tas_space

Debug Config Ui: Clean-up and nits

CMakeLists.txt

@@ -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")
 
@@ -579,20 +586,23 @@ 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 (${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
-            Include(FindPkgConfig REQUIRED)
-            pkg_check_modules(LIBVA libva)
-        endif()
         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)
-            set(FFmpeg_LIBVA_LIBRARIES
+            list(APPEND FFmpeg_HWACCEL_LIBRARIES
                 ${LIBDRM_LIBRARIES}
                 ${X11_LIBRARIES}
                 ${LIBVA-DRM_LIBRARIES}
@@ -602,11 +612,56 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                 --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 ${VDPAU_VERSION} found")
+        else()
+            list(APPEND FFmpeg_HWACCEL_FLAGS --disable-vdpau)
+        endif()
+
         # `configure` parameters builds only exactly what yuzu needs from FFmpeg
         # `--disable-vdpau` is needed to avoid linking issues
         add_custom_command(
@@ -624,7 +679,6 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                     --disable-network
                     --disable-postproc
                     --disable-swresample
-                    --disable-vdpau
                     --enable-decoder=h264
                     --enable-decoder=vp9
                     --cc="${CMAKE_C_COMPILER}"
@@ -653,15 +707,26 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                 ${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 ${FFmpeg_BUILD_LIBRARIES} ffmpeg-configure)
         link_libraries(${FFmpeg_LIBVA_LIBRARIES})
-        set(FFmpeg_LIBRARIES ${FFmpeg_LIBVA_LIBRARIES} ${FFmpeg_BUILD_LIBRARIES}
+        set(FFmpeg_LIBRARIES ${FFmpeg_BUILD_LIBRARIES} ${FFmpeg_HWACCEL_LIBRARIES}
             CACHE PATH "Paths to FFmpeg libraries" FORCE)
         unset(FFmpeg_BUILD_LIBRARIES)
-        unset(FFmpeg_LIBVA_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 ${FFmpeg_VERSION}")
@@ -670,12 +735,13 @@ if (YUZU_USE_BUNDLED_FFMPEG)
         endif()
     else() # WIN32
         # Use yuzu FFmpeg binaries
-        set(FFmpeg_EXT_NAME "ffmpeg-4.3.1")
+        set(FFmpeg_EXT_NAME "ffmpeg-4.4")
         set(FFmpeg_PATH "${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

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/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/fs/fs_paths.h

@@ -21,6 +21,7 @@
 #define SCREENSHOTS_DIR "screenshots"
 #define SDMC_DIR "sdmc"
 #define SHADER_DIR "shader"
+#define TAS_DIR "tas"
 
 // yuzu-specific files
 

src/common/fs/path_util.cpp

@@ -116,6 +116,7 @@ private:
         GenerateYuzuPath(YuzuPath::ScreenshotsDir, yuzu_path / SCREENSHOTS_DIR);
         GenerateYuzuPath(YuzuPath::SDMCDir, yuzu_path / SDMC_DIR);
         GenerateYuzuPath(YuzuPath::ShaderDir, yuzu_path / SHADER_DIR);
+        GenerateYuzuPath(YuzuPath::TASDir, yuzu_path / TAS_DIR);
     }
 
     ~PathManagerImpl() = default;

src/common/fs/path_util.h

@@ -23,6 +23,7 @@ enum class YuzuPath {
     ScreenshotsDir, // Where yuzu screenshots are stored.
     SDMCDir,        // Where the emulated SDMC is stored.
     ShaderDir,      // Where shaders are stored.
+    TASDir,         // Where TAS scripts are stored.
 };
 
 /**

src/common/logging/backend.cpp

@@ -18,6 +18,7 @@
 #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"

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

@@ -54,15 +54,13 @@ 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());
     log_setting("Audio_OutputDevice", values.audio_device_id.GetValue());
     log_setting("DataStorage_UseVirtualSd", values.use_virtual_sd.GetValue());
     log_path("DataStorage_CacheDir", Common::FS::GetYuzuPath(Common::FS::YuzuPath::CacheDir));
@@ -73,6 +71,9 @@ void LogSettings() {
     log_setting("Debugging_ProgramArgs", values.program_args.GetValue());
     log_setting("Services_BCATBackend", values.bcat_backend.GetValue());
     log_setting("Services_BCATBoxcatLocal", values.bcat_boxcat_local.GetValue());
+    log_setting("Input_EnableMotion", values.motion_enabled.GetValue());
+    log_setting("Input_EnableVibration", values.vibration_enabled.GetValue());
+    log_setting("Input_EnableRawInput", values.enable_raw_input.GetValue());
 }
 
 bool IsConfiguringGlobal() {
@@ -113,7 +114,6 @@ void RestoreGlobalState(bool is_powered_on) {
     }
 
     // Audio
-    values.enable_audio_stretching.SetGlobal(true);
     values.volume.SetGlobal(true);
 
     // Core
@@ -137,13 +137,12 @@ 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

@@ -16,7 +16,6 @@
 
 #include "common/common_types.h"
 #include "common/settings_input.h"
-#include "input_common/udp/client.h"
 
 namespace Settings {
 
@@ -48,6 +47,12 @@ enum class FullscreenMode : u32 {
     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
@@ -409,7 +414,6 @@ struct Values {
     BasicSetting<std::string> audio_device_id{"auto", "output_device"};
     BasicSetting<std::string> sink_id{"auto", "output_engine"};
     BasicSetting<bool> audio_muted{false, "audio_muted"};
-    Setting<bool> enable_audio_stretching{true, "enable_audio_stretching"};
     RangedSetting<u8> volume{100, 0, 100, "volume"};
 
     // Core
@@ -466,7 +470,7 @@ struct Values {
     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"};
     BasicRangedSetting<u16> fps_cap{1000, 1, 1000, "fps_cap"};
@@ -475,7 +479,6 @@ struct Values {
                                                 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"};
@@ -499,14 +502,20 @@ struct Values {
 
     Setting<bool> use_docked_mode{true, "use_docked_mode"};
 
+    BasicSetting<bool> enable_raw_input{false, "enable_raw_input"};
+
     Setting<bool> vibration_enabled{true, "vibration_enabled"};
     Setting<bool> enable_accurate_vibrations{false, "enable_accurate_vibrations"};
 
     Setting<bool> motion_enabled{true, "motion_enabled"};
     BasicSetting<std::string> motion_device{"engine:motion_emu,update_period:100,sensitivity:0.01",
                                             "motion_device"};
-    BasicSetting<std::string> udp_input_servers{InputCommon::CemuhookUDP::DEFAULT_SRV,
-                                                "udp_input_servers"};
+    BasicSetting<std::string> udp_input_servers{"127.0.0.1:26760", "udp_input_servers"};
+
+    BasicSetting<bool> pause_tas_on_load{true, "pause_tas_on_load"};
+    BasicSetting<bool> tas_enable{false, "tas_enable"};
+    BasicSetting<bool> tas_loop{false, "tas_loop"};
+    BasicSetting<bool> tas_swap_controllers{true, "tas_swap_controllers"};
 
     BasicSetting<bool> mouse_panning{false, "mouse_panning"};
     BasicRangedSetting<u8> mouse_panning_sensitivity{10, 1, 100, "mouse_panning_sensitivity"};

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() {
@@ -84,7 +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(); });
         }
     }
 
@@ -95,6 +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);
@@ -123,13 +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 {
@@ -166,13 +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/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

src/core/core.cpp

@@ -83,6 +83,12 @@ FileSys::StorageId GetStorageIdForFrontendSlot(
     }
 }
 
+void KProcessDeleter(Kernel::KProcess* process) {
+    process->Destroy();
+}
+
+using KProcessPtr = std::unique_ptr<Kernel::KProcess, decltype(&KProcessDeleter)>;
+
 } // Anonymous namespace
 
 FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
@@ -233,8 +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();
@@ -247,7 +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
@@ -299,10 +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();
@@ -311,11 +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");
     }
@@ -374,6 +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{};
@@ -507,6 +512,12 @@ const ARM_Interface& System::CurrentArmInterface() const {
     return impl->kernel.CurrentPhysicalCore().ArmInterface();
 }
 
+std::size_t System::CurrentCoreIndex() const {
+    std::size_t core = impl->kernel.GetCurrentHostThreadID();
+    ASSERT(core < Core::Hardware::NUM_CPU_CORES);
+    return core;
+}
+
 Kernel::PhysicalCore& System::CurrentPhysicalCore() {
     return impl->kernel.CurrentPhysicalCore();
 }

src/core/core.h

@@ -205,6 +205,9 @@ public:
     /// Gets an ARM interface to the CPU core that is currently running
     [[nodiscard]] const ARM_Interface& CurrentArmInterface() const;
 
+    /// Gets the index of the currently running CPU core
+    [[nodiscard]] std::size_t CurrentCoreIndex() const;
+
     /// Gets the physical core for the CPU core that is currently running
     [[nodiscard]] Kernel::PhysicalCore& CurrentPhysicalCore();
 

src/core/cpu_manager.cpp

@@ -118,18 +118,17 @@ void CpuManager::MultiCoreRunGuestLoop() {
             physical_core = &kernel.CurrentPhysicalCore();
         }
         system.ExitDynarmicProfile();
-        {
-            Kernel::KScopedDisableDispatch dd(kernel);
-            physical_core->ArmInterface().ClearExclusiveState();
-        }
+        physical_core->ArmInterface().ClearExclusiveState();
+        kernel.CurrentScheduler()->RescheduleCurrentCore();
     }
 }
 
 void CpuManager::MultiCoreRunIdleThread() {
     auto& kernel = system.Kernel();
     while (true) {
-        Kernel::KScopedDisableDispatch dd(kernel);
-        kernel.CurrentPhysicalCore().Idle();
+        auto& physical_core = kernel.CurrentPhysicalCore();
+        physical_core.Idle();
+        kernel.CurrentScheduler()->RescheduleCurrentCore();
     }
 }
 
@@ -137,12 +136,12 @@ void CpuManager::MultiCoreRunSuspendThread() {
     auto& kernel = system.Kernel();
     kernel.CurrentScheduler()->OnThreadStart();
     while (true) {
-        auto core = kernel.CurrentPhysicalCoreIndex();
+        auto core = kernel.GetCurrentHostThreadID();
         auto& scheduler = *kernel.CurrentScheduler();
         Kernel::KThread* current_thread = scheduler.GetCurrentThread();
         Common::Fiber::YieldTo(current_thread->GetHostContext(), *core_data[core].host_context);
         ASSERT(scheduler.ContextSwitchPending());
-        ASSERT(core == kernel.CurrentPhysicalCoreIndex());
+        ASSERT(core == kernel.GetCurrentHostThreadID());
         scheduler.RescheduleCurrentCore();
     }
 }
@@ -348,11 +347,15 @@ void CpuManager::RunThread(std::stop_token stop_token, std::size_t core) {
             sc_sync_first_use = false;
         }
 
-        // Emulation was stopped
-        if (stop_token.stop_requested()) {
+        // Abort if emulation was killed before the session really starts
+        if (!system.IsPoweredOn()) {
             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/file_sys/kernel_executable.h

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

src/core/file_sys/vfs.cpp

@@ -273,6 +273,10 @@ VirtualFile VfsDirectory::GetFile(std::string_view name) const {
     return iter == files.end() ? nullptr : *iter;
 }
 
+FileTimeStampRaw VfsDirectory::GetFileTimeStamp([[maybe_unused]] std::string_view path) const {
+    return {};
+}
+
 VirtualDir VfsDirectory::GetSubdirectory(std::string_view name) const {
     const auto& subs = GetSubdirectories();
     const auto iter = std::find_if(subs.begin(), subs.end(),

src/core/file_sys/vfs.h

@@ -199,6 +199,9 @@ public:
     // file with name.
     virtual VirtualFile GetFile(std::string_view name) const;
 
+    // Returns a struct containing the file's timestamp.
+    virtual FileTimeStampRaw GetFileTimeStamp(std::string_view path) const;
+
     // Returns a vector containing all of the subdirectories in this directory.
     virtual std::vector<VirtualDir> GetSubdirectories() const = 0;
     // Returns the directory with name matching name. Returns nullptr if directory dosen't have a

src/core/file_sys/vfs_real.cpp

@@ -13,6 +13,13 @@
 #include "common/logging/log.h"
 #include "core/file_sys/vfs_real.h"
 
+// For FileTimeStampRaw
+#include <sys/stat.h>
+
+#ifdef _MSC_VER
+#define stat _stat64
+#endif
+
 namespace FileSys {
 
 namespace FS = Common::FS;
@@ -392,6 +399,28 @@ std::vector<VirtualFile> RealVfsDirectory::GetFiles() const {
     return IterateEntries<RealVfsFile, VfsFile>();
 }
 
+FileTimeStampRaw RealVfsDirectory::GetFileTimeStamp(std::string_view path_) const {
+    const auto full_path = FS::SanitizePath(path + '/' + std::string(path_));
+    const auto fs_path = std::filesystem::path{FS::ToU8String(full_path)};
+    struct stat file_status;
+
+#ifdef _WIN32
+    const auto stat_result = _wstat64(fs_path.c_str(), &file_status);
+#else
+    const auto stat_result = stat(fs_path.c_str(), &file_status);
+#endif
+
+    if (stat_result != 0) {
+        return {};
+    }
+
+    return {
+        .created{static_cast<u64>(file_status.st_ctime)},
+        .accessed{static_cast<u64>(file_status.st_atime)},
+        .modified{static_cast<u64>(file_status.st_mtime)},
+    };
+}
+
 std::vector<VirtualDir> RealVfsDirectory::GetSubdirectories() const {
     return IterateEntries<RealVfsDirectory, VfsDirectory>();
 }

src/core/file_sys/vfs_real.h

@@ -86,6 +86,7 @@ public:
     VirtualDir CreateDirectoryRelative(std::string_view relative_path) override;
     bool DeleteSubdirectoryRecursive(std::string_view name) override;
     std::vector<VirtualFile> GetFiles() const override;
+    FileTimeStampRaw GetFileTimeStamp(std::string_view path) const override;
     std::vector<VirtualDir> GetSubdirectories() const override;
     bool IsWritable() const override;
     bool IsReadable() const override;

src/core/file_sys/vfs_types.h

@@ -6,6 +6,8 @@
 
 #include <memory>
 
+#include "common/common_types.h"
+
 namespace FileSys {
 
 class VfsDirectory;
@@ -18,4 +20,11 @@ using VirtualDir = std::shared_ptr<VfsDirectory>;
 using VirtualFile = std::shared_ptr<VfsFile>;
 using VirtualFilesystem = std::shared_ptr<VfsFilesystem>;
 
+struct FileTimeStampRaw {
+    u64 created{};
+    u64 accessed{};
+    u64 modified{};
+    u64 padding{};
+};
+
 } // namespace FileSys

src/core/hle/api_version.h

@@ -28,13 +28,20 @@ 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 = 5;
+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/k_address_arbiter.cpp

@@ -28,7 +28,7 @@ bool ReadFromUser(Core::System& system, s32* out, VAddr address) {
 
 bool DecrementIfLessThan(Core::System& system, s32* out, VAddr address, s32 value) {
     auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    const auto current_core = system.CurrentCoreIndex();
 
     // TODO(bunnei): We should disable interrupts here via KScopedInterruptDisable.
     // TODO(bunnei): We should call CanAccessAtomic(..) here.
@@ -58,7 +58,7 @@ bool DecrementIfLessThan(Core::System& system, s32* out, VAddr address, s32 valu
 
 bool UpdateIfEqual(Core::System& system, s32* out, VAddr address, s32 value, s32 new_value) {
     auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    const auto current_core = system.CurrentCoreIndex();
 
     // TODO(bunnei): We should disable interrupts here via KScopedInterruptDisable.
     // TODO(bunnei): We should call CanAccessAtomic(..) here.

src/core/hle/kernel/k_auto_object.h

@@ -170,10 +170,6 @@ public:
         }
     }
 
-    const std::string& GetName() const {
-        return name;
-    }
-
 private:
     void RegisterWithKernel();
     void UnregisterWithKernel();

src/core/hle/kernel/k_condition_variable.cpp

@@ -35,7 +35,7 @@ bool WriteToUser(Core::System& system, VAddr address, const u32* p) {
 bool UpdateLockAtomic(Core::System& system, u32* out, VAddr address, u32 if_zero,
                       u32 new_orr_mask) {
     auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    const auto current_core = system.CurrentCoreIndex();
 
     // Load the value from the address.
     const auto expected = monitor.ExclusiveRead32(current_core, address);

src/core/hle/kernel/k_handle_table.cpp

@@ -13,7 +13,6 @@ ResultCode KHandleTable::Finalize() {
     // Get the table and clear our record of it.
     u16 saved_table_size = 0;
     {
-        KScopedDisableDispatch dd(kernel);
         KScopedSpinLock lk(m_lock);
 
         std::swap(m_table_size, saved_table_size);
@@ -44,7 +43,6 @@ bool KHandleTable::Remove(Handle handle) {
     // Find the object and free the entry.
     KAutoObject* obj = nullptr;
     {
-        KScopedDisableDispatch dd(kernel);
         KScopedSpinLock lk(m_lock);
 
         if (this->IsValidHandle(handle)) {
@@ -63,7 +61,6 @@ bool KHandleTable::Remove(Handle handle) {
 }
 
 ResultCode KHandleTable::Add(Handle* out_handle, KAutoObject* obj, u16 type) {
-    KScopedDisableDispatch dd(kernel);
     KScopedSpinLock lk(m_lock);
 
     // Never exceed our capacity.
@@ -86,7 +83,6 @@ ResultCode KHandleTable::Add(Handle* out_handle, KAutoObject* obj, u16 type) {
 }
 
 ResultCode KHandleTable::Reserve(Handle* out_handle) {
-    KScopedDisableDispatch dd(kernel);
     KScopedSpinLock lk(m_lock);
 
     // Never exceed our capacity.
@@ -97,7 +93,6 @@ ResultCode KHandleTable::Reserve(Handle* out_handle) {
 }
 
 void KHandleTable::Unreserve(Handle handle) {
-    KScopedDisableDispatch dd(kernel);
     KScopedSpinLock lk(m_lock);
 
     // Unpack the handle.
@@ -116,7 +111,6 @@ void KHandleTable::Unreserve(Handle handle) {
 }
 
 void KHandleTable::Register(Handle handle, KAutoObject* obj, u16 type) {
-    KScopedDisableDispatch dd(kernel);
     KScopedSpinLock lk(m_lock);
 
     // Unpack the handle.

src/core/hle/kernel/k_handle_table.h

@@ -69,7 +69,6 @@ public:
     template <typename T = KAutoObject>
     KScopedAutoObject<T> GetObjectWithoutPseudoHandle(Handle handle) const {
         // Lock and look up in table.
-        KScopedDisableDispatch dd(kernel);
         KScopedSpinLock lk(m_lock);
 
         if constexpr (std::is_same_v<T, KAutoObject>) {
@@ -124,7 +123,6 @@ public:
         size_t num_opened;
         {
             // Lock the table.
-            KScopedDisableDispatch dd(kernel);
             KScopedSpinLock lk(m_lock);
             for (num_opened = 0; num_opened < num_handles; num_opened++) {
                 // Get the current handle.

src/core/hle/kernel/k_process.cpp

@@ -59,7 +59,6 @@ void SetupMainThread(Core::System& system, KProcess& owner_process, u32 priority
     thread->GetContext64().cpu_registers[0] = 0;
     thread->GetContext32().cpu_registers[1] = thread_handle;
     thread->GetContext64().cpu_registers[1] = thread_handle;
-    thread->DisableDispatch();
 
     auto& kernel = system.Kernel();
     // Threads by default are dormant, wake up the main thread so it runs when the scheduler fires

src/core/hle/kernel/k_scheduler.cpp

@@ -376,18 +376,20 @@ void KScheduler::ClearSchedulerUpdateNeeded(KernelCore& kernel) {
 }
 
 void KScheduler::DisableScheduling(KernelCore& kernel) {
-    ASSERT(GetCurrentThreadPointer(kernel)->GetDisableDispatchCount() >= 0);
-    GetCurrentThreadPointer(kernel)->DisableDispatch();
+    if (auto* scheduler = kernel.CurrentScheduler(); scheduler) {
+        ASSERT(scheduler->GetCurrentThread()->GetDisableDispatchCount() >= 0);
+        scheduler->GetCurrentThread()->DisableDispatch();
+    }
 }
 
 void KScheduler::EnableScheduling(KernelCore& kernel, u64 cores_needing_scheduling) {
-    ASSERT(GetCurrentThreadPointer(kernel)->GetDisableDispatchCount() >= 1);
-
-    if (GetCurrentThreadPointer(kernel)->GetDisableDispatchCount() > 1) {
-        GetCurrentThreadPointer(kernel)->EnableDispatch();
-    } else {
-        RescheduleCores(kernel, cores_needing_scheduling);
+    if (auto* scheduler = kernel.CurrentScheduler(); scheduler) {
+        ASSERT(scheduler->GetCurrentThread()->GetDisableDispatchCount() >= 1);
+        if (scheduler->GetCurrentThread()->GetDisableDispatchCount() >= 1) {
+            scheduler->GetCurrentThread()->EnableDispatch();
+        }
     }
+    RescheduleCores(kernel, cores_needing_scheduling);
 }
 
 u64 KScheduler::UpdateHighestPriorityThreads(KernelCore& kernel) {
@@ -615,17 +617,13 @@ KScheduler::KScheduler(Core::System& system_, s32 core_id_) : system{system_}, c
     state.highest_priority_thread = nullptr;
 }
 
-void KScheduler::Finalize() {
+KScheduler::~KScheduler() {
     if (idle_thread) {
         idle_thread->Close();
         idle_thread = nullptr;
     }
 }
 
-KScheduler::~KScheduler() {
-    ASSERT(!idle_thread);
-}
-
 KThread* KScheduler::GetCurrentThread() const {
     if (auto result = current_thread.load(); result) {
         return result;
@@ -644,12 +642,10 @@ void KScheduler::RescheduleCurrentCore() {
     if (phys_core.IsInterrupted()) {
         phys_core.ClearInterrupt();
     }
-
     guard.Lock();
     if (state.needs_scheduling.load()) {
         Schedule();
     } else {
-        GetCurrentThread()->EnableDispatch();
         guard.Unlock();
     }
 }
@@ -659,33 +655,26 @@ void KScheduler::OnThreadStart() {
 }
 
 void KScheduler::Unload(KThread* thread) {
-    ASSERT(thread);
-
     LOG_TRACE(Kernel, "core {}, unload thread {}", core_id, thread ? thread->GetName() : "nullptr");
 
-    if (thread->IsCallingSvc()) {
-        thread->ClearIsCallingSvc();
+    if (thread) {
+        if (thread->IsCallingSvc()) {
+            thread->ClearIsCallingSvc();
+        }
+        if (!thread->IsTerminationRequested()) {
+            prev_thread = thread;
+
+            Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
+            cpu_core.SaveContext(thread->GetContext32());
+            cpu_core.SaveContext(thread->GetContext64());
+            // Save the TPIDR_EL0 system register in case it was modified.
+            thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+            cpu_core.ClearExclusiveState();
+        } else {
+            prev_thread = nullptr;
+        }
+        thread->context_guard.Unlock();
     }
-
-    auto& physical_core = system.Kernel().PhysicalCore(core_id);
-    if (!physical_core.IsInitialized()) {
-        return;
-    }
-
-    Core::ARM_Interface& cpu_core = physical_core.ArmInterface();
-    cpu_core.SaveContext(thread->GetContext32());
-    cpu_core.SaveContext(thread->GetContext64());
-    // Save the TPIDR_EL0 system register in case it was modified.
-    thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
-    cpu_core.ClearExclusiveState();
-
-    if (!thread->IsTerminationRequested() && thread->GetActiveCore() == core_id) {
-        prev_thread = thread;
-    } else {
-        prev_thread = nullptr;
-    }
-
-    thread->context_guard.Unlock();
 }
 
 void KScheduler::Reload(KThread* thread) {
@@ -694,6 +683,11 @@ void KScheduler::Reload(KThread* thread) {
     if (thread) {
         ASSERT_MSG(thread->GetState() == ThreadState::Runnable, "Thread must be runnable.");
 
+        auto* const thread_owner_process = thread->GetOwnerProcess();
+        if (thread_owner_process != nullptr) {
+            system.Kernel().MakeCurrentProcess(thread_owner_process);
+        }
+
         Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
         cpu_core.LoadContext(thread->GetContext32());
         cpu_core.LoadContext(thread->GetContext64());
@@ -711,7 +705,7 @@ void KScheduler::SwitchContextStep2() {
 }
 
 void KScheduler::ScheduleImpl() {
-    KThread* previous_thread = GetCurrentThread();
+    KThread* previous_thread = current_thread.load();
     KThread* next_thread = state.highest_priority_thread;
 
     state.needs_scheduling = false;
@@ -723,15 +717,10 @@ void KScheduler::ScheduleImpl() {
 
     // If we're not actually switching thread, there's nothing to do.
     if (next_thread == current_thread.load()) {
-        previous_thread->EnableDispatch();
         guard.Unlock();
         return;
     }
 
-    if (next_thread->GetCurrentCore() != core_id) {
-        next_thread->SetCurrentCore(core_id);
-    }
-
     current_thread.store(next_thread);
 
     KProcess* const previous_process = system.Kernel().CurrentProcess();
@@ -742,7 +731,11 @@ void KScheduler::ScheduleImpl() {
     Unload(previous_thread);
 
     std::shared_ptr<Common::Fiber>* old_context;
-    old_context = &previous_thread->GetHostContext();
+    if (previous_thread != nullptr) {
+        old_context = &previous_thread->GetHostContext();
+    } else {
+        old_context = &idle_thread->GetHostContext();
+    }
     guard.Unlock();
 
     Common::Fiber::YieldTo(*old_context, *switch_fiber);

src/core/hle/kernel/k_scheduler.h

@@ -33,8 +33,6 @@ public:
     explicit KScheduler(Core::System& system_, s32 core_id_);
     ~KScheduler();
 
-    void Finalize();
-
     /// Reschedules to the next available thread (call after current thread is suspended)
     void RescheduleCurrentCore();
 

src/core/hle/kernel/k_thread.cpp

@@ -14,7 +14,6 @@
 #include "common/fiber.h"
 #include "common/logging/log.h"
 #include "common/scope_exit.h"
-#include "common/settings.h"
 #include "common/thread_queue_list.h"
 #include "core/core.h"
 #include "core/cpu_manager.h"
@@ -189,7 +188,7 @@ ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_s
     // Setup the stack parameters.
     StackParameters& sp = GetStackParameters();
     sp.cur_thread = this;
-    sp.disable_count = 0;
+    sp.disable_count = 1;
     SetInExceptionHandler();
 
     // Set thread ID.
@@ -216,10 +215,9 @@ ResultCode KThread::InitializeThread(KThread* thread, KThreadFunction func, uint
     // Initialize the thread.
     R_TRY(thread->Initialize(func, arg, user_stack_top, prio, core, owner, type));
 
-    // Initialize emulation parameters.
+    // Initialize host context.
     thread->host_context =
         std::make_shared<Common::Fiber>(std::move(init_func), init_func_parameter);
-    thread->is_single_core = !Settings::values.use_multi_core.GetValue();
 
     return ResultSuccess;
 }
@@ -972,9 +970,6 @@ ResultCode KThread::Run() {
 
         // Set our state and finish.
         SetState(ThreadState::Runnable);
-
-        DisableDispatch();
-
         return ResultSuccess;
     }
 }
@@ -1059,16 +1054,4 @@ s32 GetCurrentCoreId(KernelCore& kernel) {
     return GetCurrentThread(kernel).GetCurrentCore();
 }
 
-KScopedDisableDispatch::~KScopedDisableDispatch() {
-    if (GetCurrentThread(kernel).GetDisableDispatchCount() <= 1) {
-        auto scheduler = kernel.CurrentScheduler();
-
-        if (scheduler) {
-            scheduler->RescheduleCurrentCore();
-        }
-    } else {
-        GetCurrentThread(kernel).EnableDispatch();
-    }
-}
-
 } // namespace Kernel

src/core/hle/kernel/k_thread.h

@@ -450,39 +450,16 @@ public:
         sleeping_queue = q;
     }
 
-    [[nodiscard]] bool IsKernelThread() const {
-        return GetActiveCore() == 3;
-    }
-
-    [[nodiscard]] bool IsDispatchTrackingDisabled() const {
-        return is_single_core || IsKernelThread();
-    }
-
     [[nodiscard]] s32 GetDisableDispatchCount() const {
-        if (IsDispatchTrackingDisabled()) {
-            // TODO(bunnei): Until kernel threads are emulated, we cannot enable/disable dispatch.
-            return 1;
-        }
-
         return this->GetStackParameters().disable_count;
     }
 
     void DisableDispatch() {
-        if (IsDispatchTrackingDisabled()) {
-            // TODO(bunnei): Until kernel threads are emulated, we cannot enable/disable dispatch.
-            return;
-        }
-
         ASSERT(GetCurrentThread(kernel).GetDisableDispatchCount() >= 0);
         this->GetStackParameters().disable_count++;
     }
 
     void EnableDispatch() {
-        if (IsDispatchTrackingDisabled()) {
-            // TODO(bunnei): Until kernel threads are emulated, we cannot enable/disable dispatch.
-            return;
-        }
-
         ASSERT(GetCurrentThread(kernel).GetDisableDispatchCount() > 0);
         this->GetStackParameters().disable_count--;
     }
@@ -731,7 +708,6 @@ private:
 
     // For emulation
     std::shared_ptr<Common::Fiber> host_context{};
-    bool is_single_core{};
 
     // For debugging
     std::vector<KSynchronizationObject*> wait_objects_for_debugging;
@@ -776,16 +752,4 @@ public:
     }
 };
 
-class KScopedDisableDispatch {
-public:
-    [[nodiscard]] explicit KScopedDisableDispatch(KernelCore& kernel_) : kernel{kernel_} {
-        GetCurrentThread(kernel).DisableDispatch();
-    }
-
-    ~KScopedDisableDispatch();
-
-private:
-    KernelCore& kernel;
-};
-
 } // namespace Kernel

src/core/hle/kernel/kernel.cpp

@@ -85,9 +85,8 @@ struct KernelCore::Impl {
     }
 
     void InitializeCores() {
-        for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            cores[core_id].Initialize(current_process->Is64BitProcess());
-            system.Memory().SetCurrentPageTable(*current_process, core_id);
+        for (auto& core : cores) {
+            core.Initialize(current_process->Is64BitProcess());
         }
     }
 
@@ -132,6 +131,15 @@ struct KernelCore::Impl {
         next_user_process_id = KProcess::ProcessIDMin;
         next_thread_id = 1;
 
+        for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
+            if (suspend_threads[core_id]) {
+                suspend_threads[core_id]->Close();
+                suspend_threads[core_id] = nullptr;
+            }
+
+            schedulers[core_id].reset();
+        }
+
         cores.clear();
 
         global_handle_table->Finalize();
@@ -159,16 +167,6 @@ struct KernelCore::Impl {
         CleanupObject(time_shared_mem);
         CleanupObject(system_resource_limit);
 
-        for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            if (suspend_threads[core_id]) {
-                suspend_threads[core_id]->Close();
-                suspend_threads[core_id] = nullptr;
-            }
-
-            schedulers[core_id]->Finalize();
-            schedulers[core_id].reset();
-        }
-
         // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
         next_host_thread_id = Core::Hardware::NUM_CPU_CORES;
 
@@ -259,6 +257,14 @@ struct KernelCore::Impl {
 
     void MakeCurrentProcess(KProcess* process) {
         current_process = process;
+        if (process == nullptr) {
+            return;
+        }
+
+        const u32 core_id = GetCurrentHostThreadID();
+        if (core_id < Core::Hardware::NUM_CPU_CORES) {
+            system.Memory().SetCurrentPageTable(*process, core_id);
+        }
     }
 
     static inline thread_local u32 host_thread_id = UINT32_MAX;
@@ -821,20 +827,16 @@ const Kernel::PhysicalCore& KernelCore::PhysicalCore(std::size_t id) const {
     return impl->cores[id];
 }
 
-size_t KernelCore::CurrentPhysicalCoreIndex() const {
-    const u32 core_id = impl->GetCurrentHostThreadID();
-    if (core_id >= Core::Hardware::NUM_CPU_CORES) {
-        return Core::Hardware::NUM_CPU_CORES - 1;
-    }
-    return core_id;
-}
-
 Kernel::PhysicalCore& KernelCore::CurrentPhysicalCore() {
-    return impl->cores[CurrentPhysicalCoreIndex()];
+    u32 core_id = impl->GetCurrentHostThreadID();
+    ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+    return impl->cores[core_id];
 }
 
 const Kernel::PhysicalCore& KernelCore::CurrentPhysicalCore() const {
-    return impl->cores[CurrentPhysicalCoreIndex()];
+    u32 core_id = impl->GetCurrentHostThreadID();
+    ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+    return impl->cores[core_id];
 }
 
 Kernel::KScheduler* KernelCore::CurrentScheduler() {
@@ -1027,9 +1029,6 @@ void KernelCore::Suspend(bool in_suspention) {
             impl->suspend_threads[core_id]->SetState(state);
             impl->suspend_threads[core_id]->SetWaitReasonForDebugging(
                 ThreadWaitReasonForDebugging::Suspended);
-            if (!should_suspend) {
-                impl->suspend_threads[core_id]->DisableDispatch();
-            }
         }
     }
 }
@@ -1044,11 +1043,13 @@ void KernelCore::ExceptionalExit() {
 }
 
 void KernelCore::EnterSVCProfile() {
-    impl->svc_ticks[CurrentPhysicalCoreIndex()] = MicroProfileEnter(MICROPROFILE_TOKEN(Kernel_SVC));
+    std::size_t core = impl->GetCurrentHostThreadID();
+    impl->svc_ticks[core] = MicroProfileEnter(MICROPROFILE_TOKEN(Kernel_SVC));
 }
 
 void KernelCore::ExitSVCProfile() {
-    MicroProfileLeave(MICROPROFILE_TOKEN(Kernel_SVC), impl->svc_ticks[CurrentPhysicalCoreIndex()]);
+    std::size_t core = impl->GetCurrentHostThreadID();
+    MicroProfileLeave(MICROPROFILE_TOKEN(Kernel_SVC), impl->svc_ticks[core]);
 }
 
 std::weak_ptr<Kernel::ServiceThread> KernelCore::CreateServiceThread(const std::string& name) {

src/core/hle/kernel/kernel.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <functional>
 #include <memory>
 #include <string>
 #include <unordered_map>
@@ -146,9 +147,6 @@ public:
     /// Gets the an instance of the respective physical CPU core.
     const Kernel::PhysicalCore& PhysicalCore(std::size_t id) const;
 
-    /// Gets the current physical core index for the running host thread.
-    std::size_t CurrentPhysicalCoreIndex() const;
-
     /// Gets the sole instance of the Scheduler at the current running core.
     Kernel::KScheduler* CurrentScheduler();
 

src/core/hle/kernel/svc.cpp

@@ -877,7 +877,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, Handle
             const u64 thread_ticks = current_thread->GetCpuTime();
 
             out_ticks = thread_ticks + (core_timing.GetCPUTicks() - prev_ctx_ticks);
-        } else if (same_thread && info_sub_id == system.Kernel().CurrentPhysicalCoreIndex()) {
+        } else if (same_thread && info_sub_id == system.CurrentCoreIndex()) {
             out_ticks = core_timing.GetCPUTicks() - prev_ctx_ticks;
         }
 

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"
@@ -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};
 };
 

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/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;
 }
@@ -251,6 +261,18 @@ ResultVal<FileSys::EntryType> VfsDirectoryServiceWrapper::GetEntryType(
     return FileSys::ERROR_PATH_NOT_FOUND;
 }
 
+ResultVal<FileSys::FileTimeStampRaw> VfsDirectoryServiceWrapper::GetFileTimeStampRaw(
+    const std::string& path) const {
+    auto dir = GetDirectoryRelativeWrapped(backing, Common::FS::GetParentPath(path));
+    if (dir == nullptr) {
+        return FileSys::ERROR_PATH_NOT_FOUND;
+    }
+    if (GetEntryType(path).Failed()) {
+        return FileSys::ERROR_PATH_NOT_FOUND;
+    }
+    return MakeResult(dir->GetFileTimeStamp(Common::FS::GetFilename(path)));
+}
+
 FileSystemController::FileSystemController(Core::System& system_) : system{system_} {}
 
 FileSystemController::~FileSystemController() = default;

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

@@ -240,6 +240,12 @@ public:
      */
     ResultVal<FileSys::EntryType> GetEntryType(const std::string& path) const;
 
+    /**
+     * Get the timestamp of the specified path
+     * @return The timestamp of the specified path or error code
+     */
+    ResultVal<FileSys::FileTimeStampRaw> GetFileTimeStampRaw(const std::string& path) const;
+
 private:
     FileSys::VirtualDir backing;
 };

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

@@ -326,7 +326,7 @@ public:
             {11, &IFileSystem::GetFreeSpaceSize, "GetFreeSpaceSize"},
             {12, &IFileSystem::GetTotalSpaceSize, "GetTotalSpaceSize"},
             {13, &IFileSystem::CleanDirectoryRecursively, "CleanDirectoryRecursively"},
-            {14, nullptr, "GetFileTimeStampRaw"},
+            {14, &IFileSystem::GetFileTimeStampRaw, "GetFileTimeStampRaw"},
             {15, nullptr, "QueryEntry"},
         };
         RegisterHandlers(functions);
@@ -501,6 +501,24 @@ public:
         rb.Push(size.get_total_size());
     }
 
+    void GetFileTimeStampRaw(Kernel::HLERequestContext& ctx) {
+        const auto file_buffer = ctx.ReadBuffer();
+        const std::string name = Common::StringFromBuffer(file_buffer);
+
+        LOG_WARNING(Service_FS, "(Partial Implementation) called. file={}", name);
+
+        auto result = backend.GetFileTimeStampRaw(name);
+        if (result.Failed()) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(result.Code());
+            return;
+        }
+
+        IPC::ResponseBuilder rb{ctx, 10};
+        rb.Push(ResultSuccess);
+        rb.PushRaw(*result);
+    }
+
 private:
     VfsDirectoryServiceWrapper backend;
     SizeGetter size;

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 +1,59 @@
+// 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, &IService::Match, "Match"},
+            {1, &IService::Filter, "Filter"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+private:
+    void Match(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);
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(ResultSuccess);
+        // Return false since we don't censor anything
+        rb.Push(false);
+    }
+
+    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

@@ -277,37 +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);
 
@@ -352,38 +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.");
 
-        IpConfigInfo ip_config_info{
-            .ip_address_setting{
-                .is_automatic{true},
-                .current_address{0, 0, 0, 0},
-                .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 auto iface = Network::GetSelectedNetworkInterface();
-        if (iface) {
-            ip_config_info.ip_address_setting =
-                IpAddressSetting{.is_automatic{true},
-                                 .current_address{Network::TranslateIPv4(iface->ip_address)},
-                                 .subnet_mask{Network::TranslateIPv4(iface->subnet_mask)},
-                                 .gateway{Network::TranslateIPv4(iface->gateway)}};
+        const IpConfigInfo ip_config_info = [&net_iface] {
+            if (!net_iface) {
+                return IpConfigInfo{};
+            }
 
-        } else {
-            LOG_ERROR(Service_NIFM,
-                      "Couldn't get host network configuration info, using default values");
-        }
+            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);

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/nvflinger/buffer_queue.h

@@ -42,7 +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);

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

@@ -298,7 +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/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/network/network.cpp

@@ -7,7 +7,8 @@
 #include <limits>
 #include <utility>
 #include <vector>
-#include "common/common_funcs.h"
+
+#include "common/error.h"
 
 #ifdef _WIN32
 #include <winsock2.h>
@@ -223,7 +224,7 @@ Errno GetAndLogLastError() {
     if (err == Errno::AGAIN) {
         return err;
     }
-    LOG_ERROR(Network, "Socket operation error: {}", NativeErrorToString(e));
+    LOG_ERROR(Network, "Socket operation error: {}", Common::NativeErrorToString(e));
     return err;
 }
 

src/core/network/network_interface.cpp

@@ -37,73 +37,73 @@ std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
             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) {
-            adapter_addresses.resize((buf_size / sizeof(IP_ADAPTER_ADDRESSES)) + 1);
-        } else {
+        if (ret != ERROR_BUFFER_OVERFLOW) {
             break;
         }
+
+        adapter_addresses.resize((buf_size / sizeof(IP_ADAPTER_ADDRESSES)) + 1);
     }
 
-    if (ret == NO_ERROR) {
-        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.push_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 {
+    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() {
-    std::vector<NetworkInterface> result;
-
     struct ifaddrs* ifaddr = nullptr;
 
     if (getifaddrs(&ifaddr) != 0) {
         LOG_ERROR(Network, "Failed to get network interfaces with getifaddrs: {}",
                   std::strerror(errno));
-        return result;
+        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;
@@ -117,55 +117,62 @@ std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
             continue;
         }
 
-        std::uint32_t gateway{0};
+        u32 gateway{};
+
         std::ifstream file{"/proc/net/route"};
-        if (file.is_open()) {
-
-            // 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;
-
-                std::uint32_t dest{0};
-                iss >> dest;
-                if (dest != 0) {
-                    // not the default route
-                    continue;
-                }
-
-                iss >> gateway;
-
-                std::uint16_t flags{0};
-                iss >> flags;
-
-                // flag RTF_GATEWAY (defined in <linux/route.h>)
-                if ((flags & 0x2) == 0) {
-                    continue;
-                }
-
-                gateway_found = true;
-                break;
-            }
-
-            if (!gateway_found) {
-                gateway = 0;
-            }
-        } else {
+        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;
         }
 
-        result.push_back(NetworkInterface{
+        // 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},
@@ -180,11 +187,11 @@ std::vector<NetworkInterface> GetAvailableNetworkInterfaces() {
 #endif
 
 std::optional<NetworkInterface> GetSelectedNetworkInterface() {
-    const std::string& selected_network_interface = Settings::values.network_interface.GetValue();
+    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 {};
+        return std::nullopt;
     }
 
     const auto res =
@@ -192,12 +199,12 @@ std::optional<NetworkInterface> GetSelectedNetworkInterface() {
             return iface.name == selected_network_interface;
         });
 
-    if (res != network_interfaces.end()) {
-        return *res;
-    } else {
+    if (res == network_interfaces.end()) {
         LOG_ERROR(Network, "Couldn't find selected interface \"{}\"", selected_network_interface);
-        return {};
+        return std::nullopt;
     }
+
+    return *res;
 }
 
 } // 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";
@@ -214,8 +226,6 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader,
     // Log user configuration information
     constexpr auto field_type = Telemetry::FieldType::UserConfig;
     AddField(field_type, "Audio_SinkId", Settings::values.sink_id.GetValue());
-    AddField(field_type, "Audio_EnableAudioStretching",
-             Settings::values.enable_audio_stretching.GetValue());
     AddField(field_type, "Core_UseMultiCore", Settings::values.use_multi_core.GetValue());
     AddField(field_type, "Renderer_Backend",
              TranslateRenderer(Settings::values.renderer_backend.GetValue()));
@@ -229,8 +239,8 @@ 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/CMakeLists.txt

@@ -21,6 +21,10 @@ add_library(input_common STATIC
     mouse/mouse_poller.h
     sdl/sdl.cpp
     sdl/sdl.h
+    tas/tas_input.cpp
+    tas/tas_input.h
+    tas/tas_poller.cpp
+    tas/tas_poller.h
     udp/client.cpp
     udp/client.h
     udp/protocol.cpp

src/input_common/main.cpp

@@ -5,6 +5,7 @@
 #include <memory>
 #include <thread>
 #include "common/param_package.h"
+#include "common/settings.h"
 #include "input_common/analog_from_button.h"
 #include "input_common/gcadapter/gc_adapter.h"
 #include "input_common/gcadapter/gc_poller.h"
@@ -13,6 +14,8 @@
 #include "input_common/motion_from_button.h"
 #include "input_common/mouse/mouse_input.h"
 #include "input_common/mouse/mouse_poller.h"
+#include "input_common/tas/tas_input.h"
+#include "input_common/tas/tas_poller.h"
 #include "input_common/touch_from_button.h"
 #include "input_common/udp/client.h"
 #include "input_common/udp/udp.h"
@@ -60,6 +63,12 @@ struct InputSubsystem::Impl {
         Input::RegisterFactory<Input::MotionDevice>("mouse", mousemotion);
         mousetouch = std::make_shared<MouseTouchFactory>(mouse);
         Input::RegisterFactory<Input::TouchDevice>("mouse", mousetouch);
+
+        tas = std::make_shared<TasInput::Tas>();
+        tasbuttons = std::make_shared<TasButtonFactory>(tas);
+        Input::RegisterFactory<Input::ButtonDevice>("tas", tasbuttons);
+        tasanalog = std::make_shared<TasAnalogFactory>(tas);
+        Input::RegisterFactory<Input::AnalogDevice>("tas", tasanalog);
     }
 
     void Shutdown() {
@@ -94,6 +103,12 @@ struct InputSubsystem::Impl {
         mouseanalog.reset();
         mousemotion.reset();
         mousetouch.reset();
+
+        Input::UnregisterFactory<Input::ButtonDevice>("tas");
+        Input::UnregisterFactory<Input::AnalogDevice>("tas");
+
+        tasbuttons.reset();
+        tasanalog.reset();
     }
 
     [[nodiscard]] std::vector<Common::ParamPackage> GetInputDevices() const {
@@ -101,6 +116,10 @@ struct InputSubsystem::Impl {
             Common::ParamPackage{{"display", "Any"}, {"class", "any"}},
             Common::ParamPackage{{"display", "Keyboard/Mouse"}, {"class", "keyboard"}},
         };
+        if (Settings::values.tas_enable) {
+            devices.emplace_back(
+                Common::ParamPackage{{"display", "TAS Controller"}, {"class", "tas"}});
+        }
 #ifdef HAVE_SDL2
         auto sdl_devices = sdl->GetInputDevices();
         devices.insert(devices.end(), sdl_devices.begin(), sdl_devices.end());
@@ -120,6 +139,9 @@ struct InputSubsystem::Impl {
         if (params.Get("class", "") == "gcpad") {
             return gcadapter->GetAnalogMappingForDevice(params);
         }
+        if (params.Get("class", "") == "tas") {
+            return tas->GetAnalogMappingForDevice(params);
+        }
 #ifdef HAVE_SDL2
         if (params.Get("class", "") == "sdl") {
             return sdl->GetAnalogMappingForDevice(params);
@@ -136,6 +158,9 @@ struct InputSubsystem::Impl {
         if (params.Get("class", "") == "gcpad") {
             return gcadapter->GetButtonMappingForDevice(params);
         }
+        if (params.Get("class", "") == "tas") {
+            return tas->GetButtonMappingForDevice(params);
+        }
 #ifdef HAVE_SDL2
         if (params.Get("class", "") == "sdl") {
             return sdl->GetButtonMappingForDevice(params);
@@ -174,9 +199,12 @@ struct InputSubsystem::Impl {
     std::shared_ptr<MouseAnalogFactory> mouseanalog;
     std::shared_ptr<MouseMotionFactory> mousemotion;
     std::shared_ptr<MouseTouchFactory> mousetouch;
+    std::shared_ptr<TasButtonFactory> tasbuttons;
+    std::shared_ptr<TasAnalogFactory> tasanalog;
     std::shared_ptr<CemuhookUDP::Client> udp;
     std::shared_ptr<GCAdapter::Adapter> gcadapter;
     std::shared_ptr<MouseInput::Mouse> mouse;
+    std::shared_ptr<TasInput::Tas> tas;
 };
 
 InputSubsystem::InputSubsystem() : impl{std::make_unique<Impl>()} {}
@@ -207,6 +235,14 @@ const MouseInput::Mouse* InputSubsystem::GetMouse() const {
     return impl->mouse.get();
 }
 
+TasInput::Tas* InputSubsystem::GetTas() {
+    return impl->tas.get();
+}
+
+const TasInput::Tas* InputSubsystem::GetTas() const {
+    return impl->tas.get();
+}
+
 std::vector<Common::ParamPackage> InputSubsystem::GetInputDevices() const {
     return impl->GetInputDevices();
 }
@@ -287,6 +323,22 @@ const MouseTouchFactory* InputSubsystem::GetMouseTouch() const {
     return impl->mousetouch.get();
 }
 
+TasButtonFactory* InputSubsystem::GetTasButtons() {
+    return impl->tasbuttons.get();
+}
+
+const TasButtonFactory* InputSubsystem::GetTasButtons() const {
+    return impl->tasbuttons.get();
+}
+
+TasAnalogFactory* InputSubsystem::GetTasAnalogs() {
+    return impl->tasanalog.get();
+}
+
+const TasAnalogFactory* InputSubsystem::GetTasAnalogs() const {
+    return impl->tasanalog.get();
+}
+
 void InputSubsystem::ReloadInputDevices() {
     if (!impl->udp) {
         return;

src/input_common/main.h

@@ -29,6 +29,10 @@ namespace MouseInput {
 class Mouse;
 }
 
+namespace TasInput {
+class Tas;
+}
+
 namespace InputCommon {
 namespace Polling {
 
@@ -64,6 +68,8 @@ class MouseButtonFactory;
 class MouseAnalogFactory;
 class MouseMotionFactory;
 class MouseTouchFactory;
+class TasButtonFactory;
+class TasAnalogFactory;
 class Keyboard;
 
 /**
@@ -103,6 +109,11 @@ public:
     /// Retrieves the underlying mouse device.
     [[nodiscard]] const MouseInput::Mouse* GetMouse() const;
 
+    /// Retrieves the underlying tas device.
+    [[nodiscard]] TasInput::Tas* GetTas();
+
+    /// Retrieves the underlying tas device.
+    [[nodiscard]] const TasInput::Tas* GetTas() const;
     /**
      * Returns all available input devices that this Factory can create a new device with.
      * Each returned ParamPackage should have a `display` field used for display, a class field for
@@ -144,30 +155,42 @@ public:
     /// Retrieves the underlying udp touch handler.
     [[nodiscard]] const UDPTouchFactory* GetUDPTouch() const;
 
-    /// Retrieves the underlying GameCube button handler.
+    /// Retrieves the underlying mouse button handler.
     [[nodiscard]] MouseButtonFactory* GetMouseButtons();
 
-    /// Retrieves the underlying GameCube button handler.
+    /// Retrieves the underlying mouse button handler.
     [[nodiscard]] const MouseButtonFactory* GetMouseButtons() const;
 
-    /// Retrieves the underlying udp touch handler.
+    /// Retrieves the underlying mouse analog handler.
     [[nodiscard]] MouseAnalogFactory* GetMouseAnalogs();
 
-    /// Retrieves the underlying udp touch handler.
+    /// Retrieves the underlying mouse analog handler.
     [[nodiscard]] const MouseAnalogFactory* GetMouseAnalogs() const;
 
-    /// Retrieves the underlying udp motion handler.
+    /// Retrieves the underlying mouse motion handler.
     [[nodiscard]] MouseMotionFactory* GetMouseMotions();
 
-    /// Retrieves the underlying udp motion handler.
+    /// Retrieves the underlying mouse motion handler.
     [[nodiscard]] const MouseMotionFactory* GetMouseMotions() const;
 
-    /// Retrieves the underlying udp touch handler.
+    /// Retrieves the underlying mouse touch handler.
     [[nodiscard]] MouseTouchFactory* GetMouseTouch();
 
-    /// Retrieves the underlying udp touch handler.
+    /// Retrieves the underlying mouse touch handler.
     [[nodiscard]] const MouseTouchFactory* GetMouseTouch() const;
 
+    /// Retrieves the underlying tas button handler.
+    [[nodiscard]] TasButtonFactory* GetTasButtons();
+
+    /// Retrieves the underlying tas button handler.
+    [[nodiscard]] const TasButtonFactory* GetTasButtons() const;
+
+    /// Retrieves the underlying tas analogs handler.
+    [[nodiscard]] TasAnalogFactory* GetTasAnalogs();
+
+    /// Retrieves the underlying tas analogs handler.
+    [[nodiscard]] const TasAnalogFactory* GetTasAnalogs() const;
+
     /// Reloads the input devices
     void ReloadInputDevices();
 

src/input_common/sdl/sdl_impl.cpp

@@ -21,7 +21,7 @@
 #include "common/logging/log.h"
 #include "common/math_util.h"
 #include "common/param_package.h"
-#include "common/settings_input.h"
+#include "common/settings.h"
 #include "common/threadsafe_queue.h"
 #include "core/frontend/input.h"
 #include "input_common/motion_input.h"
@@ -889,8 +889,10 @@ 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");
+    if (!Settings::values.enable_raw_input) {
+        // 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");
@@ -898,10 +900,10 @@ SDLState::SDLState() {
 
     // Tell SDL2 to use the hidapi driver. This will allow joycons to be detected as a
     // GameController and not a generic one
-    SDL_SetHint("SDL_JOYSTICK_HIDAPI_JOY_CONS", "1");
+    SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_JOY_CONS, "1");
 
     // Turn off Pro controller home led
-    SDL_SetHint("SDL_JOYSTICK_HIDAPI_SWITCH_HOME_LED", "0");
+    SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_SWITCH_HOME_LED, "0");
 
     // If the frontend is going to manage the event loop, then we don't start one here
     start_thread = SDL_WasInit(SDL_INIT_JOYSTICK) == 0;

src/input_common/tas/tas_input.cpp

@@ -0,0 +1,455 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <regex>
+
+#include "common/fs/file.h"
+#include "common/fs/fs_types.h"
+#include "common/fs/path_util.h"
+#include "common/logging/log.h"
+#include "common/settings.h"
+#include "input_common/tas/tas_input.h"
+
+namespace TasInput {
+
+// Supported keywords and buttons from a TAS file
+constexpr std::array<std::pair<std::string_view, TasButton>, 20> text_to_tas_button = {
+    std::pair{"KEY_A", TasButton::BUTTON_A},
+    {"KEY_B", TasButton::BUTTON_B},
+    {"KEY_X", TasButton::BUTTON_X},
+    {"KEY_Y", TasButton::BUTTON_Y},
+    {"KEY_LSTICK", TasButton::STICK_L},
+    {"KEY_RSTICK", TasButton::STICK_R},
+    {"KEY_L", TasButton::TRIGGER_L},
+    {"KEY_R", TasButton::TRIGGER_R},
+    {"KEY_PLUS", TasButton::BUTTON_PLUS},
+    {"KEY_MINUS", TasButton::BUTTON_MINUS},
+    {"KEY_DLEFT", TasButton::BUTTON_LEFT},
+    {"KEY_DUP", TasButton::BUTTON_UP},
+    {"KEY_DRIGHT", TasButton::BUTTON_RIGHT},
+    {"KEY_DDOWN", TasButton::BUTTON_DOWN},
+    {"KEY_SL", TasButton::BUTTON_SL},
+    {"KEY_SR", TasButton::BUTTON_SR},
+    {"KEY_CAPTURE", TasButton::BUTTON_CAPTURE},
+    {"KEY_HOME", TasButton::BUTTON_HOME},
+    {"KEY_ZL", TasButton::TRIGGER_ZL},
+    {"KEY_ZR", TasButton::TRIGGER_ZR},
+};
+
+Tas::Tas() {
+    if (!Settings::values.tas_enable) {
+        needs_reset = true;
+        return;
+    }
+    LoadTasFiles();
+}
+
+Tas::~Tas() {
+    Stop();
+};
+
+void Tas::LoadTasFiles() {
+    script_length = 0;
+    for (size_t i = 0; i < commands.size(); i++) {
+        LoadTasFile(i);
+        if (commands[i].size() > script_length) {
+            script_length = commands[i].size();
+        }
+    }
+}
+
+void Tas::LoadTasFile(size_t player_index) {
+    if (!commands[player_index].empty()) {
+        commands[player_index].clear();
+    }
+    std::string file =
+        Common::FS::ReadStringFromFile(Common::FS::GetYuzuPath(Common::FS::YuzuPath::TASDir) /
+                                           fmt::format("script0-{}.txt", player_index + 1),
+                                       Common::FS::FileType::BinaryFile);
+    std::stringstream command_line(file);
+    std::string line;
+    int frame_no = 0;
+    while (std::getline(command_line, line, '\n')) {
+        if (line.empty()) {
+            continue;
+        }
+        LOG_DEBUG(Input, "Loading line: {}", line);
+        std::smatch m;
+
+        std::stringstream linestream(line);
+        std::string segment;
+        std::vector<std::string> seglist;
+
+        while (std::getline(linestream, segment, ' ')) {
+            seglist.push_back(segment);
+        }
+
+        if (seglist.size() < 4) {
+            continue;
+        }
+
+        while (frame_no < std::stoi(seglist.at(0))) {
+            commands[player_index].push_back({});
+            frame_no++;
+        }
+
+        TASCommand command = {
+            .buttons = ReadCommandButtons(seglist.at(1)),
+            .l_axis = ReadCommandAxis(seglist.at(2)),
+            .r_axis = ReadCommandAxis(seglist.at(3)),
+        };
+        commands[player_index].push_back(command);
+        frame_no++;
+    }
+    LOG_INFO(Input, "TAS file loaded! {} frames", frame_no);
+}
+
+void Tas::WriteTasFile(std::u8string file_name) {
+    std::string output_text;
+    for (size_t frame = 0; frame < record_commands.size(); frame++) {
+        if (!output_text.empty()) {
+            output_text += "\n";
+        }
+        const TASCommand& line = record_commands[frame];
+        output_text += std::to_string(frame) + " " + WriteCommandButtons(line.buttons) + " " +
+                       WriteCommandAxis(line.l_axis) + " " + WriteCommandAxis(line.r_axis);
+    }
+    const auto bytes_written = Common::FS::WriteStringToFile(
+        Common::FS::GetYuzuPath(Common::FS::YuzuPath::TASDir) / file_name,
+        Common::FS::FileType::TextFile, output_text);
+    if (bytes_written == output_text.size()) {
+        LOG_INFO(Input, "TAS file written to file!");
+    } else {
+        LOG_ERROR(Input, "Writing the TAS-file has failed! {} / {} bytes written", bytes_written,
+                  output_text.size());
+    }
+}
+
+std::pair<float, float> Tas::FlipAxisY(std::pair<float, float> old) {
+    auto [x, y] = old;
+    return {x, -y};
+}
+
+void Tas::RecordInput(u32 buttons, const std::array<std::pair<float, float>, 2>& axes) {
+    last_input = {buttons, FlipAxisY(axes[0]), FlipAxisY(axes[1])};
+}
+
+std::tuple<TasState, size_t, size_t> Tas::GetStatus() const {
+    TasState state;
+    if (is_recording) {
+        return {TasState::Recording, 0, record_commands.size()};
+    }
+
+    if (is_running) {
+        state = TasState::Running;
+    } else {
+        state = TasState::Stopped;
+    }
+
+    return {state, current_command, script_length};
+}
+
+std::string Tas::DebugButtons(u32 buttons) const {
+    return fmt::format("{{ {} }}", TasInput::Tas::ButtonsToString(buttons));
+}
+
+std::string Tas::DebugJoystick(float x, float y) const {
+    return fmt::format("[ {} , {} ]", std::to_string(x), std::to_string(y));
+}
+
+std::string Tas::DebugInput(const TasData& data) const {
+    return fmt::format("{{ {} , {} , {} }}", DebugButtons(data.buttons),
+                       DebugJoystick(data.axis[0], data.axis[1]),
+                       DebugJoystick(data.axis[2], data.axis[3]));
+}
+
+std::string Tas::DebugInputs(const std::array<TasData, PLAYER_NUMBER>& arr) const {
+    std::string returns = "[ ";
+    for (size_t i = 0; i < arr.size(); i++) {
+        returns += DebugInput(arr[i]);
+        if (i != arr.size() - 1) {
+            returns += " , ";
+        }
+    }
+    return returns + "]";
+}
+
+std::string Tas::ButtonsToString(u32 button) const {
+    std::string returns;
+    for (auto [text_button, tas_button] : text_to_tas_button) {
+        if ((button & static_cast<u32>(tas_button)) != 0)
+            returns += fmt::format(", {}", text_button.substr(4));
+    }
+    return returns.empty() ? "" : returns.substr(2);
+}
+
+void Tas::UpdateThread() {
+    if (!Settings::values.tas_enable) {
+        if (is_running) {
+            Stop();
+        }
+        return;
+    }
+
+    if (is_recording) {
+        record_commands.push_back(last_input);
+    }
+    if (needs_reset) {
+        current_command = 0;
+        needs_reset = false;
+        LoadTasFiles();
+        LOG_DEBUG(Input, "tas_reset done");
+    }
+
+    if (!is_running) {
+        tas_data.fill({});
+        return;
+    }
+    if (current_command < script_length) {
+        LOG_DEBUG(Input, "Playing TAS {}/{}", current_command, script_length);
+        size_t frame = current_command++;
+        for (size_t i = 0; i < commands.size(); i++) {
+            if (frame < commands[i].size()) {
+                TASCommand command = commands[i][frame];
+                tas_data[i].buttons = command.buttons;
+                auto [l_axis_x, l_axis_y] = command.l_axis;
+                tas_data[i].axis[0] = l_axis_x;
+                tas_data[i].axis[1] = l_axis_y;
+                auto [r_axis_x, r_axis_y] = command.r_axis;
+                tas_data[i].axis[2] = r_axis_x;
+                tas_data[i].axis[3] = r_axis_y;
+            } else {
+                tas_data[i] = {};
+            }
+        }
+    } else {
+        is_running = Settings::values.tas_loop.GetValue();
+        current_command = 0;
+        tas_data.fill({});
+        if (!is_running) {
+            SwapToStoredController();
+        }
+    }
+    LOG_DEBUG(Input, "TAS inputs: {}", DebugInputs(tas_data));
+}
+
+TasAnalog Tas::ReadCommandAxis(const std::string& line) const {
+    std::stringstream linestream(line);
+    std::string segment;
+    std::vector<std::string> seglist;
+
+    while (std::getline(linestream, segment, ';')) {
+        seglist.push_back(segment);
+    }
+
+    const float x = std::stof(seglist.at(0)) / 32767.0f;
+    const float y = std::stof(seglist.at(1)) / 32767.0f;
+
+    return {x, y};
+}
+
+u32 Tas::ReadCommandButtons(const std::string& data) const {
+    std::stringstream button_text(data);
+    std::string line;
+    u32 buttons = 0;
+    while (std::getline(button_text, line, ';')) {
+        for (auto [text, tas_button] : text_to_tas_button) {
+            if (text == line) {
+                buttons |= static_cast<u32>(tas_button);
+                break;
+            }
+        }
+    }
+    return buttons;
+}
+
+std::string Tas::WriteCommandAxis(TasAnalog data) const {
+    auto [x, y] = data;
+    std::string line;
+    line += std::to_string(static_cast<int>(x * 32767));
+    line += ";";
+    line += std::to_string(static_cast<int>(y * 32767));
+    return line;
+}
+
+std::string Tas::WriteCommandButtons(u32 data) const {
+    if (data == 0) {
+        return "NONE";
+    }
+
+    std::string line;
+    u32 index = 0;
+    while (data > 0) {
+        if ((data & 1) == 1) {
+            for (auto [text, tas_button] : text_to_tas_button) {
+                if (tas_button == static_cast<TasButton>(1 << index)) {
+                    if (line.size() > 0) {
+                        line += ";";
+                    }
+                    line += text;
+                    break;
+                }
+            }
+        }
+        index++;
+        data >>= 1;
+    }
+    return line;
+}
+
+void Tas::StartStop() {
+    if (!Settings::values.tas_enable) {
+        return;
+    }
+    if (is_running) {
+        Stop();
+    } else {
+        is_running = true;
+        SwapToTasController();
+    }
+}
+
+void Tas::Stop() {
+    is_running = false;
+    SwapToStoredController();
+}
+
+void Tas::SwapToTasController() {
+    if (!Settings::values.tas_swap_controllers) {
+        return;
+    }
+    auto& players = Settings::values.players.GetValue();
+    for (std::size_t index = 0; index < players.size(); index++) {
+        auto& player = players[index];
+        player_mappings[index] = player;
+
+        // Only swap active controllers
+        if (!player.connected) {
+            continue;
+        }
+
+        Common::ParamPackage tas_param;
+        tas_param.Set("pad", static_cast<u8>(index));
+        auto button_mapping = GetButtonMappingForDevice(tas_param);
+        auto analog_mapping = GetAnalogMappingForDevice(tas_param);
+        auto& buttons = player.buttons;
+        auto& analogs = player.analogs;
+
+        for (std::size_t i = 0; i < buttons.size(); ++i) {
+            buttons[i] = button_mapping[static_cast<Settings::NativeButton::Values>(i)].Serialize();
+        }
+        for (std::size_t i = 0; i < analogs.size(); ++i) {
+            analogs[i] = analog_mapping[static_cast<Settings::NativeAnalog::Values>(i)].Serialize();
+        }
+    }
+    is_old_input_saved = true;
+    Settings::values.is_device_reload_pending.store(true);
+}
+
+void Tas::SwapToStoredController() {
+    if (!is_old_input_saved) {
+        return;
+    }
+    auto& players = Settings::values.players.GetValue();
+    for (std::size_t index = 0; index < players.size(); index++) {
+        players[index] = player_mappings[index];
+    }
+    is_old_input_saved = false;
+    Settings::values.is_device_reload_pending.store(true);
+}
+
+void Tas::Reset() {
+    if (!Settings::values.tas_enable) {
+        return;
+    }
+    needs_reset = true;
+}
+
+bool Tas::Record() {
+    if (!Settings::values.tas_enable) {
+        return true;
+    }
+    is_recording = !is_recording;
+    return is_recording;
+}
+
+void Tas::SaveRecording(bool overwrite_file) {
+    if (is_recording) {
+        return;
+    }
+    if (record_commands.empty()) {
+        return;
+    }
+    WriteTasFile(u8"record.txt");
+    if (overwrite_file) {
+        WriteTasFile(u8"script0-1.txt");
+    }
+    needs_reset = true;
+    record_commands.clear();
+}
+
+InputCommon::ButtonMapping Tas::GetButtonMappingForDevice(
+    const Common::ParamPackage& params) const {
+    // This list is missing ZL/ZR since those are not considered buttons.
+    // We will add those afterwards
+    // This list also excludes any button that can't be really mapped
+    static constexpr std::array<std::pair<Settings::NativeButton::Values, TasButton>, 20>
+        switch_to_tas_button = {
+            std::pair{Settings::NativeButton::A, TasButton::BUTTON_A},
+            {Settings::NativeButton::B, TasButton::BUTTON_B},
+            {Settings::NativeButton::X, TasButton::BUTTON_X},
+            {Settings::NativeButton::Y, TasButton::BUTTON_Y},
+            {Settings::NativeButton::LStick, TasButton::STICK_L},
+            {Settings::NativeButton::RStick, TasButton::STICK_R},
+            {Settings::NativeButton::L, TasButton::TRIGGER_L},
+            {Settings::NativeButton::R, TasButton::TRIGGER_R},
+            {Settings::NativeButton::Plus, TasButton::BUTTON_PLUS},
+            {Settings::NativeButton::Minus, TasButton::BUTTON_MINUS},
+            {Settings::NativeButton::DLeft, TasButton::BUTTON_LEFT},
+            {Settings::NativeButton::DUp, TasButton::BUTTON_UP},
+            {Settings::NativeButton::DRight, TasButton::BUTTON_RIGHT},
+            {Settings::NativeButton::DDown, TasButton::BUTTON_DOWN},
+            {Settings::NativeButton::SL, TasButton::BUTTON_SL},
+            {Settings::NativeButton::SR, TasButton::BUTTON_SR},
+            {Settings::NativeButton::Screenshot, TasButton::BUTTON_CAPTURE},
+            {Settings::NativeButton::Home, TasButton::BUTTON_HOME},
+            {Settings::NativeButton::ZL, TasButton::TRIGGER_ZL},
+            {Settings::NativeButton::ZR, TasButton::TRIGGER_ZR},
+        };
+
+    InputCommon::ButtonMapping mapping{};
+    for (const auto& [switch_button, tas_button] : switch_to_tas_button) {
+        Common::ParamPackage button_params({{"engine", "tas"}});
+        button_params.Set("pad", params.Get("pad", 0));
+        button_params.Set("button", static_cast<int>(tas_button));
+        mapping.insert_or_assign(switch_button, std::move(button_params));
+    }
+
+    return mapping;
+}
+
+InputCommon::AnalogMapping Tas::GetAnalogMappingForDevice(
+    const Common::ParamPackage& params) const {
+
+    InputCommon::AnalogMapping mapping = {};
+    Common::ParamPackage left_analog_params;
+    left_analog_params.Set("engine", "tas");
+    left_analog_params.Set("pad", params.Get("pad", 0));
+    left_analog_params.Set("axis_x", static_cast<int>(TasAxes::StickX));
+    left_analog_params.Set("axis_y", static_cast<int>(TasAxes::StickY));
+    mapping.insert_or_assign(Settings::NativeAnalog::LStick, std::move(left_analog_params));
+    Common::ParamPackage right_analog_params;
+    right_analog_params.Set("engine", "tas");
+    right_analog_params.Set("pad", params.Get("pad", 0));
+    right_analog_params.Set("axis_x", static_cast<int>(TasAxes::SubstickX));
+    right_analog_params.Set("axis_y", static_cast<int>(TasAxes::SubstickY));
+    mapping.insert_or_assign(Settings::NativeAnalog::RStick, std::move(right_analog_params));
+    return mapping;
+}
+
+const TasData& Tas::GetTasState(std::size_t pad) const {
+    return tas_data[pad];
+}
+} // namespace TasInput

src/input_common/tas/tas_input.h

@@ -0,0 +1,237 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+#include "common/common_types.h"
+#include "common/settings_input.h"
+#include "core/frontend/input.h"
+#include "input_common/main.h"
+
+/*
+To play back TAS scripts on Yuzu, select the folder with scripts in the configuration menu below
+Tools -> Configure TAS. The file itself has normal text format and has to be called script0-1.txt
+for controller 1, script0-2.txt for controller 2 and so forth (with max. 8 players).
+
+A script file has the same format as TAS-nx uses, so final files will look like this:
+
+1 KEY_B 0;0 0;0
+6 KEY_ZL 0;0 0;0
+41 KEY_ZL;KEY_Y 0;0 0;0
+43 KEY_X;KEY_A 32767;0 0;0
+44 KEY_A 32767;0 0;0
+45 KEY_A 32767;0 0;0
+46 KEY_A 32767;0 0;0
+47 KEY_A 32767;0 0;0
+
+After placing the file at the correct location, it can be read into Yuzu with the (default) hotkey
+CTRL+F6 (refresh). In the bottom left corner, it will display the amount of frames the script file
+has. Playback can be started or stopped using CTRL+F5.
+
+However, for playback to actually work, the correct input device has to be selected: In the Controls
+menu, select TAS from the device list for the controller that the script should be played on.
+
+Recording a new script file is really simple: Just make sure that the proper device (not TAS) is
+connected on P1, and press CTRL+F7 to start recording. When done, just press the same keystroke
+again (CTRL+F7). The new script will be saved at the location previously selected, as the filename
+record.txt.
+
+For debugging purposes, the common controller debugger can be used (View -> Debugging -> Controller
+P1).
+*/
+
+namespace TasInput {
+
+constexpr size_t PLAYER_NUMBER = 8;
+
+using TasAnalog = std::pair<float, float>;
+
+enum class TasState {
+    Running,
+    Recording,
+    Stopped,
+};
+
+enum class TasButton : u32 {
+    BUTTON_A = 1U << 0,
+    BUTTON_B = 1U << 1,
+    BUTTON_X = 1U << 2,
+    BUTTON_Y = 1U << 3,
+    STICK_L = 1U << 4,
+    STICK_R = 1U << 5,
+    TRIGGER_L = 1U << 6,
+    TRIGGER_R = 1U << 7,
+    TRIGGER_ZL = 1U << 8,
+    TRIGGER_ZR = 1U << 9,
+    BUTTON_PLUS = 1U << 10,
+    BUTTON_MINUS = 1U << 11,
+    BUTTON_LEFT = 1U << 12,
+    BUTTON_UP = 1U << 13,
+    BUTTON_RIGHT = 1U << 14,
+    BUTTON_DOWN = 1U << 15,
+    BUTTON_SL = 1U << 16,
+    BUTTON_SR = 1U << 17,
+    BUTTON_HOME = 1U << 18,
+    BUTTON_CAPTURE = 1U << 19,
+};
+
+enum class TasAxes : u8 {
+    StickX,
+    StickY,
+    SubstickX,
+    SubstickY,
+    Undefined,
+};
+
+struct TasData {
+    u32 buttons{};
+    std::array<float, 4> axis{};
+};
+
+class Tas {
+public:
+    Tas();
+    ~Tas();
+
+    // Changes the input status that will be stored in each frame
+    void RecordInput(u32 buttons, const std::array<std::pair<float, float>, 2>& axes);
+
+    // Main loop that records or executes input
+    void UpdateThread();
+
+    //  Sets the flag to start or stop the TAS command excecution and swaps controllers profiles
+    void StartStop();
+
+    //  Stop the TAS and reverts any controller profile
+    void Stop();
+
+    // Sets the flag to reload the file and start from the begining in the next update
+    void Reset();
+
+    /**
+     * Sets the flag to enable or disable recording of inputs
+     * @return Returns true if the current recording status is enabled
+     */
+    bool Record();
+
+    // Saves contents of record_commands on a file if overwrite is enabled player 1 will be
+    // overwritten with the recorded commands
+    void SaveRecording(bool overwrite_file);
+
+    /**
+     * Returns the current status values of TAS playback/recording
+     * @return Tuple of
+     * TasState indicating the current state out of Running, Recording or Stopped ;
+     * Current playback progress or amount of frames (so far) for Recording ;
+     * Total length of script file currently loaded or amount of frames (so far) for Recording
+     */
+    std::tuple<TasState, size_t, size_t> GetStatus() const;
+
+    // Retuns an array of the default button mappings
+    InputCommon::ButtonMapping GetButtonMappingForDevice(const Common::ParamPackage& params) const;
+
+    // Retuns an array of the default analog mappings
+    InputCommon::AnalogMapping GetAnalogMappingForDevice(const Common::ParamPackage& params) const;
+    [[nodiscard]] const TasData& GetTasState(std::size_t pad) const;
+
+private:
+    struct TASCommand {
+        u32 buttons{};
+        TasAnalog l_axis{};
+        TasAnalog r_axis{};
+    };
+
+    // Loads TAS files from all players
+    void LoadTasFiles();
+
+    // Loads TAS file from the specified player
+    void LoadTasFile(size_t player_index);
+
+    // Writes a TAS file from the recorded commands
+    void WriteTasFile(std::u8string file_name);
+
+    /**
+     * Parses a string containing the axis values with the following format "x;y"
+     * X and Y have a range from -32767 to 32767
+     * @return Returns a TAS analog object with axis values with range from -1.0 to 1.0
+     */
+    TasAnalog ReadCommandAxis(const std::string& line) const;
+
+    /**
+     * Parses a string containing the button values with the following format "a;b;c;d..."
+     * Each button is represented by it's text format specified in text_to_tas_button array
+     * @return Returns a u32 with each bit representing the status of a button
+     */
+    u32 ReadCommandButtons(const std::string& line) const;
+
+    /**
+     * Converts an u32 containing the button status into the text equivalent
+     * @return Returns a string with the name of the buttons to be written to the file
+     */
+    std::string WriteCommandButtons(u32 data) const;
+
+    /**
+     * Converts an TAS analog object containing the axis status into the text equivalent
+     * @return Returns a string with the value of the axis to be written to the file
+     */
+    std::string WriteCommandAxis(TasAnalog data) const;
+
+    // Inverts the Y axis polarity
+    std::pair<float, float> FlipAxisY(std::pair<float, float> old);
+
+    /**
+     * Converts an u32 containing the button status into the text equivalent
+     * @return Returns a string with the name of the buttons to be printed on console
+     */
+    std::string DebugButtons(u32 buttons) const;
+
+    /**
+     * Converts an TAS analog object containing the axis status into the text equivalent
+     * @return Returns a string with the value of the axis to be printed on console
+     */
+    std::string DebugJoystick(float x, float y) const;
+
+    /**
+     * Converts the given TAS status into the text equivalent
+     * @return Returns a string with the value of the TAS status to be printed on console
+     */
+    std::string DebugInput(const TasData& data) const;
+
+    /**
+     * Converts the given TAS status of multiple players into the text equivalent
+     * @return Returns a string with the value of the status of all TAS players to be printed on
+     * console
+     */
+    std::string DebugInputs(const std::array<TasData, PLAYER_NUMBER>& arr) const;
+
+    /**
+     * Converts an u32 containing the button status into the text equivalent
+     * @return Returns a string with the name of the buttons
+     */
+    std::string ButtonsToString(u32 button) const;
+
+    // Stores current controller configuration and sets a TAS controller for every active controller
+    // to the current config
+    void SwapToTasController();
+
+    // Sets the stored controller configuration to the current config
+    void SwapToStoredController();
+
+    size_t script_length{0};
+    std::array<TasData, PLAYER_NUMBER> tas_data;
+    bool is_old_input_saved{false};
+    bool is_recording{false};
+    bool is_running{false};
+    bool needs_reset{false};
+    std::array<std::vector<TASCommand>, PLAYER_NUMBER> commands{};
+    std::vector<TASCommand> record_commands{};
+    size_t current_command{0};
+    TASCommand last_input{}; // only used for recording
+
+    // Old settings for swapping controllers
+    std::array<Settings::PlayerInput, 10> player_mappings;
+};
+} // namespace TasInput

src/input_common/tas/tas_poller.cpp

@@ -0,0 +1,101 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <mutex>
+#include <utility>
+
+#include "common/settings.h"
+#include "common/threadsafe_queue.h"
+#include "input_common/tas/tas_input.h"
+#include "input_common/tas/tas_poller.h"
+
+namespace InputCommon {
+
+class TasButton final : public Input::ButtonDevice {
+public:
+    explicit TasButton(u32 button_, u32 pad_, const TasInput::Tas* tas_input_)
+        : button(button_), pad(pad_), tas_input(tas_input_) {}
+
+    bool GetStatus() const override {
+        return (tas_input->GetTasState(pad).buttons & button) != 0;
+    }
+
+private:
+    const u32 button;
+    const u32 pad;
+    const TasInput::Tas* tas_input;
+};
+
+TasButtonFactory::TasButtonFactory(std::shared_ptr<TasInput::Tas> tas_input_)
+    : tas_input(std::move(tas_input_)) {}
+
+std::unique_ptr<Input::ButtonDevice> TasButtonFactory::Create(const Common::ParamPackage& params) {
+    const auto button_id = params.Get("button", 0);
+    const auto pad = params.Get("pad", 0);
+
+    return std::make_unique<TasButton>(button_id, pad, tas_input.get());
+}
+
+class TasAnalog final : public Input::AnalogDevice {
+public:
+    explicit TasAnalog(u32 pad_, u32 axis_x_, u32 axis_y_, const TasInput::Tas* tas_input_)
+        : pad(pad_), axis_x(axis_x_), axis_y(axis_y_), tas_input(tas_input_) {}
+
+    float GetAxis(u32 axis) const {
+        std::lock_guard lock{mutex};
+        return tas_input->GetTasState(pad).axis.at(axis);
+    }
+
+    std::pair<float, float> GetAnalog(u32 analog_axis_x, u32 analog_axis_y) const {
+        float x = GetAxis(analog_axis_x);
+        float y = GetAxis(analog_axis_y);
+
+        // Make sure the coordinates are in the unit circle,
+        // otherwise normalize it.
+        float r = x * x + y * y;
+        if (r > 1.0f) {
+            r = std::sqrt(r);
+            x /= r;
+            y /= r;
+        }
+
+        return {x, y};
+    }
+
+    std::tuple<float, float> GetStatus() const override {
+        return GetAnalog(axis_x, axis_y);
+    }
+
+    Input::AnalogProperties GetAnalogProperties() const override {
+        return {0.0f, 1.0f, 0.5f};
+    }
+
+private:
+    const u32 pad;
+    const u32 axis_x;
+    const u32 axis_y;
+    const TasInput::Tas* tas_input;
+    mutable std::mutex mutex;
+};
+
+/// An analog device factory that creates analog devices from GC Adapter
+TasAnalogFactory::TasAnalogFactory(std::shared_ptr<TasInput::Tas> tas_input_)
+    : tas_input(std::move(tas_input_)) {}
+
+/**
+ * Creates analog device from joystick axes
+ * @param params contains parameters for creating the device:
+ *     - "port": the nth gcpad on the adapter
+ *     - "axis_x": the index of the axis to be bind as x-axis
+ *     - "axis_y": the index of the axis to be bind as y-axis
+ */
+std::unique_ptr<Input::AnalogDevice> TasAnalogFactory::Create(const Common::ParamPackage& params) {
+    const auto pad = static_cast<u32>(params.Get("pad", 0));
+    const auto axis_x = static_cast<u32>(params.Get("axis_x", 0));
+    const auto axis_y = static_cast<u32>(params.Get("axis_y", 1));
+
+    return std::make_unique<TasAnalog>(pad, axis_x, axis_y, tas_input.get());
+}
+
+} // namespace InputCommon

src/input_common/tas/tas_poller.h

@@ -0,0 +1,43 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include "core/frontend/input.h"
+#include "input_common/tas/tas_input.h"
+
+namespace InputCommon {
+
+/**
+ * A button device factory representing a tas bot. It receives tas events and forward them
+ * to all button devices it created.
+ */
+class TasButtonFactory final : public Input::Factory<Input::ButtonDevice> {
+public:
+    explicit TasButtonFactory(std::shared_ptr<TasInput::Tas> tas_input_);
+
+    /**
+     * Creates a button device from a button press
+     * @param params contains parameters for creating the device:
+     *     - "code": the code of the key to bind with the button
+     */
+    std::unique_ptr<Input::ButtonDevice> Create(const Common::ParamPackage& params) override;
+
+private:
+    std::shared_ptr<TasInput::Tas> tas_input;
+};
+
+/// An analog device factory that creates analog devices from tas
+class TasAnalogFactory final : public Input::Factory<Input::AnalogDevice> {
+public:
+    explicit TasAnalogFactory(std::shared_ptr<TasInput::Tas> tas_input_);
+
+    std::unique_ptr<Input::AnalogDevice> Create(const Common::ParamPackage& params) override;
+
+private:
+    std::shared_ptr<TasInput::Tas> tas_input;
+};
+
+} // namespace InputCommon

src/input_common/udp/client.h

@@ -21,8 +21,6 @@
 
 namespace InputCommon::CemuhookUDP {
 
-constexpr char DEFAULT_SRV[] = "127.0.0.1:26760";
-
 class Socket;
 
 namespace Response {

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:
@@ -307,6 +341,12 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
         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));
@@ -316,6 +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));
@@ -333,8 +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));
@@ -429,7 +477,13 @@ void EmitSetSampleMask(EmitContext& ctx, Id value) {
 }
 
 void EmitSetFragDepth(EmitContext& ctx, Id value) {
-    ctx.OpStore(ctx.frag_depth, value);
+    if (!ctx.runtime_info.convert_depth_mode) {
+        ctx.OpStore(ctx.frag_depth, value);
+        return;
+    }
+    const Id unit{ctx.Const(0.5f)};
+    const Id new_depth{ctx.OpFma(ctx.F32[1], value, unit, unit)};
+    ctx.OpStore(ctx.frag_depth, new_depth);
 }
 
 void EmitGetZFlag(EmitContext&) {

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/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/video_core/CMakeLists.txt

@@ -231,6 +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

@@ -5,6 +5,7 @@
 #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"
@@ -16,44 +17,28 @@ extern "C" {
 }
 
 namespace Tegra {
-#if defined(LIBVA_FOUND)
-// Hardware acceleration code from FFmpeg/doc/examples/hw_decode.c originally under MIT license
 namespace {
-constexpr std::array<const char*, 2> VAAPI_DRIVERS = {
-    "i915",
-    "amdgpu",
-};
+constexpr AVPixelFormat PREFERRED_GPU_FMT = AV_PIX_FMT_NV12;
+constexpr AVPixelFormat PREFERRED_CPU_FMT = AV_PIX_FMT_YUV420P;
 
-AVPixelFormat GetHwFormat(AVCodecContext*, const AVPixelFormat* pix_fmts) {
+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_PIX_FMT_VAAPI) {
-            return AV_PIX_FMT_VAAPI;
+        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");
-    return *pix_fmts;
-}
-
-bool CreateVaapiHwdevice(AVBufferRef** av_hw_device) {
-    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_hw_device, 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);
-            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);
-    return false;
+    av_buffer_unref(&av_codec_ctx->hw_device_ctx);
+    av_codec_ctx->pix_fmt = PREFERRED_CPU_FMT;
+    return PREFERRED_CPU_FMT;
 }
 } // namespace
-#endif
 
 void AVFrameDeleter(AVFrame* ptr) {
     av_frame_free(&ptr);
@@ -68,56 +53,110 @@ Codec::~Codec() {
         return;
     }
     // Free libav memory
-    avcodec_send_packet(av_codec_ctx, nullptr);
-    AVFrame* av_frame = av_frame_alloc();
-    avcodec_receive_frame(av_codec_ctx, av_frame);
-    avcodec_flush_buffers(av_codec_ctx);
-    av_frame_free(&av_frame);
-    avcodec_close(av_codec_ctx);
-    av_buffer_unref(&av_hw_device);
+    avcodec_free_context(&av_codec_ctx);
+    av_buffer_unref(&av_gpu_decoder);
 }
 
-void Codec::InitializeHwdec() {
-    // Prioritize integrated GPU to mitigate bandwidth bottlenecks
+bool Codec::CreateGpuAvDevice() {
 #if defined(LIBVA_FOUND)
-    if (CreateVaapiHwdevice(&av_hw_device)) {
-        const auto hw_device_ctx = av_buffer_ref(av_hw_device);
-        ASSERT_MSG(hw_device_ctx, "av_buffer_ref failed");
-        av_codec_ctx->hw_device_ctx = hw_device_ctx;
-        av_codec_ctx->get_format = GetHwFormat;
+    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;
     }
-#endif
-    // TODO more GPU accelerated decoders
+    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;
-    switch (current_codec) {
-    case NvdecCommon::VideoCodec::H264:
-        codec = AV_CODEC_ID_H264;
-        break;
-    case NvdecCommon::VideoCodec::Vp9:
-        codec = AV_CODEC_ID_VP9;
-        break;
-    default:
-        UNIMPLEMENTED_MSG("Unknown codec {}", current_codec);
+    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);
-    InitializeHwdec();
     if (!av_codec_ctx->hw_device_ctx) {
         LOG_INFO(Service_NVDRV, "Using FFmpeg software decoding");
     }
-    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);
-        av_buffer_unref(&av_hw_device);
-        return;
-    }
     initialized = true;
 }
 
@@ -133,6 +172,9 @@ void Codec::Decode() {
     if (is_first_frame) {
         Initialize();
     }
+    if (!initialized) {
+        return;
+    }
     bool vp9_hidden_frame = false;
     std::vector<u8> frame_data;
     if (current_codec == NvdecCommon::VideoCodec::H264) {
@@ -141,50 +183,48 @@ void Codec::Decode() {
         frame_data = vp9_decoder->ComposeFrameHeader(state);
         vp9_hidden_frame = vp9_decoder->WasFrameHidden();
     }
-    AVPacket packet{};
-    av_init_packet(&packet);
-    packet.data = frame_data.data();
-    packet.size = static_cast<s32>(frame_data.size());
-    if (const int ret = avcodec_send_packet(av_codec_ctx, &packet); ret) {
-        LOG_DEBUG(Service_NVDRV, "avcodec_send_packet error {}", ret);
+    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;
     }
-    AVFrame* hw_frame = av_frame_alloc();
-    AVFrame* sw_frame = hw_frame;
-    ASSERT_MSG(hw_frame, "av_frame_alloc hw_frame failed");
-    if (const int ret = avcodec_receive_frame(av_codec_ctx, hw_frame); ret) {
+    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);
-        av_frame_free(&hw_frame);
         return;
     }
-    if (!hw_frame->width || !hw_frame->height) {
+    if (initial_frame->width == 0 || initial_frame->height == 0) {
         LOG_WARNING(Service_NVDRV, "Zero width or height in frame");
-        av_frame_free(&hw_frame);
         return;
     }
-#if defined(LIBVA_FOUND)
-    // Hardware acceleration code from FFmpeg/doc/examples/hw_decode.c under MIT license
-    if (hw_frame->format == AV_PIX_FMT_VAAPI) {
-        sw_frame = av_frame_alloc();
-        ASSERT_MSG(sw_frame, "av_frame_alloc sw_frame failed");
+    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
-        sw_frame->format = AV_PIX_FMT_NV12;
-        const int transfer_data_ret = av_hwframe_transfer_data(sw_frame, hw_frame, 0);
-        ASSERT_MSG(!transfer_data_ret, "av_hwframe_transfer_data error {}", transfer_data_ret);
-        av_frame_free(&hw_frame);
+        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);
     }
-#endif
-    if (sw_frame->format != AV_PIX_FMT_YUV420P && sw_frame->format != AV_PIX_FMT_NV12) {
-        UNIMPLEMENTED_MSG("Unexpected video format from host graphics: {}", sw_frame->format);
-        av_frame_free(&sw_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(AVFramePtr{sw_frame, AVFrameDeleter});
+    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();

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"
@@ -50,18 +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 InitializeHwdec();
+    void InitializeAvCodecContext();
+
+    void InitializeGpuDecoder();
+
+    bool CreateGpuAvDevice();
 
     bool initialized{};
     NvdecCommon::VideoCodec current_codec{NvdecCommon::VideoCodec::None};
 
     AVCodec* av_codec{nullptr};
-    AVBufferRef* av_hw_device{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

@@ -742,6 +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);

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 = true,
+            .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,
         };
     }
 };
@@ -316,7 +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/engines/maxwell_3d.h

@@ -475,10 +475,10 @@ public:
 
                 // These values are used by Nouveau and some games.
                 AddGL = 0x8006,
-                SubtractGL = 0x8007,
-                ReverseSubtractGL = 0x8008,
-                MinGL = 0x800a,
-                MaxGL = 0x800b
+                MinGL = 0x8007,
+                MaxGL = 0x8008,
+                SubtractGL = 0x800a,
+                ReverseSubtractGL = 0x800b
             };
 
             enum class Factor : u32 {

src/video_core/gpu.cpp

@@ -531,14 +531,6 @@ void GPU::TriggerCpuInterrupt(const u32 syncpoint_id, const u32 value) const {
     interrupt_manager.GPUInterruptSyncpt(syncpoint_id, value);
 }
 
-void GPU::ShutDown() {
-    // Signal that threads should no longer block on syncpoint fences
-    shutting_down.store(true, std::memory_order_relaxed);
-    sync_cv.notify_all();
-
-    gpu_thread.ShutDown();
-}
-
 void GPU::OnCommandListEnd() {
     if (is_async) {
         // This command only applies to asynchronous GPU mode

src/video_core/gpu.h

@@ -219,9 +219,6 @@ public:
         return *shader_notify;
     }
 
-    // Stops the GPU execution and waits for the GPU to finish working
-    void ShutDown();
-
     /// Allows the CPU/NvFlinger to wait on the GPU before presenting a frame.
     void WaitFence(u32 syncpoint_id, u32 value);
 

src/video_core/gpu_thread.cpp

@@ -17,9 +17,9 @@
 namespace VideoCommon::GPUThread {
 
 /// Runs the GPU thread
-static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
-                      Core::Frontend::GraphicsContext& context, Tegra::DmaPusher& dma_pusher,
-                      SynchState& state) {
+static void RunThread(std::stop_token stop_token, Core::System& system,
+                      VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
+                      Tegra::DmaPusher& dma_pusher, SynchState& state) {
     std::string name = "yuzu:GPU";
     MicroProfileOnThreadCreate(name.c_str());
     SCOPE_EXIT({ MicroProfileOnThreadExit(); });
@@ -28,20 +28,14 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
     Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
     system.RegisterHostThread();
 
-    // Wait for first GPU command before acquiring the window context
-    state.queue.Wait();
-
-    // If emulation was stopped during disk shader loading, abort before trying to acquire context
-    if (!state.is_running) {
-        return;
-    }
-
     auto current_context = context.Acquire();
     VideoCore::RasterizerInterface* const rasterizer = renderer.ReadRasterizer();
 
-    CommandDataContainer next;
-    while (state.is_running) {
-        next = state.queue.PopWait();
+    while (!stop_token.stop_requested()) {
+        CommandDataContainer next = state.queue.PopWait(stop_token);
+        if (stop_token.stop_requested()) {
+            break;
+        }
         if (auto* submit_list = std::get_if<SubmitListCommand>(&next.data)) {
             dma_pusher.Push(std::move(submit_list->entries));
             dma_pusher.DispatchCalls();
@@ -55,8 +49,6 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
             rasterizer->FlushRegion(flush->addr, flush->size);
         } else if (const auto* invalidate = std::get_if<InvalidateRegionCommand>(&next.data)) {
             rasterizer->OnCPUWrite(invalidate->addr, invalidate->size);
-        } else if (std::holds_alternative<EndProcessingCommand>(next.data)) {
-            ASSERT(state.is_running == false);
         } else {
             UNREACHABLE();
         }
@@ -73,16 +65,14 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
 ThreadManager::ThreadManager(Core::System& system_, bool is_async_)
     : system{system_}, is_async{is_async_} {}
 
-ThreadManager::~ThreadManager() {
-    ShutDown();
-}
+ThreadManager::~ThreadManager() = default;
 
 void ThreadManager::StartThread(VideoCore::RendererBase& renderer,
                                 Core::Frontend::GraphicsContext& context,
                                 Tegra::DmaPusher& dma_pusher) {
     rasterizer = renderer.ReadRasterizer();
-    thread = std::thread(RunThread, std::ref(system), std::ref(renderer), std::ref(context),
-                         std::ref(dma_pusher), std::ref(state));
+    thread = std::jthread(RunThread, std::ref(system), std::ref(renderer), std::ref(context),
+                          std::ref(dma_pusher), std::ref(state));
 }
 
 void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
@@ -117,26 +107,6 @@ void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     rasterizer->OnCPUWrite(addr, size);
 }
 
-void ThreadManager::ShutDown() {
-    if (!state.is_running) {
-        return;
-    }
-
-    {
-        std::lock_guard lk(state.write_lock);
-        state.is_running = false;
-        state.cv.notify_all();
-    }
-
-    if (!thread.joinable()) {
-        return;
-    }
-
-    // Notify GPU thread that a shutdown is pending
-    PushCommand(EndProcessingCommand());
-    thread.join();
-}
-
 void ThreadManager::OnCommandListEnd() {
     PushCommand(OnCommandListEndCommand());
 }
@@ -152,9 +122,8 @@ u64 ThreadManager::PushCommand(CommandData&& command_data, bool block) {
     state.queue.Push(CommandDataContainer(std::move(command_data), fence, block));
 
     if (block) {
-        state.cv.wait(lk, [this, fence] {
-            return fence <= state.signaled_fence.load(std::memory_order_relaxed) ||
-                   !state.is_running;
+        state.cv.wait(lk, thread.get_stop_token(), [this, fence] {
+            return fence <= state.signaled_fence.load(std::memory_order_relaxed);
         });
     }
 

src/video_core/gpu_thread.h

@@ -33,9 +33,6 @@ class RendererBase;
 
 namespace VideoCommon::GPUThread {
 
-/// Command to signal to the GPU thread that processing has ended
-struct EndProcessingCommand final {};
-
 /// Command to signal to the GPU thread that a command list is ready for processing
 struct SubmitListCommand final {
     explicit SubmitListCommand(Tegra::CommandList&& entries_) : entries{std::move(entries_)} {}
@@ -83,7 +80,7 @@ struct OnCommandListEndCommand final {};
 struct GPUTickCommand final {};
 
 using CommandData =
-    std::variant<EndProcessingCommand, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
+    std::variant<std::monostate, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
                  InvalidateRegionCommand, FlushAndInvalidateRegionCommand, OnCommandListEndCommand,
                  GPUTickCommand>;
 
@@ -100,14 +97,12 @@ struct CommandDataContainer {
 
 /// Struct used to synchronize the GPU thread
 struct SynchState final {
-    std::atomic_bool is_running{true};
-
-    using CommandQueue = Common::SPSCQueue<CommandDataContainer>;
+    using CommandQueue = Common::SPSCQueue<CommandDataContainer, true>;
     std::mutex write_lock;
     CommandQueue queue;
     u64 last_fence{};
     std::atomic<u64> signaled_fence{};
-    std::condition_variable cv;
+    std::condition_variable_any cv;
 };
 
 /// Class used to manage the GPU thread
@@ -149,7 +144,7 @@ private:
     VideoCore::RasterizerInterface* rasterizer = nullptr;
 
     SynchState state;
-    std::thread thread;
+    std::jthread thread;
 };
 
 } // namespace VideoCommon::GPUThread

src/video_core/memory_manager.cpp

@@ -2,6 +2,8 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
+
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
@@ -463,6 +465,7 @@ std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
         ++page_index;
         page_offset = 0;
         remaining_size -= num_bytes;
+        old_page_addr = page_addr;
     }
     split();
     return result;

src/video_core/renderer_opengl/gl_device.h

@@ -156,6 +156,10 @@ public:
         return shader_backend;
     }
 
+    bool IsAmd() const {
+        return vendor_name == "ATI Technologies Inc.";
+    }
+
 private:
     static bool TestVariableAoffi();
     static bool TestPreciseBug();

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -219,6 +219,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
       host_info{
           .support_float16 = false,
           .support_int64 = device.HasShaderInt64(),
+          .needs_demote_reorder = device.IsAmd(),
       } {
     if (use_asynchronous_shaders) {
         workers = CreateWorkers();
@@ -292,6 +293,8 @@ void ShaderCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading,
     }};
     LoadPipelines(stop_loading, shader_cache_filename, CACHE_VERSION, load_compute, load_graphics);
 
+    LOG_INFO(Render_OpenGL, "Total Pipeline Count: {}", state.total);
+
     std::unique_lock lock{state.mutex};
     callback(VideoCore::LoadCallbackStage::Build, 0, state.total);
     state.has_loaded = true;

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -149,7 +149,7 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
         const Layout::FramebufferLayout layout = render_window.GetFramebufferLayout();
         swapchain.Create(layout.width, layout.height, is_srgb);
     };
-    if (swapchain.IsSubOptimal() || swapchain.HasColorSpaceChanged(is_srgb)) {
+    if (swapchain.NeedsRecreation(is_srgb)) {
         recreate_swapchain();
     }
     bool is_outdated;
@@ -164,7 +164,8 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
         blit_screen.Recreate();
     }
     const VkSemaphore render_semaphore = blit_screen.DrawToSwapchain(*framebuffer, use_accelerated);
-    scheduler.Flush(render_semaphore);
+    const VkSemaphore present_semaphore = swapchain.CurrentPresentSemaphore();
+    scheduler.Flush(render_semaphore, present_semaphore);
     scheduler.WaitWorker();
     swapchain.Present(render_semaphore);
 

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -159,11 +159,13 @@ VkSemaphore VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
 
         const VAddr framebuffer_addr = framebuffer.address + framebuffer.offset;
         const u8* const host_ptr = cpu_memory.GetPointer(framebuffer_addr);
-        const size_t size_bytes = GetSizeInBytes(framebuffer);
 
         // TODO(Rodrigo): Read this from HLE
         constexpr u32 block_height_log2 = 4;
         const u32 bytes_per_pixel = GetBytesPerPixel(framebuffer);
+        const u64 size_bytes{Tegra::Texture::CalculateSize(true, bytes_per_pixel,
+                                                           framebuffer.stride, framebuffer.height,
+                                                           1, block_height_log2, 0)};
         Tegra::Texture::UnswizzleTexture(
             mapped_span.subspan(image_offset, size_bytes), std::span(host_ptr, size_bytes),
             bytes_per_pixel, framebuffer.width, framebuffer.height, 1, block_height_log2, 0);
@@ -356,7 +358,7 @@ void VKBlitScreen::CreateDescriptorPool() {
 void VKBlitScreen::CreateRenderPass() {
     const VkAttachmentDescription color_attachment{
         .flags = 0,
-        .format = swapchain.GetImageFormat(),
+        .format = swapchain.GetImageViewFormat(),
         .samples = VK_SAMPLE_COUNT_1_BIT,
         .loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR,
         .storeOp = VK_ATTACHMENT_STORE_OP_STORE,

src/video_core/renderer_vulkan/vk_descriptor_pool.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <mutex>
 #include <span>
 #include <vector>
@@ -18,7 +19,6 @@ namespace Vulkan {
 // Prefer small grow rates to avoid saturating the descriptor pool with barely used pipelines
 constexpr size_t SETS_GROW_RATE = 16;
 constexpr s32 SCORE_THRESHOLD = 3;
-constexpr u32 SETS_PER_POOL = 64;
 
 struct DescriptorBank {
     DescriptorBankInfo info;
@@ -58,11 +58,12 @@ static DescriptorBankInfo MakeBankInfo(std::span<const Shader::Info> infos) {
 static void AllocatePool(const Device& device, DescriptorBank& bank) {
     std::array<VkDescriptorPoolSize, 6> pool_sizes;
     size_t pool_cursor{};
+    const u32 sets_per_pool = device.GetSetsPerPool();
     const auto add = [&](VkDescriptorType type, u32 count) {
         if (count > 0) {
             pool_sizes[pool_cursor++] = {
                 .type = type,
-                .descriptorCount = count * SETS_PER_POOL,
+                .descriptorCount = count * sets_per_pool,
             };
         }
     };
@@ -77,7 +78,7 @@ static void AllocatePool(const Device& device, DescriptorBank& bank) {
         .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO,
         .pNext = nullptr,
         .flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT,
-        .maxSets = SETS_PER_POOL,
+        .maxSets = sets_per_pool,
         .poolSizeCount = static_cast<u32>(pool_cursor),
         .pPoolSizes = std::data(pool_sizes),
     }));

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -281,7 +281,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, Tegra::Engines::Maxw
         .supported_spirv = device.IsKhrSpirv1_4Supported() ? 0x00010400U : 0x00010000U,
         .unified_descriptor_binding = true,
         .support_descriptor_aliasing = true,
-        .support_int8 = true,
+        .support_int8 = device.IsInt8Supported(),
         .support_int16 = device.IsShaderInt16Supported(),
         .support_int64 = device.IsShaderInt64Supported(),
         .support_vertex_instance_id = false,
@@ -325,6 +325,8 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, Tegra::Engines::Maxw
     host_info = Shader::HostTranslateInfo{
         .support_float16 = device.IsFloat16Supported(),
         .support_int64 = device.IsShaderInt64Supported(),
+        .needs_demote_reorder = driver_id == VK_DRIVER_ID_AMD_PROPRIETARY_KHR ||
+                                driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE_KHR,
     };
 }
 
@@ -445,6 +447,8 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
     VideoCommon::LoadPipelines(stop_loading, pipeline_cache_filename, CACHE_VERSION, load_compute,
                                load_graphics);
 
+    LOG_INFO(Render_Vulkan, "Total Pipeline Count: {}", state.total);
+
     std::unique_lock lock{state.mutex};
     callback(VideoCore::LoadCallbackStage::Build, 0, state.total);
     state.has_loaded = true;

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -228,9 +228,7 @@ void RasterizerVulkan::Clear() {
     };
 
     const u32 color_attachment = regs.clear_buffers.RT;
-    const auto attachment_aspect_mask = framebuffer->ImageRanges()[color_attachment].aspectMask;
-    const bool is_color_rt = (attachment_aspect_mask & VK_IMAGE_ASPECT_COLOR_BIT) != 0;
-    if (use_color && is_color_rt) {
+    if (use_color && framebuffer->HasAspectColorBit(color_attachment)) {
         VkClearValue clear_value;
         std::memcpy(clear_value.color.float32, regs.clear_color, sizeof(regs.clear_color));
 
@@ -248,12 +246,15 @@ void RasterizerVulkan::Clear() {
         return;
     }
     VkImageAspectFlags aspect_flags = 0;
-    if (use_depth) {
+    if (use_depth && framebuffer->HasAspectDepthBit()) {
         aspect_flags |= VK_IMAGE_ASPECT_DEPTH_BIT;
     }
-    if (use_stencil) {
+    if (use_stencil && framebuffer->HasAspectStencilBit()) {
         aspect_flags |= VK_IMAGE_ASPECT_STENCIL_BIT;
     }
+    if (aspect_flags == 0) {
+        return;
+    }
     scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil,
                       clear_rect, aspect_flags](vk::CommandBuffer cmdbuf) {
         VkClearAttachment attachment;
@@ -764,12 +765,7 @@ void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
     const Maxwell::StencilOp zpass = regs.stencil_front_op_zpass;
     const Maxwell::ComparisonOp compare = regs.stencil_front_func_func;
     if (regs.stencil_two_side_enable) {
-        scheduler.Record([fail, zfail, zpass, compare](vk::CommandBuffer cmdbuf) {
-            cmdbuf.SetStencilOpEXT(VK_STENCIL_FACE_FRONT_AND_BACK, MaxwellToVK::StencilOp(fail),
-                                   MaxwellToVK::StencilOp(zpass), MaxwellToVK::StencilOp(zfail),
-                                   MaxwellToVK::ComparisonOp(compare));
-        });
-    } else {
+        // Separate stencil op per face
         const Maxwell::StencilOp back_fail = regs.stencil_back_op_fail;
         const Maxwell::StencilOp back_zfail = regs.stencil_back_op_zfail;
         const Maxwell::StencilOp back_zpass = regs.stencil_back_op_zpass;
@@ -784,6 +780,13 @@ void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
                                    MaxwellToVK::StencilOp(back_zfail),
                                    MaxwellToVK::ComparisonOp(back_compare));
         });
+    } else {
+        // Front face defines the stencil op of both faces
+        scheduler.Record([fail, zfail, zpass, compare](vk::CommandBuffer cmdbuf) {
+            cmdbuf.SetStencilOpEXT(VK_STENCIL_FACE_FRONT_AND_BACK, MaxwellToVK::StencilOp(fail),
+                                   MaxwellToVK::StencilOp(zpass), MaxwellToVK::StencilOp(zfail),
+                                   MaxwellToVK::ComparisonOp(compare));
+        });
     }
 }