registration: Various style and documentation improvements

Fix logic in RealVfsFilesystem Create methods
Remove magic numbers
Fix regex errors

CMakeLists.txt

@@ -1,5 +1,5 @@
-cmake_minimum_required(VERSION 3.7)
-
+# CMake 3.6 required for FindBoost to define IMPORTED libs properly on unknown Boost versions
+cmake_minimum_required(VERSION 3.6)
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMakeModules")
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/externals/cmake-modules")
 include(DownloadExternals)
@@ -187,8 +187,8 @@ find_package(Threads REQUIRED)
 if (ENABLE_SDL2)
     if (YUZU_USE_BUNDLED_SDL2)
         # Detect toolchain and platform
-        if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1920) AND ARCHITECTURE_x86_64)
-            set(SDL2_VER "SDL2-2.0.8")
+        if (MSVC14 AND ARCHITECTURE_x86_64)
+            set(SDL2_VER "SDL2-2.0.5")
         else()
             message(FATAL_ERROR "No bundled SDL2 binaries for your toolchain. Disable YUZU_USE_BUNDLED_SDL2 and provide your own.")
         endif()
@@ -220,7 +220,7 @@ if (YUZU_USE_BUNDLED_UNICORN)
     if (MSVC)
         message(STATUS "unicorn not found, falling back to bundled")
         # Detect toolchain and platform
-        if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1920) AND ARCHITECTURE_x86_64)
+        if (MSVC14 AND ARCHITECTURE_x86_64)
             set(UNICORN_VER "unicorn-yuzu")
         else()
             message(FATAL_ERROR "No bundled Unicorn binaries for your toolchain. Disable YUZU_USE_BUNDLED_UNICORN and provide your own.")
@@ -279,7 +279,7 @@ endif()
 
 if (ENABLE_QT)
     if (YUZU_USE_BUNDLED_QT)
-        if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1920) AND ARCHITECTURE_x86_64)
+        if (MSVC14 AND ARCHITECTURE_x86_64)
             set(QT_VER qt-5.10.0-msvc2015_64)
         else()
             message(FATAL_ERROR "No bundled Qt binaries for your toolchain. Disable YUZU_USE_BUNDLED_QT and provide your own.")
@@ -303,7 +303,7 @@ endif()
 # ======================================
 
 IF (APPLE)
-    find_library(COCOA_LIBRARY Cocoa)           # Umbrella framework for everything GUI-related
+    FIND_LIBRARY(COCOA_LIBRARY Cocoa)           # Umbrella framework for everything GUI-related
     set(PLATFORM_LIBRARIES ${COCOA_LIBRARY} ${IOKIT_LIBRARY} ${COREVIDEO_LIBRARY})
 
     if (CMAKE_CXX_COMPILER_ID STREQUAL Clang)

src/audio_core/audio_renderer.cpp

@@ -26,18 +26,6 @@ AudioRenderer::AudioRenderer(AudioRendererParameter params,
     QueueMixedBuffer(2);
 }
 
-u32 AudioRenderer::GetSampleRate() const {
-    return worker_params.sample_rate;
-}
-
-u32 AudioRenderer::GetSampleCount() const {
-    return worker_params.sample_count;
-}
-
-u32 AudioRenderer::GetMixBufferCount() const {
-    return worker_params.mix_buffer_count;
-}
-
 std::vector<u8> AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_params) {
     // Copy UpdateDataHeader struct
     UpdateDataHeader config{};

src/audio_core/audio_renderer.h

@@ -26,7 +26,7 @@ enum class PlayState : u8 {
 struct AudioRendererParameter {
     u32_le sample_rate;
     u32_le sample_count;
-    u32_le mix_buffer_count;
+    u32_le unknown_8;
     u32_le unknown_c;
     u32_le voice_count;
     u32_le sink_count;
@@ -160,9 +160,6 @@ public:
     std::vector<u8> UpdateAudioRenderer(const std::vector<u8>& input_params);
     void QueueMixedBuffer(Buffer::Tag tag);
     void ReleaseAndQueueBuffers();
-    u32 GetSampleRate() const;
-    u32 GetSampleCount() const;
-    u32 GetMixBufferCount() const;
 
 private:
     class VoiceState {

src/audio_core/cubeb_sink.cpp

@@ -4,7 +4,6 @@
 
 #include <algorithm>
 #include <cstring>
-#include <mutex>
 
 #include "audio_core/cubeb_sink.h"
 #include "audio_core/stream.h"
@@ -67,8 +66,6 @@ public:
             return;
         }
 
-        std::lock_guard lock{queue_mutex};
-
         queue.reserve(queue.size() + samples.size() * GetNumChannels());
 
         if (is_6_channel) {
@@ -97,7 +94,6 @@ private:
     u32 num_channels{};
     bool is_6_channel{};
 
-    std::mutex queue_mutex;
     std::vector<s16> queue;
 
     static long DataCallback(cubeb_stream* stream, void* user_data, const void* input_buffer,
@@ -157,8 +153,6 @@ long SinkStreamImpl::DataCallback(cubeb_stream* stream, void* user_data, const v
         return {};
     }
 
-    std::lock_guard lock{impl->queue_mutex};
-
     const size_t frames_to_write{
         std::min(impl->queue.size() / impl->GetNumChannels(), static_cast<size_t>(num_frames))};
 

src/common/CMakeLists.txt

@@ -40,6 +40,8 @@ add_library(common STATIC
     file_util.cpp
     file_util.h
     hash.h
+    hex_util.cpp
+    hex_util.h
     logging/backend.cpp
     logging/backend.h
     logging/filter.cpp

src/common/file_util.cpp

@@ -750,6 +750,12 @@ std::string GetHactoolConfigurationPath() {
 #endif
 }
 
+std::string GetNANDRegistrationDir(bool system) {
+    if (system)
+        return GetUserPath(UserPath::NANDDir) + "system/Contents/registered/";
+    return GetUserPath(UserPath::NANDDir) + "user/Contents/registered/";
+}
+
 size_t WriteStringToFile(bool text_file, const std::string& str, const char* filename) {
     return FileUtil::IOFile(filename, text_file ? "w" : "wb").WriteBytes(str.data(), str.size());
 }

src/common/file_util.h

@@ -129,6 +129,8 @@ const std::string& GetUserPath(UserPath path, const std::string& new_path = "");
 
 std::string GetHactoolConfigurationPath();
 
+std::string GetNANDRegistrationDir(bool system = false);
+
 // Returns the path to where the sys file are
 std::string GetSysDirectory();
 

src/common/hex_util.cpp

@@ -0,0 +1,27 @@
+// Copyright 2013 Dolphin Emulator Project / 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/hex_util.h"
+
+u8 ToHexNibble(char c1) {
+    if (c1 >= 65 && c1 <= 70)
+        return c1 - 55;
+    if (c1 >= 97 && c1 <= 102)
+        return c1 - 87;
+    if (c1 >= 48 && c1 <= 57)
+        return c1 - 48;
+    throw std::logic_error("Invalid hex digit");
+}
+
+std::array<u8, 16> operator""_array16(const char* str, size_t len) {
+    if (len != 32)
+        throw std::logic_error("Not of correct size.");
+    return HexStringToArray<16>(str);
+}
+
+std::array<u8, 32> operator""_array32(const char* str, size_t len) {
+    if (len != 64)
+        throw std::logic_error("Not of correct size.");
+    return HexStringToArray<32>(str);
+}

src/common/hex_util.h

@@ -0,0 +1,37 @@
+// 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 <array>
+#include <cstddef>
+#include <string>
+#include <fmt/format.h>
+#include "common/common_types.h"
+
+u8 ToHexNibble(char c1);
+
+template <size_t Size, bool le = false>
+std::array<u8, Size> HexStringToArray(std::string_view str) {
+    std::array<u8, Size> out{};
+    if constexpr (le) {
+        for (size_t i = 2 * Size - 2; i <= 2 * Size; i -= 2)
+            out[i / 2] = (ToHexNibble(str[i]) << 4) | ToHexNibble(str[i + 1]);
+    } else {
+        for (size_t i = 0; i < 2 * Size; i += 2)
+            out[i / 2] = (ToHexNibble(str[i]) << 4) | ToHexNibble(str[i + 1]);
+    }
+    return out;
+}
+
+template <size_t Size>
+std::string HexArrayToString(std::array<u8, Size> array, bool upper = true) {
+    std::string out;
+    for (u8 c : array)
+        out += fmt::format(upper ? "{:02X}" : "{:02x}", c);
+    return out;
+}
+
+std::array<u8, 0x10> operator"" _array16(const char* str, size_t len);
+std::array<u8, 0x20> operator"" _array32(const char* str, size_t len);

src/common/thread_queue_list.h

@@ -16,7 +16,7 @@ struct ThreadQueueList {
     //               (dynamically resizable) circular buffers to remove their overhead when
     //               inserting and popping.
 
-    using Priority = unsigned int;
+    typedef unsigned int Priority;
 
     // Number of priority levels. (Valid levels are [0..NUM_QUEUES).)
     static const Priority NUM_QUEUES = N;
@@ -26,9 +26,9 @@ struct ThreadQueueList {
     }
 
     // Only for debugging, returns priority level.
-    Priority contains(const T& uid) const {
+    Priority contains(const T& uid) {
         for (Priority i = 0; i < NUM_QUEUES; ++i) {
-            const Queue& cur = queues[i];
+            Queue& cur = queues[i];
             if (std::find(cur.data.cbegin(), cur.data.cend(), uid) != cur.data.cend()) {
                 return i;
             }
@@ -37,8 +37,8 @@ struct ThreadQueueList {
         return -1;
     }
 
-    T get_first() const {
-        const Queue* cur = first;
+    T get_first() {
+        Queue* cur = first;
         while (cur != nullptr) {
             if (!cur->data.empty()) {
                 return cur->data.front();

src/core/CMakeLists.txt

@@ -20,6 +20,8 @@ add_library(core STATIC
     crypto/key_manager.h
     crypto/ctr_encryption_layer.cpp
     crypto/ctr_encryption_layer.h
+    file_sys/bis_factory.cpp
+    file_sys/bis_factory.h
     file_sys/card_image.cpp
     file_sys/card_image.h
     file_sys/content_archive.cpp
@@ -29,10 +31,14 @@ add_library(core STATIC
     file_sys/directory.h
     file_sys/errors.h
     file_sys/mode.h
+    file_sys/nca_metadata.cpp
+    file_sys/nca_metadata.h
     file_sys/partition_filesystem.cpp
     file_sys/partition_filesystem.h
     file_sys/program_metadata.cpp
     file_sys/program_metadata.h
+    file_sys/registered_cache.cpp
+    file_sys/registered_cache.h
     file_sys/romfs.cpp
     file_sys/romfs.h
     file_sys/romfs_factory.cpp
@@ -43,6 +49,8 @@ add_library(core STATIC
     file_sys/sdmc_factory.h
     file_sys/vfs.cpp
     file_sys/vfs.h
+    file_sys/vfs_concat.cpp
+    file_sys/vfs_concat.h
     file_sys/vfs_offset.cpp
     file_sys/vfs_offset.h
     file_sys/vfs_real.cpp

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -86,16 +86,7 @@ public:
     }
 
     void AddTicks(u64 ticks) override {
-        // Divide the number of ticks by the amount of CPU cores. TODO(Subv): This yields only a
-        // rough approximation of the amount of executed ticks in the system, it may be thrown off
-        // if not all cores are doing a similar amount of work. Instead of doing this, we should
-        // device a way so that timing is consistent across all cores without increasing the ticks 4
-        // times.
-        u64 amortized_ticks = (ticks - num_interpreted_instructions) / Core::NUM_CPU_CORES;
-        // Always execute at least one tick.
-        amortized_ticks = std::max<u64>(amortized_ticks, 1);
-
-        CoreTiming::AddTicks(amortized_ticks);
+        CoreTiming::AddTicks(ticks - num_interpreted_instructions);
         num_interpreted_instructions = 0;
     }
     u64 GetTicksRemaining() override {
@@ -243,7 +234,9 @@ void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
 }
 
 void ARM_Dynarmic::PrepareReschedule() {
-    jit->HaltExecution();
+    if (jit->IsExecuting()) {
+        jit->HaltExecution();
+    }
 }
 
 void ARM_Dynarmic::ClearInstructionCache() {

src/core/core.cpp

@@ -5,6 +5,7 @@
 #include <memory>
 #include <utility>
 #include "common/logging/log.h"
+#include "common/string_util.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/gdbstub/gdbstub.h"
@@ -17,6 +18,7 @@
 #include "core/hle/service/sm/sm.h"
 #include "core/loader/loader.h"
 #include "core/settings.h"
+#include "file_sys/vfs_concat.h"
 #include "file_sys/vfs_real.h"
 #include "video_core/renderer_base.h"
 #include "video_core/video_core.h"
@@ -88,8 +90,39 @@ System::ResultStatus System::SingleStep() {
     return RunLoop(false);
 }
 
+static FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
+                                                const std::string& path) {
+    // To account for split 00+01+etc files.
+    std::string dir_name;
+    std::string filename;
+    Common::SplitPath(path, &dir_name, &filename, nullptr);
+    if (filename == "00") {
+        const auto dir = vfs->OpenDirectory(dir_name, FileSys::Mode::Read);
+        std::vector<FileSys::VirtualFile> concat;
+        for (u8 i = 0; i < 0x10; ++i) {
+            auto next = dir->GetFile(fmt::format("{:02X}", i));
+            if (next != nullptr)
+                concat.push_back(std::move(next));
+            else {
+                next = dir->GetFile(fmt::format("{:02x}", i));
+                if (next != nullptr)
+                    concat.push_back(std::move(next));
+                else
+                    break;
+            }
+        }
+
+        if (concat.empty())
+            return nullptr;
+
+        return FileSys::ConcatenateFiles(concat, dir->GetName());
+    }
+
+    return vfs->OpenFile(path, FileSys::Mode::Read);
+}
+
 System::ResultStatus System::Load(Frontend::EmuWindow& emu_window, const std::string& filepath) {
-    app_loader = Loader::GetLoader(virtual_filesystem->OpenFile(filepath, FileSys::Mode::Read));
+    app_loader = Loader::GetLoader(GetGameFileFromPath(virtual_filesystem, filepath));
 
     if (!app_loader) {
         LOG_CRITICAL(Core, "Failed to obtain loader for {}!", filepath);

src/core/core_cpu.cpp

@@ -14,7 +14,6 @@
 #include "core/core_timing.h"
 #include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
-#include "core/hle/lock.h"
 #include "core/settings.h"
 
 namespace Core {
@@ -91,7 +90,6 @@ void Cpu::RunLoop(bool tight_loop) {
         LOG_TRACE(Core, "Core-{} idling", core_index);
 
         if (IsMainCore()) {
-            // TODO(Subv): Only let CoreTiming idle if all 4 cores are idling.
             CoreTiming::Idle();
             CoreTiming::Advance();
         }
@@ -127,8 +125,6 @@ void Cpu::Reschedule() {
     }
 
     reschedule_pending = false;
-    // Lock the global kernel mutex when we manipulate the HLE state
-    std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock);
     scheduler->Reschedule();
 }
 

src/core/core_cpu.h

@@ -79,7 +79,7 @@ private:
     std::shared_ptr<CpuBarrier> cpu_barrier;
     std::shared_ptr<Kernel::Scheduler> scheduler;
 
-    std::atomic<bool> reschedule_pending = false;
+    bool reschedule_pending{};
     size_t core_index;
 };
 

src/core/core_timing.cpp

@@ -135,9 +135,11 @@ void ClearPendingEvents() {
 void ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata) {
     ASSERT(event_type != nullptr);
     s64 timeout = GetTicks() + cycles_into_future;
+
     // If this event needs to be scheduled before the next advance(), force one early
     if (!is_global_timer_sane)
         ForceExceptionCheck(cycles_into_future);
+
     event_queue.emplace_back(Event{timeout, event_fifo_id++, userdata, event_type});
     std::push_heap(event_queue.begin(), event_queue.end(), std::greater<>());
 }

src/core/crypto/key_manager.cpp

@@ -10,44 +10,13 @@
 #include <string_view>
 #include "common/common_paths.h"
 #include "common/file_util.h"
+#include "common/hex_util.h"
+#include "common/logging/log.h"
 #include "core/crypto/key_manager.h"
 #include "core/settings.h"
 
 namespace Core::Crypto {
 
-static u8 ToHexNibble(char c1) {
-    if (c1 >= 65 && c1 <= 70)
-        return c1 - 55;
-    if (c1 >= 97 && c1 <= 102)
-        return c1 - 87;
-    if (c1 >= 48 && c1 <= 57)
-        return c1 - 48;
-    throw std::logic_error("Invalid hex digit");
-}
-
-template <size_t Size>
-static std::array<u8, Size> HexStringToArray(std::string_view str) {
-    std::array<u8, Size> out{};
-    for (size_t i = 0; i < 2 * Size; i += 2) {
-        auto d1 = str[i];
-        auto d2 = str[i + 1];
-        out[i / 2] = (ToHexNibble(d1) << 4) | ToHexNibble(d2);
-    }
-    return out;
-}
-
-std::array<u8, 16> operator""_array16(const char* str, size_t len) {
-    if (len != 32)
-        throw std::logic_error("Not of correct size.");
-    return HexStringToArray<16>(str);
-}
-
-std::array<u8, 32> operator""_array32(const char* str, size_t len) {
-    if (len != 64)
-        throw std::logic_error("Not of correct size.");
-    return HexStringToArray<32>(str);
-}
-
 KeyManager::KeyManager() {
     // Initialize keys
     const std::string hactool_keys_dir = FileUtil::GetHactoolConfigurationPath();

src/core/crypto/key_manager.h

@@ -87,9 +87,6 @@ struct hash<Core::Crypto::KeyIndex<KeyType>> {
 
 namespace Core::Crypto {
 
-std::array<u8, 0x10> operator"" _array16(const char* str, size_t len);
-std::array<u8, 0x20> operator"" _array32(const char* str, size_t len);
-
 class KeyManager {
 public:
     KeyManager();

src/core/file_sys/bis_factory.cpp

@@ -0,0 +1,31 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/file_sys/bis_factory.h"
+
+namespace FileSys {
+
+static VirtualDir GetOrCreateDirectory(const VirtualDir& dir, std::string_view path) {
+    const auto res = dir->GetDirectoryRelative(path);
+    if (res == nullptr)
+        return dir->CreateDirectoryRelative(path);
+    return res;
+}
+
+BISFactory::BISFactory(VirtualDir nand_root_)
+    : nand_root(std::move(nand_root_)),
+      sysnand_cache(std::make_shared<RegisteredCache>(
+          GetOrCreateDirectory(nand_root, "/system/Contents/registered"))),
+      usrnand_cache(std::make_shared<RegisteredCache>(
+          GetOrCreateDirectory(nand_root, "/user/Contents/registered"))) {}
+
+std::shared_ptr<RegisteredCache> BISFactory::GetSystemNANDContents() const {
+    return sysnand_cache;
+}
+
+std::shared_ptr<RegisteredCache> BISFactory::GetUserNANDContents() const {
+    return usrnand_cache;
+}
+
+} // namespace FileSys

src/core/file_sys/bis_factory.h

@@ -0,0 +1,30 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include "core/loader/loader.h"
+#include "registered_cache.h"
+
+namespace FileSys {
+
+/// File system interface to the Built-In Storage
+/// This is currently missing accessors to BIS partitions, but seemed like a good place for the NAND
+/// registered caches.
+class BISFactory {
+public:
+    explicit BISFactory(VirtualDir nand_root);
+
+    std::shared_ptr<RegisteredCache> GetSystemNANDContents() const;
+    std::shared_ptr<RegisteredCache> GetUserNANDContents() const;
+
+private:
+    VirtualDir nand_root;
+
+    std::shared_ptr<RegisteredCache> sysnand_cache;
+    std::shared_ptr<RegisteredCache> usrnand_cache;
+};
+
+} // namespace FileSys

src/core/file_sys/card_image.cpp

@@ -93,6 +93,10 @@ VirtualDir XCI::GetLogoPartition() const {
     return GetPartition(XCIPartition::Logo);
 }
 
+const std::vector<std::shared_ptr<NCA>>& XCI::GetNCAs() const {
+    return ncas;
+}
+
 std::shared_ptr<NCA> XCI::GetNCAByType(NCAContentType type) const {
     const auto iter =
         std::find_if(ncas.begin(), ncas.end(),
@@ -107,19 +111,19 @@ VirtualFile XCI::GetNCAFileByType(NCAContentType type) const {
     return nullptr;
 }
 
-std::vector<VirtualFile> XCI::GetFiles() const {
+std::vector<std::shared_ptr<VfsFile>> XCI::GetFiles() const {
     return {};
 }
 
-std::vector<VirtualDir> XCI::GetSubdirectories() const {
-    return {};
+std::vector<std::shared_ptr<VfsDirectory>> XCI::GetSubdirectories() const {
+    return std::vector<std::shared_ptr<VfsDirectory>>();
 }
 
 std::string XCI::GetName() const {
     return file->GetName();
 }
 
-VirtualDir XCI::GetParentDirectory() const {
+std::shared_ptr<VfsDirectory> XCI::GetParentDirectory() const {
     return file->GetContainingDirectory();
 }
 

src/core/file_sys/card_image.h

@@ -68,16 +68,17 @@ public:
     VirtualDir GetUpdatePartition() const;
     VirtualDir GetLogoPartition() const;
 
+    const std::vector<std::shared_ptr<NCA>>& GetNCAs() const;
     std::shared_ptr<NCA> GetNCAByType(NCAContentType type) const;
     VirtualFile GetNCAFileByType(NCAContentType type) const;
 
-    std::vector<VirtualFile> GetFiles() const override;
+    std::vector<std::shared_ptr<VfsFile>> GetFiles() const override;
 
-    std::vector<VirtualDir> GetSubdirectories() const override;
+    std::vector<std::shared_ptr<VfsDirectory>> GetSubdirectories() const override;
 
     std::string GetName() const override;
 
-    VirtualDir GetParentDirectory() const override;
+    std::shared_ptr<VfsDirectory> GetParentDirectory() const override;
 
 protected:
     bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;

src/core/file_sys/control_metadata.cpp

@@ -16,7 +16,7 @@ std::string LanguageEntry::GetDeveloperName() const {
     return Common::StringFromFixedZeroTerminatedBuffer(developer_name.data(), 0x100);
 }
 
-NACP::NACP(VirtualFile file_) : file(std::move(file_)), raw(std::make_unique<RawNACP>()) {
+NACP::NACP(VirtualFile file) : raw(std::make_unique<RawNACP>()) {
     file->ReadObject(raw.get());
 }
 

src/core/file_sys/control_metadata.h

@@ -81,7 +81,6 @@ public:
     std::string GetVersionString() const;
 
 private:
-    VirtualFile file;
     std::unique_ptr<RawNACP> raw;
 };
 

src/core/file_sys/nca_metadata.cpp

@@ -0,0 +1,131 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include "common/common_funcs.h"
+#include "common/logging/log.h"
+#include "common/swap.h"
+#include "content_archive.h"
+#include "core/file_sys/nca_metadata.h"
+
+namespace FileSys {
+
+bool operator>=(TitleType lhs, TitleType rhs) {
+    return static_cast<size_t>(lhs) >= static_cast<size_t>(rhs);
+}
+
+bool operator<=(TitleType lhs, TitleType rhs) {
+    return static_cast<size_t>(lhs) <= static_cast<size_t>(rhs);
+}
+
+CNMT::CNMT(VirtualFile file) {
+    if (file->ReadObject(&header) != sizeof(CNMTHeader))
+        return;
+
+    // If type is {Application, Update, AOC} has opt-header.
+    if (header.type >= TitleType::Application && header.type <= TitleType::AOC) {
+        if (file->ReadObject(&opt_header, sizeof(CNMTHeader)) != sizeof(OptionalHeader)) {
+            LOG_WARNING(Loader, "Failed to read optional header.");
+        }
+    }
+
+    for (u16 i = 0; i < header.number_content_entries; ++i) {
+        auto& next = content_records.emplace_back(ContentRecord{});
+        if (file->ReadObject(&next, sizeof(CNMTHeader) + i * sizeof(ContentRecord) +
+                                        header.table_offset) != sizeof(ContentRecord)) {
+            content_records.erase(content_records.end() - 1);
+        }
+    }
+
+    for (u16 i = 0; i < header.number_meta_entries; ++i) {
+        auto& next = meta_records.emplace_back(MetaRecord{});
+        if (file->ReadObject(&next, sizeof(CNMTHeader) + i * sizeof(MetaRecord) +
+                                        header.table_offset) != sizeof(MetaRecord)) {
+            meta_records.erase(meta_records.end() - 1);
+        }
+    }
+}
+
+CNMT::CNMT(CNMTHeader header, OptionalHeader opt_header, std::vector<ContentRecord> content_records,
+           std::vector<MetaRecord> meta_records)
+    : header(std::move(header)), opt_header(std::move(opt_header)),
+      content_records(std::move(content_records)), meta_records(std::move(meta_records)) {}
+
+u64 CNMT::GetTitleID() const {
+    return header.title_id;
+}
+
+u32 CNMT::GetTitleVersion() const {
+    return header.title_version;
+}
+
+TitleType CNMT::GetType() const {
+    return header.type;
+}
+
+const std::vector<ContentRecord>& CNMT::GetContentRecords() const {
+    return content_records;
+}
+
+const std::vector<MetaRecord>& CNMT::GetMetaRecords() const {
+    return meta_records;
+}
+
+bool CNMT::UnionRecords(const CNMT& other) {
+    bool change = false;
+    for (const auto& rec : other.content_records) {
+        const auto iter = std::find_if(content_records.begin(), content_records.end(),
+                                       [&rec](const ContentRecord& r) {
+                                           return r.nca_id == rec.nca_id && r.type == rec.type;
+                                       });
+        if (iter == content_records.end()) {
+            content_records.emplace_back(rec);
+            ++header.number_content_entries;
+            change = true;
+        }
+    }
+    for (const auto& rec : other.meta_records) {
+        const auto iter =
+            std::find_if(meta_records.begin(), meta_records.end(), [&rec](const MetaRecord& r) {
+                return r.title_id == rec.title_id && r.title_version == rec.title_version &&
+                       r.type == rec.type;
+            });
+        if (iter == meta_records.end()) {
+            meta_records.emplace_back(rec);
+            ++header.number_meta_entries;
+            change = true;
+        }
+    }
+    return change;
+}
+
+std::vector<u8> CNMT::Serialize() const {
+    const bool has_opt_header =
+        header.type >= TitleType::Application && header.type <= TitleType::AOC;
+    const auto dead_zone = header.table_offset + sizeof(CNMTHeader);
+    std::vector<u8> out(
+        std::max(sizeof(CNMTHeader) + (has_opt_header ? sizeof(OptionalHeader) : 0), dead_zone) +
+        content_records.size() * sizeof(ContentRecord) + meta_records.size() * sizeof(MetaRecord));
+    memcpy(out.data(), &header, sizeof(CNMTHeader));
+
+    // Optional Header
+    if (has_opt_header) {
+        memcpy(out.data() + sizeof(CNMTHeader), &opt_header, sizeof(OptionalHeader));
+    }
+
+    auto offset = header.table_offset;
+
+    for (const auto& rec : content_records) {
+        memcpy(out.data() + offset + sizeof(CNMTHeader), &rec, sizeof(ContentRecord));
+        offset += sizeof(ContentRecord);
+    }
+
+    for (const auto& rec : meta_records) {
+        memcpy(out.data() + offset + sizeof(CNMTHeader), &rec, sizeof(MetaRecord));
+        offset += sizeof(MetaRecord);
+    }
+
+    return out;
+}
+} // namespace FileSys

src/core/file_sys/nca_metadata.h

@@ -0,0 +1,111 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstring>
+#include <memory>
+#include <vector>
+#include "common/common_types.h"
+#include "common/swap.h"
+#include "core/file_sys/vfs.h"
+
+namespace FileSys {
+class CNMT;
+
+struct CNMTHeader;
+struct OptionalHeader;
+
+enum class TitleType : u8 {
+    SystemProgram = 0x01,
+    SystemDataArchive = 0x02,
+    SystemUpdate = 0x03,
+    FirmwarePackageA = 0x04,
+    FirmwarePackageB = 0x05,
+    Application = 0x80,
+    Update = 0x81,
+    AOC = 0x82,
+    DeltaTitle = 0x83,
+};
+
+bool operator>=(TitleType lhs, TitleType rhs);
+bool operator<=(TitleType lhs, TitleType rhs);
+
+enum class ContentRecordType : u8 {
+    Meta = 0,
+    Program = 1,
+    Data = 2,
+    Control = 3,
+    Manual = 4,
+    Legal = 5,
+    Patch = 6,
+};
+
+struct ContentRecord {
+    std::array<u8, 0x20> hash;
+    std::array<u8, 0x10> nca_id;
+    std::array<u8, 0x6> size;
+    ContentRecordType type;
+    INSERT_PADDING_BYTES(1);
+};
+static_assert(sizeof(ContentRecord) == 0x38, "ContentRecord has incorrect size.");
+
+constexpr ContentRecord EMPTY_META_CONTENT_RECORD{{}, {}, {}, ContentRecordType::Meta, {}};
+
+struct MetaRecord {
+    u64_le title_id;
+    u32_le title_version;
+    TitleType type;
+    u8 install_byte;
+    INSERT_PADDING_BYTES(2);
+};
+static_assert(sizeof(MetaRecord) == 0x10, "MetaRecord has incorrect size.");
+
+struct OptionalHeader {
+    u64_le title_id;
+    u64_le minimum_version;
+};
+static_assert(sizeof(OptionalHeader) == 0x10, "OptionalHeader has incorrect size.");
+
+struct CNMTHeader {
+    u64_le title_id;
+    u32_le title_version;
+    TitleType type;
+    INSERT_PADDING_BYTES(1);
+    u16_le table_offset;
+    u16_le number_content_entries;
+    u16_le number_meta_entries;
+    INSERT_PADDING_BYTES(12);
+};
+static_assert(sizeof(CNMTHeader) == 0x20, "CNMTHeader has incorrect size.");
+
+// A class representing the format used by NCA metadata files, typically named {}.cnmt.nca or
+// meta0.ncd. These describe which NCA's belong with which titles in the registered cache.
+class CNMT {
+public:
+    explicit CNMT(VirtualFile file);
+    CNMT(CNMTHeader header, OptionalHeader opt_header, std::vector<ContentRecord> content_records,
+         std::vector<MetaRecord> meta_records);
+
+    u64 GetTitleID() const;
+    u32 GetTitleVersion() const;
+    TitleType GetType() const;
+
+    const std::vector<ContentRecord>& GetContentRecords() const;
+    const std::vector<MetaRecord>& GetMetaRecords() const;
+
+    bool UnionRecords(const CNMT& other);
+    std::vector<u8> Serialize() const;
+
+private:
+    CNMTHeader header;
+    OptionalHeader opt_header;
+    std::vector<ContentRecord> content_records;
+    std::vector<MetaRecord> meta_records;
+
+    // TODO(DarkLordZach): According to switchbrew, for Patch-type there is additional data
+    // after the table. This is not documented, unfortunately.
+};
+
+} // namespace FileSys

src/core/file_sys/partition_filesystem.h

@@ -13,7 +13,7 @@
 #include "core/file_sys/vfs.h"
 
 namespace Loader {
-enum class ResultStatus : u16;
+enum class ResultStatus;
 }
 
 namespace FileSys {

src/core/file_sys/program_metadata.h

@@ -13,7 +13,7 @@
 #include "partition_filesystem.h"
 
 namespace Loader {
-enum class ResultStatus : u16;
+enum class ResultStatus;
 }
 
 namespace FileSys {

src/core/file_sys/registered_cache.cpp

@@ -0,0 +1,476 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <regex>
+#include <mbedtls/sha256.h>
+#include "common/assert.h"
+#include "common/hex_util.h"
+#include "common/logging/log.h"
+#include "core/crypto/encryption_layer.h"
+#include "core/file_sys/card_image.h"
+#include "core/file_sys/nca_metadata.h"
+#include "core/file_sys/registered_cache.h"
+#include "core/file_sys/vfs_concat.h"
+
+namespace FileSys {
+std::string RegisteredCacheEntry::DebugInfo() const {
+    return fmt::format("title_id={:016X}, content_type={:02X}", title_id, static_cast<u8>(type));
+}
+
+bool operator<(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs) {
+    return (lhs.title_id < rhs.title_id) || (lhs.title_id == rhs.title_id && lhs.type < rhs.type);
+}
+
+static bool FollowsTwoDigitDirFormat(std::string_view name) {
+    static const std::regex two_digit_regex("000000[0-9A-F]{2}", std::regex_constants::ECMAScript |
+                                                                     std::regex_constants::icase);
+    return std::regex_match(name.begin(), name.end(), two_digit_regex);
+}
+
+static bool FollowsNcaIdFormat(std::string_view name) {
+    static const std::regex nca_id_regex("[0-9A-F]{32}\\.nca", std::regex_constants::ECMAScript |
+                                                                   std::regex_constants::icase);
+    return name.size() == 36 && std::regex_match(name.begin(), name.end(), nca_id_regex);
+}
+
+static std::string GetRelativePathFromNcaID(const std::array<u8, 16>& nca_id, bool second_hex_upper,
+                                            bool within_two_digit) {
+    if (!within_two_digit)
+        return fmt::format("/{}.nca", HexArrayToString(nca_id, second_hex_upper));
+
+    Core::Crypto::SHA256Hash hash{};
+    mbedtls_sha256(nca_id.data(), nca_id.size(), hash.data(), 0);
+    return fmt::format("/000000{:02X}/{}.nca", hash[0], HexArrayToString(nca_id, second_hex_upper));
+}
+
+static std::string GetCNMTName(TitleType type, u64 title_id) {
+    constexpr std::array<const char*, 9> TITLE_TYPE_NAMES{
+        "SystemProgram",
+        "SystemData",
+        "SystemUpdate",
+        "BootImagePackage",
+        "BootImagePackageSafe",
+        "Application",
+        "Patch",
+        "AddOnContent",
+        "" ///< Currently unknown 'DeltaTitle'
+    };
+
+    auto index = static_cast<size_t>(type);
+    // If the index is after the jump in TitleType, subtract it out.
+    if (index >= static_cast<size_t>(TitleType::Application)) {
+        index -= static_cast<size_t>(TitleType::Application) -
+                 static_cast<size_t>(TitleType::FirmwarePackageB);
+    }
+    return fmt::format("{}_{:016x}.cnmt", TITLE_TYPE_NAMES[index], title_id);
+}
+
+static ContentRecordType GetCRTypeFromNCAType(NCAContentType type) {
+    switch (type) {
+    case NCAContentType::Program:
+        // TODO(DarkLordZach): Differentiate between Program and Patch
+        return ContentRecordType::Program;
+    case NCAContentType::Meta:
+        return ContentRecordType::Meta;
+    case NCAContentType::Control:
+        return ContentRecordType::Control;
+    case NCAContentType::Data:
+        return ContentRecordType::Data;
+    case NCAContentType::Manual:
+        // TODO(DarkLordZach): Peek at NCA contents to differentiate Manual and Legal.
+        return ContentRecordType::Manual;
+    default:
+        UNREACHABLE();
+    }
+}
+
+VirtualFile RegisteredCache::OpenFileOrDirectoryConcat(const VirtualDir& dir,
+                                                       std::string_view path) const {
+    if (dir->GetFileRelative(path) != nullptr)
+        return dir->GetFileRelative(path);
+    if (dir->GetDirectoryRelative(path) != nullptr) {
+        const auto nca_dir = dir->GetDirectoryRelative(path);
+        VirtualFile file = nullptr;
+
+        const auto files = nca_dir->GetFiles();
+        if (files.size() == 1 && files[0]->GetName() == "00") {
+            file = files[0];
+        } else {
+            std::vector<VirtualFile> concat;
+            // Since the files are a two-digit hex number, max is FF.
+            for (size_t i = 0; i < 0x100; ++i) {
+                auto next = nca_dir->GetFile(fmt::format("{:02X}", i));
+                if (next != nullptr) {
+                    concat.push_back(std::move(next));
+                } else {
+                    next = nca_dir->GetFile(fmt::format("{:02x}", i));
+                    if (next != nullptr)
+                        concat.push_back(std::move(next));
+                    else
+                        break;
+                }
+            }
+
+            if (concat.empty())
+                return nullptr;
+
+            file = FileSys::ConcatenateFiles(concat);
+        }
+
+        return file;
+    }
+    return nullptr;
+}
+
+VirtualFile RegisteredCache::GetFileAtID(NcaID id) const {
+    VirtualFile file;
+    // Try all four modes of file storage:
+    // (bit 1 = uppercase/lower, bit 0 = within a two-digit dir)
+    // 00: /000000**/{:032X}.nca
+    // 01: /{:032X}.nca
+    // 10: /000000**/{:032x}.nca
+    // 11: /{:032x}.nca
+    for (u8 i = 0; i < 4; ++i) {
+        const auto path = GetRelativePathFromNcaID(id, (i & 0b10) == 0, (i & 0b01) == 0);
+        file = OpenFileOrDirectoryConcat(dir, path);
+        if (file != nullptr)
+            return file;
+    }
+    return file;
+}
+
+static boost::optional<NcaID> CheckMapForContentRecord(
+    const boost::container::flat_map<u64, CNMT>& map, u64 title_id, ContentRecordType type) {
+    if (map.find(title_id) == map.end())
+        return boost::none;
+
+    const auto& cnmt = map.at(title_id);
+
+    const auto iter = std::find_if(cnmt.GetContentRecords().begin(), cnmt.GetContentRecords().end(),
+                                   [type](const ContentRecord& rec) { return rec.type == type; });
+    if (iter == cnmt.GetContentRecords().end())
+        return boost::none;
+
+    return boost::make_optional(iter->nca_id);
+}
+
+boost::optional<NcaID> RegisteredCache::GetNcaIDFromMetadata(u64 title_id,
+                                                             ContentRecordType type) const {
+    if (type == ContentRecordType::Meta && meta_id.find(title_id) != meta_id.end())
+        return meta_id.at(title_id);
+
+    const auto res1 = CheckMapForContentRecord(yuzu_meta, title_id, type);
+    if (res1 != boost::none)
+        return res1;
+    return CheckMapForContentRecord(meta, title_id, type);
+}
+
+std::vector<NcaID> RegisteredCache::AccumulateFiles() const {
+    std::vector<NcaID> ids;
+    for (const auto& d2_dir : dir->GetSubdirectories()) {
+        if (FollowsNcaIdFormat(d2_dir->GetName())) {
+            ids.push_back(HexStringToArray<0x10, true>(d2_dir->GetName().substr(0, 0x20)));
+            continue;
+        }
+
+        if (!FollowsTwoDigitDirFormat(d2_dir->GetName()))
+            continue;
+
+        for (const auto& nca_dir : d2_dir->GetSubdirectories()) {
+            if (!FollowsNcaIdFormat(nca_dir->GetName()))
+                continue;
+
+            ids.push_back(HexStringToArray<0x10, true>(nca_dir->GetName().substr(0, 0x20)));
+        }
+
+        for (const auto& nca_file : d2_dir->GetFiles()) {
+            if (!FollowsNcaIdFormat(nca_file->GetName()))
+                continue;
+
+            ids.push_back(HexStringToArray<0x10, true>(nca_file->GetName().substr(0, 0x20)));
+        }
+    }
+
+    for (const auto& d2_file : dir->GetFiles()) {
+        if (FollowsNcaIdFormat(d2_file->GetName()))
+            ids.push_back(HexStringToArray<0x10, true>(d2_file->GetName().substr(0, 0x20)));
+    }
+    return ids;
+}
+
+void RegisteredCache::ProcessFiles(const std::vector<NcaID>& ids) {
+    for (const auto& id : ids) {
+        const auto file = GetFileAtID(id);
+
+        if (file == nullptr)
+            continue;
+        const auto nca = std::make_shared<NCA>(parser(file, id));
+        if (nca->GetStatus() != Loader::ResultStatus::Success ||
+            nca->GetType() != NCAContentType::Meta) {
+            continue;
+        }
+
+        const auto section0 = nca->GetSubdirectories()[0];
+
+        for (const auto& file : section0->GetFiles()) {
+            if (file->GetExtension() != "cnmt")
+                continue;
+
+            meta.insert_or_assign(nca->GetTitleId(), CNMT(file));
+            meta_id.insert_or_assign(nca->GetTitleId(), id);
+            break;
+        }
+    }
+}
+
+void RegisteredCache::AccumulateYuzuMeta() {
+    const auto dir = this->dir->GetSubdirectory("yuzu_meta");
+    if (dir == nullptr)
+        return;
+
+    for (const auto& file : dir->GetFiles()) {
+        if (file->GetExtension() != "cnmt")
+            continue;
+
+        CNMT cnmt(file);
+        yuzu_meta.insert_or_assign(cnmt.GetTitleID(), std::move(cnmt));
+    }
+}
+
+void RegisteredCache::Refresh() {
+    if (dir == nullptr)
+        return;
+    const auto ids = AccumulateFiles();
+    ProcessFiles(ids);
+    AccumulateYuzuMeta();
+}
+
+RegisteredCache::RegisteredCache(VirtualDir dir_, RegisteredCacheParsingFunction parsing_function)
+    : dir(std::move(dir_)), parser(std::move(parsing_function)) {
+    Refresh();
+}
+
+bool RegisteredCache::HasEntry(u64 title_id, ContentRecordType type) const {
+    return GetEntryRaw(title_id, type) != nullptr;
+}
+
+bool RegisteredCache::HasEntry(RegisteredCacheEntry entry) const {
+    return GetEntryRaw(entry) != nullptr;
+}
+
+VirtualFile RegisteredCache::GetEntryRaw(u64 title_id, ContentRecordType type) const {
+    const auto id = GetNcaIDFromMetadata(title_id, type);
+    if (id == boost::none)
+        return nullptr;
+
+    return parser(GetFileAtID(id.get()), id.get());
+}
+
+VirtualFile RegisteredCache::GetEntryRaw(RegisteredCacheEntry entry) const {
+    return GetEntryRaw(entry.title_id, entry.type);
+}
+
+std::shared_ptr<NCA> RegisteredCache::GetEntry(u64 title_id, ContentRecordType type) const {
+    const auto raw = GetEntryRaw(title_id, type);
+    if (raw == nullptr)
+        return nullptr;
+    return std::make_shared<NCA>(raw);
+}
+
+std::shared_ptr<NCA> RegisteredCache::GetEntry(RegisteredCacheEntry entry) const {
+    return GetEntry(entry.title_id, entry.type);
+}
+
+template <typename T>
+void RegisteredCache::IterateAllMetadata(
+    std::vector<T>& out, std::function<T(const CNMT&, const ContentRecord&)> proc,
+    std::function<bool(const CNMT&, const ContentRecord&)> filter) const {
+    for (const auto& kv : meta) {
+        const auto& cnmt = kv.second;
+        if (filter(cnmt, EMPTY_META_CONTENT_RECORD))
+            out.push_back(proc(cnmt, EMPTY_META_CONTENT_RECORD));
+        for (const auto& rec : cnmt.GetContentRecords()) {
+            if (GetFileAtID(rec.nca_id) != nullptr && filter(cnmt, rec)) {
+                out.push_back(proc(cnmt, rec));
+            }
+        }
+    }
+    for (const auto& kv : yuzu_meta) {
+        const auto& cnmt = kv.second;
+        for (const auto& rec : cnmt.GetContentRecords()) {
+            if (GetFileAtID(rec.nca_id) != nullptr && filter(cnmt, rec)) {
+                out.push_back(proc(cnmt, rec));
+            }
+        }
+    }
+}
+
+std::vector<RegisteredCacheEntry> RegisteredCache::ListEntries() const {
+    std::vector<RegisteredCacheEntry> out;
+    IterateAllMetadata<RegisteredCacheEntry>(
+        out,
+        [](const CNMT& c, const ContentRecord& r) {
+            return RegisteredCacheEntry{c.GetTitleID(), r.type};
+        },
+        [](const CNMT& c, const ContentRecord& r) { return true; });
+    return out;
+}
+
+std::vector<RegisteredCacheEntry> RegisteredCache::ListEntriesFilter(
+    boost::optional<TitleType> title_type, boost::optional<ContentRecordType> record_type,
+    boost::optional<u64> title_id) const {
+    std::vector<RegisteredCacheEntry> out;
+    IterateAllMetadata<RegisteredCacheEntry>(
+        out,
+        [](const CNMT& c, const ContentRecord& r) {
+            return RegisteredCacheEntry{c.GetTitleID(), r.type};
+        },
+        [&title_type, &record_type, &title_id](const CNMT& c, const ContentRecord& r) {
+            if (title_type != boost::none && title_type.get() != c.GetType())
+                return false;
+            if (record_type != boost::none && record_type.get() != r.type)
+                return false;
+            if (title_id != boost::none && title_id.get() != c.GetTitleID())
+                return false;
+            return true;
+        });
+    return out;
+}
+
+static std::shared_ptr<NCA> GetNCAFromXCIForID(std::shared_ptr<XCI> xci, const NcaID& id) {
+    const auto filename = fmt::format("{}.nca", HexArrayToString(id, false));
+    const auto iter =
+        std::find_if(xci->GetNCAs().begin(), xci->GetNCAs().end(),
+                     [&filename](std::shared_ptr<NCA> nca) { return nca->GetName() == filename; });
+    return iter == xci->GetNCAs().end() ? nullptr : *iter;
+}
+
+InstallResult RegisteredCache::InstallEntry(std::shared_ptr<XCI> xci, bool overwrite_if_exists,
+                                            const VfsCopyFunction& copy) {
+    const auto& ncas = xci->GetNCAs();
+    const auto& meta_iter = std::find_if(ncas.begin(), ncas.end(), [](std::shared_ptr<NCA> nca) {
+        return nca->GetType() == NCAContentType::Meta;
+    });
+
+    if (meta_iter == ncas.end()) {
+        LOG_ERROR(Loader, "The XCI you are attempting to install does not have a metadata NCA and "
+                          "is therefore malformed. Double check your encryption keys.");
+        return InstallResult::ErrorMetaFailed;
+    }
+
+    // Install Metadata File
+    const auto meta_id_raw = (*meta_iter)->GetName().substr(0, 32);
+    const auto meta_id = HexStringToArray<16>(meta_id_raw);
+
+    const auto res = RawInstallNCA(*meta_iter, copy, overwrite_if_exists, meta_id);
+    if (res != InstallResult::Success)
+        return res;
+
+    // Install all the other NCAs
+    const auto section0 = (*meta_iter)->GetSubdirectories()[0];
+    const auto cnmt_file = section0->GetFiles()[0];
+    const CNMT cnmt(cnmt_file);
+    for (const auto& record : cnmt.GetContentRecords()) {
+        const auto nca = GetNCAFromXCIForID(xci, record.nca_id);
+        if (nca == nullptr)
+            return InstallResult::ErrorCopyFailed;
+        const auto res2 = RawInstallNCA(nca, copy, overwrite_if_exists, record.nca_id);
+        if (res2 != InstallResult::Success)
+            return res2;
+    }
+
+    Refresh();
+    return InstallResult::Success;
+}
+
+InstallResult RegisteredCache::InstallEntry(std::shared_ptr<NCA> nca, TitleType type,
+                                            bool overwrite_if_exists, const VfsCopyFunction& copy) {
+    CNMTHeader header{
+        nca->GetTitleId(), ///< Title ID
+        0,                 ///< Ignore/Default title version
+        type,              ///< Type
+        {},                ///< Padding
+        0x10,              ///< Default table offset
+        1,                 ///< 1 Content Entry
+        0,                 ///< No Meta Entries
+        {},                ///< Padding
+    };
+    OptionalHeader opt_header{0, 0};
+    ContentRecord c_rec{{}, {}, {}, GetCRTypeFromNCAType(nca->GetType()), {}};
+    const auto& data = nca->GetBaseFile()->ReadBytes(0x100000);
+    mbedtls_sha256(data.data(), data.size(), c_rec.hash.data(), 0);
+    memcpy(&c_rec.nca_id, &c_rec.hash, 16);
+    const CNMT new_cnmt(header, opt_header, {c_rec}, {});
+    if (!RawInstallYuzuMeta(new_cnmt))
+        return InstallResult::ErrorMetaFailed;
+    return RawInstallNCA(nca, copy, overwrite_if_exists, c_rec.nca_id);
+}
+
+InstallResult RegisteredCache::RawInstallNCA(std::shared_ptr<NCA> nca, const VfsCopyFunction& copy,
+                                             bool overwrite_if_exists,
+                                             boost::optional<NcaID> override_id) {
+    const auto in = nca->GetBaseFile();
+    Core::Crypto::SHA256Hash hash{};
+
+    // Calculate NcaID
+    // NOTE: Because computing the SHA256 of an entire NCA is quite expensive (especially if the
+    // game is massive), we're going to cheat and only hash the first MB of the NCA.
+    // Also, for XCIs the NcaID matters, so if the override id isn't none, use that.
+    NcaID id{};
+    if (override_id == boost::none) {
+        const auto& data = in->ReadBytes(0x100000);
+        mbedtls_sha256(data.data(), data.size(), hash.data(), 0);
+        memcpy(id.data(), hash.data(), 16);
+    } else {
+        id = override_id.get();
+    }
+
+    std::string path = GetRelativePathFromNcaID(id, false, true);
+
+    if (GetFileAtID(id) != nullptr && !overwrite_if_exists) {
+        LOG_WARNING(Loader, "Attempting to overwrite existing NCA. Skipping...");
+        return InstallResult::ErrorAlreadyExists;
+    }
+
+    if (GetFileAtID(id) != nullptr) {
+        LOG_WARNING(Loader, "Overwriting existing NCA...");
+        VirtualDir c_dir;
+        { c_dir = dir->GetFileRelative(path)->GetContainingDirectory(); }
+        c_dir->DeleteFile(FileUtil::GetFilename(path));
+    }
+
+    auto out = dir->CreateFileRelative(path);
+    if (out == nullptr)
+        return InstallResult::ErrorCopyFailed;
+    return copy(in, out) ? InstallResult::Success : InstallResult::ErrorCopyFailed;
+}
+
+bool RegisteredCache::RawInstallYuzuMeta(const CNMT& cnmt) {
+    // Reasoning behind this method can be found in the comment for InstallEntry, NCA overload.
+    const auto dir = this->dir->CreateDirectoryRelative("yuzu_meta");
+    const auto filename = GetCNMTName(cnmt.GetType(), cnmt.GetTitleID());
+    if (dir->GetFile(filename) == nullptr) {
+        auto out = dir->CreateFile(filename);
+        const auto buffer = cnmt.Serialize();
+        out->Resize(buffer.size());
+        out->WriteBytes(buffer);
+    } else {
+        auto out = dir->GetFile(filename);
+        CNMT old_cnmt(out);
+        // Returns true on change
+        if (old_cnmt.UnionRecords(cnmt)) {
+            out->Resize(0);
+            const auto buffer = old_cnmt.Serialize();
+            out->Resize(buffer.size());
+            out->WriteBytes(buffer);
+        }
+    }
+    Refresh();
+    return std::find_if(yuzu_meta.begin(), yuzu_meta.end(),
+                        [&cnmt](const std::pair<u64, CNMT>& kv) {
+                            return kv.second.GetType() == cnmt.GetType() &&
+                                   kv.second.GetTitleID() == cnmt.GetTitleID();
+                        }) != yuzu_meta.end();
+}
+} // namespace FileSys

src/core/file_sys/registered_cache.h

@@ -0,0 +1,124 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <functional>
+#include <map>
+#include <memory>
+#include <string>
+#include <vector>
+#include <boost/container/flat_map.hpp>
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "content_archive.h"
+#include "core/file_sys/nca_metadata.h"
+#include "core/file_sys/vfs.h"
+
+namespace FileSys {
+class XCI;
+class CNMT;
+
+using NcaID = std::array<u8, 0x10>;
+using RegisteredCacheParsingFunction = std::function<VirtualFile(const VirtualFile&, const NcaID&)>;
+using VfsCopyFunction = std::function<bool(VirtualFile, VirtualFile)>;
+
+enum class InstallResult {
+    Success,
+    ErrorAlreadyExists,
+    ErrorCopyFailed,
+    ErrorMetaFailed,
+};
+
+struct RegisteredCacheEntry {
+    u64 title_id;
+    ContentRecordType type;
+
+    std::string DebugInfo() const;
+};
+
+// boost flat_map requires operator< for O(log(n)) lookups.
+bool operator<(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs);
+
+/*
+ * A class that catalogues NCAs in the registered directory structure.
+ * Nintendo's registered format follows this structure:
+ *
+ * Root
+ *   | 000000XX <- XX is the ____ two digits of the NcaID
+ *       | <hash>.nca <- hash is the NcaID (first half of SHA256 over entire file) (folder)
+ *         | 00
+ *         | 01 <- Actual content split along 4GB boundaries. (optional)
+ *
+ * (This impl also supports substituting the nca dir for an nca file, as that's more convenient when
+ * 4GB splitting can be ignored.)
+ */
+class RegisteredCache {
+public:
+    // Parsing function defines the conversion from raw file to NCA. If there are other steps
+    // besides creating the NCA from the file (e.g. NAX0 on SD Card), that should go in a custom
+    // parsing function.
+    explicit RegisteredCache(VirtualDir dir,
+                             RegisteredCacheParsingFunction parsing_function =
+                                 [](const VirtualFile& file, const NcaID& id) { return file; });
+
+    void Refresh();
+
+    bool HasEntry(u64 title_id, ContentRecordType type) const;
+    bool HasEntry(RegisteredCacheEntry entry) const;
+
+    VirtualFile GetEntryRaw(u64 title_id, ContentRecordType type) const;
+    VirtualFile GetEntryRaw(RegisteredCacheEntry entry) const;
+
+    std::shared_ptr<NCA> GetEntry(u64 title_id, ContentRecordType type) const;
+    std::shared_ptr<NCA> GetEntry(RegisteredCacheEntry entry) const;
+
+    std::vector<RegisteredCacheEntry> ListEntries() const;
+    // If a parameter is not boost::none, it will be filtered for from all entries.
+    std::vector<RegisteredCacheEntry> ListEntriesFilter(
+        boost::optional<TitleType> title_type = boost::none,
+        boost::optional<ContentRecordType> record_type = boost::none,
+        boost::optional<u64> title_id = boost::none) const;
+
+    // Raw copies all the ncas from the xci to the csache. Does some quick checks to make sure there
+    // is a meta NCA and all of them are accessible.
+    InstallResult InstallEntry(std::shared_ptr<XCI> xci, bool overwrite_if_exists = false,
+                               const VfsCopyFunction& copy = &VfsRawCopy);
+
+    // Due to the fact that we must use Meta-type NCAs to determine the existance of files, this
+    // poses quite a challenge. Instead of creating a new meta NCA for this file, yuzu will create a
+    // dir inside the NAND called 'yuzu_meta' and store the raw CNMT there.
+    // TODO(DarkLordZach): Author real meta-type NCAs and install those.
+    InstallResult InstallEntry(std::shared_ptr<NCA> nca, TitleType type,
+                               bool overwrite_if_exists = false,
+                               const VfsCopyFunction& copy = &VfsRawCopy);
+
+private:
+    template <typename T>
+    void IterateAllMetadata(std::vector<T>& out,
+                            std::function<T(const CNMT&, const ContentRecord&)> proc,
+                            std::function<bool(const CNMT&, const ContentRecord&)> filter) const;
+    std::vector<NcaID> AccumulateFiles() const;
+    void ProcessFiles(const std::vector<NcaID>& ids);
+    void AccumulateYuzuMeta();
+    boost::optional<NcaID> GetNcaIDFromMetadata(u64 title_id, ContentRecordType type) const;
+    VirtualFile GetFileAtID(NcaID id) const;
+    VirtualFile OpenFileOrDirectoryConcat(const VirtualDir& dir, std::string_view path) const;
+    InstallResult RawInstallNCA(std::shared_ptr<NCA> nca, const VfsCopyFunction& copy,
+                                bool overwrite_if_exists,
+                                boost::optional<NcaID> override_id = boost::none);
+    bool RawInstallYuzuMeta(const CNMT& cnmt);
+
+    VirtualDir dir;
+    RegisteredCacheParsingFunction parser;
+    // maps tid -> NcaID of meta
+    boost::container::flat_map<u64, NcaID> meta_id;
+    // maps tid -> meta
+    boost::container::flat_map<u64, CNMT> meta;
+    // maps tid -> meta for CNMT in yuzu_meta
+    boost::container::flat_map<u64, CNMT> yuzu_meta;
+};
+
+} // namespace FileSys

src/core/file_sys/romfs.cpp

@@ -65,7 +65,7 @@ void ProcessFile(VirtualFile file, size_t file_offset, size_t data_offset, u32 t
         auto entry = GetEntry<FileEntry>(file, file_offset + this_file_offset);
 
         parent->AddFile(std::make_shared<OffsetVfsFile>(
-            file, entry.first.size, entry.first.offset + data_offset, entry.second, parent));
+            file, entry.first.size, entry.first.offset + data_offset, entry.second));
 
         if (entry.first.sibling == ROMFS_ENTRY_EMPTY)
             break;
@@ -79,7 +79,7 @@ void ProcessDirectory(VirtualFile file, size_t dir_offset, size_t file_offset, s
     while (true) {
         auto entry = GetEntry<DirectoryEntry>(file, dir_offset + this_dir_offset);
         auto current = std::make_shared<VectorVfsDirectory>(
-            std::vector<VirtualFile>{}, std::vector<VirtualDir>{}, parent, entry.second);
+            std::vector<VirtualFile>{}, std::vector<VirtualDir>{}, entry.second);
 
         if (entry.first.child_file != ROMFS_ENTRY_EMPTY) {
             ProcessFile(file, file_offset, data_offset, entry.first.child_file, current);
@@ -108,9 +108,9 @@ VirtualDir ExtractRomFS(VirtualFile file) {
     const u64 file_offset = header.file_meta.offset;
     const u64 dir_offset = header.directory_meta.offset + 4;
 
-    const auto root =
+    auto root =
         std::make_shared<VectorVfsDirectory>(std::vector<VirtualFile>{}, std::vector<VirtualDir>{},
-                                             file->GetContainingDirectory(), file->GetName());
+                                             file->GetName(), file->GetContainingDirectory());
 
     ProcessDirectory(file, dir_offset, file_offset, header.data_offset, 0, root);
 

src/core/file_sys/vfs.cpp

@@ -74,15 +74,15 @@ VirtualFile VfsFilesystem::CopyFile(std::string_view old_path_, std::string_view
     return new_file;
 }
 
-VirtualFile VfsFilesystem::MoveFile(std::string_view old_path, std::string_view new_path) {
-    const auto sanitized_old_path = FileUtil::SanitizePath(old_path);
-    const auto sanitized_new_path = FileUtil::SanitizePath(new_path);
+VirtualFile VfsFilesystem::MoveFile(std::string_view old_path_, std::string_view new_path_) {
+    const auto old_path = FileUtil::SanitizePath(old_path_);
+    const auto new_path = FileUtil::SanitizePath(new_path_);
 
     // Again, non-default impls are highly encouraged to provide a more optimized version of this.
-    auto out = CopyFile(sanitized_old_path, sanitized_new_path);
+    auto out = CopyFile(old_path_, new_path_);
     if (out == nullptr)
         return nullptr;
-    if (DeleteFile(sanitized_old_path))
+    if (DeleteFile(old_path))
         return out;
     return nullptr;
 }
@@ -137,15 +137,15 @@ VirtualDir VfsFilesystem::CopyDirectory(std::string_view old_path_, std::string_
     return new_dir;
 }
 
-VirtualDir VfsFilesystem::MoveDirectory(std::string_view old_path, std::string_view new_path) {
-    const auto sanitized_old_path = FileUtil::SanitizePath(old_path);
-    const auto sanitized_new_path = FileUtil::SanitizePath(new_path);
+VirtualDir VfsFilesystem::MoveDirectory(std::string_view old_path_, std::string_view new_path_) {
+    const auto old_path = FileUtil::SanitizePath(old_path_);
+    const auto new_path = FileUtil::SanitizePath(new_path_);
 
     // Non-default impls are highly encouraged to provide a more optimized version of this.
-    auto out = CopyDirectory(sanitized_old_path, sanitized_new_path);
+    auto out = CopyDirectory(old_path_, new_path_);
     if (out == nullptr)
         return nullptr;
-    if (DeleteDirectory(sanitized_old_path))
+    if (DeleteDirectory(old_path))
         return out;
     return nullptr;
 }

src/core/file_sys/vfs.h

@@ -15,9 +15,9 @@
 
 namespace FileSys {
 
-class VfsDirectory;
-class VfsFile;
-class VfsFilesystem;
+struct VfsFilesystem;
+struct VfsFile;
+struct VfsDirectory;
 
 // Convenience typedefs to use Vfs* interfaces
 using VirtualFilesystem = std::shared_ptr<VfsFilesystem>;
@@ -34,9 +34,8 @@ enum class VfsEntryType {
 // A class representing an abstract filesystem. A default implementation given the root VirtualDir
 // is provided for convenience, but if the Vfs implementation has any additional state or
 // functionality, they will need to override.
-class VfsFilesystem : NonCopyable {
-public:
-    explicit VfsFilesystem(VirtualDir root);
+struct VfsFilesystem : NonCopyable {
+    VfsFilesystem(VirtualDir root);
     virtual ~VfsFilesystem();
 
     // Gets the friendly name for the filesystem.
@@ -82,8 +81,7 @@ protected:
 };
 
 // A class representing a file in an abstract filesystem.
-class VfsFile : NonCopyable {
-public:
+struct VfsFile : NonCopyable {
     virtual ~VfsFile();
 
     // Retrieves the file name.
@@ -181,8 +179,7 @@ public:
 };
 
 // A class representing a directory in an abstract filesystem.
-class VfsDirectory : NonCopyable {
-public:
+struct VfsDirectory : NonCopyable {
     virtual ~VfsDirectory();
 
     // Retrives the file located at path as if the current directory was root. Returns nullptr if
@@ -298,8 +295,7 @@ protected:
 
 // A convenience partial-implementation of VfsDirectory that stubs out methods that should only work
 // if writable. This is to avoid redundant empty methods everywhere.
-class ReadOnlyVfsDirectory : public VfsDirectory {
-public:
+struct ReadOnlyVfsDirectory : public VfsDirectory {
     bool IsWritable() const override;
     bool IsReadable() const override;
     std::shared_ptr<VfsDirectory> CreateSubdirectory(std::string_view name) override;

src/core/file_sys/vfs_concat.cpp

@@ -0,0 +1,94 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <utility>
+
+#include "core/file_sys/vfs_concat.h"
+
+namespace FileSys {
+
+VirtualFile ConcatenateFiles(std::vector<VirtualFile> files, std::string name) {
+    if (files.empty())
+        return nullptr;
+    if (files.size() == 1)
+        return files[0];
+
+    return std::shared_ptr<VfsFile>(new ConcatenatedVfsFile(std::move(files), std::move(name)));
+}
+
+ConcatenatedVfsFile::ConcatenatedVfsFile(std::vector<VirtualFile> files_, std::string name)
+    : name(std::move(name)) {
+    size_t next_offset = 0;
+    for (const auto& file : files_) {
+        files[next_offset] = file;
+        next_offset += file->GetSize();
+    }
+}
+
+std::string ConcatenatedVfsFile::GetName() const {
+    if (files.empty())
+        return "";
+    if (!name.empty())
+        return name;
+    return files.begin()->second->GetName();
+}
+
+size_t ConcatenatedVfsFile::GetSize() const {
+    if (files.empty())
+        return 0;
+    return files.rbegin()->first + files.rbegin()->second->GetSize();
+}
+
+bool ConcatenatedVfsFile::Resize(size_t new_size) {
+    return false;
+}
+
+std::shared_ptr<VfsDirectory> ConcatenatedVfsFile::GetContainingDirectory() const {
+    if (files.empty())
+        return nullptr;
+    return files.begin()->second->GetContainingDirectory();
+}
+
+bool ConcatenatedVfsFile::IsWritable() const {
+    return false;
+}
+
+bool ConcatenatedVfsFile::IsReadable() const {
+    return true;
+}
+
+size_t ConcatenatedVfsFile::Read(u8* data, size_t length, size_t offset) const {
+    auto entry = files.end();
+    for (auto iter = files.begin(); iter != files.end(); ++iter) {
+        if (iter->first > offset) {
+            entry = --iter;
+            break;
+        }
+    }
+
+    // Check if the entry should be the last one. The loop above will make it end().
+    if (entry == files.end() && offset < files.rbegin()->first + files.rbegin()->second->GetSize())
+        --entry;
+
+    if (entry == files.end())
+        return 0;
+
+    const auto remaining = entry->second->GetSize() + offset - entry->first;
+    if (length > remaining) {
+        return entry->second->Read(data, remaining, offset - entry->first) +
+               Read(data + remaining, length - remaining, offset + remaining);
+    }
+
+    return entry->second->Read(data, length, offset - entry->first);
+}
+
+size_t ConcatenatedVfsFile::Write(const u8* data, size_t length, size_t offset) {
+    return 0;
+}
+
+bool ConcatenatedVfsFile::Rename(std::string_view name) {
+    return false;
+}
+} // namespace FileSys

src/core/file_sys/vfs_concat.h

@@ -0,0 +1,41 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <string_view>
+#include <boost/container/flat_map.hpp>
+#include "core/file_sys/vfs.h"
+
+namespace FileSys {
+
+// Wrapper function to allow for more efficient handling of files.size() == 0, 1 cases.
+VirtualFile ConcatenateFiles(std::vector<VirtualFile> files, std::string name = "");
+
+// Class that wraps multiple vfs files and concatenates them, making reads seamless. Currently
+// read-only.
+class ConcatenatedVfsFile : public VfsFile {
+    friend VirtualFile ConcatenateFiles(std::vector<VirtualFile> files, std::string name);
+
+    ConcatenatedVfsFile(std::vector<VirtualFile> files, std::string name);
+
+public:
+    std::string GetName() const override;
+    size_t GetSize() const override;
+    bool Resize(size_t new_size) override;
+    std::shared_ptr<VfsDirectory> GetContainingDirectory() const override;
+    bool IsWritable() const override;
+    bool IsReadable() const override;
+    size_t Read(u8* data, size_t length, size_t offset) const override;
+    size_t Write(const u8* data, size_t length, size_t offset) override;
+    bool Rename(std::string_view name) override;
+
+private:
+    // Maps starting offset to file -- more efficient.
+    boost::container::flat_map<u64, VirtualFile> files;
+    std::string name;
+};
+
+} // namespace FileSys

src/core/file_sys/vfs_offset.h

@@ -15,8 +15,7 @@ namespace FileSys {
 // Similar to seeking to an offset.
 // If the file is writable, operations that would write past the end of the offset file will expand
 // the size of this wrapper.
-class OffsetVfsFile : public VfsFile {
-public:
+struct OffsetVfsFile : public VfsFile {
     OffsetVfsFile(std::shared_ptr<VfsFile> file, size_t size, size_t offset = 0,
                   std::string new_name = "", VirtualDir new_parent = nullptr);
 

src/core/file_sys/vfs_real.cpp

@@ -83,8 +83,12 @@ VirtualFile RealVfsFilesystem::OpenFile(std::string_view path_, Mode perms) {
 
 VirtualFile RealVfsFilesystem::CreateFile(std::string_view path_, Mode perms) {
     const auto path = FileUtil::SanitizePath(path_, FileUtil::DirectorySeparator::PlatformDefault);
-    if (!FileUtil::Exists(path) && !FileUtil::CreateEmptyFile(path))
-        return nullptr;
+    const auto path_fwd = FileUtil::SanitizePath(path, FileUtil::DirectorySeparator::ForwardSlash);
+    if (!FileUtil::Exists(path)) {
+        FileUtil::CreateFullPath(path_fwd);
+        if (!FileUtil::CreateEmptyFile(path))
+            return nullptr;
+    }
     return OpenFile(path, perms);
 }
 
@@ -140,8 +144,12 @@ VirtualDir RealVfsFilesystem::OpenDirectory(std::string_view path_, Mode perms)
 
 VirtualDir RealVfsFilesystem::CreateDirectory(std::string_view path_, Mode perms) {
     const auto path = FileUtil::SanitizePath(path_, FileUtil::DirectorySeparator::PlatformDefault);
-    if (!FileUtil::Exists(path) && !FileUtil::CreateDir(path))
-        return nullptr;
+    const auto path_fwd = FileUtil::SanitizePath(path, FileUtil::DirectorySeparator::ForwardSlash);
+    if (!FileUtil::Exists(path)) {
+        FileUtil::CreateFullPath(path_fwd);
+        if (!FileUtil::CreateDir(path))
+            return nullptr;
+    }
     // Cannot use make_shared as RealVfsDirectory constructor is private
     return std::shared_ptr<RealVfsDirectory>(new RealVfsDirectory(*this, path, perms));
 }
@@ -306,14 +314,14 @@ RealVfsDirectory::RealVfsDirectory(RealVfsFilesystem& base_, const std::string&
 
 std::shared_ptr<VfsFile> RealVfsDirectory::GetFileRelative(std::string_view path) const {
     const auto full_path = FileUtil::SanitizePath(this->path + DIR_SEP + std::string(path));
-    if (!FileUtil::Exists(full_path))
+    if (!FileUtil::Exists(full_path) || FileUtil::IsDirectory(full_path))
         return nullptr;
     return base.OpenFile(full_path, perms);
 }
 
 std::shared_ptr<VfsDirectory> RealVfsDirectory::GetDirectoryRelative(std::string_view path) const {
     const auto full_path = FileUtil::SanitizePath(this->path + DIR_SEP + std::string(path));
-    if (!FileUtil::Exists(full_path))
+    if (!FileUtil::Exists(full_path) || !FileUtil::IsDirectory(full_path))
         return nullptr;
     return base.OpenDirectory(full_path, perms);
 }

src/core/file_sys/vfs_real.h

@@ -5,7 +5,6 @@
 #pragma once
 
 #include <string_view>
-
 #include <boost/container/flat_map.hpp>
 #include "common/file_util.h"
 #include "core/file_sys/mode.h"

src/core/file_sys/vfs_vector.cpp

@@ -8,8 +8,8 @@
 
 namespace FileSys {
 VectorVfsDirectory::VectorVfsDirectory(std::vector<VirtualFile> files_,
-                                       std::vector<VirtualDir> dirs_, VirtualDir parent_,
-                                       std::string name_)
+                                       std::vector<VirtualDir> dirs_, std::string name_,
+                                       VirtualDir parent_)
     : files(std::move(files_)), dirs(std::move(dirs_)), parent(std::move(parent_)),
       name(std::move(name_)) {}
 

src/core/file_sys/vfs_vector.h

@@ -10,11 +10,10 @@ namespace FileSys {
 
 // An implementation of VfsDirectory that maintains two vectors for subdirectories and files.
 // Vector data is supplied upon construction.
-class VectorVfsDirectory : public VfsDirectory {
-public:
+struct VectorVfsDirectory : public VfsDirectory {
     explicit VectorVfsDirectory(std::vector<VirtualFile> files = {},
-                                std::vector<VirtualDir> dirs = {}, VirtualDir parent = nullptr,
-                                std::string name = "");
+                                std::vector<VirtualDir> dirs = {}, std::string name = "",
+                                VirtualDir parent = nullptr);
 
     std::vector<std::shared_ptr<VfsFile>> GetFiles() const override;
     std::vector<std::shared_ptr<VfsDirectory>> GetSubdirectories() const override;

src/core/hle/kernel/kernel.cpp

@@ -11,7 +11,7 @@
 
 namespace Kernel {
 
-std::atomic<u32> Object::next_object_id{0};
+unsigned int Object::next_object_id;
 
 /// Initialize the kernel
 void Init() {

src/core/hle/kernel/object.h

@@ -4,7 +4,6 @@
 
 #pragma once
 
-#include <atomic>
 #include <string>
 #include <utility>
 
@@ -43,8 +42,8 @@ public:
     virtual ~Object();
 
     /// Returns a unique identifier for the object. For debugging purposes only.
-    u32 GetObjectId() const {
-        return object_id.load(std::memory_order_relaxed);
+    unsigned int GetObjectId() const {
+        return object_id;
     }
 
     virtual std::string GetTypeName() const {
@@ -62,23 +61,23 @@ public:
     bool IsWaitable() const;
 
 public:
-    static std::atomic<u32> next_object_id;
+    static unsigned int next_object_id;
 
 private:
     friend void intrusive_ptr_add_ref(Object*);
     friend void intrusive_ptr_release(Object*);
 
-    std::atomic<u32> ref_count{0};
-    std::atomic<u32> object_id{next_object_id++};
+    unsigned int ref_count = 0;
+    unsigned int object_id = next_object_id++;
 };
 
 // Special functions used by boost::instrusive_ptr to do automatic ref-counting
 inline void intrusive_ptr_add_ref(Object* object) {
-    object->ref_count.fetch_add(1, std::memory_order_relaxed);
+    ++object->ref_count;
 }
 
 inline void intrusive_ptr_release(Object* object) {
-    if (object->ref_count.fetch_sub(1, std::memory_order_acq_rel) == 1) {
+    if (--object->ref_count == 0) {
         delete object;
     }
 }

src/core/hle/kernel/scheduler.cpp

@@ -25,7 +25,7 @@ Scheduler::~Scheduler() {
     }
 }
 
-bool Scheduler::HaveReadyThreads() const {
+bool Scheduler::HaveReadyThreads() {
     std::lock_guard<std::mutex> lock(scheduler_mutex);
     return ready_queue.get_first() != nullptr;
 }

src/core/hle/kernel/scheduler.h

@@ -21,7 +21,7 @@ public:
     ~Scheduler();
 
     /// Returns whether there are any threads that are ready to run.
-    bool HaveReadyThreads() const;
+    bool HaveReadyThreads();
 
     /// Reschedules to the next available thread (call after current thread is suspended)
     void Reschedule();

src/core/hle/kernel/svc.cpp

@@ -532,6 +532,7 @@ static ResultCode CreateThread(Handle* out_handle, VAddr entry_point, u64 arg, V
     CASCADE_RESULT(thread->guest_handle, g_handle_table.Create(thread));
     *out_handle = thread->guest_handle;
 
+    Core::System::GetInstance().PrepareReschedule();
     Core::System::GetInstance().CpuCore(thread->processor_id).PrepareReschedule();
 
     LOG_TRACE(Kernel_SVC,
@@ -705,7 +706,8 @@ static ResultCode SignalProcessWideKey(VAddr condition_variable_addr, s32 target
             Handle owner_handle = static_cast<Handle>(mutex_val & Mutex::MutexOwnerMask);
             auto owner = g_handle_table.Get<Thread>(owner_handle);
             ASSERT(owner);
-            ASSERT(thread->status == ThreadStatus::WaitMutex);
+            ASSERT(thread->status != ThreadStatus::Running);
+            thread->status = ThreadStatus::WaitMutex;
             thread->wakeup_callback = nullptr;
 
             owner->AddMutexWaiter(thread);

src/core/hle/kernel/thread.cpp

@@ -23,7 +23,6 @@
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/thread.h"
-#include "core/hle/lock.h"
 #include "core/hle/result.h"
 #include "core/memory.h"
 
@@ -105,10 +104,6 @@ void ExitCurrentThread() {
  */
 static void ThreadWakeupCallback(u64 thread_handle, int cycles_late) {
     const auto proper_handle = static_cast<Handle>(thread_handle);
-
-    // Lock the global kernel mutex when we enter the kernel HLE.
-    std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock);
-
     SharedPtr<Thread> thread = wakeup_callback_handle_table.Get<Thread>(proper_handle);
     if (thread == nullptr) {
         LOG_CRITICAL(Kernel, "Callback fired for invalid thread {:08X}", proper_handle);
@@ -160,10 +155,8 @@ void Thread::WakeAfterDelay(s64 nanoseconds) {
     if (nanoseconds == -1)
         return;
 
-    // This function might be called from any thread so we have to be cautious and use the
-    // thread-safe version of ScheduleEvent.
-    CoreTiming::ScheduleEventThreadsafe(CoreTiming::nsToCycles(nanoseconds), ThreadWakeupEventType,
-                                        callback_handle);
+    CoreTiming::ScheduleEvent(CoreTiming::nsToCycles(nanoseconds), ThreadWakeupEventType,
+                              callback_handle);
 }
 
 void Thread::CancelWakeupTimer() {
@@ -426,33 +419,12 @@ VAddr Thread::GetCommandBufferAddress() const {
 }
 
 void Thread::AddMutexWaiter(SharedPtr<Thread> thread) {
-    if (thread->lock_owner == this) {
-        // If the thread is already waiting for this thread to release the mutex, ensure that the
-        // waiters list is consistent and return without doing anything.
-        auto itr = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
-        ASSERT(itr != wait_mutex_threads.end());
-        return;
-    }
-
-    // A thread can't wait on two different mutexes at the same time.
-    ASSERT(thread->lock_owner == nullptr);
-
-    // Ensure that the thread is not already in the list of mutex waiters
-    auto itr = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
-    ASSERT(itr == wait_mutex_threads.end());
-
     thread->lock_owner = this;
     wait_mutex_threads.emplace_back(std::move(thread));
     UpdatePriority();
 }
 
 void Thread::RemoveMutexWaiter(SharedPtr<Thread> thread) {
-    ASSERT(thread->lock_owner == this);
-
-    // Ensure that the thread is in the list of mutex waiters
-    auto itr = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
-    ASSERT(itr != wait_mutex_threads.end());
-
     boost::remove_erase(wait_mutex_threads, thread);
     thread->lock_owner = nullptr;
     UpdatePriority();

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

@@ -20,9 +20,9 @@ public:
     explicit IAudioRenderer(AudioCore::AudioRendererParameter audren_params)
         : ServiceFramework("IAudioRenderer") {
         static const FunctionInfo functions[] = {
-            {0, &IAudioRenderer::GetAudioRendererSampleRate, "GetAudioRendererSampleRate"},
-            {1, &IAudioRenderer::GetAudioRendererSampleCount, "GetAudioRendererSampleCount"},
-            {2, &IAudioRenderer::GetAudioRendererMixBufferCount, "GetAudioRendererMixBufferCount"},
+            {0, nullptr, "GetAudioRendererSampleRate"},
+            {1, nullptr, "GetAudioRendererSampleCount"},
+            {2, nullptr, "GetAudioRendererMixBufferCount"},
             {3, nullptr, "GetAudioRendererState"},
             {4, &IAudioRenderer::RequestUpdateAudioRenderer, "RequestUpdateAudioRenderer"},
             {5, &IAudioRenderer::StartAudioRenderer, "StartAudioRenderer"},
@@ -45,27 +45,6 @@ private:
         system_event->Signal();
     }
 
-    void GetAudioRendererSampleRate(Kernel::HLERequestContext& ctx) {
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(RESULT_SUCCESS);
-        rb.Push<u32>(renderer->GetSampleRate());
-        LOG_DEBUG(Service_Audio, "called");
-    }
-
-    void GetAudioRendererSampleCount(Kernel::HLERequestContext& ctx) {
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(RESULT_SUCCESS);
-        rb.Push<u32>(renderer->GetSampleCount());
-        LOG_DEBUG(Service_Audio, "called");
-    }
-
-    void GetAudioRendererMixBufferCount(Kernel::HLERequestContext& ctx) {
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(RESULT_SUCCESS);
-        rb.Push<u32>(renderer->GetMixBufferCount());
-        LOG_DEBUG(Service_Audio, "called");
-    }
-
     void RequestUpdateAudioRenderer(Kernel::HLERequestContext& ctx) {
         ctx.WriteBuffer(renderer->UpdateAudioRenderer(ctx.ReadBuffer()));
         IPC::ResponseBuilder rb{ctx, 2};
@@ -190,8 +169,7 @@ AudRenU::AudRenU() : ServiceFramework("audren:u") {
         {1, &AudRenU::GetAudioRendererWorkBufferSize, "GetAudioRendererWorkBufferSize"},
         {2, &AudRenU::GetAudioDevice, "GetAudioDevice"},
         {3, nullptr, "OpenAudioRendererAuto"},
-        {4, &AudRenU::GetAudioDeviceServiceWithRevisionInfo,
-         "GetAudioDeviceServiceWithRevisionInfo"},
+        {4, nullptr, "GetAudioDeviceServiceWithRevisionInfo"},
     };
     RegisterHandlers(functions);
 }
@@ -211,7 +189,7 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     auto params = rp.PopRaw<AudioCore::AudioRendererParameter>();
 
-    u64 buffer_sz = Common::AlignUp(4 * params.mix_buffer_count, 0x40);
+    u64 buffer_sz = Common::AlignUp(4 * params.unknown_8, 0x40);
     buffer_sz += params.unknown_c * 1024;
     buffer_sz += 0x940 * (params.unknown_c + 1);
     buffer_sz += 0x3F0 * params.voice_count;
@@ -219,7 +197,7 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
     buffer_sz += Common::AlignUp(8 * params.voice_count, 0x10);
     buffer_sz +=
         Common::AlignUp((0x3C0 * (params.sink_count + params.unknown_c) + 4 * params.sample_count) *
-                            (params.mix_buffer_count + 6),
+                            (params.unknown_8 + 6),
                         0x40);
 
     if (IsFeatureSupported(AudioFeatures::Splitter, params.revision)) {
@@ -275,16 +253,6 @@ void AudRenU::GetAudioDevice(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Audio, "called");
 }
 
-void AudRenU::GetAudioDeviceServiceWithRevisionInfo(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<Audio::IAudioDevice>();
-
-    LOG_WARNING(Service_Audio, "(STUBBED) called"); // TODO(ogniK): Figure out what is different
-                                                    // based on the current revision
-}
-
 bool AudRenU::IsFeatureSupported(AudioFeatures feature, u32_le revision) const {
     u32_be version_num = (revision - Common::MakeMagic('R', 'E', 'V', '0')); // Byte swap
     switch (feature) {

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

@@ -22,7 +22,6 @@ private:
     void OpenAudioRenderer(Kernel::HLERequestContext& ctx);
     void GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx);
     void GetAudioDevice(Kernel::HLERequestContext& ctx);
-    void GetAudioDeviceServiceWithRevisionInfo(Kernel::HLERequestContext& ctx);
 
     enum class AudioFeatures : u32 {
         Splitter,

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

@@ -226,6 +226,7 @@ ResultVal<FileSys::EntryType> VfsDirectoryServiceWrapper::GetEntryType(
 static std::unique_ptr<FileSys::RomFSFactory> romfs_factory;
 static std::unique_ptr<FileSys::SaveDataFactory> save_data_factory;
 static std::unique_ptr<FileSys::SDMCFactory> sdmc_factory;
+static std::unique_ptr<FileSys::BISFactory> bis_factory;
 
 ResultCode RegisterRomFS(std::unique_ptr<FileSys::RomFSFactory>&& factory) {
     ASSERT_MSG(romfs_factory == nullptr, "Tried to register a second RomFS");
@@ -248,6 +249,13 @@ ResultCode RegisterSDMC(std::unique_ptr<FileSys::SDMCFactory>&& factory) {
     return RESULT_SUCCESS;
 }
 
+ResultCode RegisterBIS(std::unique_ptr<FileSys::BISFactory>&& factory) {
+    ASSERT_MSG(bis_factory == nullptr, "Tried to register a second BIS");
+    bis_factory = std::move(factory);
+    LOG_DEBUG(Service_FS, "Registred BIS");
+    return RESULT_SUCCESS;
+}
+
 ResultVal<FileSys::VirtualFile> OpenRomFS(u64 title_id) {
     LOG_TRACE(Service_FS, "Opening RomFS for title_id={:016X}", title_id);
 
@@ -281,6 +289,14 @@ ResultVal<FileSys::VirtualDir> OpenSDMC() {
     return sdmc_factory->Open();
 }
 
+std::shared_ptr<FileSys::RegisteredCache> GetSystemNANDContents() {
+    return bis_factory->GetSystemNANDContents();
+}
+
+std::shared_ptr<FileSys::RegisteredCache> GetUserNANDContents() {
+    return bis_factory->GetUserNANDContents();
+}
+
 void RegisterFileSystems(const FileSys::VirtualFilesystem& vfs) {
     romfs_factory = nullptr;
     save_data_factory = nullptr;
@@ -291,6 +307,9 @@ void RegisterFileSystems(const FileSys::VirtualFilesystem& vfs) {
     auto sd_directory = vfs->OpenDirectory(FileUtil::GetUserPath(FileUtil::UserPath::SDMCDir),
                                            FileSys::Mode::ReadWrite);
 
+    if (bis_factory == nullptr)
+        bis_factory = std::make_unique<FileSys::BISFactory>(nand_directory);
+
     auto savedata = std::make_unique<FileSys::SaveDataFactory>(std::move(nand_directory));
     save_data_factory = std::move(savedata);
 

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

@@ -6,6 +6,7 @@
 
 #include <memory>
 #include "common/common_types.h"
+#include "core/file_sys/bis_factory.h"
 #include "core/file_sys/directory.h"
 #include "core/file_sys/mode.h"
 #include "core/file_sys/romfs_factory.h"
@@ -24,16 +25,15 @@ namespace FileSystem {
 ResultCode RegisterRomFS(std::unique_ptr<FileSys::RomFSFactory>&& factory);
 ResultCode RegisterSaveData(std::unique_ptr<FileSys::SaveDataFactory>&& factory);
 ResultCode RegisterSDMC(std::unique_ptr<FileSys::SDMCFactory>&& factory);
+ResultCode RegisterBIS(std::unique_ptr<FileSys::BISFactory>&& factory);
 
-// TODO(DarkLordZach): BIS Filesystem
-// ResultCode RegisterBIS(std::unique_ptr<FileSys::BISFactory>&& factory);
 ResultVal<FileSys::VirtualFile> OpenRomFS(u64 title_id);
 ResultVal<FileSys::VirtualDir> OpenSaveData(FileSys::SaveDataSpaceId space,
                                             FileSys::SaveDataDescriptor save_struct);
 ResultVal<FileSys::VirtualDir> OpenSDMC();
 
-// TODO(DarkLordZach): BIS Filesystem
-// ResultVal<std::unique_ptr<FileSys::FileSystemBackend>> OpenBIS();
+std::shared_ptr<FileSys::RegisteredCache> GetSystemNANDContents();
+std::shared_ptr<FileSys::RegisteredCache> GetUserNANDContents();
 
 /// Registers all Filesystem services with the specified service manager.
 void InstallInterfaces(SM::ServiceManager& service_manager, const FileSys::VirtualFilesystem& vfs);

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

@@ -26,7 +26,7 @@ public:
             {10600, nullptr, "DeclareOpenOnlinePlaySession"},
             {10601, &IFriendService::DeclareCloseOnlinePlaySession,
              "DeclareCloseOnlinePlaySession"},
-            {10610, &IFriendService::UpdateUserPresence, "UpdateUserPresence"},
+            {10610, nullptr, "UpdateUserPresence"},
             {10700, nullptr, "GetPlayHistoryRegistrationKey"},
             {10701, nullptr, "GetPlayHistoryRegistrationKeyWithNetworkServiceAccountId"},
             {10702, nullptr, "AddPlayHistory"},
@@ -99,13 +99,6 @@ private:
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
     }
-
-    void UpdateUserPresence(Kernel::HLERequestContext& ctx) {
-        // Stub used by Retro City Rampage
-        LOG_WARNING(Service_ACC, "(STUBBED) called");
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(RESULT_SUCCESS);
-    }
 };
 
 void Module::Interface::CreateFriendService(Kernel::HLERequestContext& ctx) {

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

@@ -291,7 +291,6 @@ private:
 class Hid final : public ServiceFramework<Hid> {
 public:
     Hid() : ServiceFramework("hid") {
-        // clang-format off
         static const FunctionInfo functions[] = {
             {0, &Hid::CreateAppletResource, "CreateAppletResource"},
             {1, &Hid::ActivateDebugPad, "ActivateDebugPad"},
@@ -334,13 +333,15 @@ public:
             {102, &Hid::SetSupportedNpadIdType, "SetSupportedNpadIdType"},
             {103, &Hid::ActivateNpad, "ActivateNpad"},
             {104, nullptr, "DeactivateNpad"},
-            {106, &Hid::AcquireNpadStyleSetUpdateEventHandle, "AcquireNpadStyleSetUpdateEventHandle"},
-            {107, &Hid::DisconnectNpad, "DisconnectNpad"},
+            {106, &Hid::AcquireNpadStyleSetUpdateEventHandle,
+             "AcquireNpadStyleSetUpdateEventHandle"},
+            {107, nullptr, "DisconnectNpad"},
             {108, &Hid::GetPlayerLedPattern, "GetPlayerLedPattern"},
             {109, nullptr, "ActivateNpadWithRevision"},
             {120, &Hid::SetNpadJoyHoldType, "SetNpadJoyHoldType"},
             {121, &Hid::GetNpadJoyHoldType, "GetNpadJoyHoldType"},
-            {122, &Hid::SetNpadJoyAssignmentModeSingleByDefault, "SetNpadJoyAssignmentModeSingleByDefault"},
+            {122, &Hid::SetNpadJoyAssignmentModeSingleByDefault,
+             "SetNpadJoyAssignmentModeSingleByDefault"},
             {123, nullptr, "SetNpadJoyAssignmentModeSingleByDefault"},
             {124, &Hid::SetNpadJoyAssignmentModeDual, "SetNpadJoyAssignmentModeDual"},
             {125, &Hid::MergeSingleJoyAsDualJoy, "MergeSingleJoyAsDualJoy"},
@@ -397,8 +398,6 @@ public:
             {1000, nullptr, "SetNpadCommunicationMode"},
             {1001, nullptr, "GetNpadCommunicationMode"},
         };
-        // clang-format on
-
         RegisterHandlers(functions);
 
         event = Kernel::Event::Create(Kernel::ResetType::OneShot, "hid:EventHandle");
@@ -497,12 +496,6 @@ private:
         LOG_WARNING(Service_HID, "(STUBBED) called");
     }
 
-    void DisconnectNpad(Kernel::HLERequestContext& ctx) {
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(RESULT_SUCCESS);
-        LOG_WARNING(Service_HID, "(STUBBED) called");
-    }
-
     void GetPlayerLedPattern(Kernel::HLERequestContext& ctx) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);

src/core/hle/service/service.h

@@ -23,7 +23,7 @@ class HLERequestContext;
 } // namespace Kernel
 
 namespace FileSys {
-class VfsFilesystem;
+struct VfsFilesystem;
 }
 
 namespace Service {

src/core/loader/loader.cpp

@@ -46,6 +46,8 @@ FileType IdentifyFile(FileSys::VirtualFile file) {
 FileType GuessFromFilename(const std::string& name) {
     if (name == "main")
         return FileType::DeconstructedRomDirectory;
+    if (name == "00")
+        return FileType::NCA;
 
     const std::string extension =
         Common::ToLower(std::string(FileUtil::GetExtensionFromFilename(name)));
@@ -126,7 +128,7 @@ constexpr std::array<const char*, 36> RESULT_MESSAGES{
 };
 
 std::string GetMessageForResultStatus(ResultStatus status) {
-    return GetMessageForResultStatus(static_cast<u16>(status));
+    return GetMessageForResultStatus(static_cast<size_t>(status));
 }
 
 std::string GetMessageForResultStatus(u16 status) {

src/core/loader/loader.h

@@ -56,7 +56,7 @@ FileType GuessFromFilename(const std::string& name);
 std::string GetFileTypeString(FileType type);
 
 /// Return type for functions in Loader namespace
-enum class ResultStatus : u16 {
+enum class ResultStatus {
     Success,
     ErrorAlreadyLoaded,
     ErrorNotImplemented,

src/video_core/engines/shader_bytecode.h

@@ -200,14 +200,6 @@ enum class IMinMaxExchange : u64 {
     XHi = 3,
 };
 
-enum class XmadMode : u64 {
-    None = 0,
-    CLo = 1,
-    CHi = 2,
-    CSfu = 3,
-    CBcc = 4,
-};
-
 enum class FlowCondition : u64 {
     Always = 0xF,
     Fcsm_Tr = 0x1C, // TODO(bunnei): What is this used for?
@@ -464,18 +456,6 @@ union Instruction {
         }
     } bra;
 
-    union {
-        BitField<20, 16, u64> imm20_16;
-        BitField<36, 1, u64> product_shift_left;
-        BitField<37, 1, u64> merge_37;
-        BitField<48, 1, u64> sign_a;
-        BitField<49, 1, u64> sign_b;
-        BitField<50, 3, XmadMode> mode;
-        BitField<52, 1, u64> high_b;
-        BitField<53, 1, u64> high_a;
-        BitField<56, 1, u64> merge_56;
-    } xmad;
-
     union {
         BitField<20, 14, u64> offset;
         BitField<34, 5, u64> index;
@@ -613,7 +593,6 @@ public:
         IntegerSetPredicate,
         PredicateSetPredicate,
         Conversion,
-        Xmad,
         Unknown,
     };
 
@@ -803,10 +782,10 @@ private:
             INST("010010110101----", Id::ISET_C, Type::IntegerSet, "ISET_C"),
             INST("0011011-0101----", Id::ISET_IMM, Type::IntegerSet, "ISET_IMM"),
             INST("0101000010010---", Id::PSETP, Type::PredicateSetPredicate, "PSETP"),
-            INST("0011011-00------", Id::XMAD_IMM, Type::Xmad, "XMAD_IMM"),
-            INST("0100111---------", Id::XMAD_CR, Type::Xmad, "XMAD_CR"),
-            INST("010100010-------", Id::XMAD_RC, Type::Xmad, "XMAD_RC"),
-            INST("0101101100------", Id::XMAD_RR, Type::Xmad, "XMAD_RR"),
+            INST("0011011-00------", Id::XMAD_IMM, Type::Arithmetic, "XMAD_IMM"),
+            INST("0100111---------", Id::XMAD_CR, Type::Arithmetic, "XMAD_CR"),
+            INST("010100010-------", Id::XMAD_RC, Type::Arithmetic, "XMAD_RC"),
+            INST("0101101100------", Id::XMAD_RR, Type::Arithmetic, "XMAD_RR"),
         };
 #undef INST
         std::stable_sort(table.begin(), table.end(), [](const auto& a, const auto& b) {

src/video_core/gpu.cpp

@@ -46,7 +46,6 @@ u32 RenderTargetBytesPerPixel(RenderTargetFormat format) {
     case RenderTargetFormat::RGBA32_FLOAT:
     case RenderTargetFormat::RGBA32_UINT:
         return 16;
-    case RenderTargetFormat::RGBA16_UINT:
     case RenderTargetFormat::RGBA16_FLOAT:
     case RenderTargetFormat::RG32_FLOAT:
         return 8;
@@ -68,7 +67,6 @@ u32 RenderTargetBytesPerPixel(RenderTargetFormat format) {
     case RenderTargetFormat::R16_UINT:
     case RenderTargetFormat::R16_SINT:
     case RenderTargetFormat::R16_FLOAT:
-    case RenderTargetFormat::RG8_UNORM:
     case RenderTargetFormat::RG8_SNORM:
         return 2;
     case RenderTargetFormat::R8_UNORM:

src/video_core/gpu.h

@@ -20,7 +20,6 @@ enum class RenderTargetFormat : u32 {
     NONE = 0x0,
     RGBA32_FLOAT = 0xC0,
     RGBA32_UINT = 0xC2,
-    RGBA16_UINT = 0xC9,
     RGBA16_FLOAT = 0xCA,
     RG32_FLOAT = 0xCB,
     BGRA8_UNORM = 0xCF,
@@ -36,7 +35,6 @@ enum class RenderTargetFormat : u32 {
     R11G11B10_FLOAT = 0xE0,
     R32_FLOAT = 0xE5,
     B5G6R5_UNORM = 0xE8,
-    RG8_UNORM = 0xEA,
     RG8_SNORM = 0xEB,
     R16_UNORM = 0xEE,
     R16_SNORM = 0xEF,

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -36,21 +36,30 @@ MICROPROFILE_DEFINE(OpenGL_Drawing, "OpenGL", "Drawing", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255));
 MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100));
 
-RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window)
-    : emu_window{window}, stream_buffer(GL_ARRAY_BUFFER, STREAM_BUFFER_SIZE) {
+RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window) : emu_window{window} {
     // Create sampler objects
     for (size_t i = 0; i < texture_samplers.size(); ++i) {
         texture_samplers[i].Create();
         state.texture_units[i].sampler = texture_samplers[i].sampler.handle;
     }
 
+    // Create SSBOs
+    for (size_t stage = 0; stage < ssbos.size(); ++stage) {
+        for (size_t buffer = 0; buffer < ssbos[stage].size(); ++buffer) {
+            ssbos[stage][buffer].Create();
+            state.draw.const_buffers[stage][buffer].ssbo = ssbos[stage][buffer].handle;
+        }
+    }
+
     GLint ext_num;
     glGetIntegerv(GL_NUM_EXTENSIONS, &ext_num);
     for (GLint i = 0; i < ext_num; i++) {
         const std::string_view extension{
             reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, i))};
 
-        if (extension == "GL_ARB_direct_state_access") {
+        if (extension == "GL_ARB_buffer_storage") {
+            has_ARB_buffer_storage = true;
+        } else if (extension == "GL_ARB_direct_state_access") {
             has_ARB_direct_state_access = true;
         } else if (extension == "GL_ARB_separate_shader_objects") {
             has_ARB_separate_shader_objects = true;
@@ -77,31 +86,47 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window)
 
     hw_vao.Create();
 
-    state.draw.vertex_buffer = stream_buffer.GetHandle();
+    stream_buffer = OGLStreamBuffer::MakeBuffer(has_ARB_buffer_storage, GL_ARRAY_BUFFER);
+    stream_buffer->Create(STREAM_BUFFER_SIZE, STREAM_BUFFER_SIZE / 2);
+    state.draw.vertex_buffer = stream_buffer->GetHandle();
 
     shader_program_manager = std::make_unique<GLShader::ProgramManager>();
     state.draw.shader_program = 0;
     state.draw.vertex_array = hw_vao.handle;
     state.Apply();
 
-    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, stream_buffer.GetHandle());
+    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, stream_buffer->GetHandle());
+
+    for (unsigned index = 0; index < uniform_buffers.size(); ++index) {
+        auto& buffer = uniform_buffers[index];
+        buffer.Create();
+        glBindBuffer(GL_UNIFORM_BUFFER, buffer.handle);
+        glBufferData(GL_UNIFORM_BUFFER, sizeof(GLShader::MaxwellUniformData), nullptr,
+                     GL_STREAM_COPY);
+        glBindBufferBase(GL_UNIFORM_BUFFER, index, buffer.handle);
+    }
 
     glEnable(GL_BLEND);
 
-    glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_alignment);
-
     LOG_CRITICAL(Render_OpenGL, "Sync fixed function OpenGL state here!");
 }
 
-RasterizerOpenGL::~RasterizerOpenGL() {}
+RasterizerOpenGL::~RasterizerOpenGL() {
+    if (stream_buffer != nullptr) {
+        state.draw.vertex_buffer = stream_buffer->GetHandle();
+        state.Apply();
+        stream_buffer->Release();
+    }
+}
 
 std::pair<u8*, GLintptr> RasterizerOpenGL::SetupVertexArrays(u8* array_ptr,
                                                              GLintptr buffer_offset) {
     MICROPROFILE_SCOPE(OpenGL_VAO);
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& memory_manager = Core::System::GetInstance().GPU().memory_manager;
 
     state.draw.vertex_array = hw_vao.handle;
-    state.draw.vertex_buffer = stream_buffer.GetHandle();
+    state.draw.vertex_buffer = stream_buffer->GetHandle();
     state.Apply();
 
     // Upload all guest vertex arrays sequentially to our buffer
@@ -116,15 +141,16 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupVertexArrays(u8* array_ptr,
         ASSERT(end > start);
         u64 size = end - start + 1;
 
-        GLintptr vertex_buffer_offset;
-        std::tie(array_ptr, buffer_offset, vertex_buffer_offset) =
-            UploadMemory(array_ptr, buffer_offset, start, size);
+        // Copy vertex array data
+        Memory::ReadBlock(*memory_manager->GpuToCpuAddress(start), array_ptr, size);
 
         // Bind the vertex array to the buffer at the current offset.
-        glBindVertexBuffer(index, stream_buffer.GetHandle(), vertex_buffer_offset,
-                           vertex_array.stride);
+        glBindVertexBuffer(index, stream_buffer->GetHandle(), buffer_offset, vertex_array.stride);
 
         ASSERT_MSG(vertex_array.divisor == 0, "Vertex buffer divisor unimplemented");
+
+        array_ptr += size;
+        buffer_offset += size;
     }
 
     // Use the vertex array as-is, assumes that the data is formatted correctly for OpenGL.
@@ -175,12 +201,22 @@ static GLShader::ProgramCode GetShaderProgramCode(Maxwell::ShaderProgram program
     return program_code;
 }
 
-std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr buffer_offset) {
+void RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr buffer_offset) {
+    // Helper function for uploading uniform data
+    const auto copy_buffer = [&](GLuint handle, GLintptr offset, GLsizeiptr size) {
+        if (has_ARB_direct_state_access) {
+            glCopyNamedBufferSubData(stream_buffer->GetHandle(), handle, offset, 0, size);
+        } else {
+            glBindBuffer(GL_COPY_WRITE_BUFFER, handle);
+            glCopyBufferSubData(GL_ARRAY_BUFFER, GL_COPY_WRITE_BUFFER, offset, 0, size);
+        }
+    };
+
     auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
 
     // Next available bindpoints to use when uploading the const buffers and textures to the GLSL
     // shaders. The constbuffer bindpoint starts after the shader stage configuration bind points.
-    u32 current_constbuffer_bindpoint = Tegra::Engines::Maxwell3D::Regs::MaxShaderStage;
+    u32 current_constbuffer_bindpoint = static_cast<u32>(uniform_buffers.size());
     u32 current_texture_bindpoint = 0;
 
     for (size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
@@ -192,21 +228,22 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr
             continue;
         }
 
-        std::tie(buffer_ptr, buffer_offset) =
-            AlignBuffer(buffer_ptr, buffer_offset, static_cast<size_t>(uniform_buffer_alignment));
-
         const size_t stage{index == 0 ? 0 : index - 1}; // Stage indices are 0 - 5
 
         GLShader::MaxwellUniformData ubo{};
         ubo.SetFromRegs(gpu.state.shader_stages[stage]);
         std::memcpy(buffer_ptr, &ubo, sizeof(ubo));
 
-        // Bind the buffer
-        glBindBufferRange(GL_UNIFORM_BUFFER, stage, stream_buffer.GetHandle(), buffer_offset,
-                          sizeof(ubo));
+        // Flush the buffer so that the GPU can see the data we just wrote.
+        glFlushMappedBufferRange(GL_ARRAY_BUFFER, buffer_offset, sizeof(ubo));
 
-        buffer_ptr += sizeof(ubo);
-        buffer_offset += sizeof(ubo);
+        // Upload uniform data as one UBO per stage
+        const GLintptr ubo_offset = buffer_offset;
+        copy_buffer(uniform_buffers[stage].handle, ubo_offset,
+                    sizeof(GLShader::MaxwellUniformData));
+
+        buffer_ptr += sizeof(GLShader::MaxwellUniformData);
+        buffer_offset += sizeof(GLShader::MaxwellUniformData);
 
         GLShader::ShaderSetup setup{GetShaderProgramCode(program)};
         GLShader::ShaderEntries shader_resources;
@@ -245,9 +282,9 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr
             static_cast<Maxwell::ShaderStage>(stage));
 
         // Configure the const buffers for this shader stage.
-        std::tie(buffer_ptr, buffer_offset, current_constbuffer_bindpoint) = SetupConstBuffers(
-            buffer_ptr, buffer_offset, static_cast<Maxwell::ShaderStage>(stage), gl_stage_program,
-            current_constbuffer_bindpoint, shader_resources.const_buffer_entries);
+        current_constbuffer_bindpoint =
+            SetupConstBuffers(static_cast<Maxwell::ShaderStage>(stage), gl_stage_program,
+                              current_constbuffer_bindpoint, shader_resources.const_buffer_entries);
 
         // Configure the textures for this shader stage.
         current_texture_bindpoint =
@@ -262,8 +299,6 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr
     }
 
     shader_program_manager->UseTrivialGeometryShader();
-
-    return {buffer_ptr, buffer_offset};
 }
 
 size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
@@ -397,31 +432,6 @@ void RasterizerOpenGL::Clear() {
     }
 }
 
-std::pair<u8*, GLintptr> RasterizerOpenGL::AlignBuffer(u8* buffer_ptr, GLintptr buffer_offset,
-                                                       size_t alignment) {
-    // Align the offset, not the mapped pointer
-    GLintptr offset_aligned =
-        static_cast<GLintptr>(Common::AlignUp(static_cast<size_t>(buffer_offset), alignment));
-    return {buffer_ptr + (offset_aligned - buffer_offset), offset_aligned};
-}
-
-std::tuple<u8*, GLintptr, GLintptr> RasterizerOpenGL::UploadMemory(u8* buffer_ptr,
-                                                                   GLintptr buffer_offset,
-                                                                   Tegra::GPUVAddr gpu_addr,
-                                                                   size_t size, size_t alignment) {
-    std::tie(buffer_ptr, buffer_offset) = AlignBuffer(buffer_ptr, buffer_offset, alignment);
-    GLintptr uploaded_offset = buffer_offset;
-
-    const auto& memory_manager = Core::System::GetInstance().GPU().memory_manager;
-    const boost::optional<VAddr> cpu_addr{memory_manager->GpuToCpuAddress(gpu_addr)};
-    Memory::ReadBlock(*cpu_addr, buffer_ptr, size);
-
-    buffer_ptr += size;
-    buffer_offset += size;
-
-    return {buffer_ptr, buffer_offset, uploaded_offset};
-}
-
 void RasterizerOpenGL::DrawArrays() {
     if (accelerate_draw == AccelDraw::Disabled)
         return;
@@ -446,7 +456,7 @@ void RasterizerOpenGL::DrawArrays() {
     const u64 index_buffer_size{regs.index_array.count * regs.index_array.FormatSizeInBytes()};
     const unsigned vertex_num{is_indexed ? regs.index_array.count : regs.vertex_buffer.count};
 
-    state.draw.vertex_buffer = stream_buffer.GetHandle();
+    state.draw.vertex_buffer = stream_buffer->GetHandle();
     state.Apply();
 
     size_t buffer_size = CalculateVertexArraysSize();
@@ -456,31 +466,41 @@ void RasterizerOpenGL::DrawArrays() {
     }
 
     // Uniform space for the 5 shader stages
-    buffer_size =
-        Common::AlignUp<size_t>(buffer_size, 4) +
-        (sizeof(GLShader::MaxwellUniformData) + uniform_buffer_alignment) * Maxwell::MaxShaderStage;
-
-    // Add space for at least 18 constant buffers
-    buffer_size += Maxwell::MaxConstBuffers * (MaxConstbufferSize + uniform_buffer_alignment);
+    buffer_size = Common::AlignUp<size_t>(buffer_size, 4) +
+                  sizeof(GLShader::MaxwellUniformData) * Maxwell::MaxShaderStage;
 
     u8* buffer_ptr;
     GLintptr buffer_offset;
-    std::tie(buffer_ptr, buffer_offset, std::ignore) =
-        stream_buffer.Map(static_cast<GLsizeiptr>(buffer_size), 4);
-    u8* buffer_ptr_base = buffer_ptr;
+    std::tie(buffer_ptr, buffer_offset) =
+        stream_buffer->Map(static_cast<GLsizeiptr>(buffer_size), 4);
 
-    std::tie(buffer_ptr, buffer_offset) = SetupVertexArrays(buffer_ptr, buffer_offset);
+    u8* offseted_buffer;
+    std::tie(offseted_buffer, buffer_offset) = SetupVertexArrays(buffer_ptr, buffer_offset);
+
+    offseted_buffer =
+        reinterpret_cast<u8*>(Common::AlignUp(reinterpret_cast<size_t>(offseted_buffer), 4));
+    buffer_offset = Common::AlignUp<size_t>(buffer_offset, 4);
 
     // If indexed mode, copy the index buffer
     GLintptr index_buffer_offset = 0;
     if (is_indexed) {
-        std::tie(buffer_ptr, buffer_offset, index_buffer_offset) = UploadMemory(
-            buffer_ptr, buffer_offset, regs.index_array.StartAddress(), index_buffer_size);
+        const auto& memory_manager = Core::System::GetInstance().GPU().memory_manager;
+        const boost::optional<VAddr> index_data_addr{
+            memory_manager->GpuToCpuAddress(regs.index_array.StartAddress())};
+        Memory::ReadBlock(*index_data_addr, offseted_buffer, index_buffer_size);
+
+        index_buffer_offset = buffer_offset;
+        offseted_buffer += index_buffer_size;
+        buffer_offset += index_buffer_size;
     }
 
-    std::tie(buffer_ptr, buffer_offset) = SetupShaders(buffer_ptr, buffer_offset);
+    offseted_buffer =
+        reinterpret_cast<u8*>(Common::AlignUp(reinterpret_cast<size_t>(offseted_buffer), 4));
+    buffer_offset = Common::AlignUp<size_t>(buffer_offset, 4);
 
-    stream_buffer.Unmap(buffer_ptr - buffer_ptr_base);
+    SetupShaders(offseted_buffer, buffer_offset);
+
+    stream_buffer->Unmap();
 
     shader_program_manager->ApplyTo(state);
     state.Apply();
@@ -627,23 +647,36 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr
     }
 }
 
-std::tuple<u8*, GLintptr, u32> RasterizerOpenGL::SetupConstBuffers(
-    u8* buffer_ptr, GLintptr buffer_offset, Maxwell::ShaderStage stage, GLuint program,
-    u32 current_bindpoint, const std::vector<GLShader::ConstBufferEntry>& entries) {
+u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, GLuint program,
+                                        u32 current_bindpoint,
+                                        const std::vector<GLShader::ConstBufferEntry>& entries) {
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
 
+    // Reset all buffer draw state for this stage.
+    for (auto& buffer : state.draw.const_buffers[static_cast<size_t>(stage)]) {
+        buffer.bindpoint = 0;
+        buffer.enabled = false;
+    }
+
     // Upload only the enabled buffers from the 16 constbuffers of each shader stage
     const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<size_t>(stage)];
 
     for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
         const auto& used_buffer = entries[bindpoint];
         const auto& buffer = shader_stage.const_buffers[used_buffer.GetIndex()];
+        auto& buffer_draw_state =
+            state.draw.const_buffers[static_cast<size_t>(stage)][used_buffer.GetIndex()];
 
         if (!buffer.enabled) {
             continue;
         }
 
+        buffer_draw_state.enabled = true;
+        buffer_draw_state.bindpoint = current_bindpoint + bindpoint;
+
+        boost::optional<VAddr> addr = gpu.memory_manager->GpuToCpuAddress(buffer.address);
+
         size_t size = 0;
 
         if (used_buffer.IsIndirect()) {
@@ -665,26 +698,25 @@ std::tuple<u8*, GLintptr, u32> RasterizerOpenGL::SetupConstBuffers(
         size = Common::AlignUp(size, sizeof(GLvec4));
         ASSERT_MSG(size <= MaxConstbufferSize, "Constbuffer too big");
 
-        GLintptr const_buffer_offset;
-        std::tie(buffer_ptr, buffer_offset, const_buffer_offset) =
-            UploadMemory(buffer_ptr, buffer_offset, buffer.address, size,
-                         static_cast<size_t>(uniform_buffer_alignment));
+        std::vector<u8> data(size);
+        Memory::ReadBlock(*addr, data.data(), data.size());
 
-        glBindBufferRange(GL_UNIFORM_BUFFER, current_bindpoint + bindpoint,
-                          stream_buffer.GetHandle(), const_buffer_offset, size);
+        glBindBuffer(GL_UNIFORM_BUFFER, buffer_draw_state.ssbo);
+        glBufferData(GL_UNIFORM_BUFFER, data.size(), data.data(), GL_DYNAMIC_DRAW);
+        glBindBuffer(GL_UNIFORM_BUFFER, 0);
 
         // Now configure the bindpoint of the buffer inside the shader
         const std::string buffer_name = used_buffer.GetName();
         const GLuint index =
             glGetProgramResourceIndex(program, GL_UNIFORM_BLOCK, buffer_name.c_str());
         if (index != GL_INVALID_INDEX) {
-            glUniformBlockBinding(program, index, current_bindpoint + bindpoint);
+            glUniformBlockBinding(program, index, buffer_draw_state.bindpoint);
         }
     }
 
     state.Apply();
 
-    return {buffer_ptr, buffer_offset, current_bindpoint + static_cast<u32>(entries.size())};
+    return current_bindpoint + static_cast<u32>(entries.size());
 }
 
 u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, GLuint program, u32 current_unit,

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -7,7 +7,6 @@
 #include <array>
 #include <cstddef>
 #include <memory>
-#include <tuple>
 #include <utility>
 #include <vector>
 #include <glad/glad.h>
@@ -101,10 +100,9 @@ private:
      * @param entries Vector describing the buffers that are actually used in the guest shader.
      * @returns The next available bindpoint for use in the next shader stage.
      */
-    std::tuple<u8*, GLintptr, u32> SetupConstBuffers(
-        u8* buffer_ptr, GLintptr buffer_offset, Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
-        GLuint program, u32 current_bindpoint,
-        const std::vector<GLShader::ConstBufferEntry>& entries);
+    u32 SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, GLuint program,
+                          u32 current_bindpoint,
+                          const std::vector<GLShader::ConstBufferEntry>& entries);
 
     /*
      * Configures the current textures to use for the draw command.
@@ -141,6 +139,7 @@ private:
     /// Syncs the blend state to match the guest state
     void SyncBlendState();
 
+    bool has_ARB_buffer_storage = false;
     bool has_ARB_direct_state_access = false;
     bool has_ARB_separate_shader_objects = false;
     bool has_ARB_vertex_attrib_binding = false;
@@ -156,24 +155,22 @@ private:
     OGLVertexArray hw_vao;
 
     std::array<SamplerInfo, GLShader::NumTextureSamplers> texture_samplers;
+    std::array<std::array<OGLBuffer, Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers>,
+               Tegra::Engines::Maxwell3D::Regs::MaxShaderStage>
+        ssbos;
 
     static constexpr size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
-    OGLStreamBuffer stream_buffer;
+    std::unique_ptr<OGLStreamBuffer> stream_buffer;
     OGLBuffer uniform_buffer;
     OGLFramebuffer framebuffer;
-    GLint uniform_buffer_alignment;
 
     size_t CalculateVertexArraysSize() const;
 
     std::pair<u8*, GLintptr> SetupVertexArrays(u8* array_ptr, GLintptr buffer_offset);
 
-    std::pair<u8*, GLintptr> SetupShaders(u8* buffer_ptr, GLintptr buffer_offset);
+    std::array<OGLBuffer, Tegra::Engines::Maxwell3D::Regs::MaxShaderStage> uniform_buffers;
 
-    std::pair<u8*, GLintptr> AlignBuffer(u8* buffer_ptr, GLintptr buffer_offset, size_t alignment);
-
-    std::tuple<u8*, GLintptr, GLintptr> UploadMemory(u8* buffer_ptr, GLintptr buffer_offset,
-                                                     Tegra::GPUVAddr gpu_addr, size_t size,
-                                                     size_t alignment = 4);
+    void SetupShaders(u8* buffer_ptr, GLintptr buffer_offset);
 
     enum class AccelDraw { Disabled, Arrays, Indexed };
     AccelDraw accelerate_draw = AccelDraw::Disabled;

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -101,7 +101,6 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
     {GL_R8, GL_RED, GL_UNSIGNED_BYTE, ComponentType::UNorm, false},                    // R8
     {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE, ComponentType::UInt, false},           // R8UI
     {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, ComponentType::Float, false},                 // RGBA16F
-    {GL_RGBA16UI, GL_RGBA, GL_UNSIGNED_SHORT, ComponentType::UInt, false},             // RGBA16UI
     {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, ComponentType::Float,
      false},                                                                     // R11FG11FB10F
     {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // RGBA32UI
@@ -135,7 +134,6 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
     {GL_RG16_SNORM, GL_RG, GL_SHORT, ComponentType::SNorm, false},             // RG16S
     {GL_RGB32F, GL_RGB, GL_FLOAT, ComponentType::Float, false},                // RGB32F
     {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, ComponentType::UNorm, false}, // SRGBA8
-    {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, ComponentType::UNorm, false},                       // RG8U
     {GL_RG8, GL_RG, GL_BYTE, ComponentType::SNorm, false},                                // RG8S
 
     // DepthStencil formats
@@ -236,57 +234,32 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_bu
 static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPUVAddr),
                             SurfaceParams::MaxPixelFormat>
     morton_to_gl_fns = {
-        // clang-format off
-        MortonCopy<true, PixelFormat::ABGR8U>,
-        MortonCopy<true, PixelFormat::ABGR8S>,
-        MortonCopy<true, PixelFormat::B5G6R5>,
-        MortonCopy<true, PixelFormat::A2B10G10R10>,
-        MortonCopy<true, PixelFormat::A1B5G5R5>,
-        MortonCopy<true, PixelFormat::R8>,
-        MortonCopy<true, PixelFormat::R8UI>,
-        MortonCopy<true, PixelFormat::RGBA16F>,
-        MortonCopy<true, PixelFormat::RGBA16UI>,
-        MortonCopy<true, PixelFormat::R11FG11FB10F>,
-        MortonCopy<true, PixelFormat::RGBA32UI>,
-        MortonCopy<true, PixelFormat::DXT1>,
-        MortonCopy<true, PixelFormat::DXT23>,
-        MortonCopy<true, PixelFormat::DXT45>,
-        MortonCopy<true, PixelFormat::DXN1>,
-        MortonCopy<true, PixelFormat::DXN2UNORM>,
-        MortonCopy<true, PixelFormat::DXN2SNORM>,
-        MortonCopy<true, PixelFormat::BC7U>,
-        MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
-        MortonCopy<true, PixelFormat::G8R8>,
-        MortonCopy<true, PixelFormat::BGRA8>,
-        MortonCopy<true, PixelFormat::RGBA32F>,
-        MortonCopy<true, PixelFormat::RG32F>,
-        MortonCopy<true, PixelFormat::R32F>,
-        MortonCopy<true, PixelFormat::R16F>,
-        MortonCopy<true, PixelFormat::R16UNORM>,
-        MortonCopy<true, PixelFormat::R16S>,
-        MortonCopy<true, PixelFormat::R16UI>,
-        MortonCopy<true, PixelFormat::R16I>,
-        MortonCopy<true, PixelFormat::RG16>,
-        MortonCopy<true, PixelFormat::RG16F>,
-        MortonCopy<true, PixelFormat::RG16UI>,
-        MortonCopy<true, PixelFormat::RG16I>,
-        MortonCopy<true, PixelFormat::RG16S>,
-        MortonCopy<true, PixelFormat::RGB32F>,
-        MortonCopy<true, PixelFormat::SRGBA8>,
-        MortonCopy<true, PixelFormat::RG8U>,
-        MortonCopy<true, PixelFormat::RG8S>,
-        MortonCopy<true, PixelFormat::Z24S8>,
-        MortonCopy<true, PixelFormat::S8Z24>,
-        MortonCopy<true, PixelFormat::Z32F>,
-        MortonCopy<true, PixelFormat::Z16>,
+        MortonCopy<true, PixelFormat::ABGR8U>,       MortonCopy<true, PixelFormat::ABGR8S>,
+        MortonCopy<true, PixelFormat::B5G6R5>,       MortonCopy<true, PixelFormat::A2B10G10R10>,
+        MortonCopy<true, PixelFormat::A1B5G5R5>,     MortonCopy<true, PixelFormat::R8>,
+        MortonCopy<true, PixelFormat::R8UI>,         MortonCopy<true, PixelFormat::RGBA16F>,
+        MortonCopy<true, PixelFormat::R11FG11FB10F>, MortonCopy<true, PixelFormat::RGBA32UI>,
+        MortonCopy<true, PixelFormat::DXT1>,         MortonCopy<true, PixelFormat::DXT23>,
+        MortonCopy<true, PixelFormat::DXT45>,        MortonCopy<true, PixelFormat::DXN1>,
+        MortonCopy<true, PixelFormat::DXN2UNORM>,    MortonCopy<true, PixelFormat::DXN2SNORM>,
+        MortonCopy<true, PixelFormat::BC7U>,         MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
+        MortonCopy<true, PixelFormat::G8R8>,         MortonCopy<true, PixelFormat::BGRA8>,
+        MortonCopy<true, PixelFormat::RGBA32F>,      MortonCopy<true, PixelFormat::RG32F>,
+        MortonCopy<true, PixelFormat::R32F>,         MortonCopy<true, PixelFormat::R16F>,
+        MortonCopy<true, PixelFormat::R16UNORM>,     MortonCopy<true, PixelFormat::R16S>,
+        MortonCopy<true, PixelFormat::R16UI>,        MortonCopy<true, PixelFormat::R16I>,
+        MortonCopy<true, PixelFormat::RG16>,         MortonCopy<true, PixelFormat::RG16F>,
+        MortonCopy<true, PixelFormat::RG16UI>,       MortonCopy<true, PixelFormat::RG16I>,
+        MortonCopy<true, PixelFormat::RG16S>,        MortonCopy<true, PixelFormat::RGB32F>,
+        MortonCopy<true, PixelFormat::SRGBA8>,       MortonCopy<true, PixelFormat::RG8S>,
+        MortonCopy<true, PixelFormat::Z24S8>,        MortonCopy<true, PixelFormat::S8Z24>,
+        MortonCopy<true, PixelFormat::Z32F>,         MortonCopy<true, PixelFormat::Z16>,
         MortonCopy<true, PixelFormat::Z32FS8>,
-        // clang-format on
 };
 
 static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPUVAddr),
                             SurfaceParams::MaxPixelFormat>
     gl_to_morton_fns = {
-        // clang-format off
         MortonCopy<false, PixelFormat::ABGR8U>,
         MortonCopy<false, PixelFormat::ABGR8S>,
         MortonCopy<false, PixelFormat::B5G6R5>,
@@ -295,7 +268,6 @@ static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPU
         MortonCopy<false, PixelFormat::R8>,
         MortonCopy<false, PixelFormat::R8UI>,
         MortonCopy<false, PixelFormat::RGBA16F>,
-        MortonCopy<false, PixelFormat::RGBA16UI>,
         MortonCopy<false, PixelFormat::R11FG11FB10F>,
         MortonCopy<false, PixelFormat::RGBA32UI>,
         // TODO(Subv): Swizzling DXT1/DXT23/DXT45/DXN1/DXN2/BC7U/ASTC_2D_4X4 formats is not
@@ -325,14 +297,12 @@ static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPU
         MortonCopy<false, PixelFormat::RG16S>,
         MortonCopy<false, PixelFormat::RGB32F>,
         MortonCopy<false, PixelFormat::SRGBA8>,
-        MortonCopy<false, PixelFormat::RG8U>,
         MortonCopy<false, PixelFormat::RG8S>,
         MortonCopy<false, PixelFormat::Z24S8>,
         MortonCopy<false, PixelFormat::S8Z24>,
         MortonCopy<false, PixelFormat::Z32F>,
         MortonCopy<false, PixelFormat::Z16>,
         MortonCopy<false, PixelFormat::Z32FS8>,
-        // clang-format on
 };
 
 // Allocate an uninitialized texture of appropriate size and format for the surface

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -31,45 +31,43 @@ struct SurfaceParams {
         R8 = 5,
         R8UI = 6,
         RGBA16F = 7,
-        RGBA16UI = 8,
-        R11FG11FB10F = 9,
-        RGBA32UI = 10,
-        DXT1 = 11,
-        DXT23 = 12,
-        DXT45 = 13,
-        DXN1 = 14, // This is also known as BC4
-        DXN2UNORM = 15,
-        DXN2SNORM = 16,
-        BC7U = 17,
-        ASTC_2D_4X4 = 18,
-        G8R8 = 19,
-        BGRA8 = 20,
-        RGBA32F = 21,
-        RG32F = 22,
-        R32F = 23,
-        R16F = 24,
-        R16UNORM = 25,
-        R16S = 26,
-        R16UI = 27,
-        R16I = 28,
-        RG16 = 29,
-        RG16F = 30,
-        RG16UI = 31,
-        RG16I = 32,
-        RG16S = 33,
-        RGB32F = 34,
-        SRGBA8 = 35,
-        RG8U = 36,
-        RG8S = 37,
+        R11FG11FB10F = 8,
+        RGBA32UI = 9,
+        DXT1 = 10,
+        DXT23 = 11,
+        DXT45 = 12,
+        DXN1 = 13, // This is also known as BC4
+        DXN2UNORM = 14,
+        DXN2SNORM = 15,
+        BC7U = 16,
+        ASTC_2D_4X4 = 17,
+        G8R8 = 18,
+        BGRA8 = 19,
+        RGBA32F = 20,
+        RG32F = 21,
+        R32F = 22,
+        R16F = 23,
+        R16UNORM = 24,
+        R16S = 25,
+        R16UI = 26,
+        R16I = 27,
+        RG16 = 28,
+        RG16F = 29,
+        RG16UI = 30,
+        RG16I = 31,
+        RG16S = 32,
+        RGB32F = 33,
+        SRGBA8 = 34,
+        RG8S = 35,
 
         MaxColorFormat,
 
         // DepthStencil formats
-        Z24S8 = 38,
-        S8Z24 = 39,
-        Z32F = 40,
-        Z16 = 41,
-        Z32FS8 = 42,
+        Z24S8 = 36,
+        S8Z24 = 37,
+        Z32F = 38,
+        Z16 = 39,
+        Z32FS8 = 40,
 
         MaxDepthStencilFormat,
 
@@ -115,7 +113,6 @@ struct SurfaceParams {
             1, // R8
             1, // R8UI
             1, // RGBA16F
-            1, // RGBA16UI
             1, // R11FG11FB10F
             1, // RGBA32UI
             4, // DXT1
@@ -143,7 +140,6 @@ struct SurfaceParams {
             1, // RG16S
             1, // RGB32F
             1, // SRGBA8
-            1, // RG8U
             1, // RG8S
             1, // Z24S8
             1, // S8Z24
@@ -169,7 +165,6 @@ struct SurfaceParams {
             8,   // R8
             8,   // R8UI
             64,  // RGBA16F
-            64,  // RGBA16UI
             32,  // R11FG11FB10F
             128, // RGBA32UI
             64,  // DXT1
@@ -197,7 +192,6 @@ struct SurfaceParams {
             32,  // RG16S
             96,  // RGB32F
             32,  // SRGBA8
-            16,  // RG8U
             16,  // RG8S
             32,  // Z24S8
             32,  // S8Z24
@@ -247,8 +241,6 @@ struct SurfaceParams {
             return PixelFormat::A2B10G10R10;
         case Tegra::RenderTargetFormat::RGBA16_FLOAT:
             return PixelFormat::RGBA16F;
-        case Tegra::RenderTargetFormat::RGBA16_UINT:
-            return PixelFormat::RGBA16UI;
         case Tegra::RenderTargetFormat::RGBA32_FLOAT:
             return PixelFormat::RGBA32F;
         case Tegra::RenderTargetFormat::RG32_FLOAT:
@@ -273,8 +265,6 @@ struct SurfaceParams {
             return PixelFormat::RG16;
         case Tegra::RenderTargetFormat::RG16_SNORM:
             return PixelFormat::RG16S;
-        case Tegra::RenderTargetFormat::RG8_UNORM:
-            return PixelFormat::RG8U;
         case Tegra::RenderTargetFormat::RG8_SNORM:
             return PixelFormat::RG8S;
         case Tegra::RenderTargetFormat::R16_FLOAT:
@@ -442,7 +432,6 @@ struct SurfaceParams {
         case Tegra::RenderTargetFormat::RG16_UNORM:
         case Tegra::RenderTargetFormat::R16_UNORM:
         case Tegra::RenderTargetFormat::B5G6R5_UNORM:
-        case Tegra::RenderTargetFormat::RG8_UNORM:
             return ComponentType::UNorm;
         case Tegra::RenderTargetFormat::RGBA8_SNORM:
         case Tegra::RenderTargetFormat::RG16_SNORM:
@@ -458,7 +447,6 @@ struct SurfaceParams {
         case Tegra::RenderTargetFormat::R32_FLOAT:
             return ComponentType::Float;
         case Tegra::RenderTargetFormat::RGBA32_UINT:
-        case Tegra::RenderTargetFormat::RGBA16_UINT:
         case Tegra::RenderTargetFormat::RG16_UINT:
         case Tegra::RenderTargetFormat::R8_UINT:
         case Tegra::RenderTargetFormat::R16_UINT:

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -356,13 +356,13 @@ public:
      * @param reg The register to use as the source value.
      */
     void SetOutputAttributeToRegister(Attribute::Index attribute, u64 elem, const Register& reg) {
-        std::string dest = GetOutputAttribute(attribute);
+        std::string dest = GetOutputAttribute(attribute) + GetSwizzle(elem);
         std::string src = GetRegisterAsFloat(reg);
 
         if (!dest.empty()) {
             // Can happen with unknown/unimplemented output attributes, in which case we ignore the
             // instruction for now.
-            shader.AddLine(dest + GetSwizzle(elem) + " = " + src + ';');
+            shader.AddLine(dest + " = " + src + ';');
         }
     }
 
@@ -376,8 +376,6 @@ public:
             return value;
         } else if (type == GLSLRegister::Type::Integer) {
             return "floatBitsToInt(" + value + ')';
-        } else if (type == GLSLRegister::Type::UnsignedInteger) {
-            return "floatBitsToUint(" + value + ')';
         } else {
             UNREACHABLE();
         }
@@ -1632,99 +1630,6 @@ private:
             }
             break;
         }
-        case OpCode::Type::Xmad: {
-            ASSERT_MSG(!instr.xmad.sign_a, "Unimplemented");
-            ASSERT_MSG(!instr.xmad.sign_b, "Unimplemented");
-
-            std::string op_a{regs.GetRegisterAsInteger(instr.gpr8, 0, instr.xmad.sign_a)};
-            std::string op_b;
-            std::string op_c;
-
-            // TODO(bunnei): Needs to be fixed once op_a or op_b is signed
-            ASSERT_MSG(instr.xmad.sign_a == instr.xmad.sign_b, "Unimplemented");
-            const bool is_signed{instr.xmad.sign_a == 1};
-
-            bool is_merge{};
-            switch (opcode->GetId()) {
-            case OpCode::Id::XMAD_CR: {
-                is_merge = instr.xmad.merge_56;
-                op_b += regs.GetUniform(instr.cbuf34.index, instr.cbuf34.offset,
-                                        instr.xmad.sign_b ? GLSLRegister::Type::Integer
-                                                          : GLSLRegister::Type::UnsignedInteger);
-                op_c += regs.GetRegisterAsInteger(instr.gpr39, 0, is_signed);
-                break;
-            }
-            case OpCode::Id::XMAD_RR: {
-                is_merge = instr.xmad.merge_37;
-                op_b += regs.GetRegisterAsInteger(instr.gpr20, 0, instr.xmad.sign_b);
-                op_c += regs.GetRegisterAsInteger(instr.gpr39, 0, is_signed);
-                break;
-            }
-            case OpCode::Id::XMAD_RC: {
-                op_b += regs.GetRegisterAsInteger(instr.gpr39, 0, instr.xmad.sign_b);
-                op_c += regs.GetUniform(instr.cbuf34.index, instr.cbuf34.offset,
-                                        is_signed ? GLSLRegister::Type::Integer
-                                                  : GLSLRegister::Type::UnsignedInteger);
-                break;
-            }
-            case OpCode::Id::XMAD_IMM: {
-                is_merge = instr.xmad.merge_37;
-                op_b += std::to_string(instr.xmad.imm20_16);
-                op_c += regs.GetRegisterAsInteger(instr.gpr39, 0, is_signed);
-                break;
-            }
-            default: {
-                LOG_CRITICAL(HW_GPU, "Unhandled XMAD instruction: {}", opcode->GetName());
-                UNREACHABLE();
-            }
-            }
-
-            // TODO(bunnei): Ensure this is right with signed operands
-            if (instr.xmad.high_a) {
-                op_a = "((" + op_a + ") >> 16)";
-            } else {
-                op_a = "((" + op_a + ") & 0xFFFF)";
-            }
-
-            std::string src2 = '(' + op_b + ')'; // Preserve original source 2
-            if (instr.xmad.high_b) {
-                op_b = '(' + src2 + " >> 16)";
-            } else {
-                op_b = '(' + src2 + " & 0xFFFF)";
-            }
-
-            std::string product = '(' + op_a + " * " + op_b + ')';
-            if (instr.xmad.product_shift_left) {
-                product = '(' + product + " << 16)";
-            }
-
-            switch (instr.xmad.mode) {
-            case Tegra::Shader::XmadMode::None:
-                break;
-            case Tegra::Shader::XmadMode::CLo:
-                op_c = "((" + op_c + ") & 0xFFFF)";
-                break;
-            case Tegra::Shader::XmadMode::CHi:
-                op_c = "((" + op_c + ") >> 16)";
-                break;
-            case Tegra::Shader::XmadMode::CBcc:
-                op_c = "((" + op_c + ") + (" + src2 + "<< 16))";
-                break;
-            default: {
-                LOG_CRITICAL(HW_GPU, "Unhandled XMAD mode: {}",
-                             static_cast<u32>(instr.xmad.mode.Value()));
-                UNREACHABLE();
-            }
-            }
-
-            std::string sum{'(' + product + " + " + op_c + ')'};
-            if (is_merge) {
-                sum = "((" + sum + " & 0xFFFF) | (" + src2 + "<< 16))";
-            }
-
-            regs.SetRegisterToInteger(instr.gpr0, is_signed, 0, sum, 1, 1);
-            break;
-        }
         default: {
             switch (opcode->GetId()) {
             case OpCode::Id::EXIT: {
@@ -1762,15 +1667,7 @@ private:
             }
             case OpCode::Id::KIL: {
                 ASSERT(instr.flow.cond == Tegra::Shader::FlowCondition::Always);
-
-                // Enclose "discard" in a conditional, so that GLSL compilation does not complain
-                // about unexecuted instructions that may follow this.
-                shader.AddLine("if (true) {");
-                ++shader.scope;
                 shader.AddLine("discard;");
-                --shader.scope;
-                shader.AddLine("}");
-
                 break;
             }
             case OpCode::Id::BRA: {

src/video_core/renderer_opengl/gl_state.cpp

@@ -203,6 +203,21 @@ void OpenGLState::Apply() const {
         }
     }
 
+    // Constbuffers
+    for (std::size_t stage = 0; stage < draw.const_buffers.size(); ++stage) {
+        for (std::size_t buffer_id = 0; buffer_id < draw.const_buffers[stage].size(); ++buffer_id) {
+            const auto& current = cur_state.draw.const_buffers[stage][buffer_id];
+            const auto& new_state = draw.const_buffers[stage][buffer_id];
+
+            if (current.enabled != new_state.enabled || current.bindpoint != new_state.bindpoint ||
+                current.ssbo != new_state.ssbo) {
+                if (new_state.enabled) {
+                    glBindBufferBase(GL_UNIFORM_BUFFER, new_state.bindpoint, new_state.ssbo);
+                }
+            }
+        }
+    }
+
     // Framebuffer
     if (draw.read_framebuffer != cur_state.draw.read_framebuffer) {
         glBindFramebuffer(GL_READ_FRAMEBUFFER, draw.read_framebuffer);

src/video_core/renderer_opengl/gl_state.h

@@ -119,6 +119,12 @@ public:
         GLuint uniform_buffer;   // GL_UNIFORM_BUFFER_BINDING
         GLuint shader_program;   // GL_CURRENT_PROGRAM
         GLuint program_pipeline; // GL_PROGRAM_PIPELINE_BINDING
+        struct ConstBufferConfig {
+            bool enabled = false;
+            GLuint bindpoint;
+            GLuint ssbo;
+        };
+        std::array<std::array<ConstBufferConfig, Regs::MaxConstBuffers>, 5> const_buffers;
     } draw;
 
     struct {

src/video_core/renderer_opengl/gl_stream_buffer.cpp

@@ -9,91 +9,174 @@
 #include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
 
-OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent)
-    : gl_target(target), buffer_size(size) {
-    gl_buffer.Create();
-    glBindBuffer(gl_target, gl_buffer.handle);
+class OrphanBuffer : public OGLStreamBuffer {
+public:
+    explicit OrphanBuffer(GLenum target) : OGLStreamBuffer(target) {}
+    ~OrphanBuffer() override;
 
-    GLsizeiptr allocate_size = size;
-    if (target == GL_ARRAY_BUFFER) {
-        // On AMD GPU there is a strange crash in indexed drawing. The crash happens when the buffer
-        // read position is near the end and is an out-of-bound access to the vertex buffer. This is
-        // probably a bug in the driver and is related to the usage of vec3<byte> attributes in the
-        // vertex array. Doubling the allocation size for the vertex buffer seems to avoid the
-        // crash.
-        allocate_size *= 2;
-    }
+private:
+    void Create(size_t size, size_t sync_subdivide) override;
+    void Release() override;
 
-    if (GLAD_GL_ARB_buffer_storage) {
-        persistent = true;
-        coherent = prefer_coherent;
-        GLbitfield flags =
-            GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | (coherent ? GL_MAP_COHERENT_BIT : 0);
-        glBufferStorage(gl_target, allocate_size, nullptr, flags);
-        mapped_ptr = static_cast<u8*>(glMapBufferRange(
-            gl_target, 0, buffer_size, flags | (coherent ? 0 : GL_MAP_FLUSH_EXPLICIT_BIT)));
-    } else {
-        glBufferData(gl_target, allocate_size, nullptr, GL_STREAM_DRAW);
-    }
-}
+    std::pair<u8*, GLintptr> Map(size_t size, size_t alignment) override;
+    void Unmap() override;
 
-OGLStreamBuffer::~OGLStreamBuffer() {
-    if (persistent) {
-        glBindBuffer(gl_target, gl_buffer.handle);
-        glUnmapBuffer(gl_target);
-    }
-    gl_buffer.Release();
+    std::vector<u8> data;
+};
+
+class StorageBuffer : public OGLStreamBuffer {
+public:
+    explicit StorageBuffer(GLenum target) : OGLStreamBuffer(target) {}
+    ~StorageBuffer() override;
+
+private:
+    void Create(size_t size, size_t sync_subdivide) override;
+    void Release() override;
+
+    std::pair<u8*, GLintptr> Map(size_t size, size_t alignment) override;
+    void Unmap() override;
+
+    struct Fence {
+        OGLSync sync;
+        size_t offset;
+    };
+    std::deque<Fence> head;
+    std::deque<Fence> tail;
+
+    u8* mapped_ptr;
+};
+
+OGLStreamBuffer::OGLStreamBuffer(GLenum target) {
+    gl_target = target;
 }
 
 GLuint OGLStreamBuffer::GetHandle() const {
     return gl_buffer.handle;
 }
 
-GLsizeiptr OGLStreamBuffer::GetSize() const {
-    return buffer_size;
+std::unique_ptr<OGLStreamBuffer> OGLStreamBuffer::MakeBuffer(bool storage_buffer, GLenum target) {
+    if (storage_buffer) {
+        return std::make_unique<StorageBuffer>(target);
+    }
+    return std::make_unique<OrphanBuffer>(target);
 }
 
-std::tuple<u8*, GLintptr, bool> OGLStreamBuffer::Map(GLsizeiptr size, GLintptr alignment) {
-    ASSERT(size <= buffer_size);
-    ASSERT(alignment <= buffer_size);
-    mapped_size = size;
+OrphanBuffer::~OrphanBuffer() {
+    Release();
+}
 
-    if (alignment > 0) {
-        buffer_pos = Common::AlignUp<size_t>(buffer_pos, alignment);
+void OrphanBuffer::Create(size_t size, size_t /*sync_subdivide*/) {
+    buffer_pos = 0;
+    buffer_size = size;
+    data.resize(buffer_size);
+
+    if (gl_buffer.handle == 0) {
+        gl_buffer.Create();
+        glBindBuffer(gl_target, gl_buffer.handle);
     }
 
-    bool invalidate = false;
+    glBufferData(gl_target, static_cast<GLsizeiptr>(buffer_size), nullptr, GL_STREAM_DRAW);
+}
+
+void OrphanBuffer::Release() {
+    gl_buffer.Release();
+}
+
+std::pair<u8*, GLintptr> OrphanBuffer::Map(size_t size, size_t alignment) {
+    buffer_pos = Common::AlignUp(buffer_pos, alignment);
+
     if (buffer_pos + size > buffer_size) {
-        buffer_pos = 0;
-        invalidate = true;
+        Create(std::max(buffer_size, size), 0);
+    }
 
-        if (persistent) {
-            glUnmapBuffer(gl_target);
+    mapped_size = size;
+    return std::make_pair(&data[buffer_pos], static_cast<GLintptr>(buffer_pos));
+}
+
+void OrphanBuffer::Unmap() {
+    glBufferSubData(gl_target, static_cast<GLintptr>(buffer_pos),
+                    static_cast<GLsizeiptr>(mapped_size), &data[buffer_pos]);
+    buffer_pos += mapped_size;
+}
+
+StorageBuffer::~StorageBuffer() {
+    Release();
+}
+
+void StorageBuffer::Create(size_t size, size_t sync_subdivide) {
+    if (gl_buffer.handle != 0)
+        return;
+
+    buffer_pos = 0;
+    buffer_size = size;
+    buffer_sync_subdivide = std::max<size_t>(sync_subdivide, 1);
+
+    gl_buffer.Create();
+    glBindBuffer(gl_target, gl_buffer.handle);
+
+    glBufferStorage(gl_target, static_cast<GLsizeiptr>(buffer_size), nullptr,
+                    GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT);
+    mapped_ptr = reinterpret_cast<u8*>(
+        glMapBufferRange(gl_target, 0, static_cast<GLsizeiptr>(buffer_size),
+                         GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | GL_MAP_FLUSH_EXPLICIT_BIT));
+}
+
+void StorageBuffer::Release() {
+    if (gl_buffer.handle == 0)
+        return;
+
+    glUnmapBuffer(gl_target);
+
+    gl_buffer.Release();
+    head.clear();
+    tail.clear();
+}
+
+std::pair<u8*, GLintptr> StorageBuffer::Map(size_t size, size_t alignment) {
+    ASSERT(size <= buffer_size);
+
+    OGLSync sync;
+
+    buffer_pos = Common::AlignUp(buffer_pos, alignment);
+    size_t effective_offset = Common::AlignDown(buffer_pos, buffer_sync_subdivide);
+
+    if (!head.empty() &&
+        (effective_offset > head.back().offset || buffer_pos + size > buffer_size)) {
+        ASSERT(head.back().sync.handle == 0);
+        head.back().sync.Create();
+    }
+
+    if (buffer_pos + size > buffer_size) {
+        if (!tail.empty()) {
+            std::swap(sync, tail.back().sync);
+            tail.clear();
         }
+        std::swap(tail, head);
+        buffer_pos = 0;
+        effective_offset = 0;
     }
 
-    if (invalidate | !persistent) {
-        GLbitfield flags = GL_MAP_WRITE_BIT | (persistent ? GL_MAP_PERSISTENT_BIT : 0) |
-                           (coherent ? GL_MAP_COHERENT_BIT : GL_MAP_FLUSH_EXPLICIT_BIT) |
-                           (invalidate ? GL_MAP_INVALIDATE_BUFFER_BIT : GL_MAP_UNSYNCHRONIZED_BIT);
-        mapped_ptr = static_cast<u8*>(
-            glMapBufferRange(gl_target, buffer_pos, buffer_size - buffer_pos, flags));
-        mapped_offset = buffer_pos;
+    while (!tail.empty() && buffer_pos + size > tail.front().offset) {
+        std::swap(sync, tail.front().sync);
+        tail.pop_front();
     }
 
-    return std::make_tuple(mapped_ptr + buffer_pos - mapped_offset, buffer_pos, invalidate);
+    if (sync.handle != 0) {
+        glClientWaitSync(sync.handle, GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED);
+        sync.Release();
+    }
+
+    if (head.empty() || effective_offset > head.back().offset) {
+        head.emplace_back();
+        head.back().offset = effective_offset;
+    }
+
+    mapped_size = size;
+    return std::make_pair(&mapped_ptr[buffer_pos], static_cast<GLintptr>(buffer_pos));
 }
 
-void OGLStreamBuffer::Unmap(GLsizeiptr size) {
-    ASSERT(size <= mapped_size);
-
-    if (!coherent && size > 0) {
-        glFlushMappedBufferRange(gl_target, buffer_pos - mapped_offset, size);
-    }
-
-    if (!persistent) {
-        glUnmapBuffer(gl_target);
-    }
-
-    buffer_pos += size;
+void StorageBuffer::Unmap() {
+    glFlushMappedBufferRange(gl_target, static_cast<GLintptr>(buffer_pos),
+                             static_cast<GLsizeiptr>(mapped_size));
+    buffer_pos += mapped_size;
 }

src/video_core/renderer_opengl/gl_stream_buffer.h

@@ -2,41 +2,35 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <tuple>
+#pragma once
+
+#include <memory>
 #include <glad/glad.h>
 #include "common/common_types.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 
 class OGLStreamBuffer : private NonCopyable {
 public:
-    explicit OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent = false);
-    ~OGLStreamBuffer();
+    explicit OGLStreamBuffer(GLenum target);
+    virtual ~OGLStreamBuffer() = default;
+
+public:
+    static std::unique_ptr<OGLStreamBuffer> MakeBuffer(bool storage_buffer, GLenum target);
+
+    virtual void Create(size_t size, size_t sync_subdivide) = 0;
+    virtual void Release() {}
 
     GLuint GetHandle() const;
-    GLsizeiptr GetSize() const;
 
-    /*
-     * Allocates a linear chunk of memory in the GPU buffer with at least "size" bytes
-     * and the optional alignment requirement.
-     * If the buffer is full, the whole buffer is reallocated which invalidates old chunks.
-     * The return values are the pointer to the new chunk, the offset within the buffer,
-     * and the invalidation flag for previous chunks.
-     * The actual used size must be specified on unmapping the chunk.
-     */
-    std::tuple<u8*, GLintptr, bool> Map(GLsizeiptr size, GLintptr alignment = 0);
+    virtual std::pair<u8*, GLintptr> Map(size_t size, size_t alignment) = 0;
+    virtual void Unmap() = 0;
 
-    void Unmap(GLsizeiptr size);
-
-private:
+protected:
     OGLBuffer gl_buffer;
     GLenum gl_target;
 
-    bool coherent = false;
-    bool persistent = false;
-
-    GLintptr buffer_pos = 0;
-    GLsizeiptr buffer_size = 0;
-    GLintptr mapped_offset = 0;
-    GLsizeiptr mapped_size = 0;
-    u8* mapped_ptr = nullptr;
+    size_t buffer_pos = 0;
+    size_t buffer_size = 0;
+    size_t buffer_sync_subdivide = 0;
+    size_t mapped_size = 0;
 };

src/video_core/renderer_opengl/maxwell_to_gl.h

@@ -91,8 +91,6 @@ inline GLenum PrimitiveTopology(Maxwell::PrimitiveTopology topology) {
     switch (topology) {
     case Maxwell::PrimitiveTopology::Points:
         return GL_POINTS;
-    case Maxwell::PrimitiveTopology::LineStrip:
-        return GL_LINE_STRIP;
     case Maxwell::PrimitiveTopology::Triangles:
         return GL_TRIANGLES;
     case Maxwell::PrimitiveTopology::TriangleStrip:

src/yuzu/game_list.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <regex>
 #include <QApplication>
 #include <QDir>
 #include <QFileInfo>
@@ -402,12 +403,72 @@ void GameList::RefreshGameDirectory() {
     }
 }
 
-void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsigned int recursion) {
-    boost::container::flat_map<u64, std::shared_ptr<FileSys::NCA>> nca_control_map;
+static void GetMetadataFromControlNCA(const std::shared_ptr<FileSys::NCA>& nca,
+                                      std::vector<u8>& icon, std::string& name) {
+    const auto control_dir = FileSys::ExtractRomFS(nca->GetRomFS());
+    if (control_dir == nullptr)
+        return;
 
-    const auto nca_control_callback =
-        [this, &nca_control_map](u64* num_entries_out, const std::string& directory,
-                                 const std::string& virtual_name) -> bool {
+    const auto nacp_file = control_dir->GetFile("control.nacp");
+    if (nacp_file == nullptr)
+        return;
+    FileSys::NACP nacp(nacp_file);
+    name = nacp.GetApplicationName();
+
+    FileSys::VirtualFile icon_file = nullptr;
+    for (const auto& language : FileSys::LANGUAGE_NAMES) {
+        icon_file = control_dir->GetFile("icon_" + std::string(language) + ".dat");
+        if (icon_file != nullptr) {
+            icon = icon_file->ReadAllBytes();
+            break;
+        }
+    }
+}
+
+void GameListWorker::AddInstalledTitlesToGameList() {
+    const auto usernand = Service::FileSystem::GetUserNANDContents();
+    const auto installed_games = usernand->ListEntriesFilter(FileSys::TitleType::Application,
+                                                             FileSys::ContentRecordType::Program);
+
+    for (const auto& game : installed_games) {
+        const auto& file = usernand->GetEntryRaw(game);
+        std::unique_ptr<Loader::AppLoader> loader = Loader::GetLoader(file);
+        if (!loader)
+            continue;
+
+        std::vector<u8> icon;
+        std::string name;
+        u64 program_id;
+        loader->ReadProgramId(program_id);
+
+        const auto& control =
+            usernand->GetEntry(game.title_id, FileSys::ContentRecordType::Control);
+        if (control != nullptr)
+            GetMetadataFromControlNCA(control, icon, name);
+        emit EntryReady({
+            new GameListItemPath(
+                FormatGameName(file->GetFullPath()), icon, QString::fromStdString(name),
+                QString::fromStdString(Loader::GetFileTypeString(loader->GetFileType())),
+                program_id),
+            new GameListItem(
+                QString::fromStdString(Loader::GetFileTypeString(loader->GetFileType()))),
+            new GameListItemSize(file->GetSize()),
+        });
+    }
+
+    const auto control_data = usernand->ListEntriesFilter(FileSys::TitleType::Application,
+                                                          FileSys::ContentRecordType::Control);
+
+    for (const auto& entry : control_data) {
+        const auto nca = usernand->GetEntry(entry);
+        if (nca != nullptr)
+            nca_control_map.insert_or_assign(entry.title_id, nca);
+    }
+}
+
+void GameListWorker::FillControlMap(const std::string& dir_path) {
+    const auto nca_control_callback = [this](u64* num_entries_out, const std::string& directory,
+                                             const std::string& virtual_name) -> bool {
         std::string physical_name = directory + DIR_SEP + virtual_name;
 
         if (stop_processing)
@@ -425,10 +486,11 @@ void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsign
     };
 
     FileUtil::ForeachDirectoryEntry(nullptr, dir_path, nca_control_callback);
+}
 
-    const auto callback = [this, recursion,
-                           &nca_control_map](u64* num_entries_out, const std::string& directory,
-                                             const std::string& virtual_name) -> bool {
+void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsigned int recursion) {
+    const auto callback = [this, recursion](u64* num_entries_out, const std::string& directory,
+                                            const std::string& virtual_name) -> bool {
         std::string physical_name = directory + DIR_SEP + virtual_name;
 
         if (stop_processing)
@@ -458,20 +520,7 @@ void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsign
                 // Use from metadata pool.
                 if (nca_control_map.find(program_id) != nca_control_map.end()) {
                     const auto nca = nca_control_map[program_id];
-                    const auto control_dir = FileSys::ExtractRomFS(nca->GetRomFS());
-
-                    const auto nacp_file = control_dir->GetFile("control.nacp");
-                    FileSys::NACP nacp(nacp_file);
-                    name = nacp.GetApplicationName();
-
-                    FileSys::VirtualFile icon_file = nullptr;
-                    for (const auto& language : FileSys::LANGUAGE_NAMES) {
-                        icon_file = control_dir->GetFile("icon_" + std::string(language) + ".dat");
-                        if (icon_file != nullptr) {
-                            icon = icon_file->ReadAllBytes();
-                            break;
-                        }
-                    }
+                    GetMetadataFromControlNCA(nca, icon, name);
                 }
             }
 
@@ -498,7 +547,10 @@ void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsign
 void GameListWorker::run() {
     stop_processing = false;
     watch_list.append(dir_path);
+    FillControlMap(dir_path.toStdString());
+    AddInstalledTitlesToGameList();
     AddFstEntriesToGameList(dir_path.toStdString(), deep_scan ? 256 : 0);
+    nca_control_map.clear();
     emit Finished(watch_list);
 }
 

src/yuzu/game_list_p.h

@@ -163,10 +163,13 @@ signals:
 
 private:
     FileSys::VirtualFilesystem vfs;
+    std::map<u64, std::shared_ptr<FileSys::NCA>> nca_control_map;
     QStringList watch_list;
     QString dir_path;
     bool deep_scan;
     std::atomic_bool stop_processing;
 
+    void AddInstalledTitlesToGameList();
+    void FillControlMap(const std::string& dir_path);
     void AddFstEntriesToGameList(const std::string& dir_path, unsigned int recursion = 0);
 };

src/yuzu/main.cpp

@@ -24,6 +24,7 @@
 #include "common/string_util.h"
 #include "core/core.h"
 #include "core/crypto/key_manager.h"
+#include "core/file_sys/card_image.h"
 #include "core/file_sys/vfs_real.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/loader/loader.h"
@@ -114,6 +115,9 @@ GMainWindow::GMainWindow()
                        .arg(Common::g_build_name, Common::g_scm_branch, Common::g_scm_desc));
     show();
 
+    // Necessary to load titles from nand in gamelist.
+    Service::FileSystem::RegisterBIS(std::make_unique<FileSys::BISFactory>(vfs->OpenDirectory(
+        FileUtil::GetUserPath(FileUtil::UserPath::NANDDir), FileSys::Mode::ReadWrite)));
     game_list->PopulateAsync(UISettings::values.gamedir, UISettings::values.gamedir_deepscan);
 
     // Show one-time "callout" messages to the user
@@ -309,6 +313,8 @@ void GMainWindow::ConnectMenuEvents() {
     // File
     connect(ui.action_Load_File, &QAction::triggered, this, &GMainWindow::OnMenuLoadFile);
     connect(ui.action_Load_Folder, &QAction::triggered, this, &GMainWindow::OnMenuLoadFolder);
+    connect(ui.action_Install_File_NAND, &QAction::triggered, this,
+            &GMainWindow::OnMenuInstallToNAND);
     connect(ui.action_Select_Game_List_Root, &QAction::triggered, this,
             &GMainWindow::OnMenuSelectGameListRoot);
     connect(ui.action_Exit, &QAction::triggered, this, &QMainWindow::close);
@@ -612,6 +618,143 @@ void GMainWindow::OnMenuLoadFolder() {
     }
 }
 
+void GMainWindow::OnMenuInstallToNAND() {
+    const QString file_filter =
+        tr("Installable Switch File (*.nca *.xci);;Nintendo Content Archive (*.nca);;NX Cartridge "
+           "Image (*.xci)");
+    QString filename = QFileDialog::getOpenFileName(this, tr("Install File"),
+                                                    UISettings::values.roms_path, file_filter);
+
+    const auto qt_raw_copy = [this](FileSys::VirtualFile src, FileSys::VirtualFile dest) {
+        if (src == nullptr || dest == nullptr)
+            return false;
+        if (!dest->Resize(src->GetSize()))
+            return false;
+
+        QProgressDialog progress(fmt::format("Installing file \"{}\"...", src->GetName()).c_str(),
+                                 "Cancel", 0, src->GetSize() / 0x1000, this);
+        progress.setWindowModality(Qt::WindowModal);
+
+        std::array<u8, 0x1000> buffer{};
+        for (size_t i = 0; i < src->GetSize(); i += 0x1000) {
+            if (progress.wasCanceled()) {
+                dest->Resize(0);
+                return false;
+            }
+
+            progress.setValue(i / 0x1000);
+            const auto read = src->Read(buffer.data(), buffer.size(), i);
+            dest->Write(buffer.data(), read, i);
+        }
+
+        return true;
+    };
+
+    const auto success = [this]() {
+        QMessageBox::information(this, tr("Successfully Installed"),
+                                 tr("The file was successfully installed."));
+        game_list->PopulateAsync(UISettings::values.gamedir, UISettings::values.gamedir_deepscan);
+    };
+
+    const auto failed = [this]() {
+        QMessageBox::warning(
+            this, tr("Failed to Install"),
+            tr("There was an error while attempting to install the provided file. It "
+               "could have an incorrect format or be missing metadata. Please "
+               "double-check your file and try again."));
+    };
+
+    const auto overwrite = [this]() {
+        return QMessageBox::question(this, "Failed to Install",
+                                     "The file you are attempting to install already exists "
+                                     "in the cache. Would you like to overwrite it?") ==
+               QMessageBox::Yes;
+    };
+
+    if (!filename.isEmpty()) {
+        if (filename.endsWith("xci", Qt::CaseInsensitive)) {
+            const auto xci = std::make_shared<FileSys::XCI>(
+                vfs->OpenFile(filename.toStdString(), FileSys::Mode::Read));
+            if (xci->GetStatus() != Loader::ResultStatus::Success) {
+                failed();
+                return;
+            }
+            const auto res =
+                Service::FileSystem::GetUserNANDContents()->InstallEntry(xci, false, qt_raw_copy);
+            if (res == FileSys::InstallResult::Success) {
+                success();
+            } else {
+                if (res == FileSys::InstallResult::ErrorAlreadyExists) {
+                    if (overwrite()) {
+                        const auto res2 = Service::FileSystem::GetUserNANDContents()->InstallEntry(
+                            xci, true, qt_raw_copy);
+                        if (res2 == FileSys::InstallResult::Success) {
+                            success();
+                        } else {
+                            failed();
+                        }
+                    }
+                } else {
+                    failed();
+                }
+            }
+        } else {
+            const auto nca = std::make_shared<FileSys::NCA>(
+                vfs->OpenFile(filename.toStdString(), FileSys::Mode::Read));
+            if (nca->GetStatus() != Loader::ResultStatus::Success) {
+                failed();
+                return;
+            }
+
+            static const QStringList tt_options{"System Application",
+                                                "System Archive",
+                                                "System Application Update",
+                                                "Firmware Package (Type A)",
+                                                "Firmware Package (Type B)",
+                                                "Game",
+                                                "Game Update",
+                                                "Game DLC",
+                                                "Delta Title"};
+            bool ok;
+            const auto item = QInputDialog::getItem(
+                this, tr("Select NCA Install Type..."),
+                tr("Please select the type of title you would like to install this NCA as:\n(In "
+                   "most instances, the default 'Game' is fine.)"),
+                tt_options, 5, false, &ok);
+
+            auto index = tt_options.indexOf(item);
+            if (!ok || index == -1) {
+                QMessageBox::warning(this, tr("Failed to Install"),
+                                     tr("The title type you selected for the NCA is invalid."));
+                return;
+            }
+
+            if (index >= 5)
+                index += 0x7B;
+
+            const auto res = Service::FileSystem::GetUserNANDContents()->InstallEntry(
+                nca, static_cast<FileSys::TitleType>(index), false, qt_raw_copy);
+            if (res == FileSys::InstallResult::Success) {
+                success();
+            } else {
+                if (res == FileSys::InstallResult::ErrorAlreadyExists) {
+                    if (overwrite()) {
+                        const auto res2 = Service::FileSystem::GetUserNANDContents()->InstallEntry(
+                            nca, static_cast<FileSys::TitleType>(index), true, qt_raw_copy);
+                        if (res2 == FileSys::InstallResult::Success) {
+                            success();
+                        } else {
+                            failed();
+                        }
+                    }
+                } else {
+                    failed();
+                }
+            }
+        }
+    }
+}
+
 void GMainWindow::OnMenuSelectGameListRoot() {
     QString dir_path = QFileDialog::getExistingDirectory(this, tr("Select Directory"));
     if (!dir_path.isEmpty()) {

src/yuzu/main.h

@@ -125,6 +125,7 @@ private slots:
     void OnGameListOpenSaveFolder(u64 program_id);
     void OnMenuLoadFile();
     void OnMenuLoadFolder();
+    void OnMenuInstallToNAND();
     /// Called whenever a user selects the "File->Select Game List Root" menu item
     void OnMenuSelectGameListRoot();
     void OnMenuRecentFile();

src/yuzu/main.ui

@@ -57,6 +57,8 @@
       <string>Recent Files</string>
      </property>
     </widget>
+     <addaction name="action_Install_File_NAND" />
+     <addaction name="separator"/>
     <addaction name="action_Load_File"/>
     <addaction name="action_Load_Folder"/>
     <addaction name="separator"/>
@@ -102,6 +104,11 @@
    <addaction name="menu_View"/>
    <addaction name="menu_Help"/>
   </widget>
+   <action name="action_Install_File_NAND">
+     <property name="text">
+       <string>Install File to NAND...</string>
+     </property>
+   </action>
   <action name="action_Load_File">
    <property name="text">
     <string>Load File...</string>