vk_shader_decompiler: Implement flow primitives

sirit is a runtime assembler for SPIR-V

.gitmodules

@@ -40,3 +40,9 @@
 [submodule "Vulkan-Headers"]
     path = externals/Vulkan-Headers
     url = https://github.com/KhronosGroup/Vulkan-Headers.git
+[submodule "externals/zstd"]
+    path = externals/zstd
+    url = https://github.com/facebook/zstd
+[submodule "sirit"]
+    path = externals/sirit
+    url = https://github.com/ReinUsesLisp/sirit

CMakeLists.txt

@@ -309,7 +309,7 @@ if (CLANG_FORMAT)
     set(CCOMMENT "Running clang format against all the .h and .cpp files in src/")
     if (WIN32)
         add_custom_target(clang-format
-            COMMAND powershell.exe -Command "Get-ChildItem ${SRCS}/* -Include *.cpp,*.h -Recurse | Foreach {${CLANG_FORMAT} -i $_.fullname}"
+            COMMAND powershell.exe -Command "Get-ChildItem '${SRCS}/*' -Include *.cpp,*.h -Recurse | Foreach {&'${CLANG_FORMAT}' -i $_.fullname}"
             COMMENT ${CCOMMENT})
     elseif(MINGW)
         add_custom_target(clang-format

externals/CMakeLists.txt

@@ -49,6 +49,10 @@ add_subdirectory(open_source_archives EXCLUDE_FROM_ALL)
 add_library(unicorn-headers INTERFACE)
 target_include_directories(unicorn-headers INTERFACE ./unicorn/include)
 
+# Zstandard
+add_subdirectory(zstd/build/cmake EXCLUDE_FROM_ALL)
+target_include_directories(libzstd_static INTERFACE ./zstd/lib)
+
 # SoundTouch
 add_subdirectory(soundtouch)
 
@@ -68,6 +72,11 @@ if (USE_DISCORD_PRESENCE)
     target_include_directories(discord-rpc INTERFACE ./discord-rpc/include)
 endif()
 
+# Sirit
+if (ENABLE_VULKAN)
+    add_subdirectory(sirit)
+endif()
+
 if (ENABLE_WEB_SERVICE)
     # LibreSSL
     set(LIBRESSL_SKIP_INSTALL ON CACHE BOOL "")

src/common/CMakeLists.txt

@@ -125,6 +125,8 @@ add_library(common STATIC
     uint128.h
     vector_math.h
     web_result.h
+    zstd_compression.cpp
+    zstd_compression.h
 )
 
 if(ARCHITECTURE_x86_64)
@@ -138,4 +140,4 @@ endif()
 create_target_directory_groups(common)
 
 target_link_libraries(common PUBLIC Boost::boost fmt microprofile)
-target_link_libraries(common PRIVATE lz4_static)
+target_link_libraries(common PRIVATE lz4_static libzstd_static)

src/common/assert.h

@@ -57,3 +57,21 @@ __declspec(noinline, noreturn)
 
 #define UNIMPLEMENTED_IF(cond) ASSERT_MSG(!(cond), "Unimplemented code!")
 #define UNIMPLEMENTED_IF_MSG(cond, ...) ASSERT_MSG(!(cond), __VA_ARGS__)
+
+// If the assert is ignored, execute _b_
+#define ASSERT_OR_EXECUTE(_a_, _b_)                                                                \
+    do {                                                                                           \
+        ASSERT(_a_);                                                                               \
+        if (!(_a_)) {                                                                              \
+            _b_                                                                                    \
+        }                                                                                          \
+    } while (0)
+
+// If the assert is ignored, execute _b_
+#define ASSERT_OR_EXECUTE_MSG(_a_, _b_, ...)                                                       \
+    do {                                                                                           \
+        ASSERT_MSG(_a_, __VA_ARGS__);                                                              \
+        if (!(_a_)) {                                                                              \
+            _b_                                                                                    \
+        }                                                                                          \
+    } while (0)

src/common/bit_util.h

@@ -32,7 +32,7 @@ inline u32 CountLeadingZeroes32(u32 value) {
     return 32;
 }
 
-inline u64 CountLeadingZeroes64(u64 value) {
+inline u32 CountLeadingZeroes64(u64 value) {
     unsigned long leading_zero = 0;
 
     if (_BitScanReverse64(&leading_zero, value) != 0) {
@@ -47,15 +47,15 @@ inline u32 CountLeadingZeroes32(u32 value) {
         return 32;
     }
 
-    return __builtin_clz(value);
+    return static_cast<u32>(__builtin_clz(value));
 }
 
-inline u64 CountLeadingZeroes64(u64 value) {
+inline u32 CountLeadingZeroes64(u64 value) {
     if (value == 0) {
         return 64;
     }
 
-    return __builtin_clzll(value);
+    return static_cast<u32>(__builtin_clzll(value));
 }
 #endif
 
@@ -70,7 +70,7 @@ inline u32 CountTrailingZeroes32(u32 value) {
     return 32;
 }
 
-inline u64 CountTrailingZeroes64(u64 value) {
+inline u32 CountTrailingZeroes64(u64 value) {
     unsigned long trailing_zero = 0;
 
     if (_BitScanForward64(&trailing_zero, value) != 0) {
@@ -85,15 +85,15 @@ inline u32 CountTrailingZeroes32(u32 value) {
         return 32;
     }
 
-    return __builtin_ctz(value);
+    return static_cast<u32>(__builtin_ctz(value));
 }
 
-inline u64 CountTrailingZeroes64(u64 value) {
+inline u32 CountTrailingZeroes64(u64 value) {
     if (value == 0) {
         return 64;
     }
 
-    return __builtin_ctzll(value);
+    return static_cast<u32>(__builtin_ctzll(value));
 }
 #endif
 

src/common/multi_level_queue.h

@@ -72,7 +72,7 @@ public:
                 u64 prios = mlq.used_priorities;
                 prios &= ~((1ULL << (current_priority + 1)) - 1);
                 if (prios == 0) {
-                    current_priority = mlq.depth();
+                    current_priority = static_cast<u32>(mlq.depth());
                 } else {
                     current_priority = CountTrailingZeroes64(prios);
                     it = GetBeginItForPrio();

src/common/zstd_compression.cpp

@@ -0,0 +1,53 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <zstd.h>
+
+#include "common/assert.h"
+#include "common/zstd_compression.h"
+
+namespace Common::Compression {
+
+std::vector<u8> CompressDataZSTD(const u8* source, std::size_t source_size, s32 compression_level) {
+    compression_level = std::clamp(compression_level, 1, ZSTD_maxCLevel());
+
+    const std::size_t max_compressed_size = ZSTD_compressBound(source_size);
+    std::vector<u8> compressed(max_compressed_size);
+
+    const std::size_t compressed_size =
+        ZSTD_compress(compressed.data(), compressed.size(), source, source_size, compression_level);
+
+    if (ZSTD_isError(compressed_size)) {
+        // Compression failed
+        return {};
+    }
+
+    compressed.resize(compressed_size);
+
+    return compressed;
+}
+
+std::vector<u8> CompressDataZSTDDefault(const u8* source, std::size_t source_size) {
+    return CompressDataZSTD(source, source_size, ZSTD_CLEVEL_DEFAULT);
+}
+
+std::vector<u8> DecompressDataZSTD(const std::vector<u8>& compressed) {
+    const std::size_t decompressed_size =
+        ZSTD_getDecompressedSize(compressed.data(), compressed.size());
+    std::vector<u8> decompressed(decompressed_size);
+
+    const std::size_t uncompressed_result_size = ZSTD_decompress(
+        decompressed.data(), decompressed.size(), compressed.data(), compressed.size());
+
+    if (decompressed_size != uncompressed_result_size || ZSTD_isError(uncompressed_result_size)) {
+        // Decompression failed
+        return {};
+    }
+    return decompressed;
+}
+
+} // namespace Common::Compression

src/common/zstd_compression.h

@@ -0,0 +1,42 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <vector>
+
+#include "common/common_types.h"
+
+namespace Common::Compression {
+
+/**
+ * Compresses a source memory region with Zstandard and returns the compressed data in a vector.
+ *
+ * @param source the uncompressed source memory region.
+ * @param source_size the size in bytes of the uncompressed source memory region.
+ * @param compression_level the used compression level. Should be between 1 and 22.
+ *
+ * @return the compressed data.
+ */
+std::vector<u8> CompressDataZSTD(const u8* source, std::size_t source_size, s32 compression_level);
+
+/**
+ * Compresses a source memory region with Zstandard with the default compression level and returns
+ * the compressed data in a vector.
+ *
+ * @param source the uncompressed source memory region.
+ * @param source_size the size in bytes of the uncompressed source memory region.
+ *
+ * @return the compressed data.
+ */
+std::vector<u8> CompressDataZSTDDefault(const u8* source, std::size_t source_size);
+
+/**
+ * Decompresses a source memory region with Zstandard and returns the uncompressed data in a vector.
+ *
+ * @param compressed the compressed source memory region.
+ *
+ * @return the decompressed data.
+ */
+std::vector<u8> DecompressDataZSTD(const std::vector<u8>& compressed);
+
+} // namespace Common::Compression

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -163,7 +163,6 @@ MICROPROFILE_DEFINE(ARM_Jit_Dynarmic, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64)
 
 void ARM_Dynarmic::Run() {
     MICROPROFILE_SCOPE(ARM_Jit_Dynarmic);
-    ASSERT(Memory::GetCurrentPageTable() == current_page_table);
 
     jit->Run();
 }
@@ -278,7 +277,6 @@ void ARM_Dynarmic::ClearExclusiveState() {
 
 void ARM_Dynarmic::PageTableChanged() {
     jit = MakeJit();
-    current_page_table = Memory::GetCurrentPageTable();
 }
 
 DynarmicExclusiveMonitor::DynarmicExclusiveMonitor(std::size_t core_count) : monitor(core_count) {}

src/core/arm/dynarmic/arm_dynarmic.h

@@ -12,10 +12,6 @@
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
 
-namespace Common {
-struct PageTable;
-}
-
 namespace Core::Timing {
 class CoreTiming;
 }
@@ -69,8 +65,6 @@ private:
     std::size_t core_index;
     Timing::CoreTiming& core_timing;
     DynarmicExclusiveMonitor& exclusive_monitor;
-
-    Common::PageTable* current_page_table = nullptr;
 };
 
 class DynarmicExclusiveMonitor final : public ExclusiveMonitor {

src/core/arm/unicorn/arm_unicorn.cpp

@@ -192,12 +192,13 @@ void ARM_Unicorn::ExecuteInstructions(int num_instructions) {
     CHECKED(uc_emu_start(uc, GetPC(), 1ULL << 63, 0, num_instructions));
     core_timing.AddTicks(num_instructions);
     if (GDBStub::IsServerEnabled()) {
-        if (last_bkpt_hit) {
+        if (last_bkpt_hit && last_bkpt.type == GDBStub::BreakpointType::Execute) {
             uc_reg_write(uc, UC_ARM64_REG_PC, &last_bkpt.address);
         }
+
         Kernel::Thread* thread = Kernel::GetCurrentThread();
         SaveContext(thread->GetContext());
-        if (last_bkpt_hit || GDBStub::GetCpuStepFlag()) {
+        if (last_bkpt_hit || GDBStub::IsMemoryBreak() || GDBStub::GetCpuStepFlag()) {
             last_bkpt_hit = false;
             GDBStub::Break();
             GDBStub::SendTrap(thread, 5);

src/core/arm/unicorn/arm_unicorn.h

@@ -50,7 +50,7 @@ private:
     uc_engine* uc{};
     Timing::CoreTiming& core_timing;
     GDBStub::BreakpointAddress last_bkpt{};
-    bool last_bkpt_hit;
+    bool last_bkpt_hit = false;
 };
 
 } // namespace Core

src/core/core.cpp

@@ -17,6 +17,7 @@
 #include "core/core_timing.h"
 #include "core/cpu_core_manager.h"
 #include "core/file_sys/mode.h"
+#include "core/file_sys/registered_cache.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
 #include "core/gdbstub/gdbstub.h"
@@ -108,6 +109,8 @@ struct System::Impl {
         // Create a default fs if one doesn't already exist.
         if (virtual_filesystem == nullptr)
             virtual_filesystem = std::make_shared<FileSys::RealVfsFilesystem>();
+        if (content_provider == nullptr)
+            content_provider = std::make_unique<FileSys::ContentProviderUnion>();
 
         /// Create default implementations of applets if one is not provided.
         if (profile_selector == nullptr)
@@ -249,6 +252,8 @@ struct System::Impl {
     Kernel::KernelCore kernel;
     /// RealVfsFilesystem instance
     FileSys::VirtualFilesystem virtual_filesystem;
+    /// ContentProviderUnion instance
+    std::unique_ptr<FileSys::ContentProviderUnion> content_provider;
     /// AppLoader used to load the current executing application
     std::unique_ptr<Loader::AppLoader> app_loader;
     std::unique_ptr<VideoCore::RendererBase> renderer;
@@ -488,6 +493,27 @@ const Frontend::SoftwareKeyboardApplet& System::GetSoftwareKeyboard() const {
     return *impl->software_keyboard;
 }
 
+void System::SetContentProvider(std::unique_ptr<FileSys::ContentProviderUnion> provider) {
+    impl->content_provider = std::move(provider);
+}
+
+FileSys::ContentProvider& System::GetContentProvider() {
+    return *impl->content_provider;
+}
+
+const FileSys::ContentProvider& System::GetContentProvider() const {
+    return *impl->content_provider;
+}
+
+void System::RegisterContentProvider(FileSys::ContentProviderUnionSlot slot,
+                                     FileSys::ContentProvider* provider) {
+    impl->content_provider->SetSlot(slot, provider);
+}
+
+void System::ClearContentProvider(FileSys::ContentProviderUnionSlot slot) {
+    impl->content_provider->ClearSlot(slot);
+}
+
 void System::SetWebBrowser(std::unique_ptr<Frontend::WebBrowserApplet> applet) {
     impl->web_browser = std::move(applet);
 }

src/core/core.h

@@ -21,6 +21,9 @@ class WebBrowserApplet;
 
 namespace FileSys {
 class CheatList;
+class ContentProvider;
+class ContentProviderUnion;
+enum class ContentProviderUnionSlot;
 class VfsFilesystem;
 } // namespace FileSys
 
@@ -270,6 +273,17 @@ public:
     Frontend::WebBrowserApplet& GetWebBrowser();
     const Frontend::WebBrowserApplet& GetWebBrowser() const;
 
+    void SetContentProvider(std::unique_ptr<FileSys::ContentProviderUnion> provider);
+
+    FileSys::ContentProvider& GetContentProvider();
+
+    const FileSys::ContentProvider& GetContentProvider() const;
+
+    void RegisterContentProvider(FileSys::ContentProviderUnionSlot slot,
+                                 FileSys::ContentProvider* provider);
+
+    void ClearContentProvider(FileSys::ContentProviderUnionSlot slot);
+
 private:
     System();
 

src/core/crypto/key_manager.cpp

@@ -22,6 +22,7 @@
 #include "common/file_util.h"
 #include "common/hex_util.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/crypto/aes_util.h"
 #include "core/crypto/key_manager.h"
 #include "core/crypto/partition_data_manager.h"
@@ -794,7 +795,7 @@ void KeyManager::DeriveBase() {
 
 void KeyManager::DeriveETicket(PartitionDataManager& data) {
     // ETicket keys
-    const auto es = Service::FileSystem::GetUnionContents().GetEntry(
+    const auto es = Core::System::GetInstance().GetContentProvider().GetEntry(
         0x0100000000000033, FileSys::ContentRecordType::Program);
 
     if (es == nullptr)

src/core/file_sys/patch_manager.cpp

@@ -10,6 +10,7 @@
 #include "common/file_util.h"
 #include "common/hex_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/ips_layer.h"
@@ -69,7 +70,7 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
         }
     }
 
-    const auto installed = Service::FileSystem::GetUnionContents();
+    const auto& installed = Core::System::GetInstance().GetContentProvider();
 
     const auto& disabled = Settings::values.disabled_addons[title_id];
     const auto update_disabled =
@@ -155,7 +156,7 @@ std::vector<VirtualFile> PatchManager::CollectPatches(const std::vector<VirtualD
     return out;
 }
 
-std::vector<u8> PatchManager::PatchNSO(const std::vector<u8>& nso) const {
+std::vector<u8> PatchManager::PatchNSO(const std::vector<u8>& nso, const std::string& name) const {
     if (nso.size() < sizeof(Loader::NSOHeader)) {
         return nso;
     }
@@ -171,18 +172,19 @@ std::vector<u8> PatchManager::PatchNSO(const std::vector<u8>& nso) const {
     const auto build_id = build_id_raw.substr(0, build_id_raw.find_last_not_of('0') + 1);
 
     if (Settings::values.dump_nso) {
-        LOG_INFO(Loader, "Dumping NSO for build_id={}, title_id={:016X}", build_id, title_id);
+        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);
         if (dump_dir != nullptr) {
             const auto nso_dir = GetOrCreateDirectoryRelative(dump_dir, "/nso");
-            const auto file = nso_dir->CreateFile(fmt::format("{}.nso", build_id));
+            const auto file = nso_dir->CreateFile(fmt::format("{}-{}.nso", name, build_id));
 
             file->Resize(nso.size());
             file->WriteBytes(nso);
         }
     }
 
-    LOG_INFO(Loader, "Patching NSO for build_id={}", build_id);
+    LOG_INFO(Loader, "Patching NSO for name={}, build_id={}", name, build_id);
 
     const auto load_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
     auto patch_dirs = load_dir->GetSubdirectories();
@@ -345,7 +347,7 @@ VirtualFile PatchManager::PatchRomFS(VirtualFile romfs, u64 ivfc_offset, Content
     if (romfs == nullptr)
         return romfs;
 
-    const auto installed = Service::FileSystem::GetUnionContents();
+    const auto& installed = Core::System::GetInstance().GetContentProvider();
 
     // Game Updates
     const auto update_tid = GetUpdateTitleID(title_id);
@@ -392,7 +394,7 @@ static bool IsDirValidAndNonEmpty(const VirtualDir& dir) {
 std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNames(
     VirtualFile update_raw) const {
     std::map<std::string, std::string, std::less<>> out;
-    const auto installed = Service::FileSystem::GetUnionContents();
+    const auto& installed = Core::System::GetInstance().GetContentProvider();
     const auto& disabled = Settings::values.disabled_addons[title_id];
 
     // Game Updates
@@ -466,10 +468,10 @@ std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNam
 
     // DLC
     const auto dlc_entries = installed.ListEntriesFilter(TitleType::AOC, ContentRecordType::Data);
-    std::vector<RegisteredCacheEntry> dlc_match;
+    std::vector<ContentProviderEntry> dlc_match;
     dlc_match.reserve(dlc_entries.size());
     std::copy_if(dlc_entries.begin(), dlc_entries.end(), std::back_inserter(dlc_match),
-                 [this, &installed](const RegisteredCacheEntry& entry) {
+                 [this, &installed](const ContentProviderEntry& entry) {
                      return (entry.title_id & DLC_BASE_TITLE_ID_MASK) == title_id &&
                             installed.GetEntry(entry)->GetStatus() == Loader::ResultStatus::Success;
                  });
@@ -492,7 +494,7 @@ std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNam
 }
 
 std::pair<std::unique_ptr<NACP>, VirtualFile> PatchManager::GetControlMetadata() const {
-    const auto installed{Service::FileSystem::GetUnionContents()};
+    const auto& installed = Core::System::GetInstance().GetContentProvider();
 
     const auto base_control_nca = installed.GetEntry(title_id, ContentRecordType::Control);
     if (base_control_nca == nullptr)

src/core/file_sys/patch_manager.h

@@ -44,7 +44,7 @@ public:
     // Currently tracked NSO patches:
     // - IPS
     // - IPSwitch
-    std::vector<u8> PatchNSO(const std::vector<u8>& nso) const;
+    std::vector<u8> PatchNSO(const std::vector<u8>& nso, const std::string& name) const;
 
     // Checks to see if PatchNSO() will have any effect given the NSO's build ID.
     // Used to prevent expensive copies in NSO loader.

src/core/file_sys/registered_cache.cpp

@@ -23,19 +23,19 @@ namespace FileSys {
 // The size of blocks to use when vfs raw copying into nand.
 constexpr size_t VFS_RC_LARGE_COPY_BLOCK = 0x400000;
 
-std::string RegisteredCacheEntry::DebugInfo() const {
+std::string ContentProviderEntry::DebugInfo() const {
     return fmt::format("title_id={:016X}, content_type={:02X}", title_id, static_cast<u8>(type));
 }
 
-bool operator<(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs) {
+bool operator<(const ContentProviderEntry& lhs, const ContentProviderEntry& rhs) {
     return (lhs.title_id < rhs.title_id) || (lhs.title_id == rhs.title_id && lhs.type < rhs.type);
 }
 
-bool operator==(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs) {
+bool operator==(const ContentProviderEntry& lhs, const ContentProviderEntry& rhs) {
     return std::tie(lhs.title_id, lhs.type) == std::tie(rhs.title_id, rhs.type);
 }
 
-bool operator!=(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs) {
+bool operator!=(const ContentProviderEntry& lhs, const ContentProviderEntry& rhs) {
     return !operator==(lhs, rhs);
 }
 
@@ -84,7 +84,7 @@ static std::string GetCNMTName(TitleType type, u64 title_id) {
     return fmt::format("{}_{:016x}.cnmt", TITLE_TYPE_NAMES[index], title_id);
 }
 
-static ContentRecordType GetCRTypeFromNCAType(NCAContentType type) {
+ContentRecordType GetCRTypeFromNCAType(NCAContentType type) {
     switch (type) {
     case NCAContentType::Program:
         // TODO(DarkLordZach): Differentiate between Program and Patch
@@ -104,6 +104,28 @@ static ContentRecordType GetCRTypeFromNCAType(NCAContentType type) {
     }
 }
 
+ContentProvider::~ContentProvider() = default;
+
+bool ContentProvider::HasEntry(ContentProviderEntry entry) const {
+    return HasEntry(entry.title_id, entry.type);
+}
+
+VirtualFile ContentProvider::GetEntryUnparsed(ContentProviderEntry entry) const {
+    return GetEntryUnparsed(entry.title_id, entry.type);
+}
+
+VirtualFile ContentProvider::GetEntryRaw(ContentProviderEntry entry) const {
+    return GetEntryRaw(entry.title_id, entry.type);
+}
+
+std::unique_ptr<NCA> ContentProvider::GetEntry(ContentProviderEntry entry) const {
+    return GetEntry(entry.title_id, entry.type);
+}
+
+std::vector<ContentProviderEntry> ContentProvider::ListEntries() const {
+    return ListEntriesFilter(std::nullopt, std::nullopt, std::nullopt);
+}
+
 VirtualFile RegisteredCache::OpenFileOrDirectoryConcat(const VirtualDir& dir,
                                                        std::string_view path) const {
     const auto file = dir->GetFileRelative(path);
@@ -161,8 +183,8 @@ VirtualFile RegisteredCache::GetFileAtID(NcaID id) const {
     return file;
 }
 
-static std::optional<NcaID> CheckMapForContentRecord(
-    const boost::container::flat_map<u64, CNMT>& map, u64 title_id, ContentRecordType type) {
+static std::optional<NcaID> CheckMapForContentRecord(const std::map<u64, CNMT>& map, u64 title_id,
+                                                     ContentRecordType type) {
     if (map.find(title_id) == map.end())
         return {};
 
@@ -268,7 +290,7 @@ void RegisteredCache::Refresh() {
     AccumulateYuzuMeta();
 }
 
-RegisteredCache::RegisteredCache(VirtualDir dir_, RegisteredCacheParsingFunction parsing_function)
+RegisteredCache::RegisteredCache(VirtualDir dir_, ContentProviderParsingFunction parsing_function)
     : dir(std::move(dir_)), parser(std::move(parsing_function)) {
     Refresh();
 }
@@ -279,19 +301,11 @@ bool RegisteredCache::HasEntry(u64 title_id, ContentRecordType type) const {
     return GetEntryRaw(title_id, type) != nullptr;
 }
 
-bool RegisteredCache::HasEntry(RegisteredCacheEntry entry) const {
-    return GetEntryRaw(entry) != nullptr;
-}
-
 VirtualFile RegisteredCache::GetEntryUnparsed(u64 title_id, ContentRecordType type) const {
     const auto id = GetNcaIDFromMetadata(title_id, type);
     return id ? GetFileAtID(*id) : nullptr;
 }
 
-VirtualFile RegisteredCache::GetEntryUnparsed(RegisteredCacheEntry entry) const {
-    return GetEntryUnparsed(entry.title_id, entry.type);
-}
-
 std::optional<u32> RegisteredCache::GetEntryVersion(u64 title_id) const {
     const auto meta_iter = meta.find(title_id);
     if (meta_iter != meta.end())
@@ -309,10 +323,6 @@ VirtualFile RegisteredCache::GetEntryRaw(u64 title_id, ContentRecordType type) c
     return id ? parser(GetFileAtID(*id), *id) : nullptr;
 }
 
-VirtualFile RegisteredCache::GetEntryRaw(RegisteredCacheEntry entry) const {
-    return GetEntryRaw(entry.title_id, entry.type);
-}
-
 std::unique_ptr<NCA> RegisteredCache::GetEntry(u64 title_id, ContentRecordType type) const {
     const auto raw = GetEntryRaw(title_id, type);
     if (raw == nullptr)
@@ -320,10 +330,6 @@ std::unique_ptr<NCA> RegisteredCache::GetEntry(u64 title_id, ContentRecordType t
     return std::make_unique<NCA>(raw, nullptr, 0, keys);
 }
 
-std::unique_ptr<NCA> RegisteredCache::GetEntry(RegisteredCacheEntry entry) const {
-    return GetEntry(entry.title_id, entry.type);
-}
-
 template <typename T>
 void RegisteredCache::IterateAllMetadata(
     std::vector<T>& out, std::function<T(const CNMT&, const ContentRecord&)> proc,
@@ -348,25 +354,14 @@ void RegisteredCache::IterateAllMetadata(
     }
 }
 
-std::vector<RegisteredCacheEntry> RegisteredCache::ListEntries() const {
-    std::vector<RegisteredCacheEntry> out;
-    IterateAllMetadata<RegisteredCacheEntry>(
-        out,
-        [](const CNMT& c, const ContentRecord& r) {
-            return RegisteredCacheEntry{c.GetTitleID(), r.type};
-        },
-        [](const CNMT& c, const ContentRecord& r) { return true; });
-    return out;
-}
-
-std::vector<RegisteredCacheEntry> RegisteredCache::ListEntriesFilter(
+std::vector<ContentProviderEntry> RegisteredCache::ListEntriesFilter(
     std::optional<TitleType> title_type, std::optional<ContentRecordType> record_type,
     std::optional<u64> title_id) const {
-    std::vector<RegisteredCacheEntry> out;
-    IterateAllMetadata<RegisteredCacheEntry>(
+    std::vector<ContentProviderEntry> out;
+    IterateAllMetadata<ContentProviderEntry>(
         out,
         [](const CNMT& c, const ContentRecord& r) {
-            return RegisteredCacheEntry{c.GetTitleID(), r.type};
+            return ContentProviderEntry{c.GetTitleID(), r.type};
         },
         [&title_type, &record_type, &title_id](const CNMT& c, const ContentRecord& r) {
             if (title_type && *title_type != c.GetType())
@@ -521,37 +516,56 @@ bool RegisteredCache::RawInstallYuzuMeta(const CNMT& cnmt) {
                         }) != yuzu_meta.end();
 }
 
-RegisteredCacheUnion::RegisteredCacheUnion(std::vector<RegisteredCache*> caches)
-    : caches(std::move(caches)) {}
+ContentProviderUnion::~ContentProviderUnion() = default;
 
-void RegisteredCacheUnion::Refresh() {
-    for (const auto& c : caches)
-        c->Refresh();
+void ContentProviderUnion::SetSlot(ContentProviderUnionSlot slot, ContentProvider* provider) {
+    providers[slot] = provider;
 }
 
-bool RegisteredCacheUnion::HasEntry(u64 title_id, ContentRecordType type) const {
-    return std::any_of(caches.begin(), caches.end(), [title_id, type](const auto& cache) {
-        return cache->HasEntry(title_id, type);
-    });
+void ContentProviderUnion::ClearSlot(ContentProviderUnionSlot slot) {
+    providers[slot] = nullptr;
 }
 
-bool RegisteredCacheUnion::HasEntry(RegisteredCacheEntry entry) const {
-    return HasEntry(entry.title_id, entry.type);
+void ContentProviderUnion::Refresh() {
+    for (auto& provider : providers) {
+        if (provider.second == nullptr)
+            continue;
+
+        provider.second->Refresh();
+    }
 }
 
-std::optional<u32> RegisteredCacheUnion::GetEntryVersion(u64 title_id) const {
-    for (const auto& c : caches) {
-        const auto res = c->GetEntryVersion(title_id);
-        if (res)
+bool ContentProviderUnion::HasEntry(u64 title_id, ContentRecordType type) const {
+    for (const auto& provider : providers) {
+        if (provider.second == nullptr)
+            continue;
+
+        if (provider.second->HasEntry(title_id, type))
+            return true;
+    }
+
+    return false;
+}
+
+std::optional<u32> ContentProviderUnion::GetEntryVersion(u64 title_id) const {
+    for (const auto& provider : providers) {
+        if (provider.second == nullptr)
+            continue;
+
+        const auto res = provider.second->GetEntryVersion(title_id);
+        if (res != std::nullopt)
             return res;
     }
 
-    return {};
+    return std::nullopt;
 }
 
-VirtualFile RegisteredCacheUnion::GetEntryUnparsed(u64 title_id, ContentRecordType type) const {
-    for (const auto& c : caches) {
-        const auto res = c->GetEntryUnparsed(title_id, type);
+VirtualFile ContentProviderUnion::GetEntryUnparsed(u64 title_id, ContentRecordType type) const {
+    for (const auto& provider : providers) {
+        if (provider.second == nullptr)
+            continue;
+
+        const auto res = provider.second->GetEntryUnparsed(title_id, type);
         if (res != nullptr)
             return res;
     }
@@ -559,13 +573,12 @@ VirtualFile RegisteredCacheUnion::GetEntryUnparsed(u64 title_id, ContentRecordTy
     return nullptr;
 }
 
-VirtualFile RegisteredCacheUnion::GetEntryUnparsed(RegisteredCacheEntry entry) const {
-    return GetEntryUnparsed(entry.title_id, entry.type);
-}
+VirtualFile ContentProviderUnion::GetEntryRaw(u64 title_id, ContentRecordType type) const {
+    for (const auto& provider : providers) {
+        if (provider.second == nullptr)
+            continue;
 
-VirtualFile RegisteredCacheUnion::GetEntryRaw(u64 title_id, ContentRecordType type) const {
-    for (const auto& c : caches) {
-        const auto res = c->GetEntryRaw(title_id, type);
+        const auto res = provider.second->GetEntryRaw(title_id, type);
         if (res != nullptr)
             return res;
     }
@@ -573,30 +586,30 @@ VirtualFile RegisteredCacheUnion::GetEntryRaw(u64 title_id, ContentRecordType ty
     return nullptr;
 }
 
-VirtualFile RegisteredCacheUnion::GetEntryRaw(RegisteredCacheEntry entry) const {
-    return GetEntryRaw(entry.title_id, entry.type);
+std::unique_ptr<NCA> ContentProviderUnion::GetEntry(u64 title_id, ContentRecordType type) const {
+    for (const auto& provider : providers) {
+        if (provider.second == nullptr)
+            continue;
+
+        auto res = provider.second->GetEntry(title_id, type);
+        if (res != nullptr)
+            return res;
+    }
+
+    return nullptr;
 }
 
-std::unique_ptr<NCA> RegisteredCacheUnion::GetEntry(u64 title_id, ContentRecordType type) const {
-    const auto raw = GetEntryRaw(title_id, type);
-    if (raw == nullptr)
-        return nullptr;
-    return std::make_unique<NCA>(raw);
-}
+std::vector<ContentProviderEntry> ContentProviderUnion::ListEntriesFilter(
+    std::optional<TitleType> title_type, std::optional<ContentRecordType> record_type,
+    std::optional<u64> title_id) const {
+    std::vector<ContentProviderEntry> out;
 
-std::unique_ptr<NCA> RegisteredCacheUnion::GetEntry(RegisteredCacheEntry entry) const {
-    return GetEntry(entry.title_id, entry.type);
-}
+    for (const auto& provider : providers) {
+        if (provider.second == nullptr)
+            continue;
 
-std::vector<RegisteredCacheEntry> RegisteredCacheUnion::ListEntries() const {
-    std::vector<RegisteredCacheEntry> out;
-    for (const auto& c : caches) {
-        c->IterateAllMetadata<RegisteredCacheEntry>(
-            out,
-            [](const CNMT& c, const ContentRecord& r) {
-                return RegisteredCacheEntry{c.GetTitleID(), r.type};
-            },
-            [](const CNMT& c, const ContentRecord& r) { return true; });
+        const auto vec = provider.second->ListEntriesFilter(title_type, record_type, title_id);
+        std::copy(vec.begin(), vec.end(), std::back_inserter(out));
     }
 
     std::sort(out.begin(), out.end());
@@ -604,25 +617,87 @@ std::vector<RegisteredCacheEntry> RegisteredCacheUnion::ListEntries() const {
     return out;
 }
 
-std::vector<RegisteredCacheEntry> RegisteredCacheUnion::ListEntriesFilter(
+std::vector<std::pair<ContentProviderUnionSlot, ContentProviderEntry>>
+ContentProviderUnion::ListEntriesFilterOrigin(std::optional<ContentProviderUnionSlot> origin,
+                                              std::optional<TitleType> title_type,
+                                              std::optional<ContentRecordType> record_type,
+                                              std::optional<u64> title_id) const {
+    std::vector<std::pair<ContentProviderUnionSlot, ContentProviderEntry>> out;
+
+    for (const auto& provider : providers) {
+        if (provider.second == nullptr)
+            continue;
+
+        if (origin.has_value() && *origin != provider.first)
+            continue;
+
+        const auto vec = provider.second->ListEntriesFilter(title_type, record_type, title_id);
+        std::transform(vec.begin(), vec.end(), std::back_inserter(out),
+                       [&provider](const ContentProviderEntry& entry) {
+                           return std::make_pair(provider.first, entry);
+                       });
+    }
+
+    std::sort(out.begin(), out.end());
+    out.erase(std::unique(out.begin(), out.end()), out.end());
+    return out;
+}
+
+ManualContentProvider::~ManualContentProvider() = default;
+
+void ManualContentProvider::AddEntry(TitleType title_type, ContentRecordType content_type,
+                                     u64 title_id, VirtualFile file) {
+    entries.insert_or_assign({title_type, content_type, title_id}, file);
+}
+
+void ManualContentProvider::ClearAllEntries() {
+    entries.clear();
+}
+
+void ManualContentProvider::Refresh() {}
+
+bool ManualContentProvider::HasEntry(u64 title_id, ContentRecordType type) const {
+    return GetEntryRaw(title_id, type) != nullptr;
+}
+
+std::optional<u32> ManualContentProvider::GetEntryVersion(u64 title_id) const {
+    return std::nullopt;
+}
+
+VirtualFile ManualContentProvider::GetEntryUnparsed(u64 title_id, ContentRecordType type) const {
+    return GetEntryRaw(title_id, type);
+}
+
+VirtualFile ManualContentProvider::GetEntryRaw(u64 title_id, ContentRecordType type) const {
+    const auto iter =
+        std::find_if(entries.begin(), entries.end(), [title_id, type](const auto& entry) {
+            const auto [title_type, content_type, e_title_id] = entry.first;
+            return content_type == type && e_title_id == title_id;
+        });
+    if (iter == entries.end())
+        return nullptr;
+    return iter->second;
+}
+
+std::unique_ptr<NCA> ManualContentProvider::GetEntry(u64 title_id, ContentRecordType type) const {
+    const auto res = GetEntryRaw(title_id, type);
+    if (res == nullptr)
+        return nullptr;
+    return std::make_unique<NCA>(res, nullptr, 0, keys);
+}
+
+std::vector<ContentProviderEntry> ManualContentProvider::ListEntriesFilter(
     std::optional<TitleType> title_type, std::optional<ContentRecordType> record_type,
     std::optional<u64> title_id) const {
-    std::vector<RegisteredCacheEntry> out;
-    for (const auto& c : caches) {
-        c->IterateAllMetadata<RegisteredCacheEntry>(
-            out,
-            [](const CNMT& c, const ContentRecord& r) {
-                return RegisteredCacheEntry{c.GetTitleID(), r.type};
-            },
-            [&title_type, &record_type, &title_id](const CNMT& c, const ContentRecord& r) {
-                if (title_type && *title_type != c.GetType())
-                    return false;
-                if (record_type && *record_type != r.type)
-                    return false;
-                if (title_id && *title_id != c.GetTitleID())
-                    return false;
-                return true;
-            });
+    std::vector<ContentProviderEntry> out;
+
+    for (const auto& entry : entries) {
+        const auto [e_title_type, e_content_type, e_title_id] = entry.first;
+        if ((title_type == std::nullopt || e_title_type == *title_type) &&
+            (record_type == std::nullopt || e_content_type == *record_type) &&
+            (title_id == std::nullopt || e_title_id == *title_id)) {
+            out.emplace_back(ContentProviderEntry{e_title_id, e_content_type});
+        }
     }
 
     std::sort(out.begin(), out.end());

src/core/file_sys/registered_cache.h

@@ -21,12 +21,13 @@ class NSP;
 class XCI;
 
 enum class ContentRecordType : u8;
+enum class NCAContentType : u8;
 enum class TitleType : u8;
 
 struct ContentRecord;
 
 using NcaID = std::array<u8, 0x10>;
-using RegisteredCacheParsingFunction = std::function<VirtualFile(const VirtualFile&, const NcaID&)>;
+using ContentProviderParsingFunction = std::function<VirtualFile(const VirtualFile&, const NcaID&)>;
 using VfsCopyFunction = std::function<bool(const VirtualFile&, const VirtualFile&, size_t)>;
 
 enum class InstallResult {
@@ -36,7 +37,7 @@ enum class InstallResult {
     ErrorMetaFailed,
 };
 
-struct RegisteredCacheEntry {
+struct ContentProviderEntry {
     u64 title_id;
     ContentRecordType type;
 
@@ -47,12 +48,46 @@ constexpr u64 GetUpdateTitleID(u64 base_title_id) {
     return base_title_id | 0x800;
 }
 
+ContentRecordType GetCRTypeFromNCAType(NCAContentType type);
+
 // boost flat_map requires operator< for O(log(n)) lookups.
-bool operator<(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs);
+bool operator<(const ContentProviderEntry& lhs, const ContentProviderEntry& rhs);
 
 // std unique requires operator== to identify duplicates.
-bool operator==(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs);
-bool operator!=(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs);
+bool operator==(const ContentProviderEntry& lhs, const ContentProviderEntry& rhs);
+bool operator!=(const ContentProviderEntry& lhs, const ContentProviderEntry& rhs);
+
+class ContentProvider {
+public:
+    virtual ~ContentProvider();
+
+    virtual void Refresh() = 0;
+
+    virtual bool HasEntry(u64 title_id, ContentRecordType type) const = 0;
+    virtual bool HasEntry(ContentProviderEntry entry) const;
+
+    virtual std::optional<u32> GetEntryVersion(u64 title_id) const = 0;
+
+    virtual VirtualFile GetEntryUnparsed(u64 title_id, ContentRecordType type) const = 0;
+    virtual VirtualFile GetEntryUnparsed(ContentProviderEntry entry) const;
+
+    virtual VirtualFile GetEntryRaw(u64 title_id, ContentRecordType type) const = 0;
+    virtual VirtualFile GetEntryRaw(ContentProviderEntry entry) const;
+
+    virtual std::unique_ptr<NCA> GetEntry(u64 title_id, ContentRecordType type) const = 0;
+    virtual std::unique_ptr<NCA> GetEntry(ContentProviderEntry entry) const;
+
+    virtual std::vector<ContentProviderEntry> ListEntries() const;
+
+    // If a parameter is not std::nullopt, it will be filtered for from all entries.
+    virtual std::vector<ContentProviderEntry> ListEntriesFilter(
+        std::optional<TitleType> title_type = {}, std::optional<ContentRecordType> record_type = {},
+        std::optional<u64> title_id = {}) const = 0;
+
+protected:
+    // A single instance of KeyManager to be used by GetEntry()
+    Core::Crypto::KeyManager keys;
+};
 
 /*
  * A class that catalogues NCAs in the registered directory structure.
@@ -67,39 +102,32 @@ bool operator!=(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs
  * (This impl also supports substituting the nca dir for an nca file, as that's more convenient
  * when 4GB splitting can be ignored.)
  */
-class RegisteredCache {
-    friend class RegisteredCacheUnion;
-
+class RegisteredCache : public ContentProvider {
 public:
     // Parsing function defines the conversion from raw file to NCA. If there are other steps
     // besides creating the NCA from the file (e.g. NAX0 on SD Card), that should go in a custom
     // parsing function.
     explicit RegisteredCache(VirtualDir dir,
-                             RegisteredCacheParsingFunction parsing_function =
+                             ContentProviderParsingFunction parsing_function =
                                  [](const VirtualFile& file, const NcaID& id) { return file; });
-    ~RegisteredCache();
+    ~RegisteredCache() override;
 
-    void Refresh();
+    void Refresh() override;
 
-    bool HasEntry(u64 title_id, ContentRecordType type) const;
-    bool HasEntry(RegisteredCacheEntry entry) const;
+    bool HasEntry(u64 title_id, ContentRecordType type) const override;
 
-    std::optional<u32> GetEntryVersion(u64 title_id) const;
+    std::optional<u32> GetEntryVersion(u64 title_id) const override;
 
-    VirtualFile GetEntryUnparsed(u64 title_id, ContentRecordType type) const;
-    VirtualFile GetEntryUnparsed(RegisteredCacheEntry entry) const;
+    VirtualFile GetEntryUnparsed(u64 title_id, ContentRecordType type) const override;
 
-    VirtualFile GetEntryRaw(u64 title_id, ContentRecordType type) const;
-    VirtualFile GetEntryRaw(RegisteredCacheEntry entry) const;
+    VirtualFile GetEntryRaw(u64 title_id, ContentRecordType type) const override;
 
-    std::unique_ptr<NCA> GetEntry(u64 title_id, ContentRecordType type) const;
-    std::unique_ptr<NCA> GetEntry(RegisteredCacheEntry entry) const;
+    std::unique_ptr<NCA> GetEntry(u64 title_id, ContentRecordType type) const override;
 
-    std::vector<RegisteredCacheEntry> ListEntries() const;
     // If a parameter is not std::nullopt, it will be filtered for from all entries.
-    std::vector<RegisteredCacheEntry> ListEntriesFilter(
+    std::vector<ContentProviderEntry> ListEntriesFilter(
         std::optional<TitleType> title_type = {}, std::optional<ContentRecordType> record_type = {},
-        std::optional<u64> title_id = {}) const;
+        std::optional<u64> title_id = {}) const override;
 
     // Raw copies all the ncas from the xci/nsp to the csache. Does some quick checks to make sure
     // there is a meta NCA and all of them are accessible.
@@ -131,46 +159,70 @@ private:
     bool RawInstallYuzuMeta(const CNMT& cnmt);
 
     VirtualDir dir;
-    RegisteredCacheParsingFunction parser;
-    Core::Crypto::KeyManager keys;
+    ContentProviderParsingFunction parser;
 
     // maps tid -> NcaID of meta
-    boost::container::flat_map<u64, NcaID> meta_id;
+    std::map<u64, NcaID> meta_id;
     // maps tid -> meta
-    boost::container::flat_map<u64, CNMT> meta;
+    std::map<u64, CNMT> meta;
     // maps tid -> meta for CNMT in yuzu_meta
-    boost::container::flat_map<u64, CNMT> yuzu_meta;
+    std::map<u64, CNMT> yuzu_meta;
 };
 
-// Combines multiple RegisteredCaches (i.e. SysNAND, UserNAND, SDMC) into one interface.
-class RegisteredCacheUnion {
+enum class ContentProviderUnionSlot {
+    SysNAND,        ///< System NAND
+    UserNAND,       ///< User NAND
+    SDMC,           ///< SD Card
+    FrontendManual, ///< Frontend-defined game list or similar
+};
+
+// Combines multiple ContentProvider(s) (i.e. SysNAND, UserNAND, SDMC) into one interface.
+class ContentProviderUnion : public ContentProvider {
 public:
-    explicit RegisteredCacheUnion(std::vector<RegisteredCache*> caches);
+    ~ContentProviderUnion() override;
 
-    void Refresh();
+    void SetSlot(ContentProviderUnionSlot slot, ContentProvider* provider);
+    void ClearSlot(ContentProviderUnionSlot slot);
 
-    bool HasEntry(u64 title_id, ContentRecordType type) const;
-    bool HasEntry(RegisteredCacheEntry entry) const;
+    void Refresh() override;
+    bool HasEntry(u64 title_id, ContentRecordType type) const override;
+    std::optional<u32> GetEntryVersion(u64 title_id) const override;
+    VirtualFile GetEntryUnparsed(u64 title_id, ContentRecordType type) const override;
+    VirtualFile GetEntryRaw(u64 title_id, ContentRecordType type) const override;
+    std::unique_ptr<NCA> GetEntry(u64 title_id, ContentRecordType type) const override;
+    std::vector<ContentProviderEntry> ListEntriesFilter(
+        std::optional<TitleType> title_type, std::optional<ContentRecordType> record_type,
+        std::optional<u64> title_id) const override;
 
-    std::optional<u32> GetEntryVersion(u64 title_id) const;
-
-    VirtualFile GetEntryUnparsed(u64 title_id, ContentRecordType type) const;
-    VirtualFile GetEntryUnparsed(RegisteredCacheEntry entry) const;
-
-    VirtualFile GetEntryRaw(u64 title_id, ContentRecordType type) const;
-    VirtualFile GetEntryRaw(RegisteredCacheEntry entry) const;
-
-    std::unique_ptr<NCA> GetEntry(u64 title_id, ContentRecordType type) const;
-    std::unique_ptr<NCA> GetEntry(RegisteredCacheEntry entry) const;
-
-    std::vector<RegisteredCacheEntry> ListEntries() const;
-    // If a parameter is not std::nullopt, it will be filtered for from all entries.
-    std::vector<RegisteredCacheEntry> ListEntriesFilter(
+    std::vector<std::pair<ContentProviderUnionSlot, ContentProviderEntry>> ListEntriesFilterOrigin(
+        std::optional<ContentProviderUnionSlot> origin = {},
         std::optional<TitleType> title_type = {}, std::optional<ContentRecordType> record_type = {},
         std::optional<u64> title_id = {}) const;
 
 private:
-    std::vector<RegisteredCache*> caches;
+    std::map<ContentProviderUnionSlot, ContentProvider*> providers;
+};
+
+class ManualContentProvider : public ContentProvider {
+public:
+    ~ManualContentProvider() override;
+
+    void AddEntry(TitleType title_type, ContentRecordType content_type, u64 title_id,
+                  VirtualFile file);
+    void ClearAllEntries();
+
+    void Refresh() override;
+    bool HasEntry(u64 title_id, ContentRecordType type) const override;
+    std::optional<u32> GetEntryVersion(u64 title_id) const override;
+    VirtualFile GetEntryUnparsed(u64 title_id, ContentRecordType type) const override;
+    VirtualFile GetEntryRaw(u64 title_id, ContentRecordType type) const override;
+    std::unique_ptr<NCA> GetEntry(u64 title_id, ContentRecordType type) const override;
+    std::vector<ContentProviderEntry> ListEntriesFilter(
+        std::optional<TitleType> title_type, std::optional<ContentRecordType> record_type,
+        std::optional<u64> title_id) const override;
+
+private:
+    std::map<std::tuple<TitleType, ContentRecordType, u64>, VirtualFile> entries;
 };
 
 } // namespace FileSys

src/core/file_sys/romfs_factory.cpp

@@ -48,7 +48,7 @@ ResultVal<VirtualFile> RomFSFactory::Open(u64 title_id, StorageId storage, Conte
 
     switch (storage) {
     case StorageId::None:
-        res = Service::FileSystem::GetUnionContents().GetEntry(title_id, type);
+        res = Core::System::GetInstance().GetContentProvider().GetEntry(title_id, type);
         break;
     case StorageId::NandSystem:
         res = Service::FileSystem::GetSystemNANDContents()->GetEntry(title_id, type);

src/core/file_sys/submission_package.cpp

@@ -143,11 +143,12 @@ std::multimap<u64, std::shared_ptr<NCA>> NSP::GetNCAsByTitleID() const {
     return out;
 }
 
-std::map<u64, std::map<ContentRecordType, std::shared_ptr<NCA>>> NSP::GetNCAs() const {
+std::map<u64, std::map<std::pair<TitleType, ContentRecordType>, std::shared_ptr<NCA>>>
+NSP::GetNCAs() const {
     return ncas;
 }
 
-std::shared_ptr<NCA> NSP::GetNCA(u64 title_id, ContentRecordType type) const {
+std::shared_ptr<NCA> NSP::GetNCA(u64 title_id, ContentRecordType type, TitleType title_type) const {
     if (extracted)
         LOG_WARNING(Service_FS, "called on an NSP that is of type extracted.");
 
@@ -155,14 +156,14 @@ std::shared_ptr<NCA> NSP::GetNCA(u64 title_id, ContentRecordType type) const {
     if (title_id_iter == ncas.end())
         return nullptr;
 
-    const auto type_iter = title_id_iter->second.find(type);
+    const auto type_iter = title_id_iter->second.find({title_type, type});
     if (type_iter == title_id_iter->second.end())
         return nullptr;
 
     return type_iter->second;
 }
 
-VirtualFile NSP::GetNCAFile(u64 title_id, ContentRecordType type) const {
+VirtualFile NSP::GetNCAFile(u64 title_id, ContentRecordType type, TitleType title_type) const {
     if (extracted)
         LOG_WARNING(Service_FS, "called on an NSP that is of type extracted.");
     const auto nca = GetNCA(title_id, type);
@@ -240,7 +241,7 @@ void NSP::ReadNCAs(const std::vector<VirtualFile>& files) {
             const CNMT cnmt(inner_file);
             auto& ncas_title = ncas[cnmt.GetTitleID()];
 
-            ncas_title[ContentRecordType::Meta] = nca;
+            ncas_title[{cnmt.GetType(), ContentRecordType::Meta}] = nca;
             for (const auto& rec : cnmt.GetContentRecords()) {
                 const auto id_string = Common::HexArrayToString(rec.nca_id, false);
                 const auto next_file = pfs->GetFile(fmt::format("{}.nca", id_string));
@@ -258,7 +259,7 @@ void NSP::ReadNCAs(const std::vector<VirtualFile>& files) {
                 if (next_nca->GetStatus() == Loader::ResultStatus::Success ||
                     (next_nca->GetStatus() == Loader::ResultStatus::ErrorMissingBKTRBaseRomFS &&
                      (cnmt.GetTitleID() & 0x800) != 0)) {
-                    ncas_title[rec.type] = std::move(next_nca);
+                    ncas_title[{cnmt.GetType(), rec.type}] = std::move(next_nca);
                 }
             }
 

src/core/file_sys/submission_package.h

@@ -42,9 +42,12 @@ public:
     // Type 0 Only (Collection of NCAs + Certificate + Ticket + Meta XML)
     std::vector<std::shared_ptr<NCA>> GetNCAsCollapsed() const;
     std::multimap<u64, std::shared_ptr<NCA>> GetNCAsByTitleID() const;
-    std::map<u64, std::map<ContentRecordType, std::shared_ptr<NCA>>> GetNCAs() const;
-    std::shared_ptr<NCA> GetNCA(u64 title_id, ContentRecordType type) const;
-    VirtualFile GetNCAFile(u64 title_id, ContentRecordType type) const;
+    std::map<u64, std::map<std::pair<TitleType, ContentRecordType>, std::shared_ptr<NCA>>> GetNCAs()
+        const;
+    std::shared_ptr<NCA> GetNCA(u64 title_id, ContentRecordType type,
+                                TitleType title_type = TitleType::Application) const;
+    VirtualFile GetNCAFile(u64 title_id, ContentRecordType type,
+                           TitleType title_type = TitleType::Application) const;
     std::vector<Core::Crypto::Key128> GetTitlekey() const;
 
     std::vector<VirtualFile> GetFiles() const override;
@@ -67,7 +70,7 @@ private:
 
     std::shared_ptr<PartitionFilesystem> pfs;
     // Map title id -> {map type -> NCA}
-    std::map<u64, std::map<ContentRecordType, std::shared_ptr<NCA>>> ncas;
+    std::map<u64, std::map<std::pair<TitleType, ContentRecordType>, std::shared_ptr<NCA>>> ncas;
     std::vector<VirtualFile> ticket_files;
 
     Core::Crypto::KeyManager keys;

src/core/gdbstub/gdbstub.cpp

@@ -1030,7 +1030,7 @@ static void Step() {
 
 /// Tell the CPU if we hit a memory breakpoint.
 bool IsMemoryBreak() {
-    if (IsConnected()) {
+    if (!IsConnected()) {
         return false;
     }
 

src/core/hle/ipc_helpers.h

@@ -139,10 +139,8 @@ public:
             context->AddDomainObject(std::move(iface));
         } else {
             auto& kernel = Core::System::GetInstance().Kernel();
-            auto sessions =
+            auto [server, client] =
                 Kernel::ServerSession::CreateSessionPair(kernel, iface->GetServiceName());
-            auto server = std::get<Kernel::SharedPtr<Kernel::ServerSession>>(sessions);
-            auto client = std::get<Kernel::SharedPtr<Kernel::ClientSession>>(sessions);
             iface->ClientConnected(server);
             context->AddMoveObject(std::move(client));
         }

src/core/hle/kernel/client_port.cpp

@@ -2,8 +2,6 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <tuple>
-
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
 #include "core/hle/kernel/errors.h"
@@ -31,18 +29,18 @@ ResultVal<SharedPtr<ClientSession>> ClientPort::Connect() {
     active_sessions++;
 
     // Create a new session pair, let the created sessions inherit the parent port's HLE handler.
-    auto sessions = ServerSession::CreateSessionPair(kernel, server_port->GetName(), this);
+    auto [server, client] = ServerSession::CreateSessionPair(kernel, server_port->GetName(), this);
 
     if (server_port->HasHLEHandler()) {
-        server_port->GetHLEHandler()->ClientConnected(std::get<SharedPtr<ServerSession>>(sessions));
+        server_port->GetHLEHandler()->ClientConnected(server);
     } else {
-        server_port->AppendPendingSession(std::get<SharedPtr<ServerSession>>(sessions));
+        server_port->AppendPendingSession(server);
     }
 
     // Wake the threads waiting on the ServerPort
     server_port->WakeupAllWaitingThreads();
 
-    return MakeResult(std::get<SharedPtr<ClientSession>>(sessions));
+    return MakeResult(client);
 }
 
 void ClientPort::ConnectionClosed() {

src/core/hle/kernel/kernel.cpp

@@ -21,6 +21,7 @@
 #include "core/hle/kernel/thread.h"
 #include "core/hle/lock.h"
 #include "core/hle/result.h"
+#include "core/memory.h"
 
 namespace Kernel {
 
@@ -181,6 +182,7 @@ void KernelCore::AppendNewProcess(SharedPtr<Process> process) {
 
 void KernelCore::MakeCurrentProcess(Process* process) {
     impl->current_process = process;
+    Memory::SetCurrentPageTable(&process->VMManager().page_table);
 }
 
 Process* KernelCore::CurrentProcess() {

src/core/hle/kernel/process.cpp

@@ -32,9 +32,6 @@ namespace {
  * @param priority The priority to give the main thread
  */
 void SetupMainThread(Process& owner_process, KernelCore& kernel, VAddr entry_point, u32 priority) {
-    // Setup page table so we can write to memory
-    Memory::SetCurrentPageTable(&owner_process.VMManager().page_table);
-
     // Initialize new "main" thread
     const VAddr stack_top = owner_process.VMManager().GetTLSIORegionEndAddress();
     auto thread_res = Thread::Create(kernel, "main", entry_point, priority, 0,
@@ -109,6 +106,8 @@ ResultCode Process::LoadFromMetadata(const FileSys::ProgramMetadata& metadata) {
     is_64bit_process = metadata.Is64BitProgram();
 
     vm_manager.Reset(metadata.GetAddressSpaceType());
+    // Ensure that the potentially resized page table is seen by CPU backends.
+    Memory::SetCurrentPageTable(&vm_manager.page_table);
 
     const auto& caps = metadata.GetKernelCapabilities();
     const auto capability_init_result =

src/core/hle/kernel/scheduler.cpp

@@ -101,7 +101,6 @@ void Scheduler::SwitchContext(Thread* new_thread) {
         auto* const thread_owner_process = current_thread->GetOwnerProcess();
         if (previous_process != thread_owner_process) {
             system.Kernel().MakeCurrentProcess(thread_owner_process);
-            Memory::SetCurrentPageTable(&thread_owner_process->VMManager().page_table);
         }
 
         cpu_core.LoadContext(new_thread->GetContext());

src/core/hle/kernel/server_port.cpp

@@ -39,9 +39,8 @@ void ServerPort::Acquire(Thread* thread) {
     ASSERT_MSG(!ShouldWait(thread), "object unavailable!");
 }
 
-std::tuple<SharedPtr<ServerPort>, SharedPtr<ClientPort>> ServerPort::CreatePortPair(
-    KernelCore& kernel, u32 max_sessions, std::string name) {
-
+ServerPort::PortPair ServerPort::CreatePortPair(KernelCore& kernel, u32 max_sessions,
+                                                std::string name) {
     SharedPtr<ServerPort> server_port(new ServerPort(kernel));
     SharedPtr<ClientPort> client_port(new ClientPort(kernel));
 
@@ -51,7 +50,7 @@ std::tuple<SharedPtr<ServerPort>, SharedPtr<ClientPort>> ServerPort::CreatePortP
     client_port->max_sessions = max_sessions;
     client_port->active_sessions = 0;
 
-    return std::make_tuple(std::move(server_port), std::move(client_port));
+    return std::make_pair(std::move(server_port), std::move(client_port));
 }
 
 } // namespace Kernel

src/core/hle/kernel/server_port.h

@@ -6,7 +6,7 @@
 
 #include <memory>
 #include <string>
-#include <tuple>
+#include <utility>
 #include <vector>
 #include "common/common_types.h"
 #include "core/hle/kernel/object.h"
@@ -23,6 +23,7 @@ class SessionRequestHandler;
 class ServerPort final : public WaitObject {
 public:
     using HLEHandler = std::shared_ptr<SessionRequestHandler>;
+    using PortPair = std::pair<SharedPtr<ServerPort>, SharedPtr<ClientPort>>;
 
     /**
      * Creates a pair of ServerPort and an associated ClientPort.
@@ -32,8 +33,8 @@ public:
      * @param name Optional name of the ports
      * @return The created port tuple
      */
-    static std::tuple<SharedPtr<ServerPort>, SharedPtr<ClientPort>> CreatePortPair(
-        KernelCore& kernel, u32 max_sessions, std::string name = "UnknownPort");
+    static PortPair CreatePortPair(KernelCore& kernel, u32 max_sessions,
+                                   std::string name = "UnknownPort");
 
     std::string GetTypeName() const override {
         return "ServerPort";

src/core/hle/kernel/server_session.cpp

@@ -28,11 +28,9 @@ ServerSession::~ServerSession() {
     // the emulated application.
 
     // Decrease the port's connection count.
-    if (parent->port)
+    if (parent->port) {
         parent->port->ConnectionClosed();
-
-    // TODO(Subv): Wake up all the ClientSession's waiting threads and set
-    // the SendSyncRequest result to 0xC920181A.
+    }
 
     parent->server = nullptr;
 }
@@ -74,9 +72,6 @@ void ServerSession::ClientDisconnected() {
         handler->ClientDisconnected(this);
     }
 
-    // TODO(Subv): Force a wake up of all the ServerSession's waiting threads and set
-    // their WaitSynchronization result to 0xC920181A.
-
     // Clean up the list of client threads with pending requests, they are unneeded now that the
     // client endpoint is closed.
     pending_requesting_threads.clear();
@@ -204,6 +199,6 @@ ServerSession::SessionPair ServerSession::CreateSessionPair(KernelCore& kernel,
     client_session->parent = parent;
     server_session->parent = parent;
 
-    return std::make_tuple(std::move(server_session), std::move(client_session));
+    return std::make_pair(std::move(server_session), std::move(client_session));
 }
 } // namespace Kernel

src/core/hle/kernel/server_session.h

@@ -6,6 +6,7 @@
 
 #include <memory>
 #include <string>
+#include <utility>
 #include <vector>
 
 #include "core/hle/kernel/object.h"
@@ -41,6 +42,10 @@ public:
         return "ServerSession";
     }
 
+    std::string GetName() const override {
+        return name;
+    }
+
     static const HandleType HANDLE_TYPE = HandleType::ServerSession;
     HandleType GetHandleType() const override {
         return HANDLE_TYPE;
@@ -54,7 +59,7 @@ public:
         return parent.get();
     }
 
-    using SessionPair = std::tuple<SharedPtr<ServerSession>, SharedPtr<ClientSession>>;
+    using SessionPair = std::pair<SharedPtr<ServerSession>, SharedPtr<ClientSession>>;
 
     /**
      * Creates a pair of ServerSession and an associated ClientSession.

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

@@ -86,7 +86,7 @@ static FileSys::VirtualFile GetManualRomFS() {
     if (loader.ReadManualRomFS(out) == Loader::ResultStatus::Success)
         return out;
 
-    const auto& installed{FileSystem::GetUnionContents()};
+    const auto& installed{Core::System::GetInstance().GetContentProvider()};
     const auto res = installed.GetEntry(Core::System::GetInstance().CurrentProcess()->GetTitleID(),
                                         FileSys::ContentRecordType::Manual);
 

src/core/hle/service/aoc/aoc_u.cpp

@@ -33,11 +33,11 @@ static bool CheckAOCTitleIDMatchesBase(u64 title_id, u64 base) {
 
 static std::vector<u64> AccumulateAOCTitleIDs() {
     std::vector<u64> add_on_content;
-    const auto rcu = FileSystem::GetUnionContents();
+    const auto& rcu = Core::System::GetInstance().GetContentProvider();
     const auto list =
         rcu.ListEntriesFilter(FileSys::TitleType::AOC, FileSys::ContentRecordType::Data);
     std::transform(list.begin(), list.end(), std::back_inserter(add_on_content),
-                   [](const FileSys::RegisteredCacheEntry& rce) { return rce.title_id; });
+                   [](const FileSys::ContentProviderEntry& rce) { return rce.title_id; });
     add_on_content.erase(
         std::remove_if(
             add_on_content.begin(), add_on_content.end(),

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

@@ -391,11 +391,6 @@ void WriteSaveDataSize(FileSys::SaveDataType type, u64 title_id, u128 user_id,
         save_data_factory->WriteSaveDataSize(type, title_id, user_id, new_value);
 }
 
-FileSys::RegisteredCacheUnion GetUnionContents() {
-    return FileSys::RegisteredCacheUnion{
-        {GetSystemNANDContents(), GetUserNANDContents(), GetSDMCContents()}};
-}
-
 FileSys::RegisteredCache* GetSystemNANDContents() {
     LOG_TRACE(Service_FS, "Opening System NAND Contents");
 
@@ -460,6 +455,10 @@ void CreateFactories(FileSys::VfsFilesystem& vfs, bool overwrite) {
     if (bis_factory == nullptr) {
         bis_factory =
             std::make_unique<FileSys::BISFactory>(nand_directory, load_directory, dump_directory);
+        Core::System::GetInstance().RegisterContentProvider(
+            FileSys::ContentProviderUnionSlot::SysNAND, bis_factory->GetSystemNANDContents());
+        Core::System::GetInstance().RegisterContentProvider(
+            FileSys::ContentProviderUnionSlot::UserNAND, bis_factory->GetUserNANDContents());
     }
 
     if (save_data_factory == nullptr) {
@@ -468,6 +467,8 @@ void CreateFactories(FileSys::VfsFilesystem& vfs, bool overwrite) {
 
     if (sdmc_factory == nullptr) {
         sdmc_factory = std::make_unique<FileSys::SDMCFactory>(std::move(sd_directory));
+        Core::System::GetInstance().RegisterContentProvider(FileSys::ContentProviderUnionSlot::SDMC,
+                                                            sdmc_factory->GetSDMCContents());
     }
 }
 

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

@@ -54,8 +54,6 @@ FileSys::SaveDataSize ReadSaveDataSize(FileSys::SaveDataType type, u64 title_id,
 void WriteSaveDataSize(FileSys::SaveDataType type, u64 title_id, u128 user_id,
                        FileSys::SaveDataSize new_value);
 
-FileSys::RegisteredCacheUnion GetUnionContents();
-
 FileSys::RegisteredCache* GetSystemNANDContents();
 FileSys::RegisteredCache* GetUserNANDContents();
 FileSys::RegisteredCache* GetSDMCContents();

src/core/loader/nso.cpp

@@ -21,6 +21,8 @@
 #include "core/memory.h"
 #include "core/settings.h"
 
+#pragma optimize("", off)
+
 namespace Loader {
 namespace {
 struct MODHeader {
@@ -136,13 +138,13 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::Process& process,
 
     // Apply patches if necessary
     if (pm && (pm->HasNSOPatch(nso_header.build_id) || Settings::values.dump_nso)) {
-        std::vector<u8> pi_header(sizeof(NSOHeader) + program_image.size());
+        std::vector<u8> pi_header;
         pi_header.insert(pi_header.begin(), reinterpret_cast<u8*>(&nso_header),
                          reinterpret_cast<u8*>(&nso_header) + sizeof(NSOHeader));
         pi_header.insert(pi_header.begin() + sizeof(NSOHeader), program_image.begin(),
                          program_image.end());
 
-        pi_header = pm->PatchNSO(pi_header);
+        pi_header = pm->PatchNSO(pi_header, file.GetName());
 
         std::copy(pi_header.begin() + sizeof(NSOHeader), pi_header.end(), program_image.begin());
     }

src/core/memory.cpp

@@ -38,10 +38,6 @@ void SetCurrentPageTable(Common::PageTable* page_table) {
     }
 }
 
-Common::PageTable* GetCurrentPageTable() {
-    return current_page_table;
-}
-
 static void MapPages(Common::PageTable& page_table, VAddr base, u64 size, u8* memory,
                      Common::PageType type) {
     LOG_DEBUG(HW_Memory, "Mapping {} onto {:016X}-{:016X}", fmt::ptr(memory), base * PAGE_SIZE,

src/core/memory.h

@@ -28,16 +28,6 @@ constexpr u64 PAGE_MASK = PAGE_SIZE - 1;
 
 /// Virtual user-space memory regions
 enum : VAddr {
-    /// Read-only page containing kernel and system configuration values.
-    CONFIG_MEMORY_VADDR = 0x1FF80000,
-    CONFIG_MEMORY_SIZE = 0x00001000,
-    CONFIG_MEMORY_VADDR_END = CONFIG_MEMORY_VADDR + CONFIG_MEMORY_SIZE,
-
-    /// Usually read-only page containing mostly values read from hardware.
-    SHARED_PAGE_VADDR = 0x1FF81000,
-    SHARED_PAGE_SIZE = 0x00001000,
-    SHARED_PAGE_VADDR_END = SHARED_PAGE_VADDR + SHARED_PAGE_SIZE,
-
     /// TLS (Thread-Local Storage) related.
     TLS_ENTRY_SIZE = 0x200,
 
@@ -50,9 +40,8 @@ enum : VAddr {
     KERNEL_REGION_END = KERNEL_REGION_VADDR + KERNEL_REGION_SIZE,
 };
 
-/// Currently active page table
+/// Changes the currently active page table.
 void SetCurrentPageTable(Common::PageTable* page_table);
-Common::PageTable* GetCurrentPageTable();
 
 /// Determines if the given VAddr is valid for the specified process.
 bool IsValidVirtualAddress(const Kernel::Process& process, VAddr vaddr);

src/tests/core/arm/arm_test_common.cpp

@@ -17,7 +17,6 @@ TestEnvironment::TestEnvironment(bool mutable_memory_)
     : mutable_memory(mutable_memory_),
       test_memory(std::make_shared<TestMemory>(this)), kernel{Core::System::GetInstance()} {
     auto process = Kernel::Process::Create(Core::System::GetInstance(), "");
-    kernel.MakeCurrentProcess(process.get());
     page_table = &process->VMManager().page_table;
 
     std::fill(page_table->pointers.begin(), page_table->pointers.end(), nullptr);
@@ -28,7 +27,7 @@ TestEnvironment::TestEnvironment(bool mutable_memory_)
     Memory::MapIoRegion(*page_table, 0x00000000, 0x80000000, test_memory);
     Memory::MapIoRegion(*page_table, 0x80000000, 0x80000000, test_memory);
 
-    Memory::SetCurrentPageTable(page_table);
+    kernel.MakeCurrentProcess(process.get());
 }
 
 TestEnvironment::~TestEnvironment() {

src/video_core/CMakeLists.txt

@@ -127,12 +127,14 @@ if (ENABLE_VULKAN)
         renderer_vulkan/vk_sampler_cache.h
         renderer_vulkan/vk_scheduler.cpp
         renderer_vulkan/vk_scheduler.h
+        renderer_vulkan/vk_shader_decompiler.cpp
+        renderer_vulkan/vk_shader_decompiler.h
         renderer_vulkan/vk_stream_buffer.cpp
         renderer_vulkan/vk_stream_buffer.h
         renderer_vulkan/vk_swapchain.cpp
         renderer_vulkan/vk_swapchain.h)
 
-    target_include_directories(video_core PRIVATE ../../externals/Vulkan-Headers/include)
+    target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)
     target_compile_definitions(video_core PRIVATE HAS_VULKAN)
 endif()
 
@@ -140,3 +142,6 @@ create_target_directory_groups(video_core)
 
 target_link_libraries(video_core PUBLIC common core)
 target_link_libraries(video_core PRIVATE glad)
+if (ENABLE_VULKAN)
+    target_link_libraries(video_core PRIVATE sirit)
+endif()

src/video_core/dma_pusher.cpp

@@ -8,6 +8,7 @@
 #include "video_core/dma_pusher.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
 
 namespace Tegra {
 

src/video_core/engines/fermi_2d.cpp

@@ -6,12 +6,13 @@
 #include "common/logging/log.h"
 #include "common/math_util.h"
 #include "video_core/engines/fermi_2d.h"
+#include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 
 namespace Tegra::Engines {
 
 Fermi2D::Fermi2D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager)
-    : memory_manager(memory_manager), rasterizer{rasterizer} {}
+    : rasterizer{rasterizer}, memory_manager{memory_manager} {}
 
 void Fermi2D::CallMethod(const GPU::MethodCall& method_call) {
     ASSERT_MSG(method_call.method < Regs::NUM_REGS,

src/video_core/engines/fermi_2d.h

@@ -10,7 +10,10 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"
-#include "video_core/memory_manager.h"
+
+namespace Tegra {
+class MemoryManager;
+}
 
 namespace VideoCore {
 class RasterizerInterface;
@@ -115,10 +118,9 @@ public:
         };
     } regs{};
 
-    MemoryManager& memory_manager;
-
 private:
     VideoCore::RasterizerInterface& rasterizer;
+    MemoryManager& memory_manager;
 
     /// Performs the copy from the source surface to the destination surface as configured in the
     /// registers.

src/video_core/engines/kepler_compute.h

@@ -9,7 +9,10 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"
-#include "video_core/memory_manager.h"
+
+namespace Tegra {
+class MemoryManager;
+}
 
 namespace Tegra::Engines {
 
@@ -40,10 +43,11 @@ public:
     static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32),
                   "KeplerCompute Regs has wrong size");
 
-    MemoryManager& memory_manager;
-
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
+
+private:
+    MemoryManager& memory_manager;
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \

src/video_core/engines/kepler_memory.cpp

@@ -5,9 +5,9 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "core/memory.h"
 #include "video_core/engines/kepler_memory.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
 
@@ -15,7 +15,7 @@ namespace Tegra::Engines {
 
 KeplerMemory::KeplerMemory(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                            MemoryManager& memory_manager)
-    : system{system}, memory_manager(memory_manager), rasterizer{rasterizer} {}
+    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager} {}
 
 KeplerMemory::~KeplerMemory() = default;
 

src/video_core/engines/kepler_memory.h

@@ -10,12 +10,15 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"
-#include "video_core/memory_manager.h"
 
 namespace Core {
 class System;
 }
 
+namespace Tegra {
+class MemoryManager;
+}
+
 namespace VideoCore {
 class RasterizerInterface;
 }
@@ -82,8 +85,8 @@ public:
 
 private:
     Core::System& system;
-    MemoryManager& memory_manager;
     VideoCore::RasterizerInterface& rasterizer;
+    MemoryManager& memory_manager;
 
     void ProcessData(u32 data);
 };

src/video_core/engines/maxwell_3d.cpp

@@ -7,11 +7,10 @@
 #include "common/assert.h"
 #include "core/core.h"
 #include "core/core_timing.h"
-#include "core/memory.h"
 #include "video_core/debug_utils/debug_utils.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/renderer_base.h"
 #include "video_core/textures/texture.h"
 
 namespace Tegra::Engines {
@@ -21,8 +20,8 @@ constexpr u32 MacroRegistersStart = 0xE00;
 
 Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                      MemoryManager& memory_manager)
-    : memory_manager(memory_manager), system{system}, rasterizer{rasterizer},
-      macro_interpreter(*this) {
+    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager}, macro_interpreter{
+                                                                                  *this} {
     InitializeRegisterDefaults();
 }
 
@@ -250,6 +249,10 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
         ProcessQueryGet();
         break;
     }
+    case MAXWELL3D_REG_INDEX(sync_info): {
+        ProcessSyncPoint();
+        break;
+    }
     default:
         break;
     }
@@ -327,6 +330,14 @@ void Maxwell3D::ProcessQueryGet() {
     }
 }
 
+void Maxwell3D::ProcessSyncPoint() {
+    const u32 sync_point = regs.sync_info.sync_point.Value();
+    const u32 increment = regs.sync_info.increment.Value();
+    const u32 cache_flush = regs.sync_info.unknown.Value();
+    LOG_DEBUG(HW_GPU, "Syncpoint set {}, increment: {}, unk: {}", sync_point, increment,
+              cache_flush);
+}
+
 void Maxwell3D::DrawArrays() {
     LOG_DEBUG(HW_GPU, "called, topology={}, count={}", static_cast<u32>(regs.draw.topology.Value()),
               regs.vertex_buffer.count);

src/video_core/engines/maxwell_3d.h

@@ -16,13 +16,16 @@
 #include "common/math_util.h"
 #include "video_core/gpu.h"
 #include "video_core/macro_interpreter.h"
-#include "video_core/memory_manager.h"
 #include "video_core/textures/texture.h"
 
 namespace Core {
 class System;
 }
 
+namespace Tegra {
+class MemoryManager;
+}
+
 namespace VideoCore {
 class RasterizerInterface;
 }
@@ -576,7 +579,17 @@ public:
                     u32 bind;
                 } macros;
 
-                INSERT_PADDING_WORDS(0x188);
+                INSERT_PADDING_WORDS(0x69);
+
+                struct {
+                    union {
+                        BitField<0, 16, u32> sync_point;
+                        BitField<16, 1, u32> unknown;
+                        BitField<20, 1, u32> increment;
+                    };
+                } sync_info;
+
+                INSERT_PADDING_WORDS(0x11E);
 
                 u32 tfb_enabled;
 
@@ -1093,7 +1106,6 @@ public:
     };
 
     State state{};
-    MemoryManager& memory_manager;
 
     struct DirtyFlags {
         std::bitset<8> color_buffer{0xFF};
@@ -1141,6 +1153,8 @@ private:
 
     VideoCore::RasterizerInterface& rasterizer;
 
+    MemoryManager& memory_manager;
+
     /// Start offsets of each macro in macro_memory
     std::unordered_map<u32, u32> macro_offsets;
 
@@ -1180,6 +1194,9 @@ private:
     /// Handles a write to the QUERY_GET register.
     void ProcessQueryGet();
 
+    /// Handles writes to syncing register.
+    void ProcessSyncPoint();
+
     /// Handles a write to the CB_DATA[i] register.
     void ProcessCBData(u32 value);
 
@@ -1195,6 +1212,7 @@ private:
                   "Field " #field_name " has invalid position")
 
 ASSERT_REG_POSITION(macros, 0x45);
+ASSERT_REG_POSITION(sync_info, 0xB2);
 ASSERT_REG_POSITION(tfb_enabled, 0x1D1);
 ASSERT_REG_POSITION(rt, 0x200);
 ASSERT_REG_POSITION(viewport_transform, 0x280);

src/video_core/engines/maxwell_dma.cpp

@@ -5,9 +5,9 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "core/memory.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/maxwell_dma.h"
+#include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
 #include "video_core/textures/decoders.h"
@@ -16,7 +16,7 @@ namespace Tegra::Engines {
 
 MaxwellDMA::MaxwellDMA(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                        MemoryManager& memory_manager)
-    : memory_manager(memory_manager), system{system}, rasterizer{rasterizer} {}
+    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager} {}
 
 void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) {
     ASSERT_MSG(method_call.method < Regs::NUM_REGS,

src/video_core/engines/maxwell_dma.h

@@ -10,12 +10,15 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"
-#include "video_core/memory_manager.h"
 
 namespace Core {
 class System;
 }
 
+namespace Tegra {
+class MemoryManager;
+}
+
 namespace VideoCore {
 class RasterizerInterface;
 }
@@ -139,13 +142,13 @@ public:
         };
     } regs{};
 
-    MemoryManager& memory_manager;
-
 private:
     Core::System& system;
 
     VideoCore::RasterizerInterface& rasterizer;
 
+    MemoryManager& memory_manager;
+
     /// Performs the copy from the source buffer to the destination buffer as configured in the
     /// registers.
     void HandleCopy();

src/video_core/engines/shader_bytecode.h

@@ -1238,13 +1238,16 @@ union Instruction {
 
     union {
         BitField<20, 16, u64> imm20_16;
+        BitField<35, 1, u64> high_b_rr; // used on RR
         BitField<36, 1, u64> product_shift_left;
         BitField<37, 1, u64> merge_37;
         BitField<48, 1, u64> sign_a;
         BitField<49, 1, u64> sign_b;
+        BitField<50, 2, XmadMode> mode_cbf; // used by CR, RC
         BitField<50, 3, XmadMode> mode;
         BitField<52, 1, u64> high_b;
         BitField<53, 1, u64> high_a;
+        BitField<55, 1, u64> product_shift_left_second; // used on CR
         BitField<56, 1, u64> merge_56;
     } xmad;
 
@@ -1662,7 +1665,7 @@ private:
             INST("0011011-11110---", Id::BFI_IMM_R, Type::Bfi, "BFI_IMM_R"),
             INST("0100110001000---", Id::LOP_C, Type::ArithmeticInteger, "LOP_C"),
             INST("0101110001000---", Id::LOP_R, Type::ArithmeticInteger, "LOP_R"),
-            INST("0011100001000---", Id::LOP_IMM, Type::ArithmeticInteger, "LOP_IMM"),
+            INST("0011100-01000---", Id::LOP_IMM, Type::ArithmeticInteger, "LOP_IMM"),
             INST("000001----------", Id::LOP32I, Type::ArithmeticIntegerImmediate, "LOP32I"),
             INST("0000001---------", Id::LOP3_C, Type::ArithmeticInteger, "LOP3_C"),
             INST("0101101111100---", Id::LOP3_R, Type::ArithmeticInteger, "LOP3_R"),

src/video_core/gpu.cpp

@@ -31,7 +31,7 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
 
 GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer) : renderer{renderer} {
     auto& rasterizer{renderer.Rasterizer()};
-    memory_manager = std::make_unique<Tegra::MemoryManager>();
+    memory_manager = std::make_unique<Tegra::MemoryManager>(rasterizer);
     dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
     maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager);
     fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer, *memory_manager);

src/video_core/memory_manager.cpp

@@ -5,16 +5,13 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
-#include "core/core.h"
 #include "core/memory.h"
-#include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/renderer_base.h"
 
 namespace Tegra {
 
-MemoryManager::MemoryManager() {
+MemoryManager::MemoryManager(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {
     std::fill(page_table.pointers.begin(), page_table.pointers.end(), nullptr);
     std::fill(page_table.attributes.begin(), page_table.attributes.end(),
               Common::PageType::Unmapped);
@@ -70,8 +67,7 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) {
     const u64 aligned_size{Common::AlignUp(size, page_size)};
     const CacheAddr cache_addr{ToCacheAddr(GetPointer(gpu_addr))};
 
-    Core::System::GetInstance().Renderer().Rasterizer().FlushAndInvalidateRegion(cache_addr,
-                                                                                 aligned_size);
+    rasterizer.FlushAndInvalidateRegion(cache_addr, aligned_size);
     UnmapRange(gpu_addr, aligned_size);
 
     return gpu_addr;
@@ -204,14 +200,85 @@ const u8* MemoryManager::GetPointer(GPUVAddr addr) const {
 }
 
 void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, std::size_t size) const {
-    std::memcpy(dest_buffer, GetPointer(src_addr), size);
+    std::size_t remaining_size{size};
+    std::size_t page_index{src_addr >> page_bits};
+    std::size_t page_offset{src_addr & page_mask};
+
+    while (remaining_size > 0) {
+        const std::size_t copy_amount{
+            std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
+
+        switch (page_table.attributes[page_index]) {
+        case Common::PageType::Memory: {
+            const u8* src_ptr{page_table.pointers[page_index] + page_offset};
+            rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount);
+            std::memcpy(dest_buffer, src_ptr, copy_amount);
+            break;
+        }
+        default:
+            UNREACHABLE();
+        }
+
+        page_index++;
+        page_offset = 0;
+        dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
+        remaining_size -= copy_amount;
+    }
 }
+
 void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, std::size_t size) {
-    std::memcpy(GetPointer(dest_addr), src_buffer, size);
+    std::size_t remaining_size{size};
+    std::size_t page_index{dest_addr >> page_bits};
+    std::size_t page_offset{dest_addr & page_mask};
+
+    while (remaining_size > 0) {
+        const std::size_t copy_amount{
+            std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
+
+        switch (page_table.attributes[page_index]) {
+        case Common::PageType::Memory: {
+            u8* dest_ptr{page_table.pointers[page_index] + page_offset};
+            rasterizer.InvalidateRegion(ToCacheAddr(dest_ptr), copy_amount);
+            std::memcpy(dest_ptr, src_buffer, copy_amount);
+            break;
+        }
+        default:
+            UNREACHABLE();
+        }
+
+        page_index++;
+        page_offset = 0;
+        src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
+        remaining_size -= copy_amount;
+    }
 }
 
 void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size) {
-    std::memcpy(GetPointer(dest_addr), GetPointer(src_addr), size);
+    std::size_t remaining_size{size};
+    std::size_t page_index{src_addr >> page_bits};
+    std::size_t page_offset{src_addr & page_mask};
+
+    while (remaining_size > 0) {
+        const std::size_t copy_amount{
+            std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
+
+        switch (page_table.attributes[page_index]) {
+        case Common::PageType::Memory: {
+            const u8* src_ptr{page_table.pointers[page_index] + page_offset};
+            rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount);
+            WriteBlock(dest_addr, src_ptr, copy_amount);
+            break;
+        }
+        default:
+            UNREACHABLE();
+        }
+
+        page_index++;
+        page_offset = 0;
+        dest_addr += static_cast<VAddr>(copy_amount);
+        src_addr += static_cast<VAddr>(copy_amount);
+        remaining_size -= copy_amount;
+    }
 }
 
 void MemoryManager::MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type,
@@ -351,7 +418,7 @@ MemoryManager::VMAIter MemoryManager::CarveVMA(GPUVAddr base, u64 size) {
     const VirtualMemoryArea& vma{vma_handle->second};
     if (vma.type == VirtualMemoryArea::Type::Mapped) {
         // Region is already allocated
-        return {};
+        return vma_handle;
     }
 
     const VAddr start_in_vma{base - vma.base};

src/video_core/memory_manager.h

@@ -10,6 +10,10 @@
 #include "common/common_types.h"
 #include "common/page_table.h"
 
+namespace VideoCore {
+class RasterizerInterface;
+}
+
 namespace Tegra {
 
 /**
@@ -43,7 +47,7 @@ struct VirtualMemoryArea {
 
 class MemoryManager final {
 public:
-    MemoryManager();
+    MemoryManager(VideoCore::RasterizerInterface& rasterizer);
 
     GPUVAddr AllocateSpace(u64 size, u64 align);
     GPUVAddr AllocateSpace(GPUVAddr addr, u64 size, u64 align);
@@ -144,6 +148,7 @@ private:
 
     Common::PageTable page_table{page_bits};
     VMAMap vma_map;
+    VideoCore::RasterizerInterface& rasterizer;
 };
 
 } // namespace Tegra

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -7,6 +7,7 @@
 
 #include "common/alignment.h"
 #include "core/core.h"
+#include "video_core/memory_manager.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 

src/video_core/renderer_opengl/gl_global_cache.cpp

@@ -6,6 +6,7 @@
 
 #include "common/logging/log.h"
 #include "core/core.h"
+#include "video_core/memory_manager.h"
 #include "video_core/renderer_opengl/gl_global_cache.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -299,6 +299,10 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     BaseBindings base_bindings;
     std::array<bool, Maxwell::NumClipDistances> clip_distances{};
 
+    // Prepare packed bindings
+    bind_ubo_pushbuffer.Setup(base_bindings.cbuf);
+    bind_ssbo_pushbuffer.Setup(base_bindings.gmem);
+
     for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
         const auto& shader_config = gpu.regs.shader_config[index];
         const Maxwell::ShaderProgram program{static_cast<Maxwell::ShaderProgram>(index)};
@@ -321,8 +325,8 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
             &ubo, sizeof(ubo), static_cast<std::size_t>(uniform_buffer_alignment));
 
         // Bind the emulation info buffer
-        glBindBufferRange(GL_UNIFORM_BUFFER, base_bindings.cbuf, buffer_cache.GetHandle(), offset,
-                          static_cast<GLsizeiptr>(sizeof(ubo)));
+        bind_ubo_pushbuffer.Push(buffer_cache.GetHandle(), offset,
+                                 static_cast<GLsizeiptr>(sizeof(ubo)));
 
         Shader shader{shader_cache.GetStageProgram(program)};
         const auto [program_handle, next_bindings] =
@@ -366,6 +370,9 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
         base_bindings = next_bindings;
     }
 
+    bind_ubo_pushbuffer.Bind();
+    bind_ssbo_pushbuffer.Bind();
+
     SyncClipEnabled(clip_distances);
 
     gpu.dirty_flags.shaders = false;
@@ -900,23 +907,14 @@ void RasterizerOpenGL::SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::Shader
     const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
     const auto& entries = shader->GetShaderEntries().const_buffers;
 
-    constexpr u64 max_binds = Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers;
-    std::array<GLuint, max_binds> bind_buffers;
-    std::array<GLintptr, max_binds> bind_offsets;
-    std::array<GLsizeiptr, max_binds> bind_sizes;
-
-    ASSERT_MSG(entries.size() <= max_binds, "Exceeded expected number of binding points.");
-
     // Upload only the enabled buffers from the 16 constbuffers of each shader stage
     for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
         const auto& used_buffer = entries[bindpoint];
         const auto& buffer = shader_stage.const_buffers[used_buffer.GetIndex()];
 
         if (!buffer.enabled) {
-            // With disabled buffers set values as zero to unbind them
-            bind_buffers[bindpoint] = 0;
-            bind_offsets[bindpoint] = 0;
-            bind_sizes[bindpoint] = 0;
+            // Set values to zero to unbind buffers
+            bind_ubo_pushbuffer.Push(0, 0, 0);
             continue;
         }
 
@@ -944,30 +942,19 @@ void RasterizerOpenGL::SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::Shader
         const GLintptr const_buffer_offset = buffer_cache.UploadMemory(
             buffer.address, size, static_cast<std::size_t>(uniform_buffer_alignment));
 
-        // Prepare values for multibind
-        bind_buffers[bindpoint] = buffer_cache.GetHandle();
-        bind_offsets[bindpoint] = const_buffer_offset;
-        bind_sizes[bindpoint] = size;
+        bind_ubo_pushbuffer.Push(buffer_cache.GetHandle(), const_buffer_offset, size);
     }
-
-    // The first binding is reserved for emulation values
-    const GLuint ubo_base_binding = base_bindings.cbuf + 1;
-    glBindBuffersRange(GL_UNIFORM_BUFFER, ubo_base_binding, static_cast<GLsizei>(entries.size()),
-                       bind_buffers.data(), bind_offsets.data(), bind_sizes.data());
 }
 
 void RasterizerOpenGL::SetupGlobalRegions(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
                                           const Shader& shader, GLenum primitive_mode,
                                           BaseBindings base_bindings) {
-    // TODO(Rodrigo): Use ARB_multi_bind here
     const auto& entries = shader->GetShaderEntries().global_memory_entries;
-
-    for (u32 bindpoint = 0; bindpoint < static_cast<u32>(entries.size()); ++bindpoint) {
-        const auto& entry = entries[bindpoint];
-        const u32 current_bindpoint = base_bindings.gmem + bindpoint;
-        const auto& region = global_cache.GetGlobalRegion(entry, stage);
-
-        glBindBufferBase(GL_SHADER_STORAGE_BUFFER, current_bindpoint, region->GetBufferHandle());
+    for (std::size_t bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
+        const auto& entry{entries[bindpoint]};
+        const auto& region{global_cache.GetGlobalRegion(entry, stage)};
+        bind_ssbo_pushbuffer.Push(region->GetBufferHandle(), 0,
+                                  static_cast<GLsizeiptr>(region->GetSizeInBytes()));
     }
 }
 

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -28,6 +28,7 @@
 #include "video_core/renderer_opengl/gl_shader_cache.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/utils.h"
 
 namespace Core {
 class System;
@@ -229,6 +230,9 @@ private:
     PrimitiveAssembler primitive_assembler{buffer_cache};
     GLint uniform_buffer_alignment;
 
+    BindBuffersRangePushBuffer bind_ubo_pushbuffer{GL_UNIFORM_BUFFER};
+    BindBuffersRangePushBuffer bind_ssbo_pushbuffer{GL_SHADER_STORAGE_BUFFER};
+
     std::size_t CalculateVertexArraysSize() const;
 
     std::size_t CalculateIndexBufferSize() const;

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -15,6 +15,7 @@
 #include "core/hle/kernel/process.h"
 #include "core/settings.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/memory_manager.h"
 #include "video_core/morton.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"
@@ -661,8 +662,8 @@ void CachedSurface::FlushGLBuffer() {
     gl_buffer[0].resize(GetSizeInBytes());
 
     const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
-    // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
-    ASSERT(params.width * GetBytesPerPixel(params.pixel_format) % 4 == 0);
+    const u32 align = std::clamp(params.RowAlign(0), 1U, 8U);
+    glPixelStorei(GL_PACK_ALIGNMENT, align);
     glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.width));
     ASSERT(!tuple.compressed);
     glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
@@ -707,8 +708,8 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
 
     const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
 
-    // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
-    ASSERT(params.MipWidth(mip_map) * GetBytesPerPixel(params.pixel_format) % 4 == 0);
+    const u32 align = std::clamp(params.RowAlign(mip_map), 1U, 8U);
+    glPixelStorei(GL_UNPACK_ALIGNMENT, align);
     glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.MipWidth(mip_map)));
 
     const auto image_size = static_cast<GLsizei>(params.GetMipmapSizeGL(mip_map, false));

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -11,6 +11,7 @@
 #include <vector>
 
 #include "common/alignment.h"
+#include "common/bit_util.h"
 #include "common/common_types.h"
 #include "common/hash.h"
 #include "common/math_util.h"
@@ -205,6 +206,13 @@ struct SurfaceParams {
         return bd;
     }
 
+    u32 RowAlign(u32 mip_level) const {
+        const u32 m_width = MipWidth(mip_level);
+        const u32 bytes_per_pixel = GetBytesPerPixel(pixel_format);
+        const u32 l2 = Common::CountTrailingZeroes32(m_width * bytes_per_pixel);
+        return (1U << l2);
+    }
+
     /// Creates SurfaceParams from a texture configuration
     static SurfaceParams CreateForTexture(const Tegra::Texture::FullTextureInfo& config,
                                           const GLShader::SamplerEntry& entry);

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -7,6 +7,7 @@
 #include "common/hash.h"
 #include "core/core.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/memory_manager.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_cache.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
@@ -39,6 +40,10 @@ GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) {
 /// Gets the shader program code from memory for the specified address
 ProgramCode GetShaderCode(const u8* host_ptr) {
     ProgramCode program_code(VideoCommon::Shader::MAX_PROGRAM_LENGTH);
+    ASSERT_OR_EXECUTE(host_ptr != nullptr, {
+        std::fill(program_code.begin(), program_code.end(), 0);
+        return program_code;
+    });
     std::memcpy(program_code.data(), host_ptr, program_code.size() * sizeof(u64));
     return program_code;
 }

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -21,6 +21,8 @@
 
 namespace OpenGL::GLShader {
 
+namespace {
+
 using Tegra::Shader::Attribute;
 using Tegra::Shader::AttributeUse;
 using Tegra::Shader::Header;
@@ -34,14 +36,18 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 using ShaderStage = Tegra::Engines::Maxwell3D::Regs::ShaderStage;
 using Operation = const OperationNode&;
 
+enum class Type { Bool, Bool2, Float, Int, Uint, HalfFloat };
+
+struct TextureAoffi {};
+using TextureArgument = std::pair<Type, Node>;
+using TextureIR = std::variant<TextureAoffi, TextureArgument>;
+
 enum : u32 { POSITION_VARYING_LOCATION = 0, GENERIC_VARYING_START_LOCATION = 1 };
 constexpr u32 MAX_CONSTBUFFER_ELEMENTS =
     static_cast<u32>(RasterizerOpenGL::MaxConstbufferSize) / (4 * sizeof(float));
 constexpr u32 MAX_GLOBALMEMORY_ELEMENTS =
     static_cast<u32>(RasterizerOpenGL::MaxGlobalMemorySize) / sizeof(float);
 
-enum class Type { Bool, Bool2, Float, Int, Uint, HalfFloat };
-
 class ShaderWriter {
 public:
     void AddExpression(std::string_view text) {
@@ -91,7 +97,7 @@ private:
 };
 
 /// Generates code to use for a swizzle operation.
-static std::string GetSwizzle(u32 elem) {
+std::string GetSwizzle(u32 elem) {
     ASSERT(elem <= 3);
     std::string swizzle = ".";
     swizzle += "xyzw"[elem];
@@ -99,7 +105,7 @@ static std::string GetSwizzle(u32 elem) {
 }
 
 /// Translate topology
-static std::string GetTopologyName(Tegra::Shader::OutputTopology topology) {
+std::string GetTopologyName(Tegra::Shader::OutputTopology topology) {
     switch (topology) {
     case Tegra::Shader::OutputTopology::PointList:
         return "points";
@@ -114,7 +120,7 @@ static std::string GetTopologyName(Tegra::Shader::OutputTopology topology) {
 }
 
 /// Returns true if an object has to be treated as precise
-static bool IsPrecise(Operation operand) {
+bool IsPrecise(Operation operand) {
     const auto& meta = operand.GetMeta();
 
     if (const auto arithmetic = std::get_if<MetaArithmetic>(&meta)) {
@@ -126,7 +132,7 @@ static bool IsPrecise(Operation operand) {
     return false;
 }
 
-static bool IsPrecise(Node node) {
+bool IsPrecise(Node node) {
     if (const auto operation = std::get_if<OperationNode>(node)) {
         return IsPrecise(*operation);
     }
@@ -546,8 +552,7 @@ private:
             } else if (std::holds_alternative<OperationNode>(*offset)) {
                 // Indirect access
                 const std::string final_offset = code.GenerateTemporary();
-                code.AddLine("uint " + final_offset + " = (ftou(" + Visit(offset) + ") / 4) & " +
-                             std::to_string(MAX_CONSTBUFFER_ELEMENTS - 1) + ';');
+                code.AddLine("uint " + final_offset + " = (ftou(" + Visit(offset) + ") / 4);");
                 return fmt::format("{}[{} / 4][{} % 4]", GetConstBuffer(cbuf->GetIndex()),
                                    final_offset, final_offset);
 
@@ -723,8 +728,8 @@ private:
                                                          result_type));
     }
 
-    std::string GenerateTexture(Operation operation, const std::string& func,
-                                const std::vector<std::pair<Type, Node>>& extras) {
+    std::string GenerateTexture(Operation operation, const std::string& function_suffix,
+                                const std::vector<TextureIR>& extras) {
         constexpr std::array<const char*, 4> coord_constructors = {"float", "vec2", "vec3", "vec4"};
 
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
@@ -734,11 +739,11 @@ private:
         const bool has_array = meta->sampler.IsArray();
         const bool has_shadow = meta->sampler.IsShadow();
 
-        std::string expr = func;
-        expr += '(';
-        expr += GetSampler(meta->sampler);
-        expr += ", ";
-
+        std::string expr = "texture" + function_suffix;
+        if (!meta->aoffi.empty()) {
+            expr += "Offset";
+        }
+        expr += '(' + GetSampler(meta->sampler) + ", ";
         expr += coord_constructors.at(count + (has_array ? 1 : 0) + (has_shadow ? 1 : 0) - 1);
         expr += '(';
         for (std::size_t i = 0; i < count; ++i) {
@@ -756,38 +761,76 @@ private:
         }
         expr += ')';
 
-        for (const auto& extra_pair : extras) {
-            const auto [type, operand] = extra_pair;
-            if (operand == nullptr) {
-                continue;
-            }
-            expr += ", ";
-
-            switch (type) {
-            case Type::Int:
-                if (const auto immediate = std::get_if<ImmediateNode>(operand)) {
-                    // Inline the string as an immediate integer in GLSL (some extra arguments are
-                    // required to be constant)
-                    expr += std::to_string(static_cast<s32>(immediate->GetValue()));
-                } else {
-                    expr += "ftoi(" + Visit(operand) + ')';
-                }
-                break;
-            case Type::Float:
-                expr += Visit(operand);
-                break;
-            default: {
-                const auto type_int = static_cast<u32>(type);
-                UNIMPLEMENTED_MSG("Unimplemented extra type={}", type_int);
-                expr += '0';
-                break;
-            }
+        for (const auto& variant : extras) {
+            if (const auto argument = std::get_if<TextureArgument>(&variant)) {
+                expr += GenerateTextureArgument(*argument);
+            } else if (std::get_if<TextureAoffi>(&variant)) {
+                expr += GenerateTextureAoffi(meta->aoffi);
+            } else {
+                UNREACHABLE();
             }
         }
 
         return expr + ')';
     }
 
+    std::string GenerateTextureArgument(TextureArgument argument) {
+        const auto [type, operand] = argument;
+        if (operand == nullptr) {
+            return {};
+        }
+
+        std::string expr = ", ";
+        switch (type) {
+        case Type::Int:
+            if (const auto immediate = std::get_if<ImmediateNode>(operand)) {
+                // Inline the string as an immediate integer in GLSL (some extra arguments are
+                // required to be constant)
+                expr += std::to_string(static_cast<s32>(immediate->GetValue()));
+            } else {
+                expr += "ftoi(" + Visit(operand) + ')';
+            }
+            break;
+        case Type::Float:
+            expr += Visit(operand);
+            break;
+        default: {
+            const auto type_int = static_cast<u32>(type);
+            UNIMPLEMENTED_MSG("Unimplemented extra type={}", type_int);
+            expr += '0';
+            break;
+        }
+        }
+        return expr;
+    }
+
+    std::string GenerateTextureAoffi(const std::vector<Node>& aoffi) {
+        if (aoffi.empty()) {
+            return {};
+        }
+        constexpr std::array<const char*, 3> coord_constructors = {"int", "ivec2", "ivec3"};
+        std::string expr = ", ";
+        expr += coord_constructors.at(aoffi.size() - 1);
+        expr += '(';
+
+        for (std::size_t index = 0; index < aoffi.size(); ++index) {
+            const auto operand{aoffi.at(index)};
+            if (const auto immediate = std::get_if<ImmediateNode>(operand)) {
+                // Inline the string as an immediate integer in GLSL (AOFFI arguments are required
+                // to be constant by the standard).
+                expr += std::to_string(static_cast<s32>(immediate->GetValue()));
+            } else {
+                expr += "ftoi(" + Visit(operand) + ')';
+            }
+            if (index + 1 < aoffi.size()) {
+                expr += ", ";
+            }
+        }
+        expr += ')';
+
+        return expr;
+    }
+
     std::string Assign(Operation operation) {
         const Node dest = operation[0];
         const Node src = operation[1];
@@ -1164,7 +1207,8 @@ private:
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        std::string expr = GenerateTexture(operation, "texture", {{Type::Float, meta->bias}});
+        std::string expr = GenerateTexture(
+            operation, "", {TextureAoffi{}, TextureArgument{Type::Float, meta->bias}});
         if (meta->sampler.IsShadow()) {
             expr = "vec4(" + expr + ')';
         }
@@ -1175,7 +1219,8 @@ private:
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        std::string expr = GenerateTexture(operation, "textureLod", {{Type::Float, meta->lod}});
+        std::string expr = GenerateTexture(
+            operation, "Lod", {TextureArgument{Type::Float, meta->lod}, TextureAoffi{}});
         if (meta->sampler.IsShadow()) {
             expr = "vec4(" + expr + ')';
         }
@@ -1187,7 +1232,8 @@ private:
         ASSERT(meta);
 
         const auto type = meta->sampler.IsShadow() ? Type::Float : Type::Int;
-        return GenerateTexture(operation, "textureGather", {{type, meta->component}}) +
+        return GenerateTexture(operation, "Gather",
+                               {TextureArgument{type, meta->component}, TextureAoffi{}}) +
                GetSwizzle(meta->element);
     }
 
@@ -1217,8 +1263,8 @@ private:
         ASSERT(meta);
 
         if (meta->element < 2) {
-            return "itof(int((" + GenerateTexture(operation, "textureQueryLod", {}) +
-                   " * vec2(256))" + GetSwizzle(meta->element) + "))";
+            return "itof(int((" + GenerateTexture(operation, "QueryLod", {}) + " * vec2(256))" +
+                   GetSwizzle(meta->element) + "))";
         }
         return "0";
     }
@@ -1571,6 +1617,8 @@ private:
     ShaderWriter code;
 };
 
+} // Anonymous namespace
+
 std::string GetCommonDeclarations() {
     const auto cbuf = std::to_string(MAX_CONSTBUFFER_ELEMENTS);
     const auto gmem = std::to_string(MAX_GLOBALMEMORY_ELEMENTS);

src/video_core/renderer_opengl/gl_shader_disk_cache.cpp

@@ -10,8 +10,8 @@
 #include "common/common_types.h"
 #include "common/file_util.h"
 #include "common/logging/log.h"
-#include "common/lz4_compression.h"
 #include "common/scm_rev.h"
+#include "common/zstd_compression.h"
 
 #include "core/core.h"
 #include "core/hle/kernel/process.h"
@@ -259,7 +259,7 @@ ShaderDiskCacheOpenGL::LoadPrecompiledFile(FileUtil::IOFile& file) {
                 return {};
             }
 
-            dump.binary = Common::Compression::DecompressDataLZ4(compressed_binary, binary_length);
+            dump.binary = Common::Compression::DecompressDataZSTD(compressed_binary);
             if (dump.binary.empty()) {
                 return {};
             }
@@ -288,7 +288,7 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         return {};
     }
 
-    const std::vector<u8> code = Common::Compression::DecompressDataLZ4(compressed_code, code_size);
+    const std::vector<u8> code = Common::Compression::DecompressDataZSTD(compressed_code);
     if (code.empty()) {
         return {};
     }
@@ -474,8 +474,8 @@ void ShaderDiskCacheOpenGL::SaveDecompiled(u64 unique_identifier, const std::str
     if (!IsUsable())
         return;
 
-    const std::vector<u8> compressed_code{Common::Compression::CompressDataLZ4HC(
-        reinterpret_cast<const u8*>(code.data()), code.size(), 9)};
+    const std::vector<u8> compressed_code{Common::Compression::CompressDataZSTDDefault(
+        reinterpret_cast<const u8*>(code.data()), code.size())};
     if (compressed_code.empty()) {
         LOG_ERROR(Render_OpenGL, "Failed to compress GLSL code - skipping shader {:016x}",
                   unique_identifier);
@@ -506,7 +506,7 @@ void ShaderDiskCacheOpenGL::SaveDump(const ShaderDiskCacheUsage& usage, GLuint p
     glGetProgramBinary(program, binary_length, nullptr, &binary_format, binary.data());
 
     const std::vector<u8> compressed_binary =
-        Common::Compression::CompressDataLZ4HC(binary.data(), binary.size(), 9);
+        Common::Compression::CompressDataZSTDDefault(binary.data(), binary.size());
 
     if (compressed_binary.empty()) {
         LOG_ERROR(Render_OpenGL, "Failed to compress binary program in shader={:016x}",

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -62,7 +62,6 @@ public:
         UpdatePipeline();
         state.draw.shader_program = 0;
         state.draw.program_pipeline = pipeline.handle;
-        state.geometry_shaders.enabled = (gs != 0);
     }
 
 private:

src/video_core/renderer_opengl/gl_state.cpp

@@ -10,16 +10,62 @@
 
 namespace OpenGL {
 
-OpenGLState OpenGLState::cur_state;
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
+OpenGLState OpenGLState::cur_state;
 bool OpenGLState::s_rgb_used;
 
+namespace {
+
+template <typename T>
+bool UpdateValue(T& current_value, const T new_value) {
+    const bool changed = current_value != new_value;
+    current_value = new_value;
+    return changed;
+}
+
+template <typename T1, typename T2>
+bool UpdateTie(T1 current_value, const T2 new_value) {
+    const bool changed = current_value != new_value;
+    current_value = new_value;
+    return changed;
+}
+
+void Enable(GLenum cap, bool enable) {
+    if (enable) {
+        glEnable(cap);
+    } else {
+        glDisable(cap);
+    }
+}
+
+void Enable(GLenum cap, GLuint index, bool enable) {
+    if (enable) {
+        glEnablei(cap, index);
+    } else {
+        glDisablei(cap, index);
+    }
+}
+
+void Enable(GLenum cap, bool& current_value, bool new_value) {
+    if (UpdateValue(current_value, new_value))
+        Enable(cap, new_value);
+}
+
+void Enable(GLenum cap, GLuint index, bool& current_value, bool new_value) {
+    if (UpdateValue(current_value, new_value))
+        Enable(cap, index, new_value);
+}
+
+} // namespace
+
 OpenGLState::OpenGLState() {
     // These all match default OpenGL values
-    geometry_shaders.enabled = false;
     framebuffer_srgb.enabled = false;
+
     multisample_control.alpha_to_coverage = false;
     multisample_control.alpha_to_one = false;
+
     cull.enabled = false;
     cull.mode = GL_BACK;
     cull.front_face = GL_CCW;
@@ -30,14 +76,15 @@ OpenGLState::OpenGLState() {
 
     primitive_restart.enabled = false;
     primitive_restart.index = 0;
+
     for (auto& item : color_mask) {
         item.red_enabled = GL_TRUE;
         item.green_enabled = GL_TRUE;
         item.blue_enabled = GL_TRUE;
         item.alpha_enabled = GL_TRUE;
     }
-    stencil.test_enabled = false;
-    auto reset_stencil = [](auto& config) {
+
+    const auto ResetStencil = [](auto& config) {
         config.test_func = GL_ALWAYS;
         config.test_ref = 0;
         config.test_mask = 0xFFFFFFFF;
@@ -46,8 +93,10 @@ OpenGLState::OpenGLState() {
         config.action_depth_pass = GL_KEEP;
         config.action_stencil_fail = GL_KEEP;
     };
-    reset_stencil(stencil.front);
-    reset_stencil(stencil.back);
+    stencil.test_enabled = false;
+    ResetStencil(stencil.front);
+    ResetStencil(stencil.back);
+
     for (auto& item : viewports) {
         item.x = 0;
         item.y = 0;
@@ -61,6 +110,7 @@ OpenGLState::OpenGLState() {
         item.scissor.width = 0;
         item.scissor.height = 0;
     }
+
     for (auto& item : blend) {
         item.enabled = true;
         item.rgb_equation = GL_FUNC_ADD;
@@ -70,11 +120,14 @@ OpenGLState::OpenGLState() {
         item.src_a_func = GL_ONE;
         item.dst_a_func = GL_ZERO;
     }
+
     independant_blend.enabled = false;
+
     blend_color.red = 0.0f;
     blend_color.green = 0.0f;
     blend_color.blue = 0.0f;
     blend_color.alpha = 0.0f;
+
     logic_op.enabled = false;
     logic_op.operation = GL_COPY;
 
@@ -91,9 +144,12 @@ OpenGLState::OpenGLState() {
     clip_distance = {};
 
     point.size = 1;
+
     fragment_color_clamp.enabled = false;
+
     depth_clamp.far_plane = false;
     depth_clamp.near_plane = false;
+
     polygon_offset.fill_enable = false;
     polygon_offset.line_enable = false;
     polygon_offset.point_enable = false;
@@ -103,132 +159,380 @@ OpenGLState::OpenGLState() {
 }
 
 void OpenGLState::ApplyDefaultState() {
+    glEnable(GL_BLEND);
     glDisable(GL_FRAMEBUFFER_SRGB);
     glDisable(GL_CULL_FACE);
     glDisable(GL_DEPTH_TEST);
     glDisable(GL_PRIMITIVE_RESTART);
     glDisable(GL_STENCIL_TEST);
-    glEnable(GL_BLEND);
     glDisable(GL_COLOR_LOGIC_OP);
     glDisable(GL_SCISSOR_TEST);
 }
 
+void OpenGLState::ApplyFramebufferState() const {
+    if (UpdateValue(cur_state.draw.read_framebuffer, draw.read_framebuffer)) {
+        glBindFramebuffer(GL_READ_FRAMEBUFFER, draw.read_framebuffer);
+    }
+    if (UpdateValue(cur_state.draw.draw_framebuffer, draw.draw_framebuffer)) {
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, draw.draw_framebuffer);
+    }
+}
+
+void OpenGLState::ApplyVertexArrayState() const {
+    if (UpdateValue(cur_state.draw.vertex_array, draw.vertex_array)) {
+        glBindVertexArray(draw.vertex_array);
+    }
+}
+
+void OpenGLState::ApplyShaderProgram() const {
+    if (UpdateValue(cur_state.draw.shader_program, draw.shader_program)) {
+        glUseProgram(draw.shader_program);
+    }
+}
+
+void OpenGLState::ApplyProgramPipeline() const {
+    if (UpdateValue(cur_state.draw.program_pipeline, draw.program_pipeline)) {
+        glBindProgramPipeline(draw.program_pipeline);
+    }
+}
+
+void OpenGLState::ApplyClipDistances() const {
+    for (std::size_t i = 0; i < clip_distance.size(); ++i) {
+        Enable(GL_CLIP_DISTANCE0 + static_cast<GLenum>(i), cur_state.clip_distance[i],
+               clip_distance[i]);
+    }
+}
+
+void OpenGLState::ApplyPointSize() const {
+    if (UpdateValue(cur_state.point.size, point.size)) {
+        glPointSize(point.size);
+    }
+}
+
+void OpenGLState::ApplyFragmentColorClamp() const {
+    if (UpdateValue(cur_state.fragment_color_clamp.enabled, fragment_color_clamp.enabled)) {
+        glClampColor(GL_CLAMP_FRAGMENT_COLOR_ARB,
+                     fragment_color_clamp.enabled ? GL_TRUE : GL_FALSE);
+    }
+}
+
+void OpenGLState::ApplyMultisample() const {
+    Enable(GL_SAMPLE_ALPHA_TO_COVERAGE, cur_state.multisample_control.alpha_to_coverage,
+           multisample_control.alpha_to_coverage);
+    Enable(GL_SAMPLE_ALPHA_TO_ONE, cur_state.multisample_control.alpha_to_one,
+           multisample_control.alpha_to_one);
+}
+
+void OpenGLState::ApplyDepthClamp() const {
+    if (depth_clamp.far_plane == cur_state.depth_clamp.far_plane &&
+        depth_clamp.near_plane == cur_state.depth_clamp.near_plane) {
+        return;
+    }
+    cur_state.depth_clamp = depth_clamp;
+
+    UNIMPLEMENTED_IF_MSG(depth_clamp.far_plane != depth_clamp.near_plane,
+                         "Unimplemented Depth Clamp Separation!");
+
+    Enable(GL_DEPTH_CLAMP, depth_clamp.far_plane || depth_clamp.near_plane);
+}
+
 void OpenGLState::ApplySRgb() const {
-    if (framebuffer_srgb.enabled != cur_state.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);
-        }
+    if (cur_state.framebuffer_srgb.enabled == framebuffer_srgb.enabled)
+        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);
     }
 }
 
 void OpenGLState::ApplyCulling() const {
-    if (cull.enabled != cur_state.cull.enabled) {
-        if (cull.enabled) {
-            glEnable(GL_CULL_FACE);
-        } else {
-            glDisable(GL_CULL_FACE);
-        }
-    }
+    Enable(GL_CULL_FACE, cur_state.cull.enabled, cull.enabled);
 
-    if (cull.mode != cur_state.cull.mode) {
+    if (UpdateValue(cur_state.cull.mode, cull.mode)) {
         glCullFace(cull.mode);
     }
 
-    if (cull.front_face != cur_state.cull.front_face) {
+    if (UpdateValue(cur_state.cull.front_face, cull.front_face)) {
         glFrontFace(cull.front_face);
     }
 }
 
 void OpenGLState::ApplyColorMask() const {
-    if (independant_blend.enabled) {
-        for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
-            const auto& updated = color_mask[i];
-            const auto& current = cur_state.color_mask[i];
-            if (updated.red_enabled != current.red_enabled ||
-                updated.green_enabled != current.green_enabled ||
-                updated.blue_enabled != current.blue_enabled ||
-                updated.alpha_enabled != current.alpha_enabled) {
-                glColorMaski(static_cast<GLuint>(i), updated.red_enabled, updated.green_enabled,
-                             updated.blue_enabled, updated.alpha_enabled);
-            }
-        }
-    } else {
-        const auto& updated = color_mask[0];
-        const auto& current = cur_state.color_mask[0];
+    for (std::size_t i = 0; i < Maxwell::NumRenderTargets; ++i) {
+        const auto& updated = color_mask[i];
+        auto& current = cur_state.color_mask[i];
         if (updated.red_enabled != current.red_enabled ||
             updated.green_enabled != current.green_enabled ||
             updated.blue_enabled != current.blue_enabled ||
             updated.alpha_enabled != current.alpha_enabled) {
-            glColorMask(updated.red_enabled, updated.green_enabled, updated.blue_enabled,
-                        updated.alpha_enabled);
+            current = updated;
+            glColorMaski(static_cast<GLuint>(i), updated.red_enabled, updated.green_enabled,
+                         updated.blue_enabled, updated.alpha_enabled);
         }
     }
 }
 
 void OpenGLState::ApplyDepth() const {
-    if (depth.test_enabled != cur_state.depth.test_enabled) {
-        if (depth.test_enabled) {
-            glEnable(GL_DEPTH_TEST);
-        } else {
-            glDisable(GL_DEPTH_TEST);
-        }
-    }
+    Enable(GL_DEPTH_TEST, cur_state.depth.test_enabled, depth.test_enabled);
 
-    if (depth.test_func != cur_state.depth.test_func) {
+    if (cur_state.depth.test_func != depth.test_func) {
+        cur_state.depth.test_func = depth.test_func;
         glDepthFunc(depth.test_func);
     }
 
-    if (depth.write_mask != cur_state.depth.write_mask) {
+    if (cur_state.depth.write_mask != depth.write_mask) {
+        cur_state.depth.write_mask = depth.write_mask;
         glDepthMask(depth.write_mask);
     }
 }
 
 void OpenGLState::ApplyPrimitiveRestart() const {
-    if (primitive_restart.enabled != cur_state.primitive_restart.enabled) {
-        if (primitive_restart.enabled) {
-            glEnable(GL_PRIMITIVE_RESTART);
-        } else {
-            glDisable(GL_PRIMITIVE_RESTART);
-        }
-    }
+    Enable(GL_PRIMITIVE_RESTART, cur_state.primitive_restart.enabled, primitive_restart.enabled);
 
-    if (primitive_restart.index != cur_state.primitive_restart.index) {
+    if (cur_state.primitive_restart.index != primitive_restart.index) {
+        cur_state.primitive_restart.index = primitive_restart.index;
         glPrimitiveRestartIndex(primitive_restart.index);
     }
 }
 
 void OpenGLState::ApplyStencilTest() const {
-    if (stencil.test_enabled != cur_state.stencil.test_enabled) {
-        if (stencil.test_enabled) {
-            glEnable(GL_STENCIL_TEST);
-        } else {
-            glDisable(GL_STENCIL_TEST);
-        }
-    }
+    Enable(GL_STENCIL_TEST, cur_state.stencil.test_enabled, stencil.test_enabled);
 
-    const auto ConfigStencil = [](GLenum face, const auto& config, const auto& prev_config) {
-        if (config.test_func != prev_config.test_func || config.test_ref != prev_config.test_ref ||
-            config.test_mask != prev_config.test_mask) {
+    const auto ConfigStencil = [](GLenum face, const auto& config, auto& current) {
+        if (current.test_func != config.test_func || current.test_ref != config.test_ref ||
+            current.test_mask != config.test_mask) {
+            current.test_func = config.test_func;
+            current.test_ref = config.test_ref;
+            current.test_mask = config.test_mask;
             glStencilFuncSeparate(face, config.test_func, config.test_ref, config.test_mask);
         }
-        if (config.action_depth_fail != prev_config.action_depth_fail ||
-            config.action_depth_pass != prev_config.action_depth_pass ||
-            config.action_stencil_fail != prev_config.action_stencil_fail) {
+        if (current.action_depth_fail != config.action_depth_fail ||
+            current.action_depth_pass != config.action_depth_pass ||
+            current.action_stencil_fail != config.action_stencil_fail) {
+            current.action_depth_fail = config.action_depth_fail;
+            current.action_depth_pass = config.action_depth_pass;
+            current.action_stencil_fail = config.action_stencil_fail;
             glStencilOpSeparate(face, config.action_stencil_fail, config.action_depth_fail,
                                 config.action_depth_pass);
         }
-        if (config.write_mask != prev_config.write_mask) {
+        if (current.write_mask != config.write_mask) {
+            current.write_mask = config.write_mask;
             glStencilMaskSeparate(face, config.write_mask);
         }
     };
     ConfigStencil(GL_FRONT, stencil.front, cur_state.stencil.front);
     ConfigStencil(GL_BACK, stencil.back, cur_state.stencil.back);
 }
-// Viewport does not affects glClearBuffer so emulate viewport using scissor test
+
+void OpenGLState::ApplyViewport() const {
+    for (GLuint i = 0; i < static_cast<GLuint>(Maxwell::NumViewports); ++i) {
+        const auto& updated = viewports[i];
+        auto& current = cur_state.viewports[i];
+
+        if (current.x != updated.x || current.y != updated.y || current.width != updated.width ||
+            current.height != updated.height) {
+            current.x = updated.x;
+            current.y = updated.y;
+            current.width = updated.width;
+            current.height = updated.height;
+            glViewportIndexedf(i, static_cast<GLfloat>(updated.x), static_cast<GLfloat>(updated.y),
+                               static_cast<GLfloat>(updated.width),
+                               static_cast<GLfloat>(updated.height));
+        }
+        if (current.depth_range_near != updated.depth_range_near ||
+            current.depth_range_far != updated.depth_range_far) {
+            current.depth_range_near = updated.depth_range_near;
+            current.depth_range_far = updated.depth_range_far;
+            glDepthRangeIndexed(i, updated.depth_range_near, updated.depth_range_far);
+        }
+
+        Enable(GL_SCISSOR_TEST, i, current.scissor.enabled, updated.scissor.enabled);
+
+        if (current.scissor.x != updated.scissor.x || current.scissor.y != updated.scissor.y ||
+            current.scissor.width != updated.scissor.width ||
+            current.scissor.height != updated.scissor.height) {
+            current.scissor.x = updated.scissor.x;
+            current.scissor.y = updated.scissor.y;
+            current.scissor.width = updated.scissor.width;
+            current.scissor.height = updated.scissor.height;
+            glScissorIndexed(i, updated.scissor.x, updated.scissor.y, updated.scissor.width,
+                             updated.scissor.height);
+        }
+    }
+}
+
+void OpenGLState::ApplyGlobalBlending() const {
+    const Blend& updated = blend[0];
+    Blend& current = cur_state.blend[0];
+
+    Enable(GL_BLEND, current.enabled, updated.enabled);
+
+    if (current.src_rgb_func != updated.src_rgb_func ||
+        current.dst_rgb_func != updated.dst_rgb_func || current.src_a_func != updated.src_a_func ||
+        current.dst_a_func != updated.dst_a_func) {
+        current.src_rgb_func = updated.src_rgb_func;
+        current.dst_rgb_func = updated.dst_rgb_func;
+        current.src_a_func = updated.src_a_func;
+        current.dst_a_func = updated.dst_a_func;
+        glBlendFuncSeparate(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
+                            updated.dst_a_func);
+    }
+
+    if (current.rgb_equation != updated.rgb_equation || current.a_equation != updated.a_equation) {
+        current.rgb_equation = updated.rgb_equation;
+        current.a_equation = updated.a_equation;
+        glBlendEquationSeparate(updated.rgb_equation, updated.a_equation);
+    }
+}
+
+void OpenGLState::ApplyTargetBlending(std::size_t target, bool force) const {
+    const Blend& updated = blend[target];
+    Blend& current = cur_state.blend[target];
+
+    if (current.enabled != updated.enabled || force) {
+        current.enabled = updated.enabled;
+        Enable(GL_BLEND, static_cast<GLuint>(target), updated.enabled);
+    }
+
+    if (UpdateTie(std::tie(current.src_rgb_func, current.dst_rgb_func, current.src_a_func,
+                           current.dst_a_func),
+                  std::tie(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
+                           updated.dst_a_func))) {
+        glBlendFuncSeparatei(static_cast<GLuint>(target), updated.src_rgb_func,
+                             updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func);
+    }
+
+    if (UpdateTie(std::tie(current.rgb_equation, current.a_equation),
+                  std::tie(updated.rgb_equation, updated.a_equation))) {
+        glBlendEquationSeparatei(static_cast<GLuint>(target), updated.rgb_equation,
+                                 updated.a_equation);
+    }
+}
+
+void OpenGLState::ApplyBlending() const {
+    if (independant_blend.enabled) {
+        const bool force = independant_blend.enabled != cur_state.independant_blend.enabled;
+        for (std::size_t target = 0; target < Maxwell::NumRenderTargets; ++target) {
+            ApplyTargetBlending(target, force);
+        }
+    } else {
+        ApplyGlobalBlending();
+    }
+    cur_state.independant_blend.enabled = independant_blend.enabled;
+
+    if (UpdateTie(
+            std::tie(cur_state.blend_color.red, cur_state.blend_color.green,
+                     cur_state.blend_color.blue, cur_state.blend_color.alpha),
+            std::tie(blend_color.red, blend_color.green, blend_color.blue, blend_color.alpha))) {
+        glBlendColor(blend_color.red, blend_color.green, blend_color.blue, blend_color.alpha);
+    }
+}
+
+void OpenGLState::ApplyLogicOp() const {
+    Enable(GL_COLOR_LOGIC_OP, cur_state.logic_op.enabled, logic_op.enabled);
+
+    if (UpdateValue(cur_state.logic_op.operation, logic_op.operation)) {
+        glLogicOp(logic_op.operation);
+    }
+}
+
+void OpenGLState::ApplyPolygonOffset() const {
+    Enable(GL_POLYGON_OFFSET_FILL, cur_state.polygon_offset.fill_enable,
+           polygon_offset.fill_enable);
+    Enable(GL_POLYGON_OFFSET_LINE, cur_state.polygon_offset.line_enable,
+           polygon_offset.line_enable);
+    Enable(GL_POLYGON_OFFSET_POINT, cur_state.polygon_offset.point_enable,
+           polygon_offset.point_enable);
+
+    if (UpdateTie(std::tie(cur_state.polygon_offset.factor, cur_state.polygon_offset.units,
+                           cur_state.polygon_offset.clamp),
+                  std::tie(polygon_offset.factor, polygon_offset.units, polygon_offset.clamp))) {
+        if (GLAD_GL_EXT_polygon_offset_clamp && polygon_offset.clamp != 0) {
+            glPolygonOffsetClamp(polygon_offset.factor, polygon_offset.units, polygon_offset.clamp);
+        } else {
+            UNIMPLEMENTED_IF_MSG(polygon_offset.clamp != 0,
+                                 "Unimplemented Depth polygon offset clamp.");
+            glPolygonOffset(polygon_offset.factor, polygon_offset.units);
+        }
+    }
+}
+
+void OpenGLState::ApplyTextures() const {
+    bool has_delta{};
+    std::size_t first{};
+    std::size_t last{};
+    std::array<GLuint, Maxwell::NumTextureSamplers> textures;
+
+    for (std::size_t i = 0; i < std::size(texture_units); ++i) {
+        const auto& texture_unit = texture_units[i];
+        auto& cur_state_texture_unit = cur_state.texture_units[i];
+        textures[i] = texture_unit.texture;
+        if (cur_state_texture_unit.texture == textures[i])
+            continue;
+        cur_state_texture_unit.texture = textures[i];
+        if (!has_delta) {
+            first = i;
+            has_delta = true;
+        }
+        last = i;
+    }
+    if (has_delta) {
+        glBindTextures(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
+                       textures.data() + first);
+    }
+}
+
+void OpenGLState::ApplySamplers() const {
+    bool has_delta{};
+    std::size_t first{};
+    std::size_t last{};
+    std::array<GLuint, Maxwell::NumTextureSamplers> samplers;
+
+    for (std::size_t i = 0; i < std::size(samplers); ++i) {
+        if (cur_state.texture_units[i].sampler == texture_units[i].sampler)
+            continue;
+        cur_state.texture_units[i].sampler = texture_units[i].sampler;
+        samplers[i] = texture_units[i].sampler;
+        if (!has_delta) {
+            first = i;
+            has_delta = true;
+        }
+        last = i;
+    }
+    if (has_delta) {
+        glBindSamplers(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
+                       samplers.data() + first);
+    }
+}
+
+void OpenGLState::Apply() const {
+    ApplyFramebufferState();
+    ApplyVertexArrayState();
+    ApplyShaderProgram();
+    ApplyProgramPipeline();
+    ApplyClipDistances();
+    ApplyPointSize();
+    ApplyFragmentColorClamp();
+    ApplyMultisample();
+    ApplyDepthClamp();
+    ApplyColorMask();
+    ApplyViewport();
+    ApplyStencilTest();
+    ApplySRgb();
+    ApplyCulling();
+    ApplyDepth();
+    ApplyPrimitiveRestart();
+    ApplyBlending();
+    ApplyLogicOp();
+    ApplyTextures();
+    ApplySamplers();
+    ApplyPolygonOffset();
+}
+
 void OpenGLState::EmulateViewportWithScissor() {
     auto& current = viewports[0];
     if (current.scissor.enabled) {
@@ -251,332 +555,6 @@ void OpenGLState::EmulateViewportWithScissor() {
     }
 }
 
-void OpenGLState::ApplyViewport() const {
-    if (geometry_shaders.enabled) {
-        for (GLuint i = 0; i < static_cast<GLuint>(Tegra::Engines::Maxwell3D::Regs::NumViewports);
-             i++) {
-            const auto& current = cur_state.viewports[i];
-            const auto& updated = viewports[i];
-            if (updated.x != current.x || updated.y != current.y ||
-                updated.width != current.width || updated.height != current.height) {
-                glViewportIndexedf(
-                    i, static_cast<GLfloat>(updated.x), static_cast<GLfloat>(updated.y),
-                    static_cast<GLfloat>(updated.width), static_cast<GLfloat>(updated.height));
-            }
-            if (updated.depth_range_near != current.depth_range_near ||
-                updated.depth_range_far != current.depth_range_far) {
-                glDepthRangeIndexed(i, updated.depth_range_near, updated.depth_range_far);
-            }
-
-            if (updated.scissor.enabled != current.scissor.enabled) {
-                if (updated.scissor.enabled) {
-                    glEnablei(GL_SCISSOR_TEST, i);
-                } else {
-                    glDisablei(GL_SCISSOR_TEST, i);
-                }
-            }
-
-            if (updated.scissor.x != current.scissor.x || updated.scissor.y != current.scissor.y ||
-                updated.scissor.width != current.scissor.width ||
-                updated.scissor.height != current.scissor.height) {
-                glScissorIndexed(i, updated.scissor.x, updated.scissor.y, updated.scissor.width,
-                                 updated.scissor.height);
-            }
-        }
-    } else {
-        const auto& current = cur_state.viewports[0];
-        const auto& updated = viewports[0];
-        if (updated.x != current.x || updated.y != current.y || updated.width != current.width ||
-            updated.height != current.height) {
-            glViewport(updated.x, updated.y, updated.width, updated.height);
-        }
-
-        if (updated.depth_range_near != current.depth_range_near ||
-            updated.depth_range_far != current.depth_range_far) {
-            glDepthRange(updated.depth_range_near, updated.depth_range_far);
-        }
-
-        if (updated.scissor.enabled != current.scissor.enabled) {
-            if (updated.scissor.enabled) {
-                glEnable(GL_SCISSOR_TEST);
-            } else {
-                glDisable(GL_SCISSOR_TEST);
-            }
-        }
-
-        if (updated.scissor.x != current.scissor.x || updated.scissor.y != current.scissor.y ||
-            updated.scissor.width != current.scissor.width ||
-            updated.scissor.height != current.scissor.height) {
-            glScissor(updated.scissor.x, updated.scissor.y, updated.scissor.width,
-                      updated.scissor.height);
-        }
-    }
-}
-
-void OpenGLState::ApplyGlobalBlending() const {
-    const Blend& current = cur_state.blend[0];
-    const Blend& updated = blend[0];
-    if (updated.enabled != current.enabled) {
-        if (updated.enabled) {
-            glEnable(GL_BLEND);
-        } else {
-            glDisable(GL_BLEND);
-        }
-    }
-    if (!updated.enabled) {
-        return;
-    }
-    if (updated.src_rgb_func != current.src_rgb_func ||
-        updated.dst_rgb_func != current.dst_rgb_func || updated.src_a_func != current.src_a_func ||
-        updated.dst_a_func != current.dst_a_func) {
-        glBlendFuncSeparate(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
-                            updated.dst_a_func);
-    }
-
-    if (updated.rgb_equation != current.rgb_equation || updated.a_equation != current.a_equation) {
-        glBlendEquationSeparate(updated.rgb_equation, updated.a_equation);
-    }
-}
-
-void OpenGLState::ApplyTargetBlending(std::size_t target, bool force) const {
-    const Blend& updated = blend[target];
-    const Blend& current = cur_state.blend[target];
-    if (updated.enabled != current.enabled || force) {
-        if (updated.enabled) {
-            glEnablei(GL_BLEND, static_cast<GLuint>(target));
-        } else {
-            glDisablei(GL_BLEND, static_cast<GLuint>(target));
-        }
-    }
-
-    if (updated.src_rgb_func != current.src_rgb_func ||
-        updated.dst_rgb_func != current.dst_rgb_func || updated.src_a_func != current.src_a_func ||
-        updated.dst_a_func != current.dst_a_func) {
-        glBlendFuncSeparatei(static_cast<GLuint>(target), updated.src_rgb_func,
-                             updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func);
-    }
-
-    if (updated.rgb_equation != current.rgb_equation || updated.a_equation != current.a_equation) {
-        glBlendEquationSeparatei(static_cast<GLuint>(target), updated.rgb_equation,
-                                 updated.a_equation);
-    }
-}
-
-void OpenGLState::ApplyBlending() const {
-    if (independant_blend.enabled) {
-        for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
-            ApplyTargetBlending(i,
-                                independant_blend.enabled != cur_state.independant_blend.enabled);
-        }
-    } else {
-        ApplyGlobalBlending();
-    }
-    if (blend_color.red != cur_state.blend_color.red ||
-        blend_color.green != cur_state.blend_color.green ||
-        blend_color.blue != cur_state.blend_color.blue ||
-        blend_color.alpha != cur_state.blend_color.alpha) {
-        glBlendColor(blend_color.red, blend_color.green, blend_color.blue, blend_color.alpha);
-    }
-}
-
-void OpenGLState::ApplyLogicOp() const {
-    if (logic_op.enabled != cur_state.logic_op.enabled) {
-        if (logic_op.enabled) {
-            glEnable(GL_COLOR_LOGIC_OP);
-        } else {
-            glDisable(GL_COLOR_LOGIC_OP);
-        }
-    }
-
-    if (logic_op.operation != cur_state.logic_op.operation) {
-        glLogicOp(logic_op.operation);
-    }
-}
-
-void OpenGLState::ApplyPolygonOffset() const {
-    const bool fill_enable_changed =
-        polygon_offset.fill_enable != cur_state.polygon_offset.fill_enable;
-    const bool line_enable_changed =
-        polygon_offset.line_enable != cur_state.polygon_offset.line_enable;
-    const bool point_enable_changed =
-        polygon_offset.point_enable != cur_state.polygon_offset.point_enable;
-    const bool factor_changed = polygon_offset.factor != cur_state.polygon_offset.factor;
-    const bool units_changed = polygon_offset.units != cur_state.polygon_offset.units;
-    const bool clamp_changed = polygon_offset.clamp != cur_state.polygon_offset.clamp;
-
-    if (fill_enable_changed) {
-        if (polygon_offset.fill_enable) {
-            glEnable(GL_POLYGON_OFFSET_FILL);
-        } else {
-            glDisable(GL_POLYGON_OFFSET_FILL);
-        }
-    }
-
-    if (line_enable_changed) {
-        if (polygon_offset.line_enable) {
-            glEnable(GL_POLYGON_OFFSET_LINE);
-        } else {
-            glDisable(GL_POLYGON_OFFSET_LINE);
-        }
-    }
-
-    if (point_enable_changed) {
-        if (polygon_offset.point_enable) {
-            glEnable(GL_POLYGON_OFFSET_POINT);
-        } else {
-            glDisable(GL_POLYGON_OFFSET_POINT);
-        }
-    }
-
-    if (factor_changed || units_changed || clamp_changed) {
-        if (GLAD_GL_EXT_polygon_offset_clamp && polygon_offset.clamp != 0) {
-            glPolygonOffsetClamp(polygon_offset.factor, polygon_offset.units, polygon_offset.clamp);
-        } else {
-            glPolygonOffset(polygon_offset.factor, polygon_offset.units);
-            UNIMPLEMENTED_IF_MSG(polygon_offset.clamp != 0,
-                                 "Unimplemented Depth polygon offset clamp.");
-        }
-    }
-}
-
-void OpenGLState::ApplyTextures() const {
-    bool has_delta{};
-    std::size_t first{};
-    std::size_t last{};
-    std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> textures;
-
-    for (std::size_t i = 0; i < std::size(texture_units); ++i) {
-        const auto& texture_unit = texture_units[i];
-        const auto& cur_state_texture_unit = cur_state.texture_units[i];
-        textures[i] = texture_unit.texture;
-
-        if (textures[i] != cur_state_texture_unit.texture) {
-            if (!has_delta) {
-                first = i;
-                has_delta = true;
-            }
-            last = i;
-        }
-    }
-
-    if (has_delta) {
-        glBindTextures(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
-                       textures.data() + first);
-    }
-}
-
-void OpenGLState::ApplySamplers() const {
-    bool has_delta{};
-    std::size_t first{};
-    std::size_t last{};
-    std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> samplers;
-    for (std::size_t i = 0; i < std::size(samplers); ++i) {
-        samplers[i] = texture_units[i].sampler;
-        if (samplers[i] != cur_state.texture_units[i].sampler) {
-            if (!has_delta) {
-                first = i;
-                has_delta = true;
-            }
-            last = i;
-        }
-    }
-    if (has_delta) {
-        glBindSamplers(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
-                       samplers.data() + first);
-    }
-}
-
-void OpenGLState::ApplyFramebufferState() const {
-    if (draw.read_framebuffer != cur_state.draw.read_framebuffer) {
-        glBindFramebuffer(GL_READ_FRAMEBUFFER, draw.read_framebuffer);
-    }
-    if (draw.draw_framebuffer != cur_state.draw.draw_framebuffer) {
-        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, draw.draw_framebuffer);
-    }
-}
-
-void OpenGLState::ApplyVertexArrayState() const {
-    if (draw.vertex_array != cur_state.draw.vertex_array) {
-        glBindVertexArray(draw.vertex_array);
-    }
-}
-
-void OpenGLState::ApplyDepthClamp() const {
-    if (depth_clamp.far_plane == cur_state.depth_clamp.far_plane &&
-        depth_clamp.near_plane == cur_state.depth_clamp.near_plane) {
-        return;
-    }
-    UNIMPLEMENTED_IF_MSG(depth_clamp.far_plane != depth_clamp.near_plane,
-                         "Unimplemented Depth Clamp Separation!");
-
-    if (depth_clamp.far_plane || depth_clamp.near_plane) {
-        glEnable(GL_DEPTH_CLAMP);
-    } else {
-        glDisable(GL_DEPTH_CLAMP);
-    }
-}
-
-void OpenGLState::Apply() const {
-    ApplyFramebufferState();
-    ApplyVertexArrayState();
-
-    // Shader program
-    if (draw.shader_program != cur_state.draw.shader_program) {
-        glUseProgram(draw.shader_program);
-    }
-
-    // Program pipeline
-    if (draw.program_pipeline != cur_state.draw.program_pipeline) {
-        glBindProgramPipeline(draw.program_pipeline);
-    }
-    // Clip distance
-    for (std::size_t i = 0; i < clip_distance.size(); ++i) {
-        if (clip_distance[i] != cur_state.clip_distance[i]) {
-            if (clip_distance[i]) {
-                glEnable(GL_CLIP_DISTANCE0 + static_cast<GLenum>(i));
-            } else {
-                glDisable(GL_CLIP_DISTANCE0 + static_cast<GLenum>(i));
-            }
-        }
-    }
-    // Point
-    if (point.size != cur_state.point.size) {
-        glPointSize(point.size);
-    }
-    if (fragment_color_clamp.enabled != cur_state.fragment_color_clamp.enabled) {
-        glClampColor(GL_CLAMP_FRAGMENT_COLOR_ARB,
-                     fragment_color_clamp.enabled ? GL_TRUE : GL_FALSE);
-    }
-    if (multisample_control.alpha_to_coverage != cur_state.multisample_control.alpha_to_coverage) {
-        if (multisample_control.alpha_to_coverage) {
-            glEnable(GL_SAMPLE_ALPHA_TO_COVERAGE);
-        } else {
-            glDisable(GL_SAMPLE_ALPHA_TO_COVERAGE);
-        }
-    }
-    if (multisample_control.alpha_to_one != cur_state.multisample_control.alpha_to_one) {
-        if (multisample_control.alpha_to_one) {
-            glEnable(GL_SAMPLE_ALPHA_TO_ONE);
-        } else {
-            glDisable(GL_SAMPLE_ALPHA_TO_ONE);
-        }
-    }
-    ApplyDepthClamp();
-    ApplyColorMask();
-    ApplyViewport();
-    ApplyStencilTest();
-    ApplySRgb();
-    ApplyCulling();
-    ApplyDepth();
-    ApplyPrimitiveRestart();
-    ApplyBlending();
-    ApplyLogicOp();
-    ApplyTextures();
-    ApplySamplers();
-    ApplyPolygonOffset();
-    cur_state = *this;
-}
-
 OpenGLState& OpenGLState::UnbindTexture(GLuint handle) {
     for (auto& unit : texture_units) {
         if (unit.texture == handle) {

src/video_core/renderer_opengl/gl_state.h

@@ -53,10 +53,6 @@ public:
         bool near_plane;
     } depth_clamp; // GL_DEPTH_CLAMP
 
-    struct {
-        bool enabled; // viewports arrays are only supported when geometry shaders are enabled.
-    } geometry_shaders;
-
     struct {
         bool enabled;      // GL_CULL_FACE
         GLenum mode;       // GL_CULL_FACE_MODE
@@ -184,34 +180,26 @@ public:
     static OpenGLState GetCurState() {
         return cur_state;
     }
+
     static bool GetsRGBUsed() {
         return s_rgb_used;
     }
+
     static void ClearsRGBUsed() {
         s_rgb_used = false;
     }
+
     /// Apply this state as the current OpenGL state
     void Apply() const;
-    /// Apply only the state affecting the framebuffer
-    void ApplyFramebufferState() const;
-    /// Apply only the state affecting the vertex array
-    void ApplyVertexArrayState() const;
-    /// Set the initial OpenGL state
-    static void ApplyDefaultState();
-    /// Resets any references to the given resource
-    OpenGLState& UnbindTexture(GLuint handle);
-    OpenGLState& ResetSampler(GLuint handle);
-    OpenGLState& ResetProgram(GLuint handle);
-    OpenGLState& ResetPipeline(GLuint handle);
-    OpenGLState& ResetVertexArray(GLuint handle);
-    OpenGLState& ResetFramebuffer(GLuint handle);
-    void EmulateViewportWithScissor();
 
-private:
-    static OpenGLState cur_state;
-    // Workaround for sRGB problems caused by
-    // QT not supporting srgb output
-    static bool s_rgb_used;
+    void ApplyFramebufferState() const;
+    void ApplyVertexArrayState() const;
+    void ApplyShaderProgram() const;
+    void ApplyProgramPipeline() const;
+    void ApplyClipDistances() const;
+    void ApplyPointSize() const;
+    void ApplyFragmentColorClamp() const;
+    void ApplyMultisample() const;
     void ApplySRgb() const;
     void ApplyCulling() const;
     void ApplyColorMask() const;
@@ -227,6 +215,26 @@ private:
     void ApplySamplers() const;
     void ApplyDepthClamp() const;
     void ApplyPolygonOffset() const;
+
+    /// Set the initial OpenGL state
+    static void ApplyDefaultState();
+
+    /// Resets any references to the given resource
+    OpenGLState& UnbindTexture(GLuint handle);
+    OpenGLState& ResetSampler(GLuint handle);
+    OpenGLState& ResetProgram(GLuint handle);
+    OpenGLState& ResetPipeline(GLuint handle);
+    OpenGLState& ResetVertexArray(GLuint handle);
+    OpenGLState& ResetFramebuffer(GLuint handle);
+
+    /// Viewport does not affects glClearBuffer so emulate viewport using scissor test
+    void EmulateViewportWithScissor();
+
+private:
+    static OpenGLState cur_state;
+
+    // Workaround for sRGB problems caused by QT not supporting srgb output
+    static bool s_rgb_used;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/utils.cpp

@@ -5,11 +5,39 @@
 #include <string>
 #include <fmt/format.h>
 #include <glad/glad.h>
+#include "common/assert.h"
 #include "common/common_types.h"
 #include "video_core/renderer_opengl/utils.h"
 
 namespace OpenGL {
 
+BindBuffersRangePushBuffer::BindBuffersRangePushBuffer(GLenum target) : target{target} {}
+
+BindBuffersRangePushBuffer::~BindBuffersRangePushBuffer() = default;
+
+void BindBuffersRangePushBuffer::Setup(GLuint first_) {
+    first = first_;
+    buffers.clear();
+    offsets.clear();
+    sizes.clear();
+}
+
+void BindBuffersRangePushBuffer::Push(GLuint buffer, GLintptr offset, GLsizeiptr size) {
+    buffers.push_back(buffer);
+    offsets.push_back(offset);
+    sizes.push_back(size);
+}
+
+void BindBuffersRangePushBuffer::Bind() const {
+    const std::size_t count{buffers.size()};
+    DEBUG_ASSERT(count == offsets.size() && count == sizes.size());
+    if (count == 0) {
+        return;
+    }
+    glBindBuffersRange(target, first, static_cast<GLsizei>(count), buffers.data(), offsets.data(),
+                       sizes.data());
+}
+
 void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr, std::string extra_info) {
     if (!GLAD_GL_KHR_debug) {
         return; // We don't need to throw an error as this is just for debugging

src/video_core/renderer_opengl/utils.h

@@ -5,11 +5,31 @@
 #pragma once
 
 #include <string>
+#include <vector>
 #include <glad/glad.h>
 #include "common/common_types.h"
 
 namespace OpenGL {
 
+class BindBuffersRangePushBuffer {
+public:
+    BindBuffersRangePushBuffer(GLenum target);
+    ~BindBuffersRangePushBuffer();
+
+    void Setup(GLuint first_);
+
+    void Push(GLuint buffer, GLintptr offset, GLsizeiptr size);
+
+    void Bind() const;
+
+private:
+    GLenum target;
+    GLuint first;
+    std::vector<GLuint> buffers;
+    std::vector<GLintptr> offsets;
+    std::vector<GLsizeiptr> sizes;
+};
+
 void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr, std::string extra_info = "");
 
 } // namespace OpenGL

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -10,6 +10,7 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "core/memory.h"
+#include "video_core/memory_manager.h"
 #include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"

src/video_core/renderer_vulkan/vk_shader_decompiler.h

@@ -0,0 +1,80 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+#include <set>
+#include <utility>
+#include <vector>
+
+#include <sirit/sirit.h>
+
+#include "common/common_types.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+class ShaderIR;
+}
+
+namespace Vulkan::VKShader {
+
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+
+using SamplerEntry = VideoCommon::Shader::Sampler;
+
+constexpr u32 DESCRIPTOR_SET = 0;
+
+class ConstBufferEntry : public VideoCommon::Shader::ConstBuffer {
+public:
+    explicit constexpr ConstBufferEntry(const VideoCommon::Shader::ConstBuffer& entry, u32 index)
+        : VideoCommon::Shader::ConstBuffer{entry}, index{index} {}
+
+    constexpr u32 GetIndex() const {
+        return index;
+    }
+
+private:
+    u32 index{};
+};
+
+class GlobalBufferEntry {
+public:
+    explicit GlobalBufferEntry(u32 cbuf_index, u32 cbuf_offset)
+        : cbuf_index{cbuf_index}, cbuf_offset{cbuf_offset} {}
+
+    u32 GetCbufIndex() const {
+        return cbuf_index;
+    }
+
+    u32 GetCbufOffset() const {
+        return cbuf_offset;
+    }
+
+private:
+    u32 cbuf_index{};
+    u32 cbuf_offset{};
+};
+
+struct ShaderEntries {
+    u32 const_buffers_base_binding{};
+    u32 global_buffers_base_binding{};
+    u32 samplers_base_binding{};
+    std::vector<ConstBufferEntry> const_buffers;
+    std::vector<GlobalBufferEntry> global_buffers;
+    std::vector<SamplerEntry> samplers;
+    std::set<u32> attributes;
+    std::array<bool, Maxwell::NumClipDistances> clip_distances{};
+    std::size_t shader_length{};
+    Sirit::Id entry_function{};
+    std::vector<Sirit::Id> interfaces;
+};
+
+using DecompilerResult = std::pair<std::unique_ptr<Sirit::Module>, ShaderEntries>;
+
+DecompilerResult Decompile(const VideoCommon::Shader::ShaderIR& ir, Maxwell::ShaderStage stage);
+
+} // namespace Vulkan::VKShader

src/video_core/shader/decode/texture.cpp

@@ -7,7 +7,9 @@
 #include <fmt/format.h>
 
 #include "common/assert.h"
+#include "common/bit_field.h"
 #include "common/common_types.h"
+#include "common/logging/log.h"
 #include "video_core/engines/shader_bytecode.h"
 #include "video_core/shader/shader_ir.h"
 
@@ -41,19 +43,18 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
 
     switch (opcode->get().GetId()) {
     case OpCode::Id::TEX: {
-        UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(TextureMiscMode::AOFFI),
-                             "AOFFI is not implemented");
-
         if (instr.tex.UsesMiscMode(TextureMiscMode::NODEP)) {
             LOG_WARNING(HW_GPU, "TEX.NODEP implementation is incomplete");
         }
 
         const TextureType texture_type{instr.tex.texture_type};
         const bool is_array = instr.tex.array != 0;
+        const bool is_aoffi = instr.tex.UsesMiscMode(TextureMiscMode::AOFFI);
         const bool depth_compare = instr.tex.UsesMiscMode(TextureMiscMode::DC);
         const auto process_mode = instr.tex.GetTextureProcessMode();
         WriteTexInstructionFloat(
-            bb, instr, GetTexCode(instr, texture_type, process_mode, depth_compare, is_array));
+            bb, instr,
+            GetTexCode(instr, texture_type, process_mode, depth_compare, is_array, is_aoffi));
         break;
     }
     case OpCode::Id::TEXS: {
@@ -78,8 +79,6 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
     }
     case OpCode::Id::TLD4: {
         ASSERT(instr.tld4.array == 0);
-        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::AOFFI),
-                             "AOFFI is not implemented");
         UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::NDV),
                              "NDV is not implemented");
         UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::PTP),
@@ -92,8 +91,9 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         const auto texture_type = instr.tld4.texture_type.Value();
         const bool depth_compare = instr.tld4.UsesMiscMode(TextureMiscMode::DC);
         const bool is_array = instr.tld4.array != 0;
-        WriteTexInstructionFloat(bb, instr,
-                                 GetTld4Code(instr, texture_type, depth_compare, is_array));
+        const bool is_aoffi = instr.tld4.UsesMiscMode(TextureMiscMode::AOFFI);
+        WriteTexInstructionFloat(
+            bb, instr, GetTld4Code(instr, texture_type, depth_compare, is_array, is_aoffi));
         break;
     }
     case OpCode::Id::TLD4S: {
@@ -127,7 +127,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         Node4 values;
         for (u32 element = 0; element < values.size(); ++element) {
             auto coords_copy = coords;
-            MetaTexture meta{sampler, {}, {}, {}, {}, component, element};
+            MetaTexture meta{sampler, {}, {}, {}, {}, {}, component, element};
             values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
         }
 
@@ -152,7 +152,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
                 if (!instr.txq.IsComponentEnabled(element)) {
                     continue;
                 }
-                MetaTexture meta{sampler, {}, {}, {}, {}, {}, element};
+                MetaTexture meta{sampler, {}, {}, {}, {}, {}, {}, element};
                 const Node value =
                     Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8));
                 SetTemporal(bb, indexer++, value);
@@ -202,7 +202,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
 
         for (u32 element = 0; element < 2; ++element) {
             auto params = coords;
-            MetaTexture meta{sampler, {}, {}, {}, {}, {}, element};
+            MetaTexture meta{sampler, {}, {}, {}, {}, {}, {}, element};
             const Node value = Operation(OperationCode::TextureQueryLod, meta, std::move(params));
             SetTemporal(bb, element, value);
         }
@@ -325,7 +325,8 @@ void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
 
 Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
                                TextureProcessMode process_mode, std::vector<Node> coords,
-                               Node array, Node depth_compare, u32 bias_offset) {
+                               Node array, Node depth_compare, u32 bias_offset,
+                               std::vector<Node> aoffi) {
     const bool is_array = array;
     const bool is_shadow = depth_compare;
 
@@ -374,7 +375,7 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
     Node4 values;
     for (u32 element = 0; element < values.size(); ++element) {
         auto copy_coords = coords;
-        MetaTexture meta{sampler, array, depth_compare, bias, lod, {}, element};
+        MetaTexture meta{sampler, array, depth_compare, aoffi, bias, lod, {}, element};
         values[element] = Operation(read_method, meta, std::move(copy_coords));
     }
 
@@ -382,9 +383,15 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
 }
 
 Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
-                           TextureProcessMode process_mode, bool depth_compare, bool is_array) {
-    const bool lod_bias_enabled =
-        (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
+                           TextureProcessMode process_mode, bool depth_compare, bool is_array,
+                           bool is_aoffi) {
+    const bool lod_bias_enabled{
+        (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ)};
+
+    u64 parameter_register = instr.gpr20.Value();
+    if (lod_bias_enabled) {
+        ++parameter_register;
+    }
 
     const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
         texture_type, depth_compare, is_array, lod_bias_enabled, 4, 5);
@@ -404,15 +411,19 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
 
     const Node array = is_array ? GetRegister(array_register) : nullptr;
 
+    std::vector<Node> aoffi;
+    if (is_aoffi) {
+        aoffi = GetAoffiCoordinates(GetRegister(parameter_register++), coord_count, false);
+    }
+
     Node dc{};
     if (depth_compare) {
         // Depth is always stored in the register signaled by gpr20 or in the next register if lod
         // or bias are used
-        const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
-        dc = GetRegister(depth_register);
+        dc = GetRegister(parameter_register++);
     }
 
-    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0);
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0, aoffi);
 }
 
 Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
@@ -448,11 +459,11 @@ Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
         dc = GetRegister(depth_register);
     }
 
-    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset);
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset, {});
 }
 
 Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,
-                            bool is_array) {
+                            bool is_array, bool is_aoffi) {
     const std::size_t coord_count = GetCoordCount(texture_type);
     const std::size_t total_coord_count = coord_count + (is_array ? 1 : 0);
     const std::size_t total_reg_count = total_coord_count + (depth_compare ? 1 : 0);
@@ -463,15 +474,27 @@ Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool de
     const u64 coord_register = array_register + (is_array ? 1 : 0);
 
     std::vector<Node> coords;
-    for (size_t i = 0; i < coord_count; ++i)
+    for (std::size_t i = 0; i < coord_count; ++i) {
         coords.push_back(GetRegister(coord_register + i));
+    }
+
+    u64 parameter_register = instr.gpr20.Value();
+    std::vector<Node> aoffi;
+    if (is_aoffi) {
+        aoffi = GetAoffiCoordinates(GetRegister(parameter_register++), coord_count, true);
+    }
+
+    Node dc{};
+    if (depth_compare) {
+        dc = GetRegister(parameter_register++);
+    }
 
     const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
 
     Node4 values;
     for (u32 element = 0; element < values.size(); ++element) {
         auto coords_copy = coords;
-        MetaTexture meta{sampler, GetRegister(array_register), {}, {}, {}, {}, element};
+        MetaTexture meta{sampler, GetRegister(array_register), dc, aoffi, {}, {}, {}, element};
         values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
     }
 
@@ -507,7 +530,7 @@ Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is
     Node4 values;
     for (u32 element = 0; element < values.size(); ++element) {
         auto coords_copy = coords;
-        MetaTexture meta{sampler, array, {}, {}, lod, {}, element};
+        MetaTexture meta{sampler, array, {}, {}, {}, lod, {}, element};
         values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy));
     }
     return values;
@@ -531,4 +554,45 @@ std::tuple<std::size_t, std::size_t> ShaderIR::ValidateAndGetCoordinateElement(
     return {coord_count, total_coord_count};
 }
 
-} // namespace VideoCommon::Shader
+std::vector<Node> ShaderIR::GetAoffiCoordinates(Node aoffi_reg, std::size_t coord_count,
+                                                bool is_tld4) {
+    const auto [coord_offsets, size, wrap_value,
+                diff_value] = [is_tld4]() -> std::tuple<std::array<u32, 3>, u32, s32, s32> {
+        if (is_tld4) {
+            return {{0, 8, 16}, 6, 32, 64};
+        } else {
+            return {{0, 4, 8}, 4, 8, 16};
+        }
+    }();
+    const u32 mask = (1U << size) - 1;
+
+    std::vector<Node> aoffi;
+    aoffi.reserve(coord_count);
+
+    const auto aoffi_immediate{
+        TrackImmediate(aoffi_reg, global_code, static_cast<s64>(global_code.size()))};
+    if (!aoffi_immediate) {
+        // Variable access, not supported on AMD.
+        LOG_WARNING(HW_GPU,
+                    "AOFFI constant folding failed, some hardware might have graphical issues");
+        for (std::size_t coord = 0; coord < coord_count; ++coord) {
+            const Node value = BitfieldExtract(aoffi_reg, coord_offsets.at(coord), size);
+            const Node condition =
+                Operation(OperationCode::LogicalIGreaterEqual, value, Immediate(wrap_value));
+            const Node negative = Operation(OperationCode::IAdd, value, Immediate(-diff_value));
+            aoffi.push_back(Operation(OperationCode::Select, condition, negative, value));
+        }
+        return aoffi;
+    }
+
+    for (std::size_t coord = 0; coord < coord_count; ++coord) {
+        s32 value = (*aoffi_immediate >> coord_offsets.at(coord)) & mask;
+        if (value >= wrap_value) {
+            value -= diff_value;
+        }
+        aoffi.push_back(Immediate(value));
+    }
+    return aoffi;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/xmad.cpp

@@ -29,39 +29,55 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     const bool is_signed_b = instr.xmad.sign_b == 1;
     const bool is_signed_c = is_signed_a;
 
-    auto [is_merge, op_b, op_c] = [&]() -> std::tuple<bool, Node, Node> {
+    auto [is_merge, is_psl, is_high_b, mode, op_b,
+          op_c] = [&]() -> std::tuple<bool, bool, bool, Tegra::Shader::XmadMode, Node, Node> {
         switch (opcode->get().GetId()) {
         case OpCode::Id::XMAD_CR:
             return {instr.xmad.merge_56,
+                    instr.xmad.product_shift_left_second,
+                    instr.xmad.high_b,
+                    instr.xmad.mode_cbf,
                     GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset()),
                     GetRegister(instr.gpr39)};
         case OpCode::Id::XMAD_RR:
-            return {instr.xmad.merge_37, GetRegister(instr.gpr20), GetRegister(instr.gpr39)};
+            return {instr.xmad.merge_37, instr.xmad.product_shift_left, instr.xmad.high_b_rr,
+                    instr.xmad.mode,     GetRegister(instr.gpr20),      GetRegister(instr.gpr39)};
         case OpCode::Id::XMAD_RC:
-            return {false, GetRegister(instr.gpr39),
+            return {false,
+                    false,
+                    instr.xmad.high_b,
+                    instr.xmad.mode_cbf,
+                    GetRegister(instr.gpr39),
                     GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset())};
         case OpCode::Id::XMAD_IMM:
-            return {instr.xmad.merge_37, Immediate(static_cast<u32>(instr.xmad.imm20_16)),
+            return {instr.xmad.merge_37,
+                    instr.xmad.product_shift_left,
+                    false,
+                    instr.xmad.mode,
+                    Immediate(static_cast<u32>(instr.xmad.imm20_16)),
                     GetRegister(instr.gpr39)};
         }
         UNIMPLEMENTED_MSG("Unhandled XMAD instruction: {}", opcode->get().GetName());
-        return {false, Immediate(0), Immediate(0)};
+        return {false, false, false, Tegra::Shader::XmadMode::None, Immediate(0), Immediate(0)};
     }();
 
     op_a = BitfieldExtract(op_a, instr.xmad.high_a ? 16 : 0, 16);
 
     const Node original_b = op_b;
-    op_b = BitfieldExtract(op_b, instr.xmad.high_b ? 16 : 0, 16);
+    op_b = BitfieldExtract(op_b, is_high_b ? 16 : 0, 16);
 
     // TODO(Rodrigo): Use an appropiate sign for this operation
     Node product = Operation(OperationCode::IMul, NO_PRECISE, op_a, op_b);
-    if (instr.xmad.product_shift_left) {
+    if (is_psl) {
         product = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, product, Immediate(16));
     }
+    SetTemporal(bb, 0, product);
+    product = GetTemporal(0);
 
     const Node original_c = op_c;
+    const Tegra::Shader::XmadMode set_mode = mode; // Workaround to clang compile error
     op_c = [&]() {
-        switch (instr.xmad.mode) {
+        switch (set_mode) {
         case Tegra::Shader::XmadMode::None:
             return original_c;
         case Tegra::Shader::XmadMode::CLo:
@@ -80,8 +96,13 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
         }
     }();
 
+    SetTemporal(bb, 1, op_c);
+    op_c = GetTemporal(1);
+
     // TODO(Rodrigo): Use an appropiate sign for this operation
     Node sum = Operation(OperationCode::IAdd, product, op_c);
+    SetTemporal(bb, 2, sum);
+    sum = GetTemporal(2);
     if (is_merge) {
         const Node a = BitfieldExtract(sum, 0, 16);
         const Node b =
@@ -95,4 +116,4 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     return pc;
 }
 
-} // namespace VideoCommon::Shader
+} // namespace VideoCommon::Shader

src/video_core/shader/shader_ir.h

@@ -7,6 +7,7 @@
 #include <array>
 #include <cstring>
 #include <map>
+#include <optional>
 #include <set>
 #include <string>
 #include <tuple>
@@ -290,6 +291,7 @@ struct MetaTexture {
     const Sampler& sampler;
     Node array{};
     Node depth_compare{};
+    std::vector<Node> aoffi;
     Node bias{};
     Node lod{};
     Node component{};
@@ -741,14 +743,14 @@ private:
 
     Node4 GetTexCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
                      Tegra::Shader::TextureProcessMode process_mode, bool depth_compare,
-                     bool is_array);
+                     bool is_array, bool is_aoffi);
 
     Node4 GetTexsCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
                       Tegra::Shader::TextureProcessMode process_mode, bool depth_compare,
                       bool is_array);
 
     Node4 GetTld4Code(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
-                      bool depth_compare, bool is_array);
+                      bool depth_compare, bool is_array, bool is_aoffi);
 
     Node4 GetTldsCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
                       bool is_array);
@@ -757,9 +759,11 @@ private:
         Tegra::Shader::TextureType texture_type, bool depth_compare, bool is_array,
         bool lod_bias_enabled, std::size_t max_coords, std::size_t max_inputs);
 
+    std::vector<Node> GetAoffiCoordinates(Node aoffi_reg, std::size_t coord_count, bool is_tld4);
+
     Node4 GetTextureCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
                          Tegra::Shader::TextureProcessMode process_mode, std::vector<Node> coords,
-                         Node array, Node depth_compare, u32 bias_offset);
+                         Node array, Node depth_compare, u32 bias_offset, std::vector<Node> aoffi);
 
     Node GetVideoOperand(Node op, bool is_chunk, bool is_signed, Tegra::Shader::VideoType type,
                          u64 byte_height);
@@ -773,6 +777,8 @@ private:
 
     Node TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor);
 
+    std::optional<u32> TrackImmediate(Node tracked, const NodeBlock& code, s64 cursor);
+
     std::pair<Node, s64> TrackRegister(const GprNode* tracked, const NodeBlock& code, s64 cursor);
 
     template <typename... T>

src/video_core/shader/track.cpp

@@ -6,6 +6,7 @@
 #include <utility>
 #include <variant>
 
+#include "common/common_types.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace VideoCommon::Shader {
@@ -14,7 +15,7 @@ namespace {
 std::pair<Node, s64> FindOperation(const NodeBlock& code, s64 cursor,
                                    OperationCode operation_code) {
     for (; cursor >= 0; --cursor) {
-        const Node node = code[cursor];
+        const Node node = code.at(cursor);
         if (const auto operation = std::get_if<OperationNode>(node)) {
             if (operation->GetCode() == operation_code)
                 return {node, cursor};
@@ -64,6 +65,20 @@ Node ShaderIR::TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor) {
     return nullptr;
 }
 
+std::optional<u32> ShaderIR::TrackImmediate(Node tracked, const NodeBlock& code, s64 cursor) {
+    // Reduce the cursor in one to avoid infinite loops when the instruction sets the same register
+    // that it uses as operand
+    const auto [found, found_cursor] =
+        TrackRegister(&std::get<GprNode>(*tracked), code, cursor - 1);
+    if (!found) {
+        return {};
+    }
+    if (const auto immediate = std::get_if<ImmediateNode>(found)) {
+        return immediate->GetValue();
+    }
+    return {};
+}
+
 std::pair<Node, s64> ShaderIR::TrackRegister(const GprNode* tracked, const NodeBlock& code,
                                              s64 cursor) {
     for (; cursor >= 0; --cursor) {

src/video_core/textures/convert.cpp

@@ -10,6 +10,7 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
+#include "video_core/surface.h"
 #include "video_core/textures/astc.h"
 #include "video_core/textures/convert.h"
 

src/video_core/textures/convert.h

@@ -5,7 +5,10 @@
 #pragma once
 
 #include "common/common_types.h"
-#include "video_core/surface.h"
+
+namespace VideoCore::Surface {
+enum class PixelFormat;
+}
 
 namespace Tegra::Texture {
 

src/video_core/textures/texture.h

@@ -7,9 +7,7 @@
 #include <array>
 #include "common/assert.h"
 #include "common/bit_field.h"
-#include "common/common_funcs.h"
 #include "common/common_types.h"
-#include "video_core/memory_manager.h"
 
 namespace Tegra::Texture {
 

src/yuzu/CMakeLists.txt

@@ -31,6 +31,8 @@ add_executable(yuzu
     configuration/configure_general.h
     configuration/configure_graphics.cpp
     configuration/configure_graphics.h
+    configuration/configure_hotkeys.cpp
+    configuration/configure_hotkeys.h
     configuration/configure_input.cpp
     configuration/configure_input.h
     configuration/configure_input_player.cpp
@@ -76,6 +78,8 @@ add_executable(yuzu
     ui_settings.h
     util/limitable_input_dialog.cpp
     util/limitable_input_dialog.h
+    util/sequence_dialog/sequence_dialog.cpp
+    util/sequence_dialog/sequence_dialog.h
     util/spinbox.cpp
     util/spinbox.h
     util/util.cpp
@@ -93,6 +97,7 @@ set(UIS
     configuration/configure_gamelist.ui
     configuration/configure_general.ui
     configuration/configure_graphics.ui
+    configuration/configure_hotkeys.ui
     configuration/configure_input.ui
     configuration/configure_input_player.ui
     configuration/configure_input_simple.ui
@@ -103,7 +108,6 @@ set(UIS
     configuration/configure_touchscreen_advanced.ui
     configuration/configure_web.ui
     compatdb.ui
-    hotkeys.ui
     loading_screen.ui
     main.ui
 )

src/yuzu/applets/profile_select.cpp

@@ -4,6 +4,7 @@
 
 #include <mutex>
 #include <QDialogButtonBox>
+#include <QHeaderView>
 #include <QLabel>
 #include <QLineEdit>
 #include <QScrollArea>

src/yuzu/applets/profile_select.h

@@ -7,6 +7,7 @@
 #include <vector>
 #include <QDialog>
 #include <QList>
+#include <QTreeView>
 #include "core/frontend/applets/profile_select.h"
 
 class GMainWindow;
@@ -16,7 +17,6 @@ class QLabel;
 class QScrollArea;
 class QStandardItem;
 class QStandardItemModel;
-class QTreeView;
 class QVBoxLayout;
 
 class QtProfileSelectionDialog final : public QDialog {

src/yuzu/configuration/config.cpp

@@ -2,6 +2,8 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <array>
+#include <QKeySequence>
 #include <QSettings>
 #include "common/file_util.h"
 #include "configure_input_simple.h"
@@ -9,7 +11,6 @@
 #include "core/hle/service/hid/controllers/npad.h"
 #include "input_common/main.h"
 #include "yuzu/configuration/config.h"
-#include "yuzu/ui_settings.h"
 
 Config::Config() {
     // TODO: Don't hardcode the path; let the frontend decide where to put the config files.
@@ -17,7 +18,6 @@ Config::Config() {
     FileUtil::CreateFullPath(qt_config_loc);
     qt_config =
         std::make_unique<QSettings>(QString::fromStdString(qt_config_loc), QSettings::IniFormat);
-
     Reload();
 }
 
@@ -205,6 +205,27 @@ const std::array<int, Settings::NativeKeyboard::NumKeyboardMods> Config::default
     Qt::Key_Control, Qt::Key_Shift, Qt::Key_AltGr, Qt::Key_ApplicationRight,
 };
 
+// This shouldn't have anything except static initializers (no functions). So
+// QKeySequnce(...).toString() is NOT ALLOWED HERE.
+// This must be in alphabetical order according to action name as it must have the same order as
+// UISetting::values.shortcuts, which is alphabetically ordered.
+const std::array<UISettings::Shortcut, 15> Config::default_hotkeys{
+    {{"Capture Screenshot", "Main Window", {"Ctrl+P", Qt::ApplicationShortcut}},
+     {"Continue/Pause Emulation", "Main Window", {"F4", Qt::WindowShortcut}},
+     {"Decrease Speed Limit", "Main Window", {"-", Qt::ApplicationShortcut}},
+     {"Exit yuzu", "Main Window", {"Ctrl+Q", Qt::WindowShortcut}},
+     {"Exit Fullscreen", "Main Window", {"Esc", Qt::WindowShortcut}},
+     {"Fullscreen", "Main Window", {"F11", Qt::WindowShortcut}},
+     {"Increase Speed Limit", "Main Window", {"+", Qt::ApplicationShortcut}},
+     {"Load Amiibo", "Main Window", {"F2", Qt::ApplicationShortcut}},
+     {"Load File", "Main Window", {"Ctrl+O", Qt::WindowShortcut}},
+     {"Restart Emulation", "Main Window", {"F6", Qt::WindowShortcut}},
+     {"Stop Emulation", "Main Window", {"F5", Qt::WindowShortcut}},
+     {"Toggle Filter Bar", "Main Window", {"Ctrl+F", Qt::WindowShortcut}},
+     {"Toggle Speed Limit", "Main Window", {"Ctrl+Z", Qt::ApplicationShortcut}},
+     {"Toggle Status Bar", "Main Window", {"Ctrl+S", Qt::WindowShortcut}},
+     {"Change Docked Mode", "Main Window", {"F10", Qt::ApplicationShortcut}}}};
+
 void Config::ReadPlayerValues() {
     for (std::size_t p = 0; p < Settings::values.players.size(); ++p) {
         auto& player = Settings::values.players[p];
@@ -508,20 +529,15 @@ void Config::ReadValues() {
     qt_config->endGroup();
 
     qt_config->beginGroup("Shortcuts");
-    QStringList groups = qt_config->childGroups();
-    for (auto group : groups) {
+    for (auto [name, group, shortcut] : default_hotkeys) {
+        auto [keyseq, context] = shortcut;
         qt_config->beginGroup(group);
-
-        QStringList hotkeys = qt_config->childGroups();
-        for (auto hotkey : hotkeys) {
-            qt_config->beginGroup(hotkey);
-            UISettings::values.shortcuts.emplace_back(UISettings::Shortcut(
-                group + "/" + hotkey,
-                UISettings::ContextualShortcut(ReadSetting("KeySeq").toString(),
-                                               ReadSetting("Context").toInt())));
-            qt_config->endGroup();
-        }
-
+        qt_config->beginGroup(name);
+        UISettings::values.shortcuts.push_back(
+            {name,
+             group,
+             {ReadSetting("KeySeq", keyseq).toString(), ReadSetting("Context", context).toInt()}});
+        qt_config->endGroup();
         qt_config->endGroup();
     }
     qt_config->endGroup();
@@ -758,9 +774,16 @@ void Config::SaveValues() {
     qt_config->endGroup();
 
     qt_config->beginGroup("Shortcuts");
-    for (auto shortcut : UISettings::values.shortcuts) {
-        WriteSetting(shortcut.first + "/KeySeq", shortcut.second.first);
-        WriteSetting(shortcut.first + "/Context", shortcut.second.second);
+    // Lengths of UISettings::values.shortcuts & default_hotkeys are same.
+    // However, their ordering must also be the same.
+    for (std::size_t i = 0; i < default_hotkeys.size(); i++) {
+        auto [name, group, shortcut] = UISettings::values.shortcuts[i];
+        qt_config->beginGroup(group);
+        qt_config->beginGroup(name);
+        WriteSetting("KeySeq", shortcut.first, default_hotkeys[i].shortcut.first);
+        WriteSetting("Context", shortcut.second, default_hotkeys[i].shortcut.second);
+        qt_config->endGroup();
+        qt_config->endGroup();
     }
     qt_config->endGroup();
 

src/yuzu/configuration/config.h

@@ -9,6 +9,7 @@
 #include <string>
 #include <QVariant>
 #include "core/settings.h"
+#include "yuzu/ui_settings.h"
 
 class QSettings;
 
@@ -47,6 +48,8 @@ private:
     void WriteSetting(const QString& name, const QVariant& value);
     void WriteSetting(const QString& name, const QVariant& value, const QVariant& default_value);
 
+    static const std::array<UISettings::Shortcut, 15> default_hotkeys;
+
     std::unique_ptr<QSettings> qt_config;
     std::string qt_config_loc;
 };

src/yuzu/configuration/configure.ui

@@ -7,9 +7,15 @@
     <x>0</x>
     <y>0</y>
     <width>382</width>
-    <height>241</height>
+    <height>650</height>
    </rect>
   </property>
+  <property name="minimumSize">
+   <size>
+    <width>0</width>
+    <height>650</height>
+   </size>
+  </property>
   <property name="windowTitle">
    <string>yuzu Configuration</string>
   </property>
@@ -62,6 +68,11 @@
          <string>Input</string>
         </attribute>
        </widget>
+       <widget class="ConfigureHotkeys" name="hotkeysTab">
+        <attribute name="title">
+         <string>Hotkeys</string>
+        </attribute>
+       </widget>
        <widget class="ConfigureGraphics" name="graphicsTab">
         <attribute name="title">
          <string>Graphics</string>
@@ -150,6 +161,12 @@
    <header>configuration/configure_input_simple.h</header>
    <container>1</container>
   </customwidget>
+  <customwidget>
+   <class>ConfigureHotkeys</class>
+   <extends>QWidget</extends>
+   <header>configuration/configure_hotkeys.h</header>
+   <container>1</container>
+  </customwidget>
  </customwidgets>
  <resources/>
  <connections>

src/yuzu/configuration/configure_dialog.cpp

@@ -8,20 +8,22 @@
 #include "ui_configure.h"
 #include "yuzu/configuration/config.h"
 #include "yuzu/configuration/configure_dialog.h"
+#include "yuzu/configuration/configure_input_player.h"
 #include "yuzu/hotkeys.h"
 
-ConfigureDialog::ConfigureDialog(QWidget* parent, const HotkeyRegistry& registry)
-    : QDialog(parent), ui(new Ui::ConfigureDialog) {
+ConfigureDialog::ConfigureDialog(QWidget* parent, HotkeyRegistry& registry)
+    : QDialog(parent), registry(registry), ui(new Ui::ConfigureDialog) {
     ui->setupUi(this);
-    ui->generalTab->PopulateHotkeyList(registry);
+    ui->hotkeysTab->Populate(registry);
     this->setConfiguration();
     this->PopulateSelectionList();
     connect(ui->selectorList, &QListWidget::itemSelectionChanged, this,
             &ConfigureDialog::UpdateVisibleTabs);
-
     adjustSize();
-
     ui->selectorList->setCurrentRow(0);
+
+    // Synchronise lists upon initialisation
+    ui->hotkeysTab->EmitHotkeysChanged();
 }
 
 ConfigureDialog::~ConfigureDialog() = default;
@@ -34,6 +36,7 @@ void ConfigureDialog::applyConfiguration() {
     ui->systemTab->applyConfiguration();
     ui->profileManagerTab->applyConfiguration();
     ui->inputTab->applyConfiguration();
+    ui->hotkeysTab->applyConfiguration(registry);
     ui->graphicsTab->applyConfiguration();
     ui->audioTab->applyConfiguration();
     ui->debugTab->applyConfiguration();
@@ -47,7 +50,7 @@ void ConfigureDialog::PopulateSelectionList() {
         {{tr("General"), {tr("General"), tr("Web"), tr("Debug"), tr("Game List")}},
          {tr("System"), {tr("System"), tr("Profiles"), tr("Audio")}},
          {tr("Graphics"), {tr("Graphics")}},
-         {tr("Controls"), {tr("Input")}}}};
+         {tr("Controls"), {tr("Input"), tr("Hotkeys")}}}};
 
     for (const auto& entry : items) {
         auto* const item = new QListWidgetItem(entry.first);
@@ -66,6 +69,7 @@ void ConfigureDialog::UpdateVisibleTabs() {
                                                  {tr("System"), ui->systemTab},
                                                  {tr("Profiles"), ui->profileManagerTab},
                                                  {tr("Input"), ui->inputTab},
+                                                 {tr("Hotkeys"), ui->hotkeysTab},
                                                  {tr("Graphics"), ui->graphicsTab},
                                                  {tr("Audio"), ui->audioTab},
                                                  {tr("Debug"), ui->debugTab},

src/yuzu/configuration/configure_dialog.h

@@ -17,7 +17,7 @@ class ConfigureDialog : public QDialog {
     Q_OBJECT
 
 public:
-    explicit ConfigureDialog(QWidget* parent, const HotkeyRegistry& registry);
+    explicit ConfigureDialog(QWidget* parent, HotkeyRegistry& registry);
     ~ConfigureDialog() override;
 
     void applyConfiguration();
@@ -28,4 +28,5 @@ private:
     void PopulateSelectionList();
 
     std::unique_ptr<Ui::ConfigureDialog> ui;
+    HotkeyRegistry& registry;
 };

src/yuzu/configuration/configure_general.cpp

@@ -35,10 +35,6 @@ void ConfigureGeneral::setConfiguration() {
     ui->use_cpu_jit->setChecked(Settings::values.use_cpu_jit);
 }
 
-void ConfigureGeneral::PopulateHotkeyList(const HotkeyRegistry& registry) {
-    ui->widget->Populate(registry);
-}
-
 void ConfigureGeneral::applyConfiguration() {
     UISettings::values.gamedir_deepscan = ui->toggle_deepscan->isChecked();
     UISettings::values.confirm_before_closing = ui->toggle_check_exit->isChecked();

src/yuzu/configuration/configure_general.h

@@ -20,7 +20,6 @@ public:
     explicit ConfigureGeneral(QWidget* parent = nullptr);
     ~ConfigureGeneral() override;
 
-    void PopulateHotkeyList(const HotkeyRegistry& registry);
     void applyConfiguration();
 
 private:

src/yuzu/configuration/configure_general.ui

@@ -97,22 +97,6 @@
        </layout>
       </widget>
      </item>
-     <item>
-      <widget class="QGroupBox" name="HotKeysGroupBox">
-       <property name="title">
-        <string>Hotkeys</string>
-       </property>
-       <layout class="QHBoxLayout" name="HotKeysHorizontalLayout">
-        <item>
-         <layout class="QVBoxLayout" name="HotKeysVerticalLayout">
-          <item>
-           <widget class="GHotkeysDialog" name="widget" native="true"/>
-          </item>
-         </layout>
-        </item>
-       </layout>
-      </widget>
-     </item>
      <item>
       <spacer name="verticalSpacer">
        <property name="orientation">
@@ -130,14 +114,6 @@
    </item>
   </layout>
  </widget>
- <customwidgets>
-  <customwidget>
-   <class>GHotkeysDialog</class>
-   <extends>QWidget</extends>
-   <header>hotkeys.h</header>
-   <container>1</container>
-  </customwidget>
- </customwidgets>
  <resources/>
  <connections/>
 </ui>

src/yuzu/configuration/configure_hotkeys.cpp

@@ -0,0 +1,121 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <QMessageBox>
+#include <QStandardItemModel>
+#include "core/settings.h"
+#include "ui_configure_hotkeys.h"
+#include "yuzu/configuration/configure_hotkeys.h"
+#include "yuzu/hotkeys.h"
+#include "yuzu/util/sequence_dialog/sequence_dialog.h"
+
+ConfigureHotkeys::ConfigureHotkeys(QWidget* parent)
+    : QWidget(parent), ui(std::make_unique<Ui::ConfigureHotkeys>()) {
+    ui->setupUi(this);
+    setFocusPolicy(Qt::ClickFocus);
+
+    model = new QStandardItemModel(this);
+    model->setColumnCount(3);
+    model->setHorizontalHeaderLabels({tr("Action"), tr("Hotkey"), tr("Context")});
+
+    connect(ui->hotkey_list, &QTreeView::doubleClicked, this, &ConfigureHotkeys::Configure);
+    ui->hotkey_list->setModel(model);
+
+    // TODO(Kloen): Make context configurable as well (hiding the column for now)
+    ui->hotkey_list->hideColumn(2);
+
+    ui->hotkey_list->setColumnWidth(0, 200);
+    ui->hotkey_list->resizeColumnToContents(1);
+}
+
+ConfigureHotkeys::~ConfigureHotkeys() = default;
+
+void ConfigureHotkeys::EmitHotkeysChanged() {
+    emit HotkeysChanged(GetUsedKeyList());
+}
+
+QList<QKeySequence> ConfigureHotkeys::GetUsedKeyList() const {
+    QList<QKeySequence> list;
+    for (int r = 0; r < model->rowCount(); r++) {
+        const QStandardItem* parent = model->item(r, 0);
+        for (int r2 = 0; r2 < parent->rowCount(); r2++) {
+            const QStandardItem* keyseq = parent->child(r2, 1);
+            list << QKeySequence::fromString(keyseq->text(), QKeySequence::NativeText);
+        }
+    }
+    return list;
+}
+
+void ConfigureHotkeys::Populate(const HotkeyRegistry& registry) {
+    for (const auto& group : registry.hotkey_groups) {
+        auto* parent_item = new QStandardItem(group.first);
+        parent_item->setEditable(false);
+        for (const auto& hotkey : group.second) {
+            auto* action = new QStandardItem(hotkey.first);
+            auto* keyseq =
+                new QStandardItem(hotkey.second.keyseq.toString(QKeySequence::NativeText));
+            action->setEditable(false);
+            keyseq->setEditable(false);
+            parent_item->appendRow({action, keyseq});
+        }
+        model->appendRow(parent_item);
+    }
+
+    ui->hotkey_list->expandAll();
+}
+
+void ConfigureHotkeys::Configure(QModelIndex index) {
+    if (index.parent() == QModelIndex())
+        return;
+
+    index = index.sibling(index.row(), 1);
+    auto* model = ui->hotkey_list->model();
+    auto previous_key = model->data(index);
+
+    auto* hotkey_dialog = new SequenceDialog;
+    int return_code = hotkey_dialog->exec();
+
+    auto key_sequence = hotkey_dialog->GetSequence();
+
+    if (return_code == QDialog::Rejected || key_sequence.isEmpty())
+        return;
+
+    if (IsUsedKey(key_sequence) && key_sequence != QKeySequence(previous_key.toString())) {
+        QMessageBox::critical(this, tr("Error in inputted key"),
+                              tr("You're using a key that's already bound."));
+    } else {
+        model->setData(index, key_sequence.toString(QKeySequence::NativeText));
+        EmitHotkeysChanged();
+    }
+}
+
+bool ConfigureHotkeys::IsUsedKey(QKeySequence key_sequence) {
+    return GetUsedKeyList().contains(key_sequence);
+}
+
+void ConfigureHotkeys::applyConfiguration(HotkeyRegistry& registry) {
+    for (int key_id = 0; key_id < model->rowCount(); key_id++) {
+        const QStandardItem* parent = model->item(key_id, 0);
+        for (int key_column_id = 0; key_column_id < parent->rowCount(); key_column_id++) {
+            const QStandardItem* action = parent->child(key_column_id, 0);
+            const QStandardItem* keyseq = parent->child(key_column_id, 1);
+            for (auto& [group, sub_actions] : registry.hotkey_groups) {
+                if (group != parent->text())
+                    continue;
+                for (auto& [action_name, hotkey] : sub_actions) {
+                    if (action_name != action->text())
+                        continue;
+                    hotkey.keyseq = QKeySequence(keyseq->text());
+                }
+            }
+        }
+    }
+
+    registry.SaveHotkeys();
+    Settings::Apply();
+}
+
+void ConfigureHotkeys::retranslateUi() {
+    ui->retranslateUi(this);
+}

src/yuzu/configuration/configure_hotkeys.h

@@ -0,0 +1,48 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <QWidget>
+#include "core/settings.h"
+
+namespace Ui {
+class ConfigureHotkeys;
+}
+
+class HotkeyRegistry;
+class QStandardItemModel;
+
+class ConfigureHotkeys : public QWidget {
+    Q_OBJECT
+
+public:
+    explicit ConfigureHotkeys(QWidget* parent = nullptr);
+    ~ConfigureHotkeys() override;
+
+    void applyConfiguration(HotkeyRegistry& registry);
+    void retranslateUi();
+
+    void EmitHotkeysChanged();
+
+    /**
+     * Populates the hotkey list widget using data from the provided registry.
+     * Called everytime the Configure dialog is opened.
+     * @param registry The HotkeyRegistry whose data is used to populate the list.
+     */
+    void Populate(const HotkeyRegistry& registry);
+
+signals:
+    void HotkeysChanged(QList<QKeySequence> new_key_list);
+
+private:
+    void Configure(QModelIndex index);
+    bool IsUsedKey(QKeySequence key_sequence);
+    QList<QKeySequence> GetUsedKeyList() const;
+
+    std::unique_ptr<Ui::ConfigureHotkeys> ui;
+
+    QStandardItemModel* model;
+};

src/yuzu/configuration/configure_hotkeys.ui

@@ -0,0 +1,42 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<ui version="4.0">
+ <class>ConfigureHotkeys</class>
+ <widget class="QWidget" name="ConfigureHotkeys">
+  <property name="geometry">
+   <rect>
+    <x>0</x>
+    <y>0</y>
+    <width>363</width>
+    <height>388</height>
+   </rect>
+  </property>
+  <property name="windowTitle">
+   <string>Hotkey Settings</string>
+  </property>
+  <layout class="QVBoxLayout" name="verticalLayout">
+   <item>
+    <layout class="QVBoxLayout" name="verticalLayout_2">
+     <item>
+      <widget class="QLabel" name="label_2">
+       <property name="text">
+        <string>Double-click on a binding to change it.</string>
+       </property>
+      </widget>
+     </item>
+     <item>
+      <widget class="QTreeView" name="hotkey_list">
+       <property name="editTriggers">
+        <set>QAbstractItemView::NoEditTriggers</set>
+       </property>
+       <property name="sortingEnabled">
+        <bool>false</bool>
+       </property>
+      </widget>
+     </item>
+    </layout>
+   </item>
+  </layout>
+ </widget>
+ <resources/>
+ <connections/>
+</ui>

src/yuzu/game_list.cpp

@@ -18,6 +18,7 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/file_sys/patch_manager.h"
+#include "core/file_sys/registered_cache.h"
 #include "yuzu/compatibility_list.h"
 #include "yuzu/game_list.h"
 #include "yuzu/game_list_p.h"
@@ -193,8 +194,9 @@ void GameList::onFilterCloseClicked() {
     main_window->filterBarSetChecked(false);
 }
 
-GameList::GameList(FileSys::VirtualFilesystem vfs, GMainWindow* parent)
-    : QWidget{parent}, vfs(std::move(vfs)) {
+GameList::GameList(FileSys::VirtualFilesystem vfs, FileSys::ManualContentProvider* provider,
+                   GMainWindow* parent)
+    : QWidget{parent}, vfs(std::move(vfs)), provider(provider) {
     watcher = new QFileSystemWatcher(this);
     connect(watcher, &QFileSystemWatcher::directoryChanged, this, &GameList::RefreshGameDirectory);
 
@@ -432,7 +434,8 @@ void GameList::PopulateAsync(const QString& dir_path, bool deep_scan) {
 
     emit ShouldCancelWorker();
 
-    GameListWorker* worker = new GameListWorker(vfs, dir_path, deep_scan, compatibility_list);
+    GameListWorker* worker =
+        new GameListWorker(vfs, provider, dir_path, deep_scan, compatibility_list);
 
     connect(worker, &GameListWorker::EntryReady, this, &GameList::AddEntry, Qt::QueuedConnection);
     connect(worker, &GameListWorker::Finished, this, &GameList::DonePopulating,

src/yuzu/game_list.h

@@ -26,8 +26,9 @@ class GameListSearchField;
 class GMainWindow;
 
 namespace FileSys {
+class ManualContentProvider;
 class VfsFilesystem;
-}
+} // namespace FileSys
 
 enum class GameListOpenTarget {
     SaveData,
@@ -47,7 +48,8 @@ public:
         COLUMN_COUNT, // Number of columns
     };
 
-    explicit GameList(std::shared_ptr<FileSys::VfsFilesystem> vfs, GMainWindow* parent = nullptr);
+    explicit GameList(std::shared_ptr<FileSys::VfsFilesystem> vfs,
+                      FileSys::ManualContentProvider* provider, GMainWindow* parent = nullptr);
     ~GameList() override;
 
     void clearFilter();
@@ -86,6 +88,7 @@ private:
     void RefreshGameDirectory();
 
     std::shared_ptr<FileSys::VfsFilesystem> vfs;
+    FileSys::ManualContentProvider* provider;
     GameListSearchField* search_field;
     GMainWindow* main_window = nullptr;
     QVBoxLayout* layout = nullptr;

src/yuzu/game_list_worker.cpp

@@ -12,12 +12,15 @@
 
 #include "common/common_paths.h"
 #include "common/file_util.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"
 #include "core/file_sys/mode.h"
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/patch_manager.h"
 #include "core/file_sys/registered_cache.h"
+#include "core/file_sys/submission_package.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/loader/loader.h"
 #include "yuzu/compatibility_list.h"
@@ -119,20 +122,25 @@ QList<QStandardItem*> MakeGameListEntry(const std::string& path, const std::stri
 }
 } // Anonymous namespace
 
-GameListWorker::GameListWorker(FileSys::VirtualFilesystem vfs, QString dir_path, bool deep_scan,
-                               const CompatibilityList& compatibility_list)
-    : vfs(std::move(vfs)), dir_path(std::move(dir_path)), deep_scan(deep_scan),
+GameListWorker::GameListWorker(FileSys::VirtualFilesystem vfs,
+                               FileSys::ManualContentProvider* provider, QString dir_path,
+                               bool deep_scan, const CompatibilityList& compatibility_list)
+    : vfs(std::move(vfs)), provider(provider), dir_path(std::move(dir_path)), deep_scan(deep_scan),
       compatibility_list(compatibility_list) {}
 
 GameListWorker::~GameListWorker() = default;
 
-void GameListWorker::AddInstalledTitlesToGameList() {
-    const auto cache = Service::FileSystem::GetUnionContents();
-    const auto installed_games = cache.ListEntriesFilter(FileSys::TitleType::Application,
-                                                         FileSys::ContentRecordType::Program);
+void GameListWorker::AddTitlesToGameList() {
+    const auto& cache = dynamic_cast<FileSys::ContentProviderUnion&>(
+        Core::System::GetInstance().GetContentProvider());
+    const auto installed_games = cache.ListEntriesFilterOrigin(
+        std::nullopt, FileSys::TitleType::Application, FileSys::ContentRecordType::Program);
 
-    for (const auto& game : installed_games) {
-        const auto file = cache.GetEntryUnparsed(game);
+    for (const auto& [slot, game] : installed_games) {
+        if (slot == FileSys::ContentProviderUnionSlot::FrontendManual)
+            continue;
+
+        const auto file = cache.GetEntryUnparsed(game.title_id, game.type);
         std::unique_ptr<Loader::AppLoader> loader = Loader::GetLoader(file);
         if (!loader)
             continue;
@@ -150,45 +158,13 @@ void GameListWorker::AddInstalledTitlesToGameList() {
         emit EntryReady(MakeGameListEntry(file->GetFullPath(), name, icon, *loader, program_id,
                                           compatibility_list, patch));
     }
-
-    const auto control_data = cache.ListEntriesFilter(FileSys::TitleType::Application,
-                                                      FileSys::ContentRecordType::Control);
-
-    for (const auto& entry : control_data) {
-        auto nca = cache.GetEntry(entry);
-        if (nca != nullptr) {
-            nca_control_map.insert_or_assign(entry.title_id, std::move(nca));
-        }
-    }
 }
 
-void GameListWorker::FillControlMap(const std::string& dir_path) {
-    const auto nca_control_callback = [this](u64* num_entries_out, const std::string& directory,
-                                             const std::string& virtual_name) -> bool {
-        if (stop_processing) {
-            // Breaks the callback loop
-            return false;
-        }
-
-        const std::string physical_name = directory + DIR_SEP + virtual_name;
-        const QFileInfo file_info(QString::fromStdString(physical_name));
-        if (!file_info.isDir() && file_info.suffix() == QStringLiteral("nca")) {
-            auto nca =
-                std::make_unique<FileSys::NCA>(vfs->OpenFile(physical_name, FileSys::Mode::Read));
-            if (nca->GetType() == FileSys::NCAContentType::Control) {
-                const u64 title_id = nca->GetTitleId();
-                nca_control_map.insert_or_assign(title_id, std::move(nca));
-            }
-        }
-        return true;
-    };
-
-    FileUtil::ForeachDirectoryEntry(nullptr, dir_path, nca_control_callback);
-}
-
-void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsigned int recursion) {
-    const auto callback = [this, recursion](u64* num_entries_out, const std::string& directory,
-                                            const std::string& virtual_name) -> bool {
+void GameListWorker::ScanFileSystem(ScanTarget target, const std::string& dir_path,
+                                    unsigned int recursion) {
+    const auto callback = [this, target, recursion](u64* num_entries_out,
+                                                    const std::string& directory,
+                                                    const std::string& virtual_name) -> bool {
         if (stop_processing) {
             // Breaks the callback loop.
             return false;
@@ -198,7 +174,8 @@ void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsign
         const bool is_dir = FileUtil::IsDirectory(physical_name);
         if (!is_dir &&
             (HasSupportedFileExtension(physical_name) || IsExtractedNCAMain(physical_name))) {
-            auto loader = Loader::GetLoader(vfs->OpenFile(physical_name, FileSys::Mode::Read));
+            const auto file = vfs->OpenFile(physical_name, FileSys::Mode::Read);
+            auto loader = Loader::GetLoader(file);
             if (!loader) {
                 return true;
             }
@@ -209,31 +186,42 @@ void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsign
                 return true;
             }
 
-            std::vector<u8> icon;
-            const auto res1 = loader->ReadIcon(icon);
-
             u64 program_id = 0;
             const auto res2 = loader->ReadProgramId(program_id);
 
-            std::string name = " ";
-            const auto res3 = loader->ReadTitle(name);
-
-            const FileSys::PatchManager patch{program_id};
-
-            if (res1 != Loader::ResultStatus::Success && res3 != Loader::ResultStatus::Success &&
-                res2 == Loader::ResultStatus::Success) {
-                // Use from metadata pool.
-                if (nca_control_map.find(program_id) != nca_control_map.end()) {
-                    const auto& nca = nca_control_map[program_id];
-                    GetMetadataFromControlNCA(patch, *nca, icon, name);
+            if (target == ScanTarget::FillManualContentProvider) {
+                if (res2 == Loader::ResultStatus::Success && file_type == Loader::FileType::NCA) {
+                    provider->AddEntry(FileSys::TitleType::Application,
+                                       FileSys::GetCRTypeFromNCAType(FileSys::NCA{file}.GetType()),
+                                       program_id, file);
+                } else if (res2 == Loader::ResultStatus::Success &&
+                           (file_type == Loader::FileType::XCI ||
+                            file_type == Loader::FileType::NSP)) {
+                    const auto nsp = file_type == Loader::FileType::NSP
+                                         ? std::make_shared<FileSys::NSP>(file)
+                                         : FileSys::XCI{file}.GetSecurePartitionNSP();
+                    for (const auto& title : nsp->GetNCAs()) {
+                        for (const auto& entry : title.second) {
+                            provider->AddEntry(entry.first.first, entry.first.second, title.first,
+                                               entry.second->GetBaseFile());
+                        }
+                    }
                 }
-            }
+            } else {
+                std::vector<u8> icon;
+                const auto res1 = loader->ReadIcon(icon);
 
-            emit EntryReady(MakeGameListEntry(physical_name, name, icon, *loader, program_id,
-                                              compatibility_list, patch));
+                std::string name = " ";
+                const auto res3 = loader->ReadTitle(name);
+
+                const FileSys::PatchManager patch{program_id};
+
+                emit EntryReady(MakeGameListEntry(physical_name, name, icon, *loader, program_id,
+                                                  compatibility_list, patch));
+            }
         } else if (is_dir && recursion > 0) {
             watch_list.append(QString::fromStdString(physical_name));
-            AddFstEntriesToGameList(physical_name, recursion - 1);
+            ScanFileSystem(target, physical_name, recursion - 1);
         }
 
         return true;
@@ -245,10 +233,11 @@ void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsign
 void GameListWorker::run() {
     stop_processing = false;
     watch_list.append(dir_path);
-    FillControlMap(dir_path.toStdString());
-    AddInstalledTitlesToGameList();
-    AddFstEntriesToGameList(dir_path.toStdString(), deep_scan ? 256 : 0);
-    nca_control_map.clear();
+    provider->ClearAllEntries();
+    ScanFileSystem(ScanTarget::FillManualContentProvider, dir_path.toStdString(),
+                   deep_scan ? 256 : 0);
+    AddTitlesToGameList();
+    ScanFileSystem(ScanTarget::PopulateGameList, dir_path.toStdString(), deep_scan ? 256 : 0);
     emit Finished(watch_list);
 }
 

src/yuzu/game_list_worker.h

@@ -33,7 +33,8 @@ class GameListWorker : public QObject, public QRunnable {
     Q_OBJECT
 
 public:
-    GameListWorker(std::shared_ptr<FileSys::VfsFilesystem> vfs, QString dir_path, bool deep_scan,
+    GameListWorker(std::shared_ptr<FileSys::VfsFilesystem> vfs,
+                   FileSys::ManualContentProvider* provider, QString dir_path, bool deep_scan,
                    const CompatibilityList& compatibility_list);
     ~GameListWorker() override;
 
@@ -58,12 +59,17 @@ signals:
     void Finished(QStringList watch_list);
 
 private:
-    void AddInstalledTitlesToGameList();
-    void FillControlMap(const std::string& dir_path);
-    void AddFstEntriesToGameList(const std::string& dir_path, unsigned int recursion = 0);
+    void AddTitlesToGameList();
+
+    enum class ScanTarget {
+        FillManualContentProvider,
+        PopulateGameList,
+    };
+
+    void ScanFileSystem(ScanTarget target, const std::string& dir_path, unsigned int recursion = 0);
 
     std::shared_ptr<FileSys::VfsFilesystem> vfs;
-    std::map<u64, std::unique_ptr<FileSys::NCA>> nca_control_map;
+    FileSys::ManualContentProvider* provider;
     QStringList watch_list;
     QString dir_path;
     bool deep_scan;