gl_shader_decompiler: Implement SEL instruction.

Improvements to rasterizer cache

src/common/CMakeLists.txt

@@ -63,7 +63,6 @@ add_library(common STATIC
     string_util.cpp
     string_util.h
     swap.h
-    synchronized_wrapper.h
     telemetry.cpp
     telemetry.h
     thread.cpp

src/common/common_paths.h

@@ -26,7 +26,7 @@
 #define USA_DIR "USA"
 #define JAP_DIR "JAP"
 
-// Subdirs in the User dir returned by GetUserPath(D_USER_IDX)
+// Subdirs in the User dir returned by GetUserPath(UserPath::UserDir)
 #define CONFIG_DIR "config"
 #define CACHE_DIR "cache"
 #define SDMC_DIR "sdmc"
@@ -35,11 +35,11 @@
 #define LOG_DIR "log"
 
 // Filenames
-// Files in the directory returned by GetUserPath(D_CONFIG_IDX)
+// Files in the directory returned by GetUserPath(UserPath::ConfigDir)
 #define EMU_CONFIG "emu.ini"
 #define DEBUGGER_CONFIG "debugger.ini"
 #define LOGGER_CONFIG "logger.ini"
-// Files in the directory returned by GetUserPath(D_LOGS_IDX)
+// Files in the directory returned by GetUserPath(UserPath::LogDir)
 #define LOG_FILE "yuzu_log.txt"
 
 // Sys files

src/common/file_util.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <sstream>
+#include <unordered_map>
 #include "common/assert.h"
 #include "common/common_funcs.h"
 #include "common/common_paths.h"
@@ -681,67 +682,68 @@ std::string GetSysDirectory() {
 
 // Returns a string with a yuzu data dir or file in the user's home
 // directory. To be used in "multi-user" mode (that is, installed).
-const std::string& GetUserPath(const unsigned int DirIDX, const std::string& newPath) {
-    static std::string paths[NUM_PATH_INDICES];
+const std::string& GetUserPath(UserPath path, const std::string& new_path) {
+    static std::unordered_map<UserPath, std::string> paths;
+    auto& user_path = paths[UserPath::UserDir];
 
     // Set up all paths and files on the first run
-    if (paths[D_USER_IDX].empty()) {
+    if (user_path.empty()) {
 #ifdef _WIN32
-        paths[D_USER_IDX] = GetExeDirectory() + DIR_SEP USERDATA_DIR DIR_SEP;
-        if (!FileUtil::IsDirectory(paths[D_USER_IDX])) {
-            paths[D_USER_IDX] = AppDataRoamingDirectory() + DIR_SEP EMU_DATA_DIR DIR_SEP;
+        user_path = GetExeDirectory() + DIR_SEP USERDATA_DIR DIR_SEP;
+        if (!FileUtil::IsDirectory(user_path)) {
+            user_path = AppDataRoamingDirectory() + DIR_SEP EMU_DATA_DIR DIR_SEP;
         } else {
             LOG_INFO(Common_Filesystem, "Using the local user directory");
         }
 
-        paths[D_CONFIG_IDX] = paths[D_USER_IDX] + CONFIG_DIR DIR_SEP;
-        paths[D_CACHE_IDX] = paths[D_USER_IDX] + CACHE_DIR DIR_SEP;
+        paths.emplace(UserPath::ConfigDir, user_path + CONFIG_DIR DIR_SEP);
+        paths.emplace(UserPath::CacheDir, user_path + CACHE_DIR DIR_SEP);
 #else
         if (FileUtil::Exists(ROOT_DIR DIR_SEP USERDATA_DIR)) {
-            paths[D_USER_IDX] = ROOT_DIR DIR_SEP USERDATA_DIR DIR_SEP;
-            paths[D_CONFIG_IDX] = paths[D_USER_IDX] + CONFIG_DIR DIR_SEP;
-            paths[D_CACHE_IDX] = paths[D_USER_IDX] + CACHE_DIR DIR_SEP;
+            user_path = ROOT_DIR DIR_SEP USERDATA_DIR DIR_SEP;
+            paths.emplace(UserPath::ConfigDir, user_path + CONFIG_DIR DIR_SEP);
+            paths.emplace(UserPath::CacheDir, user_path + CACHE_DIR DIR_SEP);
         } else {
             std::string data_dir = GetUserDirectory("XDG_DATA_HOME");
             std::string config_dir = GetUserDirectory("XDG_CONFIG_HOME");
             std::string cache_dir = GetUserDirectory("XDG_CACHE_HOME");
 
-            paths[D_USER_IDX] = data_dir + DIR_SEP EMU_DATA_DIR DIR_SEP;
-            paths[D_CONFIG_IDX] = config_dir + DIR_SEP EMU_DATA_DIR DIR_SEP;
-            paths[D_CACHE_IDX] = cache_dir + DIR_SEP EMU_DATA_DIR DIR_SEP;
+            user_path = data_dir + DIR_SEP EMU_DATA_DIR DIR_SEP;
+            paths.emplace(UserPath::ConfigDir, config_dir + DIR_SEP EMU_DATA_DIR DIR_SEP);
+            paths.emplace(UserPath::CacheDir, cache_dir + DIR_SEP EMU_DATA_DIR DIR_SEP);
         }
 #endif
-        paths[D_SDMC_IDX] = paths[D_USER_IDX] + SDMC_DIR DIR_SEP;
-        paths[D_NAND_IDX] = paths[D_USER_IDX] + NAND_DIR DIR_SEP;
-        paths[D_SYSDATA_IDX] = paths[D_USER_IDX] + SYSDATA_DIR DIR_SEP;
+        paths.emplace(UserPath::SDMCDir, user_path + SDMC_DIR DIR_SEP);
+        paths.emplace(UserPath::NANDDir, user_path + NAND_DIR DIR_SEP);
+        paths.emplace(UserPath::SysDataDir, user_path + SYSDATA_DIR DIR_SEP);
         // TODO: Put the logs in a better location for each OS
-        paths[D_LOGS_IDX] = paths[D_USER_IDX] + LOG_DIR DIR_SEP;
+        paths.emplace(UserPath::LogDir, user_path + LOG_DIR DIR_SEP);
     }
 
-    if (!newPath.empty()) {
-        if (!FileUtil::IsDirectory(newPath)) {
-            LOG_ERROR(Common_Filesystem, "Invalid path specified {}", newPath);
-            return paths[DirIDX];
+    if (!new_path.empty()) {
+        if (!FileUtil::IsDirectory(new_path)) {
+            LOG_ERROR(Common_Filesystem, "Invalid path specified {}", new_path);
+            return paths[path];
         } else {
-            paths[DirIDX] = newPath;
+            paths[path] = new_path;
         }
 
-        switch (DirIDX) {
-        case D_ROOT_IDX:
-            paths[D_USER_IDX] = paths[D_ROOT_IDX] + DIR_SEP;
+        switch (path) {
+        case UserPath::RootDir:
+            user_path = paths[UserPath::RootDir] + DIR_SEP;
             break;
 
-        case D_USER_IDX:
-            paths[D_USER_IDX] = paths[D_ROOT_IDX] + DIR_SEP;
-            paths[D_CONFIG_IDX] = paths[D_USER_IDX] + CONFIG_DIR DIR_SEP;
-            paths[D_CACHE_IDX] = paths[D_USER_IDX] + CACHE_DIR DIR_SEP;
-            paths[D_SDMC_IDX] = paths[D_USER_IDX] + SDMC_DIR DIR_SEP;
-            paths[D_NAND_IDX] = paths[D_USER_IDX] + NAND_DIR DIR_SEP;
+        case UserPath::UserDir:
+            user_path = paths[UserPath::RootDir] + DIR_SEP;
+            paths[UserPath::ConfigDir] = user_path + CONFIG_DIR DIR_SEP;
+            paths[UserPath::CacheDir] = user_path + CACHE_DIR DIR_SEP;
+            paths[UserPath::SDMCDir] = user_path + SDMC_DIR DIR_SEP;
+            paths[UserPath::NANDDir] = user_path + NAND_DIR DIR_SEP;
             break;
         }
     }
 
-    return paths[DirIDX];
+    return paths[path];
 }
 
 size_t WriteStringToFile(bool text_file, const std::string& str, const char* filename) {
@@ -836,8 +838,7 @@ std::string GetPathWithoutTop(std::string path) {
     }
     const auto name_bck_index = path.find_first_of('\\');
     const auto name_fwd_index = path.find_first_of('/');
-    return path.substr(std::min<size_t>(name_bck_index, name_fwd_index) + 1);
-    return path.substr(std::min<size_t>(name_bck_index, name_fwd_index) + 1);
+    return path.substr(std::min(name_bck_index, name_fwd_index) + 1);
 }
 
 std::string GetFilename(std::string path) {

src/common/file_util.h

@@ -16,21 +16,20 @@
 #include "common/string_util.h"
 #endif
 
-// User directory indices for GetUserPath
-enum {
-    D_USER_IDX,
-    D_ROOT_IDX,
-    D_CONFIG_IDX,
-    D_CACHE_IDX,
-    D_SDMC_IDX,
-    D_NAND_IDX,
-    D_SYSDATA_IDX,
-    D_LOGS_IDX,
-    NUM_PATH_INDICES
-};
-
 namespace FileUtil {
 
+// User paths for GetUserPath
+enum class UserPath {
+    CacheDir,
+    ConfigDir,
+    LogDir,
+    NANDDir,
+    RootDir,
+    SDMCDir,
+    SysDataDir,
+    UserDir,
+};
+
 // FileSystem tree node/
 struct FSTEntry {
     bool isDirectory;
@@ -123,7 +122,7 @@ bool SetCurrentDir(const std::string& directory);
 
 // Returns a pointer to a string with a yuzu data dir in the user's home
 // directory. To be used in "multi-user" mode (that is, installed).
-const std::string& GetUserPath(const unsigned int DirIDX, const std::string& newPath = "");
+const std::string& GetUserPath(UserPath path, const std::string& new_path = "");
 
 // Returns the path to where the sys file are
 std::string GetSysDirectory();

src/common/logging/backend.cpp

@@ -3,19 +3,20 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
-#include <array>
+#include <atomic>
 #include <chrono>
 #include <climits>
 #include <condition_variable>
 #include <memory>
+#include <mutex>
 #include <thread>
+#include <vector>
 #ifdef _WIN32
 #include <share.h> // For _SH_DENYWR
 #else
 #define _SH_DENYWR 0
 #endif
 #include "common/assert.h"
-#include "common/common_funcs.h" // snprintf compatibility define
 #include "common/logging/backend.h"
 #include "common/logging/log.h"
 #include "common/logging/text_formatter.h"
@@ -48,11 +49,11 @@ public:
         backends.push_back(std::move(backend));
     }
 
-    void RemoveBackend(const std::string& backend_name) {
+    void RemoveBackend(std::string_view backend_name) {
         std::lock_guard<std::mutex> lock(writing_mutex);
-        auto it = std::remove_if(backends.begin(), backends.end(), [&backend_name](const auto& i) {
-            return !strcmp(i->GetName(), backend_name.c_str());
-        });
+        const auto it =
+            std::remove_if(backends.begin(), backends.end(),
+                           [&backend_name](const auto& i) { return backend_name == i->GetName(); });
         backends.erase(it, backends.end());
     }
 
@@ -64,10 +65,10 @@ public:
         filter = f;
     }
 
-    Backend* GetBackend(const std::string& backend_name) {
-        auto it = std::find_if(backends.begin(), backends.end(), [&backend_name](const auto& i) {
-            return !strcmp(i->GetName(), backend_name.c_str());
-        });
+    Backend* GetBackend(std::string_view backend_name) {
+        const auto it =
+            std::find_if(backends.begin(), backends.end(),
+                         [&backend_name](const auto& i) { return backend_name == i->GetName(); });
         if (it == backends.end())
             return nullptr;
         return it->get();
@@ -265,11 +266,11 @@ void AddBackend(std::unique_ptr<Backend> backend) {
     Impl::Instance().AddBackend(std::move(backend));
 }
 
-void RemoveBackend(const std::string& backend_name) {
+void RemoveBackend(std::string_view backend_name) {
     Impl::Instance().RemoveBackend(backend_name);
 }
 
-Backend* GetBackend(const std::string& backend_name) {
+Backend* GetBackend(std::string_view backend_name) {
     return Impl::Instance().GetBackend(backend_name);
 }
 

src/common/logging/backend.h

@@ -4,10 +4,9 @@
 #pragma once
 
 #include <chrono>
-#include <cstdarg>
 #include <memory>
 #include <string>
-#include <utility>
+#include <string_view>
 #include "common/file_util.h"
 #include "common/logging/filter.h"
 #include "common/logging/log.h"
@@ -106,9 +105,9 @@ private:
 
 void AddBackend(std::unique_ptr<Backend> backend);
 
-void RemoveBackend(const std::string& backend_name);
+void RemoveBackend(std::string_view backend_name);
 
-Backend* GetBackend(const std::string& backend_name);
+Backend* GetBackend(std::string_view backend_name);
 
 /**
  * Returns the name of the passed log class as a C-string. Subclasses are separated by periods

src/common/logging/filter.cpp

@@ -8,39 +8,9 @@
 #include "common/string_util.h"
 
 namespace Log {
-
-Filter::Filter(Level default_level) {
-    ResetAll(default_level);
-}
-
-void Filter::ResetAll(Level level) {
-    class_levels.fill(level);
-}
-
-void Filter::SetClassLevel(Class log_class, Level level) {
-    class_levels[static_cast<size_t>(log_class)] = level;
-}
-
-void Filter::ParseFilterString(const std::string& filter_str) {
-    auto clause_begin = filter_str.cbegin();
-    while (clause_begin != filter_str.cend()) {
-        auto clause_end = std::find(clause_begin, filter_str.cend(), ' ');
-
-        // If clause isn't empty
-        if (clause_end != clause_begin) {
-            ParseFilterRule(clause_begin, clause_end);
-        }
-
-        if (clause_end != filter_str.cend()) {
-            // Skip over the whitespace
-            ++clause_end;
-        }
-        clause_begin = clause_end;
-    }
-}
-
+namespace {
 template <typename It>
-static Level GetLevelByName(const It begin, const It end) {
+Level GetLevelByName(const It begin, const It end) {
     for (u8 i = 0; i < static_cast<u8>(Level::Count); ++i) {
         const char* level_name = GetLevelName(static_cast<Level>(i));
         if (Common::ComparePartialString(begin, end, level_name)) {
@@ -51,7 +21,7 @@ static Level GetLevelByName(const It begin, const It end) {
 }
 
 template <typename It>
-static Class GetClassByName(const It begin, const It end) {
+Class GetClassByName(const It begin, const It end) {
     for (ClassType i = 0; i < static_cast<ClassType>(Class::Count); ++i) {
         const char* level_name = GetLogClassName(static_cast<Class>(i));
         if (Common::ComparePartialString(begin, end, level_name)) {
@@ -61,8 +31,8 @@ static Class GetClassByName(const It begin, const It end) {
     return Class::Count;
 }
 
-bool Filter::ParseFilterRule(const std::string::const_iterator begin,
-                             const std::string::const_iterator end) {
+template <typename Iterator>
+bool ParseFilterRule(Filter& instance, Iterator begin, Iterator end) {
     auto level_separator = std::find(begin, end, ':');
     if (level_separator == end) {
         LOG_ERROR(Log, "Invalid log filter. Must specify a log level after `:`: {}",
@@ -77,7 +47,7 @@ bool Filter::ParseFilterRule(const std::string::const_iterator begin,
     }
 
     if (Common::ComparePartialString(begin, level_separator, "*")) {
-        ResetAll(level);
+        instance.ResetAll(level);
         return true;
     }
 
@@ -87,9 +57,40 @@ bool Filter::ParseFilterRule(const std::string::const_iterator begin,
         return false;
     }
 
-    SetClassLevel(log_class, level);
+    instance.SetClassLevel(log_class, level);
     return true;
 }
+} // Anonymous namespace
+
+Filter::Filter(Level default_level) {
+    ResetAll(default_level);
+}
+
+void Filter::ResetAll(Level level) {
+    class_levels.fill(level);
+}
+
+void Filter::SetClassLevel(Class log_class, Level level) {
+    class_levels[static_cast<size_t>(log_class)] = level;
+}
+
+void Filter::ParseFilterString(std::string_view filter_view) {
+    auto clause_begin = filter_view.cbegin();
+    while (clause_begin != filter_view.cend()) {
+        auto clause_end = std::find(clause_begin, filter_view.cend(), ' ');
+
+        // If clause isn't empty
+        if (clause_end != clause_begin) {
+            ParseFilterRule(*this, clause_begin, clause_end);
+        }
+
+        if (clause_end != filter_view.cend()) {
+            // Skip over the whitespace
+            ++clause_end;
+        }
+        clause_begin = clause_end;
+    }
+}
 
 bool Filter::CheckMessage(Class log_class, Level level) const {
     return static_cast<u8>(level) >= static_cast<u8>(class_levels[static_cast<size_t>(log_class)]);

src/common/logging/filter.h

@@ -6,7 +6,7 @@
 
 #include <array>
 #include <cstddef>
-#include <string>
+#include <string_view>
 #include "common/logging/log.h"
 
 namespace Log {
@@ -40,9 +40,7 @@ public:
      *  - `Service:Info` -- Sets the level of Service to Info.
      *  - `Service.FS:Trace` -- Sets the level of the Service.FS class to Trace.
      */
-    void ParseFilterString(const std::string& filter_str);
-    bool ParseFilterRule(const std::string::const_iterator start,
-                         const std::string::const_iterator end);
+    void ParseFilterString(std::string_view filter_view);
 
     /// Matches class/level combination against the filter, returning true if it passed.
     bool CheckMessage(Class log_class, Level level) const;

src/common/param_package.cpp

@@ -3,7 +3,9 @@
 // Refer to the license.txt file included.
 
 #include <array>
+#include <utility>
 #include <vector>
+
 #include "common/logging/log.h"
 #include "common/param_package.h"
 #include "common/string_util.h"
@@ -12,10 +14,11 @@ namespace Common {
 
 constexpr char KEY_VALUE_SEPARATOR = ':';
 constexpr char PARAM_SEPARATOR = ',';
+
 constexpr char ESCAPE_CHARACTER = '$';
-const std::string KEY_VALUE_SEPARATOR_ESCAPE{ESCAPE_CHARACTER, '0'};
-const std::string PARAM_SEPARATOR_ESCAPE{ESCAPE_CHARACTER, '1'};
-const std::string ESCAPE_CHARACTER_ESCAPE{ESCAPE_CHARACTER, '2'};
+constexpr char KEY_VALUE_SEPARATOR_ESCAPE[] = "$0";
+constexpr char PARAM_SEPARATOR_ESCAPE[] = "$1";
+constexpr char ESCAPE_CHARACTER_ESCAPE[] = "$2";
 
 ParamPackage::ParamPackage(const std::string& serialized) {
     std::vector<std::string> pairs;
@@ -35,7 +38,7 @@ ParamPackage::ParamPackage(const std::string& serialized) {
             part = Common::ReplaceAll(part, ESCAPE_CHARACTER_ESCAPE, {ESCAPE_CHARACTER});
         }
 
-        Set(key_value[0], key_value[1]);
+        Set(key_value[0], std::move(key_value[1]));
     }
 }
 
@@ -101,16 +104,16 @@ float ParamPackage::Get(const std::string& key, float default_value) const {
     }
 }
 
-void ParamPackage::Set(const std::string& key, const std::string& value) {
-    data[key] = value;
+void ParamPackage::Set(const std::string& key, std::string value) {
+    data.insert_or_assign(key, std::move(value));
 }
 
 void ParamPackage::Set(const std::string& key, int value) {
-    data[key] = std::to_string(value);
+    data.insert_or_assign(key, std::to_string(value));
 }
 
 void ParamPackage::Set(const std::string& key, float value) {
-    data[key] = std::to_string(value);
+    data.insert_or_assign(key, std::to_string(value));
 }
 
 bool ParamPackage::Has(const std::string& key) const {

src/common/param_package.h

@@ -28,7 +28,7 @@ public:
     std::string Get(const std::string& key, const std::string& default_value) const;
     int Get(const std::string& key, int default_value) const;
     float Get(const std::string& key, float default_value) const;
-    void Set(const std::string& key, const std::string& value);
+    void Set(const std::string& key, std::string value);
     void Set(const std::string& key, int value);
     void Set(const std::string& key, float value);
     bool Has(const std::string& key) const;

src/common/synchronized_wrapper.h

@@ -1,85 +0,0 @@
-// Copyright 2015 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <algorithm>
-#include <mutex>
-
-namespace Common {
-
-template <typename T>
-class SynchronizedWrapper;
-
-/**
- * Synchronized reference, that keeps a SynchronizedWrapper's mutex locked during its lifetime. This
- * greatly reduces the chance that someone will access the wrapped resource without locking the
- * mutex.
- */
-template <typename T>
-class SynchronizedRef {
-public:
-    SynchronizedRef(SynchronizedWrapper<T>& wrapper) : wrapper(&wrapper) {
-        wrapper.mutex.lock();
-    }
-
-    SynchronizedRef(SynchronizedRef&) = delete;
-    SynchronizedRef(SynchronizedRef&& o) : wrapper(o.wrapper) {
-        o.wrapper = nullptr;
-    }
-
-    ~SynchronizedRef() {
-        if (wrapper)
-            wrapper->mutex.unlock();
-    }
-
-    SynchronizedRef& operator=(SynchronizedRef&) = delete;
-    SynchronizedRef& operator=(SynchronizedRef&& o) {
-        std::swap(wrapper, o.wrapper);
-        return *this;
-    }
-
-    T& operator*() {
-        return wrapper->data;
-    }
-    const T& operator*() const {
-        return wrapper->data;
-    }
-
-    T* operator->() {
-        return &wrapper->data;
-    }
-    const T* operator->() const {
-        return &wrapper->data;
-    }
-
-private:
-    SynchronizedWrapper<T>* wrapper;
-};
-
-/**
- * Wraps an object, only allowing access to it via a locking reference wrapper. Good to ensure no
- * one forgets to lock a mutex before acessing an object. To access the wrapped object construct a
- * SyncronizedRef on this wrapper. Inspired by Rust's Mutex type
- * (http://doc.rust-lang.org/std/sync/struct.Mutex.html).
- */
-template <typename T>
-class SynchronizedWrapper {
-public:
-    template <typename... Args>
-    SynchronizedWrapper(Args&&... args) : data(std::forward<Args>(args)...) {}
-
-    SynchronizedRef<T> Lock() {
-        return {*this};
-    }
-
-private:
-    template <typename U>
-    friend class SynchronizedRef;
-
-    std::mutex mutex;
-    T data;
-};
-
-} // namespace Common

src/core/arm/arm_interface.h

@@ -20,9 +20,6 @@ public:
         u64 cpsr;
         std::array<u128, 32> fpu_registers;
         u64 fpscr;
-
-        // TODO(bunnei): Fix once we have proper support for tpidrro_el0, etc. in the JIT
-        VAddr tls_address;
     };
 
     /// Runs the CPU until an event happens
@@ -104,6 +101,10 @@ public:
 
     virtual void SetTlsAddress(VAddr address) = 0;
 
+    virtual u64 GetTPIDR_EL0() const = 0;
+
+    virtual void SetTPIDR_EL0(u64 value) = 0;
+
     /**
      * Saves the current CPU context
      * @param ctx Thread context to save

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -196,6 +196,14 @@ void ARM_Dynarmic::SetTlsAddress(u64 address) {
     cb->tpidrro_el0 = address;
 }
 
+u64 ARM_Dynarmic::GetTPIDR_EL0() const {
+    return cb->tpidr_el0;
+}
+
+void ARM_Dynarmic::SetTPIDR_EL0(u64 value) {
+    cb->tpidr_el0 = value;
+}
+
 void ARM_Dynarmic::SaveContext(ARM_Interface::ThreadContext& ctx) {
     ctx.cpu_registers = jit->GetRegisters();
     ctx.sp = jit->GetSP();
@@ -203,7 +211,6 @@ void ARM_Dynarmic::SaveContext(ARM_Interface::ThreadContext& ctx) {
     ctx.cpsr = jit->GetPstate();
     ctx.fpu_registers = jit->GetVectors();
     ctx.fpscr = jit->GetFpcr();
-    ctx.tls_address = cb->tpidrro_el0;
 }
 
 void ARM_Dynarmic::LoadContext(const ARM_Interface::ThreadContext& ctx) {
@@ -213,7 +220,6 @@ void ARM_Dynarmic::LoadContext(const ARM_Interface::ThreadContext& ctx) {
     jit->SetPstate(static_cast<u32>(ctx.cpsr));
     jit->SetVectors(ctx.fpu_registers);
     jit->SetFpcr(static_cast<u32>(ctx.fpscr));
-    cb->tpidrro_el0 = ctx.tls_address;
 }
 
 void ARM_Dynarmic::PrepareReschedule() {

src/core/arm/dynarmic/arm_dynarmic.h

@@ -34,6 +34,8 @@ public:
     void SetCPSR(u32 cpsr) override;
     VAddr GetTlsAddress() const override;
     void SetTlsAddress(VAddr address) override;
+    void SetTPIDR_EL0(u64 value) override;
+    u64 GetTPIDR_EL0() const override;
 
     void SaveContext(ThreadContext& ctx) override;
     void LoadContext(const ThreadContext& ctx) override;

src/core/arm/unicorn/arm_unicorn.cpp

@@ -169,6 +169,16 @@ void ARM_Unicorn::SetTlsAddress(VAddr base) {
     CHECKED(uc_reg_write(uc, UC_ARM64_REG_TPIDRRO_EL0, &base));
 }
 
+u64 ARM_Unicorn::GetTPIDR_EL0() const {
+    u64 value{};
+    CHECKED(uc_reg_read(uc, UC_ARM64_REG_TPIDR_EL0, &value));
+    return value;
+}
+
+void ARM_Unicorn::SetTPIDR_EL0(u64 value) {
+    CHECKED(uc_reg_write(uc, UC_ARM64_REG_TPIDR_EL0, &value));
+}
+
 void ARM_Unicorn::Run() {
     if (GDBStub::IsServerEnabled()) {
         ExecuteInstructions(std::max(4000000, 0));
@@ -220,8 +230,6 @@ void ARM_Unicorn::SaveContext(ARM_Interface::ThreadContext& ctx) {
 
     CHECKED(uc_reg_read_batch(uc, uregs, tregs, 31));
 
-    ctx.tls_address = GetTlsAddress();
-
     for (int i = 0; i < 32; ++i) {
         uregs[i] = UC_ARM64_REG_Q0 + i;
         tregs[i] = &ctx.fpu_registers[i];
@@ -249,8 +257,6 @@ void ARM_Unicorn::LoadContext(const ARM_Interface::ThreadContext& ctx) {
 
     CHECKED(uc_reg_write_batch(uc, uregs, tregs, 31));
 
-    SetTlsAddress(ctx.tls_address);
-
     for (auto i = 0; i < 32; ++i) {
         uregs[i] = UC_ARM64_REG_Q0 + i;
         tregs[i] = (void*)&ctx.fpu_registers[i];

src/core/arm/unicorn/arm_unicorn.h

@@ -28,6 +28,8 @@ public:
     void SetCPSR(u32 cpsr) override;
     VAddr GetTlsAddress() const override;
     void SetTlsAddress(VAddr address) override;
+    void SetTPIDR_EL0(u64 value) override;
+    u64 GetTPIDR_EL0() const override;
     void SaveContext(ThreadContext& ctx) override;
     void LoadContext(const ThreadContext& ctx) override;
     void PrepareReschedule() override;

src/core/file_sys/errors.h

@@ -20,13 +20,13 @@ enum {
 constexpr ResultCode ERROR_PATH_NOT_FOUND(ErrorModule::FS, ErrCodes::NotFound);
 
 // TODO(bunnei): Replace these with correct errors for Switch OS
-constexpr ResultCode ERROR_INVALID_PATH(ResultCode(-1));
-constexpr ResultCode ERROR_UNSUPPORTED_OPEN_FLAGS(ResultCode(-1));
-constexpr ResultCode ERROR_INVALID_OPEN_FLAGS(ResultCode(-1));
-constexpr ResultCode ERROR_FILE_NOT_FOUND(ResultCode(-1));
-constexpr ResultCode ERROR_UNEXPECTED_FILE_OR_DIRECTORY(ResultCode(-1));
-constexpr ResultCode ERROR_DIRECTORY_ALREADY_EXISTS(ResultCode(-1));
-constexpr ResultCode ERROR_FILE_ALREADY_EXISTS(ResultCode(-1));
-constexpr ResultCode ERROR_DIRECTORY_NOT_EMPTY(ResultCode(-1));
+constexpr ResultCode ERROR_INVALID_PATH(-1);
+constexpr ResultCode ERROR_UNSUPPORTED_OPEN_FLAGS(-1);
+constexpr ResultCode ERROR_INVALID_OPEN_FLAGS(-1);
+constexpr ResultCode ERROR_FILE_NOT_FOUND(-1);
+constexpr ResultCode ERROR_UNEXPECTED_FILE_OR_DIRECTORY(-1);
+constexpr ResultCode ERROR_DIRECTORY_ALREADY_EXISTS(-1);
+constexpr ResultCode ERROR_FILE_ALREADY_EXISTS(-1);
+constexpr ResultCode ERROR_DIRECTORY_NOT_EMPTY(-1);
 
 } // namespace FileSys

src/core/file_sys/partition_filesystem.cpp

@@ -2,7 +2,12 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
+#include <cstddef>
+#include <cstring>
+#include <iterator>
 #include <utility>
+
 #include "common/file_util.h"
 #include "common/logging/log.h"
 #include "core/file_sys/partition_filesystem.h"
@@ -99,14 +104,15 @@ void PartitionFilesystem::PrintDebugInfo() const {
 }
 
 bool PartitionFilesystem::ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) {
-    auto iter = std::find(pfs_files.begin(), pfs_files.end(), file);
+    const auto iter = std::find(pfs_files.begin(), pfs_files.end(), file);
     if (iter == pfs_files.end())
         return false;
 
-    pfs_files[iter - pfs_files.begin()] = pfs_files.back();
+    const std::ptrdiff_t offset = std::distance(pfs_files.begin(), iter);
+    pfs_files[offset] = pfs_files.back();
     pfs_files.pop_back();
 
-    pfs_dirs.emplace_back(dir);
+    pfs_dirs.emplace_back(std::move(dir));
 
     return true;
 }

src/core/file_sys/vfs.cpp

@@ -42,7 +42,7 @@ bool VfsFile::WriteByte(u8 data, size_t offset) {
     return Write(&data, 1, offset) == 1;
 }
 
-size_t VfsFile::WriteBytes(std::vector<u8> data, size_t offset) {
+size_t VfsFile::WriteBytes(const std::vector<u8>& data, size_t offset) {
     return Write(data.data(), data.size(), offset);
 }
 

src/core/file_sys/vfs.h

@@ -59,8 +59,7 @@ struct VfsFile : NonCopyable {
     // Returns the number of bytes (sizeof(T)*number_elements) read successfully.
     template <typename T>
     size_t ReadArray(T* data, size_t number_elements, size_t offset = 0) const {
-        static_assert(std::is_trivially_copyable<T>::value,
-                      "Data type must be trivially copyable.");
+        static_assert(std::is_trivially_copyable_v<T>, "Data type must be trivially copyable.");
 
         return Read(reinterpret_cast<u8*>(data), number_elements * sizeof(T), offset);
     }
@@ -69,8 +68,7 @@ struct VfsFile : NonCopyable {
     // Returns the number of bytes read successfully.
     template <typename T>
     size_t ReadBytes(T* data, size_t size, size_t offset = 0) const {
-        static_assert(std::is_trivially_copyable<T>::value,
-                      "Data type must be trivially copyable.");
+        static_assert(std::is_trivially_copyable_v<T>, "Data type must be trivially copyable.");
         return Read(reinterpret_cast<u8*>(data), size, offset);
     }
 
@@ -78,8 +76,7 @@ struct VfsFile : NonCopyable {
     // Returns the number of bytes read successfully (sizeof(T)).
     template <typename T>
     size_t ReadObject(T* data, size_t offset = 0) const {
-        static_assert(std::is_trivially_copyable<T>::value,
-                      "Data type must be trivially copyable.");
+        static_assert(std::is_trivially_copyable_v<T>, "Data type must be trivially copyable.");
         return Read(reinterpret_cast<u8*>(data), sizeof(T), offset);
     }
 
@@ -88,33 +85,29 @@ struct VfsFile : NonCopyable {
     virtual bool WriteByte(u8 data, size_t offset = 0);
     // Writes a vector of bytes to offset in file and returns the number of bytes successfully
     // written.
-    virtual size_t WriteBytes(std::vector<u8> data, size_t offset = 0);
+    virtual size_t WriteBytes(const std::vector<u8>& data, size_t offset = 0);
 
     // Writes an array of type T, size number_elements to offset in file.
     // Returns the number of bytes (sizeof(T)*number_elements) written successfully.
     template <typename T>
-    size_t WriteArray(T* data, size_t number_elements, size_t offset = 0) {
-        static_assert(std::is_trivially_copyable<T>::value,
-                      "Data type must be trivially copyable.");
-
+    size_t WriteArray(const T* data, size_t number_elements, size_t offset = 0) {
+        static_assert(std::is_trivially_copyable_v<T>, "Data type must be trivially copyable.");
         return Write(data, number_elements * sizeof(T), offset);
     }
 
     // Writes size bytes starting at memory location data to offset in file.
     // Returns the number of bytes written successfully.
     template <typename T>
-    size_t WriteBytes(T* data, size_t size, size_t offset = 0) {
-        static_assert(std::is_trivially_copyable<T>::value,
-                      "Data type must be trivially copyable.");
-        return Write(reinterpret_cast<u8*>(data), size, offset);
+    size_t WriteBytes(const T* data, size_t size, size_t offset = 0) {
+        static_assert(std::is_trivially_copyable_v<T>, "Data type must be trivially copyable.");
+        return Write(reinterpret_cast<const u8*>(data), size, offset);
     }
 
     // Writes one object of type T to offset in file.
     // Returns the number of bytes written successfully (sizeof(T)).
     template <typename T>
     size_t WriteObject(const T& data, size_t offset = 0) {
-        static_assert(std::is_trivially_copyable<T>::value,
-                      "Data type must be trivially copyable.");
+        static_assert(std::is_trivially_copyable_v<T>, "Data type must be trivially copyable.");
         return Write(&data, sizeof(T), offset);
     }
 

src/core/file_sys/vfs_offset.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <utility>
 
 #include "core/file_sys/vfs_offset.h"
@@ -75,7 +76,7 @@ bool OffsetVfsFile::WriteByte(u8 data, size_t r_offset) {
     return false;
 }
 
-size_t OffsetVfsFile::WriteBytes(std::vector<u8> data, size_t r_offset) {
+size_t OffsetVfsFile::WriteBytes(const std::vector<u8>& data, size_t r_offset) {
     return file->Write(data.data(), TrimToFit(data.size(), r_offset), offset + r_offset);
 }
 
@@ -88,7 +89,7 @@ size_t OffsetVfsFile::GetOffset() const {
 }
 
 size_t OffsetVfsFile::TrimToFit(size_t r_size, size_t r_offset) const {
-    return std::max<size_t>(std::min<size_t>(size - r_offset, r_size), 0);
+    return std::clamp(r_size, size_t{0}, size - r_offset);
 }
 
 } // namespace FileSys

src/core/file_sys/vfs_offset.h

@@ -28,7 +28,7 @@ struct OffsetVfsFile : public VfsFile {
     std::vector<u8> ReadBytes(size_t size, size_t offset) const override;
     std::vector<u8> ReadAllBytes() const override;
     bool WriteByte(u8 data, size_t offset) override;
-    size_t WriteBytes(std::vector<u8> data, size_t offset) override;
+    size_t WriteBytes(const std::vector<u8>& data, size_t offset) override;
 
     bool Rename(const std::string& name) override;
 

src/core/file_sys/vfs_real.cpp

@@ -2,6 +2,11 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
+#include <cstddef>
+#include <iterator>
+#include <utility>
+
 #include "common/common_paths.h"
 #include "common/logging/log.h"
 #include "core/file_sys/vfs_real.h"
@@ -104,11 +109,11 @@ RealVfsDirectory::RealVfsDirectory(const std::string& path_, Mode perms_)
 }
 
 std::vector<std::shared_ptr<VfsFile>> RealVfsDirectory::GetFiles() const {
-    return std::vector<std::shared_ptr<VfsFile>>(files);
+    return files;
 }
 
 std::vector<std::shared_ptr<VfsDirectory>> RealVfsDirectory::GetSubdirectories() const {
-    return std::vector<std::shared_ptr<VfsDirectory>>(subdirectories);
+    return subdirectories;
 }
 
 bool RealVfsDirectory::IsWritable() const {
@@ -163,14 +168,15 @@ bool RealVfsDirectory::Rename(const std::string& name) {
 }
 
 bool RealVfsDirectory::ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) {
-    auto iter = std::find(files.begin(), files.end(), file);
+    const auto iter = std::find(files.begin(), files.end(), file);
     if (iter == files.end())
         return false;
 
-    files[iter - files.begin()] = files.back();
+    const std::ptrdiff_t offset = std::distance(files.begin(), iter);
+    files[offset] = files.back();
     files.pop_back();
 
-    subdirectories.emplace_back(dir);
+    subdirectories.emplace_back(std::move(dir));
 
     return true;
 }

src/core/hle/ipc_helpers.h

@@ -174,6 +174,25 @@ public:
     template <typename First, typename... Other>
     void Push(const First& first_value, const Other&... other_values);
 
+    /**
+     * Helper function for pushing strongly-typed enumeration values.
+     *
+     * @tparam Enum The enumeration type to be pushed
+     *
+     * @param value The value to push.
+     *
+     * @note The underlying size of the enumeration type is the size of the
+     *       data that gets pushed. e.g. "enum class SomeEnum : u16" will
+     *       push a u16-sized amount of data.
+     */
+    template <typename Enum>
+    void PushEnum(Enum value) {
+        static_assert(std::is_enum_v<Enum>, "T must be an enum type within a PushEnum call.");
+        static_assert(!std::is_convertible_v<Enum, int>,
+                      "enum type in PushEnum must be a strongly typed enum.");
+        Push(static_cast<std::underlying_type_t<Enum>>(value));
+    }
+
     /**
      * @brief Copies the content of the given trivially copyable class to the buffer as a normal
      * param

src/core/hle/kernel/scheduler.cpp

@@ -56,6 +56,8 @@ void Scheduler::SwitchContext(Thread* new_thread) {
     if (previous_thread) {
         previous_thread->last_running_ticks = CoreTiming::GetTicks();
         cpu_core->SaveContext(previous_thread->context);
+        // Save the TPIDR_EL0 system register in case it was modified.
+        previous_thread->tpidr_el0 = cpu_core->GetTPIDR_EL0();
 
         if (previous_thread->status == ThreadStatus::Running) {
             // This is only the case when a reschedule is triggered without the current thread
@@ -87,6 +89,7 @@ void Scheduler::SwitchContext(Thread* new_thread) {
 
         cpu_core->LoadContext(new_thread->context);
         cpu_core->SetTlsAddress(new_thread->GetTLSAddress());
+        cpu_core->SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
         cpu_core->ClearExclusiveState();
     } else {
         current_thread = nullptr;

src/core/hle/kernel/thread.cpp

@@ -312,6 +312,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point,
     thread->status = ThreadStatus::Dormant;
     thread->entry_point = entry_point;
     thread->stack_top = stack_top;
+    thread->tpidr_el0 = 0;
     thread->nominal_priority = thread->current_priority = priority;
     thread->last_running_ticks = CoreTiming::GetTicks();
     thread->processor_id = processor_id;

src/core/hle/kernel/thread.h

@@ -182,6 +182,14 @@ public:
         return tls_address;
     }
 
+    /*
+     * Returns the value of the TPIDR_EL0 Read/Write system register for this thread.
+     * @returns The value of the TPIDR_EL0 register.
+     */
+    u64 GetTPIDR_EL0() const {
+        return tpidr_el0;
+    }
+
     /*
      * Returns the address of the current thread's command buffer, located in the TLS.
      * @returns VAddr of the thread's command buffer.
@@ -213,6 +221,7 @@ public:
     s32 processor_id;
 
     VAddr tls_address; ///< Virtual address of the Thread Local Storage of the thread
+    u64 tpidr_el0;     ///< TPIDR_EL0 read/write system register.
 
     SharedPtr<Process> owner_process; ///< Process that owns this thread
 

src/core/hle/service/apm/interface.cpp

@@ -20,6 +20,21 @@ public:
     }
 
 private:
+    enum class PerformanceConfiguration : u32 {
+        Config1 = 0x00010000,
+        Config2 = 0x00010001,
+        Config3 = 0x00010002,
+        Config4 = 0x00020000,
+        Config5 = 0x00020001,
+        Config6 = 0x00020002,
+        Config7 = 0x00020003,
+        Config8 = 0x00020004,
+        Config9 = 0x00020005,
+        Config10 = 0x00020006,
+        Config11 = 0x92220007,
+        Config12 = 0x92220008,
+    };
+
     void SetPerformanceConfiguration(Kernel::HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
 
@@ -40,7 +55,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
-        rb.Push<u32>(0); // Performance configuration
+        rb.Push<u32>(static_cast<u32>(PerformanceConfiguration::Config1));
 
         LOG_WARNING(Service_APM, "(STUBBED) called mode={}", static_cast<u32>(mode));
     }

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

@@ -272,9 +272,9 @@ void RegisterFileSystems() {
     sdmc_factory = nullptr;
 
     auto nand_directory = std::make_shared<FileSys::RealVfsDirectory>(
-        FileUtil::GetUserPath(D_NAND_IDX), FileSys::Mode::Write);
+        FileUtil::GetUserPath(FileUtil::UserPath::NANDDir), FileSys::Mode::Write);
     auto sd_directory = std::make_shared<FileSys::RealVfsDirectory>(
-        FileUtil::GetUserPath(D_SDMC_IDX), FileSys::Mode::Write);
+        FileUtil::GetUserPath(FileUtil::UserPath::SDMCDir), FileSys::Mode::Write);
 
     auto savedata = std::make_unique<FileSys::SaveDataFactory>(std::move(nand_directory));
     save_data_factory = std::move(savedata);

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

@@ -42,7 +42,7 @@ PL_U::PL_U() : ServiceFramework("pl:u") {
     RegisterHandlers(functions);
 
     // Attempt to load shared font data from disk
-    const std::string filepath{FileUtil::GetUserPath(D_SYSDATA_IDX) + SHARED_FONT};
+    const std::string filepath{FileUtil::GetUserPath(FileUtil::UserPath::SysDataDir) + SHARED_FONT};
     FileUtil::CreateFullPath(filepath); // Create path if not already created
     FileUtil::IOFile file(filepath, "rb");
 

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

@@ -42,6 +42,9 @@ u32 nvhost_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u
         if (command.cmd == NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO) {
             return SubmitGPFIFO(input, output);
         }
+        if (command.cmd == NVGPU_IOCTL_CHANNEL_KICKOFF_PB) {
+            return KickoffPB(input, output);
+        }
     }
 
     UNIMPLEMENTED_MSG("Unimplemented ioctl");
@@ -127,14 +130,37 @@ u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& outp
     IoctlSubmitGpfifo params{};
     std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
     LOG_WARNING(Service_NVDRV, "(STUBBED) called, gpfifo={:X}, num_entries={:X}, flags={:X}",
-                params.gpfifo, params.num_entries, params.flags);
+                params.address, params.num_entries, params.flags);
 
     auto entries = std::vector<IoctlGpfifoEntry>();
     entries.resize(params.num_entries);
     std::memcpy(&entries[0], &input.data()[sizeof(IoctlSubmitGpfifo)],
                 params.num_entries * sizeof(IoctlGpfifoEntry));
     for (auto entry : entries) {
-        VAddr va_addr = entry.Address();
+        Tegra::GPUVAddr va_addr = entry.Address();
+        Core::System::GetInstance().GPU().ProcessCommandList(va_addr, entry.sz);
+    }
+    params.fence_out.id = 0;
+    params.fence_out.value = 0;
+    std::memcpy(output.data(), &params, output.size());
+    return 0;
+}
+
+u32 nvhost_gpu::KickoffPB(const std::vector<u8>& input, std::vector<u8>& output) {
+    if (input.size() < sizeof(IoctlSubmitGpfifo)) {
+        UNIMPLEMENTED();
+    }
+    IoctlSubmitGpfifo params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
+    LOG_WARNING(Service_NVDRV, "(STUBBED) called, gpfifo={:X}, num_entries={:X}, flags={:X}",
+                params.address, params.num_entries, params.flags);
+
+    std::vector<IoctlGpfifoEntry> entries(params.num_entries);
+    Memory::ReadBlock(params.address, entries.data(),
+                      params.num_entries * sizeof(IoctlGpfifoEntry));
+
+    for (auto entry : entries) {
+        Tegra::GPUVAddr va_addr = entry.Address();
         Core::System::GetInstance().GPU().ProcessCommandList(va_addr, entry.sz);
     }
     params.fence_out.id = 0;

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

@@ -15,6 +15,7 @@ namespace Service::Nvidia::Devices {
 class nvmap;
 constexpr u32 NVGPU_IOCTL_MAGIC('H');
 constexpr u32 NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO(0x8);
+constexpr u32 NVGPU_IOCTL_CHANNEL_KICKOFF_PB(0x1b);
 
 class nvhost_gpu final : public nvdevice {
 public:
@@ -158,14 +159,14 @@ private:
             BitField<31, 1, u32_le> unk2;
         };
 
-        VAddr Address() const {
-            return (static_cast<VAddr>(gpu_va_hi) << 32) | entry0;
+        Tegra::GPUVAddr Address() const {
+            return (static_cast<Tegra::GPUVAddr>(gpu_va_hi) << 32) | entry0;
         }
     };
     static_assert(sizeof(IoctlGpfifoEntry) == 8, "IoctlGpfifoEntry is incorrect size");
 
     struct IoctlSubmitGpfifo {
-        u64_le gpfifo;      // (ignored) pointer to gpfifo fence structs
+        u64_le address;     // pointer to gpfifo entry structs
         u32_le num_entries; // number of fence objects being submitted
         u32_le flags;
         IoctlFence fence_out; // returned new fence object for others to wait on
@@ -193,6 +194,7 @@ private:
     u32 AllocGPFIFOEx2(const std::vector<u8>& input, std::vector<u8>& output);
     u32 AllocateObjectContext(const std::vector<u8>& input, std::vector<u8>& output);
     u32 SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 KickoffPB(const std::vector<u8>& input, std::vector<u8>& output);
     u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
     u32 ChannelSetTimeout(const std::vector<u8>& input, std::vector<u8>& output);
 

src/core/hle/service/nvdrv/interface.cpp

@@ -101,7 +101,7 @@ NVDRV::NVDRV(std::shared_ptr<Module> nvdrv, const char* name)
         {8, &NVDRV::SetClientPID, "SetClientPID"},
         {9, nullptr, "DumpGraphicsMemoryInfo"},
         {10, nullptr, "InitializeDevtools"},
-        {11, nullptr, "Ioctl2"},
+        {11, &NVDRV::Ioctl, "Ioctl2"},
         {12, nullptr, "Ioctl3"},
         {13, &NVDRV::FinishInitialize, "FinishInitialize"},
     };

src/core/loader/deconstructed_rom_directory.cpp

@@ -83,16 +83,13 @@ ResultStatus AppLoader_DeconstructedRomDirectory::Load(
     VAddr next_load_addr{Memory::PROCESS_IMAGE_VADDR};
     for (const auto& module : {"rtld", "main", "subsdk0", "subsdk1", "subsdk2", "subsdk3",
                                "subsdk4", "subsdk5", "subsdk6", "subsdk7", "sdk"}) {
-        const VAddr load_addr = next_load_addr;
         const FileSys::VirtualFile module_file = dir->GetFile(module);
-        if (module_file != nullptr)
+        if (module_file != nullptr) {
+            const VAddr load_addr = next_load_addr;
             next_load_addr = AppLoader_NSO::LoadModule(module_file, load_addr);
-        if (next_load_addr) {
             LOG_DEBUG(Loader, "loaded module {} @ 0x{:X}", module, load_addr);
             // Register module with GDBStub
             GDBStub::RegisterModule(module, load_addr, next_load_addr - 1, false);
-        } else {
-            next_load_addr = load_addr;
         }
     }
 

src/core/telemetry_session.cpp

@@ -37,7 +37,8 @@ static u64 GenerateTelemetryId() {
 
 u64 GetTelemetryId() {
     u64 telemetry_id{};
-    static const std::string& filename{FileUtil::GetUserPath(D_CONFIG_IDX) + "telemetry_id"};
+    static const std::string& filename{FileUtil::GetUserPath(FileUtil::UserPath::ConfigDir) +
+                                       "telemetry_id"};
 
     if (FileUtil::Exists(filename)) {
         FileUtil::IOFile file(filename, "rb");
@@ -61,7 +62,8 @@ u64 GetTelemetryId() {
 
 u64 RegenerateTelemetryId() {
     const u64 new_telemetry_id{GenerateTelemetryId()};
-    static const std::string& filename{FileUtil::GetUserPath(D_CONFIG_IDX) + "telemetry_id"};
+    static const std::string& filename{FileUtil::GetUserPath(FileUtil::UserPath::ConfigDir) +
+                                       "telemetry_id"};
 
     FileUtil::IOFile file(filename, "wb");
     if (!file.IsOpen()) {

src/tests/core/arm/arm_test_common.cpp

@@ -10,8 +10,6 @@
 
 namespace ArmTests {
 
-static Memory::PageTable* page_table = nullptr;
-
 TestEnvironment::TestEnvironment(bool mutable_memory_)
     : mutable_memory(mutable_memory_), test_memory(std::make_shared<TestMemory>(this)) {
 
@@ -67,10 +65,13 @@ boost::optional<bool> TestEnvironment::TestMemory::IsValidAddress(VAddr addr) {
 }
 
 boost::optional<u8> TestEnvironment::TestMemory::Read8(VAddr addr) {
-    auto iter = data.find(addr);
+    const auto iter = data.find(addr);
+
     if (iter == data.end()) {
-        return addr; // Some arbitrary data
+        // Some arbitrary data
+        return static_cast<u8>(addr);
     }
+
     return iter->second;
 }
 

src/tests/core/arm/arm_test_common.h

@@ -2,6 +2,8 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#pragma once
+
 #include <tuple>
 #include <unordered_map>
 #include <vector>
@@ -9,6 +11,10 @@
 #include "common/common_types.h"
 #include "core/memory_hook.h"
 
+namespace Memory {
+struct PageTable;
+}
+
 namespace ArmTests {
 
 struct WriteRecord {
@@ -79,6 +85,7 @@ private:
     bool mutable_memory;
     std::shared_ptr<TestMemory> test_memory;
     std::vector<WriteRecord> write_records;
+    Memory::PageTable* page_table = nullptr;
 };
 
 } // namespace ArmTests

src/video_core/engines/fermi_2d.cpp

@@ -6,8 +6,7 @@
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/textures/decoders.h"
 
-namespace Tegra {
-namespace Engines {
+namespace Tegra::Engines {
 
 Fermi2D::Fermi2D(MemoryManager& memory_manager) : memory_manager(memory_manager) {}
 
@@ -69,5 +68,4 @@ void Fermi2D::HandleSurfaceCopy() {
     }
 }
 
-} // namespace Engines
-} // namespace Tegra
+} // namespace Tegra::Engines

src/video_core/engines/fermi_2d.h

@@ -12,8 +12,7 @@
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 
-namespace Tegra {
-namespace Engines {
+namespace Tegra::Engines {
 
 #define FERMI2D_REG_INDEX(field_name)                                                              \
     (offsetof(Tegra::Engines::Fermi2D::Regs, field_name) / sizeof(u32))
@@ -110,5 +109,4 @@ ASSERT_REG_POSITION(operation, 0xAB);
 ASSERT_REG_POSITION(trigger, 0xB5);
 #undef ASSERT_REG_POSITION
 
-} // namespace Engines
-} // namespace Tegra
+} // namespace Tegra::Engines

src/video_core/engines/maxwell_3d.h

@@ -17,8 +17,7 @@
 #include "video_core/memory_manager.h"
 #include "video_core/textures/texture.h"
 
-namespace Tegra {
-namespace Engines {
+namespace Tegra::Engines {
 
 #define MAXWELL3D_REG_INDEX(field_name)                                                            \
     (offsetof(Tegra::Engines::Maxwell3D::Regs, field_name) / sizeof(u32))
@@ -488,7 +487,12 @@ public:
                     };
                 } rt_control;
 
-                INSERT_PADDING_WORDS(0x2B);
+                INSERT_PADDING_WORDS(0x2);
+
+                u32 zeta_width;
+                u32 zeta_height;
+
+                INSERT_PADDING_WORDS(0x27);
 
                 u32 depth_test_enable;
 
@@ -541,7 +545,11 @@ public:
 
                 u32 vb_element_base;
 
-                INSERT_PADDING_WORDS(0x49);
+                INSERT_PADDING_WORDS(0x40);
+
+                u32 zeta_enable;
+
+                INSERT_PADDING_WORDS(0x8);
 
                 struct {
                     u32 tsc_address_high;
@@ -866,6 +874,8 @@ ASSERT_REG_POSITION(clear_depth, 0x364);
 ASSERT_REG_POSITION(zeta, 0x3F8);
 ASSERT_REG_POSITION(vertex_attrib_format[0], 0x458);
 ASSERT_REG_POSITION(rt_control, 0x487);
+ASSERT_REG_POSITION(zeta_width, 0x48a);
+ASSERT_REG_POSITION(zeta_height, 0x48b);
 ASSERT_REG_POSITION(depth_test_enable, 0x4B3);
 ASSERT_REG_POSITION(independent_blend_enable, 0x4B9);
 ASSERT_REG_POSITION(depth_write_enabled, 0x4BA);
@@ -875,6 +885,7 @@ ASSERT_REG_POSITION(blend, 0x4CF);
 ASSERT_REG_POSITION(stencil, 0x4E0);
 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);
@@ -898,5 +909,4 @@ ASSERT_REG_POSITION(tex_info_buffers.size[0], 0xD2F);
 
 #undef ASSERT_REG_POSITION
 
-} // namespace Engines
-} // namespace Tegra
+} // namespace Tegra::Engines

src/video_core/engines/maxwell_compute.h

@@ -6,8 +6,7 @@
 
 #include "common/common_types.h"
 
-namespace Tegra {
-namespace Engines {
+namespace Tegra::Engines {
 
 class MaxwellCompute final {
 public:
@@ -18,5 +17,4 @@ public:
     void WriteReg(u32 method, u32 value);
 };
 
-} // namespace Engines
-} // namespace Tegra
+} // namespace Tegra::Engines

src/video_core/engines/maxwell_dma.h

@@ -12,8 +12,7 @@
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 
-namespace Tegra {
-namespace Engines {
+namespace Tegra::Engines {
 
 class MaxwellDMA final {
 public:
@@ -151,5 +150,4 @@ ASSERT_REG_POSITION(src_params, 0x1CA);
 
 #undef ASSERT_REG_POSITION
 
-} // namespace Engines
-} // namespace Tegra
+} // namespace Tegra::Engines

src/video_core/engines/shader_bytecode.h

@@ -15,8 +15,7 @@
 #include "common/bit_field.h"
 #include "common/common_types.h"
 
-namespace Tegra {
-namespace Shader {
+namespace Tegra::Shader {
 
 struct Register {
     /// Number of registers
@@ -109,8 +108,7 @@ union Sampler {
     u64 value{};
 };
 
-} // namespace Shader
-} // namespace Tegra
+} // namespace Tegra::Shader
 
 namespace std {
 
@@ -127,8 +125,7 @@ struct make_unsigned<Tegra::Shader::Register> {
 
 } // namespace std
 
-namespace Tegra {
-namespace Shader {
+namespace Tegra::Shader {
 
 enum class Pred : u64 {
     UnusedIndex = 0x7,
@@ -291,6 +288,11 @@ union Instruction {
         BitField<49, 1, u64> negate_a;
     } alu_integer;
 
+    union {
+        BitField<39, 3, u64> pred;
+        BitField<42, 1, u64> neg_pred;
+    } sel;
+
     union {
         BitField<39, 3, u64> pred;
         BitField<42, 1, u64> negate_pred;
@@ -516,6 +518,9 @@ public:
         ISCADD_C, // Scale and Add
         ISCADD_R,
         ISCADD_IMM,
+        SEL_C,
+        SEL_R,
+        SEL_IMM,
         MUFU,  // Multi-Function Operator
         RRO_C, // Range Reduction Operator
         RRO_R,
@@ -716,6 +721,9 @@ private:
             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("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("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"),
             INST("0100110010010---", Id::RRO_C, Type::Arithmetic, "RRO_C"),
             INST("0101110010010---", Id::RRO_R, Type::Arithmetic, "RRO_R"),
@@ -784,5 +792,4 @@ private:
     }
 };
 
-} // namespace Shader
-} // namespace Tegra
+} // namespace Tegra::Shader

src/video_core/gpu.cpp

@@ -20,7 +20,11 @@ GPU::GPU() {
 
 GPU::~GPU() = default;
 
-const Tegra::Engines::Maxwell3D& GPU::Get3DEngine() const {
+const Engines::Maxwell3D& GPU::Maxwell3D() const {
+    return *maxwell_3d;
+}
+
+Engines::Maxwell3D& GPU::Maxwell3D() {
     return *maxwell_3d;
 }
 

src/video_core/gpu.h

@@ -93,15 +93,14 @@ 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.
-    const Engines::Maxwell3D& Get3DEngine() const;
+    Engines::Maxwell3D& Maxwell3D();
 
     std::unique_ptr<MemoryManager> memory_manager;
 
-    Engines::Maxwell3D& Maxwell3D() {
-        return *maxwell_3d;
-    }
-
 private:
     /// Writes a single register in the engine bound to the specified subchannel
     void WriteReg(u32 method, u32 subchannel, u32 value, u32 remaining_params);

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -387,7 +387,7 @@ void RasterizerOpenGL::Clear() {
     }
     if (regs.clear_buffers.Z) {
         clear_mask |= GL_DEPTH_BUFFER_BIT;
-        use_depth_fb = true;
+        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.
@@ -413,11 +413,13 @@ void RasterizerOpenGL::Clear() {
     glClear(clear_mask);
 
     // Mark framebuffer surfaces as dirty
-    if (dirty_color_surface != nullptr) {
-        res_cache.MarkSurfaceAsDirty(dirty_color_surface);
-    }
-    if (dirty_depth_surface != nullptr) {
-        res_cache.MarkSurfaceAsDirty(dirty_depth_surface);
+    if (Settings::values.use_accurate_framebuffers) {
+        if (dirty_color_surface != nullptr) {
+            res_cache.FlushSurface(dirty_color_surface);
+        }
+        if (dirty_depth_surface != nullptr) {
+            res_cache.FlushSurface(dirty_depth_surface);
+        }
     }
 }
 
@@ -431,7 +433,7 @@ void RasterizerOpenGL::DrawArrays() {
     ScopeAcquireGLContext acquire_context;
 
     auto [dirty_color_surface, dirty_depth_surface] =
-        ConfigureFramebuffers(true, regs.zeta.Address() != 0);
+        ConfigureFramebuffers(true, regs.zeta.Address() != 0 && regs.zeta_enable != 0);
 
     SyncDepthTestState();
     SyncBlendState();
@@ -520,11 +522,13 @@ void RasterizerOpenGL::DrawArrays() {
     state.Apply();
 
     // Mark framebuffer surfaces as dirty
-    if (dirty_color_surface != nullptr) {
-        res_cache.MarkSurfaceAsDirty(dirty_color_surface);
-    }
-    if (dirty_depth_surface != nullptr) {
-        res_cache.MarkSurfaceAsDirty(dirty_depth_surface);
+    if (Settings::values.use_accurate_framebuffers) {
+        if (dirty_color_surface != nullptr) {
+            res_cache.FlushSurface(dirty_color_surface);
+        }
+        if (dirty_depth_surface != nullptr) {
+            res_cache.FlushSurface(dirty_depth_surface);
+        }
     }
 }
 
@@ -634,8 +638,8 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr
 u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, GLuint program,
                                         u32 current_bindpoint,
                                         const std::vector<GLShader::ConstBufferEntry>& entries) {
-    auto& gpu = Core::System::GetInstance().GPU();
-    auto& maxwell3d = gpu.Get3DEngine();
+    const auto& gpu = Core::System::GetInstance().GPU();
+    const auto& maxwell3d = gpu.Maxwell3D();
 
     // Reset all buffer draw state for this stage.
     for (auto& buffer : state.draw.const_buffers[static_cast<size_t>(stage)]) {
@@ -644,7 +648,7 @@ u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, GLuint progr
     }
 
     // Upload only the enabled buffers from the 16 constbuffers of each shader stage
-    auto& shader_stage = maxwell3d.state.shader_stages[static_cast<size_t>(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];
@@ -700,8 +704,8 @@ u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, GLuint progr
 
 u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, GLuint program, u32 current_unit,
                                     const std::vector<GLShader::SamplerEntry>& entries) {
-    auto& gpu = Core::System::GetInstance().GPU();
-    auto& maxwell3d = gpu.Get3DEngine();
+    const auto& gpu = Core::System::GetInstance().GPU();
+    const auto& maxwell3d = gpu.Maxwell3D();
 
     ASSERT_MSG(current_unit + entries.size() <= std::size(state.texture_units),
                "Exceeded the number of active textures.");

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -65,9 +65,9 @@ struct FormatTuple {
     return params;
 }
 
-/*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(
-    const Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig& config, Tegra::GPUVAddr zeta_address,
-    Tegra::DepthFormat format) {
+/*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(u32 zeta_width, u32 zeta_height,
+                                                             Tegra::GPUVAddr zeta_address,
+                                                             Tegra::DepthFormat format) {
 
     SurfaceParams params{};
     params.addr = zeta_address;
@@ -77,9 +77,9 @@ struct FormatTuple {
     params.component_type = ComponentTypeFromDepthFormat(format);
     params.type = GetFormatType(params.pixel_format);
     params.size_in_bytes = params.SizeInBytes();
-    params.width = config.width;
-    params.height = config.height;
-    params.unaligned_height = config.height;
+    params.width = zeta_width;
+    params.height = zeta_height;
+    params.unaligned_height = zeta_height;
     params.size_in_bytes = params.SizeInBytes();
     return params;
 }
@@ -254,6 +254,60 @@ static void AllocateSurfaceTexture(GLuint texture, const FormatTuple& format_tup
     cur_state.Apply();
 }
 
+static bool BlitTextures(GLuint src_tex, const MathUtil::Rectangle<u32>& src_rect, GLuint dst_tex,
+                         const MathUtil::Rectangle<u32>& dst_rect, SurfaceType type,
+                         GLuint read_fb_handle, GLuint draw_fb_handle) {
+    OpenGLState prev_state{OpenGLState::GetCurState()};
+    SCOPE_EXIT({ prev_state.Apply(); });
+
+    OpenGLState state;
+    state.draw.read_framebuffer = read_fb_handle;
+    state.draw.draw_framebuffer = draw_fb_handle;
+    state.Apply();
+
+    u32 buffers{};
+
+    if (type == SurfaceType::ColorTexture) {
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, src_tex,
+                               0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
+                               0);
+
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex,
+                               0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
+                               0);
+
+        buffers = GL_COLOR_BUFFER_BIT;
+    } else if (type == SurfaceType::Depth) {
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, src_tex, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
+
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, dst_tex, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
+
+        buffers = GL_DEPTH_BUFFER_BIT;
+    } else if (type == SurfaceType::DepthStencil) {
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                               src_tex, 0);
+
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                               dst_tex, 0);
+
+        buffers = GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
+    }
+
+    glBlitFramebuffer(src_rect.left, src_rect.bottom, src_rect.right, src_rect.top, dst_rect.left,
+                      dst_rect.bottom, dst_rect.right, dst_rect.top, buffers,
+                      buffers == GL_COLOR_BUFFER_BIT ? GL_LINEAR : GL_NEAREST);
+
+    return true;
+}
+
 CachedSurface::CachedSurface(const SurfaceParams& params) : params(params) {
     texture.Create();
     const auto& rect{params.GetRect()};
@@ -519,8 +573,8 @@ SurfaceSurfaceRect_Tuple RasterizerCacheOpenGL::GetFramebufferSurfaces(
     }
 
     if (using_depth_fb) {
-        depth_params =
-            SurfaceParams::CreateForDepthBuffer(regs.rt[0], regs.zeta.Address(), regs.zeta.format);
+        depth_params = SurfaceParams::CreateForDepthBuffer(regs.zeta_width, regs.zeta_height,
+                                                           regs.zeta.Address(), regs.zeta.format);
     }
 
     MathUtil::Rectangle<u32> color_rect{};
@@ -565,17 +619,9 @@ void RasterizerCacheOpenGL::LoadSurface(const Surface& surface) {
     surface->UploadGLTexture(read_framebuffer.handle, draw_framebuffer.handle);
 }
 
-void RasterizerCacheOpenGL::MarkSurfaceAsDirty(const Surface& surface) {
-    if (Settings::values.use_accurate_framebuffers) {
-        // If enabled, always flush dirty surfaces
-        surface->DownloadGLTexture(read_framebuffer.handle, draw_framebuffer.handle);
-        surface->FlushGLBuffer();
-    } else {
-        // Otherwise, don't mark surfaces that we write to as cached, because the resulting loads
-        // and flushes are very slow and do not seem to improve accuracy
-        const auto& params{surface->GetSurfaceParams()};
-        Memory::RasterizerMarkRegionCached(params.addr, params.size_in_bytes, false);
-    }
+void RasterizerCacheOpenGL::FlushSurface(const Surface& surface) {
+    surface->DownloadGLTexture(read_framebuffer.handle, draw_framebuffer.handle);
+    surface->FlushGLBuffer();
 }
 
 Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params) {
@@ -588,25 +634,53 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params) {
     if (gpu.memory_manager->GpuToCpuAddress(params.addr) == boost::none)
         return {};
 
-    // Check for an exact match in existing surfaces
-    const auto& surface_key{SurfaceKey::Create(params)};
-    const auto& search{surface_cache.find(surface_key)};
+    // 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;
         if (Settings::values.use_accurate_framebuffers) {
-            // Reload the surface from Switch memory
-            LoadSurface(surface);
+            // If use_accurate_framebuffers is enabled, always load from memory
+            FlushSurface(surface);
+            UnregisterSurface(surface);
+        } else if (surface->GetSurfaceParams() != params) {
+            // If surface parameters changed, recreate the surface from the old one
+            return RecreateSurface(surface, params);
+        } else {
+            // Use the cached surface as-is
+            return surface;
         }
-    } else {
-        surface = std::make_shared<CachedSurface>(params);
-        RegisterSurface(surface);
-        LoadSurface(surface);
     }
 
+    // No surface found - create a new one
+    surface = std::make_shared<CachedSurface>(params);
+    RegisterSurface(surface);
+    LoadSurface(surface);
+
     return surface;
 }
 
+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(params.pixel_format == new_params.pixel_format);
+    ASSERT(params.component_type == new_params.component_type);
+
+    // Create a new surface with the new parameters, and blit the previous surface to it
+    Surface new_surface{std::make_shared<CachedSurface>(new_params)};
+    BlitTextures(surface->Texture().handle, params.GetRect(), new_surface->Texture().handle,
+                 new_surface->GetSurfaceParams().GetRect(), params.type, read_framebuffer.handle,
+                 draw_framebuffer.handle);
+
+    // Update cache accordingly
+    UnregisterSurface(surface);
+    RegisterSurface(new_surface);
+
+    return new_surface;
+}
+
 Surface RasterizerCacheOpenGL::TryFindFramebufferSurface(VAddr cpu_addr) const {
     // Tries to find the GPU address of a framebuffer based on the CPU address. This is because
     // final output framebuffers are specified by CPU address, but internally our GPU cache uses
@@ -652,22 +726,20 @@ void RasterizerCacheOpenGL::InvalidateRegion(Tegra::GPUVAddr addr, size_t size)
 
 void RasterizerCacheOpenGL::RegisterSurface(const Surface& surface) {
     const auto& params{surface->GetSurfaceParams()};
-    const auto& surface_key{SurfaceKey::Create(params)};
-    const auto& search{surface_cache.find(surface_key)};
+    const auto& search{surface_cache.find(params.addr)};
 
     if (search != surface_cache.end()) {
         // Registered already
         return;
     }
 
-    surface_cache[surface_key] = surface;
+    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& surface_key{SurfaceKey::Create(params)};
-    const auto& search{surface_cache.find(surface_key)};
+    const auto& search{surface_cache.find(params.addr)};
 
     if (search == surface_cache.end()) {
         // Unregistered already

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -10,7 +10,6 @@
 #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/renderer_opengl/gl_resource_manager.h"
@@ -137,6 +136,7 @@ struct SurfaceParams {
         ASSERT(static_cast<size_t>(format) < bpp_table.size());
         return bpp_table[static_cast<size_t>(format)];
     }
+
     u32 GetFormatBpp() const {
         return GetFormatBpp(pixel_format);
     }
@@ -365,9 +365,21 @@ struct SurfaceParams {
         const Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig& config);
 
     /// Creates SurfaceParams for a depth buffer configuration
-    static SurfaceParams CreateForDepthBuffer(
-        const Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig& config,
-        Tegra::GPUVAddr zeta_address, Tegra::DepthFormat format);
+    static SurfaceParams CreateForDepthBuffer(u32 zeta_width, u32 zeta_height,
+                                              Tegra::GPUVAddr zeta_address,
+                                              Tegra::DepthFormat format);
+
+    bool operator==(const SurfaceParams& other) const {
+        return std::tie(addr, is_tiled, block_height, pixel_format, component_type, type, width,
+                        height, unaligned_height, size_in_bytes) ==
+               std::tie(other.addr, other.is_tiled, other.block_height, other.pixel_format,
+                        other.component_type, other.type, other.width, other.height,
+                        other.unaligned_height, other.size_in_bytes);
+    }
+
+    bool operator!=(const SurfaceParams& other) const {
+        return !operator==(other);
+    }
 
     Tegra::GPUVAddr addr;
     bool is_tiled;
@@ -381,24 +393,6 @@ struct SurfaceParams {
     size_t size_in_bytes;
 };
 
-/// Hashable variation of SurfaceParams, used for a key in the surface cache
-struct SurfaceKey : Common::HashableStruct<SurfaceParams> {
-    static SurfaceKey Create(const SurfaceParams& params) {
-        SurfaceKey res;
-        res.state = params;
-        return res;
-    }
-};
-
-namespace std {
-template <>
-struct hash<SurfaceKey> {
-    size_t operator()(const SurfaceKey& k) const {
-        return k.Hash();
-    }
-};
-} // namespace std
-
 class CachedSurface final {
 public:
     CachedSurface(const SurfaceParams& params);
@@ -444,8 +438,8 @@ public:
     SurfaceSurfaceRect_Tuple GetFramebufferSurfaces(bool using_color_fb, bool using_depth_fb,
                                                     const MathUtil::Rectangle<s32>& viewport);
 
-    /// Marks the specified surface as "dirty", in that it is out of sync with Switch memory
-    void MarkSurfaceAsDirty(const Surface& surface);
+    /// Flushes the surface to Switch memory
+    void FlushSurface(const Surface& surface);
 
     /// Tries to find a framebuffer GPU address based on the provided CPU address
     Surface TryFindFramebufferSurface(VAddr cpu_addr) const;
@@ -460,6 +454,9 @@ private:
     void LoadSurface(const Surface& surface);
     Surface GetSurface(const SurfaceParams& params);
 
+    /// 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);
 
@@ -469,7 +466,7 @@ private:
     /// Increase/decrease the number of surface in pages touching the specified region
     void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta);
 
-    std::unordered_map<SurfaceKey, Surface> surface_cache;
+    std::unordered_map<Tegra::GPUVAddr, Surface> surface_cache;
     PageMap cached_pages;
 
     OGLFramebuffer read_framebuffer;

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -12,8 +12,7 @@
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
 
-namespace GLShader {
-namespace Decompiler {
+namespace GLShader::Decompiler {
 
 using Tegra::Shader::Attribute;
 using Tegra::Shader::Instruction;
@@ -1140,6 +1139,15 @@ private:
                                           "((" + op_a + " << " + shift + ") + " + op_b + ')', 1, 1);
                 break;
             }
+            case OpCode::Id::SEL_C:
+            case OpCode::Id::SEL_R:
+            case OpCode::Id::SEL_IMM: {
+                std::string condition =
+                    GetPredicateCondition(instr.sel.pred, instr.sel.neg_pred != 0);
+                regs.SetRegisterToInteger(instr.gpr0, true, 0,
+                                          '(' + condition + ") ? " + op_a + " : " + op_b, 1, 1);
+                break;
+            }
             case OpCode::Id::LOP_C:
             case OpCode::Id::LOP_R:
             case OpCode::Id::LOP_IMM: {
@@ -1845,5 +1853,4 @@ boost::optional<ProgramResult> DecompileProgram(const ProgramCode& program_code,
     return boost::none;
 }
 
-} // namespace Decompiler
-} // namespace GLShader
+} // namespace GLShader::Decompiler

src/video_core/renderer_opengl/gl_shader_decompiler.h

@@ -12,8 +12,7 @@
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
 
-namespace GLShader {
-namespace Decompiler {
+namespace GLShader::Decompiler {
 
 using Tegra::Engines::Maxwell3D;
 
@@ -23,5 +22,4 @@ boost::optional<ProgramResult> DecompileProgram(const ProgramCode& program_code,
                                                 Maxwell3D::Regs::ShaderStage stage,
                                                 const std::string& suffix);
 
-} // namespace Decompiler
-} // namespace GLShader
+} // namespace GLShader::Decompiler

src/video_core/textures/decoders.cpp

@@ -10,8 +10,7 @@
 #include "video_core/textures/decoders.h"
 #include "video_core/textures/texture.h"
 
-namespace Tegra {
-namespace Texture {
+namespace Tegra::Texture {
 
 /**
  * Calculates the offset of an (x, y) position within a swizzled texture.
@@ -186,5 +185,4 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat
     return rgba_data;
 }
 
-} // namespace Texture
-} // namespace Tegra
+} // namespace Tegra::Texture

src/video_core/textures/decoders.h

@@ -8,8 +8,7 @@
 #include "common/common_types.h"
 #include "video_core/textures/texture.h"
 
-namespace Tegra {
-namespace Texture {
+namespace Tegra::Texture {
 
 /**
  * Unswizzles a swizzled texture without changing its format.
@@ -33,5 +32,4 @@ void CopySwizzledData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_bytes_
 std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat format, u32 width,
                               u32 height);
 
-} // namespace Texture
-} // namespace Tegra
+} // namespace Tegra::Texture

src/video_core/textures/texture.h

@@ -10,8 +10,7 @@
 #include "common/common_types.h"
 #include "video_core/memory_manager.h"
 
-namespace Tegra {
-namespace Texture {
+namespace Tegra::Texture {
 
 enum class TextureFormat : u32 {
     R32_G32_B32_A32 = 0x01,
@@ -260,5 +259,4 @@ struct FullTextureInfo {
 /// Returns the number of bytes per pixel of the input texture format.
 u32 BytesPerPixel(TextureFormat format);
 
-} // namespace Texture
-} // namespace Tegra
+} // namespace Tegra::Texture

src/yuzu/configuration/config.cpp

@@ -10,7 +10,7 @@
 
 Config::Config() {
     // TODO: Don't hardcode the path; let the frontend decide where to put the config files.
-    qt_config_loc = FileUtil::GetUserPath(D_CONFIG_IDX) + "qt-config.ini";
+    qt_config_loc = FileUtil::GetUserPath(FileUtil::UserPath::ConfigDir) + "qt-config.ini";
     FileUtil::CreateFullPath(qt_config_loc);
     qt_config = new QSettings(QString::fromStdString(qt_config_loc), QSettings::IniFormat);
 

src/yuzu/configuration/configure_debug.cpp

@@ -19,7 +19,7 @@ ConfigureDebug::ConfigureDebug(QWidget* parent) : QWidget(parent), ui(new Ui::Co
     ui->setupUi(this);
     this->setConfiguration();
     connect(ui->open_log_button, &QPushButton::pressed, []() {
-        QString path = QString::fromStdString(FileUtil::GetUserPath(D_LOGS_IDX));
+        QString path = QString::fromStdString(FileUtil::GetUserPath(FileUtil::UserPath::LogDir));
         QDesktopServices::openUrl(QUrl::fromLocalFile(path));
     });
 }

src/yuzu/debugger/graphics/graphics_surface.cpp

@@ -336,9 +336,9 @@ void GraphicsSurfaceWidget::OnUpdate() {
         // TODO: Store a reference to the registers in the debug context instead of accessing them
         // directly...
 
-        auto& registers = gpu.Get3DEngine().regs;
-        auto& rt = registers.rt[static_cast<size_t>(surface_source) -
-                                static_cast<size_t>(Source::RenderTarget0)];
+        const auto& registers = gpu.Maxwell3D().regs;
+        const auto& rt = registers.rt[static_cast<size_t>(surface_source) -
+                                      static_cast<size_t>(Source::RenderTarget0)];
 
         surface_address = rt.Address();
         surface_width = rt.width;

src/yuzu/main.cpp

@@ -909,6 +909,16 @@ 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(); });
@@ -927,13 +937,7 @@ int main(int argc, char* argv[]) {
 
     GMainWindow main_window;
     // After settings have been loaded by GMainWindow, apply the filter
-    Log::Filter log_filter;
-    log_filter.ParseFilterString(Settings::values.log_filter);
-    Log::SetGlobalFilter(log_filter);
-    FileUtil::CreateFullPath(FileUtil::GetUserPath(D_LOGS_IDX));
-    Log::AddBackend(
-        std::make_unique<Log::FileBackend>(FileUtil::GetUserPath(D_LOGS_IDX) + LOG_FILE));
-
+    InitializeLogging();
     main_window.show();
     return app.exec();
 }

src/yuzu_cmd/config.cpp

@@ -15,7 +15,7 @@
 
 Config::Config() {
     // TODO: Don't hardcode the path; let the frontend decide where to put the config files.
-    sdl2_config_loc = FileUtil::GetUserPath(D_CONFIG_IDX) + "sdl2-config.ini";
+    sdl2_config_loc = FileUtil::GetUserPath(FileUtil::UserPath::ConfigDir) + "sdl2-config.ini";
     sdl2_config = std::make_unique<INIReader>(sdl2_config_loc);
 
     Reload();

src/yuzu_cmd/yuzu.cpp

@@ -56,6 +56,18 @@ static void PrintVersion() {
     std::cout << "yuzu " << Common::g_scm_branch << " " << Common::g_scm_desc << std::endl;
 }
 
+static void InitializeLogging() {
+    Log::Filter log_filter(Log::Level::Debug);
+    log_filter.ParseFilterString(Settings::values.log_filter);
+    Log::SetGlobalFilter(log_filter);
+
+    Log::AddBackend(std::make_unique<Log::ColorConsoleBackend>());
+
+    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));
+}
+
 /// Application entry point
 int main(int argc, char** argv) {
     Config config;
@@ -124,14 +136,7 @@ int main(int argc, char** argv) {
     LocalFree(argv_w);
 #endif
 
-    Log::Filter log_filter(Log::Level::Debug);
-    log_filter.ParseFilterString(Settings::values.log_filter);
-    Log::SetGlobalFilter(log_filter);
-
-    Log::AddBackend(std::make_unique<Log::ColorConsoleBackend>());
-    FileUtil::CreateFullPath(FileUtil::GetUserPath(D_LOGS_IDX));
-    Log::AddBackend(
-        std::make_unique<Log::FileBackend>(FileUtil::GetUserPath(D_LOGS_IDX) + LOG_FILE));
+    InitializeLogging();
 
     MicroProfileOnThreadCreate("EmuThread");
     SCOPE_EXIT({ MicroProfileShutdown(); });