implement invert

gpu: Make memory_manager private

src/common/file_util.cpp

@@ -764,7 +764,7 @@ size_t ReadFileToString(bool text_file, const char* filename, std::string& str)
     IOFile file(filename, text_file ? "r" : "rb");
 
     if (!file.IsOpen())
-        return false;
+        return 0;
 
     str.resize(static_cast<u32>(file.GetSize()));
     return file.ReadArray(&str[0], str.size());

src/common/hex_util.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include "common/hex_util.h"
+#include "common/logging/log.h"
 
 namespace Common {
 
@@ -13,18 +14,29 @@ u8 ToHexNibble(char c1) {
         return c1 - 87;
     if (c1 >= 48 && c1 <= 57)
         return c1 - 48;
-    throw std::logic_error("Invalid hex digit");
+    LOG_ERROR(Common, "Invalid hex digit: 0x{:02X}", c1);
+    return 0;
 }
 
 std::array<u8, 16> operator""_array16(const char* str, size_t len) {
-    if (len != 32)
-        throw std::logic_error("Not of correct size.");
+    if (len != 32) {
+        LOG_ERROR(Common,
+                  "Attempting to parse string to array that is not of correct size (expected=32, "
+                  "actual={}).",
+                  len);
+        return {};
+    }
     return HexStringToArray<16>(str);
 }
 
 std::array<u8, 32> operator""_array32(const char* str, size_t len) {
-    if (len != 64)
-        throw std::logic_error("Not of correct size.");
+    if (len != 64) {
+        LOG_ERROR(Common,
+                  "Attempting to parse string to array that is not of correct size (expected=64, "
+                  "actual={}).",
+                  len);
+        return {};
+    }
     return HexStringToArray<32>(str);
 }
 

src/core/CMakeLists.txt

@@ -20,6 +20,8 @@ add_library(core STATIC
     crypto/key_manager.h
     crypto/ctr_encryption_layer.cpp
     crypto/ctr_encryption_layer.h
+    crypto/xts_encryption_layer.cpp
+    crypto/xts_encryption_layer.h
     file_sys/bis_factory.cpp
     file_sys/bis_factory.h
     file_sys/card_image.cpp
@@ -57,6 +59,8 @@ add_library(core STATIC
     file_sys/vfs_real.h
     file_sys/vfs_vector.cpp
     file_sys/vfs_vector.h
+    file_sys/xts_archive.cpp
+    file_sys/xts_archive.h
     frontend/emu_window.cpp
     frontend/emu_window.h
     frontend/framebuffer_layout.cpp
@@ -347,6 +351,8 @@ add_library(core STATIC
     loader/linker.h
     loader/loader.cpp
     loader/loader.h
+    loader/nax.cpp
+    loader/nax.h
     loader/nca.cpp
     loader/nca.h
     loader/nro.cpp

src/core/arm/arm_interface.h

@@ -8,6 +8,8 @@
 #include "common/common_types.h"
 #include "core/hle/kernel/vm_manager.h"
 
+namespace Core {
+
 /// Generic ARM11 CPU interface
 class ARM_Interface : NonCopyable {
 public:
@@ -122,3 +124,5 @@ public:
     /// Prepare core for thread reschedule (if needed to correctly handle state)
     virtual void PrepareReschedule() = 0;
 };
+
+} // namespace Core

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -14,6 +14,8 @@
 #include "core/hle/kernel/svc.h"
 #include "core/memory.h"
 
+namespace Core {
+
 using Vector = Dynarmic::A64::Vector;
 
 class ARM_Dynarmic_Callbacks : public Dynarmic::A64::UserCallbacks {
@@ -300,3 +302,5 @@ bool DynarmicExclusiveMonitor::ExclusiveWrite128(size_t core_index, VAddr vaddr,
         Memory::Write64(vaddr, value[1]);
     });
 }
+
+} // namespace Core

src/core/arm/dynarmic/arm_dynarmic.h

@@ -12,6 +12,8 @@
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
 
+namespace Core {
+
 class ARM_Dynarmic_Callbacks;
 class DynarmicExclusiveMonitor;
 
@@ -81,3 +83,5 @@ private:
     friend class ARM_Dynarmic;
     Dynarmic::A64::ExclusiveMonitor monitor;
 };
+
+} // namespace Core

src/core/arm/exclusive_monitor.cpp

@@ -4,4 +4,8 @@
 
 #include "core/arm/exclusive_monitor.h"
 
+namespace Core {
+
 ExclusiveMonitor::~ExclusiveMonitor() = default;
+
+} // namespace Core

src/core/arm/exclusive_monitor.h

@@ -6,6 +6,8 @@
 
 #include "common/common_types.h"
 
+namespace Core {
+
 class ExclusiveMonitor {
 public:
     virtual ~ExclusiveMonitor();
@@ -19,3 +21,5 @@ public:
     virtual bool ExclusiveWrite64(size_t core_index, VAddr vaddr, u64 value) = 0;
     virtual bool ExclusiveWrite128(size_t core_index, VAddr vaddr, u128 value) = 0;
 };
+
+} // namespace Core

src/core/arm/unicorn/arm_unicorn.cpp

@@ -11,6 +11,8 @@
 #include "core/core_timing.h"
 #include "core/hle/kernel/svc.h"
 
+namespace Core {
+
 // Load Unicorn DLL once on Windows using RAII
 #ifdef _MSC_VER
 #include <unicorn_dynload.h>
@@ -211,7 +213,7 @@ void ARM_Unicorn::ExecuteInstructions(int num_instructions) {
     }
 }
 
-void ARM_Unicorn::SaveContext(ARM_Interface::ThreadContext& ctx) {
+void ARM_Unicorn::SaveContext(ThreadContext& ctx) {
     int uregs[32];
     void* tregs[32];
 
@@ -238,7 +240,7 @@ void ARM_Unicorn::SaveContext(ARM_Interface::ThreadContext& ctx) {
     CHECKED(uc_reg_read_batch(uc, uregs, tregs, 32));
 }
 
-void ARM_Unicorn::LoadContext(const ARM_Interface::ThreadContext& ctx) {
+void ARM_Unicorn::LoadContext(const ThreadContext& ctx) {
     int uregs[32];
     void* tregs[32];
 
@@ -277,3 +279,5 @@ void ARM_Unicorn::RecordBreak(GDBStub::BreakpointAddress bkpt) {
     last_bkpt = bkpt;
     last_bkpt_hit = true;
 }
+
+} // namespace Core

src/core/arm/unicorn/arm_unicorn.h

@@ -9,6 +9,8 @@
 #include "core/arm/arm_interface.h"
 #include "core/gdbstub/gdbstub.h"
 
+namespace Core {
+
 class ARM_Unicorn final : public ARM_Interface {
 public:
     ARM_Unicorn();
@@ -46,3 +48,5 @@ private:
     GDBStub::BreakpointAddress last_bkpt{};
     bool last_bkpt_hit;
 };
+
+} // namespace Core

src/core/core.cpp

@@ -135,8 +135,7 @@ System::ResultStatus System::Load(Frontend::EmuWindow& emu_window, const std::st
         LOG_CRITICAL(Core, "Failed to determine system mode (Error {})!",
                      static_cast<int>(system_mode.second));
 
-        if (system_mode.second != Loader::ResultStatus::Success)
-            return ResultStatus::ErrorSystemMode;
+        return ResultStatus::ErrorSystemMode;
     }
 
     ResultStatus init_result{Init(emu_window)};
@@ -148,14 +147,12 @@ System::ResultStatus System::Load(Frontend::EmuWindow& emu_window, const std::st
     }
 
     const Loader::ResultStatus load_result{app_loader->Load(current_process)};
-    if (Loader::ResultStatus::Success != load_result) {
+    if (load_result != Loader::ResultStatus::Success) {
         LOG_CRITICAL(Core, "Failed to load ROM (Error {})!", static_cast<int>(load_result));
         System::Shutdown();
 
-        if (load_result != Loader::ResultStatus::Success) {
-            return static_cast<ResultStatus>(static_cast<u32>(ResultStatus::ErrorLoader) +
-                                             static_cast<u32>(load_result));
-        }
+        return static_cast<ResultStatus>(static_cast<u32>(ResultStatus::ErrorLoader) +
+                                         static_cast<u32>(load_result));
     }
     status = ResultStatus::Success;
     return status;

src/core/core.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <map>
 #include <memory>
 #include <string>
 #include <thread>
@@ -22,8 +23,6 @@
 #include "video_core/debug_utils/debug_utils.h"
 #include "video_core/gpu.h"
 
-class ARM_Interface;
-
 namespace Core::Frontend {
 class EmuWindow;
 }
@@ -38,6 +37,8 @@ class RendererBase;
 
 namespace Core {
 
+class ARM_Interface;
+
 class System {
 public:
     System(const System&) = delete;
@@ -187,6 +188,13 @@ public:
         return current_process;
     }
 
+    /// Gets the name of the current game
+    Loader::ResultStatus GetGameName(std::string& out) const {
+        if (app_loader == nullptr)
+            return Loader::ResultStatus::ErrorNotInitialized;
+        return app_loader->ReadTitle(out);
+    }
+
     PerfStats perf_stats;
     FrameLimiter frame_limiter;
 

src/core/core_cpu.h

@@ -12,14 +12,14 @@
 #include "common/common_types.h"
 #include "core/arm/exclusive_monitor.h"
 
-class ARM_Interface;
-
 namespace Kernel {
 class Scheduler;
 }
 
 namespace Core {
 
+class ARM_Interface;
+
 constexpr unsigned NUM_CPU_CORES{4};
 
 class CpuBarrier {

src/core/crypto/aes_util.cpp

@@ -99,10 +99,7 @@ void AESCipher<Key, KeySize>::Transcode(const u8* src, size_t size, u8* dest, Op
 template <typename Key, size_t KeySize>
 void AESCipher<Key, KeySize>::XTSTranscode(const u8* src, size_t size, u8* dest, size_t sector_id,
                                            size_t sector_size, Op op) {
-    if (size % sector_size > 0) {
-        LOG_CRITICAL(Crypto, "Data size must be a multiple of sector size.");
-        return;
-    }
+    ASSERT_MSG(size % sector_size == 0, "XTS decryption size must be a multiple of sector size.");
 
     for (size_t i = 0; i < size; i += sector_size) {
         SetIV(CalculateNintendoTweak(sector_id++));
@@ -112,4 +109,4 @@ void AESCipher<Key, KeySize>::XTSTranscode(const u8* src, size_t size, u8* dest,
 
 template class AESCipher<Key128>;
 template class AESCipher<Key256>;
-} // namespace Core::Crypto
+} // namespace Core::Crypto

src/core/crypto/ctr_encryption_layer.cpp

@@ -20,10 +20,8 @@ size_t CTREncryptionLayer::Read(u8* data, size_t length, size_t offset) const {
     if (sector_offset == 0) {
         UpdateIV(base_offset + offset);
         std::vector<u8> raw = base->ReadBytes(length, offset);
-        if (raw.size() != length)
-            return Read(data, raw.size(), offset);
-        cipher.Transcode(raw.data(), length, data, Op::Decrypt);
-        return length;
+        cipher.Transcode(raw.data(), raw.size(), data, Op::Decrypt);
+        return raw.size();
     }
 
     // offset does not fall on block boundary (0x10)
@@ -34,7 +32,7 @@ size_t CTREncryptionLayer::Read(u8* data, size_t length, size_t offset) const {
 
     if (length + sector_offset < 0x10) {
         std::memcpy(data, block.data() + sector_offset, std::min<u64>(length, read));
-        return read;
+        return std::min<u64>(length, read);
     }
     std::memcpy(data, block.data() + sector_offset, read);
     return read + Read(data + read, length - read, offset + read);

src/core/crypto/key_manager.cpp

@@ -12,11 +12,112 @@
 #include "common/file_util.h"
 #include "common/hex_util.h"
 #include "common/logging/log.h"
+#include "core/crypto/aes_util.h"
 #include "core/crypto/key_manager.h"
 #include "core/settings.h"
 
 namespace Core::Crypto {
 
+Key128 GenerateKeyEncryptionKey(Key128 source, Key128 master, Key128 kek_seed, Key128 key_seed) {
+    Key128 out{};
+
+    AESCipher<Key128> cipher1(master, Mode::ECB);
+    cipher1.Transcode(kek_seed.data(), kek_seed.size(), out.data(), Op::Decrypt);
+    AESCipher<Key128> cipher2(out, Mode::ECB);
+    cipher2.Transcode(source.data(), source.size(), out.data(), Op::Decrypt);
+
+    if (key_seed != Key128{}) {
+        AESCipher<Key128> cipher3(out, Mode::ECB);
+        cipher3.Transcode(key_seed.data(), key_seed.size(), out.data(), Op::Decrypt);
+    }
+
+    return out;
+}
+
+boost::optional<Key128> DeriveSDSeed() {
+    const FileUtil::IOFile save_43(FileUtil::GetUserPath(FileUtil::UserPath::NANDDir) +
+                                       "/system/save/8000000000000043",
+                                   "rb+");
+    if (!save_43.IsOpen())
+        return boost::none;
+    const FileUtil::IOFile sd_private(
+        FileUtil::GetUserPath(FileUtil::UserPath::SDMCDir) + "/Nintendo/Contents/private", "rb+");
+    if (!sd_private.IsOpen())
+        return boost::none;
+
+    sd_private.Seek(0, SEEK_SET);
+    std::array<u8, 0x10> private_seed{};
+    if (sd_private.ReadBytes(private_seed.data(), private_seed.size()) != 0x10)
+        return boost::none;
+
+    std::array<u8, 0x10> buffer{};
+    size_t offset = 0;
+    for (; offset + 0x10 < save_43.GetSize(); ++offset) {
+        save_43.Seek(offset, SEEK_SET);
+        save_43.ReadBytes(buffer.data(), buffer.size());
+        if (buffer == private_seed)
+            break;
+    }
+
+    if (offset + 0x10 >= save_43.GetSize())
+        return boost::none;
+
+    Key128 seed{};
+    save_43.Seek(offset + 0x10, SEEK_SET);
+    save_43.ReadBytes(seed.data(), seed.size());
+    return seed;
+}
+
+Loader::ResultStatus DeriveSDKeys(std::array<Key256, 2>& sd_keys, const KeyManager& keys) {
+    if (!keys.HasKey(S128KeyType::Source, static_cast<u64>(SourceKeyType::SDKEK)))
+        return Loader::ResultStatus::ErrorMissingSDKEKSource;
+    if (!keys.HasKey(S128KeyType::Source, static_cast<u64>(SourceKeyType::AESKEKGeneration)))
+        return Loader::ResultStatus::ErrorMissingAESKEKGenerationSource;
+    if (!keys.HasKey(S128KeyType::Source, static_cast<u64>(SourceKeyType::AESKeyGeneration)))
+        return Loader::ResultStatus::ErrorMissingAESKeyGenerationSource;
+
+    const auto sd_kek_source =
+        keys.GetKey(S128KeyType::Source, static_cast<u64>(SourceKeyType::SDKEK));
+    const auto aes_kek_gen =
+        keys.GetKey(S128KeyType::Source, static_cast<u64>(SourceKeyType::AESKEKGeneration));
+    const auto aes_key_gen =
+        keys.GetKey(S128KeyType::Source, static_cast<u64>(SourceKeyType::AESKeyGeneration));
+    const auto master_00 = keys.GetKey(S128KeyType::Master);
+    const auto sd_kek =
+        GenerateKeyEncryptionKey(sd_kek_source, master_00, aes_kek_gen, aes_key_gen);
+
+    if (!keys.HasKey(S128KeyType::SDSeed))
+        return Loader::ResultStatus::ErrorMissingSDSeed;
+    const auto sd_seed = keys.GetKey(S128KeyType::SDSeed);
+
+    if (!keys.HasKey(S256KeyType::SDKeySource, static_cast<u64>(SDKeyType::Save)))
+        return Loader::ResultStatus::ErrorMissingSDSaveKeySource;
+    if (!keys.HasKey(S256KeyType::SDKeySource, static_cast<u64>(SDKeyType::NCA)))
+        return Loader::ResultStatus::ErrorMissingSDNCAKeySource;
+
+    std::array<Key256, 2> sd_key_sources{
+        keys.GetKey(S256KeyType::SDKeySource, static_cast<u64>(SDKeyType::Save)),
+        keys.GetKey(S256KeyType::SDKeySource, static_cast<u64>(SDKeyType::NCA)),
+    };
+
+    // Combine sources and seed
+    for (auto& source : sd_key_sources) {
+        for (size_t i = 0; i < source.size(); ++i)
+            source[i] ^= sd_seed[i & 0xF];
+    }
+
+    AESCipher<Key128> cipher(sd_kek, Mode::ECB);
+    // The transform manipulates sd_keys as part of the Transcode, so the return/output is
+    // unnecessary. This does not alter sd_keys_sources.
+    std::transform(sd_key_sources.begin(), sd_key_sources.end(), sd_keys.begin(),
+                   sd_key_sources.begin(), [&cipher](const Key256& source, Key256& out) {
+                       cipher.Transcode(source.data(), source.size(), out.data(), Op::Decrypt);
+                       return source; ///< Return unaltered source to satisfy output requirement.
+                   });
+
+    return Loader::ResultStatus::Success;
+}
+
 KeyManager::KeyManager() {
     // Initialize keys
     const std::string hactool_keys_dir = FileUtil::GetHactoolConfigurationPath();
@@ -24,12 +125,15 @@ KeyManager::KeyManager() {
     if (Settings::values.use_dev_keys) {
         dev_mode = true;
         AttemptLoadKeyFile(yuzu_keys_dir, hactool_keys_dir, "dev.keys", false);
+        AttemptLoadKeyFile(yuzu_keys_dir, yuzu_keys_dir, "dev.keys_autogenerated", false);
     } else {
         dev_mode = false;
         AttemptLoadKeyFile(yuzu_keys_dir, hactool_keys_dir, "prod.keys", false);
+        AttemptLoadKeyFile(yuzu_keys_dir, yuzu_keys_dir, "prod.keys_autogenerated", false);
     }
 
     AttemptLoadKeyFile(yuzu_keys_dir, hactool_keys_dir, "title.keys", true);
+    AttemptLoadKeyFile(yuzu_keys_dir, yuzu_keys_dir, "title.keys_autogenerated", true);
 }
 
 void KeyManager::LoadFromFile(const std::string& filename, bool is_title_keys) {
@@ -56,17 +160,17 @@ void KeyManager::LoadFromFile(const std::string& filename, bool is_title_keys) {
             u128 rights_id{};
             std::memcpy(rights_id.data(), rights_id_raw.data(), rights_id_raw.size());
             Key128 key = Common::HexStringToArray<16>(out[1]);
-            SetKey(S128KeyType::Titlekey, key, rights_id[1], rights_id[0]);
+            s128_keys[{S128KeyType::Titlekey, rights_id[1], rights_id[0]}] = key;
         } else {
             std::transform(out[0].begin(), out[0].end(), out[0].begin(), ::tolower);
             if (s128_file_id.find(out[0]) != s128_file_id.end()) {
                 const auto index = s128_file_id.at(out[0]);
                 Key128 key = Common::HexStringToArray<16>(out[1]);
-                SetKey(index.type, key, index.field1, index.field2);
+                s128_keys[{index.type, index.field1, index.field2}] = key;
             } else if (s256_file_id.find(out[0]) != s256_file_id.end()) {
                 const auto index = s256_file_id.at(out[0]);
                 Key256 key = Common::HexStringToArray<32>(out[1]);
-                SetKey(index.type, key, index.field1, index.field2);
+                s256_keys[{index.type, index.field1, index.field2}] = key;
             }
         }
     }
@@ -100,11 +204,50 @@ Key256 KeyManager::GetKey(S256KeyType id, u64 field1, u64 field2) const {
     return s256_keys.at({id, field1, field2});
 }
 
+template <size_t Size>
+void KeyManager::WriteKeyToFile(bool title_key, std::string_view keyname,
+                                const std::array<u8, Size>& key) {
+    const std::string yuzu_keys_dir = FileUtil::GetUserPath(FileUtil::UserPath::KeysDir);
+    std::string filename = "title.keys_autogenerated";
+    if (!title_key)
+        filename = dev_mode ? "dev.keys_autogenerated" : "prod.keys_autogenerated";
+    const auto add_info_text = !FileUtil::Exists(yuzu_keys_dir + DIR_SEP + filename);
+    FileUtil::CreateFullPath(yuzu_keys_dir + DIR_SEP + filename);
+    std::ofstream file(yuzu_keys_dir + DIR_SEP + filename, std::ios::app);
+    if (!file.is_open())
+        return;
+    if (add_info_text) {
+        file
+            << "# This file is autogenerated by Yuzu\n"
+            << "# It serves to store keys that were automatically generated from the normal keys\n"
+            << "# If you are experiencing issues involving keys, it may help to delete this file\n";
+    }
+
+    file << fmt::format("\n{} = {}", keyname, Common::HexArrayToString(key));
+    AttemptLoadKeyFile(yuzu_keys_dir, yuzu_keys_dir, filename, title_key);
+}
+
 void KeyManager::SetKey(S128KeyType id, Key128 key, u64 field1, u64 field2) {
+    const auto iter = std::find_if(
+        s128_file_id.begin(), s128_file_id.end(),
+        [&id, &field1, &field2](const std::pair<std::string, KeyIndex<S128KeyType>> elem) {
+            return std::tie(elem.second.type, elem.second.field1, elem.second.field2) ==
+                   std::tie(id, field1, field2);
+        });
+    if (iter != s128_file_id.end())
+        WriteKeyToFile(id == S128KeyType::Titlekey, iter->first, key);
     s128_keys[{id, field1, field2}] = key;
 }
 
 void KeyManager::SetKey(S256KeyType id, Key256 key, u64 field1, u64 field2) {
+    const auto iter = std::find_if(
+        s256_file_id.begin(), s256_file_id.end(),
+        [&id, &field1, &field2](const std::pair<std::string, KeyIndex<S256KeyType>> elem) {
+            return std::tie(elem.second.type, elem.second.field1, elem.second.field2) ==
+                   std::tie(id, field1, field2);
+        });
+    if (iter != s256_file_id.end())
+        WriteKeyToFile(false, iter->first, key);
     s256_keys[{id, field1, field2}] = key;
 }
 
@@ -125,7 +268,16 @@ bool KeyManager::KeyFileExists(bool title) {
            FileUtil::Exists(yuzu_keys_dir + DIR_SEP + "prod.keys");
 }
 
-const std::unordered_map<std::string, KeyIndex<S128KeyType>> KeyManager::s128_file_id = {
+void KeyManager::DeriveSDSeedLazy() {
+    if (HasKey(S128KeyType::SDSeed))
+        return;
+
+    const auto res = DeriveSDSeed();
+    if (res != boost::none)
+        SetKey(S128KeyType::SDSeed, res.get());
+}
+
+const boost::container::flat_map<std::string, KeyIndex<S128KeyType>> KeyManager::s128_file_id = {
     {"master_key_00", {S128KeyType::Master, 0, 0}},
     {"master_key_01", {S128KeyType::Master, 1, 0}},
     {"master_key_02", {S128KeyType::Master, 2, 0}},
@@ -167,11 +319,17 @@ const std::unordered_map<std::string, KeyIndex<S128KeyType>> KeyManager::s128_fi
     {"key_area_key_system_02", {S128KeyType::KeyArea, 2, static_cast<u64>(KeyAreaKeyType::System)}},
     {"key_area_key_system_03", {S128KeyType::KeyArea, 3, static_cast<u64>(KeyAreaKeyType::System)}},
     {"key_area_key_system_04", {S128KeyType::KeyArea, 4, static_cast<u64>(KeyAreaKeyType::System)}},
+    {"sd_card_kek_source", {S128KeyType::Source, static_cast<u64>(SourceKeyType::SDKEK), 0}},
+    {"aes_kek_generation_source",
+     {S128KeyType::Source, static_cast<u64>(SourceKeyType::AESKEKGeneration), 0}},
+    {"aes_key_generation_source",
+     {S128KeyType::Source, static_cast<u64>(SourceKeyType::AESKeyGeneration), 0}},
+    {"sd_seed", {S128KeyType::SDSeed, 0, 0}},
 };
 
-const std::unordered_map<std::string, KeyIndex<S256KeyType>> KeyManager::s256_file_id = {
+const boost::container::flat_map<std::string, KeyIndex<S256KeyType>> KeyManager::s256_file_id = {
     {"header_key", {S256KeyType::Header, 0, 0}},
-    {"sd_card_save_key", {S256KeyType::SDSave, 0, 0}},
-    {"sd_card_nca_key", {S256KeyType::SDNCA, 0, 0}},
+    {"sd_card_save_key_source", {S256KeyType::SDKeySource, static_cast<u64>(SDKeyType::Save), 0}},
+    {"sd_card_nca_key_source", {S256KeyType::SDKeySource, static_cast<u64>(SDKeyType::NCA), 0}},
 };
 } // namespace Core::Crypto

src/core/crypto/key_manager.h

@@ -6,11 +6,13 @@
 
 #include <array>
 #include <string>
+#include <string_view>
 #include <type_traits>
-#include <unordered_map>
 #include <vector>
+#include <boost/container/flat_map.hpp>
 #include <fmt/format.h>
 #include "common/common_types.h"
+#include "core/loader/loader.h"
 
 namespace Core::Crypto {
 
@@ -22,9 +24,8 @@ static_assert(sizeof(Key128) == 16, "Key128 must be 128 bytes big.");
 static_assert(sizeof(Key256) == 32, "Key128 must be 128 bytes big.");
 
 enum class S256KeyType : u64 {
-    Header, //
-    SDSave, //
-    SDNCA,  //
+    Header,      //
+    SDKeySource, // f1=SDKeyType
 };
 
 enum class S128KeyType : u64 {
@@ -36,6 +37,7 @@ enum class S128KeyType : u64 {
     KeyArea,       // f1=crypto revision f2=type {app, ocean, system}
     SDSeed,        //
     Titlekey,      // f1=rights id LSB f2=rights id MSB
+    Source,        // f1=source type, f2= sub id
 };
 
 enum class KeyAreaKeyType : u8 {
@@ -44,6 +46,17 @@ enum class KeyAreaKeyType : u8 {
     System,
 };
 
+enum class SourceKeyType : u8 {
+    SDKEK,
+    AESKEKGeneration,
+    AESKeyGeneration,
+};
+
+enum class SDKeyType : u8 {
+    Save,
+    NCA,
+};
+
 template <typename KeyType>
 struct KeyIndex {
     KeyType type;
@@ -59,34 +72,12 @@ struct KeyIndex {
     }
 };
 
-// The following two (== and hash) are so KeyIndex can be a key in unordered_map
-
+// boost flat_map requires operator< for O(log(n)) lookups.
 template <typename KeyType>
-bool operator==(const KeyIndex<KeyType>& lhs, const KeyIndex<KeyType>& rhs) {
-    return std::tie(lhs.type, lhs.field1, lhs.field2) == std::tie(rhs.type, rhs.field1, rhs.field2);
+bool operator<(const KeyIndex<KeyType>& lhs, const KeyIndex<KeyType>& rhs) {
+    return std::tie(lhs.type, lhs.field1, lhs.field2) < std::tie(rhs.type, rhs.field1, rhs.field2);
 }
 
-template <typename KeyType>
-bool operator!=(const KeyIndex<KeyType>& lhs, const KeyIndex<KeyType>& rhs) {
-    return !operator==(lhs, rhs);
-}
-
-} // namespace Core::Crypto
-
-namespace std {
-template <typename KeyType>
-struct hash<Core::Crypto::KeyIndex<KeyType>> {
-    size_t operator()(const Core::Crypto::KeyIndex<KeyType>& k) const {
-        using std::hash;
-
-        return ((hash<u64>()(static_cast<u64>(k.type)) ^ (hash<u64>()(k.field1) << 1)) >> 1) ^
-               (hash<u64>()(k.field2) << 1);
-    }
-};
-} // namespace std
-
-namespace Core::Crypto {
-
 class KeyManager {
 public:
     KeyManager();
@@ -102,16 +93,27 @@ public:
 
     static bool KeyFileExists(bool title);
 
+    // Call before using the sd seed to attempt to derive it if it dosen't exist. Needs system save
+    // 8*43 and the private file to exist.
+    void DeriveSDSeedLazy();
+
 private:
-    std::unordered_map<KeyIndex<S128KeyType>, Key128> s128_keys;
-    std::unordered_map<KeyIndex<S256KeyType>, Key256> s256_keys;
+    boost::container::flat_map<KeyIndex<S128KeyType>, Key128> s128_keys;
+    boost::container::flat_map<KeyIndex<S256KeyType>, Key256> s256_keys;
 
     bool dev_mode;
     void LoadFromFile(const std::string& filename, bool is_title_keys);
     void AttemptLoadKeyFile(const std::string& dir1, const std::string& dir2,
                             const std::string& filename, bool title);
+    template <size_t Size>
+    void WriteKeyToFile(bool title_key, std::string_view keyname, const std::array<u8, Size>& key);
 
-    static const std::unordered_map<std::string, KeyIndex<S128KeyType>> s128_file_id;
-    static const std::unordered_map<std::string, KeyIndex<S256KeyType>> s256_file_id;
+    static const boost::container::flat_map<std::string, KeyIndex<S128KeyType>> s128_file_id;
+    static const boost::container::flat_map<std::string, KeyIndex<S256KeyType>> s256_file_id;
 };
+
+Key128 GenerateKeyEncryptionKey(Key128 source, Key128 master, Key128 kek_seed, Key128 key_seed);
+boost::optional<Key128> DeriveSDSeed();
+Loader::ResultStatus DeriveSDKeys(std::array<Key256, 2>& sd_keys, const KeyManager& keys);
+
 } // namespace Core::Crypto

src/core/crypto/xts_encryption_layer.cpp

@@ -0,0 +1,58 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstring>
+#include "common/assert.h"
+#include "core/crypto/xts_encryption_layer.h"
+
+namespace Core::Crypto {
+
+constexpr u64 XTS_SECTOR_SIZE = 0x4000;
+
+XTSEncryptionLayer::XTSEncryptionLayer(FileSys::VirtualFile base_, Key256 key_)
+    : EncryptionLayer(std::move(base_)), cipher(key_, Mode::XTS) {}
+
+size_t XTSEncryptionLayer::Read(u8* data, size_t length, size_t offset) const {
+    if (length == 0)
+        return 0;
+
+    const auto sector_offset = offset & 0x3FFF;
+    if (sector_offset == 0) {
+        if (length % XTS_SECTOR_SIZE == 0) {
+            std::vector<u8> raw = base->ReadBytes(length, offset);
+            cipher.XTSTranscode(raw.data(), raw.size(), data, offset / XTS_SECTOR_SIZE,
+                                XTS_SECTOR_SIZE, Op::Decrypt);
+            return raw.size();
+        }
+        if (length > XTS_SECTOR_SIZE) {
+            const auto rem = length % XTS_SECTOR_SIZE;
+            const auto read = length - rem;
+            return Read(data, read, offset) + Read(data + read, rem, offset + read);
+        }
+        std::vector<u8> buffer = base->ReadBytes(XTS_SECTOR_SIZE, offset);
+        if (buffer.size() < XTS_SECTOR_SIZE)
+            buffer.resize(XTS_SECTOR_SIZE);
+        cipher.XTSTranscode(buffer.data(), buffer.size(), buffer.data(), offset / XTS_SECTOR_SIZE,
+                            XTS_SECTOR_SIZE, Op::Decrypt);
+        std::memcpy(data, buffer.data(), std::min(buffer.size(), length));
+        return std::min(buffer.size(), length);
+    }
+
+    // offset does not fall on block boundary (0x4000)
+    std::vector<u8> block = base->ReadBytes(0x4000, offset - sector_offset);
+    if (block.size() < XTS_SECTOR_SIZE)
+        block.resize(XTS_SECTOR_SIZE);
+    cipher.XTSTranscode(block.data(), block.size(), block.data(),
+                        (offset - sector_offset) / XTS_SECTOR_SIZE, XTS_SECTOR_SIZE, Op::Decrypt);
+    const size_t read = XTS_SECTOR_SIZE - sector_offset;
+
+    if (length + sector_offset < XTS_SECTOR_SIZE) {
+        std::memcpy(data, block.data() + sector_offset, std::min<u64>(length, read));
+        return std::min<u64>(length, read);
+    }
+    std::memcpy(data, block.data() + sector_offset, read);
+    return read + Read(data + read, length - read, offset + read);
+}
+} // namespace Core::Crypto

src/core/crypto/xts_encryption_layer.h

@@ -0,0 +1,25 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/crypto/aes_util.h"
+#include "core/crypto/encryption_layer.h"
+#include "core/crypto/key_manager.h"
+
+namespace Core::Crypto {
+
+// Sits on top of a VirtualFile and provides XTS-mode AES decription.
+class XTSEncryptionLayer : public EncryptionLayer {
+public:
+    XTSEncryptionLayer(FileSys::VirtualFile base, Key256 key);
+
+    size_t Read(u8* data, size_t length, size_t offset) const override;
+
+private:
+    // Must be mutable as operations modify cipher contexts.
+    mutable AESCipher<Key256> cipher;
+};
+
+} // namespace Core::Crypto

src/core/file_sys/bis_factory.cpp

@@ -6,19 +6,12 @@
 
 namespace FileSys {
 
-static VirtualDir GetOrCreateDirectory(const VirtualDir& dir, std::string_view path) {
-    const auto res = dir->GetDirectoryRelative(path);
-    if (res == nullptr)
-        return dir->CreateDirectoryRelative(path);
-    return res;
-}
-
 BISFactory::BISFactory(VirtualDir nand_root_)
     : nand_root(std::move(nand_root_)),
       sysnand_cache(std::make_shared<RegisteredCache>(
-          GetOrCreateDirectory(nand_root, "/system/Contents/registered"))),
+          GetOrCreateDirectoryRelative(nand_root, "/system/Contents/registered"))),
       usrnand_cache(std::make_shared<RegisteredCache>(
-          GetOrCreateDirectory(nand_root, "/user/Contents/registered"))) {}
+          GetOrCreateDirectoryRelative(nand_root, "/user/Contents/registered"))) {}
 
 std::shared_ptr<RegisteredCache> BISFactory::GetSystemNANDContents() const {
     return sysnand_cache;

src/core/file_sys/card_image.cpp

@@ -43,6 +43,8 @@ XCI::XCI(VirtualFile file_) : file(std::move(file_)), partitions(0x4) {
             partitions[static_cast<size_t>(partition)] = std::make_shared<PartitionFilesystem>(raw);
     }
 
+    program_nca_status = Loader::ResultStatus::ErrorXCIMissingProgramNCA;
+
     auto result = AddNCAFromPartition(XCIPartition::Secure);
     if (result != Loader::ResultStatus::Success) {
         status = result;
@@ -76,6 +78,10 @@ Loader::ResultStatus XCI::GetStatus() const {
     return status;
 }
 
+Loader::ResultStatus XCI::GetProgramNCAStatus() const {
+    return program_nca_status;
+}
+
 VirtualDir XCI::GetPartition(XCIPartition partition) const {
     return partitions[static_cast<size_t>(partition)];
 }
@@ -143,6 +149,12 @@ Loader::ResultStatus XCI::AddNCAFromPartition(XCIPartition part) {
         if (file->GetExtension() != "nca")
             continue;
         auto nca = std::make_shared<NCA>(file);
+        // TODO(DarkLordZach): Add proper Rev1+ Support
+        if (nca->IsUpdate())
+            continue;
+        if (nca->GetType() == NCAContentType::Program) {
+            program_nca_status = nca->GetStatus();
+        }
         if (nca->GetStatus() == Loader::ResultStatus::Success) {
             ncas.push_back(std::move(nca));
         } else {

src/core/file_sys/card_image.h

@@ -59,6 +59,7 @@ public:
     explicit XCI(VirtualFile file);
 
     Loader::ResultStatus GetStatus() const;
+    Loader::ResultStatus GetProgramNCAStatus() const;
 
     u8 GetFormatVersion() const;
 
@@ -90,6 +91,7 @@ private:
     GamecardHeader header{};
 
     Loader::ResultStatus status;
+    Loader::ResultStatus program_nca_status;
 
     std::vector<VirtualDir> partitions;
     std::vector<std::shared_ptr<NCA>> ncas;

src/core/file_sys/content_archive.cpp

@@ -178,7 +178,7 @@ VirtualFile NCA::Decrypt(NCASectionHeader s_header, VirtualFile in, u64 starting
             return std::static_pointer_cast<VfsFile>(out);
         }
     case NCASectionCryptoType::XTS:
-        // TODO(DarkLordZach): Implement XTSEncryptionLayer.
+        // TODO(DarkLordZach): Find a test case for XTS-encrypted NCAs
     default:
         LOG_ERROR(Crypto, "called with unhandled crypto type={:02X}",
                   static_cast<u8>(s_header.raw.header.crypto_type));
@@ -258,6 +258,10 @@ NCA::NCA(VirtualFile file_) : file(std::move(file_)) {
         file->ReadBytes(sections.data(), length_sections, SECTION_HEADER_OFFSET);
     }
 
+    is_update = std::find_if(sections.begin(), sections.end(), [](const NCASectionHeader& header) {
+                    return header.raw.header.crypto_type == NCASectionCryptoType::BKTR;
+                }) != sections.end();
+
     for (std::ptrdiff_t i = 0; i < number_sections; ++i) {
         auto section = sections[i];
 
@@ -358,6 +362,10 @@ VirtualFile NCA::GetBaseFile() const {
     return file;
 }
 
+bool NCA::IsUpdate() const {
+    return is_update;
+}
+
 bool NCA::ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) {
     return false;
 }

src/core/file_sys/content_archive.h

@@ -93,6 +93,8 @@ public:
 
     VirtualFile GetBaseFile() const;
 
+    bool IsUpdate() const;
+
 protected:
     bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;
 
@@ -111,6 +113,7 @@ private:
 
     NCAHeader header{};
     bool has_rights_id{};
+    bool is_update{};
 
     Loader::ResultStatus status{};
 

src/core/file_sys/registered_cache.cpp

@@ -216,11 +216,11 @@ void RegisteredCache::ProcessFiles(const std::vector<NcaID>& ids) {
 
         const auto section0 = nca->GetSubdirectories()[0];
 
-        for (const auto& file : section0->GetFiles()) {
-            if (file->GetExtension() != "cnmt")
+        for (const auto& section0_file : section0->GetFiles()) {
+            if (section0_file->GetExtension() != "cnmt")
                 continue;
 
-            meta.insert_or_assign(nca->GetTitleId(), CNMT(file));
+            meta.insert_or_assign(nca->GetTitleId(), CNMT(section0_file));
             meta_id.insert_or_assign(nca->GetTitleId(), id);
             break;
         }
@@ -254,6 +254,8 @@ RegisteredCache::RegisteredCache(VirtualDir dir_, RegisteredCacheParsingFunction
     Refresh();
 }
 
+RegisteredCache::~RegisteredCache() = default;
+
 bool RegisteredCache::HasEntry(u64 title_id, ContentRecordType type) const {
     return GetEntryRaw(title_id, type) != nullptr;
 }
@@ -262,6 +264,18 @@ 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);
+    if (id == boost::none)
+        return nullptr;
+
+    return GetFileAtID(id.get());
+}
+
+VirtualFile RegisteredCache::GetEntryUnparsed(RegisteredCacheEntry entry) const {
+    return GetEntryUnparsed(entry.title_id, entry.type);
+}
+
 VirtualFile RegisteredCache::GetEntryRaw(u64 title_id, ContentRecordType type) const {
     const auto id = GetNcaIDFromMetadata(title_id, type);
     if (id == boost::none)

src/core/file_sys/registered_cache.h

@@ -63,12 +63,16 @@ public:
     explicit RegisteredCache(VirtualDir dir,
                              RegisteredCacheParsingFunction parsing_function =
                                  [](const VirtualFile& file, const NcaID& id) { return file; });
+    ~RegisteredCache();
 
     void Refresh();
 
     bool HasEntry(u64 title_id, ContentRecordType type) const;
     bool HasEntry(RegisteredCacheEntry entry) 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;
 

src/core/file_sys/sdmc_factory.cpp

@@ -3,14 +3,27 @@
 // Refer to the license.txt file included.
 
 #include <memory>
+#include "core/file_sys/registered_cache.h"
 #include "core/file_sys/sdmc_factory.h"
+#include "core/file_sys/xts_archive.h"
 
 namespace FileSys {
 
-SDMCFactory::SDMCFactory(VirtualDir dir) : dir(std::move(dir)) {}
+SDMCFactory::SDMCFactory(VirtualDir dir_)
+    : dir(std::move(dir_)), contents(std::make_shared<RegisteredCache>(
+                                GetOrCreateDirectoryRelative(dir, "/Nintendo/Contents/registered"),
+                                [](const VirtualFile& file, const NcaID& id) {
+                                    return std::make_shared<NAX>(file, id)->GetDecrypted();
+                                })) {}
+
+SDMCFactory::~SDMCFactory() = default;
 
 ResultVal<VirtualDir> SDMCFactory::Open() {
     return MakeResult<VirtualDir>(dir);
 }
 
+std::shared_ptr<RegisteredCache> SDMCFactory::GetSDMCContents() const {
+    return contents;
+}
+
 } // namespace FileSys

src/core/file_sys/sdmc_factory.h

@@ -4,20 +4,27 @@
 
 #pragma once
 
+#include <memory>
 #include "core/file_sys/vfs.h"
 #include "core/hle/result.h"
 
 namespace FileSys {
 
+class RegisteredCache;
+
 /// File system interface to the SDCard archive
 class SDMCFactory {
 public:
     explicit SDMCFactory(VirtualDir dir);
+    ~SDMCFactory();
 
     ResultVal<VirtualDir> Open();
+    std::shared_ptr<RegisteredCache> GetSDMCContents() const;
 
 private:
     VirtualDir dir;
+
+    std::shared_ptr<RegisteredCache> contents;
 };
 
 } // namespace FileSys

src/core/file_sys/vfs.cpp

@@ -462,4 +462,11 @@ bool VfsRawCopy(VirtualFile src, VirtualFile dest) {
     std::vector<u8> data = src->ReadAllBytes();
     return dest->WriteBytes(data, 0) == data.size();
 }
+
+VirtualDir GetOrCreateDirectoryRelative(const VirtualDir& rel, std::string_view path) {
+    const auto res = rel->GetDirectoryRelative(path);
+    if (res == nullptr)
+        return rel->CreateDirectoryRelative(path);
+    return res;
+}
 } // namespace FileSys

src/core/file_sys/vfs.h

@@ -318,4 +318,8 @@ bool DeepEquals(const VirtualFile& file1, const VirtualFile& file2, size_t block
 // directory of src/dest.
 bool VfsRawCopy(VirtualFile src, VirtualFile dest);
 
+// Checks if the directory at path relative to rel exists. If it does, returns that. If it does not
+// it attempts to create it and returns the new dir or nullptr on failure.
+VirtualDir GetOrCreateDirectoryRelative(const VirtualDir& rel, std::string_view path);
+
 } // namespace FileSys

src/core/file_sys/vfs_real.cpp

@@ -341,7 +341,6 @@ std::shared_ptr<VfsFile> RealVfsDirectory::CreateFileRelative(std::string_view p
 
 std::shared_ptr<VfsDirectory> RealVfsDirectory::CreateDirectoryRelative(std::string_view path) {
     const auto full_path = FileUtil::SanitizePath(this->path + DIR_SEP + std::string(path));
-    auto parent = std::string(FileUtil::GetParentPath(full_path));
     return base.CreateDirectory(full_path, perms);
 }
 

src/core/file_sys/xts_archive.cpp

@@ -0,0 +1,169 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <cstring>
+#include <regex>
+#include <string>
+#include <mbedtls/md.h>
+#include <mbedtls/sha256.h>
+#include "common/assert.h"
+#include "common/hex_util.h"
+#include "common/logging/log.h"
+#include "core/crypto/aes_util.h"
+#include "core/crypto/xts_encryption_layer.h"
+#include "core/file_sys/partition_filesystem.h"
+#include "core/file_sys/vfs_offset.h"
+#include "core/file_sys/xts_archive.h"
+#include "core/loader/loader.h"
+
+namespace FileSys {
+
+constexpr u64 NAX_HEADER_PADDING_SIZE = 0x4000;
+
+template <typename SourceData, typename SourceKey, typename Destination>
+static bool CalculateHMAC256(Destination* out, const SourceKey* key, size_t key_length,
+                             const SourceData* data, size_t data_length) {
+    mbedtls_md_context_t context;
+    mbedtls_md_init(&context);
+
+    const auto key_f = reinterpret_cast<const u8*>(key);
+    const std::vector<u8> key_v(key_f, key_f + key_length);
+
+    if (mbedtls_md_setup(&context, mbedtls_md_info_from_type(MBEDTLS_MD_SHA256), 1) ||
+        mbedtls_md_hmac_starts(&context, reinterpret_cast<const u8*>(key), key_length) ||
+        mbedtls_md_hmac_update(&context, reinterpret_cast<const u8*>(data), data_length) ||
+        mbedtls_md_hmac_finish(&context, reinterpret_cast<u8*>(out))) {
+        mbedtls_md_free(&context);
+        return false;
+    }
+
+    mbedtls_md_free(&context);
+    return true;
+}
+
+NAX::NAX(VirtualFile file_) : file(std::move(file_)), header(std::make_unique<NAXHeader>()) {
+    std::string path = FileUtil::SanitizePath(file->GetFullPath());
+    static const std::regex nax_path_regex("/registered/(000000[0-9A-F]{2})/([0-9A-F]{32})\\.nca",
+                                           std::regex_constants::ECMAScript |
+                                               std::regex_constants::icase);
+    std::smatch match;
+    if (!std::regex_search(path, match, nax_path_regex)) {
+        status = Loader::ResultStatus::ErrorBadNAXFilePath;
+        return;
+    }
+
+    std::string two_dir = match[1];
+    std::string nca_id = match[2];
+    std::transform(two_dir.begin(), two_dir.end(), two_dir.begin(), ::toupper);
+    std::transform(nca_id.begin(), nca_id.end(), nca_id.begin(), ::tolower);
+
+    status = Parse(fmt::format("/registered/{}/{}.nca", two_dir, nca_id));
+}
+
+NAX::NAX(VirtualFile file_, std::array<u8, 0x10> nca_id)
+    : file(std::move(file_)), header(std::make_unique<NAXHeader>()) {
+    Core::Crypto::SHA256Hash hash{};
+    mbedtls_sha256(nca_id.data(), nca_id.size(), hash.data(), 0);
+    status = Parse(fmt::format("/registered/000000{:02X}/{}.nca", hash[0],
+                               Common::HexArrayToString(nca_id, false)));
+}
+
+Loader::ResultStatus NAX::Parse(std::string_view path) {
+    if (file->ReadObject(header.get()) != sizeof(NAXHeader))
+        return Loader::ResultStatus::ErrorBadNAXHeader;
+
+    if (header->magic != Common::MakeMagic('N', 'A', 'X', '0'))
+        return Loader::ResultStatus::ErrorBadNAXHeader;
+
+    if (file->GetSize() < NAX_HEADER_PADDING_SIZE + header->file_size)
+        return Loader::ResultStatus::ErrorIncorrectNAXFileSize;
+
+    keys.DeriveSDSeedLazy();
+    std::array<Core::Crypto::Key256, 2> sd_keys{};
+    const auto sd_keys_res = Core::Crypto::DeriveSDKeys(sd_keys, keys);
+    if (sd_keys_res != Loader::ResultStatus::Success) {
+        return sd_keys_res;
+    }
+
+    const auto enc_keys = header->key_area;
+
+    size_t i = 0;
+    for (; i < sd_keys.size(); ++i) {
+        std::array<Core::Crypto::Key128, 2> nax_keys{};
+        if (!CalculateHMAC256(nax_keys.data(), sd_keys[i].data(), 0x10, std::string(path).c_str(),
+                              path.size())) {
+            return Loader::ResultStatus::ErrorNAXKeyHMACFailed;
+        }
+
+        for (size_t j = 0; j < nax_keys.size(); ++j) {
+            Core::Crypto::AESCipher<Core::Crypto::Key128> cipher(nax_keys[j],
+                                                                 Core::Crypto::Mode::ECB);
+            cipher.Transcode(enc_keys[j].data(), 0x10, header->key_area[j].data(),
+                             Core::Crypto::Op::Decrypt);
+        }
+
+        Core::Crypto::SHA256Hash validation{};
+        if (!CalculateHMAC256(validation.data(), &header->magic, 0x60, sd_keys[i].data() + 0x10,
+                              0x10)) {
+            return Loader::ResultStatus::ErrorNAXValidationHMACFailed;
+        }
+        if (header->hmac == validation)
+            break;
+    }
+
+    if (i == 2) {
+        return Loader::ResultStatus::ErrorNAXKeyDerivationFailed;
+    }
+
+    type = static_cast<NAXContentType>(i);
+
+    Core::Crypto::Key256 final_key{};
+    std::memcpy(final_key.data(), &header->key_area, final_key.size());
+    const auto enc_file =
+        std::make_shared<OffsetVfsFile>(file, header->file_size, NAX_HEADER_PADDING_SIZE);
+    dec_file = std::make_shared<Core::Crypto::XTSEncryptionLayer>(enc_file, final_key);
+
+    return Loader::ResultStatus::Success;
+}
+
+Loader::ResultStatus NAX::GetStatus() const {
+    return status;
+}
+
+VirtualFile NAX::GetDecrypted() const {
+    return dec_file;
+}
+
+std::shared_ptr<NCA> NAX::AsNCA() const {
+    if (type == NAXContentType::NCA)
+        return std::make_shared<NCA>(GetDecrypted());
+    return nullptr;
+}
+
+NAXContentType NAX::GetContentType() const {
+    return type;
+}
+
+std::vector<std::shared_ptr<VfsFile>> NAX::GetFiles() const {
+    return {dec_file};
+}
+
+std::vector<std::shared_ptr<VfsDirectory>> NAX::GetSubdirectories() const {
+    return {};
+}
+
+std::string NAX::GetName() const {
+    return file->GetName();
+}
+
+std::shared_ptr<VfsDirectory> NAX::GetParentDirectory() const {
+    return file->GetContainingDirectory();
+}
+
+bool NAX::ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) {
+    return false;
+}
+} // namespace FileSys

src/core/file_sys/xts_archive.h

@@ -0,0 +1,69 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <vector>
+#include "common/common_types.h"
+#include "common/swap.h"
+#include "core/crypto/key_manager.h"
+#include "core/file_sys/content_archive.h"
+#include "core/file_sys/vfs.h"
+#include "core/loader/loader.h"
+
+namespace FileSys {
+
+struct NAXHeader {
+    std::array<u8, 0x20> hmac;
+    u64_le magic;
+    std::array<Core::Crypto::Key128, 2> key_area;
+    u64_le file_size;
+    INSERT_PADDING_BYTES(0x30);
+};
+static_assert(sizeof(NAXHeader) == 0x80, "NAXHeader has incorrect size.");
+
+enum class NAXContentType : u8 {
+    Save = 0,
+    NCA = 1,
+};
+
+class NAX : public ReadOnlyVfsDirectory {
+public:
+    explicit NAX(VirtualFile file);
+    explicit NAX(VirtualFile file, std::array<u8, 0x10> nca_id);
+
+    Loader::ResultStatus GetStatus() const;
+
+    VirtualFile GetDecrypted() const;
+
+    std::shared_ptr<NCA> AsNCA() const;
+
+    NAXContentType GetContentType() const;
+
+    std::vector<std::shared_ptr<VfsFile>> GetFiles() const override;
+
+    std::vector<std::shared_ptr<VfsDirectory>> GetSubdirectories() const override;
+
+    std::string GetName() const override;
+
+    std::shared_ptr<VfsDirectory> GetParentDirectory() const override;
+
+protected:
+    bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;
+
+private:
+    Loader::ResultStatus Parse(std::string_view path);
+
+    std::unique_ptr<NAXHeader> header;
+
+    VirtualFile file;
+    Loader::ResultStatus status;
+    NAXContentType type;
+
+    VirtualFile dec_file;
+
+    Core::Crypto::KeyManager keys;
+};
+} // namespace FileSys

src/core/hle/kernel/errors.h

@@ -11,17 +11,16 @@ namespace Kernel {
 namespace ErrCodes {
 enum {
     // TODO(Subv): Remove these 3DS OS error codes.
-    OutOfHandles = 19,
     SessionClosedByRemote = 26,
-    PortNameTooLong = 30,
     NoPendingSessions = 35,
-    WrongPermission = 46,
     InvalidBufferDescriptor = 48,
-    MaxConnectionsReached = 52,
 
     // Confirmed Switch OS error codes
+    MaxConnectionsReached = 7,
     InvalidAddress = 102,
+    HandleTableFull = 105,
     InvalidMemoryState = 106,
+    InvalidMemoryPermissions = 108,
     InvalidProcessorId = 113,
     InvalidHandle = 114,
     InvalidCombination = 116,
@@ -30,6 +29,7 @@ enum {
     TooLarge = 119,
     InvalidEnumValue = 120,
     InvalidState = 125,
+    ResourceLimitExceeded = 132,
 };
 }
 
@@ -37,18 +37,21 @@ enum {
 // double check that the code matches before re-using the constant.
 
 // TODO(bunnei): Replace these with correct errors for Switch OS
-constexpr ResultCode ERR_OUT_OF_HANDLES(-1);
+constexpr ResultCode ERR_HANDLE_TABLE_FULL(ErrorModule::Kernel, ErrCodes::HandleTableFull);
 constexpr ResultCode ERR_SESSION_CLOSED_BY_REMOTE(-1);
-constexpr ResultCode ERR_PORT_NAME_TOO_LONG(-1);
-constexpr ResultCode ERR_WRONG_PERMISSION(-1);
-constexpr ResultCode ERR_MAX_CONNECTIONS_REACHED(-1);
+constexpr ResultCode ERR_PORT_NAME_TOO_LONG(ErrorModule::Kernel, ErrCodes::TooLarge);
+constexpr ResultCode ERR_MAX_CONNECTIONS_REACHED(ErrorModule::Kernel,
+                                                 ErrCodes::MaxConnectionsReached);
 constexpr ResultCode ERR_INVALID_ENUM_VALUE(ErrorModule::Kernel, ErrCodes::InvalidEnumValue);
 constexpr ResultCode ERR_INVALID_ENUM_VALUE_FND(-1);
 constexpr ResultCode ERR_INVALID_COMBINATION(-1);
-constexpr ResultCode ERR_INVALID_COMBINATION_KERNEL(-1);
+constexpr ResultCode ERR_INVALID_COMBINATION_KERNEL(ErrorModule::Kernel,
+                                                    ErrCodes::InvalidCombination);
 constexpr ResultCode ERR_OUT_OF_MEMORY(-1);
 constexpr ResultCode ERR_INVALID_ADDRESS(ErrorModule::Kernel, ErrCodes::InvalidAddress);
 constexpr ResultCode ERR_INVALID_ADDRESS_STATE(ErrorModule::Kernel, ErrCodes::InvalidMemoryState);
+constexpr ResultCode ERR_INVALID_MEMORY_PERMISSIONS(ErrorModule::Kernel,
+                                                    ErrCodes::InvalidMemoryPermissions);
 constexpr ResultCode ERR_INVALID_HANDLE(ErrorModule::Kernel, ErrCodes::InvalidHandle);
 constexpr ResultCode ERR_INVALID_STATE(ErrorModule::Kernel, ErrCodes::InvalidState);
 constexpr ResultCode ERR_INVALID_POINTER(-1);

src/core/hle/kernel/handle_table.cpp

@@ -26,7 +26,7 @@ ResultVal<Handle> HandleTable::Create(SharedPtr<Object> obj) {
     u16 slot = next_free_slot;
     if (slot >= generations.size()) {
         LOG_ERROR(Kernel, "Unable to allocate Handle, too many slots in use.");
-        return ERR_OUT_OF_HANDLES;
+        return ERR_HANDLE_TABLE_FULL;
     }
     next_free_slot = generations[slot];
 

src/core/hle/kernel/handle_table.h

@@ -47,7 +47,7 @@ public:
     /**
      * Allocates a handle for the given object.
      * @return The created Handle or one of the following errors:
-     *           - `ERR_OUT_OF_HANDLES`: the maximum number of handles has been exceeded.
+     *           - `ERR_HANDLE_TABLE_FULL`: the maximum number of handles has been exceeded.
      */
     ResultVal<Handle> Create(SharedPtr<Object> obj);
 

src/core/hle/kernel/scheduler.cpp

@@ -17,7 +17,7 @@ namespace Kernel {
 
 std::mutex Scheduler::scheduler_mutex;
 
-Scheduler::Scheduler(ARM_Interface* cpu_core) : cpu_core(cpu_core) {}
+Scheduler::Scheduler(Core::ARM_Interface* cpu_core) : cpu_core(cpu_core) {}
 
 Scheduler::~Scheduler() {
     for (auto& thread : thread_list) {

src/core/hle/kernel/scheduler.h

@@ -11,13 +11,15 @@
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/thread.h"
 
+namespace Core {
 class ARM_Interface;
+}
 
 namespace Kernel {
 
 class Scheduler final {
 public:
-    explicit Scheduler(ARM_Interface* cpu_core);
+    explicit Scheduler(Core::ARM_Interface* cpu_core);
     ~Scheduler();
 
     /// Returns whether there are any threads that are ready to run.
@@ -70,7 +72,7 @@ private:
 
     SharedPtr<Thread> current_thread = nullptr;
 
-    ARM_Interface* cpu_core;
+    Core::ARM_Interface* cpu_core;
 
     static std::mutex scheduler_mutex;
 };

src/core/hle/kernel/shared_memory.cpp

@@ -101,7 +101,7 @@ ResultCode SharedMemory::Map(Process* target_process, VAddr address, MemoryPermi
         static_cast<u32>(this->permissions) & ~static_cast<u32>(other_permissions)) {
         LOG_ERROR(Kernel, "cannot map id={}, address=0x{:X} name={}, permissions don't match",
                   GetObjectId(), address, name);
-        return ERR_WRONG_PERMISSION;
+        return ERR_INVALID_MEMORY_PERMISSIONS;
     }
 
     VAddr target_address = address;

src/core/hle/kernel/svc.cpp

@@ -319,8 +319,7 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
         *result = Core::CurrentProcess()->is_virtual_address_memory_enabled;
         break;
     case GetInfoType::TitleId:
-        LOG_WARNING(Kernel_SVC, "(STUBBED) Attempted to query titleid, returned 0");
-        *result = 0;
+        *result = Core::CurrentProcess()->program_id;
         break;
     case GetInfoType::PrivilegedProcessId:
         LOG_WARNING(Kernel_SVC,

src/core/hle/kernel/thread.cpp

@@ -283,9 +283,9 @@ static std::tuple<std::size_t, std::size_t, bool> GetFreeThreadLocalSlot(
  * @param entry_point Address of entry point for execution
  * @param arg User argument for thread
  */
-static void ResetThreadContext(ARM_Interface::ThreadContext& context, VAddr stack_top,
+static void ResetThreadContext(Core::ARM_Interface::ThreadContext& context, VAddr stack_top,
                                VAddr entry_point, u64 arg) {
-    memset(&context, 0, sizeof(ARM_Interface::ThreadContext));
+    memset(&context, 0, sizeof(Core::ARM_Interface::ThreadContext));
 
     context.cpu_registers[0] = arg;
     context.pc = entry_point;

src/core/hle/kernel/thread.h

@@ -204,7 +204,7 @@ public:
         return status == ThreadStatus::WaitSynchAll;
     }
 
-    ARM_Interface::ThreadContext context;
+    Core::ARM_Interface::ThreadContext context;
 
     u32 thread_id;
 

src/core/hle/result.h

@@ -227,7 +227,7 @@ public:
         }
     }
 
-    ResultVal(ResultVal&& o) : result_code(o.result_code) {
+    ResultVal(ResultVal&& o) noexcept : result_code(o.result_code) {
         if (!o.empty()) {
             new (&object) T(std::move(o.object));
         }

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

@@ -18,6 +18,7 @@
 #include "core/hle/service/apm/apm.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/nvflinger/nvflinger.h"
+#include "core/hle/service/pm/pm.h"
 #include "core/hle/service/set/set.h"
 #include "core/settings.h"
 
@@ -309,7 +310,7 @@ ICommonStateGetter::ICommonStateGetter() : ServiceFramework("ICommonStateGetter"
         {5, &ICommonStateGetter::GetOperationMode, "GetOperationMode"},
         {6, &ICommonStateGetter::GetPerformanceMode, "GetPerformanceMode"},
         {7, nullptr, "GetCradleStatus"},
-        {8, nullptr, "GetBootMode"},
+        {8, &ICommonStateGetter::GetBootMode, "GetBootMode"},
         {9, &ICommonStateGetter::GetCurrentFocusState, "GetCurrentFocusState"},
         {10, nullptr, "RequestToAcquireSleepLock"},
         {11, nullptr, "ReleaseSleepLock"},
@@ -334,6 +335,15 @@ ICommonStateGetter::ICommonStateGetter() : ServiceFramework("ICommonStateGetter"
     event = Kernel::Event::Create(Kernel::ResetType::OneShot, "ICommonStateGetter:Event");
 }
 
+void ICommonStateGetter::GetBootMode(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+
+    rb.Push<u8>(static_cast<u8>(Service::PM::SystemBootMode::Normal)); // Normal boot mode
+
+    LOG_DEBUG(Service_AM, "called");
+}
+
 void ICommonStateGetter::GetEventHandle(Kernel::HLERequestContext& ctx) {
     event->Signal();
 

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

@@ -116,6 +116,7 @@ private:
     void GetDefaultDisplayResolutionChangeEvent(Kernel::HLERequestContext& ctx);
     void GetOperationMode(Kernel::HLERequestContext& ctx);
     void GetPerformanceMode(Kernel::HLERequestContext& ctx);
+    void GetBootMode(Kernel::HLERequestContext& ctx);
 
     Kernel::SharedPtr<Kernel::Event> event;
 };

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

@@ -254,7 +254,7 @@ ResultCode RegisterSDMC(std::unique_ptr<FileSys::SDMCFactory>&& factory) {
 ResultCode RegisterBIS(std::unique_ptr<FileSys::BISFactory>&& factory) {
     ASSERT_MSG(bis_factory == nullptr, "Tried to register a second BIS");
     bis_factory = std::move(factory);
-    LOG_DEBUG(Service_FS, "Registred BIS");
+    LOG_DEBUG(Service_FS, "Registered BIS");
     return RESULT_SUCCESS;
 }
 
@@ -305,17 +305,38 @@ ResultVal<FileSys::VirtualDir> OpenSDMC() {
 }
 
 std::shared_ptr<FileSys::RegisteredCache> GetSystemNANDContents() {
+    LOG_TRACE(Service_FS, "Opening System NAND Contents");
+
+    if (bis_factory == nullptr)
+        return nullptr;
+
     return bis_factory->GetSystemNANDContents();
 }
 
 std::shared_ptr<FileSys::RegisteredCache> GetUserNANDContents() {
+    LOG_TRACE(Service_FS, "Opening User NAND Contents");
+
+    if (bis_factory == nullptr)
+        return nullptr;
+
     return bis_factory->GetUserNANDContents();
 }
 
-void RegisterFileSystems(const FileSys::VirtualFilesystem& vfs) {
-    romfs_factory = nullptr;
-    save_data_factory = nullptr;
-    sdmc_factory = nullptr;
+std::shared_ptr<FileSys::RegisteredCache> GetSDMCContents() {
+    LOG_TRACE(Service_FS, "Opening SDMC Contents");
+
+    if (sdmc_factory == nullptr)
+        return nullptr;
+
+    return sdmc_factory->GetSDMCContents();
+}
+
+void CreateFactories(const FileSys::VirtualFilesystem& vfs, bool overwrite) {
+    if (overwrite) {
+        bis_factory = nullptr;
+        save_data_factory = nullptr;
+        sdmc_factory = nullptr;
+    }
 
     auto nand_directory = vfs->OpenDirectory(FileUtil::GetUserPath(FileUtil::UserPath::NANDDir),
                                              FileSys::Mode::ReadWrite);
@@ -324,16 +345,15 @@ void RegisterFileSystems(const FileSys::VirtualFilesystem& vfs) {
 
     if (bis_factory == nullptr)
         bis_factory = std::make_unique<FileSys::BISFactory>(nand_directory);
-
-    auto savedata = std::make_unique<FileSys::SaveDataFactory>(std::move(nand_directory));
-    save_data_factory = std::move(savedata);
-
-    auto sdcard = std::make_unique<FileSys::SDMCFactory>(std::move(sd_directory));
-    sdmc_factory = std::move(sdcard);
+    if (save_data_factory == nullptr)
+        save_data_factory = std::make_unique<FileSys::SaveDataFactory>(std::move(nand_directory));
+    if (sdmc_factory == nullptr)
+        sdmc_factory = std::make_unique<FileSys::SDMCFactory>(std::move(sd_directory));
 }
 
 void InstallInterfaces(SM::ServiceManager& service_manager, const FileSys::VirtualFilesystem& vfs) {
-    RegisterFileSystems(vfs);
+    romfs_factory = nullptr;
+    CreateFactories(vfs, false);
     std::make_shared<FSP_LDR>()->InstallAsService(service_manager);
     std::make_shared<FSP_PR>()->InstallAsService(service_manager);
     std::make_shared<FSP_SRV>()->InstallAsService(service_manager);

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

@@ -46,8 +46,12 @@ ResultVal<FileSys::VirtualDir> OpenSDMC();
 
 std::shared_ptr<FileSys::RegisteredCache> GetSystemNANDContents();
 std::shared_ptr<FileSys::RegisteredCache> GetUserNANDContents();
+std::shared_ptr<FileSys::RegisteredCache> GetSDMCContents();
+
+// Creates the SaveData, SDMC, and BIS Factories. Should be called once and before any function
+// above is called.
+void CreateFactories(const FileSys::VirtualFilesystem& vfs, bool overwrite = true);
 
-/// Registers all Filesystem services with the specified service manager.
 void InstallInterfaces(SM::ServiceManager& service_manager, const FileSys::VirtualFilesystem& vfs);
 
 // A class that wraps a VfsDirectory with methods that return ResultVal and ResultCode instead of

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

@@ -17,6 +17,7 @@
 #include "core/hle/service/hid/irs.h"
 #include "core/hle/service/hid/xcd.h"
 #include "core/hle/service/service.h"
+#include "core/settings.h"
 
 namespace Service::HID {
 

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

@@ -4,8 +4,10 @@
 
 #pragma once
 
+#include <array>
+#include "common/bit_field.h"
+#include "common/common_types.h"
 #include "core/hle/service/service.h"
-#include "core/settings.h"
 
 namespace Service::HID {
 

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

@@ -35,6 +35,14 @@ static constexpr std::array<std::pair<FontArchives, const char*>, 7> SHARED_FONT
     std::make_pair(FontArchives::Extension, "nintendo_ext_003.bfttf"),
     std::make_pair(FontArchives::Extension, "nintendo_ext2_003.bfttf")};
 
+static constexpr std::array<const char*, 7> SHARED_FONTS_TTF{"FontStandard.ttf",
+                                                             "FontChineseSimplified.ttf",
+                                                             "FontExtendedChineseSimplified.ttf",
+                                                             "FontChineseTraditional.ttf",
+                                                             "FontKorean.ttf",
+                                                             "FontNintendoExtended.ttf",
+                                                             "FontNintendoExtended2.ttf"};
+
 // The below data is specific to shared font data dumped from Switch on f/w 2.2
 // Virtual address and offsets/sizes likely will vary by dump
 static constexpr VAddr SHARED_FONT_MEM_VADDR{0x00000009d3016000ULL};
@@ -76,6 +84,17 @@ void DecryptSharedFont(const std::vector<u32>& input, std::vector<u8>& output, s
     offset += transformed_font.size() * sizeof(u32);
 }
 
+static void EncryptSharedFont(const std::vector<u8>& input, std::vector<u8>& output,
+                              size_t& offset) {
+    ASSERT_MSG(offset + input.size() + 8 < SHARED_FONT_MEM_SIZE, "Shared fonts exceeds 17mb!");
+    const u32 KEY = EXPECTED_MAGIC ^ EXPECTED_RESULT;
+    std::memcpy(output.data() + offset, &EXPECTED_RESULT, sizeof(u32)); // Magic header
+    const u32 ENC_SIZE = static_cast<u32>(input.size()) ^ KEY;
+    std::memcpy(output.data() + offset + sizeof(u32), &ENC_SIZE, sizeof(u32));
+    std::memcpy(output.data() + offset + (sizeof(u32) * 2), input.data(), input.size());
+    offset += input.size() + (sizeof(u32) * 2);
+}
+
 static u32 GetU32Swapped(const u8* data) {
     u32 value;
     std::memcpy(&value, data, sizeof(value));
@@ -109,10 +128,10 @@ PL_U::PL_U() : ServiceFramework("pl:u") {
     RegisterHandlers(functions);
     // Attempt to load shared font data from disk
     const auto nand = FileSystem::GetSystemNANDContents();
+    size_t offset = 0;
     // Rebuild shared fonts from data ncas
     if (nand->HasEntry(static_cast<u64>(FontArchives::Standard),
                        FileSys::ContentRecordType::Data)) {
-        size_t offset = 0;
         shared_font = std::make_shared<std::vector<u8>>(SHARED_FONT_MEM_SIZE);
         for (auto font : SHARED_FONTS) {
             const auto nca =
@@ -152,18 +171,45 @@ PL_U::PL_U() : ServiceFramework("pl:u") {
             DecryptSharedFont(font_data_u32, *shared_font, offset);
             SHARED_FONT_REGIONS.push_back(region);
         }
+
     } else {
-        const std::string filepath{FileUtil::GetUserPath(FileUtil::UserPath::SysDataDir) +
-                                   SHARED_FONT};
+        shared_font = std::make_shared<std::vector<u8>>(
+            SHARED_FONT_MEM_SIZE); // Shared memory needs to always be allocated and a fixed size
+
+        const std::string user_path = FileUtil::GetUserPath(FileUtil::UserPath::SysDataDir);
+        const std::string filepath{user_path + SHARED_FONT};
+
         // Create path if not already created
         if (!FileUtil::CreateFullPath(filepath)) {
             LOG_ERROR(Service_NS, "Failed to create sharedfonts path \"{}\"!", filepath);
             return;
         }
+
+        bool using_ttf = false;
+        for (const char* font_ttf : SHARED_FONTS_TTF) {
+            if (FileUtil::Exists(user_path + font_ttf)) {
+                using_ttf = true;
+                FileUtil::IOFile file(user_path + font_ttf, "rb");
+                if (file.IsOpen()) {
+                    std::vector<u8> ttf_bytes(file.GetSize());
+                    file.ReadBytes<u8>(ttf_bytes.data(), ttf_bytes.size());
+                    FontRegion region{
+                        static_cast<u32>(offset + 8),
+                        static_cast<u32>(ttf_bytes.size())}; // Font offset and size do not account
+                                                             // for the header
+                    EncryptSharedFont(ttf_bytes, *shared_font, offset);
+                    SHARED_FONT_REGIONS.push_back(region);
+                } else {
+                    LOG_WARNING(Service_NS, "Unable to load font: {}", font_ttf);
+                }
+            } else if (using_ttf) {
+                LOG_WARNING(Service_NS, "Unable to find font: {}", font_ttf);
+            }
+        }
+        if (using_ttf)
+            return;
         FileUtil::IOFile file(filepath, "rb");
 
-        shared_font = std::make_shared<std::vector<u8>>(
-            SHARED_FONT_MEM_SIZE); // Shared memory needs to always be allocated and a fixed size
         if (file.IsOpen()) {
             // Read shared font data
             ASSERT(file.GetSize() == SHARED_FONT_MEM_SIZE);

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

@@ -56,9 +56,9 @@ u32 nvhost_as_gpu::AllocateSpace(const std::vector<u8>& input, std::vector<u8>&
     auto& gpu = Core::System::GetInstance().GPU();
     const u64 size{static_cast<u64>(params.pages) * static_cast<u64>(params.page_size)};
     if (params.flags & 1) {
-        params.offset = gpu.memory_manager->AllocateSpace(params.offset, size, 1);
+        params.offset = gpu.MemoryManager().AllocateSpace(params.offset, size, 1);
     } else {
-        params.offset = gpu.memory_manager->AllocateSpace(size, params.align);
+        params.offset = gpu.MemoryManager().AllocateSpace(size, params.align);
     }
 
     std::memcpy(output.data(), &params, output.size());
@@ -88,7 +88,7 @@ u32 nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& output)
         u64 size = static_cast<u64>(entry.pages) << 0x10;
         ASSERT(size <= object->size);
 
-        Tegra::GPUVAddr returned = gpu.memory_manager->MapBufferEx(object->addr, offset, size);
+        Tegra::GPUVAddr returned = gpu.MemoryManager().MapBufferEx(object->addr, offset, size);
         ASSERT(returned == offset);
     }
     std::memcpy(output.data(), entries.data(), output.size());
@@ -125,9 +125,9 @@ u32 nvhost_as_gpu::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& ou
     auto& gpu = Core::System::GetInstance().GPU();
 
     if (params.flags & 1) {
-        params.offset = gpu.memory_manager->MapBufferEx(object->addr, params.offset, object->size);
+        params.offset = gpu.MemoryManager().MapBufferEx(object->addr, params.offset, object->size);
     } else {
-        params.offset = gpu.memory_manager->MapBufferEx(object->addr, object->size);
+        params.offset = gpu.MemoryManager().MapBufferEx(object->addr, object->size);
     }
 
     // Create a new mapping entry for this operation.
@@ -161,7 +161,7 @@ u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& ou
                                                                      itr->second.size);
 
     auto& gpu = system_instance.GPU();
-    params.offset = gpu.memory_manager->UnmapBuffer(params.offset, itr->second.size);
+    params.offset = gpu.MemoryManager().UnmapBuffer(params.offset, itr->second.size);
 
     buffer_mappings.erase(itr->second.offset);
 

src/core/hle/service/pm/pm.cpp

@@ -2,6 +2,8 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/service/pm/pm.h"
 #include "core/hle/service/service.h"
 
 namespace Service::PM {
@@ -10,11 +12,20 @@ class BootMode final : public ServiceFramework<BootMode> {
 public:
     explicit BootMode() : ServiceFramework{"pm:bm"} {
         static const FunctionInfo functions[] = {
-            {0, nullptr, "GetBootMode"},
+            {0, &BootMode::GetBootMode, "GetBootMode"},
             {1, nullptr, "SetMaintenanceBoot"},
         };
         RegisterHandlers(functions);
     }
+
+private:
+    void GetBootMode(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u32>(static_cast<u32>(SystemBootMode::Normal)); // Normal boot mode
+
+        LOG_DEBUG(Service_PM, "called");
+    }
 };
 
 class DebugMonitor final : public ServiceFramework<DebugMonitor> {

src/core/hle/service/pm/pm.h

@@ -9,7 +9,7 @@ class ServiceManager;
 }
 
 namespace Service::PM {
-
+enum class SystemBootMode : u32 { Normal = 0, Maintenance = 1 };
 /// Registers all PM services with the specified service manager.
 void InstallInterfaces(SM::ServiceManager& service_manager);
 

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

@@ -32,24 +32,59 @@ constexpr std::array<LanguageCode, 17> available_language_codes = {{
     LanguageCode::ZH_HANT,
 }};
 
+constexpr size_t pre4_0_0_max_entries = 0xF;
+constexpr size_t post4_0_0_max_entries = 0x40;
+
 LanguageCode GetLanguageCodeFromIndex(size_t index) {
     return available_language_codes.at(index);
 }
 
-void SET::GetAvailableLanguageCodes(Kernel::HLERequestContext& ctx) {
-    ctx.WriteBuffer(available_language_codes);
+template <size_t size>
+static std::array<LanguageCode, size> MakeLanguageCodeSubset() {
+    std::array<LanguageCode, size> arr;
+    std::copy_n(available_language_codes.begin(), size, arr.begin());
+    return arr;
+}
 
+static void PushResponseLanguageCode(Kernel::HLERequestContext& ctx, size_t max_size) {
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-    rb.Push(static_cast<u32>(available_language_codes.size()));
+    if (available_language_codes.size() > max_size)
+        rb.Push(static_cast<u32>(max_size));
+    else
+        rb.Push(static_cast<u32>(available_language_codes.size()));
+}
+
+void SET::GetAvailableLanguageCodes(Kernel::HLERequestContext& ctx) {
+    if (available_language_codes.size() > pre4_0_0_max_entries)
+        ctx.WriteBuffer(MakeLanguageCodeSubset<pre4_0_0_max_entries>());
+    else
+        ctx.WriteBuffer(available_language_codes);
+
+    PushResponseLanguageCode(ctx, pre4_0_0_max_entries);
+
+    LOG_DEBUG(Service_SET, "called");
+}
+
+void SET::GetAvailableLanguageCodes2(Kernel::HLERequestContext& ctx) {
+    if (available_language_codes.size() > post4_0_0_max_entries)
+        ctx.WriteBuffer(MakeLanguageCodeSubset<post4_0_0_max_entries>());
+    else
+        ctx.WriteBuffer(available_language_codes);
+
+    PushResponseLanguageCode(ctx, post4_0_0_max_entries);
 
     LOG_DEBUG(Service_SET, "called");
 }
 
 void SET::GetAvailableLanguageCodeCount(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 3};
-    rb.Push(RESULT_SUCCESS);
-    rb.Push(static_cast<u32>(available_language_codes.size()));
+    PushResponseLanguageCode(ctx, pre4_0_0_max_entries);
+
+    LOG_DEBUG(Service_SET, "called");
+}
+
+void SET::GetAvailableLanguageCodeCount2(Kernel::HLERequestContext& ctx) {
+    PushResponseLanguageCode(ctx, post4_0_0_max_entries);
 
     LOG_DEBUG(Service_SET, "called");
 }
@@ -69,8 +104,8 @@ SET::SET() : ServiceFramework("set") {
         {2, nullptr, "MakeLanguageCode"},
         {3, &SET::GetAvailableLanguageCodeCount, "GetAvailableLanguageCodeCount"},
         {4, nullptr, "GetRegionCode"},
-        {5, &SET::GetAvailableLanguageCodes, "GetAvailableLanguageCodes2"},
-        {6, &SET::GetAvailableLanguageCodeCount, "GetAvailableLanguageCodeCount2"},
+        {5, &SET::GetAvailableLanguageCodes2, "GetAvailableLanguageCodes2"},
+        {6, &SET::GetAvailableLanguageCodeCount2, "GetAvailableLanguageCodeCount2"},
         {7, nullptr, "GetKeyCodeMap"},
         {8, nullptr, "GetQuestFlag"},
     };

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

@@ -38,7 +38,9 @@ public:
 private:
     void GetLanguageCode(Kernel::HLERequestContext& ctx);
     void GetAvailableLanguageCodes(Kernel::HLERequestContext& ctx);
+    void GetAvailableLanguageCodes2(Kernel::HLERequestContext& ctx);
     void GetAvailableLanguageCodeCount(Kernel::HLERequestContext& ctx);
+    void GetAvailableLanguageCodeCount2(Kernel::HLERequestContext& ctx);
 };
 
 } // namespace Service::Set

src/core/loader/loader.cpp

@@ -11,6 +11,7 @@
 #include "core/hle/kernel/process.h"
 #include "core/loader/deconstructed_rom_directory.h"
 #include "core/loader/elf.h"
+#include "core/loader/nax.h"
 #include "core/loader/nca.h"
 #include "core/loader/nro.h"
 #include "core/loader/nso.h"
@@ -32,6 +33,7 @@ FileType IdentifyFile(FileSys::VirtualFile file) {
     CHECK_TYPE(NRO)
     CHECK_TYPE(NCA)
     CHECK_TYPE(XCI)
+    CHECK_TYPE(NAX)
 
 #undef CHECK_TYPE
 
@@ -73,6 +75,8 @@ std::string GetFileTypeString(FileType type) {
         return "NCA";
     case FileType::XCI:
         return "XCI";
+    case FileType::NAX:
+        return "NAX";
     case FileType::DeconstructedRomDirectory:
         return "Directory";
     case FileType::Error:
@@ -83,7 +87,7 @@ std::string GetFileTypeString(FileType type) {
     return "unknown";
 }
 
-constexpr std::array<const char*, 36> RESULT_MESSAGES{
+constexpr std::array<const char*, 49> RESULT_MESSAGES{
     "The operation completed successfully.",
     "The loader requested to load is already loaded.",
     "The operation is not implemented.",
@@ -120,6 +124,19 @@ constexpr std::array<const char*, 36> RESULT_MESSAGES{
     "There was a general error loading the NRO into emulated memory.",
     "There is no icon available.",
     "There is no control data available.",
+    "The NAX file has a bad header.",
+    "The NAX file has incorrect size as determined by the header.",
+    "The HMAC to generated the NAX decryption keys failed.",
+    "The HMAC to validate the NAX decryption keys failed.",
+    "The NAX key derivation failed.",
+    "The NAX file cannot be interpreted as an NCA file.",
+    "The NAX file has an incorrect path.",
+    "The SD seed could not be found or derived.",
+    "The SD KEK Source could not be found.",
+    "The AES KEK Generation Source could not be found.",
+    "The AES Key Generation Source could not be found.",
+    "The SD Save Key Source could not be found.",
+    "The SD NCA Key Source could not be found.",
 };
 
 std::ostream& operator<<(std::ostream& os, ResultStatus status) {
@@ -150,13 +167,18 @@ static std::unique_ptr<AppLoader> GetFileLoader(FileSys::VirtualFile file, FileT
     case FileType::NRO:
         return std::make_unique<AppLoader_NRO>(std::move(file));
 
-    // NX NCA file format.
+    // NX NCA (Nintendo Content Archive) file format.
     case FileType::NCA:
         return std::make_unique<AppLoader_NCA>(std::move(file));
 
+    // NX XCI (nX Card Image) file format.
     case FileType::XCI:
         return std::make_unique<AppLoader_XCI>(std::move(file));
 
+    // NX NAX (NintendoAesXts) file format.
+    case FileType::NAX:
+        return std::make_unique<AppLoader_NAX>(std::move(file));
+
     // NX deconstructed ROM directory.
     case FileType::DeconstructedRomDirectory:
         return std::make_unique<AppLoader_DeconstructedRomDirectory>(std::move(file));
@@ -170,7 +192,8 @@ std::unique_ptr<AppLoader> GetLoader(FileSys::VirtualFile file) {
     FileType type = IdentifyFile(file);
     FileType filename_type = GuessFromFilename(file->GetName());
 
-    if (type != filename_type) {
+    // Special case: 00 is either a NCA or NAX.
+    if (type != filename_type && !(file->GetName() == "00" && type == FileType::NAX)) {
         LOG_WARNING(Loader, "File {} has a different type than its extension.", file->GetName());
         if (FileType::Unknown == type)
             type = filename_type;

src/core/loader/loader.h

@@ -32,6 +32,7 @@ enum class FileType {
     NRO,
     NCA,
     XCI,
+    NAX,
     DeconstructedRomDirectory,
 };
 
@@ -93,6 +94,19 @@ enum class ResultStatus : u16 {
     ErrorLoadingNRO,
     ErrorNoIcon,
     ErrorNoControl,
+    ErrorBadNAXHeader,
+    ErrorIncorrectNAXFileSize,
+    ErrorNAXKeyHMACFailed,
+    ErrorNAXValidationHMACFailed,
+    ErrorNAXKeyDerivationFailed,
+    ErrorNAXInconvertibleToNCA,
+    ErrorBadNAXFilePath,
+    ErrorMissingSDSeed,
+    ErrorMissingSDKEKSource,
+    ErrorMissingAESKEKGenerationSource,
+    ErrorMissingAESKeyGenerationSource,
+    ErrorMissingSDSaveKeySource,
+    ErrorMissingSDNCAKeySource,
 };
 
 std::ostream& operator<<(std::ostream& os, ResultStatus status);

src/core/loader/nax.cpp

@@ -0,0 +1,66 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/logging/log.h"
+#include "core/file_sys/content_archive.h"
+#include "core/file_sys/romfs.h"
+#include "core/file_sys/xts_archive.h"
+#include "core/hle/kernel/process.h"
+#include "core/loader/nax.h"
+#include "core/loader/nca.h"
+
+namespace Loader {
+
+AppLoader_NAX::AppLoader_NAX(FileSys::VirtualFile file)
+    : AppLoader(file), nax(std::make_unique<FileSys::NAX>(file)),
+      nca_loader(std::make_unique<AppLoader_NCA>(nax->GetDecrypted())) {}
+
+AppLoader_NAX::~AppLoader_NAX() = default;
+
+FileType AppLoader_NAX::IdentifyType(const FileSys::VirtualFile& file) {
+    FileSys::NAX nax(file);
+
+    if (nax.GetStatus() == ResultStatus::Success && nax.AsNCA() != nullptr &&
+        nax.AsNCA()->GetStatus() == ResultStatus::Success) {
+        return FileType::NAX;
+    }
+
+    return FileType::Error;
+}
+
+ResultStatus AppLoader_NAX::Load(Kernel::SharedPtr<Kernel::Process>& process) {
+    if (is_loaded) {
+        return ResultStatus::ErrorAlreadyLoaded;
+    }
+
+    if (nax->GetStatus() != ResultStatus::Success)
+        return nax->GetStatus();
+
+    const auto nca = nax->AsNCA();
+    if (nca == nullptr) {
+        if (!Core::Crypto::KeyManager::KeyFileExists(false))
+            return ResultStatus::ErrorMissingProductionKeyFile;
+        return ResultStatus::ErrorNAXInconvertibleToNCA;
+    }
+
+    if (nca->GetStatus() != ResultStatus::Success)
+        return nca->GetStatus();
+
+    const auto result = nca_loader->Load(process);
+    if (result != ResultStatus::Success)
+        return result;
+
+    is_loaded = true;
+
+    return ResultStatus::Success;
+}
+
+ResultStatus AppLoader_NAX::ReadRomFS(FileSys::VirtualFile& dir) {
+    return nca_loader->ReadRomFS(dir);
+}
+
+ResultStatus AppLoader_NAX::ReadProgramId(u64& out_program_id) {
+    return nca_loader->ReadProgramId(out_program_id);
+}
+} // namespace Loader

src/core/loader/nax.h

@@ -0,0 +1,48 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include "common/common_types.h"
+#include "core/loader/loader.h"
+
+namespace FileSys {
+
+class NAX;
+
+} // namespace FileSys
+
+namespace Loader {
+
+class AppLoader_NCA;
+
+/// Loads a NAX file
+class AppLoader_NAX final : public AppLoader {
+public:
+    explicit AppLoader_NAX(FileSys::VirtualFile file);
+    ~AppLoader_NAX() override;
+
+    /**
+     * Returns the type of the file
+     * @param file std::shared_ptr<VfsFile> open file
+     * @return FileType found, or FileType::Error if this loader doesn't know it
+     */
+    static FileType IdentifyType(const FileSys::VirtualFile& file);
+
+    FileType GetFileType() override {
+        return IdentifyType(file);
+    }
+
+    ResultStatus Load(Kernel::SharedPtr<Kernel::Process>& process) override;
+
+    ResultStatus ReadRomFS(FileSys::VirtualFile& dir) override;
+    ResultStatus ReadProgramId(u64& out_program_id) override;
+
+private:
+    std::unique_ptr<FileSys::NAX> nax;
+    std::unique_ptr<AppLoader_NCA> nca_loader;
+};
+
+} // namespace Loader

src/core/loader/xci.cpp

@@ -61,11 +61,12 @@ ResultStatus AppLoader_XCI::Load(Kernel::SharedPtr<Kernel::Process>& process) {
     if (xci->GetStatus() != ResultStatus::Success)
         return xci->GetStatus();
 
-    if (xci->GetNCAFileByType(FileSys::NCAContentType::Program) == nullptr) {
-        if (!Core::Crypto::KeyManager::KeyFileExists(false))
-            return ResultStatus::ErrorMissingProductionKeyFile;
-        return ResultStatus::ErrorXCIMissingProgramNCA;
-    }
+    if (xci->GetProgramNCAStatus() != ResultStatus::Success)
+        return xci->GetProgramNCAStatus();
+
+    const auto nca = xci->GetNCAFileByType(FileSys::NCAContentType::Program);
+    if (nca == nullptr && !Core::Crypto::KeyManager::KeyFileExists(false))
+        return ResultStatus::ErrorMissingProductionKeyFile;
 
     auto result = nca_loader->Load(process);
     if (result != ResultStatus::Success)

src/core/memory.cpp

@@ -264,7 +264,7 @@ void RasterizerMarkRegionCached(Tegra::GPUVAddr gpu_addr, u64 size, bool cached)
     u64 num_pages = ((gpu_addr + size - 1) >> PAGE_BITS) - (gpu_addr >> PAGE_BITS) + 1;
     for (unsigned i = 0; i < num_pages; ++i, gpu_addr += PAGE_SIZE) {
         boost::optional<VAddr> maybe_vaddr =
-            Core::System::GetInstance().GPU().memory_manager->GpuToCpuAddress(gpu_addr);
+            Core::System::GetInstance().GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
         // The GPU <-> CPU virtual memory mapping is not 1:1
         if (!maybe_vaddr) {
             LOG_ERROR(HW_Memory,
@@ -346,7 +346,7 @@ void RasterizerFlushVirtualRegion(VAddr start, u64 size, FlushMode mode) {
         const VAddr overlap_end = std::min(end, region_end);
 
         const std::vector<Tegra::GPUVAddr> gpu_addresses =
-            system_instance.GPU().memory_manager->CpuToGpuAddress(overlap_start);
+            system_instance.GPU().MemoryManager().CpuToGpuAddress(overlap_start);
 
         if (gpu_addresses.empty()) {
             return;

src/video_core/CMakeLists.txt

@@ -18,6 +18,7 @@ add_library(video_core STATIC
     macro_interpreter.h
     memory_manager.cpp
     memory_manager.h
+    rasterizer_cache.h
     rasterizer_interface.h
     renderer_base.cpp
     renderer_base.h
@@ -26,6 +27,8 @@ add_library(video_core STATIC
     renderer_opengl/gl_rasterizer_cache.cpp
     renderer_opengl/gl_rasterizer_cache.h
     renderer_opengl/gl_resource_manager.h
+    renderer_opengl/gl_shader_cache.cpp
+    renderer_opengl/gl_shader_cache.h
     renderer_opengl/gl_shader_decompiler.cpp
     renderer_opengl/gl_shader_decompiler.h
     renderer_opengl/gl_shader_gen.cpp

src/video_core/debug_utils/debug_utils.cpp

@@ -2,23 +2,8 @@
 // Licensed under GPLv2
 // Refer to the license.txt file included.
 
-#include <algorithm>
-#include <condition_variable>
-#include <cstdint>
-#include <cstring>
-#include <fstream>
-#include <map>
 #include <mutex>
-#include <string>
 
-#include "common/assert.h"
-#include "common/bit_field.h"
-#include "common/color.h"
-#include "common/common_types.h"
-#include "common/file_util.h"
-#include "common/logging/log.h"
-#include "common/math_util.h"
-#include "common/vector_math.h"
 #include "video_core/debug_utils/debug_utils.h"
 
 namespace Tegra {

src/video_core/debug_utils/debug_utils.h

@@ -4,19 +4,11 @@
 
 #pragma once
 
-#include <algorithm>
 #include <array>
 #include <condition_variable>
-#include <iterator>
 #include <list>
-#include <map>
 #include <memory>
 #include <mutex>
-#include <string>
-#include <utility>
-#include <vector>
-#include "common/common_types.h"
-#include "common/vector_math.h"
 
 namespace Tegra {
 
@@ -46,7 +38,7 @@ public:
     class BreakPointObserver {
     public:
         /// Constructs the object such that it observes events of the given DebugContext.
-        BreakPointObserver(std::shared_ptr<DebugContext> debug_context)
+        explicit BreakPointObserver(std::shared_ptr<DebugContext> debug_context)
             : context_weak(debug_context) {
             std::unique_lock<std::mutex> lock(debug_context->breakpoint_mutex);
             debug_context->breakpoint_observers.push_back(this);
@@ -141,8 +133,8 @@ public:
     }
 
     // TODO: Evaluate if access to these members should be hidden behind a public interface.
-    std::array<BreakPoint, (int)Event::NumEvents> breakpoints;
-    Event active_breakpoint;
+    std::array<BreakPoint, static_cast<int>(Event::NumEvents)> breakpoints;
+    Event active_breakpoint{};
     bool at_breakpoint = false;
 
 private:

src/video_core/engines/maxwell_3d.cpp

@@ -218,10 +218,6 @@ void Maxwell3D::DrawArrays() {
         debug_context->OnEvent(Tegra::DebugContext::Event::IncomingPrimitiveBatch, nullptr);
     }
 
-    if (debug_context) {
-        debug_context->OnEvent(Tegra::DebugContext::Event::FinishedPrimitiveBatch, nullptr);
-    }
-
     // Both instance configuration registers can not be set at the same time.
     ASSERT_MSG(!regs.draw.instance_next || !regs.draw.instance_cont,
                "Illegal combination of instancing parameters");
@@ -237,6 +233,10 @@ void Maxwell3D::DrawArrays() {
     const bool is_indexed{regs.index_array.count && !regs.vertex_buffer.count};
     rasterizer.AccelerateDrawBatch(is_indexed);
 
+    if (debug_context) {
+        debug_context->OnEvent(Tegra::DebugContext::Event::FinishedPrimitiveBatch, nullptr);
+    }
+
     // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
     // the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
     // it's possible that it is incorrect and that there is some other register used to specify the

src/video_core/engines/maxwell_3d.h

@@ -330,6 +330,17 @@ public:
             Set = 0x150F,
         };
 
+        enum class StencilOp : u32 {
+            Keep = 1,
+            Zero = 2,
+            Replace = 3,
+            Incr = 4,
+            Decr = 5,
+            Invert = 6,
+            IncrWrap = 7,
+            DecrWrap = 8,
+        };
+
         struct Cull {
             enum class FrontFace : u32 {
                 ClockWise = 0x0900,
@@ -508,8 +519,16 @@ public:
 
                 float clear_color[4];
                 float clear_depth;
+                INSERT_PADDING_WORDS(0x3);
+                s32 clear_stencil;
 
-                INSERT_PADDING_WORDS(0x93);
+                INSERT_PADDING_WORDS(0x6C);
+
+                s32 stencil_back_func_ref;
+                u32 stencil_back_mask;
+                u32 stencil_back_func_mask;
+
+                INSERT_PADDING_WORDS(0x20);
 
                 struct {
                     u32 address_high;
@@ -573,16 +592,14 @@ public:
                     u32 enable[NumRenderTargets];
                 } blend;
 
-                struct {
-                    u32 enable;
-                    u32 front_op_fail;
-                    u32 front_op_zfail;
-                    u32 front_op_zpass;
-                    u32 front_func_func;
-                    u32 front_func_ref;
-                    u32 front_func_mask;
-                    u32 front_mask;
-                } stencil;
+                u32 stencil_enable;
+                StencilOp stencil_front_op_fail;
+                StencilOp stencil_front_op_zfail;
+                StencilOp stencil_front_op_zpass;
+                ComparisonOp stencil_front_func_func;
+                s32 stencil_front_func_ref;
+                u32 stencil_front_func_mask;
+                u32 stencil_front_mask;
 
                 INSERT_PADDING_WORDS(0x3);
 
@@ -626,13 +643,11 @@ public:
 
                 INSERT_PADDING_WORDS(0x5);
 
-                struct {
-                    u32 enable;
-                    u32 back_op_fail;
-                    u32 back_op_zfail;
-                    u32 back_op_zpass;
-                    u32 back_func_func;
-                } stencil_two_side;
+                u32 stencil_two_side_enable;
+                StencilOp stencil_back_op_fail;
+                StencilOp stencil_back_op_zfail;
+                StencilOp stencil_back_op_zpass;
+                ComparisonOp stencil_back_func_func;
 
                 INSERT_PADDING_WORDS(0x17);
 
@@ -944,6 +959,10 @@ ASSERT_REG_POSITION(viewport, 0x300);
 ASSERT_REG_POSITION(vertex_buffer, 0x35D);
 ASSERT_REG_POSITION(clear_color[0], 0x360);
 ASSERT_REG_POSITION(clear_depth, 0x364);
+ASSERT_REG_POSITION(clear_stencil, 0x368);
+ASSERT_REG_POSITION(stencil_back_func_ref, 0x3D5);
+ASSERT_REG_POSITION(stencil_back_mask, 0x3D6);
+ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7);
 ASSERT_REG_POSITION(zeta, 0x3F8);
 ASSERT_REG_POSITION(vertex_attrib_format[0], 0x458);
 ASSERT_REG_POSITION(rt_control, 0x487);
@@ -955,13 +974,24 @@ ASSERT_REG_POSITION(depth_write_enabled, 0x4BA);
 ASSERT_REG_POSITION(d3d_cull_mode, 0x4C2);
 ASSERT_REG_POSITION(depth_test_func, 0x4C3);
 ASSERT_REG_POSITION(blend, 0x4CF);
-ASSERT_REG_POSITION(stencil, 0x4E0);
+ASSERT_REG_POSITION(stencil_enable, 0x4E0);
+ASSERT_REG_POSITION(stencil_front_op_fail, 0x4E1);
+ASSERT_REG_POSITION(stencil_front_op_zfail, 0x4E2);
+ASSERT_REG_POSITION(stencil_front_op_zpass, 0x4E3);
+ASSERT_REG_POSITION(stencil_front_func_func, 0x4E4);
+ASSERT_REG_POSITION(stencil_front_func_ref, 0x4E5);
+ASSERT_REG_POSITION(stencil_front_func_mask, 0x4E6);
+ASSERT_REG_POSITION(stencil_front_mask, 0x4E7);
 ASSERT_REG_POSITION(screen_y_control, 0x4EB);
 ASSERT_REG_POSITION(vb_element_base, 0x50D);
 ASSERT_REG_POSITION(zeta_enable, 0x54E);
 ASSERT_REG_POSITION(tsc, 0x557);
 ASSERT_REG_POSITION(tic, 0x55D);
-ASSERT_REG_POSITION(stencil_two_side, 0x565);
+ASSERT_REG_POSITION(stencil_two_side_enable, 0x565);
+ASSERT_REG_POSITION(stencil_back_op_fail, 0x566);
+ASSERT_REG_POSITION(stencil_back_op_zfail, 0x567);
+ASSERT_REG_POSITION(stencil_back_op_zpass, 0x568);
+ASSERT_REG_POSITION(stencil_back_func_func, 0x569);
 ASSERT_REG_POSITION(point_coord_replace, 0x581);
 ASSERT_REG_POSITION(code_address, 0x582);
 ASSERT_REG_POSITION(draw, 0x585);

src/video_core/engines/shader_bytecode.h

@@ -280,6 +280,19 @@ union Instruction {
             BitField<56, 1, u64> invert_b;
         } lop32i;
 
+        union {
+            BitField<28, 8, u64> imm_lut28;
+            BitField<48, 8, u64> imm_lut48;
+
+            u32 GetImmLut28() const {
+                return static_cast<u32>(imm_lut28);
+            }
+
+            u32 GetImmLut48() const {
+                return static_cast<u32>(imm_lut48);
+            }
+        } lop3;
+
         u32 GetImm20_19() const {
             u32 imm{static_cast<u32>(imm20_19)};
             imm <<= 12;
@@ -313,6 +326,10 @@ union Instruction {
         BitField<49, 1, u64> negate_a;
     } alu_integer;
 
+    union {
+        BitField<40, 1, u64> invert;
+    } popc;
+
     union {
         BitField<39, 3, u64> pred;
         BitField<42, 1, u64> neg_pred;
@@ -623,10 +640,16 @@ public:
         IADD_C,
         IADD_R,
         IADD_IMM,
+        IADD3_C,
+        IADD3_R,
+        IADD3_IMM,
         IADD32I,
         ISCADD_C, // Scale and Add
         ISCADD_R,
         ISCADD_IMM,
+        POPC_C,
+        POPC_R,
+        POPC_IMM,
         SEL_C,
         SEL_R,
         SEL_IMM,
@@ -650,6 +673,9 @@ public:
         LOP_R,
         LOP_IMM,
         LOP32I,
+        LOP3_C,
+        LOP3_R,
+        LOP3_IMM,
         MOV_C,
         MOV_R,
         MOV_IMM,
@@ -838,13 +864,19 @@ private:
             INST("0100110000010---", Id::IADD_C, Type::ArithmeticInteger, "IADD_C"),
             INST("0101110000010---", Id::IADD_R, Type::ArithmeticInteger, "IADD_R"),
             INST("0011100-00010---", Id::IADD_IMM, Type::ArithmeticInteger, "IADD_IMM"),
+            INST("010011001100----", Id::IADD3_C, Type::ArithmeticInteger, "IADD3_C"),
+            INST("010111001100----", Id::IADD3_R, Type::ArithmeticInteger, "IADD3_R"),
+            INST("0011100-1100----", Id::IADD3_IMM, Type::ArithmeticInteger, "IADD3_IMM"),
             INST("0001110---------", Id::IADD32I, Type::ArithmeticIntegerImmediate, "IADD32I"),
             INST("0100110000011---", Id::ISCADD_C, Type::ArithmeticInteger, "ISCADD_C"),
             INST("0101110000011---", Id::ISCADD_R, Type::ArithmeticInteger, "ISCADD_R"),
             INST("0011100-00011---", Id::ISCADD_IMM, Type::ArithmeticInteger, "ISCADD_IMM"),
+            INST("0100110000001---", Id::POPC_C, Type::ArithmeticInteger, "POPC_C"),
+            INST("0101110000001---", Id::POPC_R, Type::ArithmeticInteger, "POPC_R"),
+            INST("0011100-00001---", Id::POPC_IMM, Type::ArithmeticInteger, "POPC_IMM"),
             INST("0100110010100---", Id::SEL_C, Type::ArithmeticInteger, "SEL_C"),
             INST("0101110010100---", Id::SEL_R, Type::ArithmeticInteger, "SEL_R"),
-            INST("0011100010100---", Id::SEL_IMM, Type::ArithmeticInteger, "SEL_IMM"),
+            INST("0011100-10100---", Id::SEL_IMM, Type::ArithmeticInteger, "SEL_IMM"),
             INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"),
             INST("0100110010010---", Id::RRO_C, Type::Arithmetic, "RRO_C"),
             INST("0101110010010---", Id::RRO_R, Type::Arithmetic, "RRO_R"),
@@ -872,6 +904,9 @@ private:
             INST("0101110001000---", Id::LOP_R, Type::ArithmeticInteger, "LOP_R"),
             INST("0011100001000---", 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"),
+            INST("0011110---------", Id::LOP3_IMM, Type::ArithmeticInteger, "LOP3_IMM"),
             INST("0100110001001---", Id::SHL_C, Type::Shift, "SHL_C"),
             INST("0101110001001---", Id::SHL_R, Type::Shift, "SHL_R"),
             INST("0011100-01001---", Id::SHL_IMM, Type::Shift, "SHL_IMM"),

src/video_core/gpu.cpp

@@ -22,7 +22,7 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
 }
 
 GPU::GPU(VideoCore::RasterizerInterface& rasterizer) {
-    memory_manager = std::make_unique<MemoryManager>();
+    memory_manager = std::make_unique<Tegra::MemoryManager>();
     maxwell_3d = std::make_unique<Engines::Maxwell3D>(rasterizer, *memory_manager);
     fermi_2d = std::make_unique<Engines::Fermi2D>(*memory_manager);
     maxwell_compute = std::make_unique<Engines::MaxwellCompute>();
@@ -31,12 +31,20 @@ GPU::GPU(VideoCore::RasterizerInterface& rasterizer) {
 
 GPU::~GPU() = default;
 
+Engines::Maxwell3D& GPU::Maxwell3D() {
+    return *maxwell_3d;
+}
+
 const Engines::Maxwell3D& GPU::Maxwell3D() const {
     return *maxwell_3d;
 }
 
-Engines::Maxwell3D& GPU::Maxwell3D() {
-    return *maxwell_3d;
+MemoryManager& GPU::MemoryManager() {
+    return *memory_manager;
+}
+
+const MemoryManager& GPU::MemoryManager() const {
+    return *memory_manager;
 }
 
 u32 RenderTargetBytesPerPixel(RenderTargetFormat format) {

src/video_core/gpu.h

@@ -117,18 +117,24 @@ public:
     /// Processes a command list stored at the specified address in GPU memory.
     void ProcessCommandList(GPUVAddr address, u32 size);
 
-    /// Returns a const reference to the Maxwell3D GPU engine.
-    const Engines::Maxwell3D& Maxwell3D() const;
-
     /// Returns a reference to the Maxwell3D GPU engine.
     Engines::Maxwell3D& Maxwell3D();
 
-    std::unique_ptr<MemoryManager> memory_manager;
+    /// Returns a const reference to the Maxwell3D GPU engine.
+    const Engines::Maxwell3D& Maxwell3D() const;
+
+    /// Returns a reference to the GPU memory manager.
+    Tegra::MemoryManager& MemoryManager();
+
+    /// Returns a const reference to the GPU memory manager.
+    const Tegra::MemoryManager& MemoryManager() const;
 
 private:
     /// Writes a single register in the engine bound to the specified subchannel
     void WriteReg(u32 method, u32 subchannel, u32 value, u32 remaining_params);
 
+    std::unique_ptr<Tegra::MemoryManager> memory_manager;
+
     /// Mapping of command subchannels to their bound engine ids.
     std::unordered_map<u32, EngineID> bound_engines;
 

src/video_core/rasterizer_cache.h

@@ -0,0 +1,116 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <unordered_map>
+#include <boost/icl/interval_map.hpp>
+#include <boost/range/iterator_range.hpp>
+
+#include "common/common_types.h"
+#include "core/memory.h"
+#include "video_core/memory_manager.h"
+
+template <class T>
+class RasterizerCache : NonCopyable {
+public:
+    /// Mark the specified region as being invalidated
+    void InvalidateRegion(Tegra::GPUVAddr region_addr, size_t region_size) {
+        for (auto iter = cached_objects.cbegin(); iter != cached_objects.cend();) {
+            const auto& object{iter->second};
+
+            ++iter;
+
+            if (object->GetAddr() <= (region_addr + region_size) &&
+                region_addr <= (object->GetAddr() + object->GetSizeInBytes())) {
+                // Regions overlap, so invalidate
+                Unregister(object);
+            }
+        }
+    }
+
+protected:
+    /// Tries to get an object from the cache with the specified address
+    T TryGet(Tegra::GPUVAddr addr) const {
+        const auto& search{cached_objects.find(addr)};
+        if (search != cached_objects.end()) {
+            return search->second;
+        }
+
+        return nullptr;
+    }
+
+    /// Gets a reference to the cache
+    const std::unordered_map<Tegra::GPUVAddr, T>& GetCache() const {
+        return cached_objects;
+    }
+
+    /// Register an object into the cache
+    void Register(const T& object) {
+        const auto& search{cached_objects.find(object->GetAddr())};
+        if (search != cached_objects.end()) {
+            // Registered already
+            return;
+        }
+
+        cached_objects[object->GetAddr()] = object;
+        UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), 1);
+    }
+
+    /// Unregisters an object from the cache
+    void Unregister(const T& object) {
+        const auto& search{cached_objects.find(object->GetAddr())};
+        if (search == cached_objects.end()) {
+            // Unregistered already
+            return;
+        }
+
+        UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), -1);
+        cached_objects.erase(search);
+    }
+
+private:
+    using PageMap = boost::icl::interval_map<u64, int>;
+
+    template <typename Map, typename Interval>
+    constexpr auto RangeFromInterval(Map& map, const Interval& interval) {
+        return boost::make_iterator_range(map.equal_range(interval));
+    }
+
+    /// Increase/decrease the number of object in pages touching the specified region
+    void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta) {
+        const u64 page_start{addr >> Tegra::MemoryManager::PAGE_BITS};
+        const u64 page_end{(addr + size) >> Tegra::MemoryManager::PAGE_BITS};
+
+        // Interval maps will erase segments if count reaches 0, so if delta is negative we have to
+        // subtract after iterating
+        const auto pages_interval = PageMap::interval_type::right_open(page_start, page_end);
+        if (delta > 0)
+            cached_pages.add({pages_interval, delta});
+
+        for (const auto& pair : RangeFromInterval(cached_pages, pages_interval)) {
+            const auto interval = pair.first & pages_interval;
+            const int count = pair.second;
+
+            const Tegra::GPUVAddr interval_start_addr = boost::icl::first(interval)
+                                                        << Tegra::MemoryManager::PAGE_BITS;
+            const Tegra::GPUVAddr interval_end_addr = boost::icl::last_next(interval)
+                                                      << Tegra::MemoryManager::PAGE_BITS;
+            const u64 interval_size = interval_end_addr - interval_start_addr;
+
+            if (delta > 0 && count == delta)
+                Memory::RasterizerMarkRegionCached(interval_start_addr, interval_size, true);
+            else if (delta < 0 && count == -delta)
+                Memory::RasterizerMarkRegionCached(interval_start_addr, interval_size, false);
+            else
+                ASSERT(count >= 0);
+        }
+
+        if (delta < 0)
+            cached_pages.add({pages_interval, delta});
+    }
+
+    std::unordered_map<Tegra::GPUVAddr, T> cached_objects;
+    PageMap cached_pages;
+};

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -14,6 +14,7 @@
 #include "common/logging/log.h"
 #include "common/math_util.h"
 #include "common/microprofile.h"
+#include "common/scope_exit.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
 #include "core/hle/kernel/process.h"
@@ -177,19 +178,6 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupVertexArrays(u8* array_ptr,
     return {array_ptr, buffer_offset};
 }
 
-static GLShader::ProgramCode GetShaderProgramCode(Maxwell::ShaderProgram program) {
-    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
-
-    // Fetch program code from memory
-    GLShader::ProgramCode program_code;
-    auto& shader_config = gpu.regs.shader_config[static_cast<size_t>(program)];
-    const u64 gpu_address{gpu.regs.code_address.CodeAddress() + shader_config.offset};
-    const boost::optional<VAddr> cpu_address{gpu.memory_manager.GpuToCpuAddress(gpu_address)};
-    Memory::ReadBlock(*cpu_address, program_code.data(), program_code.size() * sizeof(u64));
-
-    return program_code;
-}
-
 std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr buffer_offset) {
     auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
 
@@ -223,31 +211,16 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr
         buffer_ptr += sizeof(ubo);
         buffer_offset += sizeof(ubo);
 
-        GLShader::ShaderSetup setup{GetShaderProgramCode(program)};
-        GLShader::ShaderEntries shader_resources;
+        Shader shader{shader_cache.GetStageProgram(program)};
 
         switch (program) {
-        case Maxwell::ShaderProgram::VertexA: {
-            // VertexB is always enabled, so when VertexA is enabled, we have two vertex shaders.
-            // Conventional HW does not support this, so we combine VertexA and VertexB into one
-            // stage here.
-            setup.SetProgramB(GetShaderProgramCode(Maxwell::ShaderProgram::VertexB));
-            GLShader::MaxwellVSConfig vs_config{setup};
-            shader_resources =
-                shader_program_manager->UseProgrammableVertexShader(vs_config, setup);
-            break;
-        }
-
+        case Maxwell::ShaderProgram::VertexA:
         case Maxwell::ShaderProgram::VertexB: {
-            GLShader::MaxwellVSConfig vs_config{setup};
-            shader_resources =
-                shader_program_manager->UseProgrammableVertexShader(vs_config, setup);
+            shader_program_manager->UseProgrammableVertexShader(shader->GetProgramHandle());
             break;
         }
         case Maxwell::ShaderProgram::Fragment: {
-            GLShader::MaxwellFSConfig fs_config{setup};
-            shader_resources =
-                shader_program_manager->UseProgrammableFragmentShader(fs_config, setup);
+            shader_program_manager->UseProgrammableFragmentShader(shader->GetProgramHandle());
             break;
         }
         default:
@@ -256,18 +229,14 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr
             UNREACHABLE();
         }
 
-        GLuint gl_stage_program = shader_program_manager->GetCurrentProgramStage(
-            static_cast<Maxwell::ShaderStage>(stage));
-
         // Configure the const buffers for this shader stage.
-        std::tie(buffer_ptr, buffer_offset, current_constbuffer_bindpoint) = SetupConstBuffers(
-            buffer_ptr, buffer_offset, static_cast<Maxwell::ShaderStage>(stage), gl_stage_program,
-            current_constbuffer_bindpoint, shader_resources.const_buffer_entries);
+        std::tie(buffer_ptr, buffer_offset, current_constbuffer_bindpoint) =
+            SetupConstBuffers(buffer_ptr, buffer_offset, static_cast<Maxwell::ShaderStage>(stage),
+                              shader, current_constbuffer_bindpoint);
 
         // Configure the textures for this shader stage.
-        current_texture_bindpoint =
-            SetupTextures(static_cast<Maxwell::ShaderStage>(stage), gl_stage_program,
-                          current_texture_bindpoint, shader_resources.texture_samplers);
+        current_texture_bindpoint = SetupTextures(static_cast<Maxwell::ShaderStage>(stage), shader,
+                                                  current_texture_bindpoint);
 
         // When VertexA is enabled, we have dual vertex shaders
         if (program == Maxwell::ShaderProgram::VertexA) {
@@ -315,16 +284,14 @@ std::pair<Surface, Surface> RasterizerOpenGL::ConfigureFramebuffers(bool using_c
         using_color_fb = false;
     }
 
-    // TODO(bunnei): Implement this
-    const bool has_stencil = false;
-
+    const bool has_stencil = regs.stencil_enable;
     const bool write_color_fb =
         state.color_mask.red_enabled == GL_TRUE || state.color_mask.green_enabled == GL_TRUE ||
         state.color_mask.blue_enabled == GL_TRUE || state.color_mask.alpha_enabled == GL_TRUE;
 
     const bool write_depth_fb =
         (state.depth.test_enabled && state.depth.write_mask == GL_TRUE) ||
-        (has_stencil && state.stencil.test_enabled && state.stencil.write_mask != 0);
+        (has_stencil && (state.stencil.front.write_mask || state.stencil.back.write_mask));
 
     Surface color_surface;
     Surface depth_surface;
@@ -364,41 +331,70 @@ std::pair<Surface, Surface> RasterizerOpenGL::ConfigureFramebuffers(bool using_c
 }
 
 void RasterizerOpenGL::Clear() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto prev_state{state};
+    SCOPE_EXIT({ prev_state.Apply(); });
 
+    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
     bool use_color_fb = false;
     bool use_depth_fb = false;
 
-    GLbitfield clear_mask = 0;
-    if (regs.clear_buffers.R && regs.clear_buffers.G && regs.clear_buffers.B &&
+    OpenGLState clear_state;
+    clear_state.draw.draw_framebuffer = state.draw.draw_framebuffer;
+    clear_state.color_mask.red_enabled = regs.clear_buffers.R ? GL_TRUE : GL_FALSE;
+    clear_state.color_mask.green_enabled = regs.clear_buffers.G ? GL_TRUE : GL_FALSE;
+    clear_state.color_mask.blue_enabled = regs.clear_buffers.B ? GL_TRUE : GL_FALSE;
+    clear_state.color_mask.alpha_enabled = regs.clear_buffers.A ? GL_TRUE : GL_FALSE;
+
+    GLbitfield clear_mask{};
+    if (regs.clear_buffers.R || regs.clear_buffers.G || regs.clear_buffers.B ||
         regs.clear_buffers.A) {
-        clear_mask |= GL_COLOR_BUFFER_BIT;
-        use_color_fb = true;
+        if (regs.clear_buffers.RT == 0) {
+            // We only support clearing the first color attachment for now
+            clear_mask |= GL_COLOR_BUFFER_BIT;
+            use_color_fb = true;
+        } else {
+            // TODO(subv): Add support for the other color attachments
+            LOG_CRITICAL(HW_GPU, "Clear unimplemented for RT {}", regs.clear_buffers.RT);
+        }
     }
     if (regs.clear_buffers.Z) {
+        ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear Z but buffer is not enabled!");
+        use_depth_fb = true;
         clear_mask |= GL_DEPTH_BUFFER_BIT;
-        use_depth_fb = regs.zeta_enable != 0;
 
         // Always enable the depth write when clearing the depth buffer. The depth write mask is
         // ignored when clearing the buffer in the Switch, but OpenGL obeys it so we set it to true.
-        state.depth.test_enabled = true;
-        state.depth.write_mask = GL_TRUE;
-        state.depth.test_func = GL_ALWAYS;
-        state.Apply();
+        clear_state.depth.test_enabled = true;
+        clear_state.depth.test_func = GL_ALWAYS;
+    }
+    if (regs.clear_buffers.S) {
+        ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear stencil but buffer is not enabled!");
+        use_depth_fb = true;
+        clear_mask |= GL_STENCIL_BUFFER_BIT;
+        clear_state.stencil.test_enabled = true;
     }
 
-    if (clear_mask == 0)
+    if (!use_color_fb && !use_depth_fb) {
+        // No color surface nor depth/stencil surface are enabled
         return;
+    }
+
+    if (clear_mask == 0) {
+        // No clear mask is enabled
+        return;
+    }
 
     ScopeAcquireGLContext acquire_context{emu_window};
 
     auto [dirty_color_surface, dirty_depth_surface] =
         ConfigureFramebuffers(use_color_fb, use_depth_fb, false);
 
-    // TODO(Subv): Support clearing only partial colors.
+    clear_state.Apply();
+
     glClearColor(regs.clear_color[0], regs.clear_color[1], regs.clear_color[2],
                  regs.clear_color[3]);
     glClearDepth(regs.clear_depth);
+    glClearStencil(regs.clear_stencil);
 
     glClear(clear_mask);
 
@@ -428,8 +424,8 @@ std::tuple<u8*, GLintptr, GLintptr> RasterizerOpenGL::UploadMemory(u8* buffer_pt
     std::tie(buffer_ptr, buffer_offset) = AlignBuffer(buffer_ptr, buffer_offset, alignment);
     GLintptr uploaded_offset = buffer_offset;
 
-    const auto& memory_manager = Core::System::GetInstance().GPU().memory_manager;
-    const boost::optional<VAddr> cpu_addr{memory_manager->GpuToCpuAddress(gpu_addr)};
+    auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
+    const boost::optional<VAddr> cpu_addr{memory_manager.GpuToCpuAddress(gpu_addr)};
     Memory::ReadBlock(*cpu_addr, buffer_ptr, size);
 
     buffer_ptr += size;
@@ -451,6 +447,7 @@ void RasterizerOpenGL::DrawArrays() {
         ConfigureFramebuffers(true, regs.zeta.Address() != 0 && regs.zeta_enable != 0, true);
 
     SyncDepthTestState();
+    SyncStencilTestState();
     SyncBlendState();
     SyncLogicOpState();
     SyncCullMode();
@@ -461,7 +458,6 @@ void RasterizerOpenGL::DrawArrays() {
     // Draw the vertex batch
     const bool is_indexed = accelerate_draw == AccelDraw::Indexed;
     const u64 index_buffer_size{regs.index_array.count * regs.index_array.FormatSizeInBytes()};
-    const unsigned vertex_num{is_indexed ? regs.index_array.count : regs.vertex_buffer.count};
 
     state.draw.vertex_buffer = stream_buffer.GetHandle();
     state.Apply();
@@ -542,23 +538,21 @@ void RasterizerOpenGL::NotifyMaxwellRegisterChanged(u32 method) {}
 
 void RasterizerOpenGL::FlushAll() {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-    res_cache.FlushRegion(0, Kernel::VMManager::MAX_ADDRESS);
 }
 
 void RasterizerOpenGL::FlushRegion(Tegra::GPUVAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-    res_cache.FlushRegion(addr, size);
 }
 
 void RasterizerOpenGL::InvalidateRegion(Tegra::GPUVAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     res_cache.InvalidateRegion(addr, size);
+    shader_cache.InvalidateRegion(addr, size);
 }
 
 void RasterizerOpenGL::FlushAndInvalidateRegion(Tegra::GPUVAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-    res_cache.FlushRegion(addr, size);
-    res_cache.InvalidateRegion(addr, size);
+    InvalidateRegion(addr, size);
 }
 
 bool RasterizerOpenGL::AccelerateDisplayTransfer(const void* config) {
@@ -643,15 +637,17 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr
     }
 }
 
-std::tuple<u8*, GLintptr, u32> RasterizerOpenGL::SetupConstBuffers(
-    u8* buffer_ptr, GLintptr buffer_offset, Maxwell::ShaderStage stage, GLuint program,
-    u32 current_bindpoint, const std::vector<GLShader::ConstBufferEntry>& entries) {
+std::tuple<u8*, GLintptr, u32> RasterizerOpenGL::SetupConstBuffers(u8* buffer_ptr,
+                                                                   GLintptr buffer_offset,
+                                                                   Maxwell::ShaderStage stage,
+                                                                   Shader& shader,
+                                                                   u32 current_bindpoint) {
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
+    const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<size_t>(stage)];
+    const auto& entries = shader->GetShaderEntries().const_buffer_entries;
 
     // Upload only the enabled buffers from the 16 constbuffers of each shader stage
-    const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<size_t>(stage)];
-
     for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
         const auto& used_buffer = entries[bindpoint];
         const auto& buffer = shader_stage.const_buffers[used_buffer.GetIndex()];
@@ -690,12 +686,9 @@ std::tuple<u8*, GLintptr, u32> RasterizerOpenGL::SetupConstBuffers(
                           stream_buffer.GetHandle(), const_buffer_offset, size);
 
         // Now configure the bindpoint of the buffer inside the shader
-        const std::string buffer_name = used_buffer.GetName();
-        const GLuint index =
-            glGetProgramResourceIndex(program, GL_UNIFORM_BLOCK, buffer_name.c_str());
-        if (index != GL_INVALID_INDEX) {
-            glUniformBlockBinding(program, index, current_bindpoint + bindpoint);
-        }
+        glUniformBlockBinding(shader->GetProgramHandle(),
+                              shader->GetProgramResourceIndex(used_buffer.GetName()),
+                              current_bindpoint + bindpoint);
     }
 
     state.Apply();
@@ -703,10 +696,10 @@ std::tuple<u8*, GLintptr, u32> RasterizerOpenGL::SetupConstBuffers(
     return {buffer_ptr, buffer_offset, current_bindpoint + static_cast<u32>(entries.size())};
 }
 
-u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, GLuint program, u32 current_unit,
-                                    const std::vector<GLShader::SamplerEntry>& entries) {
+u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader, u32 current_unit) {
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
+    const auto& entries = shader->GetShaderEntries().texture_samplers;
 
     ASSERT_MSG(current_unit + entries.size() <= std::size(state.texture_units),
                "Exceeded the number of active textures.");
@@ -716,12 +709,9 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, GLuint program,
         u32 current_bindpoint = current_unit + bindpoint;
 
         // Bind the uniform to the sampler.
-        GLint uniform = glGetUniformLocation(program, entry.GetName().c_str());
-        if (uniform == -1) {
-            continue;
-        }
 
-        glProgramUniform1i(program, uniform, current_bindpoint);
+        glProgramUniform1i(shader->GetProgramHandle(), shader->GetUniformLocation(entry.GetName()),
+                           current_bindpoint);
 
         const auto texture = maxwell3d.GetStageTexture(entry.GetStage(), entry.GetOffset());
 
@@ -841,6 +831,34 @@ void RasterizerOpenGL::SyncDepthTestState() {
     state.depth.test_func = MaxwellToGL::ComparisonOp(regs.depth_test_func);
 }
 
+void RasterizerOpenGL::SyncStencilTestState() {
+    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    state.stencil.test_enabled = regs.stencil_enable != 0;
+
+    if (!regs.stencil_enable) {
+        return;
+    }
+
+    // TODO(bunnei): Verify behavior when this is not set
+    ASSERT(regs.stencil_two_side_enable);
+
+    state.stencil.front.test_func = MaxwellToGL::ComparisonOp(regs.stencil_front_func_func);
+    state.stencil.front.test_ref = regs.stencil_front_func_ref;
+    state.stencil.front.test_mask = regs.stencil_front_func_mask;
+    state.stencil.front.action_stencil_fail = MaxwellToGL::StencilOp(regs.stencil_front_op_fail);
+    state.stencil.front.action_depth_fail = MaxwellToGL::StencilOp(regs.stencil_front_op_zfail);
+    state.stencil.front.action_depth_pass = MaxwellToGL::StencilOp(regs.stencil_front_op_zpass);
+    state.stencil.front.write_mask = regs.stencil_front_mask;
+
+    state.stencil.back.test_func = MaxwellToGL::ComparisonOp(regs.stencil_back_func_func);
+    state.stencil.back.test_ref = regs.stencil_back_func_ref;
+    state.stencil.back.test_mask = regs.stencil_back_func_mask;
+    state.stencil.back.action_stencil_fail = MaxwellToGL::StencilOp(regs.stencil_back_op_fail);
+    state.stencil.back.action_depth_fail = MaxwellToGL::StencilOp(regs.stencil_back_op_zfail);
+    state.stencil.back.action_depth_pass = MaxwellToGL::StencilOp(regs.stencil_back_op_zpass);
+    state.stencil.back.write_mask = regs.stencil_back_mask;
+}
+
 void RasterizerOpenGL::SyncBlendState() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
 
@@ -872,7 +890,8 @@ void RasterizerOpenGL::SyncLogicOpState() {
     if (!state.logic_op.enabled)
         return;
 
-    ASSERT_MSG(regs.blend.enable == 0, "Blending and logic op can't be enabled at the same time.");
+    ASSERT_MSG(regs.blend.enable[0] == 0,
+               "Blending and logic op can't be enabled at the same time.");
 
     state.logic_op.operation = MaxwellToGL::LogicOp(regs.logic_op.operation);
 }

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -17,6 +17,7 @@
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
+#include "video_core/renderer_opengl/gl_shader_cache.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
@@ -99,26 +100,23 @@ private:
     /*
      * Configures the current constbuffers to use for the draw command.
      * @param stage The shader stage to configure buffers for.
-     * @param program The OpenGL program object that contains the specified stage.
+     * @param shader The shader object that contains the specified stage.
      * @param current_bindpoint The offset at which to start counting new buffer bindpoints.
-     * @param entries Vector describing the buffers that are actually used in the guest shader.
      * @returns The next available bindpoint for use in the next shader stage.
      */
     std::tuple<u8*, GLintptr, u32> SetupConstBuffers(
         u8* buffer_ptr, GLintptr buffer_offset, Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
-        GLuint program, u32 current_bindpoint,
-        const std::vector<GLShader::ConstBufferEntry>& entries);
+        Shader& shader, u32 current_bindpoint);
 
     /*
      * Configures the current textures to use for the draw command.
      * @param stage The shader stage to configure textures for.
-     * @param program The OpenGL program object that contains the specified stage.
+     * @param shader The shader object that contains the specified stage.
      * @param current_unit The offset at which to start counting unused texture units.
-     * @param entries Vector describing the textures that are actually used in the guest shader.
      * @returns The next available bindpoint for use in the next shader stage.
      */
-    u32 SetupTextures(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, GLuint program,
-                      u32 current_unit, const std::vector<GLShader::SamplerEntry>& entries);
+    u32 SetupTextures(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, Shader& shader,
+                      u32 current_unit);
 
     /// Syncs the viewport to match the guest state
     void SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect);
@@ -141,6 +139,9 @@ private:
     /// Syncs the depth test state to match the guest state
     void SyncDepthTestState();
 
+    /// Syncs the stencil test state to match the guest state
+    void SyncStencilTestState();
+
     /// Syncs the blend state to match the guest state
     void SyncBlendState();
 
@@ -154,6 +155,7 @@ private:
     OpenGLState state;
 
     RasterizerCacheOpenGL res_cache;
+    ShaderCacheOpenGL shader_cache;
 
     Core::Frontend::EmuWindow& emu_window;
 

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -168,8 +168,8 @@ static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType
 }
 
 VAddr SurfaceParams::GetCpuAddr() const {
-    const auto& gpu = Core::System::GetInstance().GPU();
-    return *gpu.memory_manager->GpuToCpuAddress(addr);
+    auto& gpu = Core::System::GetInstance().GPU();
+    return *gpu.MemoryManager().GpuToCpuAddress(addr);
 }
 
 static bool IsPixelFormatASTC(PixelFormat format) {
@@ -220,14 +220,14 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_bu
                 Tegra::GPUVAddr addr) {
     constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
     constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format);
-    const auto& gpu = Core::System::GetInstance().GPU();
+    auto& gpu = Core::System::GetInstance().GPU();
 
     if (morton_to_gl) {
         // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
         // pixel values.
         const u32 tile_size{IsFormatBCn(format) ? 4U : 1U};
         const std::vector<u8> data =
-            Tegra::Texture::UnswizzleTexture(*gpu.memory_manager->GpuToCpuAddress(addr), tile_size,
+            Tegra::Texture::UnswizzleTexture(*gpu.MemoryManager().GpuToCpuAddress(addr), tile_size,
                                              bytes_per_pixel, stride, height, block_height);
         const size_t size_to_copy{std::min(gl_buffer.size(), data.size())};
         gl_buffer.assign(data.begin(), data.begin() + size_to_copy);
@@ -237,7 +237,7 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_bu
         LOG_WARNING(Render_OpenGL, "need to use correct swizzle/GOB parameters!");
         VideoCore::MortonCopyPixels128(
             stride, height, bytes_per_pixel, gl_bytes_per_pixel,
-            Memory::GetPointer(*gpu.memory_manager->GpuToCpuAddress(addr)), gl_buffer.data(),
+            Memory::GetPointer(*gpu.MemoryManager().GpuToCpuAddress(addr)), gl_buffer.data(),
             morton_to_gl);
     }
 }
@@ -677,12 +677,6 @@ RasterizerCacheOpenGL::RasterizerCacheOpenGL() {
     draw_framebuffer.Create();
 }
 
-RasterizerCacheOpenGL::~RasterizerCacheOpenGL() {
-    while (!surface_cache.empty()) {
-        UnregisterSurface(surface_cache.begin()->second);
-    }
-}
-
 Surface RasterizerCacheOpenGL::GetTextureSurface(const Tegra::Texture::FullTextureInfo& config) {
     return GetSurface(SurfaceParams::CreateForTexture(config));
 }
@@ -760,37 +754,48 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool pres
         return {};
     }
 
-    const auto& gpu = Core::System::GetInstance().GPU();
+    auto& gpu = Core::System::GetInstance().GPU();
     // Don't try to create any entries in the cache if the address of the texture is invalid.
-    if (gpu.memory_manager->GpuToCpuAddress(params.addr) == boost::none)
+    if (gpu.MemoryManager().GpuToCpuAddress(params.addr) == boost::none)
         return {};
 
     // Look up surface in the cache based on address
-    const auto& search{surface_cache.find(params.addr)};
-    Surface surface;
-    if (search != surface_cache.end()) {
-        surface = search->second;
+    Surface surface{TryGet(params.addr)};
+    if (surface) {
         if (Settings::values.use_accurate_framebuffers) {
             // If use_accurate_framebuffers is enabled, always load from memory
             FlushSurface(surface);
-            UnregisterSurface(surface);
+            Unregister(surface);
         } else if (surface->GetSurfaceParams().IsCompatibleSurface(params)) {
             // Use the cached surface as-is
             return surface;
         } else if (preserve_contents) {
             // If surface parameters changed and we care about keeping the previous data, recreate
             // the surface from the old one
-            return RecreateSurface(surface, params);
+            Unregister(surface);
+            Surface new_surface{RecreateSurface(surface, params)};
+            Register(new_surface);
+            return new_surface;
         } else {
             // Delete the old surface before creating a new one to prevent collisions.
-            UnregisterSurface(surface);
+            Unregister(surface);
         }
     }
 
+    // Try to get a previously reserved surface
+    surface = TryGetReservedSurface(params);
+
     // No surface found - create a new one
-    surface = std::make_shared<CachedSurface>(params);
-    RegisterSurface(surface);
-    LoadSurface(surface);
+    if (!surface) {
+        surface = std::make_shared<CachedSurface>(params);
+        ReserveSurface(surface);
+        Register(surface);
+    }
+
+    // Only load surface from memory if we care about the contents
+    if (preserve_contents) {
+        LoadSurface(surface);
+    }
 
     return surface;
 }
@@ -799,13 +804,18 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
                                                const SurfaceParams& new_params) {
     // Verify surface is compatible for blitting
     const auto& params{surface->GetSurfaceParams()};
-    ASSERT(params.type == new_params.type);
-    ASSERT_MSG(params.GetCompressionFactor(params.pixel_format) == 1,
-               "Compressed texture reinterpretation is not supported");
 
     // Create a new surface with the new parameters, and blit the previous surface to it
     Surface new_surface{std::make_shared<CachedSurface>(new_params)};
 
+    // If format is unchanged, we can do a faster blit without reinterpreting pixel data
+    if (params.pixel_format == new_params.pixel_format) {
+        BlitTextures(surface->Texture().handle, params.GetRect(), new_surface->Texture().handle,
+                     new_surface->GetSurfaceParams().GetRect(), params.type,
+                     read_framebuffer.handle, draw_framebuffer.handle);
+        return new_surface;
+    }
+
     auto source_format = GetFormatTuple(params.pixel_format, params.component_type);
     auto dest_format = GetFormatTuple(new_params.pixel_format, new_params.component_type);
 
@@ -818,9 +828,13 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
 
     glBindBuffer(GL_PIXEL_PACK_BUFFER, pbo.handle);
     glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW_ARB);
-    glGetTextureImage(surface->Texture().handle, 0, source_format.format, source_format.type,
-                      params.SizeInBytes(), nullptr);
-
+    if (source_format.compressed) {
+        glGetCompressedTextureImage(surface->Texture().handle, 0,
+                                    static_cast<GLsizei>(params.SizeInBytes()), nullptr);
+    } else {
+        glGetTextureImage(surface->Texture().handle, 0, source_format.format, source_format.type,
+                          static_cast<GLsizei>(params.SizeInBytes()), nullptr);
+    }
     // If the new texture is bigger than the previous one, we need to fill in the rest with data
     // from the CPU.
     if (params.SizeInBytes() < new_params.SizeInBytes()) {
@@ -834,7 +848,7 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
                                  "reinterpretation but the texture is tiled.");
         }
         size_t remaining_size = new_params.SizeInBytes() - params.SizeInBytes();
-        auto address = Core::System::GetInstance().GPU().memory_manager->GpuToCpuAddress(
+        auto address = Core::System::GetInstance().GPU().MemoryManager().GpuToCpuAddress(
             new_params.addr + params.SizeInBytes());
         std::vector<u8> data(remaining_size);
         Memory::ReadBlock(*address, data.data(), data.size());
@@ -846,17 +860,21 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
     const auto& dest_rect{new_params.GetRect()};
 
     glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo.handle);
-    glTextureSubImage2D(
-        new_surface->Texture().handle, 0, 0, 0, static_cast<GLsizei>(dest_rect.GetWidth()),
-        static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format, dest_format.type, nullptr);
+    if (dest_format.compressed) {
+        glCompressedTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0,
+                                  static_cast<GLsizei>(dest_rect.GetWidth()),
+                                  static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format,
+                                  static_cast<GLsizei>(new_params.SizeInBytes()), nullptr);
+    } else {
+        glTextureSubImage2D(new_surface->Texture().handle, 0, 0, 0,
+                            static_cast<GLsizei>(dest_rect.GetWidth()),
+                            static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format,
+                            dest_format.type, nullptr);
+    }
     glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
 
     pbo.Release();
 
-    // Update cache accordingly
-    UnregisterSurface(surface);
-    RegisterSurface(new_surface);
-
     return new_surface;
 }
 
@@ -868,7 +886,7 @@ Surface RasterizerCacheOpenGL::TryFindFramebufferSurface(VAddr cpu_addr) const {
     // framebuffer overlaps surfaces.
 
     std::vector<Surface> surfaces;
-    for (const auto& surface : surface_cache) {
+    for (const auto& surface : GetCache()) {
         const auto& params = surface.second->GetSurfaceParams();
         const VAddr surface_cpu_addr = params.GetCpuAddr();
         if (cpu_addr >= surface_cpu_addr && cpu_addr < (surface_cpu_addr + params.size_in_bytes)) {
@@ -886,88 +904,19 @@ Surface RasterizerCacheOpenGL::TryFindFramebufferSurface(VAddr cpu_addr) const {
     return surfaces[0];
 }
 
-void RasterizerCacheOpenGL::FlushRegion(Tegra::GPUVAddr /*addr*/, size_t /*size*/) {
-    // TODO(bunnei): This is unused in the current implementation of the rasterizer cache. We should
-    // probably implement this in the future, but for now, the `use_accurate_framebufers` setting
-    // can be used to always flush.
+void RasterizerCacheOpenGL::ReserveSurface(const Surface& surface) {
+    const auto& surface_reserve_key{SurfaceReserveKey::Create(surface->GetSurfaceParams())};
+    surface_reserve[surface_reserve_key] = surface;
 }
 
-void RasterizerCacheOpenGL::InvalidateRegion(Tegra::GPUVAddr addr, size_t size) {
-    for (auto iter = surface_cache.cbegin(); iter != surface_cache.cend();) {
-        const auto& surface{iter->second};
-        const auto& params{surface->GetSurfaceParams()};
-
-        ++iter;
-
-        if (params.IsOverlappingRegion(addr, size)) {
-            UnregisterSurface(surface);
-        }
+Surface RasterizerCacheOpenGL::TryGetReservedSurface(const SurfaceParams& params) {
+    const auto& surface_reserve_key{SurfaceReserveKey::Create(params)};
+    auto search{surface_reserve.find(surface_reserve_key)};
+    if (search != surface_reserve.end()) {
+        Register(search->second);
+        return search->second;
     }
-}
-
-void RasterizerCacheOpenGL::RegisterSurface(const Surface& surface) {
-    const auto& params{surface->GetSurfaceParams()};
-    const auto& search{surface_cache.find(params.addr)};
-
-    if (search != surface_cache.end()) {
-        // Registered already
-        return;
-    }
-
-    surface_cache[params.addr] = surface;
-    UpdatePagesCachedCount(params.addr, params.size_in_bytes, 1);
-}
-
-void RasterizerCacheOpenGL::UnregisterSurface(const Surface& surface) {
-    const auto& params{surface->GetSurfaceParams()};
-    const auto& search{surface_cache.find(params.addr)};
-
-    if (search == surface_cache.end()) {
-        // Unregistered already
-        return;
-    }
-
-    UpdatePagesCachedCount(params.addr, params.size_in_bytes, -1);
-    surface_cache.erase(search);
-}
-
-template <typename Map, typename Interval>
-constexpr auto RangeFromInterval(Map& map, const Interval& interval) {
-    return boost::make_iterator_range(map.equal_range(interval));
-}
-
-void RasterizerCacheOpenGL::UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta) {
-    const u64 num_pages = ((addr + size - 1) >> Tegra::MemoryManager::PAGE_BITS) -
-                          (addr >> Tegra::MemoryManager::PAGE_BITS) + 1;
-    const u64 page_start = addr >> Tegra::MemoryManager::PAGE_BITS;
-    const u64 page_end = page_start + num_pages;
-
-    // Interval maps will erase segments if count reaches 0, so if delta is negative we have to
-    // subtract after iterating
-    const auto pages_interval = PageMap::interval_type::right_open(page_start, page_end);
-    if (delta > 0)
-        cached_pages.add({pages_interval, delta});
-
-    for (const auto& pair : RangeFromInterval(cached_pages, pages_interval)) {
-        const auto interval = pair.first & pages_interval;
-        const int count = pair.second;
-
-        const Tegra::GPUVAddr interval_start_addr = boost::icl::first(interval)
-                                                    << Tegra::MemoryManager::PAGE_BITS;
-        const Tegra::GPUVAddr interval_end_addr = boost::icl::last_next(interval)
-                                                  << Tegra::MemoryManager::PAGE_BITS;
-        const u64 interval_size = interval_end_addr - interval_start_addr;
-
-        if (delta > 0 && count == delta)
-            Memory::RasterizerMarkRegionCached(interval_start_addr, interval_size, true);
-        else if (delta < 0 && count == -delta)
-            Memory::RasterizerMarkRegionCached(interval_start_addr, interval_size, false);
-        else
-            ASSERT(count >= 0);
-    }
-
-    if (delta < 0)
-        cached_pages.add({pages_interval, delta});
+    return {};
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -8,11 +8,12 @@
 #include <map>
 #include <memory>
 #include <vector>
-#include <boost/icl/interval_map.hpp>
 
 #include "common/common_types.h"
+#include "common/hash.h"
 #include "common/math_util.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/textures/texture.h"
 
@@ -21,7 +22,6 @@ namespace OpenGL {
 class CachedSurface;
 using Surface = std::shared_ptr<CachedSurface>;
 using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, MathUtil::Rectangle<u32>>;
-using PageMap = boost::icl::interval_map<u64, int>;
 
 struct SurfaceParams {
     enum class PixelFormat {
@@ -631,11 +631,6 @@ struct SurfaceParams {
     /// Returns the CPU virtual address for this surface
     VAddr GetCpuAddr() const;
 
-    /// Returns true if the specified region overlaps with this surface's region in Switch memory
-    bool IsOverlappingRegion(Tegra::GPUVAddr region_addr, size_t region_size) const {
-        return addr <= (region_addr + region_size) && region_addr <= (addr + size_in_bytes);
-    }
-
     /// Creates SurfaceParams from a texture configuration
     static SurfaceParams CreateForTexture(const Tegra::Texture::FullTextureInfo& config);
 
@@ -682,10 +677,39 @@ struct SurfaceParams {
     u32 cache_height;
 };
 
+}; // namespace OpenGL
+
+/// Hashable variation of SurfaceParams, used for a key in the surface cache
+struct SurfaceReserveKey : Common::HashableStruct<OpenGL::SurfaceParams> {
+    static SurfaceReserveKey Create(const OpenGL::SurfaceParams& params) {
+        SurfaceReserveKey res;
+        res.state = params;
+        return res;
+    }
+};
+namespace std {
+template <>
+struct hash<SurfaceReserveKey> {
+    size_t operator()(const SurfaceReserveKey& k) const {
+        return k.Hash();
+    }
+};
+} // namespace std
+
+namespace OpenGL {
+
 class CachedSurface final {
 public:
     CachedSurface(const SurfaceParams& params);
 
+    Tegra::GPUVAddr GetAddr() const {
+        return params.addr;
+    }
+
+    size_t GetSizeInBytes() const {
+        return params.size_in_bytes;
+    }
+
     const OGLTexture& Texture() const {
         return texture;
     }
@@ -715,10 +739,9 @@ private:
     SurfaceParams params;
 };
 
-class RasterizerCacheOpenGL final : NonCopyable {
+class RasterizerCacheOpenGL final : public RasterizerCache<Surface> {
 public:
     RasterizerCacheOpenGL();
-    ~RasterizerCacheOpenGL();
 
     /// Get a surface based on the texture configuration
     Surface GetTextureSurface(const Tegra::Texture::FullTextureInfo& config);
@@ -733,12 +756,6 @@ public:
     /// Tries to find a framebuffer GPU address based on the provided CPU address
     Surface TryFindFramebufferSurface(VAddr cpu_addr) const;
 
-    /// Write any cached resources overlapping the region back to memory (if dirty)
-    void FlushRegion(Tegra::GPUVAddr addr, size_t size);
-
-    /// Mark the specified region as being invalidated
-    void InvalidateRegion(Tegra::GPUVAddr addr, size_t size);
-
 private:
     void LoadSurface(const Surface& surface);
     Surface GetSurface(const SurfaceParams& params, bool preserve_contents = true);
@@ -746,17 +763,16 @@ private:
     /// Recreates a surface with new parameters
     Surface RecreateSurface(const Surface& surface, const SurfaceParams& new_params);
 
-    /// Register surface into the cache
-    void RegisterSurface(const Surface& surface);
+    /// Reserves a unique surface that can be reused later
+    void ReserveSurface(const Surface& surface);
 
-    /// Remove surface from the cache
-    void UnregisterSurface(const Surface& surface);
+    /// Tries to get a reserved surface for the specified parameters
+    Surface TryGetReservedSurface(const SurfaceParams& params);
 
-    /// Increase/decrease the number of surface in pages touching the specified region
-    void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta);
-
-    std::unordered_map<Tegra::GPUVAddr, Surface> surface_cache;
-    PageMap cached_pages;
+    /// The surface reserve is a "backup" cache, this is where we put unique surfaces that have
+    /// previously been used. This is to prevent surfaces from being constantly created and
+    /// destroyed when used with different surface parameters.
+    std::unordered_map<SurfaceReserveKey, Surface> surface_reserve;
 
     OGLFramebuffer read_framebuffer;
     OGLFramebuffer draw_framebuffer;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -0,0 +1,130 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "core/memory.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/renderer_opengl/gl_shader_cache.h"
+#include "video_core/renderer_opengl/gl_shader_manager.h"
+
+namespace OpenGL {
+
+/// Gets the address for the specified shader stage program
+static Tegra::GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) {
+    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    auto& shader_config = gpu.regs.shader_config[static_cast<size_t>(program)];
+
+    return gpu.regs.code_address.CodeAddress() + shader_config.offset;
+}
+
+/// Gets the shader program code from memory for the specified address
+static GLShader::ProgramCode GetShaderCode(Tegra::GPUVAddr addr) {
+    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+
+    GLShader::ProgramCode program_code(GLShader::MAX_PROGRAM_CODE_LENGTH);
+    const boost::optional<VAddr> cpu_address{gpu.memory_manager.GpuToCpuAddress(addr)};
+    Memory::ReadBlock(*cpu_address, program_code.data(), program_code.size() * sizeof(u64));
+
+    return program_code;
+}
+
+/// Helper function to set shader uniform block bindings for a single shader stage
+static void SetShaderUniformBlockBinding(GLuint shader, const char* name,
+                                         Maxwell::ShaderStage binding, size_t expected_size) {
+    const GLuint ub_index = glGetUniformBlockIndex(shader, name);
+    if (ub_index == GL_INVALID_INDEX) {
+        return;
+    }
+
+    GLint ub_size = 0;
+    glGetActiveUniformBlockiv(shader, ub_index, GL_UNIFORM_BLOCK_DATA_SIZE, &ub_size);
+    ASSERT_MSG(static_cast<size_t>(ub_size) == expected_size,
+               "Uniform block size did not match! Got {}, expected {}", ub_size, expected_size);
+    glUniformBlockBinding(shader, ub_index, static_cast<GLuint>(binding));
+}
+
+/// Sets shader uniform block bindings for an entire shader program
+static void SetShaderUniformBlockBindings(GLuint shader) {
+    SetShaderUniformBlockBinding(shader, "vs_config", Maxwell::ShaderStage::Vertex,
+                                 sizeof(GLShader::MaxwellUniformData));
+    SetShaderUniformBlockBinding(shader, "gs_config", Maxwell::ShaderStage::Geometry,
+                                 sizeof(GLShader::MaxwellUniformData));
+    SetShaderUniformBlockBinding(shader, "fs_config", Maxwell::ShaderStage::Fragment,
+                                 sizeof(GLShader::MaxwellUniformData));
+}
+
+CachedShader::CachedShader(Tegra::GPUVAddr addr, Maxwell::ShaderProgram program_type)
+    : addr{addr}, program_type{program_type}, setup{GetShaderCode(addr)} {
+
+    GLShader::ProgramResult program_result;
+    GLenum gl_type{};
+
+    switch (program_type) {
+    case Maxwell::ShaderProgram::VertexA:
+        // VertexB is always enabled, so when VertexA is enabled, we have two vertex shaders.
+        // Conventional HW does not support this, so we combine VertexA and VertexB into one
+        // stage here.
+        setup.SetProgramB(GetShaderCode(GetShaderAddress(Maxwell::ShaderProgram::VertexB)));
+    case Maxwell::ShaderProgram::VertexB:
+        program_result = GLShader::GenerateVertexShader(setup);
+        gl_type = GL_VERTEX_SHADER;
+        break;
+    case Maxwell::ShaderProgram::Fragment:
+        program_result = GLShader::GenerateFragmentShader(setup);
+        gl_type = GL_FRAGMENT_SHADER;
+        break;
+    default:
+        LOG_CRITICAL(HW_GPU, "Unimplemented program_type={}", static_cast<u32>(program_type));
+        UNREACHABLE();
+        return;
+    }
+
+    entries = program_result.second;
+
+    OGLShader shader;
+    shader.Create(program_result.first.c_str(), gl_type);
+    program.Create(true, shader.handle);
+    SetShaderUniformBlockBindings(program.handle);
+}
+
+GLuint CachedShader::GetProgramResourceIndex(const std::string& name) {
+    auto search{resource_cache.find(name)};
+    if (search == resource_cache.end()) {
+        const GLuint index{
+            glGetProgramResourceIndex(program.handle, GL_UNIFORM_BLOCK, name.c_str())};
+        resource_cache[name] = index;
+        return index;
+    }
+
+    return search->second;
+}
+
+GLint CachedShader::GetUniformLocation(const std::string& name) {
+    auto search{uniform_cache.find(name)};
+    if (search == uniform_cache.end()) {
+        const GLint index{glGetUniformLocation(program.handle, name.c_str())};
+        uniform_cache[name] = index;
+        return index;
+    }
+
+    return search->second;
+}
+
+Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
+    const Tegra::GPUVAddr program_addr{GetShaderAddress(program)};
+
+    // Look up shader in the cache based on address
+    Shader shader{TryGet(program_addr)};
+
+    if (!shader) {
+        // No shader found - create a new one
+        shader = std::make_shared<CachedShader>(program_addr, program);
+        Register(shader);
+    }
+
+    return shader;
+}
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -0,0 +1,69 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <unordered_map>
+
+#include "common/common_types.h"
+#include "video_core/memory_manager.h"
+#include "video_core/rasterizer_cache.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+#include "video_core/renderer_opengl/gl_shader_gen.h"
+
+namespace OpenGL {
+
+class CachedShader;
+using Shader = std::shared_ptr<CachedShader>;
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+
+class CachedShader final {
+public:
+    CachedShader(Tegra::GPUVAddr addr, Maxwell::ShaderProgram program_type);
+
+    /// Gets the address of the shader in guest memory, required for cache management
+    Tegra::GPUVAddr GetAddr() const {
+        return addr;
+    }
+
+    /// Gets the size of the shader in guest memory, required for cache management
+    size_t GetSizeInBytes() const {
+        return sizeof(GLShader::ProgramCode);
+    }
+
+    /// Gets the shader entries for the shader
+    const GLShader::ShaderEntries& GetShaderEntries() const {
+        return entries;
+    }
+
+    /// Gets the GL program handle for the shader
+    GLuint GetProgramHandle() const {
+        return program.handle;
+    }
+
+    /// Gets the GL program resource location for the specified resource, caching as needed
+    GLuint GetProgramResourceIndex(const std::string& name);
+
+    /// Gets the GL uniform location for the specified resource, caching as needed
+    GLint GetUniformLocation(const std::string& name);
+
+private:
+    Tegra::GPUVAddr addr;
+    Maxwell::ShaderProgram program_type;
+    GLShader::ShaderSetup setup;
+    GLShader::ShaderEntries entries;
+    OGLProgram program;
+
+    std::unordered_map<std::string, GLuint> resource_cache;
+    std::unordered_map<std::string, GLint> uniform_cache;
+};
+
+class ShaderCacheOpenGL final : public RasterizerCache<Shader> {
+public:
+    /// Gets the current specified shader stage program
+    Shader GetStageProgram(Maxwell::ShaderProgram program);
+};
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -849,6 +849,33 @@ private:
         }
     }
 
+    void WriteLop3Instruction(Register dest, const std::string& op_a, const std::string& op_b,
+                              const std::string& op_c, const std::string& imm_lut) {
+        if (dest == Tegra::Shader::Register::ZeroIndex) {
+            return;
+        }
+
+        static constexpr std::array<const char*, 32> shift_amounts = {
+            "0",  "1",  "2",  "3",  "4",  "5",  "6",  "7",  "8",  "9",  "10",
+            "11", "12", "13", "14", "15", "16", "17", "18", "19", "20", "21",
+            "22", "23", "24", "25", "26", "27", "28", "29", "30", "31"};
+
+        std::string result;
+        result += '(';
+
+        for (size_t i = 0; i < shift_amounts.size(); ++i) {
+            if (i)
+                result += '|';
+            result += "(((" + imm_lut + " >> (((" + op_c + " >> " + shift_amounts[i] +
+                      ") & 1) | ((" + op_b + " >> " + shift_amounts[i] + ") & 1) << 1 | ((" + op_a +
+                      " >> " + shift_amounts[i] + ") & 1) << 2)) & 1) << " + shift_amounts[i] + ")";
+        }
+
+        result += ')';
+
+        regs.SetRegisterToInteger(dest, true, 0, result, 1, 1);
+    }
+
     void WriteTexsInstruction(const Instruction& instr, const std::string& coord,
                               const std::string& texture) {
         // Add an extra scope and declare the texture coords inside to prevent
@@ -1275,6 +1302,15 @@ private:
                                           "((" + op_a + " << " + shift + ") + " + op_b + ')', 1, 1);
                 break;
             }
+            case OpCode::Id::POPC_C:
+            case OpCode::Id::POPC_R:
+            case OpCode::Id::POPC_IMM: {
+                if (instr.popc.invert) {
+                    op_b = "~(" + op_b + ')';
+                }
+                regs.SetRegisterToInteger(instr.gpr0, true, 0, "bitCount(" + op_b + ')', 1, 1);
+                break;
+            }
             case OpCode::Id::SEL_C:
             case OpCode::Id::SEL_R:
             case OpCode::Id::SEL_IMM: {
@@ -1297,6 +1333,20 @@ private:
                                     instr.alu.lop.pred_result_mode, instr.alu.lop.pred48);
                 break;
             }
+            case OpCode::Id::LOP3_C:
+            case OpCode::Id::LOP3_R:
+            case OpCode::Id::LOP3_IMM: {
+                std::string op_c = regs.GetRegisterAsInteger(instr.gpr39);
+                std::string lut;
+                if (opcode->GetId() == OpCode::Id::LOP3_R) {
+                    lut = '(' + std::to_string(instr.alu.lop3.GetImmLut28()) + ')';
+                } else {
+                    lut = '(' + std::to_string(instr.alu.lop3.GetImmLut48()) + ')';
+                }
+
+                WriteLop3Instruction(instr.gpr0, op_a, op_b, op_c, lut);
+                break;
+            }
             case OpCode::Id::IMNMX_C:
             case OpCode::Id::IMNMX_R:
             case OpCode::Id::IMNMX_IMM: {

src/video_core/renderer_opengl/gl_shader_gen.cpp

@@ -13,7 +13,7 @@ using Tegra::Engines::Maxwell3D;
 
 static constexpr u32 PROGRAM_OFFSET{10};
 
-ProgramResult GenerateVertexShader(const ShaderSetup& setup, const MaxwellVSConfig& config) {
+ProgramResult GenerateVertexShader(const ShaderSetup& setup) {
     std::string out = "#version 430 core\n";
     out += "#extension GL_ARB_separate_shader_objects : enable\n\n";
     out += Decompiler::GetCommonDeclarations();
@@ -75,7 +75,7 @@ void main() {
     return {out, program.second};
 }
 
-ProgramResult GenerateFragmentShader(const ShaderSetup& setup, const MaxwellFSConfig& config) {
+ProgramResult GenerateFragmentShader(const ShaderSetup& setup) {
     std::string out = "#version 430 core\n";
     out += "#extension GL_ARB_separate_shader_objects : enable\n\n";
     out += Decompiler::GetCommonDeclarations();

src/video_core/renderer_opengl/gl_shader_gen.h

@@ -6,17 +6,14 @@
 
 #include <array>
 #include <string>
-#include <type_traits>
-#include <utility>
 #include <vector>
+
 #include "common/common_types.h"
-#include "common/hash.h"
 
 namespace OpenGL::GLShader {
 
 constexpr size_t MAX_PROGRAM_CODE_LENGTH{0x1000};
-
-using ProgramCode = std::array<u64, MAX_PROGRAM_CODE_LENGTH>;
+using ProgramCode = std::vector<u64>;
 
 class ConstBufferEntry {
     using Maxwell = Tegra::Engines::Maxwell3D::Regs;
@@ -115,8 +112,8 @@ struct ShaderEntries {
 using ProgramResult = std::pair<std::string, ShaderEntries>;
 
 struct ShaderSetup {
-    ShaderSetup(const ProgramCode& program_code) {
-        program.code = program_code;
+    explicit ShaderSetup(ProgramCode program_code) {
+        program.code = std::move(program_code);
     }
 
     struct {
@@ -124,19 +121,9 @@ struct ShaderSetup {
         ProgramCode code_b; // Used for dual vertex shaders
     } program;
 
-    bool program_code_hash_dirty = true;
-
-    u64 GetProgramCodeHash() {
-        if (program_code_hash_dirty) {
-            program_code_hash = GetNewHash();
-            program_code_hash_dirty = false;
-        }
-        return program_code_hash;
-    }
-
     /// Used in scenarios where we have a dual vertex shaders
-    void SetProgramB(const ProgramCode& program_b) {
-        program.code_b = program_b;
+    void SetProgramB(ProgramCode&& program_b) {
+        program.code_b = std::move(program_b);
         has_program_b = true;
     }
 
@@ -145,68 +132,19 @@ struct ShaderSetup {
     }
 
 private:
-    u64 GetNewHash() const {
-        if (has_program_b) {
-            // Compute hash over dual shader programs
-            return Common::ComputeHash64(&program, sizeof(program));
-        } else {
-            // Compute hash over a single shader program
-            return Common::ComputeHash64(&program.code, program.code.size());
-        }
-    }
-
-    u64 program_code_hash{};
     bool has_program_b{};
 };
 
-struct MaxwellShaderConfigCommon {
-    void Init(ShaderSetup& setup) {
-        program_hash = setup.GetProgramCodeHash();
-    }
-
-    u64 program_hash;
-};
-
-struct MaxwellVSConfig : Common::HashableStruct<MaxwellShaderConfigCommon> {
-    explicit MaxwellVSConfig(ShaderSetup& setup) {
-        state.Init(setup);
-    }
-};
-
-struct MaxwellFSConfig : Common::HashableStruct<MaxwellShaderConfigCommon> {
-    explicit MaxwellFSConfig(ShaderSetup& setup) {
-        state.Init(setup);
-    }
-};
-
 /**
  * Generates the GLSL vertex shader program source code for the given VS program
  * @returns String of the shader source code
  */
-ProgramResult GenerateVertexShader(const ShaderSetup& setup, const MaxwellVSConfig& config);
+ProgramResult GenerateVertexShader(const ShaderSetup& setup);
 
 /**
  * Generates the GLSL fragment shader program source code for the given FS program
  * @returns String of the shader source code
  */
-ProgramResult GenerateFragmentShader(const ShaderSetup& setup, const MaxwellFSConfig& config);
+ProgramResult GenerateFragmentShader(const ShaderSetup& setup);
 
 } // namespace OpenGL::GLShader
-
-namespace std {
-
-template <>
-struct hash<OpenGL::GLShader::MaxwellVSConfig> {
-    size_t operator()(const OpenGL::GLShader::MaxwellVSConfig& k) const {
-        return k.Hash();
-    }
-};
-
-template <>
-struct hash<OpenGL::GLShader::MaxwellFSConfig> {
-    size_t operator()(const OpenGL::GLShader::MaxwellFSConfig& k) const {
-        return k.Hash();
-    }
-};
-
-} // namespace std

src/video_core/renderer_opengl/gl_shader_manager.cpp

@@ -3,39 +3,10 @@
 // Refer to the license.txt file included.
 
 #include "core/core.h"
-#include "core/hle/kernel/process.h"
-#include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 
 namespace OpenGL::GLShader {
 
-namespace Impl {
-static void SetShaderUniformBlockBinding(GLuint shader, const char* name,
-                                         Maxwell3D::Regs::ShaderStage binding,
-                                         size_t expected_size) {
-    const GLuint ub_index = glGetUniformBlockIndex(shader, name);
-    if (ub_index == GL_INVALID_INDEX) {
-        return;
-    }
-
-    GLint ub_size = 0;
-    glGetActiveUniformBlockiv(shader, ub_index, GL_UNIFORM_BLOCK_DATA_SIZE, &ub_size);
-    ASSERT_MSG(static_cast<size_t>(ub_size) == expected_size,
-               "Uniform block size did not match! Got {}, expected {}", ub_size, expected_size);
-    glUniformBlockBinding(shader, ub_index, static_cast<GLuint>(binding));
-}
-
-void SetShaderUniformBlockBindings(GLuint shader) {
-    SetShaderUniformBlockBinding(shader, "vs_config", Maxwell3D::Regs::ShaderStage::Vertex,
-                                 sizeof(MaxwellUniformData));
-    SetShaderUniformBlockBinding(shader, "gs_config", Maxwell3D::Regs::ShaderStage::Geometry,
-                                 sizeof(MaxwellUniformData));
-    SetShaderUniformBlockBinding(shader, "fs_config", Maxwell3D::Regs::ShaderStage::Fragment,
-                                 sizeof(MaxwellUniformData));
-}
-
-} // namespace Impl
-
 void MaxwellUniformData::SetFromRegs(const Maxwell3D::State::ShaderStageInfo& shader_stage) {
     const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
     const auto& regs = gpu.regs;

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -4,12 +4,9 @@
 
 #pragma once
 
-#include <tuple>
-#include <unordered_map>
-#include <boost/functional/hash.hpp>
 #include <glad/glad.h>
+
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_shader_gen.h"
 #include "video_core/renderer_opengl/maxwell_to_gl.h"
 
 namespace OpenGL::GLShader {
@@ -19,10 +16,6 @@ static constexpr size_t NumTextureSamplers = 32;
 
 using Tegra::Engines::Maxwell3D;
 
-namespace Impl {
-void SetShaderUniformBlockBindings(GLuint shader);
-} // namespace Impl
-
 /// Uniform structure for the Uniform Buffer Object, all vectors must be 16-byte aligned
 // NOTE: Always keep a vec4 at the end. The GL spec is not clear whether the alignment at
 //       the end of a uniform block is included in UNIFORM_BLOCK_DATA_SIZE or not.
@@ -36,102 +29,22 @@ static_assert(sizeof(MaxwellUniformData) == 32, "MaxwellUniformData structure si
 static_assert(sizeof(MaxwellUniformData) < 16384,
               "MaxwellUniformData structure must be less than 16kb as per the OpenGL spec");
 
-class OGLShaderStage {
-public:
-    OGLShaderStage() = default;
-
-    void Create(const ProgramResult& program_result, GLenum type) {
-        OGLShader shader;
-        shader.Create(program_result.first.c_str(), type);
-        program.Create(true, shader.handle);
-        Impl::SetShaderUniformBlockBindings(program.handle);
-        entries = program_result.second;
-    }
-    GLuint GetHandle() const {
-        return program.handle;
-    }
-
-    ShaderEntries GetEntries() const {
-        return entries;
-    }
-
-private:
-    OGLProgram program;
-    ShaderEntries entries;
-};
-
-// TODO(wwylele): beautify this doc
-// This is a shader cache designed for translating PICA shader to GLSL shader.
-// The double cache is needed because diffent KeyConfigType, which includes a hash of the code
-// region (including its leftover unused code) can generate the same GLSL code.
-template <typename KeyConfigType,
-          ProgramResult (*CodeGenerator)(const ShaderSetup&, const KeyConfigType&),
-          GLenum ShaderType>
-class ShaderCache {
-public:
-    ShaderCache() = default;
-
-    using Result = std::pair<GLuint, ShaderEntries>;
-
-    Result Get(const KeyConfigType& key, const ShaderSetup& setup) {
-        auto map_it = shader_map.find(key);
-        if (map_it == shader_map.end()) {
-            ProgramResult program = CodeGenerator(setup, key);
-
-            auto [iter, new_shader] = shader_cache.emplace(program.first, OGLShaderStage{});
-            OGLShaderStage& cached_shader = iter->second;
-            if (new_shader) {
-                cached_shader.Create(program, ShaderType);
-            }
-            shader_map[key] = &cached_shader;
-            return {cached_shader.GetHandle(), program.second};
-        } else {
-            return {map_it->second->GetHandle(), map_it->second->GetEntries()};
-        }
-    }
-
-private:
-    std::unordered_map<KeyConfigType, OGLShaderStage*> shader_map;
-    std::unordered_map<std::string, OGLShaderStage> shader_cache;
-};
-
-using VertexShaders = ShaderCache<MaxwellVSConfig, &GenerateVertexShader, GL_VERTEX_SHADER>;
-
-using FragmentShaders = ShaderCache<MaxwellFSConfig, &GenerateFragmentShader, GL_FRAGMENT_SHADER>;
-
 class ProgramManager {
 public:
     ProgramManager() {
         pipeline.Create();
     }
 
-    ShaderEntries UseProgrammableVertexShader(const MaxwellVSConfig& config,
-                                              const ShaderSetup& setup) {
-        ShaderEntries result;
-        std::tie(current.vs, result) = vertex_shaders.Get(config, setup);
-        return result;
+    void UseProgrammableVertexShader(GLuint program) {
+        vs = program;
     }
 
-    ShaderEntries UseProgrammableFragmentShader(const MaxwellFSConfig& config,
-                                                const ShaderSetup& setup) {
-        ShaderEntries result;
-        std::tie(current.fs, result) = fragment_shaders.Get(config, setup);
-        return result;
-    }
-
-    GLuint GetCurrentProgramStage(Maxwell3D::Regs::ShaderStage stage) const {
-        switch (stage) {
-        case Maxwell3D::Regs::ShaderStage::Vertex:
-            return current.vs;
-        case Maxwell3D::Regs::ShaderStage::Fragment:
-            return current.fs;
-        }
-
-        UNREACHABLE();
+    void UseProgrammableFragmentShader(GLuint program) {
+        fs = program;
     }
 
     void UseTrivialGeometryShader() {
-        current.gs = 0;
+        gs = 0;
     }
 
     void ApplyTo(OpenGLState& state) {
@@ -140,35 +53,16 @@ public:
                            GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT | GL_FRAGMENT_SHADER_BIT,
                            0);
 
-        glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, current.vs);
-        glUseProgramStages(pipeline.handle, GL_GEOMETRY_SHADER_BIT, current.gs);
-        glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, current.fs);
+        glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, vs);
+        glUseProgramStages(pipeline.handle, GL_GEOMETRY_SHADER_BIT, gs);
+        glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fs);
         state.draw.shader_program = 0;
         state.draw.program_pipeline = pipeline.handle;
     }
 
 private:
-    struct ShaderTuple {
-        GLuint vs = 0, gs = 0, fs = 0;
-        bool operator==(const ShaderTuple& rhs) const {
-            return std::tie(vs, gs, fs) == std::tie(rhs.vs, rhs.gs, rhs.fs);
-        }
-        struct Hash {
-            std::size_t operator()(const ShaderTuple& tuple) const {
-                std::size_t hash = 0;
-                boost::hash_combine(hash, tuple.vs);
-                boost::hash_combine(hash, tuple.gs);
-                boost::hash_combine(hash, tuple.fs);
-                return hash;
-            }
-        };
-    };
-    ShaderTuple current;
-    VertexShaders vertex_shaders;
-    FragmentShaders fragment_shaders;
-
-    std::unordered_map<ShaderTuple, OGLProgram, ShaderTuple::Hash> program_cache;
     OGLPipeline pipeline;
+    GLuint vs{}, fs{}, gs{};
 };
 
 } // namespace OpenGL::GLShader

src/video_core/renderer_opengl/gl_state.cpp

@@ -27,13 +27,17 @@ OpenGLState::OpenGLState() {
     color_mask.alpha_enabled = GL_TRUE;
 
     stencil.test_enabled = false;
-    stencil.test_func = GL_ALWAYS;
-    stencil.test_ref = 0;
-    stencil.test_mask = 0xFF;
-    stencil.write_mask = 0xFF;
-    stencil.action_depth_fail = GL_KEEP;
-    stencil.action_depth_pass = GL_KEEP;
-    stencil.action_stencil_fail = GL_KEEP;
+    auto reset_stencil = [](auto& config) {
+        config.test_func = GL_ALWAYS;
+        config.test_ref = 0;
+        config.test_mask = 0xFFFFFFFF;
+        config.write_mask = 0xFFFFFFFF;
+        config.action_depth_fail = GL_KEEP;
+        config.action_depth_pass = GL_KEEP;
+        config.action_stencil_fail = GL_KEEP;
+    };
+    reset_stencil(stencil.front);
+    reset_stencil(stencil.back);
 
     blend.enabled = true;
     blend.rgb_equation = GL_FUNC_ADD;
@@ -129,24 +133,23 @@ void OpenGLState::Apply() const {
             glDisable(GL_STENCIL_TEST);
         }
     }
-
-    if (stencil.test_func != cur_state.stencil.test_func ||
-        stencil.test_ref != cur_state.stencil.test_ref ||
-        stencil.test_mask != cur_state.stencil.test_mask) {
-        glStencilFunc(stencil.test_func, stencil.test_ref, stencil.test_mask);
-    }
-
-    if (stencil.action_depth_fail != cur_state.stencil.action_depth_fail ||
-        stencil.action_depth_pass != cur_state.stencil.action_depth_pass ||
-        stencil.action_stencil_fail != cur_state.stencil.action_stencil_fail) {
-        glStencilOp(stencil.action_stencil_fail, stencil.action_depth_fail,
-                    stencil.action_depth_pass);
-    }
-
-    // Stencil mask
-    if (stencil.write_mask != cur_state.stencil.write_mask) {
-        glStencilMask(stencil.write_mask);
-    }
+    auto config_stencil = [](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) {
+            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) {
+            glStencilOpSeparate(face, config.action_stencil_fail, config.action_depth_fail,
+                                config.action_depth_pass);
+        }
+        if (config.write_mask != prev_config.write_mask) {
+            glStencilMaskSeparate(face, config.write_mask);
+        }
+    };
+    config_stencil(GL_FRONT, stencil.front, cur_state.stencil.front);
+    config_stencil(GL_BACK, stencil.back, cur_state.stencil.back);
 
     // Blending
     if (blend.enabled != cur_state.blend.enabled) {

src/video_core/renderer_opengl/gl_state.h

@@ -58,14 +58,16 @@ public:
     } color_mask; // GL_COLOR_WRITEMASK
 
     struct {
-        bool test_enabled;          // GL_STENCIL_TEST
-        GLenum test_func;           // GL_STENCIL_FUNC
-        GLint test_ref;             // GL_STENCIL_REF
-        GLuint test_mask;           // GL_STENCIL_VALUE_MASK
-        GLuint write_mask;          // GL_STENCIL_WRITEMASK
-        GLenum action_stencil_fail; // GL_STENCIL_FAIL
-        GLenum action_depth_fail;   // GL_STENCIL_PASS_DEPTH_FAIL
-        GLenum action_depth_pass;   // GL_STENCIL_PASS_DEPTH_PASS
+        bool test_enabled; // GL_STENCIL_TEST
+        struct {
+            GLenum test_func;           // GL_STENCIL_FUNC
+            GLint test_ref;             // GL_STENCIL_REF
+            GLuint test_mask;           // GL_STENCIL_VALUE_MASK
+            GLuint write_mask;          // GL_STENCIL_WRITEMASK
+            GLenum action_stencil_fail; // GL_STENCIL_FAIL
+            GLenum action_depth_fail;   // GL_STENCIL_PASS_DEPTH_FAIL
+            GLenum action_depth_pass;   // GL_STENCIL_PASS_DEPTH_PASS
+        } front, back;
     } stencil;
 
     struct {

src/video_core/renderer_opengl/maxwell_to_gl.h

@@ -295,6 +295,30 @@ inline GLenum ComparisonOp(Maxwell::ComparisonOp comparison) {
     return {};
 }
 
+inline GLenum StencilOp(Maxwell::StencilOp stencil) {
+    switch (stencil) {
+    case Maxwell::StencilOp::Keep:
+        return GL_KEEP;
+    case Maxwell::StencilOp::Zero:
+        return GL_ZERO;
+    case Maxwell::StencilOp::Replace:
+        return GL_REPLACE;
+    case Maxwell::StencilOp::Incr:
+        return GL_INCR;
+    case Maxwell::StencilOp::Decr:
+        return GL_DECR;
+    case Maxwell::StencilOp::Invert:
+        return GL_INVERT;
+    case Maxwell::StencilOp::IncrWrap:
+        return GL_INCR_WRAP;
+    case Maxwell::StencilOp::DecrWrap:
+        return GL_DECR_WRAP;
+    }
+    LOG_CRITICAL(Render_OpenGL, "Unimplemented stencil op={}", static_cast<u32>(stencil));
+    UNREACHABLE();
+    return {};
+}
+
 inline GLenum FrontFace(Maxwell::Cull::FrontFace front_face) {
     switch (front_face) {
     case Maxwell::Cull::FrontFace::ClockWise:

src/yuzu/configuration/config.cpp

@@ -125,9 +125,9 @@ void Config::ReadValues() {
 
     qt_config->beginGroup("UIGameList");
     UISettings::values.show_unknown = qt_config->value("show_unknown", true).toBool();
-    UISettings::values.icon_size = qt_config->value("icon_size", 48).toUInt();
-    UISettings::values.row_1_text_id = qt_config->value("row_1_text_id", 0).toUInt();
-    UISettings::values.row_2_text_id = qt_config->value("row_2_text_id", 3).toUInt();
+    UISettings::values.icon_size = qt_config->value("icon_size", 64).toUInt();
+    UISettings::values.row_1_text_id = qt_config->value("row_1_text_id", 3).toUInt();
+    UISettings::values.row_2_text_id = qt_config->value("row_2_text_id", 2).toUInt();
     qt_config->endGroup();
 
     qt_config->beginGroup("UILayout");

src/yuzu/configuration/configure_input.ui

@@ -29,24 +29,6 @@
        </property>
        <layout class="QGridLayout" name="gridLayout_6">
         <item row="0" column="0">
-         <layout class="QVBoxLayout" name="buttonMiscPlusVerticalLayout">
-          <item>
-           <widget class="QLabel" name="labelPlus">
-            <property name="text">
-             <string>Plus:</string>
-            </property>
-           </widget>
-          </item>
-          <item>
-           <widget class="QPushButton" name="buttonPlus">
-            <property name="text">
-             <string/>
-            </property>
-           </widget>
-          </item>
-         </layout>
-        </item>
-        <item row="0" column="1">
          <layout class="QVBoxLayout" name="buttonMiscMinusVerticalLayout">
           <item>
            <widget class="QLabel" name="labelMinus">
@@ -64,6 +46,24 @@
           </item>
          </layout>
         </item>
+        <item row="0" column="1">
+         <layout class="QVBoxLayout" name="buttonMiscPlusVerticalLayout">
+          <item>
+           <widget class="QLabel" name="labelPlus">
+            <property name="text">
+             <string>Plus:</string>
+            </property>
+           </widget>
+          </item>
+          <item>
+           <widget class="QPushButton" name="buttonPlus">
+            <property name="text">
+             <string/>
+            </property>
+           </widget>
+          </item>
+         </layout>
+        </item>
         <item row="1" column="0">
          <layout class="QVBoxLayout" name="buttonMiscHomeVerticalLayout">
           <item>
@@ -472,7 +472,7 @@ Capture:</string>
           </property>
          </widget>
         </item>
-        <item row="1" column="0">
+        <item row="0" column="0">
          <layout class="QVBoxLayout" name="buttonRStickLeftVerticalLayout">
           <item>
            <widget class="QLabel" name="labelRStickLeft">
@@ -490,7 +490,7 @@ Capture:</string>
           </item>
          </layout>
         </item>
-        <item row="0" column="0">
+        <item row="1" column="0">
          <layout class="QVBoxLayout" name="buttonRStickUpVerticalLayout">
           <item>
            <widget class="QLabel" name="labelRStickUp">

src/yuzu/debugger/graphics/graphics_breakpoints.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <map>
 #include <QLabel>
 #include <QMetaType>
 #include <QPushButton>

src/yuzu/debugger/graphics/graphics_surface.cpp

@@ -11,12 +11,13 @@
 #include <QPushButton>
 #include <QScrollArea>
 #include <QSpinBox>
+#include "common/vector_math.h"
 #include "core/core.h"
+#include "core/memory.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
 #include "video_core/textures/decoders.h"
 #include "video_core/textures/texture.h"
-#include "video_core/utils.h"
 #include "yuzu/debugger/graphics/graphics_surface.h"
 #include "yuzu/util/spinbox.h"
 
@@ -381,7 +382,7 @@ void GraphicsSurfaceWidget::OnUpdate() {
     // TODO: Implement a good way to visualize alpha components!
 
     QImage decoded_image(surface_width, surface_height, QImage::Format_ARGB32);
-    boost::optional<VAddr> address = gpu.memory_manager->GpuToCpuAddress(surface_address);
+    boost::optional<VAddr> address = gpu.MemoryManager().GpuToCpuAddress(surface_address);
 
     // TODO(bunnei): Will not work with BCn formats that swizzle 4x4 tiles.
     // Needs to be fixed if we plan to use this feature more, otherwise we may remove it.
@@ -442,7 +443,7 @@ void GraphicsSurfaceWidget::SaveSurface() {
             pixmap->save(&file, "PNG");
     } else if (selectedFilter == bin_filter) {
         auto& gpu = Core::System::GetInstance().GPU();
-        boost::optional<VAddr> address = gpu.memory_manager->GpuToCpuAddress(surface_address);
+        boost::optional<VAddr> address = gpu.MemoryManager().GpuToCpuAddress(surface_address);
 
         const u8* buffer = Memory::GetPointer(*address);
         ASSERT_MSG(buffer != nullptr, "Memory not accessible");

src/yuzu/game_list.cpp

@@ -426,13 +426,12 @@ static void GetMetadataFromControlNCA(const std::shared_ptr<FileSys::NCA>& nca,
     }
 }
 
-void GameListWorker::AddInstalledTitlesToGameList() {
-    const auto usernand = Service::FileSystem::GetUserNANDContents();
-    const auto installed_games = usernand->ListEntriesFilter(FileSys::TitleType::Application,
-                                                             FileSys::ContentRecordType::Program);
+void GameListWorker::AddInstalledTitlesToGameList(std::shared_ptr<FileSys::RegisteredCache> cache) {
+    const auto installed_games = cache->ListEntriesFilter(FileSys::TitleType::Application,
+                                                          FileSys::ContentRecordType::Program);
 
     for (const auto& game : installed_games) {
-        const auto& file = usernand->GetEntryRaw(game);
+        const auto& file = cache->GetEntryUnparsed(game);
         std::unique_ptr<Loader::AppLoader> loader = Loader::GetLoader(file);
         if (!loader)
             continue;
@@ -442,8 +441,7 @@ void GameListWorker::AddInstalledTitlesToGameList() {
         u64 program_id = 0;
         loader->ReadProgramId(program_id);
 
-        const auto& control =
-            usernand->GetEntry(game.title_id, FileSys::ContentRecordType::Control);
+        const auto& control = cache->GetEntry(game.title_id, FileSys::ContentRecordType::Control);
         if (control != nullptr)
             GetMetadataFromControlNCA(control, icon, name);
         emit EntryReady({
@@ -457,11 +455,11 @@ void GameListWorker::AddInstalledTitlesToGameList() {
         });
     }
 
-    const auto control_data = usernand->ListEntriesFilter(FileSys::TitleType::Application,
-                                                          FileSys::ContentRecordType::Control);
+    const auto control_data = cache->ListEntriesFilter(FileSys::TitleType::Application,
+                                                       FileSys::ContentRecordType::Control);
 
     for (const auto& entry : control_data) {
-        const auto nca = usernand->GetEntry(entry);
+        const auto nca = cache->GetEntry(entry);
         if (nca != nullptr)
             nca_control_map.insert_or_assign(entry.title_id, nca);
     }
@@ -549,7 +547,9 @@ void GameListWorker::run() {
     stop_processing = false;
     watch_list.append(dir_path);
     FillControlMap(dir_path.toStdString());
-    AddInstalledTitlesToGameList();
+    AddInstalledTitlesToGameList(Service::FileSystem::GetUserNANDContents());
+    AddInstalledTitlesToGameList(Service::FileSystem::GetSystemNANDContents());
+    AddInstalledTitlesToGameList(Service::FileSystem::GetSDMCContents());
     AddFstEntriesToGameList(dir_path.toStdString(), deep_scan ? 256 : 0);
     nca_control_map.clear();
     emit Finished(watch_list);

src/yuzu/game_list_p.h

@@ -172,7 +172,7 @@ private:
     bool deep_scan;
     std::atomic_bool stop_processing;
 
-    void AddInstalledTitlesToGameList();
+    void AddInstalledTitlesToGameList(std::shared_ptr<FileSys::RegisteredCache> cache);
     void FillControlMap(const std::string& dir_path);
     void AddFstEntriesToGameList(const std::string& dir_path, unsigned int recursion = 0);
 };

src/yuzu/main.cpp

@@ -91,9 +91,20 @@ void GMainWindow::ShowCallouts() {}
 
 const int GMainWindow::max_recent_files_item;
 
+static void InitializeLogging() {
+    Log::Filter log_filter;
+    log_filter.ParseFilterString(Settings::values.log_filter);
+    Log::SetGlobalFilter(log_filter);
+
+    const std::string& log_dir = FileUtil::GetUserPath(FileUtil::UserPath::LogDir);
+    FileUtil::CreateFullPath(log_dir);
+    Log::AddBackend(std::make_unique<Log::FileBackend>(log_dir + LOG_FILE));
+}
+
 GMainWindow::GMainWindow()
     : config(new Config()), emu_thread(nullptr),
       vfs(std::make_shared<FileSys::RealVfsFilesystem>()) {
+    InitializeLogging();
 
     debug_context = Tegra::DebugContext::Construct();
 
@@ -122,8 +133,7 @@ GMainWindow::GMainWindow()
     show();
 
     // Necessary to load titles from nand in gamelist.
-    Service::FileSystem::RegisterBIS(std::make_unique<FileSys::BISFactory>(vfs->OpenDirectory(
-        FileUtil::GetUserPath(FileUtil::UserPath::NANDDir), FileSys::Mode::ReadWrite)));
+    Service::FileSystem::CreateFactories(vfs);
     game_list->PopulateAsync(UISettings::values.gamedir, UISettings::values.gamedir_deepscan);
 
     // Show one-time "callout" messages to the user
@@ -545,6 +555,15 @@ void GMainWindow::BootGame(const QString& filename) {
     }
     status_bar_update_timer.start(2000);
 
+    std::string title_name;
+    const auto res = Core::System::GetInstance().GetGameName(title_name);
+    if (res != Loader::ResultStatus::Success)
+        title_name = FileUtil::GetFilename(filename.toStdString());
+
+    setWindowTitle(QString("yuzu %1| %4 | %2-%3")
+                       .arg(Common::g_build_name, Common::g_scm_branch, Common::g_scm_desc,
+                            QString::fromStdString(title_name)));
+
     render_window->show();
     render_window->setFocus();
 
@@ -576,6 +595,8 @@ void GMainWindow::ShutdownGame() {
     render_window->hide();
     game_list->show();
     game_list->setFilterFocus();
+    setWindowTitle(QString("yuzu %1| %2-%3")
+                       .arg(Common::g_build_name, Common::g_scm_branch, Common::g_scm_desc));
 
     // Disable status bar updates
     status_bar_update_timer.stop();
@@ -872,6 +893,7 @@ void GMainWindow::OnStartGame() {
 
     ui.action_Pause->setEnabled(true);
     ui.action_Stop->setEnabled(true);
+    ui.action_Restart->setEnabled(true);
 }
 
 void GMainWindow::OnPauseGame() {
@@ -880,7 +902,6 @@ void GMainWindow::OnPauseGame() {
     ui.action_Start->setEnabled(true);
     ui.action_Pause->setEnabled(false);
     ui.action_Stop->setEnabled(true);
-    ui.action_Restart->setEnabled(true);
 }
 
 void GMainWindow::OnStopGame() {
@@ -1173,16 +1194,6 @@ void GMainWindow::UpdateUITheme() {
 #undef main
 #endif
 
-static void InitializeLogging() {
-    Log::Filter log_filter;
-    log_filter.ParseFilterString(Settings::values.log_filter);
-    Log::SetGlobalFilter(log_filter);
-
-    const std::string& log_dir = FileUtil::GetUserPath(FileUtil::UserPath::LogDir);
-    FileUtil::CreateFullPath(log_dir);
-    Log::AddBackend(std::make_unique<Log::FileBackend>(log_dir + LOG_FILE));
-}
-
 int main(int argc, char* argv[]) {
     MicroProfileOnThreadCreate("Frontend");
     SCOPE_EXIT({ MicroProfileShutdown(); });
@@ -1200,7 +1211,6 @@ int main(int argc, char* argv[]) {
 
     GMainWindow main_window;
     // After settings have been loaded by GMainWindow, apply the filter
-    InitializeLogging();
     main_window.show();
     return app.exec();
 }

src/yuzu/main.ui

@@ -74,6 +74,7 @@
     <addaction name="action_Start"/>
     <addaction name="action_Pause"/>
     <addaction name="action_Stop"/>
+    <addaction name="action_Restart"/>
     <addaction name="separator"/>
     <addaction name="action_Configure"/>
    </widget>