"Merge PR 1340"

Implement a MME Draw commands Inliner and correct host instance drawing

README.md

@@ -6,9 +6,11 @@ yuzu emulator
 
 yuzu is an experimental open-source emulator for the Nintendo Switch from the creators of [Citra](https://citra-emu.org/).
 
-It is written in C++ with portability in mind, with builds actively maintained for Windows, Linux and macOS. The emulator is currently only useful for homebrew development and research purposes.
+It is written in C++ with portability in mind, with builds actively maintained for Windows and Linux. The emulator is capable of running several commercial games.
 
-yuzu only emulates a subset of Switch hardware and therefore is generally only useful for running/debugging homebrew applications. yuzu can boot some games, to varying degrees of success.
+yuzu only emulates a subset of Switch hardware and therefore most commercial games **do not** run at full speed or are not fully functional.
+
+Do you want to check which games are compatible and which ones are not? Please visit our [Compatibility page](https://yuzu-emu.org/game/)!
 
 yuzu is licensed under the GPLv2 (or any later version). Refer to the license.txt file included.
 

src/common/assert.h

@@ -28,18 +28,14 @@ __declspec(noinline, noreturn)
 }
 
 #define ASSERT(_a_)                                                                                \
-    do                                                                                             \
-        if (!(_a_)) {                                                                              \
-            assert_noinline_call([] { LOG_CRITICAL(Debug, "Assertion Failed!"); });                \
-        }                                                                                          \
-    while (0)
+    if (!(_a_)) {                                                                                  \
+        LOG_CRITICAL(Debug, "Assertion Failed!");                                                  \
+    }
 
 #define ASSERT_MSG(_a_, ...)                                                                       \
-    do                                                                                             \
-        if (!(_a_)) {                                                                              \
-            assert_noinline_call([&] { LOG_CRITICAL(Debug, "Assertion Failed!\n" __VA_ARGS__); }); \
-        }                                                                                          \
-    while (0)
+    if (!(_a_)) {                                                                                  \
+        LOG_CRITICAL(Debug, "Assertion Failed! " __VA_ARGS__);                                     \
+    }
 
 #define UNREACHABLE() ASSERT_MSG(false, "Unreachable code!")
 #define UNREACHABLE_MSG(...) ASSERT_MSG(false, __VA_ARGS__)

src/common/logging/backend.cpp

@@ -255,6 +255,7 @@ void DebuggerBackend::Write(const Entry& entry) {
     CLS(Input)                                                                                     \
     CLS(Network)                                                                                   \
     CLS(Loader)                                                                                    \
+    CLS(CheatEngine)                                                                               \
     CLS(Crypto)                                                                                    \
     CLS(WebService)
 

src/common/logging/log.h

@@ -117,6 +117,7 @@ enum class Class : ClassType {
     Audio_DSP,         ///< The HLE implementation of the DSP
     Audio_Sink,        ///< Emulator audio output backend
     Loader,            ///< ROM loader
+    CheatEngine,       ///< Memory manipulation and engine VM functions
     Crypto,            ///< Cryptographic engine/functions
     Input,             ///< Input emulation
     Network,           ///< Network emulation

src/core/CMakeLists.txt

@@ -33,8 +33,6 @@ add_library(core STATIC
     file_sys/bis_factory.h
     file_sys/card_image.cpp
     file_sys/card_image.h
-    file_sys/cheat_engine.cpp
-    file_sys/cheat_engine.h
     file_sys/content_archive.cpp
     file_sys/content_archive.h
     file_sys/control_metadata.cpp
@@ -477,6 +475,11 @@ add_library(core STATIC
     loader/nsp.h
     loader/xci.cpp
     loader/xci.h
+    memory/cheat_engine.cpp
+    memory/cheat_engine.h
+    memory/dmnt_cheat_types.h
+    memory/dmnt_cheat_vm.cpp
+    memory/dmnt_cheat_vm.h
     memory.cpp
     memory.h
     memory_setup.h

src/core/core.cpp

@@ -14,8 +14,14 @@
 #include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/cpu_core_manager.h"
+#include "core/file_sys/bis_factory.h"
+#include "core/file_sys/card_image.h"
 #include "core/file_sys/mode.h"
+#include "core/file_sys/patch_manager.h"
 #include "core/file_sys/registered_cache.h"
+#include "core/file_sys/romfs_factory.h"
+#include "core/file_sys/savedata_factory.h"
+#include "core/file_sys/sdmc_factory.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
 #include "core/gdbstub/gdbstub.h"
@@ -27,17 +33,17 @@
 #include "core/hle/kernel/thread.h"
 #include "core/hle/service/am/applets/applets.h"
 #include "core/hle/service/apm/controller.h"
+#include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/glue/manager.h"
 #include "core/hle/service/service.h"
 #include "core/hle/service/sm/sm.h"
 #include "core/loader/loader.h"
+#include "core/memory/cheat_engine.h"
 #include "core/perf_stats.h"
 #include "core/reporter.h"
 #include "core/settings.h"
 #include "core/telemetry_session.h"
 #include "core/tools/freezer.h"
-#include "file_sys/cheat_engine.h"
-#include "file_sys/patch_manager.h"
 #include "video_core/debug_utils/debug_utils.h"
 #include "video_core/renderer_base.h"
 #include "video_core/video_core.h"
@@ -160,10 +166,6 @@ struct System::Impl {
 
         LOG_DEBUG(Core, "Initialized OK");
 
-        // Reset counters and set time origin to current frame
-        GetAndResetPerfStats();
-        perf_stats.BeginSystemFrame();
-
         return ResultStatus::Success;
     }
 
@@ -202,10 +204,34 @@ struct System::Impl {
         gpu_core->Start();
         cpu_core_manager.StartThreads();
 
+        // Initialize cheat engine
+        if (cheat_engine) {
+            cheat_engine->Initialize();
+        }
+
         // All threads are started, begin main process execution, now that we're in the clear.
         main_process->Run(load_parameters->main_thread_priority,
                           load_parameters->main_thread_stack_size);
 
+        if (Settings::values.gamecard_inserted) {
+            if (Settings::values.gamecard_current_game) {
+                fs_controller.SetGameCard(GetGameFileFromPath(virtual_filesystem, filepath));
+            } else if (!Settings::values.gamecard_path.empty()) {
+                fs_controller.SetGameCard(
+                    GetGameFileFromPath(virtual_filesystem, Settings::values.gamecard_path));
+            }
+        }
+
+        u64 title_id{0};
+        if (app_loader->ReadProgramId(title_id) != Loader::ResultStatus::Success) {
+            LOG_ERROR(Core, "Failed to find title id for ROM (Error {})",
+                      static_cast<u32>(load_result));
+        }
+        perf_stats = std::make_unique<PerfStats>(title_id);
+        // Reset counters and set time origin to current frame
+        GetAndResetPerfStats();
+        perf_stats->BeginSystemFrame();
+
         status = ResultStatus::Success;
         return status;
     }
@@ -219,6 +245,8 @@ struct System::Impl {
                                     perf_results.game_fps);
         telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Frametime",
                                     perf_results.frametime * 1000.0);
+        telemetry_session->AddField(Telemetry::FieldType::Performance, "Mean_Frametime_MS",
+                                    perf_stats->GetMeanFrametime());
 
         is_powered_on = false;
 
@@ -229,6 +257,7 @@ struct System::Impl {
         service_manager.reset();
         cheat_engine.reset();
         telemetry_session.reset();
+        perf_stats.reset();
         gpu_core.reset();
 
         // Close all CPU/threading state
@@ -286,7 +315,7 @@ struct System::Impl {
     }
 
     PerfStatsResults GetAndResetPerfStats() {
-        return perf_stats.GetAndResetStats(core_timing.GetGlobalTimeUs());
+        return perf_stats->GetAndResetStats(core_timing.GetGlobalTimeUs());
     }
 
     Timing::CoreTiming core_timing;
@@ -295,6 +324,7 @@ struct System::Impl {
     FileSys::VirtualFilesystem virtual_filesystem;
     /// ContentProviderUnion instance
     std::unique_ptr<FileSys::ContentProviderUnion> content_provider;
+    Service::FileSystem::FileSystemController fs_controller;
     /// AppLoader used to load the current executing application
     std::unique_ptr<Loader::AppLoader> app_loader;
     std::unique_ptr<VideoCore::RendererBase> renderer;
@@ -304,7 +334,7 @@ struct System::Impl {
     CpuCoreManager cpu_core_manager;
     bool is_powered_on = false;
 
-    std::unique_ptr<FileSys::CheatEngine> cheat_engine;
+    std::unique_ptr<Memory::CheatEngine> cheat_engine;
     std::unique_ptr<Tools::Freezer> memory_freezer;
 
     /// Frontend applets
@@ -327,7 +357,7 @@ struct System::Impl {
     ResultStatus status = ResultStatus::Success;
     std::string status_details = "";
 
-    Core::PerfStats perf_stats;
+    std::unique_ptr<Core::PerfStats> perf_stats;
     Core::FrameLimiter frame_limiter;
 };
 
@@ -480,11 +510,11 @@ const Timing::CoreTiming& System::CoreTiming() const {
 }
 
 Core::PerfStats& System::GetPerfStats() {
-    return impl->perf_stats;
+    return *impl->perf_stats;
 }
 
 const Core::PerfStats& System::GetPerfStats() const {
-    return impl->perf_stats;
+    return *impl->perf_stats;
 }
 
 Core::FrameLimiter& System::FrameLimiter() {
@@ -519,13 +549,6 @@ Tegra::DebugContext* System::GetGPUDebugContext() const {
     return impl->debug_context.get();
 }
 
-void System::RegisterCheatList(const std::vector<FileSys::CheatList>& list,
-                               const std::string& build_id, VAddr code_region_start,
-                               VAddr code_region_end) {
-    impl->cheat_engine = std::make_unique<FileSys::CheatEngine>(*this, list, build_id,
-                                                                code_region_start, code_region_end);
-}
-
 void System::SetFilesystem(std::shared_ptr<FileSys::VfsFilesystem> vfs) {
     impl->virtual_filesystem = std::move(vfs);
 }
@@ -534,6 +557,13 @@ std::shared_ptr<FileSys::VfsFilesystem> System::GetFilesystem() const {
     return impl->virtual_filesystem;
 }
 
+void System::RegisterCheatList(const std::vector<Memory::CheatEntry>& list,
+                               const std::array<u8, 32>& build_id, VAddr main_region_begin,
+                               u64 main_region_size) {
+    impl->cheat_engine = std::make_unique<Memory::CheatEngine>(*this, list, build_id);
+    impl->cheat_engine->SetMainMemoryParameters(main_region_begin, main_region_size);
+}
+
 void System::SetAppletFrontendSet(Service::AM::Applets::AppletFrontendSet&& set) {
     impl->applet_manager.SetAppletFrontendSet(std::move(set));
 }
@@ -562,6 +592,14 @@ const FileSys::ContentProvider& System::GetContentProvider() const {
     return *impl->content_provider;
 }
 
+Service::FileSystem::FileSystemController& System::GetFileSystemController() {
+    return impl->fs_controller;
+}
+
+const Service::FileSystem::FileSystemController& System::GetFileSystemController() const {
+    return impl->fs_controller;
+}
+
 void System::RegisterContentProvider(FileSys::ContentProviderUnionSlot slot,
                                      FileSys::ContentProvider* provider) {
     impl->content_provider->SetSlot(slot, provider);

src/core/core.h

@@ -18,7 +18,6 @@ class EmuWindow;
 } // namespace Core::Frontend
 
 namespace FileSys {
-class CheatList;
 class ContentProvider;
 class ContentProviderUnion;
 enum class ContentProviderUnionSlot;
@@ -36,6 +35,10 @@ class AppLoader;
 enum class ResultStatus : u16;
 } // namespace Loader
 
+namespace Memory {
+struct CheatEntry;
+} // namespace Memory
+
 namespace Service {
 
 namespace AM::Applets {
@@ -47,6 +50,10 @@ namespace APM {
 class Controller;
 }
 
+namespace FileSystem {
+class FileSystemController;
+} // namespace FileSystem
+
 namespace Glue {
 class ARPManager;
 }
@@ -282,8 +289,9 @@ public:
 
     std::shared_ptr<FileSys::VfsFilesystem> GetFilesystem() const;
 
-    void RegisterCheatList(const std::vector<FileSys::CheatList>& list, const std::string& build_id,
-                           VAddr code_region_start, VAddr code_region_end);
+    void RegisterCheatList(const std::vector<Memory::CheatEntry>& list,
+                           const std::array<u8, 0x20>& build_id, VAddr main_region_begin,
+                           u64 main_region_size);
 
     void SetAppletFrontendSet(Service::AM::Applets::AppletFrontendSet&& set);
 
@@ -299,6 +307,10 @@ public:
 
     const FileSys::ContentProvider& GetContentProvider() const;
 
+    Service::FileSystem::FileSystemController& GetFileSystemController();
+
+    const Service::FileSystem::FileSystemController& GetFileSystemController() const;
+
     void RegisterContentProvider(FileSys::ContentProviderUnionSlot slot,
                                  FileSys::ContentProvider* provider);
 

src/core/crypto/partition_data_manager.cpp

@@ -480,6 +480,10 @@ void PartitionDataManager::DecryptProdInfo(std::array<u8, 0x20> bis_key) {
     prodinfo_decrypted = std::make_shared<XTSEncryptionLayer>(prodinfo, bis_key);
 }
 
+FileSys::VirtualFile PartitionDataManager::GetDecryptedProdInfo() const {
+    return prodinfo_decrypted;
+}
+
 std::array<u8, 576> PartitionDataManager::GetETicketExtendedKek() const {
     std::array<u8, 0x240> out{};
     if (prodinfo_decrypted != nullptr)

src/core/crypto/partition_data_manager.h

@@ -84,6 +84,7 @@ public:
     bool HasProdInfo() const;
     FileSys::VirtualFile GetProdInfoRaw() const;
     void DecryptProdInfo(std::array<u8, 0x20> bis_key);
+    FileSys::VirtualFile GetDecryptedProdInfo() const;
     std::array<u8, 0x240> GetETicketExtendedKek() const;
 
 private:

src/core/file_sys/bis_factory.cpp

@@ -3,8 +3,12 @@
 // Refer to the license.txt file included.
 
 #include <fmt/format.h>
+#include "common/file_util.h"
+#include "core/core.h"
 #include "core/file_sys/bis_factory.h"
+#include "core/file_sys/mode.h"
 #include "core/file_sys/registered_cache.h"
+#include "core/settings.h"
 
 namespace FileSys {
 
@@ -14,10 +18,22 @@ BISFactory::BISFactory(VirtualDir nand_root_, VirtualDir load_root_, VirtualDir
       sysnand_cache(std::make_unique<RegisteredCache>(
           GetOrCreateDirectoryRelative(nand_root, "/system/Contents/registered"))),
       usrnand_cache(std::make_unique<RegisteredCache>(
-          GetOrCreateDirectoryRelative(nand_root, "/user/Contents/registered"))) {}
+          GetOrCreateDirectoryRelative(nand_root, "/user/Contents/registered"))),
+      sysnand_placeholder(std::make_unique<PlaceholderCache>(
+          GetOrCreateDirectoryRelative(nand_root, "/system/Contents/placehld"))),
+      usrnand_placeholder(std::make_unique<PlaceholderCache>(
+          GetOrCreateDirectoryRelative(nand_root, "/user/Contents/placehld"))) {}
 
 BISFactory::~BISFactory() = default;
 
+VirtualDir BISFactory::GetSystemNANDContentDirectory() const {
+    return GetOrCreateDirectoryRelative(nand_root, "/system/Contents");
+}
+
+VirtualDir BISFactory::GetUserNANDContentDirectory() const {
+    return GetOrCreateDirectoryRelative(nand_root, "/user/Contents");
+}
+
 RegisteredCache* BISFactory::GetSystemNANDContents() const {
     return sysnand_cache.get();
 }
@@ -26,9 +42,17 @@ RegisteredCache* BISFactory::GetUserNANDContents() const {
     return usrnand_cache.get();
 }
 
+PlaceholderCache* BISFactory::GetSystemNANDPlaceholder() const {
+    return sysnand_placeholder.get();
+}
+
+PlaceholderCache* BISFactory::GetUserNANDPlaceholder() const {
+    return usrnand_placeholder.get();
+}
+
 VirtualDir BISFactory::GetModificationLoadRoot(u64 title_id) const {
     // LayeredFS doesn't work on updates and title id-less homebrew
-    if (title_id == 0 || (title_id & 0x800) > 0)
+    if (title_id == 0 || (title_id & 0xFFF) == 0x800)
         return nullptr;
     return GetOrCreateDirectoryRelative(load_root, fmt::format("/{:016X}", title_id));
 }
@@ -39,4 +63,77 @@ VirtualDir BISFactory::GetModificationDumpRoot(u64 title_id) const {
     return GetOrCreateDirectoryRelative(dump_root, fmt::format("/{:016X}", title_id));
 }
 
+VirtualDir BISFactory::OpenPartition(BisPartitionId id) const {
+    switch (id) {
+    case BisPartitionId::CalibrationFile:
+        return GetOrCreateDirectoryRelative(nand_root, "/prodinfof");
+    case BisPartitionId::SafeMode:
+        return GetOrCreateDirectoryRelative(nand_root, "/safe");
+    case BisPartitionId::System:
+        return GetOrCreateDirectoryRelative(nand_root, "/system");
+    case BisPartitionId::User:
+        return GetOrCreateDirectoryRelative(nand_root, "/user");
+    default:
+        return nullptr;
+    }
+}
+
+VirtualFile BISFactory::OpenPartitionStorage(BisPartitionId id) const {
+    Core::Crypto::KeyManager keys;
+    Core::Crypto::PartitionDataManager pdm{
+        Core::System::GetInstance().GetFilesystem()->OpenDirectory(
+            FileUtil::GetUserPath(FileUtil::UserPath::SysDataDir), Mode::Read)};
+    keys.PopulateFromPartitionData(pdm);
+
+    switch (id) {
+    case BisPartitionId::CalibrationBinary:
+        return pdm.GetDecryptedProdInfo();
+    case BisPartitionId::BootConfigAndPackage2Part1:
+    case BisPartitionId::BootConfigAndPackage2Part2:
+    case BisPartitionId::BootConfigAndPackage2Part3:
+    case BisPartitionId::BootConfigAndPackage2Part4:
+    case BisPartitionId::BootConfigAndPackage2Part5:
+    case BisPartitionId::BootConfigAndPackage2Part6: {
+        const auto new_id = static_cast<u8>(id) -
+                            static_cast<u8>(BisPartitionId::BootConfigAndPackage2Part1) +
+                            static_cast<u8>(Core::Crypto::Package2Type::NormalMain);
+        return pdm.GetPackage2Raw(static_cast<Core::Crypto::Package2Type>(new_id));
+    }
+    default:
+        return nullptr;
+    }
+}
+
+VirtualDir BISFactory::GetImageDirectory() const {
+    return GetOrCreateDirectoryRelative(nand_root, "/user/Album");
+}
+
+u64 BISFactory::GetSystemNANDFreeSpace() const {
+    const auto sys_dir = GetOrCreateDirectoryRelative(nand_root, "/system");
+    if (sys_dir == nullptr)
+        return 0;
+
+    return GetSystemNANDTotalSpace() - sys_dir->GetSize();
+}
+
+u64 BISFactory::GetSystemNANDTotalSpace() const {
+    return static_cast<u64>(Settings::values.nand_system_size);
+}
+
+u64 BISFactory::GetUserNANDFreeSpace() const {
+    const auto usr_dir = GetOrCreateDirectoryRelative(nand_root, "/user");
+    if (usr_dir == nullptr)
+        return 0;
+
+    return GetUserNANDTotalSpace() - usr_dir->GetSize();
+}
+
+u64 BISFactory::GetUserNANDTotalSpace() const {
+    return static_cast<u64>(Settings::values.nand_user_size);
+}
+
+u64 BISFactory::GetFullNANDTotalSpace() const {
+    return static_cast<u64>(Settings::values.nand_total_size);
+}
+
 } // namespace FileSys

src/core/file_sys/bis_factory.h

@@ -10,7 +10,25 @@
 
 namespace FileSys {
 
+enum class BisPartitionId : u32 {
+    UserDataRoot = 20,
+    CalibrationBinary = 27,
+    CalibrationFile = 28,
+    BootConfigAndPackage2Part1 = 21,
+    BootConfigAndPackage2Part2 = 22,
+    BootConfigAndPackage2Part3 = 23,
+    BootConfigAndPackage2Part4 = 24,
+    BootConfigAndPackage2Part5 = 25,
+    BootConfigAndPackage2Part6 = 26,
+    SafeMode = 29,
+    System = 31,
+    SystemProperEncryption = 32,
+    SystemProperPartition = 33,
+    User = 30,
+};
+
 class RegisteredCache;
+class PlaceholderCache;
 
 /// 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
@@ -20,12 +38,29 @@ public:
     explicit BISFactory(VirtualDir nand_root, VirtualDir load_root, VirtualDir dump_root);
     ~BISFactory();
 
+    VirtualDir GetSystemNANDContentDirectory() const;
+    VirtualDir GetUserNANDContentDirectory() const;
+
     RegisteredCache* GetSystemNANDContents() const;
     RegisteredCache* GetUserNANDContents() const;
 
+    PlaceholderCache* GetSystemNANDPlaceholder() const;
+    PlaceholderCache* GetUserNANDPlaceholder() const;
+
     VirtualDir GetModificationLoadRoot(u64 title_id) const;
     VirtualDir GetModificationDumpRoot(u64 title_id) const;
 
+    VirtualDir OpenPartition(BisPartitionId id) const;
+    VirtualFile OpenPartitionStorage(BisPartitionId id) const;
+
+    VirtualDir GetImageDirectory() const;
+
+    u64 GetSystemNANDFreeSpace() const;
+    u64 GetSystemNANDTotalSpace() const;
+    u64 GetUserNANDFreeSpace() const;
+    u64 GetUserNANDTotalSpace() const;
+    u64 GetFullNANDTotalSpace() const;
+
 private:
     VirtualDir nand_root;
     VirtualDir load_root;
@@ -33,6 +68,9 @@ private:
 
     std::unique_ptr<RegisteredCache> sysnand_cache;
     std::unique_ptr<RegisteredCache> usrnand_cache;
+
+    std::unique_ptr<PlaceholderCache> sysnand_placeholder;
+    std::unique_ptr<PlaceholderCache> usrnand_placeholder;
 };
 
 } // namespace FileSys

src/core/file_sys/card_image.cpp

@@ -12,12 +12,16 @@
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/partition_filesystem.h"
+#include "core/file_sys/romfs.h"
 #include "core/file_sys/submission_package.h"
+#include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_offset.h"
+#include "core/file_sys/vfs_vector.h"
 #include "core/loader/loader.h"
 
 namespace FileSys {
 
+constexpr u64 GAMECARD_CERTIFICATE_OFFSET = 0x7000;
 constexpr std::array partition_names{
     "update",
     "normal",
@@ -175,6 +179,26 @@ VirtualDir XCI::GetParentDirectory() const {
     return file->GetContainingDirectory();
 }
 
+VirtualDir XCI::ConcatenatedPseudoDirectory() {
+    const auto out = std::make_shared<VectorVfsDirectory>();
+    for (const auto& part_id : {XCIPartition::Normal, XCIPartition::Logo, XCIPartition::Secure}) {
+        const auto& part = GetPartition(part_id);
+        if (part == nullptr)
+            continue;
+
+        for (const auto& file : part->GetFiles())
+            out->AddFile(file);
+    }
+
+    return out;
+}
+
+std::array<u8, 0x200> XCI::GetCertificate() const {
+    std::array<u8, 0x200> out;
+    file->Read(out.data(), out.size(), GAMECARD_CERTIFICATE_OFFSET);
+    return out;
+}
+
 Loader::ResultStatus XCI::AddNCAFromPartition(XCIPartition part) {
     const auto partition_index = static_cast<std::size_t>(part);
     const auto& partition = partitions[partition_index];

src/core/file_sys/card_image.h

@@ -91,6 +91,8 @@ public:
     VirtualDir GetLogoPartition() const;
 
     u64 GetProgramTitleID() const;
+    u32 GetSystemUpdateVersion();
+    u64 GetSystemUpdateTitleID() const;
 
     bool HasProgramNCA() const;
     VirtualFile GetProgramNCAFile() const;
@@ -106,6 +108,11 @@ public:
 
     VirtualDir GetParentDirectory() const override;
 
+    // Creates a directory that contains all the NCAs in the gamecard
+    VirtualDir ConcatenatedPseudoDirectory();
+
+    std::array<u8, 0x200> GetCertificate() const;
+
 private:
     Loader::ResultStatus AddNCAFromPartition(XCIPartition part);
 
@@ -120,6 +127,8 @@ private:
     std::shared_ptr<NCA> program;
     std::vector<std::shared_ptr<NCA>> ncas;
 
+    u64 update_normal_partition_end;
+
     Core::Crypto::KeyManager keys;
 };
 } // namespace FileSys

src/core/file_sys/cheat_engine.cpp

@@ -1,492 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <locale>
-#include "common/hex_util.h"
-#include "common/microprofile.h"
-#include "common/swap.h"
-#include "core/core.h"
-#include "core/core_timing.h"
-#include "core/core_timing_util.h"
-#include "core/file_sys/cheat_engine.h"
-#include "core/hle/kernel/process.h"
-#include "core/hle/service/hid/controllers/npad.h"
-#include "core/hle/service/hid/hid.h"
-#include "core/hle/service/sm/sm.h"
-
-namespace FileSys {
-
-constexpr s64 CHEAT_ENGINE_TICKS = static_cast<s64>(Core::Timing::BASE_CLOCK_RATE / 60);
-constexpr u32 KEYPAD_BITMASK = 0x3FFFFFF;
-
-u64 Cheat::Address() const {
-    u64 out;
-    std::memcpy(&out, raw.data(), sizeof(u64));
-    return Common::swap64(out) & 0xFFFFFFFFFF;
-}
-
-u64 Cheat::ValueWidth(u64 offset) const {
-    return Value(offset, width);
-}
-
-u64 Cheat::Value(u64 offset, u64 width) const {
-    u64 out;
-    std::memcpy(&out, raw.data() + offset, sizeof(u64));
-    out = Common::swap64(out);
-    if (width == 8)
-        return out;
-    return out & ((1ull << (width * CHAR_BIT)) - 1);
-}
-
-u32 Cheat::KeypadValue() const {
-    u32 out;
-    std::memcpy(&out, raw.data(), sizeof(u32));
-    return Common::swap32(out) & 0x0FFFFFFF;
-}
-
-void CheatList::SetMemoryParameters(VAddr main_begin, VAddr heap_begin, VAddr main_end,
-                                    VAddr heap_end, MemoryWriter writer, MemoryReader reader) {
-    this->main_region_begin = main_begin;
-    this->main_region_end = main_end;
-    this->heap_region_begin = heap_begin;
-    this->heap_region_end = heap_end;
-    this->writer = writer;
-    this->reader = reader;
-}
-
-MICROPROFILE_DEFINE(Cheat_Engine, "Add-Ons", "Cheat Engine", MP_RGB(70, 200, 70));
-
-void CheatList::Execute() {
-    MICROPROFILE_SCOPE(Cheat_Engine);
-
-    std::fill(scratch.begin(), scratch.end(), 0);
-    in_standard = false;
-    for (std::size_t i = 0; i < master_list.size(); ++i) {
-        LOG_DEBUG(Common_Filesystem, "Executing block #{:08X} ({})", i, master_list[i].first);
-        current_block = i;
-        ExecuteBlock(master_list[i].second);
-    }
-
-    in_standard = true;
-    for (std::size_t i = 0; i < standard_list.size(); ++i) {
-        LOG_DEBUG(Common_Filesystem, "Executing block #{:08X} ({})", i, standard_list[i].first);
-        current_block = i;
-        ExecuteBlock(standard_list[i].second);
-    }
-}
-
-CheatList::CheatList(const Core::System& system_, ProgramSegment master, ProgramSegment standard)
-    : master_list{std::move(master)}, standard_list{std::move(standard)}, system{&system_} {}
-
-bool CheatList::EvaluateConditional(const Cheat& cheat) const {
-    using ComparisonFunction = bool (*)(u64, u64);
-    constexpr std::array<ComparisonFunction, 6> comparison_functions{
-        [](u64 a, u64 b) { return a > b; },  [](u64 a, u64 b) { return a >= b; },
-        [](u64 a, u64 b) { return a < b; },  [](u64 a, u64 b) { return a <= b; },
-        [](u64 a, u64 b) { return a == b; }, [](u64 a, u64 b) { return a != b; },
-    };
-
-    if (cheat.type == CodeType::ConditionalInput) {
-        const auto applet_resource =
-            system->ServiceManager().GetService<Service::HID::Hid>("hid")->GetAppletResource();
-        if (applet_resource == nullptr) {
-            LOG_WARNING(
-                Common_Filesystem,
-                "Attempted to evaluate input conditional, but applet resource is not initialized!");
-            return false;
-        }
-
-        const auto press_state =
-            applet_resource
-                ->GetController<Service::HID::Controller_NPad>(Service::HID::HidController::NPad)
-                .GetAndResetPressState();
-        return ((press_state & cheat.KeypadValue()) & KEYPAD_BITMASK) != 0;
-    }
-
-    ASSERT(cheat.type == CodeType::Conditional);
-
-    const auto offset =
-        cheat.memory_type == MemoryType::MainNSO ? main_region_begin : heap_region_begin;
-    ASSERT(static_cast<u8>(cheat.comparison_op.Value()) < 6);
-    auto* function = comparison_functions[static_cast<u8>(cheat.comparison_op.Value())];
-    const auto addr = cheat.Address() + offset;
-
-    return function(reader(cheat.width, SanitizeAddress(addr)), cheat.ValueWidth(8));
-}
-
-void CheatList::ProcessBlockPairs(const Block& block) {
-    block_pairs.clear();
-
-    u64 scope = 0;
-    std::map<u64, u64> pairs;
-
-    for (std::size_t i = 0; i < block.size(); ++i) {
-        const auto& cheat = block[i];
-
-        switch (cheat.type) {
-        case CodeType::Conditional:
-        case CodeType::ConditionalInput:
-            pairs.insert_or_assign(scope, i);
-            ++scope;
-            break;
-        case CodeType::EndConditional: {
-            --scope;
-            const auto idx = pairs.at(scope);
-            block_pairs.insert_or_assign(idx, i);
-            break;
-        }
-        case CodeType::Loop: {
-            if (cheat.end_of_loop) {
-                --scope;
-                const auto idx = pairs.at(scope);
-                block_pairs.insert_or_assign(idx, i);
-            } else {
-                pairs.insert_or_assign(scope, i);
-                ++scope;
-            }
-            break;
-        }
-        }
-    }
-}
-
-void CheatList::WriteImmediate(const Cheat& cheat) {
-    const auto offset =
-        cheat.memory_type == MemoryType::MainNSO ? main_region_begin : heap_region_begin;
-    const auto& register_3 = scratch.at(cheat.register_3);
-
-    const auto addr = cheat.Address() + offset + register_3;
-    LOG_DEBUG(Common_Filesystem, "writing value={:016X} to addr={:016X}", addr,
-              cheat.Value(8, cheat.width));
-    writer(cheat.width, SanitizeAddress(addr), cheat.ValueWidth(8));
-}
-
-void CheatList::BeginConditional(const Cheat& cheat) {
-    if (EvaluateConditional(cheat)) {
-        return;
-    }
-
-    const auto iter = block_pairs.find(current_index);
-    ASSERT(iter != block_pairs.end());
-    current_index = iter->second - 1;
-}
-
-void CheatList::EndConditional(const Cheat& cheat) {
-    LOG_DEBUG(Common_Filesystem, "Ending conditional block.");
-}
-
-void CheatList::Loop(const Cheat& cheat) {
-    if (cheat.end_of_loop.Value())
-        ASSERT(!cheat.end_of_loop.Value());
-
-    auto& register_3 = scratch.at(cheat.register_3);
-    const auto iter = block_pairs.find(current_index);
-    ASSERT(iter != block_pairs.end());
-    ASSERT(iter->first < iter->second);
-
-    const s32 initial_value = static_cast<s32>(cheat.Value(4, sizeof(s32)));
-    for (s32 i = initial_value; i >= 0; --i) {
-        register_3 = static_cast<u64>(i);
-        for (std::size_t c = iter->first + 1; c < iter->second; ++c) {
-            current_index = c;
-            ExecuteSingleCheat(
-                (in_standard ? standard_list : master_list)[current_block].second[c]);
-        }
-    }
-
-    current_index = iter->second;
-}
-
-void CheatList::LoadImmediate(const Cheat& cheat) {
-    auto& register_3 = scratch.at(cheat.register_3);
-
-    LOG_DEBUG(Common_Filesystem, "setting register={:01X} equal to value={:016X}", cheat.register_3,
-              cheat.Value(4, 8));
-    register_3 = cheat.Value(4, 8);
-}
-
-void CheatList::LoadIndexed(const Cheat& cheat) {
-    const auto offset =
-        cheat.memory_type == MemoryType::MainNSO ? main_region_begin : heap_region_begin;
-    auto& register_3 = scratch.at(cheat.register_3);
-
-    const auto addr = (cheat.load_from_register.Value() ? register_3 : offset) + cheat.Address();
-    LOG_DEBUG(Common_Filesystem, "writing indexed value to register={:01X}, addr={:016X}",
-              cheat.register_3, addr);
-    register_3 = reader(cheat.width, SanitizeAddress(addr));
-}
-
-void CheatList::StoreIndexed(const Cheat& cheat) {
-    const auto& register_3 = scratch.at(cheat.register_3);
-
-    const auto addr =
-        register_3 + (cheat.add_additional_register.Value() ? scratch.at(cheat.register_6) : 0);
-    LOG_DEBUG(Common_Filesystem, "writing value={:016X} to addr={:016X}",
-              cheat.Value(4, cheat.width), addr);
-    writer(cheat.width, SanitizeAddress(addr), cheat.ValueWidth(4));
-}
-
-void CheatList::RegisterArithmetic(const Cheat& cheat) {
-    using ArithmeticFunction = u64 (*)(u64, u64);
-    constexpr std::array<ArithmeticFunction, 5> arithmetic_functions{
-        [](u64 a, u64 b) { return a + b; },  [](u64 a, u64 b) { return a - b; },
-        [](u64 a, u64 b) { return a * b; },  [](u64 a, u64 b) { return a << b; },
-        [](u64 a, u64 b) { return a >> b; },
-    };
-
-    using ArithmeticOverflowCheck = bool (*)(u64, u64);
-    constexpr std::array<ArithmeticOverflowCheck, 5> arithmetic_overflow_checks{
-        [](u64 a, u64 b) { return a > (std::numeric_limits<u64>::max() - b); },       // a + b
-        [](u64 a, u64 b) { return a > (std::numeric_limits<u64>::max() + b); },       // a - b
-        [](u64 a, u64 b) { return a > (std::numeric_limits<u64>::max() / b); },       // a * b
-        [](u64 a, u64 b) { return b >= 64 || (a & ~((1ull << (64 - b)) - 1)) != 0; }, // a << b
-        [](u64 a, u64 b) { return b >= 64 || (a & ((1ull << b) - 1)) != 0; },         // a >> b
-    };
-
-    static_assert(sizeof(arithmetic_functions) == sizeof(arithmetic_overflow_checks),
-                  "Missing or have extra arithmetic overflow checks compared to functions!");
-
-    auto& register_3 = scratch.at(cheat.register_3);
-
-    ASSERT(static_cast<u8>(cheat.arithmetic_op.Value()) < 5);
-    auto* function = arithmetic_functions[static_cast<u8>(cheat.arithmetic_op.Value())];
-    auto* overflow_function =
-        arithmetic_overflow_checks[static_cast<u8>(cheat.arithmetic_op.Value())];
-    LOG_DEBUG(Common_Filesystem, "performing arithmetic with register={:01X}, value={:016X}",
-              cheat.register_3, cheat.ValueWidth(4));
-
-    if (overflow_function(register_3, cheat.ValueWidth(4))) {
-        LOG_WARNING(Common_Filesystem,
-                    "overflow will occur when performing arithmetic operation={:02X} with operands "
-                    "a={:016X}, b={:016X}!",
-                    static_cast<u8>(cheat.arithmetic_op.Value()), register_3, cheat.ValueWidth(4));
-    }
-
-    register_3 = function(register_3, cheat.ValueWidth(4));
-}
-
-void CheatList::BeginConditionalInput(const Cheat& cheat) {
-    if (EvaluateConditional(cheat))
-        return;
-
-    const auto iter = block_pairs.find(current_index);
-    ASSERT(iter != block_pairs.end());
-    current_index = iter->second - 1;
-}
-
-VAddr CheatList::SanitizeAddress(VAddr in) const {
-    if ((in < main_region_begin || in >= main_region_end) &&
-        (in < heap_region_begin || in >= heap_region_end)) {
-        LOG_ERROR(Common_Filesystem,
-                  "Cheat attempting to access memory at invalid address={:016X}, if this persists, "
-                  "the cheat may be incorrect. However, this may be normal early in execution if "
-                  "the game has not properly set up yet.",
-                  in);
-        return 0; ///< Invalid addresses will hard crash
-    }
-
-    return in;
-}
-
-void CheatList::ExecuteSingleCheat(const Cheat& cheat) {
-    using CheatOperationFunction = void (CheatList::*)(const Cheat&);
-    constexpr std::array<CheatOperationFunction, 9> cheat_operation_functions{
-        &CheatList::WriteImmediate,        &CheatList::BeginConditional,
-        &CheatList::EndConditional,        &CheatList::Loop,
-        &CheatList::LoadImmediate,         &CheatList::LoadIndexed,
-        &CheatList::StoreIndexed,          &CheatList::RegisterArithmetic,
-        &CheatList::BeginConditionalInput,
-    };
-
-    const auto index = static_cast<u8>(cheat.type.Value());
-    ASSERT(index < sizeof(cheat_operation_functions));
-    const auto op = cheat_operation_functions[index];
-    (this->*op)(cheat);
-}
-
-void CheatList::ExecuteBlock(const Block& block) {
-    encountered_loops.clear();
-
-    ProcessBlockPairs(block);
-    for (std::size_t i = 0; i < block.size(); ++i) {
-        current_index = i;
-        ExecuteSingleCheat(block[i]);
-        i = current_index;
-    }
-}
-
-CheatParser::~CheatParser() = default;
-
-CheatList CheatParser::MakeCheatList(const Core::System& system, CheatList::ProgramSegment master,
-                                     CheatList::ProgramSegment standard) const {
-    return {system, std::move(master), std::move(standard)};
-}
-
-TextCheatParser::~TextCheatParser() = default;
-
-CheatList TextCheatParser::Parse(const Core::System& system, const std::vector<u8>& data) const {
-    std::stringstream ss;
-    ss.write(reinterpret_cast<const char*>(data.data()), data.size());
-
-    std::vector<std::string> lines;
-    std::string stream_line;
-    while (std::getline(ss, stream_line)) {
-        // Remove a trailing \r
-        if (!stream_line.empty() && stream_line.back() == '\r')
-            stream_line.pop_back();
-        lines.push_back(std::move(stream_line));
-    }
-
-    CheatList::ProgramSegment master_list;
-    CheatList::ProgramSegment standard_list;
-
-    for (std::size_t i = 0; i < lines.size(); ++i) {
-        auto line = lines[i];
-
-        if (!line.empty() && (line[0] == '[' || line[0] == '{')) {
-            const auto master = line[0] == '{';
-            const auto begin = master ? line.find('{') : line.find('[');
-            const auto end = master ? line.rfind('}') : line.rfind(']');
-
-            ASSERT(begin != std::string::npos && end != std::string::npos);
-
-            const std::string patch_name{line.begin() + begin + 1, line.begin() + end};
-            CheatList::Block block{};
-
-            while (i < lines.size() - 1) {
-                line = lines[++i];
-                if (!line.empty() && (line[0] == '[' || line[0] == '{')) {
-                    --i;
-                    break;
-                }
-
-                if (line.size() < 8)
-                    continue;
-
-                Cheat out{};
-                out.raw = ParseSingleLineCheat(line);
-                block.push_back(out);
-            }
-
-            (master ? master_list : standard_list).emplace_back(patch_name, block);
-        }
-    }
-
-    return MakeCheatList(system, master_list, standard_list);
-}
-
-std::array<u8, 16> TextCheatParser::ParseSingleLineCheat(const std::string& line) const {
-    std::array<u8, 16> out{};
-
-    if (line.size() < 8)
-        return out;
-
-    const auto word1 = Common::HexStringToArray<sizeof(u32)>(std::string_view{line.data(), 8});
-    std::memcpy(out.data(), word1.data(), sizeof(u32));
-
-    if (line.size() < 17 || line[8] != ' ')
-        return out;
-
-    const auto word2 = Common::HexStringToArray<sizeof(u32)>(std::string_view{line.data() + 9, 8});
-    std::memcpy(out.data() + sizeof(u32), word2.data(), sizeof(u32));
-
-    if (line.size() < 26 || line[17] != ' ') {
-        // Perform shifting in case value is truncated early.
-        const auto type = static_cast<CodeType>((out[0] & 0xF0) >> 4);
-        if (type == CodeType::Loop || type == CodeType::LoadImmediate ||
-            type == CodeType::StoreIndexed || type == CodeType::RegisterArithmetic) {
-            std::memcpy(out.data() + 8, out.data() + 4, sizeof(u32));
-            std::memset(out.data() + 4, 0, sizeof(u32));
-        }
-
-        return out;
-    }
-
-    const auto word3 = Common::HexStringToArray<sizeof(u32)>(std::string_view{line.data() + 18, 8});
-    std::memcpy(out.data() + 2 * sizeof(u32), word3.data(), sizeof(u32));
-
-    if (line.size() < 35 || line[26] != ' ') {
-        // Perform shifting in case value is truncated early.
-        const auto type = static_cast<CodeType>((out[0] & 0xF0) >> 4);
-        if (type == CodeType::WriteImmediate || type == CodeType::Conditional) {
-            std::memcpy(out.data() + 12, out.data() + 8, sizeof(u32));
-            std::memset(out.data() + 8, 0, sizeof(u32));
-        }
-
-        return out;
-    }
-
-    const auto word4 = Common::HexStringToArray<sizeof(u32)>(std::string_view{line.data() + 27, 8});
-    std::memcpy(out.data() + 3 * sizeof(u32), word4.data(), sizeof(u32));
-
-    return out;
-}
-
-namespace {
-u64 MemoryReadImpl(u32 width, VAddr addr) {
-    switch (width) {
-    case 1:
-        return Memory::Read8(addr);
-    case 2:
-        return Memory::Read16(addr);
-    case 4:
-        return Memory::Read32(addr);
-    case 8:
-        return Memory::Read64(addr);
-    default:
-        UNREACHABLE();
-        return 0;
-    }
-}
-
-void MemoryWriteImpl(u32 width, VAddr addr, u64 value) {
-    switch (width) {
-    case 1:
-        Memory::Write8(addr, static_cast<u8>(value));
-        break;
-    case 2:
-        Memory::Write16(addr, static_cast<u16>(value));
-        break;
-    case 4:
-        Memory::Write32(addr, static_cast<u32>(value));
-        break;
-    case 8:
-        Memory::Write64(addr, value);
-        break;
-    default:
-        UNREACHABLE();
-    }
-}
-} // Anonymous namespace
-
-CheatEngine::CheatEngine(Core::System& system, std::vector<CheatList> cheats_,
-                         const std::string& build_id, VAddr code_region_start,
-                         VAddr code_region_end)
-    : cheats{std::move(cheats_)}, core_timing{system.CoreTiming()} {
-    event = core_timing.RegisterEvent(
-        "CheatEngine::FrameCallback::" + build_id,
-        [this](u64 userdata, s64 cycles_late) { FrameCallback(userdata, cycles_late); });
-    core_timing.ScheduleEvent(CHEAT_ENGINE_TICKS, event);
-
-    const auto& vm_manager = system.CurrentProcess()->VMManager();
-    for (auto& list : this->cheats) {
-        list.SetMemoryParameters(code_region_start, vm_manager.GetHeapRegionBaseAddress(),
-                                 code_region_end, vm_manager.GetHeapRegionEndAddress(),
-                                 &MemoryWriteImpl, &MemoryReadImpl);
-    }
-}
-
-CheatEngine::~CheatEngine() {
-    core_timing.UnscheduleEvent(event, 0);
-}
-
-void CheatEngine::FrameCallback(u64 userdata, s64 cycles_late) {
-    for (auto& list : cheats) {
-        list.Execute();
-    }
-
-    core_timing.ScheduleEvent(CHEAT_ENGINE_TICKS - cycles_late, event);
-}
-
-} // namespace FileSys

src/core/file_sys/cheat_engine.h

@@ -1,234 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <map>
-#include <set>
-#include <vector>
-#include "common/bit_field.h"
-#include "common/common_types.h"
-
-namespace Core {
-class System;
-}
-
-namespace Core::Timing {
-class CoreTiming;
-struct EventType;
-} // namespace Core::Timing
-
-namespace FileSys {
-
-enum class CodeType : u32 {
-    // 0TMR00AA AAAAAAAA YYYYYYYY YYYYYYYY
-    // Writes a T sized value Y to the address A added to the value of register R in memory domain M
-    WriteImmediate = 0,
-
-    // 1TMC00AA AAAAAAAA YYYYYYYY YYYYYYYY
-    // Compares the T sized value Y to the value at address A in memory domain M using the
-    // conditional function C. If success, continues execution. If failure, jumps to the matching
-    // EndConditional statement.
-    Conditional = 1,
-
-    // 20000000
-    // Terminates a Conditional or ConditionalInput block.
-    EndConditional = 2,
-
-    // 300R0000 VVVVVVVV
-    // Starts looping V times, storing the current count in register R.
-    // Loop block is terminated with a matching 310R0000.
-    Loop = 3,
-
-    // 400R0000 VVVVVVVV VVVVVVVV
-    // Sets the value of register R to the value V.
-    LoadImmediate = 4,
-
-    // 5TMRI0AA AAAAAAAA
-    // Sets the value of register R to the value of width T at address A in memory domain M, with
-    // the current value of R added to the address if I == 1.
-    LoadIndexed = 5,
-
-    // 6T0RIFG0 VVVVVVVV VVVVVVVV
-    // Writes the value V of width T to the memory address stored in register R. Adds the value of
-    // register G to the final calculation if F is nonzero. Increments the value of register R by T
-    // after operation if I is nonzero.
-    StoreIndexed = 6,
-
-    // 7T0RA000 VVVVVVVV
-    // Performs the arithmetic operation A on the value in register R and the value V of width T,
-    // storing the result in register R.
-    RegisterArithmetic = 7,
-
-    // 8KKKKKKK
-    // Checks to see if any of the buttons defined by the bitmask K are pressed. If any are,
-    // execution continues. If none are, execution skips to the next EndConditional command.
-    ConditionalInput = 8,
-};
-
-enum class MemoryType : u32 {
-    // Addressed relative to start of main NSO
-    MainNSO = 0,
-
-    // Addressed relative to start of heap
-    Heap = 1,
-};
-
-enum class ArithmeticOp : u32 {
-    Add = 0,
-    Sub = 1,
-    Mult = 2,
-    LShift = 3,
-    RShift = 4,
-};
-
-enum class ComparisonOp : u32 {
-    GreaterThan = 1,
-    GreaterThanEqual = 2,
-    LessThan = 3,
-    LessThanEqual = 4,
-    Equal = 5,
-    Inequal = 6,
-};
-
-union Cheat {
-    std::array<u8, 16> raw;
-
-    BitField<4, 4, CodeType> type;
-    BitField<0, 4, u32> width; // Can be 1, 2, 4, or 8. Measured in bytes.
-    BitField<0, 4, u32> end_of_loop;
-    BitField<12, 4, MemoryType> memory_type;
-    BitField<8, 4, u32> register_3;
-    BitField<8, 4, ComparisonOp> comparison_op;
-    BitField<20, 4, u32> load_from_register;
-    BitField<20, 4, u32> increment_register;
-    BitField<20, 4, ArithmeticOp> arithmetic_op;
-    BitField<16, 4, u32> add_additional_register;
-    BitField<28, 4, u32> register_6;
-
-    u64 Address() const;
-    u64 ValueWidth(u64 offset) const;
-    u64 Value(u64 offset, u64 width) const;
-    u32 KeypadValue() const;
-};
-
-class CheatParser;
-
-// Represents a full collection of cheats for a game. The Execute function should be called every
-// interval that all cheats should be executed. Clients should not directly instantiate this class
-// (hence private constructor), they should instead receive an instance from CheatParser, which
-// guarantees the list is always in an acceptable state.
-class CheatList {
-public:
-    friend class CheatParser;
-
-    using Block = std::vector<Cheat>;
-    using ProgramSegment = std::vector<std::pair<std::string, Block>>;
-
-    // (width in bytes, address, value)
-    using MemoryWriter = void (*)(u32, VAddr, u64);
-    // (width in bytes, address) -> value
-    using MemoryReader = u64 (*)(u32, VAddr);
-
-    void SetMemoryParameters(VAddr main_begin, VAddr heap_begin, VAddr main_end, VAddr heap_end,
-                             MemoryWriter writer, MemoryReader reader);
-
-    void Execute();
-
-private:
-    CheatList(const Core::System& system_, ProgramSegment master, ProgramSegment standard);
-
-    void ProcessBlockPairs(const Block& block);
-    void ExecuteSingleCheat(const Cheat& cheat);
-
-    void ExecuteBlock(const Block& block);
-
-    bool EvaluateConditional(const Cheat& cheat) const;
-
-    // Individual cheat operations
-    void WriteImmediate(const Cheat& cheat);
-    void BeginConditional(const Cheat& cheat);
-    void EndConditional(const Cheat& cheat);
-    void Loop(const Cheat& cheat);
-    void LoadImmediate(const Cheat& cheat);
-    void LoadIndexed(const Cheat& cheat);
-    void StoreIndexed(const Cheat& cheat);
-    void RegisterArithmetic(const Cheat& cheat);
-    void BeginConditionalInput(const Cheat& cheat);
-
-    VAddr SanitizeAddress(VAddr in) const;
-
-    // Master Codes are defined as codes that cannot be disabled and are run prior to all
-    // others.
-    ProgramSegment master_list;
-    // All other codes
-    ProgramSegment standard_list;
-
-    bool in_standard = false;
-
-    // 16 (0x0-0xF) scratch registers that can be used by cheats
-    std::array<u64, 16> scratch{};
-
-    MemoryWriter writer = nullptr;
-    MemoryReader reader = nullptr;
-
-    u64 main_region_begin{};
-    u64 heap_region_begin{};
-    u64 main_region_end{};
-    u64 heap_region_end{};
-
-    u64 current_block{};
-    // The current index of the cheat within the current Block
-    u64 current_index{};
-
-    // The 'stack' of the program. When a conditional or loop statement is encountered, its index is
-    // pushed onto this queue. When a end block is encountered, the condition is checked.
-    std::map<u64, u64> block_pairs;
-
-    std::set<u64> encountered_loops;
-
-    const Core::System* system;
-};
-
-// Intermediary class that parses a text file or other disk format for storing cheats into a
-// CheatList object, that can be used for execution.
-class CheatParser {
-public:
-    virtual ~CheatParser();
-
-    virtual CheatList Parse(const Core::System& system, const std::vector<u8>& data) const = 0;
-
-protected:
-    CheatList MakeCheatList(const Core::System& system_, CheatList::ProgramSegment master,
-                            CheatList::ProgramSegment standard) const;
-};
-
-// CheatParser implementation that parses text files
-class TextCheatParser final : public CheatParser {
-public:
-    ~TextCheatParser() override;
-
-    CheatList Parse(const Core::System& system, const std::vector<u8>& data) const override;
-
-private:
-    std::array<u8, 16> ParseSingleLineCheat(const std::string& line) const;
-};
-
-// Class that encapsulates a CheatList and manages its interaction with memory and CoreTiming
-class CheatEngine final {
-public:
-    CheatEngine(Core::System& system_, std::vector<CheatList> cheats_, const std::string& build_id,
-                VAddr code_region_start, VAddr code_region_end);
-    ~CheatEngine();
-
-private:
-    void FrameCallback(u64 userdata, s64 cycles_late);
-
-    std::vector<CheatList> cheats;
-
-    Core::Timing::EventType* event;
-    Core::Timing::CoreTiming& core_timing;
-};
-
-} // namespace FileSys

src/core/file_sys/content_archive.cpp

@@ -528,6 +528,14 @@ u64 NCA::GetTitleId() const {
     return header.title_id;
 }
 
+std::array<u8, 16> NCA::GetRightsId() const {
+    return header.rights_id;
+}
+
+u32 NCA::GetSDKVersion() const {
+    return header.sdk_version;
+}
+
 bool NCA::IsUpdate() const {
     return is_update;
 }

src/core/file_sys/content_archive.h

@@ -112,6 +112,8 @@ public:
 
     NCAContentType GetType() const;
     u64 GetTitleId() const;
+    std::array<u8, 0x10> GetRightsId() const;
+    u32 GetSDKVersion() const;
     bool IsUpdate() const;
 
     VirtualFile GetRomFS() const;

src/core/file_sys/patch_manager.cpp

@@ -22,6 +22,7 @@
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/loader/loader.h"
 #include "core/loader/nso.h"
+#include "core/memory/cheat_engine.h"
 #include "core/settings.h"
 
 namespace FileSys {
@@ -63,7 +64,8 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
 
     if (Settings::values.dump_exefs) {
         LOG_INFO(Loader, "Dumping ExeFS for title_id={:016X}", title_id);
-        const auto dump_dir = Service::FileSystem::GetModificationDumpRoot(title_id);
+        const auto dump_dir =
+            Core::System::GetInstance().GetFileSystemController().GetModificationDumpRoot(title_id);
         if (dump_dir != nullptr) {
             const auto exefs_dir = GetOrCreateDirectoryRelative(dump_dir, "/exefs");
             VfsRawCopyD(exefs, exefs_dir);
@@ -88,7 +90,8 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
     }
 
     // LayeredExeFS
-    const auto load_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
+    const auto load_dir =
+        Core::System::GetInstance().GetFileSystemController().GetModificationLoadRoot(title_id);
     if (load_dir != nullptr && load_dir->GetSize() > 0) {
         auto patch_dirs = load_dir->GetSubdirectories();
         std::sort(
@@ -174,7 +177,8 @@ std::vector<u8> PatchManager::PatchNSO(const std::vector<u8>& nso, const std::st
     if (Settings::values.dump_nso) {
         LOG_INFO(Loader, "Dumping NSO for name={}, build_id={}, title_id={:016X}", name, build_id,
                  title_id);
-        const auto dump_dir = Service::FileSystem::GetModificationDumpRoot(title_id);
+        const auto dump_dir =
+            Core::System::GetInstance().GetFileSystemController().GetModificationDumpRoot(title_id);
         if (dump_dir != nullptr) {
             const auto nso_dir = GetOrCreateDirectoryRelative(dump_dir, "/nso");
             const auto file = nso_dir->CreateFile(fmt::format("{}-{}.nso", name, build_id));
@@ -186,7 +190,13 @@ std::vector<u8> PatchManager::PatchNSO(const std::vector<u8>& nso, const std::st
 
     LOG_INFO(Loader, "Patching NSO for name={}, build_id={}", name, build_id);
 
-    const auto load_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
+    const auto load_dir =
+        Core::System::GetInstance().GetFileSystemController().GetModificationLoadRoot(title_id);
+    if (load_dir == nullptr) {
+        LOG_ERROR(Loader, "Cannot load mods for invalid title_id={:016X}", title_id);
+        return nso;
+    }
+
     auto patch_dirs = load_dir->GetSubdirectories();
     std::sort(patch_dirs.begin(), patch_dirs.end(),
               [](const VirtualDir& l, const VirtualDir& r) { return l->GetName() < r->GetName(); });
@@ -224,7 +234,13 @@ bool PatchManager::HasNSOPatch(const std::array<u8, 32>& build_id_) const {
 
     LOG_INFO(Loader, "Querying NSO patch existence for build_id={}", build_id);
 
-    const auto load_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
+    const auto load_dir =
+        Core::System::GetInstance().GetFileSystemController().GetModificationLoadRoot(title_id);
+    if (load_dir == nullptr) {
+        LOG_ERROR(Loader, "Cannot load mods for invalid title_id={:016X}", title_id);
+        return false;
+    }
+
     auto patch_dirs = load_dir->GetSubdirectories();
     std::sort(patch_dirs.begin(), patch_dirs.end(),
               [](const VirtualDir& l, const VirtualDir& r) { return l->GetName() < r->GetName(); });
@@ -232,9 +248,10 @@ bool PatchManager::HasNSOPatch(const std::array<u8, 32>& build_id_) const {
     return !CollectPatches(patch_dirs, build_id).empty();
 }
 
-static std::optional<CheatList> ReadCheatFileFromFolder(const Core::System& system, u64 title_id,
-                                                        const std::array<u8, 0x20>& build_id_,
-                                                        const VirtualDir& base_path, bool upper) {
+namespace {
+std::optional<std::vector<Memory::CheatEntry>> ReadCheatFileFromFolder(
+    const Core::System& system, u64 title_id, const std::array<u8, 0x20>& build_id_,
+    const VirtualDir& base_path, bool upper) {
     const auto build_id_raw = Common::HexToString(build_id_, upper);
     const auto build_id = build_id_raw.substr(0, sizeof(u64) * 2);
     const auto file = base_path->GetFile(fmt::format("{}.txt", build_id));
@@ -252,31 +269,39 @@ static std::optional<CheatList> ReadCheatFileFromFolder(const Core::System& syst
         return std::nullopt;
     }
 
-    TextCheatParser parser;
-    return parser.Parse(system, data);
+    Memory::TextCheatParser parser;
+    return parser.Parse(
+        system, std::string_view(reinterpret_cast<const char* const>(data.data()), data.size()));
 }
 
-std::vector<CheatList> PatchManager::CreateCheatList(const Core::System& system,
-                                                     const std::array<u8, 32>& build_id_) const {
-    const auto load_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
+} // Anonymous namespace
+
+std::vector<Memory::CheatEntry> PatchManager::CreateCheatList(
+    const Core::System& system, const std::array<u8, 32>& build_id_) const {
+    const auto load_dir = system.GetFileSystemController().GetModificationLoadRoot(title_id);
+    if (load_dir == nullptr) {
+        LOG_ERROR(Loader, "Cannot load mods for invalid title_id={:016X}", title_id);
+        return {};
+    }
+
     auto patch_dirs = load_dir->GetSubdirectories();
     std::sort(patch_dirs.begin(), patch_dirs.end(),
               [](const VirtualDir& l, const VirtualDir& r) { return l->GetName() < r->GetName(); });
 
-    std::vector<CheatList> out;
-    out.reserve(patch_dirs.size());
+    std::vector<Memory::CheatEntry> out;
     for (const auto& subdir : patch_dirs) {
         auto cheats_dir = subdir->GetSubdirectory("cheats");
         if (cheats_dir != nullptr) {
             auto res = ReadCheatFileFromFolder(system, title_id, build_id_, cheats_dir, true);
             if (res.has_value()) {
-                out.push_back(std::move(*res));
+                std::copy(res->begin(), res->end(), std::back_inserter(out));
                 continue;
             }
 
             res = ReadCheatFileFromFolder(system, title_id, build_id_, cheats_dir, false);
-            if (res.has_value())
-                out.push_back(std::move(*res));
+            if (res.has_value()) {
+                std::copy(res->begin(), res->end(), std::back_inserter(out));
+            }
         }
     }
 
@@ -284,7 +309,8 @@ std::vector<CheatList> PatchManager::CreateCheatList(const Core::System& system,
 }
 
 static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType type) {
-    const auto load_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
+    const auto load_dir =
+        Core::System::GetInstance().GetFileSystemController().GetModificationLoadRoot(title_id);
     if ((type != ContentRecordType::Program && type != ContentRecordType::Data) ||
         load_dir == nullptr || load_dir->GetSize() <= 0) {
         return;
@@ -393,6 +419,8 @@ static bool IsDirValidAndNonEmpty(const VirtualDir& dir) {
 
 std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNames(
     VirtualFile update_raw) const {
+    if (title_id == 0)
+        return {};
     std::map<std::string, std::string, std::less<>> out;
     const auto& installed = Core::System::GetInstance().GetContentProvider();
     const auto& disabled = Settings::values.disabled_addons[title_id];
@@ -423,7 +451,8 @@ std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNam
     }
 
     // General Mods (LayeredFS and IPS)
-    const auto mod_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
+    const auto mod_dir =
+        Core::System::GetInstance().GetFileSystemController().GetModificationLoadRoot(title_id);
     if (mod_dir != nullptr && mod_dir->GetSize() > 0) {
         for (const auto& mod : mod_dir->GetSubdirectories()) {
             std::string types;

src/core/file_sys/patch_manager.h

@@ -8,9 +8,9 @@
 #include <memory>
 #include <string>
 #include "common/common_types.h"
-#include "core/file_sys/cheat_engine.h"
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/vfs.h"
+#include "core/memory/dmnt_cheat_types.h"
 
 namespace Core {
 class System;
@@ -51,8 +51,8 @@ public:
     bool HasNSOPatch(const std::array<u8, 0x20>& build_id) const;
 
     // Creates a CheatList object with all
-    std::vector<CheatList> CreateCheatList(const Core::System& system,
-                                           const std::array<u8, 0x20>& build_id) const;
+    std::vector<Memory::CheatEntry> CreateCheatList(const Core::System& system,
+                                                    const std::array<u8, 0x20>& build_id) const;
 
     // Currently tracked RomFS patches:
     // - Game Updates

src/core/file_sys/registered_cache.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
+#include <random>
 #include <regex>
 #include <mbedtls/sha256.h>
 #include "common/assert.h"
@@ -48,18 +49,21 @@ static bool FollowsTwoDigitDirFormat(std::string_view name) {
 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 const std::regex nca_id_cnmt_regex(
+        "[0-9A-F]{32}\\.cnmt.nca", std::regex_constants::ECMAScript | std::regex_constants::icase);
+    return (name.size() == 36 && std::regex_match(name.begin(), name.end(), nca_id_regex)) ||
+           (name.size() == 41 && std::regex_match(name.begin(), name.end(), nca_id_cnmt_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", Common::HexToString(nca_id, second_hex_upper));
-    }
+                                            bool within_two_digit, bool cnmt_suffix) {
+    if (!within_two_digit)
+        return fmt::format(cnmt_suffix ? "{}.cnmt.nca" : "/{}.nca",
+                           Common::HexToString(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],
+    return fmt::format(cnmt_suffix ? "/000000{:02X}/{}.cnmt.nca" : "/000000{:02X}/{}.nca", hash[0],
                        Common::HexToString(nca_id, second_hex_upper));
 }
 
@@ -127,6 +131,156 @@ std::vector<ContentProviderEntry> ContentProvider::ListEntries() const {
     return ListEntriesFilter(std::nullopt, std::nullopt, std::nullopt);
 }
 
+PlaceholderCache::PlaceholderCache(VirtualDir dir_) : dir(std::move(dir_)) {}
+
+bool PlaceholderCache::Create(const NcaID& id, u64 size) const {
+    const auto path = GetRelativePathFromNcaID(id, false, true, false);
+
+    if (dir->GetFileRelative(path) != nullptr) {
+        return false;
+    }
+
+    Core::Crypto::SHA256Hash hash{};
+    mbedtls_sha256(id.data(), id.size(), hash.data(), 0);
+    const auto dirname = fmt::format("000000{:02X}", hash[0]);
+
+    const auto dir2 = GetOrCreateDirectoryRelative(dir, dirname);
+
+    if (dir2 == nullptr)
+        return false;
+
+    const auto file = dir2->CreateFile(fmt::format("{}.nca", Common::HexToString(id, false)));
+
+    if (file == nullptr)
+        return false;
+
+    return file->Resize(size);
+}
+
+bool PlaceholderCache::Delete(const NcaID& id) const {
+    const auto path = GetRelativePathFromNcaID(id, false, true, false);
+
+    if (dir->GetFileRelative(path) == nullptr) {
+        return false;
+    }
+
+    Core::Crypto::SHA256Hash hash{};
+    mbedtls_sha256(id.data(), id.size(), hash.data(), 0);
+    const auto dirname = fmt::format("000000{:02X}", hash[0]);
+
+    const auto dir2 = GetOrCreateDirectoryRelative(dir, dirname);
+
+    const auto res = dir2->DeleteFile(fmt::format("{}.nca", Common::HexToString(id, false)));
+
+    return res;
+}
+
+bool PlaceholderCache::Exists(const NcaID& id) const {
+    const auto path = GetRelativePathFromNcaID(id, false, true, false);
+
+    return dir->GetFileRelative(path) != nullptr;
+}
+
+bool PlaceholderCache::Write(const NcaID& id, u64 offset, const std::vector<u8>& data) const {
+    const auto path = GetRelativePathFromNcaID(id, false, true, false);
+    const auto file = dir->GetFileRelative(path);
+
+    if (file == nullptr)
+        return false;
+
+    return file->WriteBytes(data, offset) == data.size();
+}
+
+bool PlaceholderCache::Register(RegisteredCache* cache, const NcaID& placeholder,
+                                const NcaID& install) const {
+    const auto path = GetRelativePathFromNcaID(placeholder, false, true, false);
+    const auto file = dir->GetFileRelative(path);
+
+    if (file == nullptr)
+        return false;
+
+    const auto res = cache->RawInstallNCA(NCA{file}, &VfsRawCopy, false, install);
+
+    if (res != InstallResult::Success)
+        return false;
+
+    return Delete(placeholder);
+}
+
+bool PlaceholderCache::CleanAll() const {
+    return dir->GetParentDirectory()->CleanSubdirectoryRecursive(dir->GetName());
+}
+
+std::optional<std::array<u8, 0x10>> PlaceholderCache::GetRightsID(const NcaID& id) const {
+    const auto path = GetRelativePathFromNcaID(id, false, true, false);
+    const auto file = dir->GetFileRelative(path);
+
+    if (file == nullptr)
+        return std::nullopt;
+
+    NCA nca{file};
+
+    if (nca.GetStatus() != Loader::ResultStatus::Success &&
+        nca.GetStatus() != Loader::ResultStatus::ErrorMissingBKTRBaseRomFS) {
+        return std::nullopt;
+    }
+
+    const auto rights_id = nca.GetRightsId();
+    if (rights_id == NcaID{})
+        return std::nullopt;
+
+    return rights_id;
+}
+
+u64 PlaceholderCache::Size(const NcaID& id) const {
+    const auto path = GetRelativePathFromNcaID(id, false, true, false);
+    const auto file = dir->GetFileRelative(path);
+
+    if (file == nullptr)
+        return 0;
+
+    return file->GetSize();
+}
+
+bool PlaceholderCache::SetSize(const NcaID& id, u64 new_size) const {
+    const auto path = GetRelativePathFromNcaID(id, false, true, false);
+    const auto file = dir->GetFileRelative(path);
+
+    if (file == nullptr)
+        return false;
+
+    return file->Resize(new_size);
+}
+
+std::vector<NcaID> PlaceholderCache::List() const {
+    std::vector<NcaID> out;
+    for (const auto& sdir : dir->GetSubdirectories()) {
+        for (const auto& file : sdir->GetFiles()) {
+            const auto name = file->GetName();
+            if (name.length() == 36 && name[32] == '.' && name[33] == 'n' && name[34] == 'c' &&
+                name[35] == 'a') {
+                out.push_back(Common::HexStringToArray<0x10>(name.substr(0, 32)));
+            }
+        }
+    }
+    return out;
+}
+
+NcaID PlaceholderCache::Generate() {
+    std::random_device device;
+    std::mt19937 gen(device());
+    std::uniform_int_distribution<u64> distribution(1, std::numeric_limits<u64>::max());
+
+    NcaID out{};
+
+    const auto v1 = distribution(gen);
+    const auto v2 = distribution(gen);
+    std::memcpy(out.data(), &v1, sizeof(u64));
+    std::memcpy(out.data() + sizeof(u64), &v2, sizeof(u64));
+
+    return out;
+}
+
 VirtualFile RegisteredCache::OpenFileOrDirectoryConcat(const VirtualDir& dir,
                                                        std::string_view path) const {
     const auto file = dir->GetFileRelative(path);
@@ -169,14 +323,18 @@ VirtualFile RegisteredCache::OpenFileOrDirectoryConcat(const VirtualDir& dir,
 
 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);
+    // Try all five relevant modes of file storage:
+    // (bit 2 = uppercase/lower, bit 1 = within a two-digit dir, bit 0 = .cnmt suffix)
+    // 000: /000000**/{:032X}.nca
+    // 010: /{:032X}.nca
+    // 100: /000000**/{:032x}.nca
+    // 110: /{:032x}.nca
+    // 111: /{:032x}.cnmt.nca
+    for (u8 i = 0; i < 8; ++i) {
+        if ((i % 2) == 1 && i != 7)
+            continue;
+        const auto path =
+            GetRelativePathFromNcaID(id, (i & 0b100) == 0, (i & 0b010) == 0, (i & 0b001) == 0b001);
         file = OpenFileOrDirectoryConcat(dir, path);
         if (file != nullptr)
             return file;
@@ -472,7 +630,7 @@ InstallResult RegisteredCache::RawInstallNCA(const NCA& nca, const VfsCopyFuncti
         memcpy(id.data(), hash.data(), 16);
     }
 
-    std::string path = GetRelativePathFromNcaID(id, false, true);
+    std::string path = GetRelativePathFromNcaID(id, false, true, false);
 
     if (GetFileAtID(id) != nullptr && !overwrite_if_exists) {
         LOG_WARNING(Loader, "Attempting to overwrite existing NCA. Skipping...");

src/core/file_sys/registered_cache.h

@@ -25,6 +25,8 @@ enum class NCAContentType : u8;
 enum class TitleType : u8;
 
 struct ContentRecord;
+struct MetaRecord;
+class RegisteredCache;
 
 using NcaID = std::array<u8, 0x10>;
 using ContentProviderParsingFunction = std::function<VirtualFile(const VirtualFile&, const NcaID&)>;
@@ -89,6 +91,27 @@ protected:
     Core::Crypto::KeyManager keys;
 };
 
+class PlaceholderCache {
+public:
+    explicit PlaceholderCache(VirtualDir dir);
+
+    bool Create(const NcaID& id, u64 size) const;
+    bool Delete(const NcaID& id) const;
+    bool Exists(const NcaID& id) const;
+    bool Write(const NcaID& id, u64 offset, const std::vector<u8>& data) const;
+    bool Register(RegisteredCache* cache, const NcaID& placeholder, const NcaID& install) const;
+    bool CleanAll() const;
+    std::optional<std::array<u8, 0x10>> GetRightsID(const NcaID& id) const;
+    u64 Size(const NcaID& id) const;
+    bool SetSize(const NcaID& id, u64 new_size) const;
+    std::vector<NcaID> List() const;
+
+    static NcaID Generate();
+
+private:
+    VirtualDir dir;
+};
+
 /*
  * A class that catalogues NCAs in the registered directory structure.
  * Nintendo's registered format follows this structure:
@@ -103,6 +126,8 @@ protected:
  * when 4GB splitting can be ignored.)
  */
 class RegisteredCache : public ContentProvider {
+    friend class PlaceholderCache;
+
 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

src/core/file_sys/romfs_factory.cpp

@@ -7,6 +7,7 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
+#include "core/file_sys/card_image.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/patch_manager.h"
@@ -34,7 +35,7 @@ void RomFSFactory::SetPackedUpdate(VirtualFile update_raw) {
     this->update_raw = std::move(update_raw);
 }
 
-ResultVal<VirtualFile> RomFSFactory::OpenCurrentProcess() {
+ResultVal<VirtualFile> RomFSFactory::OpenCurrentProcess() const {
     if (!updatable)
         return MakeResult<VirtualFile>(file);
 
@@ -43,7 +44,8 @@ ResultVal<VirtualFile> RomFSFactory::OpenCurrentProcess() {
         patch_manager.PatchRomFS(file, ivfc_offset, ContentRecordType::Program, update_raw));
 }
 
-ResultVal<VirtualFile> RomFSFactory::Open(u64 title_id, StorageId storage, ContentRecordType type) {
+ResultVal<VirtualFile> RomFSFactory::Open(u64 title_id, StorageId storage,
+                                          ContentRecordType type) const {
     std::shared_ptr<NCA> res;
 
     switch (storage) {
@@ -51,13 +53,17 @@ ResultVal<VirtualFile> RomFSFactory::Open(u64 title_id, StorageId storage, Conte
         res = Core::System::GetInstance().GetContentProvider().GetEntry(title_id, type);
         break;
     case StorageId::NandSystem:
-        res = Service::FileSystem::GetSystemNANDContents()->GetEntry(title_id, type);
+        res =
+            Core::System::GetInstance().GetFileSystemController().GetSystemNANDContents()->GetEntry(
+                title_id, type);
         break;
     case StorageId::NandUser:
-        res = Service::FileSystem::GetUserNANDContents()->GetEntry(title_id, type);
+        res = Core::System::GetInstance().GetFileSystemController().GetUserNANDContents()->GetEntry(
+            title_id, type);
         break;
     case StorageId::SdCard:
-        res = Service::FileSystem::GetSDMCContents()->GetEntry(title_id, type);
+        res = Core::System::GetInstance().GetFileSystemController().GetSDMCContents()->GetEntry(
+            title_id, type);
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented storage_id={:02X}", static_cast<u8>(storage));

src/core/file_sys/romfs_factory.h

@@ -33,8 +33,8 @@ public:
     ~RomFSFactory();
 
     void SetPackedUpdate(VirtualFile update_raw);
-    ResultVal<VirtualFile> OpenCurrentProcess();
-    ResultVal<VirtualFile> Open(u64 title_id, StorageId storage, ContentRecordType type);
+    ResultVal<VirtualFile> OpenCurrentProcess() const;
+    ResultVal<VirtualFile> Open(u64 title_id, StorageId storage, ContentRecordType type) const;
 
 private:
     VirtualFile file;

src/core/file_sys/savedata_factory.cpp

@@ -15,22 +15,8 @@ namespace FileSys {
 
 constexpr char SAVE_DATA_SIZE_FILENAME[] = ".yuzu_save_size";
 
-std::string SaveDataDescriptor::DebugInfo() const {
-    return fmt::format("[type={:02X}, title_id={:016X}, user_id={:016X}{:016X}, save_id={:016X}, "
-                       "rank={}, index={}]",
-                       static_cast<u8>(type), title_id, user_id[1], user_id[0], save_id,
-                       static_cast<u8>(rank), index);
-}
-
-SaveDataFactory::SaveDataFactory(VirtualDir save_directory) : dir(std::move(save_directory)) {
-    // Delete all temporary storages
-    // On hardware, it is expected that temporary storage be empty at first use.
-    dir->DeleteSubdirectoryRecursive("temp");
-}
-
-SaveDataFactory::~SaveDataFactory() = default;
-
-ResultVal<VirtualDir> SaveDataFactory::Open(SaveDataSpaceId space, const SaveDataDescriptor& meta) {
+namespace {
+void PrintSaveDataDescriptorWarnings(SaveDataDescriptor meta) {
     if (meta.type == SaveDataType::SystemSaveData || meta.type == SaveDataType::SaveData) {
         if (meta.zero_1 != 0) {
             LOG_WARNING(Service_FS,
@@ -65,23 +51,51 @@ ResultVal<VirtualDir> SaveDataFactory::Open(SaveDataSpaceId space, const SaveDat
                     "non-zero ({:016X}{:016X})",
                     meta.user_id[1], meta.user_id[0]);
     }
+}
+} // Anonymous namespace
 
-    std::string save_directory =
+std::string SaveDataDescriptor::DebugInfo() const {
+    return fmt::format("[type={:02X}, title_id={:016X}, user_id={:016X}{:016X}, "
+                       "save_id={:016X}, "
+                       "rank={}, index={}]",
+                       static_cast<u8>(type), title_id, user_id[1], user_id[0], save_id,
+                       static_cast<u8>(rank), index);
+}
+
+SaveDataFactory::SaveDataFactory(VirtualDir save_directory) : dir(std::move(save_directory)) {
+    // Delete all temporary storages
+    // On hardware, it is expected that temporary storage be empty at first use.
+    dir->DeleteSubdirectoryRecursive("temp");
+}
+
+SaveDataFactory::~SaveDataFactory() = default;
+
+ResultVal<VirtualDir> SaveDataFactory::Create(SaveDataSpaceId space,
+                                              const SaveDataDescriptor& meta) const {
+    PrintSaveDataDescriptorWarnings(meta);
+
+    const auto save_directory =
         GetFullPath(space, meta.type, meta.title_id, meta.user_id, meta.save_id);
 
-    // TODO(DarkLordZach): Try to not create when opening, there are dedicated create save methods.
-    // But, user_ids don't match so this works for now.
+    auto out = dir->CreateDirectoryRelative(save_directory);
+
+    // Return an error if the save data doesn't actually exist.
+    if (out == nullptr) {
+        // TODO(DarkLordZach): Find out correct error code.
+        return ResultCode(-1);
+    }
+
+    return MakeResult<VirtualDir>(std::move(out));
+}
+
+ResultVal<VirtualDir> SaveDataFactory::Open(SaveDataSpaceId space,
+                                            const SaveDataDescriptor& meta) const {
+
+    const auto save_directory =
+        GetFullPath(space, meta.type, meta.title_id, meta.user_id, meta.save_id);
 
     auto out = dir->GetDirectoryRelative(save_directory);
 
-    if (out == nullptr) {
-        // TODO(bunnei): This is a work-around to always create a save data directory if it does not
-        // already exist. This is a hack, as we do not understand yet how this works on hardware.
-        // Without a save data directory, many games will assert on boot. This should not have any
-        // bad side-effects.
-        out = dir->CreateDirectoryRelative(save_directory);
-    }
-
     // Return an error if the save data doesn't actually exist.
     if (out == nullptr) {
         // TODO(Subv): Find out correct error code.
@@ -152,7 +166,7 @@ SaveDataSize SaveDataFactory::ReadSaveDataSize(SaveDataType type, u64 title_id,
 }
 
 void SaveDataFactory::WriteSaveDataSize(SaveDataType type, u64 title_id, u128 user_id,
-                                        SaveDataSize new_value) {
+                                        SaveDataSize new_value) const {
     const auto path = GetFullPath(SaveDataSpaceId::NandUser, type, title_id, user_id, 0);
     const auto dir = GetOrCreateDirectoryRelative(this->dir, path);
 

src/core/file_sys/savedata_factory.h

@@ -64,7 +64,8 @@ public:
     explicit SaveDataFactory(VirtualDir dir);
     ~SaveDataFactory();
 
-    ResultVal<VirtualDir> Open(SaveDataSpaceId space, const SaveDataDescriptor& meta);
+    ResultVal<VirtualDir> Create(SaveDataSpaceId space, const SaveDataDescriptor& meta) const;
+    ResultVal<VirtualDir> Open(SaveDataSpaceId space, const SaveDataDescriptor& meta) const;
 
     VirtualDir GetSaveDataSpaceDirectory(SaveDataSpaceId space) const;
 
@@ -73,7 +74,8 @@ public:
                                    u128 user_id, u64 save_id);
 
     SaveDataSize ReadSaveDataSize(SaveDataType type, u64 title_id, u128 user_id) const;
-    void WriteSaveDataSize(SaveDataType type, u64 title_id, u128 user_id, SaveDataSize new_value);
+    void WriteSaveDataSize(SaveDataType type, u64 title_id, u128 user_id,
+                           SaveDataSize new_value) const;
 
 private:
     VirtualDir dir;

src/core/file_sys/sdmc_factory.cpp

@@ -6,6 +6,7 @@
 #include "core/file_sys/registered_cache.h"
 #include "core/file_sys/sdmc_factory.h"
 #include "core/file_sys/xts_archive.h"
+#include "core/settings.h"
 
 namespace FileSys {
 
@@ -14,16 +15,38 @@ SDMCFactory::SDMCFactory(VirtualDir dir_)
                                 GetOrCreateDirectoryRelative(dir, "/Nintendo/Contents/registered"),
                                 [](const VirtualFile& file, const NcaID& id) {
                                     return NAX{file, id}.GetDecrypted();
-                                })) {}
+                                })),
+      placeholder(std::make_unique<PlaceholderCache>(
+          GetOrCreateDirectoryRelative(dir, "/Nintendo/Contents/placehld"))) {}
 
 SDMCFactory::~SDMCFactory() = default;
 
-ResultVal<VirtualDir> SDMCFactory::Open() {
+ResultVal<VirtualDir> SDMCFactory::Open() const {
     return MakeResult<VirtualDir>(dir);
 }
 
+VirtualDir SDMCFactory::GetSDMCContentDirectory() const {
+    return GetOrCreateDirectoryRelative(dir, "/Nintendo/Contents");
+}
+
 RegisteredCache* SDMCFactory::GetSDMCContents() const {
     return contents.get();
 }
 
+PlaceholderCache* SDMCFactory::GetSDMCPlaceholder() const {
+    return placeholder.get();
+}
+
+VirtualDir SDMCFactory::GetImageDirectory() const {
+    return GetOrCreateDirectoryRelative(dir, "/Nintendo/Album");
+}
+
+u64 SDMCFactory::GetSDMCFreeSpace() const {
+    return GetSDMCTotalSpace() - dir->GetSize();
+}
+
+u64 SDMCFactory::GetSDMCTotalSpace() const {
+    return static_cast<u64>(Settings::values.sdmc_size);
+}
+
 } // namespace FileSys

src/core/file_sys/sdmc_factory.h

@@ -11,6 +11,7 @@
 namespace FileSys {
 
 class RegisteredCache;
+class PlaceholderCache;
 
 /// File system interface to the SDCard archive
 class SDMCFactory {
@@ -18,13 +19,23 @@ public:
     explicit SDMCFactory(VirtualDir dir);
     ~SDMCFactory();
 
-    ResultVal<VirtualDir> Open();
+    ResultVal<VirtualDir> Open() const;
+
+    VirtualDir GetSDMCContentDirectory() const;
+
     RegisteredCache* GetSDMCContents() const;
+    PlaceholderCache* GetSDMCPlaceholder() const;
+
+    VirtualDir GetImageDirectory() const;
+
+    u64 GetSDMCFreeSpace() const;
+    u64 GetSDMCTotalSpace() const;
 
 private:
     VirtualDir dir;
 
     std::unique_ptr<RegisteredCache> contents;
+    std::unique_ptr<PlaceholderCache> placeholder;
 };
 
 } // namespace FileSys

src/core/file_sys/submission_package.cpp

@@ -14,6 +14,7 @@
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/partition_filesystem.h"
+#include "core/file_sys/program_metadata.h"
 #include "core/file_sys/submission_package.h"
 #include "core/loader/loader.h"
 
@@ -78,6 +79,10 @@ Loader::ResultStatus NSP::GetStatus() const {
 }
 
 Loader::ResultStatus NSP::GetProgramStatus(u64 title_id) const {
+    if (IsExtractedType() && GetExeFS() != nullptr && FileSys::IsDirectoryExeFS(GetExeFS())) {
+        return Loader::ResultStatus::Success;
+    }
+
     const auto iter = program_status.find(title_id);
     if (iter == program_status.end())
         return Loader::ResultStatus::ErrorNSPMissingProgramNCA;
@@ -85,12 +90,29 @@ Loader::ResultStatus NSP::GetProgramStatus(u64 title_id) const {
 }
 
 u64 NSP::GetFirstTitleID() const {
+    if (IsExtractedType()) {
+        return GetProgramTitleID();
+    }
+
     if (program_status.empty())
         return 0;
     return program_status.begin()->first;
 }
 
 u64 NSP::GetProgramTitleID() const {
+    if (IsExtractedType()) {
+        if (GetExeFS() == nullptr || !IsDirectoryExeFS(GetExeFS())) {
+            return 0;
+        }
+
+        ProgramMetadata meta;
+        if (meta.Load(GetExeFS()->GetFile("main.npdm")) == Loader::ResultStatus::Success) {
+            return meta.GetTitleID();
+        } else {
+            return 0;
+        }
+    }
+
     const auto out = GetFirstTitleID();
     if ((out & 0x800) == 0)
         return out;
@@ -102,6 +124,10 @@ u64 NSP::GetProgramTitleID() const {
 }
 
 std::vector<u64> NSP::GetTitleIDs() const {
+    if (IsExtractedType()) {
+        return {GetProgramTitleID()};
+    }
+
     std::vector<u64> out;
     out.reserve(ncas.size());
     for (const auto& kv : ncas)
@@ -222,7 +248,8 @@ void NSP::InitializeExeFSAndRomFS(const std::vector<VirtualFile>& files) {
 
 void NSP::ReadNCAs(const std::vector<VirtualFile>& files) {
     for (const auto& outer_file : files) {
-        if (outer_file->GetName().substr(outer_file->GetName().size() - 9) != ".cnmt.nca") {
+        if (outer_file->GetName().size() < 9 ||
+            outer_file->GetName().substr(outer_file->GetName().size() - 9) != ".cnmt.nca") {
             continue;
         }
 

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

@@ -31,6 +31,9 @@
 
 namespace Service::Account {
 
+constexpr ResultCode ERR_INVALID_BUFFER_SIZE{ErrorModule::Account, 30};
+constexpr ResultCode ERR_FAILED_SAVE_DATA{ErrorModule::Account, 100};
+
 static std::string GetImagePath(Common::UUID uuid) {
     return FileUtil::GetUserPath(FileUtil::UserPath::NANDDir) +
            "/system/save/8000000000000010/su/avators/" + uuid.FormatSwitch() + ".jpg";
@@ -41,20 +44,31 @@ static constexpr u32 SanitizeJPEGSize(std::size_t size) {
     return static_cast<u32>(std::min(size, max_jpeg_image_size));
 }
 
-class IProfile final : public ServiceFramework<IProfile> {
+class IProfileCommon : public ServiceFramework<IProfileCommon> {
 public:
-    explicit IProfile(Common::UUID user_id, ProfileManager& profile_manager)
-        : ServiceFramework("IProfile"), profile_manager(profile_manager), user_id(user_id) {
+    explicit IProfileCommon(const char* name, bool editor_commands, Common::UUID user_id,
+                            ProfileManager& profile_manager)
+        : ServiceFramework(name), profile_manager(profile_manager), user_id(user_id) {
         static const FunctionInfo functions[] = {
-            {0, &IProfile::Get, "Get"},
-            {1, &IProfile::GetBase, "GetBase"},
-            {10, &IProfile::GetImageSize, "GetImageSize"},
-            {11, &IProfile::LoadImage, "LoadImage"},
+            {0, &IProfileCommon::Get, "Get"},
+            {1, &IProfileCommon::GetBase, "GetBase"},
+            {10, &IProfileCommon::GetImageSize, "GetImageSize"},
+            {11, &IProfileCommon::LoadImage, "LoadImage"},
         };
+
         RegisterHandlers(functions);
+
+        if (editor_commands) {
+            static const FunctionInfo editor_functions[] = {
+                {100, &IProfileCommon::Store, "Store"},
+                {101, &IProfileCommon::StoreWithImage, "StoreWithImage"},
+            };
+
+            RegisterHandlers(editor_functions);
+        }
     }
 
-private:
+protected:
     void Get(Kernel::HLERequestContext& ctx) {
         LOG_INFO(Service_ACC, "called user_id={}", user_id.Format());
         ProfileBase profile_base{};
@@ -127,10 +141,91 @@ private:
         }
     }
 
-    const ProfileManager& profile_manager;
+    void Store(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto base = rp.PopRaw<ProfileBase>();
+
+        const auto user_data = ctx.ReadBuffer();
+
+        LOG_DEBUG(Service_ACC, "called, username='{}', timestamp={:016X}, uuid={}",
+                  Common::StringFromFixedZeroTerminatedBuffer(
+                      reinterpret_cast<const char*>(base.username.data()), base.username.size()),
+                  base.timestamp, base.user_uuid.Format());
+
+        if (user_data.size() < sizeof(ProfileData)) {
+            LOG_ERROR(Service_ACC, "ProfileData buffer too small!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_INVALID_BUFFER_SIZE);
+            return;
+        }
+
+        ProfileData data;
+        std::memcpy(&data, user_data.data(), sizeof(ProfileData));
+
+        if (!profile_manager.SetProfileBaseAndData(user_id, base, data)) {
+            LOG_ERROR(Service_ACC, "Failed to update profile data and base!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_FAILED_SAVE_DATA);
+            return;
+        }
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void StoreWithImage(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto base = rp.PopRaw<ProfileBase>();
+
+        const auto user_data = ctx.ReadBuffer();
+        const auto image_data = ctx.ReadBuffer(1);
+
+        LOG_DEBUG(Service_ACC, "called, username='{}', timestamp={:016X}, uuid={}",
+                  Common::StringFromFixedZeroTerminatedBuffer(
+                      reinterpret_cast<const char*>(base.username.data()), base.username.size()),
+                  base.timestamp, base.user_uuid.Format());
+
+        if (user_data.size() < sizeof(ProfileData)) {
+            LOG_ERROR(Service_ACC, "ProfileData buffer too small!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_INVALID_BUFFER_SIZE);
+            return;
+        }
+
+        ProfileData data;
+        std::memcpy(&data, user_data.data(), sizeof(ProfileData));
+
+        FileUtil::IOFile image(GetImagePath(user_id), "wb");
+
+        if (!image.IsOpen() || !image.Resize(image_data.size()) ||
+            image.WriteBytes(image_data.data(), image_data.size()) != image_data.size() ||
+            !profile_manager.SetProfileBaseAndData(user_id, base, data)) {
+            LOG_ERROR(Service_ACC, "Failed to update profile data, base, and image!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_FAILED_SAVE_DATA);
+            return;
+        }
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    ProfileManager& profile_manager;
     Common::UUID user_id; ///< The user id this profile refers to.
 };
 
+class IProfile final : public IProfileCommon {
+public:
+    IProfile(Common::UUID user_id, ProfileManager& profile_manager)
+        : IProfileCommon("IProfile", false, user_id, profile_manager) {}
+};
+
+class IProfileEditor final : public IProfileCommon {
+public:
+    IProfileEditor(Common::UUID user_id, ProfileManager& profile_manager)
+        : IProfileCommon("IProfileEditor", true, user_id, profile_manager) {}
+};
+
 class IManagerForApplication final : public ServiceFramework<IManagerForApplication> {
 public:
     IManagerForApplication() : ServiceFramework("IManagerForApplication") {
@@ -322,6 +417,17 @@ void Module::Interface::IsUserAccountSwitchLocked(Kernel::HLERequestContext& ctx
     rb.Push(is_locked);
 }
 
+void Module::Interface::GetProfileEditor(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    Common::UUID user_id = rp.PopRaw<Common::UUID>();
+
+    LOG_DEBUG(Service_ACC, "called, user_id={}", user_id.Format());
+
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushIpcInterface<IProfileEditor>(user_id, *profile_manager);
+}
+
 void Module::Interface::TrySelectUserWithoutInteraction(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_ACC, "called");
     // A u8 is passed into this function which we can safely ignore. It's to determine if we have

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

@@ -32,6 +32,7 @@ public:
         void IsUserRegistrationRequestPermitted(Kernel::HLERequestContext& ctx);
         void TrySelectUserWithoutInteraction(Kernel::HLERequestContext& ctx);
         void IsUserAccountSwitchLocked(Kernel::HLERequestContext& ctx);
+        void GetProfileEditor(Kernel::HLERequestContext& ctx);
 
     private:
         ResultCode InitializeApplicationInfoBase(u64 process_id);

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

@@ -41,7 +41,7 @@ ACC_SU::ACC_SU(std::shared_ptr<Module> module, std::shared_ptr<ProfileManager> p
         {202, nullptr, "CancelUserRegistration"},
         {203, nullptr, "DeleteUser"},
         {204, nullptr, "SetUserPosition"},
-        {205, nullptr, "GetProfileEditor"},
+        {205, &ACC_SU::GetProfileEditor, "GetProfileEditor"},
         {206, nullptr, "CompleteUserRegistrationForcibly"},
         {210, nullptr, "CreateFloatingRegistrationRequest"},
         {230, nullptr, "AuthenticateServiceAsync"},

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

@@ -305,6 +305,17 @@ bool ProfileManager::SetProfileBase(UUID uuid, const ProfileBase& profile_new) {
     return true;
 }
 
+bool ProfileManager::SetProfileBaseAndData(Common::UUID uuid, const ProfileBase& profile_new,
+                                           const ProfileData& data_new) {
+    const auto index = GetUserIndex(uuid);
+    if (index.has_value() && SetProfileBase(uuid, profile_new)) {
+        profiles[*index].data = data_new;
+        return true;
+    }
+
+    return false;
+}
+
 void ProfileManager::ParseUserSaveFile() {
     FileUtil::IOFile save(FileUtil::GetUserPath(FileUtil::UserPath::NANDDir) +
                               ACC_SAVE_AVATORS_BASE_PATH + "profiles.dat",

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

@@ -91,6 +91,8 @@ public:
 
     bool RemoveUser(Common::UUID uuid);
     bool SetProfileBase(Common::UUID uuid, const ProfileBase& profile_new);
+    bool SetProfileBaseAndData(Common::UUID uuid, const ProfileBase& profile_new,
+                               const ProfileData& data_new);
 
 private:
     void ParseUserSaveFile();

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

@@ -1143,13 +1143,21 @@ void IApplicationFunctions::CreateApplicationAndRequestToStartForQuest(
 
 void IApplicationFunctions::EnsureSaveData(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
-    u128 uid = rp.PopRaw<u128>(); // What does this do?
-    LOG_WARNING(Service, "(STUBBED) called uid = {:016X}{:016X}", uid[1], uid[0]);
+    u128 user_id = rp.PopRaw<u128>();
+
+    LOG_DEBUG(Service_AM, "called, uid={:016X}{:016X}", user_id[1], user_id[0]);
+
+    FileSys::SaveDataDescriptor descriptor{};
+    descriptor.title_id = Core::CurrentProcess()->GetTitleID();
+    descriptor.user_id = user_id;
+    descriptor.type = FileSys::SaveDataType::SaveData;
+    const auto res = system.GetFileSystemController().CreateSaveData(
+        FileSys::SaveDataSpaceId::NandUser, descriptor);
 
     IPC::ResponseBuilder rb{ctx, 4};
-    rb.Push(RESULT_SUCCESS);
+    rb.Push(res.Code());
     rb.Push<u64>(0);
-} // namespace Service::AM
+}
 
 void IApplicationFunctions::SetTerminateResult(Kernel::HLERequestContext& ctx) {
     // Takes an input u32 Result, no output.
@@ -1261,8 +1269,8 @@ void IApplicationFunctions::ExtendSaveData(Kernel::HLERequestContext& ctx) {
               "new_journal={:016X}",
               static_cast<u8>(type), user_id[1], user_id[0], new_normal_size, new_journal_size);
 
-    const auto title_id = system.CurrentProcess()->GetTitleID();
-    FileSystem::WriteSaveDataSize(type, title_id, user_id, {new_normal_size, new_journal_size});
+    system.GetFileSystemController().WriteSaveDataSize(
+        type, system.CurrentProcess()->GetTitleID(), user_id, {new_normal_size, new_journal_size});
 
     IPC::ResponseBuilder rb{ctx, 4};
     rb.Push(RESULT_SUCCESS);
@@ -1281,8 +1289,8 @@ void IApplicationFunctions::GetSaveDataSize(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called with type={:02X}, user_id={:016X}{:016X}", static_cast<u8>(type),
               user_id[1], user_id[0]);
 
-    const auto title_id = system.CurrentProcess()->GetTitleID();
-    const auto size = FileSystem::ReadSaveDataSize(type, title_id, user_id);
+    const auto size = system.GetFileSystemController().ReadSaveDataSize(
+        type, system.CurrentProcess()->GetTitleID(), user_id);
 
     IPC::ResponseBuilder rb{ctx, 6};
     rb.Push(RESULT_SUCCESS);

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

@@ -9,6 +9,10 @@
 #include "core/hle/service/service.h"
 
 namespace Service {
+namespace FileSystem {
+class FileSystemController;
+}
+
 namespace NVFlinger {
 class NVFlinger;
 }

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

@@ -9,6 +9,10 @@
 #include "core/hle/service/service.h"
 
 namespace Service {
+namespace FileSystem {
+class FileSystemController;
+}
+
 namespace NVFlinger {
 class NVFlinger;
 }

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

@@ -8,6 +8,7 @@
 #include "common/file_util.h"
 #include "core/core.h"
 #include "core/file_sys/bis_factory.h"
+#include "core/file_sys/card_image.h"
 #include "core/file_sys/control_metadata.h"
 #include "core/file_sys/errors.h"
 #include "core/file_sys/mode.h"
@@ -25,14 +26,10 @@
 #include "core/hle/service/filesystem/fsp_pr.h"
 #include "core/hle/service/filesystem/fsp_srv.h"
 #include "core/loader/loader.h"
+#include "core/settings.h"
 
 namespace Service::FileSystem {
 
-// Size of emulated sd card free space, reported in bytes.
-// Just using 32GB because thats reasonable
-// TODO(DarkLordZach): Eventually make this configurable in settings.
-constexpr u64 EMULATED_SD_REPORTED_SIZE = 32000000000;
-
 // A default size for normal/journal save data size if application control metadata cannot be found.
 // This should be large enough to satisfy even the most extreme requirements (~4.2GB)
 constexpr u64 SUFFICIENT_SAVE_DATA_SIZE = 0xF0000000;
@@ -226,13 +223,6 @@ ResultVal<FileSys::VirtualDir> VfsDirectoryServiceWrapper::OpenDirectory(const s
     return MakeResult(dir);
 }
 
-u64 VfsDirectoryServiceWrapper::GetFreeSpaceSize() const {
-    if (backing->IsWritable())
-        return EMULATED_SD_REPORTED_SIZE;
-
-    return 0;
-}
-
 ResultVal<FileSys::EntryType> VfsDirectoryServiceWrapper::GetEntryType(
     const std::string& path_) const {
     std::string path(FileUtil::SanitizePath(path_));
@@ -251,44 +241,39 @@ ResultVal<FileSys::EntryType> VfsDirectoryServiceWrapper::GetEntryType(
     return FileSys::ERROR_PATH_NOT_FOUND;
 }
 
-/**
- * Map of registered file systems, identified by type. Once an file system is registered here, it
- * is never removed until UnregisterFileSystems is called.
- */
-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;
+FileSystemController::FileSystemController() = default;
 
-ResultCode RegisterRomFS(std::unique_ptr<FileSys::RomFSFactory>&& factory) {
-    ASSERT_MSG(romfs_factory == nullptr, "Tried to register a second RomFS");
+FileSystemController::~FileSystemController() = default;
+
+ResultCode FileSystemController::RegisterRomFS(std::unique_ptr<FileSys::RomFSFactory>&& factory) {
     romfs_factory = std::move(factory);
     LOG_DEBUG(Service_FS, "Registered RomFS");
     return RESULT_SUCCESS;
 }
 
-ResultCode RegisterSaveData(std::unique_ptr<FileSys::SaveDataFactory>&& factory) {
-    ASSERT_MSG(romfs_factory == nullptr, "Tried to register a second save data");
+ResultCode FileSystemController::RegisterSaveData(
+    std::unique_ptr<FileSys::SaveDataFactory>&& factory) {
+    ASSERT_MSG(save_data_factory == nullptr, "Tried to register a second save data");
     save_data_factory = std::move(factory);
     LOG_DEBUG(Service_FS, "Registered save data");
     return RESULT_SUCCESS;
 }
 
-ResultCode RegisterSDMC(std::unique_ptr<FileSys::SDMCFactory>&& factory) {
+ResultCode FileSystemController::RegisterSDMC(std::unique_ptr<FileSys::SDMCFactory>&& factory) {
     ASSERT_MSG(sdmc_factory == nullptr, "Tried to register a second SDMC");
     sdmc_factory = std::move(factory);
     LOG_DEBUG(Service_FS, "Registered SDMC");
     return RESULT_SUCCESS;
 }
 
-ResultCode RegisterBIS(std::unique_ptr<FileSys::BISFactory>&& factory) {
+ResultCode FileSystemController::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, "Registered BIS");
     return RESULT_SUCCESS;
 }
 
-void SetPackedUpdate(FileSys::VirtualFile update_raw) {
+void FileSystemController::SetPackedUpdate(FileSys::VirtualFile update_raw) {
     LOG_TRACE(Service_FS, "Setting packed update for romfs");
 
     if (romfs_factory == nullptr)
@@ -297,7 +282,7 @@ void SetPackedUpdate(FileSys::VirtualFile update_raw) {
     romfs_factory->SetPackedUpdate(std::move(update_raw));
 }
 
-ResultVal<FileSys::VirtualFile> OpenRomFSCurrentProcess() {
+ResultVal<FileSys::VirtualFile> FileSystemController::OpenRomFSCurrentProcess() const {
     LOG_TRACE(Service_FS, "Opening RomFS for current process");
 
     if (romfs_factory == nullptr) {
@@ -308,8 +293,8 @@ ResultVal<FileSys::VirtualFile> OpenRomFSCurrentProcess() {
     return romfs_factory->OpenCurrentProcess();
 }
 
-ResultVal<FileSys::VirtualFile> OpenRomFS(u64 title_id, FileSys::StorageId storage_id,
-                                          FileSys::ContentRecordType type) {
+ResultVal<FileSys::VirtualFile> FileSystemController::OpenRomFS(
+    u64 title_id, FileSys::StorageId storage_id, FileSys::ContentRecordType type) const {
     LOG_TRACE(Service_FS, "Opening RomFS for title_id={:016X}, storage_id={:02X}, type={:02X}",
               title_id, static_cast<u8>(storage_id), static_cast<u8>(type));
 
@@ -321,8 +306,20 @@ ResultVal<FileSys::VirtualFile> OpenRomFS(u64 title_id, FileSys::StorageId stora
     return romfs_factory->Open(title_id, storage_id, type);
 }
 
-ResultVal<FileSys::VirtualDir> OpenSaveData(FileSys::SaveDataSpaceId space,
-                                            const FileSys::SaveDataDescriptor& descriptor) {
+ResultVal<FileSys::VirtualDir> FileSystemController::CreateSaveData(
+    FileSys::SaveDataSpaceId space, const FileSys::SaveDataDescriptor& save_struct) const {
+    LOG_TRACE(Service_FS, "Creating Save Data for space_id={:01X}, save_struct={}",
+              static_cast<u8>(space), save_struct.DebugInfo());
+
+    if (save_data_factory == nullptr) {
+        return FileSys::ERROR_ENTITY_NOT_FOUND;
+    }
+
+    return save_data_factory->Create(space, save_struct);
+}
+
+ResultVal<FileSys::VirtualDir> FileSystemController::OpenSaveData(
+    FileSys::SaveDataSpaceId space, const FileSys::SaveDataDescriptor& descriptor) const {
     LOG_TRACE(Service_FS, "Opening Save Data for space_id={:01X}, save_struct={}",
               static_cast<u8>(space), descriptor.DebugInfo());
 
@@ -333,7 +330,8 @@ ResultVal<FileSys::VirtualDir> OpenSaveData(FileSys::SaveDataSpaceId space,
     return save_data_factory->Open(space, descriptor);
 }
 
-ResultVal<FileSys::VirtualDir> OpenSaveDataSpace(FileSys::SaveDataSpaceId space) {
+ResultVal<FileSys::VirtualDir> FileSystemController::OpenSaveDataSpace(
+    FileSys::SaveDataSpaceId space) const {
     LOG_TRACE(Service_FS, "Opening Save Data Space for space_id={:01X}", static_cast<u8>(space));
 
     if (save_data_factory == nullptr) {
@@ -343,7 +341,7 @@ ResultVal<FileSys::VirtualDir> OpenSaveDataSpace(FileSys::SaveDataSpaceId space)
     return MakeResult(save_data_factory->GetSaveDataSpaceDirectory(space));
 }
 
-ResultVal<FileSys::VirtualDir> OpenSDMC() {
+ResultVal<FileSys::VirtualDir> FileSystemController::OpenSDMC() const {
     LOG_TRACE(Service_FS, "Opening SDMC");
 
     if (sdmc_factory == nullptr) {
@@ -353,7 +351,92 @@ ResultVal<FileSys::VirtualDir> OpenSDMC() {
     return sdmc_factory->Open();
 }
 
-FileSys::SaveDataSize ReadSaveDataSize(FileSys::SaveDataType type, u64 title_id, u128 user_id) {
+ResultVal<FileSys::VirtualDir> FileSystemController::OpenBISPartition(
+    FileSys::BisPartitionId id) const {
+    LOG_TRACE(Service_FS, "Opening BIS Partition with id={:08X}", static_cast<u32>(id));
+
+    if (bis_factory == nullptr) {
+        return FileSys::ERROR_ENTITY_NOT_FOUND;
+    }
+
+    auto part = bis_factory->OpenPartition(id);
+    if (part == nullptr) {
+        return FileSys::ERROR_INVALID_ARGUMENT;
+    }
+
+    return MakeResult<FileSys::VirtualDir>(std::move(part));
+}
+
+ResultVal<FileSys::VirtualFile> FileSystemController::OpenBISPartitionStorage(
+    FileSys::BisPartitionId id) const {
+    LOG_TRACE(Service_FS, "Opening BIS Partition Storage with id={:08X}", static_cast<u32>(id));
+
+    if (bis_factory == nullptr) {
+        return FileSys::ERROR_ENTITY_NOT_FOUND;
+    }
+
+    auto part = bis_factory->OpenPartitionStorage(id);
+    if (part == nullptr) {
+        return FileSys::ERROR_INVALID_ARGUMENT;
+    }
+
+    return MakeResult<FileSys::VirtualFile>(std::move(part));
+}
+
+u64 FileSystemController::GetFreeSpaceSize(FileSys::StorageId id) const {
+    switch (id) {
+    case FileSys::StorageId::None:
+    case FileSys::StorageId::GameCard:
+        return 0;
+    case FileSys::StorageId::SdCard:
+        if (sdmc_factory == nullptr)
+            return 0;
+        return sdmc_factory->GetSDMCFreeSpace();
+    case FileSys::StorageId::Host:
+        if (bis_factory == nullptr)
+            return 0;
+        return bis_factory->GetSystemNANDFreeSpace() + bis_factory->GetUserNANDFreeSpace();
+    case FileSys::StorageId::NandSystem:
+        if (bis_factory == nullptr)
+            return 0;
+        return bis_factory->GetSystemNANDFreeSpace();
+    case FileSys::StorageId::NandUser:
+        if (bis_factory == nullptr)
+            return 0;
+        return bis_factory->GetUserNANDFreeSpace();
+    }
+
+    return 0;
+}
+
+u64 FileSystemController::GetTotalSpaceSize(FileSys::StorageId id) const {
+    switch (id) {
+    case FileSys::StorageId::None:
+    case FileSys::StorageId::GameCard:
+        return 0;
+    case FileSys::StorageId::SdCard:
+        if (sdmc_factory == nullptr)
+            return 0;
+        return sdmc_factory->GetSDMCTotalSpace();
+    case FileSys::StorageId::Host:
+        if (bis_factory == nullptr)
+            return 0;
+        return bis_factory->GetFullNANDTotalSpace();
+    case FileSys::StorageId::NandSystem:
+        if (bis_factory == nullptr)
+            return 0;
+        return bis_factory->GetSystemNANDTotalSpace();
+    case FileSys::StorageId::NandUser:
+        if (bis_factory == nullptr)
+            return 0;
+        return bis_factory->GetUserNANDTotalSpace();
+    }
+
+    return 0;
+}
+
+FileSys::SaveDataSize FileSystemController::ReadSaveDataSize(FileSys::SaveDataType type,
+                                                             u64 title_id, u128 user_id) const {
     if (save_data_factory == nullptr) {
         return {0, 0};
     }
@@ -385,13 +468,32 @@ FileSys::SaveDataSize ReadSaveDataSize(FileSys::SaveDataType type, u64 title_id,
     return value;
 }
 
-void WriteSaveDataSize(FileSys::SaveDataType type, u64 title_id, u128 user_id,
-                       FileSys::SaveDataSize new_value) {
+void FileSystemController::WriteSaveDataSize(FileSys::SaveDataType type, u64 title_id, u128 user_id,
+                                             FileSys::SaveDataSize new_value) const {
     if (save_data_factory != nullptr)
         save_data_factory->WriteSaveDataSize(type, title_id, user_id, new_value);
 }
 
-FileSys::RegisteredCache* GetSystemNANDContents() {
+void FileSystemController::SetGameCard(FileSys::VirtualFile file) {
+    gamecard = std::make_unique<FileSys::XCI>(file);
+    const auto dir = gamecard->ConcatenatedPseudoDirectory();
+    gamecard_registered = std::make_unique<FileSys::RegisteredCache>(dir);
+    gamecard_placeholder = std::make_unique<FileSys::PlaceholderCache>(dir);
+}
+
+FileSys::XCI* FileSystemController::GetGameCard() const {
+    return gamecard.get();
+}
+
+FileSys::RegisteredCache* FileSystemController::GetGameCardContents() const {
+    return gamecard_registered.get();
+}
+
+FileSys::PlaceholderCache* FileSystemController::GetGameCardPlaceholder() const {
+    return gamecard_placeholder.get();
+}
+
+FileSys::RegisteredCache* FileSystemController::GetSystemNANDContents() const {
     LOG_TRACE(Service_FS, "Opening System NAND Contents");
 
     if (bis_factory == nullptr)
@@ -400,7 +502,7 @@ FileSys::RegisteredCache* GetSystemNANDContents() {
     return bis_factory->GetSystemNANDContents();
 }
 
-FileSys::RegisteredCache* GetUserNANDContents() {
+FileSys::RegisteredCache* FileSystemController::GetUserNANDContents() const {
     LOG_TRACE(Service_FS, "Opening User NAND Contents");
 
     if (bis_factory == nullptr)
@@ -409,7 +511,7 @@ FileSys::RegisteredCache* GetUserNANDContents() {
     return bis_factory->GetUserNANDContents();
 }
 
-FileSys::RegisteredCache* GetSDMCContents() {
+FileSys::RegisteredCache* FileSystemController::GetSDMCContents() const {
     LOG_TRACE(Service_FS, "Opening SDMC Contents");
 
     if (sdmc_factory == nullptr)
@@ -418,7 +520,143 @@ FileSys::RegisteredCache* GetSDMCContents() {
     return sdmc_factory->GetSDMCContents();
 }
 
-FileSys::VirtualDir GetModificationLoadRoot(u64 title_id) {
+FileSys::PlaceholderCache* FileSystemController::GetSystemNANDPlaceholder() const {
+    LOG_TRACE(Service_FS, "Opening System NAND Placeholder");
+
+    if (bis_factory == nullptr)
+        return nullptr;
+
+    return bis_factory->GetSystemNANDPlaceholder();
+}
+
+FileSys::PlaceholderCache* FileSystemController::GetUserNANDPlaceholder() const {
+    LOG_TRACE(Service_FS, "Opening User NAND Placeholder");
+
+    if (bis_factory == nullptr)
+        return nullptr;
+
+    return bis_factory->GetUserNANDPlaceholder();
+}
+
+FileSys::PlaceholderCache* FileSystemController::GetSDMCPlaceholder() const {
+    LOG_TRACE(Service_FS, "Opening SDMC Placeholder");
+
+    if (sdmc_factory == nullptr)
+        return nullptr;
+
+    return sdmc_factory->GetSDMCPlaceholder();
+}
+
+FileSys::RegisteredCache* FileSystemController::GetRegisteredCacheForStorage(
+    FileSys::StorageId id) const {
+    switch (id) {
+    case FileSys::StorageId::None:
+    case FileSys::StorageId::Host:
+        UNIMPLEMENTED();
+        return nullptr;
+    case FileSys::StorageId::GameCard:
+        return GetGameCardContents();
+    case FileSys::StorageId::NandSystem:
+        return GetSystemNANDContents();
+    case FileSys::StorageId::NandUser:
+        return GetUserNANDContents();
+    case FileSys::StorageId::SdCard:
+        return GetSDMCContents();
+    }
+
+    return nullptr;
+}
+
+FileSys::PlaceholderCache* FileSystemController::GetPlaceholderCacheForStorage(
+    FileSys::StorageId id) const {
+    switch (id) {
+    case FileSys::StorageId::None:
+    case FileSys::StorageId::Host:
+        UNIMPLEMENTED();
+        return nullptr;
+    case FileSys::StorageId::GameCard:
+        return GetGameCardPlaceholder();
+    case FileSys::StorageId::NandSystem:
+        return GetSystemNANDPlaceholder();
+    case FileSys::StorageId::NandUser:
+        return GetUserNANDPlaceholder();
+    case FileSys::StorageId::SdCard:
+        return GetSDMCPlaceholder();
+    }
+
+    return nullptr;
+}
+
+FileSys::VirtualDir FileSystemController::GetSystemNANDContentDirectory() const {
+    LOG_TRACE(Service_FS, "Opening system NAND content directory");
+
+    if (bis_factory == nullptr)
+        return nullptr;
+
+    return bis_factory->GetSystemNANDContentDirectory();
+}
+
+FileSys::VirtualDir FileSystemController::GetUserNANDContentDirectory() const {
+    LOG_TRACE(Service_FS, "Opening user NAND content directory");
+
+    if (bis_factory == nullptr)
+        return nullptr;
+
+    return bis_factory->GetUserNANDContentDirectory();
+}
+
+FileSys::VirtualDir FileSystemController::GetSDMCContentDirectory() const {
+    LOG_TRACE(Service_FS, "Opening SDMC content directory");
+
+    if (sdmc_factory == nullptr)
+        return nullptr;
+
+    return sdmc_factory->GetSDMCContentDirectory();
+}
+
+FileSys::VirtualDir FileSystemController::GetNANDImageDirectory() const {
+    LOG_TRACE(Service_FS, "Opening NAND image directory");
+
+    if (bis_factory == nullptr)
+        return nullptr;
+
+    return bis_factory->GetImageDirectory();
+}
+
+FileSys::VirtualDir FileSystemController::GetSDMCImageDirectory() const {
+    LOG_TRACE(Service_FS, "Opening SDMC image directory");
+
+    if (sdmc_factory == nullptr)
+        return nullptr;
+
+    return sdmc_factory->GetImageDirectory();
+}
+
+FileSys::VirtualDir FileSystemController::GetContentDirectory(ContentStorageId id) const {
+    switch (id) {
+    case ContentStorageId::System:
+        return GetSystemNANDContentDirectory();
+    case ContentStorageId::User:
+        return GetUserNANDContentDirectory();
+    case ContentStorageId::SdCard:
+        return GetSDMCContentDirectory();
+    }
+
+    return nullptr;
+}
+
+FileSys::VirtualDir FileSystemController::GetImageDirectory(ImageDirectoryId id) const {
+    switch (id) {
+    case ImageDirectoryId::NAND:
+        return GetNANDImageDirectory();
+    case ImageDirectoryId::SdCard:
+        return GetSDMCImageDirectory();
+    }
+
+    return nullptr;
+}
+
+FileSys::VirtualDir FileSystemController::GetModificationLoadRoot(u64 title_id) const {
     LOG_TRACE(Service_FS, "Opening mod load root for tid={:016X}", title_id);
 
     if (bis_factory == nullptr)
@@ -427,7 +665,7 @@ FileSys::VirtualDir GetModificationLoadRoot(u64 title_id) {
     return bis_factory->GetModificationLoadRoot(title_id);
 }
 
-FileSys::VirtualDir GetModificationDumpRoot(u64 title_id) {
+FileSys::VirtualDir FileSystemController::GetModificationDumpRoot(u64 title_id) const {
     LOG_TRACE(Service_FS, "Opening mod dump root for tid={:016X}", title_id);
 
     if (bis_factory == nullptr)
@@ -436,7 +674,7 @@ FileSys::VirtualDir GetModificationDumpRoot(u64 title_id) {
     return bis_factory->GetModificationDumpRoot(title_id);
 }
 
-void CreateFactories(FileSys::VfsFilesystem& vfs, bool overwrite) {
+void FileSystemController::CreateFactories(FileSys::VfsFilesystem& vfs, bool overwrite) {
     if (overwrite) {
         bis_factory = nullptr;
         save_data_factory = nullptr;
@@ -473,11 +711,10 @@ void CreateFactories(FileSys::VfsFilesystem& vfs, bool overwrite) {
 }
 
 void InstallInterfaces(Core::System& system) {
-    romfs_factory = nullptr;
-    CreateFactories(*system.GetFilesystem(), false);
     std::make_shared<FSP_LDR>()->InstallAsService(system.ServiceManager());
     std::make_shared<FSP_PR>()->InstallAsService(system.ServiceManager());
-    std::make_shared<FSP_SRV>(system.GetReporter())->InstallAsService(system.ServiceManager());
+    std::make_shared<FSP_SRV>(system.GetFileSystemController(), system.GetReporter())
+        ->InstallAsService(system.ServiceManager());
 }
 
 } // namespace Service::FileSystem

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

@@ -14,10 +14,13 @@ namespace FileSys {
 class BISFactory;
 class RegisteredCache;
 class RegisteredCacheUnion;
+class PlaceholderCache;
 class RomFSFactory;
 class SaveDataFactory;
 class SDMCFactory;
+class XCI;
 
+enum class BisPartitionId : u32;
 enum class ContentRecordType : u8;
 enum class Mode : u32;
 enum class SaveDataSpaceId : u8;
@@ -36,34 +39,91 @@ class ServiceManager;
 
 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);
+enum class ContentStorageId : u32 {
+    System,
+    User,
+    SdCard,
+};
 
-void SetPackedUpdate(FileSys::VirtualFile update_raw);
-ResultVal<FileSys::VirtualFile> OpenRomFSCurrentProcess();
-ResultVal<FileSys::VirtualFile> OpenRomFS(u64 title_id, FileSys::StorageId storage_id,
-                                          FileSys::ContentRecordType type);
-ResultVal<FileSys::VirtualDir> OpenSaveData(FileSys::SaveDataSpaceId space,
-                                            const FileSys::SaveDataDescriptor& descriptor);
-ResultVal<FileSys::VirtualDir> OpenSaveDataSpace(FileSys::SaveDataSpaceId space);
-ResultVal<FileSys::VirtualDir> OpenSDMC();
+enum class ImageDirectoryId : u32 {
+    NAND,
+    SdCard,
+};
 
-FileSys::SaveDataSize ReadSaveDataSize(FileSys::SaveDataType type, u64 title_id, u128 user_id);
-void WriteSaveDataSize(FileSys::SaveDataType type, u64 title_id, u128 user_id,
-                       FileSys::SaveDataSize new_value);
+class FileSystemController {
+public:
+    FileSystemController();
+    ~FileSystemController();
 
-FileSys::RegisteredCache* GetSystemNANDContents();
-FileSys::RegisteredCache* GetUserNANDContents();
-FileSys::RegisteredCache* GetSDMCContents();
+    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);
 
-FileSys::VirtualDir GetModificationLoadRoot(u64 title_id);
-FileSys::VirtualDir GetModificationDumpRoot(u64 title_id);
+    void SetPackedUpdate(FileSys::VirtualFile update_raw);
+    ResultVal<FileSys::VirtualFile> OpenRomFSCurrentProcess() const;
+    ResultVal<FileSys::VirtualFile> OpenRomFS(u64 title_id, FileSys::StorageId storage_id,
+                                              FileSys::ContentRecordType type) const;
+    ResultVal<FileSys::VirtualDir> CreateSaveData(
+        FileSys::SaveDataSpaceId space, const FileSys::SaveDataDescriptor& save_struct) const;
+    ResultVal<FileSys::VirtualDir> OpenSaveData(
+        FileSys::SaveDataSpaceId space, const FileSys::SaveDataDescriptor& save_struct) const;
+    ResultVal<FileSys::VirtualDir> OpenSaveDataSpace(FileSys::SaveDataSpaceId space) const;
+    ResultVal<FileSys::VirtualDir> OpenSDMC() const;
+    ResultVal<FileSys::VirtualDir> OpenBISPartition(FileSys::BisPartitionId id) const;
+    ResultVal<FileSys::VirtualFile> OpenBISPartitionStorage(FileSys::BisPartitionId id) const;
 
-// Creates the SaveData, SDMC, and BIS Factories. Should be called once and before any function
-// above is called.
-void CreateFactories(FileSys::VfsFilesystem& vfs, bool overwrite = true);
+    u64 GetFreeSpaceSize(FileSys::StorageId id) const;
+    u64 GetTotalSpaceSize(FileSys::StorageId id) const;
+
+    FileSys::SaveDataSize ReadSaveDataSize(FileSys::SaveDataType type, u64 title_id,
+                                           u128 user_id) const;
+    void WriteSaveDataSize(FileSys::SaveDataType type, u64 title_id, u128 user_id,
+                           FileSys::SaveDataSize new_value) const;
+
+    void SetGameCard(FileSys::VirtualFile file);
+    FileSys::XCI* GetGameCard() const;
+
+    FileSys::RegisteredCache* GetSystemNANDContents() const;
+    FileSys::RegisteredCache* GetUserNANDContents() const;
+    FileSys::RegisteredCache* GetSDMCContents() const;
+    FileSys::RegisteredCache* GetGameCardContents() const;
+
+    FileSys::PlaceholderCache* GetSystemNANDPlaceholder() const;
+    FileSys::PlaceholderCache* GetUserNANDPlaceholder() const;
+    FileSys::PlaceholderCache* GetSDMCPlaceholder() const;
+    FileSys::PlaceholderCache* GetGameCardPlaceholder() const;
+
+    FileSys::RegisteredCache* GetRegisteredCacheForStorage(FileSys::StorageId id) const;
+    FileSys::PlaceholderCache* GetPlaceholderCacheForStorage(FileSys::StorageId id) const;
+
+    FileSys::VirtualDir GetSystemNANDContentDirectory() const;
+    FileSys::VirtualDir GetUserNANDContentDirectory() const;
+    FileSys::VirtualDir GetSDMCContentDirectory() const;
+
+    FileSys::VirtualDir GetNANDImageDirectory() const;
+    FileSys::VirtualDir GetSDMCImageDirectory() const;
+
+    FileSys::VirtualDir GetContentDirectory(ContentStorageId id) const;
+    FileSys::VirtualDir GetImageDirectory(ImageDirectoryId id) const;
+
+    FileSys::VirtualDir GetModificationLoadRoot(u64 title_id) const;
+    FileSys::VirtualDir GetModificationDumpRoot(u64 title_id) const;
+
+    // Creates the SaveData, SDMC, and BIS Factories. Should be called once and before any function
+    // above is called.
+    void CreateFactories(FileSys::VfsFilesystem& vfs, bool overwrite = true);
+
+private:
+    std::unique_ptr<FileSys::RomFSFactory> romfs_factory;
+    std::unique_ptr<FileSys::SaveDataFactory> save_data_factory;
+    std::unique_ptr<FileSys::SDMCFactory> sdmc_factory;
+    std::unique_ptr<FileSys::BISFactory> bis_factory;
+
+    std::unique_ptr<FileSys::XCI> gamecard;
+    std::unique_ptr<FileSys::RegisteredCache> gamecard_registered;
+    std::unique_ptr<FileSys::PlaceholderCache> gamecard_placeholder;
+};
 
 void InstallInterfaces(Core::System& system);
 
@@ -159,12 +219,6 @@ public:
      */
     ResultVal<FileSys::VirtualDir> OpenDirectory(const std::string& path);
 
-    /**
-     * Get the free space
-     * @return The number of free bytes in the archive
-     */
-    u64 GetFreeSpaceSize() const;
-
     /**
      * Get the type of the specified path
      * @return The type of the specified path or error code

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

@@ -19,6 +19,7 @@
 #include "core/file_sys/mode.h"
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/patch_manager.h"
+#include "core/file_sys/romfs_factory.h"
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/system_archive/system_archive.h"
 #include "core/file_sys/vfs.h"
@@ -30,6 +31,18 @@
 
 namespace Service::FileSystem {
 
+struct SizeGetter {
+    std::function<u64()> get_free_size;
+    std::function<u64()> get_total_size;
+
+    static SizeGetter FromStorageId(const FileSystemController& fsc, FileSys::StorageId id) {
+        return {
+            [&fsc, id] { return fsc.GetFreeSpaceSize(id); },
+            [&fsc, id] { return fsc.GetTotalSpaceSize(id); },
+        };
+    }
+};
+
 enum class FileSystemType : u8 {
     Invalid0 = 0,
     Invalid1 = 1,
@@ -289,8 +302,8 @@ private:
 
 class IFileSystem final : public ServiceFramework<IFileSystem> {
 public:
-    explicit IFileSystem(FileSys::VirtualDir backend)
-        : ServiceFramework("IFileSystem"), backend(std::move(backend)) {
+    explicit IFileSystem(FileSys::VirtualDir backend, SizeGetter size)
+        : ServiceFramework("IFileSystem"), backend(std::move(backend)), size(std::move(size)) {
         static const FunctionInfo functions[] = {
             {0, &IFileSystem::CreateFile, "CreateFile"},
             {1, &IFileSystem::DeleteFile, "DeleteFile"},
@@ -467,14 +480,31 @@ public:
         rb.Push(RESULT_SUCCESS);
     }
 
+    void GetFreeSpaceSize(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_FS, "called");
+
+        IPC::ResponseBuilder rb{ctx, 4};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push(size.get_free_size());
+    }
+
+    void GetTotalSpaceSize(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_FS, "called");
+
+        IPC::ResponseBuilder rb{ctx, 4};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push(size.get_total_size());
+    }
+
 private:
     VfsDirectoryServiceWrapper backend;
+    SizeGetter size;
 };
 
 class ISaveDataInfoReader final : public ServiceFramework<ISaveDataInfoReader> {
 public:
-    explicit ISaveDataInfoReader(FileSys::SaveDataSpaceId space)
-        : ServiceFramework("ISaveDataInfoReader") {
+    explicit ISaveDataInfoReader(FileSys::SaveDataSpaceId space, FileSystemController& fsc)
+        : ServiceFramework("ISaveDataInfoReader"), fsc(fsc) {
         static const FunctionInfo functions[] = {
             {0, &ISaveDataInfoReader::ReadSaveDataInfo, "ReadSaveDataInfo"},
         };
@@ -520,8 +550,13 @@ private:
     }
 
     void FindAllSaves(FileSys::SaveDataSpaceId space) {
-        const auto save_root = OpenSaveDataSpace(space);
-        ASSERT(save_root.Succeeded());
+        const auto save_root = fsc.OpenSaveDataSpace(space);
+
+        if (save_root.Failed() || *save_root == nullptr) {
+            LOG_ERROR(Service_FS, "The save root for the space_id={:02X} was invalid!",
+                      static_cast<u8>(space));
+            return;
+        }
 
         for (const auto& type : (*save_root)->GetSubdirectories()) {
             if (type->GetName() == "save") {
@@ -610,11 +645,13 @@ private:
     };
     static_assert(sizeof(SaveDataInfo) == 0x60, "SaveDataInfo has incorrect size.");
 
+    FileSystemController& fsc;
     std::vector<SaveDataInfo> info;
     u64 next_entry_index = 0;
 };
 
-FSP_SRV::FSP_SRV(const Core::Reporter& reporter) : ServiceFramework("fsp-srv"), reporter(reporter) {
+FSP_SRV::FSP_SRV(FileSystemController& fsc, const Core::Reporter& reporter)
+    : ServiceFramework("fsp-srv"), fsc(fsc), reporter(reporter) {
     // clang-format off
     static const FunctionInfo functions[] = {
         {0, nullptr, "OpenFileSystem"},
@@ -754,7 +791,8 @@ void FSP_SRV::OpenFileSystemWithPatch(Kernel::HLERequestContext& ctx) {
 void FSP_SRV::OpenSdCardFileSystem(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_FS, "called");
 
-    IFileSystem filesystem(OpenSDMC().Unwrap());
+    IFileSystem filesystem(fsc.OpenSDMC().Unwrap(),
+                           SizeGetter::FromStorageId(fsc, FileSys::StorageId::SdCard));
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
@@ -768,8 +806,10 @@ void FSP_SRV::CreateSaveDataFileSystem(Kernel::HLERequestContext& ctx) {
     auto save_create_struct = rp.PopRaw<std::array<u8, 0x40>>();
     u128 uid = rp.PopRaw<u128>();
 
-    LOG_WARNING(Service_FS, "(STUBBED) called save_struct = {}, uid = {:016X}{:016X}",
-                save_struct.DebugInfo(), uid[1], uid[0]);
+    LOG_DEBUG(Service_FS, "called save_struct = {}, uid = {:016X}{:016X}", save_struct.DebugInfo(),
+              uid[1], uid[0]);
+
+    fsc.CreateSaveData(FileSys::SaveDataSpaceId::NandUser, save_struct);
 
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(RESULT_SUCCESS);
@@ -786,14 +826,24 @@ void FSP_SRV::OpenSaveDataFileSystem(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto parameters = rp.PopRaw<Parameters>();
 
-    auto dir = OpenSaveData(parameters.save_data_space_id, parameters.descriptor);
+    auto dir = fsc.OpenSaveData(parameters.save_data_space_id, parameters.descriptor);
     if (dir.Failed()) {
         IPC::ResponseBuilder rb{ctx, 2, 0, 0};
         rb.Push(FileSys::ERROR_ENTITY_NOT_FOUND);
         return;
     }
 
-    IFileSystem filesystem(std::move(dir.Unwrap()));
+    FileSys::StorageId id;
+    if (parameters.save_data_space_id == FileSys::SaveDataSpaceId::NandUser) {
+        id = FileSys::StorageId::NandUser;
+    } else if (parameters.save_data_space_id == FileSys::SaveDataSpaceId::SdCardSystem ||
+               parameters.save_data_space_id == FileSys::SaveDataSpaceId::SdCardUser) {
+        id = FileSys::StorageId::SdCard;
+    } else {
+        id = FileSys::StorageId::NandSystem;
+    }
+
+    IFileSystem filesystem(std::move(dir.Unwrap()), SizeGetter::FromStorageId(fsc, id));
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
@@ -812,7 +862,7 @@ void FSP_SRV::OpenSaveDataInfoReaderBySaveDataSpaceId(Kernel::HLERequestContext&
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISaveDataInfoReader>(std::make_shared<ISaveDataInfoReader>(space));
+    rb.PushIpcInterface<ISaveDataInfoReader>(std::make_shared<ISaveDataInfoReader>(space, fsc));
 }
 
 void FSP_SRV::SetGlobalAccessLogMode(Kernel::HLERequestContext& ctx) {
@@ -836,7 +886,7 @@ void FSP_SRV::GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx) {
 void FSP_SRV::OpenDataStorageByCurrentProcess(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_FS, "called");
 
-    auto romfs = OpenRomFSCurrentProcess();
+    auto romfs = fsc.OpenRomFSCurrentProcess();
     if (romfs.Failed()) {
         // TODO (bunnei): Find the right error code to use here
         LOG_CRITICAL(Service_FS, "no file system interface available!");
@@ -861,7 +911,7 @@ void FSP_SRV::OpenDataStorageByDataId(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_FS, "called with storage_id={:02X}, unknown={:08X}, title_id={:016X}",
               static_cast<u8>(storage_id), unknown, title_id);
 
-    auto data = OpenRomFS(title_id, storage_id, FileSys::ContentRecordType::Data);
+    auto data = fsc.OpenRomFS(title_id, storage_id, FileSys::ContentRecordType::Data);
 
     if (data.Failed()) {
         const auto archive = FileSys::SystemArchive::SynthesizeSystemArchive(title_id);

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

@@ -32,7 +32,7 @@ enum class LogMode : u32 {
 
 class FSP_SRV final : public ServiceFramework<FSP_SRV> {
 public:
-    explicit FSP_SRV(const Core::Reporter& reporter);
+    explicit FSP_SRV(FileSystemController& fsc, const Core::Reporter& reporter);
     ~FSP_SRV() override;
 
 private:
@@ -51,6 +51,8 @@ private:
     void OutputAccessLogToSdCard(Kernel::HLERequestContext& ctx);
     void GetAccessLogVersionInfo(Kernel::HLERequestContext& ctx);
 
+    FileSystemController& fsc;
+
     FileSys::VirtualFile romfs;
     u64 current_process_id = 0;
     u32 access_log_program_index = 0;

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

@@ -617,7 +617,7 @@ public:
     }
 };
 
-void InstallInterfaces(SM::ServiceManager& service_manager) {
+void InstallInterfaces(SM::ServiceManager& service_manager, FileSystem::FileSystemController& fsc) {
     std::make_shared<NS>("ns:am2")->InstallAsService(service_manager);
     std::make_shared<NS>("ns:ec")->InstallAsService(service_manager);
     std::make_shared<NS>("ns:rid")->InstallAsService(service_manager);
@@ -628,7 +628,7 @@ void InstallInterfaces(SM::ServiceManager& service_manager) {
     std::make_shared<NS_SU>()->InstallAsService(service_manager);
     std::make_shared<NS_VM>()->InstallAsService(service_manager);
 
-    std::make_shared<PL_U>()->InstallAsService(service_manager);
+    std::make_shared<PL_U>(fsc)->InstallAsService(service_manager);
 }
 
 } // namespace Service::NS

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

@@ -6,7 +6,13 @@
 
 #include "core/hle/service/service.h"
 
-namespace Service::NS {
+namespace Service {
+
+namespace FileSystem {
+class FileSystemController;
+} // namespace FileSystem
+
+namespace NS {
 
 class IAccountProxyInterface final : public ServiceFramework<IAccountProxyInterface> {
 public:
@@ -91,6 +97,8 @@ private:
 };
 
 /// Registers all NS services with the specified service manager.
-void InstallInterfaces(SM::ServiceManager& service_manager);
+void InstallInterfaces(SM::ServiceManager& service_manager, FileSystem::FileSystemController& fsc);
 
-} // namespace Service::NS
+} // namespace NS
+
+} // namespace Service

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

@@ -150,7 +150,8 @@ struct PL_U::Impl {
     std::vector<FontRegion> shared_font_regions;
 };
 
-PL_U::PL_U() : ServiceFramework("pl:u"), impl{std::make_unique<Impl>()} {
+PL_U::PL_U(FileSystem::FileSystemController& fsc)
+    : ServiceFramework("pl:u"), impl{std::make_unique<Impl>()} {
     static const FunctionInfo functions[] = {
         {0, &PL_U::RequestLoad, "RequestLoad"},
         {1, &PL_U::GetLoadState, "GetLoadState"},
@@ -161,7 +162,7 @@ PL_U::PL_U() : ServiceFramework("pl:u"), impl{std::make_unique<Impl>()} {
     };
     RegisterHandlers(functions);
     // Attempt to load shared font data from disk
-    const auto* nand = FileSystem::GetSystemNANDContents();
+    const auto* nand = fsc.GetSystemNANDContents();
     std::size_t offset = 0;
     // Rebuild shared fonts from data ncas
     if (nand->HasEntry(static_cast<u64>(FontArchives::Standard),

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

@@ -7,11 +7,17 @@
 #include <memory>
 #include "core/hle/service/service.h"
 
-namespace Service::NS {
+namespace Service {
+
+namespace FileSystem {
+class FileSystemController;
+} // namespace FileSystem
+
+namespace NS {
 
 class PL_U final : public ServiceFramework<PL_U> {
 public:
-    PL_U();
+    PL_U(FileSystem::FileSystemController& fsc);
     ~PL_U() override;
 
 private:
@@ -26,4 +32,6 @@ private:
     std::unique_ptr<Impl> impl;
 };
 
-} // namespace Service::NS
+} // namespace NS
+
+} // namespace Service

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

@@ -146,8 +146,8 @@ u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& outp
     }
     IoctlSubmitGpfifo params{};
     std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called, gpfifo={:X}, num_entries={:X}, flags={:X}",
-                params.address, params.num_entries, params.flags.raw);
+    LOG_TRACE(Service_NVDRV, "called, gpfifo={:X}, num_entries={:X}, flags={:X}", params.address,
+              params.num_entries, params.flags.raw);
 
     ASSERT_MSG(input.size() == sizeof(IoctlSubmitGpfifo) +
                                    params.num_entries * sizeof(Tegra::CommandListHeader),
@@ -179,8 +179,8 @@ u32 nvhost_gpu::KickoffPB(const std::vector<u8>& input, std::vector<u8>& output)
     }
     IoctlSubmitGpfifo params{};
     std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
-    LOG_WARNING(Service_NVDRV, "(STUBBED) called, gpfifo={:X}, num_entries={:X}, flags={:X}",
-                params.address, params.num_entries, params.flags.raw);
+    LOG_TRACE(Service_NVDRV, "called, gpfifo={:X}, num_entries={:X}, flags={:X}", params.address,
+              params.num_entries, params.flags.raw);
 
     Tegra::CommandList entries(params.num_entries);
     Memory::ReadBlock(params.address, entries.data(),

src/core/hle/service/prepo/prepo.cpp

@@ -2,34 +2,31 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <json.hpp>
-#include "common/file_util.h"
 #include "common/hex_util.h"
 #include "common/logging/log.h"
-#include "common/scm_rev.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/service/acc/profile_manager.h"
 #include "core/hle/service/prepo/prepo.h"
 #include "core/hle/service/service.h"
 #include "core/reporter.h"
-#include "core/settings.h"
 
 namespace Service::PlayReport {
 
 class PlayReport final : public ServiceFramework<PlayReport> {
 public:
-    explicit PlayReport(const char* name) : ServiceFramework{name} {
+    explicit PlayReport(Core::System& system, const char* name)
+        : ServiceFramework{name}, system(system) {
         // clang-format off
         static const FunctionInfo functions[] = {
-            {10100, nullptr, "SaveReportOld"},
-            {10101, &PlayReport::SaveReportWithUserOld, "SaveReportWithUserOld"},
-            {10102, nullptr, "SaveReport"},
-            {10103, nullptr, "SaveReportWithUser"},
+            {10100, &PlayReport::SaveReport<Core::Reporter::PlayReportType::Old>, "SaveReportOld"},
+            {10101, &PlayReport::SaveReportWithUser<Core::Reporter::PlayReportType::Old>, "SaveReportWithUserOld"},
+            {10102, &PlayReport::SaveReport<Core::Reporter::PlayReportType::New>, "SaveReport"},
+            {10103, &PlayReport::SaveReportWithUser<Core::Reporter::PlayReportType::New>, "SaveReportWithUser"},
             {10200, nullptr, "RequestImmediateTransmission"},
             {10300, nullptr, "GetTransmissionStatus"},
-            {20100, nullptr, "SaveSystemReport"},
-            {20101, nullptr, "SaveSystemReportWithUser"},
+            {20100, &PlayReport::SaveSystemReport, "SaveSystemReport"},
+            {20101, &PlayReport::SaveSystemReportWithUser, "SaveSystemReportWithUser"},
             {20200, nullptr, "SetOperationMode"},
             {30100, nullptr, "ClearStorage"},
             {30200, nullptr, "ClearStatistics"},
@@ -47,7 +44,28 @@ public:
     }
 
 private:
-    void SaveReportWithUserOld(Kernel::HLERequestContext& ctx) {
+    template <Core::Reporter::PlayReportType Type>
+    void SaveReport(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto process_id = rp.PopRaw<u64>();
+
+        const auto data1 = ctx.ReadBuffer(0);
+        const auto data2 = ctx.ReadBuffer(1);
+
+        LOG_DEBUG(Service_PREPO,
+                  "called, type={:02X}, process_id={:016X}, data1_size={:016X}, data2_size={:016X}",
+                  static_cast<u8>(Type), process_id, data1.size(), data2.size());
+
+        const auto& reporter{system.GetReporter()};
+        reporter.SavePlayReport(Type, system.CurrentProcess()->GetTitleID(), {data1, data2},
+                                process_id);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    template <Core::Reporter::PlayReportType Type>
+    void SaveReportWithUser(Kernel::HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
         const auto user_id = rp.PopRaw<u128>();
         const auto process_id = rp.PopRaw<u64>();
@@ -57,24 +75,65 @@ private:
 
         LOG_DEBUG(
             Service_PREPO,
-            "called, user_id={:016X}{:016X}, unk1={:016X}, data1_size={:016X}, data2_size={:016X}",
-            user_id[1], user_id[0], process_id, data1.size(), data2.size());
+            "called, type={:02X}, user_id={:016X}{:016X}, process_id={:016X}, data1_size={:016X}, "
+            "data2_size={:016X}",
+            static_cast<u8>(Type), user_id[1], user_id[0], process_id, data1.size(), data2.size());
 
-        const auto& reporter{Core::System::GetInstance().GetReporter()};
-        reporter.SavePlayReport(Core::CurrentProcess()->GetTitleID(), process_id, {data1, data2},
-                                user_id);
+        const auto& reporter{system.GetReporter()};
+        reporter.SavePlayReport(Type, system.CurrentProcess()->GetTitleID(), {data1, data2},
+                                process_id, user_id);
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
     }
+
+    void SaveSystemReport(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto title_id = rp.PopRaw<u64>();
+
+        const auto data1 = ctx.ReadBuffer(0);
+        const auto data2 = ctx.ReadBuffer(1);
+
+        LOG_DEBUG(Service_PREPO, "called, title_id={:016X}, data1_size={:016X}, data2_size={:016X}",
+                  title_id, data1.size(), data2.size());
+
+        const auto& reporter{system.GetReporter()};
+        reporter.SavePlayReport(Core::Reporter::PlayReportType::System, title_id, {data1, data2});
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void SaveSystemReportWithUser(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto user_id = rp.PopRaw<u128>();
+        const auto title_id = rp.PopRaw<u64>();
+
+        const auto data1 = ctx.ReadBuffer(0);
+        const auto data2 = ctx.ReadBuffer(1);
+
+        LOG_DEBUG(Service_PREPO,
+                  "called, user_id={:016X}{:016X}, title_id={:016X}, data1_size={:016X}, "
+                  "data2_size={:016X}",
+                  user_id[1], user_id[0], title_id, data1.size(), data2.size());
+
+        const auto& reporter{system.GetReporter()};
+        reporter.SavePlayReport(Core::Reporter::PlayReportType::System, title_id, {data1, data2},
+                                std::nullopt, user_id);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    Core::System& system;
 };
 
-void InstallInterfaces(SM::ServiceManager& service_manager) {
-    std::make_shared<PlayReport>("prepo:a")->InstallAsService(service_manager);
-    std::make_shared<PlayReport>("prepo:a2")->InstallAsService(service_manager);
-    std::make_shared<PlayReport>("prepo:m")->InstallAsService(service_manager);
-    std::make_shared<PlayReport>("prepo:s")->InstallAsService(service_manager);
-    std::make_shared<PlayReport>("prepo:u")->InstallAsService(service_manager);
+void InstallInterfaces(Core::System& system) {
+    std::make_shared<PlayReport>(system, "prepo:a")->InstallAsService(system.ServiceManager());
+    std::make_shared<PlayReport>(system, "prepo:a2")->InstallAsService(system.ServiceManager());
+    std::make_shared<PlayReport>(system, "prepo:m")->InstallAsService(system.ServiceManager());
+    std::make_shared<PlayReport>(system, "prepo:s")->InstallAsService(system.ServiceManager());
+    std::make_shared<PlayReport>(system, "prepo:u")->InstallAsService(system.ServiceManager());
 }
 
 } // namespace Service::PlayReport

src/core/hle/service/prepo/prepo.h

@@ -10,6 +10,6 @@ class ServiceManager;
 
 namespace Service::PlayReport {
 
-void InstallInterfaces(SM::ServiceManager& service_manager);
+void InstallInterfaces(Core::System& system);
 
 } // namespace Service::PlayReport

src/core/hle/service/service.cpp

@@ -199,6 +199,7 @@ void Init(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system) {
     // NVFlinger needs to be accessed by several services like Vi and AppletOE so we instantiate it
     // here and pass it into the respective InstallInterfaces functions.
     auto nv_flinger = std::make_shared<NVFlinger::NVFlinger>(system.CoreTiming());
+    system.GetFileSystemController().CreateFactories(*system.GetFilesystem(), false);
 
     SM::ServiceManager::InstallInterfaces(sm);
 
@@ -235,12 +236,12 @@ void Init(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system) {
     NIFM::InstallInterfaces(*sm);
     NIM::InstallInterfaces(*sm);
     NPNS::InstallInterfaces(*sm);
-    NS::InstallInterfaces(*sm);
+    NS::InstallInterfaces(*sm, system.GetFileSystemController());
     Nvidia::InstallInterfaces(*sm, *nv_flinger, system);
     PCIe::InstallInterfaces(*sm);
     PCTL::InstallInterfaces(*sm);
     PCV::InstallInterfaces(*sm);
-    PlayReport::InstallInterfaces(*sm);
+    PlayReport::InstallInterfaces(system);
     PM::InstallInterfaces(system);
     PSC::InstallInterfaces(*sm);
     PSM::InstallInterfaces(*sm);

src/core/hle/service/service.h

@@ -18,10 +18,6 @@ namespace Core {
 class System;
 }
 
-namespace FileSys {
-class VfsFilesystem;
-}
-
 namespace Kernel {
 class ClientPort;
 class ServerPort;
@@ -31,6 +27,10 @@ class HLERequestContext;
 
 namespace Service {
 
+namespace FileSystem {
+class FileSystemController;
+} // namespace FileSystem
+
 namespace SM {
 class ServiceManager;
 }

src/core/loader/deconstructed_rom_directory.cpp

@@ -7,6 +7,7 @@
 #include "common/common_funcs.h"
 #include "common/file_util.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/control_metadata.h"
 #include "core/file_sys/patch_manager.h"
@@ -176,7 +177,8 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
     // Register the RomFS if a ".romfs" file was found
     if (romfs_iter != files.end() && *romfs_iter != nullptr) {
         romfs = *romfs_iter;
-        Service::FileSystem::RegisterRomFS(std::make_unique<FileSys::RomFSFactory>(*this));
+        Core::System::GetInstance().GetFileSystemController().RegisterRomFS(
+            std::make_unique<FileSys::RomFSFactory>(*this));
     }
 
     is_loaded = true;

src/core/loader/nca.cpp

@@ -6,6 +6,7 @@
 
 #include "common/file_util.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/romfs_factory.h"
 #include "core/hle/kernel/process.h"
@@ -57,7 +58,8 @@ AppLoader_NCA::LoadResult AppLoader_NCA::Load(Kernel::Process& process) {
     }
 
     if (nca->GetRomFS() != nullptr && nca->GetRomFS()->GetSize() > 0) {
-        Service::FileSystem::RegisterRomFS(std::make_unique<FileSys::RomFSFactory>(*this));
+        Core::System::GetInstance().GetFileSystemController().RegisterRomFS(
+            std::make_unique<FileSys::RomFSFactory>(*this));
     }
 
     is_loaded = true;

src/core/loader/nro.cpp

@@ -10,6 +10,7 @@
 #include "common/file_util.h"
 #include "common/logging/log.h"
 #include "common/swap.h"
+#include "core/core.h"
 #include "core/file_sys/control_metadata.h"
 #include "core/file_sys/romfs_factory.h"
 #include "core/file_sys/vfs_offset.h"
@@ -214,7 +215,8 @@ AppLoader_NRO::LoadResult AppLoader_NRO::Load(Kernel::Process& process) {
     }
 
     if (romfs != nullptr) {
-        Service::FileSystem::RegisterRomFS(std::make_unique<FileSys::RomFSFactory>(*this));
+        Core::System::GetInstance().GetFileSystemController().RegisterRomFS(
+            std::make_unique<FileSys::RomFSFactory>(*this));
     }
 
     is_loaded = true;

src/core/loader/nso.cpp

@@ -152,8 +152,7 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::Process& process,
         auto& system = Core::System::GetInstance();
         const auto cheats = pm->CreateCheatList(system, nso_header.build_id);
         if (!cheats.empty()) {
-            system.RegisterCheatList(cheats, Common::HexToString(nso_header.build_id), load_base,
-                                     load_base + program_image.size());
+            system.RegisterCheatList(cheats, nso_header.build_id, load_base, image_size);
         }
     }
 

src/core/loader/nsp.cpp

@@ -5,6 +5,7 @@
 #include <vector>
 
 #include "common/common_types.h"
+#include "core/core.h"
 #include "core/file_sys/card_image.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/control_metadata.h"
@@ -26,20 +27,18 @@ AppLoader_NSP::AppLoader_NSP(FileSys::VirtualFile file)
 
     if (nsp->GetStatus() != ResultStatus::Success)
         return;
-    if (nsp->IsExtractedType())
-        return;
-
-    const auto control_nca =
-        nsp->GetNCA(nsp->GetProgramTitleID(), FileSys::ContentRecordType::Control);
-    if (control_nca == nullptr || control_nca->GetStatus() != ResultStatus::Success)
-        return;
-
-    std::tie(nacp_file, icon_file) =
-        FileSys::PatchManager(nsp->GetProgramTitleID()).ParseControlNCA(*control_nca);
 
     if (nsp->IsExtractedType()) {
         secondary_loader = std::make_unique<AppLoader_DeconstructedRomDirectory>(nsp->GetExeFS());
     } else {
+        const auto control_nca =
+            nsp->GetNCA(nsp->GetProgramTitleID(), FileSys::ContentRecordType::Control);
+        if (control_nca == nullptr || control_nca->GetStatus() != ResultStatus::Success)
+            return;
+
+        std::tie(nacp_file, icon_file) =
+            FileSys::PatchManager(nsp->GetProgramTitleID()).ParseControlNCA(*control_nca);
+
         if (title_id == 0)
             return;
 
@@ -56,11 +55,11 @@ FileType AppLoader_NSP::IdentifyType(const FileSys::VirtualFile& file) {
     if (nsp.GetStatus() == ResultStatus::Success) {
         // Extracted Type case
         if (nsp.IsExtractedType() && nsp.GetExeFS() != nullptr &&
-            FileSys::IsDirectoryExeFS(nsp.GetExeFS()) && nsp.GetRomFS() != nullptr) {
+            FileSys::IsDirectoryExeFS(nsp.GetExeFS())) {
             return FileType::NSP;
         }
 
-        // Non-Ectracted Type case
+        // Non-Extracted Type case
         if (!nsp.IsExtractedType() &&
             nsp.GetNCA(nsp.GetFirstTitleID(), FileSys::ContentRecordType::Program) != nullptr &&
             AppLoader_NCA::IdentifyType(nsp.GetNCAFile(
@@ -77,7 +76,7 @@ AppLoader_NSP::LoadResult AppLoader_NSP::Load(Kernel::Process& process) {
         return {ResultStatus::ErrorAlreadyLoaded, {}};
     }
 
-    if (title_id == 0) {
+    if (!nsp->IsExtractedType() && title_id == 0) {
         return {ResultStatus::ErrorNSPMissingProgramNCA, {}};
     }
 
@@ -91,7 +90,8 @@ AppLoader_NSP::LoadResult AppLoader_NSP::Load(Kernel::Process& process) {
         return {nsp_program_status, {}};
     }
 
-    if (nsp->GetNCA(title_id, FileSys::ContentRecordType::Program) == nullptr) {
+    if (!nsp->IsExtractedType() &&
+        nsp->GetNCA(title_id, FileSys::ContentRecordType::Program) == nullptr) {
         if (!Core::Crypto::KeyManager::KeyFileExists(false)) {
             return {ResultStatus::ErrorMissingProductionKeyFile, {}};
         }
@@ -106,7 +106,8 @@ AppLoader_NSP::LoadResult AppLoader_NSP::Load(Kernel::Process& process) {
 
     FileSys::VirtualFile update_raw;
     if (ReadUpdateRaw(update_raw) == ResultStatus::Success && update_raw != nullptr) {
-        Service::FileSystem::SetPackedUpdate(std::move(update_raw));
+        Core::System::GetInstance().GetFileSystemController().SetPackedUpdate(
+            std::move(update_raw));
     }
 
     is_loaded = true;

src/core/loader/xci.cpp

@@ -5,6 +5,7 @@
 #include <vector>
 
 #include "common/common_types.h"
+#include "core/core.h"
 #include "core/file_sys/card_image.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/control_metadata.h"
@@ -72,7 +73,8 @@ AppLoader_XCI::LoadResult AppLoader_XCI::Load(Kernel::Process& process) {
 
     FileSys::VirtualFile update_raw;
     if (ReadUpdateRaw(update_raw) == ResultStatus::Success && update_raw != nullptr) {
-        Service::FileSystem::SetPackedUpdate(std::move(update_raw));
+        Core::System::GetInstance().GetFileSystemController().SetPackedUpdate(
+            std::move(update_raw));
     }
 
     is_loaded = true;

src/core/memory.cpp

@@ -43,8 +43,13 @@ static void MapPages(Common::PageTable& page_table, VAddr base, u64 size, u8* me
 
     // During boot, current_page_table might not be set yet, in which case we need not flush
     if (Core::System::GetInstance().IsPoweredOn()) {
-        Core::System::GetInstance().GPU().FlushAndInvalidateRegion(base << PAGE_BITS,
-                                                                   size * PAGE_SIZE);
+        auto& gpu = Core::System::GetInstance().GPU();
+        for (u64 i = 0; i < size; i++) {
+            const auto page = base + i;
+            if (page_table.attributes[page] == Common::PageType::RasterizerCachedMemory) {
+                gpu.FlushAndInvalidateRegion(page << PAGE_BITS, PAGE_SIZE);
+            }
+        }
     }
 
     VAddr end = base + size;

src/core/memory/cheat_engine.cpp

@@ -0,0 +1,234 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <locale>
+#include "common/hex_util.h"
+#include "common/microprofile.h"
+#include "common/swap.h"
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/core_timing_util.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/service/hid/controllers/npad.h"
+#include "core/hle/service/hid/hid.h"
+#include "core/hle/service/sm/sm.h"
+#include "core/memory/cheat_engine.h"
+
+namespace Memory {
+
+constexpr s64 CHEAT_ENGINE_TICKS = static_cast<s64>(Core::Timing::BASE_CLOCK_RATE / 12);
+constexpr u32 KEYPAD_BITMASK = 0x3FFFFFF;
+
+StandardVmCallbacks::StandardVmCallbacks(const Core::System& system,
+                                         const CheatProcessMetadata& metadata)
+    : system(system), metadata(metadata) {}
+
+StandardVmCallbacks::~StandardVmCallbacks() = default;
+
+void StandardVmCallbacks::MemoryRead(VAddr address, void* data, u64 size) {
+    ReadBlock(SanitizeAddress(address), data, size);
+}
+
+void StandardVmCallbacks::MemoryWrite(VAddr address, const void* data, u64 size) {
+    WriteBlock(SanitizeAddress(address), data, size);
+}
+
+u64 StandardVmCallbacks::HidKeysDown() {
+    const auto applet_resource =
+        system.ServiceManager().GetService<Service::HID::Hid>("hid")->GetAppletResource();
+    if (applet_resource == nullptr) {
+        LOG_WARNING(CheatEngine,
+                    "Attempted to read input state, but applet resource is not initialized!");
+        return false;
+    }
+
+    const auto press_state =
+        applet_resource
+            ->GetController<Service::HID::Controller_NPad>(Service::HID::HidController::NPad)
+            .GetAndResetPressState();
+    return press_state & KEYPAD_BITMASK;
+}
+
+void StandardVmCallbacks::DebugLog(u8 id, u64 value) {
+    LOG_INFO(CheatEngine, "Cheat triggered DebugLog: ID '{:01X}' Value '{:016X}'", id, value);
+}
+
+void StandardVmCallbacks::CommandLog(std::string_view data) {
+    LOG_DEBUG(CheatEngine, "[DmntCheatVm]: {}",
+              data.back() == '\n' ? data.substr(0, data.size() - 1) : data);
+}
+
+VAddr StandardVmCallbacks::SanitizeAddress(VAddr in) const {
+    if ((in < metadata.main_nso_extents.base ||
+         in >= metadata.main_nso_extents.base + metadata.main_nso_extents.size) &&
+        (in < metadata.heap_extents.base ||
+         in >= metadata.heap_extents.base + metadata.heap_extents.size)) {
+        LOG_ERROR(CheatEngine,
+                  "Cheat attempting to access memory at invalid address={:016X}, if this "
+                  "persists, "
+                  "the cheat may be incorrect. However, this may be normal early in execution if "
+                  "the game has not properly set up yet.",
+                  in);
+        return 0; ///< Invalid addresses will hard crash
+    }
+
+    return in;
+}
+
+CheatParser::~CheatParser() = default;
+
+TextCheatParser::~TextCheatParser() = default;
+
+namespace {
+template <char match>
+std::string_view ExtractName(std::string_view data, std::size_t start_index) {
+    auto end_index = start_index;
+    while (data[end_index] != match) {
+        ++end_index;
+        if (end_index > data.size() ||
+            (end_index - start_index - 1) > sizeof(CheatDefinition::readable_name)) {
+            return {};
+        }
+    }
+
+    return data.substr(start_index, end_index - start_index);
+}
+} // Anonymous namespace
+
+std::vector<CheatEntry> TextCheatParser::Parse(const Core::System& system,
+                                               std::string_view data) const {
+    std::vector<CheatEntry> out(1);
+    std::optional<u64> current_entry = std::nullopt;
+
+    for (std::size_t i = 0; i < data.size(); ++i) {
+        if (::isspace(data[i])) {
+            continue;
+        }
+
+        if (data[i] == '{') {
+            current_entry = 0;
+
+            if (out[*current_entry].definition.num_opcodes > 0) {
+                return {};
+            }
+
+            const auto name = ExtractName<'}'>(data, i + 1);
+            if (name.empty()) {
+                return {};
+            }
+
+            std::memcpy(out[*current_entry].definition.readable_name.data(), name.data(),
+                        std::min<std::size_t>(out[*current_entry].definition.readable_name.size(),
+                                              name.size()));
+            out[*current_entry]
+                .definition.readable_name[out[*current_entry].definition.readable_name.size() - 1] =
+                '\0';
+
+            i += name.length() + 1;
+        } else if (data[i] == '[') {
+            current_entry = out.size();
+            out.emplace_back();
+
+            const auto name = ExtractName<']'>(data, i + 1);
+            if (name.empty()) {
+                return {};
+            }
+
+            std::memcpy(out[*current_entry].definition.readable_name.data(), name.data(),
+                        std::min<std::size_t>(out[*current_entry].definition.readable_name.size(),
+                                              name.size()));
+            out[*current_entry]
+                .definition.readable_name[out[*current_entry].definition.readable_name.size() - 1] =
+                '\0';
+
+            i += name.length() + 1;
+        } else if (::isxdigit(data[i])) {
+            if (!current_entry || out[*current_entry].definition.num_opcodes >=
+                                      out[*current_entry].definition.opcodes.size()) {
+                return {};
+            }
+
+            const auto hex = std::string(data.substr(i, 8));
+            if (!std::all_of(hex.begin(), hex.end(), ::isxdigit)) {
+                return {};
+            }
+
+            out[*current_entry].definition.opcodes[out[*current_entry].definition.num_opcodes++] =
+                std::stoul(hex, nullptr, 0x10);
+
+            i += 8;
+        } else {
+            return {};
+        }
+    }
+
+    out[0].enabled = out[0].definition.num_opcodes > 0;
+    out[0].cheat_id = 0;
+
+    for (u32 i = 1; i < out.size(); ++i) {
+        out[i].enabled = out[i].definition.num_opcodes > 0;
+        out[i].cheat_id = i;
+    }
+
+    return out;
+}
+
+CheatEngine::CheatEngine(Core::System& system, std::vector<CheatEntry> cheats,
+                         const std::array<u8, 0x20>& build_id)
+    : system{system}, core_timing{system.CoreTiming()}, vm{std::make_unique<StandardVmCallbacks>(
+                                                            system, metadata)},
+      cheats(std::move(cheats)) {
+    metadata.main_nso_build_id = build_id;
+}
+
+CheatEngine::~CheatEngine() {
+    core_timing.UnscheduleEvent(event, 0);
+}
+
+void CheatEngine::Initialize() {
+    event = core_timing.RegisterEvent(
+        "CheatEngine::FrameCallback::" + Common::HexToString(metadata.main_nso_build_id),
+        [this](u64 userdata, s64 cycles_late) { FrameCallback(userdata, cycles_late); });
+    core_timing.ScheduleEvent(CHEAT_ENGINE_TICKS, event);
+
+    metadata.process_id = system.CurrentProcess()->GetProcessID();
+    metadata.title_id = system.CurrentProcess()->GetTitleID();
+
+    const auto& vm_manager = system.CurrentProcess()->VMManager();
+    metadata.heap_extents = {vm_manager.GetHeapRegionBaseAddress(), vm_manager.GetHeapRegionSize()};
+    metadata.address_space_extents = {vm_manager.GetAddressSpaceBaseAddress(),
+                                      vm_manager.GetAddressSpaceSize()};
+    metadata.alias_extents = {vm_manager.GetMapRegionBaseAddress(), vm_manager.GetMapRegionSize()};
+
+    is_pending_reload.exchange(true);
+}
+
+void CheatEngine::SetMainMemoryParameters(VAddr main_region_begin, u64 main_region_size) {
+    metadata.main_nso_extents = {main_region_begin, main_region_size};
+}
+
+void CheatEngine::Reload(std::vector<CheatEntry> cheats) {
+    this->cheats = std::move(cheats);
+    is_pending_reload.exchange(true);
+}
+
+MICROPROFILE_DEFINE(Cheat_Engine, "Add-Ons", "Cheat Engine", MP_RGB(70, 200, 70));
+
+void CheatEngine::FrameCallback(u64 userdata, s64 cycles_late) {
+    if (is_pending_reload.exchange(false)) {
+        vm.LoadProgram(cheats);
+    }
+
+    if (vm.GetProgramSize() == 0) {
+        return;
+    }
+
+    MICROPROFILE_SCOPE(Cheat_Engine);
+
+    vm.Execute(metadata);
+
+    core_timing.ScheduleEvent(CHEAT_ENGINE_TICKS - cycles_late, event);
+}
+
+} // namespace Memory

src/core/memory/cheat_engine.h

@@ -0,0 +1,86 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+#include <vector>
+#include "common/common_types.h"
+#include "core/memory/dmnt_cheat_types.h"
+#include "core/memory/dmnt_cheat_vm.h"
+
+namespace Core {
+class System;
+}
+
+namespace Core::Timing {
+class CoreTiming;
+struct EventType;
+} // namespace Core::Timing
+
+namespace Memory {
+
+class StandardVmCallbacks : public DmntCheatVm::Callbacks {
+public:
+    StandardVmCallbacks(const Core::System& system, const CheatProcessMetadata& metadata);
+    ~StandardVmCallbacks() override;
+
+    void MemoryRead(VAddr address, void* data, u64 size) override;
+    void MemoryWrite(VAddr address, const void* data, u64 size) override;
+    u64 HidKeysDown() override;
+    void DebugLog(u8 id, u64 value) override;
+    void CommandLog(std::string_view data) override;
+
+private:
+    VAddr SanitizeAddress(VAddr address) const;
+
+    const CheatProcessMetadata& metadata;
+    const Core::System& system;
+};
+
+// Intermediary class that parses a text file or other disk format for storing cheats into a
+// CheatList object, that can be used for execution.
+class CheatParser {
+public:
+    virtual ~CheatParser();
+
+    virtual std::vector<CheatEntry> Parse(const Core::System& system,
+                                          std::string_view data) const = 0;
+};
+
+// CheatParser implementation that parses text files
+class TextCheatParser final : public CheatParser {
+public:
+    ~TextCheatParser() override;
+
+    std::vector<CheatEntry> Parse(const Core::System& system, std::string_view data) const override;
+};
+
+// Class that encapsulates a CheatList and manages its interaction with memory and CoreTiming
+class CheatEngine final {
+public:
+    CheatEngine(Core::System& system_, std::vector<CheatEntry> cheats_,
+                const std::array<u8, 0x20>& build_id);
+    ~CheatEngine();
+
+    void Initialize();
+    void SetMainMemoryParameters(VAddr main_region_begin, u64 main_region_size);
+
+    void Reload(std::vector<CheatEntry> cheats);
+
+private:
+    void FrameCallback(u64 userdata, s64 cycles_late);
+
+    DmntCheatVm vm;
+    CheatProcessMetadata metadata;
+
+    std::vector<CheatEntry> cheats;
+    std::atomic_bool is_pending_reload{false};
+
+    Core::Timing::EventType* event{};
+    Core::Timing::CoreTiming& core_timing;
+    Core::System& system;
+};
+
+} // namespace Memory

src/core/memory/dmnt_cheat_types.h

@@ -0,0 +1,58 @@
+/*
+ * Copyright (c) 2018-2019 Atmosphère-NX
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/*
+ * Adapted by DarkLordZach for use/interaction with yuzu
+ *
+ * Modifications Copyright 2019 yuzu emulator team
+ * Licensed under GPLv2 or any later version
+ * Refer to the license.txt file included.
+ */
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Memory {
+
+struct MemoryRegionExtents {
+    u64 base{};
+    u64 size{};
+};
+
+struct CheatProcessMetadata {
+    u64 process_id{};
+    u64 title_id{};
+    MemoryRegionExtents main_nso_extents{};
+    MemoryRegionExtents heap_extents{};
+    MemoryRegionExtents alias_extents{};
+    MemoryRegionExtents address_space_extents{};
+    std::array<u8, 0x20> main_nso_build_id{};
+};
+
+struct CheatDefinition {
+    std::array<char, 0x40> readable_name{};
+    u32 num_opcodes{};
+    std::array<u32, 0x100> opcodes{};
+};
+
+struct CheatEntry {
+    bool enabled{};
+    u32 cheat_id{};
+    CheatDefinition definition{};
+};
+
+} // namespace Memory

src/core/memory/dmnt_cheat_vm.h

@@ -0,0 +1,321 @@
+/*
+ * Copyright (c) 2018-2019 Atmosphère-NX
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/*
+ * Adapted by DarkLordZach for use/interaction with yuzu
+ *
+ * Modifications Copyright 2019 yuzu emulator team
+ * Licensed under GPLv2 or any later version
+ * Refer to the license.txt file included.
+ */
+
+#pragma once
+
+#include <variant>
+#include <vector>
+#include <fmt/printf.h>
+#include "common/common_types.h"
+#include "core/memory/dmnt_cheat_types.h"
+
+namespace Memory {
+
+enum class CheatVmOpcodeType : u32 {
+    StoreStatic = 0,
+    BeginConditionalBlock = 1,
+    EndConditionalBlock = 2,
+    ControlLoop = 3,
+    LoadRegisterStatic = 4,
+    LoadRegisterMemory = 5,
+    StoreStaticToAddress = 6,
+    PerformArithmeticStatic = 7,
+    BeginKeypressConditionalBlock = 8,
+
+    // These are not implemented by Gateway's VM.
+    PerformArithmeticRegister = 9,
+    StoreRegisterToAddress = 10,
+    Reserved11 = 11,
+
+    // This is a meta entry, and not a real opcode.
+    // This is to facilitate multi-nybble instruction decoding.
+    ExtendedWidth = 12,
+
+    // Extended width opcodes.
+    BeginRegisterConditionalBlock = 0xC0,
+    SaveRestoreRegister = 0xC1,
+    SaveRestoreRegisterMask = 0xC2,
+
+    // This is a meta entry, and not a real opcode.
+    // This is to facilitate multi-nybble instruction decoding.
+    DoubleExtendedWidth = 0xF0,
+
+    // Double-extended width opcodes.
+    DebugLog = 0xFFF,
+};
+
+enum class MemoryAccessType : u32 {
+    MainNso = 0,
+    Heap = 1,
+};
+
+enum class ConditionalComparisonType : u32 {
+    GT = 1,
+    GE = 2,
+    LT = 3,
+    LE = 4,
+    EQ = 5,
+    NE = 6,
+};
+
+enum class RegisterArithmeticType : u32 {
+    Addition = 0,
+    Subtraction = 1,
+    Multiplication = 2,
+    LeftShift = 3,
+    RightShift = 4,
+
+    // These are not supported by Gateway's VM.
+    LogicalAnd = 5,
+    LogicalOr = 6,
+    LogicalNot = 7,
+    LogicalXor = 8,
+
+    None = 9,
+};
+
+enum class StoreRegisterOffsetType : u32 {
+    None = 0,
+    Reg = 1,
+    Imm = 2,
+    MemReg = 3,
+    MemImm = 4,
+    MemImmReg = 5,
+};
+
+enum class CompareRegisterValueType : u32 {
+    MemoryRelAddr = 0,
+    MemoryOfsReg = 1,
+    RegisterRelAddr = 2,
+    RegisterOfsReg = 3,
+    StaticValue = 4,
+    OtherRegister = 5,
+};
+
+enum class SaveRestoreRegisterOpType : u32 {
+    Restore = 0,
+    Save = 1,
+    ClearSaved = 2,
+    ClearRegs = 3,
+};
+
+enum class DebugLogValueType : u32 {
+    MemoryRelAddr = 0,
+    MemoryOfsReg = 1,
+    RegisterRelAddr = 2,
+    RegisterOfsReg = 3,
+    RegisterValue = 4,
+};
+
+union VmInt {
+    u8 bit8;
+    u16 bit16;
+    u32 bit32;
+    u64 bit64;
+};
+
+struct StoreStaticOpcode {
+    u32 bit_width{};
+    MemoryAccessType mem_type{};
+    u32 offset_register{};
+    u64 rel_address{};
+    VmInt value{};
+};
+
+struct BeginConditionalOpcode {
+    u32 bit_width{};
+    MemoryAccessType mem_type{};
+    ConditionalComparisonType cond_type{};
+    u64 rel_address{};
+    VmInt value{};
+};
+
+struct EndConditionalOpcode {};
+
+struct ControlLoopOpcode {
+    bool start_loop{};
+    u32 reg_index{};
+    u32 num_iters{};
+};
+
+struct LoadRegisterStaticOpcode {
+    u32 reg_index{};
+    u64 value{};
+};
+
+struct LoadRegisterMemoryOpcode {
+    u32 bit_width{};
+    MemoryAccessType mem_type{};
+    u32 reg_index{};
+    bool load_from_reg{};
+    u64 rel_address{};
+};
+
+struct StoreStaticToAddressOpcode {
+    u32 bit_width{};
+    u32 reg_index{};
+    bool increment_reg{};
+    bool add_offset_reg{};
+    u32 offset_reg_index{};
+    u64 value{};
+};
+
+struct PerformArithmeticStaticOpcode {
+    u32 bit_width{};
+    u32 reg_index{};
+    RegisterArithmeticType math_type{};
+    u32 value{};
+};
+
+struct BeginKeypressConditionalOpcode {
+    u32 key_mask{};
+};
+
+struct PerformArithmeticRegisterOpcode {
+    u32 bit_width{};
+    RegisterArithmeticType math_type{};
+    u32 dst_reg_index{};
+    u32 src_reg_1_index{};
+    u32 src_reg_2_index{};
+    bool has_immediate{};
+    VmInt value{};
+};
+
+struct StoreRegisterToAddressOpcode {
+    u32 bit_width{};
+    u32 str_reg_index{};
+    u32 addr_reg_index{};
+    bool increment_reg{};
+    StoreRegisterOffsetType ofs_type{};
+    MemoryAccessType mem_type{};
+    u32 ofs_reg_index{};
+    u64 rel_address{};
+};
+
+struct BeginRegisterConditionalOpcode {
+    u32 bit_width{};
+    ConditionalComparisonType cond_type{};
+    u32 val_reg_index{};
+    CompareRegisterValueType comp_type{};
+    MemoryAccessType mem_type{};
+    u32 addr_reg_index{};
+    u32 other_reg_index{};
+    u32 ofs_reg_index{};
+    u64 rel_address{};
+    VmInt value{};
+};
+
+struct SaveRestoreRegisterOpcode {
+    u32 dst_index{};
+    u32 src_index{};
+    SaveRestoreRegisterOpType op_type{};
+};
+
+struct SaveRestoreRegisterMaskOpcode {
+    SaveRestoreRegisterOpType op_type{};
+    std::array<bool, 0x10> should_operate{};
+};
+
+struct DebugLogOpcode {
+    u32 bit_width{};
+    u32 log_id{};
+    DebugLogValueType val_type{};
+    MemoryAccessType mem_type{};
+    u32 addr_reg_index{};
+    u32 val_reg_index{};
+    u32 ofs_reg_index{};
+    u64 rel_address{};
+};
+
+struct UnrecognizedInstruction {
+    CheatVmOpcodeType opcode{};
+};
+
+struct CheatVmOpcode {
+    bool begin_conditional_block{};
+    std::variant<StoreStaticOpcode, BeginConditionalOpcode, EndConditionalOpcode, ControlLoopOpcode,
+                 LoadRegisterStaticOpcode, LoadRegisterMemoryOpcode, StoreStaticToAddressOpcode,
+                 PerformArithmeticStaticOpcode, BeginKeypressConditionalOpcode,
+                 PerformArithmeticRegisterOpcode, StoreRegisterToAddressOpcode,
+                 BeginRegisterConditionalOpcode, SaveRestoreRegisterOpcode,
+                 SaveRestoreRegisterMaskOpcode, DebugLogOpcode, UnrecognizedInstruction>
+        opcode{};
+};
+
+class DmntCheatVm {
+public:
+    /// Helper Type for DmntCheatVm <=> yuzu Interface
+    class Callbacks {
+    public:
+        virtual ~Callbacks();
+
+        virtual void MemoryRead(VAddr address, void* data, u64 size) = 0;
+        virtual void MemoryWrite(VAddr address, const void* data, u64 size) = 0;
+
+        virtual u64 HidKeysDown() = 0;
+
+        virtual void DebugLog(u8 id, u64 value) = 0;
+        virtual void CommandLog(std::string_view data) = 0;
+    };
+
+    static constexpr std::size_t MaximumProgramOpcodeCount = 0x400;
+    static constexpr std::size_t NumRegisters = 0x10;
+
+    explicit DmntCheatVm(std::unique_ptr<Callbacks> callbacks);
+    ~DmntCheatVm();
+
+    std::size_t GetProgramSize() const {
+        return this->num_opcodes;
+    }
+
+    bool LoadProgram(const std::vector<CheatEntry>& cheats);
+    void Execute(const CheatProcessMetadata& metadata);
+
+private:
+    std::unique_ptr<Callbacks> callbacks;
+
+    std::size_t num_opcodes = 0;
+    std::size_t instruction_ptr = 0;
+    std::size_t condition_depth = 0;
+    bool decode_success = false;
+    std::array<u32, MaximumProgramOpcodeCount> program{};
+    std::array<u64, NumRegisters> registers{};
+    std::array<u64, NumRegisters> saved_values{};
+    std::array<std::size_t, NumRegisters> loop_tops{};
+
+    bool DecodeNextOpcode(CheatVmOpcode& out);
+    void SkipConditionalBlock();
+    void ResetState();
+
+    // For implementing the DebugLog opcode.
+    void DebugLog(u32 log_id, u64 value);
+
+    void LogOpcode(const CheatVmOpcode& opcode);
+
+    static u64 GetVmInt(VmInt value, u32 bit_width);
+    static u64 GetCheatProcessAddress(const CheatProcessMetadata& metadata,
+                                      MemoryAccessType mem_type, u64 rel_address);
+};
+
+}; // namespace Memory

src/core/perf_stats.cpp

@@ -4,8 +4,14 @@
 
 #include <algorithm>
 #include <chrono>
+#include <iterator>
 #include <mutex>
+#include <numeric>
+#include <sstream>
 #include <thread>
+#include <fmt/chrono.h>
+#include <fmt/format.h>
+#include "common/file_util.h"
 #include "common/math_util.h"
 #include "core/perf_stats.h"
 #include "core/settings.h"
@@ -15,8 +21,31 @@ using DoubleSecs = std::chrono::duration<double, std::chrono::seconds::period>;
 using std::chrono::duration_cast;
 using std::chrono::microseconds;
 
+// Purposefully ignore the first five frames, as there's a significant amount of overhead in
+// booting that we shouldn't account for
+constexpr std::size_t IgnoreFrames = 5;
+
 namespace Core {
 
+PerfStats::PerfStats(u64 title_id) : title_id(title_id) {}
+
+PerfStats::~PerfStats() {
+    if (!Settings::values.record_frame_times || title_id == 0) {
+        return;
+    }
+
+    const std::time_t t = std::time(nullptr);
+    std::ostringstream stream;
+    std::copy(perf_history.begin() + IgnoreFrames, perf_history.begin() + current_index,
+              std::ostream_iterator<double>(stream, "\n"));
+    const std::string& path = FileUtil::GetUserPath(FileUtil::UserPath::LogDir);
+    // %F Date format expanded is "%Y-%m-%d"
+    const std::string filename =
+        fmt::format("{}/{:%F-%H-%M}_{:016X}.csv", path, *std::localtime(&t), title_id);
+    FileUtil::IOFile file(filename, "w");
+    file.WriteString(stream.str());
+}
+
 void PerfStats::BeginSystemFrame() {
     std::lock_guard lock{object_mutex};
 
@@ -27,7 +56,12 @@ void PerfStats::EndSystemFrame() {
     std::lock_guard lock{object_mutex};
 
     auto frame_end = Clock::now();
-    accumulated_frametime += frame_end - frame_begin;
+    const auto frame_time = frame_end - frame_begin;
+    if (current_index < perf_history.size()) {
+        perf_history[current_index++] =
+            std::chrono::duration<double, std::milli>(frame_time).count();
+    }
+    accumulated_frametime += frame_time;
     system_frames += 1;
 
     previous_frame_length = frame_end - previous_frame_end;
@@ -40,6 +74,17 @@ void PerfStats::EndGameFrame() {
     game_frames += 1;
 }
 
+double PerfStats::GetMeanFrametime() {
+    std::lock_guard lock{object_mutex};
+
+    if (current_index <= IgnoreFrames) {
+        return 0;
+    }
+    const double sum = std::accumulate(perf_history.begin() + IgnoreFrames,
+                                       perf_history.begin() + current_index, 0);
+    return sum / (current_index - IgnoreFrames);
+}
+
 PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us) {
     std::lock_guard lock{object_mutex};
 

src/core/perf_stats.h

@@ -4,7 +4,9 @@
 
 #pragma once
 
+#include <array>
 #include <chrono>
+#include <cstddef>
 #include <mutex>
 #include "common/common_types.h"
 
@@ -27,6 +29,10 @@ struct PerfStatsResults {
  */
 class PerfStats {
 public:
+    explicit PerfStats(u64 title_id);
+
+    ~PerfStats();
+
     using Clock = std::chrono::high_resolution_clock;
 
     void BeginSystemFrame();
@@ -35,6 +41,11 @@ public:
 
     PerfStatsResults GetAndResetStats(std::chrono::microseconds current_system_time_us);
 
+    /**
+     * Returns the Arthimetic Mean of all frametime values stored in the performance history.
+     */
+    double GetMeanFrametime();
+
     /**
      * Gets the ratio between walltime and the emulated time of the previous system frame. This is
      * useful for scaling inputs or outputs moving between the two time domains.
@@ -42,7 +53,15 @@ public:
     double GetLastFrameTimeScale();
 
 private:
-    std::mutex object_mutex;
+    std::mutex object_mutex{};
+
+    /// Title ID for the game that is running. 0 if there is no game running yet
+    u64 title_id{0};
+    /// Current index for writing to the perf_history array
+    std::size_t current_index{0};
+    /// Stores an hour of historical frametime data useful for processing and tracking performance
+    /// regressions with code changes.
+    std::array<double, 216000> perf_history = {};
 
     /// Point when the cumulative counters were reset
     Clock::time_point reset_point = Clock::now();

src/core/reporter.cpp

@@ -304,8 +304,8 @@ void Reporter::SaveUnimplementedAppletReport(
     SaveToFile(std::move(out), GetPath("unimpl_applet_report", title_id, timestamp));
 }
 
-void Reporter::SavePlayReport(u64 title_id, u64 process_id, std::vector<std::vector<u8>> data,
-                              std::optional<u128> user_id) const {
+void Reporter::SavePlayReport(PlayReportType type, u64 title_id, std::vector<std::vector<u8>> data,
+                              std::optional<u64> process_id, std::optional<u128> user_id) const {
     if (!IsReportingEnabled()) {
         return;
     }
@@ -321,7 +321,11 @@ void Reporter::SavePlayReport(u64 title_id, u64 process_id, std::vector<std::vec
         data_out.push_back(Common::HexToString(d));
     }
 
-    out["play_report_process_id"] = fmt::format("{:016X}", process_id);
+    if (process_id.has_value()) {
+        out["play_report_process_id"] = fmt::format("{:016X}", *process_id);
+    }
+
+    out["play_report_type"] = fmt::format("{:02}", static_cast<u8>(type));
     out["play_report_data"] = std::move(data_out);
 
     SaveToFile(std::move(out), GetPath("play_report", title_id, timestamp));

src/core/reporter.h

@@ -46,8 +46,14 @@ public:
                                        std::vector<std::vector<u8>> normal_channel,
                                        std::vector<std::vector<u8>> interactive_channel) const;
 
-    void SavePlayReport(u64 title_id, u64 process_id, std::vector<std::vector<u8>> data,
-                        std::optional<u128> user_id = {}) const;
+    enum class PlayReportType {
+        Old,
+        New,
+        System,
+    };
+
+    void SavePlayReport(PlayReportType type, u64 title_id, std::vector<std::vector<u8>> data,
+                        std::optional<u64> process_id = {}, std::optional<u128> user_id = {}) const;
 
     void SaveErrorReport(u64 title_id, ResultCode result,
                          std::optional<std::string> custom_text_main = {},

src/core/settings.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/file_util.h"
 #include "core/core.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hle/service/hid/hid.h"
@@ -97,8 +98,8 @@ void LogSettings() {
     LogSetting("Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);
     LogSetting("Audio_OutputDevice", Settings::values.audio_device_id);
     LogSetting("DataStorage_UseVirtualSd", Settings::values.use_virtual_sd);
-    LogSetting("DataStorage_NandDir", Settings::values.nand_dir);
-    LogSetting("DataStorage_SdmcDir", Settings::values.sdmc_dir);
+    LogSetting("DataStorage_NandDir", FileUtil::GetUserPath(FileUtil::UserPath::NANDDir));
+    LogSetting("DataStorage_SdmcDir", FileUtil::GetUserPath(FileUtil::UserPath::SDMCDir));
     LogSetting("Debugging_UseGdbstub", Settings::values.use_gdbstub);
     LogSetting("Debugging_GdbstubPort", Settings::values.gdbstub_port);
     LogSetting("Debugging_ProgramArgs", Settings::values.program_args);

src/core/settings.h

@@ -346,6 +346,31 @@ struct TouchscreenInput {
     u32 rotation_angle;
 };
 
+enum class NANDTotalSize : u64 {
+    S29_1GB = 0x747C00000ULL,
+};
+
+enum class NANDUserSize : u64 {
+    S26GB = 0x680000000ULL,
+};
+
+enum class NANDSystemSize : u64 {
+    S2_5GB = 0xA0000000,
+};
+
+enum class SDMCSize : u64 {
+    S1GB = 0x40000000,
+    S2GB = 0x80000000,
+    S4GB = 0x100000000ULL,
+    S8GB = 0x200000000ULL,
+    S16GB = 0x400000000ULL,
+    S32GB = 0x800000000ULL,
+    S64GB = 0x1000000000ULL,
+    S128GB = 0x2000000000ULL,
+    S256GB = 0x4000000000ULL,
+    S1TB = 0x10000000000ULL,
+};
+
 struct Values {
     // System
     bool use_docked_mode;
@@ -382,8 +407,13 @@ struct Values {
 
     // Data Storage
     bool use_virtual_sd;
-    std::string nand_dir;
-    std::string sdmc_dir;
+    bool gamecard_inserted;
+    bool gamecard_current_game;
+    std::string gamecard_path;
+    NANDTotalSize nand_total_size;
+    NANDSystemSize nand_system_size;
+    NANDUserSize nand_user_size;
+    SDMCSize sdmc_size;
 
     // Renderer
     float resolution_factor;
@@ -409,6 +439,7 @@ struct Values {
     float volume;
 
     // Debugging
+    bool record_frame_times;
     bool use_gdbstub;
     u16 gdbstub_port;
     std::string program_args;

src/video_core/engines/fermi_2d.cpp

@@ -29,8 +29,8 @@ void Fermi2D::CallMethod(const GPU::MethodCall& method_call) {
 }
 
 void Fermi2D::HandleSurfaceCopy() {
-    LOG_WARNING(HW_GPU, "Requested a surface copy with operation {}",
-                static_cast<u32>(regs.operation));
+    LOG_DEBUG(HW_GPU, "Requested a surface copy with operation {}",
+              static_cast<u32>(regs.operation));
 
     // TODO(Subv): Only raw copies are implemented.
     ASSERT(regs.operation == Operation::SrcCopy);

src/video_core/engines/maxwell_3d.cpp

@@ -89,6 +89,13 @@ void Maxwell3D::InitializeRegisterDefaults() {
 
     // Commercial games seem to assume this value is enabled and nouveau sets this value manually.
     regs.rt_separate_frag_data = 1;
+
+    // Some games (like Super Mario Odyssey) assume that SRGB is enabled.
+    regs.framebuffer_srgb = 1;
+    mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_end_gl)] = true;
+    mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_begin_gl)] = true;
+    mme_inline[MAXWELL3D_REG_INDEX(vertex_buffer.count)] = true;
+    mme_inline[MAXWELL3D_REG_INDEX(index_array.count)] = true;
 }
 
 #define DIRTY_REGS_POS(field_name) (offsetof(Maxwell3D::DirtyRegs, field_name))
@@ -253,6 +260,9 @@ void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u3
 
     // Execute the current macro.
     macro_interpreter.Execute(macro_positions[entry], num_parameters, parameters);
+    if (mme_draw.current_mode != MMEDrawMode::Undefined) {
+        FlushMMEInlineDraw();
+    }
 }
 
 void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
@@ -329,6 +339,10 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
         ProcessMacroBind(method_call.argument);
         break;
     }
+    case MAXWELL3D_REG_INDEX(firmware[4]): {
+        ProcessFirmwareCall4();
+        break;
+    }
     case MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]):
     case MAXWELL3D_REG_INDEX(const_buffer.cb_data[1]):
     case MAXWELL3D_REG_INDEX(const_buffer.cb_data[2]):
@@ -409,6 +423,97 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
+void Maxwell3D::StepInstance(const MMEDrawMode expected_mode, const u32 count) {
+    if (mme_draw.current_mode == MMEDrawMode::Undefined) {
+        if (mme_draw.gl_begin_consume) {
+            mme_draw.current_mode = expected_mode;
+            mme_draw.current_count = count;
+            mme_draw.instance_count = 1;
+            mme_draw.gl_begin_consume = false;
+            mme_draw.gl_end_count = 0;
+        }
+        return;
+    } else {
+        if (mme_draw.current_mode == expected_mode && count == mme_draw.current_count &&
+            mme_draw.instance_mode && mme_draw.gl_begin_consume) {
+            mme_draw.instance_count++;
+            mme_draw.gl_begin_consume = false;
+            return;
+        } else {
+            FlushMMEInlineDraw();
+        }
+    }
+    // Tail call in case it needs to retry.
+    StepInstance(expected_mode, count);
+}
+
+void Maxwell3D::CallMethodFromMME(const GPU::MethodCall& method_call) {
+    const u32 method = method_call.method;
+    if (mme_inline[method]) {
+        regs.reg_array[method] = method_call.argument;
+        if (method == MAXWELL3D_REG_INDEX(vertex_buffer.count) ||
+            method == MAXWELL3D_REG_INDEX(index_array.count)) {
+            const MMEDrawMode expected_mode = method == MAXWELL3D_REG_INDEX(vertex_buffer.count)
+                                                  ? MMEDrawMode::Array
+                                                  : MMEDrawMode::Indexed;
+            StepInstance(expected_mode, method_call.argument);
+        } else if (method == MAXWELL3D_REG_INDEX(draw.vertex_begin_gl)) {
+            mme_draw.instance_mode =
+                (regs.draw.instance_next != 0) || (regs.draw.instance_cont != 0);
+            mme_draw.gl_begin_consume = true;
+        } else {
+            mme_draw.gl_end_count++;
+        }
+    } else {
+        if (mme_draw.current_mode != MMEDrawMode::Undefined) {
+            FlushMMEInlineDraw();
+        }
+        CallMethod(method_call);
+    }
+}
+
+void Maxwell3D::FlushMMEInlineDraw() {
+    LOG_DEBUG(HW_GPU, "called, topology={}, count={}", static_cast<u32>(regs.draw.topology.Value()),
+              regs.vertex_buffer.count);
+    ASSERT_MSG(!(regs.index_array.count && regs.vertex_buffer.count), "Both indexed and direct?");
+    ASSERT(mme_draw.instance_count == mme_draw.gl_end_count);
+
+    auto debug_context = system.GetGPUDebugContext();
+
+    if (debug_context) {
+        debug_context->OnEvent(Tegra::DebugContext::Event::IncomingPrimitiveBatch, nullptr);
+    }
+
+    // Both instance configuration registers can not be set at the same time.
+    ASSERT_MSG(!regs.draw.instance_next || !regs.draw.instance_cont,
+               "Illegal combination of instancing parameters");
+
+    const bool is_indexed = mme_draw.current_mode == MMEDrawMode::Indexed;
+    if (ShouldExecute()) {
+        rasterizer.DrawMultiBatch(is_indexed);
+    }
+
+    if (debug_context) {
+        debug_context->OnEvent(Tegra::DebugContext::Event::FinishedPrimitiveBatch, nullptr);
+    }
+
+    // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
+    // the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
+    // it's possible that it is incorrect and that there is some other register used to specify the
+    // drawing mode.
+    if (is_indexed) {
+        regs.index_array.count = 0;
+    } else {
+        regs.vertex_buffer.count = 0;
+    }
+    mme_draw.current_mode = MMEDrawMode::Undefined;
+    mme_draw.current_count = 0;
+    mme_draw.instance_count = 0;
+    mme_draw.instance_mode = false;
+    mme_draw.gl_begin_consume = false;
+    mme_draw.gl_end_count = 0;
+}
+
 void Maxwell3D::ProcessMacroUpload(u32 data) {
     ASSERT_MSG(regs.macros.upload_address < macro_memory.size(),
                "upload_address exceeded macro_memory size!");
@@ -419,6 +524,14 @@ void Maxwell3D::ProcessMacroBind(u32 data) {
     macro_positions[regs.macros.entry++] = data;
 }
 
+void Maxwell3D::ProcessFirmwareCall4() {
+    LOG_WARNING(HW_GPU, "(STUBBED) called");
+
+    // Firmware call 4 is a blob that changes some registers depending on its parameters.
+    // These registers don't affect emulation and so are stubbed by setting 0xd00 to 1.
+    regs.reg_array[0xd00] = 1;
+}
+
 void Maxwell3D::ProcessQueryGet() {
     const GPUVAddr sequence_address{regs.query.QueryAddress()};
     // Since the sequence address is given as a GPU VAddr, we have to convert it to an application
@@ -526,7 +639,7 @@ void Maxwell3D::ProcessSyncPoint() {
 }
 
 void Maxwell3D::DrawArrays() {
-    LOG_DEBUG(HW_GPU, "called, topology={}, count={}", static_cast<u32>(regs.draw.topology.Value()),
+    LOG_TRACE(HW_GPU, "called, topology={}, count={}", static_cast<u32>(regs.draw.topology.Value()),
               regs.vertex_buffer.count);
     ASSERT_MSG(!(regs.index_array.count && regs.vertex_buffer.count), "Both indexed and direct?");
 
@@ -549,7 +662,9 @@ void Maxwell3D::DrawArrays() {
     }
 
     const bool is_indexed{regs.index_array.count && !regs.vertex_buffer.count};
-    rasterizer.AccelerateDrawBatch(is_indexed);
+    if (ShouldExecute()) {
+        rasterizer.DrawBatch(is_indexed);
+    }
 
     if (debug_context) {
         debug_context->OnEvent(Tegra::DebugContext::Event::FinishedPrimitiveBatch, nullptr);

src/video_core/engines/maxwell_3d.h

@@ -811,8 +811,9 @@ public:
                 INSERT_PADDING_WORDS(0x21);
 
                 u32 vb_element_base;
+                u32 vb_base_instance;
 
-                INSERT_PADDING_WORDS(0x36);
+                INSERT_PADDING_WORDS(0x35);
 
                 union {
                     BitField<0, 1, u32> c0;
@@ -1089,7 +1090,9 @@ public:
                     INSERT_PADDING_WORDS(14);
                 } shader_config[MaxShaderProgram];
 
-                INSERT_PADDING_WORDS(0x80);
+                INSERT_PADDING_WORDS(0x60);
+
+                u32 firmware[0x20];
 
                 struct {
                     u32 cb_size;
@@ -1236,6 +1239,11 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write the value to the register identified by method.
+    void CallMethodFromMME(const GPU::MethodCall& method_call);
+
+    void FlushMMEInlineDraw();
+
     /// Given a Texture Handle, returns the TSC and TIC entries.
     Texture::FullTextureInfo GetTextureInfo(const Texture::TextureHandle tex_handle,
                                             std::size_t offset) const;
@@ -1261,6 +1269,21 @@ public:
         return execute_on;
     }
 
+    enum class MMEDrawMode : u32 {
+        Undefined,
+        Array,
+        Indexed,
+    };
+
+    struct MMEDrawState {
+        MMEDrawMode current_mode{MMEDrawMode::Undefined};
+        u32 current_count{};
+        u32 instance_count{};
+        bool instance_mode{};
+        bool gl_begin_consume{};
+        u32 gl_end_count{};
+    } mme_draw;
+
 private:
     void InitializeRegisterDefaults();
 
@@ -1273,6 +1296,8 @@ private:
     /// Start offsets of each macro in macro_memory
     std::array<u32, 0x80> macro_positions = {};
 
+    std::array<bool, Regs::NUM_REGS> mme_inline{};
+
     /// Memory for macro code
     MacroMemory macro_memory;
 
@@ -1319,6 +1344,9 @@ private:
     /// Handles writes to the macro bind register.
     void ProcessMacroBind(u32 data);
 
+    /// Handles firmware blob 4
+    void ProcessFirmwareCall4();
+
     /// Handles a write to the CLEAR_BUFFERS register.
     void ProcessClearBuffers();
 
@@ -1341,6 +1369,9 @@ private:
 
     /// Handles a write to the VERTEX_END_GL register, triggering a draw.
     void DrawArrays();
+
+    // Handles a instance drawcall from MME
+    void StepInstance(MMEDrawMode expected_mode, u32 count);
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
@@ -1397,6 +1428,7 @@ ASSERT_REG_POSITION(stencil_front_mask, 0x4E7);
 ASSERT_REG_POSITION(frag_color_clamp, 0x4EA);
 ASSERT_REG_POSITION(screen_y_control, 0x4EB);
 ASSERT_REG_POSITION(vb_element_base, 0x50D);
+ASSERT_REG_POSITION(vb_base_instance, 0x50E);
 ASSERT_REG_POSITION(clip_distance_enabled, 0x544);
 ASSERT_REG_POSITION(point_size, 0x546);
 ASSERT_REG_POSITION(zeta_enable, 0x54E);
@@ -1431,6 +1463,7 @@ ASSERT_REG_POSITION(vertex_array[0], 0x700);
 ASSERT_REG_POSITION(independent_blend, 0x780);
 ASSERT_REG_POSITION(vertex_array_limit[0], 0x7C0);
 ASSERT_REG_POSITION(shader_config[0], 0x800);
+ASSERT_REG_POSITION(firmware, 0x8C0);
 ASSERT_REG_POSITION(const_buffer, 0x8E0);
 ASSERT_REG_POSITION(cb_bind[0], 0x904);
 ASSERT_REG_POSITION(tex_cb_index, 0x982);

src/video_core/engines/shader_bytecode.h

@@ -544,6 +544,35 @@ enum class VoteOperation : u64 {
     Eq = 2,  // allThreadsEqualNV
 };
 
+enum class ImageAtomicSize : u64 {
+    U32 = 0,
+    S32 = 1,
+    U64 = 2,
+    F32 = 3,
+    S64 = 5,
+    SD32 = 6,
+    SD64 = 7,
+};
+
+enum class ImageAtomicOperation : u64 {
+    Add = 0,
+    Min = 1,
+    Max = 2,
+    Inc = 3,
+    Dec = 4,
+    And = 5,
+    Or = 6,
+    Xor = 7,
+    Exch = 8,
+};
+
+enum class ShuffleOperation : u64 {
+    Idx = 0,  // shuffleNV
+    Up = 1,   // shuffleUpNV
+    Down = 2, // shuffleDownNV
+    Bfly = 3, // shuffleXorNV
+};
+
 union Instruction {
     Instruction& operator=(const Instruction& instr) {
         value = instr.value;
@@ -577,6 +606,15 @@ union Instruction {
         BitField<42, 1, u64> negate_value;
     } vote;
 
+    union {
+        BitField<30, 2, ShuffleOperation> operation;
+        BitField<48, 3, u64> pred48;
+        BitField<28, 1, u64> is_index_imm;
+        BitField<29, 1, u64> is_mask_imm;
+        BitField<20, 5, u64> index_imm;
+        BitField<34, 13, u64> mask_imm;
+    } shfl;
+
     union {
         BitField<8, 8, Register> gpr;
         BitField<20, 24, s64> offset;
@@ -911,6 +949,11 @@ union Instruction {
         BitField<49, 3, PredCondition> cond;
     } isetp;
 
+    union {
+        BitField<48, 1, u64> is_signed;
+        BitField<49, 3, PredCondition> cond;
+    } icmp;
+
     union {
         BitField<0, 3, u64> pred0;
         BitField<3, 3, u64> pred3;
@@ -1391,6 +1434,14 @@ union Instruction {
         }
     } sust;
 
+    union {
+        BitField<28, 1, u64> is_ba;
+        BitField<51, 3, ImageAtomicSize> size;
+        BitField<33, 3, ImageType> image_type;
+        BitField<29, 4, ImageAtomicOperation> operation;
+        BitField<49, 2, OutOfBoundsStore> out_of_bounds_store;
+    } suatom_d;
+
     union {
         BitField<20, 24, u64> target;
         BitField<5, 1, u64> constant_buffer;
@@ -1512,6 +1563,7 @@ public:
         BRK,
         DEPBAR,
         VOTE,
+        SHFL,
         BFE_C,
         BFE_R,
         BFE_IMM,
@@ -1543,6 +1595,7 @@ public:
         TMML_B, // Texture Mip Map Level
         TMML,   // Texture Mip Map Level
         SUST,   // Surface Store
+        SUATOM, // Surface Atomic Operation
         EXIT,
         NOP,
         IPA,
@@ -1597,6 +1650,10 @@ public:
         SEL_C,
         SEL_R,
         SEL_IMM,
+        ICMP_RC,
+        ICMP_R,
+        ICMP_CR,
+        ICMP_IMM,
         MUFU,  // Multi-Function Operator
         RRO_C, // Range Reduction Operator
         RRO_R,
@@ -1802,6 +1859,7 @@ private:
             INST("111000110000----", Id::EXIT, Type::Flow, "EXIT"),
             INST("1111000011110---", Id::DEPBAR, Type::Synch, "DEPBAR"),
             INST("0101000011011---", Id::VOTE, Type::Warp, "VOTE"),
+            INST("1110111100010---", Id::SHFL, Type::Warp, "SHFL"),
             INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"),
             INST("1110111101001---", Id::LD_S, Type::Memory, "LD_S"),
             INST("1110111101000---", Id::LD_L, Type::Memory, "LD_L"),
@@ -1826,6 +1884,7 @@ private:
             INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"),
             INST("1101111101011---", Id::TMML, Type::Texture, "TMML"),
             INST("11101011001-----", Id::SUST, Type::Image, "SUST"),
+            INST("1110101000------", Id::SUATOM, Type::Image, "SUATOM_D"),
             INST("0101000010110---", Id::NOP, Type::Trivial, "NOP"),
             INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
             INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),
@@ -1860,6 +1919,10 @@ private:
             INST("0100110010100---", Id::SEL_C, Type::ArithmeticInteger, "SEL_C"),
             INST("0101110010100---", Id::SEL_R, Type::ArithmeticInteger, "SEL_R"),
             INST("0011100-10100---", Id::SEL_IMM, Type::ArithmeticInteger, "SEL_IMM"),
+            INST("010100110100----", Id::ICMP_RC, Type::ArithmeticInteger, "ICMP_RC"),
+            INST("010110110100----", Id::ICMP_R, Type::ArithmeticInteger, "ICMP_R"),
+            INST("010010110100----", Id::ICMP_CR, Type::ArithmeticInteger, "ICMP_CR"),
+            INST("0011011-0100----", Id::ICMP_IMM, Type::ArithmeticInteger, "ICMP_IMM"),
             INST("0101101111011---", Id::LEA_R2, Type::ArithmeticInteger, "LEA_R2"),
             INST("0101101111010---", Id::LEA_R1, Type::ArithmeticInteger, "LEA_R1"),
             INST("001101101101----", Id::LEA_IMM, Type::ArithmeticInteger, "LEA_IMM"),

src/video_core/gpu.cpp

@@ -122,6 +122,7 @@ u32 RenderTargetBytesPerPixel(RenderTargetFormat format) {
     case RenderTargetFormat::RGBA16_UINT:
     case RenderTargetFormat::RGBA16_UNORM:
     case RenderTargetFormat::RGBA16_FLOAT:
+    case RenderTargetFormat::RGBX16_FLOAT:
     case RenderTargetFormat::RG32_FLOAT:
     case RenderTargetFormat::RG32_UINT:
         return 8;

src/video_core/gpu.h

@@ -42,6 +42,7 @@ enum class RenderTargetFormat : u32 {
     RGBA16_FLOAT = 0xCA,
     RG32_FLOAT = 0xCB,
     RG32_UINT = 0xCD,
+    RGBX16_FLOAT = 0xCE,
     BGRA8_UNORM = 0xCF,
     BGRA8_SRGB = 0xD0,
     RGB10_A2_UNORM = 0xD1,

src/video_core/macro_interpreter.cpp

@@ -131,9 +131,7 @@ bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {
 
     // An instruction with the Exit flag will not actually
     // cause an exit if it's executed inside a delay slot.
-    // TODO(Blinkhawk): Reversed to always exit. The behavior explained above requires further
-    // testing on the MME code.
-    if (opcode.is_exit) {
+    if (opcode.is_exit && !is_delay_slot) {
         // Exit has a delay slot, execute the next instruction
         Step(offset, true);
         return false;
@@ -259,7 +257,7 @@ void MacroInterpreter::SetMethodAddress(u32 address) {
 }
 
 void MacroInterpreter::Send(u32 value) {
-    maxwell3d.CallMethod({method_address.address, value});
+    maxwell3d.CallMethodFromMME({method_address.address, value});
     // Increment the method address by the method increment.
     method_address.address.Assign(method_address.address.Value() +
                                   method_address.increment.Value());

src/video_core/morton.cpp

@@ -83,6 +83,7 @@ static constexpr ConversionArray morton_to_linear_fns = {
     MortonCopy<true, PixelFormat::RG8U>,
     MortonCopy<true, PixelFormat::RG8S>,
     MortonCopy<true, PixelFormat::RG32UI>,
+    MortonCopy<true, PixelFormat::RGBX16F>,
     MortonCopy<true, PixelFormat::R32UI>,
     MortonCopy<true, PixelFormat::ASTC_2D_8X8>,
     MortonCopy<true, PixelFormat::ASTC_2D_8X5>,
@@ -151,6 +152,7 @@ static constexpr ConversionArray linear_to_morton_fns = {
     MortonCopy<false, PixelFormat::RG8U>,
     MortonCopy<false, PixelFormat::RG8S>,
     MortonCopy<false, PixelFormat::RG32UI>,
+    MortonCopy<false, PixelFormat::RGBX16F>,
     MortonCopy<false, PixelFormat::R32UI>,
     nullptr,
     nullptr,

src/video_core/rasterizer_interface.h

@@ -29,7 +29,10 @@ public:
     virtual ~RasterizerInterface() {}
 
     /// Draw the current batch of vertex arrays
-    virtual void DrawArrays() = 0;
+    virtual bool DrawBatch(bool is_indexed) = 0;
+
+    /// Draw the current batch of multiple instances of vertex arrays
+    virtual bool DrawMultiBatch(bool is_indexed) = 0;
 
     /// Clear the current framebuffer
     virtual void Clear() = 0;
@@ -69,10 +72,6 @@ public:
         return false;
     }
 
-    virtual bool AccelerateDrawBatch(bool is_indexed) {
-        return false;
-    }
-
     /// Increase/decrease the number of object in pages touching the specified region
     virtual void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {}
 

src/video_core/renderer_opengl/gl_framebuffer_cache.cpp

@@ -35,21 +35,16 @@ OGLFramebuffer FramebufferCacheOpenGL::CreateFramebuffer(const FramebufferCacheK
     local_state.draw.draw_framebuffer = framebuffer.handle;
     local_state.ApplyFramebufferState();
 
-    if (key.is_single_buffer) {
-        if (key.color_attachments[0] != GL_NONE && key.colors[0]) {
-            key.colors[0]->Attach(key.color_attachments[0], GL_DRAW_FRAMEBUFFER);
-            glDrawBuffer(key.color_attachments[0]);
-        } else {
-            glDrawBuffer(GL_NONE);
-        }
-    } else {
-        for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) {
-            if (key.colors[index]) {
-                key.colors[index]->Attach(GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(index),
-                                          GL_DRAW_FRAMEBUFFER);
-            }
+    for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) {
+        if (key.colors[index]) {
+            key.colors[index]->Attach(GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(index),
+                                      GL_DRAW_FRAMEBUFFER);
         }
+    }
+    if (key.colors_count) {
         glDrawBuffers(key.colors_count, key.color_attachments.data());
+    } else {
+        glDrawBuffer(GL_NONE);
     }
 
     if (key.zeta) {
@@ -67,9 +62,9 @@ std::size_t FramebufferCacheKey::Hash() const {
 }
 
 bool FramebufferCacheKey::operator==(const FramebufferCacheKey& rhs) const {
-    return std::tie(is_single_buffer, stencil_enable, colors_count, color_attachments, colors,
-                    zeta) == std::tie(rhs.is_single_buffer, rhs.stencil_enable, rhs.colors_count,
-                                      rhs.color_attachments, rhs.colors, rhs.zeta);
+    return std::tie(stencil_enable, colors_count, color_attachments, colors, zeta) ==
+           std::tie(rhs.stencil_enable, rhs.colors_count, rhs.color_attachments, rhs.colors,
+                    rhs.zeta);
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_framebuffer_cache.h

@@ -19,7 +19,6 @@
 namespace OpenGL {
 
 struct alignas(sizeof(u64)) FramebufferCacheKey {
-    bool is_single_buffer = false;
     bool stencil_enable = false;
     u16 colors_count = 0;
 

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -49,40 +49,6 @@ MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100));
 MICROPROFILE_DEFINE(OpenGL_PrimitiveAssembly, "OpenGL", "Prim Asmbl", MP_RGB(255, 100, 100));
 
-struct DrawParameters {
-    GLenum primitive_mode;
-    GLsizei count;
-    GLint current_instance;
-    bool use_indexed;
-
-    GLint vertex_first;
-
-    GLenum index_format;
-    GLint base_vertex;
-    GLintptr index_buffer_offset;
-
-    void DispatchDraw() const {
-        if (use_indexed) {
-            const auto index_buffer_ptr = reinterpret_cast<const void*>(index_buffer_offset);
-            if (current_instance > 0) {
-                glDrawElementsInstancedBaseVertexBaseInstance(primitive_mode, count, index_format,
-                                                              index_buffer_ptr, 1, base_vertex,
-                                                              current_instance);
-            } else {
-                glDrawElementsBaseVertex(primitive_mode, count, index_format, index_buffer_ptr,
-                                         base_vertex);
-            }
-        } else {
-            if (current_instance > 0) {
-                glDrawArraysInstancedBaseInstance(primitive_mode, vertex_first, count, 1,
-                                                  current_instance);
-            } else {
-                glDrawArrays(primitive_mode, vertex_first, count);
-            }
-        }
-    }
-};
-
 static std::size_t GetConstBufferSize(const Tegra::Engines::ConstBufferInfo& buffer,
                                       const GLShader::ConstBufferEntry& entry) {
     if (!entry.IsIndirect()) {
@@ -270,29 +236,6 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() {
     return offset;
 }
 
-DrawParameters RasterizerOpenGL::SetupDraw(GLintptr index_buffer_offset) {
-    const auto& gpu = system.GPU().Maxwell3D();
-    const auto& regs = gpu.regs;
-    const bool is_indexed = accelerate_draw == AccelDraw::Indexed;
-
-    DrawParameters params{};
-    params.current_instance = gpu.state.current_instance;
-
-    params.use_indexed = is_indexed;
-    params.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology);
-
-    if (is_indexed) {
-        params.index_format = MaxwellToGL::IndexFormat(regs.index_array.format);
-        params.count = regs.index_array.count;
-        params.index_buffer_offset = index_buffer_offset;
-        params.base_vertex = static_cast<GLint>(regs.vb_element_base);
-    } else {
-        params.count = regs.vertex_buffer.count;
-        params.vertex_first = regs.vertex_buffer.first;
-    }
-    return params;
-}
-
 void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     MICROPROFILE_SCOPE(OpenGL_Shader);
     auto& gpu = system.GPU().Maxwell3D();
@@ -399,12 +342,6 @@ std::size_t RasterizerOpenGL::CalculateIndexBufferSize() const {
            static_cast<std::size_t>(regs.index_array.FormatSizeInBytes());
 }
 
-bool RasterizerOpenGL::AccelerateDrawBatch(bool is_indexed) {
-    accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays;
-    DrawArrays();
-    return true;
-}
-
 template <typename Map, typename Interval>
 static constexpr auto RangeFromInterval(Map& map, const Interval& interval) {
     return boost::make_iterator_range(map.equal_range(interval));
@@ -445,107 +382,51 @@ void RasterizerOpenGL::LoadDiskResources(const std::atomic_bool& stop_loading,
     shader_cache.LoadDiskCache(stop_loading, callback);
 }
 
-std::pair<bool, bool> RasterizerOpenGL::ConfigureFramebuffers(
-    OpenGLState& current_state, bool using_color_fb, bool using_depth_fb, bool preserve_contents,
-    std::optional<std::size_t> single_color_target) {
+void RasterizerOpenGL::ConfigureFramebuffers() {
     MICROPROFILE_SCOPE(OpenGL_Framebuffer);
     auto& gpu = system.GPU().Maxwell3D();
-    const auto& regs = gpu.regs;
-
-    const FramebufferConfigState fb_config_state{using_color_fb, using_depth_fb, preserve_contents,
-                                                 single_color_target};
-    if (fb_config_state == current_framebuffer_config_state && !gpu.dirty.render_settings) {
-        // Only skip if the previous ConfigureFramebuffers call was from the same kind (multiple or
-        // single color targets). This is done because the guest registers may not change but the
-        // host framebuffer may contain different attachments
-        return current_depth_stencil_usage;
+    if (!gpu.dirty.render_settings) {
+        return;
     }
     gpu.dirty.render_settings = false;
-    current_framebuffer_config_state = fb_config_state;
 
     texture_cache.GuardRenderTargets(true);
 
-    View depth_surface{};
-    if (using_depth_fb) {
-        depth_surface = texture_cache.GetDepthBufferSurface(preserve_contents);
-    } else {
-        texture_cache.SetEmptyDepthBuffer();
-    }
+    View depth_surface = texture_cache.GetDepthBufferSurface(true);
 
+    const auto& regs = gpu.regs;
+    state.framebuffer_srgb.enabled = regs.framebuffer_srgb != 0;
     UNIMPLEMENTED_IF(regs.rt_separate_frag_data == 0);
 
     // Bind the framebuffer surfaces
-    current_state.framebuffer_srgb.enabled = regs.framebuffer_srgb != 0;
-
     FramebufferCacheKey fbkey;
+    for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) {
+        View color_surface{texture_cache.GetColorBufferSurface(index, true)};
 
-    if (using_color_fb) {
-        if (single_color_target) {
-            // Used when just a single color attachment is enabled, e.g. for clearing a color buffer
-            View color_surface{
-                texture_cache.GetColorBufferSurface(*single_color_target, preserve_contents)};
-
-            if (color_surface) {
-                // Assume that a surface will be written to if it is used as a framebuffer, even if
-                // the shader doesn't actually write to it.
-                texture_cache.MarkColorBufferInUse(*single_color_target);
-                // Workaround for and issue in nvidia drivers
-                // https://devtalk.nvidia.com/default/topic/776591/opengl/gl_framebuffer_srgb-functions-incorrectly/
-                state.framebuffer_srgb.enabled |= color_surface->GetSurfaceParams().srgb_conversion;
-            }
-
-            fbkey.is_single_buffer = true;
-            fbkey.color_attachments[0] =
-                GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(*single_color_target);
-            fbkey.colors[0] = color_surface;
-            for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) {
-                if (index != *single_color_target) {
-                    texture_cache.SetEmptyColorBuffer(index);
-                }
-            }
-        } else {
-            // Multiple color attachments are enabled
-            for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) {
-                View color_surface{texture_cache.GetColorBufferSurface(index, preserve_contents)};
-
-                if (color_surface) {
-                    // Assume that a surface will be written to if it is used as a framebuffer, even
-                    // if the shader doesn't actually write to it.
-                    texture_cache.MarkColorBufferInUse(index);
-                    // Enable sRGB only for supported formats
-                    // Workaround for and issue in nvidia drivers
-                    // https://devtalk.nvidia.com/default/topic/776591/opengl/gl_framebuffer_srgb-functions-incorrectly/
-                    state.framebuffer_srgb.enabled |=
-                        color_surface->GetSurfaceParams().srgb_conversion;
-                }
-
-                fbkey.color_attachments[index] =
-                    GL_COLOR_ATTACHMENT0 + regs.rt_control.GetMap(index);
-                fbkey.colors[index] = color_surface;
-            }
-            fbkey.is_single_buffer = false;
-            fbkey.colors_count = regs.rt_control.count;
+        if (color_surface) {
+            // Assume that a surface will be written to if it is used as a framebuffer, even
+            // if the shader doesn't actually write to it.
+            texture_cache.MarkColorBufferInUse(index);
         }
-    } else {
-        // No color attachments are enabled - leave them as zero
-        fbkey.is_single_buffer = true;
+
+        fbkey.color_attachments[index] = GL_COLOR_ATTACHMENT0 + regs.rt_control.GetMap(index);
+        fbkey.colors[index] = std::move(color_surface);
     }
+    fbkey.colors_count = regs.rt_control.count;
 
     if (depth_surface) {
         // Assume that a surface will be written to if it is used as a framebuffer, even if
         // the shader doesn't actually write to it.
         texture_cache.MarkDepthBufferInUse();
 
-        fbkey.zeta = depth_surface;
         fbkey.stencil_enable = depth_surface->GetSurfaceParams().type == SurfaceType::DepthStencil;
+        fbkey.zeta = std::move(depth_surface);
     }
 
     texture_cache.GuardRenderTargets(false);
 
-    current_state.draw.draw_framebuffer = framebuffer_cache.GetFramebuffer(fbkey);
-    SyncViewport(current_state);
-
-    return current_depth_stencil_usage = {static_cast<bool>(depth_surface), fbkey.stencil_enable};
+    state.draw.draw_framebuffer = framebuffer_cache.GetFramebuffer(fbkey);
+    SyncViewport(state);
 }
 
 void RasterizerOpenGL::ConfigureClearFramebuffer(OpenGLState& current_state, bool using_color_fb,
@@ -696,17 +577,9 @@ void RasterizerOpenGL::Clear() {
     }
 }
 
-void RasterizerOpenGL::DrawArrays() {
-    if (accelerate_draw == AccelDraw::Disabled)
-        return;
-
-    MICROPROFILE_SCOPE(OpenGL_Drawing);
+void RasterizerOpenGL::DrawPrelude() {
     auto& gpu = system.GPU().Maxwell3D();
 
-    if (!gpu.ShouldExecute()) {
-        return;
-    }
-
     SyncColorMask();
     SyncFragmentColorClampState();
     SyncMultiSampleState();
@@ -751,10 +624,7 @@ void RasterizerOpenGL::DrawArrays() {
     // Upload vertex and index data.
     SetupVertexBuffer(vao);
     SetupVertexInstances(vao);
-    const GLintptr index_buffer_offset = SetupIndexBuffer();
-
-    // Setup draw parameters. It will automatically choose what glDraw* method to use.
-    const DrawParameters params = SetupDraw(index_buffer_offset);
+    index_buffer_offset = SetupIndexBuffer();
 
     // Prepare packed bindings.
     bind_ubo_pushbuffer.Setup(0);
@@ -762,10 +632,11 @@ void RasterizerOpenGL::DrawArrays() {
 
     // Setup shaders and their used resources.
     texture_cache.GuardSamplers(true);
-    SetupShaders(params.primitive_mode);
+    const auto primitive_mode = MaxwellToGL::PrimitiveTopology(gpu.regs.draw.topology);
+    SetupShaders(primitive_mode);
     texture_cache.GuardSamplers(false);
 
-    ConfigureFramebuffers(state);
+    ConfigureFramebuffers();
 
     // Signal the buffer cache that we are not going to upload more things.
     const bool invalidate = buffer_cache.Unmap();
@@ -786,11 +657,107 @@ void RasterizerOpenGL::DrawArrays() {
     if (texture_cache.TextureBarrier()) {
         glTextureBarrier();
     }
+}
 
-    params.DispatchDraw();
+struct DrawParams {
+    bool is_indexed{};
+    bool is_instanced{};
+    GLenum primitive_mode{};
+    GLint count{};
+    GLint base_vertex{};
 
+    // Indexed settings
+    GLenum index_format{};
+    GLintptr index_buffer_offset{};
+
+    // Instanced setting
+    GLint num_instances{};
+    GLint base_instance{};
+
+    void DispatchDraw() {
+        if (is_indexed) {
+            const auto index_buffer_ptr = reinterpret_cast<const void*>(index_buffer_offset);
+            if (is_instanced) {
+                glDrawElementsInstancedBaseVertexBaseInstance(primitive_mode, count, index_format,
+                                                              index_buffer_ptr, num_instances,
+                                                              base_vertex, base_instance);
+            } else {
+                glDrawElementsBaseVertex(primitive_mode, count, index_format, index_buffer_ptr,
+                                         base_vertex);
+            }
+        } else {
+            if (is_instanced) {
+                glDrawArraysInstancedBaseInstance(primitive_mode, base_vertex, count, num_instances,
+                                                  base_instance);
+            } else {
+                glDrawArrays(primitive_mode, base_vertex, count);
+            }
+        }
+    }
+};
+
+bool RasterizerOpenGL::DrawBatch(bool is_indexed) {
+    accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays;
+
+    MICROPROFILE_SCOPE(OpenGL_Drawing);
+
+    DrawPrelude();
+
+    auto& maxwell3d = system.GPU().Maxwell3D();
+    const auto& regs = maxwell3d.regs;
+    const auto current_instance = maxwell3d.state.current_instance;
+    DrawParams draw_call{};
+    draw_call.is_indexed = is_indexed;
+    draw_call.num_instances = static_cast<GLint>(1);
+    draw_call.base_instance = static_cast<GLint>(current_instance);
+    draw_call.is_instanced = current_instance > 0;
+    draw_call.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology);
+    if (draw_call.is_indexed) {
+        draw_call.count = static_cast<GLint>(regs.index_array.count);
+        draw_call.base_vertex = static_cast<GLint>(regs.vb_element_base);
+        draw_call.index_format = MaxwellToGL::IndexFormat(regs.index_array.format);
+        draw_call.index_buffer_offset = index_buffer_offset;
+    } else {
+        draw_call.count = static_cast<GLint>(regs.vertex_buffer.count);
+        draw_call.base_vertex = static_cast<GLint>(regs.vertex_buffer.first);
+    }
+    draw_call.DispatchDraw();
+
+    maxwell3d.dirty.memory_general = false;
     accelerate_draw = AccelDraw::Disabled;
-    gpu.dirty.memory_general = false;
+    return true;
+}
+
+bool RasterizerOpenGL::DrawMultiBatch(bool is_indexed) {
+    accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays;
+
+    MICROPROFILE_SCOPE(OpenGL_Drawing);
+
+    DrawPrelude();
+
+    auto& maxwell3d = system.GPU().Maxwell3D();
+    const auto& regs = maxwell3d.regs;
+    const auto& draw_setup = maxwell3d.mme_draw;
+    DrawParams draw_call{};
+    draw_call.is_indexed = is_indexed;
+    draw_call.num_instances = static_cast<GLint>(draw_setup.instance_count);
+    draw_call.base_instance = static_cast<GLint>(regs.vb_base_instance);
+    draw_call.is_instanced = draw_setup.instance_count > 1;
+    draw_call.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology);
+    if (draw_call.is_indexed) {
+        draw_call.count = static_cast<GLint>(regs.index_array.count);
+        draw_call.base_vertex = static_cast<GLint>(regs.vb_element_base);
+        draw_call.index_format = MaxwellToGL::IndexFormat(regs.index_array.format);
+        draw_call.index_buffer_offset = index_buffer_offset;
+    } else {
+        draw_call.count = static_cast<GLint>(regs.vertex_buffer.count);
+        draw_call.base_vertex = static_cast<GLint>(regs.vertex_buffer.first);
+    }
+    draw_call.DispatchDraw();
+
+    maxwell3d.dirty.memory_general = false;
+    accelerate_draw = AccelDraw::Disabled;
+    return true;
 }
 
 void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
@@ -906,6 +873,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
     }
 
     screen_info.display_texture = surface->GetTexture();
+    screen_info.display_srgb = surface->GetSurfaceParams().srgb_conversion;
 
     return true;
 }

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -57,7 +57,8 @@ public:
                               ScreenInfo& info);
     ~RasterizerOpenGL() override;
 
-    void DrawArrays() override;
+    bool DrawBatch(bool is_indexed) override;
+    bool DrawMultiBatch(bool is_indexed) override;
     void Clear() override;
     void DispatchCompute(GPUVAddr code_addr) override;
     void FlushAll() override;
@@ -71,45 +72,13 @@ public:
                                const Tegra::Engines::Fermi2D::Config& copy_config) override;
     bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
                            u32 pixel_stride) override;
-    bool AccelerateDrawBatch(bool is_indexed) override;
     void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) override;
     void LoadDiskResources(const std::atomic_bool& stop_loading,
                            const VideoCore::DiskResourceLoadCallback& callback) override;
 
 private:
-    struct FramebufferConfigState {
-        bool using_color_fb{};
-        bool using_depth_fb{};
-        bool preserve_contents{};
-        std::optional<std::size_t> single_color_target;
-
-        bool operator==(const FramebufferConfigState& rhs) const {
-            return std::tie(using_color_fb, using_depth_fb, preserve_contents,
-                            single_color_target) == std::tie(rhs.using_color_fb, rhs.using_depth_fb,
-                                                             rhs.preserve_contents,
-                                                             rhs.single_color_target);
-        }
-        bool operator!=(const FramebufferConfigState& rhs) const {
-            return !operator==(rhs);
-        }
-    };
-
-    /**
-     * Configures the color and depth framebuffer states.
-     *
-     * @param current_state       The current OpenGL state.
-     * @param using_color_fb      If true, configure color framebuffers.
-     * @param using_depth_fb      If true, configure the depth/stencil framebuffer.
-     * @param preserve_contents   If true, tries to preserve data from a previously used
-     *                            framebuffer.
-     * @param single_color_target Specifies if a single color buffer target should be used.
-     *
-     * @returns If depth (first) or stencil (second) are being stored in the bound zeta texture
-     *          (requires using_depth_fb to be true)
-     */
-    std::pair<bool, bool> ConfigureFramebuffers(
-        OpenGLState& current_state, bool using_color_fb = true, bool using_depth_fb = true,
-        bool preserve_contents = true, std::optional<std::size_t> single_color_target = {});
+    /// Configures the color and depth framebuffer states.
+    void ConfigureFramebuffers();
 
     void ConfigureClearFramebuffer(OpenGLState& current_state, bool using_color_fb,
                                    bool using_depth_fb, bool using_stencil_fb);
@@ -136,6 +105,9 @@ private:
     void SetupGlobalMemory(const GLShader::GlobalMemoryEntry& entry, GPUVAddr gpu_addr,
                            std::size_t size);
 
+    /// Syncs all the state, shaders, render targets and textures setting before a draw call.
+    void DrawPrelude();
+
     /// Configures the current textures to use for the draw command. Returns shaders texture buffer
     /// usage.
     TextureBufferUsage SetupDrawTextures(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
@@ -228,9 +200,6 @@ private:
              OGLVertexArray>
         vertex_array_cache;
 
-    FramebufferConfigState current_framebuffer_config_state;
-    std::pair<bool, bool> current_depth_stencil_usage{};
-
     static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
     OGLBufferCache buffer_cache;
 
@@ -250,7 +219,7 @@ private:
 
     GLintptr SetupIndexBuffer();
 
-    DrawParameters SetupDraw(GLintptr index_buffer_offset);
+    GLintptr index_buffer_offset;
 
     void SetupShaders(GLenum primitive_mode);
 

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -214,7 +214,8 @@ CachedProgram SpecializeShader(const std::string& code, const GLShader::ShaderEn
     std::string source = "#version 430 core\n"
                          "#extension GL_ARB_separate_shader_objects : enable\n"
                          "#extension GL_NV_gpu_shader5 : enable\n"
-                         "#extension GL_NV_shader_thread_group : enable\n";
+                         "#extension GL_NV_shader_thread_group : enable\n"
+                         "#extension GL_NV_shader_thread_shuffle : enable\n";
     if (entries.shader_viewport_layer_array) {
         source += "#extension GL_ARB_shader_viewport_layer_array : enable\n";
     }

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -325,6 +325,7 @@ public:
         DeclareRegisters();
         DeclarePredicates();
         DeclareLocalMemory();
+        DeclareSharedMemory();
         DeclareInternalFlags();
         DeclareInputAttributes();
         DeclareOutputAttributes();
@@ -461,6 +462,14 @@ private:
             code.AddLine("float gl_PointSize;");
         }
 
+        if (ir.UsesInstanceId()) {
+            code.AddLine("int gl_InstanceID;");
+        }
+
+        if (ir.UsesVertexId()) {
+            code.AddLine("int gl_VertexID;");
+        }
+
         --code.scope;
         code.AddLine("}};");
         code.AddNewLine();
@@ -499,6 +508,13 @@ private:
         code.AddNewLine();
     }
 
+    void DeclareSharedMemory() {
+        if (stage != ProgramType::Compute) {
+            return;
+        }
+        code.AddLine("shared uint {}[];", GetSharedMemory());
+    }
+
     void DeclareInternalFlags() {
         for (u32 flag = 0; flag < static_cast<u32>(InternalFlag::Amount); flag++) {
             const auto flag_code = static_cast<InternalFlag>(flag);
@@ -706,7 +722,7 @@ private:
     void DeclareImages() {
         const auto& images{ir.GetImages()};
         for (const auto& [offset, image] : images) {
-            const std::string image_type = [&]() {
+            const char* image_type = [&] {
                 switch (image.GetType()) {
                 case Tegra::Shader::ImageType::Texture1D:
                     return "image1D";
@@ -725,6 +741,23 @@ private:
                     return "image1D";
                 }
             }();
+
+            const auto [type_prefix, format] = [&]() -> std::pair<const char*, const char*> {
+                if (!image.IsSizeKnown()) {
+                    return {"", ""};
+                }
+                switch (image.GetSize()) {
+                case Tegra::Shader::ImageAtomicSize::U32:
+                    return {"u", "r32ui, "};
+                case Tegra::Shader::ImageAtomicSize::S32:
+                    return {"i", "r32i, "};
+                default:
+                    UNIMPLEMENTED_MSG("Unimplemented atomic size={}",
+                                      static_cast<u32>(image.GetSize()));
+                    return {"", ""};
+                }
+            }();
+
             std::string qualifier = "coherent volatile";
             if (image.IsRead() && !image.IsWritten()) {
                 qualifier += " readonly";
@@ -864,6 +897,12 @@ private:
                 Type::Uint};
         }
 
+        if (const auto smem = std::get_if<SmemNode>(&*node)) {
+            return {
+                fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
+                Type::Uint};
+        }
+
         if (const auto internal_flag = std::get_if<InternalFlagNode>(&*node)) {
             return {GetInternalFlag(internal_flag->GetFlag()), Type::Bool};
         }
@@ -933,7 +972,7 @@ private:
             switch (element) {
             case 2:
                 // Config pack's first value is instance_id.
-                return {"config_pack[0]", Type::Uint};
+                return {"gl_InstanceID", Type::Int};
             case 3:
                 return {"gl_VertexID", Type::Int};
             }
@@ -990,10 +1029,10 @@ private:
         return {std::move(temporary), value.GetType()};
     }
 
-    Expression GetOutputAttribute(const AbufNode* abuf) {
+    std::optional<Expression> GetOutputAttribute(const AbufNode* abuf) {
         switch (const auto attribute = abuf->GetIndex()) {
         case Attribute::Index::Position:
-            return {"gl_Position"s + GetSwizzle(abuf->GetElement()), Type::Float};
+            return {{"gl_Position"s + GetSwizzle(abuf->GetElement()), Type::Float}};
         case Attribute::Index::LayerViewportPointSize:
             switch (abuf->GetElement()) {
             case 0:
@@ -1003,25 +1042,25 @@ private:
                 if (IsVertexShader(stage) && !device.HasVertexViewportLayer()) {
                     return {};
                 }
-                return {"gl_Layer", Type::Int};
+                return {{"gl_Layer", Type::Int}};
             case 2:
                 if (IsVertexShader(stage) && !device.HasVertexViewportLayer()) {
                     return {};
                 }
-                return {"gl_ViewportIndex", Type::Int};
+                return {{"gl_ViewportIndex", Type::Int}};
             case 3:
                 UNIMPLEMENTED_MSG("Requires some state changes for gl_PointSize to work in shader");
-                return {"gl_PointSize", Type::Float};
+                return {{"gl_PointSize", Type::Float}};
             }
             return {};
         case Attribute::Index::ClipDistances0123:
-            return {fmt::format("gl_ClipDistance[{}]", abuf->GetElement()), Type::Float};
+            return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement()), Type::Float}};
         case Attribute::Index::ClipDistances4567:
-            return {fmt::format("gl_ClipDistance[{}]", abuf->GetElement() + 4), Type::Float};
+            return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement() + 4), Type::Float}};
         default:
             if (IsGenericAttribute(attribute)) {
-                return {GetOutputAttribute(attribute) + GetSwizzle(abuf->GetElement()),
-                        Type::Float};
+                return {
+                    {GetOutputAttribute(attribute) + GetSwizzle(abuf->GetElement()), Type::Float}};
             }
             UNIMPLEMENTED_MSG("Unhandled output attribute: {}", static_cast<u32>(attribute));
             return {};
@@ -1180,6 +1219,74 @@ private:
         return expr;
     }
 
+    std::string BuildIntegerCoordinates(Operation operation) {
+        constexpr std::array constructors{"int(", "ivec2(", "ivec3(", "ivec4("};
+        const std::size_t coords_count{operation.GetOperandsCount()};
+        std::string expr = constructors.at(coords_count - 1);
+        for (std::size_t i = 0; i < coords_count; ++i) {
+            expr += VisitOperand(operation, i).AsInt();
+            if (i + 1 < coords_count) {
+                expr += ", ";
+            }
+        }
+        expr += ')';
+        return expr;
+    }
+
+    std::string BuildImageValues(Operation operation) {
+        const auto meta{std::get<MetaImage>(operation.GetMeta())};
+        const auto [constructors, type] = [&]() -> std::pair<std::array<const char*, 4>, Type> {
+            constexpr std::array float_constructors{"float", "vec2", "vec3", "vec4"};
+            if (!meta.image.IsSizeKnown()) {
+                return {float_constructors, Type::Float};
+            }
+            switch (meta.image.GetSize()) {
+            case Tegra::Shader::ImageAtomicSize::U32:
+                return {{"uint", "uvec2", "uvec3", "uvec4"}, Type::Uint};
+            case Tegra::Shader::ImageAtomicSize::S32:
+                return {{"int", "ivec2", "ivec3", "ivec4"}, Type::Uint};
+            default:
+                UNIMPLEMENTED_MSG("Unimplemented image size={}",
+                                  static_cast<u32>(meta.image.GetSize()));
+                return {float_constructors, Type::Float};
+            }
+        }();
+
+        const std::size_t values_count{meta.values.size()};
+        std::string expr = fmt::format("{}(", constructors.at(values_count - 1));
+        for (std::size_t i = 0; i < values_count; ++i) {
+            expr += Visit(meta.values.at(i)).As(type);
+            if (i + 1 < values_count) {
+                expr += ", ";
+            }
+        }
+        expr += ')';
+        return expr;
+    }
+
+    Expression AtomicImage(Operation operation, const char* opname) {
+        constexpr std::array constructors{"int(", "ivec2(", "ivec3(", "ivec4("};
+        const auto meta{std::get<MetaImage>(operation.GetMeta())};
+        ASSERT(meta.values.size() == 1);
+        ASSERT(meta.image.IsSizeKnown());
+
+        const auto type = [&]() {
+            switch (const auto size = meta.image.GetSize()) {
+            case Tegra::Shader::ImageAtomicSize::U32:
+                return Type::Uint;
+            case Tegra::Shader::ImageAtomicSize::S32:
+                return Type::Int;
+            default:
+                UNIMPLEMENTED_MSG("Unimplemented image size={}", static_cast<u32>(size));
+                return Type::Uint;
+            }
+        }();
+
+        return {fmt::format("{}({}, {}, {})", opname, GetImage(meta.image),
+                            BuildIntegerCoordinates(operation), Visit(meta.values[0]).As(type)),
+                type};
+    }
+
     Expression Assign(Operation operation) {
         const Node& dest = operation[0];
         const Node& src = operation[1];
@@ -1193,7 +1300,11 @@ private:
             target = {GetRegister(gpr->GetIndex()), Type::Float};
         } else if (const auto abuf = std::get_if<AbufNode>(&*dest)) {
             UNIMPLEMENTED_IF(abuf->IsPhysicalBuffer());
-            target = GetOutputAttribute(abuf);
+            auto output = GetOutputAttribute(abuf);
+            if (!output) {
+                return {};
+            }
+            target = std::move(*output);
         } else if (const auto lmem = std::get_if<LmemNode>(&*dest)) {
             if (stage == ProgramType::Compute) {
                 LOG_WARNING(Render_OpenGL, "Local memory is stubbed on compute shaders");
@@ -1201,6 +1312,11 @@ private:
             target = {
                 fmt::format("{}[{} >> 2]", GetLocalMemory(), Visit(lmem->GetAddress()).AsUint()),
                 Type::Uint};
+        } else if (const auto smem = std::get_if<SmemNode>(&*dest)) {
+            ASSERT(stage == ProgramType::Compute);
+            target = {
+                fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
+                Type::Uint};
         } else if (const auto gmem = std::get_if<GmemNode>(&*dest)) {
             const std::string real = Visit(gmem->GetRealAddress()).AsUint();
             const std::string base = Visit(gmem->GetBaseAddress()).AsUint();
@@ -1694,38 +1810,39 @@ private:
     }
 
     Expression ImageStore(Operation operation) {
-        constexpr std::array constructors{"int(", "ivec2(", "ivec3(", "ivec4("};
         const auto meta{std::get<MetaImage>(operation.GetMeta())};
-
-        std::string expr = "imageStore(";
-        expr += GetImage(meta.image);
-        expr += ", ";
-
-        const std::size_t coords_count{operation.GetOperandsCount()};
-        expr += constructors.at(coords_count - 1);
-        for (std::size_t i = 0; i < coords_count; ++i) {
-            expr += VisitOperand(operation, i).AsInt();
-            if (i + 1 < coords_count) {
-                expr += ", ";
-            }
-        }
-        expr += "), ";
-
-        const std::size_t values_count{meta.values.size()};
-        UNIMPLEMENTED_IF(values_count != 4);
-        expr += "vec4(";
-        for (std::size_t i = 0; i < values_count; ++i) {
-            expr += Visit(meta.values.at(i)).AsFloat();
-            if (i + 1 < values_count) {
-                expr += ", ";
-            }
-        }
-        expr += "));";
-
-        code.AddLine(expr);
+        code.AddLine("imageStore({}, {}, {});", GetImage(meta.image),
+                     BuildIntegerCoordinates(operation), BuildImageValues(operation));
         return {};
     }
 
+    Expression AtomicImageAdd(Operation operation) {
+        return AtomicImage(operation, "imageAtomicAdd");
+    }
+
+    Expression AtomicImageMin(Operation operation) {
+        return AtomicImage(operation, "imageAtomicMin");
+    }
+
+    Expression AtomicImageMax(Operation operation) {
+        return AtomicImage(operation, "imageAtomicMax");
+    }
+    Expression AtomicImageAnd(Operation operation) {
+        return AtomicImage(operation, "imageAtomicAnd");
+    }
+
+    Expression AtomicImageOr(Operation operation) {
+        return AtomicImage(operation, "imageAtomicOr");
+    }
+
+    Expression AtomicImageXor(Operation operation) {
+        return AtomicImage(operation, "imageAtomicXor");
+    }
+
+    Expression AtomicImageExchange(Operation operation) {
+        return AtomicImage(operation, "imageAtomicExchange");
+    }
+
     Expression Branch(Operation operation) {
         const auto target = std::get_if<ImmediateNode>(&*operation[0]);
         UNIMPLEMENTED_IF(!target);
@@ -1848,8 +1965,7 @@ private:
     Expression BallotThread(Operation operation) {
         const std::string value = VisitOperand(operation, 0).AsBool();
         if (!device.HasWarpIntrinsics()) {
-            LOG_ERROR(Render_OpenGL,
-                      "Nvidia warp intrinsics are not available and its required by a shader");
+            LOG_ERROR(Render_OpenGL, "Nvidia vote intrinsics are required by this shader");
             // Stub on non-Nvidia devices by simulating all threads voting the same as the active
             // one.
             return {fmt::format("({} ? 0xFFFFFFFFU : 0U)", value), Type::Uint};
@@ -1860,8 +1976,7 @@ private:
     Expression Vote(Operation operation, const char* func) {
         const std::string value = VisitOperand(operation, 0).AsBool();
         if (!device.HasWarpIntrinsics()) {
-            LOG_ERROR(Render_OpenGL,
-                      "Nvidia vote intrinsics are not available and its required by a shader");
+            LOG_ERROR(Render_OpenGL, "Nvidia vote intrinsics are required by this shader");
             // Stub with a warp size of one.
             return {value, Type::Bool};
         }
@@ -1878,15 +1993,54 @@ private:
 
     Expression VoteEqual(Operation operation) {
         if (!device.HasWarpIntrinsics()) {
-            LOG_ERROR(Render_OpenGL,
-                      "Nvidia vote intrinsics are not available and its required by a shader");
-            // We must return true here since a stub for a theoretical warp size of 1 will always
-            // return an equal result for all its votes.
+            LOG_ERROR(Render_OpenGL, "Nvidia vote intrinsics are required by this shader");
+            // We must return true here since a stub for a theoretical warp size of 1.
+            // This will always return an equal result across all votes.
             return {"true", Type::Bool};
         }
         return Vote(operation, "allThreadsEqualNV");
     }
 
+    template <const std::string_view& func>
+    Expression Shuffle(Operation operation) {
+        const std::string value = VisitOperand(operation, 0).AsFloat();
+        if (!device.HasWarpIntrinsics()) {
+            LOG_ERROR(Render_OpenGL, "Nvidia shuffle intrinsics are required by this shader");
+            // On a "single-thread" device we are either on the same thread or out of bounds. Both
+            // cases return the passed value.
+            return {value, Type::Float};
+        }
+
+        const std::string index = VisitOperand(operation, 1).AsUint();
+        const std::string width = VisitOperand(operation, 2).AsUint();
+        return {fmt::format("{}({}, {}, {})", func, value, index, width), Type::Float};
+    }
+
+    template <const std::string_view& func>
+    Expression InRangeShuffle(Operation operation) {
+        const std::string index = VisitOperand(operation, 0).AsUint();
+        const std::string width = VisitOperand(operation, 1).AsUint();
+        if (!device.HasWarpIntrinsics()) {
+            // On a "single-thread" device we are only in bounds when the requested index is 0.
+            return {fmt::format("({} == 0U)", index), Type::Bool};
+        }
+
+        const std::string in_range = code.GenerateTemporary();
+        code.AddLine("bool {};", in_range);
+        code.AddLine("{}(0U, {}, {}, {});", func, index, width, in_range);
+        return {in_range, Type::Bool};
+    }
+
+    struct Func final {
+        Func() = delete;
+        ~Func() = delete;
+
+        static constexpr std::string_view ShuffleIndexed = "shuffleNV";
+        static constexpr std::string_view ShuffleUp = "shuffleUpNV";
+        static constexpr std::string_view ShuffleDown = "shuffleDownNV";
+        static constexpr std::string_view ShuffleButterfly = "shuffleXorNV";
+    };
+
     static constexpr std::array operation_decompilers = {
         &GLSLDecompiler::Assign,
 
@@ -2019,6 +2173,13 @@ private:
         &GLSLDecompiler::TexelFetch,
 
         &GLSLDecompiler::ImageStore,
+        &GLSLDecompiler::AtomicImageAdd,
+        &GLSLDecompiler::AtomicImageMin,
+        &GLSLDecompiler::AtomicImageMax,
+        &GLSLDecompiler::AtomicImageAnd,
+        &GLSLDecompiler::AtomicImageOr,
+        &GLSLDecompiler::AtomicImageXor,
+        &GLSLDecompiler::AtomicImageExchange,
 
         &GLSLDecompiler::Branch,
         &GLSLDecompiler::BranchIndirect,
@@ -2042,6 +2203,16 @@ private:
         &GLSLDecompiler::VoteAll,
         &GLSLDecompiler::VoteAny,
         &GLSLDecompiler::VoteEqual,
+
+        &GLSLDecompiler::Shuffle<Func::ShuffleIndexed>,
+        &GLSLDecompiler::Shuffle<Func::ShuffleUp>,
+        &GLSLDecompiler::Shuffle<Func::ShuffleDown>,
+        &GLSLDecompiler::Shuffle<Func::ShuffleButterfly>,
+
+        &GLSLDecompiler::InRangeShuffle<Func::ShuffleIndexed>,
+        &GLSLDecompiler::InRangeShuffle<Func::ShuffleUp>,
+        &GLSLDecompiler::InRangeShuffle<Func::ShuffleDown>,
+        &GLSLDecompiler::InRangeShuffle<Func::ShuffleButterfly>,
     };
     static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
 
@@ -2082,6 +2253,10 @@ private:
         return "lmem_" + suffix;
     }
 
+    std::string GetSharedMemory() const {
+        return fmt::format("smem_{}", suffix);
+    }
+
     std::string GetInternalFlag(InternalFlag flag) const {
         constexpr std::array InternalFlagNames = {"zero_flag", "sign_flag", "carry_flag",
                                                   "overflow_flag"};

src/video_core/renderer_opengl/gl_shader_disk_cache.cpp

@@ -341,16 +341,22 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         u64 index{};
         u32 type{};
         u8 is_bindless{};
-        u8 is_read{};
         u8 is_written{};
+        u8 is_read{};
+        u8 is_size_known{};
+        u32 size{};
         if (!LoadObjectFromPrecompiled(offset) || !LoadObjectFromPrecompiled(index) ||
             !LoadObjectFromPrecompiled(type) || !LoadObjectFromPrecompiled(is_bindless) ||
-            !LoadObjectFromPrecompiled(is_read) || !LoadObjectFromPrecompiled(is_written)) {
+            !LoadObjectFromPrecompiled(is_written) || !LoadObjectFromPrecompiled(is_read) ||
+            !LoadObjectFromPrecompiled(is_size_known) || !LoadObjectFromPrecompiled(size)) {
             return {};
         }
-        entry.entries.images.emplace_back(static_cast<u64>(offset), static_cast<std::size_t>(index),
-                                          static_cast<Tegra::Shader::ImageType>(type),
-                                          is_bindless != 0, is_written != 0, is_read != 0);
+        entry.entries.images.emplace_back(
+            static_cast<std::size_t>(offset), static_cast<std::size_t>(index),
+            static_cast<Tegra::Shader::ImageType>(type), is_bindless != 0, is_written != 0,
+            is_read != 0,
+            is_size_known ? std::make_optional(static_cast<Tegra::Shader::ImageAtomicSize>(size))
+                          : std::nullopt);
     }
 
     u32 global_memory_count{};
@@ -429,12 +435,14 @@ bool ShaderDiskCacheOpenGL::SaveDecompiledFile(u64 unique_identifier, const std:
         return false;
     }
     for (const auto& image : entries.images) {
+        const u32 size = image.IsSizeKnown() ? static_cast<u32>(image.GetSize()) : 0U;
         if (!SaveObjectToPrecompiled(static_cast<u64>(image.GetOffset())) ||
             !SaveObjectToPrecompiled(static_cast<u64>(image.GetIndex())) ||
             !SaveObjectToPrecompiled(static_cast<u32>(image.GetType())) ||
             !SaveObjectToPrecompiled(static_cast<u8>(image.IsBindless() ? 1 : 0)) ||
+            !SaveObjectToPrecompiled(static_cast<u8>(image.IsWritten() ? 1 : 0)) ||
             !SaveObjectToPrecompiled(static_cast<u8>(image.IsRead() ? 1 : 0)) ||
-            !SaveObjectToPrecompiled(static_cast<u8>(image.IsWritten() ? 1 : 0))) {
+            !SaveObjectToPrecompiled(image.IsSizeKnown()) || !SaveObjectToPrecompiled(size)) {
             return false;
         }
     }

src/video_core/renderer_opengl/gl_state.cpp

@@ -16,7 +16,6 @@ namespace OpenGL {
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
 OpenGLState OpenGLState::cur_state;
-bool OpenGLState::s_rgb_used;
 
 namespace {
 
@@ -282,8 +281,6 @@ void OpenGLState::ApplySRgb() const {
         return;
     cur_state.framebuffer_srgb.enabled = framebuffer_srgb.enabled;
     if (framebuffer_srgb.enabled) {
-        // Track if sRGB is used
-        s_rgb_used = true;
         glEnable(GL_FRAMEBUFFER_SRGB);
     } else {
         glDisable(GL_FRAMEBUFFER_SRGB);

src/video_core/renderer_opengl/gl_state.h

@@ -175,14 +175,6 @@ public:
         return cur_state;
     }
 
-    static bool GetsRGBUsed() {
-        return s_rgb_used;
-    }
-
-    static void ClearsRGBUsed() {
-        s_rgb_used = false;
-    }
-
     void SetDefaultViewports();
     /// Apply this state as the current OpenGL state
     void Apply();
@@ -253,8 +245,6 @@ public:
 private:
     static OpenGLState cur_state;
 
-    // Workaround for sRGB problems caused by QT not supporting srgb output
-    static bool s_rgb_used;
     struct {
         bool blend_state;
         bool stencil_state;

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -97,6 +97,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format
     {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, ComponentType::UNorm, false},            // RG8U
     {GL_RG8, GL_RG, GL_BYTE, ComponentType::SNorm, false},                     // RG8S
     {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false},   // RG32UI
+    {GL_RGB16F, GL_RGBA16, GL_HALF_FLOAT, ComponentType::Float, false},        // RGBX16F
     {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false},   // R32UI
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false},        // ASTC_2D_8X8
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false},        // ASTC_2D_8X5

src/video_core/renderer_opengl/maxwell_to_gl.h

@@ -145,7 +145,7 @@ inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode,
         case Tegra::Texture::TextureMipmapFilter::None:
             return GL_LINEAR;
         case Tegra::Texture::TextureMipmapFilter::Nearest:
-            return GL_NEAREST_MIPMAP_LINEAR;
+            return GL_LINEAR_MIPMAP_NEAREST;
         case Tegra::Texture::TextureMipmapFilter::Linear:
             return GL_LINEAR_MIPMAP_LINEAR;
         }
@@ -157,7 +157,7 @@ inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode,
         case Tegra::Texture::TextureMipmapFilter::Nearest:
             return GL_NEAREST_MIPMAP_NEAREST;
         case Tegra::Texture::TextureMipmapFilter::Linear:
-            return GL_LINEAR_MIPMAP_NEAREST;
+            return GL_NEAREST_MIPMAP_LINEAR;
         }
     }
     }

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -264,7 +264,6 @@ void RendererOpenGL::CreateRasterizer() {
     if (rasterizer) {
         return;
     }
-    OpenGLState::ClearsRGBUsed();
     rasterizer = std::make_unique<RasterizerOpenGL>(system, emu_window, screen_info);
 }
 
@@ -343,9 +342,7 @@ void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x,
     }};
 
     state.textures[0] = screen_info.display_texture;
-    // Workaround brigthness problems in SMO by enabling sRGB in the final output
-    // if it has been used in the frame. Needed because of this bug in QT: QTBUG-50987
-    state.framebuffer_srgb.enabled = OpenGLState::GetsRGBUsed();
+    state.framebuffer_srgb.enabled = screen_info.display_srgb;
     state.AllDirty();
     state.Apply();
     glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), vertices.data());
@@ -355,8 +352,6 @@ void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x,
     state.textures[0] = 0;
     state.AllDirty();
     state.Apply();
-    // Clear sRGB state for the next frame
-    OpenGLState::ClearsRGBUsed();
 }
 
 /**
@@ -406,8 +401,8 @@ void RendererOpenGL::CaptureScreenshot() {
     GLuint renderbuffer;
     glGenRenderbuffers(1, &renderbuffer);
     glBindRenderbuffer(GL_RENDERBUFFER, renderbuffer);
-    glRenderbufferStorage(GL_RENDERBUFFER, state.GetsRGBUsed() ? GL_SRGB8 : GL_RGB8, layout.width,
-                          layout.height);
+    glRenderbufferStorage(GL_RENDERBUFFER, screen_info.display_srgb ? GL_SRGB8 : GL_RGB8,
+                          layout.width, layout.height);
     glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, renderbuffer);
 
     DrawScreen(layout);

src/video_core/renderer_opengl/renderer_opengl.h

@@ -38,7 +38,8 @@ struct TextureInfo {
 
 /// Structure used for storing information about the display target for the Switch screen
 struct ScreenInfo {
-    GLuint display_texture;
+    GLuint display_texture{};
+    bool display_srgb{};
     const Common::Rectangle<float> display_texcoords{0.0f, 0.0f, 1.0f, 1.0f};
     TextureInfo texture;
 };

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -143,6 +143,7 @@ static constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex
     {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG8U
     {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG8S
     {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG32UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBX16F
     {vk::Format::eUndefined, ComponentType::Invalid, false},           // R32UI
     {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_8X8
     {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_8X5

src/video_core/renderer_vulkan/vk_device.cpp

@@ -2,9 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <map>
+#include <bitset>
 #include <optional>
 #include <set>
+#include <string_view>
 #include <vector>
 #include "common/assert.h"
 #include "video_core/renderer_vulkan/declarations.h"
@@ -12,13 +13,32 @@
 
 namespace Vulkan {
 
+namespace {
+
+template <typename T>
+void SetNext(void**& next, T& data) {
+    *next = &data;
+    next = &data.pNext;
+}
+
+template <typename T>
+T GetFeatures(vk::PhysicalDevice physical, vk::DispatchLoaderDynamic dldi) {
+    vk::PhysicalDeviceFeatures2 features;
+    T extension_features;
+    features.pNext = &extension_features;
+    physical.getFeatures2(&features, dldi);
+    return extension_features;
+}
+
+} // Anonymous namespace
+
 namespace Alternatives {
 
-constexpr std::array<vk::Format, 3> Depth24UnormS8Uint = {
-    vk::Format::eD32SfloatS8Uint, vk::Format::eD16UnormS8Uint, {}};
-constexpr std::array<vk::Format, 3> Depth16UnormS8Uint = {
-    vk::Format::eD24UnormS8Uint, vk::Format::eD32SfloatS8Uint, {}};
-constexpr std::array<vk::Format, 2> Astc = {vk::Format::eA8B8G8R8UnormPack32, {}};
+constexpr std::array Depth24UnormS8Uint = {vk::Format::eD32SfloatS8Uint,
+                                           vk::Format::eD16UnormS8Uint, vk::Format{}};
+constexpr std::array Depth16UnormS8Uint = {vk::Format::eD24UnormS8Uint,
+                                           vk::Format::eD32SfloatS8Uint, vk::Format{}};
+constexpr std::array Astc = {vk::Format::eA8B8G8R8UnormPack32, vk::Format{}};
 
 } // namespace Alternatives
 
@@ -58,16 +78,53 @@ VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice phy
 VKDevice::~VKDevice() = default;
 
 bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance) {
-    vk::PhysicalDeviceFeatures device_features;
-    device_features.vertexPipelineStoresAndAtomics = true;
-    device_features.independentBlend = true;
-    device_features.textureCompressionASTC_LDR = is_optimal_astc_supported;
-
     const auto queue_cis = GetDeviceQueueCreateInfos();
-    const std::vector<const char*> extensions = LoadExtensions(dldi);
-    const vk::DeviceCreateInfo device_ci({}, static_cast<u32>(queue_cis.size()), queue_cis.data(),
-                                         0, nullptr, static_cast<u32>(extensions.size()),
-                                         extensions.data(), &device_features);
+    const std::vector extensions = LoadExtensions(dldi);
+
+    vk::PhysicalDeviceFeatures2 features2;
+    void** next = &features2.pNext;
+    auto& features = features2.features;
+    features.vertexPipelineStoresAndAtomics = true;
+    features.independentBlend = true;
+    features.depthClamp = true;
+    features.samplerAnisotropy = true;
+    features.largePoints = true;
+    features.textureCompressionASTC_LDR = is_optimal_astc_supported;
+
+    vk::PhysicalDeviceVertexAttributeDivisorFeaturesEXT vertex_divisor;
+    vertex_divisor.vertexAttributeInstanceRateDivisor = true;
+    vertex_divisor.vertexAttributeInstanceRateZeroDivisor = true;
+    SetNext(next, vertex_divisor);
+
+    vk::PhysicalDeviceFloat16Int8FeaturesKHR float16_int8;
+    if (is_float16_supported) {
+        float16_int8.shaderFloat16 = true;
+        SetNext(next, float16_int8);
+    } else {
+        LOG_INFO(Render_Vulkan, "Device doesn't support float16 natively");
+    }
+
+    vk::PhysicalDeviceUniformBufferStandardLayoutFeaturesKHR std430_layout;
+    if (khr_uniform_buffer_standard_layout) {
+        std430_layout.uniformBufferStandardLayout = true;
+        SetNext(next, std430_layout);
+    } else {
+        LOG_INFO(Render_Vulkan, "Device doesn't support packed UBOs");
+    }
+
+    vk::PhysicalDeviceIndexTypeUint8FeaturesEXT index_type_uint8;
+    if (ext_index_type_uint8) {
+        index_type_uint8.indexTypeUint8 = true;
+        SetNext(next, index_type_uint8);
+    } else {
+        LOG_INFO(Render_Vulkan, "Device doesn't support uint8 indexes");
+    }
+
+    vk::DeviceCreateInfo device_ci({}, static_cast<u32>(queue_cis.size()), queue_cis.data(), 0,
+                                   nullptr, static_cast<u32>(extensions.size()), extensions.data(),
+                                   nullptr);
+    device_ci.pNext = &features2;
+
     vk::Device dummy_logical;
     if (physical.createDevice(&device_ci, nullptr, &dummy_logical, dldi) != vk::Result::eSuccess) {
         LOG_CRITICAL(Render_Vulkan, "Logical device failed to be created!");
@@ -78,6 +135,17 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan
     logical = UniqueDevice(
         dummy_logical, vk::ObjectDestroy<vk::NoParent, vk::DispatchLoaderDynamic>(nullptr, dld));
 
+    if (khr_driver_properties) {
+        vk::PhysicalDeviceDriverPropertiesKHR driver;
+        vk::PhysicalDeviceProperties2 properties;
+        properties.pNext = &driver;
+        physical.getProperties2(&properties, dld);
+        driver_id = driver.driverID;
+        LOG_INFO(Render_Vulkan, "Driver: {} {}", driver.driverName, driver.driverInfo);
+    } else {
+        LOG_INFO(Render_Vulkan, "Driver: Unknown");
+    }
+
     graphics_queue = logical->getQueue(graphics_family, 0, dld);
     present_queue = logical->getQueue(present_family, 0, dld);
     return true;
@@ -92,20 +160,19 @@ vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format,
     // The wanted format is not supported by hardware, search for alternatives
     const vk::Format* alternatives = GetFormatAlternatives(wanted_format);
     if (alternatives == nullptr) {
-        LOG_CRITICAL(Render_Vulkan,
-                     "Format={} with usage={} and type={} has no defined alternatives and host "
-                     "hardware does not support it",
-                     vk::to_string(wanted_format), vk::to_string(wanted_usage),
-                     static_cast<u32>(format_type));
-        UNREACHABLE();
+        UNREACHABLE_MSG("Format={} with usage={} and type={} has no defined alternatives and host "
+                        "hardware does not support it",
+                        vk::to_string(wanted_format), vk::to_string(wanted_usage),
+                        static_cast<u32>(format_type));
         return wanted_format;
     }
 
     std::size_t i = 0;
     for (vk::Format alternative = alternatives[0]; alternative != vk::Format{};
          alternative = alternatives[++i]) {
-        if (!IsFormatSupported(alternative, wanted_usage, format_type))
+        if (!IsFormatSupported(alternative, wanted_usage, format_type)) {
             continue;
+        }
         LOG_WARNING(Render_Vulkan,
                     "Emulating format={} with alternative format={} with usage={} and type={}",
                     static_cast<u32>(wanted_format), static_cast<u32>(alternative),
@@ -114,12 +181,10 @@ vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format,
     }
 
     // No alternatives found, panic
-    LOG_CRITICAL(Render_Vulkan,
-                 "Format={} with usage={} and type={} is not supported by the host hardware and "
-                 "doesn't support any of the alternatives",
-                 static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage),
-                 static_cast<u32>(format_type));
-    UNREACHABLE();
+    UNREACHABLE_MSG("Format={} with usage={} and type={} is not supported by the host hardware and "
+                    "doesn't support any of the alternatives",
+                    static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage),
+                    static_cast<u32>(format_type));
     return wanted_format;
 }
 
@@ -132,7 +197,7 @@ bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features
         vk::FormatFeatureFlagBits::eSampledImage | vk::FormatFeatureFlagBits::eBlitSrc |
         vk::FormatFeatureFlagBits::eBlitDst | vk::FormatFeatureFlagBits::eTransferSrc |
         vk::FormatFeatureFlagBits::eTransferDst};
-    constexpr std::array<vk::Format, 9> astc_formats = {
+    constexpr std::array astc_formats = {
         vk::Format::eAstc4x4UnormBlock, vk::Format::eAstc4x4SrgbBlock,
         vk::Format::eAstc8x8SrgbBlock,  vk::Format::eAstc8x6SrgbBlock,
         vk::Format::eAstc5x4SrgbBlock,  vk::Format::eAstc5x5UnormBlock,
@@ -151,76 +216,120 @@ bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlag
                                  FormatType format_type) const {
     const auto it = format_properties.find(wanted_format);
     if (it == format_properties.end()) {
-        LOG_CRITICAL(Render_Vulkan, "Unimplemented format query={}", vk::to_string(wanted_format));
-        UNREACHABLE();
+        UNIMPLEMENTED_MSG("Unimplemented format query={}", vk::to_string(wanted_format));
         return true;
     }
-    const vk::FormatFeatureFlags supported_usage = GetFormatFeatures(it->second, format_type);
+    const auto supported_usage = GetFormatFeatures(it->second, format_type);
     return (supported_usage & wanted_usage) == wanted_usage;
 }
 
 bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
                           vk::SurfaceKHR surface) {
-    bool has_swapchain{};
+    LOG_INFO(Render_Vulkan, "{}", physical.getProperties(dldi).deviceName);
+    bool is_suitable = true;
+
+    constexpr std::array required_extensions = {VK_KHR_SWAPCHAIN_EXTENSION_NAME,
+                                                VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME};
+    std::bitset<required_extensions.size()> available_extensions{};
+
     for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
-        has_swapchain |= prop.extensionName == std::string(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
+        for (std::size_t i = 0; i < required_extensions.size(); ++i) {
+            if (available_extensions[i]) {
+                continue;
+            }
+            available_extensions[i] =
+                required_extensions[i] == std::string_view{prop.extensionName};
+        }
     }
-    if (!has_swapchain) {
-        // The device doesn't support creating swapchains.
-        return false;
+    if (!available_extensions.all()) {
+        for (std::size_t i = 0; i < required_extensions.size(); ++i) {
+            if (available_extensions[i]) {
+                continue;
+            }
+            LOG_INFO(Render_Vulkan, "Missing required extension: {}", required_extensions[i]);
+            is_suitable = false;
+        }
     }
 
     bool has_graphics{}, has_present{};
     const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
     for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
         const auto& family = queue_family_properties[i];
-        if (family.queueCount == 0)
+        if (family.queueCount == 0) {
             continue;
-
+        }
         has_graphics |=
             (family.queueFlags & vk::QueueFlagBits::eGraphics) != static_cast<vk::QueueFlagBits>(0);
         has_present |= physical.getSurfaceSupportKHR(i, surface, dldi) != 0;
     }
     if (!has_graphics || !has_present) {
-        // The device doesn't have a graphics and present queue.
-        return false;
+        LOG_INFO(Render_Vulkan, "Device lacks a graphics and present queue");
+        is_suitable = false;
     }
 
     // TODO(Rodrigo): Check if the device matches all requeriments.
     const auto properties{physical.getProperties(dldi)};
-    const auto limits{properties.limits};
-    if (limits.maxUniformBufferRange < 65536) {
-        return false;
+    const auto& limits{properties.limits};
+
+    constexpr u32 required_ubo_size = 65536;
+    if (limits.maxUniformBufferRange < required_ubo_size) {
+        LOG_INFO(Render_Vulkan, "Device UBO size {} is too small, {} is required)",
+                 limits.maxUniformBufferRange, required_ubo_size);
+        is_suitable = false;
     }
 
-    const vk::PhysicalDeviceFeatures features{physical.getFeatures(dldi)};
-    if (!features.vertexPipelineStoresAndAtomics || !features.independentBlend) {
-        return false;
+    const auto features{physical.getFeatures(dldi)};
+    const std::array feature_report = {
+        std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"),
+        std::make_pair(features.independentBlend, "independentBlend"),
+        std::make_pair(features.depthClamp, "depthClamp"),
+        std::make_pair(features.samplerAnisotropy, "samplerAnisotropy"),
+        std::make_pair(features.largePoints, "largePoints"),
+    };
+    for (const auto& [supported, name] : feature_report) {
+        if (supported) {
+            continue;
+        }
+        LOG_INFO(Render_Vulkan, "Missing required feature: {}", name);
+        is_suitable = false;
     }
 
-    // Device is suitable.
-    return true;
+    return is_suitable;
 }
 
 std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynamic& dldi) {
     std::vector<const char*> extensions;
-    extensions.reserve(2);
+    extensions.reserve(7);
     extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
+    extensions.push_back(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME);
 
     const auto Test = [&](const vk::ExtensionProperties& extension,
                           std::optional<std::reference_wrapper<bool>> status, const char* name,
-                          u32 revision) {
-        if (extension.extensionName != std::string(name)) {
+                          bool push) {
+        if (extension.extensionName != std::string_view(name)) {
             return;
         }
-        extensions.push_back(name);
+        if (push) {
+            extensions.push_back(name);
+        }
         if (status) {
             status->get() = true;
         }
     };
 
+    bool khr_shader_float16_int8{};
     for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
-        Test(extension, ext_scalar_block_layout, VK_EXT_SCALAR_BLOCK_LAYOUT_EXTENSION_NAME, 1);
+        Test(extension, khr_uniform_buffer_standard_layout,
+             VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true);
+        Test(extension, ext_index_type_uint8, VK_EXT_INDEX_TYPE_UINT8_EXTENSION_NAME, true);
+        Test(extension, khr_driver_properties, VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME, true);
+        Test(extension, khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, false);
+    }
+
+    if (khr_shader_float16_int8) {
+        is_float16_supported =
+            GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dldi).shaderFloat16;
+        extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME);
     }
 
     return extensions;
@@ -250,9 +359,10 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
 }
 
 void VKDevice::SetupProperties(const vk::DispatchLoaderDynamic& dldi) {
-    const vk::PhysicalDeviceProperties props = physical.getProperties(dldi);
+    const auto props = physical.getProperties(dldi);
     device_type = props.deviceType;
     uniform_buffer_alignment = static_cast<u64>(props.limits.minUniformBufferOffsetAlignment);
+    storage_buffer_alignment = static_cast<u64>(props.limits.minStorageBufferOffsetAlignment);
     max_storage_buffer_range = static_cast<u64>(props.limits.maxStorageBufferRange);
 }
 
@@ -273,42 +383,53 @@ std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() con
     return queue_cis;
 }
 
-std::map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
+std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
     const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) {
-    static constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32,
-                                        vk::Format::eB5G6R5UnormPack16,
-                                        vk::Format::eA2B10G10R10UnormPack32,
-                                        vk::Format::eR32G32B32A32Sfloat,
-                                        vk::Format::eR16G16Unorm,
-                                        vk::Format::eR16G16Snorm,
-                                        vk::Format::eR8G8B8A8Srgb,
-                                        vk::Format::eR8Unorm,
-                                        vk::Format::eB10G11R11UfloatPack32,
-                                        vk::Format::eR32Sfloat,
-                                        vk::Format::eR16Sfloat,
-                                        vk::Format::eR16G16B16A16Sfloat,
-                                        vk::Format::eD32Sfloat,
-                                        vk::Format::eD16Unorm,
-                                        vk::Format::eD16UnormS8Uint,
-                                        vk::Format::eD24UnormS8Uint,
-                                        vk::Format::eD32SfloatS8Uint,
-                                        vk::Format::eBc1RgbaUnormBlock,
-                                        vk::Format::eBc2UnormBlock,
-                                        vk::Format::eBc3UnormBlock,
-                                        vk::Format::eBc4UnormBlock,
-                                        vk::Format::eBc5UnormBlock,
-                                        vk::Format::eBc5SnormBlock,
-                                        vk::Format::eBc7UnormBlock,
-                                        vk::Format::eAstc4x4UnormBlock,
-                                        vk::Format::eAstc4x4SrgbBlock,
-                                        vk::Format::eAstc8x8SrgbBlock,
-                                        vk::Format::eAstc8x6SrgbBlock,
-                                        vk::Format::eAstc5x4SrgbBlock,
-                                        vk::Format::eAstc5x5UnormBlock,
-                                        vk::Format::eAstc5x5SrgbBlock,
-                                        vk::Format::eAstc10x8UnormBlock,
-                                        vk::Format::eAstc10x8SrgbBlock};
-    std::map<vk::Format, vk::FormatProperties> format_properties;
+    constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32,
+                                 vk::Format::eA8B8G8R8SnormPack32,
+                                 vk::Format::eA8B8G8R8SrgbPack32,
+                                 vk::Format::eB5G6R5UnormPack16,
+                                 vk::Format::eA2B10G10R10UnormPack32,
+                                 vk::Format::eR32G32B32A32Sfloat,
+                                 vk::Format::eR16G16B16A16Uint,
+                                 vk::Format::eR16G16Unorm,
+                                 vk::Format::eR16G16Snorm,
+                                 vk::Format::eR16G16Sfloat,
+                                 vk::Format::eR16Unorm,
+                                 vk::Format::eR8G8B8A8Srgb,
+                                 vk::Format::eR8G8Unorm,
+                                 vk::Format::eR8G8Snorm,
+                                 vk::Format::eR8Unorm,
+                                 vk::Format::eB10G11R11UfloatPack32,
+                                 vk::Format::eR32Sfloat,
+                                 vk::Format::eR16Sfloat,
+                                 vk::Format::eR16G16B16A16Sfloat,
+                                 vk::Format::eB8G8R8A8Unorm,
+                                 vk::Format::eD32Sfloat,
+                                 vk::Format::eD16Unorm,
+                                 vk::Format::eD16UnormS8Uint,
+                                 vk::Format::eD24UnormS8Uint,
+                                 vk::Format::eD32SfloatS8Uint,
+                                 vk::Format::eBc1RgbaUnormBlock,
+                                 vk::Format::eBc2UnormBlock,
+                                 vk::Format::eBc3UnormBlock,
+                                 vk::Format::eBc4UnormBlock,
+                                 vk::Format::eBc5UnormBlock,
+                                 vk::Format::eBc5SnormBlock,
+                                 vk::Format::eBc7UnormBlock,
+                                 vk::Format::eBc1RgbaSrgbBlock,
+                                 vk::Format::eBc3SrgbBlock,
+                                 vk::Format::eBc7SrgbBlock,
+                                 vk::Format::eAstc4x4UnormBlock,
+                                 vk::Format::eAstc4x4SrgbBlock,
+                                 vk::Format::eAstc8x8SrgbBlock,
+                                 vk::Format::eAstc8x6SrgbBlock,
+                                 vk::Format::eAstc5x4SrgbBlock,
+                                 vk::Format::eAstc5x5UnormBlock,
+                                 vk::Format::eAstc5x5SrgbBlock,
+                                 vk::Format::eAstc10x8UnormBlock,
+                                 vk::Format::eAstc10x8SrgbBlock};
+    std::unordered_map<vk::Format, vk::FormatProperties> format_properties;
     for (const auto format : formats) {
         format_properties.emplace(format, physical.getFormatProperties(format, dldi));
     }

src/video_core/renderer_vulkan/vk_device.h

@@ -4,7 +4,7 @@
 
 #pragma once
 
-#include <map>
+#include <unordered_map>
 #include <vector>
 #include "common/common_types.h"
 #include "video_core/renderer_vulkan/declarations.h"
@@ -69,16 +69,26 @@ public:
         return present_family;
     }
 
-    /// Returns if the device is integrated with the host CPU.
+    /// Returns true if the device is integrated with the host CPU.
     bool IsIntegrated() const {
         return device_type == vk::PhysicalDeviceType::eIntegratedGpu;
     }
 
+    /// Returns the driver ID.
+    vk::DriverIdKHR GetDriverID() const {
+        return driver_id;
+    }
+
     /// Returns uniform buffer alignment requeriment.
     u64 GetUniformBufferAlignment() const {
         return uniform_buffer_alignment;
     }
 
+    /// Returns storage alignment requeriment.
+    u64 GetStorageBufferAlignment() const {
+        return storage_buffer_alignment;
+    }
+
     /// Returns the maximum range for storage buffers.
     u64 GetMaxStorageBufferRange() const {
         return max_storage_buffer_range;
@@ -89,9 +99,19 @@ public:
         return is_optimal_astc_supported;
     }
 
+    /// Returns true if the device supports float16 natively
+    bool IsFloat16Supported() const {
+        return is_float16_supported;
+    }
+
     /// Returns true if the device supports VK_EXT_scalar_block_layout.
-    bool IsExtScalarBlockLayoutSupported() const {
-        return ext_scalar_block_layout;
+    bool IsKhrUniformBufferStandardLayoutSupported() const {
+        return khr_uniform_buffer_standard_layout;
+    }
+
+    /// Returns true if the device supports VK_EXT_index_type_uint8.
+    bool IsExtIndexTypeUint8Supported() const {
+        return ext_index_type_uint8;
     }
 
     /// Checks if the physical device is suitable.
@@ -123,22 +143,28 @@ private:
                            FormatType format_type) const;
 
     /// Returns the device properties for Vulkan formats.
-    static std::map<vk::Format, vk::FormatProperties> GetFormatProperties(
+    static std::unordered_map<vk::Format, vk::FormatProperties> GetFormatProperties(
         const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical);
 
-    const vk::PhysicalDevice physical;  ///< Physical device.
-    vk::DispatchLoaderDynamic dld;      ///< Device function pointers.
-    UniqueDevice logical;               ///< Logical device.
-    vk::Queue graphics_queue;           ///< Main graphics queue.
-    vk::Queue present_queue;            ///< Main present queue.
-    u32 graphics_family{};              ///< Main graphics queue family index.
-    u32 present_family{};               ///< Main present queue family index.
-    vk::PhysicalDeviceType device_type; ///< Physical device type.
-    u64 uniform_buffer_alignment{};     ///< Uniform buffer alignment requeriment.
-    u64 max_storage_buffer_range{};     ///< Max storage buffer size.
-    bool is_optimal_astc_supported{};   ///< Support for native ASTC.
-    bool ext_scalar_block_layout{};     ///< Support for VK_EXT_scalar_block_layout.
-    std::map<vk::Format, vk::FormatProperties> format_properties; ///< Format properties dictionary.
+    const vk::PhysicalDevice physical;         ///< Physical device.
+    vk::DispatchLoaderDynamic dld;             ///< Device function pointers.
+    UniqueDevice logical;                      ///< Logical device.
+    vk::Queue graphics_queue;                  ///< Main graphics queue.
+    vk::Queue present_queue;                   ///< Main present queue.
+    u32 graphics_family{};                     ///< Main graphics queue family index.
+    u32 present_family{};                      ///< Main present queue family index.
+    vk::PhysicalDeviceType device_type;        ///< Physical device type.
+    vk::DriverIdKHR driver_id{};               ///< Driver ID.
+    u64 uniform_buffer_alignment{};            ///< Uniform buffer alignment requeriment.
+    u64 storage_buffer_alignment{};            ///< Storage buffer alignment requeriment.
+    u64 max_storage_buffer_range{};            ///< Max storage buffer size.
+    bool is_optimal_astc_supported{};          ///< Support for native ASTC.
+    bool is_float16_supported{};               ///< Support for float16 arithmetics.
+    bool khr_uniform_buffer_standard_layout{}; ///< Support for std430 on UBOs.
+    bool ext_index_type_uint8{};               ///< Support for VK_EXT_index_type_uint8.
+    bool khr_driver_properties{};              ///< Support for VK_KHR_driver_properties.
+    std::unordered_map<vk::Format, vk::FormatProperties>
+        format_properties; ///< Format properties dictionary.
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_shader_decompiler.cpp

@@ -370,8 +370,8 @@ private:
         u32 binding = const_buffers_base_binding;
         for (const auto& entry : ir.GetConstantBuffers()) {
             const auto [index, size] = entry;
-            const Id type =
-                device.IsExtScalarBlockLayoutSupported() ? t_cbuf_scalar_ubo : t_cbuf_std140_ubo;
+            const Id type = device.IsKhrUniformBufferStandardLayoutSupported() ? t_cbuf_scalar_ubo
+                                                                               : t_cbuf_std140_ubo;
             const Id id = OpVariable(type, spv::StorageClass::Uniform);
             AddGlobalVariable(Name(id, fmt::format("cbuf_{}", index)));
 
@@ -565,7 +565,7 @@ private:
             const Id buffer_id = constant_buffers.at(cbuf->GetIndex());
 
             Id pointer{};
-            if (device.IsExtScalarBlockLayoutSupported()) {
+            if (device.IsKhrUniformBufferStandardLayoutSupported()) {
                 const Id buffer_offset = Emit(OpShiftRightLogical(
                     t_uint, BitcastTo<Type::Uint>(Visit(offset)), Constant(t_uint, 2u)));
                 pointer = Emit(
@@ -944,6 +944,41 @@ private:
         return {};
     }
 
+    Id AtomicImageAdd(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id AtomicImageMin(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id AtomicImageMax(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id AtomicImageAnd(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id AtomicImageOr(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id AtomicImageXor(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id AtomicImageExchange(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
     Id Branch(Operation operation) {
         const auto target = std::get_if<ImmediateNode>(&*operation[0]);
         UNIMPLEMENTED_IF(!target);
@@ -1092,6 +1127,46 @@ private:
         return {};
     }
 
+    Id ShuffleIndexed(Operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id ShuffleUp(Operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id ShuffleDown(Operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id ShuffleButterfly(Operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id InRangeShuffleIndexed(Operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id InRangeShuffleUp(Operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id InRangeShuffleDown(Operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+    Id InRangeShuffleButterfly(Operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
     Id DeclareBuiltIn(spv::BuiltIn builtin, spv::StorageClass storage, Id type,
                       const std::string& name) {
         const Id id = OpVariable(type, storage);
@@ -1366,6 +1441,13 @@ private:
         &SPIRVDecompiler::TexelFetch,
 
         &SPIRVDecompiler::ImageStore,
+        &SPIRVDecompiler::AtomicImageAdd,
+        &SPIRVDecompiler::AtomicImageMin,
+        &SPIRVDecompiler::AtomicImageMax,
+        &SPIRVDecompiler::AtomicImageAnd,
+        &SPIRVDecompiler::AtomicImageOr,
+        &SPIRVDecompiler::AtomicImageXor,
+        &SPIRVDecompiler::AtomicImageExchange,
 
         &SPIRVDecompiler::Branch,
         &SPIRVDecompiler::BranchIndirect,
@@ -1389,6 +1471,16 @@ private:
         &SPIRVDecompiler::VoteAll,
         &SPIRVDecompiler::VoteAny,
         &SPIRVDecompiler::VoteEqual,
+
+        &SPIRVDecompiler::ShuffleIndexed,
+        &SPIRVDecompiler::ShuffleUp,
+        &SPIRVDecompiler::ShuffleDown,
+        &SPIRVDecompiler::ShuffleButterfly,
+
+        &SPIRVDecompiler::InRangeShuffleIndexed,
+        &SPIRVDecompiler::InRangeShuffleUp,
+        &SPIRVDecompiler::InRangeShuffleDown,
+        &SPIRVDecompiler::InRangeShuffleButterfly,
     };
     static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
 

src/video_core/shader/decode/arithmetic_integer.cpp

@@ -138,6 +138,35 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
         SetRegister(bb, instr.gpr0, value);
         break;
     }
+    case OpCode::Id::ICMP_CR:
+    case OpCode::Id::ICMP_R:
+    case OpCode::Id::ICMP_RC:
+    case OpCode::Id::ICMP_IMM: {
+        const Node zero = Immediate(0);
+
+        const auto [op_b, test] = [&]() -> std::pair<Node, Node> {
+            switch (opcode->get().GetId()) {
+            case OpCode::Id::ICMP_CR:
+                return {GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset),
+                        GetRegister(instr.gpr39)};
+            case OpCode::Id::ICMP_R:
+                return {GetRegister(instr.gpr20), GetRegister(instr.gpr39)};
+            case OpCode::Id::ICMP_RC:
+                return {GetRegister(instr.gpr39),
+                        GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset)};
+            case OpCode::Id::ICMP_IMM:
+                return {Immediate(instr.alu.GetSignedImm20_20()), GetRegister(instr.gpr39)};
+            default:
+                UNREACHABLE();
+                return {zero, zero};
+            }
+        }();
+        const Node op_a = GetRegister(instr.gpr8);
+        const Node comparison =
+            GetPredicateComparisonInteger(instr.icmp.cond, instr.icmp.is_signed != 0, test, zero);
+        SetRegister(bb, instr.gpr0, Operation(OperationCode::Select, comparison, op_a, op_b));
+        break;
+    }
     case OpCode::Id::LOP_C:
     case OpCode::Id::LOP_R:
     case OpCode::Id::LOP_IMM: {

src/video_core/shader/decode/image.cpp

@@ -44,7 +44,6 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
     switch (opcode->get().GetId()) {
     case OpCode::Id::SUST: {
         UNIMPLEMENTED_IF(instr.sust.mode != Tegra::Shader::SurfaceDataMode::P);
-        UNIMPLEMENTED_IF(instr.sust.image_type == Tegra::Shader::ImageType::TextureBuffer);
         UNIMPLEMENTED_IF(instr.sust.out_of_bounds_store != Tegra::Shader::OutOfBoundsStore::Ignore);
         UNIMPLEMENTED_IF(instr.sust.component_mask_selector != 0xf); // Ensure we have an RGBA store
 
@@ -66,8 +65,46 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
         image.MarkWrite();
 
         MetaImage meta{image, values};
-        const Node store{Operation(OperationCode::ImageStore, meta, std::move(coords))};
-        bb.push_back(store);
+        bb.push_back(Operation(OperationCode::ImageStore, meta, std::move(coords)));
+        break;
+    }
+    case OpCode::Id::SUATOM: {
+        UNIMPLEMENTED_IF(instr.suatom_d.is_ba != 0);
+
+        Node value = GetRegister(instr.gpr0);
+
+        std::vector<Node> coords;
+        const std::size_t num_coords{GetImageTypeNumCoordinates(instr.sust.image_type)};
+        for (std::size_t i = 0; i < num_coords; ++i) {
+            coords.push_back(GetRegister(instr.gpr8.Value() + i));
+        }
+
+        const OperationCode operation_code = [instr] {
+            switch (instr.suatom_d.operation) {
+            case Tegra::Shader::ImageAtomicOperation::Add:
+                return OperationCode::AtomicImageAdd;
+            case Tegra::Shader::ImageAtomicOperation::Min:
+                return OperationCode::AtomicImageMin;
+            case Tegra::Shader::ImageAtomicOperation::Max:
+                return OperationCode::AtomicImageMax;
+            case Tegra::Shader::ImageAtomicOperation::And:
+                return OperationCode::AtomicImageAnd;
+            case Tegra::Shader::ImageAtomicOperation::Or:
+                return OperationCode::AtomicImageOr;
+            case Tegra::Shader::ImageAtomicOperation::Xor:
+                return OperationCode::AtomicImageXor;
+            case Tegra::Shader::ImageAtomicOperation::Exch:
+                return OperationCode::AtomicImageExchange;
+            default:
+                UNIMPLEMENTED_MSG("Unimplemented operation={}",
+                                  static_cast<u32>(instr.suatom_d.operation.Value()));
+                return OperationCode::AtomicImageAdd;
+            }
+        }();
+
+        const auto& image{GetImage(instr.image, instr.suatom_d.image_type, instr.suatom_d.size)};
+        MetaImage meta{image, {std::move(value)}};
+        SetRegister(bb, instr.gpr0, Operation(operation_code, meta, std::move(coords)));
         break;
     }
     default:
@@ -77,38 +114,51 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
     return pc;
 }
 
-Image& ShaderIR::GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type) {
-    const auto offset{static_cast<u64>(image.index.Value())};
-
-    // If this image has already been used, return the existing mapping.
-    const auto it = used_images.find(offset);
-    if (it != used_images.end()) {
-        ASSERT(it->second.GetType() == type);
-        return it->second;
+Image& ShaderIR::GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type,
+                          std::optional<Tegra::Shader::ImageAtomicSize> size) {
+    const auto offset{static_cast<std::size_t>(image.index.Value())};
+    if (const auto image = TryUseExistingImage(offset, type, size)) {
+        return *image;
     }
 
-    // Otherwise create a new mapping for this image.
     const std::size_t next_index{used_images.size()};
-    return used_images.emplace(offset, Image{offset, next_index, type}).first->second;
+    return used_images.emplace(offset, Image{offset, next_index, type, size}).first->second;
 }
 
-Image& ShaderIR::GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type) {
+Image& ShaderIR::GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type,
+                                  std::optional<Tegra::Shader::ImageAtomicSize> size) {
     const Node image_register{GetRegister(reg)};
     const auto [base_image, cbuf_index, cbuf_offset]{
         TrackCbuf(image_register, global_code, static_cast<s64>(global_code.size()))};
     const auto cbuf_key{(static_cast<u64>(cbuf_index) << 32) | static_cast<u64>(cbuf_offset)};
 
-    // If this image has already been used, return the existing mapping.
-    const auto it = used_images.find(cbuf_key);
-    if (it != used_images.end()) {
-        ASSERT(it->second.GetType() == type);
-        return it->second;
+    if (const auto image = TryUseExistingImage(cbuf_key, type, size)) {
+        return *image;
     }
 
-    // Otherwise create a new mapping for this image.
     const std::size_t next_index{used_images.size()};
-    return used_images.emplace(cbuf_key, Image{cbuf_index, cbuf_offset, next_index, type})
+    return used_images.emplace(cbuf_key, Image{cbuf_index, cbuf_offset, next_index, type, size})
         .first->second;
 }
 
+Image* ShaderIR::TryUseExistingImage(u64 offset, Tegra::Shader::ImageType type,
+                                     std::optional<Tegra::Shader::ImageAtomicSize> size) {
+    auto it = used_images.find(offset);
+    if (it == used_images.end()) {
+        return nullptr;
+    }
+    auto& image = it->second;
+    ASSERT(image.GetType() == type);
+
+    if (size) {
+        // We know the size, if it's known it has to be the same as before, otherwise we can set it.
+        if (image.IsSizeKnown()) {
+            ASSERT(image.GetSize() == size);
+        } else {
+            image.SetSize(*size);
+        }
+    }
+    return &image;
+}
+
 } // namespace VideoCommon::Shader

src/video_core/shader/decode/memory.cpp

@@ -35,7 +35,7 @@ u32 GetUniformTypeElementsCount(Tegra::Shader::UniformType uniform_type) {
         return 1;
     }
 }
-} // namespace
+} // Anonymous namespace
 
 u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
@@ -106,16 +106,17 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         }
         break;
     }
-    case OpCode::Id::LD_L: {
-        LOG_DEBUG(HW_GPU, "LD_L cache management mode: {}",
-                  static_cast<u64>(instr.ld_l.unknown.Value()));
-
-        const auto GetLmem = [&](s32 offset) {
+    case OpCode::Id::LD_L:
+        LOG_DEBUG(HW_GPU, "LD_L cache management mode: {}", static_cast<u64>(instr.ld_l.unknown));
+        [[fallthrough]];
+    case OpCode::Id::LD_S: {
+        const auto GetMemory = [&](s32 offset) {
             ASSERT(offset % 4 == 0);
             const Node immediate_offset = Immediate(static_cast<s32>(instr.smem_imm) + offset);
             const Node address = Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8),
                                            immediate_offset);
-            return GetLocalMemory(address);
+            return opcode->get().GetId() == OpCode::Id::LD_S ? GetSharedMemory(address)
+                                                             : GetLocalMemory(address);
         };
 
         switch (instr.ldst_sl.type.Value()) {
@@ -135,14 +136,16 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
                     return 0;
                 }
             }();
-            for (u32 i = 0; i < count; ++i)
-                SetTemporary(bb, i, GetLmem(i * 4));
-            for (u32 i = 0; i < count; ++i)
+            for (u32 i = 0; i < count; ++i) {
+                SetTemporary(bb, i, GetMemory(i * 4));
+            }
+            for (u32 i = 0; i < count; ++i) {
                 SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i));
+            }
             break;
         }
         default:
-            UNIMPLEMENTED_MSG("LD_L Unhandled type: {}",
+            UNIMPLEMENTED_MSG("{} Unhandled type: {}", opcode->get().GetName(),
                               static_cast<u32>(instr.ldst_sl.type.Value()));
         }
         break;
@@ -209,27 +212,34 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
 
         break;
     }
-    case OpCode::Id::ST_L: {
+    case OpCode::Id::ST_L:
         LOG_DEBUG(HW_GPU, "ST_L cache management mode: {}",
                   static_cast<u64>(instr.st_l.cache_management.Value()));
-
-        const auto GetLmemAddr = [&](s32 offset) {
+        [[fallthrough]];
+    case OpCode::Id::ST_S: {
+        const auto GetAddress = [&](s32 offset) {
             ASSERT(offset % 4 == 0);
             const Node immediate = Immediate(static_cast<s32>(instr.smem_imm) + offset);
             return Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8), immediate);
         };
 
+        const auto set_memory = opcode->get().GetId() == OpCode::Id::ST_L
+                                    ? &ShaderIR::SetLocalMemory
+                                    : &ShaderIR::SetSharedMemory;
+
         switch (instr.ldst_sl.type.Value()) {
         case Tegra::Shader::StoreType::Bits128:
-            SetLocalMemory(bb, GetLmemAddr(12), GetRegister(instr.gpr0.Value() + 3));
-            SetLocalMemory(bb, GetLmemAddr(8), GetRegister(instr.gpr0.Value() + 2));
+            (this->*set_memory)(bb, GetAddress(12), GetRegister(instr.gpr0.Value() + 3));
+            (this->*set_memory)(bb, GetAddress(8), GetRegister(instr.gpr0.Value() + 2));
+            [[fallthrough]];
         case Tegra::Shader::StoreType::Bits64:
-            SetLocalMemory(bb, GetLmemAddr(4), GetRegister(instr.gpr0.Value() + 1));
+            (this->*set_memory)(bb, GetAddress(4), GetRegister(instr.gpr0.Value() + 1));
+            [[fallthrough]];
         case Tegra::Shader::StoreType::Bits32:
-            SetLocalMemory(bb, GetLmemAddr(0), GetRegister(instr.gpr0));
+            (this->*set_memory)(bb, GetAddress(0), GetRegister(instr.gpr0));
             break;
         default:
-            UNIMPLEMENTED_MSG("ST_L Unhandled type: {}",
+            UNIMPLEMENTED_MSG("{} unhandled type: {}", opcode->get().GetName(),
                               static_cast<u32>(instr.ldst_sl.type.Value()));
         }
         break;

src/video_core/shader/decode/warp.cpp

@@ -13,6 +13,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::Pred;
+using Tegra::Shader::ShuffleOperation;
 using Tegra::Shader::VoteOperation;
 
 namespace {
@@ -44,6 +45,52 @@ u32 ShaderIR::DecodeWarp(NodeBlock& bb, u32 pc) {
         SetPredicate(bb, instr.vote.dest_pred, vote);
         break;
     }
+    case OpCode::Id::SHFL: {
+        Node mask = instr.shfl.is_mask_imm ? Immediate(static_cast<u32>(instr.shfl.mask_imm))
+                                           : GetRegister(instr.gpr39);
+        Node width = [&] {
+            // Convert the obscure SHFL mask back into GL_NV_shader_thread_shuffle's width. This has
+            // been done reversing Nvidia's math. It won't work on all cases due to SHFL having
+            // different parameters that don't properly map to GLSL's interface, but it should work
+            // for cases emitted by Nvidia's compiler.
+            if (instr.shfl.operation == ShuffleOperation::Up) {
+                return Operation(
+                    OperationCode::ILogicalShiftRight,
+                    Operation(OperationCode::IAdd, std::move(mask), Immediate(-0x2000)),
+                    Immediate(8));
+            } else {
+                return Operation(OperationCode::ILogicalShiftRight,
+                                 Operation(OperationCode::IAdd, Immediate(0x201F),
+                                           Operation(OperationCode::INegate, std::move(mask))),
+                                 Immediate(8));
+            }
+        }();
+
+        const auto [operation, in_range] = [instr]() -> std::pair<OperationCode, OperationCode> {
+            switch (instr.shfl.operation) {
+            case ShuffleOperation::Idx:
+                return {OperationCode::ShuffleIndexed, OperationCode::InRangeShuffleIndexed};
+            case ShuffleOperation::Up:
+                return {OperationCode::ShuffleUp, OperationCode::InRangeShuffleUp};
+            case ShuffleOperation::Down:
+                return {OperationCode::ShuffleDown, OperationCode::InRangeShuffleDown};
+            case ShuffleOperation::Bfly:
+                return {OperationCode::ShuffleButterfly, OperationCode::InRangeShuffleButterfly};
+            }
+            UNREACHABLE_MSG("Invalid SHFL operation: {}",
+                            static_cast<u64>(instr.shfl.operation.Value()));
+            return {};
+        }();
+
+        // Setting the predicate before the register is intentional to avoid overwriting.
+        Node index = instr.shfl.is_index_imm ? Immediate(static_cast<u32>(instr.shfl.index_imm))
+                                             : GetRegister(instr.gpr20);
+        SetPredicate(bb, instr.shfl.pred48, Operation(in_range, index, width));
+        SetRegister(
+            bb, instr.gpr0,
+            Operation(operation, GetRegister(instr.gpr8), std::move(index), std::move(width)));
+        break;
+    }
     default:
         UNIMPLEMENTED_MSG("Unhandled warp instruction: {}", opcode->get().GetName());
         break;

src/video_core/shader/node.h

@@ -7,6 +7,7 @@
 #include <array>
 #include <cstddef>
 #include <memory>
+#include <optional>
 #include <string>
 #include <tuple>
 #include <utility>
@@ -148,7 +149,14 @@ enum class OperationCode {
     TextureQueryLod,        /// (MetaTexture, float[N] coords) -> float4
     TexelFetch,             /// (MetaTexture, int[N], int) -> float4
 
-    ImageStore, /// (MetaImage, float[N] coords) -> void
+    ImageStore,          /// (MetaImage, int[N] values) -> void
+    AtomicImageAdd,      /// (MetaImage, int[N] coords) -> void
+    AtomicImageMin,      /// (MetaImage, int[N] coords) -> void
+    AtomicImageMax,      /// (MetaImage, int[N] coords) -> void
+    AtomicImageAnd,      /// (MetaImage, int[N] coords) -> void
+    AtomicImageOr,       /// (MetaImage, int[N] coords) -> void
+    AtomicImageXor,      /// (MetaImage, int[N] coords) -> void
+    AtomicImageExchange, /// (MetaImage, int[N] coords) -> void
 
     Branch,         /// (uint branch_target) -> void
     BranchIndirect, /// (uint branch_target) -> void
@@ -173,6 +181,16 @@ enum class OperationCode {
     VoteAny,      /// (bool) -> bool
     VoteEqual,    /// (bool) -> bool
 
+    ShuffleIndexed,   /// (uint value, uint index, uint width) -> uint
+    ShuffleUp,        /// (uint value, uint index, uint width) -> uint
+    ShuffleDown,      /// (uint value, uint index, uint width) -> uint
+    ShuffleButterfly, /// (uint value, uint index, uint width) -> uint
+
+    InRangeShuffleIndexed,   /// (uint index, uint width) -> bool
+    InRangeShuffleUp,        /// (uint index, uint width) -> bool
+    InRangeShuffleDown,      /// (uint index, uint width) -> bool
+    InRangeShuffleButterfly, /// (uint index, uint width) -> bool
+
     Amount,
 };
 
@@ -198,12 +216,13 @@ class PredicateNode;
 class AbufNode;
 class CbufNode;
 class LmemNode;
+class SmemNode;
 class GmemNode;
 class CommentNode;
 
 using NodeData =
     std::variant<OperationNode, ConditionalNode, GprNode, ImmediateNode, InternalFlagNode,
-                 PredicateNode, AbufNode, CbufNode, LmemNode, GmemNode, CommentNode>;
+                 PredicateNode, AbufNode, CbufNode, LmemNode, SmemNode, GmemNode, CommentNode>;
 using Node = std::shared_ptr<NodeData>;
 using Node4 = std::array<Node, 4>;
 using NodeBlock = std::vector<Node>;
@@ -275,25 +294,32 @@ private:
 
 class Image final {
 public:
-    constexpr explicit Image(u64 offset, std::size_t index, Tegra::Shader::ImageType type)
-        : offset{offset}, index{index}, type{type}, is_bindless{false} {}
+    constexpr explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type,
+                             std::optional<Tegra::Shader::ImageAtomicSize> size)
+        : offset{offset}, index{index}, type{type}, is_bindless{false}, size{size} {}
 
     constexpr explicit Image(u32 cbuf_index, u32 cbuf_offset, std::size_t index,
-                             Tegra::Shader::ImageType type)
+                             Tegra::Shader::ImageType type,
+                             std::optional<Tegra::Shader::ImageAtomicSize> size)
         : offset{(static_cast<u64>(cbuf_index) << 32) | cbuf_offset}, index{index}, type{type},
-          is_bindless{true} {}
+          is_bindless{true}, size{size} {}
 
     constexpr explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type,
-                             bool is_bindless, bool is_written, bool is_read)
+                             bool is_bindless, bool is_written, bool is_read,
+                             std::optional<Tegra::Shader::ImageAtomicSize> size)
         : offset{offset}, index{index}, type{type}, is_bindless{is_bindless},
-          is_written{is_written}, is_read{is_read} {}
+          is_written{is_written}, is_read{is_read}, size{size} {}
+
+    void MarkWrite() {
+        is_written = true;
+    }
 
     void MarkRead() {
         is_read = true;
     }
 
-    void MarkWrite() {
-        is_written = true;
+    void SetSize(Tegra::Shader::ImageAtomicSize size_) {
+        size = size_;
     }
 
     constexpr std::size_t GetOffset() const {
@@ -312,25 +338,39 @@ public:
         return is_bindless;
     }
 
-    constexpr bool IsRead() const {
-        return is_read;
-    }
-
     constexpr bool IsWritten() const {
         return is_written;
     }
 
+    constexpr bool IsRead() const {
+        return is_read;
+    }
+
     constexpr std::pair<u32, u32> GetBindlessCBuf() const {
         return {static_cast<u32>(offset >> 32), static_cast<u32>(offset)};
     }
 
+    constexpr bool IsSizeKnown() const {
+        return size.has_value();
+    }
+
+    constexpr Tegra::Shader::ImageAtomicSize GetSize() const {
+        return size.value();
+    }
+
+    constexpr bool operator<(const Image& rhs) const {
+        return std::tie(offset, index, type, size, is_bindless) <
+               std::tie(rhs.offset, rhs.index, rhs.type, rhs.size, rhs.is_bindless);
+    }
+
 private:
     u64 offset{};
     std::size_t index{};
     Tegra::Shader::ImageType type{};
     bool is_bindless{};
-    bool is_read{};
     bool is_written{};
+    bool is_read{};
+    std::optional<Tegra::Shader::ImageAtomicSize> size{};
 };
 
 struct GlobalMemoryBase {
@@ -554,6 +594,19 @@ private:
     Node address;
 };
 
+/// Shared memory node
+class SmemNode final {
+public:
+    explicit SmemNode(Node address) : address{std::move(address)} {}
+
+    const Node& GetAddress() const {
+        return address;
+    }
+
+private:
+    Node address;
+};
+
 /// Global memory node
 class GmemNode final {
 public: