yuzu/configuration: Only assert that all buttons exist when we are handling the click for a button device

This fixes failed assertions that were present in yuzu master code for 18 months.

.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:

CMakeModules/CopyYuzuQt5Deps.cmake

@@ -6,9 +6,9 @@ function(copy_yuzu_Qt5_deps target_dir)
     set(Qt5_STYLES_DIR "${Qt5_DIR}/../../../plugins/styles/")
     set(Qt5_IMAGEFORMATS_DIR "${Qt5_DIR}/../../../plugins/imageformats/")
     set(Qt5_RESOURCES_DIR "${Qt5_DIR}/../../../resources/")
-    set(PLATFORMS ${DLL_DEST}platforms/)
-    set(STYLES ${DLL_DEST}styles/)
-    set(IMAGEFORMATS ${DLL_DEST}imageformats/)
+    set(PLATFORMS ${DLL_DEST}plugins/platforms/)
+    set(STYLES ${DLL_DEST}plugins/styles/)
+    set(IMAGEFORMATS ${DLL_DEST}plugins/imageformats/)
     windows_copy_files(${target_dir} ${Qt5_DLL_DIR} ${DLL_DEST}
         icudt*.dll
         icuin*.dll
@@ -42,11 +42,15 @@ function(copy_yuzu_Qt5_deps target_dir)
             icudtl.dat
         )
     endif ()
-
     windows_copy_files(yuzu ${Qt5_PLATFORMS_DIR} ${PLATFORMS} qwindows$<$<CONFIG:Debug>:d>.*)
     windows_copy_files(yuzu ${Qt5_STYLES_DIR} ${STYLES} qwindowsvistastyle$<$<CONFIG:Debug>:d>.*)
     windows_copy_files(yuzu ${Qt5_IMAGEFORMATS_DIR} ${IMAGEFORMATS}
         qjpeg$<$<CONFIG:Debug>:d>.*
         qgif$<$<CONFIG:Debug>:d>.*
         )
+    # Create an empty qt.conf file. Qt will detect that this file exists, and use the folder that its in as the root folder.
+    # This way it'll look for plugins in the root/plugins/ folder
+    add_custom_command(TARGET yuzu POST_BUILD
+        COMMAND ${CMAKE_COMMAND} -E touch ${DLL_DEST}qt.conf
+    )
 endfunction(copy_yuzu_Qt5_deps)

README.md

@@ -2,6 +2,7 @@ yuzu emulator
 =============
 [![Travis CI Build Status](https://travis-ci.com/yuzu-emu/yuzu.svg?branch=master)](https://travis-ci.com/yuzu-emu/yuzu)
 [![Azure Mainline CI Build Status](https://dev.azure.com/yuzu-emu/yuzu/_apis/build/status/yuzu%20mainline?branchName=master)](https://dev.azure.com/yuzu-emu/yuzu/)
+[![Discord](https://img.shields.io/discord/398318088170242053?color=%237289DA&label=yuzu&logo=discord&logoColor=white)](https://discord.gg/XQV6dn9)
 
 yuzu is an experimental open-source emulator for the Nintendo Switch from the creators of [Citra](https://citra-emu.org/).
 
@@ -21,7 +22,7 @@ For development discussion, please join us on [Discord](https://discord.gg/XQV6d
 
 Most of the development happens on GitHub. It's also where [our central repository](https://github.com/yuzu-emu/yuzu) is hosted.
 
-If you want to contribute please take a look at the [Contributor's Guide](CONTRIBUTING.md) and [Developer Information](https://github.com/yuzu-emu/yuzu/wiki/Developer-Information). You should also contact any of the developers on Discord in order to know about the current state of the emulator.
+If you want to contribute please take a look at the [Contributor's Guide](https://github.com/yuzu-emu/yuzu/wiki/Contributing) and [Developer Information](https://github.com/yuzu-emu/yuzu/wiki/Developer-Information). You should also contact any of the developers on Discord in order to know about the current state of the emulator.
 
 ### Building
 

dist/license.md

@@ -2,8 +2,8 @@ The icons in this folder and its subfolders have the following licenses:
 
 Icon Name | License | Origin/Author
 --- | --- | ---
-qt_themes/default/icons/16x16/checked.png | Free for non-commercial use
-qt_themes/default/icons/16x16/failed.png | Free for non-commercial use
+qt_themes/default/icons/16x16/checked.png | CC BY-ND 3.0 | https://icons8.com
+qt_themes/default/icons/16x16/failed.png | CC BY-ND 3.0 | https://icons8.com
 qt_themes/default/icons/16x16/lock.png | CC BY-ND 3.0 | https://icons8.com
 qt_themes/default/icons/256x256/plus_folder.png | CC BY-ND 3.0 | https://icons8.com
 qt_themes/default/icons/48x48/bad_folder.png | CC BY-ND 3.0 | https://icons8.com
@@ -11,8 +11,6 @@ qt_themes/default/icons/48x48/chip.png | CC BY-ND 3.0 | https://icons8.com
 qt_themes/default/icons/48x48/folder.png | CC BY-ND 3.0 | https://icons8.com
 qt_themes/default/icons/48x48/plus.png | CC0 1.0 | Designed by BreadFish64 from the Citra team
 qt_themes/default/icons/48x48/sd_card.png | CC BY-ND 3.0 | https://icons8.com
-qt_themes/qdarkstyle/icons/16x16/checked.png | Free for non-commercial use
-qt_themes/qdarkstyle/icons/16x16/failed.png | Free for non-commercial use
 qt_themes/qdarkstyle/icons/16x16/lock.png | CC BY-ND 3.0 | https://icons8.com
 qt_themes/qdarkstyle/icons/256x256/plus_folder.png | CC BY-ND 3.0 | https://icons8.com
 qt_themes/qdarkstyle/icons/48x48/bad_folder.png | CC BY-ND 3.0 | https://icons8.com

externals/microprofile/microprofile.h

@@ -828,7 +828,7 @@ inline MicroProfileLogEntry MicroProfileMakeLogIndex(uint64_t nBegin, MicroProfi
     MicroProfileLogEntry Entry =  (nBegin<<62) | ((0x3fff&nToken)<<48) | (MP_LOG_TICK_MASK&nTick);
     int t = MicroProfileLogType(Entry);
     uint64_t nTimerIndex = MicroProfileLogTimerIndex(Entry);
-    MP_ASSERT(t == nBegin);
+    MP_ASSERT((uint64_t)t == nBegin);
     MP_ASSERT(nTimerIndex == (nToken&0x3fff));
     return Entry;
 
@@ -1556,10 +1556,10 @@ void MicroProfileFlip()
 
         pFramePut->nFrameStartCpu = MP_TICK();
         pFramePut->nFrameStartGpu = (uint32_t)MicroProfileGpuInsertTimeStamp();
-        if(pFrameNext->nFrameStartGpu != (uint64_t)-1)
+        if(pFrameNext->nFrameStartGpu != -1)
             pFrameNext->nFrameStartGpu = MicroProfileGpuGetTimeStamp((uint32_t)pFrameNext->nFrameStartGpu);
 
-        if(pFrameCurrent->nFrameStartGpu == (uint64_t)-1)
+        if(pFrameCurrent->nFrameStartGpu == -1)
             pFrameCurrent->nFrameStartGpu = pFrameNext->nFrameStartGpu + 1;
 
         uint64_t nFrameStartCpu = pFrameCurrent->nFrameStartCpu;

license.txt

@@ -343,8 +343,8 @@ The icons used in this project have the following licenses:
 
 Icon Name                                   | License       | Origin/Author
 ---                                         | ---           | ---
-checked.png                                 | Free for non-commercial use
-failed.png                                  | Free for non-commercial use
+checked.png                                 | CC BY-ND 3.0  | https://icons8.com
+failed.png                                  | CC BY-ND 3.0  | https://icons8.com
 lock.png                                    | CC BY-ND 3.0  | https://icons8.com
 plus_folder.png (Default, Dark)             | CC BY-ND 3.0  | https://icons8.com
 bad_folder.png (Default, Dark)              | CC BY-ND 3.0  | https://icons8.com

src/audio_core/audio_renderer.h

@@ -189,7 +189,7 @@ struct UpdateDataHeader {
     UpdateDataHeader() {}
 
     explicit UpdateDataHeader(const AudioRendererParameter& config) {
-        revision = Common::MakeMagic('R', 'E', 'V', '4'); // 5.1.0 Revision
+        revision = Common::MakeMagic('R', 'E', 'V', '8'); // 9.2.0 Revision
         behavior_size = 0xb0;
         memory_pools_size = (config.effect_count + (config.voice_count * 4)) * 0x10;
         voices_size = config.voice_count * 0x10;

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

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"
@@ -166,15 +165,14 @@ struct System::Impl {
         service_manager = std::make_shared<Service::SM::ServiceManager>();
 
         Service::Init(service_manager, system);
-        GDBStub::Init();
+        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/frontend/emu_window.h

@@ -13,19 +13,39 @@
 namespace Core::Frontend {
 
 /**
- * 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 +66,7 @@ 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 {
@@ -60,17 +80,9 @@ public:
     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;

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/gdbstub/gdbstub.cpp

@@ -141,6 +141,7 @@ constexpr char target_xml[] =
 )";
 
 int gdbserver_socket = -1;
+bool defer_start = false;
 
 u8 command_buffer[GDB_BUFFER_SIZE];
 u32 command_length;
@@ -1166,6 +1167,9 @@ static void RemoveBreakpoint() {
 
 void HandlePacket() {
     if (!IsConnected()) {
+        if (defer_start) {
+            ToggleServer(true);
+        }
         return;
     }
 
@@ -1256,6 +1260,10 @@ void ToggleServer(bool status) {
     }
 }
 
+void DeferStart() {
+    defer_start = true;
+}
+
 static void Init(u16 port) {
     if (!server_enabled) {
         // Set the halt loop to false in case the user enabled the gdbstub mid-execution.
@@ -1341,6 +1349,7 @@ void Shutdown() {
     if (!server_enabled) {
         return;
     }
+    defer_start = false;
 
     LOG_INFO(Debug_GDBStub, "Stopping GDB ...");
     if (gdbserver_socket != -1) {

src/core/gdbstub/gdbstub.h

@@ -43,6 +43,13 @@ void ToggleServer(bool status);
 /// Start the gdbstub server.
 void Init();
 
+/**
+ * Defer initialization of the gdbstub to the first packet processing functions.
+ * This avoids a case where the gdbstub thread is frozen after initialization
+ * and fails to respond in time to packets.
+ */
+void DeferStart();
+
 /// Stop gdbstub server.
 void Shutdown();
 

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

@@ -52,6 +52,11 @@ enum class LaunchParameterKind : u32 {
     AccountPreselectedUser = 2,
 };
 
+enum class VrMode : u8 {
+    Disabled = 0,
+    Enabled = 1,
+};
+
 constexpr u32 LAUNCH_PARAMETER_ACCOUNT_PRESELECTED_USER_MAGIC = 0xC79497CA;
 
 struct LaunchParameterAccountPreselectedUser {
@@ -605,11 +610,11 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system,
         {30, nullptr, "GetHomeButtonReaderLockAccessor"},
         {31, nullptr, "GetReaderLockAccessorEx"},
         {40, nullptr, "GetCradleFwVersion"},
-        {50, nullptr, "IsVrModeEnabled"},
-        {51, nullptr, "SetVrModeEnabled"},
+        {50, &ICommonStateGetter::IsVrModeEnabled, "IsVrModeEnabled"},
+        {51, &ICommonStateGetter::SetVrModeEnabled, "SetVrModeEnabled"},
         {52, &ICommonStateGetter::SetLcdBacklighOffEnabled, "SetLcdBacklighOffEnabled"},
         {53, nullptr, "BeginVrModeEx"},
-        {54, nullptr, "EndVrModeEx"},
+        {54, &ICommonStateGetter::EndVrModeEx, "EndVrModeEx"},
         {55, nullptr, "IsInControllerFirmwareUpdateSection"},
         {60, &ICommonStateGetter::GetDefaultDisplayResolution, "GetDefaultDisplayResolution"},
         {61, &ICommonStateGetter::GetDefaultDisplayResolutionChangeEvent, "GetDefaultDisplayResolutionChangeEvent"},
@@ -672,6 +677,30 @@ void ICommonStateGetter::GetCurrentFocusState(Kernel::HLERequestContext& ctx) {
     rb.Push(static_cast<u8>(FocusState::InFocus));
 }
 
+void ICommonStateGetter::IsVrModeEnabled(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushEnum(VrMode::Disabled);
+}
+
+void ICommonStateGetter::SetVrModeEnabled(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto is_vr_mode_enabled = rp.Pop<bool>();
+
+    LOG_WARNING(Service_AM, "(STUBBED) called. is_vr_mode_enabled={}", is_vr_mode_enabled);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    if (!is_vr_mode_enabled) {
+        rb.Push(RESULT_SUCCESS);
+    } else {
+        // TODO: Find better error code for this
+        UNIMPLEMENTED_MSG("is_vr_mode_enabled={}", is_vr_mode_enabled);
+        rb.Push(RESULT_UNKNOWN);
+    }
+}
+
 void ICommonStateGetter::SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto is_lcd_backlight_off_enabled = rp.Pop<bool>();
@@ -683,6 +712,13 @@ void ICommonStateGetter::SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx
     rb.Push(RESULT_SUCCESS);
 }
 
+void ICommonStateGetter::EndVrModeEx(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 void ICommonStateGetter::GetDefaultDisplayResolutionChangeEvent(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called");
 

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

@@ -182,7 +182,10 @@ private:
     void GetOperationMode(Kernel::HLERequestContext& ctx);
     void GetPerformanceMode(Kernel::HLERequestContext& ctx);
     void GetBootMode(Kernel::HLERequestContext& ctx);
+    void IsVrModeEnabled(Kernel::HLERequestContext& ctx);
+    void SetVrModeEnabled(Kernel::HLERequestContext& ctx);
     void SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx);
+    void EndVrModeEx(Kernel::HLERequestContext& ctx);
     void GetDefaultDisplayResolution(Kernel::HLERequestContext& ctx);
     void SetCpuBoostMode(Kernel::HLERequestContext& ctx);
 

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

@@ -254,6 +254,12 @@ void WebBrowser::Execute() {
 
     if (status != RESULT_SUCCESS) {
         complete = true;
+
+        // This is a workaround in order not to softlock yuzu when an error happens during the
+        // webapplet init. In order to avoid an svcBreak, the status is set to RESULT_SUCCESS
+        Finalize();
+        status = RESULT_SUCCESS;
+
         return;
     }
 

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/hid/hid.cpp

@@ -235,7 +235,7 @@ Hid::Hid(Core::System& system) : ServiceFramework("hid"), system(system) {
         {303, nullptr, "ActivateSevenSixAxisSensor"},
         {304, nullptr, "StartSevenSixAxisSensor"},
         {305, nullptr, "StopSevenSixAxisSensor"},
-        {306, nullptr, "InitializeSevenSixAxisSensor"},
+        {306, &Hid::InitializeSevenSixAxisSensor, "InitializeSevenSixAxisSensor"},
         {307, nullptr, "FinalizeSevenSixAxisSensor"},
         {308, nullptr, "SetSevenSixAxisSensorFusionStrength"},
         {309, nullptr, "GetSevenSixAxisSensorFusionStrength"},
@@ -853,6 +853,13 @@ void Hid::SetPalmaBoostMode(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
+void Hid::InitializeSevenSixAxisSensor(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service_HID, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 class HidDbg final : public ServiceFramework<HidDbg> {
 public:
     explicit HidDbg() : ServiceFramework{"hid:dbg"} {

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

@@ -128,6 +128,7 @@ private:
     void StopSixAxisSensor(Kernel::HLERequestContext& ctx);
     void SetIsPalmaAllConnectable(Kernel::HLERequestContext& ctx);
     void SetPalmaBoostMode(Kernel::HLERequestContext& ctx);
+    void InitializeSevenSixAxisSensor(Kernel::HLERequestContext& ctx);
 
     std::shared_ptr<IAppletResource> applet_resource;
     Core::System& system;

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/set/set.cpp

@@ -111,6 +111,14 @@ void SET::GetLanguageCode(Kernel::HLERequestContext& ctx) {
     rb.PushEnum(available_language_codes[Settings::values.language_index]);
 }
 
+void SET::GetRegionCode(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_SET, "called");
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+    rb.Push(Settings::values.region_index);
+}
+
 SET::SET() : ServiceFramework("set") {
     // clang-format off
     static const FunctionInfo functions[] = {
@@ -118,7 +126,7 @@ SET::SET() : ServiceFramework("set") {
         {1, &SET::GetAvailableLanguageCodes, "GetAvailableLanguageCodes"},
         {2, &SET::MakeLanguageCode, "MakeLanguageCode"},
         {3, &SET::GetAvailableLanguageCodeCount, "GetAvailableLanguageCodeCount"},
-        {4, nullptr, "GetRegionCode"},
+        {4, &SET::GetRegionCode, "GetRegionCode"},
         {5, &SET::GetAvailableLanguageCodes2, "GetAvailableLanguageCodes2"},
         {6, &SET::GetAvailableLanguageCodeCount2, "GetAvailableLanguageCodeCount2"},
         {7, nullptr, "GetKeyCodeMap"},

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

@@ -43,6 +43,7 @@ private:
     void GetAvailableLanguageCodeCount(Kernel::HLERequestContext& ctx);
     void GetAvailableLanguageCodeCount2(Kernel::HLERequestContext& ctx);
     void GetQuestFlag(Kernel::HLERequestContext& ctx);
+    void GetRegionCode(Kernel::HLERequestContext& ctx);
 };
 
 } // namespace Service::Set

src/core/hle/service/sm/controller.cpp

@@ -44,7 +44,7 @@ void Controller::QueryPointerBufferSize(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-    rb.Push<u16>(0x500);
+    rb.Push<u16>(0x1000);
 }
 
 Controller::Controller() : ServiceFramework("IpcController") {

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

@@ -30,7 +30,7 @@ Time::Time(std::shared_ptr<Module> module, Core::System& system, const char* nam
         {400, &Time::GetClockSnapshot, "GetClockSnapshot"},
         {401, &Time::GetClockSnapshotFromSystemClockContext, "GetClockSnapshotFromSystemClockContext"},
         {500, nullptr, "CalculateStandardUserSystemClockDifferenceByUser"},
-        {501, nullptr, "CalculateSpanBetween"},
+        {501, &Time::CalculateSpanBetween, "CalculateSpanBetween"},
     };
     // clang-format on
 

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

@@ -308,6 +308,35 @@ void Module::Interface::GetClockSnapshotFromSystemClockContext(Kernel::HLEReques
     ctx.WriteBuffer(&clock_snapshot, sizeof(Clock::ClockSnapshot));
 }
 
+void Module::Interface::CalculateSpanBetween(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>();
+
+    Clock::TimeSpanType time_span_type{};
+    s64 span{};
+    if (const ResultCode result{snapshot_a.steady_clock_time_point.GetSpanBetween(
+            snapshot_b.steady_clock_time_point, span)};
+        result != RESULT_SUCCESS) {
+        if (snapshot_a.network_time && snapshot_b.network_time) {
+            time_span_type =
+                Clock::TimeSpanType::FromSeconds(snapshot_b.network_time - snapshot_a.network_time);
+        } else {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_TIME_NOT_FOUND);
+            return;
+        }
+    } else {
+        time_span_type = Clock::TimeSpanType::FromSeconds(span);
+    }
+
+    IPC::ResponseBuilder rb{ctx, (sizeof(s64) / 4) + 2};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushRaw(time_span_type.nanoseconds);
+}
+
 void Module::Interface::GetSharedMemoryNativeHandle(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 1};

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 CalculateSpanBetween(Kernel::HLERequestContext& ctx);
         void GetSharedMemoryNativeHandle(Kernel::HLERequestContext& ctx);
 
     private:

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

@@ -53,7 +53,7 @@ static std::vector<std::string> BuildLocationNameCache(Core::System& system) {
         return {};
     }
 
-    std::vector<char> raw_data(binary_list->GetSize());
+    std::vector<char> raw_data(binary_list->GetSize() + 1);
     binary_list->ReadBytes<char>(raw_data.data(), binary_list->GetSize());
 
     std::stringstream data_stream{raw_data.data()};

src/core/settings.cpp

@@ -86,6 +86,7 @@ void LogSettings() {
     LogSetting("System_RngSeed", Settings::values.rng_seed.value_or(0));
     LogSetting("System_CurrentUser", Settings::values.current_user);
     LogSetting("System_LanguageIndex", Settings::values.language_index);
+    LogSetting("System_RegionIndex", Settings::values.region_index);
     LogSetting("Core_UseMultiCore", Settings::values.use_multi_core);
     LogSetting("Renderer_UseResolutionFactor", Settings::values.resolution_factor);
     LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);

src/core/settings.h

@@ -387,6 +387,8 @@ struct Values {
 
     s32 current_user;
     s32 language_index;
+    s32 region_index;
+    s32 sound_index;
 
     // Controls
     std::array<PlayerInput, 10> players;

src/input_common/udp/client.cpp

@@ -35,7 +35,7 @@ public:
           pad_index(pad_index) {
         boost::system::error_code ec{};
         auto ipv4 = boost::asio::ip::make_address_v4(host, ec);
-        if (ec.failed()) {
+        if (ec.value() != boost::system::errc::success) {
             LOG_ERROR(Input, "Invalid IPv4 address \"{}\" provided to socket", host);
             ipv4 = boost::asio::ip::address_v4{};
         }

src/input_common/udp/udp.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <mutex>
+#include <optional>
 #include <tuple>
 
 #include "common/param_package.h"
@@ -44,7 +45,7 @@ public:
     std::unique_ptr<Input::TouchDevice> Create(const Common::ParamPackage& params) override {
         {
             std::lock_guard guard(status->update_mutex);
-            status->touch_calibration.emplace();
+            status->touch_calibration = DeviceStatus::CalibrationData{};
             // These default values work well for DS4 but probably not other touch inputs
             status->touch_calibration->min_x = params.Get("min_x", 100);
             status->touch_calibration->min_y = params.Get("min_y", 50);

src/video_core/CMakeLists.txt

@@ -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/dirty_flags.cpp

@@ -15,14 +15,6 @@ namespace VideoCommon::Dirty {
 
 using Tegra::Engines::Maxwell3D;
 
-void SetupCommonOnWriteStores(Tegra::Engines::Maxwell3D::DirtyState::Flags& store) {
-    store[RenderTargets] = true;
-    store[ZetaBuffer] = true;
-    for (std::size_t i = 0; i < Maxwell3D::Regs::NumRenderTargets; ++i) {
-        store[ColorBuffer0 + i] = true;
-    }
-}
-
 void SetupDirtyRenderTargets(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables) {
     static constexpr std::size_t num_per_rt = NUM(rt[0]);
     static constexpr std::size_t begin = OFF(rt);

src/video_core/dirty_flags.h

@@ -44,8 +44,6 @@ void FillBlock(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables, std::size_
     FillBlock(tables[1], begin, num, index_b);
 }
 
-void SetupCommonOnWriteStores(Tegra::Engines::Maxwell3D::DirtyState::Flags& store);
-
 void SetupDirtyRenderTargets(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables);
 
 } // namespace VideoCommon::Dirty

src/video_core/engines/const_buffer_engine_interface.h

@@ -18,10 +18,14 @@ struct SamplerDescriptor {
     union {
         u32 raw = 0;
         BitField<0, 2, Tegra::Shader::TextureType> texture_type;
-        BitField<2, 3, Tegra::Texture::ComponentType> component_type;
+        BitField<2, 3, Tegra::Texture::ComponentType> r_type;
         BitField<5, 1, u32> is_array;
         BitField<6, 1, u32> is_buffer;
         BitField<7, 1, u32> is_shadow;
+        BitField<8, 3, Tegra::Texture::ComponentType> g_type;
+        BitField<11, 3, Tegra::Texture::ComponentType> b_type;
+        BitField<14, 3, Tegra::Texture::ComponentType> a_type;
+        BitField<17, 7, Tegra::Texture::TextureFormat> format;
     };
 
     bool operator==(const SamplerDescriptor& rhs) const noexcept {
@@ -36,9 +40,11 @@ struct SamplerDescriptor {
         using Tegra::Shader::TextureType;
         SamplerDescriptor result;
 
-        // This is going to be used to determine the shading language type.
-        // Because of that we don't care about all component types on color textures.
-        result.component_type.Assign(tic.r_type.Value());
+        result.format.Assign(tic.format.Value());
+        result.r_type.Assign(tic.r_type.Value());
+        result.g_type.Assign(tic.g_type.Value());
+        result.b_type.Assign(tic.b_type.Value());
+        result.a_type.Assign(tic.a_type.Value());
 
         switch (tic.texture_type.Value()) {
         case Tegra::Texture::TextureType::Texture1D:

src/video_core/engines/kepler_compute.cpp

@@ -119,14 +119,6 @@ Texture::TICEntry KeplerCompute::GetTICEntry(u32 tic_index) const {
     Texture::TICEntry tic_entry;
     memory_manager.ReadBlockUnsafe(tic_address_gpu, &tic_entry, sizeof(Texture::TICEntry));
 
-    const auto r_type{tic_entry.r_type.Value()};
-    const auto g_type{tic_entry.g_type.Value()};
-    const auto b_type{tic_entry.b_type.Value()};
-    const auto a_type{tic_entry.a_type.Value()};
-
-    // TODO(Subv): Different data types for separate components are not supported
-    DEBUG_ASSERT(r_type == g_type && r_type == b_type && r_type == a_type);
-
     return tic_entry;
 }
 

src/video_core/engines/maxwell_3d.cpp

@@ -98,6 +98,8 @@ void Maxwell3D::InitializeRegisterDefaults() {
     regs.framebuffer_srgb = 1;
     regs.front_face = Maxwell3D::Regs::FrontFace::ClockWise;
 
+    shadow_state = regs;
+
     mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_end_gl)] = true;
     mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_begin_gl)] = true;
     mme_inline[MAXWELL3D_REG_INDEX(vertex_buffer.count)] = true;
@@ -160,8 +162,17 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
     ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid Maxwell3D register, increase the size of the Regs structure");
 
-    if (regs.reg_array[method] != method_call.argument) {
-        regs.reg_array[method] = method_call.argument;
+    u32 arg = method_call.argument;
+    // Keep track of the register value in shadow_state when requested.
+    if (shadow_state.shadow_ram_control == Regs::ShadowRamControl::Track ||
+        shadow_state.shadow_ram_control == Regs::ShadowRamControl::TrackWithFilter) {
+        shadow_state.reg_array[method] = arg;
+    } else if (shadow_state.shadow_ram_control == Regs::ShadowRamControl::Replay) {
+        arg = shadow_state.reg_array[method];
+    }
+
+    if (regs.reg_array[method] != arg) {
+        regs.reg_array[method] = arg;
 
         for (const auto& table : dirty.tables) {
             dirty.flags[table[method]] = true;
@@ -169,12 +180,16 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
     }
 
     switch (method) {
+    case MAXWELL3D_REG_INDEX(shadow_ram_control): {
+        shadow_state.shadow_ram_control = static_cast<Regs::ShadowRamControl>(method_call.argument);
+        break;
+    }
     case MAXWELL3D_REG_INDEX(macros.data): {
-        ProcessMacroUpload(method_call.argument);
+        ProcessMacroUpload(arg);
         break;
     }
     case MAXWELL3D_REG_INDEX(macros.bind): {
-        ProcessMacroBind(method_call.argument);
+        ProcessMacroBind(arg);
         break;
     }
     case MAXWELL3D_REG_INDEX(firmware[4]): {
@@ -250,7 +265,7 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
     }
     case MAXWELL3D_REG_INDEX(data_upload): {
         const bool is_last_call = method_call.IsLastCall();
-        upload_state.ProcessData(method_call.argument, is_last_call);
+        upload_state.ProcessData(arg, is_last_call);
         if (is_last_call) {
             OnMemoryWrite();
         }

src/video_core/engines/maxwell_3d.h

@@ -531,6 +531,17 @@ public:
             Fill = 0x1b02,
         };
 
+        enum class ShadowRamControl : u32 {
+            // write value to shadow ram
+            Track = 0,
+            // write value to shadow ram ( with validation ??? )
+            TrackWithFilter = 1,
+            // only write to real hw register
+            Passthrough = 2,
+            // write value from shadow ram to real hw register
+            Replay = 3,
+        };
+
         struct RenderTargetConfig {
             u32 address_high;
             u32 address_low;
@@ -674,7 +685,9 @@ public:
                     u32 bind;
                 } macros;
 
-                INSERT_UNION_PADDING_WORDS(0x17);
+                ShadowRamControl shadow_ram_control;
+
+                INSERT_UNION_PADDING_WORDS(0x16);
 
                 Upload::Registers upload;
                 struct {
@@ -1263,7 +1276,12 @@ public:
             };
             std::array<u32, NUM_REGS> reg_array;
         };
-    } regs{};
+    };
+
+    Regs regs{};
+
+    /// Store temporary hw register values, used by some calls to restore state after a operation
+    Regs shadow_state;
 
     static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32), "Maxwell3D Regs has wrong size");
     static_assert(std::is_trivially_copyable_v<Regs>, "Maxwell3D Regs must be trivially copyable");
@@ -1458,6 +1476,7 @@ private:
                   "Field " #field_name " has invalid position")
 
 ASSERT_REG_POSITION(macros, 0x45);
+ASSERT_REG_POSITION(shadow_ram_control, 0x49);
 ASSERT_REG_POSITION(upload, 0x60);
 ASSERT_REG_POSITION(exec_upload, 0x6C);
 ASSERT_REG_POSITION(data_upload, 0x6D);

src/video_core/engines/shader_bytecode.h

@@ -82,6 +82,10 @@ union Attribute {
         Position = 7,
         Attribute_0 = 8,
         Attribute_31 = 39,
+        FrontColor = 40,
+        FrontSecondaryColor = 41,
+        BackColor = 42,
+        BackSecondaryColor = 43,
         ClipDistances0123 = 44,
         ClipDistances4567 = 45,
         PointCoord = 46,
@@ -89,6 +93,8 @@ union Attribute {
         // shader, and a tuple of (TessCoord.x, TessCoord.y, TessCoord.z, ~) when inside a Tess Eval
         // shader.
         TessCoordInstanceIDVertexID = 47,
+        TexCoord_0 = 48,
+        TexCoord_7 = 55,
         // This attribute contains a tuple of (Unk, Unk, Unk, gl_FrontFacing) when inside a fragment
         // shader. It is unknown what the other values contain.
         FrontFacing = 63,
@@ -225,18 +231,6 @@ enum class AtomicOp : u64 {
     Or = 6,
     Xor = 7,
     Exch = 8,
-};
-
-enum class GlobalAtomicOp : u64 {
-    Add = 0,
-    Min = 1,
-    Max = 2,
-    Inc = 3,
-    Dec = 4,
-    And = 5,
-    Or = 6,
-    Xor = 7,
-    Exch = 8,
     SafeAdd = 10,
 };
 
@@ -911,14 +905,9 @@ union Instruction {
     } fadd32i;
 
     union {
-        BitField<20, 8, u64> shift_position;
-        BitField<28, 8, u64> shift_length;
-        BitField<48, 1, u64> negate_b;
-        BitField<49, 1, u64> negate_a;
-
-        u64 GetLeftShiftValue() const {
-            return 32 - (shift_position + shift_length);
-        }
+        BitField<40, 1, u64> brev;
+        BitField<47, 1, u64> rd_cc;
+        BitField<48, 1, u64> is_signed;
     } bfe;
 
     union {
@@ -1000,7 +989,7 @@ union Instruction {
     } stg;
 
     union {
-        BitField<52, 4, GlobalAtomicOp> operation;
+        BitField<52, 4, AtomicOp> operation;
         BitField<49, 3, GlobalAtomicType> type;
         BitField<28, 20, s64> offset;
     } atom;

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;
@@ -39,6 +42,7 @@ enum class RenderTargetFormat : u32 {
     RGBA32_FLOAT = 0xC0,
     RGBA32_UINT = 0xC2,
     RGBA16_UNORM = 0xC6,
+    RGBA16_SNORM = 0xC7,
     RGBA16_UINT = 0xC9,
     RGBA16_FLOAT = 0xCA,
     RG32_FLOAT = 0xCB,
@@ -128,7 +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();
 
@@ -173,6 +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;
 
@@ -286,7 +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 +10,16 @@
 
 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}, gpu_context(std::move(context)),
+      cpu_context(renderer->GetRenderWindow().CreateSharedContext()) {}
 
 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) {

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
@@ -16,7 +20,8 @@ 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;
@@ -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);
+    renderer->Rasterizer().FlushRegion(addr, size);
 }
 
 void GPUSynch::InvalidateRegion(CacheAddr addr, u64 size) {
-    renderer.Rasterizer().InvalidateRegion(addr, size);
+    renderer->Rasterizer().InvalidateRegion(addr, size);
 }
 
 void GPUSynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
-    renderer.Rasterizer().FlushAndInvalidateRegion(addr, 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
@@ -15,7 +19,8 @@ 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;
@@ -29,6 +34,9 @@ public:
 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) {

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
 
@@ -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);

src/video_core/morton.cpp

@@ -51,6 +51,7 @@ static constexpr ConversionArray morton_to_linear_fns = {
     MortonCopy<true, PixelFormat::R8UI>,
     MortonCopy<true, PixelFormat::RGBA16F>,
     MortonCopy<true, PixelFormat::RGBA16U>,
+    MortonCopy<true, PixelFormat::RGBA16S>,
     MortonCopy<true, PixelFormat::RGBA16UI>,
     MortonCopy<true, PixelFormat::R11FG11FB10F>,
     MortonCopy<true, PixelFormat::RGBA32UI>,
@@ -131,6 +132,7 @@ static constexpr ConversionArray linear_to_morton_fns = {
     MortonCopy<false, PixelFormat::R8U>,
     MortonCopy<false, PixelFormat::R8UI>,
     MortonCopy<false, PixelFormat::RGBA16F>,
+    MortonCopy<false, PixelFormat::RGBA16S>,
     MortonCopy<false, PixelFormat::RGBA16U>,
     MortonCopy<false, PixelFormat::RGBA16UI>,
     MortonCopy<false, PixelFormat::R11FG11FB10F>,

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_rasterizer.cpp

@@ -93,10 +93,6 @@ void oglEnable(GLenum cap, bool state) {
     (state ? glEnable : glDisable)(cap);
 }
 
-void oglEnablei(GLenum cap, bool state, GLuint index) {
-    (state ? glEnablei : glDisablei)(cap, index);
-}
-
 } // Anonymous namespace
 
 RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
@@ -448,6 +444,7 @@ void RasterizerOpenGL::Clear() {
     }
 
     SyncRasterizeEnable();
+    SyncStencilTestState();
 
     if (regs.clear_flags.scissor) {
         SyncScissorTest();
@@ -478,7 +475,6 @@ void RasterizerOpenGL::Clear() {
 void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
     auto& gpu = system.GPU().Maxwell3D();
-    const auto& regs = gpu.regs;
 
     query_cache.UpdateCounters();
 
@@ -529,7 +525,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     // Upload vertex and index data.
     SetupVertexBuffer();
     SetupVertexInstances();
-    GLintptr index_buffer_offset;
+    GLintptr index_buffer_offset = 0;
     if (is_indexed) {
         index_buffer_offset = SetupIndexBuffer();
     }
@@ -555,7 +551,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     ConfigureFramebuffers();
 
     // Signal the buffer cache that we are not going to upload more things.
-    const bool invalidate = buffer_cache.Unmap();
+    buffer_cache.Unmap();
 
     // Now that we are no longer uploading data, we can safely bind the buffers to OpenGL.
     vertex_array_pushbuffer.Bind();
@@ -938,13 +934,15 @@ void RasterizerOpenGL::SyncViewport() {
             }
             flags[Dirty::Viewport0 + i] = false;
 
-            const Common::Rectangle<f32> rect{regs.viewport_transform[i].GetRect()};
+            const auto& src = regs.viewport_transform[i];
+            const Common::Rectangle<f32> rect{src.GetRect()};
             glViewportIndexedf(static_cast<GLuint>(i), rect.left, rect.bottom, rect.GetWidth(),
                                rect.GetHeight());
 
-            const auto& src = regs.viewports[i];
-            glDepthRangeIndexed(static_cast<GLuint>(i), static_cast<GLdouble>(src.depth_range_near),
-                                static_cast<GLdouble>(src.depth_range_far));
+            const GLdouble reduce_z = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne;
+            const GLdouble near_depth = src.translate_z - src.scale_z * reduce_z;
+            const GLdouble far_depth = src.translate_z + src.scale_z;
+            glDepthRangeIndexed(static_cast<GLuint>(i), near_depth, far_depth);
         }
     }
 }
@@ -1055,12 +1053,8 @@ 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),

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -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) {

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -366,10 +366,19 @@ constexpr bool IsGenericAttribute(Attribute::Index index) {
     return index >= Attribute::Index::Attribute_0 && index <= Attribute::Index::Attribute_31;
 }
 
+constexpr bool IsLegacyTexCoord(Attribute::Index index) {
+    return static_cast<int>(index) >= static_cast<int>(Attribute::Index::TexCoord_0) &&
+           static_cast<int>(index) <= static_cast<int>(Attribute::Index::TexCoord_7);
+}
+
 constexpr Attribute::Index ToGenericAttribute(u64 value) {
     return static_cast<Attribute::Index>(value + static_cast<u64>(Attribute::Index::Attribute_0));
 }
 
+constexpr int GetLegacyTexCoordIndex(Attribute::Index index) {
+    return static_cast<int>(index) - static_cast<int>(Attribute::Index::TexCoord_0);
+}
+
 u32 GetGenericAttributeIndex(Attribute::Index index) {
     ASSERT(IsGenericAttribute(index));
     return static_cast<u32>(index) - static_cast<u32>(Attribute::Index::Attribute_0);
@@ -393,10 +402,6 @@ std::string FlowStackTopName(MetaStackClass stack) {
     return fmt::format("{}_flow_stack_top", GetFlowStackPrefix(stack));
 }
 
-[[deprecated]] constexpr bool IsVertexShader(ShaderType stage) {
-    return stage == ShaderType::Vertex;
-}
-
 struct GenericVaryingDescription {
     std::string name;
     u8 first_element = 0;
@@ -502,7 +507,7 @@ private:
         if (!identifier.empty()) {
             code.AddLine("// {}", identifier);
         }
-        code.AddLine("#version 440 core");
+        code.AddLine("#version 440 {}", ir.UsesLegacyVaryings() ? "compatibility" : "core");
         code.AddLine("#extension GL_ARB_separate_shader_objects : enable");
         if (device.HasShaderBallot()) {
             code.AddLine("#extension GL_ARB_shader_ballot : require");
@@ -529,8 +534,9 @@ private:
     }
 
     void DeclareVertex() {
-        if (!IsVertexShader(stage))
+        if (stage != ShaderType::Vertex) {
             return;
+        }
 
         DeclareVertexRedeclarations();
     }
@@ -564,6 +570,16 @@ private:
         if (stage != ShaderType::Fragment) {
             return;
         }
+        if (ir.UsesLegacyVaryings()) {
+            code.AddLine("in gl_PerFragment {{");
+            ++code.scope;
+            code.AddLine("vec4 gl_TexCoord[8];");
+            code.AddLine("vec4 gl_Color;");
+            code.AddLine("vec4 gl_SecondaryColor;");
+            --code.scope;
+            code.AddLine("}};");
+        }
+
         for (u32 rt = 0; rt < Maxwell::NumRenderTargets; ++rt) {
             code.AddLine("layout (location = {}) out vec4 frag_color{};", rt, rt);
         }
@@ -602,14 +618,14 @@ private:
                 break;
             }
         }
-        if (!IsVertexShader(stage) || device.HasVertexViewportLayer()) {
+        if (stage != ShaderType::Vertex || device.HasVertexViewportLayer()) {
             if (ir.UsesLayer()) {
                 code.AddLine("int gl_Layer;");
             }
             if (ir.UsesViewportIndex()) {
                 code.AddLine("int gl_ViewportIndex;");
             }
-        } else if ((ir.UsesLayer() || ir.UsesViewportIndex()) && IsVertexShader(stage) &&
+        } else if ((ir.UsesLayer() || ir.UsesViewportIndex()) && stage == ShaderType::Vertex &&
                    !device.HasVertexViewportLayer()) {
             LOG_ERROR(
                 Render_OpenGL,
@@ -620,12 +636,12 @@ private:
             code.AddLine("float gl_PointSize;");
         }
 
-        if (ir.UsesInstanceId()) {
-            code.AddLine("int gl_InstanceID;");
-        }
-
-        if (ir.UsesVertexId()) {
-            code.AddLine("int gl_VertexID;");
+        if (ir.UsesLegacyVaryings()) {
+            code.AddLine("vec4 gl_TexCoord[8];");
+            code.AddLine("vec4 gl_FrontColor;");
+            code.AddLine("vec4 gl_FrontSecondaryColor;");
+            code.AddLine("vec4 gl_BackColor;");
+            code.AddLine("vec4 gl_BackSecondaryColor;");
         }
 
         --code.scope;
@@ -1131,6 +1147,10 @@ private:
             default:
                 UNREACHABLE();
             }
+        case Attribute::Index::FrontColor:
+            return {"gl_Color"s + GetSwizzle(element), Type::Float};
+        case Attribute::Index::FrontSecondaryColor:
+            return {"gl_SecondaryColor"s + GetSwizzle(element), Type::Float};
         case Attribute::Index::PointCoord:
             switch (element) {
             case 0:
@@ -1147,7 +1167,7 @@ private:
             // TODO(Subv): Find out what the values are for the first two elements when inside a
             // vertex shader, and what's the value of the fourth element when inside a Tess Eval
             // shader.
-            ASSERT(IsVertexShader(stage));
+            ASSERT(stage == ShaderType::Vertex);
             switch (element) {
             case 2:
                 // Config pack's first value is instance_id.
@@ -1171,6 +1191,12 @@ private:
                 return {GeometryPass(GetGenericInputAttribute(attribute)) + GetSwizzle(element),
                         Type::Float};
             }
+            if (IsLegacyTexCoord(attribute)) {
+                UNIMPLEMENTED_IF(stage == ShaderType::Geometry);
+                return {fmt::format("gl_TexCoord[{}]{}", GetLegacyTexCoordIndex(attribute),
+                                    GetSwizzle(element)),
+                        Type::Float};
+            }
             break;
         }
         UNIMPLEMENTED_MSG("Unhandled input attribute: {}", static_cast<u32>(attribute));
@@ -1209,21 +1235,22 @@ private:
     }
 
     std::optional<Expression> GetOutputAttribute(const AbufNode* abuf) {
+        const u32 element = abuf->GetElement();
         switch (const auto attribute = abuf->GetIndex()) {
         case Attribute::Index::Position:
-            return {{"gl_Position"s + GetSwizzle(abuf->GetElement()), Type::Float}};
+            return {{"gl_Position"s + GetSwizzle(element), Type::Float}};
         case Attribute::Index::LayerViewportPointSize:
-            switch (abuf->GetElement()) {
+            switch (element) {
             case 0:
                 UNIMPLEMENTED();
                 return {};
             case 1:
-                if (IsVertexShader(stage) && !device.HasVertexViewportLayer()) {
+                if (stage == ShaderType::Vertex && !device.HasVertexViewportLayer()) {
                     return {};
                 }
                 return {{"gl_Layer", Type::Int}};
             case 2:
-                if (IsVertexShader(stage) && !device.HasVertexViewportLayer()) {
+                if (stage == ShaderType::Vertex && !device.HasVertexViewportLayer()) {
                     return {};
                 }
                 return {{"gl_ViewportIndex", Type::Int}};
@@ -1231,13 +1258,26 @@ private:
                 return {{"gl_PointSize", Type::Float}};
             }
             return {};
+        case Attribute::Index::FrontColor:
+            return {{"gl_FrontColor"s + GetSwizzle(element), Type::Float}};
+        case Attribute::Index::FrontSecondaryColor:
+            return {{"gl_FrontSecondaryColor"s + GetSwizzle(element), Type::Float}};
+        case Attribute::Index::BackColor:
+            return {{"gl_BackColor"s + GetSwizzle(element), Type::Float}};
+        case Attribute::Index::BackSecondaryColor:
+            return {{"gl_BackSecondaryColor"s + GetSwizzle(element), Type::Float}};
         case Attribute::Index::ClipDistances0123:
-            return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement()), Type::Float}};
+            return {{fmt::format("gl_ClipDistance[{}]", element), Type::Float}};
         case Attribute::Index::ClipDistances4567:
-            return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement() + 4), Type::Float}};
+            return {{fmt::format("gl_ClipDistance[{}]", element + 4), Type::Float}};
         default:
             if (IsGenericAttribute(attribute)) {
-                return {{GetGenericOutputAttribute(attribute, abuf->GetElement()), Type::Float}};
+                return {{GetGenericOutputAttribute(attribute, element), Type::Float}};
+            }
+            if (IsLegacyTexCoord(attribute)) {
+                return {{fmt::format("gl_TexCoord[{}]{}", GetLegacyTexCoordIndex(attribute),
+                                     GetSwizzle(element)),
+                         Type::Float}};
             }
             UNIMPLEMENTED_MSG("Unhandled output attribute: {}", static_cast<u32>(attribute));
             return {};
@@ -2009,16 +2049,19 @@ private:
         expr += GetSampler(meta->sampler);
         expr += ", ";
 
-        expr += constructors.at(operation.GetOperandsCount() - 1);
+        expr += constructors.at(operation.GetOperandsCount() + (meta->array ? 1 : 0) - 1);
         expr += '(';
         for (std::size_t i = 0; i < count; ++i) {
-            expr += VisitOperand(operation, i).AsInt();
-            const std::size_t next = i + 1;
-            if (next == count)
-                expr += ')';
-            else if (next < count)
+            if (i > 0) {
                 expr += ", ";
+            }
+            expr += VisitOperand(operation, i).AsInt();
         }
+        if (meta->array) {
+            expr += ", ";
+            expr += Visit(meta->array).AsInt();
+        }
+        expr += ')';
 
         if (meta->lod && !meta->sampler.IsBuffer()) {
             expr += ", ";
@@ -2071,6 +2114,10 @@ private:
 
     template <const std::string_view& opname, Type type>
     Expression Atomic(Operation operation) {
+        if ((opname == Func::Min || opname == Func::Max) && type == Type::Int) {
+            UNIMPLEMENTED_MSG("Unimplemented Min & Max for atomic operations");
+            return {};
+        }
         return {fmt::format("atomic{}({}, {})", opname, Visit(operation[0]).GetCode(),
                             Visit(operation[1]).As(type)),
                 type};
@@ -2264,6 +2311,8 @@ private:
         ~Func() = delete;
 
         static constexpr std::string_view Add = "Add";
+        static constexpr std::string_view Min = "Min";
+        static constexpr std::string_view Max = "Max";
         static constexpr std::string_view And = "And";
         static constexpr std::string_view Or = "Or";
         static constexpr std::string_view Xor = "Xor";
@@ -2414,7 +2463,21 @@ private:
         &GLSLDecompiler::AtomicImage<Func::Xor>,
         &GLSLDecompiler::AtomicImage<Func::Exchange>,
 
+        &GLSLDecompiler::Atomic<Func::Exchange, Type::Uint>,
         &GLSLDecompiler::Atomic<Func::Add, Type::Uint>,
+        &GLSLDecompiler::Atomic<Func::Min, Type::Uint>,
+        &GLSLDecompiler::Atomic<Func::Max, Type::Uint>,
+        &GLSLDecompiler::Atomic<Func::And, Type::Uint>,
+        &GLSLDecompiler::Atomic<Func::Or, Type::Uint>,
+        &GLSLDecompiler::Atomic<Func::Xor, Type::Uint>,
+
+        &GLSLDecompiler::Atomic<Func::Exchange, Type::Int>,
+        &GLSLDecompiler::Atomic<Func::Add, Type::Int>,
+        &GLSLDecompiler::Atomic<Func::Min, Type::Int>,
+        &GLSLDecompiler::Atomic<Func::Max, Type::Int>,
+        &GLSLDecompiler::Atomic<Func::And, Type::Int>,
+        &GLSLDecompiler::Atomic<Func::Or, Type::Int>,
+        &GLSLDecompiler::Atomic<Func::Xor, Type::Int>,
 
         &GLSLDecompiler::Branch,
         &GLSLDecompiler::BranchIndirect,
@@ -2529,7 +2592,7 @@ private:
     }
 
     u32 GetNumPhysicalInputAttributes() const {
-        return IsVertexShader(stage) ? GetNumPhysicalAttributes() : GetNumPhysicalVaryings();
+        return stage == ShaderType::Vertex ? GetNumPhysicalAttributes() : GetNumPhysicalVaryings();
     }
 
     u32 GetNumPhysicalAttributes() const {

src/video_core/renderer_opengl/gl_state_tracker.cpp

@@ -238,7 +238,6 @@ void StateTracker::Initialize() {
     SetupDirtyMisc(tables);
 
     auto& store = dirty.on_write_stores;
-    SetupCommonOnWriteStores(store);
     store[VertexBuffers] = true;
     for (std::size_t i = 0; i < Regs::NumVertexArrays; ++i) {
         store[VertexBuffer0 + i] = true;

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -53,6 +53,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format
     {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

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -5,8 +5,11 @@
 #include <algorithm>
 #include <cstddef>
 #include <cstdlib>
+#include <cstring>
 #include <memory>
+
 #include <glad/glad.h>
+
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/microprofile.h"
@@ -25,8 +28,8 @@
 
 namespace OpenGL {
 
-// 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.
+namespace {
+
 constexpr std::size_t SWAP_CHAIN_SIZE = 3;
 
 struct Frame {
@@ -41,6 +44,159 @@ struct Frame {
     bool is_srgb{};                   /// Framebuffer is sRGB or RGB
 };
 
+constexpr char VERTEX_SHADER[] = R"(
+#version 430 core
+
+out gl_PerVertex {
+    vec4 gl_Position;
+};
+
+layout (location = 0) in vec2 vert_position;
+layout (location = 1) in vec2 vert_tex_coord;
+layout (location = 0) out vec2 frag_tex_coord;
+
+// This is a truncated 3x3 matrix for 2D transformations:
+// The upper-left 2x2 submatrix performs scaling/rotation/mirroring.
+// The third column performs translation.
+// The third row could be used for projection, which we don't need in 2D. It hence is assumed to
+// implicitly be [0, 0, 1]
+layout (location = 0) uniform mat3x2 modelview_matrix;
+
+void main() {
+    // Multiply input position by the rotscale part of the matrix and then manually translate by
+    // the last column. This is equivalent to using a full 3x3 matrix and expanding the vector
+    // to `vec3(vert_position.xy, 1.0)`
+    gl_Position = vec4(mat2(modelview_matrix) * vert_position + modelview_matrix[2], 0.0, 1.0);
+    frag_tex_coord = vert_tex_coord;
+}
+)";
+
+constexpr char FRAGMENT_SHADER[] = R"(
+#version 430 core
+
+layout (location = 0) in vec2 frag_tex_coord;
+layout (location = 0) out vec4 color;
+
+layout (binding = 0) uniform sampler2D color_texture;
+
+void main() {
+    color = vec4(texture(color_texture, frag_tex_coord).rgb, 1.0f);
+}
+)";
+
+constexpr GLint PositionLocation = 0;
+constexpr GLint TexCoordLocation = 1;
+constexpr GLint ModelViewMatrixLocation = 0;
+
+struct ScreenRectVertex {
+    constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v)
+        : position{{static_cast<GLfloat>(x), static_cast<GLfloat>(y)}}, tex_coord{{u, v}} {}
+
+    std::array<GLfloat, 2> position;
+    std::array<GLfloat, 2> tex_coord;
+};
+
+/// Returns true if any debug tool is attached
+bool HasDebugTool() {
+    const bool nsight = std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED");
+    if (nsight) {
+        return true;
+    }
+
+    GLint num_extensions;
+    glGetIntegerv(GL_NUM_EXTENSIONS, &num_extensions);
+    for (GLuint index = 0; index < static_cast<GLuint>(num_extensions); ++index) {
+        const auto name = reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, index));
+        if (!std::strcmp(name, "GL_EXT_debug_tool")) {
+            return true;
+        }
+    }
+    return false;
+}
+
+/**
+ * Defines a 1:1 pixel ortographic projection matrix with (0,0) on the top-left
+ * corner and (width, height) on the lower-bottom.
+ *
+ * The projection part of the matrix is trivial, hence these operations are represented
+ * by a 3x2 matrix.
+ */
+std::array<GLfloat, 3 * 2> MakeOrthographicMatrix(float width, float height) {
+    std::array<GLfloat, 3 * 2> matrix; // Laid out in column-major order
+
+    // clang-format off
+    matrix[0] = 2.f / width; matrix[2] =  0.f;          matrix[4] = -1.f;
+    matrix[1] = 0.f;         matrix[3] = -2.f / height; matrix[5] =  1.f;
+    // Last matrix row is implicitly assumed to be [0, 0, 1].
+    // clang-format on
+
+    return matrix;
+}
+
+const char* GetSource(GLenum source) {
+    switch (source) {
+    case GL_DEBUG_SOURCE_API:
+        return "API";
+    case GL_DEBUG_SOURCE_WINDOW_SYSTEM:
+        return "WINDOW_SYSTEM";
+    case GL_DEBUG_SOURCE_SHADER_COMPILER:
+        return "SHADER_COMPILER";
+    case GL_DEBUG_SOURCE_THIRD_PARTY:
+        return "THIRD_PARTY";
+    case GL_DEBUG_SOURCE_APPLICATION:
+        return "APPLICATION";
+    case GL_DEBUG_SOURCE_OTHER:
+        return "OTHER";
+    default:
+        UNREACHABLE();
+        return "Unknown source";
+    }
+}
+
+const char* GetType(GLenum type) {
+    switch (type) {
+    case GL_DEBUG_TYPE_ERROR:
+        return "ERROR";
+    case GL_DEBUG_TYPE_DEPRECATED_BEHAVIOR:
+        return "DEPRECATED_BEHAVIOR";
+    case GL_DEBUG_TYPE_UNDEFINED_BEHAVIOR:
+        return "UNDEFINED_BEHAVIOR";
+    case GL_DEBUG_TYPE_PORTABILITY:
+        return "PORTABILITY";
+    case GL_DEBUG_TYPE_PERFORMANCE:
+        return "PERFORMANCE";
+    case GL_DEBUG_TYPE_OTHER:
+        return "OTHER";
+    case GL_DEBUG_TYPE_MARKER:
+        return "MARKER";
+    default:
+        UNREACHABLE();
+        return "Unknown type";
+    }
+}
+
+void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severity, GLsizei length,
+                           const GLchar* message, const void* user_param) {
+    const char format[] = "{} {} {}: {}";
+    const char* const str_source = GetSource(source);
+    const char* const str_type = GetType(type);
+
+    switch (severity) {
+    case GL_DEBUG_SEVERITY_HIGH:
+        LOG_CRITICAL(Render_OpenGL, format, str_source, str_type, id, message);
+        break;
+    case GL_DEBUG_SEVERITY_MEDIUM:
+        LOG_WARNING(Render_OpenGL, format, str_source, str_type, id, message);
+        break;
+    case GL_DEBUG_SEVERITY_NOTIFICATION:
+    case GL_DEBUG_SEVERITY_LOW:
+        LOG_DEBUG(Render_OpenGL, format, str_source, str_type, id, message);
+        break;
+    }
+}
+
+} // Anonymous namespace
+
 /**
  * For smooth Vsync rendering, we want to always present the latest frame that the core generates,
  * but also make sure that rendering happens at the pace that the frontend dictates. This is a
@@ -157,146 +313,10 @@ public:
     }
 };
 
-namespace {
-
-constexpr char VERTEX_SHADER[] = R"(
-#version 430 core
-
-out gl_PerVertex {
-    vec4 gl_Position;
-};
-
-layout (location = 0) in vec2 vert_position;
-layout (location = 1) in vec2 vert_tex_coord;
-layout (location = 0) out vec2 frag_tex_coord;
-
-// This is a truncated 3x3 matrix for 2D transformations:
-// The upper-left 2x2 submatrix performs scaling/rotation/mirroring.
-// The third column performs translation.
-// The third row could be used for projection, which we don't need in 2D. It hence is assumed to
-// implicitly be [0, 0, 1]
-layout (location = 0) uniform mat3x2 modelview_matrix;
-
-void main() {
-    // Multiply input position by the rotscale part of the matrix and then manually translate by
-    // the last column. This is equivalent to using a full 3x3 matrix and expanding the vector
-    // to `vec3(vert_position.xy, 1.0)`
-    gl_Position = vec4(mat2(modelview_matrix) * vert_position + modelview_matrix[2], 0.0, 1.0);
-    frag_tex_coord = vert_tex_coord;
-}
-)";
-
-constexpr char FRAGMENT_SHADER[] = R"(
-#version 430 core
-
-layout (location = 0) in vec2 frag_tex_coord;
-layout (location = 0) out vec4 color;
-
-layout (binding = 0) uniform sampler2D color_texture;
-
-void main() {
-    color = vec4(texture(color_texture, frag_tex_coord).rgb, 1.0f);
-}
-)";
-
-constexpr GLint PositionLocation = 0;
-constexpr GLint TexCoordLocation = 1;
-constexpr GLint ModelViewMatrixLocation = 0;
-
-struct ScreenRectVertex {
-    constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v)
-        : position{{static_cast<GLfloat>(x), static_cast<GLfloat>(y)}}, tex_coord{{u, v}} {}
-
-    std::array<GLfloat, 2> position;
-    std::array<GLfloat, 2> tex_coord;
-};
-
-/**
- * Defines a 1:1 pixel ortographic projection matrix with (0,0) on the top-left
- * corner and (width, height) on the lower-bottom.
- *
- * The projection part of the matrix is trivial, hence these operations are represented
- * by a 3x2 matrix.
- */
-std::array<GLfloat, 3 * 2> MakeOrthographicMatrix(float width, float height) {
-    std::array<GLfloat, 3 * 2> matrix; // Laid out in column-major order
-
-    // clang-format off
-    matrix[0] = 2.f / width; matrix[2] =  0.f;          matrix[4] = -1.f;
-    matrix[1] = 0.f;         matrix[3] = -2.f / height; matrix[5] =  1.f;
-    // Last matrix row is implicitly assumed to be [0, 0, 1].
-    // clang-format on
-
-    return matrix;
-}
-
-const char* GetSource(GLenum source) {
-    switch (source) {
-    case GL_DEBUG_SOURCE_API:
-        return "API";
-    case GL_DEBUG_SOURCE_WINDOW_SYSTEM:
-        return "WINDOW_SYSTEM";
-    case GL_DEBUG_SOURCE_SHADER_COMPILER:
-        return "SHADER_COMPILER";
-    case GL_DEBUG_SOURCE_THIRD_PARTY:
-        return "THIRD_PARTY";
-    case GL_DEBUG_SOURCE_APPLICATION:
-        return "APPLICATION";
-    case GL_DEBUG_SOURCE_OTHER:
-        return "OTHER";
-    default:
-        UNREACHABLE();
-        return "Unknown source";
-    }
-}
-
-const char* GetType(GLenum type) {
-    switch (type) {
-    case GL_DEBUG_TYPE_ERROR:
-        return "ERROR";
-    case GL_DEBUG_TYPE_DEPRECATED_BEHAVIOR:
-        return "DEPRECATED_BEHAVIOR";
-    case GL_DEBUG_TYPE_UNDEFINED_BEHAVIOR:
-        return "UNDEFINED_BEHAVIOR";
-    case GL_DEBUG_TYPE_PORTABILITY:
-        return "PORTABILITY";
-    case GL_DEBUG_TYPE_PERFORMANCE:
-        return "PERFORMANCE";
-    case GL_DEBUG_TYPE_OTHER:
-        return "OTHER";
-    case GL_DEBUG_TYPE_MARKER:
-        return "MARKER";
-    default:
-        UNREACHABLE();
-        return "Unknown type";
-    }
-}
-
-void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severity, GLsizei length,
-                           const GLchar* message, const void* user_param) {
-    const char format[] = "{} {} {}: {}";
-    const char* const str_source = GetSource(source);
-    const char* const str_type = GetType(type);
-
-    switch (severity) {
-    case GL_DEBUG_SEVERITY_HIGH:
-        LOG_CRITICAL(Render_OpenGL, format, str_source, str_type, id, message);
-        break;
-    case GL_DEBUG_SEVERITY_MEDIUM:
-        LOG_WARNING(Render_OpenGL, format, str_source, str_type, id, message);
-        break;
-    case GL_DEBUG_SEVERITY_NOTIFICATION:
-    case GL_DEBUG_SEVERITY_LOW:
-        LOG_DEBUG(Render_OpenGL, format, str_source, str_type, id, message);
-        break;
-    }
-}
-
-} // Anonymous namespace
-
-RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system)
+RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system,
+                               Core::Frontend::GraphicsContext& context)
     : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system},
-      frame_mailbox{std::make_unique<FrameMailbox>()} {}
+      frame_mailbox{}, context{context}, has_debug_tool{HasDebugTool()} {}
 
 RendererOpenGL::~RendererOpenGL() = default;
 
@@ -304,8 +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;
     }
@@ -361,6 +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) {
@@ -428,6 +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);
 
@@ -640,12 +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
@@ -673,6 +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_vulkan/maxwell_to_vk.cpp

@@ -125,6 +125,7 @@ struct FormatTuple {
     {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
@@ -256,6 +257,8 @@ vk::ShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage) {
         return vk::ShaderStageFlagBits::eGeometry;
     case Tegra::Engines::ShaderType::Fragment:
         return vk::ShaderStageFlagBits::eFragment;
+    case Tegra::Engines::ShaderType::Compute:
+        return vk::ShaderStageFlagBits::eCompute;
     }
     UNIMPLEMENTED_MSG("Unimplemented shader stage={}", static_cast<u32>(stage));
     return {};
@@ -331,6 +334,8 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
             return vk::Format::eR16G16B16Unorm;
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
             return vk::Format::eR16G16B16A16Unorm;
+        case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+            return vk::Format::eA2B10G10R10UnormPack32;
         default:
             break;
         }
@@ -364,6 +369,10 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
             return vk::Format::eR8G8B8A8Uint;
         case Maxwell::VertexAttribute::Size::Size_32:
             return vk::Format::eR32Uint;
+        case Maxwell::VertexAttribute::Size::Size_32_32:
+            return vk::Format::eR32G32Uint;
+        case Maxwell::VertexAttribute::Size::Size_32_32_32:
+            return vk::Format::eR32G32B32Uint;
         case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
             return vk::Format::eR32G32B32A32Uint;
         default:
@@ -392,6 +401,26 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
         }
         break;
     case Maxwell::VertexAttribute::Type::SignedScaled:
+        switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
+            return vk::Format::eR8Sscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8:
+            return vk::Format::eR8G8Sscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+            return vk::Format::eR8G8B8Sscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return vk::Format::eR8G8B8A8Sscaled;
+        case Maxwell::VertexAttribute::Size::Size_16:
+            return vk::Format::eR16Sscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+            return vk::Format::eR16G16Sscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+            return vk::Format::eR16G16B16Sscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return vk::Format::eR16G16B16A16Sscaled;
+        default:
+            break;
+        }
         break;
     case Maxwell::VertexAttribute::Type::Float:
         switch (size) {

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -141,8 +141,9 @@ 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() {

src/video_core/renderer_vulkan/renderer_vulkan.h

@@ -42,7 +42,7 @@ 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;
 
 private:
     std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback(

src/video_core/renderer_vulkan/vk_device.cpp

@@ -535,6 +535,7 @@ std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperti
                                         vk::Format::eR32G32Sfloat,
                                         vk::Format::eR32G32Uint,
                                         vk::Format::eR16G16B16A16Uint,
+                                        vk::Format::eR16G16B16A16Snorm,
                                         vk::Format::eR16G16B16A16Unorm,
                                         vk::Format::eR16G16Unorm,
                                         vk::Format::eR16G16Snorm,

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -179,10 +179,11 @@ Tegra::Engines::ConstBufferEngineInterface& CachedShader::GetEngine(
 VKPipelineCache::VKPipelineCache(Core::System& system, RasterizerVulkan& rasterizer,
                                  const VKDevice& device, VKScheduler& scheduler,
                                  VKDescriptorPool& descriptor_pool,
-                                 VKUpdateDescriptorQueue& update_descriptor_queue)
+                                 VKUpdateDescriptorQueue& update_descriptor_queue,
+                                 VKRenderPassCache& renderpass_cache)
     : RasterizerCache{rasterizer}, system{system}, device{device}, scheduler{scheduler},
       descriptor_pool{descriptor_pool}, update_descriptor_queue{update_descriptor_queue},
-      renderpass_cache(device) {}
+      renderpass_cache{renderpass_cache} {}
 
 VKPipelineCache::~VKPipelineCache() = default;
 
@@ -191,7 +192,6 @@ std::array<Shader, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() {
 
     std::array<Shader, Maxwell::MaxShaderProgram> shaders;
     for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
-        const auto& shader_config = gpu.regs.shader_config[index];
         const auto program{static_cast<Maxwell::ShaderProgram>(index)};
 
         // Skip stages that are not enabled

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -161,7 +161,8 @@ public:
     explicit VKPipelineCache(Core::System& system, RasterizerVulkan& rasterizer,
                              const VKDevice& device, VKScheduler& scheduler,
                              VKDescriptorPool& descriptor_pool,
-                             VKUpdateDescriptorQueue& update_descriptor_queue);
+                             VKUpdateDescriptorQueue& update_descriptor_queue,
+                             VKRenderPassCache& renderpass_cache);
     ~VKPipelineCache();
 
     std::array<Shader, Maxwell::MaxShaderProgram> GetShaders();
@@ -184,8 +185,7 @@ private:
     VKScheduler& scheduler;
     VKDescriptorPool& descriptor_pool;
     VKUpdateDescriptorQueue& update_descriptor_queue;
-
-    VKRenderPassCache renderpass_cache;
+    VKRenderPassCache& renderpass_cache;
 
     std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
 

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -287,12 +287,13 @@ RasterizerVulkan::RasterizerVulkan(Core::System& system, Core::Frontend::EmuWind
       screen_info{screen_info}, device{device}, resource_manager{resource_manager},
       memory_manager{memory_manager}, state_tracker{state_tracker}, scheduler{scheduler},
       staging_pool(device, memory_manager, scheduler), descriptor_pool(device),
-      update_descriptor_queue(device, scheduler),
+      update_descriptor_queue(device, scheduler), renderpass_cache(device),
       quad_array_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
       uint8_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
       texture_cache(system, *this, device, resource_manager, memory_manager, scheduler,
                     staging_pool),
-      pipeline_cache(system, *this, device, scheduler, descriptor_pool, update_descriptor_queue),
+      pipeline_cache(system, *this, device, scheduler, descriptor_pool, update_descriptor_queue,
+                     renderpass_cache),
       buffer_cache(*this, system, device, memory_manager, scheduler, staging_pool),
       sampler_cache(device), query_cache(system, *this, device, scheduler) {
     scheduler.SetQueryCache(query_cache);
@@ -365,13 +366,16 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
 void RasterizerVulkan::Clear() {
     MICROPROFILE_SCOPE(Vulkan_Clearing);
 
-    query_cache.UpdateCounters();
-
     const auto& gpu = system.GPU().Maxwell3D();
     if (!system.GPU().Maxwell3D().ShouldExecute()) {
         return;
     }
 
+    sampled_views.clear();
+    image_views.clear();
+
+    query_cache.UpdateCounters();
+
     const auto& regs = gpu.regs;
     const bool use_color = regs.clear_buffers.R || regs.clear_buffers.G || regs.clear_buffers.B ||
                            regs.clear_buffers.A;
@@ -380,52 +384,54 @@ void RasterizerVulkan::Clear() {
     if (!use_color && !use_depth && !use_stencil) {
         return;
     }
-    // Clearing images requires to be out of a renderpass
-    scheduler.RequestOutsideRenderPassOperationContext();
 
-    // TODO(Rodrigo): Implement clears rendering a quad or using beginning a renderpass.
+    [[maybe_unused]] const auto texceptions = UpdateAttachments();
+    DEBUG_ASSERT(texceptions.none());
+    SetupImageTransitions(0, color_attachments, zeta_attachment);
+
+    const vk::RenderPass renderpass = renderpass_cache.GetRenderPass(GetRenderPassParams(0));
+    const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass);
+    scheduler.RequestRenderpass({renderpass, framebuffer, {{0, 0}, render_area}, 0, nullptr});
+
+    const auto& scissor = regs.scissor_test[0];
+    const vk::Offset2D scissor_offset(scissor.min_x, scissor.min_y);
+    vk::Extent2D scissor_extent{scissor.max_x - scissor.min_x, scissor.max_y - scissor.min_y};
+    scissor_extent.width = std::min(scissor_extent.width, render_area.width);
+    scissor_extent.height = std::min(scissor_extent.height, render_area.height);
+
+    const u32 layer = regs.clear_buffers.layer;
+    const vk::ClearRect clear_rect({scissor_offset, scissor_extent}, layer, 1);
 
     if (use_color) {
-        View color_view;
-        {
-            MICROPROFILE_SCOPE(Vulkan_RenderTargets);
-            color_view = texture_cache.GetColorBufferSurface(regs.clear_buffers.RT.Value(), false);
-        }
-
-        color_view->Transition(vk::ImageLayout::eTransferDstOptimal,
-                               vk::PipelineStageFlagBits::eTransfer,
-                               vk::AccessFlagBits::eTransferWrite);
-
         const std::array clear_color = {regs.clear_color[0], regs.clear_color[1],
                                         regs.clear_color[2], regs.clear_color[3]};
-        const vk::ClearColorValue clear(clear_color);
-        scheduler.Record([image = color_view->GetImage(),
-                          subresource = color_view->GetImageSubresourceRange(),
-                          clear](auto cmdbuf, auto& dld) {
-            cmdbuf.clearColorImage(image, vk::ImageLayout::eTransferDstOptimal, clear, subresource,
-                                   dld);
+        const vk::ClearValue clear_value{clear_color};
+        const u32 color_attachment = regs.clear_buffers.RT;
+        scheduler.Record([color_attachment, clear_value, clear_rect](auto cmdbuf, auto& dld) {
+            const vk::ClearAttachment attachment(vk::ImageAspectFlagBits::eColor, color_attachment,
+                                                 clear_value);
+            cmdbuf.clearAttachments(1, &attachment, 1, &clear_rect, dld);
         });
     }
-    if (use_depth || use_stencil) {
-        View zeta_surface;
-        {
-            MICROPROFILE_SCOPE(Vulkan_RenderTargets);
-            zeta_surface = texture_cache.GetDepthBufferSurface(false);
-        }
 
-        zeta_surface->Transition(vk::ImageLayout::eTransferDstOptimal,
-                                 vk::PipelineStageFlagBits::eTransfer,
-                                 vk::AccessFlagBits::eTransferWrite);
-
-        const vk::ClearDepthStencilValue clear(regs.clear_depth,
-                                               static_cast<u32>(regs.clear_stencil));
-        scheduler.Record([image = zeta_surface->GetImage(),
-                          subresource = zeta_surface->GetImageSubresourceRange(),
-                          clear](auto cmdbuf, auto& dld) {
-            cmdbuf.clearDepthStencilImage(image, vk::ImageLayout::eTransferDstOptimal, clear,
-                                          subresource, dld);
-        });
+    if (!use_depth && !use_stencil) {
+        return;
     }
+    vk::ImageAspectFlags aspect_flags;
+    if (use_depth) {
+        aspect_flags |= vk::ImageAspectFlagBits::eDepth;
+    }
+    if (use_stencil) {
+        aspect_flags |= vk::ImageAspectFlagBits::eStencil;
+    }
+
+    scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil,
+                      clear_rect, aspect_flags](auto cmdbuf, auto& dld) {
+        const vk::ClearDepthStencilValue clear_zeta(clear_depth, clear_stencil);
+        const vk::ClearValue clear_value{clear_zeta};
+        const vk::ClearAttachment attachment(aspect_flags, 0, clear_value);
+        cmdbuf.clearAttachments(1, &attachment, 1, &clear_rect, dld);
+    });
 }
 
 void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
@@ -542,8 +548,6 @@ bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config,
 
     // Verify that the cached surface is the same size and format as the requested framebuffer
     const auto& params{surface->GetSurfaceParams()};
-    const auto& pixel_format{
-        VideoCore::Surface::PixelFormatFromGPUPixelFormat(config.pixel_format)};
     ASSERT_MSG(params.width == config.width, "Framebuffer width is different");
     ASSERT_MSG(params.height == config.height, "Framebuffer height is different");
 
@@ -1151,7 +1155,7 @@ std::size_t RasterizerVulkan::CalculateVertexArraysSize() const {
         // This implementation assumes that all attributes are used in the shader.
         const GPUVAddr start{regs.vertex_array[index].StartAddress()};
         const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
-        DEBUG_ASSERT(end > start);
+        DEBUG_ASSERT(end >= start);
 
         size += (end - start + 1) * regs.vertex_array[index].enable;
     }

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -253,6 +253,7 @@ private:
     VKStagingBufferPool staging_pool;
     VKDescriptorPool descriptor_pool;
     VKUpdateDescriptorQueue update_descriptor_queue;
+    VKRenderPassCache renderpass_cache;
     QuadArrayPass quad_array_pass;
     Uint8Pass uint8_pass;
 

src/video_core/renderer_vulkan/vk_shader_decompiler.cpp

@@ -1941,7 +1941,11 @@ private:
         return {};
     }
 
-    Expression AtomicAdd(Operation operation) {
+    template <Id (Module::*func)(Id, Id, Id, Id, Id), Type result_type,
+              Type value_type = result_type>
+    Expression Atomic(Operation operation) {
+        const Id type_def = GetTypeDefinition(result_type);
+
         Id pointer;
         if (const auto smem = std::get_if<SmemNode>(&*operation[0])) {
             pointer = GetSharedMemoryPointer(*smem);
@@ -1949,14 +1953,15 @@ private:
             pointer = GetGlobalMemoryPointer(*gmem);
         } else {
             UNREACHABLE();
-            return {Constant(t_uint, 0), Type::Uint};
+            return {Constant(type_def, 0), result_type};
         }
 
-        const Id scope = Constant(t_uint, static_cast<u32>(spv::Scope::Device));
-        const Id semantics = Constant(t_uint, 0U);
+        const Id value = As(Visit(operation[1]), value_type);
 
-        const Id value = AsUint(Visit(operation[1]));
-        return {OpAtomicIAdd(t_uint, pointer, scope, semantics, value), Type::Uint};
+        const Id scope = Constant(t_uint, static_cast<u32>(spv::Scope::Device));
+        const Id semantics = Constant(type_def, 0);
+
+        return {(this->*func)(type_def, pointer, scope, semantics, value), result_type};
     }
 
     Expression Branch(Operation operation) {
@@ -2545,7 +2550,21 @@ private:
         &SPIRVDecompiler::AtomicImageXor,
         &SPIRVDecompiler::AtomicImageExchange,
 
-        &SPIRVDecompiler::AtomicAdd,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicExchange, Type::Uint>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicIAdd, Type::Uint>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicUMin, Type::Uint>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicUMax, Type::Uint>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicAnd, Type::Uint>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicOr, Type::Uint>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicXor, Type::Uint>,
+
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicExchange, Type::Int>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicIAdd, Type::Int>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicSMin, Type::Int>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicSMax, Type::Int>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicAnd, Type::Int>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicOr, Type::Int>,
+        &SPIRVDecompiler::Atomic<&Module::OpAtomicXor, Type::Int>,
 
         &SPIRVDecompiler::Branch,
         &SPIRVDecompiler::BranchIndirect,

src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp

@@ -100,7 +100,6 @@ void VKStagingBufferPool::ReleaseCache(bool host_visible) {
 }
 
 u64 VKStagingBufferPool::ReleaseLevel(StagingBuffersCache& cache, std::size_t log2) {
-    static constexpr u64 epochs_to_destroy = 180;
     static constexpr std::size_t deletions_per_tick = 16;
 
     auto& staging = cache[log2];
@@ -108,6 +107,7 @@ u64 VKStagingBufferPool::ReleaseLevel(StagingBuffersCache& cache, std::size_t lo
     const std::size_t old_size = entries.size();
 
     const auto is_deleteable = [this](const auto& entry) {
+        static constexpr u64 epochs_to_destroy = 180;
         return entry.last_epoch + epochs_to_destroy < epoch && !entry.watch.IsUsed();
     };
     const std::size_t begin_offset = staging.delete_index;

src/video_core/renderer_vulkan/vk_state_tracker.cpp

@@ -90,8 +90,6 @@ void StateTracker::Initialize() {
     SetupDirtyBlendConstants(tables);
     SetupDirtyDepthBounds(tables);
     SetupDirtyStencilProperties(tables);
-
-    SetupCommonOnWriteStores(dirty.on_write_stores);
 }
 
 void StateTracker::InvalidateCommandBufferState() {

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -52,6 +52,9 @@ vk::ImageType SurfaceTargetToImage(SurfaceTarget target) {
         return vk::ImageType::e2D;
     case SurfaceTarget::Texture3D:
         return vk::ImageType::e3D;
+    case SurfaceTarget::TextureBuffer:
+        UNREACHABLE();
+        return {};
     }
     UNREACHABLE_MSG("Unknown texture target={}", static_cast<u32>(target));
     return {};
@@ -273,7 +276,6 @@ void CachedSurface::UploadImage(const std::vector<u8>& staging_buffer) {
 
     for (u32 level = 0; level < params.num_levels; ++level) {
         vk::BufferImageCopy copy = GetBufferImageCopy(level);
-        const auto& dld = device.GetDispatchLoader();
         if (image->GetAspectMask() ==
             (vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil)) {
             vk::BufferImageCopy depth = copy;
@@ -422,7 +424,6 @@ void VKTextureCache::ImageCopy(Surface& src_surface, Surface& dst_surface,
         dst_base_layer, num_layers, copy_params.dest_level, 1, vk::PipelineStageFlagBits::eTransfer,
         vk::AccessFlagBits::eTransferWrite, vk::ImageLayout::eTransferDstOptimal);
 
-    const auto& dld{device.GetDispatchLoader()};
     const vk::ImageSubresourceLayers src_subresource(
         src_surface->GetAspectMask(), copy_params.source_level, copy_params.source_z, num_layers);
     const vk::ImageSubresourceLayers dst_subresource(
@@ -458,7 +459,6 @@ void VKTextureCache::ImageBlit(View& src_view, View& dst_view,
                              dst_view->GetImageSubresourceLayers(), {dst_top_left, dst_bot_right});
     const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear;
 
-    const auto& dld{device.GetDispatchLoader()};
     scheduler.Record([src_image = src_view->GetImage(), dst_image = dst_view->GetImage(), blit,
                       is_linear](auto cmdbuf, auto& dld) {
         cmdbuf.blitImage(src_image, vk::ImageLayout::eTransferSrcOptimal, dst_image,

src/video_core/renderer_vulkan/wrapper.cpp

@@ -0,0 +1,750 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <exception>
+#include <memory>
+#include <optional>
+#include <utility>
+#include <vector>
+
+#include "common/common_types.h"
+
+#include "video_core/renderer_vulkan/wrapper.h"
+
+namespace Vulkan::vk {
+
+namespace {
+
+template <typename T>
+bool Proc(T& result, const InstanceDispatch& dld, const char* proc_name,
+          VkInstance instance = nullptr) noexcept {
+    result = reinterpret_cast<T>(dld.vkGetInstanceProcAddr(instance, proc_name));
+    return result != nullptr;
+}
+
+template <typename T>
+void Proc(T& result, const DeviceDispatch& dld, const char* proc_name, VkDevice device) noexcept {
+    result = reinterpret_cast<T>(dld.vkGetDeviceProcAddr(device, proc_name));
+}
+
+void Load(VkDevice device, DeviceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name, device)
+    X(vkAcquireNextImageKHR);
+    X(vkAllocateCommandBuffers);
+    X(vkAllocateDescriptorSets);
+    X(vkAllocateMemory);
+    X(vkBeginCommandBuffer);
+    X(vkBindBufferMemory);
+    X(vkBindImageMemory);
+    X(vkCmdBeginQuery);
+    X(vkCmdBeginRenderPass);
+    X(vkCmdBeginTransformFeedbackEXT);
+    X(vkCmdBindDescriptorSets);
+    X(vkCmdBindIndexBuffer);
+    X(vkCmdBindPipeline);
+    X(vkCmdBindTransformFeedbackBuffersEXT);
+    X(vkCmdBindVertexBuffers);
+    X(vkCmdBlitImage);
+    X(vkCmdClearAttachments);
+    X(vkCmdCopyBuffer);
+    X(vkCmdCopyBufferToImage);
+    X(vkCmdCopyImage);
+    X(vkCmdCopyImageToBuffer);
+    X(vkCmdDispatch);
+    X(vkCmdDraw);
+    X(vkCmdDrawIndexed);
+    X(vkCmdEndQuery);
+    X(vkCmdEndRenderPass);
+    X(vkCmdEndTransformFeedbackEXT);
+    X(vkCmdFillBuffer);
+    X(vkCmdPipelineBarrier);
+    X(vkCmdPushConstants);
+    X(vkCmdSetBlendConstants);
+    X(vkCmdSetCheckpointNV);
+    X(vkCmdSetDepthBias);
+    X(vkCmdSetDepthBounds);
+    X(vkCmdSetScissor);
+    X(vkCmdSetStencilCompareMask);
+    X(vkCmdSetStencilReference);
+    X(vkCmdSetStencilWriteMask);
+    X(vkCmdSetViewport);
+    X(vkCreateBuffer);
+    X(vkCreateBufferView);
+    X(vkCreateCommandPool);
+    X(vkCreateComputePipelines);
+    X(vkCreateDescriptorPool);
+    X(vkCreateDescriptorSetLayout);
+    X(vkCreateDescriptorUpdateTemplateKHR);
+    X(vkCreateFence);
+    X(vkCreateFramebuffer);
+    X(vkCreateGraphicsPipelines);
+    X(vkCreateImage);
+    X(vkCreateImageView);
+    X(vkCreatePipelineLayout);
+    X(vkCreateQueryPool);
+    X(vkCreateRenderPass);
+    X(vkCreateSampler);
+    X(vkCreateSemaphore);
+    X(vkCreateShaderModule);
+    X(vkCreateSwapchainKHR);
+    X(vkDestroyBuffer);
+    X(vkDestroyBufferView);
+    X(vkDestroyCommandPool);
+    X(vkDestroyDescriptorPool);
+    X(vkDestroyDescriptorSetLayout);
+    X(vkDestroyDescriptorUpdateTemplateKHR);
+    X(vkDestroyFence);
+    X(vkDestroyFramebuffer);
+    X(vkDestroyImage);
+    X(vkDestroyImageView);
+    X(vkDestroyPipeline);
+    X(vkDestroyPipelineLayout);
+    X(vkDestroyQueryPool);
+    X(vkDestroyRenderPass);
+    X(vkDestroySampler);
+    X(vkDestroySemaphore);
+    X(vkDestroyShaderModule);
+    X(vkDestroySwapchainKHR);
+    X(vkDeviceWaitIdle);
+    X(vkEndCommandBuffer);
+    X(vkFreeCommandBuffers);
+    X(vkFreeDescriptorSets);
+    X(vkFreeMemory);
+    X(vkGetBufferMemoryRequirements);
+    X(vkGetDeviceQueue);
+    X(vkGetFenceStatus);
+    X(vkGetImageMemoryRequirements);
+    X(vkGetQueryPoolResults);
+    X(vkGetQueueCheckpointDataNV);
+    X(vkMapMemory);
+    X(vkQueueSubmit);
+    X(vkResetFences);
+    X(vkResetQueryPoolEXT);
+    X(vkUnmapMemory);
+    X(vkUpdateDescriptorSetWithTemplateKHR);
+    X(vkUpdateDescriptorSets);
+    X(vkWaitForFences);
+#undef X
+}
+
+} // Anonymous namespace
+
+bool Load(InstanceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name)
+    return X(vkCreateInstance) && X(vkEnumerateInstanceExtensionProperties);
+#undef X
+}
+
+bool Load(VkInstance instance, InstanceDispatch& dld) noexcept {
+#define X(name) Proc(dld.name, dld, #name, instance)
+    // These functions may fail to load depending on the enabled extensions.
+    // Don't return a failure on these.
+    X(vkCreateDebugUtilsMessengerEXT);
+    X(vkDestroyDebugUtilsMessengerEXT);
+    X(vkDestroySurfaceKHR);
+    X(vkGetPhysicalDeviceFeatures2KHR);
+    X(vkGetPhysicalDeviceProperties2KHR);
+    X(vkGetPhysicalDeviceSurfaceCapabilitiesKHR);
+    X(vkGetPhysicalDeviceSurfaceFormatsKHR);
+    X(vkGetPhysicalDeviceSurfacePresentModesKHR);
+    X(vkGetPhysicalDeviceSurfaceSupportKHR);
+    X(vkGetSwapchainImagesKHR);
+    X(vkQueuePresentKHR);
+
+    return X(vkCreateDevice) && X(vkDestroyDevice) && X(vkDestroyDevice) &&
+           X(vkEnumerateDeviceExtensionProperties) && X(vkEnumeratePhysicalDevices) &&
+           X(vkGetDeviceProcAddr) && X(vkGetPhysicalDeviceFormatProperties) &&
+           X(vkGetPhysicalDeviceMemoryProperties) && X(vkGetPhysicalDeviceProperties) &&
+           X(vkGetPhysicalDeviceQueueFamilyProperties);
+#undef X
+}
+
+const char* Exception::what() const noexcept {
+    return ToString(result);
+}
+
+const char* ToString(VkResult result) noexcept {
+    switch (result) {
+    case VkResult::VK_SUCCESS:
+        return "VK_SUCCESS";
+    case VkResult::VK_NOT_READY:
+        return "VK_NOT_READY";
+    case VkResult::VK_TIMEOUT:
+        return "VK_TIMEOUT";
+    case VkResult::VK_EVENT_SET:
+        return "VK_EVENT_SET";
+    case VkResult::VK_EVENT_RESET:
+        return "VK_EVENT_RESET";
+    case VkResult::VK_INCOMPLETE:
+        return "VK_INCOMPLETE";
+    case VkResult::VK_ERROR_OUT_OF_HOST_MEMORY:
+        return "VK_ERROR_OUT_OF_HOST_MEMORY";
+    case VkResult::VK_ERROR_OUT_OF_DEVICE_MEMORY:
+        return "VK_ERROR_OUT_OF_DEVICE_MEMORY";
+    case VkResult::VK_ERROR_INITIALIZATION_FAILED:
+        return "VK_ERROR_INITIALIZATION_FAILED";
+    case VkResult::VK_ERROR_DEVICE_LOST:
+        return "VK_ERROR_DEVICE_LOST";
+    case VkResult::VK_ERROR_MEMORY_MAP_FAILED:
+        return "VK_ERROR_MEMORY_MAP_FAILED";
+    case VkResult::VK_ERROR_LAYER_NOT_PRESENT:
+        return "VK_ERROR_LAYER_NOT_PRESENT";
+    case VkResult::VK_ERROR_EXTENSION_NOT_PRESENT:
+        return "VK_ERROR_EXTENSION_NOT_PRESENT";
+    case VkResult::VK_ERROR_FEATURE_NOT_PRESENT:
+        return "VK_ERROR_FEATURE_NOT_PRESENT";
+    case VkResult::VK_ERROR_INCOMPATIBLE_DRIVER:
+        return "VK_ERROR_INCOMPATIBLE_DRIVER";
+    case VkResult::VK_ERROR_TOO_MANY_OBJECTS:
+        return "VK_ERROR_TOO_MANY_OBJECTS";
+    case VkResult::VK_ERROR_FORMAT_NOT_SUPPORTED:
+        return "VK_ERROR_FORMAT_NOT_SUPPORTED";
+    case VkResult::VK_ERROR_FRAGMENTED_POOL:
+        return "VK_ERROR_FRAGMENTED_POOL";
+    case VkResult::VK_ERROR_OUT_OF_POOL_MEMORY:
+        return "VK_ERROR_OUT_OF_POOL_MEMORY";
+    case VkResult::VK_ERROR_INVALID_EXTERNAL_HANDLE:
+        return "VK_ERROR_INVALID_EXTERNAL_HANDLE";
+    case VkResult::VK_ERROR_SURFACE_LOST_KHR:
+        return "VK_ERROR_SURFACE_LOST_KHR";
+    case VkResult::VK_ERROR_NATIVE_WINDOW_IN_USE_KHR:
+        return "VK_ERROR_NATIVE_WINDOW_IN_USE_KHR";
+    case VkResult::VK_SUBOPTIMAL_KHR:
+        return "VK_SUBOPTIMAL_KHR";
+    case VkResult::VK_ERROR_OUT_OF_DATE_KHR:
+        return "VK_ERROR_OUT_OF_DATE_KHR";
+    case VkResult::VK_ERROR_INCOMPATIBLE_DISPLAY_KHR:
+        return "VK_ERROR_INCOMPATIBLE_DISPLAY_KHR";
+    case VkResult::VK_ERROR_VALIDATION_FAILED_EXT:
+        return "VK_ERROR_VALIDATION_FAILED_EXT";
+    case VkResult::VK_ERROR_INVALID_SHADER_NV:
+        return "VK_ERROR_INVALID_SHADER_NV";
+    case VkResult::VK_ERROR_INVALID_DRM_FORMAT_MODIFIER_PLANE_LAYOUT_EXT:
+        return "VK_ERROR_INVALID_DRM_FORMAT_MODIFIER_PLANE_LAYOUT_EXT";
+    case VkResult::VK_ERROR_FRAGMENTATION_EXT:
+        return "VK_ERROR_FRAGMENTATION_EXT";
+    case VkResult::VK_ERROR_NOT_PERMITTED_EXT:
+        return "VK_ERROR_NOT_PERMITTED_EXT";
+    case VkResult::VK_ERROR_INVALID_DEVICE_ADDRESS_EXT:
+        return "VK_ERROR_INVALID_DEVICE_ADDRESS_EXT";
+    case VkResult::VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT:
+        return "VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT";
+    }
+    return "Unknown";
+}
+
+void Destroy(VkInstance instance, const InstanceDispatch& dld) noexcept {
+    dld.vkDestroyInstance(instance, nullptr);
+}
+
+void Destroy(VkDevice device, const InstanceDispatch& dld) noexcept {
+    dld.vkDestroyDevice(device, nullptr);
+}
+
+void Destroy(VkDevice device, VkBuffer handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyBuffer(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkBufferView handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyBufferView(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkCommandPool handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyCommandPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorPool handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyDescriptorPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorSetLayout handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyDescriptorSetLayout(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDescriptorUpdateTemplateKHR handle,
+             const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyDescriptorUpdateTemplateKHR(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkDeviceMemory handle, const DeviceDispatch& dld) noexcept {
+    dld.vkFreeMemory(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkFence handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyFence(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkFramebuffer handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyFramebuffer(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkImage handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyImage(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkImageView handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyImageView(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkPipeline handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyPipeline(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkPipelineLayout handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyPipelineLayout(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkQueryPool handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyQueryPool(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkRenderPass handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyRenderPass(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSampler handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroySampler(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSwapchainKHR handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroySwapchainKHR(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkSemaphore handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroySemaphore(device, handle, nullptr);
+}
+
+void Destroy(VkDevice device, VkShaderModule handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyShaderModule(device, handle, nullptr);
+}
+
+void Destroy(VkInstance instance, VkDebugUtilsMessengerEXT handle,
+             const InstanceDispatch& dld) noexcept {
+    dld.vkDestroyDebugUtilsMessengerEXT(instance, handle, nullptr);
+}
+
+void Destroy(VkInstance instance, VkSurfaceKHR handle, const InstanceDispatch& dld) noexcept {
+    dld.vkDestroySurfaceKHR(instance, handle, nullptr);
+}
+
+VkResult Free(VkDevice device, VkDescriptorPool handle, Span<VkDescriptorSet> sets,
+              const DeviceDispatch& dld) noexcept {
+    return dld.vkFreeDescriptorSets(device, handle, sets.size(), sets.data());
+}
+
+VkResult Free(VkDevice device, VkCommandPool handle, Span<VkCommandBuffer> buffers,
+              const DeviceDispatch& dld) noexcept {
+    dld.vkFreeCommandBuffers(device, handle, buffers.size(), buffers.data());
+    return VK_SUCCESS;
+}
+
+Instance Instance::Create(Span<const char*> layers, Span<const char*> extensions,
+                          InstanceDispatch& dld) noexcept {
+    VkApplicationInfo application_info;
+    application_info.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO;
+    application_info.pNext = nullptr;
+    application_info.pApplicationName = "yuzu Emulator";
+    application_info.applicationVersion = VK_MAKE_VERSION(0, 1, 0);
+    application_info.pEngineName = "yuzu Emulator";
+    application_info.engineVersion = VK_MAKE_VERSION(0, 1, 0);
+    application_info.apiVersion = VK_API_VERSION_1_1;
+
+    VkInstanceCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.pApplicationInfo = &application_info;
+    ci.enabledLayerCount = layers.size();
+    ci.ppEnabledLayerNames = layers.data();
+    ci.enabledExtensionCount = extensions.size();
+    ci.ppEnabledExtensionNames = extensions.data();
+
+    VkInstance instance;
+    if (dld.vkCreateInstance(&ci, nullptr, &instance) != VK_SUCCESS) {
+        // Failed to create the instance.
+        return {};
+    }
+    if (!Proc(dld.vkDestroyInstance, dld, "vkDestroyInstance", instance)) {
+        // We successfully created an instance but the destroy function couldn't be loaded.
+        // This is a good moment to panic.
+        return {};
+    }
+
+    return Instance(instance, dld);
+}
+
+std::optional<std::vector<VkPhysicalDevice>> Instance::EnumeratePhysicalDevices() {
+    u32 num;
+    if (dld->vkEnumeratePhysicalDevices(handle, &num, nullptr) != VK_SUCCESS) {
+        return std::nullopt;
+    }
+    std::vector<VkPhysicalDevice> physical_devices(num);
+    if (dld->vkEnumeratePhysicalDevices(handle, &num, physical_devices.data()) != VK_SUCCESS) {
+        return std::nullopt;
+    }
+    return physical_devices;
+}
+
+DebugCallback Instance::TryCreateDebugCallback(
+    PFN_vkDebugUtilsMessengerCallbackEXT callback) noexcept {
+    VkDebugUtilsMessengerCreateInfoEXT ci;
+    ci.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.messageSeverity = VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT |
+                         VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT |
+                         VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT |
+                         VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT;
+    ci.messageType = VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT |
+                     VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT |
+                     VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT;
+    ci.pfnUserCallback = callback;
+    ci.pUserData = nullptr;
+
+    VkDebugUtilsMessengerEXT messenger;
+    if (dld->vkCreateDebugUtilsMessengerEXT(handle, &ci, nullptr, &messenger) != VK_SUCCESS) {
+        return {};
+    }
+    return DebugCallback(messenger, handle, *dld);
+}
+
+std::vector<VkCheckpointDataNV> Queue::GetCheckpointDataNV(const DeviceDispatch& dld) const {
+    if (!dld.vkGetQueueCheckpointDataNV) {
+        return {};
+    }
+    u32 num;
+    dld.vkGetQueueCheckpointDataNV(queue, &num, nullptr);
+    std::vector<VkCheckpointDataNV> checkpoints(num);
+    dld.vkGetQueueCheckpointDataNV(queue, &num, checkpoints.data());
+    return checkpoints;
+}
+
+void Buffer::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
+    Check(dld->vkBindBufferMemory(owner, handle, memory, offset));
+}
+
+void Image::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
+    Check(dld->vkBindImageMemory(owner, handle, memory, offset));
+}
+
+DescriptorSets DescriptorPool::Allocate(const VkDescriptorSetAllocateInfo& ai) const {
+    const std::size_t num = ai.descriptorSetCount;
+    std::unique_ptr sets = std::make_unique<VkDescriptorSet[]>(num);
+    switch (const VkResult result = dld->vkAllocateDescriptorSets(owner, &ai, sets.get())) {
+    case VK_SUCCESS:
+        return DescriptorSets(std::move(sets), num, owner, handle, *dld);
+    case VK_ERROR_OUT_OF_POOL_MEMORY:
+        return {};
+    default:
+        throw Exception(result);
+    }
+}
+
+CommandBuffers CommandPool::Allocate(std::size_t num_buffers, VkCommandBufferLevel level) const {
+    VkCommandBufferAllocateInfo ai;
+    ai.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
+    ai.pNext = nullptr;
+    ai.commandPool = handle;
+    ai.level = level;
+    ai.commandBufferCount = static_cast<u32>(num_buffers);
+
+    std::unique_ptr buffers = std::make_unique<VkCommandBuffer[]>(num_buffers);
+    switch (const VkResult result = dld->vkAllocateCommandBuffers(owner, &ai, buffers.get())) {
+    case VK_SUCCESS:
+        return CommandBuffers(std::move(buffers), num_buffers, owner, handle, *dld);
+    case VK_ERROR_OUT_OF_POOL_MEMORY:
+        return {};
+    default:
+        throw Exception(result);
+    }
+}
+
+std::vector<VkImage> SwapchainKHR::GetImages() const {
+    u32 num;
+    Check(dld->vkGetSwapchainImagesKHR(owner, handle, &num, nullptr));
+    std::vector<VkImage> images(num);
+    Check(dld->vkGetSwapchainImagesKHR(owner, handle, &num, images.data()));
+    return images;
+}
+
+Device Device::Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
+                      Span<const char*> enabled_extensions,
+                      const VkPhysicalDeviceFeatures2& enabled_features,
+                      DeviceDispatch& dld) noexcept {
+    VkDeviceCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO;
+    ci.pNext = &enabled_features;
+    ci.flags = 0;
+    ci.queueCreateInfoCount = queues_ci.size();
+    ci.pQueueCreateInfos = queues_ci.data();
+    ci.enabledLayerCount = 0;
+    ci.ppEnabledLayerNames = nullptr;
+    ci.enabledExtensionCount = enabled_extensions.size();
+    ci.ppEnabledExtensionNames = enabled_extensions.data();
+    ci.pEnabledFeatures = nullptr;
+
+    VkDevice device;
+    if (dld.vkCreateDevice(physical_device, &ci, nullptr, &device) != VK_SUCCESS) {
+        return {};
+    }
+    Load(device, dld);
+    return Device(device, dld);
+}
+
+Queue Device::GetQueue(u32 family_index) const noexcept {
+    VkQueue queue;
+    dld->vkGetDeviceQueue(handle, family_index, 0, &queue);
+    return Queue(queue, *dld);
+}
+
+Buffer Device::CreateBuffer(const VkBufferCreateInfo& ci) const {
+    VkBuffer object;
+    Check(dld->vkCreateBuffer(handle, &ci, nullptr, &object));
+    return Buffer(object, handle, *dld);
+}
+
+BufferView Device::CreateBufferView(const VkBufferViewCreateInfo& ci) const {
+    VkBufferView object;
+    Check(dld->vkCreateBufferView(handle, &ci, nullptr, &object));
+    return BufferView(object, handle, *dld);
+}
+
+Image Device::CreateImage(const VkImageCreateInfo& ci) const {
+    VkImage object;
+    Check(dld->vkCreateImage(handle, &ci, nullptr, &object));
+    return Image(object, handle, *dld);
+}
+
+ImageView Device::CreateImageView(const VkImageViewCreateInfo& ci) const {
+    VkImageView object;
+    Check(dld->vkCreateImageView(handle, &ci, nullptr, &object));
+    return ImageView(object, handle, *dld);
+}
+
+Semaphore Device::CreateSemaphore() const {
+    VkSemaphoreCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+
+    VkSemaphore object;
+    Check(dld->vkCreateSemaphore(handle, &ci, nullptr, &object));
+    return Semaphore(object, handle, *dld);
+}
+
+Fence Device::CreateFence(const VkFenceCreateInfo& ci) const {
+    VkFence object;
+    Check(dld->vkCreateFence(handle, &ci, nullptr, &object));
+    return Fence(object, handle, *dld);
+}
+
+DescriptorPool Device::CreateDescriptorPool(const VkDescriptorPoolCreateInfo& ci) const {
+    VkDescriptorPool object;
+    Check(dld->vkCreateDescriptorPool(handle, &ci, nullptr, &object));
+    return DescriptorPool(object, handle, *dld);
+}
+
+RenderPass Device::CreateRenderPass(const VkRenderPassCreateInfo& ci) const {
+    VkRenderPass object;
+    Check(dld->vkCreateRenderPass(handle, &ci, nullptr, &object));
+    return RenderPass(object, handle, *dld);
+}
+
+DescriptorSetLayout Device::CreateDescriptorSetLayout(
+    const VkDescriptorSetLayoutCreateInfo& ci) const {
+    VkDescriptorSetLayout object;
+    Check(dld->vkCreateDescriptorSetLayout(handle, &ci, nullptr, &object));
+    return DescriptorSetLayout(object, handle, *dld);
+}
+
+PipelineLayout Device::CreatePipelineLayout(const VkPipelineLayoutCreateInfo& ci) const {
+    VkPipelineLayout object;
+    Check(dld->vkCreatePipelineLayout(handle, &ci, nullptr, &object));
+    return PipelineLayout(object, handle, *dld);
+}
+
+Pipeline Device::CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci) const {
+    VkPipeline object;
+    Check(dld->vkCreateGraphicsPipelines(handle, nullptr, 1, &ci, nullptr, &object));
+    return Pipeline(object, handle, *dld);
+}
+
+Pipeline Device::CreateComputePipeline(const VkComputePipelineCreateInfo& ci) const {
+    VkPipeline object;
+    Check(dld->vkCreateComputePipelines(handle, nullptr, 1, &ci, nullptr, &object));
+    return Pipeline(object, handle, *dld);
+}
+
+Sampler Device::CreateSampler(const VkSamplerCreateInfo& ci) const {
+    VkSampler object;
+    Check(dld->vkCreateSampler(handle, &ci, nullptr, &object));
+    return Sampler(object, handle, *dld);
+}
+
+Framebuffer Device::CreateFramebuffer(const VkFramebufferCreateInfo& ci) const {
+    VkFramebuffer object;
+    Check(dld->vkCreateFramebuffer(handle, &ci, nullptr, &object));
+    return Framebuffer(object, handle, *dld);
+}
+
+CommandPool Device::CreateCommandPool(const VkCommandPoolCreateInfo& ci) const {
+    VkCommandPool object;
+    Check(dld->vkCreateCommandPool(handle, &ci, nullptr, &object));
+    return CommandPool(object, handle, *dld);
+}
+
+DescriptorUpdateTemplateKHR Device::CreateDescriptorUpdateTemplateKHR(
+    const VkDescriptorUpdateTemplateCreateInfoKHR& ci) const {
+    VkDescriptorUpdateTemplateKHR object;
+    Check(dld->vkCreateDescriptorUpdateTemplateKHR(handle, &ci, nullptr, &object));
+    return DescriptorUpdateTemplateKHR(object, handle, *dld);
+}
+
+QueryPool Device::CreateQueryPool(const VkQueryPoolCreateInfo& ci) const {
+    VkQueryPool object;
+    Check(dld->vkCreateQueryPool(handle, &ci, nullptr, &object));
+    return QueryPool(object, handle, *dld);
+}
+
+ShaderModule Device::CreateShaderModule(const VkShaderModuleCreateInfo& ci) const {
+    VkShaderModule object;
+    Check(dld->vkCreateShaderModule(handle, &ci, nullptr, &object));
+    return ShaderModule(object, handle, *dld);
+}
+
+SwapchainKHR Device::CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const {
+    VkSwapchainKHR object;
+    Check(dld->vkCreateSwapchainKHR(handle, &ci, nullptr, &object));
+    return SwapchainKHR(object, handle, *dld);
+}
+
+DeviceMemory Device::TryAllocateMemory(const VkMemoryAllocateInfo& ai) const noexcept {
+    VkDeviceMemory memory;
+    if (dld->vkAllocateMemory(handle, &ai, nullptr, &memory) != VK_SUCCESS) {
+        return {};
+    }
+    return DeviceMemory(memory, handle, *dld);
+}
+
+DeviceMemory Device::AllocateMemory(const VkMemoryAllocateInfo& ai) const {
+    VkDeviceMemory memory;
+    Check(dld->vkAllocateMemory(handle, &ai, nullptr, &memory));
+    return DeviceMemory(memory, handle, *dld);
+}
+
+VkMemoryRequirements Device::GetBufferMemoryRequirements(VkBuffer buffer) const noexcept {
+    VkMemoryRequirements requirements;
+    dld->vkGetBufferMemoryRequirements(handle, buffer, &requirements);
+    return requirements;
+}
+
+VkMemoryRequirements Device::GetImageMemoryRequirements(VkImage image) const noexcept {
+    VkMemoryRequirements requirements;
+    dld->vkGetImageMemoryRequirements(handle, image, &requirements);
+    return requirements;
+}
+
+void Device::UpdateDescriptorSets(Span<VkWriteDescriptorSet> writes,
+                                  Span<VkCopyDescriptorSet> copies) const noexcept {
+    dld->vkUpdateDescriptorSets(handle, writes.size(), writes.data(), copies.size(), copies.data());
+}
+
+VkPhysicalDeviceProperties PhysicalDevice::GetProperties() const noexcept {
+    VkPhysicalDeviceProperties properties;
+    dld->vkGetPhysicalDeviceProperties(physical_device, &properties);
+    return properties;
+}
+
+void PhysicalDevice::GetProperties2KHR(VkPhysicalDeviceProperties2KHR& properties) const noexcept {
+    dld->vkGetPhysicalDeviceProperties2KHR(physical_device, &properties);
+}
+
+VkPhysicalDeviceFeatures PhysicalDevice::GetFeatures() const noexcept {
+    VkPhysicalDeviceFeatures2KHR features2;
+    features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2_KHR;
+    features2.pNext = nullptr;
+    dld->vkGetPhysicalDeviceFeatures2KHR(physical_device, &features2);
+    return features2.features;
+}
+
+void PhysicalDevice::GetFeatures2KHR(VkPhysicalDeviceFeatures2KHR& features) const noexcept {
+    dld->vkGetPhysicalDeviceFeatures2KHR(physical_device, &features);
+}
+
+VkFormatProperties PhysicalDevice::GetFormatProperties(VkFormat format) const noexcept {
+    VkFormatProperties properties;
+    dld->vkGetPhysicalDeviceFormatProperties(physical_device, format, &properties);
+    return properties;
+}
+
+std::vector<VkExtensionProperties> PhysicalDevice::EnumerateDeviceExtensionProperties() const {
+    u32 num;
+    dld->vkEnumerateDeviceExtensionProperties(physical_device, nullptr, &num, nullptr);
+    std::vector<VkExtensionProperties> properties(num);
+    dld->vkEnumerateDeviceExtensionProperties(physical_device, nullptr, &num, properties.data());
+    return properties;
+}
+
+std::vector<VkQueueFamilyProperties> PhysicalDevice::GetQueueFamilyProperties() const {
+    u32 num;
+    dld->vkGetPhysicalDeviceQueueFamilyProperties(physical_device, &num, nullptr);
+    std::vector<VkQueueFamilyProperties> properties(num);
+    dld->vkGetPhysicalDeviceQueueFamilyProperties(physical_device, &num, properties.data());
+    return properties;
+}
+
+bool PhysicalDevice::GetSurfaceSupportKHR(u32 queue_family_index, VkSurfaceKHR surface) const {
+    VkBool32 supported;
+    Check(dld->vkGetPhysicalDeviceSurfaceSupportKHR(physical_device, queue_family_index, surface,
+                                                    &supported));
+    return supported == VK_TRUE;
+}
+
+VkSurfaceCapabilitiesKHR PhysicalDevice::GetSurfaceCapabilitiesKHR(VkSurfaceKHR surface) const
+    noexcept {
+    VkSurfaceCapabilitiesKHR capabilities;
+    Check(dld->vkGetPhysicalDeviceSurfaceCapabilitiesKHR(physical_device, surface, &capabilities));
+    return capabilities;
+}
+
+std::vector<VkSurfaceFormatKHR> PhysicalDevice::GetSurfaceFormatsKHR(VkSurfaceKHR surface) const {
+    u32 num;
+    Check(dld->vkGetPhysicalDeviceSurfaceFormatsKHR(physical_device, surface, &num, nullptr));
+    std::vector<VkSurfaceFormatKHR> formats(num);
+    Check(
+        dld->vkGetPhysicalDeviceSurfaceFormatsKHR(physical_device, surface, &num, formats.data()));
+    return formats;
+}
+
+std::vector<VkPresentModeKHR> PhysicalDevice::GetSurfacePresentModesKHR(
+    VkSurfaceKHR surface) const {
+    u32 num;
+    Check(dld->vkGetPhysicalDeviceSurfacePresentModesKHR(physical_device, surface, &num, nullptr));
+    std::vector<VkPresentModeKHR> modes(num);
+    Check(dld->vkGetPhysicalDeviceSurfacePresentModesKHR(physical_device, surface, &num,
+                                                         modes.data()));
+    return modes;
+}
+
+VkPhysicalDeviceMemoryProperties PhysicalDevice::GetMemoryProperties() const noexcept {
+    VkPhysicalDeviceMemoryProperties properties;
+    dld->vkGetPhysicalDeviceMemoryProperties(physical_device, &properties);
+    return properties;
+}
+
+std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProperties(
+    const InstanceDispatch& dld) {
+    u32 num;
+    if (dld.vkEnumerateInstanceExtensionProperties(nullptr, &num, nullptr) != VK_SUCCESS) {
+        return std::nullopt;
+    }
+    std::vector<VkExtensionProperties> properties(num);
+    if (dld.vkEnumerateInstanceExtensionProperties(nullptr, &num, properties.data()) !=
+        VK_SUCCESS) {
+        return std::nullopt;
+    }
+    return properties;
+}
+
+} // namespace Vulkan::vk

src/video_core/renderer_vulkan/wrapper.h

@@ -0,0 +1,987 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <exception>
+#include <iterator>
+#include <limits>
+#include <memory>
+#include <optional>
+#include <type_traits>
+#include <utility>
+#include <vector>
+
+#define VK_NO_PROTOTYPES
+#include <vulkan/vulkan.h>
+
+#include "common/common_types.h"
+
+namespace Vulkan::vk {
+
+/**
+ * Span for Vulkan arrays.
+ * Based on std::span but optimized for array access instead of iterators.
+ * Size returns uint32_t instead of size_t to ease interaction with Vulkan functions.
+ */
+template <typename T>
+class Span {
+public:
+    using value_type = T;
+    using size_type = u32;
+    using difference_type = std::ptrdiff_t;
+    using reference = const T&;
+    using const_reference = const T&;
+    using pointer = const T*;
+    using const_pointer = const T*;
+    using iterator = const T*;
+    using const_iterator = const T*;
+
+    /// Construct an empty span.
+    constexpr Span() noexcept = default;
+
+    /// Construct a span from a single element.
+    constexpr Span(const T& value) noexcept : ptr{&value}, num{1} {}
+
+    /// Construct a span from a range.
+    template <typename Range>
+    // requires std::data(const Range&)
+    // requires std::size(const Range&)
+    constexpr Span(const Range& range) : ptr{std::data(range)}, num{std::size(range)} {}
+
+    /// Construct a span from a pointer and a size.
+    /// This is inteded for subranges.
+    constexpr Span(const T* ptr, std::size_t num) noexcept : ptr{ptr}, num{num} {}
+
+    /// Returns the data pointer by the span.
+    constexpr const T* data() const noexcept {
+        return ptr;
+    }
+
+    /// Returns the number of elements in the span.
+    /// @note Returns a 32 bits integer because most Vulkan functions expect this type.
+    constexpr u32 size() const noexcept {
+        return static_cast<u32>(num);
+    }
+
+    /// Returns true when the span is empty.
+    constexpr bool empty() const noexcept {
+        return num == 0;
+    }
+
+    /// Returns a reference to the element in the passed index.
+    /// @pre: index < size()
+    constexpr const T& operator[](std::size_t index) const noexcept {
+        return ptr[index];
+    }
+
+    /// Returns an iterator to the beginning of the span.
+    constexpr const T* begin() const noexcept {
+        return ptr;
+    }
+
+    /// Returns an iterator to the end of the span.
+    constexpr const T* end() const noexcept {
+        return ptr + num;
+    }
+
+    /// Returns an iterator to the beginning of the span.
+    constexpr const T* cbegin() const noexcept {
+        return ptr;
+    }
+
+    /// Returns an iterator to the end of the span.
+    constexpr const T* cend() const noexcept {
+        return ptr + num;
+    }
+
+private:
+    const T* ptr = nullptr;
+    std::size_t num = 0;
+};
+
+/// Vulkan exception generated from a VkResult.
+class Exception final : public std::exception {
+public:
+    /// Construct the exception with a result.
+    /// @pre result != VK_SUCCESS
+    explicit Exception(VkResult result_) : result{result_} {}
+    virtual ~Exception() = default;
+
+    const char* what() const noexcept override;
+
+private:
+    VkResult result;
+};
+
+/// Converts a VkResult enum into a rodata string
+const char* ToString(VkResult) noexcept;
+
+/// Throws a Vulkan exception if result is not success.
+inline void Check(VkResult result) {
+    if (result != VK_SUCCESS) {
+        throw Exception(result);
+    }
+}
+
+/// Throws a Vulkan exception if result is an error.
+/// @return result
+inline VkResult Filter(VkResult result) {
+    if (result < 0) {
+        throw Exception(result);
+    }
+    return result;
+}
+
+/// Table holding Vulkan instance function pointers.
+struct InstanceDispatch {
+    PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr;
+
+    PFN_vkCreateInstance vkCreateInstance;
+    PFN_vkDestroyInstance vkDestroyInstance;
+    PFN_vkEnumerateInstanceExtensionProperties vkEnumerateInstanceExtensionProperties;
+
+    PFN_vkCreateDebugUtilsMessengerEXT vkCreateDebugUtilsMessengerEXT;
+    PFN_vkCreateDevice vkCreateDevice;
+    PFN_vkDestroyDebugUtilsMessengerEXT vkDestroyDebugUtilsMessengerEXT;
+    PFN_vkDestroyDevice vkDestroyDevice;
+    PFN_vkDestroySurfaceKHR vkDestroySurfaceKHR;
+    PFN_vkEnumerateDeviceExtensionProperties vkEnumerateDeviceExtensionProperties;
+    PFN_vkEnumeratePhysicalDevices vkEnumeratePhysicalDevices;
+    PFN_vkGetDeviceProcAddr vkGetDeviceProcAddr;
+    PFN_vkGetPhysicalDeviceFeatures2KHR vkGetPhysicalDeviceFeatures2KHR;
+    PFN_vkGetPhysicalDeviceFormatProperties vkGetPhysicalDeviceFormatProperties;
+    PFN_vkGetPhysicalDeviceMemoryProperties vkGetPhysicalDeviceMemoryProperties;
+    PFN_vkGetPhysicalDeviceProperties vkGetPhysicalDeviceProperties;
+    PFN_vkGetPhysicalDeviceProperties2KHR vkGetPhysicalDeviceProperties2KHR;
+    PFN_vkGetPhysicalDeviceQueueFamilyProperties vkGetPhysicalDeviceQueueFamilyProperties;
+    PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR vkGetPhysicalDeviceSurfaceCapabilitiesKHR;
+    PFN_vkGetPhysicalDeviceSurfaceFormatsKHR vkGetPhysicalDeviceSurfaceFormatsKHR;
+    PFN_vkGetPhysicalDeviceSurfacePresentModesKHR vkGetPhysicalDeviceSurfacePresentModesKHR;
+    PFN_vkGetPhysicalDeviceSurfaceSupportKHR vkGetPhysicalDeviceSurfaceSupportKHR;
+    PFN_vkGetSwapchainImagesKHR vkGetSwapchainImagesKHR;
+    PFN_vkQueuePresentKHR vkQueuePresentKHR;
+};
+
+/// Table holding Vulkan device function pointers.
+struct DeviceDispatch : public InstanceDispatch {
+    PFN_vkAcquireNextImageKHR vkAcquireNextImageKHR;
+    PFN_vkAllocateCommandBuffers vkAllocateCommandBuffers;
+    PFN_vkAllocateDescriptorSets vkAllocateDescriptorSets;
+    PFN_vkAllocateMemory vkAllocateMemory;
+    PFN_vkBeginCommandBuffer vkBeginCommandBuffer;
+    PFN_vkBindBufferMemory vkBindBufferMemory;
+    PFN_vkBindImageMemory vkBindImageMemory;
+    PFN_vkCmdBeginQuery vkCmdBeginQuery;
+    PFN_vkCmdBeginRenderPass vkCmdBeginRenderPass;
+    PFN_vkCmdBeginTransformFeedbackEXT vkCmdBeginTransformFeedbackEXT;
+    PFN_vkCmdBindDescriptorSets vkCmdBindDescriptorSets;
+    PFN_vkCmdBindIndexBuffer vkCmdBindIndexBuffer;
+    PFN_vkCmdBindPipeline vkCmdBindPipeline;
+    PFN_vkCmdBindTransformFeedbackBuffersEXT vkCmdBindTransformFeedbackBuffersEXT;
+    PFN_vkCmdBindVertexBuffers vkCmdBindVertexBuffers;
+    PFN_vkCmdBlitImage vkCmdBlitImage;
+    PFN_vkCmdClearAttachments vkCmdClearAttachments;
+    PFN_vkCmdCopyBuffer vkCmdCopyBuffer;
+    PFN_vkCmdCopyBufferToImage vkCmdCopyBufferToImage;
+    PFN_vkCmdCopyImage vkCmdCopyImage;
+    PFN_vkCmdCopyImageToBuffer vkCmdCopyImageToBuffer;
+    PFN_vkCmdDispatch vkCmdDispatch;
+    PFN_vkCmdDraw vkCmdDraw;
+    PFN_vkCmdDrawIndexed vkCmdDrawIndexed;
+    PFN_vkCmdEndQuery vkCmdEndQuery;
+    PFN_vkCmdEndRenderPass vkCmdEndRenderPass;
+    PFN_vkCmdEndTransformFeedbackEXT vkCmdEndTransformFeedbackEXT;
+    PFN_vkCmdFillBuffer vkCmdFillBuffer;
+    PFN_vkCmdPipelineBarrier vkCmdPipelineBarrier;
+    PFN_vkCmdPushConstants vkCmdPushConstants;
+    PFN_vkCmdSetBlendConstants vkCmdSetBlendConstants;
+    PFN_vkCmdSetCheckpointNV vkCmdSetCheckpointNV;
+    PFN_vkCmdSetDepthBias vkCmdSetDepthBias;
+    PFN_vkCmdSetDepthBounds vkCmdSetDepthBounds;
+    PFN_vkCmdSetScissor vkCmdSetScissor;
+    PFN_vkCmdSetStencilCompareMask vkCmdSetStencilCompareMask;
+    PFN_vkCmdSetStencilReference vkCmdSetStencilReference;
+    PFN_vkCmdSetStencilWriteMask vkCmdSetStencilWriteMask;
+    PFN_vkCmdSetViewport vkCmdSetViewport;
+    PFN_vkCreateBuffer vkCreateBuffer;
+    PFN_vkCreateBufferView vkCreateBufferView;
+    PFN_vkCreateCommandPool vkCreateCommandPool;
+    PFN_vkCreateComputePipelines vkCreateComputePipelines;
+    PFN_vkCreateDescriptorPool vkCreateDescriptorPool;
+    PFN_vkCreateDescriptorSetLayout vkCreateDescriptorSetLayout;
+    PFN_vkCreateDescriptorUpdateTemplateKHR vkCreateDescriptorUpdateTemplateKHR;
+    PFN_vkCreateFence vkCreateFence;
+    PFN_vkCreateFramebuffer vkCreateFramebuffer;
+    PFN_vkCreateGraphicsPipelines vkCreateGraphicsPipelines;
+    PFN_vkCreateImage vkCreateImage;
+    PFN_vkCreateImageView vkCreateImageView;
+    PFN_vkCreatePipelineLayout vkCreatePipelineLayout;
+    PFN_vkCreateQueryPool vkCreateQueryPool;
+    PFN_vkCreateRenderPass vkCreateRenderPass;
+    PFN_vkCreateSampler vkCreateSampler;
+    PFN_vkCreateSemaphore vkCreateSemaphore;
+    PFN_vkCreateShaderModule vkCreateShaderModule;
+    PFN_vkCreateSwapchainKHR vkCreateSwapchainKHR;
+    PFN_vkDestroyBuffer vkDestroyBuffer;
+    PFN_vkDestroyBufferView vkDestroyBufferView;
+    PFN_vkDestroyCommandPool vkDestroyCommandPool;
+    PFN_vkDestroyDescriptorPool vkDestroyDescriptorPool;
+    PFN_vkDestroyDescriptorSetLayout vkDestroyDescriptorSetLayout;
+    PFN_vkDestroyDescriptorUpdateTemplateKHR vkDestroyDescriptorUpdateTemplateKHR;
+    PFN_vkDestroyFence vkDestroyFence;
+    PFN_vkDestroyFramebuffer vkDestroyFramebuffer;
+    PFN_vkDestroyImage vkDestroyImage;
+    PFN_vkDestroyImageView vkDestroyImageView;
+    PFN_vkDestroyPipeline vkDestroyPipeline;
+    PFN_vkDestroyPipelineLayout vkDestroyPipelineLayout;
+    PFN_vkDestroyQueryPool vkDestroyQueryPool;
+    PFN_vkDestroyRenderPass vkDestroyRenderPass;
+    PFN_vkDestroySampler vkDestroySampler;
+    PFN_vkDestroySemaphore vkDestroySemaphore;
+    PFN_vkDestroyShaderModule vkDestroyShaderModule;
+    PFN_vkDestroySwapchainKHR vkDestroySwapchainKHR;
+    PFN_vkDeviceWaitIdle vkDeviceWaitIdle;
+    PFN_vkEndCommandBuffer vkEndCommandBuffer;
+    PFN_vkFreeCommandBuffers vkFreeCommandBuffers;
+    PFN_vkFreeDescriptorSets vkFreeDescriptorSets;
+    PFN_vkFreeMemory vkFreeMemory;
+    PFN_vkGetBufferMemoryRequirements vkGetBufferMemoryRequirements;
+    PFN_vkGetDeviceQueue vkGetDeviceQueue;
+    PFN_vkGetFenceStatus vkGetFenceStatus;
+    PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements;
+    PFN_vkGetQueryPoolResults vkGetQueryPoolResults;
+    PFN_vkGetQueueCheckpointDataNV vkGetQueueCheckpointDataNV;
+    PFN_vkMapMemory vkMapMemory;
+    PFN_vkQueueSubmit vkQueueSubmit;
+    PFN_vkResetFences vkResetFences;
+    PFN_vkResetQueryPoolEXT vkResetQueryPoolEXT;
+    PFN_vkUnmapMemory vkUnmapMemory;
+    PFN_vkUpdateDescriptorSetWithTemplateKHR vkUpdateDescriptorSetWithTemplateKHR;
+    PFN_vkUpdateDescriptorSets vkUpdateDescriptorSets;
+    PFN_vkWaitForFences vkWaitForFences;
+};
+
+/// Loads instance agnostic function pointers.
+/// @return True on success, false on error.
+bool Load(InstanceDispatch&) noexcept;
+
+/// Loads instance function pointers.
+/// @return True on success, false on error.
+bool Load(VkInstance, InstanceDispatch&) noexcept;
+
+void Destroy(VkInstance, const InstanceDispatch&) noexcept;
+void Destroy(VkDevice, const InstanceDispatch&) noexcept;
+
+void Destroy(VkDevice, VkBuffer, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkBufferView, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkCommandPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorSetLayout, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDescriptorUpdateTemplateKHR, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkDeviceMemory, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkFence, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkFramebuffer, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkImage, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkImageView, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkPipeline, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkPipelineLayout, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkQueryPool, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkRenderPass, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSampler, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSwapchainKHR, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkSemaphore, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkShaderModule, const DeviceDispatch&) noexcept;
+void Destroy(VkInstance, VkDebugUtilsMessengerEXT, const InstanceDispatch&) noexcept;
+void Destroy(VkInstance, VkSurfaceKHR, const InstanceDispatch&) noexcept;
+
+VkResult Free(VkDevice, VkDescriptorPool, Span<VkDescriptorSet>, const DeviceDispatch&) noexcept;
+VkResult Free(VkDevice, VkCommandPool, Span<VkCommandBuffer>, const DeviceDispatch&) noexcept;
+
+template <typename Type, typename OwnerType, typename Dispatch>
+class Handle;
+
+/// Handle with an owning type.
+/// Analogue to std::unique_ptr.
+template <typename Type, typename OwnerType, typename Dispatch>
+class Handle {
+public:
+    /// Construct a handle and hold it's ownership.
+    explicit Handle(Type handle_, OwnerType owner_, const Dispatch& dld_) noexcept
+        : handle{handle_}, owner{owner_}, dld{&dld_} {}
+
+    /// Construct an empty handle.
+    Handle() = default;
+
+    /// Copying Vulkan objects is not supported and will never be.
+    Handle(const Handle&) = delete;
+    Handle& operator=(const Handle&) = delete;
+
+    /// Construct a handle transfering the ownership from another handle.
+    Handle(Handle&& rhs) noexcept
+        : handle{std::exchange(rhs.handle, nullptr)}, owner{rhs.owner}, dld{rhs.dld} {}
+
+    /// Assign the current handle transfering the ownership from another handle.
+    /// Destroys any previously held object.
+    Handle& operator=(Handle&& rhs) noexcept {
+        Release();
+        handle = std::exchange(rhs.handle, nullptr);
+        owner = rhs.owner;
+        dld = rhs.dld;
+        return *this;
+    }
+
+    /// Destroys the current handle if it existed.
+    ~Handle() noexcept {
+        Release();
+    }
+
+    /// Destroys any held object.
+    void reset() noexcept {
+        Release();
+        handle = nullptr;
+    }
+
+    /// Returns the address of the held object.
+    /// Intended for Vulkan structures that expect a pointer to an array.
+    const Type* address() const noexcept {
+        return std::addressof(handle);
+    }
+
+    /// Returns the held Vulkan handle.
+    Type operator*() const noexcept {
+        return handle;
+    }
+
+    /// Returns true when there's a held object.
+    explicit operator bool() const noexcept {
+        return handle != nullptr;
+    }
+
+protected:
+    Type handle = nullptr;
+    OwnerType owner = nullptr;
+    const Dispatch* dld = nullptr;
+
+private:
+    /// Destroys the held object if it exists.
+    void Release() noexcept {
+        if (handle) {
+            Destroy(owner, handle, *dld);
+        }
+    }
+};
+
+/// Dummy type used to specify a handle has no owner.
+struct NoOwner {};
+
+/// Handle without an owning type.
+/// Analogue to std::unique_ptr
+template <typename Type, typename Dispatch>
+class Handle<Type, NoOwner, Dispatch> {
+public:
+    /// Construct a handle and hold it's ownership.
+    explicit Handle(Type handle_, const Dispatch& dld_) noexcept : handle{handle_}, dld{&dld_} {}
+
+    /// Construct an empty handle.
+    Handle() noexcept = default;
+
+    /// Copying Vulkan objects is not supported and will never be.
+    Handle(const Handle&) = delete;
+    Handle& operator=(const Handle&) = delete;
+
+    /// Construct a handle transfering ownership from another handle.
+    Handle(Handle&& rhs) noexcept : handle{std::exchange(rhs.handle, nullptr)}, dld{rhs.dld} {}
+
+    /// Assign the current handle transfering the ownership from another handle.
+    /// Destroys any previously held object.
+    Handle& operator=(Handle&& rhs) noexcept {
+        Release();
+        handle = std::exchange(rhs.handle, nullptr);
+        dld = rhs.dld;
+        return *this;
+    }
+
+    /// Destroys the current handle if it existed.
+    ~Handle() noexcept {
+        Release();
+    }
+
+    /// Destroys any held object.
+    void reset() noexcept {
+        Release();
+        handle = nullptr;
+    }
+
+    /// Returns the address of the held object.
+    /// Intended for Vulkan structures that expect a pointer to an array.
+    const Type* address() const noexcept {
+        return std::addressof(handle);
+    }
+
+    /// Returns the held Vulkan handle.
+    Type operator*() const noexcept {
+        return handle;
+    }
+
+    /// Returns true when there's a held object.
+    operator bool() const noexcept {
+        return handle != nullptr;
+    }
+
+protected:
+    Type handle = nullptr;
+    const Dispatch* dld = nullptr;
+
+private:
+    /// Destroys the held object if it exists.
+    void Release() noexcept {
+        if (handle) {
+            Destroy(handle, *dld);
+        }
+    }
+};
+
+/// Array of a pool allocation.
+/// Analogue to std::vector
+template <typename AllocationType, typename PoolType>
+class PoolAllocations {
+public:
+    /// Construct an empty allocation.
+    PoolAllocations() = default;
+
+    /// Construct an allocation. Errors are reported through IsOutOfPoolMemory().
+    explicit PoolAllocations(std::unique_ptr<AllocationType[]> allocations, std::size_t num,
+                             VkDevice device, PoolType pool, const DeviceDispatch& dld) noexcept
+        : allocations{std::move(allocations)}, num{num}, device{device}, pool{pool}, dld{&dld} {}
+
+    /// Copying Vulkan allocations is not supported and will never be.
+    PoolAllocations(const PoolAllocations&) = delete;
+    PoolAllocations& operator=(const PoolAllocations&) = delete;
+
+    /// Construct an allocation transfering ownership from another allocation.
+    PoolAllocations(PoolAllocations&& rhs) noexcept
+        : allocations{std::move(rhs.allocations)}, num{rhs.num}, device{rhs.device}, pool{rhs.pool},
+          dld{rhs.dld} {}
+
+    /// Assign an allocation transfering ownership from another allocation.
+    /// Releases any previously held allocation.
+    PoolAllocations& operator=(PoolAllocations&& rhs) noexcept {
+        Release();
+        allocations = std::move(rhs.allocations);
+        num = rhs.num;
+        device = rhs.device;
+        pool = rhs.pool;
+        dld = rhs.dld;
+        return *this;
+    }
+
+    /// Destroys any held allocation.
+    ~PoolAllocations() {
+        Release();
+    }
+
+    /// Returns the number of allocations.
+    std::size_t size() const noexcept {
+        return num;
+    }
+
+    /// Returns a pointer to the array of allocations.
+    AllocationType const* data() const noexcept {
+        return allocations.get();
+    }
+
+    /// Returns the allocation in the specified index.
+    /// @pre index < size()
+    AllocationType operator[](std::size_t index) const noexcept {
+        return allocations[index];
+    }
+
+    /// True when a pool fails to construct.
+    bool IsOutOfPoolMemory() const noexcept {
+        return !device;
+    }
+
+private:
+    /// Destroys the held allocations if they exist.
+    void Release() noexcept {
+        if (!allocations) {
+            return;
+        }
+        const Span<AllocationType> span(allocations.get(), num);
+        const VkResult result = Free(device, pool, span, *dld);
+        // There's no way to report errors from a destructor.
+        if (result != VK_SUCCESS) {
+            std::terminate();
+        }
+    }
+
+    std::unique_ptr<AllocationType[]> allocations;
+    std::size_t num = 0;
+    VkDevice device = nullptr;
+    PoolType pool = nullptr;
+    const DeviceDispatch* dld = nullptr;
+};
+
+using BufferView = Handle<VkBufferView, VkDevice, DeviceDispatch>;
+using DebugCallback = Handle<VkDebugUtilsMessengerEXT, VkInstance, InstanceDispatch>;
+using DescriptorSetLayout = Handle<VkDescriptorSetLayout, VkDevice, DeviceDispatch>;
+using DescriptorUpdateTemplateKHR = Handle<VkDescriptorUpdateTemplateKHR, VkDevice, DeviceDispatch>;
+using Framebuffer = Handle<VkFramebuffer, VkDevice, DeviceDispatch>;
+using ImageView = Handle<VkImageView, VkDevice, DeviceDispatch>;
+using Pipeline = Handle<VkPipeline, VkDevice, DeviceDispatch>;
+using PipelineLayout = Handle<VkPipelineLayout, VkDevice, DeviceDispatch>;
+using QueryPool = Handle<VkQueryPool, VkDevice, DeviceDispatch>;
+using RenderPass = Handle<VkRenderPass, VkDevice, DeviceDispatch>;
+using Sampler = Handle<VkSampler, VkDevice, DeviceDispatch>;
+using Semaphore = Handle<VkSemaphore, VkDevice, DeviceDispatch>;
+using ShaderModule = Handle<VkShaderModule, VkDevice, DeviceDispatch>;
+using SurfaceKHR = Handle<VkSurfaceKHR, VkInstance, InstanceDispatch>;
+
+using DescriptorSets = PoolAllocations<VkDescriptorSet, VkDescriptorPool>;
+using CommandBuffers = PoolAllocations<VkCommandBuffer, VkCommandPool>;
+
+/// Vulkan instance owning handle.
+class Instance : public Handle<VkInstance, NoOwner, InstanceDispatch> {
+    using Handle<VkInstance, NoOwner, InstanceDispatch>::Handle;
+
+public:
+    /// Creates a Vulkan instance. Use "operator bool" for error handling.
+    static Instance Create(Span<const char*> layers, Span<const char*> extensions,
+                           InstanceDispatch& dld) noexcept;
+
+    /// Enumerates physical devices.
+    /// @return Physical devices and an empty handle on failure.
+    std::optional<std::vector<VkPhysicalDevice>> EnumeratePhysicalDevices();
+
+    /// Tries to create a debug callback messenger. Returns an empty handle on failure.
+    DebugCallback TryCreateDebugCallback(PFN_vkDebugUtilsMessengerCallbackEXT callback) noexcept;
+};
+
+class Queue {
+public:
+    /// Construct an empty queue handle.
+    constexpr Queue() noexcept = default;
+
+    /// Construct a queue handle.
+    constexpr Queue(VkQueue queue, const DeviceDispatch& dld) noexcept : queue{queue}, dld{&dld} {}
+
+    /// Returns the checkpoint data.
+    /// @note Returns an empty vector when the function pointer is not present.
+    std::vector<VkCheckpointDataNV> GetCheckpointDataNV(const DeviceDispatch& dld) const;
+
+    void Submit(Span<VkSubmitInfo> submit_infos, VkFence fence) const {
+        Check(dld->vkQueueSubmit(queue, submit_infos.size(), submit_infos.data(), fence));
+    }
+
+    VkResult Present(const VkPresentInfoKHR& present_info) const noexcept {
+        return dld->vkQueuePresentKHR(queue, &present_info);
+    }
+
+private:
+    VkQueue queue = nullptr;
+    const DeviceDispatch* dld = nullptr;
+};
+
+class Buffer : public Handle<VkBuffer, VkDevice, DeviceDispatch> {
+    using Handle<VkBuffer, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    /// Attaches a memory allocation.
+    void BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const;
+};
+
+class Image : public Handle<VkImage, VkDevice, DeviceDispatch> {
+    using Handle<VkImage, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    /// Attaches a memory allocation.
+    void BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const;
+};
+
+class DeviceMemory : public Handle<VkDeviceMemory, VkDevice, DeviceDispatch> {
+    using Handle<VkDeviceMemory, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    u8* Map(VkDeviceSize offset, VkDeviceSize size) const {
+        void* data;
+        Check(dld->vkMapMemory(owner, handle, offset, size, 0, &data));
+        return static_cast<u8*>(data);
+    }
+
+    void Unmap() const noexcept {
+        dld->vkUnmapMemory(owner, handle);
+    }
+};
+
+class Fence : public Handle<VkFence, VkDevice, DeviceDispatch> {
+    using Handle<VkFence, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    VkResult Wait(u64 timeout = std::numeric_limits<u64>::max()) const noexcept {
+        return dld->vkWaitForFences(owner, 1, &handle, true, timeout);
+    }
+
+    VkResult GetStatus() const noexcept {
+        return dld->vkGetFenceStatus(owner, handle);
+    }
+
+    void Reset() const {
+        Check(dld->vkResetFences(owner, 1, &handle));
+    }
+};
+
+class DescriptorPool : public Handle<VkDescriptorPool, VkDevice, DeviceDispatch> {
+    using Handle<VkDescriptorPool, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    DescriptorSets Allocate(const VkDescriptorSetAllocateInfo& ai) const;
+};
+
+class CommandPool : public Handle<VkCommandPool, VkDevice, DeviceDispatch> {
+    using Handle<VkCommandPool, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    CommandBuffers Allocate(std::size_t num_buffers,
+                            VkCommandBufferLevel level = VK_COMMAND_BUFFER_LEVEL_PRIMARY) const;
+};
+
+class SwapchainKHR : public Handle<VkSwapchainKHR, VkDevice, DeviceDispatch> {
+    using Handle<VkSwapchainKHR, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    std::vector<VkImage> GetImages() const;
+};
+
+class Device : public Handle<VkDevice, NoOwner, DeviceDispatch> {
+    using Handle<VkDevice, NoOwner, DeviceDispatch>::Handle;
+
+public:
+    static Device Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
+                         Span<const char*> enabled_extensions,
+                         const VkPhysicalDeviceFeatures2& enabled_features,
+                         DeviceDispatch& dld) noexcept;
+
+    Queue GetQueue(u32 family_index) const noexcept;
+
+    Buffer CreateBuffer(const VkBufferCreateInfo& ci) const;
+
+    BufferView CreateBufferView(const VkBufferViewCreateInfo& ci) const;
+
+    Image CreateImage(const VkImageCreateInfo& ci) const;
+
+    ImageView CreateImageView(const VkImageViewCreateInfo& ci) const;
+
+    Semaphore CreateSemaphore() const;
+
+    Fence CreateFence(const VkFenceCreateInfo& ci) const;
+
+    DescriptorPool CreateDescriptorPool(const VkDescriptorPoolCreateInfo& ci) const;
+
+    RenderPass CreateRenderPass(const VkRenderPassCreateInfo& ci) const;
+
+    DescriptorSetLayout CreateDescriptorSetLayout(const VkDescriptorSetLayoutCreateInfo& ci) const;
+
+    PipelineLayout CreatePipelineLayout(const VkPipelineLayoutCreateInfo& ci) const;
+
+    Pipeline CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci) const;
+
+    Pipeline CreateComputePipeline(const VkComputePipelineCreateInfo& ci) const;
+
+    Sampler CreateSampler(const VkSamplerCreateInfo& ci) const;
+
+    Framebuffer CreateFramebuffer(const VkFramebufferCreateInfo& ci) const;
+
+    CommandPool CreateCommandPool(const VkCommandPoolCreateInfo& ci) const;
+
+    DescriptorUpdateTemplateKHR CreateDescriptorUpdateTemplateKHR(
+        const VkDescriptorUpdateTemplateCreateInfoKHR& ci) const;
+
+    QueryPool CreateQueryPool(const VkQueryPoolCreateInfo& ci) const;
+
+    ShaderModule CreateShaderModule(const VkShaderModuleCreateInfo& ci) const;
+
+    SwapchainKHR CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const;
+
+    DeviceMemory TryAllocateMemory(const VkMemoryAllocateInfo& ai) const noexcept;
+
+    DeviceMemory AllocateMemory(const VkMemoryAllocateInfo& ai) const;
+
+    VkMemoryRequirements GetBufferMemoryRequirements(VkBuffer buffer) const noexcept;
+
+    VkMemoryRequirements GetImageMemoryRequirements(VkImage image) const noexcept;
+
+    void UpdateDescriptorSets(Span<VkWriteDescriptorSet> writes,
+                              Span<VkCopyDescriptorSet> copies) const noexcept;
+
+    void UpdateDescriptorSet(VkDescriptorSet set, VkDescriptorUpdateTemplateKHR update_template,
+                             const void* data) const noexcept {
+        dld->vkUpdateDescriptorSetWithTemplateKHR(handle, set, update_template, data);
+    }
+
+    VkResult AcquireNextImageKHR(VkSwapchainKHR swapchain, u64 timeout, VkSemaphore semaphore,
+                                 VkFence fence, u32* image_index) const noexcept {
+        return dld->vkAcquireNextImageKHR(handle, swapchain, timeout, semaphore, fence,
+                                          image_index);
+    }
+
+    VkResult WaitIdle() const noexcept {
+        return dld->vkDeviceWaitIdle(handle);
+    }
+
+    void ResetQueryPoolEXT(VkQueryPool query_pool, u32 first, u32 count) const noexcept {
+        dld->vkResetQueryPoolEXT(handle, query_pool, first, count);
+    }
+
+    void GetQueryResults(VkQueryPool query_pool, u32 first, u32 count, std::size_t data_size,
+                         void* data, VkDeviceSize stride, VkQueryResultFlags flags) const {
+        Check(dld->vkGetQueryPoolResults(handle, query_pool, first, count, data_size, data, stride,
+                                         flags));
+    }
+
+    template <typename T>
+    T GetQueryResult(VkQueryPool query_pool, u32 first, VkQueryResultFlags flags) const {
+        static_assert(std::is_trivially_copyable_v<T>);
+        T value;
+        GetQueryResults(query_pool, first, 1, sizeof(T), &value, sizeof(T), flags);
+        return value;
+    }
+};
+
+class PhysicalDevice {
+public:
+    constexpr PhysicalDevice() noexcept = default;
+
+    constexpr PhysicalDevice(VkPhysicalDevice physical_device, const InstanceDispatch& dld) noexcept
+        : physical_device{physical_device}, dld{&dld} {}
+
+    constexpr operator VkPhysicalDevice() const noexcept {
+        return physical_device;
+    }
+
+    VkPhysicalDeviceProperties GetProperties() const noexcept;
+
+    void GetProperties2KHR(VkPhysicalDeviceProperties2KHR&) const noexcept;
+
+    VkPhysicalDeviceFeatures GetFeatures() const noexcept;
+
+    void GetFeatures2KHR(VkPhysicalDeviceFeatures2KHR&) const noexcept;
+
+    VkFormatProperties GetFormatProperties(VkFormat) const noexcept;
+
+    std::vector<VkExtensionProperties> EnumerateDeviceExtensionProperties() const;
+
+    std::vector<VkQueueFamilyProperties> GetQueueFamilyProperties() const;
+
+    bool GetSurfaceSupportKHR(u32 queue_family_index, VkSurfaceKHR) const;
+
+    VkSurfaceCapabilitiesKHR GetSurfaceCapabilitiesKHR(VkSurfaceKHR) const noexcept;
+
+    std::vector<VkSurfaceFormatKHR> GetSurfaceFormatsKHR(VkSurfaceKHR) const;
+
+    std::vector<VkPresentModeKHR> GetSurfacePresentModesKHR(VkSurfaceKHR) const;
+
+    VkPhysicalDeviceMemoryProperties GetMemoryProperties() const noexcept;
+
+private:
+    VkPhysicalDevice physical_device = nullptr;
+    const InstanceDispatch* dld = nullptr;
+};
+
+class CommandBuffer {
+public:
+    CommandBuffer() noexcept = default;
+
+    explicit CommandBuffer(VkCommandBuffer handle, const DeviceDispatch& dld) noexcept
+        : handle{handle}, dld{&dld} {}
+
+    const VkCommandBuffer* address() const noexcept {
+        return &handle;
+    }
+
+    void Begin(const VkCommandBufferBeginInfo& begin_info) const {
+        Check(dld->vkBeginCommandBuffer(handle, &begin_info));
+    }
+
+    void End() const {
+        Check(dld->vkEndCommandBuffer(handle));
+    }
+
+    void BeginRenderPass(const VkRenderPassBeginInfo& renderpass_bi,
+                         VkSubpassContents contents) const noexcept {
+        dld->vkCmdBeginRenderPass(handle, &renderpass_bi, contents);
+    }
+
+    void EndRenderPass() const noexcept {
+        dld->vkCmdEndRenderPass(handle);
+    }
+
+    void BeginQuery(VkQueryPool query_pool, u32 query, VkQueryControlFlags flags) const noexcept {
+        dld->vkCmdBeginQuery(handle, query_pool, query, flags);
+    }
+
+    void EndQuery(VkQueryPool query_pool, u32 query) const noexcept {
+        dld->vkCmdEndQuery(handle, query_pool, query);
+    }
+
+    void BindDescriptorSets(VkPipelineBindPoint bind_point, VkPipelineLayout layout, u32 first,
+                            Span<VkDescriptorSet> sets, Span<u32> dynamic_offsets) const noexcept {
+        dld->vkCmdBindDescriptorSets(handle, bind_point, layout, first, sets.size(), sets.data(),
+                                     dynamic_offsets.size(), dynamic_offsets.data());
+    }
+
+    void BindPipeline(VkPipelineBindPoint bind_point, VkPipeline pipeline) const noexcept {
+        dld->vkCmdBindPipeline(handle, bind_point, pipeline);
+    }
+
+    void BindIndexBuffer(VkBuffer buffer, VkDeviceSize offset, VkIndexType index_type) const
+        noexcept {
+        dld->vkCmdBindIndexBuffer(handle, buffer, offset, index_type);
+    }
+
+    void BindVertexBuffers(u32 first, u32 count, const VkBuffer* buffers,
+                           const VkDeviceSize* offsets) const noexcept {
+        dld->vkCmdBindVertexBuffers(handle, first, count, buffers, offsets);
+    }
+
+    void BindVertexBuffer(u32 binding, VkBuffer buffer, VkDeviceSize offset) const noexcept {
+        BindVertexBuffers(binding, 1, &buffer, &offset);
+    }
+
+    void Draw(u32 vertex_count, u32 instance_count, u32 first_vertex, u32 first_instance) const
+        noexcept {
+        dld->vkCmdDraw(handle, vertex_count, instance_count, first_vertex, first_instance);
+    }
+
+    void DrawIndexed(u32 index_count, u32 instance_count, u32 first_index, u32 vertex_offset,
+                     u32 first_instance) const noexcept {
+        dld->vkCmdDrawIndexed(handle, index_count, instance_count, first_index, vertex_offset,
+                              first_instance);
+    }
+
+    void ClearAttachments(Span<VkClearAttachment> attachments, Span<VkClearRect> rects) const
+        noexcept {
+        dld->vkCmdClearAttachments(handle, attachments.size(), attachments.data(), rects.size(),
+                                   rects.data());
+    }
+
+    void BlitImage(VkImage src_image, VkImageLayout src_layout, VkImage dst_image,
+                   VkImageLayout dst_layout, Span<VkImageBlit> regions, VkFilter filter) const
+        noexcept {
+        dld->vkCmdBlitImage(handle, src_image, src_layout, dst_image, dst_layout, regions.size(),
+                            regions.data(), filter);
+    }
+
+    void Dispatch(u32 x, u32 y, u32 z) const noexcept {
+        dld->vkCmdDispatch(handle, x, y, z);
+    }
+
+    void PipelineBarrier(VkPipelineStageFlags src_stage_mask, VkPipelineStageFlags dst_stage_mask,
+                         VkDependencyFlags dependency_flags, Span<VkMemoryBarrier> memory_barriers,
+                         Span<VkBufferMemoryBarrier> buffer_barriers,
+                         Span<VkImageMemoryBarrier> image_barriers) const noexcept {
+        dld->vkCmdPipelineBarrier(handle, src_stage_mask, dst_stage_mask, dependency_flags,
+                                  memory_barriers.size(), memory_barriers.data(),
+                                  buffer_barriers.size(), buffer_barriers.data(),
+                                  image_barriers.size(), image_barriers.data());
+    }
+
+    void CopyBufferToImage(VkBuffer src_buffer, VkImage dst_image, VkImageLayout dst_image_layout,
+                           Span<VkBufferImageCopy> regions) const noexcept {
+        dld->vkCmdCopyBufferToImage(handle, src_buffer, dst_image, dst_image_layout, regions.size(),
+                                    regions.data());
+    }
+
+    void CopyBuffer(VkBuffer src_buffer, VkBuffer dst_buffer, Span<VkBufferCopy> regions) const
+        noexcept {
+        dld->vkCmdCopyBuffer(handle, src_buffer, dst_buffer, regions.size(), regions.data());
+    }
+
+    void CopyImage(VkImage src_image, VkImageLayout src_layout, VkImage dst_image,
+                   VkImageLayout dst_layout, Span<VkImageCopy> regions) const noexcept {
+        dld->vkCmdCopyImage(handle, src_image, src_layout, dst_image, dst_layout, regions.size(),
+                            regions.data());
+    }
+
+    void CopyImageToBuffer(VkImage src_image, VkImageLayout src_layout, VkBuffer dst_buffer,
+                           Span<VkBufferImageCopy> regions) const noexcept {
+        dld->vkCmdCopyImageToBuffer(handle, src_image, src_layout, dst_buffer, regions.size(),
+                                    regions.data());
+    }
+
+    void FillBuffer(VkBuffer dst_buffer, VkDeviceSize dst_offset, VkDeviceSize size, u32 data) const
+        noexcept {
+        dld->vkCmdFillBuffer(handle, dst_buffer, dst_offset, size, data);
+    }
+
+    void PushConstants(VkPipelineLayout layout, VkShaderStageFlags flags, u32 offset, u32 size,
+                       const void* values) const noexcept {
+        dld->vkCmdPushConstants(handle, layout, flags, offset, size, values);
+    }
+
+    void SetCheckpointNV(const void* checkpoint_marker) const noexcept {
+        dld->vkCmdSetCheckpointNV(handle, checkpoint_marker);
+    }
+
+    void SetViewport(u32 first, Span<VkViewport> viewports) const noexcept {
+        dld->vkCmdSetViewport(handle, first, viewports.size(), viewports.data());
+    }
+
+    void SetScissor(u32 first, Span<VkRect2D> scissors) const noexcept {
+        dld->vkCmdSetScissor(handle, first, scissors.size(), scissors.data());
+    }
+
+    void SetBlendConstants(const float blend_constants[4]) const noexcept {
+        dld->vkCmdSetBlendConstants(handle, blend_constants);
+    }
+
+    void SetStencilCompareMask(VkStencilFaceFlags face_mask, u32 compare_mask) const noexcept {
+        dld->vkCmdSetStencilCompareMask(handle, face_mask, compare_mask);
+    }
+
+    void SetStencilReference(VkStencilFaceFlags face_mask, u32 reference) const noexcept {
+        dld->vkCmdSetStencilReference(handle, face_mask, reference);
+    }
+
+    void SetStencilWriteMask(VkStencilFaceFlags face_mask, u32 write_mask) const noexcept {
+        dld->vkCmdSetStencilWriteMask(handle, face_mask, write_mask);
+    }
+
+    void SetDepthBias(float constant_factor, float clamp, float slope_factor) const noexcept {
+        dld->vkCmdSetDepthBias(handle, constant_factor, clamp, slope_factor);
+    }
+
+    void SetDepthBounds(float min_depth_bounds, float max_depth_bounds) const noexcept {
+        dld->vkCmdSetDepthBounds(handle, min_depth_bounds, max_depth_bounds);
+    }
+
+    void BindTransformFeedbackBuffersEXT(u32 first, u32 count, const VkBuffer* buffers,
+                                         const VkDeviceSize* offsets,
+                                         const VkDeviceSize* sizes) const noexcept {
+        dld->vkCmdBindTransformFeedbackBuffersEXT(handle, first, count, buffers, offsets, sizes);
+    }
+
+    void BeginTransformFeedbackEXT(u32 first_counter_buffer, u32 counter_buffers_count,
+                                   const VkBuffer* counter_buffers,
+                                   const VkDeviceSize* counter_buffer_offsets) const noexcept {
+        dld->vkCmdBeginTransformFeedbackEXT(handle, first_counter_buffer, counter_buffers_count,
+                                            counter_buffers, counter_buffer_offsets);
+    }
+
+    void EndTransformFeedbackEXT(u32 first_counter_buffer, u32 counter_buffers_count,
+                                 const VkBuffer* counter_buffers,
+                                 const VkDeviceSize* counter_buffer_offsets) const noexcept {
+        dld->vkCmdEndTransformFeedbackEXT(handle, first_counter_buffer, counter_buffers_count,
+                                          counter_buffers, counter_buffer_offsets);
+    }
+
+private:
+    VkCommandBuffer handle;
+    const DeviceDispatch* dld;
+};
+
+std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProperties(
+    const InstanceDispatch& dld);
+
+} // namespace Vulkan::vk

src/video_core/shader/decode/arithmetic_integer.cpp

@@ -235,34 +235,30 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
     case OpCode::Id::LEA_IMM:
     case OpCode::Id::LEA_RZ:
     case OpCode::Id::LEA_HI: {
-        const auto [op_a, op_b, op_c] = [&]() -> std::tuple<Node, Node, Node> {
+        auto [op_a, op_b, op_c] = [&]() -> std::tuple<Node, Node, Node> {
             switch (opcode->get().GetId()) {
             case OpCode::Id::LEA_R2: {
                 return {GetRegister(instr.gpr20), GetRegister(instr.gpr39),
                         Immediate(static_cast<u32>(instr.lea.r2.entry_a))};
             }
-
             case OpCode::Id::LEA_R1: {
                 const bool neg = instr.lea.r1.neg != 0;
                 return {GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
                         GetRegister(instr.gpr20),
                         Immediate(static_cast<u32>(instr.lea.r1.entry_a))};
             }
-
             case OpCode::Id::LEA_IMM: {
                 const bool neg = instr.lea.imm.neg != 0;
                 return {Immediate(static_cast<u32>(instr.lea.imm.entry_a)),
                         GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
                         Immediate(static_cast<u32>(instr.lea.imm.entry_b))};
             }
-
             case OpCode::Id::LEA_RZ: {
                 const bool neg = instr.lea.rz.neg != 0;
                 return {GetConstBuffer(instr.lea.rz.cb_index, instr.lea.rz.cb_offset),
                         GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
                         Immediate(static_cast<u32>(instr.lea.rz.entry_a))};
             }
-
             case OpCode::Id::LEA_HI:
             default:
                 UNIMPLEMENTED_MSG("Unhandled LEA subinstruction: {}", opcode->get().GetName());
@@ -275,12 +271,9 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
         UNIMPLEMENTED_IF_MSG(instr.lea.pred48 != static_cast<u64>(Pred::UnusedIndex),
                              "Unhandled LEA Predicate");
 
-        const Node shifted_c =
-            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, Immediate(1), op_c);
-        const Node mul_bc = Operation(OperationCode::IMul, NO_PRECISE, op_b, shifted_c);
-        const Node value = Operation(OperationCode::IAdd, NO_PRECISE, op_a, mul_bc);
-
-        SetRegister(bb, instr.gpr0, value);
+        Node value = Operation(OperationCode::ILogicalShiftLeft, std::move(op_a), std::move(op_c));
+        value = Operation(OperationCode::IAdd, std::move(op_b), std::move(value));
+        SetRegister(bb, instr.gpr0, std::move(value));
 
         break;
     }

src/video_core/shader/decode/bfe.cpp

@@ -17,33 +17,60 @@ u32 ShaderIR::DecodeBfe(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
-    UNIMPLEMENTED_IF(instr.bfe.negate_b);
-
     Node op_a = GetRegister(instr.gpr8);
-    op_a = GetOperandAbsNegInteger(op_a, false, instr.bfe.negate_a, false);
+    Node op_b = [&] {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::BFE_R:
+            return GetRegister(instr.gpr20);
+        case OpCode::Id::BFE_C:
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
+        case OpCode::Id::BFE_IMM:
+            return Immediate(instr.alu.GetSignedImm20_20());
+        default:
+            UNREACHABLE();
+            return Immediate(0);
+        }
+    }();
 
-    switch (opcode->get().GetId()) {
-    case OpCode::Id::BFE_IMM: {
-        UNIMPLEMENTED_IF_MSG(instr.generates_cc,
-                             "Condition codes generation in BFE is not implemented");
+    UNIMPLEMENTED_IF_MSG(instr.bfe.rd_cc, "Condition codes in BFE is not implemented");
 
-        const Node inner_shift_imm = Immediate(static_cast<u32>(instr.bfe.GetLeftShiftValue()));
-        const Node outer_shift_imm =
-            Immediate(static_cast<u32>(instr.bfe.GetLeftShiftValue() + instr.bfe.shift_position));
+    const bool is_signed = instr.bfe.is_signed;
 
-        const Node inner_shift =
-            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, op_a, inner_shift_imm);
-        const Node outer_shift =
-            Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, inner_shift, outer_shift_imm);
-
-        SetInternalFlagsFromInteger(bb, outer_shift, instr.generates_cc);
-        SetRegister(bb, instr.gpr0, outer_shift);
-        break;
-    }
-    default:
-        UNIMPLEMENTED_MSG("Unhandled BFE instruction: {}", opcode->get().GetName());
+    // using reverse parallel method in
+    // https://graphics.stanford.edu/~seander/bithacks.html#ReverseParallel
+    // note for later if possible to implement faster method.
+    if (instr.bfe.brev) {
+        const auto swap = [&](u32 s, u32 mask) {
+            Node v1 =
+                SignedOperation(OperationCode::ILogicalShiftRight, is_signed, op_a, Immediate(s));
+            if (mask != 0) {
+                v1 = SignedOperation(OperationCode::IBitwiseAnd, is_signed, std::move(v1),
+                                     Immediate(mask));
+            }
+            Node v2 = op_a;
+            if (mask != 0) {
+                v2 = SignedOperation(OperationCode::IBitwiseAnd, is_signed, std::move(v2),
+                                     Immediate(mask));
+            }
+            v2 = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed, std::move(v2),
+                                 Immediate(s));
+            return SignedOperation(OperationCode::IBitwiseOr, is_signed, std::move(v1),
+                                   std::move(v2));
+        };
+        op_a = swap(1, 0x55555555U);
+        op_a = swap(2, 0x33333333U);
+        op_a = swap(4, 0x0F0F0F0FU);
+        op_a = swap(8, 0x00FF00FFU);
+        op_a = swap(16, 0);
     }
 
+    const auto offset = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_b,
+                                        Immediate(0), Immediate(8));
+    const auto bits = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_b,
+                                      Immediate(8), Immediate(8));
+    auto result = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_a, offset, bits);
+    SetRegister(bb, instr.gpr0, std::move(result));
+
     return pc;
 }
 

src/video_core/shader/decode/conversion.cpp

@@ -138,18 +138,23 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
 
         value = GetOperandAbsNegFloat(value, instr.conversion.abs_a, instr.conversion.negate_a);
 
-        value = [&]() {
+        value = [&] {
+            if (instr.conversion.src_size != instr.conversion.dst_size) {
+                // Rounding operations only matter when the source and destination conversion size
+                // is the same.
+                return value;
+            }
             switch (instr.conversion.f2f.GetRoundingMode()) {
             case Tegra::Shader::F2fRoundingOp::None:
                 return value;
             case Tegra::Shader::F2fRoundingOp::Round:
-                return Operation(OperationCode::FRoundEven, PRECISE, value);
+                return Operation(OperationCode::FRoundEven, value);
             case Tegra::Shader::F2fRoundingOp::Floor:
-                return Operation(OperationCode::FFloor, PRECISE, value);
+                return Operation(OperationCode::FFloor, value);
             case Tegra::Shader::F2fRoundingOp::Ceil:
-                return Operation(OperationCode::FCeil, PRECISE, value);
+                return Operation(OperationCode::FCeil, value);
             case Tegra::Shader::F2fRoundingOp::Trunc:
-                return Operation(OperationCode::FTrunc, PRECISE, value);
+                return Operation(OperationCode::FTrunc, value);
             default:
                 UNIMPLEMENTED_MSG("Unimplemented F2F rounding mode {}",
                                   static_cast<u32>(instr.conversion.f2f.rounding.Value()));

src/video_core/shader/decode/memory.cpp

@@ -19,7 +19,6 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::AtomicOp;
 using Tegra::Shader::AtomicType;
 using Tegra::Shader::Attribute;
-using Tegra::Shader::GlobalAtomicOp;
 using Tegra::Shader::GlobalAtomicType;
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
@@ -28,6 +27,31 @@ using Tegra::Shader::StoreType;
 
 namespace {
 
+Node GetAtomOperation(AtomicOp op, bool is_signed, Node memory, Node data) {
+    const OperationCode operation_code = [op] {
+        switch (op) {
+        case AtomicOp::Add:
+            return OperationCode::AtomicIAdd;
+        case AtomicOp::Min:
+            return OperationCode::AtomicIMin;
+        case AtomicOp::Max:
+            return OperationCode::AtomicIMax;
+        case AtomicOp::And:
+            return OperationCode::AtomicIAnd;
+        case AtomicOp::Or:
+            return OperationCode::AtomicIOr;
+        case AtomicOp::Xor:
+            return OperationCode::AtomicIXor;
+        case AtomicOp::Exch:
+            return OperationCode::AtomicIExchange;
+        default:
+            UNIMPLEMENTED_MSG("op={}", static_cast<int>(op));
+            return OperationCode::AtomicIAdd;
+        }
+    }();
+    return SignedOperation(operation_code, is_signed, std::move(memory), std::move(data));
+}
+
 bool IsUnaligned(Tegra::Shader::UniformType uniform_type) {
     return uniform_type == Tegra::Shader::UniformType::UnsignedByte ||
            uniform_type == Tegra::Shader::UniformType::UnsignedShort;
@@ -363,10 +387,13 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         break;
     }
     case OpCode::Id::ATOM: {
-        UNIMPLEMENTED_IF_MSG(instr.atom.operation != GlobalAtomicOp::Add, "operation={}",
-                             static_cast<int>(instr.atom.operation.Value()));
-        UNIMPLEMENTED_IF_MSG(instr.atom.type != GlobalAtomicType::S32, "type={}",
-                             static_cast<int>(instr.atom.type.Value()));
+        UNIMPLEMENTED_IF_MSG(instr.atom.operation == AtomicOp::Inc ||
+                                 instr.atom.operation == AtomicOp::Dec ||
+                                 instr.atom.operation == AtomicOp::SafeAdd,
+                             "operation={}", static_cast<int>(instr.atom.operation.Value()));
+        UNIMPLEMENTED_IF_MSG(instr.atom.type == GlobalAtomicType::S64 ||
+                                 instr.atom.type == GlobalAtomicType::U64,
+                             "type={}", static_cast<int>(instr.atom.type.Value()));
 
         const auto [real_address, base_address, descriptor] =
             TrackGlobalMemory(bb, instr, true, true);
@@ -375,25 +402,29 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
             break;
         }
 
+        const bool is_signed =
+            instr.atoms.type == AtomicType::S32 || instr.atoms.type == AtomicType::S64;
         Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor);
-        Node value = Operation(OperationCode::AtomicAdd, std::move(gmem), GetRegister(instr.gpr20));
+        Node value = GetAtomOperation(static_cast<AtomicOp>(instr.atom.operation), is_signed, gmem,
+                                      GetRegister(instr.gpr20));
         SetRegister(bb, instr.gpr0, std::move(value));
         break;
     }
     case OpCode::Id::ATOMS: {
-        UNIMPLEMENTED_IF_MSG(instr.atoms.operation != AtomicOp::Add, "operation={}",
-                             static_cast<int>(instr.atoms.operation.Value()));
-        UNIMPLEMENTED_IF_MSG(instr.atoms.type != AtomicType::U32, "type={}",
-                             static_cast<int>(instr.atoms.type.Value()));
-
+        UNIMPLEMENTED_IF_MSG(instr.atoms.operation == AtomicOp::Inc ||
+                                 instr.atoms.operation == AtomicOp::Dec,
+                             "operation={}", static_cast<int>(instr.atoms.operation.Value()));
+        UNIMPLEMENTED_IF_MSG(instr.atoms.type == AtomicType::S64 ||
+                                 instr.atoms.type == AtomicType::U64,
+                             "type={}", static_cast<int>(instr.atoms.type.Value()));
+        const bool is_signed =
+            instr.atoms.type == AtomicType::S32 || instr.atoms.type == AtomicType::S64;
         const s32 offset = instr.atoms.GetImmediateOffset();
         Node address = GetRegister(instr.gpr8);
         address = Operation(OperationCode::IAdd, std::move(address), Immediate(offset));
-
-        Node memory = GetSharedMemory(std::move(address));
-        Node data = GetRegister(instr.gpr20);
-
-        Node value = Operation(OperationCode::AtomicAdd, std::move(memory), std::move(data));
+        Node value =
+            GetAtomOperation(static_cast<AtomicOp>(instr.atoms.operation), is_signed,
+                             GetSharedMemory(std::move(address)), GetRegister(instr.gpr20));
         SetRegister(bb, instr.gpr0, std::move(value));
         break;
     }

src/video_core/shader/decode/xmad.cpp

@@ -12,6 +12,7 @@ namespace VideoCommon::Shader {
 
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
+using Tegra::Shader::PredCondition;
 
 u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
@@ -30,7 +31,7 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     const bool is_signed_b = instr.xmad.sign_b == 1;
     const bool is_signed_c = is_signed_a;
 
-    auto [is_merge, is_psl, is_high_b, mode, op_b,
+    auto [is_merge, is_psl, is_high_b, mode, op_b_binding,
           op_c] = [&]() -> std::tuple<bool, bool, bool, Tegra::Shader::XmadMode, Node, Node> {
         switch (opcode->get().GetId()) {
         case OpCode::Id::XMAD_CR:
@@ -63,15 +64,19 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
         }
     }();
 
-    op_a = BitfieldExtract(op_a, instr.xmad.high_a ? 16 : 0, 16);
+    op_a = SignedOperation(OperationCode::IBitfieldExtract, is_signed_a, std::move(op_a),
+                           instr.xmad.high_a ? Immediate(16) : Immediate(0), Immediate(16));
 
-    const Node original_b = op_b;
-    op_b = BitfieldExtract(op_b, is_high_b ? 16 : 0, 16);
+    const Node original_b = op_b_binding;
+    const Node op_b =
+        SignedOperation(OperationCode::IBitfieldExtract, is_signed_b, std::move(op_b_binding),
+                        is_high_b ? Immediate(16) : Immediate(0), Immediate(16));
 
-    // TODO(Rodrigo): Use an appropiate sign for this operation
-    Node product = Operation(OperationCode::IMul, NO_PRECISE, op_a, op_b);
+    // we already check sign_a and sign_b is difference or not before so just use one in here.
+    Node product = SignedOperation(OperationCode::IMul, is_signed_a, op_a, op_b);
     if (is_psl) {
-        product = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, product, Immediate(16));
+        product =
+            SignedOperation(OperationCode::ILogicalShiftLeft, is_signed_a, product, Immediate(16));
     }
     SetTemporary(bb, 0, product);
     product = GetTemporary(0);
@@ -88,12 +93,40 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
             return BitfieldExtract(original_c, 16, 16);
         case Tegra::Shader::XmadMode::CBcc: {
             const Node shifted_b = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed_b,
-                                                   NO_PRECISE, original_b, Immediate(16));
-            return SignedOperation(OperationCode::IAdd, is_signed_c, NO_PRECISE, original_c,
-                                   shifted_b);
+                                                   original_b, Immediate(16));
+            return SignedOperation(OperationCode::IAdd, is_signed_c, original_c, shifted_b);
+        }
+        case Tegra::Shader::XmadMode::CSfu: {
+            const Node comp_a = GetPredicateComparisonInteger(PredCondition::Equal, is_signed_a,
+                                                              op_a, Immediate(0));
+            const Node comp_b = GetPredicateComparisonInteger(PredCondition::Equal, is_signed_b,
+                                                              op_b, Immediate(0));
+            const Node comp = Operation(OperationCode::LogicalOr, comp_a, comp_b);
+
+            const Node comp_minus_a = GetPredicateComparisonInteger(
+                PredCondition::NotEqual, is_signed_a,
+                SignedOperation(OperationCode::IBitwiseAnd, is_signed_a, op_a,
+                                Immediate(0x80000000)),
+                Immediate(0));
+            const Node comp_minus_b = GetPredicateComparisonInteger(
+                PredCondition::NotEqual, is_signed_b,
+                SignedOperation(OperationCode::IBitwiseAnd, is_signed_b, op_b,
+                                Immediate(0x80000000)),
+                Immediate(0));
+
+            Node new_c = Operation(
+                OperationCode::Select, comp_minus_a,
+                SignedOperation(OperationCode::IAdd, is_signed_c, original_c, Immediate(-65536)),
+                original_c);
+            new_c = Operation(
+                OperationCode::Select, comp_minus_b,
+                SignedOperation(OperationCode::IAdd, is_signed_c, new_c, Immediate(-65536)),
+                std::move(new_c));
+
+            return Operation(OperationCode::Select, comp, original_c, std::move(new_c));
         }
         default:
-            UNIMPLEMENTED_MSG("Unhandled XMAD mode: {}", static_cast<u32>(instr.xmad.mode.Value()));
+            UNREACHABLE();
             return Immediate(0);
         }
     }();
@@ -102,18 +135,19 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     op_c = GetTemporary(1);
 
     // TODO(Rodrigo): Use an appropiate sign for this operation
-    Node sum = Operation(OperationCode::IAdd, product, op_c);
+    Node sum = SignedOperation(OperationCode::IAdd, is_signed_a, product, std::move(op_c));
     SetTemporary(bb, 2, sum);
     sum = GetTemporary(2);
     if (is_merge) {
-        const Node a = BitfieldExtract(sum, 0, 16);
-        const Node b =
-            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, original_b, Immediate(16));
-        sum = Operation(OperationCode::IBitwiseOr, NO_PRECISE, a, b);
+        const Node a = SignedOperation(OperationCode::IBitfieldExtract, is_signed_a, std::move(sum),
+                                       Immediate(0), Immediate(16));
+        const Node b = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed_b, original_b,
+                                       Immediate(16));
+        sum = SignedOperation(OperationCode::IBitwiseOr, is_signed_a, a, b);
     }
 
     SetInternalFlagsFromInteger(bb, sum, instr.generates_cc);
-    SetRegister(bb, instr.gpr0, sum);
+    SetRegister(bb, instr.gpr0, std::move(sum));
 
     return pc;
 }

src/video_core/shader/node.h

@@ -162,7 +162,21 @@ enum class OperationCode {
     AtomicImageXor,      /// (MetaImage, int[N] coords) -> void
     AtomicImageExchange, /// (MetaImage, int[N] coords) -> void
 
-    AtomicAdd, /// (memory, {u}int) -> {u}int
+    AtomicUExchange, /// (memory, uint) -> uint
+    AtomicUAdd,      /// (memory, uint) -> uint
+    AtomicUMin,      /// (memory, uint) -> uint
+    AtomicUMax,      /// (memory, uint) -> uint
+    AtomicUAnd,      /// (memory, uint) -> uint
+    AtomicUOr,       /// (memory, uint) -> uint
+    AtomicUXor,      /// (memory, uint) -> uint
+
+    AtomicIExchange, /// (memory, int) -> int
+    AtomicIAdd,      /// (memory, int) -> int
+    AtomicIMin,      /// (memory, int) -> int
+    AtomicIMax,      /// (memory, int) -> int
+    AtomicIAnd,      /// (memory, int) -> int
+    AtomicIOr,       /// (memory, int) -> int
+    AtomicIXor,      /// (memory, int) -> int
 
     Branch,         /// (uint branch_target) -> void
     BranchIndirect, /// (uint branch_target) -> void

src/video_core/shader/node_helper.cpp

@@ -68,6 +68,8 @@ OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed)
         return OperationCode::UBitwiseXor;
     case OperationCode::IBitwiseNot:
         return OperationCode::UBitwiseNot;
+    case OperationCode::IBitfieldExtract:
+        return OperationCode::UBitfieldExtract;
     case OperationCode::IBitfieldInsert:
         return OperationCode::UBitfieldInsert;
     case OperationCode::IBitCount:
@@ -84,6 +86,20 @@ OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed)
         return OperationCode::LogicalUNotEqual;
     case OperationCode::LogicalIGreaterEqual:
         return OperationCode::LogicalUGreaterEqual;
+    case OperationCode::AtomicIExchange:
+        return OperationCode::AtomicUExchange;
+    case OperationCode::AtomicIAdd:
+        return OperationCode::AtomicUAdd;
+    case OperationCode::AtomicIMin:
+        return OperationCode::AtomicUMin;
+    case OperationCode::AtomicIMax:
+        return OperationCode::AtomicUMax;
+    case OperationCode::AtomicIAnd:
+        return OperationCode::AtomicUAnd;
+    case OperationCode::AtomicIOr:
+        return OperationCode::AtomicUOr;
+    case OperationCode::AtomicIXor:
+        return OperationCode::AtomicUXor;
     case OperationCode::INegate:
         UNREACHABLE_MSG("Can't negate an unsigned integer");
         return {};

src/video_core/shader/shader_ir.cpp

@@ -96,6 +96,7 @@ Node ShaderIR::GetPredicate(bool immediate) {
 }
 
 Node ShaderIR::GetInputAttribute(Attribute::Index index, u64 element, Node buffer) {
+    MarkAttributeUsage(index, element);
     used_input_attributes.emplace(index);
     return MakeNode<AbufNode>(index, static_cast<u32>(element), std::move(buffer));
 }
@@ -106,42 +107,8 @@ Node ShaderIR::GetPhysicalInputAttribute(Tegra::Shader::Register physical_addres
 }
 
 Node ShaderIR::GetOutputAttribute(Attribute::Index index, u64 element, Node buffer) {
-    if (index == Attribute::Index::LayerViewportPointSize) {
-        switch (element) {
-        case 0:
-            UNIMPLEMENTED();
-            break;
-        case 1:
-            uses_layer = true;
-            break;
-        case 2:
-            uses_viewport_index = true;
-            break;
-        case 3:
-            uses_point_size = true;
-            break;
-        }
-    }
-    if (index == Attribute::Index::TessCoordInstanceIDVertexID) {
-        switch (element) {
-        case 2:
-            uses_instance_id = true;
-            break;
-        case 3:
-            uses_vertex_id = true;
-            break;
-        default:
-            break;
-        }
-    }
-    if (index == Attribute::Index::ClipDistances0123 ||
-        index == Attribute::Index::ClipDistances4567) {
-        const auto clip_index =
-            static_cast<u32>((index == Attribute::Index::ClipDistances4567 ? 1 : 0) + element);
-        used_clip_distances.at(clip_index) = true;
-    }
+    MarkAttributeUsage(index, element);
     used_output_attributes.insert(index);
-
     return MakeNode<AbufNode>(index, static_cast<u32>(element), std::move(buffer));
 }
 
@@ -452,6 +419,54 @@ Node ShaderIR::BitfieldInsert(Node base, Node insert, u32 offset, u32 bits) {
                      Immediate(bits));
 }
 
+void ShaderIR::MarkAttributeUsage(Attribute::Index index, u64 element) {
+    switch (index) {
+    case Attribute::Index::LayerViewportPointSize:
+        switch (element) {
+        case 0:
+            UNIMPLEMENTED();
+            break;
+        case 1:
+            uses_layer = true;
+            break;
+        case 2:
+            uses_viewport_index = true;
+            break;
+        case 3:
+            uses_point_size = true;
+            break;
+        }
+        break;
+    case Attribute::Index::TessCoordInstanceIDVertexID:
+        switch (element) {
+        case 2:
+            uses_instance_id = true;
+            break;
+        case 3:
+            uses_vertex_id = true;
+            break;
+        }
+        break;
+    case Attribute::Index::ClipDistances0123:
+    case Attribute::Index::ClipDistances4567: {
+        const u64 clip_index = (index == Attribute::Index::ClipDistances4567 ? 4 : 0) + element;
+        used_clip_distances.at(clip_index) = true;
+        break;
+    }
+    case Attribute::Index::FrontColor:
+    case Attribute::Index::FrontSecondaryColor:
+    case Attribute::Index::BackColor:
+    case Attribute::Index::BackSecondaryColor:
+        uses_legacy_varyings = true;
+        break;
+    default:
+        if (index >= Attribute::Index::TexCoord_0 && index <= Attribute::Index::TexCoord_7) {
+            uses_legacy_varyings = true;
+        }
+        break;
+    }
+}
+
 std::size_t ShaderIR::DeclareAmend(Node new_amend) {
     const std::size_t id = amend_code.size();
     amend_code.push_back(new_amend);

src/video_core/shader/shader_ir.h

@@ -137,6 +137,10 @@ public:
         return uses_vertex_id;
     }
 
+    bool UsesLegacyVaryings() const {
+        return uses_legacy_varyings;
+    }
+
     bool UsesWarps() const {
         return uses_warps;
     }
@@ -343,6 +347,9 @@ private:
     /// Inserts a sequence of bits from a node
     Node BitfieldInsert(Node base, Node insert, u32 offset, u32 bits);
 
+    /// Marks the usage of a input or output attribute.
+    void MarkAttributeUsage(Tegra::Shader::Attribute::Index index, u64 element);
+
     void WriteTexInstructionFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
                                   const Node4& components);
 
@@ -443,6 +450,7 @@ private:
     bool uses_physical_attributes{}; // Shader uses AL2P or physical attribute read/writes
     bool uses_instance_id{};
     bool uses_vertex_id{};
+    bool uses_legacy_varyings{};
     bool uses_warps{};
     bool uses_indexed_samplers{};
 

src/video_core/surface.cpp

@@ -111,6 +111,8 @@ PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format)
         return PixelFormat::RGBA16F;
     case Tegra::RenderTargetFormat::RGBA16_UNORM:
         return PixelFormat::RGBA16U;
+    case Tegra::RenderTargetFormat::RGBA16_SNORM:
+        return PixelFormat::RGBA16S;
     case Tegra::RenderTargetFormat::RGBA16_UINT:
         return PixelFormat::RGBA16UI;
     case Tegra::RenderTargetFormat::RGBA32_FLOAT:

src/video_core/surface.h

@@ -25,82 +25,83 @@ enum class PixelFormat {
     R8UI = 7,
     RGBA16F = 8,
     RGBA16U = 9,
-    RGBA16UI = 10,
-    R11FG11FB10F = 11,
-    RGBA32UI = 12,
-    DXT1 = 13,
-    DXT23 = 14,
-    DXT45 = 15,
-    DXN1 = 16, // This is also known as BC4
-    DXN2UNORM = 17,
-    DXN2SNORM = 18,
-    BC7U = 19,
-    BC6H_UF16 = 20,
-    BC6H_SF16 = 21,
-    ASTC_2D_4X4 = 22,
-    BGRA8 = 23,
-    RGBA32F = 24,
-    RG32F = 25,
-    R32F = 26,
-    R16F = 27,
-    R16U = 28,
-    R16S = 29,
-    R16UI = 30,
-    R16I = 31,
-    RG16 = 32,
-    RG16F = 33,
-    RG16UI = 34,
-    RG16I = 35,
-    RG16S = 36,
-    RGB32F = 37,
-    RGBA8_SRGB = 38,
-    RG8U = 39,
-    RG8S = 40,
-    RG32UI = 41,
-    RGBX16F = 42,
-    R32UI = 43,
-    R32I = 44,
-    ASTC_2D_8X8 = 45,
-    ASTC_2D_8X5 = 46,
-    ASTC_2D_5X4 = 47,
-    BGRA8_SRGB = 48,
-    DXT1_SRGB = 49,
-    DXT23_SRGB = 50,
-    DXT45_SRGB = 51,
-    BC7U_SRGB = 52,
-    R4G4B4A4U = 53,
-    ASTC_2D_4X4_SRGB = 54,
-    ASTC_2D_8X8_SRGB = 55,
-    ASTC_2D_8X5_SRGB = 56,
-    ASTC_2D_5X4_SRGB = 57,
-    ASTC_2D_5X5 = 58,
-    ASTC_2D_5X5_SRGB = 59,
-    ASTC_2D_10X8 = 60,
-    ASTC_2D_10X8_SRGB = 61,
-    ASTC_2D_6X6 = 62,
-    ASTC_2D_6X6_SRGB = 63,
-    ASTC_2D_10X10 = 64,
-    ASTC_2D_10X10_SRGB = 65,
-    ASTC_2D_12X12 = 66,
-    ASTC_2D_12X12_SRGB = 67,
-    ASTC_2D_8X6 = 68,
-    ASTC_2D_8X6_SRGB = 69,
-    ASTC_2D_6X5 = 70,
-    ASTC_2D_6X5_SRGB = 71,
-    E5B9G9R9F = 72,
+    RGBA16S = 10,
+    RGBA16UI = 11,
+    R11FG11FB10F = 12,
+    RGBA32UI = 13,
+    DXT1 = 14,
+    DXT23 = 15,
+    DXT45 = 16,
+    DXN1 = 17, // This is also known as BC4
+    DXN2UNORM = 18,
+    DXN2SNORM = 19,
+    BC7U = 20,
+    BC6H_UF16 = 21,
+    BC6H_SF16 = 22,
+    ASTC_2D_4X4 = 23,
+    BGRA8 = 24,
+    RGBA32F = 25,
+    RG32F = 26,
+    R32F = 27,
+    R16F = 28,
+    R16U = 29,
+    R16S = 30,
+    R16UI = 31,
+    R16I = 32,
+    RG16 = 33,
+    RG16F = 34,
+    RG16UI = 35,
+    RG16I = 36,
+    RG16S = 37,
+    RGB32F = 38,
+    RGBA8_SRGB = 39,
+    RG8U = 40,
+    RG8S = 41,
+    RG32UI = 42,
+    RGBX16F = 43,
+    R32UI = 44,
+    R32I = 45,
+    ASTC_2D_8X8 = 46,
+    ASTC_2D_8X5 = 47,
+    ASTC_2D_5X4 = 48,
+    BGRA8_SRGB = 49,
+    DXT1_SRGB = 50,
+    DXT23_SRGB = 51,
+    DXT45_SRGB = 52,
+    BC7U_SRGB = 53,
+    R4G4B4A4U = 54,
+    ASTC_2D_4X4_SRGB = 55,
+    ASTC_2D_8X8_SRGB = 56,
+    ASTC_2D_8X5_SRGB = 57,
+    ASTC_2D_5X4_SRGB = 58,
+    ASTC_2D_5X5 = 59,
+    ASTC_2D_5X5_SRGB = 60,
+    ASTC_2D_10X8 = 61,
+    ASTC_2D_10X8_SRGB = 62,
+    ASTC_2D_6X6 = 63,
+    ASTC_2D_6X6_SRGB = 64,
+    ASTC_2D_10X10 = 65,
+    ASTC_2D_10X10_SRGB = 66,
+    ASTC_2D_12X12 = 67,
+    ASTC_2D_12X12_SRGB = 68,
+    ASTC_2D_8X6 = 69,
+    ASTC_2D_8X6_SRGB = 70,
+    ASTC_2D_6X5 = 71,
+    ASTC_2D_6X5_SRGB = 72,
+    E5B9G9R9F = 73,
 
     MaxColorFormat,
 
     // Depth formats
-    Z32F = 73,
-    Z16 = 74,
+    Z32F = 74,
+    Z16 = 75,
 
     MaxDepthFormat,
 
     // DepthStencil formats
-    Z24S8 = 75,
-    S8Z24 = 76,
-    Z32FS8 = 77,
+    Z24S8 = 76,
+    S8Z24 = 77,
+    Z32FS8 = 78,
 
     MaxDepthStencilFormat,
 
@@ -138,6 +139,7 @@ constexpr std::array<u32, MaxPixelFormat> compression_factor_shift_table = {{
     0, // R8UI
     0, // RGBA16F
     0, // RGBA16U
+    0, // RGBA16S
     0, // RGBA16UI
     0, // R11FG11FB10F
     0, // RGBA32UI
@@ -235,6 +237,7 @@ constexpr std::array<u32, MaxPixelFormat> block_width_table = {{
     1,  // R8UI
     1,  // RGBA16F
     1,  // RGBA16U
+    1,  // RGBA16S
     1,  // RGBA16UI
     1,  // R11FG11FB10F
     1,  // RGBA32UI
@@ -324,6 +327,7 @@ constexpr std::array<u32, MaxPixelFormat> block_height_table = {{
     1,  // R8UI
     1,  // RGBA16F
     1,  // RGBA16U
+    1,  // RGBA16S
     1,  // RGBA16UI
     1,  // R11FG11FB10F
     1,  // RGBA32UI
@@ -413,6 +417,7 @@ constexpr std::array<u32, MaxPixelFormat> bpp_table = {{
     8,   // R8UI
     64,  // RGBA16F
     64,  // RGBA16U
+    64,  // RGBA16S
     64,  // RGBA16UI
     32,  // R11FG11FB10F
     128, // RGBA32UI
@@ -517,6 +522,7 @@ constexpr std::array<SurfaceCompression, MaxPixelFormat> compression_type_table
     SurfaceCompression::None,       // R8UI
     SurfaceCompression::None,       // RGBA16F
     SurfaceCompression::None,       // RGBA16U
+    SurfaceCompression::None,       // RGBA16S
     SurfaceCompression::None,       // RGBA16UI
     SurfaceCompression::None,       // R11FG11FB10F
     SurfaceCompression::None,       // RGBA32UI

src/video_core/texture_cache/format_lookup_table.cpp

@@ -41,7 +41,7 @@ struct Table {
     ComponentType alpha_component;
     bool is_srgb;
 };
-constexpr std::array<Table, 75> DefinitionTable = {{
+constexpr std::array<Table, 76> DefinitionTable = {{
     {TextureFormat::A8R8G8B8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ABGR8U},
     {TextureFormat::A8R8G8B8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::ABGR8S},
     {TextureFormat::A8R8G8B8, C, UINT, UINT, UINT, UINT, PixelFormat::ABGR8UI},
@@ -61,6 +61,7 @@ constexpr std::array<Table, 75> DefinitionTable = {{
     {TextureFormat::G8R8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RG8U},
     {TextureFormat::G8R8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RG8S},
 
+    {TextureFormat::R16_G16_B16_A16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RGBA16S},
     {TextureFormat::R16_G16_B16_A16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RGBA16U},
     {TextureFormat::R16_G16_B16_A16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::RGBA16F},
     {TextureFormat::R16_G16_B16_A16, C, UINT, UINT, UINT, UINT, PixelFormat::RGBA16UI},