1) {
+                // Skip reordering nested demote branches.
+                continue;
+            }
+            for (const IR::Inst& inst : node.data.block->Instructions()) {
+                const IR::Opcode op{inst.GetOpcode()};
+                if (op == IR::Opcode::DemoteToHelperInvocation) {
+                    demote_blocks.push_back(node.data.block);
+                    break;
+                }
+                if (op == IR::Opcode::Epilogue) {
+                    ++num_epilogues;
+                }
+            }
+        }
+        if (demote_blocks.size() == 0) {
+            return;
+        }
+        if (num_epilogues > 1) {
+            LOG_DEBUG(Shader, "Combining demotes with more than one return is not implemented.");
+            return;
+        }
+        s64 last_iterator_offset{};
+        auto& asl{syntax_list};
+        for (const IR::Block* demote_block : demote_blocks) {
+            const auto start_it{asl.begin() + last_iterator_offset};
+            auto asl_it{std::find_if(start_it, asl.end(), [&](const IR::AbstractSyntaxNode& asn) {
+                return asn.type == Type::If && asn.data.if_node.body == demote_block;
+            })};
+            if (asl_it == asl.end()) {
+                // Demote without a conditional branch.
+                // No need to proceed since all fragment instances will be demoted regardless.
+                return;
+            }
+            const IR::Block* const end_if = asl_it->data.if_node.merge;
+            demote_conds.push_back(asl_it->data.if_node.cond);
+            last_iterator_offset = std::distance(asl.begin(), asl_it);
+
+            asl_it = asl.erase(asl_it);
+            asl_it = std::find_if(asl_it, asl.end(), [&](const IR::AbstractSyntaxNode& asn) {
+                return asn.type == Type::Block && asn.data.block == demote_block;
+            });
+
+            asl_it = asl.erase(asl_it);
+            asl_it = std::find_if(asl_it, asl.end(), [&](const IR::AbstractSyntaxNode& asn) {
+                return asn.type == Type::EndIf && asn.data.end_if.merge == end_if;
+            });
+            asl_it = asl.erase(asl_it);
+        }
+        const auto epilogue_func{[](const IR::AbstractSyntaxNode& asn) {
+            if (asn.type != Type::Block) {
+                return false;
+            }
+            for (const auto& inst : asn.data.block->Instructions()) {
+                if (inst.GetOpcode() == IR::Opcode::Epilogue) {
+                    return true;
+                }
+            }
+            return false;
+        }};
+        const auto reverse_it{std::find_if(asl.rbegin(), asl.rend(), epilogue_func)};
+        const auto return_block_it{(reverse_it + 1).base()};
+
+        IR::IREmitter ir{*(return_block_it - 1)->data.block};
+        IR::U1 cond(IR::Value(false));
+        for (const auto& demote_cond : demote_conds) {
+            cond = ir.LogicalOr(cond, demote_cond);
+        }
+        cond.Inst()->DestructiveAddUsage(1);
+
+        IR::AbstractSyntaxNode demote_if_node{};
+        demote_if_node.type = Type::If;
+        demote_if_node.data.if_node.cond = cond;
+        demote_if_node.data.if_node.body = demote_blocks[0];
+        demote_if_node.data.if_node.merge = return_block_it->data.block;
+
+        IR::AbstractSyntaxNode demote_node{};
+        demote_node.type = Type::Block;
+        demote_node.data.block = demote_blocks[0];
+
+        IR::AbstractSyntaxNode demote_endif_node{};
+        demote_endif_node.type = Type::EndIf;
+        demote_endif_node.data.end_if.merge = return_block_it->data.block;
+
+        asl.insert(return_block_it, demote_endif_node);
+        asl.insert(return_block_it, demote_node);
+        asl.insert(return_block_it, demote_if_node);
+    }
+
     ObjectPool<Statement>& stmt_pool;
     ObjectPool<IR::Inst>& inst_pool;
     ObjectPool<IR::Block>& block_pool;
     Environment& env;
     IR::AbstractSyntaxList& syntax_list;
+    bool uses_demote_to_helper{};
 
 // TODO: C++20 Remove this when all compilers support constexpr std::vector
 #if __cpp_lib_constexpr_vector >= 201907
@@ -871,12 +989,13 @@ private:
 } // Anonymous namespace
 
 IR::AbstractSyntaxList BuildASL(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
-                                Environment& env, Flow::CFG& cfg) {
+                                Environment& env, Flow::CFG& cfg,
+                                const HostTranslateInfo& host_info) {
     ObjectPool<Statement> stmt_pool{64};
     GotoPass goto_pass{cfg, stmt_pool};
     Statement& root{goto_pass.RootStatement()};
     IR::AbstractSyntaxList syntax_list;
-    TranslatePass{inst_pool, block_pool, stmt_pool, env, root, syntax_list};
+    TranslatePass{inst_pool, block_pool, stmt_pool, env, root, syntax_list, host_info};
     return syntax_list;
 }
 

src/shader_recompiler/frontend/maxwell/structured_control_flow.h

@@ -11,10 +11,13 @@
 #include "shader_recompiler/frontend/maxwell/control_flow.h"
 #include "shader_recompiler/object_pool.h"
 
-namespace Shader::Maxwell {
+namespace Shader {
+struct HostTranslateInfo;
+namespace Maxwell {
 
 [[nodiscard]] IR::AbstractSyntaxList BuildASL(ObjectPool<IR::Inst>& inst_pool,
                                               ObjectPool<IR::Block>& block_pool, Environment& env,
-                                              Flow::CFG& cfg);
+                                              Flow::CFG& cfg, const HostTranslateInfo& host_info);
 
-} // namespace Shader::Maxwell
+} // namespace Maxwell
+} // namespace Shader

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -130,7 +130,7 @@ void AddNVNStorageBuffers(IR::Program& program) {
 IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
                              Environment& env, Flow::CFG& cfg, const HostTranslateInfo& host_info) {
     IR::Program program;
-    program.syntax_list = BuildASL(inst_pool, block_pool, env, cfg);
+    program.syntax_list = BuildASL(inst_pool, block_pool, env, cfg, host_info);
     program.blocks = GenerateBlocks(program.syntax_list);
     program.post_order_blocks = PostOrder(program.syntax_list.front());
     program.stage = env.ShaderStage();

src/shader_recompiler/host_translate_info.h

@@ -11,8 +11,9 @@ namespace Shader {
 
 /// Misc information about the host
 struct HostTranslateInfo {
-    bool support_float16{}; ///< True when the device supports 16-bit floats
-    bool support_int64{};   ///< True when the device supports 64-bit integers
+    bool support_float16{};      ///< True when the device supports 16-bit floats
+    bool support_int64{};        ///< True when the device supports 64-bit integers
+    bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
 };
 
 } // namespace Shader

src/shader_recompiler/ir_opt/constant_propagation_pass.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
+#include <functional>
 #include <tuple>
 #include <type_traits>
 
@@ -88,6 +89,26 @@ bool FoldWhenAllImmediates(IR::Inst& inst, Func&& func) {
     return true;
 }
 
+/// Return true when all values in a range are equal
+template <typename Range>
+bool AreEqual(const Range& range) {
+    auto resolver{[](const auto& value) { return value.Resolve(); }};
+    auto equal{[](const IR::Value& lhs, const IR::Value& rhs) {
+        if (lhs == rhs) {
+            return true;
+        }
+        // Not equal, but try to match if they read the same constant buffer
+        if (!lhs.IsImmediate() && !rhs.IsImmediate() &&
+            lhs.Inst()->GetOpcode() == IR::Opcode::GetCbufU32 &&
+            rhs.Inst()->GetOpcode() == IR::Opcode::GetCbufU32 &&
+            lhs.Inst()->Arg(0) == rhs.Inst()->Arg(0) && lhs.Inst()->Arg(1) == rhs.Inst()->Arg(1)) {
+            return true;
+        }
+        return false;
+    }};
+    return std::ranges::adjacent_find(range, std::not_fn(equal), resolver) == std::end(range);
+}
+
 void FoldGetRegister(IR::Inst& inst) {
     if (inst.Arg(0).Reg() == IR::Reg::RZ) {
         inst.ReplaceUsesWith(IR::Value{u32{0}});
@@ -100,6 +121,157 @@ void FoldGetPred(IR::Inst& inst) {
     }
 }
 
+/// Replaces the XMAD pattern generated by an integer FMA
+bool FoldXmadMultiplyAdd(IR::Block& block, IR::Inst& inst) {
+    /*
+     * We are looking for this specific pattern:
+     *   %6      = BitFieldUExtract %op_b, #0, #16
+     *   %7      = BitFieldUExtract %op_a', #16, #16
+     *   %8      = IMul32 %6, %7
+     *   %10     = BitFieldUExtract %op_a', #0, #16
+     *   %11     = BitFieldInsert %8, %10, #16, #16
+     *   %15     = BitFieldUExtract %op_b, #0, #16
+     *   %16     = BitFieldUExtract %op_a, #0, #16
+     *   %17     = IMul32 %15, %16
+     *   %18     = IAdd32 %17, %op_c
+     *   %22     = BitFieldUExtract %op_b, #16, #16
+     *   %23     = BitFieldUExtract %11, #16, #16
+     *   %24     = IMul32 %22, %23
+     *   %25     = ShiftLeftLogical32 %24, #16
+     *   %26     = ShiftLeftLogical32 %11, #16
+     *   %27     = IAdd32 %26, %18
+     *   %result = IAdd32 %25, %27
+     *
+     * And replace it with:
+     *  %temp   = IMul32 %op_a, %op_b
+     *  %result = IAdd32 %temp, %op_c
+     *
+     * This optimization has been proven safe by Nvidia's compiler logic being reversed.
+     * (If Nvidia generates this code from 'fma(a, b, c)', we can do the same in the reverse order.)
+     */
+    const IR::Value zero{0u};
+    const IR::Value sixteen{16u};
+    IR::Inst* const _25{inst.Arg(0).TryInstRecursive()};
+    IR::Inst* const _27{inst.Arg(1).TryInstRecursive()};
+    if (!_25 || !_27) {
+        return false;
+    }
+    if (_27->GetOpcode() != IR::Opcode::IAdd32) {
+        return false;
+    }
+    if (_25->GetOpcode() != IR::Opcode::ShiftLeftLogical32 || _25->Arg(1) != sixteen) {
+        return false;
+    }
+    IR::Inst* const _24{_25->Arg(0).TryInstRecursive()};
+    if (!_24 || _24->GetOpcode() != IR::Opcode::IMul32) {
+        return false;
+    }
+    IR::Inst* const _22{_24->Arg(0).TryInstRecursive()};
+    IR::Inst* const _23{_24->Arg(1).TryInstRecursive()};
+    if (!_22 || !_23) {
+        return false;
+    }
+    if (_22->GetOpcode() != IR::Opcode::BitFieldUExtract) {
+        return false;
+    }
+    if (_23->GetOpcode() != IR::Opcode::BitFieldUExtract) {
+        return false;
+    }
+    if (_22->Arg(1) != sixteen || _22->Arg(2) != sixteen) {
+        return false;
+    }
+    if (_23->Arg(1) != sixteen || _23->Arg(2) != sixteen) {
+        return false;
+    }
+    IR::Inst* const _11{_23->Arg(0).TryInstRecursive()};
+    if (!_11 || _11->GetOpcode() != IR::Opcode::BitFieldInsert) {
+        return false;
+    }
+    if (_11->Arg(2) != sixteen || _11->Arg(3) != sixteen) {
+        return false;
+    }
+    IR::Inst* const _8{_11->Arg(0).TryInstRecursive()};
+    IR::Inst* const _10{_11->Arg(1).TryInstRecursive()};
+    if (!_8 || !_10) {
+        return false;
+    }
+    if (_8->GetOpcode() != IR::Opcode::IMul32) {
+        return false;
+    }
+    if (_10->GetOpcode() != IR::Opcode::BitFieldUExtract) {
+        return false;
+    }
+    IR::Inst* const _6{_8->Arg(0).TryInstRecursive()};
+    IR::Inst* const _7{_8->Arg(1).TryInstRecursive()};
+    if (!_6 || !_7) {
+        return false;
+    }
+    if (_6->GetOpcode() != IR::Opcode::BitFieldUExtract) {
+        return false;
+    }
+    if (_7->GetOpcode() != IR::Opcode::BitFieldUExtract) {
+        return false;
+    }
+    if (_6->Arg(1) != zero || _6->Arg(2) != sixteen) {
+        return false;
+    }
+    if (_7->Arg(1) != sixteen || _7->Arg(2) != sixteen) {
+        return false;
+    }
+    IR::Inst* const _26{_27->Arg(0).TryInstRecursive()};
+    IR::Inst* const _18{_27->Arg(1).TryInstRecursive()};
+    if (!_26 || !_18) {
+        return false;
+    }
+    if (_26->GetOpcode() != IR::Opcode::ShiftLeftLogical32 || _26->Arg(1) != sixteen) {
+        return false;
+    }
+    if (_26->Arg(0).InstRecursive() != _11) {
+        return false;
+    }
+    if (_18->GetOpcode() != IR::Opcode::IAdd32) {
+        return false;
+    }
+    IR::Inst* const _17{_18->Arg(0).TryInstRecursive()};
+    if (!_17 || _17->GetOpcode() != IR::Opcode::IMul32) {
+        return false;
+    }
+    IR::Inst* const _15{_17->Arg(0).TryInstRecursive()};
+    IR::Inst* const _16{_17->Arg(1).TryInstRecursive()};
+    if (!_15 || !_16) {
+        return false;
+    }
+    if (_15->GetOpcode() != IR::Opcode::BitFieldUExtract) {
+        return false;
+    }
+    if (_16->GetOpcode() != IR::Opcode::BitFieldUExtract) {
+        return false;
+    }
+    if (_15->Arg(1) != zero || _16->Arg(1) != zero || _10->Arg(1) != zero) {
+        return false;
+    }
+    if (_15->Arg(2) != sixteen || _16->Arg(2) != sixteen || _10->Arg(2) != sixteen) {
+        return false;
+    }
+    const std::array<IR::Value, 3> op_as{
+        _7->Arg(0).Resolve(),
+        _16->Arg(0).Resolve(),
+        _10->Arg(0).Resolve(),
+    };
+    const std::array<IR::Value, 3> op_bs{
+        _22->Arg(0).Resolve(),
+        _6->Arg(0).Resolve(),
+        _15->Arg(0).Resolve(),
+    };
+    const IR::U32 op_c{_18->Arg(1)};
+    if (!AreEqual(op_as) || !AreEqual(op_bs)) {
+        return false;
+    }
+    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
+    inst.ReplaceUsesWith(ir.IAdd(ir.IMul(IR::U32{op_as[0]}, IR::U32{op_bs[1]}), op_c));
+    return true;
+}
+
 /// Replaces the pattern generated by two XMAD multiplications
 bool FoldXmadMultiply(IR::Block& block, IR::Inst& inst) {
     /*
@@ -116,33 +288,31 @@ bool FoldXmadMultiply(IR::Block& block, IR::Inst& inst) {
      *
      * This optimization has been proven safe by LLVM and MSVC.
      */
-    const IR::Value lhs_arg{inst.Arg(0)};
-    const IR::Value rhs_arg{inst.Arg(1)};
-    if (lhs_arg.IsImmediate() || rhs_arg.IsImmediate()) {
+    IR::Inst* const lhs_shl{inst.Arg(0).TryInstRecursive()};
+    IR::Inst* const rhs_mul{inst.Arg(1).TryInstRecursive()};
+    if (!lhs_shl || !rhs_mul) {
         return false;
     }
-    IR::Inst* const lhs_shl{lhs_arg.InstRecursive()};
     if (lhs_shl->GetOpcode() != IR::Opcode::ShiftLeftLogical32 ||
         lhs_shl->Arg(1) != IR::Value{16U}) {
         return false;
     }
-    if (lhs_shl->Arg(0).IsImmediate()) {
+    IR::Inst* const lhs_mul{lhs_shl->Arg(0).TryInstRecursive()};
+    if (!lhs_mul) {
         return false;
     }
-    IR::Inst* const lhs_mul{lhs_shl->Arg(0).InstRecursive()};
-    IR::Inst* const rhs_mul{rhs_arg.InstRecursive()};
     if (lhs_mul->GetOpcode() != IR::Opcode::IMul32 || rhs_mul->GetOpcode() != IR::Opcode::IMul32) {
         return false;
     }
-    if (lhs_mul->Arg(1).Resolve() != rhs_mul->Arg(1).Resolve()) {
-        return false;
-    }
     const IR::U32 factor_b{lhs_mul->Arg(1)};
-    if (lhs_mul->Arg(0).IsImmediate() || rhs_mul->Arg(0).IsImmediate()) {
+    if (factor_b.Resolve() != rhs_mul->Arg(1).Resolve()) {
+        return false;
+    }
+    IR::Inst* const lhs_bfe{lhs_mul->Arg(0).TryInstRecursive()};
+    IR::Inst* const rhs_bfe{rhs_mul->Arg(0).TryInstRecursive()};
+    if (!lhs_bfe || !rhs_bfe) {
         return false;
     }
-    IR::Inst* const lhs_bfe{lhs_mul->Arg(0).InstRecursive()};
-    IR::Inst* const rhs_bfe{rhs_mul->Arg(0).InstRecursive()};
     if (lhs_bfe->GetOpcode() != IR::Opcode::BitFieldUExtract) {
         return false;
     }
@@ -155,10 +325,10 @@ bool FoldXmadMultiply(IR::Block& block, IR::Inst& inst) {
     if (rhs_bfe->Arg(1) != IR::Value{0U} || rhs_bfe->Arg(2) != IR::Value{16U}) {
         return false;
     }
-    if (lhs_bfe->Arg(0).Resolve() != rhs_bfe->Arg(0).Resolve()) {
+    const IR::U32 factor_a{lhs_bfe->Arg(0)};
+    if (factor_a.Resolve() != rhs_bfe->Arg(0).Resolve()) {
         return false;
     }
-    const IR::U32 factor_a{lhs_bfe->Arg(0)};
     IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     inst.ReplaceUsesWith(ir.IMul(factor_a, factor_b));
     return true;
@@ -181,6 +351,9 @@ void FoldAdd(IR::Block& block, IR::Inst& inst) {
         if (FoldXmadMultiply(block, inst)) {
             return;
         }
+        if (FoldXmadMultiplyAdd(block, inst)) {
+            return;
+        }
     }
 }
 
@@ -476,6 +649,10 @@ void ConstantPropagation(IR::Block& block, IR::Inst& inst) {
         return FoldInverseFunc(inst, IR::Opcode::UnpackHalf2x16);
     case IR::Opcode::UnpackHalf2x16:
         return FoldInverseFunc(inst, IR::Opcode::PackHalf2x16);
+    case IR::Opcode::PackFloat2x16:
+        return FoldInverseFunc(inst, IR::Opcode::UnpackFloat2x16);
+    case IR::Opcode::UnpackFloat2x16:
+        return FoldInverseFunc(inst, IR::Opcode::PackFloat2x16);
     case IR::Opcode::SelectU1:
     case IR::Opcode::SelectU8:
     case IR::Opcode::SelectU16:

src/tests/common/param_package.cpp

@@ -4,11 +4,13 @@
 
 #include <catch2/catch.hpp>
 #include <math.h>
+#include "common/logging/backend.h"
 #include "common/param_package.h"
 
 namespace Common {
 
 TEST_CASE("ParamPackage", "[common]") {
+    Common::Log::DisableLoggingInTests();
     ParamPackage original{
         {"abc", "xyz"},
         {"def", "42"},

src/video_core/CMakeLists.txt

@@ -1,5 +1,10 @@
 add_subdirectory(host_shaders)
 
+if(LIBVA_FOUND)
+    set_source_files_properties(command_classes/codecs/codec.cpp
+        PROPERTIES COMPILE_DEFINITIONS LIBVA_FOUND=1)
+endif()
+
 add_library(video_core STATIC
     buffer_cache/buffer_base.h
     buffer_cache/buffer_cache.cpp
@@ -92,6 +97,7 @@ add_library(video_core STATIC
     renderer_opengl/gl_stream_buffer.h
     renderer_opengl/gl_texture_cache.cpp
     renderer_opengl/gl_texture_cache.h
+    renderer_opengl/gl_texture_cache_base.cpp
     renderer_opengl/gl_query_cache.cpp
     renderer_opengl/gl_query_cache.h
     renderer_opengl/maxwell_to_gl.h
@@ -106,6 +112,8 @@ add_library(video_core STATIC
     renderer_vulkan/maxwell_to_vk.cpp
     renderer_vulkan/maxwell_to_vk.h
     renderer_vulkan/pipeline_helper.h
+    renderer_vulkan/pipeline_statistics.cpp
+    renderer_vulkan/pipeline_statistics.h
     renderer_vulkan/renderer_vulkan.h
     renderer_vulkan/renderer_vulkan.cpp
     renderer_vulkan/vk_blit_screen.cpp
@@ -148,6 +156,7 @@ add_library(video_core STATIC
     renderer_vulkan/vk_swapchain.h
     renderer_vulkan/vk_texture_cache.cpp
     renderer_vulkan/vk_texture_cache.h
+    renderer_vulkan/vk_texture_cache_base.cpp
     renderer_vulkan/vk_update_descriptor.cpp
     renderer_vulkan/vk_update_descriptor.h
     shader_cache.cpp
@@ -179,6 +188,7 @@ add_library(video_core STATIC
     texture_cache/samples_helper.h
     texture_cache/slot_vector.h
     texture_cache/texture_cache.h
+    texture_cache/texture_cache_base.h
     texture_cache/types.h
     texture_cache/util.cpp
     texture_cache/util.h
@@ -221,6 +231,7 @@ endif()
 
 target_include_directories(video_core PRIVATE ${FFmpeg_INCLUDE_DIR})
 target_link_libraries(video_core PRIVATE ${FFmpeg_LIBRARIES})
+target_link_options(video_core PRIVATE ${FFmpeg_LDFLAGS})
 
 add_dependencies(video_core host_shaders)
 target_include_directories(video_core PRIVATE ${HOST_SHADERS_INCLUDE})

src/video_core/buffer_cache/buffer_base.h

@@ -261,16 +261,6 @@ public:
         stream_score += score;
     }
 
-    /// Sets the new frame tick
-    void SetFrameTick(u64 new_frame_tick) noexcept {
-        frame_tick = new_frame_tick;
-    }
-
-    /// Returns the new frame tick
-    [[nodiscard]] u64 FrameTick() const noexcept {
-        return frame_tick;
-    }
-
     /// Returns the likeliness of this being a stream buffer
     [[nodiscard]] int StreamScore() const noexcept {
         return stream_score;
@@ -307,6 +297,14 @@ public:
         return words.size_bytes;
     }
 
+    size_t getLRUID() const noexcept {
+        return lru_id;
+    }
+
+    void setLRUID(size_t lru_id_) {
+        lru_id = lru_id_;
+    }
+
 private:
     template <Type type>
     u64* Array() noexcept {
@@ -603,9 +601,9 @@ private:
     RasterizerInterface* rasterizer = nullptr;
     VAddr cpu_addr = 0;
     Words words;
-    u64 frame_tick = 0;
     BufferFlagBits flags{};
     int stream_score = 0;
+    size_t lru_id = SIZE_MAX;
 };
 
 } // namespace VideoCommon

src/video_core/buffer_cache/buffer_cache.h

@@ -20,6 +20,7 @@
 #include "common/common_types.h"
 #include "common/div_ceil.h"
 #include "common/literals.h"
+#include "common/lru_cache.h"
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
 #include "common/settings.h"
@@ -330,7 +331,7 @@ private:
     template <bool insert>
     void ChangeRegister(BufferId buffer_id);
 
-    void TouchBuffer(Buffer& buffer) const noexcept;
+    void TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept;
 
     bool SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size);
 
@@ -428,7 +429,11 @@ private:
     size_t immediate_buffer_capacity = 0;
     std::unique_ptr<u8[]> immediate_buffer_alloc;
 
-    typename SlotVector<Buffer>::Iterator deletion_iterator;
+    struct LRUItemParams {
+        using ObjectType = BufferId;
+        using TickType = u64;
+    };
+    Common::LeastRecentlyUsedCache<LRUItemParams> lru_cache;
     u64 frame_tick = 0;
     u64 total_used_memory = 0;
 
@@ -445,7 +450,6 @@ BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
       kepler_compute{kepler_compute_}, gpu_memory{gpu_memory_}, cpu_memory{cpu_memory_} {
     // Ensure the first slot is used for the null buffer
     void(slot_buffers.insert(runtime, NullBufferParams{}));
-    deletion_iterator = slot_buffers.end();
     common_ranges.clear();
 }
 
@@ -454,20 +458,17 @@ void BufferCache<P>::RunGarbageCollector() {
     const bool aggressive_gc = total_used_memory >= CRITICAL_MEMORY;
     const u64 ticks_to_destroy = aggressive_gc ? 60 : 120;
     int num_iterations = aggressive_gc ? 64 : 32;
-    for (; num_iterations > 0; --num_iterations) {
-        if (deletion_iterator == slot_buffers.end()) {
-            deletion_iterator = slot_buffers.begin();
+    const auto clean_up = [this, &num_iterations](BufferId buffer_id) {
+        if (num_iterations == 0) {
+            return true;
         }
-        ++deletion_iterator;
-        if (deletion_iterator == slot_buffers.end()) {
-            break;
-        }
-        const auto [buffer_id, buffer] = *deletion_iterator;
-        if (buffer->FrameTick() + ticks_to_destroy < frame_tick) {
-            DownloadBufferMemory(*buffer);
-            DeleteBuffer(buffer_id);
-        }
-    }
+        --num_iterations;
+        auto& buffer = slot_buffers[buffer_id];
+        DownloadBufferMemory(buffer);
+        DeleteBuffer(buffer_id);
+        return false;
+    };
+    lru_cache.ForEachItemBelow(frame_tick - ticks_to_destroy, clean_up);
 }
 
 template <class P>
@@ -485,7 +486,7 @@ void BufferCache<P>::TickFrame() {
     const bool skip_preferred = hits * 256 < shots * 251;
     uniform_buffer_skip_cache_size = skip_preferred ? DEFAULT_SKIP_CACHE_SIZE : 0;
 
-    if (Settings::values.use_caches_gc.GetValue() && total_used_memory >= EXPECTED_MEMORY) {
+    if (total_used_memory >= EXPECTED_MEMORY) {
         RunGarbageCollector();
     }
     ++frame_tick;
@@ -954,7 +955,7 @@ bool BufferCache<P>::IsRegionCpuModified(VAddr addr, size_t size) {
 template <class P>
 void BufferCache<P>::BindHostIndexBuffer() {
     Buffer& buffer = slot_buffers[index_buffer.buffer_id];
-    TouchBuffer(buffer);
+    TouchBuffer(buffer, index_buffer.buffer_id);
     const u32 offset = buffer.Offset(index_buffer.cpu_addr);
     const u32 size = index_buffer.size;
     SynchronizeBuffer(buffer, index_buffer.cpu_addr, size);
@@ -975,7 +976,7 @@ void BufferCache<P>::BindHostVertexBuffers() {
     for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
         const Binding& binding = vertex_buffers[index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer);
+        TouchBuffer(buffer, binding.buffer_id);
         SynchronizeBuffer(buffer, binding.cpu_addr, binding.size);
         if (!flags[Dirty::VertexBuffer0 + index]) {
             continue;
@@ -1011,7 +1012,7 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
     const VAddr cpu_addr = binding.cpu_addr;
     const u32 size = std::min(binding.size, (*uniform_buffer_sizes)[stage][index]);
     Buffer& buffer = slot_buffers[binding.buffer_id];
-    TouchBuffer(buffer);
+    TouchBuffer(buffer, binding.buffer_id);
     const bool use_fast_buffer = binding.buffer_id != NULL_BUFFER_ID &&
                                  size <= uniform_buffer_skip_cache_size &&
                                  !buffer.IsRegionGpuModified(cpu_addr, size);
@@ -1083,7 +1084,7 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
     ForEachEnabledBit(enabled_storage_buffers[stage], [&](u32 index) {
         const Binding& binding = storage_buffers[stage][index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer);
+        TouchBuffer(buffer, binding.buffer_id);
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
@@ -1128,7 +1129,7 @@ void BufferCache<P>::BindHostTransformFeedbackBuffers() {
     for (u32 index = 0; index < NUM_TRANSFORM_FEEDBACK_BUFFERS; ++index) {
         const Binding& binding = transform_feedback_buffers[index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer);
+        TouchBuffer(buffer, binding.buffer_id);
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
@@ -1148,7 +1149,7 @@ void BufferCache<P>::BindHostComputeUniformBuffers() {
     ForEachEnabledBit(enabled_compute_uniform_buffer_mask, [&](u32 index) {
         const Binding& binding = compute_uniform_buffers[index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer);
+        TouchBuffer(buffer, binding.buffer_id);
         const u32 size = std::min(binding.size, (*compute_uniform_buffer_sizes)[index]);
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
@@ -1168,7 +1169,7 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
     ForEachEnabledBit(enabled_compute_storage_buffers, [&](u32 index) {
         const Binding& binding = compute_storage_buffers[index];
         Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer);
+        TouchBuffer(buffer, binding.buffer_id);
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
@@ -1513,11 +1514,11 @@ BufferId BufferCache<P>::CreateBuffer(VAddr cpu_addr, u32 wanted_size) {
     const OverlapResult overlap = ResolveOverlaps(cpu_addr, wanted_size);
     const u32 size = static_cast<u32>(overlap.end - overlap.begin);
     const BufferId new_buffer_id = slot_buffers.insert(runtime, rasterizer, overlap.begin, size);
-    TouchBuffer(slot_buffers[new_buffer_id]);
     for (const BufferId overlap_id : overlap.ids) {
         JoinOverlap(new_buffer_id, overlap_id, !overlap.has_stream_leap);
     }
     Register(new_buffer_id);
+    TouchBuffer(slot_buffers[new_buffer_id], new_buffer_id);
     return new_buffer_id;
 }
 
@@ -1534,12 +1535,14 @@ void BufferCache<P>::Unregister(BufferId buffer_id) {
 template <class P>
 template <bool insert>
 void BufferCache<P>::ChangeRegister(BufferId buffer_id) {
-    const Buffer& buffer = slot_buffers[buffer_id];
+    Buffer& buffer = slot_buffers[buffer_id];
     const auto size = buffer.SizeBytes();
     if (insert) {
         total_used_memory += Common::AlignUp(size, 1024);
+        buffer.setLRUID(lru_cache.Insert(buffer_id, frame_tick));
     } else {
         total_used_memory -= Common::AlignUp(size, 1024);
+        lru_cache.Free(buffer.getLRUID());
     }
     const VAddr cpu_addr_begin = buffer.CpuAddr();
     const VAddr cpu_addr_end = cpu_addr_begin + size;
@@ -1555,8 +1558,10 @@ void BufferCache<P>::ChangeRegister(BufferId buffer_id) {
 }
 
 template <class P>
-void BufferCache<P>::TouchBuffer(Buffer& buffer) const noexcept {
-    buffer.SetFrameTick(frame_tick);
+void BufferCache<P>::TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept {
+    if (buffer_id != NULL_BUFFER_ID) {
+        lru_cache.Touch(buffer.getLRUID(), frame_tick);
+    }
 }
 
 template <class P>

src/video_core/command_classes/codecs/codec.cpp

@@ -2,10 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <cstring>
 #include <fstream>
 #include <vector>
 #include "common/assert.h"
+#include "common/settings.h"
 #include "video_core/command_classes/codecs/codec.h"
 #include "video_core/command_classes/codecs/h264.h"
 #include "video_core/command_classes/codecs/vp9.h"
@@ -17,10 +17,31 @@ extern "C" {
 }
 
 namespace Tegra {
+namespace {
+constexpr AVPixelFormat PREFERRED_GPU_FMT = AV_PIX_FMT_NV12;
+constexpr AVPixelFormat PREFERRED_CPU_FMT = AV_PIX_FMT_YUV420P;
+
+void AVPacketDeleter(AVPacket* ptr) {
+    av_packet_free(&ptr);
+}
+
+using AVPacketPtr = std::unique_ptr<AVPacket, decltype(&AVPacketDeleter)>;
+
+AVPixelFormat GetGpuFormat(AVCodecContext* av_codec_ctx, const AVPixelFormat* pix_fmts) {
+    for (const AVPixelFormat* p = pix_fmts; *p != AV_PIX_FMT_NONE; ++p) {
+        if (*p == av_codec_ctx->pix_fmt) {
+            return av_codec_ctx->pix_fmt;
+        }
+    }
+    LOG_INFO(Service_NVDRV, "Could not find compatible GPU AV format, falling back to CPU");
+    av_buffer_unref(&av_codec_ctx->hw_device_ctx);
+    av_codec_ctx->pix_fmt = PREFERRED_CPU_FMT;
+    return PREFERRED_CPU_FMT;
+}
+} // namespace
 
 void AVFrameDeleter(AVFrame* ptr) {
-    av_frame_unref(ptr);
-    av_free(ptr);
+    av_frame_free(&ptr);
 }
 
 Codec::Codec(GPU& gpu_, const NvdecCommon::NvdecRegisters& regs)
@@ -32,43 +53,111 @@ Codec::~Codec() {
         return;
     }
     // Free libav memory
-    AVFrame* av_frame{nullptr};
-    avcodec_send_packet(av_codec_ctx, nullptr);
-    av_frame = av_frame_alloc();
-    avcodec_receive_frame(av_codec_ctx, av_frame);
-    avcodec_flush_buffers(av_codec_ctx);
+    avcodec_free_context(&av_codec_ctx);
+    av_buffer_unref(&av_gpu_decoder);
+}
 
-    av_frame_unref(av_frame);
-    av_free(av_frame);
-    avcodec_close(av_codec_ctx);
+bool Codec::CreateGpuAvDevice() {
+#if defined(LIBVA_FOUND)
+    static constexpr std::array<const char*, 3> VAAPI_DRIVERS = {
+        "i915",
+        "iHD",
+        "amdgpu",
+    };
+    AVDictionary* hwdevice_options = nullptr;
+    av_dict_set(&hwdevice_options, "connection_type", "drm", 0);
+    for (const auto& driver : VAAPI_DRIVERS) {
+        av_dict_set(&hwdevice_options, "kernel_driver", driver, 0);
+        const int hwdevice_error = av_hwdevice_ctx_create(&av_gpu_decoder, AV_HWDEVICE_TYPE_VAAPI,
+                                                          nullptr, hwdevice_options, 0);
+        if (hwdevice_error >= 0) {
+            LOG_INFO(Service_NVDRV, "Using VA-API with {}", driver);
+            av_dict_free(&hwdevice_options);
+            av_codec_ctx->pix_fmt = AV_PIX_FMT_VAAPI;
+            return true;
+        }
+        LOG_DEBUG(Service_NVDRV, "VA-API av_hwdevice_ctx_create failed {}", hwdevice_error);
+    }
+    LOG_DEBUG(Service_NVDRV, "VA-API av_hwdevice_ctx_create failed for all drivers");
+    av_dict_free(&hwdevice_options);
+#endif
+    static constexpr auto HW_CONFIG_METHOD = AV_CODEC_HW_CONFIG_METHOD_HW_DEVICE_CTX;
+    static constexpr std::array GPU_DECODER_TYPES{
+        AV_HWDEVICE_TYPE_CUDA,
+#ifdef _WIN32
+        AV_HWDEVICE_TYPE_D3D11VA,
+#else
+        AV_HWDEVICE_TYPE_VDPAU,
+#endif
+    };
+    for (const auto& type : GPU_DECODER_TYPES) {
+        const int hwdevice_res = av_hwdevice_ctx_create(&av_gpu_decoder, type, nullptr, nullptr, 0);
+        if (hwdevice_res < 0) {
+            LOG_DEBUG(Service_NVDRV, "{} av_hwdevice_ctx_create failed {}",
+                      av_hwdevice_get_type_name(type), hwdevice_res);
+            continue;
+        }
+        for (int i = 0;; i++) {
+            const AVCodecHWConfig* config = avcodec_get_hw_config(av_codec, i);
+            if (!config) {
+                LOG_DEBUG(Service_NVDRV, "{} decoder does not support device type {}.",
+                          av_codec->name, av_hwdevice_get_type_name(type));
+                break;
+            }
+            if (config->methods & HW_CONFIG_METHOD && config->device_type == type) {
+                av_codec_ctx->pix_fmt = config->pix_fmt;
+                LOG_INFO(Service_NVDRV, "Using {} GPU decoder", av_hwdevice_get_type_name(type));
+                return true;
+            }
+        }
+    }
+    return false;
+}
+
+void Codec::InitializeAvCodecContext() {
+    av_codec_ctx = avcodec_alloc_context3(av_codec);
+    av_opt_set(av_codec_ctx->priv_data, "tune", "zerolatency", 0);
+}
+
+void Codec::InitializeGpuDecoder() {
+    if (!CreateGpuAvDevice()) {
+        av_buffer_unref(&av_gpu_decoder);
+        return;
+    }
+    auto* hw_device_ctx = av_buffer_ref(av_gpu_decoder);
+    ASSERT_MSG(hw_device_ctx, "av_buffer_ref failed");
+    av_codec_ctx->hw_device_ctx = hw_device_ctx;
+    av_codec_ctx->get_format = GetGpuFormat;
 }
 
 void Codec::Initialize() {
-    AVCodecID codec{AV_CODEC_ID_NONE};
-    switch (current_codec) {
-    case NvdecCommon::VideoCodec::H264:
-        codec = AV_CODEC_ID_H264;
-        break;
-    case NvdecCommon::VideoCodec::Vp9:
-        codec = AV_CODEC_ID_VP9;
-        break;
-    default:
+    const AVCodecID codec = [&] {
+        switch (current_codec) {
+        case NvdecCommon::VideoCodec::H264:
+            return AV_CODEC_ID_H264;
+        case NvdecCommon::VideoCodec::Vp9:
+            return AV_CODEC_ID_VP9;
+        default:
+            UNIMPLEMENTED_MSG("Unknown codec {}", current_codec);
+            return AV_CODEC_ID_NONE;
+        }
+    }();
+    av_codec = avcodec_find_decoder(codec);
+
+    InitializeAvCodecContext();
+    if (Settings::values.nvdec_emulation.GetValue() == Settings::NvdecEmulation::GPU) {
+        InitializeGpuDecoder();
+    }
+    if (const int res = avcodec_open2(av_codec_ctx, av_codec, nullptr); res < 0) {
+        LOG_ERROR(Service_NVDRV, "avcodec_open2() Failed with result {}", res);
+        avcodec_free_context(&av_codec_ctx);
+        av_buffer_unref(&av_gpu_decoder);
         return;
     }
-    av_codec = avcodec_find_decoder(codec);
-    av_codec_ctx = avcodec_alloc_context3(av_codec);
-    av_opt_set(av_codec_ctx->priv_data, "tune", "zerolatency", 0);
-
-    // TODO(ameerj): libavcodec gpu hw acceleration
-
-    const auto av_error = avcodec_open2(av_codec_ctx, av_codec, nullptr);
-    if (av_error < 0) {
-        LOG_ERROR(Service_NVDRV, "avcodec_open2() Failed.");
-        avcodec_close(av_codec_ctx);
-        return;
+    if (!av_codec_ctx->hw_device_ctx) {
+        LOG_INFO(Service_NVDRV, "Using FFmpeg software decoding");
     }
     initialized = true;
-    return;
 }
 
 void Codec::SetTargetCodec(NvdecCommon::VideoCodec codec) {
@@ -80,36 +169,65 @@ void Codec::SetTargetCodec(NvdecCommon::VideoCodec codec) {
 
 void Codec::Decode() {
     const bool is_first_frame = !initialized;
-    if (!initialized) {
+    if (is_first_frame) {
         Initialize();
     }
-
+    if (!initialized) {
+        return;
+    }
     bool vp9_hidden_frame = false;
-    AVPacket packet{};
-    av_init_packet(&packet);
     std::vector<u8> frame_data;
-
     if (current_codec == NvdecCommon::VideoCodec::H264) {
         frame_data = h264_decoder->ComposeFrameHeader(state, is_first_frame);
     } else if (current_codec == NvdecCommon::VideoCodec::Vp9) {
         frame_data = vp9_decoder->ComposeFrameHeader(state);
         vp9_hidden_frame = vp9_decoder->WasFrameHidden();
     }
-
-    packet.data = frame_data.data();
-    packet.size = static_cast<s32>(frame_data.size());
-
-    avcodec_send_packet(av_codec_ctx, &packet);
-
-    if (!vp9_hidden_frame) {
-        // Only receive/store visible frames
-        AVFramePtr frame = AVFramePtr{av_frame_alloc(), AVFrameDeleter};
-        avcodec_receive_frame(av_codec_ctx, frame.get());
-        av_frames.push(std::move(frame));
-        // Limit queue to 10 frames. Workaround for ZLA decode and queue spam
-        if (av_frames.size() > 10) {
-            av_frames.pop();
-        }
+    AVPacketPtr packet{av_packet_alloc(), AVPacketDeleter};
+    if (!packet) {
+        LOG_ERROR(Service_NVDRV, "av_packet_alloc failed");
+        return;
+    }
+    packet->data = frame_data.data();
+    packet->size = static_cast<s32>(frame_data.size());
+    if (const int res = avcodec_send_packet(av_codec_ctx, packet.get()); res != 0) {
+        LOG_DEBUG(Service_NVDRV, "avcodec_send_packet error {}", res);
+        return;
+    }
+    // Only receive/store visible frames
+    if (vp9_hidden_frame) {
+        return;
+    }
+    AVFramePtr initial_frame{av_frame_alloc(), AVFrameDeleter};
+    AVFramePtr final_frame{nullptr, AVFrameDeleter};
+    ASSERT_MSG(initial_frame, "av_frame_alloc initial_frame failed");
+    if (const int ret = avcodec_receive_frame(av_codec_ctx, initial_frame.get()); ret) {
+        LOG_DEBUG(Service_NVDRV, "avcodec_receive_frame error {}", ret);
+        return;
+    }
+    if (initial_frame->width == 0 || initial_frame->height == 0) {
+        LOG_WARNING(Service_NVDRV, "Zero width or height in frame");
+        return;
+    }
+    if (av_codec_ctx->hw_device_ctx) {
+        final_frame = AVFramePtr{av_frame_alloc(), AVFrameDeleter};
+        ASSERT_MSG(final_frame, "av_frame_alloc final_frame failed");
+        // Can't use AV_PIX_FMT_YUV420P and share code with software decoding in vic.cpp
+        // because Intel drivers crash unless using AV_PIX_FMT_NV12
+        final_frame->format = PREFERRED_GPU_FMT;
+        const int ret = av_hwframe_transfer_data(final_frame.get(), initial_frame.get(), 0);
+        ASSERT_MSG(!ret, "av_hwframe_transfer_data error {}", ret);
+    } else {
+        final_frame = std::move(initial_frame);
+    }
+    if (final_frame->format != PREFERRED_CPU_FMT && final_frame->format != PREFERRED_GPU_FMT) {
+        UNIMPLEMENTED_MSG("Unexpected video format: {}", final_frame->format);
+        return;
+    }
+    av_frames.push(std::move(final_frame));
+    if (av_frames.size() > 10) {
+        LOG_TRACE(Service_NVDRV, "av_frames.push overflow dropped frame");
+        av_frames.pop();
     }
 }
 
@@ -119,7 +237,6 @@ AVFramePtr Codec::GetCurrentFrame() {
     if (av_frames.empty()) {
         return AVFramePtr{nullptr, AVFrameDeleter};
     }
-
     AVFramePtr frame = std::move(av_frames.front());
     av_frames.pop();
     return frame;
@@ -144,6 +261,5 @@ std::string_view Codec::GetCurrentCodecName() const {
     default:
         return "Unknown";
     }
-};
-
+}
 } // namespace Tegra

src/video_core/command_classes/codecs/codec.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <memory>
+#include <string_view>
 #include <queue>
 #include "common/common_types.h"
 #include "video_core/command_classes/nvdec_common.h"
@@ -22,7 +23,6 @@ extern "C" {
 
 namespace Tegra {
 class GPU;
-struct VicRegisters;
 
 void AVFrameDeleter(AVFrame* ptr);
 using AVFramePtr = std::unique_ptr<AVFrame, decltype(&AVFrameDeleter)>;
@@ -51,15 +51,23 @@ public:
 
     /// Returns the value of current_codec
     [[nodiscard]] NvdecCommon::VideoCodec GetCurrentCodec() const;
+
     /// Return name of the current codec
     [[nodiscard]] std::string_view GetCurrentCodecName() const;
 
 private:
+    void InitializeAvCodecContext();
+
+    void InitializeGpuDecoder();
+
+    bool CreateGpuAvDevice();
+
     bool initialized{};
     NvdecCommon::VideoCodec current_codec{NvdecCommon::VideoCodec::None};
 
     AVCodec* av_codec{nullptr};
     AVCodecContext* av_codec_ctx{nullptr};
+    AVBufferRef* av_gpu_decoder{nullptr};
 
     GPU& gpu;
     const NvdecCommon::NvdecRegisters& state;

src/video_core/command_classes/codecs/h264.cpp

@@ -95,7 +95,8 @@ const std::vector<u8>& H264::ComposeFrameHeader(const NvdecCommon::NvdecRegister
     const s32 pic_height = context.h264_parameter_set.frame_height_in_map_units /
                            (context.h264_parameter_set.frame_mbs_only_flag ? 1 : 2);
 
-    writer.WriteUe(16);
+    // TODO (ameerj): Where do we get this number, it seems to be particular for each stream
+    writer.WriteUe(6); // Max number of reference frames
     writer.WriteBit(false);
     writer.WriteUe(context.h264_parameter_set.pic_width_in_mbs - 1);
     writer.WriteUe(pic_height - 1);

src/video_core/command_classes/codecs/vp9.cpp

@@ -11,6 +11,9 @@
 
 namespace Tegra::Decoder {
 namespace {
+constexpr u32 diff_update_probability = 252;
+constexpr u32 frame_sync_code = 0x498342;
+
 // Default compressed header probabilities once frame context resets
 constexpr Vp9EntropyProbs default_probs{
     .y_mode_prob{
@@ -361,8 +364,7 @@ Vp9PictureInfo VP9::GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state)
     InsertEntropy(state.vp9_entropy_probs_offset, vp9_info.entropy);
 
     // surface_luma_offset[0:3] contains the address of the reference frame offsets in the following
-    // order: last, golden, altref, current. It may be worthwhile to track the updates done here
-    // to avoid buffering frame data needed for reference frame updating in the header composition.
+    // order: last, golden, altref, current.
     std::copy(state.surface_luma_offset.begin(), state.surface_luma_offset.begin() + 4,
               vp9_info.frame_offsets.begin());
 
@@ -384,40 +386,25 @@ Vp9FrameContainer VP9::GetCurrentFrame(const NvdecCommon::NvdecRegisters& state)
         gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, current_frame.bit_stream.data(),
                                       current_frame.info.bitstream_size);
     }
-    // Buffer two frames, saving the last show frame info
-    if (!next_next_frame.bit_stream.empty()) {
+    if (!next_frame.bit_stream.empty()) {
         Vp9FrameContainer temp{
             .info = current_frame.info,
             .bit_stream = std::move(current_frame.bit_stream),
         };
-        next_next_frame.info.show_frame = current_frame.info.last_frame_shown;
-        current_frame.info = next_next_frame.info;
-        current_frame.bit_stream = std::move(next_next_frame.bit_stream);
-        next_next_frame = std::move(temp);
-
-        if (!next_frame.bit_stream.empty()) {
-            Vp9FrameContainer temp2{
-                .info = current_frame.info,
-                .bit_stream = std::move(current_frame.bit_stream),
-            };
-            next_frame.info.show_frame = current_frame.info.last_frame_shown;
-            current_frame.info = next_frame.info;
-            current_frame.bit_stream = std::move(next_frame.bit_stream);
-            next_frame = std::move(temp2);
-        } else {
-            next_frame.info = current_frame.info;
-            next_frame.bit_stream = std::move(current_frame.bit_stream);
-        }
+        next_frame.info.show_frame = current_frame.info.last_frame_shown;
+        current_frame.info = next_frame.info;
+        current_frame.bit_stream = std::move(next_frame.bit_stream);
+        next_frame = std::move(temp);
     } else {
-        next_next_frame.info = current_frame.info;
-        next_next_frame.bit_stream = std::move(current_frame.bit_stream);
+        next_frame.info = current_frame.info;
+        next_frame.bit_stream = current_frame.bit_stream;
     }
     return current_frame;
 }
 
 std::vector<u8> VP9::ComposeCompressedHeader() {
     VpxRangeEncoder writer{};
-    const bool update_probs = current_frame_info.show_frame && !current_frame_info.is_key_frame;
+    const bool update_probs = !current_frame_info.is_key_frame && current_frame_info.show_frame;
     if (!current_frame_info.lossless) {
         if (static_cast<u32>(current_frame_info.transform_mode) >= 3) {
             writer.Write(3, 2);
@@ -613,86 +600,64 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
 
         // Reset context
         prev_frame_probs = default_probs;
-        swap_next_golden = false;
+        swap_ref_indices = false;
         loop_filter_ref_deltas.fill(0);
         loop_filter_mode_deltas.fill(0);
-
-        // allow frames offsets to stabilize before checking for golden frames
-        grace_period = 4;
-
-        // On key frames, all frame slots are set to the current frame,
-        // so the value of the selected slot doesn't really matter.
-        frame_ctxs.fill({current_frame_number, false, default_probs});
+        frame_ctxs.fill(default_probs);
 
         // intra only, meaning the frame can be recreated with no other references
         current_frame_info.intra_only = true;
-
     } else {
-
         if (!current_frame_info.show_frame) {
             uncomp_writer.WriteBit(current_frame_info.intra_only);
-            if (!current_frame_info.last_frame_was_key) {
-                swap_next_golden = !swap_next_golden;
-            }
         } else {
             current_frame_info.intra_only = false;
         }
         if (!current_frame_info.error_resilient_mode) {
             uncomp_writer.WriteU(0, 2); // Reset frame context.
         }
-
-        // Last, Golden, Altref frames
-        std::array<s32, 3> ref_frame_index{0, 1, 2};
-
-        // Set when next frame is hidden
-        // altref and golden references are swapped
-        if (swap_next_golden) {
-            ref_frame_index = std::array<s32, 3>{0, 2, 1};
+        const auto& curr_offsets = current_frame_info.frame_offsets;
+        const auto& next_offsets = next_frame.info.frame_offsets;
+        const bool ref_frames_different = curr_offsets[1] != curr_offsets[2];
+        const bool next_references_swap =
+            (next_offsets[1] == curr_offsets[2]) || (next_offsets[2] == curr_offsets[1]);
+        const bool needs_ref_swap = ref_frames_different && next_references_swap;
+        if (needs_ref_swap) {
+            swap_ref_indices = !swap_ref_indices;
         }
+        union {
+            u32 raw;
+            BitField<0, 1, u32> refresh_last;
+            BitField<1, 2, u32> refresh_golden;
+            BitField<2, 1, u32> refresh_alt;
+        } refresh_frame_flags;
 
-        // update Last Frame
-        u64 refresh_frame_flags = 1;
-
-        // golden frame may refresh, determined if the next golden frame offset is changed
-        bool golden_refresh = false;
-        if (grace_period <= 0) {
-            for (s32 index = 1; index < 3; ++index) {
-                if (current_frame_info.frame_offsets[index] !=
-                    next_frame.info.frame_offsets[index]) {
-                    current_frame_info.refresh_frame[index] = true;
-                    golden_refresh = true;
-                    grace_period = 3;
-                }
+        refresh_frame_flags.raw = 0;
+        for (u32 index = 0; index < 3; ++index) {
+            // Refresh indices that use the current frame as an index
+            if (curr_offsets[3] == next_offsets[index]) {
+                refresh_frame_flags.raw |= 1u << index;
             }
         }
-
-        if (current_frame_info.show_frame &&
-            (!next_frame.info.show_frame || next_frame.info.is_key_frame)) {
-            // Update golden frame
-            refresh_frame_flags = swap_next_golden ? 2 : 4;
+        if (swap_ref_indices) {
+            const u32 temp = refresh_frame_flags.refresh_golden;
+            refresh_frame_flags.refresh_golden.Assign(refresh_frame_flags.refresh_alt.Value());
+            refresh_frame_flags.refresh_alt.Assign(temp);
         }
-
-        if (!current_frame_info.show_frame) {
-            // Update altref
-            refresh_frame_flags = swap_next_golden ? 2 : 4;
-        } else if (golden_refresh) {
-            refresh_frame_flags = 3;
-        }
-
         if (current_frame_info.intra_only) {
             uncomp_writer.WriteU(frame_sync_code, 24);
-            uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8);
+            uncomp_writer.WriteU(refresh_frame_flags.raw, 8);
             uncomp_writer.WriteU(current_frame_info.frame_size.width - 1, 16);
             uncomp_writer.WriteU(current_frame_info.frame_size.height - 1, 16);
             uncomp_writer.WriteBit(false); // Render and frame size different.
         } else {
-            uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8);
-
-            for (s32 index = 1; index < 4; index++) {
+            const bool swap_indices = needs_ref_swap ^ swap_ref_indices;
+            const auto ref_frame_index = swap_indices ? std::array{0, 2, 1} : std::array{0, 1, 2};
+            uncomp_writer.WriteU(refresh_frame_flags.raw, 8);
+            for (size_t index = 1; index < 4; index++) {
                 uncomp_writer.WriteU(ref_frame_index[index - 1], 3);
                 uncomp_writer.WriteU(current_frame_info.ref_frame_sign_bias[index], 1);
             }
-
             uncomp_writer.WriteBit(true);  // Frame size with refs.
             uncomp_writer.WriteBit(false); // Render and frame size different.
             uncomp_writer.WriteBit(current_frame_info.allow_high_precision_mv);
@@ -714,10 +679,9 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
         frame_ctx_idx = 1;
     }
 
-    uncomp_writer.WriteU(frame_ctx_idx, 2); // Frame context index.
-    prev_frame_probs =
-        frame_ctxs[frame_ctx_idx].probs; // reference probabilities for compressed header
-    frame_ctxs[frame_ctx_idx] = {current_frame_number, false, current_frame_info.entropy};
+    uncomp_writer.WriteU(frame_ctx_idx, 2);       // Frame context index.
+    prev_frame_probs = frame_ctxs[frame_ctx_idx]; // reference probabilities for compressed header
+    frame_ctxs[frame_ctx_idx] = current_frame_info.entropy;
 
     uncomp_writer.WriteU(current_frame_info.first_level, 6);
     uncomp_writer.WriteU(current_frame_info.sharpness_level, 3);
@@ -778,6 +742,7 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
     uncomp_writer.WriteDeltaQ(current_frame_info.uv_dc_delta_q);
     uncomp_writer.WriteDeltaQ(current_frame_info.uv_ac_delta_q);
 
+    ASSERT(!current_frame_info.segment_enabled);
     uncomp_writer.WriteBit(false); // Segmentation enabled (TODO).
 
     const s32 min_tile_cols_log2 = CalcMinLog2TileCols(current_frame_info.frame_size.width);
@@ -812,7 +777,6 @@ const std::vector<u8>& VP9::ComposeFrameHeader(const NvdecCommon::NvdecRegisters
         current_frame_info = curr_frame.info;
         bitstream = std::move(curr_frame.bit_stream);
     }
-
     // The uncompressed header routine sets PrevProb parameters needed for the compressed header
     auto uncomp_writer = ComposeUncompressedHeader();
     std::vector<u8> compressed_header = ComposeCompressedHeader();
@@ -828,13 +792,6 @@ const std::vector<u8>& VP9::ComposeFrameHeader(const NvdecCommon::NvdecRegisters
               frame.begin() + uncompressed_header.size());
     std::copy(bitstream.begin(), bitstream.end(),
               frame.begin() + uncompressed_header.size() + compressed_header.size());
-
-    // keep track of frame number
-    current_frame_number++;
-    grace_period--;
-
-    // don't display hidden frames
-    hidden = !current_frame_info.show_frame;
     return frame;
 }
 

src/video_core/command_classes/codecs/vp9.h

@@ -14,7 +14,6 @@
 
 namespace Tegra {
 class GPU;
-enum class FrameType { KeyFrame = 0, InterFrame = 1 };
 namespace Decoder {
 
 /// The VpxRangeEncoder, and VpxBitStreamWriter classes are used to compose the
@@ -124,7 +123,7 @@ public:
 
     /// Returns true if the most recent frame was a hidden frame.
     [[nodiscard]] bool WasFrameHidden() const {
-        return hidden;
+        return !current_frame_info.show_frame;
     }
 
 private:
@@ -178,19 +177,12 @@ private:
     std::array<s8, 4> loop_filter_ref_deltas{};
     std::array<s8, 2> loop_filter_mode_deltas{};
 
-    bool hidden = false;
-    s64 current_frame_number = -2; // since we buffer 2 frames
-    s32 grace_period = 6;          // frame offsets need to stabilize
-    std::array<FrameContexts, 4> frame_ctxs{};
     Vp9FrameContainer next_frame{};
-    Vp9FrameContainer next_next_frame{};
-    bool swap_next_golden{};
+    std::array<Vp9EntropyProbs, 4> frame_ctxs{};
+    bool swap_ref_indices{};
 
     Vp9PictureInfo current_frame_info{};
     Vp9EntropyProbs prev_frame_probs{};
-
-    s32 diff_update_probability = 252;
-    s32 frame_sync_code = 0x498342;
 };
 
 } // namespace Decoder

src/video_core/command_classes/codecs/vp9_types.h

@@ -22,7 +22,7 @@ struct Vp9FrameDimensions {
 };
 static_assert(sizeof(Vp9FrameDimensions) == 0x8, "Vp9 Vp9FrameDimensions is an invalid size");
 
-enum FrameFlags : u32 {
+enum class FrameFlags : u32 {
     IsKeyFrame = 1 << 0,
     LastFrameIsKeyFrame = 1 << 1,
     FrameSizeChanged = 1 << 2,
@@ -30,6 +30,7 @@ enum FrameFlags : u32 {
     LastShowFrame = 1 << 4,
     IntraOnly = 1 << 5,
 };
+DECLARE_ENUM_FLAG_OPERATORS(FrameFlags)
 
 enum class TxSize {
     Tx4x4 = 0,   // 4x4 transform
@@ -92,44 +93,34 @@ struct Vp9EntropyProbs {
 static_assert(sizeof(Vp9EntropyProbs) == 0x7B4, "Vp9EntropyProbs is an invalid size");
 
 struct Vp9PictureInfo {
-    bool is_key_frame;
-    bool intra_only;
-    bool last_frame_was_key;
-    bool frame_size_changed;
-    bool error_resilient_mode;
-    bool last_frame_shown;
-    bool show_frame;
+    u32 bitstream_size;
+    std::array<u64, 4> frame_offsets;
     std::array<s8, 4> ref_frame_sign_bias;
     s32 base_q_index;
     s32 y_dc_delta_q;
     s32 uv_dc_delta_q;
     s32 uv_ac_delta_q;
-    bool lossless;
     s32 transform_mode;
-    bool allow_high_precision_mv;
     s32 interp_filter;
     s32 reference_mode;
-    s8 comp_fixed_ref;
-    std::array<s8, 2> comp_var_ref;
     s32 log2_tile_cols;
     s32 log2_tile_rows;
-    bool segment_enabled;
-    bool segment_map_update;
-    bool segment_map_temporal_update;
-    s32 segment_abs_delta;
-    std::array<u32, 8> segment_feature_enable;
-    std::array<std::array<s16, 4>, 8> segment_feature_data;
-    bool mode_ref_delta_enabled;
-    bool use_prev_in_find_mv_refs;
     std::array<s8, 4> ref_deltas;
     std::array<s8, 2> mode_deltas;
     Vp9EntropyProbs entropy;
     Vp9FrameDimensions frame_size;
     u8 first_level;
     u8 sharpness_level;
-    u32 bitstream_size;
-    std::array<u64, 4> frame_offsets;
-    std::array<bool, 4> refresh_frame;
+    bool is_key_frame;
+    bool intra_only;
+    bool last_frame_was_key;
+    bool error_resilient_mode;
+    bool last_frame_shown;
+    bool show_frame;
+    bool lossless;
+    bool allow_high_precision_mv;
+    bool segment_enabled;
+    bool mode_ref_delta_enabled;
 };
 
 struct Vp9FrameContainer {
@@ -145,7 +136,7 @@ struct PictureInfo {
     Vp9FrameDimensions golden_frame_size;  ///< 0x50
     Vp9FrameDimensions alt_frame_size;     ///< 0x58
     Vp9FrameDimensions current_frame_size; ///< 0x60
-    u32 vp9_flags;                         ///< 0x68
+    FrameFlags vp9_flags;                  ///< 0x68
     std::array<s8, 4> ref_frame_sign_bias; ///< 0x6C
     u8 first_level;                        ///< 0x70
     u8 sharpness_level;                    ///< 0x71
@@ -158,60 +149,43 @@ struct PictureInfo {
     u8 allow_high_precision_mv;            ///< 0x78
     u8 interp_filter;                      ///< 0x79
     u8 reference_mode;                     ///< 0x7A
-    s8 comp_fixed_ref;                     ///< 0x7B
-    std::array<s8, 2> comp_var_ref;        ///< 0x7C
+    INSERT_PADDING_BYTES_NOINIT(3);        ///< 0x7B
     u8 log2_tile_cols;                     ///< 0x7E
     u8 log2_tile_rows;                     ///< 0x7F
     Segmentation segmentation;             ///< 0x80
     LoopFilter loop_filter;                ///< 0xE4
-    INSERT_PADDING_BYTES_NOINIT(5);        ///< 0xEB
-    u32 surface_params;                    ///< 0xF0
-    INSERT_PADDING_WORDS_NOINIT(3);        ///< 0xF4
+    INSERT_PADDING_BYTES_NOINIT(21);       ///< 0xEB
 
     [[nodiscard]] Vp9PictureInfo Convert() const {
         return {
-            .is_key_frame = (vp9_flags & FrameFlags::IsKeyFrame) != 0,
-            .intra_only = (vp9_flags & FrameFlags::IntraOnly) != 0,
-            .last_frame_was_key = (vp9_flags & FrameFlags::LastFrameIsKeyFrame) != 0,
-            .frame_size_changed = (vp9_flags & FrameFlags::FrameSizeChanged) != 0,
-            .error_resilient_mode = (vp9_flags & FrameFlags::ErrorResilientMode) != 0,
-            .last_frame_shown = (vp9_flags & FrameFlags::LastShowFrame) != 0,
-            .show_frame = false,
+            .bitstream_size = bitstream_size,
+            .frame_offsets{},
             .ref_frame_sign_bias = ref_frame_sign_bias,
             .base_q_index = base_q_index,
             .y_dc_delta_q = y_dc_delta_q,
             .uv_dc_delta_q = uv_dc_delta_q,
             .uv_ac_delta_q = uv_ac_delta_q,
-            .lossless = lossless != 0,
             .transform_mode = tx_mode,
-            .allow_high_precision_mv = allow_high_precision_mv != 0,
             .interp_filter = interp_filter,
             .reference_mode = reference_mode,
-            .comp_fixed_ref = comp_fixed_ref,
-            .comp_var_ref = comp_var_ref,
             .log2_tile_cols = log2_tile_cols,
             .log2_tile_rows = log2_tile_rows,
-            .segment_enabled = segmentation.enabled != 0,
-            .segment_map_update = segmentation.update_map != 0,
-            .segment_map_temporal_update = segmentation.temporal_update != 0,
-            .segment_abs_delta = segmentation.abs_delta,
-            .segment_feature_enable = segmentation.feature_mask,
-            .segment_feature_data = segmentation.feature_data,
-            .mode_ref_delta_enabled = loop_filter.mode_ref_delta_enabled != 0,
-            .use_prev_in_find_mv_refs = !(vp9_flags == (FrameFlags::ErrorResilientMode)) &&
-                                        !(vp9_flags == (FrameFlags::FrameSizeChanged)) &&
-                                        !(vp9_flags == (FrameFlags::IntraOnly)) &&
-                                        (vp9_flags == (FrameFlags::LastShowFrame)) &&
-                                        !(vp9_flags == (FrameFlags::LastFrameIsKeyFrame)),
             .ref_deltas = loop_filter.ref_deltas,
             .mode_deltas = loop_filter.mode_deltas,
             .entropy{},
             .frame_size = current_frame_size,
             .first_level = first_level,
             .sharpness_level = sharpness_level,
-            .bitstream_size = bitstream_size,
-            .frame_offsets{},
-            .refresh_frame{},
+            .is_key_frame = True(vp9_flags & FrameFlags::IsKeyFrame),
+            .intra_only = True(vp9_flags & FrameFlags::IntraOnly),
+            .last_frame_was_key = True(vp9_flags & FrameFlags::LastFrameIsKeyFrame),
+            .error_resilient_mode = True(vp9_flags & FrameFlags::ErrorResilientMode),
+            .last_frame_shown = True(vp9_flags & FrameFlags::LastShowFrame),
+            .show_frame = true,
+            .lossless = lossless != 0,
+            .allow_high_precision_mv = allow_high_precision_mv != 0,
+            .segment_enabled = segmentation.enabled != 0,
+            .mode_ref_delta_enabled = loop_filter.mode_ref_delta_enabled != 0,
         };
     }
 };
@@ -296,12 +270,6 @@ struct RefPoolElement {
     bool refresh{};
 };
 
-struct FrameContexts {
-    s64 from;
-    bool adapted;
-    Vp9EntropyProbs probs;
-};
-
 #define ASSERT_POSITION(field_name, position)                                                      \
     static_assert(offsetof(Vp9EntropyProbs, field_name) == position,                               \
                   "Field " #field_name " has invalid position")
@@ -322,7 +290,6 @@ ASSERT_POSITION(last_frame_size, 0x48);
 ASSERT_POSITION(first_level, 0x70);
 ASSERT_POSITION(segmentation, 0x80);
 ASSERT_POSITION(loop_filter, 0xE4);
-ASSERT_POSITION(surface_params, 0xF0);
 #undef ASSERT_POSITION
 
 #define ASSERT_POSITION(field_name, position)                                                      \

src/video_core/command_classes/nvdec.cpp

@@ -39,7 +39,7 @@ void Nvdec::Execute() {
         codec->Decode();
         break;
     default:
-        UNIMPLEMENTED_MSG("Unknown codec {}", static_cast<u32>(codec->GetCurrentCodec()));
+        UNIMPLEMENTED_MSG("Codec {}", codec->GetCurrentCodecName());
         break;
     }
 }

src/video_core/command_classes/vic.cpp

@@ -46,11 +46,8 @@ void Vic::ProcessMethod(Method method, u32 argument) {
     case Method::SetOutputSurfaceLumaOffset:
         output_surface_luma_address = arg;
         break;
-    case Method::SetOutputSurfaceChromaUOffset:
-        output_surface_chroma_u_address = arg;
-        break;
-    case Method::SetOutputSurfaceChromaVOffset:
-        output_surface_chroma_v_address = arg;
+    case Method::SetOutputSurfaceChromaOffset:
+        output_surface_chroma_address = arg;
         break;
     default:
         break;
@@ -65,11 +62,10 @@ void Vic::Execute() {
     const VicConfig config{gpu.MemoryManager().Read<u64>(config_struct_address + 0x20)};
     const AVFramePtr frame_ptr = nvdec_processor->GetFrame();
     const auto* frame = frame_ptr.get();
-    if (!frame || frame->width == 0 || frame->height == 0) {
+    if (!frame) {
         return;
     }
-    const VideoPixelFormat pixel_format =
-        static_cast<VideoPixelFormat>(config.pixel_format.Value());
+    const auto pixel_format = static_cast<VideoPixelFormat>(config.pixel_format.Value());
     switch (pixel_format) {
     case VideoPixelFormat::BGRA8:
     case VideoPixelFormat::RGBA8: {
@@ -83,37 +79,37 @@ void Vic::Execute() {
             sws_freeContext(scaler_ctx);
             scaler_ctx = nullptr;
 
-            // FFmpeg returns all frames in YUV420, convert it into expected format
-            scaler_ctx =
-                sws_getContext(frame->width, frame->height, AV_PIX_FMT_YUV420P, frame->width,
-                               frame->height, target_format, 0, nullptr, nullptr, nullptr);
+            // Frames are decoded into either YUV420 or NV12 formats. Convert to desired format
+            scaler_ctx = sws_getContext(frame->width, frame->height,
+                                        static_cast<AVPixelFormat>(frame->format), frame->width,
+                                        frame->height, target_format, 0, nullptr, nullptr, nullptr);
 
             scaler_width = frame->width;
             scaler_height = frame->height;
         }
         // Get Converted frame
-        const std::size_t linear_size = frame->width * frame->height * 4;
+        const u32 width = static_cast<u32>(frame->width);
+        const u32 height = static_cast<u32>(frame->height);
+        const std::size_t linear_size = width * height * 4;
 
         // Only allocate frame_buffer once per stream, as the size is not expected to change
         if (!converted_frame_buffer) {
             converted_frame_buffer = AVMallocPtr{static_cast<u8*>(av_malloc(linear_size)), av_free};
         }
-
-        const int converted_stride{frame->width * 4};
+        const std::array<int, 4> converted_stride{frame->width * 4, frame->height * 4, 0, 0};
         u8* const converted_frame_buf_addr{converted_frame_buffer.get()};
 
         sws_scale(scaler_ctx, frame->data, frame->linesize, 0, frame->height,
-                  &converted_frame_buf_addr, &converted_stride);
+                  &converted_frame_buf_addr, converted_stride.data());
 
         const u32 blk_kind = static_cast<u32>(config.block_linear_kind);
         if (blk_kind != 0) {
             // swizzle pitch linear to block linear
             const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
-            const auto size = Tegra::Texture::CalculateSize(true, 4, frame->width, frame->height, 1,
-                                                            block_height, 0);
+            const auto size =
+                Tegra::Texture::CalculateSize(true, 4, width, height, 1, block_height, 0);
             luma_buffer.resize(size);
-            Tegra::Texture::SwizzleSubrect(frame->width, frame->height, frame->width * 4,
-                                           frame->width, 4, luma_buffer.data(),
+            Tegra::Texture::SwizzleSubrect(width, height, width * 4, width, 4, luma_buffer.data(),
                                            converted_frame_buffer.get(), block_height, 0, 0);
 
             gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(), size);
@@ -131,41 +127,65 @@ void Vic::Execute() {
         const std::size_t surface_height = config.surface_height_minus1 + 1;
         const auto frame_width = std::min(surface_width, static_cast<size_t>(frame->width));
         const auto frame_height = std::min(surface_height, static_cast<size_t>(frame->height));
-        const std::size_t half_width = frame_width / 2;
-        const std::size_t half_height = frame_height / 2;
-        const std::size_t aligned_width = (surface_width + 0xff) & ~0xff;
+        const std::size_t aligned_width = (surface_width + 0xff) & ~0xffUL;
 
-        const auto* luma_ptr = frame->data[0];
-        const auto* chroma_b_ptr = frame->data[1];
-        const auto* chroma_r_ptr = frame->data[2];
         const auto stride = static_cast<size_t>(frame->linesize[0]);
-        const auto half_stride = static_cast<size_t>(frame->linesize[1]);
 
         luma_buffer.resize(aligned_width * surface_height);
         chroma_buffer.resize(aligned_width * surface_height / 2);
 
         // Populate luma buffer
+        const u8* luma_src = frame->data[0];
         for (std::size_t y = 0; y < frame_height; ++y) {
             const std::size_t src = y * stride;
             const std::size_t dst = y * aligned_width;
             for (std::size_t x = 0; x < frame_width; ++x) {
-                luma_buffer[dst + x] = luma_ptr[src + x];
+                luma_buffer[dst + x] = luma_src[src + x];
             }
         }
         gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(),
                                        luma_buffer.size());
 
-        // Populate chroma buffer from both channels with interleaving.
-        for (std::size_t y = 0; y < half_height; ++y) {
-            const std::size_t src = y * half_stride;
-            const std::size_t dst = y * aligned_width;
+        // Chroma
+        const std::size_t half_height = frame_height / 2;
+        const auto half_stride = static_cast<size_t>(frame->linesize[1]);
 
-            for (std::size_t x = 0; x < half_width; ++x) {
-                chroma_buffer[dst + x * 2] = chroma_b_ptr[src + x];
-                chroma_buffer[dst + x * 2 + 1] = chroma_r_ptr[src + x];
+        switch (frame->format) {
+        case AV_PIX_FMT_YUV420P: {
+            // Frame from FFmpeg software
+            // Populate chroma buffer from both channels with interleaving.
+            const std::size_t half_width = frame_width / 2;
+            const u8* chroma_b_src = frame->data[1];
+            const u8* chroma_r_src = frame->data[2];
+            for (std::size_t y = 0; y < half_height; ++y) {
+                const std::size_t src = y * half_stride;
+                const std::size_t dst = y * aligned_width;
+
+                for (std::size_t x = 0; x < half_width; ++x) {
+                    chroma_buffer[dst + x * 2] = chroma_b_src[src + x];
+                    chroma_buffer[dst + x * 2 + 1] = chroma_r_src[src + x];
+                }
             }
+            break;
         }
-        gpu.MemoryManager().WriteBlock(output_surface_chroma_u_address, chroma_buffer.data(),
+        case AV_PIX_FMT_NV12: {
+            // Frame from VA-API hardware
+            // This is already interleaved so just copy
+            const u8* chroma_src = frame->data[1];
+            for (std::size_t y = 0; y < half_height; ++y) {
+                const std::size_t src = y * stride;
+                const std::size_t dst = y * aligned_width;
+                for (std::size_t x = 0; x < frame_width; ++x) {
+                    chroma_buffer[dst + x] = chroma_src[src + x];
+                }
+            }
+            break;
+        }
+        default:
+            UNREACHABLE();
+            break;
+        }
+        gpu.MemoryManager().WriteBlock(output_surface_chroma_address, chroma_buffer.data(),
                                        chroma_buffer.size());
         break;
     }