vk_compute_pass: Implement indexed quads

Implement indexed quads (GL_QUADS used with glDrawElements*) with a
compute pass conversion.

The compute shader converts from uint8/uint16/uint32 indices to uint32.
The format is passed through push constants to avoid having different
variants of the same shader.

- Used by Fast RMX
- Used by Xenoblade Chronicles 2 (it still has graphical due to
synchronization issues on Vulkan)

.ci/templates/build-msvc.yml

@@ -4,7 +4,7 @@ parameters:
   version: ''
 
 steps:
-- script: mkdir build && cd build && cmake -G "Visual Studio 15 2017 Win64" --config Release -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_BUNDLED_SDL2=1 -DYUZU_USE_BUNDLED_UNICORN=1 -DYUZU_USE_QT_WEB_ENGINE=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${COMPAT} -DUSE_DISCORD_PRESENCE=ON -DDISPLAY_VERSION=${{ parameters['version'] }} .. && cd ..
+- script: mkdir build && cd build && cmake -G "Visual Studio 16 2019" -A x64 --config Release -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_BUNDLED_SDL2=1 -DYUZU_USE_BUNDLED_UNICORN=1 -DYUZU_USE_QT_WEB_ENGINE=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${COMPAT} -DUSE_DISCORD_PRESENCE=ON -DDISPLAY_VERSION=${{ parameters['version'] }} .. && cd ..
   displayName: 'Configure CMake'
 - task: MSBuild@1
   displayName: 'Build'

.ci/yuzu-mainline-step2.yml

@@ -45,7 +45,7 @@ stages:
   - job: build
     displayName: 'msvc'
     pool:
-      vmImage: vs2017-win2016
+      vmImage: windows-2019
     steps:
     - template: ./templates/sync-source.yml
       parameters:

.ci/yuzu-patreon-step2.yml

@@ -22,7 +22,7 @@ stages:
   - job: build
     displayName: 'windows-msvc'
     pool:
-      vmImage: vs2017-win2016
+      vmImage: windows-2019
     steps:
     - template: ./templates/sync-source.yml
       parameters:

externals/CMakeLists.txt

@@ -3,13 +3,27 @@
 set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${PROJECT_SOURCE_DIR}/CMakeModules)
 include(DownloadExternals)
 
+# pkgconfig -- it is used to find shared libraries without cmake modules on linux systems
+find_package(PkgConfig)
+if (NOT PkgConfig_FOUND)
+    function(pkg_check_modules)
+        # STUB
+    endfunction()
+endif()
+
 # Catch
 add_library(catch-single-include INTERFACE)
 target_include_directories(catch-single-include INTERFACE catch/single_include)
 
 # libfmt
-add_subdirectory(fmt)
-add_library(fmt::fmt ALIAS fmt)
+pkg_check_modules(FMT IMPORTED_TARGET GLOBAL fmt>=6.1.0)
+if (FMT_FOUND)
+    add_library(fmt::fmt ALIAS PkgConfig::FMT)
+else()
+    message(STATUS "fmt 6.1.0 or newer not found, falling back to externals")
+    add_subdirectory(fmt)
+    add_library(fmt::fmt ALIAS fmt)
+endif()
 
 # Dynarmic
 if (ARCHITECTURE_x86_64)
@@ -30,9 +44,15 @@ add_subdirectory(glad)
 add_subdirectory(inih)
 
 # lz4
-set(LZ4_BUNDLED_MODE ON)
-add_subdirectory(lz4/contrib/cmake_unofficial EXCLUDE_FROM_ALL)
-target_include_directories(lz4_static INTERFACE ./lz4/lib)
+pkg_check_modules(LIBLZ4 IMPORTED_TARGET GLOBAL liblz4>=1.8.0)
+if (LIBLZ4_FOUND)
+    add_library(lz4_static ALIAS PkgConfig::LIBLZ4)
+else()
+    message(STATUS "liblz4 1.8.0 or newer not found, falling back to externals")
+    set(LZ4_BUNDLED_MODE ON)
+    add_subdirectory(lz4/contrib/cmake_unofficial EXCLUDE_FROM_ALL)
+    target_include_directories(lz4_static INTERFACE ./lz4/lib)
+endif()
 
 # mbedtls
 add_subdirectory(mbedtls EXCLUDE_FROM_ALL)
@@ -47,15 +67,27 @@ add_library(unicorn-headers INTERFACE)
 target_include_directories(unicorn-headers INTERFACE ./unicorn/include)
 
 # Zstandard
-add_subdirectory(zstd/build/cmake EXCLUDE_FROM_ALL)
-target_include_directories(libzstd_static INTERFACE ./zstd/lib)
+pkg_check_modules(LIBZSTD IMPORTED_TARGET GLOBAL libzstd>=1.3.8)
+if (LIBZSTD_FOUND)
+    add_library(libzstd_static ALIAS PkgConfig::LIBZSTD)
+else()
+    message(STATUS "libzstd 1.3.8 or newer not found, falling back to externals")
+    add_subdirectory(zstd/build/cmake EXCLUDE_FROM_ALL)
+    target_include_directories(libzstd_static INTERFACE ./zstd/lib)
+endif()
 
 # SoundTouch
 add_subdirectory(soundtouch)
 
 # Opus
-add_subdirectory(opus)
-target_include_directories(opus INTERFACE ./opus/include)
+pkg_check_modules(OPUS IMPORTED_TARGET GLOBAL opus>=1.3.1)
+if (OPUS_FOUND)
+    add_library(opus ALIAS PkgConfig::OPUS)
+else()
+    message(STATUS "opus 1.3.1 or newer not found, falling back to externals")
+    add_subdirectory(opus)
+    target_include_directories(opus INTERFACE ./opus/include)
+endif()
 
 # Cubeb
 if(ENABLE_CUBEB)
@@ -75,18 +107,35 @@ if (ENABLE_VULKAN)
 endif()
 
 # zlib
-add_subdirectory(zlib EXCLUDE_FROM_ALL)
-set(ZLIB_LIBRARIES z)
+find_package(ZLIB 1.2.11)
+if (NOT ZLIB_FOUND)
+    message(STATUS "zlib 1.2.11 or newer not found, falling back to externals")
+    add_subdirectory(zlib EXCLUDE_FROM_ALL)
+    set(ZLIB_LIBRARIES z)
+endif()
 
 # libzip
-add_subdirectory(libzip EXCLUDE_FROM_ALL)
+pkg_check_modules(LIBZIP IMPORTED_TARGET GLOBAL libzip>=1.5.3)
+if (LIBZIP_FOUND)
+    add_library(zip ALIAS PkgConfig::LIBZIP)
+else()
+    message(STATUS "libzip 1.5.3 or newer not found, falling back to externals")
+    add_subdirectory(libzip EXCLUDE_FROM_ALL)
+endif()
 
 if (ENABLE_WEB_SERVICE)
     # LibreSSL
-    set(LIBRESSL_SKIP_INSTALL ON CACHE BOOL "")
-    add_subdirectory(libressl EXCLUDE_FROM_ALL)
-    target_include_directories(ssl INTERFACE ./libressl/include)
-	target_compile_definitions(ssl PRIVATE -DHAVE_INET_NTOP)
+    find_package(OpenSSL COMPONENTS Crypto SSL)
+    if (NOT OpenSSL_FOUND)
+        message(STATUS "OpenSSL not found, falling back to externals")
+        set(LIBRESSL_SKIP_INSTALL ON CACHE BOOL "")
+        add_subdirectory(libressl EXCLUDE_FROM_ALL)
+        target_include_directories(ssl INTERFACE ./libressl/include)
+        target_compile_definitions(ssl PRIVATE -DHAVE_INET_NTOP)
+        get_directory_property(OPENSSL_LIBRARIES
+            DIRECTORY libressl
+            DEFINITION OPENSSL_LIBS)
+    endif()
 
     # lurlparser
     add_subdirectory(lurlparser EXCLUDE_FROM_ALL)
@@ -94,6 +143,8 @@ if (ENABLE_WEB_SERVICE)
     # httplib
     add_library(httplib INTERFACE)
     target_include_directories(httplib INTERFACE ./httplib)
+    target_compile_definitions(httplib INTERFACE -DCPPHTTPLIB_OPENSSL_SUPPORT)
+    target_link_libraries(httplib INTERFACE ${OPENSSL_LIBRARIES})
 
     # JSON
     add_library(json-headers INTERFACE)

src/CMakeLists.txt

@@ -53,6 +53,7 @@ if (MSVC)
 else()
     add_compile_options(
         -Wall
+        -Werror=reorder
         -Wno-attributes
     )
 

src/common/CMakeLists.txt

@@ -106,6 +106,8 @@ add_library(common STATIC
     common_funcs.h
     common_paths.h
     common_types.h
+    dynamic_library.cpp
+    dynamic_library.h
     file_util.cpp
     file_util.h
     hash.h

src/common/dynamic_library.cpp

@@ -0,0 +1,106 @@
+// Copyright 2019 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <string>
+#include <utility>
+
+#include <fmt/format.h>
+
+#include "common/dynamic_library.h"
+
+#ifdef _WIN32
+#include <windows.h>
+#else
+#include <dlfcn.h>
+#endif
+
+namespace Common {
+
+DynamicLibrary::DynamicLibrary() = default;
+
+DynamicLibrary::DynamicLibrary(const char* filename) {
+    Open(filename);
+}
+
+DynamicLibrary::DynamicLibrary(DynamicLibrary&& rhs) noexcept
+    : handle{std::exchange(rhs.handle, nullptr)} {}
+
+DynamicLibrary& DynamicLibrary::operator=(DynamicLibrary&& rhs) noexcept {
+    Close();
+    handle = std::exchange(rhs.handle, nullptr);
+    return *this;
+}
+
+DynamicLibrary::~DynamicLibrary() {
+    Close();
+}
+
+std::string DynamicLibrary::GetUnprefixedFilename(const char* filename) {
+#if defined(_WIN32)
+    return std::string(filename) + ".dll";
+#elif defined(__APPLE__)
+    return std::string(filename) + ".dylib";
+#else
+    return std::string(filename) + ".so";
+#endif
+}
+
+std::string DynamicLibrary::GetVersionedFilename(const char* libname, int major, int minor) {
+#if defined(_WIN32)
+    if (major >= 0 && minor >= 0)
+        return fmt::format("{}-{}-{}.dll", libname, major, minor);
+    else if (major >= 0)
+        return fmt::format("{}-{}.dll", libname, major);
+    else
+        return fmt::format("{}.dll", libname);
+#elif defined(__APPLE__)
+    const char* prefix = std::strncmp(libname, "lib", 3) ? "lib" : "";
+    if (major >= 0 && minor >= 0)
+        return fmt::format("{}{}.{}.{}.dylib", prefix, libname, major, minor);
+    else if (major >= 0)
+        return fmt::format("{}{}.{}.dylib", prefix, libname, major);
+    else
+        return fmt::format("{}{}.dylib", prefix, libname);
+#else
+    const char* prefix = std::strncmp(libname, "lib", 3) ? "lib" : "";
+    if (major >= 0 && minor >= 0)
+        return fmt::format("{}{}.so.{}.{}", prefix, libname, major, minor);
+    else if (major >= 0)
+        return fmt::format("{}{}.so.{}", prefix, libname, major);
+    else
+        return fmt::format("{}{}.so", prefix, libname);
+#endif
+}
+
+bool DynamicLibrary::Open(const char* filename) {
+#ifdef _WIN32
+    handle = reinterpret_cast<void*>(LoadLibraryA(filename));
+#else
+    handle = dlopen(filename, RTLD_NOW);
+#endif
+    return handle != nullptr;
+}
+
+void DynamicLibrary::Close() {
+    if (!IsOpen())
+        return;
+
+#ifdef _WIN32
+    FreeLibrary(reinterpret_cast<HMODULE>(handle));
+#else
+    dlclose(handle);
+#endif
+    handle = nullptr;
+}
+
+void* DynamicLibrary::GetSymbolAddress(const char* name) const {
+#ifdef _WIN32
+    return reinterpret_cast<void*>(GetProcAddress(reinterpret_cast<HMODULE>(handle), name));
+#else
+    return reinterpret_cast<void*>(dlsym(handle, name));
+#endif
+}
+
+} // namespace Common

src/common/dynamic_library.h

@@ -0,0 +1,75 @@
+// Copyright 2019 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <string>
+
+namespace Common {
+
+/**
+ * Provides a platform-independent interface for loading a dynamic library and retrieving symbols.
+ * The interface maintains an internal reference count to allow one handle to be shared between
+ * multiple users.
+ */
+class DynamicLibrary final {
+public:
+    /// Default constructor, does not load a library.
+    explicit DynamicLibrary();
+
+    /// Automatically loads the specified library. Call IsOpen() to check validity before use.
+    explicit DynamicLibrary(const char* filename);
+
+    /// Moves the library.
+    DynamicLibrary(DynamicLibrary&&) noexcept;
+    DynamicLibrary& operator=(DynamicLibrary&&) noexcept;
+
+    /// Delete copies, we can't copy a dynamic library.
+    DynamicLibrary(const DynamicLibrary&) = delete;
+    DynamicLibrary& operator=(const DynamicLibrary&) = delete;
+
+    /// Closes the library.
+    ~DynamicLibrary();
+
+    /// Returns the specified library name with the platform-specific suffix added.
+    static std::string GetUnprefixedFilename(const char* filename);
+
+    /// Returns the specified library name in platform-specific format.
+    /// Major/minor versions will not be included if set to -1.
+    /// If libname already contains the "lib" prefix, it will not be added again.
+    /// Windows: LIBNAME-MAJOR-MINOR.dll
+    /// Linux: libLIBNAME.so.MAJOR.MINOR
+    /// Mac: libLIBNAME.MAJOR.MINOR.dylib
+    static std::string GetVersionedFilename(const char* libname, int major = -1, int minor = -1);
+
+    /// Returns true if a module is loaded, otherwise false.
+    bool IsOpen() const {
+        return handle != nullptr;
+    }
+
+    /// Loads (or replaces) the handle with the specified library file name.
+    /// Returns true if the library was loaded and can be used.
+    bool Open(const char* filename);
+
+    /// Unloads the library, any function pointers from this library are no longer valid.
+    void Close();
+
+    /// Returns the address of the specified symbol (function or variable) as an untyped pointer.
+    /// If the specified symbol does not exist in this library, nullptr is returned.
+    void* GetSymbolAddress(const char* name) const;
+
+    /// Obtains the address of the specified symbol, automatically casting to the correct type.
+    /// Returns true if the symbol was found and assigned, otherwise false.
+    template <typename T>
+    bool GetSymbol(const char* name, T* ptr) const {
+        *ptr = reinterpret_cast<T>(GetSymbolAddress(name));
+        return *ptr != nullptr;
+    }
+
+private:
+    /// Platform-dependent data type representing a dynamic library handle.
+    void* handle = nullptr;
+};
+
+} // namespace Common

src/common/file_util.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <array>
+#include <limits>
 #include <memory>
 #include <sstream>
 #include <unordered_map>
@@ -530,11 +531,11 @@ void CopyDir(const std::string& source_path, const std::string& dest_path) {
 std::optional<std::string> GetCurrentDir() {
 // Get the current working directory (getcwd uses malloc)
 #ifdef _WIN32
-    wchar_t* dir;
-    if (!(dir = _wgetcwd(nullptr, 0))) {
+    wchar_t* dir = _wgetcwd(nullptr, 0);
+    if (!dir) {
 #else
-    char* dir;
-    if (!(dir = getcwd(nullptr, 0))) {
+    char* dir = getcwd(nullptr, 0);
+    if (!dir) {
 #endif
         LOG_ERROR(Common_Filesystem, "GetCurrentDirectory failed: {}", GetLastErrorMsg());
         return {};
@@ -918,19 +919,22 @@ void IOFile::Swap(IOFile& other) noexcept {
 
 bool IOFile::Open(const std::string& filename, const char openmode[], int flags) {
     Close();
+    bool m_good;
 #ifdef _WIN32
     if (flags != 0) {
         m_file = _wfsopen(Common::UTF8ToUTF16W(filename).c_str(),
                           Common::UTF8ToUTF16W(openmode).c_str(), flags);
+        m_good = m_file != nullptr;
     } else {
-        _wfopen_s(&m_file, Common::UTF8ToUTF16W(filename).c_str(),
-                  Common::UTF8ToUTF16W(openmode).c_str());
+        m_good = _wfopen_s(&m_file, Common::UTF8ToUTF16W(filename).c_str(),
+                           Common::UTF8ToUTF16W(openmode).c_str()) == 0;
     }
 #else
-    m_file = fopen(filename.c_str(), openmode);
+    m_file = std::fopen(filename.c_str(), openmode);
+    m_good = m_file != nullptr;
 #endif
 
-    return IsOpen();
+    return m_good;
 }
 
 bool IOFile::Close() {
@@ -956,7 +960,7 @@ u64 IOFile::Tell() const {
     if (IsOpen())
         return ftello(m_file);
 
-    return -1;
+    return std::numeric_limits<u64>::max();
 }
 
 bool IOFile::Flush() {

src/common/thread.cpp

@@ -28,11 +28,8 @@ namespace Common {
 #ifdef _MSC_VER
 
 // Sets the debugger-visible name of the current thread.
-// Uses undocumented (actually, it is now documented) trick.
-// http://msdn.microsoft.com/library/default.asp?url=/library/en-us/vsdebug/html/vxtsksettingthreadname.asp
-
-// This is implemented much nicer in upcoming msvc++, see:
-// http://msdn.microsoft.com/en-us/library/xcb2z8hs(VS.100).aspx
+// Uses trick documented in:
+// https://docs.microsoft.com/en-us/visualstudio/debugger/how-to-set-a-thread-name-in-native-code
 void SetCurrentThreadName(const char* name) {
     static const DWORD MS_VC_EXCEPTION = 0x406D1388;
 
@@ -47,7 +44,7 @@ void SetCurrentThreadName(const char* name) {
 
     info.dwType = 0x1000;
     info.szName = name;
-    info.dwThreadID = -1; // dwThreadID;
+    info.dwThreadID = std::numeric_limits<DWORD>::max();
     info.dwFlags = 0;
 
     __try {

src/core/CMakeLists.txt

@@ -131,8 +131,6 @@ add_library(core STATIC
     frontend/framebuffer_layout.cpp
     frontend/framebuffer_layout.h
     frontend/input.h
-    frontend/scope_acquire_context.cpp
-    frontend/scope_acquire_context.h
     gdbstub/gdbstub.cpp
     gdbstub/gdbstub.h
     hardware_interrupt_manager.cpp
@@ -287,6 +285,18 @@ add_library(core STATIC
     hle/service/btm/btm.h
     hle/service/caps/caps.cpp
     hle/service/caps/caps.h
+    hle/service/caps/caps_a.cpp
+    hle/service/caps/caps_a.h
+    hle/service/caps/caps_c.cpp
+    hle/service/caps/caps_c.h
+    hle/service/caps/caps_u.cpp
+    hle/service/caps/caps_u.h
+    hle/service/caps/caps_sc.cpp
+    hle/service/caps/caps_sc.h
+    hle/service/caps/caps_ss.cpp
+    hle/service/caps/caps_ss.h
+    hle/service/caps/caps_su.cpp
+    hle/service/caps/caps_su.h
     hle/service/erpt/erpt.cpp
     hle/service/erpt/erpt.h
     hle/service/es/es.cpp
@@ -581,11 +591,8 @@ target_link_libraries(core PUBLIC common PRIVATE audio_core video_core)
 target_link_libraries(core PUBLIC Boost::boost PRIVATE fmt json-headers mbedtls opus unicorn)
 
 if (YUZU_ENABLE_BOXCAT)
-    get_directory_property(OPENSSL_LIBS
-        DIRECTORY ${PROJECT_SOURCE_DIR}/externals/libressl
-        DEFINITION OPENSSL_LIBS)
-    target_compile_definitions(core PRIVATE -DCPPHTTPLIB_OPENSSL_SUPPORT -DYUZU_ENABLE_BOXCAT)
-    target_link_libraries(core PRIVATE httplib json-headers ${OPENSSL_LIBS} zip)
+    target_compile_definitions(core PRIVATE -DYUZU_ENABLE_BOXCAT)
+    target_link_libraries(core PRIVATE httplib json-headers zip)
 endif()
 
 if (ENABLE_WEB_SERVICE)

src/core/core.cpp

@@ -24,7 +24,6 @@
 #include "core/file_sys/sdmc_factory.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
-#include "core/frontend/scope_acquire_context.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hardware_interrupt_manager.h"
 #include "core/hle/kernel/client_port.h"
@@ -168,13 +167,12 @@ struct System::Impl {
         Service::Init(service_manager, system);
         GDBStub::DeferStart();
 
-        renderer = VideoCore::CreateRenderer(emu_window, system);
-        if (!renderer->Init()) {
+        interrupt_manager = std::make_unique<Core::Hardware::InterruptManager>(system);
+        gpu_core = VideoCore::CreateGPU(emu_window, system);
+        if (!gpu_core) {
             return ResultStatus::ErrorVideoCore;
         }
-        interrupt_manager = std::make_unique<Core::Hardware::InterruptManager>(system);
-        gpu_core = VideoCore::CreateGPU(system);
-        renderer->Rasterizer().SetupDirtyFlags();
+        gpu_core->Renderer().Rasterizer().SetupDirtyFlags();
 
         is_powered_on = true;
         exit_lock = false;
@@ -186,8 +184,6 @@ struct System::Impl {
 
     ResultStatus Load(System& system, Frontend::EmuWindow& emu_window,
                       const std::string& filepath) {
-        Core::Frontend::ScopeAcquireContext acquire_context{emu_window};
-
         app_loader = Loader::GetLoader(GetGameFileFromPath(virtual_filesystem, filepath));
         if (!app_loader) {
             LOG_CRITICAL(Core, "Failed to obtain loader for {}!", filepath);
@@ -216,10 +212,6 @@ struct System::Impl {
         AddGlueRegistrationForProcess(*app_loader, *main_process);
         kernel.MakeCurrentProcess(main_process.get());
 
-        // Main process has been loaded and been made current.
-        // Begin GPU and CPU execution.
-        gpu_core->Start();
-
         // Initialize cheat engine
         if (cheat_engine) {
             cheat_engine->Initialize();
@@ -277,7 +269,6 @@ struct System::Impl {
         }
 
         // Shutdown emulation session
-        renderer.reset();
         GDBStub::Shutdown();
         Service::Shutdown();
         service_manager.reset();
@@ -353,7 +344,6 @@ struct System::Impl {
     Service::FileSystem::FileSystemController fs_controller;
     /// AppLoader used to load the current executing application
     std::unique_ptr<Loader::AppLoader> app_loader;
-    std::unique_ptr<VideoCore::RendererBase> renderer;
     std::unique_ptr<Tegra::GPU> gpu_core;
     std::unique_ptr<Hardware::InterruptManager> interrupt_manager;
     Memory::Memory memory;
@@ -536,11 +526,11 @@ const Core::Hardware::InterruptManager& System::InterruptManager() const {
 }
 
 VideoCore::RendererBase& System::Renderer() {
-    return *impl->renderer;
+    return impl->gpu_core->Renderer();
 }
 
 const VideoCore::RendererBase& System::Renderer() const {
-    return *impl->renderer;
+    return impl->gpu_core->Renderer();
 }
 
 Kernel::KernelCore& System::Kernel() {

src/core/file_sys/patch_manager.cpp

@@ -348,6 +348,12 @@ static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType t
         if (ext_dir != nullptr)
             layers_ext.push_back(std::move(ext_dir));
     }
+
+    // When there are no layers to apply, return early as there is no need to rebuild the RomFS
+    if (layers.empty() && layers_ext.empty()) {
+        return;
+    }
+
     layers.push_back(std::move(extracted));
 
     auto layered = LayeredVfsDirectory::MakeLayeredDirectory(std::move(layers));

src/core/file_sys/romfs.cpp

@@ -5,6 +5,7 @@
 #include <memory>
 
 #include "common/common_types.h"
+#include "common/string_util.h"
 #include "common/swap.h"
 #include "core/file_sys/fsmitm_romfsbuild.h"
 #include "core/file_sys/romfs.h"
@@ -126,7 +127,7 @@ VirtualDir ExtractRomFS(VirtualFile file, RomFSExtractionType type) {
         return out->GetSubdirectories().front();
 
     while (out->GetSubdirectories().size() == 1 && out->GetFiles().empty()) {
-        if (out->GetSubdirectories().front()->GetName() == "data" &&
+        if (Common::ToLower(out->GetSubdirectories().front()->GetName()) == "data" &&
             type == RomFSExtractionType::Truncated)
             break;
         out = out->GetSubdirectories().front();

src/core/frontend/emu_window.h

@@ -12,20 +12,49 @@
 
 namespace Core::Frontend {
 
+/// Information for the Graphics Backends signifying what type of screen pointer is in
+/// WindowInformation
+enum class WindowSystemType {
+    Headless,
+    Windows,
+    X11,
+    Wayland,
+};
+
 /**
- * Represents a graphics context that can be used for background computation or drawing. If the
- * graphics backend doesn't require the context, then the implementation of these methods can be
- * stubs
+ * Represents a drawing context that supports graphics operations.
  */
 class GraphicsContext {
 public:
     virtual ~GraphicsContext();
 
+    /// Inform the driver to swap the front/back buffers and present the current image
+    virtual void SwapBuffers() {}
+
     /// Makes the graphics context current for the caller thread
-    virtual void MakeCurrent() = 0;
+    virtual void MakeCurrent() {}
 
     /// Releases (dunno if this is the "right" word) the context from the caller thread
-    virtual void DoneCurrent() = 0;
+    virtual void DoneCurrent() {}
+
+    class Scoped {
+    public:
+        explicit Scoped(GraphicsContext& context_) : context(context_) {
+            context.MakeCurrent();
+        }
+        ~Scoped() {
+            context.DoneCurrent();
+        }
+
+    private:
+        GraphicsContext& context;
+    };
+
+    /// Calls MakeCurrent on the context and calls DoneCurrent when the scope for the returned value
+    /// ends
+    Scoped Acquire() {
+        return Scoped{*this};
+    }
 };
 
 /**
@@ -46,7 +75,7 @@ public:
  * - DO NOT TREAT THIS CLASS AS A GUI TOOLKIT ABSTRACTION LAYER. That's not what it is. Please
  *   re-read the upper points again and think about it if you don't see this.
  */
-class EmuWindow : public GraphicsContext {
+class EmuWindow {
 public:
     /// Data structure to store emuwindow configuration
     struct WindowConfig {
@@ -56,29 +85,34 @@ public:
         std::pair<unsigned, unsigned> min_client_area_size;
     };
 
+    /// Data describing host window system information
+    struct WindowSystemInfo {
+        // Window system type. Determines which GL context or Vulkan WSI is used.
+        WindowSystemType type = WindowSystemType::Headless;
+
+        // Connection to a display server. This is used on X11 and Wayland platforms.
+        void* display_connection = nullptr;
+
+        // Render surface. This is a pointer to the native window handle, which depends
+        // on the platform. e.g. HWND for Windows, Window for X11. If the surface is
+        // set to nullptr, the video backend will run in headless mode.
+        void* render_surface = nullptr;
+
+        // Scale of the render surface. For hidpi systems, this will be >1.
+        float render_surface_scale = 1.0f;
+    };
+
     /// Polls window events
     virtual void PollEvents() = 0;
 
     /**
-     * Returns a GraphicsContext that the frontend provides that is shared with the emu window. This
-     * context can be used from other threads for background graphics computation. If the frontend
-     * is using a graphics backend that doesn't need anything specific to run on a different thread,
-     * then it can use a stubbed implemenation for GraphicsContext.
-     *
-     * If the return value is null, then the core should assume that the frontend cannot provide a
-     * Shared Context
+     * Returns a GraphicsContext that the frontend provides to be used for rendering.
      */
-    virtual std::unique_ptr<GraphicsContext> CreateSharedContext() const {
-        return nullptr;
-    }
+    virtual std::unique_ptr<GraphicsContext> CreateSharedContext() const = 0;
 
     /// Returns if window is shown (not minimized)
     virtual bool IsShown() const = 0;
 
-    /// Retrieves Vulkan specific handlers from the window
-    virtual void RetrieveVulkanHandlers(void* get_instance_proc_addr, void* instance,
-                                        void* surface) const = 0;
-
     /**
      * Signal that a touch pressed event has occurred (e.g. mouse click pressed)
      * @param framebuffer_x Framebuffer x-coordinate that was pressed
@@ -115,6 +149,13 @@ public:
         config = val;
     }
 
+    /**
+     * Returns system information about the drawing area.
+     */
+    const WindowSystemInfo& GetWindowInfo() const {
+        return window_info;
+    }
+
     /**
      * Gets the framebuffer layout (width, height, and screen regions)
      * @note This method is thread-safe
@@ -130,7 +171,7 @@ public:
     void UpdateCurrentFramebufferLayout(unsigned width, unsigned height);
 
 protected:
-    EmuWindow();
+    explicit EmuWindow();
     virtual ~EmuWindow();
 
     /**
@@ -167,6 +208,8 @@ protected:
         client_area_height = size.second;
     }
 
+    WindowSystemInfo window_info;
+
 private:
     /**
      * Handler called when the minimal client area was requested to be changed via SetConfig.

src/core/frontend/scope_acquire_context.cpp

@@ -1,18 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "core/frontend/emu_window.h"
-#include "core/frontend/scope_acquire_context.h"
-
-namespace Core::Frontend {
-
-ScopeAcquireContext::ScopeAcquireContext(Core::Frontend::GraphicsContext& context)
-    : context{context} {
-    context.MakeCurrent();
-}
-ScopeAcquireContext::~ScopeAcquireContext() {
-    context.DoneCurrent();
-}
-
-} // namespace Core::Frontend

src/core/frontend/scope_acquire_context.h

@@ -1,23 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include "common/common_types.h"
-
-namespace Core::Frontend {
-
-class GraphicsContext;
-
-/// Helper class to acquire/release window context within a given scope
-class ScopeAcquireContext : NonCopyable {
-public:
-    explicit ScopeAcquireContext(Core::Frontend::GraphicsContext& context);
-    ~ScopeAcquireContext();
-
-private:
-    Core::Frontend::GraphicsContext& context;
-};
-
-} // namespace Core::Frontend

src/core/hle/kernel/kernel.cpp

@@ -103,7 +103,7 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] s64 cycles_
 
 struct KernelCore::Impl {
     explicit Impl(Core::System& system, KernelCore& kernel)
-        : system{system}, global_scheduler{kernel}, synchronization{system}, time_manager{system} {}
+        : global_scheduler{kernel}, synchronization{system}, time_manager{system}, system{system} {}
 
     void Initialize(KernelCore& kernel) {
         Shutdown();

src/core/hle/service/caps/caps.cpp

@@ -2,168 +2,24 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <memory>
-
 #include "core/hle/service/caps/caps.h"
+#include "core/hle/service/caps/caps_a.h"
+#include "core/hle/service/caps/caps_c.h"
+#include "core/hle/service/caps/caps_sc.h"
+#include "core/hle/service/caps/caps_ss.h"
+#include "core/hle/service/caps/caps_su.h"
+#include "core/hle/service/caps/caps_u.h"
 #include "core/hle/service/service.h"
-#include "core/hle/service/sm/sm.h"
 
 namespace Service::Capture {
 
-class CAPS_A final : public ServiceFramework<CAPS_A> {
-public:
-    explicit CAPS_A() : ServiceFramework{"caps:a"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {0, nullptr, "GetAlbumFileCount"},
-            {1, nullptr, "GetAlbumFileList"},
-            {2, nullptr, "LoadAlbumFile"},
-            {3, nullptr, "DeleteAlbumFile"},
-            {4, nullptr, "StorageCopyAlbumFile"},
-            {5, nullptr, "IsAlbumMounted"},
-            {6, nullptr, "GetAlbumUsage"},
-            {7, nullptr, "GetAlbumFileSize"},
-            {8, nullptr, "LoadAlbumFileThumbnail"},
-            {9, nullptr, "LoadAlbumScreenShotImage"},
-            {10, nullptr, "LoadAlbumScreenShotThumbnailImage"},
-            {11, nullptr, "GetAlbumEntryFromApplicationAlbumEntry"},
-            {12, nullptr, "Unknown12"},
-            {13, nullptr, "Unknown13"},
-            {14, nullptr, "Unknown14"},
-            {15, nullptr, "Unknown15"},
-            {16, nullptr, "Unknown16"},
-            {17, nullptr, "Unknown17"},
-            {18, nullptr, "Unknown18"},
-            {202, nullptr, "SaveEditedScreenShot"},
-            {301, nullptr, "GetLastThumbnail"},
-            {401, nullptr, "GetAutoSavingStorage"},
-            {501, nullptr, "GetRequiredStorageSpaceSizeToCopyAll"},
-            {1001, nullptr, "Unknown1001"},
-            {1002, nullptr, "Unknown1002"},
-            {1003, nullptr, "Unknown1003"},
-            {8001, nullptr, "ForceAlbumUnmounted"},
-            {8002, nullptr, "ResetAlbumMountStatus"},
-            {8011, nullptr, "RefreshAlbumCache"},
-            {8012, nullptr, "GetAlbumCache"},
-            {8013, nullptr, "Unknown8013"},
-            {8021, nullptr, "GetAlbumEntryFromApplicationAlbumEntryAruid"},
-            {10011, nullptr, "SetInternalErrorConversionEnabled"},
-            {50000, nullptr, "Unknown50000"},
-            {60002, nullptr, "Unknown60002"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
-class CAPS_C final : public ServiceFramework<CAPS_C> {
-public:
-    explicit CAPS_C() : ServiceFramework{"caps:c"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {33, nullptr, "Unknown33"},
-            {2001, nullptr, "Unknown2001"},
-            {2002, nullptr, "Unknown2002"},
-            {2011, nullptr, "Unknown2011"},
-            {2012, nullptr, "Unknown2012"},
-            {2013, nullptr, "Unknown2013"},
-            {2014, nullptr, "Unknown2014"},
-            {2101, nullptr, "Unknown2101"},
-            {2102, nullptr, "Unknown2102"},
-            {2201, nullptr, "Unknown2201"},
-            {2301, nullptr, "Unknown2301"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
-class CAPS_SC final : public ServiceFramework<CAPS_SC> {
-public:
-    explicit CAPS_SC() : ServiceFramework{"caps:sc"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {1, nullptr, "Unknown1"},
-            {2, nullptr, "Unknown2"},
-            {1001, nullptr, "Unknown3"},
-            {1002, nullptr, "Unknown4"},
-            {1003, nullptr, "Unknown5"},
-            {1011, nullptr, "Unknown6"},
-            {1012, nullptr, "Unknown7"},
-            {1201, nullptr, "Unknown8"},
-            {1202, nullptr, "Unknown9"},
-            {1203, nullptr, "Unknown10"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
-class CAPS_SS final : public ServiceFramework<CAPS_SS> {
-public:
-    explicit CAPS_SS() : ServiceFramework{"caps:ss"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {201, nullptr, "Unknown1"},
-            {202, nullptr, "Unknown2"},
-            {203, nullptr, "Unknown3"},
-            {204, nullptr, "Unknown4"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
-class CAPS_SU final : public ServiceFramework<CAPS_SU> {
-public:
-    explicit CAPS_SU() : ServiceFramework{"caps:su"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {201, nullptr, "SaveScreenShot"},
-            {203, nullptr, "SaveScreenShotEx0"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
-class CAPS_U final : public ServiceFramework<CAPS_U> {
-public:
-    explicit CAPS_U() : ServiceFramework{"caps:u"} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {32, nullptr, "SetShimLibraryVersion"},
-            {102, nullptr, "GetAlbumFileListByAruid"},
-            {103, nullptr, "DeleteAlbumFileByAruid"},
-            {104, nullptr, "GetAlbumFileSizeByAruid"},
-            {105, nullptr, "DeleteAlbumFileByAruidForDebug"},
-            {110, nullptr, "LoadAlbumScreenShotImageByAruid"},
-            {120, nullptr, "LoadAlbumScreenShotThumbnailImageByAruid"},
-            {130, nullptr, "PrecheckToCreateContentsByAruid"},
-            {140, nullptr, "GetAlbumFileList1AafeAruidDeprecated"},
-            {141, nullptr, "GetAlbumFileList2AafeUidAruidDeprecated"},
-            {142, nullptr, "GetAlbumFileList3AaeAruid"},
-            {143, nullptr, "GetAlbumFileList4AaeUidAruid"},
-            {60002, nullptr, "OpenAccessorSessionForApplication"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-};
-
 void InstallInterfaces(SM::ServiceManager& sm) {
     std::make_shared<CAPS_A>()->InstallAsService(sm);
     std::make_shared<CAPS_C>()->InstallAsService(sm);
+    std::make_shared<CAPS_U>()->InstallAsService(sm);
     std::make_shared<CAPS_SC>()->InstallAsService(sm);
     std::make_shared<CAPS_SS>()->InstallAsService(sm);
     std::make_shared<CAPS_SU>()->InstallAsService(sm);
-    std::make_shared<CAPS_U>()->InstallAsService(sm);
 }
 
 } // namespace Service::Capture

src/core/hle/service/caps/caps.h

@@ -4,12 +4,83 @@
 
 #pragma once
 
+#include "core/hle/service/service.h"
+
 namespace Service::SM {
 class ServiceManager;
 }
 
 namespace Service::Capture {
 
+enum AlbumImageOrientation {
+    Orientation0 = 0,
+    Orientation1 = 1,
+    Orientation2 = 2,
+    Orientation3 = 3,
+};
+
+enum AlbumReportOption {
+    Disable = 0,
+    Enable = 1,
+};
+
+enum ContentType : u8 {
+    Screenshot = 0,
+    Movie = 1,
+    ExtraMovie = 3,
+};
+
+enum AlbumStorage : u8 {
+    NAND = 0,
+    SD = 1,
+};
+
+struct AlbumFileDateTime {
+    u16 year;
+    u8 month;
+    u8 day;
+    u8 hour;
+    u8 minute;
+    u8 second;
+    u8 uid;
+};
+
+struct AlbumEntry {
+    u64 size;
+    u64 application_id;
+    AlbumFileDateTime datetime;
+    AlbumStorage storage;
+    ContentType content;
+    u8 padding[6];
+};
+
+struct AlbumFileEntry {
+    u64 size;
+    u64 hash;
+    AlbumFileDateTime datetime;
+    AlbumStorage storage;
+    ContentType content;
+    u8 padding[5];
+    u8 unknown;
+};
+
+struct ApplicationAlbumEntry {
+    u64 size;
+    u64 hash;
+    AlbumFileDateTime datetime;
+    AlbumStorage storage;
+    ContentType content;
+    u8 padding[5];
+    u8 unknown;
+};
+
+struct ApplicationAlbumFileEntry {
+    ApplicationAlbumEntry entry;
+    AlbumFileDateTime datetime;
+    u64 unknown;
+};
+
+/// Registers all Capture services with the specified service manager.
 void InstallInterfaces(SM::ServiceManager& sm);
 
 } // namespace Service::Capture

src/core/hle/service/caps/caps_a.cpp

@@ -0,0 +1,78 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/service/caps/caps_a.h"
+
+namespace Service::Capture {
+
+class IAlbumAccessorSession final : public ServiceFramework<IAlbumAccessorSession> {
+public:
+    explicit IAlbumAccessorSession() : ServiceFramework{"IAlbumAccessorSession"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {2001, nullptr, "OpenAlbumMovieReadStream"},
+            {2002, nullptr, "CloseAlbumMovieReadStream"},
+            {2003, nullptr, "GetAlbumMovieReadStreamMovieDataSize"},
+            {2004, nullptr, "ReadMovieDataFromAlbumMovieReadStream"},
+            {2005, nullptr, "GetAlbumMovieReadStreamBrokenReason"},
+            {2006, nullptr, "GetAlbumMovieReadStreamImageDataSize"},
+            {2007, nullptr, "ReadImageDataFromAlbumMovieReadStream"},
+            {2008, nullptr, "ReadFileAttributeFromAlbumMovieReadStream"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+};
+
+CAPS_A::CAPS_A() : ServiceFramework("caps:a") {
+    // clang-format off
+    static const FunctionInfo functions[] = {
+        {0, nullptr, "GetAlbumFileCount"},
+        {1, nullptr, "GetAlbumFileList"},
+        {2, nullptr, "LoadAlbumFile"},
+        {3, nullptr, "DeleteAlbumFile"},
+        {4, nullptr, "StorageCopyAlbumFile"},
+        {5, nullptr, "IsAlbumMounted"},
+        {6, nullptr, "GetAlbumUsage"},
+        {7, nullptr, "GetAlbumFileSize"},
+        {8, nullptr, "LoadAlbumFileThumbnail"},
+        {9, nullptr, "LoadAlbumScreenShotImage"},
+        {10, nullptr, "LoadAlbumScreenShotThumbnailImage"},
+        {11, nullptr, "GetAlbumEntryFromApplicationAlbumEntry"},
+        {12, nullptr, "LoadAlbumScreenShotImageEx"},
+        {13, nullptr, "LoadAlbumScreenShotThumbnailImageEx"},
+        {14, nullptr, "LoadAlbumScreenShotImageEx0"},
+        {15, nullptr, "GetAlbumUsage3"},
+        {16, nullptr, "GetAlbumMountResult"},
+        {17, nullptr, "GetAlbumUsage16"},
+        {18, nullptr, "Unknown18"},
+        {100, nullptr, "GetAlbumFileCountEx0"},
+        {101, nullptr, "GetAlbumFileListEx0"},
+        {202, nullptr, "SaveEditedScreenShot"},
+        {301, nullptr, "GetLastThumbnail"},
+        {302, nullptr, "GetLastOverlayMovieThumbnail"},
+        {401, nullptr, "GetAutoSavingStorage"},
+        {501, nullptr, "GetRequiredStorageSpaceSizeToCopyAll"},
+        {1001, nullptr, "LoadAlbumScreenShotThumbnailImageEx0"},
+        {1002, nullptr, "LoadAlbumScreenShotImageEx1"},
+        {1003, nullptr, "LoadAlbumScreenShotThumbnailImageEx1"},
+        {8001, nullptr, "ForceAlbumUnmounted"},
+        {8002, nullptr, "ResetAlbumMountStatus"},
+        {8011, nullptr, "RefreshAlbumCache"},
+        {8012, nullptr, "GetAlbumCache"},
+        {8013, nullptr, "GetAlbumCacheEx"},
+        {8021, nullptr, "GetAlbumEntryFromApplicationAlbumEntryAruid"},
+        {10011, nullptr, "SetInternalErrorConversionEnabled"},
+        {50000, nullptr, "LoadMakerNoteInfoForDebug"},
+        {60002, nullptr, "OpenAccessorSession"},
+    };
+    // clang-format on
+
+    RegisterHandlers(functions);
+}
+
+CAPS_A::~CAPS_A() = default;
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_a.h

@@ -0,0 +1,21 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::Capture {
+
+class CAPS_A final : public ServiceFramework<CAPS_A> {
+public:
+    explicit CAPS_A();
+    ~CAPS_A() override;
+};
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_c.cpp

@@ -0,0 +1,75 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/service/caps/caps_c.h"
+
+namespace Service::Capture {
+
+class IAlbumControlSession final : public ServiceFramework<IAlbumControlSession> {
+public:
+    explicit IAlbumControlSession() : ServiceFramework{"IAlbumControlSession"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {2001, nullptr, "OpenAlbumMovieReadStream"},
+            {2002, nullptr, "CloseAlbumMovieReadStream"},
+            {2003, nullptr, "GetAlbumMovieReadStreamMovieDataSize"},
+            {2004, nullptr, "ReadMovieDataFromAlbumMovieReadStream"},
+            {2005, nullptr, "GetAlbumMovieReadStreamBrokenReason"},
+            {2006, nullptr, "GetAlbumMovieReadStreamImageDataSize"},
+            {2007, nullptr, "ReadImageDataFromAlbumMovieReadStream"},
+            {2008, nullptr, "ReadFileAttributeFromAlbumMovieReadStream"},
+            {2401, nullptr, "OpenAlbumMovieWriteStream"},
+            {2402, nullptr, "FinishAlbumMovieWriteStream"},
+            {2403, nullptr, "CommitAlbumMovieWriteStream"},
+            {2404, nullptr, "DiscardAlbumMovieWriteStream"},
+            {2405, nullptr, "DiscardAlbumMovieWriteStreamNoDelete"},
+            {2406, nullptr, "CommitAlbumMovieWriteStreamEx"},
+            {2411, nullptr, "StartAlbumMovieWriteStreamDataSection"},
+            {2412, nullptr, "EndAlbumMovieWriteStreamDataSection"},
+            {2413, nullptr, "StartAlbumMovieWriteStreamMetaSection"},
+            {2414, nullptr, "EndAlbumMovieWriteStreamMetaSection"},
+            {2421, nullptr, "ReadDataFromAlbumMovieWriteStream"},
+            {2422, nullptr, "WriteDataToAlbumMovieWriteStream"},
+            {2424, nullptr, "WriteMetaToAlbumMovieWriteStream"},
+            {2431, nullptr, "GetAlbumMovieWriteStreamBrokenReason"},
+            {2433, nullptr, "GetAlbumMovieWriteStreamDataSize"},
+            {2434, nullptr, "SetAlbumMovieWriteStreamDataSize"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+};
+
+CAPS_C::CAPS_C() : ServiceFramework("caps:c") {
+    // clang-format off
+    static const FunctionInfo functions[] = {
+        {1, nullptr, "CaptureRawImage"},
+        {2, nullptr, "CaptureRawImageWithTimeout"},
+        {33, nullptr, "Unknown33"},
+        {1001, nullptr, "RequestTakingScreenShot"},
+        {1002, nullptr, "RequestTakingScreenShotWithTimeout"},
+        {1011, nullptr, "NotifyTakingScreenShotRefused"},
+        {2001, nullptr, "NotifyAlbumStorageIsAvailable"},
+        {2002, nullptr, "NotifyAlbumStorageIsUnavailable"},
+        {2011, nullptr, "RegisterAppletResourceUserId"},
+        {2012, nullptr, "UnregisterAppletResourceUserId"},
+        {2013, nullptr, "GetApplicationIdFromAruid"},
+        {2014, nullptr, "CheckApplicationIdRegistered"},
+        {2101, nullptr, "GenerateCurrentAlbumFileId"},
+        {2102, nullptr, "GenerateApplicationAlbumEntry"},
+        {2201, nullptr, "SaveAlbumScreenShotFile"},
+        {2202, nullptr, "SaveAlbumScreenShotFileEx"},
+        {2301, nullptr, "SetOverlayScreenShotThumbnailData"},
+        {2302, nullptr, "SetOverlayMovieThumbnailData"},
+        {60001, nullptr, "OpenControlSession"},
+    };
+    // clang-format on
+
+    RegisterHandlers(functions);
+}
+
+CAPS_C::~CAPS_C() = default;
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_c.h

@@ -0,0 +1,21 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::Capture {
+
+class CAPS_C final : public ServiceFramework<CAPS_C> {
+public:
+    explicit CAPS_C();
+    ~CAPS_C() override;
+};
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_sc.cpp

@@ -0,0 +1,40 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/service/caps/caps_sc.h"
+
+namespace Service::Capture {
+
+CAPS_SC::CAPS_SC() : ServiceFramework("caps:sc") {
+    // clang-format off
+    static const FunctionInfo functions[] = {
+        {1, nullptr, "CaptureRawImage"},
+        {2, nullptr, "CaptureRawImageWithTimeout"},
+        {3, nullptr, "AttachSharedBuffer"},
+        {5, nullptr, "CaptureRawImageToAttachedSharedBuffer"},
+        {210, nullptr, "Unknown210"},
+        {1001, nullptr, "RequestTakingScreenShot"},
+        {1002, nullptr, "RequestTakingScreenShotWithTimeout"},
+        {1003, nullptr, "RequestTakingScreenShotEx"},
+        {1004, nullptr, "RequestTakingScreenShotEx1"},
+        {1009, nullptr, "CancelTakingScreenShot"},
+        {1010, nullptr, "SetTakingScreenShotCancelState"},
+        {1011, nullptr, "NotifyTakingScreenShotRefused"},
+        {1012, nullptr, "NotifyTakingScreenShotFailed"},
+        {1101, nullptr, "SetupOverlayMovieThumbnail"},
+        {1106, nullptr, "Unknown1106"},
+        {1107, nullptr, "Unknown1107"},
+        {1201, nullptr, "OpenRawScreenShotReadStream"},
+        {1202, nullptr, "CloseRawScreenShotReadStream"},
+        {1203, nullptr, "ReadRawScreenShotReadStream"},
+        {1204, nullptr, "Unknown1204"},
+    };
+    // clang-format on
+
+    RegisterHandlers(functions);
+}
+
+CAPS_SC::~CAPS_SC() = default;
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_sc.h

@@ -0,0 +1,21 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::Capture {
+
+class CAPS_SC final : public ServiceFramework<CAPS_SC> {
+public:
+    explicit CAPS_SC();
+    ~CAPS_SC() override;
+};
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_ss.cpp

@@ -0,0 +1,26 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/service/caps/caps_ss.h"
+
+namespace Service::Capture {
+
+CAPS_SS::CAPS_SS() : ServiceFramework("caps:ss") {
+    // clang-format off
+    static const FunctionInfo functions[] = {
+        {201, nullptr, "SaveScreenShot"},
+        {202, nullptr, "SaveEditedScreenShot"},
+        {203, nullptr, "SaveScreenShotEx0"},
+        {204, nullptr, "SaveEditedScreenShotEx0"},
+        {206, nullptr, "Unknown206"},
+        {208, nullptr, "SaveScreenShotOfMovieEx1"},
+    };
+    // clang-format on
+
+    RegisterHandlers(functions);
+}
+
+CAPS_SS::~CAPS_SS() = default;
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_ss.h

@@ -0,0 +1,21 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::Capture {
+
+class CAPS_SS final : public ServiceFramework<CAPS_SS> {
+public:
+    explicit CAPS_SS();
+    ~CAPS_SS() override;
+};
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_su.cpp

@@ -0,0 +1,22 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/service/caps/caps_su.h"
+
+namespace Service::Capture {
+
+CAPS_SU::CAPS_SU() : ServiceFramework("caps:su") {
+    // clang-format off
+    static const FunctionInfo functions[] = {
+        {201, nullptr, "SaveScreenShot"},
+        {203, nullptr, "SaveScreenShotEx0"},
+    };
+    // clang-format on
+
+    RegisterHandlers(functions);
+}
+
+CAPS_SU::~CAPS_SU() = default;
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_su.h

@@ -0,0 +1,21 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::Capture {
+
+class CAPS_SU final : public ServiceFramework<CAPS_SU> {
+public:
+    explicit CAPS_SU();
+    ~CAPS_SU() override;
+};
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_u.cpp

@@ -0,0 +1,76 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/logging/log.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/service/caps/caps.h"
+#include "core/hle/service/caps/caps_u.h"
+
+namespace Service::Capture {
+
+class IAlbumAccessorApplicationSession final
+    : public ServiceFramework<IAlbumAccessorApplicationSession> {
+public:
+    explicit IAlbumAccessorApplicationSession()
+        : ServiceFramework{"IAlbumAccessorApplicationSession"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {2001, nullptr, "OpenAlbumMovieReadStream"},
+            {2002, nullptr, "CloseAlbumMovieReadStream"},
+            {2003, nullptr, "GetAlbumMovieReadStreamMovieDataSize"},
+            {2004, nullptr, "ReadMovieDataFromAlbumMovieReadStream"},
+            {2005, nullptr, "GetAlbumMovieReadStreamBrokenReason"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+};
+
+CAPS_U::CAPS_U() : ServiceFramework("caps:u") {
+    // clang-format off
+    static const FunctionInfo functions[] = {
+        {31, nullptr, "GetShimLibraryVersion"},
+        {32, nullptr, "SetShimLibraryVersion"},
+        {102, &CAPS_U::GetAlbumContentsFileListForApplication, "GetAlbumContentsFileListForApplication"},
+        {103, nullptr, "DeleteAlbumContentsFileForApplication"},
+        {104, nullptr, "GetAlbumContentsFileSizeForApplication"},
+        {105, nullptr, "DeleteAlbumFileByAruidForDebug"},
+        {110, nullptr, "LoadAlbumContentsFileScreenShotImageForApplication"},
+        {120, nullptr, "LoadAlbumContentsFileThumbnailImageForApplication"},
+        {130, nullptr, "PrecheckToCreateContentsForApplication"},
+        {140, nullptr, "GetAlbumFileList1AafeAruidDeprecated"},
+        {141, nullptr, "GetAlbumFileList2AafeUidAruidDeprecated"},
+        {142, nullptr, "GetAlbumFileList3AaeAruid"},
+        {143, nullptr, "GetAlbumFileList4AaeUidAruid"},
+        {60002, nullptr, "OpenAccessorSessionForApplication"},
+    };
+    // clang-format on
+
+    RegisterHandlers(functions);
+}
+
+CAPS_U::~CAPS_U() = default;
+
+void CAPS_U::GetAlbumContentsFileListForApplication(Kernel::HLERequestContext& ctx) {
+    // Takes a type-0x6 output buffer containing an array of ApplicationAlbumFileEntry, a PID, an
+    // u8 ContentType, two s64s, and an u64 AppletResourceUserId. Returns an output u64 for total
+    // output entries (which is copied to a s32 by official SW).
+    IPC::RequestParser rp{ctx};
+    [[maybe_unused]] const auto application_album_file_entries = rp.PopRaw<std::array<u8, 0x30>>();
+    const auto pid = rp.Pop<s32>();
+    const auto content_type = rp.PopRaw<ContentType>();
+    [[maybe_unused]] const auto start_datetime = rp.PopRaw<AlbumFileDateTime>();
+    [[maybe_unused]] const auto end_datetime = rp.PopRaw<AlbumFileDateTime>();
+    const auto applet_resource_user_id = rp.Pop<u64>();
+    LOG_WARNING(Service_Capture,
+                "(STUBBED) called. pid={}, content_type={}, applet_resource_user_id={}", pid,
+                content_type, applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+    rb.Push<s32>(0);
+}
+
+} // namespace Service::Capture

src/core/hle/service/caps/caps_u.h

@@ -0,0 +1,24 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::Capture {
+
+class CAPS_U final : public ServiceFramework<CAPS_U> {
+public:
+    explicit CAPS_U();
+    ~CAPS_U() override;
+
+private:
+    void GetAlbumContentsFileListForApplication(Kernel::HLERequestContext& ctx);
+};
+
+} // namespace Service::Capture

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

@@ -27,7 +27,7 @@ public:
             {10110, nullptr, "GetFriendProfileImage"},
             {10200, nullptr, "SendFriendRequestForApplication"},
             {10211, nullptr, "AddFacedFriendRequestForApplication"},
-            {10400, nullptr, "GetBlockedUserListIds"},
+            {10400, &IFriendService::GetBlockedUserListIds, "GetBlockedUserListIds"},
             {10500, nullptr, "GetProfileList"},
             {10600, nullptr, "DeclareOpenOnlinePlaySession"},
             {10601, &IFriendService::DeclareCloseOnlinePlaySession, "DeclareCloseOnlinePlaySession"},
@@ -121,6 +121,15 @@ private:
     };
     static_assert(sizeof(SizedFriendFilter) == 0x10, "SizedFriendFilter is an invalid size");
 
+    void GetBlockedUserListIds(Kernel::HLERequestContext& ctx) {
+        // This is safe to stub, as there should be no adverse consequences from reporting no
+        // blocked users.
+        LOG_WARNING(Service_ACC, "(STUBBED) called");
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u32>(0); // Indicates there are no blocked users
+    }
+
     void DeclareCloseOnlinePlaySession(Kernel::HLERequestContext& ctx) {
         // Stub used by Splatoon 2
         LOG_WARNING(Service_ACC, "(STUBBED) called");

src/core/hle/service/ldr/ldr.cpp

@@ -342,17 +342,27 @@ public:
             return;
         }
 
-        ASSERT(
-            vm_manager
-                .MirrorMemory(*map_address, nro_address, nro_size, Kernel::MemoryState::ModuleCode)
-                .IsSuccess());
+        // Mark text and read-only region as ModuleCode
+        ASSERT(vm_manager
+                   .MirrorMemory(*map_address, nro_address, header.text_size + header.ro_size,
+                                 Kernel::MemoryState::ModuleCode)
+                   .IsSuccess());
+        // Mark read/write region as ModuleCodeData, which is necessary if this region is used for
+        // TransferMemory (e.g. Final Fantasy VIII Remastered does this)
+        ASSERT(vm_manager
+                   .MirrorMemory(*map_address + header.rw_offset, nro_address + header.rw_offset,
+                                 header.rw_size, Kernel::MemoryState::ModuleCodeData)
+                   .IsSuccess());
+        // Revoke permissions from the old memory region
         ASSERT(vm_manager.ReprotectRange(nro_address, nro_size, Kernel::VMAPermission::None)
                    .IsSuccess());
 
         if (bss_size > 0) {
+            // Mark BSS region as ModuleCodeData, which is necessary if this region is used for
+            // TransferMemory (e.g. Final Fantasy VIII Remastered does this)
             ASSERT(vm_manager
                        .MirrorMemory(*map_address + nro_size, bss_address, bss_size,
-                                     Kernel::MemoryState::ModuleCode)
+                                     Kernel::MemoryState::ModuleCodeData)
                        .IsSuccess());
             ASSERT(vm_manager.ReprotectRange(bss_address, bss_size, Kernel::VMAPermission::None)
                        .IsSuccess());

src/core/hle/service/nvflinger/buffer_queue.cpp

@@ -28,6 +28,7 @@ void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer)
     buffer.slot = slot;
     buffer.igbp_buffer = igbp_buffer;
     buffer.status = Buffer::Status::Free;
+    free_buffers.push_back(slot);
 
     queue.emplace_back(buffer);
     buffer_wait_event.writable->Signal();
@@ -35,16 +36,37 @@ void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer)
 
 std::optional<std::pair<u32, Service::Nvidia::MultiFence*>> BufferQueue::DequeueBuffer(u32 width,
                                                                                        u32 height) {
-    auto itr = std::find_if(queue.begin(), queue.end(), [&](const Buffer& buffer) {
-        // Only consider free buffers. Buffers become free once again after they've been Acquired
-        // and Released by the compositor, see the NVFlinger::Compose method.
-        if (buffer.status != Buffer::Status::Free) {
-            return false;
-        }
 
-        // Make sure that the parameters match.
-        return buffer.igbp_buffer.width == width && buffer.igbp_buffer.height == height;
-    });
+    if (free_buffers.empty()) {
+        return {};
+    }
+
+    auto f_itr = free_buffers.begin();
+    auto itr = queue.end();
+
+    while (f_itr != free_buffers.end()) {
+        auto slot = *f_itr;
+        itr = std::find_if(queue.begin(), queue.end(), [&](const Buffer& buffer) {
+            // Only consider free buffers. Buffers become free once again after they've been
+            // Acquired and Released by the compositor, see the NVFlinger::Compose method.
+            if (buffer.status != Buffer::Status::Free) {
+                return false;
+            }
+
+            if (buffer.slot != slot) {
+                return false;
+            }
+
+            // Make sure that the parameters match.
+            return buffer.igbp_buffer.width == width && buffer.igbp_buffer.height == height;
+        });
+
+        if (itr != queue.end()) {
+            free_buffers.erase(f_itr);
+            break;
+        }
+        ++f_itr;
+    }
 
     if (itr == queue.end()) {
         return {};
@@ -99,10 +121,18 @@ void BufferQueue::ReleaseBuffer(u32 slot) {
     ASSERT(itr != queue.end());
     ASSERT(itr->status == Buffer::Status::Acquired);
     itr->status = Buffer::Status::Free;
+    free_buffers.push_back(slot);
 
     buffer_wait_event.writable->Signal();
 }
 
+void BufferQueue::Disconnect() {
+    queue.clear();
+    queue_sequence.clear();
+    id = 1;
+    layer_id = 1;
+}
+
 u32 BufferQueue::Query(QueryType type) {
     LOG_WARNING(Service, "(STUBBED) called type={}", static_cast<u32>(type));
 

src/core/hle/service/nvflinger/buffer_queue.h

@@ -87,6 +87,7 @@ public:
                      Service::Nvidia::MultiFence& multi_fence);
     std::optional<std::reference_wrapper<const Buffer>> AcquireBuffer();
     void ReleaseBuffer(u32 slot);
+    void Disconnect();
     u32 Query(QueryType type);
 
     u32 GetId() const {
@@ -101,6 +102,7 @@ private:
     u32 id;
     u64 layer_id;
 
+    std::list<u32> free_buffers;
     std::vector<Buffer> queue;
     std::list<u32> queue_sequence;
     Kernel::EventPair buffer_wait_event;

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

@@ -29,7 +29,7 @@ Time::Time(std::shared_ptr<Module> module, Core::System& system, const char* nam
         {300, &Time::CalculateMonotonicSystemClockBaseTimePoint, "CalculateMonotonicSystemClockBaseTimePoint"},
         {400, &Time::GetClockSnapshot, "GetClockSnapshot"},
         {401, &Time::GetClockSnapshotFromSystemClockContext, "GetClockSnapshotFromSystemClockContext"},
-        {500, nullptr, "CalculateStandardUserSystemClockDifferenceByUser"},
+        {500, &Time::CalculateStandardUserSystemClockDifferenceByUser, "CalculateStandardUserSystemClockDifferenceByUser"},
         {501, &Time::CalculateSpanBetween, "CalculateSpanBetween"},
     };
     // clang-format on

src/core/hle/service/time/time.cpp

@@ -308,6 +308,29 @@ void Module::Interface::GetClockSnapshotFromSystemClockContext(Kernel::HLEReques
     ctx.WriteBuffer(&clock_snapshot, sizeof(Clock::ClockSnapshot));
 }
 
+void Module::Interface::CalculateStandardUserSystemClockDifferenceByUser(
+    Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_Time, "called");
+
+    IPC::RequestParser rp{ctx};
+    const auto snapshot_a = rp.PopRaw<Clock::ClockSnapshot>();
+    const auto snapshot_b = rp.PopRaw<Clock::ClockSnapshot>();
+
+    auto time_span_type{Clock::TimeSpanType::FromSeconds(snapshot_b.user_context.offset -
+                                                         snapshot_a.user_context.offset)};
+
+    if ((snapshot_b.user_context.steady_time_point.clock_source_id !=
+         snapshot_a.user_context.steady_time_point.clock_source_id) ||
+        (snapshot_b.is_automatic_correction_enabled &&
+         snapshot_a.is_automatic_correction_enabled)) {
+        time_span_type.nanoseconds = 0;
+    }
+
+    IPC::ResponseBuilder rb{ctx, (sizeof(s64) / 4) + 2};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushRaw(time_span_type.nanoseconds);
+}
+
 void Module::Interface::CalculateSpanBetween(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
 

src/core/hle/service/time/time.h

@@ -32,6 +32,7 @@ public:
         void CalculateMonotonicSystemClockBaseTimePoint(Kernel::HLERequestContext& ctx);
         void GetClockSnapshot(Kernel::HLERequestContext& ctx);
         void GetClockSnapshotFromSystemClockContext(Kernel::HLERequestContext& ctx);
+        void CalculateStandardUserSystemClockDifferenceByUser(Kernel::HLERequestContext& ctx);
         void CalculateSpanBetween(Kernel::HLERequestContext& ctx);
         void GetSharedMemoryNativeHandle(Kernel::HLERequestContext& ctx);
 

src/core/hle/service/vi/vi.cpp

@@ -513,7 +513,8 @@ private:
 
         auto& buffer_queue = nv_flinger->FindBufferQueue(id);
 
-        if (transaction == TransactionId::Connect) {
+        switch (transaction) {
+        case TransactionId::Connect: {
             IGBPConnectRequestParcel request{ctx.ReadBuffer()};
             IGBPConnectResponseParcel response{
                 static_cast<u32>(static_cast<u32>(DisplayResolution::UndockedWidth) *
@@ -521,14 +522,18 @@ private:
                 static_cast<u32>(static_cast<u32>(DisplayResolution::UndockedHeight) *
                                  Settings::values.resolution_factor)};
             ctx.WriteBuffer(response.Serialize());
-        } else if (transaction == TransactionId::SetPreallocatedBuffer) {
+            break;
+        }
+        case TransactionId::SetPreallocatedBuffer: {
             IGBPSetPreallocatedBufferRequestParcel request{ctx.ReadBuffer()};
 
             buffer_queue.SetPreallocatedBuffer(request.data.slot, request.buffer);
 
             IGBPSetPreallocatedBufferResponseParcel response{};
             ctx.WriteBuffer(response.Serialize());
-        } else if (transaction == TransactionId::DequeueBuffer) {
+            break;
+        }
+        case TransactionId::DequeueBuffer: {
             IGBPDequeueBufferRequestParcel request{ctx.ReadBuffer()};
             const u32 width{request.data.width};
             const u32 height{request.data.height};
@@ -556,14 +561,18 @@ private:
                     },
                     buffer_queue.GetWritableBufferWaitEvent());
             }
-        } else if (transaction == TransactionId::RequestBuffer) {
+            break;
+        }
+        case TransactionId::RequestBuffer: {
             IGBPRequestBufferRequestParcel request{ctx.ReadBuffer()};
 
             auto& buffer = buffer_queue.RequestBuffer(request.slot);
 
             IGBPRequestBufferResponseParcel response{buffer};
             ctx.WriteBuffer(response.Serialize());
-        } else if (transaction == TransactionId::QueueBuffer) {
+            break;
+        }
+        case TransactionId::QueueBuffer: {
             IGBPQueueBufferRequestParcel request{ctx.ReadBuffer()};
 
             buffer_queue.QueueBuffer(request.data.slot, request.data.transform,
@@ -572,7 +581,9 @@ private:
 
             IGBPQueueBufferResponseParcel response{1280, 720};
             ctx.WriteBuffer(response.Serialize());
-        } else if (transaction == TransactionId::Query) {
+            break;
+        }
+        case TransactionId::Query: {
             IGBPQueryRequestParcel request{ctx.ReadBuffer()};
 
             const u32 value =
@@ -580,15 +591,30 @@ private:
 
             IGBPQueryResponseParcel response{value};
             ctx.WriteBuffer(response.Serialize());
-        } else if (transaction == TransactionId::CancelBuffer) {
+            break;
+        }
+        case TransactionId::CancelBuffer: {
             LOG_CRITICAL(Service_VI, "(STUBBED) called, transaction=CancelBuffer");
-        } else if (transaction == TransactionId::Disconnect ||
-                   transaction == TransactionId::DetachBuffer) {
+            break;
+        }
+        case TransactionId::Disconnect: {
+            LOG_WARNING(Service_VI, "(STUBBED) called, transaction=Disconnect");
+            const auto buffer = ctx.ReadBuffer();
+
+            buffer_queue.Disconnect();
+
+            IGBPEmptyResponseParcel response{};
+            ctx.WriteBuffer(response.Serialize());
+            break;
+        }
+        case TransactionId::DetachBuffer: {
             const auto buffer = ctx.ReadBuffer();
 
             IGBPEmptyResponseParcel response{};
             ctx.WriteBuffer(response.Serialize());
-        } else {
+            break;
+        }
+        default:
             ASSERT_MSG(false, "Unimplemented");
         }
 

src/core/memory.cpp

@@ -242,7 +242,52 @@ struct Memory::Impl {
             }
             case Common::PageType::RasterizerCachedMemory: {
                 const u8* const host_ptr = GetPointerFromVMA(process, current_vaddr);
-                system.GPU().FlushRegion(ToCacheAddr(host_ptr), copy_amount);
+                system.GPU().FlushRegion(current_vaddr, copy_amount);
+                std::memcpy(dest_buffer, host_ptr, copy_amount);
+                break;
+            }
+            default:
+                UNREACHABLE();
+            }
+
+            page_index++;
+            page_offset = 0;
+            dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
+            remaining_size -= copy_amount;
+        }
+    }
+
+    void ReadBlockUnsafe(const Kernel::Process& process, const VAddr src_addr, void* dest_buffer,
+                         const std::size_t size) {
+        const auto& page_table = process.VMManager().page_table;
+
+        std::size_t remaining_size = size;
+        std::size_t page_index = src_addr >> PAGE_BITS;
+        std::size_t page_offset = src_addr & PAGE_MASK;
+
+        while (remaining_size > 0) {
+            const std::size_t copy_amount =
+                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
+            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
+
+            switch (page_table.attributes[page_index]) {
+            case Common::PageType::Unmapped: {
+                LOG_ERROR(HW_Memory,
+                          "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
+                          current_vaddr, src_addr, size);
+                std::memset(dest_buffer, 0, copy_amount);
+                break;
+            }
+            case Common::PageType::Memory: {
+                DEBUG_ASSERT(page_table.pointers[page_index]);
+
+                const u8* const src_ptr =
+                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
+                std::memcpy(dest_buffer, src_ptr, copy_amount);
+                break;
+            }
+            case Common::PageType::RasterizerCachedMemory: {
+                const u8* const host_ptr = GetPointerFromVMA(process, current_vaddr);
                 std::memcpy(dest_buffer, host_ptr, copy_amount);
                 break;
             }
@@ -261,6 +306,10 @@ struct Memory::Impl {
         ReadBlock(*system.CurrentProcess(), src_addr, dest_buffer, size);
     }
 
+    void ReadBlockUnsafe(const VAddr src_addr, void* dest_buffer, const std::size_t size) {
+        ReadBlockUnsafe(*system.CurrentProcess(), src_addr, dest_buffer, size);
+    }
+
     void WriteBlock(const Kernel::Process& process, const VAddr dest_addr, const void* src_buffer,
                     const std::size_t size) {
         const auto& page_table = process.VMManager().page_table;
@@ -290,7 +339,50 @@ struct Memory::Impl {
             }
             case Common::PageType::RasterizerCachedMemory: {
                 u8* const host_ptr = GetPointerFromVMA(process, current_vaddr);
-                system.GPU().InvalidateRegion(ToCacheAddr(host_ptr), copy_amount);
+                system.GPU().InvalidateRegion(current_vaddr, copy_amount);
+                std::memcpy(host_ptr, src_buffer, copy_amount);
+                break;
+            }
+            default:
+                UNREACHABLE();
+            }
+
+            page_index++;
+            page_offset = 0;
+            src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
+            remaining_size -= copy_amount;
+        }
+    }
+
+    void WriteBlockUnsafe(const Kernel::Process& process, const VAddr dest_addr,
+                          const void* src_buffer, const std::size_t size) {
+        const auto& page_table = process.VMManager().page_table;
+        std::size_t remaining_size = size;
+        std::size_t page_index = dest_addr >> PAGE_BITS;
+        std::size_t page_offset = dest_addr & PAGE_MASK;
+
+        while (remaining_size > 0) {
+            const std::size_t copy_amount =
+                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
+            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
+
+            switch (page_table.attributes[page_index]) {
+            case Common::PageType::Unmapped: {
+                LOG_ERROR(HW_Memory,
+                          "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
+                          current_vaddr, dest_addr, size);
+                break;
+            }
+            case Common::PageType::Memory: {
+                DEBUG_ASSERT(page_table.pointers[page_index]);
+
+                u8* const dest_ptr =
+                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
+                std::memcpy(dest_ptr, src_buffer, copy_amount);
+                break;
+            }
+            case Common::PageType::RasterizerCachedMemory: {
+                u8* const host_ptr = GetPointerFromVMA(process, current_vaddr);
                 std::memcpy(host_ptr, src_buffer, copy_amount);
                 break;
             }
@@ -309,6 +401,10 @@ struct Memory::Impl {
         WriteBlock(*system.CurrentProcess(), dest_addr, src_buffer, size);
     }
 
+    void WriteBlockUnsafe(const VAddr dest_addr, const void* src_buffer, const std::size_t size) {
+        WriteBlockUnsafe(*system.CurrentProcess(), dest_addr, src_buffer, size);
+    }
+
     void ZeroBlock(const Kernel::Process& process, const VAddr dest_addr, const std::size_t size) {
         const auto& page_table = process.VMManager().page_table;
         std::size_t remaining_size = size;
@@ -337,7 +433,7 @@ struct Memory::Impl {
             }
             case Common::PageType::RasterizerCachedMemory: {
                 u8* const host_ptr = GetPointerFromVMA(process, current_vaddr);
-                system.GPU().InvalidateRegion(ToCacheAddr(host_ptr), copy_amount);
+                system.GPU().InvalidateRegion(current_vaddr, copy_amount);
                 std::memset(host_ptr, 0, copy_amount);
                 break;
             }
@@ -384,7 +480,7 @@ struct Memory::Impl {
             }
             case Common::PageType::RasterizerCachedMemory: {
                 const u8* const host_ptr = GetPointerFromVMA(process, current_vaddr);
-                system.GPU().FlushRegion(ToCacheAddr(host_ptr), copy_amount);
+                system.GPU().FlushRegion(current_vaddr, copy_amount);
                 WriteBlock(process, dest_addr, host_ptr, copy_amount);
                 break;
             }
@@ -545,7 +641,7 @@ struct Memory::Impl {
             break;
         case Common::PageType::RasterizerCachedMemory: {
             const u8* const host_ptr = GetPointerFromVMA(vaddr);
-            system.GPU().FlushRegion(ToCacheAddr(host_ptr), sizeof(T));
+            system.GPU().FlushRegion(vaddr, sizeof(T));
             T value;
             std::memcpy(&value, host_ptr, sizeof(T));
             return value;
@@ -587,7 +683,7 @@ struct Memory::Impl {
             break;
         case Common::PageType::RasterizerCachedMemory: {
             u8* const host_ptr{GetPointerFromVMA(vaddr)};
-            system.GPU().InvalidateRegion(ToCacheAddr(host_ptr), sizeof(T));
+            system.GPU().InvalidateRegion(vaddr, sizeof(T));
             std::memcpy(host_ptr, &data, sizeof(T));
             break;
         }
@@ -696,6 +792,15 @@ void Memory::ReadBlock(const VAddr src_addr, void* dest_buffer, const std::size_
     impl->ReadBlock(src_addr, dest_buffer, size);
 }
 
+void Memory::ReadBlockUnsafe(const Kernel::Process& process, const VAddr src_addr,
+                             void* dest_buffer, const std::size_t size) {
+    impl->ReadBlockUnsafe(process, src_addr, dest_buffer, size);
+}
+
+void Memory::ReadBlockUnsafe(const VAddr src_addr, void* dest_buffer, const std::size_t size) {
+    impl->ReadBlockUnsafe(src_addr, dest_buffer, size);
+}
+
 void Memory::WriteBlock(const Kernel::Process& process, VAddr dest_addr, const void* src_buffer,
                         std::size_t size) {
     impl->WriteBlock(process, dest_addr, src_buffer, size);
@@ -705,6 +810,16 @@ void Memory::WriteBlock(const VAddr dest_addr, const void* src_buffer, const std
     impl->WriteBlock(dest_addr, src_buffer, size);
 }
 
+void Memory::WriteBlockUnsafe(const Kernel::Process& process, VAddr dest_addr,
+                              const void* src_buffer, std::size_t size) {
+    impl->WriteBlockUnsafe(process, dest_addr, src_buffer, size);
+}
+
+void Memory::WriteBlockUnsafe(const VAddr dest_addr, const void* src_buffer,
+                              const std::size_t size) {
+    impl->WriteBlockUnsafe(dest_addr, src_buffer, size);
+}
+
 void Memory::ZeroBlock(const Kernel::Process& process, VAddr dest_addr, std::size_t size) {
     impl->ZeroBlock(process, dest_addr, size);
 }

src/core/memory.h

@@ -294,6 +294,27 @@ public:
     void ReadBlock(const Kernel::Process& process, VAddr src_addr, void* dest_buffer,
                    std::size_t size);
 
+    /**
+     * Reads a contiguous block of bytes from a specified process' address space.
+     * This unsafe version does not trigger GPU flushing.
+     *
+     * @param process     The process to read the data from.
+     * @param src_addr    The virtual address to begin reading from.
+     * @param dest_buffer The buffer to place the read bytes into.
+     * @param size        The amount of data to read, in bytes.
+     *
+     * @note If a size of 0 is specified, then this function reads nothing and
+     *       no attempts to access memory are made at all.
+     *
+     * @pre dest_buffer must be at least size bytes in length, otherwise a
+     *      buffer overrun will occur.
+     *
+     * @post The range [dest_buffer, size) contains the read bytes from the
+     *       process' address space.
+     */
+    void ReadBlockUnsafe(const Kernel::Process& process, VAddr src_addr, void* dest_buffer,
+                         std::size_t size);
+
     /**
      * Reads a contiguous block of bytes from the current process' address space.
      *
@@ -312,6 +333,25 @@ public:
      */
     void ReadBlock(VAddr src_addr, void* dest_buffer, std::size_t size);
 
+    /**
+     * Reads a contiguous block of bytes from the current process' address space.
+     * This unsafe version does not trigger GPU flushing.
+     *
+     * @param src_addr    The virtual address to begin reading from.
+     * @param dest_buffer The buffer to place the read bytes into.
+     * @param size        The amount of data to read, in bytes.
+     *
+     * @note If a size of 0 is specified, then this function reads nothing and
+     *       no attempts to access memory are made at all.
+     *
+     * @pre dest_buffer must be at least size bytes in length, otherwise a
+     *      buffer overrun will occur.
+     *
+     * @post The range [dest_buffer, size) contains the read bytes from the
+     *       current process' address space.
+     */
+    void ReadBlockUnsafe(VAddr src_addr, void* dest_buffer, std::size_t size);
+
     /**
      * Writes a range of bytes into a given process' address space at the specified
      * virtual address.
@@ -335,6 +375,26 @@ public:
     void WriteBlock(const Kernel::Process& process, VAddr dest_addr, const void* src_buffer,
                     std::size_t size);
 
+    /**
+     * Writes a range of bytes into a given process' address space at the specified
+     * virtual address.
+     * This unsafe version does not invalidate GPU Memory.
+     *
+     * @param process    The process to write data into the address space of.
+     * @param dest_addr  The destination virtual address to begin writing the data at.
+     * @param src_buffer The data to write into the process' address space.
+     * @param size       The size of the data to write, in bytes.
+     *
+     * @post The address range [dest_addr, size) in the process' address space
+     *       contains the data that was within src_buffer.
+     *
+     * @post If an attempt is made to write into an unmapped region of memory, the writes
+     *       will be ignored and an error will be logged.
+     *
+     */
+    void WriteBlockUnsafe(const Kernel::Process& process, VAddr dest_addr, const void* src_buffer,
+                          std::size_t size);
+
     /**
      * Writes a range of bytes into the current process' address space at the specified
      * virtual address.
@@ -356,6 +416,24 @@ public:
      */
     void WriteBlock(VAddr dest_addr, const void* src_buffer, std::size_t size);
 
+    /**
+     * Writes a range of bytes into the current process' address space at the specified
+     * virtual address.
+     * This unsafe version does not invalidate GPU Memory.
+     *
+     * @param dest_addr  The destination virtual address to begin writing the data at.
+     * @param src_buffer The data to write into the current process' address space.
+     * @param size       The size of the data to write, in bytes.
+     *
+     * @post The address range [dest_addr, size) in the current process' address space
+     *       contains the data that was within src_buffer.
+     *
+     * @post If an attempt is made to write into an unmapped region of memory, the writes
+     *       will be ignored and an error will be logged.
+     *
+     */
+    void WriteBlockUnsafe(VAddr dest_addr, const void* src_buffer, std::size_t size);
+
     /**
      * Fills the specified address range within a process' address space with zeroes.
      *

src/input_common/CMakeLists.txt

@@ -27,4 +27,4 @@ if(SDL2_FOUND)
 endif()
 
 create_target_directory_groups(input_common)
-target_link_libraries(input_common PUBLIC core PRIVATE common ${Boost_LIBRARIES})
+target_link_libraries(input_common PUBLIC core PRIVATE common Boost::boost)

src/video_core/CMakeLists.txt

@@ -148,6 +148,7 @@ add_library(video_core STATIC
     textures/convert.h
     textures/decoders.cpp
     textures/decoders.h
+    textures/texture.cpp
     textures/texture.h
     video_core.cpp
     video_core.h
@@ -155,7 +156,6 @@ add_library(video_core STATIC
 
 if (ENABLE_VULKAN)
     target_sources(video_core PRIVATE
-        renderer_vulkan/declarations.h
         renderer_vulkan/fixed_pipeline_state.cpp
         renderer_vulkan/fixed_pipeline_state.h
         renderer_vulkan/maxwell_to_vk.cpp
@@ -210,6 +210,8 @@ if (ENABLE_VULKAN)
         renderer_vulkan/vk_texture_cache.h
         renderer_vulkan/vk_update_descriptor.cpp
         renderer_vulkan/vk_update_descriptor.h
+        renderer_vulkan/wrapper.cpp
+        renderer_vulkan/wrapper.h
     )
 
     target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)

src/video_core/buffer_cache/buffer_block.h

@@ -15,37 +15,29 @@ namespace VideoCommon {
 
 class BufferBlock {
 public:
-    bool Overlaps(const CacheAddr start, const CacheAddr end) const {
-        return (cache_addr < end) && (cache_addr_end > start);
+    bool Overlaps(const VAddr start, const VAddr end) const {
+        return (cpu_addr < end) && (cpu_addr_end > start);
     }
 
-    bool IsInside(const CacheAddr other_start, const CacheAddr other_end) const {
-        return cache_addr <= other_start && other_end <= cache_addr_end;
+    bool IsInside(const VAddr other_start, const VAddr other_end) const {
+        return cpu_addr <= other_start && other_end <= cpu_addr_end;
     }
 
-    u8* GetWritableHostPtr() const {
-        return FromCacheAddr(cache_addr);
+    std::size_t GetOffset(const VAddr in_addr) {
+        return static_cast<std::size_t>(in_addr - cpu_addr);
     }
 
-    u8* GetWritableHostPtr(std::size_t offset) const {
-        return FromCacheAddr(cache_addr + offset);
+    VAddr GetCpuAddr() const {
+        return cpu_addr;
     }
 
-    std::size_t GetOffset(const CacheAddr in_addr) {
-        return static_cast<std::size_t>(in_addr - cache_addr);
+    VAddr GetCpuAddrEnd() const {
+        return cpu_addr_end;
     }
 
-    CacheAddr GetCacheAddr() const {
-        return cache_addr;
-    }
-
-    CacheAddr GetCacheAddrEnd() const {
-        return cache_addr_end;
-    }
-
-    void SetCacheAddr(const CacheAddr new_addr) {
-        cache_addr = new_addr;
-        cache_addr_end = new_addr + size;
+    void SetCpuAddr(const VAddr new_addr) {
+        cpu_addr = new_addr;
+        cpu_addr_end = new_addr + size;
     }
 
     std::size_t GetSize() const {
@@ -61,14 +53,14 @@ public:
     }
 
 protected:
-    explicit BufferBlock(CacheAddr cache_addr, const std::size_t size) : size{size} {
-        SetCacheAddr(cache_addr);
+    explicit BufferBlock(VAddr cpu_addr, const std::size_t size) : size{size} {
+        SetCpuAddr(cpu_addr);
     }
     ~BufferBlock() = default;
 
 private:
-    CacheAddr cache_addr{};
-    CacheAddr cache_addr_end{};
+    VAddr cpu_addr{};
+    VAddr cpu_addr_end{};
     std::size_t size{};
     u64 epoch{};
 };

src/video_core/buffer_cache/buffer_cache.h

@@ -19,6 +19,7 @@
 #include "common/alignment.h"
 #include "common/common_types.h"
 #include "core/core.h"
+#include "core/memory.h"
 #include "video_core/buffer_cache/buffer_block.h"
 #include "video_core/buffer_cache/map_interval.h"
 #include "video_core/memory_manager.h"
@@ -28,37 +29,54 @@ namespace VideoCommon {
 
 using MapInterval = std::shared_ptr<MapIntervalBase>;
 
-template <typename TBuffer, typename TBufferType, typename StreamBuffer>
+template <typename OwnerBuffer, typename BufferType, typename StreamBuffer>
 class BufferCache {
 public:
-    using BufferInfo = std::pair<const TBufferType*, u64>;
+    using BufferInfo = std::pair<BufferType, u64>;
 
     BufferInfo UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment = 4,
                             bool is_written = false, bool use_fast_cbuf = false) {
         std::lock_guard lock{mutex};
 
-        auto& memory_manager = system.GPU().MemoryManager();
-        const auto host_ptr = memory_manager.GetPointer(gpu_addr);
-        if (!host_ptr) {
+        const std::optional<VAddr> cpu_addr_opt =
+            system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
+
+        if (!cpu_addr_opt) {
             return {GetEmptyBuffer(size), 0};
         }
-        const auto cache_addr = ToCacheAddr(host_ptr);
+
+        VAddr cpu_addr = *cpu_addr_opt;
 
         // Cache management is a big overhead, so only cache entries with a given size.
         // TODO: Figure out which size is the best for given games.
         constexpr std::size_t max_stream_size = 0x800;
         if (use_fast_cbuf || size < max_stream_size) {
-            if (!is_written && !IsRegionWritten(cache_addr, cache_addr + size - 1)) {
+            if (!is_written && !IsRegionWritten(cpu_addr, cpu_addr + size - 1)) {
+                auto& memory_manager = system.GPU().MemoryManager();
                 if (use_fast_cbuf) {
-                    return ConstBufferUpload(host_ptr, size);
+                    if (memory_manager.IsGranularRange(gpu_addr, size)) {
+                        const auto host_ptr = memory_manager.GetPointer(gpu_addr);
+                        return ConstBufferUpload(host_ptr, size);
+                    } else {
+                        staging_buffer.resize(size);
+                        memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
+                        return ConstBufferUpload(staging_buffer.data(), size);
+                    }
                 } else {
-                    return StreamBufferUpload(host_ptr, size, alignment);
+                    if (memory_manager.IsGranularRange(gpu_addr, size)) {
+                        const auto host_ptr = memory_manager.GetPointer(gpu_addr);
+                        return StreamBufferUpload(host_ptr, size, alignment);
+                    } else {
+                        staging_buffer.resize(size);
+                        memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
+                        return StreamBufferUpload(staging_buffer.data(), size, alignment);
+                    }
                 }
             }
         }
 
-        auto block = GetBlock(cache_addr, size);
-        auto map = MapAddress(block, gpu_addr, cache_addr, size);
+        auto block = GetBlock(cpu_addr, size);
+        auto map = MapAddress(block, gpu_addr, cpu_addr, size);
         if (is_written) {
             map->MarkAsModified(true, GetModifiedTicks());
             if (!map->IsWritten()) {
@@ -71,9 +89,7 @@ public:
             }
         }
 
-        const u64 offset = static_cast<u64>(block->GetOffset(cache_addr));
-
-        return {ToHandle(block), offset};
+        return {ToHandle(block), static_cast<u64>(block->GetOffset(cpu_addr))};
     }
 
     /// Uploads from a host memory. Returns the OpenGL buffer where it's located and its offset.
@@ -112,7 +128,7 @@ public:
     }
 
     /// Write any cached resources overlapping the specified region back to memory
-    void FlushRegion(CacheAddr addr, std::size_t size) {
+    void FlushRegion(VAddr addr, std::size_t size) {
         std::lock_guard lock{mutex};
 
         std::vector<MapInterval> objects = GetMapsInRange(addr, size);
@@ -127,7 +143,7 @@ public:
     }
 
     /// Mark the specified region as being invalidated
-    void InvalidateRegion(CacheAddr addr, u64 size) {
+    void InvalidateRegion(VAddr addr, u64 size) {
         std::lock_guard lock{mutex};
 
         std::vector<MapInterval> objects = GetMapsInRange(addr, size);
@@ -138,7 +154,7 @@ public:
         }
     }
 
-    virtual const TBufferType* GetEmptyBuffer(std::size_t size) = 0;
+    virtual BufferType GetEmptyBuffer(std::size_t size) = 0;
 
 protected:
     explicit BufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
@@ -148,19 +164,19 @@ protected:
 
     ~BufferCache() = default;
 
-    virtual const TBufferType* ToHandle(const TBuffer& storage) = 0;
+    virtual BufferType ToHandle(const OwnerBuffer& storage) = 0;
 
     virtual void WriteBarrier() = 0;
 
-    virtual TBuffer CreateBlock(CacheAddr cache_addr, std::size_t size) = 0;
+    virtual OwnerBuffer CreateBlock(VAddr cpu_addr, std::size_t size) = 0;
 
-    virtual void UploadBlockData(const TBuffer& buffer, std::size_t offset, std::size_t size,
+    virtual void UploadBlockData(const OwnerBuffer& buffer, std::size_t offset, std::size_t size,
                                  const u8* data) = 0;
 
-    virtual void DownloadBlockData(const TBuffer& buffer, std::size_t offset, std::size_t size,
+    virtual void DownloadBlockData(const OwnerBuffer& buffer, std::size_t offset, std::size_t size,
                                    u8* data) = 0;
 
-    virtual void CopyBlock(const TBuffer& src, const TBuffer& dst, std::size_t src_offset,
+    virtual void CopyBlock(const OwnerBuffer& src, const OwnerBuffer& dst, std::size_t src_offset,
                            std::size_t dst_offset, std::size_t size) = 0;
 
     virtual BufferInfo ConstBufferUpload(const void* raw_pointer, std::size_t size) {
@@ -169,20 +185,17 @@ protected:
 
     /// Register an object into the cache
     void Register(const MapInterval& new_map, bool inherit_written = false) {
-        const CacheAddr cache_ptr = new_map->GetStart();
-        const std::optional<VAddr> cpu_addr =
-            system.GPU().MemoryManager().GpuToCpuAddress(new_map->GetGpuAddress());
-        if (!cache_ptr || !cpu_addr) {
+        const VAddr cpu_addr = new_map->GetStart();
+        if (!cpu_addr) {
             LOG_CRITICAL(HW_GPU, "Failed to register buffer with unmapped gpu_address 0x{:016x}",
                          new_map->GetGpuAddress());
             return;
         }
         const std::size_t size = new_map->GetEnd() - new_map->GetStart();
-        new_map->SetCpuAddress(*cpu_addr);
         new_map->MarkAsRegistered(true);
         const IntervalType interval{new_map->GetStart(), new_map->GetEnd()};
         mapped_addresses.insert({interval, new_map});
-        rasterizer.UpdatePagesCachedCount(*cpu_addr, size, 1);
+        rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
         if (inherit_written) {
             MarkRegionAsWritten(new_map->GetStart(), new_map->GetEnd() - 1);
             new_map->MarkAsWritten(true);
@@ -192,7 +205,7 @@ protected:
     /// Unregisters an object from the cache
     void Unregister(MapInterval& map) {
         const std::size_t size = map->GetEnd() - map->GetStart();
-        rasterizer.UpdatePagesCachedCount(map->GetCpuAddress(), size, -1);
+        rasterizer.UpdatePagesCachedCount(map->GetStart(), size, -1);
         map->MarkAsRegistered(false);
         if (map->IsWritten()) {
             UnmarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
@@ -202,32 +215,38 @@ protected:
     }
 
 private:
-    MapInterval CreateMap(const CacheAddr start, const CacheAddr end, const GPUVAddr gpu_addr) {
+    MapInterval CreateMap(const VAddr start, const VAddr end, const GPUVAddr gpu_addr) {
         return std::make_shared<MapIntervalBase>(start, end, gpu_addr);
     }
 
-    MapInterval MapAddress(const TBuffer& block, const GPUVAddr gpu_addr,
-                           const CacheAddr cache_addr, const std::size_t size) {
-
-        std::vector<MapInterval> overlaps = GetMapsInRange(cache_addr, size);
+    MapInterval MapAddress(const OwnerBuffer& block, const GPUVAddr gpu_addr, const VAddr cpu_addr,
+                           const std::size_t size) {
+        std::vector<MapInterval> overlaps = GetMapsInRange(cpu_addr, size);
         if (overlaps.empty()) {
-            const CacheAddr cache_addr_end = cache_addr + size;
-            MapInterval new_map = CreateMap(cache_addr, cache_addr_end, gpu_addr);
-            u8* host_ptr = FromCacheAddr(cache_addr);
-            UploadBlockData(block, block->GetOffset(cache_addr), size, host_ptr);
+            auto& memory_manager = system.GPU().MemoryManager();
+            const VAddr cpu_addr_end = cpu_addr + size;
+            MapInterval new_map = CreateMap(cpu_addr, cpu_addr_end, gpu_addr);
+            if (memory_manager.IsGranularRange(gpu_addr, size)) {
+                u8* host_ptr = memory_manager.GetPointer(gpu_addr);
+                UploadBlockData(block, block->GetOffset(cpu_addr), size, host_ptr);
+            } else {
+                staging_buffer.resize(size);
+                memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
+                UploadBlockData(block, block->GetOffset(cpu_addr), size, staging_buffer.data());
+            }
             Register(new_map);
             return new_map;
         }
 
-        const CacheAddr cache_addr_end = cache_addr + size;
+        const VAddr cpu_addr_end = cpu_addr + size;
         if (overlaps.size() == 1) {
             MapInterval& current_map = overlaps[0];
-            if (current_map->IsInside(cache_addr, cache_addr_end)) {
+            if (current_map->IsInside(cpu_addr, cpu_addr_end)) {
                 return current_map;
             }
         }
-        CacheAddr new_start = cache_addr;
-        CacheAddr new_end = cache_addr_end;
+        VAddr new_start = cpu_addr;
+        VAddr new_end = cpu_addr_end;
         bool write_inheritance = false;
         bool modified_inheritance = false;
         // Calculate new buffer parameters
@@ -237,7 +256,7 @@ private:
             write_inheritance |= overlap->IsWritten();
             modified_inheritance |= overlap->IsModified();
         }
-        GPUVAddr new_gpu_addr = gpu_addr + new_start - cache_addr;
+        GPUVAddr new_gpu_addr = gpu_addr + new_start - cpu_addr;
         for (auto& overlap : overlaps) {
             Unregister(overlap);
         }
@@ -250,7 +269,7 @@ private:
         return new_map;
     }
 
-    void UpdateBlock(const TBuffer& block, CacheAddr start, CacheAddr end,
+    void UpdateBlock(const OwnerBuffer& block, VAddr start, VAddr end,
                      std::vector<MapInterval>& overlaps) {
         const IntervalType base_interval{start, end};
         IntervalSet interval_set{};
@@ -262,13 +281,15 @@ private:
         for (auto& interval : interval_set) {
             std::size_t size = interval.upper() - interval.lower();
             if (size > 0) {
-                u8* host_ptr = FromCacheAddr(interval.lower());
-                UploadBlockData(block, block->GetOffset(interval.lower()), size, host_ptr);
+                staging_buffer.resize(size);
+                system.Memory().ReadBlockUnsafe(interval.lower(), staging_buffer.data(), size);
+                UploadBlockData(block, block->GetOffset(interval.lower()), size,
+                                staging_buffer.data());
             }
         }
     }
 
-    std::vector<MapInterval> GetMapsInRange(CacheAddr addr, std::size_t size) {
+    std::vector<MapInterval> GetMapsInRange(VAddr addr, std::size_t size) {
         if (size == 0) {
             return {};
         }
@@ -289,9 +310,10 @@ private:
 
     void FlushMap(MapInterval map) {
         std::size_t size = map->GetEnd() - map->GetStart();
-        TBuffer block = blocks[map->GetStart() >> block_page_bits];
-        u8* host_ptr = FromCacheAddr(map->GetStart());
-        DownloadBlockData(block, block->GetOffset(map->GetStart()), size, host_ptr);
+        OwnerBuffer block = blocks[map->GetStart() >> block_page_bits];
+        staging_buffer.resize(size);
+        DownloadBlockData(block, block->GetOffset(map->GetStart()), size, staging_buffer.data());
+        system.Memory().WriteBlockUnsafe(map->GetStart(), staging_buffer.data(), size);
         map->MarkAsModified(false, 0);
     }
 
@@ -303,7 +325,7 @@ private:
 
         buffer_ptr += size;
         buffer_offset += size;
-        return {&stream_buffer_handle, uploaded_offset};
+        return {stream_buffer_handle, uploaded_offset};
     }
 
     void AlignBuffer(std::size_t alignment) {
@@ -313,17 +335,17 @@ private:
         buffer_offset = offset_aligned;
     }
 
-    TBuffer EnlargeBlock(TBuffer buffer) {
+    OwnerBuffer EnlargeBlock(OwnerBuffer buffer) {
         const std::size_t old_size = buffer->GetSize();
         const std::size_t new_size = old_size + block_page_size;
-        const CacheAddr cache_addr = buffer->GetCacheAddr();
-        TBuffer new_buffer = CreateBlock(cache_addr, new_size);
+        const VAddr cpu_addr = buffer->GetCpuAddr();
+        OwnerBuffer new_buffer = CreateBlock(cpu_addr, new_size);
         CopyBlock(buffer, new_buffer, 0, 0, old_size);
         buffer->SetEpoch(epoch);
         pending_destruction.push_back(buffer);
-        const CacheAddr cache_addr_end = cache_addr + new_size - 1;
-        u64 page_start = cache_addr >> block_page_bits;
-        const u64 page_end = cache_addr_end >> block_page_bits;
+        const VAddr cpu_addr_end = cpu_addr + new_size - 1;
+        u64 page_start = cpu_addr >> block_page_bits;
+        const u64 page_end = cpu_addr_end >> block_page_bits;
         while (page_start <= page_end) {
             blocks[page_start] = new_buffer;
             ++page_start;
@@ -331,23 +353,23 @@ private:
         return new_buffer;
     }
 
-    TBuffer MergeBlocks(TBuffer first, TBuffer second) {
+    OwnerBuffer MergeBlocks(OwnerBuffer first, OwnerBuffer second) {
         const std::size_t size_1 = first->GetSize();
         const std::size_t size_2 = second->GetSize();
-        const CacheAddr first_addr = first->GetCacheAddr();
-        const CacheAddr second_addr = second->GetCacheAddr();
-        const CacheAddr new_addr = std::min(first_addr, second_addr);
+        const VAddr first_addr = first->GetCpuAddr();
+        const VAddr second_addr = second->GetCpuAddr();
+        const VAddr new_addr = std::min(first_addr, second_addr);
         const std::size_t new_size = size_1 + size_2;
-        TBuffer new_buffer = CreateBlock(new_addr, new_size);
+        OwnerBuffer new_buffer = CreateBlock(new_addr, new_size);
         CopyBlock(first, new_buffer, 0, new_buffer->GetOffset(first_addr), size_1);
         CopyBlock(second, new_buffer, 0, new_buffer->GetOffset(second_addr), size_2);
         first->SetEpoch(epoch);
         second->SetEpoch(epoch);
         pending_destruction.push_back(first);
         pending_destruction.push_back(second);
-        const CacheAddr cache_addr_end = new_addr + new_size - 1;
+        const VAddr cpu_addr_end = new_addr + new_size - 1;
         u64 page_start = new_addr >> block_page_bits;
-        const u64 page_end = cache_addr_end >> block_page_bits;
+        const u64 page_end = cpu_addr_end >> block_page_bits;
         while (page_start <= page_end) {
             blocks[page_start] = new_buffer;
             ++page_start;
@@ -355,18 +377,18 @@ private:
         return new_buffer;
     }
 
-    TBuffer GetBlock(const CacheAddr cache_addr, const std::size_t size) {
-        TBuffer found{};
-        const CacheAddr cache_addr_end = cache_addr + size - 1;
-        u64 page_start = cache_addr >> block_page_bits;
-        const u64 page_end = cache_addr_end >> block_page_bits;
+    OwnerBuffer GetBlock(const VAddr cpu_addr, const std::size_t size) {
+        OwnerBuffer found;
+        const VAddr cpu_addr_end = cpu_addr + size - 1;
+        u64 page_start = cpu_addr >> block_page_bits;
+        const u64 page_end = cpu_addr_end >> block_page_bits;
         while (page_start <= page_end) {
             auto it = blocks.find(page_start);
             if (it == blocks.end()) {
                 if (found) {
                     found = EnlargeBlock(found);
                 } else {
-                    const CacheAddr start_addr = (page_start << block_page_bits);
+                    const VAddr start_addr = (page_start << block_page_bits);
                     found = CreateBlock(start_addr, block_page_size);
                     blocks[page_start] = found;
                 }
@@ -386,7 +408,7 @@ private:
         return found;
     }
 
-    void MarkRegionAsWritten(const CacheAddr start, const CacheAddr end) {
+    void MarkRegionAsWritten(const VAddr start, const VAddr end) {
         u64 page_start = start >> write_page_bit;
         const u64 page_end = end >> write_page_bit;
         while (page_start <= page_end) {
@@ -400,7 +422,7 @@ private:
         }
     }
 
-    void UnmarkRegionAsWritten(const CacheAddr start, const CacheAddr end) {
+    void UnmarkRegionAsWritten(const VAddr start, const VAddr end) {
         u64 page_start = start >> write_page_bit;
         const u64 page_end = end >> write_page_bit;
         while (page_start <= page_end) {
@@ -416,7 +438,7 @@ private:
         }
     }
 
-    bool IsRegionWritten(const CacheAddr start, const CacheAddr end) const {
+    bool IsRegionWritten(const VAddr start, const VAddr end) const {
         u64 page_start = start >> write_page_bit;
         const u64 page_end = end >> write_page_bit;
         while (page_start <= page_end) {
@@ -432,7 +454,7 @@ private:
     Core::System& system;
 
     std::unique_ptr<StreamBuffer> stream_buffer;
-    TBufferType stream_buffer_handle{};
+    BufferType stream_buffer_handle{};
 
     bool invalidated = false;
 
@@ -440,8 +462,8 @@ private:
     u64 buffer_offset = 0;
     u64 buffer_offset_base = 0;
 
-    using IntervalSet = boost::icl::interval_set<CacheAddr>;
-    using IntervalCache = boost::icl::interval_map<CacheAddr, MapInterval>;
+    using IntervalSet = boost::icl::interval_set<VAddr>;
+    using IntervalCache = boost::icl::interval_map<VAddr, MapInterval>;
     using IntervalType = typename IntervalCache::interval_type;
     IntervalCache mapped_addresses;
 
@@ -450,12 +472,14 @@ private:
 
     static constexpr u64 block_page_bits = 21;
     static constexpr u64 block_page_size = 1ULL << block_page_bits;
-    std::unordered_map<u64, TBuffer> blocks;
+    std::unordered_map<u64, OwnerBuffer> blocks;
 
-    std::list<TBuffer> pending_destruction;
+    std::list<OwnerBuffer> pending_destruction;
     u64 epoch = 0;
     u64 modified_ticks = 0;
 
+    std::vector<u8> staging_buffer;
+
     std::recursive_mutex mutex;
 };
 

src/video_core/buffer_cache/map_interval.h

@@ -11,7 +11,7 @@ namespace VideoCommon {
 
 class MapIntervalBase {
 public:
-    MapIntervalBase(const CacheAddr start, const CacheAddr end, const GPUVAddr gpu_addr)
+    MapIntervalBase(const VAddr start, const VAddr end, const GPUVAddr gpu_addr)
         : start{start}, end{end}, gpu_addr{gpu_addr} {}
 
     void SetCpuAddress(VAddr new_cpu_addr) {
@@ -26,7 +26,7 @@ public:
         return gpu_addr;
     }
 
-    bool IsInside(const CacheAddr other_start, const CacheAddr other_end) const {
+    bool IsInside(const VAddr other_start, const VAddr other_end) const {
         return (start <= other_start && other_end <= end);
     }
 
@@ -46,11 +46,11 @@ public:
         return is_registered;
     }
 
-    CacheAddr GetStart() const {
+    VAddr GetStart() const {
         return start;
     }
 
-    CacheAddr GetEnd() const {
+    VAddr GetEnd() const {
         return end;
     }
 
@@ -76,8 +76,8 @@ public:
     }
 
 private:
-    CacheAddr start;
-    CacheAddr end;
+    VAddr start;
+    VAddr end;
     GPUVAddr gpu_addr;
     VAddr cpu_addr{};
     bool is_written{};

src/video_core/engines/maxwell_3d.h

@@ -303,6 +303,10 @@ public:
                 return (type == Type::SignedNorm) || (type == Type::UnsignedNorm);
             }
 
+            bool IsConstant() const {
+                return constant;
+            }
+
             bool IsValid() const {
                 return size != Size::Invalid;
             }
@@ -312,6 +316,35 @@ public:
             }
         };
 
+        struct MsaaSampleLocation {
+            union {
+                BitField<0, 4, u32> x0;
+                BitField<4, 4, u32> y0;
+                BitField<8, 4, u32> x1;
+                BitField<12, 4, u32> y1;
+                BitField<16, 4, u32> x2;
+                BitField<20, 4, u32> y2;
+                BitField<24, 4, u32> x3;
+                BitField<28, 4, u32> y3;
+            };
+
+            constexpr std::pair<u32, u32> Location(int index) const {
+                switch (index) {
+                case 0:
+                    return {x0, y0};
+                case 1:
+                    return {x1, y1};
+                case 2:
+                    return {x2, y2};
+                case 3:
+                    return {x3, y3};
+                default:
+                    UNREACHABLE();
+                    return {0, 0};
+                }
+            }
+        };
+
         enum class DepthMode : u32 {
             MinusOneToOne = 0,
             ZeroToOne = 1,
@@ -793,7 +826,13 @@ public:
 
                 u32 rt_separate_frag_data;
 
-                INSERT_UNION_PADDING_WORDS(0xC);
+                INSERT_UNION_PADDING_WORDS(0x1);
+
+                u32 multisample_raster_enable;
+                u32 multisample_raster_samples;
+                std::array<u32, 4> multisample_sample_mask;
+
+                INSERT_UNION_PADDING_WORDS(0x5);
 
                 struct {
                     u32 address_high;
@@ -830,7 +869,16 @@ public:
 
                 std::array<VertexAttribute, NumVertexAttributes> vertex_attrib_format;
 
-                INSERT_UNION_PADDING_WORDS(0xF);
+                std::array<MsaaSampleLocation, 4> multisample_sample_locations;
+
+                INSERT_UNION_PADDING_WORDS(0x2);
+
+                union {
+                    BitField<0, 1, u32> enable;
+                    BitField<4, 3, u32> target;
+                } multisample_coverage_to_color;
+
+                INSERT_UNION_PADDING_WORDS(0x8);
 
                 struct {
                     union {
@@ -922,7 +970,10 @@ public:
                     BitField<4, 1, u32> triangle_rast_flip;
                 } screen_y_control;
 
-                INSERT_UNION_PADDING_WORDS(0x21);
+                float line_width_smooth;
+                float line_width_aliased;
+
+                INSERT_UNION_PADDING_WORDS(0x1F);
 
                 u32 vb_element_base;
                 u32 vb_base_instance;
@@ -943,7 +994,7 @@ public:
 
                 CounterReset counter_reset;
 
-                INSERT_UNION_PADDING_WORDS(0x1);
+                u32 multisample_enable;
 
                 u32 zeta_enable;
 
@@ -980,7 +1031,7 @@ public:
 
                 float polygon_offset_factor;
 
-                INSERT_UNION_PADDING_WORDS(0x1);
+                u32 line_smooth_enable;
 
                 struct {
                     u32 tic_address_high;
@@ -1007,7 +1058,11 @@ public:
 
                 float polygon_offset_units;
 
-                INSERT_UNION_PADDING_WORDS(0x11);
+                INSERT_UNION_PADDING_WORDS(0x4);
+
+                Tegra::Texture::MsaaMode multisample_mode;
+
+                INSERT_UNION_PADDING_WORDS(0xC);
 
                 union {
                     BitField<2, 1, u32> coord_origin;
@@ -1507,12 +1562,17 @@ 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(color_mask_common, 0x3E4);
-ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
 ASSERT_REG_POSITION(depth_bounds, 0x3E7);
+ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
+ASSERT_REG_POSITION(multisample_raster_enable, 0x3ED);
+ASSERT_REG_POSITION(multisample_raster_samples, 0x3EE);
+ASSERT_REG_POSITION(multisample_sample_mask, 0x3EF);
 ASSERT_REG_POSITION(zeta, 0x3F8);
 ASSERT_REG_POSITION(clear_flags, 0x43E);
 ASSERT_REG_POSITION(fill_rectangle, 0x44F);
 ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
+ASSERT_REG_POSITION(multisample_sample_locations, 0x478);
+ASSERT_REG_POSITION(multisample_coverage_to_color, 0x47E);
 ASSERT_REG_POSITION(rt_control, 0x487);
 ASSERT_REG_POSITION(zeta_width, 0x48a);
 ASSERT_REG_POSITION(zeta_height, 0x48b);
@@ -1538,6 +1598,8 @@ ASSERT_REG_POSITION(stencil_front_func_mask, 0x4E6);
 ASSERT_REG_POSITION(stencil_front_mask, 0x4E7);
 ASSERT_REG_POSITION(frag_color_clamp, 0x4EA);
 ASSERT_REG_POSITION(screen_y_control, 0x4EB);
+ASSERT_REG_POSITION(line_width_smooth, 0x4EC);
+ASSERT_REG_POSITION(line_width_aliased, 0x4ED);
 ASSERT_REG_POSITION(vb_element_base, 0x50D);
 ASSERT_REG_POSITION(vb_base_instance, 0x50E);
 ASSERT_REG_POSITION(clip_distance_enabled, 0x544);
@@ -1545,11 +1607,13 @@ ASSERT_REG_POSITION(samplecnt_enable, 0x545);
 ASSERT_REG_POSITION(point_size, 0x546);
 ASSERT_REG_POSITION(point_sprite_enable, 0x548);
 ASSERT_REG_POSITION(counter_reset, 0x54C);
+ASSERT_REG_POSITION(multisample_enable, 0x54D);
 ASSERT_REG_POSITION(zeta_enable, 0x54E);
 ASSERT_REG_POSITION(multisample_control, 0x54F);
 ASSERT_REG_POSITION(condition, 0x554);
 ASSERT_REG_POSITION(tsc, 0x557);
-ASSERT_REG_POSITION(polygon_offset_factor, 0x55b);
+ASSERT_REG_POSITION(polygon_offset_factor, 0x55B);
+ASSERT_REG_POSITION(line_smooth_enable, 0x55C);
 ASSERT_REG_POSITION(tic, 0x55D);
 ASSERT_REG_POSITION(stencil_two_side_enable, 0x565);
 ASSERT_REG_POSITION(stencil_back_op_fail, 0x566);
@@ -1558,6 +1622,7 @@ ASSERT_REG_POSITION(stencil_back_op_zpass, 0x568);
 ASSERT_REG_POSITION(stencil_back_func_func, 0x569);
 ASSERT_REG_POSITION(framebuffer_srgb, 0x56E);
 ASSERT_REG_POSITION(polygon_offset_units, 0x56F);
+ASSERT_REG_POSITION(multisample_mode, 0x574);
 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

@@ -290,6 +290,23 @@ enum class VmadShr : u64 {
     Shr15 = 2,
 };
 
+enum class VmnmxType : u64 {
+    Bits8,
+    Bits16,
+    Bits32,
+};
+
+enum class VmnmxOperation : u64 {
+    Mrg_16H = 0,
+    Mrg_16L = 1,
+    Mrg_8B0 = 2,
+    Mrg_8B2 = 3,
+    Acc = 4,
+    Min = 5,
+    Max = 6,
+    Nop = 7,
+};
+
 enum class XmadMode : u64 {
     None = 0,
     CLo = 1,
@@ -988,6 +1005,12 @@ union Instruction {
         BitField<46, 2, u64> cache_mode;
     } stg;
 
+    union {
+        BitField<23, 3, AtomicOp> operation;
+        BitField<48, 1, u64> extended;
+        BitField<20, 3, GlobalAtomicType> type;
+    } red;
+
     union {
         BitField<52, 4, AtomicOp> operation;
         BitField<49, 3, GlobalAtomicType> type;
@@ -1650,6 +1673,42 @@ union Instruction {
         BitField<47, 1, u64> cc;
     } vmad;
 
+    union {
+        BitField<54, 1, u64> is_dest_signed;
+        BitField<48, 1, u64> is_src_a_signed;
+        BitField<49, 1, u64> is_src_b_signed;
+        BitField<37, 2, u64> src_format_a;
+        BitField<29, 2, u64> src_format_b;
+        BitField<56, 1, u64> mx;
+        BitField<55, 1, u64> sat;
+        BitField<36, 2, u64> selector_a;
+        BitField<28, 2, u64> selector_b;
+        BitField<50, 1, u64> is_op_b_register;
+        BitField<51, 3, VmnmxOperation> operation;
+
+        VmnmxType SourceFormatA() const {
+            switch (src_format_a) {
+            case 0b11:
+                return VmnmxType::Bits32;
+            case 0b10:
+                return VmnmxType::Bits16;
+            default:
+                return VmnmxType::Bits8;
+            }
+        }
+
+        VmnmxType SourceFormatB() const {
+            switch (src_format_b) {
+            case 0b11:
+                return VmnmxType::Bits32;
+            case 0b10:
+                return VmnmxType::Bits16;
+            default:
+                return VmnmxType::Bits8;
+            }
+        }
+    } vmnmx;
+
     union {
         BitField<20, 16, u64> imm20_16;
         BitField<35, 1, u64> high_b_rr; // used on RR
@@ -1712,6 +1771,7 @@ public:
         BRK,
         DEPBAR,
         VOTE,
+        VOTE_VTG,
         SHFL,
         FSWZADD,
         BFE_C,
@@ -1733,6 +1793,7 @@ public:
         ST_S,
         ST,    // Store in generic memory
         STG,   // Store in global memory
+        RED,   // Reduction operation
         ATOM,  // Atomic operation on global memory
         ATOMS, // Atomic operation on shared memory
         AL2P,  // Transforms attribute memory into physical memory
@@ -1758,9 +1819,11 @@ public:
         IPA,
         OUT_R, // Emit vertex/primitive
         ISBERD,
+        BAR,
         MEMBAR,
         VMAD,
         VSETP,
+        VMNMX,
         FFMA_IMM, // Fused Multiply and Add
         FFMA_CR,
         FFMA_RC,
@@ -1815,7 +1878,8 @@ public:
         ICMP_R,
         ICMP_CR,
         ICMP_IMM,
-        FCMP_R,
+        FCMP_RR,
+        FCMP_RC,
         MUFU,  // Multi-Function Operator
         RRO_C, // Range Reduction Operator
         RRO_R,
@@ -1842,7 +1906,7 @@ public:
         MOV_C,
         MOV_R,
         MOV_IMM,
-        MOV_SYS,
+        S2R,
         MOV32_IMM,
         SHL_C,
         SHL_R,
@@ -2026,6 +2090,7 @@ private:
             INST("111000110000----", Id::EXIT, Type::Flow, "EXIT"),
             INST("1111000011110---", Id::DEPBAR, Type::Synch, "DEPBAR"),
             INST("0101000011011---", Id::VOTE, Type::Warp, "VOTE"),
+            INST("0101000011100---", Id::VOTE_VTG, Type::Warp, "VOTE_VTG"),
             INST("1110111100010---", Id::SHFL, Type::Warp, "SHFL"),
             INST("0101000011111---", Id::FSWZADD, Type::Warp, "FSWZADD"),
             INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"),
@@ -2039,6 +2104,7 @@ private:
             INST("1110111101010---", Id::ST_L, Type::Memory, "ST_L"),
             INST("101-------------", Id::ST, Type::Memory, "ST"),
             INST("1110111011011---", Id::STG, Type::Memory, "STG"),
+            INST("1110101111111---", Id::RED, Type::Memory, "RED"),
             INST("11101101--------", Id::ATOM, Type::Memory, "ATOM"),
             INST("11101100--------", Id::ATOMS, Type::Memory, "ATOMS"),
             INST("1110111110100---", Id::AL2P, Type::Memory, "AL2P"),
@@ -2063,9 +2129,11 @@ private:
             INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
             INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),
             INST("1110111111010---", Id::ISBERD, Type::Trivial, "ISBERD"),
+            INST("1111000010101---", Id::BAR, Type::Trivial, "BAR"),
             INST("1110111110011---", Id::MEMBAR, Type::Trivial, "MEMBAR"),
             INST("01011111--------", Id::VMAD, Type::Video, "VMAD"),
             INST("0101000011110---", Id::VSETP, Type::Video, "VSETP"),
+            INST("0011101---------", Id::VMNMX, Type::Video, "VMNMX"),
             INST("0011001-1-------", Id::FFMA_IMM, Type::Ffma, "FFMA_IMM"),
             INST("010010011-------", Id::FFMA_CR, Type::Ffma, "FFMA_CR"),
             INST("010100011-------", Id::FFMA_RC, Type::Ffma, "FFMA_RC"),
@@ -2120,7 +2188,8 @@ private:
             INST("0101110100100---", Id::HSETP2_R, Type::HalfSetPredicate, "HSETP2_R"),
             INST("0111111-0-------", Id::HSETP2_IMM, Type::HalfSetPredicate, "HSETP2_IMM"),
             INST("0101110100011---", Id::HSET2_R, Type::HalfSet, "HSET2_R"),
-            INST("010110111010----", Id::FCMP_R, Type::Arithmetic, "FCMP_R"),
+            INST("010110111010----", Id::FCMP_RR, Type::Arithmetic, "FCMP_RR"),
+            INST("010010111010----", Id::FCMP_RC, Type::Arithmetic, "FCMP_RC"),
             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"),
@@ -2134,7 +2203,7 @@ private:
             INST("0100110010011---", Id::MOV_C, Type::Arithmetic, "MOV_C"),
             INST("0101110010011---", Id::MOV_R, Type::Arithmetic, "MOV_R"),
             INST("0011100-10011---", Id::MOV_IMM, Type::Arithmetic, "MOV_IMM"),
-            INST("1111000011001---", Id::MOV_SYS, Type::Trivial, "MOV_SYS"),
+            INST("1111000011001---", Id::S2R, Type::Trivial, "S2R"),
             INST("000000010000----", Id::MOV32_IMM, Type::ArithmeticImmediate, "MOV32_IMM"),
             INST("0100110001100---", Id::FMNMX_C, Type::Arithmetic, "FMNMX_C"),
             INST("0101110001100---", Id::FMNMX_R, Type::Arithmetic, "FMNMX_R"),
@@ -2166,7 +2235,7 @@ private:
             INST("0011011-11111---", Id::SHF_LEFT_IMM, Type::Shift, "SHF_LEFT_IMM"),
             INST("0100110011100---", Id::I2I_C, Type::Conversion, "I2I_C"),
             INST("0101110011100---", Id::I2I_R, Type::Conversion, "I2I_R"),
-            INST("0011101-11100---", Id::I2I_IMM, Type::Conversion, "I2I_IMM"),
+            INST("0011100-11100---", Id::I2I_IMM, Type::Conversion, "I2I_IMM"),
             INST("0100110010111---", Id::I2F_C, Type::Conversion, "I2F_C"),
             INST("0101110010111---", Id::I2F_R, Type::Conversion, "I2F_R"),
             INST("0011100-10111---", Id::I2F_IMM, Type::Conversion, "I2F_IMM"),

src/video_core/engines/shader_header.h

@@ -4,6 +4,9 @@
 
 #pragma once
 
+#include <array>
+#include <optional>
+
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
@@ -16,7 +19,7 @@ enum class OutputTopology : u32 {
     TriangleStrip = 7,
 };
 
-enum class AttributeUse : u8 {
+enum class PixelImap : u8 {
     Unused = 0,
     Constant = 1,
     Perspective = 2,
@@ -24,7 +27,7 @@ enum class AttributeUse : u8 {
 };
 
 // Documentation in:
-// http://download.nvidia.com/open-gpu-doc/Shader-Program-Header/1/Shader-Program-Header.html#ImapTexture
+// http://download.nvidia.com/open-gpu-doc/Shader-Program-Header/1/Shader-Program-Header.html
 struct Header {
     union {
         BitField<0, 5, u32> sph_type;
@@ -59,8 +62,8 @@ struct Header {
     union {
         BitField<0, 12, u32> max_output_vertices;
         BitField<12, 8, u32> store_req_start; // NOTE: not used by geometry shaders.
-        BitField<24, 4, u32> reserved;
-        BitField<12, 8, u32> store_req_end; // NOTE: not used by geometry shaders.
+        BitField<20, 4, u32> reserved;
+        BitField<24, 8, u32> store_req_end; // NOTE: not used by geometry shaders.
     } common4{};
 
     union {
@@ -93,17 +96,20 @@ struct Header {
         struct {
             INSERT_UNION_PADDING_BYTES(3); // ImapSystemValuesA
             INSERT_UNION_PADDING_BYTES(1); // ImapSystemValuesB
+
             union {
-                BitField<0, 2, AttributeUse> x;
-                BitField<2, 2, AttributeUse> y;
-                BitField<4, 2, AttributeUse> w;
-                BitField<6, 2, AttributeUse> z;
+                BitField<0, 2, PixelImap> x;
+                BitField<2, 2, PixelImap> y;
+                BitField<4, 2, PixelImap> z;
+                BitField<6, 2, PixelImap> w;
                 u8 raw;
             } imap_generic_vector[32];
+
             INSERT_UNION_PADDING_BYTES(2);  // ImapColor
             INSERT_UNION_PADDING_BYTES(2);  // ImapSystemValuesC
             INSERT_UNION_PADDING_BYTES(10); // ImapFixedFncTexture[10]
             INSERT_UNION_PADDING_BYTES(2);  // ImapReserved
+
             struct {
                 u32 target;
                 union {
@@ -112,31 +118,30 @@ struct Header {
                     BitField<2, 30, u32> reserved;
                 };
             } omap;
+
             bool IsColorComponentOutputEnabled(u32 render_target, u32 component) const {
                 const u32 bit = render_target * 4 + component;
                 return omap.target & (1 << bit);
             }
-            AttributeUse GetAttributeIndexUse(u32 attribute, u32 index) const {
-                return static_cast<AttributeUse>(
-                    (imap_generic_vector[attribute].raw >> (index * 2)) & 0x03);
-            }
-            AttributeUse GetAttributeUse(u32 attribute) const {
-                AttributeUse result = AttributeUse::Unused;
-                for (u32 i = 0; i < 4; i++) {
-                    const auto index = GetAttributeIndexUse(attribute, i);
-                    if (index == AttributeUse::Unused) {
+
+            PixelImap GetPixelImap(u32 attribute) const {
+                const auto get_index = [this, attribute](u32 index) {
+                    return static_cast<PixelImap>(
+                        (imap_generic_vector[attribute].raw >> (index * 2)) & 3);
+                };
+
+                std::optional<PixelImap> result;
+                for (u32 component = 0; component < 4; ++component) {
+                    const PixelImap index = get_index(component);
+                    if (index == PixelImap::Unused) {
                         continue;
                     }
-                    if (result == AttributeUse::Unused || result == index) {
-                        result = index;
-                        continue;
-                    }
-                    LOG_CRITICAL(HW_GPU, "Generic Attribute Conflict in Interpolation Mode");
-                    if (index == AttributeUse::Perspective) {
-                        result = index;
+                    if (result && result != index) {
+                        LOG_CRITICAL(HW_GPU, "Generic attribute conflict in interpolation mode");
                     }
+                    result = index;
                 }
-                return result;
+                return result.value_or(PixelImap::Unused);
             }
         } ps;
 

src/video_core/gpu.cpp

@@ -7,6 +7,7 @@
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
+#include "core/frontend/emu_window.h"
 #include "core/memory.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/engines/kepler_compute.h"
@@ -16,14 +17,15 @@
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 #include "video_core/renderer_base.h"
+#include "video_core/video_core.h"
 
 namespace Tegra {
 
 MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192));
 
-GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async)
-    : system{system}, renderer{renderer}, is_async{is_async} {
-    auto& rasterizer{renderer.Rasterizer()};
+GPU::GPU(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer_, bool is_async)
+    : system{system}, renderer{std::move(renderer_)}, is_async{is_async} {
+    auto& rasterizer{renderer->Rasterizer()};
     memory_manager = std::make_unique<Tegra::MemoryManager>(system, rasterizer);
     dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
     maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager);
@@ -137,7 +139,7 @@ u64 GPU::GetTicks() const {
 }
 
 void GPU::FlushCommands() {
-    renderer.Rasterizer().FlushCommands();
+    renderer->Rasterizer().FlushCommands();
 }
 
 // Note that, traditionally, methods are treated as 4-byte addressable locations, and hence

src/video_core/gpu.h

@@ -25,8 +25,11 @@ inline u8* FromCacheAddr(CacheAddr cache_addr) {
 }
 
 namespace Core {
-class System;
+namespace Frontend {
+class EmuWindow;
 }
+class System;
+} // namespace Core
 
 namespace VideoCore {
 class RendererBase;
@@ -129,7 +132,8 @@ class MemoryManager;
 
 class GPU {
 public:
-    explicit GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async);
+    explicit GPU(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer,
+                 bool is_async);
 
     virtual ~GPU();
 
@@ -174,6 +178,14 @@ public:
     /// Returns a reference to the GPU DMA pusher.
     Tegra::DmaPusher& DmaPusher();
 
+    VideoCore::RendererBase& Renderer() {
+        return *renderer;
+    }
+
+    const VideoCore::RendererBase& Renderer() const {
+        return *renderer;
+    }
+
     // Waits for the GPU to finish working
     virtual void WaitIdle() const = 0;
 
@@ -258,13 +270,13 @@ public:
     virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    virtual void FlushRegion(CacheAddr addr, u64 size) = 0;
+    virtual void FlushRegion(VAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    virtual void InvalidateRegion(CacheAddr addr, u64 size) = 0;
+    virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
-    virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0;
+    virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
 
 protected:
     virtual void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const = 0;
@@ -287,7 +299,7 @@ private:
 protected:
     std::unique_ptr<Tegra::DmaPusher> dma_pusher;
     Core::System& system;
-    VideoCore::RendererBase& renderer;
+    std::unique_ptr<VideoCore::RendererBase> renderer;
 
 private:
     std::unique_ptr<Tegra::MemoryManager> memory_manager;

src/video_core/gpu_asynch.cpp

@@ -10,13 +10,17 @@
 
 namespace VideoCommon {
 
-GPUAsynch::GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer)
-    : GPU(system, renderer, true), gpu_thread{system} {}
+GPUAsynch::GPUAsynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer_,
+                     std::unique_ptr<Core::Frontend::GraphicsContext>&& context)
+    : GPU(system, std::move(renderer_), true), gpu_thread{system},
+      cpu_context(renderer->GetRenderWindow().CreateSharedContext()),
+      gpu_context(std::move(context)) {}
 
 GPUAsynch::~GPUAsynch() = default;
 
 void GPUAsynch::Start() {
-    gpu_thread.StartThread(renderer, *dma_pusher);
+    cpu_context->MakeCurrent();
+    gpu_thread.StartThread(*renderer, *gpu_context, *dma_pusher);
 }
 
 void GPUAsynch::PushGPUEntries(Tegra::CommandList&& entries) {
@@ -27,15 +31,15 @@ void GPUAsynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     gpu_thread.SwapBuffers(framebuffer);
 }
 
-void GPUAsynch::FlushRegion(CacheAddr addr, u64 size) {
+void GPUAsynch::FlushRegion(VAddr addr, u64 size) {
     gpu_thread.FlushRegion(addr, size);
 }
 
-void GPUAsynch::InvalidateRegion(CacheAddr addr, u64 size) {
+void GPUAsynch::InvalidateRegion(VAddr addr, u64 size) {
     gpu_thread.InvalidateRegion(addr, size);
 }
 
-void GPUAsynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+void GPUAsynch::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     gpu_thread.FlushAndInvalidateRegion(addr, size);
 }
 

src/video_core/gpu_asynch.h

@@ -7,6 +7,10 @@
 #include "video_core/gpu.h"
 #include "video_core/gpu_thread.h"
 
+namespace Core::Frontend {
+class GraphicsContext;
+}
+
 namespace VideoCore {
 class RendererBase;
 } // namespace VideoCore
@@ -16,15 +20,16 @@ namespace VideoCommon {
 /// Implementation of GPU interface that runs the GPU asynchronously
 class GPUAsynch final : public Tegra::GPU {
 public:
-    explicit GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer);
+    explicit GPUAsynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer,
+                       std::unique_ptr<Core::Frontend::GraphicsContext>&& context);
     ~GPUAsynch() override;
 
     void Start() override;
     void PushGPUEntries(Tegra::CommandList&& entries) override;
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-    void FlushRegion(CacheAddr addr, u64 size) override;
-    void InvalidateRegion(CacheAddr addr, u64 size) override;
-    void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+    void FlushRegion(VAddr addr, u64 size) override;
+    void InvalidateRegion(VAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void WaitIdle() const override;
 
 protected:
@@ -32,6 +37,8 @@ protected:
 
 private:
     GPUThread::ThreadManager gpu_thread;
+    std::unique_ptr<Core::Frontend::GraphicsContext> cpu_context;
+    std::unique_ptr<Core::Frontend::GraphicsContext> gpu_context;
 };
 
 } // namespace VideoCommon

src/video_core/gpu_synch.cpp

@@ -7,12 +7,15 @@
 
 namespace VideoCommon {
 
-GPUSynch::GPUSynch(Core::System& system, VideoCore::RendererBase& renderer)
-    : GPU(system, renderer, false) {}
+GPUSynch::GPUSynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer,
+                   std::unique_ptr<Core::Frontend::GraphicsContext>&& context)
+    : GPU(system, std::move(renderer), false), context{std::move(context)} {}
 
 GPUSynch::~GPUSynch() = default;
 
-void GPUSynch::Start() {}
+void GPUSynch::Start() {
+    context->MakeCurrent();
+}
 
 void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
     dma_pusher->Push(std::move(entries));
@@ -20,19 +23,19 @@ void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
 }
 
 void GPUSynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
-    renderer.SwapBuffers(framebuffer);
+    renderer->SwapBuffers(framebuffer);
 }
 
-void GPUSynch::FlushRegion(CacheAddr addr, u64 size) {
-    renderer.Rasterizer().FlushRegion(addr, size);
+void GPUSynch::FlushRegion(VAddr addr, u64 size) {
+    renderer->Rasterizer().FlushRegion(addr, size);
 }
 
-void GPUSynch::InvalidateRegion(CacheAddr addr, u64 size) {
-    renderer.Rasterizer().InvalidateRegion(addr, size);
+void GPUSynch::InvalidateRegion(VAddr addr, u64 size) {
+    renderer->Rasterizer().InvalidateRegion(addr, size);
 }
 
-void GPUSynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
-    renderer.Rasterizer().FlushAndInvalidateRegion(addr, size);
+void GPUSynch::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+    renderer->Rasterizer().FlushAndInvalidateRegion(addr, size);
 }
 
 } // namespace VideoCommon

src/video_core/gpu_synch.h

@@ -6,6 +6,10 @@
 
 #include "video_core/gpu.h"
 
+namespace Core::Frontend {
+class GraphicsContext;
+}
+
 namespace VideoCore {
 class RendererBase;
 } // namespace VideoCore
@@ -15,20 +19,24 @@ namespace VideoCommon {
 /// Implementation of GPU interface that runs the GPU synchronously
 class GPUSynch final : public Tegra::GPU {
 public:
-    explicit GPUSynch(Core::System& system, VideoCore::RendererBase& renderer);
+    explicit GPUSynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer,
+                      std::unique_ptr<Core::Frontend::GraphicsContext>&& context);
     ~GPUSynch() override;
 
     void Start() override;
     void PushGPUEntries(Tegra::CommandList&& entries) override;
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-    void FlushRegion(CacheAddr addr, u64 size) override;
-    void InvalidateRegion(CacheAddr addr, u64 size) override;
-    void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+    void FlushRegion(VAddr addr, u64 size) override;
+    void InvalidateRegion(VAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void WaitIdle() const override {}
 
 protected:
     void TriggerCpuInterrupt([[maybe_unused]] u32 syncpoint_id,
                              [[maybe_unused]] u32 value) const override {}
+
+private:
+    std::unique_ptr<Core::Frontend::GraphicsContext> context;
 };
 
 } // namespace VideoCommon

src/video_core/gpu_thread.cpp

@@ -5,7 +5,7 @@
 #include "common/assert.h"
 #include "common/microprofile.h"
 #include "core/core.h"
-#include "core/frontend/scope_acquire_context.h"
+#include "core/frontend/emu_window.h"
 #include "video_core/dma_pusher.h"
 #include "video_core/gpu.h"
 #include "video_core/gpu_thread.h"
@@ -14,8 +14,8 @@
 namespace VideoCommon::GPUThread {
 
 /// Runs the GPU thread
-static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher,
-                      SynchState& state) {
+static void RunThread(VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
+                      Tegra::DmaPusher& dma_pusher, SynchState& state) {
     MicroProfileOnThreadCreate("GpuThread");
 
     // Wait for first GPU command before acquiring the window context
@@ -27,7 +27,7 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
         return;
     }
 
-    Core::Frontend::ScopeAcquireContext acquire_context{renderer.GetRenderWindow()};
+    auto current_context = context.Acquire();
 
     CommandDataContainer next;
     while (state.is_running) {
@@ -62,8 +62,11 @@ ThreadManager::~ThreadManager() {
     thread.join();
 }
 
-void ThreadManager::StartThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher) {
-    thread = std::thread{RunThread, std::ref(renderer), std::ref(dma_pusher), std::ref(state)};
+void ThreadManager::StartThread(VideoCore::RendererBase& renderer,
+                                Core::Frontend::GraphicsContext& context,
+                                Tegra::DmaPusher& dma_pusher) {
+    thread = std::thread{RunThread, std::ref(renderer), std::ref(context), std::ref(dma_pusher),
+                         std::ref(state)};
 }
 
 void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
@@ -74,15 +77,15 @@ void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     PushCommand(SwapBuffersCommand(framebuffer ? std::make_optional(*framebuffer) : std::nullopt));
 }
 
-void ThreadManager::FlushRegion(CacheAddr addr, u64 size) {
+void ThreadManager::FlushRegion(VAddr addr, u64 size) {
     PushCommand(FlushRegionCommand(addr, size));
 }
 
-void ThreadManager::InvalidateRegion(CacheAddr addr, u64 size) {
+void ThreadManager::InvalidateRegion(VAddr addr, u64 size) {
     system.Renderer().Rasterizer().InvalidateRegion(addr, size);
 }
 
-void ThreadManager::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     // Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important
     InvalidateRegion(addr, size);
 }

src/video_core/gpu_thread.h

@@ -10,7 +10,6 @@
 #include <optional>
 #include <thread>
 #include <variant>
-
 #include "common/threadsafe_queue.h"
 #include "video_core/gpu.h"
 
@@ -20,6 +19,9 @@ class DmaPusher;
 } // namespace Tegra
 
 namespace Core {
+namespace Frontend {
+class GraphicsContext;
+}
 class System;
 } // namespace Core
 
@@ -45,26 +47,26 @@ struct SwapBuffersCommand final {
 
 /// Command to signal to the GPU thread to flush a region
 struct FlushRegionCommand final {
-    explicit constexpr FlushRegionCommand(CacheAddr addr, u64 size) : addr{addr}, size{size} {}
+    explicit constexpr FlushRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {}
 
-    CacheAddr addr;
+    VAddr addr;
     u64 size;
 };
 
 /// Command to signal to the GPU thread to invalidate a region
 struct InvalidateRegionCommand final {
-    explicit constexpr InvalidateRegionCommand(CacheAddr addr, u64 size) : addr{addr}, size{size} {}
+    explicit constexpr InvalidateRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {}
 
-    CacheAddr addr;
+    VAddr addr;
     u64 size;
 };
 
 /// Command to signal to the GPU thread to flush and invalidate a region
 struct FlushAndInvalidateRegionCommand final {
-    explicit constexpr FlushAndInvalidateRegionCommand(CacheAddr addr, u64 size)
+    explicit constexpr FlushAndInvalidateRegionCommand(VAddr addr, u64 size)
         : addr{addr}, size{size} {}
 
-    CacheAddr addr;
+    VAddr addr;
     u64 size;
 };
 
@@ -99,7 +101,8 @@ public:
     ~ThreadManager();
 
     /// Creates and starts the GPU thread.
-    void StartThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher);
+    void StartThread(VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
+                     Tegra::DmaPusher& dma_pusher);
 
     /// Push GPU command entries to be processed
     void SubmitList(Tegra::CommandList&& entries);
@@ -108,13 +111,13 @@ public:
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    void FlushRegion(CacheAddr addr, u64 size);
+    void FlushRegion(VAddr addr, u64 size);
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    void InvalidateRegion(CacheAddr addr, u64 size);
+    void InvalidateRegion(VAddr addr, u64 size);
 
     /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
-    void FlushAndInvalidateRegion(CacheAddr addr, u64 size);
+    void FlushAndInvalidateRegion(VAddr addr, u64 size);
 
     // Wait until the gpu thread is idle.
     void WaitIdle() const;

src/video_core/memory_manager.cpp

@@ -81,12 +81,11 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) {
     ASSERT((gpu_addr & page_mask) == 0);
 
     const u64 aligned_size{Common::AlignUp(size, page_size)};
-    const CacheAddr cache_addr{ToCacheAddr(GetPointer(gpu_addr))};
     const auto cpu_addr = GpuToCpuAddress(gpu_addr);
     ASSERT(cpu_addr);
 
     // Flush and invalidate through the GPU interface, to be asynchronous if possible.
-    system.GPU().FlushAndInvalidateRegion(cache_addr, aligned_size);
+    system.GPU().FlushAndInvalidateRegion(*cpu_addr, aligned_size);
 
     UnmapRange(gpu_addr, aligned_size);
     ASSERT(system.CurrentProcess()
@@ -140,11 +139,11 @@ T MemoryManager::Read(GPUVAddr addr) const {
         return {};
     }
 
-    const u8* page_pointer{page_table.pointers[addr >> page_bits]};
+    const u8* page_pointer{GetPointer(addr)};
     if (page_pointer) {
         // NOTE: Avoid adding any extra logic to this fast-path block
         T value;
-        std::memcpy(&value, &page_pointer[addr & page_mask], sizeof(T));
+        std::memcpy(&value, page_pointer, sizeof(T));
         return value;
     }
 
@@ -167,10 +166,10 @@ void MemoryManager::Write(GPUVAddr addr, T data) {
         return;
     }
 
-    u8* page_pointer{page_table.pointers[addr >> page_bits]};
+    u8* page_pointer{GetPointer(addr)};
     if (page_pointer) {
         // NOTE: Avoid adding any extra logic to this fast-path block
-        std::memcpy(&page_pointer[addr & page_mask], &data, sizeof(T));
+        std::memcpy(page_pointer, &data, sizeof(T));
         return;
     }
 
@@ -201,9 +200,12 @@ u8* MemoryManager::GetPointer(GPUVAddr addr) {
         return {};
     }
 
-    u8* const page_pointer{page_table.pointers[addr >> page_bits]};
-    if (page_pointer != nullptr) {
-        return page_pointer + (addr & page_mask);
+    auto& memory = system.Memory();
+
+    const VAddr page_addr{page_table.backing_addr[addr >> page_bits]};
+
+    if (page_addr != 0) {
+        return memory.GetPointer(page_addr + (addr & page_mask));
     }
 
     LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr);
@@ -215,9 +217,12 @@ const u8* MemoryManager::GetPointer(GPUVAddr addr) const {
         return {};
     }
 
-    const u8* const page_pointer{page_table.pointers[addr >> page_bits]};
-    if (page_pointer != nullptr) {
-        return page_pointer + (addr & page_mask);
+    const auto& memory = system.Memory();
+
+    const VAddr page_addr{page_table.backing_addr[addr >> page_bits]};
+
+    if (page_addr != 0) {
+        return memory.GetPointer(page_addr + (addr & page_mask));
     }
 
     LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr);
@@ -238,17 +243,19 @@ void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, const std::s
     std::size_t page_index{src_addr >> page_bits};
     std::size_t page_offset{src_addr & page_mask};
 
+    auto& memory = system.Memory();
+
     while (remaining_size > 0) {
         const std::size_t copy_amount{
             std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
 
         switch (page_table.attributes[page_index]) {
         case Common::PageType::Memory: {
-            const u8* src_ptr{page_table.pointers[page_index] + page_offset};
+            const VAddr src_addr{page_table.backing_addr[page_index] + page_offset};
             // Flush must happen on the rasterizer interface, such that memory is always synchronous
             // when it is read (even when in asynchronous GPU mode). Fixes Dead Cells title menu.
-            rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount);
-            std::memcpy(dest_buffer, src_ptr, copy_amount);
+            rasterizer.FlushRegion(src_addr, copy_amount);
+            memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount);
             break;
         }
         default:
@@ -268,13 +275,15 @@ void MemoryManager::ReadBlockUnsafe(GPUVAddr src_addr, void* dest_buffer,
     std::size_t page_index{src_addr >> page_bits};
     std::size_t page_offset{src_addr & page_mask};
 
+    auto& memory = system.Memory();
+
     while (remaining_size > 0) {
         const std::size_t copy_amount{
             std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
         const u8* page_pointer = page_table.pointers[page_index];
         if (page_pointer) {
-            const u8* src_ptr{page_pointer + page_offset};
-            std::memcpy(dest_buffer, src_ptr, copy_amount);
+            const VAddr src_addr{page_table.backing_addr[page_index] + page_offset};
+            memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount);
         } else {
             std::memset(dest_buffer, 0, copy_amount);
         }
@@ -290,17 +299,19 @@ void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, const
     std::size_t page_index{dest_addr >> page_bits};
     std::size_t page_offset{dest_addr & page_mask};
 
+    auto& memory = system.Memory();
+
     while (remaining_size > 0) {
         const std::size_t copy_amount{
             std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
 
         switch (page_table.attributes[page_index]) {
         case Common::PageType::Memory: {
-            u8* dest_ptr{page_table.pointers[page_index] + page_offset};
+            const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset};
             // Invalidate must happen on the rasterizer interface, such that memory is always
             // synchronous when it is written (even when in asynchronous GPU mode).
-            rasterizer.InvalidateRegion(ToCacheAddr(dest_ptr), copy_amount);
-            std::memcpy(dest_ptr, src_buffer, copy_amount);
+            rasterizer.InvalidateRegion(dest_addr, copy_amount);
+            memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount);
             break;
         }
         default:
@@ -320,13 +331,15 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer,
     std::size_t page_index{dest_addr >> page_bits};
     std::size_t page_offset{dest_addr & page_mask};
 
+    auto& memory = system.Memory();
+
     while (remaining_size > 0) {
         const std::size_t copy_amount{
             std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
         u8* page_pointer = page_table.pointers[page_index];
         if (page_pointer) {
-            u8* dest_ptr{page_pointer + page_offset};
-            std::memcpy(dest_ptr, src_buffer, copy_amount);
+            const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset};
+            memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount);
         }
         page_index++;
         page_offset = 0;
@@ -336,33 +349,9 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer,
 }
 
 void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, const std::size_t size) {
-    std::size_t remaining_size{size};
-    std::size_t page_index{src_addr >> page_bits};
-    std::size_t page_offset{src_addr & page_mask};
-
-    while (remaining_size > 0) {
-        const std::size_t copy_amount{
-            std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)};
-
-        switch (page_table.attributes[page_index]) {
-        case Common::PageType::Memory: {
-            // Flush must happen on the rasterizer interface, such that memory is always synchronous
-            // when it is copied (even when in asynchronous GPU mode).
-            const u8* src_ptr{page_table.pointers[page_index] + page_offset};
-            rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount);
-            WriteBlock(dest_addr, src_ptr, copy_amount);
-            break;
-        }
-        default:
-            UNREACHABLE();
-        }
-
-        page_index++;
-        page_offset = 0;
-        dest_addr += static_cast<VAddr>(copy_amount);
-        src_addr += static_cast<VAddr>(copy_amount);
-        remaining_size -= copy_amount;
-    }
+    std::vector<u8> tmp_buffer(size);
+    ReadBlock(src_addr, tmp_buffer.data(), size);
+    WriteBlock(dest_addr, tmp_buffer.data(), size);
 }
 
 void MemoryManager::CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, const std::size_t size) {
@@ -371,6 +360,12 @@ void MemoryManager::CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, const
     WriteBlockUnsafe(dest_addr, tmp_buffer.data(), size);
 }
 
+bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) {
+    const VAddr addr = page_table.backing_addr[gpu_addr >> page_bits];
+    const std::size_t page = (addr & Memory::PAGE_MASK) + size;
+    return page <= Memory::PAGE_SIZE;
+}
+
 void MemoryManager::MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type,
                              VAddr backing_addr) {
     LOG_DEBUG(HW_GPU, "Mapping {} onto {:016X}-{:016X}", fmt::ptr(memory), base * page_size,

src/video_core/memory_manager.h

@@ -97,6 +97,11 @@ public:
     void WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer, std::size_t size);
     void CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size);
 
+    /**
+     * IsGranularRange checks if a gpu region can be simply read with a pointer
+     */
+    bool IsGranularRange(GPUVAddr gpu_addr, std::size_t size);
+
 private:
     using VMAMap = std::map<GPUVAddr, VirtualMemoryArea>;
     using VMAHandle = VMAMap::const_iterator;

src/video_core/query_cache.h

@@ -98,12 +98,12 @@ public:
                                                       static_cast<QueryCache&>(*this),
                                                       VideoCore::QueryType::SamplesPassed}}} {}
 
-    void InvalidateRegion(CacheAddr addr, std::size_t size) {
+    void InvalidateRegion(VAddr addr, std::size_t size) {
         std::unique_lock lock{mutex};
         FlushAndRemoveRegion(addr, size);
     }
 
-    void FlushRegion(CacheAddr addr, std::size_t size) {
+    void FlushRegion(VAddr addr, std::size_t size) {
         std::unique_lock lock{mutex};
         FlushAndRemoveRegion(addr, size);
     }
@@ -117,14 +117,16 @@ public:
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) {
         std::unique_lock lock{mutex};
         auto& memory_manager = system.GPU().MemoryManager();
-        const auto host_ptr = memory_manager.GetPointer(gpu_addr);
+        const std::optional<VAddr> cpu_addr_opt = memory_manager.GpuToCpuAddress(gpu_addr);
+        ASSERT(cpu_addr_opt);
+        VAddr cpu_addr = *cpu_addr_opt;
 
-        CachedQuery* query = TryGet(ToCacheAddr(host_ptr));
+        CachedQuery* query = TryGet(cpu_addr);
         if (!query) {
-            const auto cpu_addr = memory_manager.GpuToCpuAddress(gpu_addr);
-            ASSERT_OR_EXECUTE(cpu_addr, return;);
+            ASSERT_OR_EXECUTE(cpu_addr_opt, return;);
+            const auto host_ptr = memory_manager.GetPointer(gpu_addr);
 
-            query = Register(type, *cpu_addr, host_ptr, timestamp.has_value());
+            query = Register(type, cpu_addr, host_ptr, timestamp.has_value());
         }
 
         query->BindCounter(Stream(type).Current(), timestamp);
@@ -173,11 +175,11 @@ protected:
 
 private:
     /// Flushes a memory range to guest memory and removes it from the cache.
-    void FlushAndRemoveRegion(CacheAddr addr, std::size_t size) {
+    void FlushAndRemoveRegion(VAddr addr, std::size_t size) {
         const u64 addr_begin = static_cast<u64>(addr);
         const u64 addr_end = addr_begin + static_cast<u64>(size);
         const auto in_range = [addr_begin, addr_end](CachedQuery& query) {
-            const u64 cache_begin = query.GetCacheAddr();
+            const u64 cache_begin = query.GetCpuAddr();
             const u64 cache_end = cache_begin + query.SizeInBytes();
             return cache_begin < addr_end && addr_begin < cache_end;
         };
@@ -193,7 +195,7 @@ private:
                 if (!in_range(query)) {
                     continue;
                 }
-                rasterizer.UpdatePagesCachedCount(query.CpuAddr(), query.SizeInBytes(), -1);
+                rasterizer.UpdatePagesCachedCount(query.GetCpuAddr(), query.SizeInBytes(), -1);
                 query.Flush();
             }
             contents.erase(std::remove_if(std::begin(contents), std::end(contents), in_range),
@@ -204,22 +206,21 @@ private:
     /// Registers the passed parameters as cached and returns a pointer to the stored cached query.
     CachedQuery* Register(VideoCore::QueryType type, VAddr cpu_addr, u8* host_ptr, bool timestamp) {
         rasterizer.UpdatePagesCachedCount(cpu_addr, CachedQuery::SizeInBytes(timestamp), 1);
-        const u64 page = static_cast<u64>(ToCacheAddr(host_ptr)) >> PAGE_SHIFT;
+        const u64 page = static_cast<u64>(cpu_addr) >> PAGE_SHIFT;
         return &cached_queries[page].emplace_back(static_cast<QueryCache&>(*this), type, cpu_addr,
                                                   host_ptr);
     }
 
     /// Tries to a get a cached query. Returns nullptr on failure.
-    CachedQuery* TryGet(CacheAddr addr) {
+    CachedQuery* TryGet(VAddr addr) {
         const u64 page = static_cast<u64>(addr) >> PAGE_SHIFT;
         const auto it = cached_queries.find(page);
         if (it == std::end(cached_queries)) {
             return nullptr;
         }
         auto& contents = it->second;
-        const auto found =
-            std::find_if(std::begin(contents), std::end(contents),
-                         [addr](auto& query) { return query.GetCacheAddr() == addr; });
+        const auto found = std::find_if(std::begin(contents), std::end(contents),
+                                        [addr](auto& query) { return query.GetCpuAddr() == addr; });
         return found != std::end(contents) ? &*found : nullptr;
     }
 
@@ -323,14 +324,10 @@ public:
         timestamp = timestamp_;
     }
 
-    VAddr CpuAddr() const noexcept {
+    VAddr GetCpuAddr() const noexcept {
         return cpu_addr;
     }
 
-    CacheAddr GetCacheAddr() const noexcept {
-        return ToCacheAddr(host_ptr);
-    }
-
     u64 SizeInBytes() const noexcept {
         return SizeInBytes(timestamp.has_value());
     }

src/video_core/rasterizer_cache.h

@@ -18,22 +18,14 @@
 
 class RasterizerCacheObject {
 public:
-    explicit RasterizerCacheObject(const u8* host_ptr)
-        : host_ptr{host_ptr}, cache_addr{ToCacheAddr(host_ptr)} {}
+    explicit RasterizerCacheObject(const VAddr cpu_addr) : cpu_addr{cpu_addr} {}
 
     virtual ~RasterizerCacheObject();
 
-    CacheAddr GetCacheAddr() const {
-        return cache_addr;
+    VAddr GetCpuAddr() const {
+        return cpu_addr;
     }
 
-    const u8* GetHostPtr() const {
-        return host_ptr;
-    }
-
-    /// Gets the address of the shader in guest memory, required for cache management
-    virtual VAddr GetCpuAddr() const = 0;
-
     /// Gets the size of the shader in guest memory, required for cache management
     virtual std::size_t GetSizeInBytes() const = 0;
 
@@ -68,8 +60,7 @@ private:
     bool is_registered{};      ///< Whether the object is currently registered with the cache
     bool is_dirty{};           ///< Whether the object is dirty (out of sync with guest memory)
     u64 last_modified_ticks{}; ///< When the object was last modified, used for in-order flushing
-    const u8* host_ptr{};      ///< Pointer to the memory backing this cached region
-    CacheAddr cache_addr{};    ///< Cache address memory, unique from emulated virtual address space
+    VAddr cpu_addr{};          ///< Cpu address memory, unique from emulated virtual address space
 };
 
 template <class T>
@@ -80,7 +71,7 @@ public:
     explicit RasterizerCache(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {}
 
     /// Write any cached resources overlapping the specified region back to memory
-    void FlushRegion(CacheAddr addr, std::size_t size) {
+    void FlushRegion(VAddr addr, std::size_t size) {
         std::lock_guard lock{mutex};
 
         const auto& objects{GetSortedObjectsFromRegion(addr, size)};
@@ -90,7 +81,7 @@ public:
     }
 
     /// Mark the specified region as being invalidated
-    void InvalidateRegion(CacheAddr addr, u64 size) {
+    void InvalidateRegion(VAddr addr, u64 size) {
         std::lock_guard lock{mutex};
 
         const auto& objects{GetSortedObjectsFromRegion(addr, size)};
@@ -114,27 +105,20 @@ public:
 
 protected:
     /// Tries to get an object from the cache with the specified cache address
-    T TryGet(CacheAddr addr) const {
+    T TryGet(VAddr addr) const {
         const auto iter = map_cache.find(addr);
         if (iter != map_cache.end())
             return iter->second;
         return nullptr;
     }
 
-    T TryGet(const void* addr) const {
-        const auto iter = map_cache.find(ToCacheAddr(addr));
-        if (iter != map_cache.end())
-            return iter->second;
-        return nullptr;
-    }
-
     /// Register an object into the cache
     virtual void Register(const T& object) {
         std::lock_guard lock{mutex};
 
         object->SetIsRegistered(true);
         interval_cache.add({GetInterval(object), ObjectSet{object}});
-        map_cache.insert({object->GetCacheAddr(), object});
+        map_cache.insert({object->GetCpuAddr(), object});
         rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), 1);
     }
 
@@ -144,7 +128,7 @@ protected:
 
         object->SetIsRegistered(false);
         rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), -1);
-        const CacheAddr addr = object->GetCacheAddr();
+        const VAddr addr = object->GetCpuAddr();
         interval_cache.subtract({GetInterval(object), ObjectSet{object}});
         map_cache.erase(addr);
     }
@@ -173,7 +157,7 @@ protected:
 
 private:
     /// Returns a list of cached objects from the specified memory region, ordered by access time
-    std::vector<T> GetSortedObjectsFromRegion(CacheAddr addr, u64 size) {
+    std::vector<T> GetSortedObjectsFromRegion(VAddr addr, u64 size) {
         if (size == 0) {
             return {};
         }
@@ -197,13 +181,13 @@ private:
     }
 
     using ObjectSet = std::set<T>;
-    using ObjectCache = std::unordered_map<CacheAddr, T>;
-    using IntervalCache = boost::icl::interval_map<CacheAddr, ObjectSet>;
+    using ObjectCache = std::unordered_map<VAddr, T>;
+    using IntervalCache = boost::icl::interval_map<VAddr, ObjectSet>;
     using ObjectInterval = typename IntervalCache::interval_type;
 
     static auto GetInterval(const T& object) {
-        return ObjectInterval::right_open(object->GetCacheAddr(),
-                                          object->GetCacheAddr() + object->GetSizeInBytes());
+        return ObjectInterval::right_open(object->GetCpuAddr(),
+                                          object->GetCpuAddr() + object->GetSizeInBytes());
     }
 
     ObjectCache map_cache;

src/video_core/rasterizer_interface.h

@@ -53,14 +53,14 @@ public:
     virtual void FlushAll() = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    virtual void FlushRegion(CacheAddr addr, u64 size) = 0;
+    virtual void FlushRegion(VAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    virtual void InvalidateRegion(CacheAddr addr, u64 size) = 0;
+    virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     /// and invalidated
-    virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0;
+    virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
 
     /// Notify the rasterizer to send all written commands to the host GPU.
     virtual void FlushCommands() = 0;

src/video_core/renderer_base.h

@@ -46,7 +46,8 @@ public:
 
     /// Draws the latest frame to the window waiting timeout_ms for a frame to arrive (Renderer
     /// specific implementation)
-    virtual void TryPresent(int timeout_ms) = 0;
+    /// Returns true if a frame was drawn
+    virtual bool TryPresent(int timeout_ms) = 0;
 
     // Getter/setter functions:
     // ------------------------

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -21,8 +21,8 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
 MICROPROFILE_DEFINE(OpenGL_Buffer_Download, "OpenGL", "Buffer Download", MP_RGB(192, 192, 128));
 
-CachedBufferBlock::CachedBufferBlock(CacheAddr cache_addr, const std::size_t size)
-    : VideoCommon::BufferBlock{cache_addr, size} {
+CachedBufferBlock::CachedBufferBlock(VAddr cpu_addr, const std::size_t size)
+    : VideoCommon::BufferBlock{cpu_addr, size} {
     gl_buffer.Create();
     glNamedBufferData(gl_buffer.handle, static_cast<GLsizeiptr>(size), nullptr, GL_DYNAMIC_DRAW);
 }
@@ -47,41 +47,39 @@ OGLBufferCache::~OGLBufferCache() {
     glDeleteBuffers(static_cast<GLsizei>(std::size(cbufs)), std::data(cbufs));
 }
 
-Buffer OGLBufferCache::CreateBlock(CacheAddr cache_addr, std::size_t size) {
-    return std::make_shared<CachedBufferBlock>(cache_addr, size);
+Buffer OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
+    return std::make_shared<CachedBufferBlock>(cpu_addr, size);
 }
 
 void OGLBufferCache::WriteBarrier() {
     glMemoryBarrier(GL_ALL_BARRIER_BITS);
 }
 
-const GLuint* OGLBufferCache::ToHandle(const Buffer& buffer) {
+GLuint OGLBufferCache::ToHandle(const Buffer& buffer) {
     return buffer->GetHandle();
 }
 
-const GLuint* OGLBufferCache::GetEmptyBuffer(std::size_t) {
-    static const GLuint null_buffer = 0;
-    return &null_buffer;
+GLuint OGLBufferCache::GetEmptyBuffer(std::size_t) {
+    return 0;
 }
 
 void OGLBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                                      const u8* data) {
-    glNamedBufferSubData(*buffer->GetHandle(), static_cast<GLintptr>(offset),
+    glNamedBufferSubData(buffer->GetHandle(), static_cast<GLintptr>(offset),
                          static_cast<GLsizeiptr>(size), data);
 }
 
 void OGLBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                                        u8* data) {
     MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
-    glGetNamedBufferSubData(*buffer->GetHandle(), static_cast<GLintptr>(offset),
+    glGetNamedBufferSubData(buffer->GetHandle(), static_cast<GLintptr>(offset),
                             static_cast<GLsizeiptr>(size), data);
 }
 
 void OGLBufferCache::CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
                                std::size_t dst_offset, std::size_t size) {
-    glCopyNamedBufferSubData(*src->GetHandle(), *dst->GetHandle(),
-                             static_cast<GLintptr>(src_offset), static_cast<GLintptr>(dst_offset),
-                             static_cast<GLsizeiptr>(size));
+    glCopyNamedBufferSubData(src->GetHandle(), dst->GetHandle(), static_cast<GLintptr>(src_offset),
+                             static_cast<GLintptr>(dst_offset), static_cast<GLsizeiptr>(size));
 }
 
 OGLBufferCache::BufferInfo OGLBufferCache::ConstBufferUpload(const void* raw_pointer,
@@ -89,7 +87,7 @@ OGLBufferCache::BufferInfo OGLBufferCache::ConstBufferUpload(const void* raw_poi
     DEBUG_ASSERT(cbuf_cursor < std::size(cbufs));
     const GLuint& cbuf = cbufs[cbuf_cursor++];
     glNamedBufferSubData(cbuf, 0, static_cast<GLsizeiptr>(size), raw_pointer);
-    return {&cbuf, 0};
+    return {cbuf, 0};
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -31,15 +31,15 @@ using GenericBufferCache = VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuf
 
 class CachedBufferBlock : public VideoCommon::BufferBlock {
 public:
-    explicit CachedBufferBlock(CacheAddr cache_addr, const std::size_t size);
+    explicit CachedBufferBlock(VAddr cpu_addr, const std::size_t size);
     ~CachedBufferBlock();
 
-    const GLuint* GetHandle() const {
-        return &gl_buffer.handle;
+    GLuint GetHandle() const {
+        return gl_buffer.handle;
     }
 
 private:
-    OGLBuffer gl_buffer{};
+    OGLBuffer gl_buffer;
 };
 
 class OGLBufferCache final : public GenericBufferCache {
@@ -48,19 +48,19 @@ public:
                             const Device& device, std::size_t stream_size);
     ~OGLBufferCache();
 
-    const GLuint* GetEmptyBuffer(std::size_t) override;
+    GLuint GetEmptyBuffer(std::size_t) override;
 
     void Acquire() noexcept {
         cbuf_cursor = 0;
     }
 
 protected:
-    Buffer CreateBlock(CacheAddr cache_addr, std::size_t size) override;
+    Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override;
+
+    GLuint ToHandle(const Buffer& buffer) override;
 
     void WriteBarrier() override;
 
-    const GLuint* ToHandle(const Buffer& buffer) override;
-
     void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                          const u8* data) override;
 

src/video_core/renderer_opengl/gl_device.cpp

@@ -87,7 +87,7 @@ u32 Extract(u32& base, u32& num, u32 amount, std::optional<GLenum> limit = {}) {
 std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> BuildBaseBindings() noexcept {
     std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> bindings;
 
-    static std::array<std::size_t, 5> stage_swizzle = {0, 1, 2, 3, 4};
+    static constexpr std::array<std::size_t, 5> stage_swizzle{0, 1, 2, 3, 4};
     const u32 total_ubos = GetInteger<u32>(GL_MAX_UNIFORM_BUFFER_BINDINGS);
     const u32 total_ssbos = GetInteger<u32>(GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS);
     const u32 total_samplers = GetInteger<u32>(GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS);
@@ -131,6 +131,31 @@ std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> BuildBaseBindin
     return bindings;
 }
 
+bool IsASTCSupported() {
+    static constexpr std::array formats = {
+        GL_COMPRESSED_RGBA_ASTC_4x4_KHR,           GL_COMPRESSED_RGBA_ASTC_5x4_KHR,
+        GL_COMPRESSED_RGBA_ASTC_5x5_KHR,           GL_COMPRESSED_RGBA_ASTC_6x5_KHR,
+        GL_COMPRESSED_RGBA_ASTC_6x6_KHR,           GL_COMPRESSED_RGBA_ASTC_8x5_KHR,
+        GL_COMPRESSED_RGBA_ASTC_8x6_KHR,           GL_COMPRESSED_RGBA_ASTC_8x8_KHR,
+        GL_COMPRESSED_RGBA_ASTC_10x5_KHR,          GL_COMPRESSED_RGBA_ASTC_10x6_KHR,
+        GL_COMPRESSED_RGBA_ASTC_10x8_KHR,          GL_COMPRESSED_RGBA_ASTC_10x10_KHR,
+        GL_COMPRESSED_RGBA_ASTC_12x10_KHR,         GL_COMPRESSED_RGBA_ASTC_12x12_KHR,
+        GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR,   GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x4_KHR,
+        GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR,   GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x5_KHR,
+        GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR,   GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x5_KHR,
+        GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR,   GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x8_KHR,
+        GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x5_KHR,  GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x6_KHR,
+        GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR,  GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR,
+        GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x10_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR,
+    };
+    return std::find_if_not(formats.begin(), formats.end(), [](GLenum format) {
+               GLint supported;
+               glGetInternalformativ(GL_TEXTURE_2D, format, GL_INTERNALFORMAT_SUPPORTED, 1,
+                                     &supported);
+               return supported == GL_TRUE;
+           }) == formats.end();
+}
+
 } // Anonymous namespace
 
 Device::Device() : base_bindings{BuildBaseBindings()} {
@@ -152,6 +177,7 @@ Device::Device() : base_bindings{BuildBaseBindings()} {
     has_shader_ballot = GLAD_GL_ARB_shader_ballot;
     has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
     has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
+    has_astc = IsASTCSupported();
     has_variable_aoffi = TestVariableAoffi();
     has_component_indexing_bug = is_amd;
     has_precise_bug = TestPreciseBug();

src/video_core/renderer_opengl/gl_device.h

@@ -64,6 +64,10 @@ public:
         return has_image_load_formatted;
     }
 
+    bool HasASTC() const {
+        return has_astc;
+    }
+
     bool HasVariableAoffi() const {
         return has_variable_aoffi;
     }
@@ -97,6 +101,7 @@ private:
     bool has_shader_ballot{};
     bool has_vertex_viewport_layer{};
     bool has_image_load_formatted{};
+    bool has_astc{};
     bool has_variable_aoffi{};
     bool has_component_indexing_bug{};
     bool has_precise_bug{};

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -140,8 +140,8 @@ void RasterizerOpenGL::SetupVertexFormat() {
         const auto attrib = gpu.regs.vertex_attrib_format[index];
         const auto gl_index = static_cast<GLuint>(index);
 
-        // Ignore invalid attributes.
-        if (!attrib.IsValid()) {
+        // Disable constant attributes.
+        if (attrib.IsConstant()) {
             glDisableVertexAttribArray(gl_index);
             continue;
         }
@@ -188,10 +188,8 @@ void RasterizerOpenGL::SetupVertexBuffer() {
         ASSERT(end > start);
         const u64 size = end - start + 1;
         const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
-
-        // Bind the vertex array to the buffer at the current offset.
-        vertex_array_pushbuffer.SetVertexBuffer(static_cast<GLuint>(index), vertex_buffer,
-                                                vertex_buffer_offset, vertex_array.stride);
+        glBindVertexBuffer(static_cast<GLuint>(index), vertex_buffer, vertex_buffer_offset,
+                           vertex_array.stride);
     }
 }
 
@@ -222,7 +220,7 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() {
     const auto& regs = system.GPU().Maxwell3D().regs;
     const std::size_t size = CalculateIndexBufferSize();
     const auto [buffer, offset] = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
-    vertex_array_pushbuffer.SetIndexBuffer(buffer);
+    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer);
     return offset;
 }
 
@@ -345,7 +343,7 @@ void RasterizerOpenGL::ConfigureFramebuffers() {
 
     texture_cache.GuardRenderTargets(true);
 
-    View depth_surface = texture_cache.GetDepthBufferSurface(true);
+    View depth_surface = texture_cache.GetDepthBufferSurface();
 
     const auto& regs = gpu.regs;
     UNIMPLEMENTED_IF(regs.rt_separate_frag_data == 0);
@@ -354,7 +352,7 @@ void RasterizerOpenGL::ConfigureFramebuffers() {
     FramebufferCacheKey key;
     const auto colors_count = static_cast<std::size_t>(regs.rt_control.count);
     for (std::size_t index = 0; index < colors_count; ++index) {
-        View color_surface{texture_cache.GetColorBufferSurface(index, true)};
+        View color_surface{texture_cache.GetColorBufferSurface(index)};
         if (!color_surface) {
             continue;
         }
@@ -386,11 +384,14 @@ void RasterizerOpenGL::ConfigureClearFramebuffer(bool using_color_fb, bool using
     texture_cache.GuardRenderTargets(true);
     View color_surface;
     if (using_color_fb) {
-        color_surface = texture_cache.GetColorBufferSurface(regs.clear_buffers.RT, false);
+        const std::size_t index = regs.clear_buffers.RT;
+        color_surface = texture_cache.GetColorBufferSurface(index);
+        texture_cache.MarkColorBufferInUse(index);
     }
     View depth_surface;
     if (using_depth_fb || using_stencil_fb) {
-        depth_surface = texture_cache.GetDepthBufferSurface(false);
+        depth_surface = texture_cache.GetDepthBufferSurface();
+        texture_cache.MarkDepthBufferInUse();
     }
     texture_cache.GuardRenderTargets(false);
 
@@ -444,6 +445,7 @@ void RasterizerOpenGL::Clear() {
     }
 
     SyncRasterizeEnable();
+    SyncStencilTestState();
 
     if (regs.clear_flags.scissor) {
         SyncScissorTest();
@@ -492,6 +494,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     SyncPrimitiveRestart();
     SyncScissorTest();
     SyncPointState();
+    SyncLineState();
     SyncPolygonOffset();
     SyncAlphaTest();
     SyncFramebufferSRGB();
@@ -519,7 +522,6 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
 
     // Prepare vertex array format.
     SetupVertexFormat();
-    vertex_array_pushbuffer.Setup();
 
     // Upload vertex and index data.
     SetupVertexBuffer();
@@ -529,17 +531,13 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
         index_buffer_offset = SetupIndexBuffer();
     }
 
-    // Prepare packed bindings.
-    bind_ubo_pushbuffer.Setup();
-    bind_ssbo_pushbuffer.Setup();
-
     // Setup emulation uniform buffer.
     GLShader::MaxwellUniformData ubo;
     ubo.SetFromRegs(gpu);
     const auto [buffer, offset] =
         buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
-    bind_ubo_pushbuffer.Push(EmulationUniformBlockBinding, buffer, offset,
-                             static_cast<GLsizeiptr>(sizeof(ubo)));
+    glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, buffer, offset,
+                      static_cast<GLsizeiptr>(sizeof(ubo)));
 
     // Setup shaders and their used resources.
     texture_cache.GuardSamplers(true);
@@ -552,11 +550,6 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     // Signal the buffer cache that we are not going to upload more things.
     buffer_cache.Unmap();
 
-    // Now that we are no longer uploading data, we can safely bind the buffers to OpenGL.
-    vertex_array_pushbuffer.Bind();
-    bind_ubo_pushbuffer.Bind();
-    bind_ssbo_pushbuffer.Bind();
-
     program_manager.BindGraphicsPipeline();
 
     if (texture_cache.TextureBarrier()) {
@@ -625,17 +618,11 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
         (Maxwell::MaxConstBufferSize + device.GetUniformBufferAlignment());
     buffer_cache.Map(buffer_size);
 
-    bind_ubo_pushbuffer.Setup();
-    bind_ssbo_pushbuffer.Setup();
-
     SetupComputeConstBuffers(kernel);
     SetupComputeGlobalMemory(kernel);
 
     buffer_cache.Unmap();
 
-    bind_ubo_pushbuffer.Bind();
-    bind_ssbo_pushbuffer.Bind();
-
     const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
     glDispatchCompute(launch_desc.grid_dim_x, launch_desc.grid_dim_y, launch_desc.grid_dim_z);
     ++num_queued_commands;
@@ -652,9 +639,9 @@ void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCore::QueryType type,
 
 void RasterizerOpenGL::FlushAll() {}
 
-void RasterizerOpenGL::FlushRegion(CacheAddr addr, u64 size) {
+void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-    if (!addr || !size) {
+    if (addr == 0 || size == 0) {
         return;
     }
     texture_cache.FlushRegion(addr, size);
@@ -662,9 +649,9 @@ void RasterizerOpenGL::FlushRegion(CacheAddr addr, u64 size) {
     query_cache.FlushRegion(addr, size);
 }
 
-void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) {
+void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-    if (!addr || !size) {
+    if (addr == 0 || size == 0) {
         return;
     }
     texture_cache.InvalidateRegion(addr, size);
@@ -673,7 +660,7 @@ void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) {
     query_cache.InvalidateRegion(addr, size);
 }
 
-void RasterizerOpenGL::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     if (Settings::values.use_accurate_gpu_emulation) {
         FlushRegion(addr, size);
     }
@@ -712,8 +699,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
 
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
 
-    const auto surface{
-        texture_cache.TryFindFramebufferSurface(system.Memory().GetPointer(framebuffer_addr))};
+    const auto surface{texture_cache.TryFindFramebufferSurface(framebuffer_addr)};
     if (!surface) {
         return {};
     }
@@ -767,8 +753,8 @@ void RasterizerOpenGL::SetupConstBuffer(u32 binding, const Tegra::Engines::Const
                                         const ConstBufferEntry& entry) {
     if (!buffer.enabled) {
         // Set values to zero to unbind buffers
-        bind_ubo_pushbuffer.Push(binding, buffer_cache.GetEmptyBuffer(sizeof(float)), 0,
-                                 sizeof(float));
+        glBindBufferRange(GL_UNIFORM_BUFFER, binding, buffer_cache.GetEmptyBuffer(sizeof(float)), 0,
+                          sizeof(float));
         return;
     }
 
@@ -779,7 +765,7 @@ void RasterizerOpenGL::SetupConstBuffer(u32 binding, const Tegra::Engines::Const
     const auto alignment = device.GetUniformBufferAlignment();
     const auto [cbuf, offset] = buffer_cache.UploadMemory(buffer.address, size, alignment, false,
                                                           device.HasFastBufferSubData());
-    bind_ubo_pushbuffer.Push(binding, cbuf, offset, size);
+    glBindBufferRange(GL_UNIFORM_BUFFER, binding, cbuf, offset, size);
 }
 
 void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, const Shader& shader) {
@@ -815,7 +801,8 @@ void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& e
     const auto alignment{device.GetShaderStorageBufferAlignment()};
     const auto [ssbo, buffer_offset] =
         buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.IsWritten());
-    bind_ssbo_pushbuffer.Push(binding, ssbo, buffer_offset, static_cast<GLsizeiptr>(size));
+    glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, ssbo, buffer_offset,
+                      static_cast<GLsizeiptr>(size));
 }
 
 void RasterizerOpenGL::SetupDrawTextures(std::size_t stage_index, const Shader& shader) {
@@ -1052,12 +1039,8 @@ void RasterizerOpenGL::SyncStencilTestState() {
     flags[Dirty::StencilTest] = false;
 
     const auto& regs = gpu.regs;
-    if (!regs.stencil_enable) {
-        glDisable(GL_STENCIL_TEST);
-        return;
-    }
+    oglEnable(GL_STENCIL_TEST, regs.stencil_enable);
 
-    glEnable(GL_STENCIL_TEST);
     glStencilFuncSeparate(GL_FRONT, MaxwellToGL::ComparisonOp(regs.stencil_front_func_func),
                           regs.stencil_front_func_ref, regs.stencil_front_func_mask);
     glStencilOpSeparate(GL_FRONT, MaxwellToGL::StencilOp(regs.stencil_front_op_fail),
@@ -1312,6 +1295,19 @@ void RasterizerOpenGL::SyncPointState() {
     glDisable(GL_PROGRAM_POINT_SIZE);
 }
 
+void RasterizerOpenGL::SyncLineState() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::LineWidth]) {
+        return;
+    }
+    flags[Dirty::LineWidth] = false;
+
+    const auto& regs = gpu.regs;
+    oglEnable(GL_LINE_SMOOTH, regs.line_smooth_enable);
+    glLineWidth(regs.line_smooth_enable ? regs.line_width_smooth : regs.line_width_aliased);
+}
+
 void RasterizerOpenGL::SyncPolygonOffset() {
     auto& gpu = system.GPU().Maxwell3D();
     auto& flags = gpu.dirty.flags;
@@ -1419,7 +1415,7 @@ void RasterizerOpenGL::EndTransformFeedback() {
         const GPUVAddr gpu_addr = binding.Address();
         const std::size_t size = binding.buffer_size;
         const auto [dest_buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
-        glCopyNamedBufferSubData(handle, *dest_buffer, 0, offset, static_cast<GLsizeiptr>(size));
+        glCopyNamedBufferSubData(handle, dest_buffer, 0, offset, static_cast<GLsizeiptr>(size));
     }
 }
 

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -65,9 +65,9 @@ public:
     void ResetCounter(VideoCore::QueryType type) override;
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
     void FlushAll() override;
-    void FlushRegion(CacheAddr addr, u64 size) override;
-    void InvalidateRegion(CacheAddr addr, u64 size) override;
-    void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+    void FlushRegion(VAddr addr, u64 size) override;
+    void InvalidateRegion(VAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void FlushCommands() override;
     void TickFrame() override;
     bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
@@ -171,6 +171,9 @@ private:
     /// Syncs the point state to match the guest state
     void SyncPointState();
 
+    /// Syncs the line state to match the guest state
+    void SyncLineState();
+
     /// Syncs the rasterizer enable state to match the guest state
     void SyncRasterizeEnable();
 
@@ -228,9 +231,7 @@ private:
     static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
     OGLBufferCache buffer_cache;
 
-    VertexArrayPushBuffer vertex_array_pushbuffer{state_tracker};
-    BindBuffersRangePushBuffer bind_ubo_pushbuffer{GL_UNIFORM_BUFFER};
-    BindBuffersRangePushBuffer bind_ssbo_pushbuffer{GL_SHADER_STORAGE_BUFFER};
+    GLint vertex_binding = 0;
 
     std::array<OGLBuffer, Tegra::Engines::Maxwell3D::Regs::NumTransformFeedbackBuffers>
         transform_feedback_buffers;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -34,6 +34,8 @@
 namespace OpenGL {
 
 using Tegra::Engines::ShaderType;
+using VideoCommon::Shader::CompileDepth;
+using VideoCommon::Shader::CompilerSettings;
 using VideoCommon::Shader::ProgramCode;
 using VideoCommon::Shader::Registry;
 using VideoCommon::Shader::ShaderIR;
@@ -43,7 +45,7 @@ namespace {
 constexpr u32 STAGE_MAIN_OFFSET = 10;
 constexpr u32 KERNEL_MAIN_OFFSET = 0;
 
-constexpr VideoCommon::Shader::CompilerSettings COMPILER_SETTINGS{};
+constexpr CompilerSettings COMPILER_SETTINGS{CompileDepth::FullDecompile};
 
 /// Gets the address for the specified shader stage program
 GPUVAddr GetShaderAddress(Core::System& system, Maxwell::ShaderProgram program) {
@@ -214,11 +216,11 @@ std::unordered_set<GLenum> GetSupportedFormats() {
 
 } // Anonymous namespace
 
-CachedShader::CachedShader(const u8* host_ptr, VAddr cpu_addr, std::size_t size_in_bytes,
+CachedShader::CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
                            std::shared_ptr<VideoCommon::Shader::Registry> registry,
                            ShaderEntries entries, std::shared_ptr<OGLProgram> program)
-    : RasterizerCacheObject{host_ptr}, registry{std::move(registry)}, entries{std::move(entries)},
-      cpu_addr{cpu_addr}, size_in_bytes{size_in_bytes}, program{std::move(program)} {}
+    : RasterizerCacheObject{cpu_addr}, registry{std::move(registry)}, entries{std::move(entries)},
+      size_in_bytes{size_in_bytes}, program{std::move(program)} {}
 
 CachedShader::~CachedShader() = default;
 
@@ -254,9 +256,8 @@ Shader CachedShader::CreateStageFromMemory(const ShaderParameters& params,
     entry.bindless_samplers = registry->GetBindlessSamplers();
     params.disk_cache.SaveEntry(std::move(entry));
 
-    return std::shared_ptr<CachedShader>(new CachedShader(params.host_ptr, params.cpu_addr,
-                                                          size_in_bytes, std::move(registry),
-                                                          MakeEntries(ir), std::move(program)));
+    return std::shared_ptr<CachedShader>(new CachedShader(
+        params.cpu_addr, size_in_bytes, std::move(registry), MakeEntries(ir), std::move(program)));
 }
 
 Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code) {
@@ -279,17 +280,16 @@ Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, Prog
     entry.bindless_samplers = registry->GetBindlessSamplers();
     params.disk_cache.SaveEntry(std::move(entry));
 
-    return std::shared_ptr<CachedShader>(new CachedShader(params.host_ptr, params.cpu_addr,
-                                                          size_in_bytes, std::move(registry),
-                                                          MakeEntries(ir), std::move(program)));
+    return std::shared_ptr<CachedShader>(new CachedShader(
+        params.cpu_addr, size_in_bytes, std::move(registry), MakeEntries(ir), std::move(program)));
 }
 
 Shader CachedShader::CreateFromCache(const ShaderParameters& params,
                                      const PrecompiledShader& precompiled_shader,
                                      std::size_t size_in_bytes) {
-    return std::shared_ptr<CachedShader>(new CachedShader(
-        params.host_ptr, params.cpu_addr, size_in_bytes, precompiled_shader.registry,
-        precompiled_shader.entries, precompiled_shader.program));
+    return std::shared_ptr<CachedShader>(
+        new CachedShader(params.cpu_addr, size_in_bytes, precompiled_shader.registry,
+                         precompiled_shader.entries, precompiled_shader.program));
 }
 
 ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
@@ -327,8 +327,7 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
 
     const auto worker = [&](Core::Frontend::GraphicsContext* context, std::size_t begin,
                             std::size_t end) {
-        context->MakeCurrent();
-        SCOPE_EXIT({ return context->DoneCurrent(); });
+        const auto scope = context->Acquire();
 
         for (std::size_t i = begin; i < end; ++i) {
             if (stop_loading) {
@@ -450,12 +449,14 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
     const GPUVAddr address{GetShaderAddress(system, program)};
 
     // Look up shader in the cache based on address
-    const auto host_ptr{memory_manager.GetPointer(address)};
-    Shader shader{TryGet(host_ptr)};
+    const auto cpu_addr{memory_manager.GpuToCpuAddress(address)};
+    Shader shader{cpu_addr ? TryGet(*cpu_addr) : nullptr};
     if (shader) {
         return last_shaders[static_cast<std::size_t>(program)] = shader;
     }
 
+    const auto host_ptr{memory_manager.GetPointer(address)};
+
     // No shader found - create a new one
     ProgramCode code{GetShaderCode(memory_manager, address, host_ptr)};
     ProgramCode code_b;
@@ -466,9 +467,9 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
 
     const auto unique_identifier = GetUniqueIdentifier(
         GetShaderType(program), program == Maxwell::ShaderProgram::VertexA, code, code_b);
-    const auto cpu_addr{*memory_manager.GpuToCpuAddress(address)};
-    const ShaderParameters params{system,   disk_cache, device,
-                                  cpu_addr, host_ptr,   unique_identifier};
+
+    const ShaderParameters params{system,    disk_cache, device,
+                                  *cpu_addr, host_ptr,   unique_identifier};
 
     const auto found = runtime_cache.find(unique_identifier);
     if (found == runtime_cache.end()) {
@@ -485,18 +486,20 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
 
 Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
     auto& memory_manager{system.GPU().MemoryManager()};
-    const auto host_ptr{memory_manager.GetPointer(code_addr)};
-    auto kernel = TryGet(host_ptr);
+    const auto cpu_addr{memory_manager.GpuToCpuAddress(code_addr)};
+
+    auto kernel = cpu_addr ? TryGet(*cpu_addr) : nullptr;
     if (kernel) {
         return kernel;
     }
 
+    const auto host_ptr{memory_manager.GetPointer(code_addr)};
     // No kernel found, create a new one
     auto code{GetShaderCode(memory_manager, code_addr, host_ptr)};
     const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
-    const auto cpu_addr{*memory_manager.GpuToCpuAddress(code_addr)};
-    const ShaderParameters params{system,   disk_cache, device,
-                                  cpu_addr, host_ptr,   unique_identifier};
+
+    const ShaderParameters params{system,    disk_cache, device,
+                                  *cpu_addr, host_ptr,   unique_identifier};
 
     const auto found = runtime_cache.find(unique_identifier);
     if (found == runtime_cache.end()) {

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -65,11 +65,6 @@ public:
     /// Gets the GL program handle for the shader
     GLuint GetHandle() const;
 
-    /// Returns the guest CPU address of the shader
-    VAddr GetCpuAddr() const override {
-        return cpu_addr;
-    }
-
     /// Returns the size in bytes of the shader
     std::size_t GetSizeInBytes() const override {
         return size_in_bytes;
@@ -90,13 +85,12 @@ public:
                                   std::size_t size_in_bytes);
 
 private:
-    explicit CachedShader(const u8* host_ptr, VAddr cpu_addr, std::size_t size_in_bytes,
+    explicit CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
                           std::shared_ptr<VideoCommon::Shader::Registry> registry,
                           ShaderEntries entries, std::shared_ptr<OGLProgram> program);
 
     std::shared_ptr<VideoCommon::Shader::Registry> registry;
     ShaderEntries entries;
-    VAddr cpu_addr = 0;
     std::size_t size_in_bytes = 0;
     std::shared_ptr<OGLProgram> program;
 };

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -31,11 +31,11 @@ namespace {
 
 using Tegra::Engines::ShaderType;
 using Tegra::Shader::Attribute;
-using Tegra::Shader::AttributeUse;
 using Tegra::Shader::Header;
 using Tegra::Shader::IpaInterpMode;
 using Tegra::Shader::IpaMode;
 using Tegra::Shader::IpaSampleMode;
+using Tegra::Shader::PixelImap;
 using Tegra::Shader::Register;
 using VideoCommon::Shader::BuildTransformFeedback;
 using VideoCommon::Shader::Registry;
@@ -702,20 +702,19 @@ private:
         code.AddNewLine();
     }
 
-    std::string GetInputFlags(AttributeUse attribute) {
+    const char* GetInputFlags(PixelImap attribute) {
         switch (attribute) {
-        case AttributeUse::Perspective:
-            // Default, Smooth
-            return {};
-        case AttributeUse::Constant:
-            return "flat ";
-        case AttributeUse::ScreenLinear:
-            return "noperspective ";
-        default:
-        case AttributeUse::Unused:
-            UNIMPLEMENTED_MSG("Unknown attribute usage index={}", static_cast<u32>(attribute));
-            return {};
+        case PixelImap::Perspective:
+            return "smooth";
+        case PixelImap::Constant:
+            return "flat";
+        case PixelImap::ScreenLinear:
+            return "noperspective";
+        case PixelImap::Unused:
+            break;
         }
+        UNIMPLEMENTED_MSG("Unknown attribute usage index={}", static_cast<int>(attribute));
+        return {};
     }
 
     void DeclareInputAttributes() {
@@ -749,8 +748,8 @@ private:
 
         std::string suffix;
         if (stage == ShaderType::Fragment) {
-            const auto input_mode{header.ps.GetAttributeUse(location)};
-            if (skip_unused && input_mode == AttributeUse::Unused) {
+            const auto input_mode{header.ps.GetPixelImap(location)};
+            if (input_mode == PixelImap::Unused) {
                 return;
             }
             suffix = GetInputFlags(input_mode);
@@ -927,7 +926,7 @@ private:
                 const u32 address{generic_base + index * generic_stride + element * element_stride};
 
                 const bool declared = stage != ShaderType::Fragment ||
-                                      header.ps.GetAttributeUse(index) != AttributeUse::Unused;
+                                      header.ps.GetPixelImap(index) != PixelImap::Unused;
                 const std::string value =
                     declared ? ReadAttribute(attribute, element).AsFloat() : "0.0f";
                 code.AddLine("case 0x{:X}U: return {};", address, value);
@@ -1142,8 +1141,7 @@ private:
                                     GetSwizzle(element)),
                         Type::Float};
             case ShaderType::Fragment:
-                return {element == 3 ? "1.0f" : ("gl_FragCoord"s + GetSwizzle(element)),
-                        Type::Float};
+                return {"gl_FragCoord"s + GetSwizzle(element), Type::Float};
             default:
                 UNREACHABLE();
             }
@@ -1821,15 +1819,17 @@ private:
     }
 
     Expression HMergeH0(Operation operation) {
-        std::string dest = VisitOperand(operation, 0).AsUint();
-        std::string src = VisitOperand(operation, 1).AsUint();
-        return {fmt::format("(({} & 0x0000FFFFU) | ({} & 0xFFFF0000U))", src, dest), Type::Uint};
+        const std::string dest = VisitOperand(operation, 0).AsUint();
+        const std::string src = VisitOperand(operation, 1).AsUint();
+        return {fmt::format("vec2(unpackHalf2x16({}).x, unpackHalf2x16({}).y)", src, dest),
+                Type::HalfFloat};
     }
 
     Expression HMergeH1(Operation operation) {
-        std::string dest = VisitOperand(operation, 0).AsUint();
-        std::string src = VisitOperand(operation, 1).AsUint();
-        return {fmt::format("(({} & 0x0000FFFFU) | ({} & 0xFFFF0000U))", dest, src), Type::Uint};
+        const std::string dest = VisitOperand(operation, 0).AsUint();
+        const std::string src = VisitOperand(operation, 1).AsUint();
+        return {fmt::format("vec2(unpackHalf2x16({}).x, unpackHalf2x16({}).y)", dest, src),
+                Type::HalfFloat};
     }
 
     Expression HPack2(Operation operation) {
@@ -2119,8 +2119,14 @@ private:
             return {};
         }
         return {fmt::format("atomic{}({}, {})", opname, Visit(operation[0]).GetCode(),
-                            Visit(operation[1]).As(type)),
-                type};
+                            Visit(operation[1]).AsUint()),
+                Type::Uint};
+    }
+
+    template <const std::string_view& opname, Type type>
+    Expression Reduce(Operation operation) {
+        code.AddLine("{};", Atomic<opname, type>(operation).GetCode());
+        return {};
     }
 
     Expression Branch(Operation operation) {
@@ -2479,6 +2485,20 @@ private:
         &GLSLDecompiler::Atomic<Func::Or, Type::Int>,
         &GLSLDecompiler::Atomic<Func::Xor, Type::Int>,
 
+        &GLSLDecompiler::Reduce<Func::Add, Type::Uint>,
+        &GLSLDecompiler::Reduce<Func::Min, Type::Uint>,
+        &GLSLDecompiler::Reduce<Func::Max, Type::Uint>,
+        &GLSLDecompiler::Reduce<Func::And, Type::Uint>,
+        &GLSLDecompiler::Reduce<Func::Or, Type::Uint>,
+        &GLSLDecompiler::Reduce<Func::Xor, Type::Uint>,
+
+        &GLSLDecompiler::Reduce<Func::Add, Type::Int>,
+        &GLSLDecompiler::Reduce<Func::Min, Type::Int>,
+        &GLSLDecompiler::Reduce<Func::Max, Type::Int>,
+        &GLSLDecompiler::Reduce<Func::And, Type::Int>,
+        &GLSLDecompiler::Reduce<Func::Or, Type::Int>,
+        &GLSLDecompiler::Reduce<Func::Xor, Type::Int>,
+
         &GLSLDecompiler::Branch,
         &GLSLDecompiler::BranchIndirect,
         &GLSLDecompiler::PushFlowStack,

src/video_core/renderer_opengl/gl_state_tracker.cpp

@@ -185,6 +185,12 @@ void SetupDirtyPointSize(Tables& tables) {
     tables[0][OFF(point_sprite_enable)] = PointSize;
 }
 
+void SetupDirtyLineWidth(Tables& tables) {
+    tables[0][OFF(line_width_smooth)] = LineWidth;
+    tables[0][OFF(line_width_aliased)] = LineWidth;
+    tables[0][OFF(line_smooth_enable)] = LineWidth;
+}
+
 void SetupDirtyClipControl(Tables& tables) {
     auto& table = tables[0];
     table[OFF(screen_y_control)] = ClipControl;
@@ -233,6 +239,7 @@ void StateTracker::Initialize() {
     SetupDirtyLogicOp(tables);
     SetupDirtyFragmentClampColor(tables);
     SetupDirtyPointSize(tables);
+    SetupDirtyLineWidth(tables);
     SetupDirtyClipControl(tables);
     SetupDirtyDepthClampEnabled(tables);
     SetupDirtyMisc(tables);

src/video_core/renderer_opengl/gl_state_tracker.h

@@ -78,6 +78,7 @@ enum : u8 {
     LogicOp,
     FragmentClampColor,
     PointSize,
+    LineWidth,
     ClipControl,
     DepthClampEnabled,
 

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -24,7 +24,6 @@ using Tegra::Texture::SwizzleSource;
 using VideoCore::MortonSwizzleMode;
 
 using VideoCore::Surface::PixelFormat;
-using VideoCore::Surface::SurfaceCompression;
 using VideoCore::Surface::SurfaceTarget;
 using VideoCore::Surface::SurfaceType;
 
@@ -37,102 +36,100 @@ namespace {
 
 struct FormatTuple {
     GLint internal_format;
-    GLenum format;
-    GLenum type;
-    bool compressed;
+    GLenum format = GL_NONE;
+    GLenum type = GL_NONE;
 };
 
 constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format_tuples = {{
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false},                        // ABGR8U
-    {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE, false},                                      // ABGR8S
-    {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE, false},                         // ABGR8UI
-    {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, false},                        // B5G6R5U
-    {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, false},                  // A2B10G10R10U
-    {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, false},                    // A1B5G5R5U
-    {GL_R8, GL_RED, GL_UNSIGNED_BYTE, false},                                       // R8U
-    {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE, false},                             // R8UI
-    {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, false},                                    // RGBA16F
-    {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT, false},                                 // RGBA16U
-    {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT, false},                                    // RGBA16S
-    {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT, false},                       // RGBA16UI
-    {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, false},            // R11FG11FB10F
-    {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, false},                         // RGBA32UI
-    {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true},     // DXT1
-    {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true},     // DXT23
-    {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true},     // DXT45
-    {GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, true},               // DXN1
-    {GL_COMPRESSED_RG_RGTC2, GL_RG, GL_UNSIGNED_INT_8_8_8_8, true},                 // DXN2UNORM
-    {GL_COMPRESSED_SIGNED_RG_RGTC2, GL_RG, GL_INT, true},                           // DXN2SNORM
-    {GL_COMPRESSED_RGBA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true},        // BC7U
-    {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true}, // BC6H_UF16
-    {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true},   // BC6H_SF16
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_4X4
-    {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE, false},                                   // BGRA8
-    {GL_RGBA32F, GL_RGBA, GL_FLOAT, false},                                         // RGBA32F
-    {GL_RG32F, GL_RG, GL_FLOAT, false},                                             // RG32F
-    {GL_R32F, GL_RED, GL_FLOAT, false},                                             // R32F
-    {GL_R16F, GL_RED, GL_HALF_FLOAT, false},                                        // R16F
-    {GL_R16, GL_RED, GL_UNSIGNED_SHORT, false},                                     // R16U
-    {GL_R16_SNORM, GL_RED, GL_SHORT, false},                                        // R16S
-    {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT, false},                           // R16UI
-    {GL_R16I, GL_RED_INTEGER, GL_SHORT, false},                                     // R16I
-    {GL_RG16, GL_RG, GL_UNSIGNED_SHORT, false},                                     // RG16
-    {GL_RG16F, GL_RG, GL_HALF_FLOAT, false},                                        // RG16F
-    {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT, false},                           // RG16UI
-    {GL_RG16I, GL_RG_INTEGER, GL_SHORT, false},                                     // RG16I
-    {GL_RG16_SNORM, GL_RG, GL_SHORT, false},                                        // RG16S
-    {GL_RGB32F, GL_RGB, GL_FLOAT, false},                                           // RGB32F
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false},                 // RGBA8_SRGB
-    {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, false},                                       // RG8U
-    {GL_RG8_SNORM, GL_RG, GL_BYTE, false},                                          // RG8S
-    {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, false},                             // RG32UI
-    {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT, false},                                     // RGBX16F
-    {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, false},                             // R32UI
-    {GL_R32I, GL_RED_INTEGER, GL_INT, false},                                       // R32I
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_8X8
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_8X5
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_5X4
-    {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE, false},                            // BGRA8
+    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV},             // ABGR8U
+    {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE},                           // ABGR8S
+    {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE},              // ABGR8UI
+    {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV},             // B5G6R5U
+    {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV},       // A2B10G10R10U
+    {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV},         // A1B5G5R5U
+    {GL_R8, GL_RED, GL_UNSIGNED_BYTE},                            // R8U
+    {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE},                  // R8UI
+    {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT},                         // RGBA16F
+    {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT},                      // RGBA16U
+    {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT},                         // RGBA16S
+    {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT},            // RGBA16UI
+    {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV}, // R11FG11FB10F
+    {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT},              // RGBA32UI
+    {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT},                           // DXT1
+    {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT},                           // DXT23
+    {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT},                           // DXT45
+    {GL_COMPRESSED_RED_RGTC1},                                    // DXN1
+    {GL_COMPRESSED_RG_RGTC2},                                     // DXN2UNORM
+    {GL_COMPRESSED_SIGNED_RG_RGTC2},                              // DXN2SNORM
+    {GL_COMPRESSED_RGBA_BPTC_UNORM},                              // BC7U
+    {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT},                      // BC6H_UF16
+    {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT},                        // BC6H_SF16
+    {GL_COMPRESSED_RGBA_ASTC_4x4_KHR},                            // ASTC_2D_4X4
+    {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE},                        // BGRA8
+    {GL_RGBA32F, GL_RGBA, GL_FLOAT},                              // RGBA32F
+    {GL_RG32F, GL_RG, GL_FLOAT},                                  // RG32F
+    {GL_R32F, GL_RED, GL_FLOAT},                                  // R32F
+    {GL_R16F, GL_RED, GL_HALF_FLOAT},                             // R16F
+    {GL_R16, GL_RED, GL_UNSIGNED_SHORT},                          // R16U
+    {GL_R16_SNORM, GL_RED, GL_SHORT},                             // R16S
+    {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT},                // R16UI
+    {GL_R16I, GL_RED_INTEGER, GL_SHORT},                          // R16I
+    {GL_RG16, GL_RG, GL_UNSIGNED_SHORT},                          // RG16
+    {GL_RG16F, GL_RG, GL_HALF_FLOAT},                             // RG16F
+    {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT},                // RG16UI
+    {GL_RG16I, GL_RG_INTEGER, GL_SHORT},                          // RG16I
+    {GL_RG16_SNORM, GL_RG, GL_SHORT},                             // RG16S
+    {GL_RGB32F, GL_RGB, GL_FLOAT},                                // RGB32F
+    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV},      // RGBA8_SRGB
+    {GL_RG8, GL_RG, GL_UNSIGNED_BYTE},                            // RG8U
+    {GL_RG8_SNORM, GL_RG, GL_BYTE},                               // RG8S
+    {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT},                  // RG32UI
+    {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT},                          // RGBX16F
+    {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT},                  // R32UI
+    {GL_R32I, GL_RED_INTEGER, GL_INT},                            // R32I
+    {GL_COMPRESSED_RGBA_ASTC_8x8_KHR},                            // ASTC_2D_8X8
+    {GL_COMPRESSED_RGBA_ASTC_8x5_KHR},                            // ASTC_2D_8X5
+    {GL_COMPRESSED_RGBA_ASTC_5x4_KHR},                            // ASTC_2D_5X4
+    {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE},                 // BGRA8
     // Compressed sRGB formats
-    {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT1_SRGB
-    {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT23_SRGB
-    {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT45_SRGB
-    {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true},    // BC7U_SRGB
-    {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV, false},                        // R4G4B4A4U
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_4X4_SRGB
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_8X8_SRGB
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_8X5_SRGB
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_5X4_SRGB
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},             // ASTC_2D_5X5
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_5X5_SRGB
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},             // ASTC_2D_10X8
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_10X8_SRGB
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},             // ASTC_2D_6X6
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_6X6_SRGB
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},             // ASTC_2D_10X10
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_10X10_SRGB
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},             // ASTC_2D_12X12
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_12X12_SRGB
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},             // ASTC_2D_8X6
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_8X6_SRGB
-    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},             // ASTC_2D_6X5
-    {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false},      // ASTC_2D_6X5_SRGB
-    {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV, false}, // E5B9G9R9F
+    {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT},           // DXT1_SRGB
+    {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT},           // DXT23_SRGB
+    {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT},           // DXT45_SRGB
+    {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM},              // BC7U_SRGB
+    {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV}, // R4G4B4A4U
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR},          // ASTC_2D_4X4_SRGB
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x8_KHR},          // ASTC_2D_8X8_SRGB
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x5_KHR},          // ASTC_2D_8X5_SRGB
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x4_KHR},          // ASTC_2D_5X4_SRGB
+    {GL_COMPRESSED_RGBA_ASTC_5x5_KHR},                  // ASTC_2D_5X5
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR},          // ASTC_2D_5X5_SRGB
+    {GL_COMPRESSED_RGBA_ASTC_10x8_KHR},                 // ASTC_2D_10X8
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR},         // ASTC_2D_10X8_SRGB
+    {GL_COMPRESSED_RGBA_ASTC_6x6_KHR},                  // ASTC_2D_6X6
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR},          // ASTC_2D_6X6_SRGB
+    {GL_COMPRESSED_RGBA_ASTC_10x10_KHR},                // ASTC_2D_10X10
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR},        // ASTC_2D_10X10_SRGB
+    {GL_COMPRESSED_RGBA_ASTC_12x12_KHR},                // ASTC_2D_12X12
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR},        // ASTC_2D_12X12_SRGB
+    {GL_COMPRESSED_RGBA_ASTC_8x6_KHR},                  // ASTC_2D_8X6
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR},          // ASTC_2D_8X6_SRGB
+    {GL_COMPRESSED_RGBA_ASTC_6x5_KHR},                  // ASTC_2D_6X5
+    {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x5_KHR},          // ASTC_2D_6X5_SRGB
+    {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV},  // E5B9G9R9F
 
     // Depth formats
-    {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT, false},         // Z32F
-    {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT, false}, // Z16
+    {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT},         // Z32F
+    {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // Z16
 
     // DepthStencil formats
-    {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, false},               // Z24S8
-    {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, false},               // S8Z24
-    {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV, false}, // Z32FS8
+    {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8},               // Z24S8
+    {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8},               // S8Z24
+    {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV}, // Z32FS8
 }};
 
 const FormatTuple& GetFormatTuple(PixelFormat pixel_format) {
     ASSERT(static_cast<std::size_t>(pixel_format) < tex_format_tuples.size());
-    const auto& format{tex_format_tuples[static_cast<std::size_t>(pixel_format)]};
-    return format;
+    return tex_format_tuples[static_cast<std::size_t>(pixel_format)];
 }
 
 GLenum GetTextureTarget(const SurfaceTarget& target) {
@@ -242,13 +239,20 @@ OGLTexture CreateTexture(const SurfaceParams& params, GLenum target, GLenum inte
 
 } // Anonymous namespace
 
-CachedSurface::CachedSurface(const GPUVAddr gpu_addr, const SurfaceParams& params)
-    : VideoCommon::SurfaceBase<View>(gpu_addr, params) {
-    const auto& tuple{GetFormatTuple(params.pixel_format)};
-    internal_format = tuple.internal_format;
-    format = tuple.format;
-    type = tuple.type;
-    is_compressed = tuple.compressed;
+CachedSurface::CachedSurface(const GPUVAddr gpu_addr, const SurfaceParams& params,
+                             bool is_astc_supported)
+    : VideoCommon::SurfaceBase<View>(gpu_addr, params, is_astc_supported) {
+    if (is_converted) {
+        internal_format = params.srgb_conversion ? GL_SRGB8_ALPHA8 : GL_RGBA8;
+        format = GL_RGBA;
+        type = GL_UNSIGNED_BYTE;
+    } else {
+        const auto& tuple{GetFormatTuple(params.pixel_format)};
+        internal_format = tuple.internal_format;
+        format = tuple.format;
+        type = tuple.type;
+        is_compressed = params.IsCompressed();
+    }
     target = GetTextureTarget(params.target);
     texture = CreateTexture(params, target, internal_format, texture_buffer);
     DecorateSurfaceName();
@@ -264,7 +268,7 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) {
 
     if (params.IsBuffer()) {
         glGetNamedBufferSubData(texture_buffer.handle, 0,
-                                static_cast<GLsizeiptr>(params.GetHostSizeInBytes()),
+                                static_cast<GLsizeiptr>(params.GetHostSizeInBytes(false)),
                                 staging_buffer.data());
         return;
     }
@@ -272,9 +276,10 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) {
     SCOPE_EXIT({ glPixelStorei(GL_PACK_ROW_LENGTH, 0); });
 
     for (u32 level = 0; level < params.emulated_levels; ++level) {
-        glPixelStorei(GL_PACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level)));
+        glPixelStorei(GL_PACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level, is_converted)));
         glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.GetMipWidth(level)));
-        const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level);
+        const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted);
+
         u8* const mip_data = staging_buffer.data() + mip_offset;
         const GLsizei size = static_cast<GLsizei>(params.GetHostMipmapSize(level));
         if (is_compressed) {
@@ -294,14 +299,10 @@ void CachedSurface::UploadTexture(const std::vector<u8>& staging_buffer) {
 }
 
 void CachedSurface::UploadTextureMipmap(u32 level, const std::vector<u8>& staging_buffer) {
-    glPixelStorei(GL_UNPACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level)));
+    glPixelStorei(GL_UNPACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level, is_converted)));
     glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.GetMipWidth(level)));
 
-    auto compression_type = params.GetCompressionType();
-
-    const std::size_t mip_offset = compression_type == SurfaceCompression::Converted
-                                       ? params.GetConvertedMipmapOffset(level)
-                                       : params.GetHostMipmapLevelOffset(level);
+    const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted);
     const u8* buffer{staging_buffer.data() + mip_offset};
     if (is_compressed) {
         const auto image_size{static_cast<GLsizei>(params.GetHostMipmapSize(level))};
@@ -410,14 +411,13 @@ CachedSurfaceView::~CachedSurfaceView() = default;
 void CachedSurfaceView::Attach(GLenum attachment, GLenum target) const {
     ASSERT(params.num_levels == 1);
 
-    const GLuint texture = surface.GetTexture();
     if (params.num_layers > 1) {
         // Layered framebuffer attachments
         UNIMPLEMENTED_IF(params.base_layer != 0);
 
         switch (params.target) {
         case SurfaceTarget::Texture2DArray:
-            glFramebufferTexture(target, attachment, texture, params.base_level);
+            glFramebufferTexture(target, attachment, GetTexture(), 0);
             break;
         default:
             UNIMPLEMENTED();
@@ -426,6 +426,7 @@ void CachedSurfaceView::Attach(GLenum attachment, GLenum target) const {
     }
 
     const GLenum view_target = surface.GetTarget();
+    const GLuint texture = surface.GetTexture();
     switch (surface.GetSurfaceParams().target) {
     case SurfaceTarget::Texture1D:
         glFramebufferTexture1D(target, attachment, view_target, texture, params.base_level);
@@ -482,7 +483,7 @@ OGLTextureView CachedSurfaceView::CreateTextureView() const {
 TextureCacheOpenGL::TextureCacheOpenGL(Core::System& system,
                                        VideoCore::RasterizerInterface& rasterizer,
                                        const Device& device, StateTracker& state_tracker)
-    : TextureCacheBase{system, rasterizer}, state_tracker{state_tracker} {
+    : TextureCacheBase{system, rasterizer, device.HasASTC()}, state_tracker{state_tracker} {
     src_framebuffer.Create();
     dst_framebuffer.Create();
 }
@@ -490,7 +491,7 @@ TextureCacheOpenGL::TextureCacheOpenGL(Core::System& system,
 TextureCacheOpenGL::~TextureCacheOpenGL() = default;
 
 Surface TextureCacheOpenGL::CreateSurface(GPUVAddr gpu_addr, const SurfaceParams& params) {
-    return std::make_shared<CachedSurface>(gpu_addr, params);
+    return std::make_shared<CachedSurface>(gpu_addr, params, is_astc_supported);
 }
 
 void TextureCacheOpenGL::ImageCopy(Surface& src_surface, Surface& dst_surface,
@@ -596,7 +597,7 @@ void TextureCacheOpenGL::BufferCopy(Surface& src_surface, Surface& dst_surface)
 
     glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle);
 
-    if (source_format.compressed) {
+    if (src_surface->IsCompressed()) {
         glGetCompressedTextureImage(src_surface->GetTexture(), 0, static_cast<GLsizei>(source_size),
                                     nullptr);
     } else {
@@ -610,7 +611,7 @@ void TextureCacheOpenGL::BufferCopy(Surface& src_surface, Surface& dst_surface)
     const GLsizei width = static_cast<GLsizei>(dst_params.width);
     const GLsizei height = static_cast<GLsizei>(dst_params.height);
     const GLsizei depth = static_cast<GLsizei>(dst_params.depth);
-    if (dest_format.compressed) {
+    if (dst_surface->IsCompressed()) {
         LOG_CRITICAL(HW_GPU, "Compressed buffer copy is unimplemented!");
         UNREACHABLE();
     } else {

src/video_core/renderer_opengl/gl_texture_cache.h

@@ -37,7 +37,7 @@ class CachedSurface final : public VideoCommon::SurfaceBase<View> {
     friend CachedSurfaceView;
 
 public:
-    explicit CachedSurface(GPUVAddr gpu_addr, const SurfaceParams& params);
+    explicit CachedSurface(GPUVAddr gpu_addr, const SurfaceParams& params, bool is_astc_supported);
     ~CachedSurface();
 
     void UploadTexture(const std::vector<u8>& staging_buffer) override;
@@ -51,6 +51,10 @@ public:
         return texture.handle;
     }
 
+    bool IsCompressed() const {
+        return is_compressed;
+    }
+
 protected:
     void DecorateSurfaceName() override;
 

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -30,8 +30,6 @@ namespace OpenGL {
 
 namespace {
 
-// If the size of this is too small, it ends up creating a soft cap on FPS as the renderer will have
-// to wait on available presentation frames.
 constexpr std::size_t SWAP_CHAIN_SIZE = 3;
 
 struct Frame {
@@ -214,7 +212,7 @@ public:
     std::deque<Frame*> present_queue;
     Frame* previous_frame{};
 
-    FrameMailbox() : has_debug_tool{HasDebugTool()} {
+    FrameMailbox() {
         for (auto& frame : swap_chain) {
             free_queue.push(&frame);
         }
@@ -285,13 +283,9 @@ public:
         std::unique_lock lock{swap_chain_lock};
         present_queue.push_front(frame);
         present_cv.notify_one();
-
-        DebugNotifyNextFrame();
     }
 
     Frame* TryGetPresentFrame(int timeout_ms) {
-        DebugWaitForNextFrame();
-
         std::unique_lock lock{swap_chain_lock};
         // wait for new entries in the present_queue
         present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
@@ -317,38 +311,12 @@ public:
         previous_frame = frame;
         return frame;
     }
-
-private:
-    std::mutex debug_synch_mutex;
-    std::condition_variable debug_synch_condition;
-    std::atomic_int frame_for_debug{};
-    const bool has_debug_tool; // When true, using a GPU debugger, so keep frames in lock-step
-
-    /// Signal that a new frame is available (called from GPU thread)
-    void DebugNotifyNextFrame() {
-        if (!has_debug_tool) {
-            return;
-        }
-        frame_for_debug++;
-        std::lock_guard lock{debug_synch_mutex};
-        debug_synch_condition.notify_one();
-    }
-
-    /// Wait for a new frame to be available (called from presentation thread)
-    void DebugWaitForNextFrame() {
-        if (!has_debug_tool) {
-            return;
-        }
-        const int last_frame = frame_for_debug;
-        std::unique_lock lock{debug_synch_mutex};
-        debug_synch_condition.wait(lock,
-                                   [this, last_frame] { return frame_for_debug > last_frame; });
-    }
 };
 
-RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system)
-    : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system},
-      frame_mailbox{std::make_unique<FrameMailbox>()} {}
+RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system,
+                               Core::Frontend::GraphicsContext& context)
+    : RendererBase{emu_window}, emu_window{emu_window}, system{system}, context{context},
+      has_debug_tool{HasDebugTool()} {}
 
 RendererOpenGL::~RendererOpenGL() = default;
 
@@ -356,8 +324,6 @@ MICROPROFILE_DEFINE(OpenGL_RenderFrame, "OpenGL", "Render Frame", MP_RGB(128, 12
 MICROPROFILE_DEFINE(OpenGL_WaitPresent, "OpenGL", "Wait For Present", MP_RGB(128, 128, 128));
 
 void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
-    render_window.PollEvents();
-
     if (!framebuffer) {
         return;
     }
@@ -413,6 +379,13 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
         m_current_frame++;
         rasterizer->TickFrame();
     }
+
+    render_window.PollEvents();
+    if (has_debug_tool) {
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0);
+        Present(0);
+        context.SwapBuffers();
+    }
 }
 
 void RendererOpenGL::PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer) {
@@ -480,6 +453,8 @@ void RendererOpenGL::LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color
 }
 
 void RendererOpenGL::InitOpenGLObjects() {
+    frame_mailbox = std::make_unique<FrameMailbox>();
+
     glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
                  0.0f);
 
@@ -692,12 +667,21 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
 }
 
-void RendererOpenGL::TryPresent(int timeout_ms) {
+bool RendererOpenGL::TryPresent(int timeout_ms) {
+    if (has_debug_tool) {
+        LOG_DEBUG(Render_OpenGL,
+                  "Skipping presentation because we are presenting on the main context");
+        return false;
+    }
+    return Present(timeout_ms);
+}
+
+bool RendererOpenGL::Present(int timeout_ms) {
     const auto& layout = render_window.GetFramebufferLayout();
     auto frame = frame_mailbox->TryGetPresentFrame(timeout_ms);
     if (!frame) {
         LOG_DEBUG(Render_OpenGL, "TryGetPresentFrame returned no frame to present");
-        return;
+        return false;
     }
 
     // Clearing before a full overwrite of a fbo can signal to drivers that they can avoid a
@@ -725,6 +709,7 @@ void RendererOpenGL::TryPresent(int timeout_ms) {
     glFlush();
 
     glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
+    return true;
 }
 
 void RendererOpenGL::RenderScreenshot() {

src/video_core/renderer_opengl/renderer_opengl.h

@@ -55,13 +55,14 @@ class FrameMailbox;
 
 class RendererOpenGL final : public VideoCore::RendererBase {
 public:
-    explicit RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system);
+    explicit RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system,
+                            Core::Frontend::GraphicsContext& context);
     ~RendererOpenGL() override;
 
     bool Init() override;
     void ShutDown() override;
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-    void TryPresent(int timeout_ms) override;
+    bool TryPresent(int timeout_ms) override;
 
 private:
     /// Initializes the OpenGL state and creates persistent objects.
@@ -89,8 +90,11 @@ private:
 
     void PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer);
 
+    bool Present(int timeout_ms);
+
     Core::Frontend::EmuWindow& emu_window;
     Core::System& system;
+    Core::Frontend::GraphicsContext& context;
 
     StateTracker state_tracker{system};
 
@@ -115,6 +119,8 @@ private:
 
     /// Frame presentation mailbox
     std::unique_ptr<FrameMailbox> frame_mailbox;
+
+    bool has_debug_tool = false;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/utils.cpp

@@ -14,68 +14,6 @@
 
 namespace OpenGL {
 
-struct VertexArrayPushBuffer::Entry {
-    GLuint binding_index{};
-    const GLuint* buffer{};
-    GLintptr offset{};
-    GLsizei stride{};
-};
-
-VertexArrayPushBuffer::VertexArrayPushBuffer(StateTracker& state_tracker)
-    : state_tracker{state_tracker} {}
-
-VertexArrayPushBuffer::~VertexArrayPushBuffer() = default;
-
-void VertexArrayPushBuffer::Setup() {
-    index_buffer = nullptr;
-    vertex_buffers.clear();
-}
-
-void VertexArrayPushBuffer::SetIndexBuffer(const GLuint* buffer) {
-    index_buffer = buffer;
-}
-
-void VertexArrayPushBuffer::SetVertexBuffer(GLuint binding_index, const GLuint* buffer,
-                                            GLintptr offset, GLsizei stride) {
-    vertex_buffers.push_back(Entry{binding_index, buffer, offset, stride});
-}
-
-void VertexArrayPushBuffer::Bind() {
-    if (index_buffer) {
-        state_tracker.BindIndexBuffer(*index_buffer);
-    }
-
-    for (const auto& entry : vertex_buffers) {
-        glBindVertexBuffer(entry.binding_index, *entry.buffer, entry.offset, entry.stride);
-    }
-}
-
-struct BindBuffersRangePushBuffer::Entry {
-    GLuint binding;
-    const GLuint* buffer;
-    GLintptr offset;
-    GLsizeiptr size;
-};
-
-BindBuffersRangePushBuffer::BindBuffersRangePushBuffer(GLenum target) : target{target} {}
-
-BindBuffersRangePushBuffer::~BindBuffersRangePushBuffer() = default;
-
-void BindBuffersRangePushBuffer::Setup() {
-    entries.clear();
-}
-
-void BindBuffersRangePushBuffer::Push(GLuint binding, const GLuint* buffer, GLintptr offset,
-                                      GLsizeiptr size) {
-    entries.push_back(Entry{binding, buffer, offset, size});
-}
-
-void BindBuffersRangePushBuffer::Bind() {
-    for (const Entry& entry : entries) {
-        glBindBufferRange(target, entry.binding, *entry.buffer, entry.offset, entry.size);
-    }
-}
-
 void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr, std::string_view extra_info) {
     if (!GLAD_GL_KHR_debug) {
         // We don't need to throw an error as this is just for debugging

src/video_core/renderer_opengl/utils.h

@@ -11,49 +11,6 @@
 
 namespace OpenGL {
 
-class StateTracker;
-
-class VertexArrayPushBuffer final {
-public:
-    explicit VertexArrayPushBuffer(StateTracker& state_tracker);
-    ~VertexArrayPushBuffer();
-
-    void Setup();
-
-    void SetIndexBuffer(const GLuint* buffer);
-
-    void SetVertexBuffer(GLuint binding_index, const GLuint* buffer, GLintptr offset,
-                         GLsizei stride);
-
-    void Bind();
-
-private:
-    struct Entry;
-
-    StateTracker& state_tracker;
-
-    const GLuint* index_buffer{};
-    std::vector<Entry> vertex_buffers;
-};
-
-class BindBuffersRangePushBuffer final {
-public:
-    explicit BindBuffersRangePushBuffer(GLenum target);
-    ~BindBuffersRangePushBuffer();
-
-    void Setup();
-
-    void Push(GLuint binding, const GLuint* buffer, GLintptr offset, GLsizeiptr size);
-
-    void Bind();
-
-private:
-    struct Entry;
-
-    GLenum target;
-    std::vector<Entry> entries;
-};
-
 void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr, std::string_view extra_info = {});
 
 } // namespace OpenGL

src/video_core/renderer_vulkan/declarations.h

@@ -1,58 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-namespace vk {
-class DispatchLoaderDynamic;
-}
-
-namespace Vulkan {
-constexpr vk::DispatchLoaderDynamic* dont_use_me_dld = nullptr;
-}
-
-#define VULKAN_HPP_DEFAULT_DISPATCHER (*::Vulkan::dont_use_me_dld)
-#define VULKAN_HPP_ENABLE_DYNAMIC_LOADER_TOOL 0
-#define VULKAN_HPP_DISPATCH_LOADER_DYNAMIC 1
-#include <vulkan/vulkan.hpp>
-
-namespace Vulkan {
-
-// vulkan.hpp unique handlers use DispatchLoaderStatic
-template <typename T>
-using UniqueHandle = vk::UniqueHandle<T, vk::DispatchLoaderDynamic>;
-
-using UniqueAccelerationStructureNV = UniqueHandle<vk::AccelerationStructureNV>;
-using UniqueBuffer = UniqueHandle<vk::Buffer>;
-using UniqueBufferView = UniqueHandle<vk::BufferView>;
-using UniqueCommandBuffer = UniqueHandle<vk::CommandBuffer>;
-using UniqueCommandPool = UniqueHandle<vk::CommandPool>;
-using UniqueDescriptorPool = UniqueHandle<vk::DescriptorPool>;
-using UniqueDescriptorSet = UniqueHandle<vk::DescriptorSet>;
-using UniqueDescriptorSetLayout = UniqueHandle<vk::DescriptorSetLayout>;
-using UniqueDescriptorUpdateTemplate = UniqueHandle<vk::DescriptorUpdateTemplate>;
-using UniqueDevice = UniqueHandle<vk::Device>;
-using UniqueDeviceMemory = UniqueHandle<vk::DeviceMemory>;
-using UniqueEvent = UniqueHandle<vk::Event>;
-using UniqueFence = UniqueHandle<vk::Fence>;
-using UniqueFramebuffer = UniqueHandle<vk::Framebuffer>;
-using UniqueImage = UniqueHandle<vk::Image>;
-using UniqueImageView = UniqueHandle<vk::ImageView>;
-using UniqueIndirectCommandsLayoutNVX = UniqueHandle<vk::IndirectCommandsLayoutNVX>;
-using UniqueObjectTableNVX = UniqueHandle<vk::ObjectTableNVX>;
-using UniquePipeline = UniqueHandle<vk::Pipeline>;
-using UniquePipelineCache = UniqueHandle<vk::PipelineCache>;
-using UniquePipelineLayout = UniqueHandle<vk::PipelineLayout>;
-using UniqueQueryPool = UniqueHandle<vk::QueryPool>;
-using UniqueRenderPass = UniqueHandle<vk::RenderPass>;
-using UniqueSampler = UniqueHandle<vk::Sampler>;
-using UniqueSamplerYcbcrConversion = UniqueHandle<vk::SamplerYcbcrConversion>;
-using UniqueSemaphore = UniqueHandle<vk::Semaphore>;
-using UniqueShaderModule = UniqueHandle<vk::ShaderModule>;
-using UniqueSwapchainKHR = UniqueHandle<vk::SwapchainKHR>;
-using UniqueValidationCacheEXT = UniqueHandle<vk::ValidationCacheEXT>;
-using UniqueDebugReportCallbackEXT = UniqueHandle<vk::DebugReportCallbackEXT>;
-using UniqueDebugUtilsMessengerEXT = UniqueHandle<vk::DebugUtilsMessengerEXT>;
-
-} // namespace Vulkan

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -2,13 +2,15 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <iterator>
+
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "video_core/engines/maxwell_3d.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/maxwell_to_vk.h"
 #include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 #include "video_core/surface.h"
 
 namespace Vulkan::MaxwellToVK {
@@ -17,88 +19,89 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
 namespace Sampler {
 
-vk::Filter Filter(Tegra::Texture::TextureFilter filter) {
+VkFilter Filter(Tegra::Texture::TextureFilter filter) {
     switch (filter) {
     case Tegra::Texture::TextureFilter::Linear:
-        return vk::Filter::eLinear;
+        return VK_FILTER_LINEAR;
     case Tegra::Texture::TextureFilter::Nearest:
-        return vk::Filter::eNearest;
+        return VK_FILTER_NEAREST;
     }
     UNIMPLEMENTED_MSG("Unimplemented sampler filter={}", static_cast<u32>(filter));
     return {};
 }
 
-vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter) {
+VkSamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter) {
     switch (mipmap_filter) {
     case Tegra::Texture::TextureMipmapFilter::None:
         // TODO(Rodrigo): None seems to be mapped to OpenGL's mag and min filters without mipmapping
         // (e.g. GL_NEAREST and GL_LINEAR). Vulkan doesn't have such a thing, find out if we have to
         // use an image view with a single mipmap level to emulate this.
-        return vk::SamplerMipmapMode::eLinear;
+        return VK_SAMPLER_MIPMAP_MODE_LINEAR;
+        ;
     case Tegra::Texture::TextureMipmapFilter::Linear:
-        return vk::SamplerMipmapMode::eLinear;
+        return VK_SAMPLER_MIPMAP_MODE_LINEAR;
     case Tegra::Texture::TextureMipmapFilter::Nearest:
-        return vk::SamplerMipmapMode::eNearest;
+        return VK_SAMPLER_MIPMAP_MODE_NEAREST;
     }
     UNIMPLEMENTED_MSG("Unimplemented sampler mipmap mode={}", static_cast<u32>(mipmap_filter));
     return {};
 }
 
-vk::SamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode,
-                                Tegra::Texture::TextureFilter filter) {
+VkSamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode,
+                              Tegra::Texture::TextureFilter filter) {
     switch (wrap_mode) {
     case Tegra::Texture::WrapMode::Wrap:
-        return vk::SamplerAddressMode::eRepeat;
+        return VK_SAMPLER_ADDRESS_MODE_REPEAT;
     case Tegra::Texture::WrapMode::Mirror:
-        return vk::SamplerAddressMode::eMirroredRepeat;
+        return VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT;
     case Tegra::Texture::WrapMode::ClampToEdge:
-        return vk::SamplerAddressMode::eClampToEdge;
+        return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE;
     case Tegra::Texture::WrapMode::Border:
-        return vk::SamplerAddressMode::eClampToBorder;
+        return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER;
     case Tegra::Texture::WrapMode::Clamp:
-        if (device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) {
+        if (device.GetDriverID() == VK_DRIVER_ID_NVIDIA_PROPRIETARY_KHR) {
             // Nvidia's Vulkan driver defaults to GL_CLAMP on invalid enumerations, we can hack this
             // by sending an invalid enumeration.
-            return static_cast<vk::SamplerAddressMode>(0xcafe);
+            return static_cast<VkSamplerAddressMode>(0xcafe);
         }
         // TODO(Rodrigo): Emulate GL_CLAMP properly on other vendors
         switch (filter) {
         case Tegra::Texture::TextureFilter::Nearest:
-            return vk::SamplerAddressMode::eClampToEdge;
+            return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE;
         case Tegra::Texture::TextureFilter::Linear:
-            return vk::SamplerAddressMode::eClampToBorder;
+            return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER;
         }
         UNREACHABLE();
-        return vk::SamplerAddressMode::eClampToEdge;
+        return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE;
     case Tegra::Texture::WrapMode::MirrorOnceClampToEdge:
-        return vk::SamplerAddressMode::eMirrorClampToEdge;
+        return VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE;
     case Tegra::Texture::WrapMode::MirrorOnceBorder:
         UNIMPLEMENTED();
-        return vk::SamplerAddressMode::eMirrorClampToEdge;
+        return VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE;
     default:
         UNIMPLEMENTED_MSG("Unimplemented wrap mode={}", static_cast<u32>(wrap_mode));
         return {};
     }
 }
 
-vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) {
+VkCompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) {
     switch (depth_compare_func) {
     case Tegra::Texture::DepthCompareFunc::Never:
-        return vk::CompareOp::eNever;
+        return VK_COMPARE_OP_NEVER;
     case Tegra::Texture::DepthCompareFunc::Less:
-        return vk::CompareOp::eLess;
+        return VK_COMPARE_OP_LESS;
     case Tegra::Texture::DepthCompareFunc::LessEqual:
-        return vk::CompareOp::eLessOrEqual;
+        return VK_COMPARE_OP_LESS_OR_EQUAL;
     case Tegra::Texture::DepthCompareFunc::Equal:
-        return vk::CompareOp::eEqual;
+        return VK_COMPARE_OP_EQUAL;
     case Tegra::Texture::DepthCompareFunc::NotEqual:
-        return vk::CompareOp::eNotEqual;
+        return VK_COMPARE_OP_NOT_EQUAL;
     case Tegra::Texture::DepthCompareFunc::Greater:
-        return vk::CompareOp::eGreater;
+        return VK_COMPARE_OP_GREATER;
     case Tegra::Texture::DepthCompareFunc::GreaterEqual:
-        return vk::CompareOp::eGreaterOrEqual;
+        return VK_COMPARE_OP_GREATER_OR_EQUAL;
     case Tegra::Texture::DepthCompareFunc::Always:
-        return vk::CompareOp::eAlways;
+        return VK_COMPARE_OP_ALWAYS;
     }
     UNIMPLEMENTED_MSG("Unimplemented sampler depth compare function={}",
                       static_cast<u32>(depth_compare_func));
@@ -112,92 +115,92 @@ namespace {
 enum : u32 { Attachable = 1, Storage = 2 };
 
 struct FormatTuple {
-    vk::Format format; ///< Vulkan format
-    int usage;         ///< Describes image format usage
+    VkFormat format; ///< Vulkan format
+    int usage = 0;   ///< Describes image format usage
 } constexpr tex_format_tuples[] = {
-    {vk::Format::eA8B8G8R8UnormPack32, Attachable | Storage},    // ABGR8U
-    {vk::Format::eA8B8G8R8SnormPack32, Attachable | Storage},    // ABGR8S
-    {vk::Format::eA8B8G8R8UintPack32, Attachable | Storage},     // ABGR8UI
-    {vk::Format::eB5G6R5UnormPack16, {}},                        // B5G6R5U
-    {vk::Format::eA2B10G10R10UnormPack32, Attachable | Storage}, // A2B10G10R10U
-    {vk::Format::eA1R5G5B5UnormPack16, Attachable},              // A1B5G5R5U (flipped with swizzle)
-    {vk::Format::eR8Unorm, Attachable | Storage},                // R8U
-    {vk::Format::eR8Uint, Attachable | Storage},                 // R8UI
-    {vk::Format::eR16G16B16A16Sfloat, Attachable | Storage},     // RGBA16F
-    {vk::Format::eR16G16B16A16Unorm, Attachable | Storage},      // RGBA16U
-    {vk::Format::eR16G16B16A16Snorm, Attachable | Storage},      // RGBA16S
-    {vk::Format::eR16G16B16A16Uint, Attachable | Storage},       // RGBA16UI
-    {vk::Format::eB10G11R11UfloatPack32, Attachable | Storage},  // R11FG11FB10F
-    {vk::Format::eR32G32B32A32Uint, Attachable | Storage},       // RGBA32UI
-    {vk::Format::eBc1RgbaUnormBlock, {}},                        // DXT1
-    {vk::Format::eBc2UnormBlock, {}},                            // DXT23
-    {vk::Format::eBc3UnormBlock, {}},                            // DXT45
-    {vk::Format::eBc4UnormBlock, {}},                            // DXN1
-    {vk::Format::eBc5UnormBlock, {}},                            // DXN2UNORM
-    {vk::Format::eBc5SnormBlock, {}},                            // DXN2SNORM
-    {vk::Format::eBc7UnormBlock, {}},                            // BC7U
-    {vk::Format::eBc6HUfloatBlock, {}},                          // BC6H_UF16
-    {vk::Format::eBc6HSfloatBlock, {}},                          // BC6H_SF16
-    {vk::Format::eAstc4x4UnormBlock, {}},                        // ASTC_2D_4X4
-    {vk::Format::eB8G8R8A8Unorm, {}},                            // BGRA8
-    {vk::Format::eR32G32B32A32Sfloat, Attachable | Storage},     // RGBA32F
-    {vk::Format::eR32G32Sfloat, Attachable | Storage},           // RG32F
-    {vk::Format::eR32Sfloat, Attachable | Storage},              // R32F
-    {vk::Format::eR16Sfloat, Attachable | Storage},              // R16F
-    {vk::Format::eR16Unorm, Attachable | Storage},               // R16U
-    {vk::Format::eUndefined, {}},                                // R16S
-    {vk::Format::eUndefined, {}},                                // R16UI
-    {vk::Format::eUndefined, {}},                                // R16I
-    {vk::Format::eR16G16Unorm, Attachable | Storage},            // RG16
-    {vk::Format::eR16G16Sfloat, Attachable | Storage},           // RG16F
-    {vk::Format::eUndefined, {}},                                // RG16UI
-    {vk::Format::eUndefined, {}},                                // RG16I
-    {vk::Format::eR16G16Snorm, Attachable | Storage},            // RG16S
-    {vk::Format::eUndefined, {}},                                // RGB32F
-    {vk::Format::eR8G8B8A8Srgb, Attachable},                     // RGBA8_SRGB
-    {vk::Format::eR8G8Unorm, Attachable | Storage},              // RG8U
-    {vk::Format::eR8G8Snorm, Attachable | Storage},              // RG8S
-    {vk::Format::eR32G32Uint, Attachable | Storage},             // RG32UI
-    {vk::Format::eUndefined, {}},                                // RGBX16F
-    {vk::Format::eR32Uint, Attachable | Storage},                // R32UI
-    {vk::Format::eR32Sint, Attachable | Storage},                // R32I
-    {vk::Format::eAstc8x8UnormBlock, {}},                        // ASTC_2D_8X8
-    {vk::Format::eUndefined, {}},                                // ASTC_2D_8X5
-    {vk::Format::eUndefined, {}},                                // ASTC_2D_5X4
-    {vk::Format::eUndefined, {}},                                // BGRA8_SRGB
-    {vk::Format::eBc1RgbaSrgbBlock, {}},                         // DXT1_SRGB
-    {vk::Format::eBc2SrgbBlock, {}},                             // DXT23_SRGB
-    {vk::Format::eBc3SrgbBlock, {}},                             // DXT45_SRGB
-    {vk::Format::eBc7SrgbBlock, {}},                             // BC7U_SRGB
-    {vk::Format::eR4G4B4A4UnormPack16, Attachable},              // R4G4B4A4U
-    {vk::Format::eAstc4x4SrgbBlock, {}},                         // ASTC_2D_4X4_SRGB
-    {vk::Format::eAstc8x8SrgbBlock, {}},                         // ASTC_2D_8X8_SRGB
-    {vk::Format::eAstc8x5SrgbBlock, {}},                         // ASTC_2D_8X5_SRGB
-    {vk::Format::eAstc5x4SrgbBlock, {}},                         // ASTC_2D_5X4_SRGB
-    {vk::Format::eAstc5x5UnormBlock, {}},                        // ASTC_2D_5X5
-    {vk::Format::eAstc5x5SrgbBlock, {}},                         // ASTC_2D_5X5_SRGB
-    {vk::Format::eAstc10x8UnormBlock, {}},                       // ASTC_2D_10X8
-    {vk::Format::eAstc10x8SrgbBlock, {}},                        // ASTC_2D_10X8_SRGB
-    {vk::Format::eAstc6x6UnormBlock, {}},                        // ASTC_2D_6X6
-    {vk::Format::eAstc6x6SrgbBlock, {}},                         // ASTC_2D_6X6_SRGB
-    {vk::Format::eAstc10x10UnormBlock, {}},                      // ASTC_2D_10X10
-    {vk::Format::eAstc10x10SrgbBlock, {}},                       // ASTC_2D_10X10_SRGB
-    {vk::Format::eAstc12x12UnormBlock, {}},                      // ASTC_2D_12X12
-    {vk::Format::eAstc12x12SrgbBlock, {}},                       // ASTC_2D_12X12_SRGB
-    {vk::Format::eAstc8x6UnormBlock, {}},                        // ASTC_2D_8X6
-    {vk::Format::eAstc8x6SrgbBlock, {}},                         // ASTC_2D_8X6_SRGB
-    {vk::Format::eAstc6x5UnormBlock, {}},                        // ASTC_2D_6X5
-    {vk::Format::eAstc6x5SrgbBlock, {}},                         // ASTC_2D_6X5_SRGB
-    {vk::Format::eE5B9G9R9UfloatPack32, {}},                     // E5B9G9R9F
+    {VK_FORMAT_A8B8G8R8_UNORM_PACK32, Attachable | Storage},    // ABGR8U
+    {VK_FORMAT_A8B8G8R8_SNORM_PACK32, Attachable | Storage},    // ABGR8S
+    {VK_FORMAT_A8B8G8R8_UINT_PACK32, Attachable | Storage},     // ABGR8UI
+    {VK_FORMAT_B5G6R5_UNORM_PACK16},                            // B5G6R5U
+    {VK_FORMAT_A2B10G10R10_UNORM_PACK32, Attachable | Storage}, // A2B10G10R10U
+    {VK_FORMAT_A1R5G5B5_UNORM_PACK16, Attachable},              // A1B5G5R5U (flipped with swizzle)
+    {VK_FORMAT_R8_UNORM, Attachable | Storage},                 // R8U
+    {VK_FORMAT_R8_UINT, Attachable | Storage},                  // R8UI
+    {VK_FORMAT_R16G16B16A16_SFLOAT, Attachable | Storage},      // RGBA16F
+    {VK_FORMAT_R16G16B16A16_UNORM, Attachable | Storage},       // RGBA16U
+    {VK_FORMAT_R16G16B16A16_SNORM, Attachable | Storage},       // RGBA16S
+    {VK_FORMAT_R16G16B16A16_UINT, Attachable | Storage},        // RGBA16UI
+    {VK_FORMAT_B10G11R11_UFLOAT_PACK32, Attachable | Storage},  // R11FG11FB10F
+    {VK_FORMAT_R32G32B32A32_UINT, Attachable | Storage},        // RGBA32UI
+    {VK_FORMAT_BC1_RGBA_UNORM_BLOCK},                           // DXT1
+    {VK_FORMAT_BC2_UNORM_BLOCK},                                // DXT23
+    {VK_FORMAT_BC3_UNORM_BLOCK},                                // DXT45
+    {VK_FORMAT_BC4_UNORM_BLOCK},                                // DXN1
+    {VK_FORMAT_BC5_UNORM_BLOCK},                                // DXN2UNORM
+    {VK_FORMAT_BC5_SNORM_BLOCK},                                // DXN2SNORM
+    {VK_FORMAT_BC7_UNORM_BLOCK},                                // BC7U
+    {VK_FORMAT_BC6H_UFLOAT_BLOCK},                              // BC6H_UF16
+    {VK_FORMAT_BC6H_SFLOAT_BLOCK},                              // BC6H_SF16
+    {VK_FORMAT_ASTC_4x4_UNORM_BLOCK},                           // ASTC_2D_4X4
+    {VK_FORMAT_B8G8R8A8_UNORM},                                 // BGRA8
+    {VK_FORMAT_R32G32B32A32_SFLOAT, Attachable | Storage},      // RGBA32F
+    {VK_FORMAT_R32G32_SFLOAT, Attachable | Storage},            // RG32F
+    {VK_FORMAT_R32_SFLOAT, Attachable | Storage},               // R32F
+    {VK_FORMAT_R16_SFLOAT, Attachable | Storage},               // R16F
+    {VK_FORMAT_R16_UNORM, Attachable | Storage},                // R16U
+    {VK_FORMAT_UNDEFINED},                                      // R16S
+    {VK_FORMAT_UNDEFINED},                                      // R16UI
+    {VK_FORMAT_UNDEFINED},                                      // R16I
+    {VK_FORMAT_R16G16_UNORM, Attachable | Storage},             // RG16
+    {VK_FORMAT_R16G16_SFLOAT, Attachable | Storage},            // RG16F
+    {VK_FORMAT_UNDEFINED},                                      // RG16UI
+    {VK_FORMAT_UNDEFINED},                                      // RG16I
+    {VK_FORMAT_R16G16_SNORM, Attachable | Storage},             // RG16S
+    {VK_FORMAT_UNDEFINED},                                      // RGB32F
+    {VK_FORMAT_R8G8B8A8_SRGB, Attachable},                      // RGBA8_SRGB
+    {VK_FORMAT_R8G8_UNORM, Attachable | Storage},               // RG8U
+    {VK_FORMAT_R8G8_SNORM, Attachable | Storage},               // RG8S
+    {VK_FORMAT_R32G32_UINT, Attachable | Storage},              // RG32UI
+    {VK_FORMAT_UNDEFINED},                                      // RGBX16F
+    {VK_FORMAT_R32_UINT, Attachable | Storage},                 // R32UI
+    {VK_FORMAT_R32_SINT, Attachable | Storage},                 // R32I
+    {VK_FORMAT_ASTC_8x8_UNORM_BLOCK},                           // ASTC_2D_8X8
+    {VK_FORMAT_UNDEFINED},                                      // ASTC_2D_8X5
+    {VK_FORMAT_UNDEFINED},                                      // ASTC_2D_5X4
+    {VK_FORMAT_UNDEFINED},                                      // BGRA8_SRGB
+    {VK_FORMAT_BC1_RGBA_SRGB_BLOCK},                            // DXT1_SRGB
+    {VK_FORMAT_BC2_SRGB_BLOCK},                                 // DXT23_SRGB
+    {VK_FORMAT_BC3_SRGB_BLOCK},                                 // DXT45_SRGB
+    {VK_FORMAT_BC7_SRGB_BLOCK},                                 // BC7U_SRGB
+    {VK_FORMAT_R4G4B4A4_UNORM_PACK16, Attachable},              // R4G4B4A4U
+    {VK_FORMAT_ASTC_4x4_SRGB_BLOCK},                            // ASTC_2D_4X4_SRGB
+    {VK_FORMAT_ASTC_8x8_SRGB_BLOCK},                            // ASTC_2D_8X8_SRGB
+    {VK_FORMAT_ASTC_8x5_SRGB_BLOCK},                            // ASTC_2D_8X5_SRGB
+    {VK_FORMAT_ASTC_5x4_SRGB_BLOCK},                            // ASTC_2D_5X4_SRGB
+    {VK_FORMAT_ASTC_5x5_UNORM_BLOCK},                           // ASTC_2D_5X5
+    {VK_FORMAT_ASTC_5x5_SRGB_BLOCK},                            // ASTC_2D_5X5_SRGB
+    {VK_FORMAT_ASTC_10x8_UNORM_BLOCK},                          // ASTC_2D_10X8
+    {VK_FORMAT_ASTC_10x8_SRGB_BLOCK},                           // ASTC_2D_10X8_SRGB
+    {VK_FORMAT_ASTC_6x6_UNORM_BLOCK},                           // ASTC_2D_6X6
+    {VK_FORMAT_ASTC_6x6_SRGB_BLOCK},                            // ASTC_2D_6X6_SRGB
+    {VK_FORMAT_ASTC_10x10_UNORM_BLOCK},                         // ASTC_2D_10X10
+    {VK_FORMAT_ASTC_10x10_SRGB_BLOCK},                          // ASTC_2D_10X10_SRGB
+    {VK_FORMAT_ASTC_12x12_UNORM_BLOCK},                         // ASTC_2D_12X12
+    {VK_FORMAT_ASTC_12x12_SRGB_BLOCK},                          // ASTC_2D_12X12_SRGB
+    {VK_FORMAT_ASTC_8x6_UNORM_BLOCK},                           // ASTC_2D_8X6
+    {VK_FORMAT_ASTC_8x6_SRGB_BLOCK},                            // ASTC_2D_8X6_SRGB
+    {VK_FORMAT_ASTC_6x5_UNORM_BLOCK},                           // ASTC_2D_6X5
+    {VK_FORMAT_ASTC_6x5_SRGB_BLOCK},                            // ASTC_2D_6X5_SRGB
+    {VK_FORMAT_E5B9G9R9_UFLOAT_PACK32},                         // E5B9G9R9F
 
     // Depth formats
-    {vk::Format::eD32Sfloat, Attachable}, // Z32F
-    {vk::Format::eD16Unorm, Attachable},  // Z16
+    {VK_FORMAT_D32_SFLOAT, Attachable}, // Z32F
+    {VK_FORMAT_D16_UNORM, Attachable},  // Z16
 
     // DepthStencil formats
-    {vk::Format::eD24UnormS8Uint, Attachable},  // Z24S8
-    {vk::Format::eD24UnormS8Uint, Attachable},  // S8Z24 (emulated)
-    {vk::Format::eD32SfloatS8Uint, Attachable}, // Z32FS8
+    {VK_FORMAT_D24_UNORM_S8_UINT, Attachable},  // Z24S8
+    {VK_FORMAT_D24_UNORM_S8_UINT, Attachable},  // S8Z24 (emulated)
+    {VK_FORMAT_D32_SFLOAT_S8_UINT, Attachable}, // Z32FS8
 };
 static_assert(std::size(tex_format_tuples) == VideoCore::Surface::MaxPixelFormat);
 
@@ -212,106 +215,106 @@ FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFo
     ASSERT(static_cast<std::size_t>(pixel_format) < std::size(tex_format_tuples));
 
     auto tuple = tex_format_tuples[static_cast<std::size_t>(pixel_format)];
-    if (tuple.format == vk::Format::eUndefined) {
+    if (tuple.format == VK_FORMAT_UNDEFINED) {
         UNIMPLEMENTED_MSG("Unimplemented texture format with pixel format={}",
                           static_cast<u32>(pixel_format));
-        return {vk::Format::eA8B8G8R8UnormPack32, true, true};
+        return {VK_FORMAT_A8B8G8R8_UNORM_PACK32, true, true};
     }
 
     // Use ABGR8 on hardware that doesn't support ASTC natively
     if (!device.IsOptimalAstcSupported() && VideoCore::Surface::IsPixelFormatASTC(pixel_format)) {
         tuple.format = VideoCore::Surface::IsPixelFormatSRGB(pixel_format)
-                           ? vk::Format::eA8B8G8R8SrgbPack32
-                           : vk::Format::eA8B8G8R8UnormPack32;
+                           ? VK_FORMAT_A8B8G8R8_SRGB_PACK32
+                           : VK_FORMAT_A8B8G8R8_UNORM_PACK32;
     }
     const bool attachable = tuple.usage & Attachable;
     const bool storage = tuple.usage & Storage;
 
-    vk::FormatFeatureFlags usage;
+    VkFormatFeatureFlags usage;
     if (format_type == FormatType::Buffer) {
-        usage = vk::FormatFeatureFlagBits::eStorageTexelBuffer |
-                vk::FormatFeatureFlagBits::eUniformTexelBuffer;
+        usage =
+            VK_FORMAT_FEATURE_STORAGE_TEXEL_BUFFER_BIT | VK_FORMAT_FEATURE_UNIFORM_TEXEL_BUFFER_BIT;
     } else {
-        usage = vk::FormatFeatureFlagBits::eSampledImage | vk::FormatFeatureFlagBits::eTransferDst |
-                vk::FormatFeatureFlagBits::eTransferSrc;
+        usage = VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_TRANSFER_DST_BIT |
+                VK_FORMAT_FEATURE_TRANSFER_SRC_BIT;
         if (attachable) {
-            usage |= IsZetaFormat(pixel_format) ? vk::FormatFeatureFlagBits::eDepthStencilAttachment
-                                                : vk::FormatFeatureFlagBits::eColorAttachment;
+            usage |= IsZetaFormat(pixel_format) ? VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT
+                                                : VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT;
         }
         if (storage) {
-            usage |= vk::FormatFeatureFlagBits::eStorageImage;
+            usage |= VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT;
         }
     }
     return {device.GetSupportedFormat(tuple.format, usage, format_type), attachable, storage};
 }
 
-vk::ShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage) {
+VkShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage) {
     switch (stage) {
     case Tegra::Engines::ShaderType::Vertex:
-        return vk::ShaderStageFlagBits::eVertex;
+        return VK_SHADER_STAGE_VERTEX_BIT;
     case Tegra::Engines::ShaderType::TesselationControl:
-        return vk::ShaderStageFlagBits::eTessellationControl;
+        return VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT;
     case Tegra::Engines::ShaderType::TesselationEval:
-        return vk::ShaderStageFlagBits::eTessellationEvaluation;
+        return VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT;
     case Tegra::Engines::ShaderType::Geometry:
-        return vk::ShaderStageFlagBits::eGeometry;
+        return VK_SHADER_STAGE_GEOMETRY_BIT;
     case Tegra::Engines::ShaderType::Fragment:
-        return vk::ShaderStageFlagBits::eFragment;
+        return VK_SHADER_STAGE_FRAGMENT_BIT;
     case Tegra::Engines::ShaderType::Compute:
-        return vk::ShaderStageFlagBits::eCompute;
+        return VK_SHADER_STAGE_COMPUTE_BIT;
     }
     UNIMPLEMENTED_MSG("Unimplemented shader stage={}", static_cast<u32>(stage));
     return {};
 }
 
-vk::PrimitiveTopology PrimitiveTopology([[maybe_unused]] const VKDevice& device,
-                                        Maxwell::PrimitiveTopology topology) {
+VkPrimitiveTopology PrimitiveTopology([[maybe_unused]] const VKDevice& device,
+                                      Maxwell::PrimitiveTopology topology) {
     switch (topology) {
     case Maxwell::PrimitiveTopology::Points:
-        return vk::PrimitiveTopology::ePointList;
+        return VK_PRIMITIVE_TOPOLOGY_POINT_LIST;
     case Maxwell::PrimitiveTopology::Lines:
-        return vk::PrimitiveTopology::eLineList;
+        return VK_PRIMITIVE_TOPOLOGY_LINE_LIST;
     case Maxwell::PrimitiveTopology::LineStrip:
-        return vk::PrimitiveTopology::eLineStrip;
+        return VK_PRIMITIVE_TOPOLOGY_LINE_STRIP;
     case Maxwell::PrimitiveTopology::Triangles:
-        return vk::PrimitiveTopology::eTriangleList;
+        return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST;
     case Maxwell::PrimitiveTopology::TriangleStrip:
-        return vk::PrimitiveTopology::eTriangleStrip;
+        return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP;
     case Maxwell::PrimitiveTopology::TriangleFan:
-        return vk::PrimitiveTopology::eTriangleFan;
+        return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN;
     case Maxwell::PrimitiveTopology::Quads:
         // TODO(Rodrigo): Use VK_PRIMITIVE_TOPOLOGY_QUAD_LIST_EXT whenever it releases
-        return vk::PrimitiveTopology::eTriangleList;
+        return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST;
     case Maxwell::PrimitiveTopology::Patches:
-        return vk::PrimitiveTopology::ePatchList;
+        return VK_PRIMITIVE_TOPOLOGY_PATCH_LIST;
     default:
         UNIMPLEMENTED_MSG("Unimplemented topology={}", static_cast<u32>(topology));
         return {};
     }
 }
 
-vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) {
+VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) {
     switch (type) {
     case Maxwell::VertexAttribute::Type::SignedNorm:
         switch (size) {
         case Maxwell::VertexAttribute::Size::Size_8:
-            return vk::Format::eR8Snorm;
+            return VK_FORMAT_R8_SNORM;
         case Maxwell::VertexAttribute::Size::Size_8_8:
-            return vk::Format::eR8G8Snorm;
+            return VK_FORMAT_R8G8_SNORM;
         case Maxwell::VertexAttribute::Size::Size_8_8_8:
-            return vk::Format::eR8G8B8Snorm;
+            return VK_FORMAT_R8G8B8_SNORM;
         case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
-            return vk::Format::eR8G8B8A8Snorm;
+            return VK_FORMAT_R8G8B8A8_SNORM;
         case Maxwell::VertexAttribute::Size::Size_16:
-            return vk::Format::eR16Snorm;
+            return VK_FORMAT_R16_SNORM;
         case Maxwell::VertexAttribute::Size::Size_16_16:
-            return vk::Format::eR16G16Snorm;
+            return VK_FORMAT_R16G16_SNORM;
         case Maxwell::VertexAttribute::Size::Size_16_16_16:
-            return vk::Format::eR16G16B16Snorm;
+            return VK_FORMAT_R16G16B16_SNORM;
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
-            return vk::Format::eR16G16B16A16Snorm;
+            return VK_FORMAT_R16G16B16A16_SNORM;
         case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
-            return vk::Format::eA2B10G10R10SnormPack32;
+            return VK_FORMAT_A2B10G10R10_SNORM_PACK32;
         default:
             break;
         }
@@ -319,23 +322,23 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
     case Maxwell::VertexAttribute::Type::UnsignedNorm:
         switch (size) {
         case Maxwell::VertexAttribute::Size::Size_8:
-            return vk::Format::eR8Unorm;
+            return VK_FORMAT_R8_UNORM;
         case Maxwell::VertexAttribute::Size::Size_8_8:
-            return vk::Format::eR8G8Unorm;
+            return VK_FORMAT_R8G8_UNORM;
         case Maxwell::VertexAttribute::Size::Size_8_8_8:
-            return vk::Format::eR8G8B8Unorm;
+            return VK_FORMAT_R8G8B8_UNORM;
         case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
-            return vk::Format::eR8G8B8A8Unorm;
+            return VK_FORMAT_R8G8B8A8_UNORM;
         case Maxwell::VertexAttribute::Size::Size_16:
-            return vk::Format::eR16Unorm;
+            return VK_FORMAT_R16_UNORM;
         case Maxwell::VertexAttribute::Size::Size_16_16:
-            return vk::Format::eR16G16Unorm;
+            return VK_FORMAT_R16G16_UNORM;
         case Maxwell::VertexAttribute::Size::Size_16_16_16:
-            return vk::Format::eR16G16B16Unorm;
+            return VK_FORMAT_R16G16B16_UNORM;
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
-            return vk::Format::eR16G16B16A16Unorm;
+            return VK_FORMAT_R16G16B16A16_UNORM;
         case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
-            return vk::Format::eA2B10G10R10UnormPack32;
+            return VK_FORMAT_A2B10G10R10_UNORM_PACK32;
         default:
             break;
         }
@@ -343,59 +346,69 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
     case Maxwell::VertexAttribute::Type::SignedInt:
         switch (size) {
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
-            return vk::Format::eR16G16B16A16Sint;
+            return VK_FORMAT_R16G16B16A16_SINT;
         case Maxwell::VertexAttribute::Size::Size_8:
-            return vk::Format::eR8Sint;
+            return VK_FORMAT_R8_SINT;
         case Maxwell::VertexAttribute::Size::Size_8_8:
-            return vk::Format::eR8G8Sint;
+            return VK_FORMAT_R8G8_SINT;
         case Maxwell::VertexAttribute::Size::Size_8_8_8:
-            return vk::Format::eR8G8B8Sint;
+            return VK_FORMAT_R8G8B8_SINT;
         case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
-            return vk::Format::eR8G8B8A8Sint;
+            return VK_FORMAT_R8G8B8A8_SINT;
         case Maxwell::VertexAttribute::Size::Size_32:
-            return vk::Format::eR32Sint;
+            return VK_FORMAT_R32_SINT;
         default:
             break;
         }
+        break;
     case Maxwell::VertexAttribute::Type::UnsignedInt:
         switch (size) {
         case Maxwell::VertexAttribute::Size::Size_8:
-            return vk::Format::eR8Uint;
+            return VK_FORMAT_R8_UINT;
         case Maxwell::VertexAttribute::Size::Size_8_8:
-            return vk::Format::eR8G8Uint;
+            return VK_FORMAT_R8G8_UINT;
         case Maxwell::VertexAttribute::Size::Size_8_8_8:
-            return vk::Format::eR8G8B8Uint;
+            return VK_FORMAT_R8G8B8_UINT;
         case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
-            return vk::Format::eR8G8B8A8Uint;
+            return VK_FORMAT_R8G8B8A8_UINT;
+        case Maxwell::VertexAttribute::Size::Size_16:
+            return VK_FORMAT_R16_UINT;
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+            return VK_FORMAT_R16G16_UINT;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+            return VK_FORMAT_R16G16B16_UINT;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return VK_FORMAT_R16G16B16A16_UINT;
         case Maxwell::VertexAttribute::Size::Size_32:
-            return vk::Format::eR32Uint;
+            return VK_FORMAT_R32_UINT;
         case Maxwell::VertexAttribute::Size::Size_32_32:
-            return vk::Format::eR32G32Uint;
+            return VK_FORMAT_R32G32_UINT;
         case Maxwell::VertexAttribute::Size::Size_32_32_32:
-            return vk::Format::eR32G32B32Uint;
+            return VK_FORMAT_R32G32B32_UINT;
         case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
-            return vk::Format::eR32G32B32A32Uint;
+            return VK_FORMAT_R32G32B32A32_UINT;
         default:
             break;
         }
+        break;
     case Maxwell::VertexAttribute::Type::UnsignedScaled:
         switch (size) {
         case Maxwell::VertexAttribute::Size::Size_8:
-            return vk::Format::eR8Uscaled;
+            return VK_FORMAT_R8_USCALED;
         case Maxwell::VertexAttribute::Size::Size_8_8:
-            return vk::Format::eR8G8Uscaled;
+            return VK_FORMAT_R8G8_USCALED;
         case Maxwell::VertexAttribute::Size::Size_8_8_8:
-            return vk::Format::eR8G8B8Uscaled;
+            return VK_FORMAT_R8G8B8_USCALED;
         case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
-            return vk::Format::eR8G8B8A8Uscaled;
+            return VK_FORMAT_R8G8B8A8_USCALED;
         case Maxwell::VertexAttribute::Size::Size_16:
-            return vk::Format::eR16Uscaled;
+            return VK_FORMAT_R16_USCALED;
         case Maxwell::VertexAttribute::Size::Size_16_16:
-            return vk::Format::eR16G16Uscaled;
+            return VK_FORMAT_R16G16_USCALED;
         case Maxwell::VertexAttribute::Size::Size_16_16_16:
-            return vk::Format::eR16G16B16Uscaled;
+            return VK_FORMAT_R16G16B16_USCALED;
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
-            return vk::Format::eR16G16B16A16Uscaled;
+            return VK_FORMAT_R16G16B16A16_USCALED;
         default:
             break;
         }
@@ -403,21 +416,21 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
     case Maxwell::VertexAttribute::Type::SignedScaled:
         switch (size) {
         case Maxwell::VertexAttribute::Size::Size_8:
-            return vk::Format::eR8Sscaled;
+            return VK_FORMAT_R8_SSCALED;
         case Maxwell::VertexAttribute::Size::Size_8_8:
-            return vk::Format::eR8G8Sscaled;
+            return VK_FORMAT_R8G8_SSCALED;
         case Maxwell::VertexAttribute::Size::Size_8_8_8:
-            return vk::Format::eR8G8B8Sscaled;
+            return VK_FORMAT_R8G8B8_SSCALED;
         case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
-            return vk::Format::eR8G8B8A8Sscaled;
+            return VK_FORMAT_R8G8B8A8_SSCALED;
         case Maxwell::VertexAttribute::Size::Size_16:
-            return vk::Format::eR16Sscaled;
+            return VK_FORMAT_R16_SSCALED;
         case Maxwell::VertexAttribute::Size::Size_16_16:
-            return vk::Format::eR16G16Sscaled;
+            return VK_FORMAT_R16G16_SSCALED;
         case Maxwell::VertexAttribute::Size::Size_16_16_16:
-            return vk::Format::eR16G16B16Sscaled;
+            return VK_FORMAT_R16G16B16_SSCALED;
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
-            return vk::Format::eR16G16B16A16Sscaled;
+            return VK_FORMAT_R16G16B16A16_SSCALED;
         default:
             break;
         }
@@ -425,21 +438,21 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
     case Maxwell::VertexAttribute::Type::Float:
         switch (size) {
         case Maxwell::VertexAttribute::Size::Size_32:
-            return vk::Format::eR32Sfloat;
+            return VK_FORMAT_R32_SFLOAT;
         case Maxwell::VertexAttribute::Size::Size_32_32:
-            return vk::Format::eR32G32Sfloat;
+            return VK_FORMAT_R32G32_SFLOAT;
         case Maxwell::VertexAttribute::Size::Size_32_32_32:
-            return vk::Format::eR32G32B32Sfloat;
+            return VK_FORMAT_R32G32B32_SFLOAT;
         case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
-            return vk::Format::eR32G32B32A32Sfloat;
+            return VK_FORMAT_R32G32B32A32_SFLOAT;
         case Maxwell::VertexAttribute::Size::Size_16:
-            return vk::Format::eR16Sfloat;
+            return VK_FORMAT_R16_SFLOAT;
         case Maxwell::VertexAttribute::Size::Size_16_16:
-            return vk::Format::eR16G16Sfloat;
+            return VK_FORMAT_R16G16_SFLOAT;
         case Maxwell::VertexAttribute::Size::Size_16_16_16:
-            return vk::Format::eR16G16B16Sfloat;
+            return VK_FORMAT_R16G16B16_SFLOAT;
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
-            return vk::Format::eR16G16B16A16Sfloat;
+            return VK_FORMAT_R16G16B16A16_SFLOAT;
         default:
             break;
         }
@@ -450,210 +463,210 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
     return {};
 }
 
-vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison) {
+VkCompareOp ComparisonOp(Maxwell::ComparisonOp comparison) {
     switch (comparison) {
     case Maxwell::ComparisonOp::Never:
     case Maxwell::ComparisonOp::NeverOld:
-        return vk::CompareOp::eNever;
+        return VK_COMPARE_OP_NEVER;
     case Maxwell::ComparisonOp::Less:
     case Maxwell::ComparisonOp::LessOld:
-        return vk::CompareOp::eLess;
+        return VK_COMPARE_OP_LESS;
     case Maxwell::ComparisonOp::Equal:
     case Maxwell::ComparisonOp::EqualOld:
-        return vk::CompareOp::eEqual;
+        return VK_COMPARE_OP_EQUAL;
     case Maxwell::ComparisonOp::LessEqual:
     case Maxwell::ComparisonOp::LessEqualOld:
-        return vk::CompareOp::eLessOrEqual;
+        return VK_COMPARE_OP_LESS_OR_EQUAL;
     case Maxwell::ComparisonOp::Greater:
     case Maxwell::ComparisonOp::GreaterOld:
-        return vk::CompareOp::eGreater;
+        return VK_COMPARE_OP_GREATER;
     case Maxwell::ComparisonOp::NotEqual:
     case Maxwell::ComparisonOp::NotEqualOld:
-        return vk::CompareOp::eNotEqual;
+        return VK_COMPARE_OP_NOT_EQUAL;
     case Maxwell::ComparisonOp::GreaterEqual:
     case Maxwell::ComparisonOp::GreaterEqualOld:
-        return vk::CompareOp::eGreaterOrEqual;
+        return VK_COMPARE_OP_GREATER_OR_EQUAL;
     case Maxwell::ComparisonOp::Always:
     case Maxwell::ComparisonOp::AlwaysOld:
-        return vk::CompareOp::eAlways;
+        return VK_COMPARE_OP_ALWAYS;
     }
     UNIMPLEMENTED_MSG("Unimplemented comparison op={}", static_cast<u32>(comparison));
     return {};
 }
 
-vk::IndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format) {
+VkIndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format) {
     switch (index_format) {
     case Maxwell::IndexFormat::UnsignedByte:
         if (!device.IsExtIndexTypeUint8Supported()) {
             UNIMPLEMENTED_MSG("Native uint8 indices are not supported on this device");
-            return vk::IndexType::eUint16;
+            return VK_INDEX_TYPE_UINT16;
         }
-        return vk::IndexType::eUint8EXT;
+        return VK_INDEX_TYPE_UINT8_EXT;
     case Maxwell::IndexFormat::UnsignedShort:
-        return vk::IndexType::eUint16;
+        return VK_INDEX_TYPE_UINT16;
     case Maxwell::IndexFormat::UnsignedInt:
-        return vk::IndexType::eUint32;
+        return VK_INDEX_TYPE_UINT32;
     }
     UNIMPLEMENTED_MSG("Unimplemented index_format={}", static_cast<u32>(index_format));
     return {};
 }
 
-vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op) {
+VkStencilOp StencilOp(Maxwell::StencilOp stencil_op) {
     switch (stencil_op) {
     case Maxwell::StencilOp::Keep:
     case Maxwell::StencilOp::KeepOGL:
-        return vk::StencilOp::eKeep;
+        return VK_STENCIL_OP_KEEP;
     case Maxwell::StencilOp::Zero:
     case Maxwell::StencilOp::ZeroOGL:
-        return vk::StencilOp::eZero;
+        return VK_STENCIL_OP_ZERO;
     case Maxwell::StencilOp::Replace:
     case Maxwell::StencilOp::ReplaceOGL:
-        return vk::StencilOp::eReplace;
+        return VK_STENCIL_OP_REPLACE;
     case Maxwell::StencilOp::Incr:
     case Maxwell::StencilOp::IncrOGL:
-        return vk::StencilOp::eIncrementAndClamp;
+        return VK_STENCIL_OP_INCREMENT_AND_CLAMP;
     case Maxwell::StencilOp::Decr:
     case Maxwell::StencilOp::DecrOGL:
-        return vk::StencilOp::eDecrementAndClamp;
+        return VK_STENCIL_OP_DECREMENT_AND_CLAMP;
     case Maxwell::StencilOp::Invert:
     case Maxwell::StencilOp::InvertOGL:
-        return vk::StencilOp::eInvert;
+        return VK_STENCIL_OP_INVERT;
     case Maxwell::StencilOp::IncrWrap:
     case Maxwell::StencilOp::IncrWrapOGL:
-        return vk::StencilOp::eIncrementAndWrap;
+        return VK_STENCIL_OP_INCREMENT_AND_WRAP;
     case Maxwell::StencilOp::DecrWrap:
     case Maxwell::StencilOp::DecrWrapOGL:
-        return vk::StencilOp::eDecrementAndWrap;
+        return VK_STENCIL_OP_DECREMENT_AND_WRAP;
     }
     UNIMPLEMENTED_MSG("Unimplemented stencil op={}", static_cast<u32>(stencil_op));
     return {};
 }
 
-vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation) {
+VkBlendOp BlendEquation(Maxwell::Blend::Equation equation) {
     switch (equation) {
     case Maxwell::Blend::Equation::Add:
     case Maxwell::Blend::Equation::AddGL:
-        return vk::BlendOp::eAdd;
+        return VK_BLEND_OP_ADD;
     case Maxwell::Blend::Equation::Subtract:
     case Maxwell::Blend::Equation::SubtractGL:
-        return vk::BlendOp::eSubtract;
+        return VK_BLEND_OP_SUBTRACT;
     case Maxwell::Blend::Equation::ReverseSubtract:
     case Maxwell::Blend::Equation::ReverseSubtractGL:
-        return vk::BlendOp::eReverseSubtract;
+        return VK_BLEND_OP_REVERSE_SUBTRACT;
     case Maxwell::Blend::Equation::Min:
     case Maxwell::Blend::Equation::MinGL:
-        return vk::BlendOp::eMin;
+        return VK_BLEND_OP_MIN;
     case Maxwell::Blend::Equation::Max:
     case Maxwell::Blend::Equation::MaxGL:
-        return vk::BlendOp::eMax;
+        return VK_BLEND_OP_MAX;
     }
     UNIMPLEMENTED_MSG("Unimplemented blend equation={}", static_cast<u32>(equation));
     return {};
 }
 
-vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor) {
+VkBlendFactor BlendFactor(Maxwell::Blend::Factor factor) {
     switch (factor) {
     case Maxwell::Blend::Factor::Zero:
     case Maxwell::Blend::Factor::ZeroGL:
-        return vk::BlendFactor::eZero;
+        return VK_BLEND_FACTOR_ZERO;
     case Maxwell::Blend::Factor::One:
     case Maxwell::Blend::Factor::OneGL:
-        return vk::BlendFactor::eOne;
+        return VK_BLEND_FACTOR_ONE;
     case Maxwell::Blend::Factor::SourceColor:
     case Maxwell::Blend::Factor::SourceColorGL:
-        return vk::BlendFactor::eSrcColor;
+        return VK_BLEND_FACTOR_SRC_COLOR;
     case Maxwell::Blend::Factor::OneMinusSourceColor:
     case Maxwell::Blend::Factor::OneMinusSourceColorGL:
-        return vk::BlendFactor::eOneMinusSrcColor;
+        return VK_BLEND_FACTOR_ONE_MINUS_SRC_COLOR;
     case Maxwell::Blend::Factor::SourceAlpha:
     case Maxwell::Blend::Factor::SourceAlphaGL:
-        return vk::BlendFactor::eSrcAlpha;
+        return VK_BLEND_FACTOR_SRC_ALPHA;
     case Maxwell::Blend::Factor::OneMinusSourceAlpha:
     case Maxwell::Blend::Factor::OneMinusSourceAlphaGL:
-        return vk::BlendFactor::eOneMinusSrcAlpha;
+        return VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA;
     case Maxwell::Blend::Factor::DestAlpha:
     case Maxwell::Blend::Factor::DestAlphaGL:
-        return vk::BlendFactor::eDstAlpha;
+        return VK_BLEND_FACTOR_DST_ALPHA;
     case Maxwell::Blend::Factor::OneMinusDestAlpha:
     case Maxwell::Blend::Factor::OneMinusDestAlphaGL:
-        return vk::BlendFactor::eOneMinusDstAlpha;
+        return VK_BLEND_FACTOR_ONE_MINUS_DST_ALPHA;
     case Maxwell::Blend::Factor::DestColor:
     case Maxwell::Blend::Factor::DestColorGL:
-        return vk::BlendFactor::eDstColor;
+        return VK_BLEND_FACTOR_DST_COLOR;
     case Maxwell::Blend::Factor::OneMinusDestColor:
     case Maxwell::Blend::Factor::OneMinusDestColorGL:
-        return vk::BlendFactor::eOneMinusDstColor;
+        return VK_BLEND_FACTOR_ONE_MINUS_DST_COLOR;
     case Maxwell::Blend::Factor::SourceAlphaSaturate:
     case Maxwell::Blend::Factor::SourceAlphaSaturateGL:
-        return vk::BlendFactor::eSrcAlphaSaturate;
+        return VK_BLEND_FACTOR_SRC_ALPHA_SATURATE;
     case Maxwell::Blend::Factor::Source1Color:
     case Maxwell::Blend::Factor::Source1ColorGL:
-        return vk::BlendFactor::eSrc1Color;
+        return VK_BLEND_FACTOR_SRC1_COLOR;
     case Maxwell::Blend::Factor::OneMinusSource1Color:
     case Maxwell::Blend::Factor::OneMinusSource1ColorGL:
-        return vk::BlendFactor::eOneMinusSrc1Color;
+        return VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR;
     case Maxwell::Blend::Factor::Source1Alpha:
     case Maxwell::Blend::Factor::Source1AlphaGL:
-        return vk::BlendFactor::eSrc1Alpha;
+        return VK_BLEND_FACTOR_SRC1_ALPHA;
     case Maxwell::Blend::Factor::OneMinusSource1Alpha:
     case Maxwell::Blend::Factor::OneMinusSource1AlphaGL:
-        return vk::BlendFactor::eOneMinusSrc1Alpha;
+        return VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA;
     case Maxwell::Blend::Factor::ConstantColor:
     case Maxwell::Blend::Factor::ConstantColorGL:
-        return vk::BlendFactor::eConstantColor;
+        return VK_BLEND_FACTOR_CONSTANT_COLOR;
     case Maxwell::Blend::Factor::OneMinusConstantColor:
     case Maxwell::Blend::Factor::OneMinusConstantColorGL:
-        return vk::BlendFactor::eOneMinusConstantColor;
+        return VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_COLOR;
     case Maxwell::Blend::Factor::ConstantAlpha:
     case Maxwell::Blend::Factor::ConstantAlphaGL:
-        return vk::BlendFactor::eConstantAlpha;
+        return VK_BLEND_FACTOR_CONSTANT_ALPHA;
     case Maxwell::Blend::Factor::OneMinusConstantAlpha:
     case Maxwell::Blend::Factor::OneMinusConstantAlphaGL:
-        return vk::BlendFactor::eOneMinusConstantAlpha;
+        return VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_ALPHA;
     }
     UNIMPLEMENTED_MSG("Unimplemented blend factor={}", static_cast<u32>(factor));
     return {};
 }
 
-vk::FrontFace FrontFace(Maxwell::FrontFace front_face) {
+VkFrontFace FrontFace(Maxwell::FrontFace front_face) {
     switch (front_face) {
     case Maxwell::FrontFace::ClockWise:
-        return vk::FrontFace::eClockwise;
+        return VK_FRONT_FACE_CLOCKWISE;
     case Maxwell::FrontFace::CounterClockWise:
-        return vk::FrontFace::eCounterClockwise;
+        return VK_FRONT_FACE_COUNTER_CLOCKWISE;
     }
     UNIMPLEMENTED_MSG("Unimplemented front face={}", static_cast<u32>(front_face));
     return {};
 }
 
-vk::CullModeFlags CullFace(Maxwell::CullFace cull_face) {
+VkCullModeFlags CullFace(Maxwell::CullFace cull_face) {
     switch (cull_face) {
     case Maxwell::CullFace::Front:
-        return vk::CullModeFlagBits::eFront;
+        return VK_CULL_MODE_FRONT_BIT;
     case Maxwell::CullFace::Back:
-        return vk::CullModeFlagBits::eBack;
+        return VK_CULL_MODE_BACK_BIT;
     case Maxwell::CullFace::FrontAndBack:
-        return vk::CullModeFlagBits::eFrontAndBack;
+        return VK_CULL_MODE_FRONT_AND_BACK;
     }
     UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face));
     return {};
 }
 
-vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) {
+VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) {
     switch (swizzle) {
     case Tegra::Texture::SwizzleSource::Zero:
-        return vk::ComponentSwizzle::eZero;
+        return VK_COMPONENT_SWIZZLE_ZERO;
     case Tegra::Texture::SwizzleSource::R:
-        return vk::ComponentSwizzle::eR;
+        return VK_COMPONENT_SWIZZLE_R;
     case Tegra::Texture::SwizzleSource::G:
-        return vk::ComponentSwizzle::eG;
+        return VK_COMPONENT_SWIZZLE_G;
     case Tegra::Texture::SwizzleSource::B:
-        return vk::ComponentSwizzle::eB;
+        return VK_COMPONENT_SWIZZLE_B;
     case Tegra::Texture::SwizzleSource::A:
-        return vk::ComponentSwizzle::eA;
+        return VK_COMPONENT_SWIZZLE_A;
     case Tegra::Texture::SwizzleSource::OneInt:
     case Tegra::Texture::SwizzleSource::OneFloat:
-        return vk::ComponentSwizzle::eOne;
+        return VK_COMPONENT_SWIZZLE_ONE;
     }
     UNIMPLEMENTED_MSG("Unimplemented swizzle source={}", static_cast<u32>(swizzle));
     return {};

src/video_core/renderer_vulkan/maxwell_to_vk.h

@@ -6,8 +6,8 @@
 
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 #include "video_core/surface.h"
 #include "video_core/textures/texture.h"
 
@@ -18,46 +18,45 @@ using PixelFormat = VideoCore::Surface::PixelFormat;
 
 namespace Sampler {
 
-vk::Filter Filter(Tegra::Texture::TextureFilter filter);
+VkFilter Filter(Tegra::Texture::TextureFilter filter);
 
-vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter);
+VkSamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter);
 
-vk::SamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode,
-                                Tegra::Texture::TextureFilter filter);
+VkSamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode,
+                              Tegra::Texture::TextureFilter filter);
 
-vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func);
+VkCompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func);
 
 } // namespace Sampler
 
 struct FormatInfo {
-    vk::Format format;
+    VkFormat format;
     bool attachable;
     bool storage;
 };
 
 FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFormat pixel_format);
 
-vk::ShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage);
+VkShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage);
 
-vk::PrimitiveTopology PrimitiveTopology(const VKDevice& device,
-                                        Maxwell::PrimitiveTopology topology);
+VkPrimitiveTopology PrimitiveTopology(const VKDevice& device, Maxwell::PrimitiveTopology topology);
 
-vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size);
+VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size);
 
-vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison);
+VkCompareOp ComparisonOp(Maxwell::ComparisonOp comparison);
 
-vk::IndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format);
+VkIndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format);
 
-vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op);
+VkStencilOp StencilOp(Maxwell::StencilOp stencil_op);
 
-vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation);
+VkBlendOp BlendEquation(Maxwell::Blend::Equation equation);
 
-vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor);
+VkBlendFactor BlendFactor(Maxwell::Blend::Factor factor);
 
-vk::FrontFace FrontFace(Maxwell::FrontFace front_face);
+VkFrontFace FrontFace(Maxwell::FrontFace front_face);
 
-vk::CullModeFlags CullFace(Maxwell::CullFace cull_face);
+VkCullModeFlags CullFace(Maxwell::CullFace cull_face);
 
-vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle);
+VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle);
 
 } // namespace Vulkan::MaxwellToVK

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -2,13 +2,18 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
+#include <array>
+#include <cstring>
 #include <memory>
 #include <optional>
+#include <string>
 #include <vector>
 
 #include <fmt/format.h>
 
 #include "common/assert.h"
+#include "common/dynamic_library.h"
 #include "common/logging/log.h"
 #include "common/telemetry.h"
 #include "core/core.h"
@@ -19,7 +24,6 @@
 #include "core/settings.h"
 #include "core/telemetry_session.h"
 #include "video_core/gpu.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/renderer_vulkan.h"
 #include "video_core/renderer_vulkan/vk_blit_screen.h"
 #include "video_core/renderer_vulkan/vk_device.h"
@@ -29,30 +33,145 @@
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_state_tracker.h"
 #include "video_core/renderer_vulkan/vk_swapchain.h"
+#include "video_core/renderer_vulkan/wrapper.h"
+
+// Include these late to avoid polluting previous headers
+#ifdef _WIN32
+#include <windows.h>
+// ensure include order
+#include <vulkan/vulkan_win32.h>
+#endif
+
+#ifdef __linux__
+#include <X11/Xlib.h>
+#include <vulkan/vulkan_wayland.h>
+#include <vulkan/vulkan_xlib.h>
+#endif
 
 namespace Vulkan {
 
 namespace {
 
-VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity_,
+using Core::Frontend::WindowSystemType;
+
+VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity,
                        VkDebugUtilsMessageTypeFlagsEXT type,
                        const VkDebugUtilsMessengerCallbackDataEXT* data,
                        [[maybe_unused]] void* user_data) {
-    const vk::DebugUtilsMessageSeverityFlagBitsEXT severity{severity_};
     const char* message{data->pMessage};
 
-    if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eError) {
+    if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT) {
         LOG_CRITICAL(Render_Vulkan, "{}", message);
-    } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eWarning) {
+    } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT) {
         LOG_WARNING(Render_Vulkan, "{}", message);
-    } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eInfo) {
+    } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT) {
         LOG_INFO(Render_Vulkan, "{}", message);
-    } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eVerbose) {
+    } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT) {
         LOG_DEBUG(Render_Vulkan, "{}", message);
     }
     return VK_FALSE;
 }
 
+Common::DynamicLibrary OpenVulkanLibrary() {
+    Common::DynamicLibrary library;
+#ifdef __APPLE__
+    // Check if a path to a specific Vulkan library has been specified.
+    char* libvulkan_env = getenv("LIBVULKAN_PATH");
+    if (!libvulkan_env || !library.Open(libvulkan_env)) {
+        // Use the libvulkan.dylib from the application bundle.
+        std::string filename = File::GetBundleDirectory() + "/Contents/Frameworks/libvulkan.dylib";
+        library.Open(filename.c_str());
+    }
+#else
+    std::string filename = Common::DynamicLibrary::GetVersionedFilename("vulkan", 1);
+    if (!library.Open(filename.c_str())) {
+        // Android devices may not have libvulkan.so.1, only libvulkan.so.
+        filename = Common::DynamicLibrary::GetVersionedFilename("vulkan");
+        library.Open(filename.c_str());
+    }
+#endif
+    return library;
+}
+
+vk::Instance CreateInstance(Common::DynamicLibrary& library, vk::InstanceDispatch& dld,
+                            WindowSystemType window_type = WindowSystemType::Headless,
+                            bool enable_layers = false) {
+    if (!library.IsOpen()) {
+        LOG_ERROR(Render_Vulkan, "Vulkan library not available");
+        return {};
+    }
+    if (!library.GetSymbol("vkGetInstanceProcAddr", &dld.vkGetInstanceProcAddr)) {
+        LOG_ERROR(Render_Vulkan, "vkGetInstanceProcAddr not present in Vulkan");
+        return {};
+    }
+    if (!vk::Load(dld)) {
+        LOG_ERROR(Render_Vulkan, "Failed to load Vulkan function pointers");
+        return {};
+    }
+
+    std::vector<const char*> extensions;
+    extensions.reserve(6);
+    switch (window_type) {
+    case Core::Frontend::WindowSystemType::Headless:
+        break;
+#ifdef _WIN32
+    case Core::Frontend::WindowSystemType::Windows:
+        extensions.push_back(VK_KHR_WIN32_SURFACE_EXTENSION_NAME);
+        break;
+#endif
+#ifdef __linux__
+    case Core::Frontend::WindowSystemType::X11:
+        extensions.push_back(VK_KHR_XLIB_SURFACE_EXTENSION_NAME);
+        break;
+    case Core::Frontend::WindowSystemType::Wayland:
+        extensions.push_back(VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME);
+        break;
+#endif
+    default:
+        LOG_ERROR(Render_Vulkan, "Presentation not supported on this platform");
+        break;
+    }
+    if (window_type != Core::Frontend::WindowSystemType::Headless) {
+        extensions.push_back(VK_KHR_SURFACE_EXTENSION_NAME);
+    }
+    if (enable_layers) {
+        extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
+    }
+    extensions.push_back(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME);
+
+    const std::optional properties = vk::EnumerateInstanceExtensionProperties(dld);
+    if (!properties) {
+        LOG_ERROR(Render_Vulkan, "Failed to query extension properties");
+        return {};
+    }
+
+    for (const char* extension : extensions) {
+        const auto it =
+            std::find_if(properties->begin(), properties->end(), [extension](const auto& prop) {
+                return !std::strcmp(extension, prop.extensionName);
+            });
+        if (it == properties->end()) {
+            LOG_ERROR(Render_Vulkan, "Required instance extension {} is not available", extension);
+            return {};
+        }
+    }
+
+    static constexpr std::array layers_data{"VK_LAYER_LUNARG_standard_validation"};
+    vk::Span<const char*> layers = layers_data;
+    if (!enable_layers) {
+        layers = {};
+    }
+    vk::Instance instance = vk::Instance::Create(layers, extensions, dld);
+    if (!instance) {
+        LOG_ERROR(Render_Vulkan, "Failed to create Vulkan instance");
+        return {};
+    }
+    if (!vk::Load(*instance, dld)) {
+        LOG_ERROR(Render_Vulkan, "Failed to load Vulkan instance function pointers");
+    }
+    return instance;
+}
+
 std::string GetReadableVersion(u32 version) {
     return fmt::format("{}.{}.{}", VK_VERSION_MAJOR(version), VK_VERSION_MINOR(version),
                        VK_VERSION_PATCH(version));
@@ -63,14 +182,14 @@ std::string GetDriverVersion(const VKDevice& device) {
     // https://github.com/SaschaWillems/vulkan.gpuinfo.org/blob/5dddea46ea1120b0df14eef8f15ff8e318e35462/functions.php#L308-L314
     const u32 version = device.GetDriverVersion();
 
-    if (device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) {
+    if (device.GetDriverID() == VK_DRIVER_ID_NVIDIA_PROPRIETARY_KHR) {
         const u32 major = (version >> 22) & 0x3ff;
         const u32 minor = (version >> 14) & 0x0ff;
         const u32 secondary = (version >> 6) & 0x0ff;
         const u32 tertiary = version & 0x003f;
         return fmt::format("{}.{}.{}.{}", major, minor, secondary, tertiary);
     }
-    if (device.GetDriverID() == vk::DriverIdKHR::eIntelProprietaryWindows) {
+    if (device.GetDriverID() == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS_KHR) {
         const u32 major = version >> 14;
         const u32 minor = version & 0x3fff;
         return fmt::format("{}.{}", major, minor);
@@ -141,32 +260,18 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     render_window.PollEvents();
 }
 
-void RendererVulkan::TryPresent(int /*timeout_ms*/) {
+bool RendererVulkan::TryPresent(int /*timeout_ms*/) {
     // TODO (bunnei): ImplementMe
+    return true;
 }
 
 bool RendererVulkan::Init() {
-    PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr{};
-    render_window.RetrieveVulkanHandlers(&vkGetInstanceProcAddr, &instance, &surface);
-    const vk::DispatchLoaderDynamic dldi(instance, vkGetInstanceProcAddr);
-
-    std::optional<vk::DebugUtilsMessengerEXT> callback;
-    if (Settings::values.renderer_debug && dldi.vkCreateDebugUtilsMessengerEXT) {
-        callback = CreateDebugCallback(dldi);
-        if (!callback) {
-            return false;
-        }
-    }
-
-    if (!PickDevices(dldi)) {
-        if (callback) {
-            instance.destroy(*callback, nullptr, dldi);
-        }
+    library = OpenVulkanLibrary();
+    instance = CreateInstance(library, dld, render_window.GetWindowInfo().type,
+                              Settings::values.renderer_debug);
+    if (!instance || !CreateDebugCallback() || !CreateSurface() || !PickDevices()) {
         return false;
     }
-    debug_callback = UniqueDebugUtilsMessengerEXT(
-        *callback, vk::ObjectDestroy<vk::Instance, vk::DispatchLoaderDynamic>(
-                       instance, nullptr, device->GetDispatchLoader()));
 
     Report();
 
@@ -175,7 +280,7 @@ bool RendererVulkan::Init() {
     resource_manager = std::make_unique<VKResourceManager>(*device);
 
     const auto& framebuffer = render_window.GetFramebufferLayout();
-    swapchain = std::make_unique<VKSwapchain>(surface, *device);
+    swapchain = std::make_unique<VKSwapchain>(*surface, *device);
     swapchain->Create(framebuffer.width, framebuffer.height, false);
 
     state_tracker = std::make_unique<StateTracker>(system);
@@ -197,10 +302,8 @@ void RendererVulkan::ShutDown() {
     if (!device) {
         return;
     }
-    const auto dev = device->GetLogical();
-    const auto& dld = device->GetDispatchLoader();
-    if (dev && dld.vkDeviceWaitIdle) {
-        dev.waitIdle(dld);
+    if (const auto& dev = device->GetLogical()) {
+        dev.WaitIdle();
     }
 
     rasterizer.reset();
@@ -212,44 +315,94 @@ void RendererVulkan::ShutDown() {
     device.reset();
 }
 
-std::optional<vk::DebugUtilsMessengerEXT> RendererVulkan::CreateDebugCallback(
-    const vk::DispatchLoaderDynamic& dldi) {
-    const vk::DebugUtilsMessengerCreateInfoEXT callback_ci(
-        {},
-        vk::DebugUtilsMessageSeverityFlagBitsEXT::eError |
-            vk::DebugUtilsMessageSeverityFlagBitsEXT::eWarning |
-            vk::DebugUtilsMessageSeverityFlagBitsEXT::eInfo |
-            vk::DebugUtilsMessageSeverityFlagBitsEXT::eVerbose,
-        vk::DebugUtilsMessageTypeFlagBitsEXT::eGeneral |
-            vk::DebugUtilsMessageTypeFlagBitsEXT::eValidation |
-            vk::DebugUtilsMessageTypeFlagBitsEXT::ePerformance,
-        &DebugCallback, nullptr);
-    vk::DebugUtilsMessengerEXT callback;
-    if (instance.createDebugUtilsMessengerEXT(&callback_ci, nullptr, &callback, dldi) !=
-        vk::Result::eSuccess) {
-        LOG_ERROR(Render_Vulkan, "Failed to create debug callback");
-        return {};
+bool RendererVulkan::CreateDebugCallback() {
+    if (!Settings::values.renderer_debug) {
+        return true;
     }
-    return callback;
+    debug_callback = instance.TryCreateDebugCallback(DebugCallback);
+    if (!debug_callback) {
+        LOG_ERROR(Render_Vulkan, "Failed to create debug callback");
+        return false;
+    }
+    return true;
 }
 
-bool RendererVulkan::PickDevices(const vk::DispatchLoaderDynamic& dldi) {
-    const auto devices = instance.enumeratePhysicalDevices(dldi);
+bool RendererVulkan::CreateSurface() {
+    [[maybe_unused]] const auto& window_info = render_window.GetWindowInfo();
+    VkSurfaceKHR unsafe_surface = nullptr;
+
+#ifdef _WIN32
+    if (window_info.type == Core::Frontend::WindowSystemType::Windows) {
+        const HWND hWnd = static_cast<HWND>(window_info.render_surface);
+        const VkWin32SurfaceCreateInfoKHR win32_ci{VK_STRUCTURE_TYPE_WIN32_SURFACE_CREATE_INFO_KHR,
+                                                   nullptr, 0, nullptr, hWnd};
+        const auto vkCreateWin32SurfaceKHR = reinterpret_cast<PFN_vkCreateWin32SurfaceKHR>(
+            dld.vkGetInstanceProcAddr(*instance, "vkCreateWin32SurfaceKHR"));
+        if (!vkCreateWin32SurfaceKHR ||
+            vkCreateWin32SurfaceKHR(*instance, &win32_ci, nullptr, &unsafe_surface) != VK_SUCCESS) {
+            LOG_ERROR(Render_Vulkan, "Failed to initialize Win32 surface");
+            return false;
+        }
+    }
+#endif
+#ifdef __linux__
+    if (window_info.type == Core::Frontend::WindowSystemType::X11) {
+        const VkXlibSurfaceCreateInfoKHR xlib_ci{
+            VK_STRUCTURE_TYPE_XLIB_SURFACE_CREATE_INFO_KHR, nullptr, 0,
+            static_cast<Display*>(window_info.display_connection),
+            reinterpret_cast<Window>(window_info.render_surface)};
+        const auto vkCreateXlibSurfaceKHR = reinterpret_cast<PFN_vkCreateXlibSurfaceKHR>(
+            dld.vkGetInstanceProcAddr(*instance, "vkCreateXlibSurfaceKHR"));
+        if (!vkCreateXlibSurfaceKHR ||
+            vkCreateXlibSurfaceKHR(*instance, &xlib_ci, nullptr, &unsafe_surface) != VK_SUCCESS) {
+            LOG_ERROR(Render_Vulkan, "Failed to initialize Xlib surface");
+            return false;
+        }
+    }
+    if (window_info.type == Core::Frontend::WindowSystemType::Wayland) {
+        const VkWaylandSurfaceCreateInfoKHR wayland_ci{
+            VK_STRUCTURE_TYPE_WAYLAND_SURFACE_CREATE_INFO_KHR, nullptr, 0,
+            static_cast<wl_display*>(window_info.display_connection),
+            static_cast<wl_surface*>(window_info.render_surface)};
+        const auto vkCreateWaylandSurfaceKHR = reinterpret_cast<PFN_vkCreateWaylandSurfaceKHR>(
+            dld.vkGetInstanceProcAddr(*instance, "vkCreateWaylandSurfaceKHR"));
+        if (!vkCreateWaylandSurfaceKHR ||
+            vkCreateWaylandSurfaceKHR(*instance, &wayland_ci, nullptr, &unsafe_surface) !=
+                VK_SUCCESS) {
+            LOG_ERROR(Render_Vulkan, "Failed to initialize Wayland surface");
+            return false;
+        }
+    }
+#endif
+    if (!unsafe_surface) {
+        LOG_ERROR(Render_Vulkan, "Presentation not supported on this platform");
+        return false;
+    }
+
+    surface = vk::SurfaceKHR(unsafe_surface, *instance, dld);
+    return true;
+}
+
+bool RendererVulkan::PickDevices() {
+    const auto devices = instance.EnumeratePhysicalDevices();
+    if (!devices) {
+        LOG_ERROR(Render_Vulkan, "Failed to enumerate physical devices");
+        return false;
+    }
 
-    // TODO(Rodrigo): Choose device from config file
     const s32 device_index = Settings::values.vulkan_device;
-    if (device_index < 0 || device_index >= static_cast<s32>(devices.size())) {
+    if (device_index < 0 || device_index >= static_cast<s32>(devices->size())) {
         LOG_ERROR(Render_Vulkan, "Invalid device index {}!", device_index);
         return false;
     }
-    const vk::PhysicalDevice physical_device = devices[device_index];
-
-    if (!VKDevice::IsSuitable(dldi, physical_device, surface)) {
+    const vk::PhysicalDevice physical_device((*devices)[static_cast<std::size_t>(device_index)],
+                                             dld);
+    if (!VKDevice::IsSuitable(physical_device, *surface)) {
         return false;
     }
 
-    device = std::make_unique<VKDevice>(dldi, physical_device, surface);
-    return device->Create(dldi, instance);
+    device = std::make_unique<VKDevice>(*instance, physical_device, *surface, dld);
+    return device->Create();
 }
 
 void RendererVulkan::Report() const {
@@ -275,4 +428,25 @@ void RendererVulkan::Report() const {
     telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions);
 }
 
+std::vector<std::string> RendererVulkan::EnumerateDevices() {
+    vk::InstanceDispatch dld;
+    Common::DynamicLibrary library = OpenVulkanLibrary();
+    vk::Instance instance = CreateInstance(library, dld);
+    if (!instance) {
+        return {};
+    }
+
+    const std::optional physical_devices = instance.EnumeratePhysicalDevices();
+    if (!physical_devices) {
+        return {};
+    }
+
+    std::vector<std::string> names;
+    names.reserve(physical_devices->size());
+    for (const auto& device : *physical_devices) {
+        names.push_back(vk::PhysicalDevice(device, dld).GetProperties().deviceName);
+    }
+    return names;
+}
+
 } // namespace Vulkan

src/video_core/renderer_vulkan/renderer_vulkan.h

@@ -6,10 +6,13 @@
 
 #include <memory>
 #include <optional>
+#include <string>
 #include <vector>
 
+#include "common/dynamic_library.h"
+
 #include "video_core/renderer_base.h"
-#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Core {
 class System;
@@ -42,24 +45,30 @@ public:
     bool Init() override;
     void ShutDown() override;
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-    void TryPresent(int timeout_ms) override;
+    bool TryPresent(int timeout_ms) override;
+
+    static std::vector<std::string> EnumerateDevices();
 
 private:
-    std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback(
-        const vk::DispatchLoaderDynamic& dldi);
+    bool CreateDebugCallback();
 
-    bool PickDevices(const vk::DispatchLoaderDynamic& dldi);
+    bool CreateSurface();
+
+    bool PickDevices();
 
     void Report() const;
 
     Core::System& system;
 
+    Common::DynamicLibrary library;
+    vk::InstanceDispatch dld;
+
     vk::Instance instance;
     vk::SurfaceKHR surface;
 
     VKScreenInfo screen_info;
 
-    UniqueDebugUtilsMessengerEXT debug_callback;
+    vk::DebugCallback debug_callback;
     std::unique_ptr<VKDevice> device;
     std::unique_ptr<VKSwapchain> swapchain;
     std::unique_ptr<VKMemoryManager> memory_manager;

src/video_core/renderer_vulkan/shaders/quad_indexed.comp

@@ -0,0 +1,50 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+/*
+ * Build instructions:
+ * $ glslangValidator -V quad_indexed.comp -o output.spv
+ * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
+ * $ xxd -i optimized.spv
+ *
+ * Then copy that bytecode to the C++ file
+ */
+
+#version 460 core
+
+layout (local_size_x = 1024) in;
+
+layout (std430, set = 0, binding = 0) readonly buffer InputBuffer {
+    uint input_indexes[];
+};
+
+layout (std430, set = 0, binding = 1) writeonly buffer OutputBuffer {
+    uint output_indexes[];
+};
+
+layout (push_constant) uniform PushConstants {
+    uint base_vertex;
+    int index_shift; // 0: uint8, 1: uint16, 2: uint32
+};
+
+void main() {
+    int primitive = int(gl_GlobalInvocationID.x);
+    if (primitive * 6 >= output_indexes.length()) {
+        return;
+    }
+
+    int index_size = 8 << index_shift;
+    int flipped_shift = 2 - index_shift;
+    int mask = (1 << flipped_shift) - 1;
+
+    const int quad_swizzle[6] = int[](0, 1, 2, 0, 2, 3);
+    for (uint vertex = 0; vertex < 6; ++vertex) {
+        int offset = primitive * 4 + quad_swizzle[vertex];
+        int int_offset = offset >> flipped_shift;
+        int bit_offset = (offset & mask) * index_size;
+        uint packed_input = input_indexes[int_offset];
+        uint index = bitfieldExtract(packed_input, bit_offset, index_size);
+        output_indexes[primitive * 6 + vertex] = index + base_vertex;
+    }
+}

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -20,7 +20,6 @@
 #include "video_core/gpu.h"
 #include "video_core/morton.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/renderer_vulkan.h"
 #include "video_core/renderer_vulkan/vk_blit_screen.h"
 #include "video_core/renderer_vulkan/vk_device.h"
@@ -30,6 +29,7 @@
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_shader_util.h"
 #include "video_core/renderer_vulkan/vk_swapchain.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 #include "video_core/surface.h"
 
 namespace Vulkan {
@@ -140,16 +140,25 @@ struct ScreenRectVertex {
     std::array<f32, 2> position;
     std::array<f32, 2> tex_coord;
 
-    static vk::VertexInputBindingDescription GetDescription() {
-        return vk::VertexInputBindingDescription(0, sizeof(ScreenRectVertex),
-                                                 vk::VertexInputRate::eVertex);
+    static VkVertexInputBindingDescription GetDescription() {
+        VkVertexInputBindingDescription description;
+        description.binding = 0;
+        description.stride = sizeof(ScreenRectVertex);
+        description.inputRate = VK_VERTEX_INPUT_RATE_VERTEX;
+        return description;
     }
 
-    static std::array<vk::VertexInputAttributeDescription, 2> GetAttributes() {
-        return {vk::VertexInputAttributeDescription(0, 0, vk::Format::eR32G32Sfloat,
-                                                    offsetof(ScreenRectVertex, position)),
-                vk::VertexInputAttributeDescription(1, 0, vk::Format::eR32G32Sfloat,
-                                                    offsetof(ScreenRectVertex, tex_coord))};
+    static std::array<VkVertexInputAttributeDescription, 2> GetAttributes() {
+        std::array<VkVertexInputAttributeDescription, 2> attributes;
+        attributes[0].location = 0;
+        attributes[0].binding = 0;
+        attributes[0].format = VK_FORMAT_R32G32_SFLOAT;
+        attributes[0].offset = offsetof(ScreenRectVertex, position);
+        attributes[1].location = 1;
+        attributes[1].binding = 0;
+        attributes[1].format = VK_FORMAT_R32G32_SFLOAT;
+        attributes[1].offset = offsetof(ScreenRectVertex, tex_coord);
+        return attributes;
     }
 };
 
@@ -172,16 +181,16 @@ std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) {
            static_cast<std::size_t>(framebuffer.height) * GetBytesPerPixel(framebuffer);
 }
 
-vk::Format GetFormat(const Tegra::FramebufferConfig& framebuffer) {
+VkFormat GetFormat(const Tegra::FramebufferConfig& framebuffer) {
     switch (framebuffer.pixel_format) {
     case Tegra::FramebufferConfig::PixelFormat::ABGR8:
-        return vk::Format::eA8B8G8R8UnormPack32;
+        return VK_FORMAT_A8B8G8R8_UNORM_PACK32;
     case Tegra::FramebufferConfig::PixelFormat::RGB565:
-        return vk::Format::eR5G6B5UnormPack16;
+        return VK_FORMAT_R5G6B5_UNORM_PACK16;
     default:
         UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}",
                           static_cast<u32>(framebuffer.pixel_format));
-        return vk::Format::eA8B8G8R8UnormPack32;
+        return VK_FORMAT_A8B8G8R8_UNORM_PACK32;
     }
 }
 
@@ -219,8 +228,8 @@ void VKBlitScreen::Recreate() {
     CreateDynamicResources();
 }
 
-std::tuple<VKFence&, vk::Semaphore> VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
-                                                       bool use_accelerated) {
+std::tuple<VKFence&, VkSemaphore> VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
+                                                     bool use_accelerated) {
     RefreshResources(framebuffer);
 
     // Finish any pending renderpass
@@ -255,46 +264,76 @@ std::tuple<VKFence&, vk::Semaphore> VKBlitScreen::Draw(const Tegra::FramebufferC
                                  framebuffer.stride, block_height_log2, framebuffer.height, 0, 1, 1,
                                  map.GetAddress() + image_offset, host_ptr);
 
-        blit_image->Transition(0, 1, 0, 1, vk::PipelineStageFlagBits::eTransfer,
-                               vk::AccessFlagBits::eTransferWrite,
-                               vk::ImageLayout::eTransferDstOptimal);
+        blit_image->Transition(0, 1, 0, 1, VK_PIPELINE_STAGE_TRANSFER_BIT,
+                               VK_ACCESS_TRANSFER_WRITE_BIT, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
 
-        const vk::BufferImageCopy copy(image_offset, 0, 0,
-                                       {vk::ImageAspectFlagBits::eColor, 0, 0, 1}, {0, 0, 0},
-                                       {framebuffer.width, framebuffer.height, 1});
-        scheduler.Record([buffer_handle = *buffer, image = blit_image->GetHandle(),
-                          copy](auto cmdbuf, auto& dld) {
-            cmdbuf.copyBufferToImage(buffer_handle, image, vk::ImageLayout::eTransferDstOptimal,
-                                     {copy}, dld);
-        });
+        VkBufferImageCopy copy;
+        copy.bufferOffset = image_offset;
+        copy.bufferRowLength = 0;
+        copy.bufferImageHeight = 0;
+        copy.imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
+        copy.imageSubresource.mipLevel = 0;
+        copy.imageSubresource.baseArrayLayer = 0;
+        copy.imageSubresource.layerCount = 1;
+        copy.imageOffset.x = 0;
+        copy.imageOffset.y = 0;
+        copy.imageOffset.z = 0;
+        copy.imageExtent.width = framebuffer.width;
+        copy.imageExtent.height = framebuffer.height;
+        copy.imageExtent.depth = 1;
+        scheduler.Record(
+            [buffer = *buffer, image = *blit_image->GetHandle(), copy](vk::CommandBuffer cmdbuf) {
+                cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy);
+            });
     }
     map.Release();
 
-    blit_image->Transition(0, 1, 0, 1, vk::PipelineStageFlagBits::eFragmentShader,
-                           vk::AccessFlagBits::eShaderRead,
-                           vk::ImageLayout::eShaderReadOnlyOptimal);
+    blit_image->Transition(0, 1, 0, 1, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT,
+                           VK_ACCESS_SHADER_READ_BIT, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
 
     scheduler.Record([renderpass = *renderpass, framebuffer = *framebuffers[image_index],
                       descriptor_set = descriptor_sets[image_index], buffer = *buffer,
                       size = swapchain.GetSize(), pipeline = *pipeline,
-                      layout = *pipeline_layout](auto cmdbuf, auto& dld) {
-        const vk::ClearValue clear_color{std::array{0.0f, 0.0f, 0.0f, 1.0f}};
-        const vk::RenderPassBeginInfo renderpass_bi(renderpass, framebuffer, {{0, 0}, size}, 1,
-                                                    &clear_color);
+                      layout = *pipeline_layout](vk::CommandBuffer cmdbuf) {
+        VkClearValue clear_color;
+        clear_color.color.float32[0] = 0.0f;
+        clear_color.color.float32[1] = 0.0f;
+        clear_color.color.float32[2] = 0.0f;
+        clear_color.color.float32[3] = 0.0f;
 
-        cmdbuf.beginRenderPass(renderpass_bi, vk::SubpassContents::eInline, dld);
-        cmdbuf.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline, dld);
-        cmdbuf.setViewport(
-            0,
-            {{0.0f, 0.0f, static_cast<f32>(size.width), static_cast<f32>(size.height), 0.0f, 1.0f}},
-            dld);
-        cmdbuf.setScissor(0, {{{0, 0}, size}}, dld);
+        VkRenderPassBeginInfo renderpass_bi;
+        renderpass_bi.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
+        renderpass_bi.pNext = nullptr;
+        renderpass_bi.renderPass = renderpass;
+        renderpass_bi.framebuffer = framebuffer;
+        renderpass_bi.renderArea.offset.x = 0;
+        renderpass_bi.renderArea.offset.y = 0;
+        renderpass_bi.renderArea.extent = size;
+        renderpass_bi.clearValueCount = 1;
+        renderpass_bi.pClearValues = &clear_color;
 
-        cmdbuf.bindVertexBuffers(0, {buffer}, {offsetof(BufferData, vertices)}, dld);
-        cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eGraphics, layout, 0, {descriptor_set}, {},
-                                  dld);
-        cmdbuf.draw(4, 1, 0, 0, dld);
-        cmdbuf.endRenderPass(dld);
+        VkViewport viewport;
+        viewport.x = 0.0f;
+        viewport.y = 0.0f;
+        viewport.width = static_cast<float>(size.width);
+        viewport.height = static_cast<float>(size.height);
+        viewport.minDepth = 0.0f;
+        viewport.maxDepth = 1.0f;
+
+        VkRect2D scissor;
+        scissor.offset.x = 0;
+        scissor.offset.y = 0;
+        scissor.extent = size;
+
+        cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+        cmdbuf.SetViewport(0, viewport);
+        cmdbuf.SetScissor(0, scissor);
+
+        cmdbuf.BindVertexBuffer(0, buffer, offsetof(BufferData, vertices));
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set, {});
+        cmdbuf.Draw(4, 1, 0, 0);
+        cmdbuf.EndRenderPass();
     });
 
     return {scheduler.GetFence(), *semaphores[image_index]};
@@ -334,165 +373,297 @@ void VKBlitScreen::CreateShaders() {
 }
 
 void VKBlitScreen::CreateSemaphores() {
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-
     semaphores.resize(image_count);
-    for (std::size_t i = 0; i < image_count; ++i) {
-        semaphores[i] = dev.createSemaphoreUnique({}, nullptr, dld);
-    }
+    std::generate(semaphores.begin(), semaphores.end(),
+                  [this] { return device.GetLogical().CreateSemaphore(); });
 }
 
 void VKBlitScreen::CreateDescriptorPool() {
-    const std::array<vk::DescriptorPoolSize, 2> pool_sizes{
-        vk::DescriptorPoolSize{vk::DescriptorType::eUniformBuffer, static_cast<u32>(image_count)},
-        vk::DescriptorPoolSize{vk::DescriptorType::eCombinedImageSampler,
-                               static_cast<u32>(image_count)}};
-    const vk::DescriptorPoolCreateInfo pool_ci(
-        {}, static_cast<u32>(image_count), static_cast<u32>(pool_sizes.size()), pool_sizes.data());
-    const auto dev = device.GetLogical();
-    descriptor_pool = dev.createDescriptorPoolUnique(pool_ci, nullptr, device.GetDispatchLoader());
+    std::array<VkDescriptorPoolSize, 2> pool_sizes;
+    pool_sizes[0].type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
+    pool_sizes[0].descriptorCount = static_cast<u32>(image_count);
+    pool_sizes[1].type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
+    pool_sizes[1].descriptorCount = static_cast<u32>(image_count);
+
+    VkDescriptorPoolCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT;
+    ci.maxSets = static_cast<u32>(image_count);
+    ci.poolSizeCount = static_cast<u32>(pool_sizes.size());
+    ci.pPoolSizes = pool_sizes.data();
+    descriptor_pool = device.GetLogical().CreateDescriptorPool(ci);
 }
 
 void VKBlitScreen::CreateRenderPass() {
-    const vk::AttachmentDescription color_attachment(
-        {}, swapchain.GetImageFormat(), vk::SampleCountFlagBits::e1, vk::AttachmentLoadOp::eClear,
-        vk::AttachmentStoreOp::eStore, vk::AttachmentLoadOp::eDontCare,
-        vk::AttachmentStoreOp::eDontCare, vk::ImageLayout::eUndefined,
-        vk::ImageLayout::ePresentSrcKHR);
+    VkAttachmentDescription color_attachment;
+    color_attachment.flags = 0;
+    color_attachment.format = swapchain.GetImageFormat();
+    color_attachment.samples = VK_SAMPLE_COUNT_1_BIT;
+    color_attachment.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
+    color_attachment.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
+    color_attachment.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
+    color_attachment.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
+    color_attachment.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
+    color_attachment.finalLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR;
 
-    const vk::AttachmentReference color_attachment_ref(0, vk::ImageLayout::eColorAttachmentOptimal);
+    VkAttachmentReference color_attachment_ref;
+    color_attachment_ref.attachment = 0;
+    color_attachment_ref.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
 
-    const vk::SubpassDescription subpass_description({}, vk::PipelineBindPoint::eGraphics, 0,
-                                                     nullptr, 1, &color_attachment_ref, nullptr,
-                                                     nullptr, 0, nullptr);
+    VkSubpassDescription subpass_description;
+    subpass_description.flags = 0;
+    subpass_description.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
+    subpass_description.inputAttachmentCount = 0;
+    subpass_description.pInputAttachments = nullptr;
+    subpass_description.colorAttachmentCount = 1;
+    subpass_description.pColorAttachments = &color_attachment_ref;
+    subpass_description.pResolveAttachments = nullptr;
+    subpass_description.pDepthStencilAttachment = nullptr;
+    subpass_description.preserveAttachmentCount = 0;
+    subpass_description.pPreserveAttachments = nullptr;
 
-    const vk::SubpassDependency dependency(
-        VK_SUBPASS_EXTERNAL, 0, vk::PipelineStageFlagBits::eColorAttachmentOutput,
-        vk::PipelineStageFlagBits::eColorAttachmentOutput, {},
-        vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite, {});
+    VkSubpassDependency dependency;
+    dependency.srcSubpass = VK_SUBPASS_EXTERNAL;
+    dependency.dstSubpass = 0;
+    dependency.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
+    dependency.dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
+    dependency.srcAccessMask = 0;
+    dependency.dstAccessMask =
+        VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
+    dependency.dependencyFlags = 0;
 
-    const vk::RenderPassCreateInfo renderpass_ci({}, 1, &color_attachment, 1, &subpass_description,
-                                                 1, &dependency);
+    VkRenderPassCreateInfo renderpass_ci;
+    renderpass_ci.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO;
+    renderpass_ci.pNext = nullptr;
+    renderpass_ci.flags = 0;
+    renderpass_ci.attachmentCount = 1;
+    renderpass_ci.pAttachments = &color_attachment;
+    renderpass_ci.subpassCount = 1;
+    renderpass_ci.pSubpasses = &subpass_description;
+    renderpass_ci.dependencyCount = 1;
+    renderpass_ci.pDependencies = &dependency;
 
-    const auto dev = device.GetLogical();
-    renderpass = dev.createRenderPassUnique(renderpass_ci, nullptr, device.GetDispatchLoader());
+    renderpass = device.GetLogical().CreateRenderPass(renderpass_ci);
 }
 
 void VKBlitScreen::CreateDescriptorSetLayout() {
-    const std::array<vk::DescriptorSetLayoutBinding, 2> layout_bindings{
-        vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eUniformBuffer, 1,
-                                       vk::ShaderStageFlagBits::eVertex, nullptr),
-        vk::DescriptorSetLayoutBinding(1, vk::DescriptorType::eCombinedImageSampler, 1,
-                                       vk::ShaderStageFlagBits::eFragment, nullptr)};
-    const vk::DescriptorSetLayoutCreateInfo descriptor_layout_ci(
-        {}, static_cast<u32>(layout_bindings.size()), layout_bindings.data());
+    std::array<VkDescriptorSetLayoutBinding, 2> layout_bindings;
+    layout_bindings[0].binding = 0;
+    layout_bindings[0].descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
+    layout_bindings[0].descriptorCount = 1;
+    layout_bindings[0].stageFlags = VK_SHADER_STAGE_VERTEX_BIT;
+    layout_bindings[0].pImmutableSamplers = nullptr;
+    layout_bindings[1].binding = 1;
+    layout_bindings[1].descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
+    layout_bindings[1].descriptorCount = 1;
+    layout_bindings[1].stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT;
+    layout_bindings[1].pImmutableSamplers = nullptr;
 
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-    descriptor_set_layout = dev.createDescriptorSetLayoutUnique(descriptor_layout_ci, nullptr, dld);
+    VkDescriptorSetLayoutCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.bindingCount = static_cast<u32>(layout_bindings.size());
+    ci.pBindings = layout_bindings.data();
+
+    descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout(ci);
 }
 
 void VKBlitScreen::CreateDescriptorSets() {
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
+    const std::vector layouts(image_count, *descriptor_set_layout);
 
-    descriptor_sets.resize(image_count);
-    for (std::size_t i = 0; i < image_count; ++i) {
-        const vk::DescriptorSetLayout layout = *descriptor_set_layout;
-        const vk::DescriptorSetAllocateInfo descriptor_set_ai(*descriptor_pool, 1, &layout);
-        const vk::Result result =
-            dev.allocateDescriptorSets(&descriptor_set_ai, &descriptor_sets[i], dld);
-        ASSERT(result == vk::Result::eSuccess);
-    }
+    VkDescriptorSetAllocateInfo ai;
+    ai.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO;
+    ai.pNext = nullptr;
+    ai.descriptorPool = *descriptor_pool;
+    ai.descriptorSetCount = static_cast<u32>(image_count);
+    ai.pSetLayouts = layouts.data();
+    descriptor_sets = descriptor_pool.Allocate(ai);
 }
 
 void VKBlitScreen::CreatePipelineLayout() {
-    const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &descriptor_set_layout.get(), 0,
-                                                          nullptr);
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-    pipeline_layout = dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld);
+    VkPipelineLayoutCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.setLayoutCount = 1;
+    ci.pSetLayouts = descriptor_set_layout.address();
+    ci.pushConstantRangeCount = 0;
+    ci.pPushConstantRanges = nullptr;
+    pipeline_layout = device.GetLogical().CreatePipelineLayout(ci);
 }
 
 void VKBlitScreen::CreateGraphicsPipeline() {
-    const std::array shader_stages = {
-        vk::PipelineShaderStageCreateInfo({}, vk::ShaderStageFlagBits::eVertex, *vertex_shader,
-                                          "main", nullptr),
-        vk::PipelineShaderStageCreateInfo({}, vk::ShaderStageFlagBits::eFragment, *fragment_shader,
-                                          "main", nullptr)};
+    std::array<VkPipelineShaderStageCreateInfo, 2> shader_stages;
+    shader_stages[0].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
+    shader_stages[0].pNext = nullptr;
+    shader_stages[0].flags = 0;
+    shader_stages[0].stage = VK_SHADER_STAGE_VERTEX_BIT;
+    shader_stages[0].module = *vertex_shader;
+    shader_stages[0].pName = "main";
+    shader_stages[0].pSpecializationInfo = nullptr;
+    shader_stages[1].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
+    shader_stages[1].pNext = nullptr;
+    shader_stages[1].flags = 0;
+    shader_stages[1].stage = VK_SHADER_STAGE_FRAGMENT_BIT;
+    shader_stages[1].module = *fragment_shader;
+    shader_stages[1].pName = "main";
+    shader_stages[1].pSpecializationInfo = nullptr;
 
     const auto vertex_binding_description = ScreenRectVertex::GetDescription();
     const auto vertex_attrs_description = ScreenRectVertex::GetAttributes();
-    const vk::PipelineVertexInputStateCreateInfo vertex_input(
-        {}, 1, &vertex_binding_description, static_cast<u32>(vertex_attrs_description.size()),
-        vertex_attrs_description.data());
 
-    const vk::PipelineInputAssemblyStateCreateInfo input_assembly(
-        {}, vk::PrimitiveTopology::eTriangleStrip, false);
+    VkPipelineVertexInputStateCreateInfo vertex_input_ci;
+    vertex_input_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO;
+    vertex_input_ci.pNext = nullptr;
+    vertex_input_ci.flags = 0;
+    vertex_input_ci.vertexBindingDescriptionCount = 1;
+    vertex_input_ci.pVertexBindingDescriptions = &vertex_binding_description;
+    vertex_input_ci.vertexAttributeDescriptionCount = u32{vertex_attrs_description.size()};
+    vertex_input_ci.pVertexAttributeDescriptions = vertex_attrs_description.data();
 
-    // Set a dummy viewport, it's going to be replaced by dynamic states.
-    const vk::Viewport viewport(0.0f, 0.0f, 1.0f, 1.0f, 0.0f, 1.0f);
-    const vk::Rect2D scissor({0, 0}, {1, 1});
+    VkPipelineInputAssemblyStateCreateInfo input_assembly_ci;
+    input_assembly_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO;
+    input_assembly_ci.pNext = nullptr;
+    input_assembly_ci.flags = 0;
+    input_assembly_ci.topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP;
+    input_assembly_ci.primitiveRestartEnable = VK_FALSE;
 
-    const vk::PipelineViewportStateCreateInfo viewport_state({}, 1, &viewport, 1, &scissor);
+    VkPipelineViewportStateCreateInfo viewport_state_ci;
+    viewport_state_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO;
+    viewport_state_ci.pNext = nullptr;
+    viewport_state_ci.flags = 0;
+    viewport_state_ci.viewportCount = 1;
+    viewport_state_ci.pViewports = nullptr;
+    viewport_state_ci.scissorCount = 1;
+    viewport_state_ci.pScissors = nullptr;
 
-    const vk::PipelineRasterizationStateCreateInfo rasterizer(
-        {}, false, false, vk::PolygonMode::eFill, vk::CullModeFlagBits::eNone,
-        vk::FrontFace::eClockwise, false, 0.0f, 0.0f, 0.0f, 1.0f);
+    VkPipelineRasterizationStateCreateInfo rasterization_ci;
+    rasterization_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO;
+    rasterization_ci.pNext = nullptr;
+    rasterization_ci.flags = 0;
+    rasterization_ci.depthClampEnable = VK_FALSE;
+    rasterization_ci.rasterizerDiscardEnable = VK_FALSE;
+    rasterization_ci.polygonMode = VK_POLYGON_MODE_FILL;
+    rasterization_ci.cullMode = VK_CULL_MODE_NONE;
+    rasterization_ci.frontFace = VK_FRONT_FACE_CLOCKWISE;
+    rasterization_ci.depthBiasEnable = VK_FALSE;
+    rasterization_ci.depthBiasConstantFactor = 0.0f;
+    rasterization_ci.depthBiasClamp = 0.0f;
+    rasterization_ci.depthBiasSlopeFactor = 0.0f;
+    rasterization_ci.lineWidth = 1.0f;
 
-    const vk::PipelineMultisampleStateCreateInfo multisampling({}, vk::SampleCountFlagBits::e1,
-                                                               false, 0.0f, nullptr, false, false);
+    VkPipelineMultisampleStateCreateInfo multisampling_ci;
+    multisampling_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO;
+    multisampling_ci.pNext = nullptr;
+    multisampling_ci.flags = 0;
+    multisampling_ci.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT;
+    multisampling_ci.sampleShadingEnable = VK_FALSE;
+    multisampling_ci.minSampleShading = 0.0f;
+    multisampling_ci.pSampleMask = nullptr;
+    multisampling_ci.alphaToCoverageEnable = VK_FALSE;
+    multisampling_ci.alphaToOneEnable = VK_FALSE;
 
-    const vk::PipelineColorBlendAttachmentState color_blend_attachment(
-        false, vk::BlendFactor::eZero, vk::BlendFactor::eZero, vk::BlendOp::eAdd,
-        vk::BlendFactor::eZero, vk::BlendFactor::eZero, vk::BlendOp::eAdd,
-        vk::ColorComponentFlagBits::eR | vk::ColorComponentFlagBits::eG |
-            vk::ColorComponentFlagBits::eB | vk::ColorComponentFlagBits::eA);
+    VkPipelineColorBlendAttachmentState color_blend_attachment;
+    color_blend_attachment.blendEnable = VK_FALSE;
+    color_blend_attachment.srcColorBlendFactor = VK_BLEND_FACTOR_ZERO;
+    color_blend_attachment.dstColorBlendFactor = VK_BLEND_FACTOR_ZERO;
+    color_blend_attachment.colorBlendOp = VK_BLEND_OP_ADD;
+    color_blend_attachment.srcAlphaBlendFactor = VK_BLEND_FACTOR_ZERO;
+    color_blend_attachment.dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO;
+    color_blend_attachment.alphaBlendOp = VK_BLEND_OP_ADD;
+    color_blend_attachment.colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT |
+                                            VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT;
 
-    const vk::PipelineColorBlendStateCreateInfo color_blending(
-        {}, false, vk::LogicOp::eCopy, 1, &color_blend_attachment, {0.0f, 0.0f, 0.0f, 0.0f});
+    VkPipelineColorBlendStateCreateInfo color_blend_ci;
+    color_blend_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO;
+    color_blend_ci.flags = 0;
+    color_blend_ci.pNext = nullptr;
+    color_blend_ci.logicOpEnable = VK_FALSE;
+    color_blend_ci.logicOp = VK_LOGIC_OP_COPY;
+    color_blend_ci.attachmentCount = 1;
+    color_blend_ci.pAttachments = &color_blend_attachment;
+    color_blend_ci.blendConstants[0] = 0.0f;
+    color_blend_ci.blendConstants[1] = 0.0f;
+    color_blend_ci.blendConstants[2] = 0.0f;
+    color_blend_ci.blendConstants[3] = 0.0f;
 
-    const std::array<vk::DynamicState, 2> dynamic_states = {vk::DynamicState::eViewport,
-                                                            vk::DynamicState::eScissor};
+    static constexpr std::array dynamic_states = {VK_DYNAMIC_STATE_VIEWPORT,
+                                                  VK_DYNAMIC_STATE_SCISSOR};
+    VkPipelineDynamicStateCreateInfo dynamic_state_ci;
+    dynamic_state_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO;
+    dynamic_state_ci.pNext = nullptr;
+    dynamic_state_ci.flags = 0;
+    dynamic_state_ci.dynamicStateCount = static_cast<u32>(dynamic_states.size());
+    dynamic_state_ci.pDynamicStates = dynamic_states.data();
 
-    const vk::PipelineDynamicStateCreateInfo dynamic_state(
-        {}, static_cast<u32>(dynamic_states.size()), dynamic_states.data());
+    VkGraphicsPipelineCreateInfo pipeline_ci;
+    pipeline_ci.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO;
+    pipeline_ci.pNext = nullptr;
+    pipeline_ci.flags = 0;
+    pipeline_ci.stageCount = static_cast<u32>(shader_stages.size());
+    pipeline_ci.pStages = shader_stages.data();
+    pipeline_ci.pVertexInputState = &vertex_input_ci;
+    pipeline_ci.pInputAssemblyState = &input_assembly_ci;
+    pipeline_ci.pTessellationState = nullptr;
+    pipeline_ci.pViewportState = &viewport_state_ci;
+    pipeline_ci.pRasterizationState = &rasterization_ci;
+    pipeline_ci.pMultisampleState = &multisampling_ci;
+    pipeline_ci.pDepthStencilState = nullptr;
+    pipeline_ci.pColorBlendState = &color_blend_ci;
+    pipeline_ci.pDynamicState = &dynamic_state_ci;
+    pipeline_ci.layout = *pipeline_layout;
+    pipeline_ci.renderPass = *renderpass;
+    pipeline_ci.subpass = 0;
+    pipeline_ci.basePipelineHandle = 0;
+    pipeline_ci.basePipelineIndex = 0;
 
-    const vk::GraphicsPipelineCreateInfo pipeline_ci(
-        {}, static_cast<u32>(shader_stages.size()), shader_stages.data(), &vertex_input,
-        &input_assembly, nullptr, &viewport_state, &rasterizer, &multisampling, nullptr,
-        &color_blending, &dynamic_state, *pipeline_layout, *renderpass, 0, nullptr, 0);
-
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-    pipeline = dev.createGraphicsPipelineUnique({}, pipeline_ci, nullptr, dld);
+    pipeline = device.GetLogical().CreateGraphicsPipeline(pipeline_ci);
 }
 
 void VKBlitScreen::CreateSampler() {
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-    const vk::SamplerCreateInfo sampler_ci(
-        {}, vk::Filter::eLinear, vk::Filter::eLinear, vk::SamplerMipmapMode::eLinear,
-        vk::SamplerAddressMode::eClampToBorder, vk::SamplerAddressMode::eClampToBorder,
-        vk::SamplerAddressMode::eClampToBorder, 0.0f, false, 0.0f, false, vk::CompareOp::eNever,
-        0.0f, 0.0f, vk::BorderColor::eFloatOpaqueBlack, false);
-    sampler = dev.createSamplerUnique(sampler_ci, nullptr, dld);
+    VkSamplerCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.magFilter = VK_FILTER_LINEAR;
+    ci.minFilter = VK_FILTER_NEAREST;
+    ci.mipmapMode = VK_SAMPLER_MIPMAP_MODE_LINEAR;
+    ci.addressModeU = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER;
+    ci.addressModeV = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER;
+    ci.addressModeW = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER;
+    ci.mipLodBias = 0.0f;
+    ci.anisotropyEnable = VK_FALSE;
+    ci.maxAnisotropy = 0.0f;
+    ci.compareEnable = VK_FALSE;
+    ci.compareOp = VK_COMPARE_OP_NEVER;
+    ci.minLod = 0.0f;
+    ci.maxLod = 0.0f;
+    ci.borderColor = VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+    ci.unnormalizedCoordinates = VK_FALSE;
+
+    sampler = device.GetLogical().CreateSampler(ci);
 }
 
 void VKBlitScreen::CreateFramebuffers() {
-    const vk::Extent2D size{swapchain.GetSize()};
-    framebuffers.clear();
+    const VkExtent2D size{swapchain.GetSize()};
     framebuffers.resize(image_count);
 
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
+    VkFramebufferCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.renderPass = *renderpass;
+    ci.attachmentCount = 1;
+    ci.width = size.width;
+    ci.height = size.height;
+    ci.layers = 1;
 
     for (std::size_t i = 0; i < image_count; ++i) {
-        const vk::ImageView image_view{swapchain.GetImageViewIndex(i)};
-        const vk::FramebufferCreateInfo framebuffer_ci({}, *renderpass, 1, &image_view, size.width,
-                                                       size.height, 1);
-        framebuffers[i] = dev.createFramebufferUnique(framebuffer_ci, nullptr, dld);
+        const VkImageView image_view{swapchain.GetImageViewIndex(i)};
+        ci.pAttachments = &image_view;
+        framebuffers[i] = device.GetLogical().CreateFramebuffer(ci);
     }
 }
 
@@ -507,54 +678,86 @@ void VKBlitScreen::ReleaseRawImages() {
 }
 
 void VKBlitScreen::CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer) {
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
+    VkBufferCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.size = CalculateBufferSize(framebuffer);
+    ci.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
+               VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT;
+    ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
+    ci.queueFamilyIndexCount = 0;
+    ci.pQueueFamilyIndices = nullptr;
 
-    const vk::BufferCreateInfo buffer_ci({}, CalculateBufferSize(framebuffer),
-                                         vk::BufferUsageFlagBits::eTransferSrc |
-                                             vk::BufferUsageFlagBits::eVertexBuffer |
-                                             vk::BufferUsageFlagBits::eUniformBuffer,
-                                         vk::SharingMode::eExclusive, 0, nullptr);
-    buffer = dev.createBufferUnique(buffer_ci, nullptr, dld);
-    buffer_commit = memory_manager.Commit(*buffer, true);
+    buffer = device.GetLogical().CreateBuffer(ci);
+    buffer_commit = memory_manager.Commit(buffer, true);
 }
 
 void VKBlitScreen::CreateRawImages(const Tegra::FramebufferConfig& framebuffer) {
     raw_images.resize(image_count);
     raw_buffer_commits.resize(image_count);
 
-    const auto format = GetFormat(framebuffer);
-    for (std::size_t i = 0; i < image_count; ++i) {
-        const vk::ImageCreateInfo image_ci(
-            {}, vk::ImageType::e2D, format, {framebuffer.width, framebuffer.height, 1}, 1, 1,
-            vk::SampleCountFlagBits::e1, vk::ImageTiling::eOptimal,
-            vk::ImageUsageFlagBits::eTransferDst | vk::ImageUsageFlagBits::eSampled,
-            vk::SharingMode::eExclusive, 0, nullptr, vk::ImageLayout::eUndefined);
+    VkImageCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.imageType = VK_IMAGE_TYPE_2D;
+    ci.format = GetFormat(framebuffer);
+    ci.extent.width = framebuffer.width;
+    ci.extent.height = framebuffer.height;
+    ci.extent.depth = 1;
+    ci.mipLevels = 1;
+    ci.arrayLayers = 1;
+    ci.samples = VK_SAMPLE_COUNT_1_BIT;
+    ci.tiling = VK_IMAGE_TILING_LINEAR;
+    ci.usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT;
+    ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
+    ci.queueFamilyIndexCount = 0;
+    ci.pQueueFamilyIndices = nullptr;
+    ci.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
 
-        raw_images[i] =
-            std::make_unique<VKImage>(device, scheduler, image_ci, vk::ImageAspectFlagBits::eColor);
+    for (std::size_t i = 0; i < image_count; ++i) {
+        raw_images[i] = std::make_unique<VKImage>(device, scheduler, ci, VK_IMAGE_ASPECT_COLOR_BIT);
         raw_buffer_commits[i] = memory_manager.Commit(raw_images[i]->GetHandle(), false);
     }
 }
 
-void VKBlitScreen::UpdateDescriptorSet(std::size_t image_index, vk::ImageView image_view) const {
-    const vk::DescriptorSet descriptor_set = descriptor_sets[image_index];
+void VKBlitScreen::UpdateDescriptorSet(std::size_t image_index, VkImageView image_view) const {
+    VkDescriptorBufferInfo buffer_info;
+    buffer_info.buffer = *buffer;
+    buffer_info.offset = offsetof(BufferData, uniform);
+    buffer_info.range = sizeof(BufferData::uniform);
 
-    const vk::DescriptorBufferInfo buffer_info(*buffer, offsetof(BufferData, uniform),
-                                               sizeof(BufferData::uniform));
-    const vk::WriteDescriptorSet ubo_write(descriptor_set, 0, 0, 1,
-                                           vk::DescriptorType::eUniformBuffer, nullptr,
-                                           &buffer_info, nullptr);
+    VkWriteDescriptorSet ubo_write;
+    ubo_write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
+    ubo_write.pNext = nullptr;
+    ubo_write.dstSet = descriptor_sets[image_index];
+    ubo_write.dstBinding = 0;
+    ubo_write.dstArrayElement = 0;
+    ubo_write.descriptorCount = 1;
+    ubo_write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
+    ubo_write.pImageInfo = nullptr;
+    ubo_write.pBufferInfo = &buffer_info;
+    ubo_write.pTexelBufferView = nullptr;
 
-    const vk::DescriptorImageInfo image_info(*sampler, image_view,
-                                             vk::ImageLayout::eShaderReadOnlyOptimal);
-    const vk::WriteDescriptorSet sampler_write(descriptor_set, 1, 0, 1,
-                                               vk::DescriptorType::eCombinedImageSampler,
-                                               &image_info, nullptr, nullptr);
+    VkDescriptorImageInfo image_info;
+    image_info.sampler = *sampler;
+    image_info.imageView = image_view;
+    image_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
 
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-    dev.updateDescriptorSets({ubo_write, sampler_write}, {}, dld);
+    VkWriteDescriptorSet sampler_write;
+    sampler_write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
+    sampler_write.pNext = nullptr;
+    sampler_write.dstSet = descriptor_sets[image_index];
+    sampler_write.dstBinding = 1;
+    sampler_write.dstArrayElement = 0;
+    sampler_write.descriptorCount = 1;
+    sampler_write.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
+    sampler_write.pImageInfo = &image_info;
+    sampler_write.pBufferInfo = nullptr;
+    sampler_write.pTexelBufferView = nullptr;
+
+    device.GetLogical().UpdateDescriptorSets(std::array{ubo_write, sampler_write}, {});
 }
 
 void VKBlitScreen::SetUniformData(BufferData& data,

src/video_core/renderer_vulkan/vk_blit_screen.h

@@ -8,9 +8,9 @@
 #include <memory>
 #include <tuple>
 
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_memory_manager.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Core {
 class System;
@@ -49,8 +49,8 @@ public:
 
     void Recreate();
 
-    std::tuple<VKFence&, vk::Semaphore> Draw(const Tegra::FramebufferConfig& framebuffer,
-                                             bool use_accelerated);
+    std::tuple<VKFence&, VkSemaphore> Draw(const Tegra::FramebufferConfig& framebuffer,
+                                           bool use_accelerated);
 
 private:
     struct BufferData;
@@ -74,7 +74,7 @@ private:
     void CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer);
     void CreateRawImages(const Tegra::FramebufferConfig& framebuffer);
 
-    void UpdateDescriptorSet(std::size_t image_index, vk::ImageView image_view) const;
+    void UpdateDescriptorSet(std::size_t image_index, VkImageView image_view) const;
     void SetUniformData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const;
     void SetVertexData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const;
 
@@ -93,23 +93,23 @@ private:
     const std::size_t image_count;
     const VKScreenInfo& screen_info;
 
-    UniqueShaderModule vertex_shader;
-    UniqueShaderModule fragment_shader;
-    UniqueDescriptorPool descriptor_pool;
-    UniqueDescriptorSetLayout descriptor_set_layout;
-    UniquePipelineLayout pipeline_layout;
-    UniquePipeline pipeline;
-    UniqueRenderPass renderpass;
-    std::vector<UniqueFramebuffer> framebuffers;
-    std::vector<vk::DescriptorSet> descriptor_sets;
-    UniqueSampler sampler;
+    vk::ShaderModule vertex_shader;
+    vk::ShaderModule fragment_shader;
+    vk::DescriptorPool descriptor_pool;
+    vk::DescriptorSetLayout descriptor_set_layout;
+    vk::PipelineLayout pipeline_layout;
+    vk::Pipeline pipeline;
+    vk::RenderPass renderpass;
+    std::vector<vk::Framebuffer> framebuffers;
+    vk::DescriptorSets descriptor_sets;
+    vk::Sampler sampler;
 
-    UniqueBuffer buffer;
+    vk::Buffer buffer;
     VKMemoryCommit buffer_commit;
 
     std::vector<std::unique_ptr<VKFenceWatch>> watches;
 
-    std::vector<UniqueSemaphore> semaphores;
+    std::vector<vk::Semaphore> semaphores;
     std::vector<std::unique_ptr<VKImage>> raw_images;
     std::vector<VKMemoryCommit> raw_buffer_commits;
     u32 raw_width = 0;

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -11,48 +11,50 @@
 #include "common/assert.h"
 #include "common/bit_util.h"
 #include "core/core.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"
 #include "video_core/renderer_vulkan/vk_device.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_stream_buffer.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
 namespace {
 
-const auto BufferUsage =
-    vk::BufferUsageFlagBits::eVertexBuffer | vk::BufferUsageFlagBits::eIndexBuffer |
-    vk::BufferUsageFlagBits::eUniformBuffer | vk::BufferUsageFlagBits::eStorageBuffer;
+constexpr VkBufferUsageFlags BUFFER_USAGE =
+    VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT |
+    VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
 
-const auto UploadPipelineStage =
-    vk::PipelineStageFlagBits::eTransfer | vk::PipelineStageFlagBits::eVertexInput |
-    vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader |
-    vk::PipelineStageFlagBits::eComputeShader;
+constexpr VkPipelineStageFlags UPLOAD_PIPELINE_STAGE =
+    VK_PIPELINE_STAGE_TRANSFER_BIT | VK_PIPELINE_STAGE_VERTEX_INPUT_BIT |
+    VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+    VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
 
-const auto UploadAccessBarriers =
-    vk::AccessFlagBits::eTransferRead | vk::AccessFlagBits::eShaderRead |
-    vk::AccessFlagBits::eUniformRead | vk::AccessFlagBits::eVertexAttributeRead |
-    vk::AccessFlagBits::eIndexRead;
+constexpr VkAccessFlags UPLOAD_ACCESS_BARRIERS =
+    VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_UNIFORM_READ_BIT |
+    VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDEX_READ_BIT;
 
-auto CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) {
-    return std::make_unique<VKStreamBuffer>(device, scheduler, BufferUsage);
+std::unique_ptr<VKStreamBuffer> CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) {
+    return std::make_unique<VKStreamBuffer>(device, scheduler, BUFFER_USAGE);
 }
 
 } // Anonymous namespace
 
 CachedBufferBlock::CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager,
-                                     CacheAddr cache_addr, std::size_t size)
-    : VideoCommon::BufferBlock{cache_addr, size} {
-    const vk::BufferCreateInfo buffer_ci({}, static_cast<vk::DeviceSize>(size),
-                                         BufferUsage | vk::BufferUsageFlagBits::eTransferSrc |
-                                             vk::BufferUsageFlagBits::eTransferDst,
-                                         vk::SharingMode::eExclusive, 0, nullptr);
+                                     VAddr cpu_addr, std::size_t size)
+    : VideoCommon::BufferBlock{cpu_addr, size} {
+    VkBufferCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.size = static_cast<VkDeviceSize>(size);
+    ci.usage = BUFFER_USAGE | VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT;
+    ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
+    ci.queueFamilyIndexCount = 0;
+    ci.pQueueFamilyIndices = nullptr;
 
-    const auto& dld{device.GetDispatchLoader()};
-    const auto dev{device.GetLogical()};
-    buffer.handle = dev.createBufferUnique(buffer_ci, nullptr, dld);
-    buffer.commit = memory_manager.Commit(*buffer.handle, false);
+    buffer.handle = device.GetLogical().CreateBuffer(ci);
+    buffer.commit = memory_manager.Commit(buffer.handle, false);
 }
 
 CachedBufferBlock::~CachedBufferBlock() = default;
@@ -60,30 +62,30 @@ CachedBufferBlock::~CachedBufferBlock() = default;
 VKBufferCache::VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
                              const VKDevice& device, VKMemoryManager& memory_manager,
                              VKScheduler& scheduler, VKStagingBufferPool& staging_pool)
-    : VideoCommon::BufferCache<Buffer, vk::Buffer, VKStreamBuffer>{rasterizer, system,
-                                                                   CreateStreamBuffer(device,
-                                                                                      scheduler)},
+    : VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer>{rasterizer, system,
+                                                                 CreateStreamBuffer(device,
+                                                                                    scheduler)},
       device{device}, memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{
                                                                                 staging_pool} {}
 
 VKBufferCache::~VKBufferCache() = default;
 
-Buffer VKBufferCache::CreateBlock(CacheAddr cache_addr, std::size_t size) {
-    return std::make_shared<CachedBufferBlock>(device, memory_manager, cache_addr, size);
+Buffer VKBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
+    return std::make_shared<CachedBufferBlock>(device, memory_manager, cpu_addr, size);
 }
 
-const vk::Buffer* VKBufferCache::ToHandle(const Buffer& buffer) {
+VkBuffer VKBufferCache::ToHandle(const Buffer& buffer) {
     return buffer->GetHandle();
 }
 
-const vk::Buffer* VKBufferCache::GetEmptyBuffer(std::size_t size) {
+VkBuffer VKBufferCache::GetEmptyBuffer(std::size_t size) {
     size = std::max(size, std::size_t(4));
     const auto& empty = staging_pool.GetUnusedBuffer(size, false);
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([size, buffer = *empty.handle](vk::CommandBuffer cmdbuf, auto& dld) {
-        cmdbuf.fillBuffer(buffer, 0, size, 0, dld);
+    scheduler.Record([size, buffer = *empty.handle](vk::CommandBuffer cmdbuf) {
+        cmdbuf.FillBuffer(buffer, 0, size, 0);
     });
-    return &*empty.handle;
+    return *empty.handle;
 }
 
 void VKBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
@@ -92,15 +94,22 @@ void VKBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, st
     std::memcpy(staging.commit->Map(size), data, size);
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset,
-                      size](auto cmdbuf, auto& dld) {
-        cmdbuf.copyBuffer(staging, buffer, {{0, offset, size}}, dld);
-        cmdbuf.pipelineBarrier(
-            vk::PipelineStageFlagBits::eTransfer, UploadPipelineStage, {}, {},
-            {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, UploadAccessBarriers,
-                                     VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer,
-                                     offset, size)},
-            {}, dld);
+    scheduler.Record([staging = *staging.handle, buffer = buffer->GetHandle(), offset,
+                      size](vk::CommandBuffer cmdbuf) {
+        cmdbuf.CopyBuffer(staging, buffer, VkBufferCopy{0, offset, size});
+
+        VkBufferMemoryBarrier barrier;
+        barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
+        barrier.pNext = nullptr;
+        barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
+        barrier.dstAccessMask = UPLOAD_ACCESS_BARRIERS;
+        barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.buffer = buffer;
+        barrier.offset = offset;
+        barrier.size = size;
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0, {},
+                               barrier, {});
     });
 }
 
@@ -108,17 +117,24 @@ void VKBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset,
                                       u8* data) {
     const auto& staging = staging_pool.GetUnusedBuffer(size, true);
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset,
-                      size](auto cmdbuf, auto& dld) {
-        cmdbuf.pipelineBarrier(
-            vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader |
-                vk::PipelineStageFlagBits::eComputeShader,
-            vk::PipelineStageFlagBits::eTransfer, {}, {},
-            {vk::BufferMemoryBarrier(vk::AccessFlagBits::eShaderWrite,
-                                     vk::AccessFlagBits::eTransferRead, VK_QUEUE_FAMILY_IGNORED,
-                                     VK_QUEUE_FAMILY_IGNORED, buffer, offset, size)},
-            {}, dld);
-        cmdbuf.copyBuffer(buffer, staging, {{offset, 0, size}}, dld);
+    scheduler.Record([staging = *staging.handle, buffer = buffer->GetHandle(), offset,
+                      size](vk::CommandBuffer cmdbuf) {
+        VkBufferMemoryBarrier barrier;
+        barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
+        barrier.pNext = nullptr;
+        barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
+        barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
+        barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.buffer = buffer;
+        barrier.offset = offset;
+        barrier.size = size;
+
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_VERTEX_SHADER_BIT |
+                                   VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+                                   VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
+                               VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {}, barrier, {});
+        cmdbuf.CopyBuffer(buffer, staging, VkBufferCopy{offset, 0, size});
     });
     scheduler.Finish();
 
@@ -128,18 +144,31 @@ void VKBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset,
 void VKBufferCache::CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
                               std::size_t dst_offset, std::size_t size) {
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([src_buffer = *src->GetHandle(), dst_buffer = *dst->GetHandle(), src_offset,
-                      dst_offset, size](auto cmdbuf, auto& dld) {
-        cmdbuf.copyBuffer(src_buffer, dst_buffer, {{src_offset, dst_offset, size}}, dld);
-        cmdbuf.pipelineBarrier(
-            vk::PipelineStageFlagBits::eTransfer, UploadPipelineStage, {}, {},
-            {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferRead,
-                                     vk::AccessFlagBits::eShaderWrite, VK_QUEUE_FAMILY_IGNORED,
-                                     VK_QUEUE_FAMILY_IGNORED, src_buffer, src_offset, size),
-             vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, UploadAccessBarriers,
-                                     VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, dst_buffer,
-                                     dst_offset, size)},
-            {}, dld);
+    scheduler.Record([src_buffer = src->GetHandle(), dst_buffer = dst->GetHandle(), src_offset,
+                      dst_offset, size](vk::CommandBuffer cmdbuf) {
+        cmdbuf.CopyBuffer(src_buffer, dst_buffer, VkBufferCopy{src_offset, dst_offset, size});
+
+        std::array<VkBufferMemoryBarrier, 2> barriers;
+        barriers[0].sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
+        barriers[0].pNext = nullptr;
+        barriers[0].srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
+        barriers[0].dstAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
+        barriers[0].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barriers[0].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barriers[0].buffer = src_buffer;
+        barriers[0].offset = src_offset;
+        barriers[0].size = size;
+        barriers[1].sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
+        barriers[1].pNext = nullptr;
+        barriers[1].srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
+        barriers[1].dstAccessMask = UPLOAD_ACCESS_BARRIERS;
+        barriers[1].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barriers[1].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barriers[1].buffer = dst_buffer;
+        barriers[1].offset = dst_offset;
+        barriers[1].size = size;
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0, {},
+                               barriers, {});
     });
 }
 

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -11,11 +11,11 @@
 #include "common/common_types.h"
 #include "video_core/buffer_cache/buffer_cache.h"
 #include "video_core/rasterizer_cache.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_memory_manager.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
 #include "video_core/renderer_vulkan/vk_stream_buffer.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Core {
 class System;
@@ -30,11 +30,11 @@ class VKScheduler;
 class CachedBufferBlock final : public VideoCommon::BufferBlock {
 public:
     explicit CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager,
-                               CacheAddr cache_addr, std::size_t size);
+                               VAddr cpu_addr, std::size_t size);
     ~CachedBufferBlock();
 
-    const vk::Buffer* GetHandle() const {
-        return &*buffer.handle;
+    VkBuffer GetHandle() const {
+        return *buffer.handle;
     }
 
 private:
@@ -43,21 +43,21 @@ private:
 
 using Buffer = std::shared_ptr<CachedBufferBlock>;
 
-class VKBufferCache final : public VideoCommon::BufferCache<Buffer, vk::Buffer, VKStreamBuffer> {
+class VKBufferCache final : public VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer> {
 public:
     explicit VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
                            const VKDevice& device, VKMemoryManager& memory_manager,
                            VKScheduler& scheduler, VKStagingBufferPool& staging_pool);
     ~VKBufferCache();
 
-    const vk::Buffer* GetEmptyBuffer(std::size_t size) override;
+    VkBuffer GetEmptyBuffer(std::size_t size) override;
 
 protected:
+    VkBuffer ToHandle(const Buffer& buffer) override;
+
     void WriteBarrier() override {}
 
-    Buffer CreateBlock(CacheAddr cache_addr, std::size_t size) override;
-
-    const vk::Buffer* ToHandle(const Buffer& buffer) override;
+    Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override;
 
     void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                          const u8* data) override;

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -10,13 +10,13 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/common_types.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
 #include "video_core/renderer_vulkan/vk_device.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
@@ -114,6 +114,35 @@ constexpr u8 quad_array[] = {
     0xf9, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x4b, 0x00, 0x00, 0x00,
     0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
 
+VkDescriptorSetLayoutBinding BuildQuadArrayPassDescriptorSetLayoutBinding() {
+    VkDescriptorSetLayoutBinding binding;
+    binding.binding = 0;
+    binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
+    binding.descriptorCount = 1;
+    binding.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
+    binding.pImmutableSamplers = nullptr;
+    return binding;
+}
+
+VkDescriptorUpdateTemplateEntryKHR BuildQuadArrayPassDescriptorUpdateTemplateEntry() {
+    VkDescriptorUpdateTemplateEntryKHR entry;
+    entry.dstBinding = 0;
+    entry.dstArrayElement = 0;
+    entry.descriptorCount = 1;
+    entry.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
+    entry.offset = 0;
+    entry.stride = sizeof(DescriptorUpdateEntry);
+    return entry;
+}
+
+VkPushConstantRange BuildComputePushConstantRange(std::size_t size) {
+    VkPushConstantRange range;
+    range.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
+    range.offset = 0;
+    range.size = static_cast<u32>(size);
+    return range;
+}
+
 // Uint8 SPIR-V module. Generated from the "shaders/" directory.
 constexpr u8 uint8_pass[] = {
     0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x2f, 0x00, 0x00, 0x00,
@@ -191,53 +220,234 @@ constexpr u8 uint8_pass[] = {
     0xf9, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00,
     0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
 
+// Quad indexed SPIR-V module. Generated from the "shaders/" directory.
+constexpr u8 QUAD_INDEXED_SPV[] = {
+    0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x7c, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c, 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30,
+    0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x0f, 0x00, 0x06, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e,
+    0x00, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x10, 0x00, 0x06, 0x00, 0x04, 0x00, 0x00, 0x00,
+    0x11, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x15, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x04, 0x00, 0x16, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x05, 0x00, 0x16, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x16, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x05, 0x00, 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x22, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x23, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x22, 0x00, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x56, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x04, 0x00, 0x00, 0x00, 0x48, 0x00, 0x04, 0x00, 0x57, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x18, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x57, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x23, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x57, 0x00, 0x00, 0x00,
+    0x03, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x59, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x59, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x72, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00,
+    0x19, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00, 0x21, 0x00, 0x03, 0x00,
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x20, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00,
+    0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00,
+    0x20, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x0a, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00,
+    0x0c, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00,
+    0x0d, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0e, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x13, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00, 0x15, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x16, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00,
+    0x20, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00,
+    0x3b, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x14, 0x00, 0x02, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x21, 0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x04, 0x00, 0x22, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x24, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x25, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x26, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x2b, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00,
+    0x3b, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x3f, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x04, 0x00, 0x41, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x42, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x43, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x09, 0x00, 0x41, 0x00, 0x00, 0x00,
+    0x44, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x00, 0x00,
+    0x42, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x00, 0x00, 0x43, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
+    0x46, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x41, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00,
+    0x56, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x57, 0x00, 0x00, 0x00,
+    0x56, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x58, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x57, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x58, 0x00, 0x00, 0x00, 0x59, 0x00, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x5b, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x69, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x70, 0x00, 0x00, 0x00,
+    0x00, 0x04, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x71, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x06, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x72, 0x00, 0x00, 0x00,
+    0x70, 0x00, 0x00, 0x00, 0x71, 0x00, 0x00, 0x00, 0x71, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
+    0xf8, 0x00, 0x02, 0x00, 0x05, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x46, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00, 0x74, 0x00, 0x00, 0x00,
+    0xf8, 0x00, 0x02, 0x00, 0x74, 0x00, 0x00, 0x00, 0xf6, 0x00, 0x04, 0x00, 0x73, 0x00, 0x00, 0x00,
+    0x76, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00, 0x75, 0x00, 0x00, 0x00,
+    0xf8, 0x00, 0x02, 0x00, 0x75, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x0e, 0x00, 0x00, 0x00,
+    0x0f, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x13, 0x00, 0x00, 0x00,
+    0x44, 0x00, 0x05, 0x00, 0x09, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00,
+    0x19, 0x00, 0x00, 0x00, 0xaf, 0x00, 0x05, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
+    0x14, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00, 0xf7, 0x00, 0x03, 0x00, 0x1e, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x04, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x00, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00,
+    0x73, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1e, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00,
+    0x26, 0x00, 0x00, 0x00, 0x27, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
+    0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00, 0x27, 0x00, 0x00, 0x00,
+    0xc4, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00,
+    0x28, 0x00, 0x00, 0x00, 0x82, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00,
+    0x2b, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00, 0xc4, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x31, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00, 0x82, 0x00, 0x05, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x32, 0x00, 0x00, 0x00, 0x31, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
+    0xf9, 0x00, 0x02, 0x00, 0x35, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x35, 0x00, 0x00, 0x00,
+    0xf5, 0x00, 0x07, 0x00, 0x09, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0x6f, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0xb0, 0x00, 0x05, 0x00,
+    0x1b, 0x00, 0x00, 0x00, 0x3c, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x00, 0x00,
+    0xf6, 0x00, 0x04, 0x00, 0x37, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0xfa, 0x00, 0x04, 0x00, 0x3c, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0x37, 0x00, 0x00, 0x00,
+    0xf8, 0x00, 0x02, 0x00, 0x36, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x40, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x3f, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00,
+    0x47, 0x00, 0x00, 0x00, 0x44, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x07, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x00, 0x00, 0x47, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x49, 0x00, 0x00, 0x00, 0x48, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x4a, 0x00, 0x00, 0x00, 0x40, 0x00, 0x00, 0x00, 0x49, 0x00, 0x00, 0x00,
+    0xc3, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x4e, 0x00, 0x00, 0x00, 0x4a, 0x00, 0x00, 0x00,
+    0x2e, 0x00, 0x00, 0x00, 0xc7, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x52, 0x00, 0x00, 0x00,
+    0x4a, 0x00, 0x00, 0x00, 0x32, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x54, 0x00, 0x00, 0x00, 0x52, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00,
+    0x5b, 0x00, 0x00, 0x00, 0x5c, 0x00, 0x00, 0x00, 0x59, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00,
+    0x4e, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x5d, 0x00, 0x00, 0x00,
+    0x5c, 0x00, 0x00, 0x00, 0xcb, 0x00, 0x06, 0x00, 0x09, 0x00, 0x00, 0x00, 0x62, 0x00, 0x00, 0x00,
+    0x5d, 0x00, 0x00, 0x00, 0x54, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x65, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x67, 0x00, 0x00, 0x00, 0x65, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00,
+    0x41, 0x00, 0x05, 0x00, 0x69, 0x00, 0x00, 0x00, 0x6a, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00,
+    0x42, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x6b, 0x00, 0x00, 0x00,
+    0x6a, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00, 0x09, 0x00, 0x00, 0x00, 0x6c, 0x00, 0x00, 0x00,
+    0x62, 0x00, 0x00, 0x00, 0x6b, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00, 0x5b, 0x00, 0x00, 0x00,
+    0x6d, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00, 0x67, 0x00, 0x00, 0x00,
+    0x3e, 0x00, 0x03, 0x00, 0x6d, 0x00, 0x00, 0x00, 0x6c, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x6f, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
+    0xf9, 0x00, 0x02, 0x00, 0x35, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x37, 0x00, 0x00, 0x00,
+    0xf9, 0x00, 0x02, 0x00, 0x73, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x76, 0x00, 0x00, 0x00,
+    0xf9, 0x00, 0x02, 0x00, 0x74, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x73, 0x00, 0x00, 0x00,
+    0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
+
+std::array<VkDescriptorSetLayoutBinding, 2> BuildInputOutputDescriptorSetBindings() {
+    std::array<VkDescriptorSetLayoutBinding, 2> bindings;
+    bindings[0].binding = 0;
+    bindings[0].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
+    bindings[0].descriptorCount = 1;
+    bindings[0].stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
+    bindings[0].pImmutableSamplers = nullptr;
+    bindings[1].binding = 1;
+    bindings[1].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
+    bindings[1].descriptorCount = 1;
+    bindings[1].stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
+    bindings[1].pImmutableSamplers = nullptr;
+    return bindings;
+}
+
+VkDescriptorUpdateTemplateEntryKHR BuildInputOutputDescriptorUpdateTemplate() {
+    VkDescriptorUpdateTemplateEntryKHR entry;
+    entry.dstBinding = 0;
+    entry.dstArrayElement = 0;
+    entry.descriptorCount = 2;
+    entry.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
+    entry.offset = 0;
+    entry.stride = sizeof(DescriptorUpdateEntry);
+    return entry;
+}
+
 } // Anonymous namespace
 
 VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool,
-                             const std::vector<vk::DescriptorSetLayoutBinding>& bindings,
-                             const std::vector<vk::DescriptorUpdateTemplateEntry>& templates,
-                             const std::vector<vk::PushConstantRange> push_constants,
-                             std::size_t code_size, const u8* code) {
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
+                             vk::Span<VkDescriptorSetLayoutBinding> bindings,
+                             vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates,
+                             vk::Span<VkPushConstantRange> push_constants, std::size_t code_size,
+                             const u8* code) {
+    VkDescriptorSetLayoutCreateInfo descriptor_layout_ci;
+    descriptor_layout_ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
+    descriptor_layout_ci.pNext = nullptr;
+    descriptor_layout_ci.flags = 0;
+    descriptor_layout_ci.bindingCount = bindings.size();
+    descriptor_layout_ci.pBindings = bindings.data();
+    descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout(descriptor_layout_ci);
 
-    const vk::DescriptorSetLayoutCreateInfo descriptor_layout_ci(
-        {}, static_cast<u32>(bindings.size()), bindings.data());
-    descriptor_set_layout = dev.createDescriptorSetLayoutUnique(descriptor_layout_ci, nullptr, dld);
-
-    const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &*descriptor_set_layout,
-                                                          static_cast<u32>(push_constants.size()),
-                                                          push_constants.data());
-    layout = dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld);
+    VkPipelineLayoutCreateInfo pipeline_layout_ci;
+    pipeline_layout_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
+    pipeline_layout_ci.pNext = nullptr;
+    pipeline_layout_ci.flags = 0;
+    pipeline_layout_ci.setLayoutCount = 1;
+    pipeline_layout_ci.pSetLayouts = descriptor_set_layout.address();
+    pipeline_layout_ci.pushConstantRangeCount = push_constants.size();
+    pipeline_layout_ci.pPushConstantRanges = push_constants.data();
+    layout = device.GetLogical().CreatePipelineLayout(pipeline_layout_ci);
 
     if (!templates.empty()) {
-        const vk::DescriptorUpdateTemplateCreateInfo template_ci(
-            {}, static_cast<u32>(templates.size()), templates.data(),
-            vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout,
-            vk::PipelineBindPoint::eGraphics, *layout, 0);
-        descriptor_template = dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld);
+        VkDescriptorUpdateTemplateCreateInfoKHR template_ci;
+        template_ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR;
+        template_ci.pNext = nullptr;
+        template_ci.flags = 0;
+        template_ci.descriptorUpdateEntryCount = templates.size();
+        template_ci.pDescriptorUpdateEntries = templates.data();
+        template_ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR;
+        template_ci.descriptorSetLayout = *descriptor_set_layout;
+        template_ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
+        template_ci.pipelineLayout = *layout;
+        template_ci.set = 0;
+        descriptor_template = device.GetLogical().CreateDescriptorUpdateTemplateKHR(template_ci);
 
         descriptor_allocator.emplace(descriptor_pool, *descriptor_set_layout);
     }
 
     auto code_copy = std::make_unique<u32[]>(code_size / sizeof(u32) + 1);
     std::memcpy(code_copy.get(), code, code_size);
-    const vk::ShaderModuleCreateInfo module_ci({}, code_size, code_copy.get());
-    module = dev.createShaderModuleUnique(module_ci, nullptr, dld);
 
-    const vk::PipelineShaderStageCreateInfo stage_ci({}, vk::ShaderStageFlagBits::eCompute, *module,
-                                                     "main", nullptr);
+    VkShaderModuleCreateInfo module_ci;
+    module_ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
+    module_ci.pNext = nullptr;
+    module_ci.flags = 0;
+    module_ci.codeSize = code_size;
+    module_ci.pCode = code_copy.get();
+    module = device.GetLogical().CreateShaderModule(module_ci);
 
-    const vk::ComputePipelineCreateInfo pipeline_ci({}, stage_ci, *layout, nullptr, 0);
-    pipeline = dev.createComputePipelineUnique(nullptr, pipeline_ci, nullptr, dld);
+    VkComputePipelineCreateInfo pipeline_ci;
+    pipeline_ci.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
+    pipeline_ci.pNext = nullptr;
+    pipeline_ci.flags = 0;
+    pipeline_ci.layout = *layout;
+    pipeline_ci.basePipelineHandle = nullptr;
+    pipeline_ci.basePipelineIndex = 0;
+
+    VkPipelineShaderStageCreateInfo& stage_ci = pipeline_ci.stage;
+    stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
+    stage_ci.pNext = nullptr;
+    stage_ci.flags = 0;
+    stage_ci.stage = VK_SHADER_STAGE_COMPUTE_BIT;
+    stage_ci.module = *module;
+    stage_ci.pName = "main";
+    stage_ci.pSpecializationInfo = nullptr;
+
+    pipeline = device.GetLogical().CreateComputePipeline(pipeline_ci);
 }
 
 VKComputePass::~VKComputePass() = default;
 
-vk::DescriptorSet VKComputePass::CommitDescriptorSet(
-    VKUpdateDescriptorQueue& update_descriptor_queue, VKFence& fence) {
+VkDescriptorSet VKComputePass::CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue,
+                                                   VKFence& fence) {
     if (!descriptor_template) {
-        return {};
+        return nullptr;
     }
     const auto set = descriptor_allocator->Commit(fence);
     update_descriptor_queue.Send(*descriptor_template, set);
@@ -248,25 +458,21 @@ QuadArrayPass::QuadArrayPass(const VKDevice& device, VKScheduler& scheduler,
                              VKDescriptorPool& descriptor_pool,
                              VKStagingBufferPool& staging_buffer_pool,
                              VKUpdateDescriptorQueue& update_descriptor_queue)
-    : VKComputePass(device, descriptor_pool,
-                    {vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eStorageBuffer, 1,
-                                                    vk::ShaderStageFlagBits::eCompute, nullptr)},
-                    {vk::DescriptorUpdateTemplateEntry(0, 0, 1, vk::DescriptorType::eStorageBuffer,
-                                                       0, sizeof(DescriptorUpdateEntry))},
-                    {vk::PushConstantRange(vk::ShaderStageFlagBits::eCompute, 0, sizeof(u32))},
-                    std::size(quad_array), quad_array),
+    : VKComputePass(device, descriptor_pool, BuildQuadArrayPassDescriptorSetLayoutBinding(),
+                    BuildQuadArrayPassDescriptorUpdateTemplateEntry(),
+                    BuildComputePushConstantRange(sizeof(u32)), std::size(quad_array), quad_array),
       scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool},
       update_descriptor_queue{update_descriptor_queue} {}
 
 QuadArrayPass::~QuadArrayPass() = default;
 
-std::pair<const vk::Buffer&, vk::DeviceSize> QuadArrayPass::Assemble(u32 num_vertices, u32 first) {
-    const u32 num_triangle_vertices = num_vertices * 6 / 4;
+std::pair<VkBuffer, VkDeviceSize> QuadArrayPass::Assemble(u32 num_vertices, u32 first) {
+    const u32 num_triangle_vertices = (num_vertices / 4) * 6;
     const std::size_t staging_size = num_triangle_vertices * sizeof(u32);
     auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false);
 
     update_descriptor_queue.Acquire();
-    update_descriptor_queue.AddBuffer(&*buffer.handle, 0, staging_size);
+    update_descriptor_queue.AddBuffer(*buffer.handle, 0, staging_size);
     const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence());
 
     scheduler.RequestOutsideRenderPassOperationContext();
@@ -274,20 +480,25 @@ std::pair<const vk::Buffer&, vk::DeviceSize> QuadArrayPass::Assemble(u32 num_ver
     ASSERT(num_vertices % 4 == 0);
     const u32 num_quads = num_vertices / 4;
     scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, num_quads,
-                      first, set](auto cmdbuf, auto& dld) {
+                      first, set](vk::CommandBuffer cmdbuf) {
         constexpr u32 dispatch_size = 1024;
-        cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline, dld);
-        cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, 0, {set}, {}, dld);
-        cmdbuf.pushConstants(layout, vk::ShaderStageFlagBits::eCompute, 0, sizeof(first), &first,
-                             dld);
-        cmdbuf.dispatch(Common::AlignUp(num_quads, dispatch_size) / dispatch_size, 1, 1, dld);
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline);
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, 0, set, {});
+        cmdbuf.PushConstants(layout, VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(first), &first);
+        cmdbuf.Dispatch(Common::AlignUp(num_quads, dispatch_size) / dispatch_size, 1, 1);
 
-        const vk::BufferMemoryBarrier barrier(
-            vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eVertexAttributeRead,
-            VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, 0,
-            static_cast<vk::DeviceSize>(num_quads) * 6 * sizeof(u32));
-        cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader,
-                               vk::PipelineStageFlagBits::eVertexInput, {}, {}, {barrier}, {}, dld);
+        VkBufferMemoryBarrier barrier;
+        barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
+        barrier.pNext = nullptr;
+        barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
+        barrier.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT;
+        barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.buffer = buffer;
+        barrier.offset = 0;
+        barrier.size = static_cast<VkDeviceSize>(num_quads) * 6 * sizeof(u32);
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
+                               VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, {}, {barrier}, {});
     });
     return {*buffer.handle, 0};
 }
@@ -295,45 +506,112 @@ std::pair<const vk::Buffer&, vk::DeviceSize> QuadArrayPass::Assemble(u32 num_ver
 Uint8Pass::Uint8Pass(const VKDevice& device, VKScheduler& scheduler,
                      VKDescriptorPool& descriptor_pool, VKStagingBufferPool& staging_buffer_pool,
                      VKUpdateDescriptorQueue& update_descriptor_queue)
-    : VKComputePass(device, descriptor_pool,
-                    {vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eStorageBuffer, 1,
-                                                    vk::ShaderStageFlagBits::eCompute, nullptr),
-                     vk::DescriptorSetLayoutBinding(1, vk::DescriptorType::eStorageBuffer, 1,
-                                                    vk::ShaderStageFlagBits::eCompute, nullptr)},
-                    {vk::DescriptorUpdateTemplateEntry(0, 0, 2, vk::DescriptorType::eStorageBuffer,
-                                                       0, sizeof(DescriptorUpdateEntry))},
-                    {}, std::size(uint8_pass), uint8_pass),
+    : VKComputePass(device, descriptor_pool, BuildInputOutputDescriptorSetBindings(),
+                    BuildInputOutputDescriptorUpdateTemplate(), {}, std::size(uint8_pass),
+                    uint8_pass),
       scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool},
       update_descriptor_queue{update_descriptor_queue} {}
 
 Uint8Pass::~Uint8Pass() = default;
 
-std::pair<const vk::Buffer*, u64> Uint8Pass::Assemble(u32 num_vertices, vk::Buffer src_buffer,
-                                                      u64 src_offset) {
+std::pair<VkBuffer, u64> Uint8Pass::Assemble(u32 num_vertices, VkBuffer src_buffer,
+                                             u64 src_offset) {
     const auto staging_size = static_cast<u32>(num_vertices * sizeof(u16));
     auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false);
 
     update_descriptor_queue.Acquire();
-    update_descriptor_queue.AddBuffer(&src_buffer, src_offset, num_vertices);
-    update_descriptor_queue.AddBuffer(&*buffer.handle, 0, staging_size);
+    update_descriptor_queue.AddBuffer(src_buffer, src_offset, num_vertices);
+    update_descriptor_queue.AddBuffer(*buffer.handle, 0, staging_size);
     const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence());
 
     scheduler.RequestOutsideRenderPassOperationContext();
     scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, set,
-                      num_vertices](auto cmdbuf, auto& dld) {
+                      num_vertices](vk::CommandBuffer cmdbuf) {
         constexpr u32 dispatch_size = 1024;
-        cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline, dld);
-        cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, 0, {set}, {}, dld);
-        cmdbuf.dispatch(Common::AlignUp(num_vertices, dispatch_size) / dispatch_size, 1, 1, dld);
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline);
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, 0, set, {});
+        cmdbuf.Dispatch(Common::AlignUp(num_vertices, dispatch_size) / dispatch_size, 1, 1);
 
-        const vk::BufferMemoryBarrier barrier(
-            vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eVertexAttributeRead,
-            VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, 0,
-            static_cast<vk::DeviceSize>(num_vertices) * sizeof(u16));
-        cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader,
-                               vk::PipelineStageFlagBits::eVertexInput, {}, {}, {barrier}, {}, dld);
+        VkBufferMemoryBarrier barrier;
+        barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
+        barrier.pNext = nullptr;
+        barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
+        barrier.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT;
+        barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.buffer = buffer;
+        barrier.offset = 0;
+        barrier.size = static_cast<VkDeviceSize>(num_vertices * sizeof(u16));
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
+                               VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, {}, barrier, {});
     });
-    return {&*buffer.handle, 0};
+    return {*buffer.handle, 0};
+}
+
+QuadIndexedPass::QuadIndexedPass(const VKDevice& device, VKScheduler& scheduler,
+                                 VKDescriptorPool& descriptor_pool,
+                                 VKStagingBufferPool& staging_buffer_pool,
+                                 VKUpdateDescriptorQueue& update_descriptor_queue)
+    : VKComputePass(device, descriptor_pool, BuildInputOutputDescriptorSetBindings(),
+                    BuildInputOutputDescriptorUpdateTemplate(),
+                    BuildComputePushConstantRange(sizeof(u32) * 2), std::size(QUAD_INDEXED_SPV),
+                    QUAD_INDEXED_SPV),
+      scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool},
+      update_descriptor_queue{update_descriptor_queue} {}
+
+QuadIndexedPass::~QuadIndexedPass() = default;
+
+std::pair<VkBuffer, u64> QuadIndexedPass::Assemble(
+    Tegra::Engines::Maxwell3D::Regs::IndexFormat index_format, u32 num_vertices, u32 base_vertex,
+    VkBuffer src_buffer, u64 src_offset) {
+    const u32 index_shift = [index_format] {
+        switch (index_format) {
+        case Tegra::Engines::Maxwell3D::Regs::IndexFormat::UnsignedByte:
+            return 0;
+        case Tegra::Engines::Maxwell3D::Regs::IndexFormat::UnsignedShort:
+            return 1;
+        case Tegra::Engines::Maxwell3D::Regs::IndexFormat::UnsignedInt:
+            return 2;
+        }
+        UNREACHABLE();
+        return 2;
+    }();
+    const u32 input_size = num_vertices << index_shift;
+    const u32 num_tri_vertices = (num_vertices / 4) * 6;
+
+    const std::size_t staging_size = num_tri_vertices * sizeof(u32);
+    auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false);
+
+    update_descriptor_queue.Acquire();
+    update_descriptor_queue.AddBuffer(src_buffer, src_offset, input_size);
+    update_descriptor_queue.AddBuffer(*buffer.handle, 0, staging_size);
+    const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence());
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, set,
+                      num_tri_vertices, base_vertex, index_shift](vk::CommandBuffer cmdbuf) {
+        static constexpr u32 dispatch_size = 1024;
+        const std::array push_constants = {base_vertex, index_shift};
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline);
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, 0, set, {});
+        cmdbuf.PushConstants(layout, VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(push_constants),
+                             &push_constants);
+        cmdbuf.Dispatch(Common::AlignUp(num_tri_vertices, dispatch_size) / dispatch_size, 1, 1);
+
+        VkBufferMemoryBarrier barrier;
+        barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
+        barrier.pNext = nullptr;
+        barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
+        barrier.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT;
+        barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+        barrier.buffer = buffer;
+        barrier.offset = 0;
+        barrier.size = static_cast<VkDeviceSize>(num_tri_vertices * sizeof(u32));
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
+                               VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, {}, barrier, {});
+    });
+    return {*buffer.handle, 0};
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pass.h

@@ -8,8 +8,9 @@
 #include <utility>
 #include <vector>
 #include "common/common_types.h"
-#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
@@ -22,24 +23,24 @@ class VKUpdateDescriptorQueue;
 class VKComputePass {
 public:
     explicit VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool,
-                           const std::vector<vk::DescriptorSetLayoutBinding>& bindings,
-                           const std::vector<vk::DescriptorUpdateTemplateEntry>& templates,
-                           const std::vector<vk::PushConstantRange> push_constants,
-                           std::size_t code_size, const u8* code);
+                           vk::Span<VkDescriptorSetLayoutBinding> bindings,
+                           vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates,
+                           vk::Span<VkPushConstantRange> push_constants, std::size_t code_size,
+                           const u8* code);
     ~VKComputePass();
 
 protected:
-    vk::DescriptorSet CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue,
-                                          VKFence& fence);
+    VkDescriptorSet CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue,
+                                        VKFence& fence);
 
-    UniqueDescriptorUpdateTemplate descriptor_template;
-    UniquePipelineLayout layout;
-    UniquePipeline pipeline;
+    vk::DescriptorUpdateTemplateKHR descriptor_template;
+    vk::PipelineLayout layout;
+    vk::Pipeline pipeline;
 
 private:
-    UniqueDescriptorSetLayout descriptor_set_layout;
+    vk::DescriptorSetLayout descriptor_set_layout;
     std::optional<DescriptorAllocator> descriptor_allocator;
-    UniqueShaderModule module;
+    vk::ShaderModule module;
 };
 
 class QuadArrayPass final : public VKComputePass {
@@ -50,7 +51,7 @@ public:
                            VKUpdateDescriptorQueue& update_descriptor_queue);
     ~QuadArrayPass();
 
-    std::pair<const vk::Buffer&, vk::DeviceSize> Assemble(u32 num_vertices, u32 first);
+    std::pair<VkBuffer, VkDeviceSize> Assemble(u32 num_vertices, u32 first);
 
 private:
     VKScheduler& scheduler;
@@ -65,8 +66,25 @@ public:
                        VKUpdateDescriptorQueue& update_descriptor_queue);
     ~Uint8Pass();
 
-    std::pair<const vk::Buffer*, u64> Assemble(u32 num_vertices, vk::Buffer src_buffer,
-                                               u64 src_offset);
+    std::pair<VkBuffer, u64> Assemble(u32 num_vertices, VkBuffer src_buffer, u64 src_offset);
+
+private:
+    VKScheduler& scheduler;
+    VKStagingBufferPool& staging_buffer_pool;
+    VKUpdateDescriptorQueue& update_descriptor_queue;
+};
+
+class QuadIndexedPass final : public VKComputePass {
+public:
+    explicit QuadIndexedPass(const VKDevice& device, VKScheduler& scheduler,
+                             VKDescriptorPool& descriptor_pool,
+                             VKStagingBufferPool& staging_buffer_pool,
+                             VKUpdateDescriptorQueue& update_descriptor_queue);
+    ~QuadIndexedPass();
+
+    std::pair<VkBuffer, u64> Assemble(Tegra::Engines::Maxwell3D::Regs::IndexFormat index_format,
+                                      u32 num_vertices, u32 base_vertex, VkBuffer src_buffer,
+                                      u64 src_offset);
 
 private:
     VKScheduler& scheduler;

src/video_core/renderer_vulkan/vk_compute_pipeline.cpp

@@ -5,7 +5,6 @@
 #include <memory>
 #include <vector>
 
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_compute_pipeline.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
 #include "video_core/renderer_vulkan/vk_device.h"
@@ -14,6 +13,7 @@
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_shader_decompiler.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
@@ -30,7 +30,7 @@ VKComputePipeline::VKComputePipeline(const VKDevice& device, VKScheduler& schedu
 
 VKComputePipeline::~VKComputePipeline() = default;
 
-vk::DescriptorSet VKComputePipeline::CommitDescriptorSet() {
+VkDescriptorSet VKComputePipeline::CommitDescriptorSet() {
     if (!descriptor_template) {
         return {};
     }
@@ -39,74 +39,109 @@ vk::DescriptorSet VKComputePipeline::CommitDescriptorSet() {
     return set;
 }
 
-UniqueDescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const {
-    std::vector<vk::DescriptorSetLayoutBinding> bindings;
+vk::DescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const {
+    std::vector<VkDescriptorSetLayoutBinding> bindings;
     u32 binding = 0;
-    const auto AddBindings = [&](vk::DescriptorType descriptor_type, std::size_t num_entries) {
+    const auto add_bindings = [&](VkDescriptorType descriptor_type, std::size_t num_entries) {
         // TODO(Rodrigo): Maybe make individual bindings here?
         for (u32 bindpoint = 0; bindpoint < static_cast<u32>(num_entries); ++bindpoint) {
-            bindings.emplace_back(binding++, descriptor_type, 1, vk::ShaderStageFlagBits::eCompute,
-                                  nullptr);
+            VkDescriptorSetLayoutBinding& entry = bindings.emplace_back();
+            entry.binding = binding++;
+            entry.descriptorType = descriptor_type;
+            entry.descriptorCount = 1;
+            entry.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
+            entry.pImmutableSamplers = nullptr;
         }
     };
-    AddBindings(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size());
-    AddBindings(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size());
-    AddBindings(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size());
-    AddBindings(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size());
-    AddBindings(vk::DescriptorType::eStorageImage, entries.images.size());
+    add_bindings(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, entries.const_buffers.size());
+    add_bindings(VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, entries.global_buffers.size());
+    add_bindings(VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER, entries.texel_buffers.size());
+    add_bindings(VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, entries.samplers.size());
+    add_bindings(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, entries.images.size());
 
-    const vk::DescriptorSetLayoutCreateInfo descriptor_set_layout_ci(
-        {}, static_cast<u32>(bindings.size()), bindings.data());
-
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-    return dev.createDescriptorSetLayoutUnique(descriptor_set_layout_ci, nullptr, dld);
+    VkDescriptorSetLayoutCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.bindingCount = static_cast<u32>(bindings.size());
+    ci.pBindings = bindings.data();
+    return device.GetLogical().CreateDescriptorSetLayout(ci);
 }
 
-UniquePipelineLayout VKComputePipeline::CreatePipelineLayout() const {
-    const vk::PipelineLayoutCreateInfo layout_ci({}, 1, &*descriptor_set_layout, 0, nullptr);
-    const auto dev = device.GetLogical();
-    return dev.createPipelineLayoutUnique(layout_ci, nullptr, device.GetDispatchLoader());
+vk::PipelineLayout VKComputePipeline::CreatePipelineLayout() const {
+    VkPipelineLayoutCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.setLayoutCount = 1;
+    ci.pSetLayouts = descriptor_set_layout.address();
+    ci.pushConstantRangeCount = 0;
+    ci.pPushConstantRanges = nullptr;
+    return device.GetLogical().CreatePipelineLayout(ci);
 }
 
-UniqueDescriptorUpdateTemplate VKComputePipeline::CreateDescriptorUpdateTemplate() const {
-    std::vector<vk::DescriptorUpdateTemplateEntry> template_entries;
+vk::DescriptorUpdateTemplateKHR VKComputePipeline::CreateDescriptorUpdateTemplate() const {
+    std::vector<VkDescriptorUpdateTemplateEntryKHR> template_entries;
     u32 binding = 0;
     u32 offset = 0;
     FillDescriptorUpdateTemplateEntries(entries, binding, offset, template_entries);
     if (template_entries.empty()) {
         // If the shader doesn't use descriptor sets, skip template creation.
-        return UniqueDescriptorUpdateTemplate{};
+        return {};
     }
 
-    const vk::DescriptorUpdateTemplateCreateInfo template_ci(
-        {}, static_cast<u32>(template_entries.size()), template_entries.data(),
-        vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout,
-        vk::PipelineBindPoint::eGraphics, *layout, DESCRIPTOR_SET);
-
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-    return dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld);
+    VkDescriptorUpdateTemplateCreateInfoKHR ci;
+    ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.descriptorUpdateEntryCount = static_cast<u32>(template_entries.size());
+    ci.pDescriptorUpdateEntries = template_entries.data();
+    ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR;
+    ci.descriptorSetLayout = *descriptor_set_layout;
+    ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
+    ci.pipelineLayout = *layout;
+    ci.set = DESCRIPTOR_SET;
+    return device.GetLogical().CreateDescriptorUpdateTemplateKHR(ci);
 }
 
-UniqueShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const {
-    const vk::ShaderModuleCreateInfo module_ci({}, code.size() * sizeof(u32), code.data());
-    const auto dev = device.GetLogical();
-    return dev.createShaderModuleUnique(module_ci, nullptr, device.GetDispatchLoader());
+vk::ShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const {
+    VkShaderModuleCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.codeSize = code.size() * sizeof(u32);
+    ci.pCode = code.data();
+    return device.GetLogical().CreateShaderModule(ci);
 }
 
-UniquePipeline VKComputePipeline::CreatePipeline() const {
-    vk::PipelineShaderStageCreateInfo shader_stage_ci({}, vk::ShaderStageFlagBits::eCompute,
-                                                      *shader_module, "main", nullptr);
-    vk::PipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci;
+vk::Pipeline VKComputePipeline::CreatePipeline() const {
+    VkComputePipelineCreateInfo ci;
+    VkPipelineShaderStageCreateInfo& stage_ci = ci.stage;
+    stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
+    stage_ci.pNext = nullptr;
+    stage_ci.flags = 0;
+    stage_ci.stage = VK_SHADER_STAGE_COMPUTE_BIT;
+    stage_ci.module = *shader_module;
+    stage_ci.pName = "main";
+    stage_ci.pSpecializationInfo = nullptr;
+
+    VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci;
+    subgroup_size_ci.sType =
+        VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_REQUIRED_SUBGROUP_SIZE_CREATE_INFO_EXT;
+    subgroup_size_ci.pNext = nullptr;
     subgroup_size_ci.requiredSubgroupSize = GuestWarpSize;
-    if (entries.uses_warps && device.IsGuestWarpSizeSupported(vk::ShaderStageFlagBits::eCompute)) {
-        shader_stage_ci.pNext = &subgroup_size_ci;
+
+    if (entries.uses_warps && device.IsGuestWarpSizeSupported(VK_SHADER_STAGE_COMPUTE_BIT)) {
+        stage_ci.pNext = &subgroup_size_ci;
     }
 
-    const vk::ComputePipelineCreateInfo create_info({}, shader_stage_ci, *layout, {}, 0);
-    const auto dev = device.GetLogical();
-    return dev.createComputePipelineUnique({}, create_info, nullptr, device.GetDispatchLoader());
+    ci.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.layout = *layout;
+    ci.basePipelineHandle = nullptr;
+    ci.basePipelineIndex = 0;
+    return device.GetLogical().CreateComputePipeline(ci);
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pipeline.h

@@ -7,9 +7,9 @@
 #include <memory>
 
 #include "common/common_types.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
 #include "video_core/renderer_vulkan/vk_shader_decompiler.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
@@ -25,42 +25,42 @@ public:
                                const SPIRVShader& shader);
     ~VKComputePipeline();
 
-    vk::DescriptorSet CommitDescriptorSet();
+    VkDescriptorSet CommitDescriptorSet();
 
-    vk::Pipeline GetHandle() const {
+    VkPipeline GetHandle() const {
         return *pipeline;
     }
 
-    vk::PipelineLayout GetLayout() const {
+    VkPipelineLayout GetLayout() const {
         return *layout;
     }
 
-    const ShaderEntries& GetEntries() {
+    const ShaderEntries& GetEntries() const {
         return entries;
     }
 
 private:
-    UniqueDescriptorSetLayout CreateDescriptorSetLayout() const;
+    vk::DescriptorSetLayout CreateDescriptorSetLayout() const;
 
-    UniquePipelineLayout CreatePipelineLayout() const;
+    vk::PipelineLayout CreatePipelineLayout() const;
 
-    UniqueDescriptorUpdateTemplate CreateDescriptorUpdateTemplate() const;
+    vk::DescriptorUpdateTemplateKHR CreateDescriptorUpdateTemplate() const;
 
-    UniqueShaderModule CreateShaderModule(const std::vector<u32>& code) const;
+    vk::ShaderModule CreateShaderModule(const std::vector<u32>& code) const;
 
-    UniquePipeline CreatePipeline() const;
+    vk::Pipeline CreatePipeline() const;
 
     const VKDevice& device;
     VKScheduler& scheduler;
     ShaderEntries entries;
 
-    UniqueDescriptorSetLayout descriptor_set_layout;
+    vk::DescriptorSetLayout descriptor_set_layout;
     DescriptorAllocator descriptor_allocator;
     VKUpdateDescriptorQueue& update_descriptor_queue;
-    UniquePipelineLayout layout;
-    UniqueDescriptorUpdateTemplate descriptor_template;
-    UniqueShaderModule shader_module;
-    UniquePipeline pipeline;
+    vk::PipelineLayout layout;
+    vk::DescriptorUpdateTemplateKHR descriptor_template;
+    vk::ShaderModule shader_module;
+    vk::Pipeline pipeline;
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_descriptor_pool.cpp

@@ -6,10 +6,10 @@
 #include <vector>
 
 #include "common/common_types.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
 #include "video_core/renderer_vulkan/vk_device.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
@@ -17,19 +17,18 @@ namespace Vulkan {
 constexpr std::size_t SETS_GROW_RATE = 0x20;
 
 DescriptorAllocator::DescriptorAllocator(VKDescriptorPool& descriptor_pool,
-                                         vk::DescriptorSetLayout layout)
+                                         VkDescriptorSetLayout layout)
     : VKFencedPool{SETS_GROW_RATE}, descriptor_pool{descriptor_pool}, layout{layout} {}
 
 DescriptorAllocator::~DescriptorAllocator() = default;
 
-vk::DescriptorSet DescriptorAllocator::Commit(VKFence& fence) {
-    return *descriptors[CommitResource(fence)];
+VkDescriptorSet DescriptorAllocator::Commit(VKFence& fence) {
+    const std::size_t index = CommitResource(fence);
+    return descriptors_allocations[index / SETS_GROW_RATE][index % SETS_GROW_RATE];
 }
 
 void DescriptorAllocator::Allocate(std::size_t begin, std::size_t end) {
-    auto new_sets = descriptor_pool.AllocateDescriptors(layout, end - begin);
-    descriptors.insert(descriptors.end(), std::make_move_iterator(new_sets.begin()),
-                       std::make_move_iterator(new_sets.end()));
+    descriptors_allocations.push_back(descriptor_pool.AllocateDescriptors(layout, end - begin));
 }
 
 VKDescriptorPool::VKDescriptorPool(const VKDevice& device)
@@ -37,53 +36,50 @@ VKDescriptorPool::VKDescriptorPool(const VKDevice& device)
 
 VKDescriptorPool::~VKDescriptorPool() = default;
 
-vk::DescriptorPool VKDescriptorPool::AllocateNewPool() {
+vk::DescriptorPool* VKDescriptorPool::AllocateNewPool() {
     static constexpr u32 num_sets = 0x20000;
-    static constexpr vk::DescriptorPoolSize pool_sizes[] = {
-        {vk::DescriptorType::eUniformBuffer, num_sets * 90},
-        {vk::DescriptorType::eStorageBuffer, num_sets * 60},
-        {vk::DescriptorType::eUniformTexelBuffer, num_sets * 64},
-        {vk::DescriptorType::eCombinedImageSampler, num_sets * 64},
-        {vk::DescriptorType::eStorageImage, num_sets * 40}};
+    static constexpr VkDescriptorPoolSize pool_sizes[] = {
+        {VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, num_sets * 90},
+        {VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, num_sets * 60},
+        {VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER, num_sets * 64},
+        {VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, num_sets * 64},
+        {VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, num_sets * 40}};
 
-    const vk::DescriptorPoolCreateInfo create_info(
-        vk::DescriptorPoolCreateFlagBits::eFreeDescriptorSet, num_sets,
-        static_cast<u32>(std::size(pool_sizes)), std::data(pool_sizes));
-    const auto dev = device.GetLogical();
-    return *pools.emplace_back(
-        dev.createDescriptorPoolUnique(create_info, nullptr, device.GetDispatchLoader()));
+    VkDescriptorPoolCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT;
+    ci.maxSets = num_sets;
+    ci.poolSizeCount = static_cast<u32>(std::size(pool_sizes));
+    ci.pPoolSizes = std::data(pool_sizes);
+    return &pools.emplace_back(device.GetLogical().CreateDescriptorPool(ci));
 }
 
-std::vector<UniqueDescriptorSet> VKDescriptorPool::AllocateDescriptors(
-    vk::DescriptorSetLayout layout, std::size_t count) {
-    std::vector layout_copies(count, layout);
-    vk::DescriptorSetAllocateInfo allocate_info(active_pool, static_cast<u32>(count),
-                                                layout_copies.data());
+vk::DescriptorSets VKDescriptorPool::AllocateDescriptors(VkDescriptorSetLayout layout,
+                                                         std::size_t count) {
+    const std::vector layout_copies(count, layout);
+    VkDescriptorSetAllocateInfo ai;
+    ai.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO;
+    ai.pNext = nullptr;
+    ai.descriptorPool = **active_pool;
+    ai.descriptorSetCount = static_cast<u32>(count);
+    ai.pSetLayouts = layout_copies.data();
 
-    std::vector<vk::DescriptorSet> sets(count);
-    const auto dev = device.GetLogical();
-    const auto& dld = device.GetDispatchLoader();
-    switch (const auto result = dev.allocateDescriptorSets(&allocate_info, sets.data(), dld)) {
-    case vk::Result::eSuccess:
-        break;
-    case vk::Result::eErrorOutOfPoolMemory:
-        active_pool = AllocateNewPool();
-        allocate_info.descriptorPool = active_pool;
-        if (dev.allocateDescriptorSets(&allocate_info, sets.data(), dld) == vk::Result::eSuccess) {
-            break;
-        }
-        [[fallthrough]];
-    default:
-        vk::throwResultException(result, "vk::Device::allocateDescriptorSetsUnique");
+    vk::DescriptorSets sets = active_pool->Allocate(ai);
+    if (!sets.IsOutOfPoolMemory()) {
+        return sets;
     }
 
-    vk::PoolFree deleter(dev, active_pool, dld);
-    std::vector<UniqueDescriptorSet> unique_sets;
-    unique_sets.reserve(count);
-    for (const auto set : sets) {
-        unique_sets.push_back(UniqueDescriptorSet{set, deleter});
+    // Our current pool is out of memory. Allocate a new one and retry
+    active_pool = AllocateNewPool();
+    ai.descriptorPool = **active_pool;
+    sets = active_pool->Allocate(ai);
+    if (!sets.IsOutOfPoolMemory()) {
+        return sets;
     }
-    return unique_sets;
+
+    // After allocating a new pool, we are out of memory again. We can't handle this from here.
+    throw vk::Exception(VK_ERROR_OUT_OF_POOL_MEMORY);
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_descriptor_pool.h

@@ -8,8 +8,8 @@
 #include <vector>
 
 #include "common/common_types.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
@@ -17,21 +17,21 @@ class VKDescriptorPool;
 
 class DescriptorAllocator final : public VKFencedPool {
 public:
-    explicit DescriptorAllocator(VKDescriptorPool& descriptor_pool, vk::DescriptorSetLayout layout);
+    explicit DescriptorAllocator(VKDescriptorPool& descriptor_pool, VkDescriptorSetLayout layout);
     ~DescriptorAllocator() override;
 
     DescriptorAllocator(const DescriptorAllocator&) = delete;
 
-    vk::DescriptorSet Commit(VKFence& fence);
+    VkDescriptorSet Commit(VKFence& fence);
 
 protected:
     void Allocate(std::size_t begin, std::size_t end) override;
 
 private:
     VKDescriptorPool& descriptor_pool;
-    const vk::DescriptorSetLayout layout;
+    const VkDescriptorSetLayout layout;
 
-    std::vector<UniqueDescriptorSet> descriptors;
+    std::vector<vk::DescriptorSets> descriptors_allocations;
 };
 
 class VKDescriptorPool final {
@@ -42,15 +42,14 @@ public:
     ~VKDescriptorPool();
 
 private:
-    vk::DescriptorPool AllocateNewPool();
+    vk::DescriptorPool* AllocateNewPool();
 
-    std::vector<UniqueDescriptorSet> AllocateDescriptors(vk::DescriptorSetLayout layout,
-                                                         std::size_t count);
+    vk::DescriptorSets AllocateDescriptors(VkDescriptorSetLayout layout, std::size_t count);
 
     const VKDevice& device;
 
-    std::vector<UniqueDescriptorPool> pools;
-    vk::DescriptorPool active_pool;
+    std::vector<vk::DescriptorPool> pools;
+    vk::DescriptorPool* active_pool;
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_device.cpp

@@ -6,14 +6,15 @@
 #include <chrono>
 #include <cstdlib>
 #include <optional>
-#include <set>
 #include <string_view>
 #include <thread>
+#include <unordered_set>
 #include <vector>
+
 #include "common/assert.h"
 #include "core/settings.h"
-#include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
@@ -21,49 +22,43 @@ namespace {
 
 namespace Alternatives {
 
-constexpr std::array Depth24UnormS8Uint = {vk::Format::eD32SfloatS8Uint,
-                                           vk::Format::eD16UnormS8Uint, vk::Format{}};
-constexpr std::array Depth16UnormS8Uint = {vk::Format::eD24UnormS8Uint,
-                                           vk::Format::eD32SfloatS8Uint, vk::Format{}};
+constexpr std::array Depth24UnormS8_UINT = {VK_FORMAT_D32_SFLOAT_S8_UINT,
+                                            VK_FORMAT_D16_UNORM_S8_UINT, VkFormat{}};
+constexpr std::array Depth16UnormS8_UINT = {VK_FORMAT_D24_UNORM_S8_UINT,
+                                            VK_FORMAT_D32_SFLOAT_S8_UINT, VkFormat{}};
 
 } // namespace Alternatives
 
+constexpr std::array REQUIRED_EXTENSIONS = {
+    VK_KHR_SWAPCHAIN_EXTENSION_NAME,
+    VK_KHR_16BIT_STORAGE_EXTENSION_NAME,
+    VK_KHR_8BIT_STORAGE_EXTENSION_NAME,
+    VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME,
+    VK_KHR_DESCRIPTOR_UPDATE_TEMPLATE_EXTENSION_NAME,
+    VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME,
+    VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME,
+    VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME,
+    VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME,
+};
+
 template <typename T>
 void SetNext(void**& next, T& data) {
     *next = &data;
     next = &data.pNext;
 }
 
-template <typename T>
-T GetFeatures(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dldi) {
-    vk::PhysicalDeviceFeatures2 features;
-    T extension_features;
-    features.pNext = &extension_features;
-    physical.getFeatures2(&features, dldi);
-    return extension_features;
-}
-
-template <typename T>
-T GetProperties(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dldi) {
-    vk::PhysicalDeviceProperties2 properties;
-    T extension_properties;
-    properties.pNext = &extension_properties;
-    physical.getProperties2(&properties, dldi);
-    return extension_properties;
-}
-
-constexpr const vk::Format* GetFormatAlternatives(vk::Format format) {
+constexpr const VkFormat* GetFormatAlternatives(VkFormat format) {
     switch (format) {
-    case vk::Format::eD24UnormS8Uint:
-        return Alternatives::Depth24UnormS8Uint.data();
-    case vk::Format::eD16UnormS8Uint:
-        return Alternatives::Depth16UnormS8Uint.data();
+    case VK_FORMAT_D24_UNORM_S8_UINT:
+        return Alternatives::Depth24UnormS8_UINT.data();
+    case VK_FORMAT_D16_UNORM_S8_UINT:
+        return Alternatives::Depth16UnormS8_UINT.data();
     default:
         return nullptr;
     }
 }
 
-vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties, FormatType format_type) {
+VkFormatFeatureFlags GetFormatFeatures(VkFormatProperties properties, FormatType format_type) {
     switch (format_type) {
     case FormatType::Linear:
         return properties.linearTilingFeatures;
@@ -76,79 +71,220 @@ vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties, Format
     }
 }
 
+std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(
+    vk::PhysicalDevice physical, const vk::InstanceDispatch& dld) {
+    static constexpr std::array formats{VK_FORMAT_A8B8G8R8_UNORM_PACK32,
+                                        VK_FORMAT_A8B8G8R8_UINT_PACK32,
+                                        VK_FORMAT_A8B8G8R8_SNORM_PACK32,
+                                        VK_FORMAT_A8B8G8R8_SRGB_PACK32,
+                                        VK_FORMAT_B5G6R5_UNORM_PACK16,
+                                        VK_FORMAT_A2B10G10R10_UNORM_PACK32,
+                                        VK_FORMAT_A1R5G5B5_UNORM_PACK16,
+                                        VK_FORMAT_R32G32B32A32_SFLOAT,
+                                        VK_FORMAT_R32G32B32A32_UINT,
+                                        VK_FORMAT_R32G32_SFLOAT,
+                                        VK_FORMAT_R32G32_UINT,
+                                        VK_FORMAT_R16G16B16A16_UINT,
+                                        VK_FORMAT_R16G16B16A16_SNORM,
+                                        VK_FORMAT_R16G16B16A16_UNORM,
+                                        VK_FORMAT_R16G16_UNORM,
+                                        VK_FORMAT_R16G16_SNORM,
+                                        VK_FORMAT_R16G16_SFLOAT,
+                                        VK_FORMAT_R16_UNORM,
+                                        VK_FORMAT_R8G8B8A8_SRGB,
+                                        VK_FORMAT_R8G8_UNORM,
+                                        VK_FORMAT_R8G8_SNORM,
+                                        VK_FORMAT_R8_UNORM,
+                                        VK_FORMAT_R8_UINT,
+                                        VK_FORMAT_B10G11R11_UFLOAT_PACK32,
+                                        VK_FORMAT_R32_SFLOAT,
+                                        VK_FORMAT_R32_UINT,
+                                        VK_FORMAT_R32_SINT,
+                                        VK_FORMAT_R16_SFLOAT,
+                                        VK_FORMAT_R16G16B16A16_SFLOAT,
+                                        VK_FORMAT_B8G8R8A8_UNORM,
+                                        VK_FORMAT_R4G4B4A4_UNORM_PACK16,
+                                        VK_FORMAT_D32_SFLOAT,
+                                        VK_FORMAT_D16_UNORM,
+                                        VK_FORMAT_D16_UNORM_S8_UINT,
+                                        VK_FORMAT_D24_UNORM_S8_UINT,
+                                        VK_FORMAT_D32_SFLOAT_S8_UINT,
+                                        VK_FORMAT_BC1_RGBA_UNORM_BLOCK,
+                                        VK_FORMAT_BC2_UNORM_BLOCK,
+                                        VK_FORMAT_BC3_UNORM_BLOCK,
+                                        VK_FORMAT_BC4_UNORM_BLOCK,
+                                        VK_FORMAT_BC5_UNORM_BLOCK,
+                                        VK_FORMAT_BC5_SNORM_BLOCK,
+                                        VK_FORMAT_BC7_UNORM_BLOCK,
+                                        VK_FORMAT_BC6H_UFLOAT_BLOCK,
+                                        VK_FORMAT_BC6H_SFLOAT_BLOCK,
+                                        VK_FORMAT_BC1_RGBA_SRGB_BLOCK,
+                                        VK_FORMAT_BC2_SRGB_BLOCK,
+                                        VK_FORMAT_BC3_SRGB_BLOCK,
+                                        VK_FORMAT_BC7_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_4x4_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_8x8_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_8x5_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_5x4_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_5x5_UNORM_BLOCK,
+                                        VK_FORMAT_ASTC_5x5_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_10x8_UNORM_BLOCK,
+                                        VK_FORMAT_ASTC_10x8_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_6x6_UNORM_BLOCK,
+                                        VK_FORMAT_ASTC_6x6_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_10x10_UNORM_BLOCK,
+                                        VK_FORMAT_ASTC_10x10_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_12x12_UNORM_BLOCK,
+                                        VK_FORMAT_ASTC_12x12_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_8x6_UNORM_BLOCK,
+                                        VK_FORMAT_ASTC_8x6_SRGB_BLOCK,
+                                        VK_FORMAT_ASTC_6x5_UNORM_BLOCK,
+                                        VK_FORMAT_ASTC_6x5_SRGB_BLOCK,
+                                        VK_FORMAT_E5B9G9R9_UFLOAT_PACK32};
+    std::unordered_map<VkFormat, VkFormatProperties> format_properties;
+    for (const auto format : formats) {
+        format_properties.emplace(format, physical.GetFormatProperties(format));
+    }
+    return format_properties;
+}
+
 } // Anonymous namespace
 
-VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
-                   vk::SurfaceKHR surface)
-    : physical{physical}, properties{physical.getProperties(dldi)},
-      format_properties{GetFormatProperties(dldi, physical)} {
-    SetupFamilies(dldi, surface);
-    SetupFeatures(dldi);
+VKDevice::VKDevice(VkInstance instance, vk::PhysicalDevice physical, VkSurfaceKHR surface,
+                   const vk::InstanceDispatch& dld)
+    : dld{dld}, physical{physical}, properties{physical.GetProperties()},
+      format_properties{GetFormatProperties(physical, dld)} {
+    SetupFamilies(surface);
+    SetupFeatures();
 }
 
 VKDevice::~VKDevice() = default;
 
-bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance) {
+bool VKDevice::Create() {
     const auto queue_cis = GetDeviceQueueCreateInfos();
-    const std::vector extensions = LoadExtensions(dldi);
+    const std::vector extensions = LoadExtensions();
 
-    vk::PhysicalDeviceFeatures2 features2;
+    VkPhysicalDeviceFeatures2 features2;
+    features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
+    features2.pNext = nullptr;
     void** next = &features2.pNext;
+
     auto& features = features2.features;
-    features.vertexPipelineStoresAndAtomics = true;
+    features.robustBufferAccess = false;
+    features.fullDrawIndexUint32 = false;
+    features.imageCubeArray = false;
     features.independentBlend = true;
-    features.depthClamp = true;
-    features.samplerAnisotropy = true;
-    features.largePoints = true;
-    features.multiViewport = true;
-    features.depthBiasClamp = true;
     features.geometryShader = true;
     features.tessellationShader = true;
+    features.sampleRateShading = false;
+    features.dualSrcBlend = false;
+    features.logicOp = false;
+    features.multiDrawIndirect = false;
+    features.drawIndirectFirstInstance = false;
+    features.depthClamp = true;
+    features.depthBiasClamp = true;
+    features.fillModeNonSolid = false;
+    features.depthBounds = false;
+    features.wideLines = false;
+    features.largePoints = true;
+    features.alphaToOne = false;
+    features.multiViewport = true;
+    features.samplerAnisotropy = true;
+    features.textureCompressionETC2 = false;
+    features.textureCompressionASTC_LDR = is_optimal_astc_supported;
+    features.textureCompressionBC = false;
     features.occlusionQueryPrecise = true;
+    features.pipelineStatisticsQuery = false;
+    features.vertexPipelineStoresAndAtomics = true;
     features.fragmentStoresAndAtomics = true;
+    features.shaderTessellationAndGeometryPointSize = false;
     features.shaderImageGatherExtended = true;
+    features.shaderStorageImageExtendedFormats = false;
+    features.shaderStorageImageMultisample = false;
     features.shaderStorageImageReadWithoutFormat = is_formatless_image_load_supported;
     features.shaderStorageImageWriteWithoutFormat = true;
-    features.textureCompressionASTC_LDR = is_optimal_astc_supported;
+    features.shaderUniformBufferArrayDynamicIndexing = false;
+    features.shaderSampledImageArrayDynamicIndexing = false;
+    features.shaderStorageBufferArrayDynamicIndexing = false;
+    features.shaderStorageImageArrayDynamicIndexing = false;
+    features.shaderClipDistance = false;
+    features.shaderCullDistance = false;
+    features.shaderFloat64 = false;
+    features.shaderInt64 = false;
+    features.shaderInt16 = false;
+    features.shaderResourceResidency = false;
+    features.shaderResourceMinLod = false;
+    features.sparseBinding = false;
+    features.sparseResidencyBuffer = false;
+    features.sparseResidencyImage2D = false;
+    features.sparseResidencyImage3D = false;
+    features.sparseResidency2Samples = false;
+    features.sparseResidency4Samples = false;
+    features.sparseResidency8Samples = false;
+    features.sparseResidency16Samples = false;
+    features.sparseResidencyAliased = false;
+    features.variableMultisampleRate = false;
+    features.inheritedQueries = false;
 
-    vk::PhysicalDevice16BitStorageFeaturesKHR bit16_storage;
+    VkPhysicalDevice16BitStorageFeaturesKHR bit16_storage;
+    bit16_storage.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES_KHR;
+    bit16_storage.pNext = nullptr;
+    bit16_storage.storageBuffer16BitAccess = false;
     bit16_storage.uniformAndStorageBuffer16BitAccess = true;
+    bit16_storage.storagePushConstant16 = false;
+    bit16_storage.storageInputOutput16 = false;
     SetNext(next, bit16_storage);
 
-    vk::PhysicalDevice8BitStorageFeaturesKHR bit8_storage;
+    VkPhysicalDevice8BitStorageFeaturesKHR bit8_storage;
+    bit8_storage.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_8BIT_STORAGE_FEATURES_KHR;
+    bit8_storage.pNext = nullptr;
+    bit8_storage.storageBuffer8BitAccess = false;
     bit8_storage.uniformAndStorageBuffer8BitAccess = true;
+    bit8_storage.storagePushConstant8 = false;
     SetNext(next, bit8_storage);
 
-    vk::PhysicalDeviceHostQueryResetFeaturesEXT host_query_reset;
+    VkPhysicalDeviceHostQueryResetFeaturesEXT host_query_reset;
+    host_query_reset.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_HOST_QUERY_RESET_FEATURES_EXT;
     host_query_reset.hostQueryReset = true;
     SetNext(next, host_query_reset);
 
-    vk::PhysicalDeviceFloat16Int8FeaturesKHR float16_int8;
+    VkPhysicalDeviceFloat16Int8FeaturesKHR float16_int8;
     if (is_float16_supported) {
+        float16_int8.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT16_INT8_FEATURES_KHR;
+        float16_int8.pNext = nullptr;
         float16_int8.shaderFloat16 = true;
+        float16_int8.shaderInt8 = false;
         SetNext(next, float16_int8);
     } else {
         LOG_INFO(Render_Vulkan, "Device doesn't support float16 natively");
     }
 
-    vk::PhysicalDeviceUniformBufferStandardLayoutFeaturesKHR std430_layout;
+    VkPhysicalDeviceUniformBufferStandardLayoutFeaturesKHR std430_layout;
     if (khr_uniform_buffer_standard_layout) {
+        std430_layout.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_UNIFORM_BUFFER_STANDARD_LAYOUT_FEATURES_KHR;
+        std430_layout.pNext = nullptr;
         std430_layout.uniformBufferStandardLayout = true;
         SetNext(next, std430_layout);
     } else {
         LOG_INFO(Render_Vulkan, "Device doesn't support packed UBOs");
     }
 
-    vk::PhysicalDeviceIndexTypeUint8FeaturesEXT index_type_uint8;
+    VkPhysicalDeviceIndexTypeUint8FeaturesEXT index_type_uint8;
     if (ext_index_type_uint8) {
+        index_type_uint8.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_INDEX_TYPE_UINT8_FEATURES_EXT;
+        index_type_uint8.pNext = nullptr;
         index_type_uint8.indexTypeUint8 = true;
         SetNext(next, index_type_uint8);
     } else {
         LOG_INFO(Render_Vulkan, "Device doesn't support uint8 indexes");
     }
 
-    vk::PhysicalDeviceTransformFeedbackFeaturesEXT transform_feedback;
+    VkPhysicalDeviceTransformFeedbackFeaturesEXT transform_feedback;
     if (ext_transform_feedback) {
+        transform_feedback.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT;
+        transform_feedback.pNext = nullptr;
         transform_feedback.transformFeedback = true;
         transform_feedback.geometryStreams = true;
         SetNext(next, transform_feedback);
@@ -160,62 +296,48 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan
         LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted");
     }
 
-    vk::DeviceCreateInfo device_ci({}, static_cast<u32>(queue_cis.size()), queue_cis.data(), 0,
-                                   nullptr, static_cast<u32>(extensions.size()), extensions.data(),
-                                   nullptr);
-    device_ci.pNext = &features2;
-
-    vk::Device dummy_logical;
-    if (physical.createDevice(&device_ci, nullptr, &dummy_logical, dldi) != vk::Result::eSuccess) {
-        LOG_CRITICAL(Render_Vulkan, "Logical device failed to be created!");
+    logical = vk::Device::Create(physical, queue_cis, extensions, features2, dld);
+    if (!logical) {
+        LOG_ERROR(Render_Vulkan, "Failed to create logical device");
         return false;
     }
 
-    dld.init(instance, dldi.vkGetInstanceProcAddr, dummy_logical, dldi.vkGetDeviceProcAddr);
-    logical = UniqueDevice(
-        dummy_logical, vk::ObjectDestroy<vk::NoParent, vk::DispatchLoaderDynamic>(nullptr, dld));
-
     CollectTelemetryParameters();
 
-    graphics_queue = logical->getQueue(graphics_family, 0, dld);
-    present_queue = logical->getQueue(present_family, 0, dld);
+    graphics_queue = logical.GetQueue(graphics_family);
+    present_queue = logical.GetQueue(present_family);
     return true;
 }
 
-vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format,
-                                        vk::FormatFeatureFlags wanted_usage,
-                                        FormatType format_type) const {
+VkFormat VKDevice::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage,
+                                      FormatType format_type) const {
     if (IsFormatSupported(wanted_format, wanted_usage, format_type)) {
         return wanted_format;
     }
     // The wanted format is not supported by hardware, search for alternatives
-    const vk::Format* alternatives = GetFormatAlternatives(wanted_format);
+    const VkFormat* alternatives = GetFormatAlternatives(wanted_format);
     if (alternatives == nullptr) {
         UNREACHABLE_MSG("Format={} with usage={} and type={} has no defined alternatives and host "
                         "hardware does not support it",
-                        vk::to_string(wanted_format), vk::to_string(wanted_usage),
-                        static_cast<u32>(format_type));
+                        wanted_format, wanted_usage, format_type);
         return wanted_format;
     }
 
     std::size_t i = 0;
-    for (vk::Format alternative = alternatives[0]; alternative != vk::Format{};
-         alternative = alternatives[++i]) {
+    for (VkFormat alternative = *alternatives; alternative; alternative = alternatives[++i]) {
         if (!IsFormatSupported(alternative, wanted_usage, format_type)) {
             continue;
         }
         LOG_WARNING(Render_Vulkan,
                     "Emulating format={} with alternative format={} with usage={} and type={}",
-                    static_cast<u32>(wanted_format), static_cast<u32>(alternative),
-                    static_cast<u32>(wanted_usage), static_cast<u32>(format_type));
+                    wanted_format, alternative, wanted_usage, format_type);
         return alternative;
     }
 
     // No alternatives found, panic
     UNREACHABLE_MSG("Format={} with usage={} and type={} is not supported by the host hardware and "
                     "doesn't support any of the alternatives",
-                    static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage),
-                    static_cast<u32>(format_type));
+                    wanted_format, wanted_usage, format_type);
     return wanted_format;
 }
 
@@ -229,35 +351,39 @@ void VKDevice::ReportLoss() const {
         return;
     }
 
-    [[maybe_unused]] const std::vector data = graphics_queue.getCheckpointDataNV(dld);
+    [[maybe_unused]] const std::vector data = graphics_queue.GetCheckpointDataNV(dld);
     // Catch here in debug builds (or with optimizations disabled) the last graphics pipeline to be
     // executed. It can be done on a debugger by evaluating the expression:
     // *(VKGraphicsPipeline*)data[0]
 }
 
-bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
-                                      const vk::DispatchLoaderDynamic& dldi) const {
+bool VKDevice::IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const {
     // Disable for now to avoid converting ASTC twice.
-    return false;
     static constexpr std::array astc_formats = {
-        vk::Format::eAstc4x4SrgbBlock,    vk::Format::eAstc8x8SrgbBlock,
-        vk::Format::eAstc8x5SrgbBlock,    vk::Format::eAstc5x4SrgbBlock,
-        vk::Format::eAstc5x5UnormBlock,   vk::Format::eAstc5x5SrgbBlock,
-        vk::Format::eAstc10x8UnormBlock,  vk::Format::eAstc10x8SrgbBlock,
-        vk::Format::eAstc6x6UnormBlock,   vk::Format::eAstc6x6SrgbBlock,
-        vk::Format::eAstc10x10UnormBlock, vk::Format::eAstc10x10SrgbBlock,
-        vk::Format::eAstc12x12UnormBlock, vk::Format::eAstc12x12SrgbBlock,
-        vk::Format::eAstc8x6UnormBlock,   vk::Format::eAstc8x6SrgbBlock,
-        vk::Format::eAstc6x5UnormBlock,   vk::Format::eAstc6x5SrgbBlock};
+        VK_FORMAT_ASTC_4x4_UNORM_BLOCK,   VK_FORMAT_ASTC_4x4_SRGB_BLOCK,
+        VK_FORMAT_ASTC_5x4_UNORM_BLOCK,   VK_FORMAT_ASTC_5x4_SRGB_BLOCK,
+        VK_FORMAT_ASTC_5x5_UNORM_BLOCK,   VK_FORMAT_ASTC_5x5_SRGB_BLOCK,
+        VK_FORMAT_ASTC_6x5_UNORM_BLOCK,   VK_FORMAT_ASTC_6x5_SRGB_BLOCK,
+        VK_FORMAT_ASTC_6x6_UNORM_BLOCK,   VK_FORMAT_ASTC_6x6_SRGB_BLOCK,
+        VK_FORMAT_ASTC_8x5_UNORM_BLOCK,   VK_FORMAT_ASTC_8x5_SRGB_BLOCK,
+        VK_FORMAT_ASTC_8x6_UNORM_BLOCK,   VK_FORMAT_ASTC_8x6_SRGB_BLOCK,
+        VK_FORMAT_ASTC_8x8_UNORM_BLOCK,   VK_FORMAT_ASTC_8x8_SRGB_BLOCK,
+        VK_FORMAT_ASTC_10x5_UNORM_BLOCK,  VK_FORMAT_ASTC_10x5_SRGB_BLOCK,
+        VK_FORMAT_ASTC_10x6_UNORM_BLOCK,  VK_FORMAT_ASTC_10x6_SRGB_BLOCK,
+        VK_FORMAT_ASTC_10x8_UNORM_BLOCK,  VK_FORMAT_ASTC_10x8_SRGB_BLOCK,
+        VK_FORMAT_ASTC_10x10_UNORM_BLOCK, VK_FORMAT_ASTC_10x10_SRGB_BLOCK,
+        VK_FORMAT_ASTC_12x10_UNORM_BLOCK, VK_FORMAT_ASTC_12x10_SRGB_BLOCK,
+        VK_FORMAT_ASTC_12x12_UNORM_BLOCK, VK_FORMAT_ASTC_12x12_SRGB_BLOCK,
+    };
     if (!features.textureCompressionASTC_LDR) {
         return false;
     }
     const auto format_feature_usage{
-        vk::FormatFeatureFlagBits::eSampledImage | vk::FormatFeatureFlagBits::eBlitSrc |
-        vk::FormatFeatureFlagBits::eBlitDst | vk::FormatFeatureFlagBits::eTransferSrc |
-        vk::FormatFeatureFlagBits::eTransferDst};
+        VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_BLIT_SRC_BIT |
+        VK_FORMAT_FEATURE_BLIT_DST_BIT | VK_FORMAT_FEATURE_TRANSFER_SRC_BIT |
+        VK_FORMAT_FEATURE_TRANSFER_DST_BIT};
     for (const auto format : astc_formats) {
-        const auto format_properties{physical.getFormatProperties(format, dldi)};
+        const auto format_properties{physical.GetFormatProperties(format)};
         if (!(format_properties.optimalTilingFeatures & format_feature_usage)) {
             return false;
         }
@@ -265,62 +391,49 @@ bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features
     return true;
 }
 
-bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
+bool VKDevice::IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage,
                                  FormatType format_type) const {
     const auto it = format_properties.find(wanted_format);
     if (it == format_properties.end()) {
-        UNIMPLEMENTED_MSG("Unimplemented format query={}", vk::to_string(wanted_format));
+        UNIMPLEMENTED_MSG("Unimplemented format query={}", wanted_format);
         return true;
     }
     const auto supported_usage = GetFormatFeatures(it->second, format_type);
     return (supported_usage & wanted_usage) == wanted_usage;
 }
 
-bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
-                          vk::SurfaceKHR surface) {
+bool VKDevice::IsSuitable(vk::PhysicalDevice physical, VkSurfaceKHR surface) {
     bool is_suitable = true;
+    std::bitset<REQUIRED_EXTENSIONS.size()> available_extensions;
 
-    constexpr std::array required_extensions = {
-        VK_KHR_SWAPCHAIN_EXTENSION_NAME,
-        VK_KHR_16BIT_STORAGE_EXTENSION_NAME,
-        VK_KHR_8BIT_STORAGE_EXTENSION_NAME,
-        VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME,
-        VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME,
-        VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME,
-        VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME,
-        VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME,
-    };
-    std::bitset<required_extensions.size()> available_extensions{};
-
-    for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
-        for (std::size_t i = 0; i < required_extensions.size(); ++i) {
+    for (const auto& prop : physical.EnumerateDeviceExtensionProperties()) {
+        for (std::size_t i = 0; i < REQUIRED_EXTENSIONS.size(); ++i) {
             if (available_extensions[i]) {
                 continue;
             }
-            available_extensions[i] =
-                required_extensions[i] == std::string_view{prop.extensionName};
+            const std::string_view name{prop.extensionName};
+            available_extensions[i] = name == REQUIRED_EXTENSIONS[i];
         }
     }
     if (!available_extensions.all()) {
-        for (std::size_t i = 0; i < required_extensions.size(); ++i) {
+        for (std::size_t i = 0; i < REQUIRED_EXTENSIONS.size(); ++i) {
             if (available_extensions[i]) {
                 continue;
             }
-            LOG_ERROR(Render_Vulkan, "Missing required extension: {}", required_extensions[i]);
+            LOG_ERROR(Render_Vulkan, "Missing required extension: {}", REQUIRED_EXTENSIONS[i]);
             is_suitable = false;
         }
     }
 
     bool has_graphics{}, has_present{};
-    const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+    const std::vector queue_family_properties = physical.GetQueueFamilyProperties();
     for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
         const auto& family = queue_family_properties[i];
         if (family.queueCount == 0) {
             continue;
         }
-        has_graphics |=
-            (family.queueFlags & vk::QueueFlagBits::eGraphics) != static_cast<vk::QueueFlagBits>(0);
-        has_present |= physical.getSurfaceSupportKHR(i, surface, dldi) != 0;
+        has_graphics |= family.queueFlags & VK_QUEUE_GRAPHICS_BIT;
+        has_present |= physical.GetSurfaceSupportKHR(i, surface);
     }
     if (!has_graphics || !has_present) {
         LOG_ERROR(Render_Vulkan, "Device lacks a graphics and present queue");
@@ -328,7 +441,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
     }
 
     // TODO(Rodrigo): Check if the device matches all requeriments.
-    const auto properties{physical.getProperties(dldi)};
+    const auto properties{physical.GetProperties()};
     const auto& limits{properties.limits};
 
     constexpr u32 required_ubo_size = 65536;
@@ -345,7 +458,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
         is_suitable = false;
     }
 
-    const auto features{physical.getFeatures(dldi)};
+    const auto features{physical.GetFeatures()};
     const std::array feature_report = {
         std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"),
         std::make_pair(features.independentBlend, "independentBlend"),
@@ -377,9 +490,9 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
     return is_suitable;
 }
 
-std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynamic& dldi) {
+std::vector<const char*> VKDevice::LoadExtensions() {
     std::vector<const char*> extensions;
-    const auto Test = [&](const vk::ExtensionProperties& extension,
+    const auto Test = [&](const VkExtensionProperties& extension,
                           std::optional<std::reference_wrapper<bool>> status, const char* name,
                           bool push) {
         if (extension.extensionName != std::string_view(name)) {
@@ -393,22 +506,13 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
         }
     };
 
-    extensions.reserve(15);
-    extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
-    extensions.push_back(VK_KHR_16BIT_STORAGE_EXTENSION_NAME);
-    extensions.push_back(VK_KHR_8BIT_STORAGE_EXTENSION_NAME);
-    extensions.push_back(VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME);
-    extensions.push_back(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME);
-    extensions.push_back(VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME);
-    extensions.push_back(VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME);
-    extensions.push_back(VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME);
+    extensions.reserve(7 + REQUIRED_EXTENSIONS.size());
+    extensions.insert(extensions.begin(), REQUIRED_EXTENSIONS.begin(), REQUIRED_EXTENSIONS.end());
 
-    [[maybe_unused]] const bool nsight =
-        std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED");
     bool has_khr_shader_float16_int8{};
     bool has_ext_subgroup_size_control{};
     bool has_ext_transform_feedback{};
-    for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
+    for (const auto& extension : physical.EnumerateDeviceExtensionProperties()) {
         Test(extension, khr_uniform_buffer_standard_layout,
              VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true);
         Test(extension, has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME,
@@ -428,38 +532,67 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
         }
     }
 
+    VkPhysicalDeviceFeatures2KHR features;
+    features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2_KHR;
+
+    VkPhysicalDeviceProperties2KHR properties;
+    properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR;
+
     if (has_khr_shader_float16_int8) {
-        is_float16_supported =
-            GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dldi).shaderFloat16;
+        VkPhysicalDeviceFloat16Int8FeaturesKHR float16_int8_features;
+        float16_int8_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT16_INT8_FEATURES_KHR;
+        float16_int8_features.pNext = nullptr;
+        features.pNext = &float16_int8_features;
+
+        physical.GetFeatures2KHR(features);
+        is_float16_supported = float16_int8_features.shaderFloat16;
         extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME);
     }
 
     if (has_ext_subgroup_size_control) {
-        const auto features =
-            GetFeatures<vk::PhysicalDeviceSubgroupSizeControlFeaturesEXT>(physical, dldi);
-        const auto properties =
-            GetProperties<vk::PhysicalDeviceSubgroupSizeControlPropertiesEXT>(physical, dldi);
+        VkPhysicalDeviceSubgroupSizeControlFeaturesEXT subgroup_features;
+        subgroup_features.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_FEATURES_EXT;
+        subgroup_features.pNext = nullptr;
+        features.pNext = &subgroup_features;
+        physical.GetFeatures2KHR(features);
 
-        is_warp_potentially_bigger = properties.maxSubgroupSize > GuestWarpSize;
+        VkPhysicalDeviceSubgroupSizeControlPropertiesEXT subgroup_properties;
+        subgroup_properties.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_PROPERTIES_EXT;
+        subgroup_properties.pNext = nullptr;
+        properties.pNext = &subgroup_properties;
+        physical.GetProperties2KHR(properties);
 
-        if (features.subgroupSizeControl && properties.minSubgroupSize <= GuestWarpSize &&
-            properties.maxSubgroupSize >= GuestWarpSize) {
+        is_warp_potentially_bigger = subgroup_properties.maxSubgroupSize > GuestWarpSize;
+
+        if (subgroup_features.subgroupSizeControl &&
+            subgroup_properties.minSubgroupSize <= GuestWarpSize &&
+            subgroup_properties.maxSubgroupSize >= GuestWarpSize) {
             extensions.push_back(VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME);
-            guest_warp_stages = properties.requiredSubgroupSizeStages;
+            guest_warp_stages = subgroup_properties.requiredSubgroupSizeStages;
         }
     } else {
         is_warp_potentially_bigger = true;
     }
 
     if (has_ext_transform_feedback) {
-        const auto features =
-            GetFeatures<vk::PhysicalDeviceTransformFeedbackFeaturesEXT>(physical, dldi);
-        const auto properties =
-            GetProperties<vk::PhysicalDeviceTransformFeedbackPropertiesEXT>(physical, dldi);
+        VkPhysicalDeviceTransformFeedbackFeaturesEXT tfb_features;
+        tfb_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT;
+        tfb_features.pNext = nullptr;
+        features.pNext = &tfb_features;
+        physical.GetFeatures2KHR(features);
 
-        if (features.transformFeedback && features.geometryStreams &&
-            properties.maxTransformFeedbackStreams >= 4 && properties.maxTransformFeedbackBuffers &&
-            properties.transformFeedbackQueries && properties.transformFeedbackDraw) {
+        VkPhysicalDeviceTransformFeedbackPropertiesEXT tfb_properties;
+        tfb_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_PROPERTIES_EXT;
+        tfb_properties.pNext = nullptr;
+        properties.pNext = &tfb_properties;
+        physical.GetProperties2KHR(properties);
+
+        if (tfb_features.transformFeedback && tfb_features.geometryStreams &&
+            tfb_properties.maxTransformFeedbackStreams >= 4 &&
+            tfb_properties.maxTransformFeedbackBuffers && tfb_properties.transformFeedbackQueries &&
+            tfb_properties.transformFeedbackDraw) {
             extensions.push_back(VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME);
             ext_transform_feedback = true;
         }
@@ -468,10 +601,10 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
     return extensions;
 }
 
-void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface) {
+void VKDevice::SetupFamilies(VkSurfaceKHR surface) {
     std::optional<u32> graphics_family_, present_family_;
 
-    const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+    const std::vector queue_family_properties = physical.GetQueueFamilyProperties();
     for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
         if (graphics_family_ && present_family_)
             break;
@@ -480,10 +613,12 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
         if (queue_family.queueCount == 0)
             continue;
 
-        if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics)
+        if (queue_family.queueFlags & VK_QUEUE_GRAPHICS_BIT) {
             graphics_family_ = i;
-        if (physical.getSurfaceSupportKHR(i, surface, dldi))
+        }
+        if (physical.GetSurfaceSupportKHR(i, surface)) {
             present_family_ = i;
+        }
     }
     ASSERT(graphics_family_ && present_family_);
 
@@ -491,111 +626,49 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
     present_family = *present_family_;
 }
 
-void VKDevice::SetupFeatures(const vk::DispatchLoaderDynamic& dldi) {
-    const auto supported_features{physical.getFeatures(dldi)};
+void VKDevice::SetupFeatures() {
+    const auto supported_features{physical.GetFeatures()};
     is_formatless_image_load_supported = supported_features.shaderStorageImageReadWithoutFormat;
-    is_optimal_astc_supported = IsOptimalAstcSupported(supported_features, dldi);
+    is_optimal_astc_supported = IsOptimalAstcSupported(supported_features);
 }
 
 void VKDevice::CollectTelemetryParameters() {
-    const auto driver = GetProperties<vk::PhysicalDeviceDriverPropertiesKHR>(physical, dld);
+    VkPhysicalDeviceDriverPropertiesKHR driver;
+    driver.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES_KHR;
+    driver.pNext = nullptr;
+
+    VkPhysicalDeviceProperties2KHR properties;
+    properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR;
+    properties.pNext = &driver;
+    physical.GetProperties2KHR(properties);
+
     driver_id = driver.driverID;
     vendor_name = driver.driverName;
 
-    const auto extensions = physical.enumerateDeviceExtensionProperties(nullptr, dld);
+    const std::vector extensions = physical.EnumerateDeviceExtensionProperties();
     reported_extensions.reserve(std::size(extensions));
     for (const auto& extension : extensions) {
         reported_extensions.push_back(extension.extensionName);
     }
 }
 
-std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const {
-    static const float QUEUE_PRIORITY = 1.0f;
+std::vector<VkDeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const {
+    static constexpr float QUEUE_PRIORITY = 1.0f;
 
-    std::set<u32> unique_queue_families = {graphics_family, present_family};
-    std::vector<vk::DeviceQueueCreateInfo> queue_cis;
+    std::unordered_set<u32> unique_queue_families = {graphics_family, present_family};
+    std::vector<VkDeviceQueueCreateInfo> queue_cis;
 
-    for (u32 queue_family : unique_queue_families)
-        queue_cis.push_back({{}, queue_family, 1, &QUEUE_PRIORITY});
+    for (const u32 queue_family : unique_queue_families) {
+        VkDeviceQueueCreateInfo& ci = queue_cis.emplace_back();
+        ci.sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO;
+        ci.pNext = nullptr;
+        ci.flags = 0;
+        ci.queueFamilyIndex = queue_family;
+        ci.queueCount = 1;
+        ci.pQueuePriorities = &QUEUE_PRIORITY;
+    }
 
     return queue_cis;
 }
 
-std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
-    const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) {
-    static constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32,
-                                        vk::Format::eA8B8G8R8UintPack32,
-                                        vk::Format::eA8B8G8R8SnormPack32,
-                                        vk::Format::eA8B8G8R8SrgbPack32,
-                                        vk::Format::eB5G6R5UnormPack16,
-                                        vk::Format::eA2B10G10R10UnormPack32,
-                                        vk::Format::eA1R5G5B5UnormPack16,
-                                        vk::Format::eR32G32B32A32Sfloat,
-                                        vk::Format::eR32G32B32A32Uint,
-                                        vk::Format::eR32G32Sfloat,
-                                        vk::Format::eR32G32Uint,
-                                        vk::Format::eR16G16B16A16Uint,
-                                        vk::Format::eR16G16B16A16Snorm,
-                                        vk::Format::eR16G16B16A16Unorm,
-                                        vk::Format::eR16G16Unorm,
-                                        vk::Format::eR16G16Snorm,
-                                        vk::Format::eR16G16Sfloat,
-                                        vk::Format::eR16Unorm,
-                                        vk::Format::eR8G8B8A8Srgb,
-                                        vk::Format::eR8G8Unorm,
-                                        vk::Format::eR8G8Snorm,
-                                        vk::Format::eR8Unorm,
-                                        vk::Format::eR8Uint,
-                                        vk::Format::eB10G11R11UfloatPack32,
-                                        vk::Format::eR32Sfloat,
-                                        vk::Format::eR32Uint,
-                                        vk::Format::eR32Sint,
-                                        vk::Format::eR16Sfloat,
-                                        vk::Format::eR16G16B16A16Sfloat,
-                                        vk::Format::eB8G8R8A8Unorm,
-                                        vk::Format::eR4G4B4A4UnormPack16,
-                                        vk::Format::eD32Sfloat,
-                                        vk::Format::eD16Unorm,
-                                        vk::Format::eD16UnormS8Uint,
-                                        vk::Format::eD24UnormS8Uint,
-                                        vk::Format::eD32SfloatS8Uint,
-                                        vk::Format::eBc1RgbaUnormBlock,
-                                        vk::Format::eBc2UnormBlock,
-                                        vk::Format::eBc3UnormBlock,
-                                        vk::Format::eBc4UnormBlock,
-                                        vk::Format::eBc5UnormBlock,
-                                        vk::Format::eBc5SnormBlock,
-                                        vk::Format::eBc7UnormBlock,
-                                        vk::Format::eBc6HUfloatBlock,
-                                        vk::Format::eBc6HSfloatBlock,
-                                        vk::Format::eBc1RgbaSrgbBlock,
-                                        vk::Format::eBc2SrgbBlock,
-                                        vk::Format::eBc3SrgbBlock,
-                                        vk::Format::eBc7SrgbBlock,
-                                        vk::Format::eAstc4x4SrgbBlock,
-                                        vk::Format::eAstc8x8SrgbBlock,
-                                        vk::Format::eAstc8x5SrgbBlock,
-                                        vk::Format::eAstc5x4SrgbBlock,
-                                        vk::Format::eAstc5x5UnormBlock,
-                                        vk::Format::eAstc5x5SrgbBlock,
-                                        vk::Format::eAstc10x8UnormBlock,
-                                        vk::Format::eAstc10x8SrgbBlock,
-                                        vk::Format::eAstc6x6UnormBlock,
-                                        vk::Format::eAstc6x6SrgbBlock,
-                                        vk::Format::eAstc10x10UnormBlock,
-                                        vk::Format::eAstc10x10SrgbBlock,
-                                        vk::Format::eAstc12x12UnormBlock,
-                                        vk::Format::eAstc12x12SrgbBlock,
-                                        vk::Format::eAstc8x6UnormBlock,
-                                        vk::Format::eAstc8x6SrgbBlock,
-                                        vk::Format::eAstc6x5UnormBlock,
-                                        vk::Format::eAstc6x5SrgbBlock,
-                                        vk::Format::eE5B9G9R9UfloatPack32};
-    std::unordered_map<vk::Format, vk::FormatProperties> format_properties;
-    for (const auto format : formats) {
-        format_properties.emplace(format, physical.getFormatProperties(format, dldi));
-    }
-    return format_properties;
-}
-
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_device.h

@@ -8,8 +8,9 @@
 #include <string_view>
 #include <unordered_map>
 #include <vector>
+
 #include "common/common_types.h"
-#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
 
@@ -22,12 +23,12 @@ const u32 GuestWarpSize = 32;
 /// Handles data specific to a physical device.
 class VKDevice final {
 public:
-    explicit VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
-                      vk::SurfaceKHR surface);
+    explicit VKDevice(VkInstance instance, vk::PhysicalDevice physical, VkSurfaceKHR surface,
+                      const vk::InstanceDispatch& dld);
     ~VKDevice();
 
     /// Initializes the device. Returns true on success.
-    bool Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance);
+    bool Create();
 
     /**
      * Returns a format supported by the device for the passed requeriments.
@@ -36,20 +37,20 @@ public:
      * @param format_type Format type usage.
      * @returns A format supported by the device.
      */
-    vk::Format GetSupportedFormat(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
-                                  FormatType format_type) const;
+    VkFormat GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage,
+                                FormatType format_type) const;
 
     /// Reports a device loss.
     void ReportLoss() const;
 
     /// Returns the dispatch loader with direct function pointers of the device.
-    const vk::DispatchLoaderDynamic& GetDispatchLoader() const {
+    const vk::DeviceDispatch& GetDispatchLoader() const {
         return dld;
     }
 
     /// Returns the logical device.
-    vk::Device GetLogical() const {
-        return logical.get();
+    const vk::Device& GetLogical() const {
+        return logical;
     }
 
     /// Returns the physical device.
@@ -79,7 +80,7 @@ public:
 
     /// Returns true if the device is integrated with the host CPU.
     bool IsIntegrated() const {
-        return properties.deviceType == vk::PhysicalDeviceType::eIntegratedGpu;
+        return properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU;
     }
 
     /// Returns the current Vulkan API version provided in Vulkan-formatted version numbers.
@@ -98,27 +99,27 @@ public:
     }
 
     /// Returns the driver ID.
-    vk::DriverIdKHR GetDriverID() const {
+    VkDriverIdKHR GetDriverID() const {
         return driver_id;
     }
 
     /// Returns uniform buffer alignment requeriment.
-    vk::DeviceSize GetUniformBufferAlignment() const {
+    VkDeviceSize GetUniformBufferAlignment() const {
         return properties.limits.minUniformBufferOffsetAlignment;
     }
 
     /// Returns storage alignment requeriment.
-    vk::DeviceSize GetStorageBufferAlignment() const {
+    VkDeviceSize GetStorageBufferAlignment() const {
         return properties.limits.minStorageBufferOffsetAlignment;
     }
 
     /// Returns the maximum range for storage buffers.
-    vk::DeviceSize GetMaxStorageBufferRange() const {
+    VkDeviceSize GetMaxStorageBufferRange() const {
         return properties.limits.maxStorageBufferRange;
     }
 
     /// Returns the maximum size for push constants.
-    vk::DeviceSize GetMaxPushConstantsSize() const {
+    VkDeviceSize GetMaxPushConstantsSize() const {
         return properties.limits.maxPushConstantsSize;
     }
 
@@ -138,8 +139,8 @@ public:
     }
 
     /// Returns true if the device can be forced to use the guest warp size.
-    bool IsGuestWarpSizeSupported(vk::ShaderStageFlagBits stage) const {
-        return (guest_warp_stages & stage) != vk::ShaderStageFlags{};
+    bool IsGuestWarpSizeSupported(VkShaderStageFlagBits stage) const {
+        return guest_warp_stages & stage;
     }
 
     /// Returns true if formatless image load is supported.
@@ -188,50 +189,44 @@ public:
     }
 
     /// Checks if the physical device is suitable.
-    static bool IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
-                           vk::SurfaceKHR surface);
+    static bool IsSuitable(vk::PhysicalDevice physical, VkSurfaceKHR surface);
 
 private:
     /// Loads extensions into a vector and stores available ones in this object.
-    std::vector<const char*> LoadExtensions(const vk::DispatchLoaderDynamic& dldi);
+    std::vector<const char*> LoadExtensions();
 
     /// Sets up queue families.
-    void SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface);
+    void SetupFamilies(VkSurfaceKHR surface);
 
     /// Sets up device features.
-    void SetupFeatures(const vk::DispatchLoaderDynamic& dldi);
+    void SetupFeatures();
 
     /// Collects telemetry information from the device.
     void CollectTelemetryParameters();
 
     /// Returns a list of queue initialization descriptors.
-    std::vector<vk::DeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
+    std::vector<VkDeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
 
     /// Returns true if ASTC textures are natively supported.
-    bool IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
-                                const vk::DispatchLoaderDynamic& dldi) const;
+    bool IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const;
 
     /// Returns true if a format is supported.
-    bool IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
+    bool IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage,
                            FormatType format_type) const;
 
-    /// Returns the device properties for Vulkan formats.
-    static std::unordered_map<vk::Format, vk::FormatProperties> GetFormatProperties(
-        const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical);
-
-    const vk::PhysicalDevice physical;        ///< Physical device.
-    vk::DispatchLoaderDynamic dld;            ///< Device function pointers.
-    vk::PhysicalDeviceProperties properties;  ///< Device properties.
-    UniqueDevice logical;                     ///< Logical device.
-    vk::Queue graphics_queue;                 ///< Main graphics queue.
-    vk::Queue present_queue;                  ///< Main present queue.
-    u32 graphics_family{};                    ///< Main graphics queue family index.
-    u32 present_family{};                     ///< Main present queue family index.
-    vk::DriverIdKHR driver_id{};              ///< Driver ID.
-    vk::ShaderStageFlags guest_warp_stages{}; ///< Stages where the guest warp size can be forced.ed
-    bool is_optimal_astc_supported{};         ///< Support for native ASTC.
-    bool is_float16_supported{};              ///< Support for float16 arithmetics.
-    bool is_warp_potentially_bigger{};        ///< Host warp size can be bigger than guest.
+    vk::DeviceDispatch dld;                 ///< Device function pointers.
+    vk::PhysicalDevice physical;            ///< Physical device.
+    VkPhysicalDeviceProperties properties;  ///< Device properties.
+    vk::Device logical;                     ///< Logical device.
+    vk::Queue graphics_queue;               ///< Main graphics queue.
+    vk::Queue present_queue;                ///< Main present queue.
+    u32 graphics_family{};                  ///< Main graphics queue family index.
+    u32 present_family{};                   ///< Main present queue family index.
+    VkDriverIdKHR driver_id{};              ///< Driver ID.
+    VkShaderStageFlags guest_warp_stages{}; ///< Stages where the guest warp size can be forced.ed
+    bool is_optimal_astc_supported{};       ///< Support for native ASTC.
+    bool is_float16_supported{};            ///< Support for float16 arithmetics.
+    bool is_warp_potentially_bigger{};      ///< Host warp size can be bigger than guest.
     bool is_formatless_image_load_supported{}; ///< Support for shader image read without format.
     bool khr_uniform_buffer_standard_layout{}; ///< Support for std430 on UBOs.
     bool ext_index_type_uint8{};               ///< Support for VK_EXT_index_type_uint8.
@@ -245,7 +240,7 @@ private:
     std::vector<std::string> reported_extensions; ///< Reported Vulkan extensions.
 
     /// Format properties dictionary.
-    std::unordered_map<vk::Format, vk::FormatProperties> format_properties;
+    std::unordered_map<VkFormat, VkFormatProperties> format_properties;
 };
 
 } // namespace Vulkan