Prepare frontend for multiple graphics APIs (#6347)

* externals: Update dynarmic

* settings: Introduce GraphicsAPI enum

* For now it's OpenGL only but will be expanded upon later

* citra_qt: Introduce backend agnostic context management

* Mostly a direct port from yuzu

* core: Simplify context acquire

* settings: Add option to create debug contexts

* renderer_opengl: Abstract initialization to Driver

* This commit also updates glad and adds some useful extensions which we will use in part 2

* Rasterizer construction is moved to the specific renderer instead of RendererBase.
  Software rendering has been disable to achieve this but will be brought back in the next commit.

* video_core: Remove Init/Shutdown methods from renderer

* The constructor and destructor can do the same job

* In addition move opengl function loading to Qt since SDL already does this. Also remove ErrorVideoCore which is never reached

* citra_qt: Decouple software renderer from opengl part 1

* citra: Decouple software renderer from opengl part 2

* android: Decouple software renderer from opengl part 3

* swrasterizer: Decouple software renderer from opengl part 4

* This commit simply enforces the renderer naming conventions in the software renderer

* video_core: Move RendererBase to VideoCore

* video_core: De-globalize screenshot state

* video_core: Pass system to the renderers

* video_core: Commonize shader uniform data

* video_core: Abstract backend agnostic rasterizer operations

* bootmanager: Remove references to OpenGL for macOS

OpenGL macOS headers definitions clash heavily with each other

* citra_qt: Proper title for api settings

* video_core: Reduce boost usage

* bootmanager: Fix hide mouse option

Remove event handlers from RenderWidget for events that are
already handled by the parent GRenderWindow.
Also enable mouse tracking on the RenderWidget.

* android: Remove software from graphics api list

* code: Address review comments

* citra: Port per-game settings read

* Having to update the default value for all backends is a pain so lets centralize it

* android: Rename to OpenGLES

---------

Co-authored-by: MerryMage <MerryMage@users.noreply.github.com>
Co-authored-by: Vitor Kiguchi <vitor-kiguchi@hotmail.com>

externals/dynarmic

@@ -1 +1 @@
-Subproject commit 9af4b970d302389829448a30608c7cb4fce9b662
+Subproject commit b3a92ab54dadd26a0c2a87d2677b80249d2e1a5a

externals/glad/Readme.md

@@ -1,5 +1,5 @@
 These files were generated by the [glad](https://github.com/Dav1dde/glad) OpenGL loader generator and have been checked in as-is. You can re-generate them using glad with the following command:
 
 ```
-python -m glad --profile core --out-path glad/ --api "gl=3.3,gles2=3.2" --generator=c
+python -m glad --profile core --out-path glad/ --api "gl=4.6,gles2=3.2" --generator=c
 ```

src/android/app/src/main/java/org/citra/citra_emu/NativeLibrary.java

@@ -496,9 +496,6 @@ public final class NativeLibrary {
         final int ErrorLoader_ErrorEncrypted = 5;
         final int ErrorLoader_ErrorInvalidFormat = 6;
         final int ErrorSystemFiles = 7;
-        final int ErrorVideoCore = 8;
-        final int ErrorVideoCore_ErrorGenericDrivers = 9;
-        final int ErrorVideoCore_ErrorBelowGL33 = 10;
         final int ShutdownRequested = 11;
         final int ErrorUnknown = 12;
 

src/android/app/src/main/java/org/citra/citra_emu/features/settings/ui/SettingsFragmentPresenter.java

@@ -355,6 +355,7 @@ public final class SettingsFragmentPresenter {
         mView.getActivity().setTitle(R.string.preferences_graphics);
 
         SettingSection rendererSection = mSettings.getSection(Settings.SECTION_RENDERER);
+        Setting graphicsApi = rendererSection.getSetting(SettingsFile.KEY_GRAPHICS_API);
         Setting resolutionFactor = rendererSection.getSetting(SettingsFile.KEY_RESOLUTION_FACTOR);
         Setting filterMode = rendererSection.getSetting(SettingsFile.KEY_FILTER_MODE);
         Setting shadersAccurateMul = rendererSection.getSetting(SettingsFile.KEY_SHADERS_ACCURATE_MUL);
@@ -371,6 +372,7 @@ public final class SettingsFragmentPresenter {
         //Setting preloadTextures = utilitySection.getSetting(SettingsFile.KEY_PRELOAD_TEXTURES);
 
         sl.add(new HeaderSetting(null, null, R.string.renderer, 0));
+        sl.add(new SingleChoiceSetting(SettingsFile.KEY_GRAPHICS_API, Settings.SECTION_RENDERER, R.string.graphics_api, 0, R.array.graphicsApiNames, R.array.graphicsApiValues, 0, graphicsApi));
         sl.add(new SliderSetting(SettingsFile.KEY_RESOLUTION_FACTOR, Settings.SECTION_RENDERER, R.string.internal_resolution, R.string.internal_resolution_description, 1, 4, "x", 1, resolutionFactor));
         sl.add(new CheckBoxSetting(SettingsFile.KEY_FILTER_MODE, Settings.SECTION_RENDERER, R.string.linear_filtering, R.string.linear_filtering_description, true, filterMode));
         sl.add(new CheckBoxSetting(SettingsFile.KEY_SHADERS_ACCURATE_MUL, Settings.SECTION_RENDERER, R.string.shaders_accurate_mul, R.string.shaders_accurate_mul_description, false, shadersAccurateMul));
@@ -409,14 +411,14 @@ public final class SettingsFragmentPresenter {
         SettingSection coreSection = mSettings.getSection(Settings.SECTION_CORE);
         SettingSection rendererSection = mSettings.getSection(Settings.SECTION_RENDERER);
         Setting useCpuJit = coreSection.getSetting(SettingsFile.KEY_CPU_JIT);
-        Setting hardwareRenderer = rendererSection.getSetting(SettingsFile.KEY_HW_RENDERER);
         Setting hardwareShader = rendererSection.getSetting(SettingsFile.KEY_HW_SHADER);
         Setting vsyncEnable = rendererSection.getSetting(SettingsFile.KEY_USE_VSYNC);
+        Setting rendererDebug = rendererSection.getSetting(SettingsFile.KEY_RENDERER_DEBUG);
 
         sl.add(new HeaderSetting(null, null, R.string.debug_warning, 0));
         sl.add(new CheckBoxSetting(SettingsFile.KEY_CPU_JIT, Settings.SECTION_CORE, R.string.cpu_jit, R.string.cpu_jit_description, true, useCpuJit, true, mView));
-        sl.add(new CheckBoxSetting(SettingsFile.KEY_HW_RENDERER, Settings.SECTION_RENDERER, R.string.hw_renderer, R.string.hw_renderer_description, true, hardwareRenderer, true, mView));
         sl.add(new CheckBoxSetting(SettingsFile.KEY_HW_SHADER, Settings.SECTION_RENDERER, R.string.hw_shaders, R.string.hw_shaders_description, true, hardwareShader, true, mView));
         sl.add(new CheckBoxSetting(SettingsFile.KEY_USE_VSYNC, Settings.SECTION_RENDERER, R.string.vsync, R.string.vsync_description, true, vsyncEnable));
+        sl.add(new CheckBoxSetting(SettingsFile.KEY_RENDERER_DEBUG, Settings.SECTION_RENDERER, R.string.renderer_debug, R.string.renderer_debug_description, false, rendererDebug));
     }
 }

src/android/app/src/main/java/org/citra/citra_emu/features/settings/utils/SettingsFile.java

@@ -44,7 +44,8 @@ public final class SettingsFile {
 
     public static final String KEY_PREMIUM = "premium";
 
-    public static final String KEY_HW_RENDERER = "use_hw_renderer";
+    public static final String KEY_GRAPHICS_API = "graphics_api";
+    public static final String KEY_RENDERER_DEBUG = "renderer_debug";
     public static final String KEY_HW_SHADER = "use_hw_shader";
     public static final String KEY_SHADERS_ACCURATE_MUL = "shaders_accurate_mul";
     public static final String KEY_USE_SHADER_JIT = "use_shader_jit";

src/android/app/src/main/jni/config.cpp

@@ -82,6 +82,30 @@ void Config::UpdateCFG() {
     cfg->UpdateConfigNANDSavegame();
 }
 
+template <>
+void Config::ReadSetting(const std::string& group, Settings::Setting<std::string>& setting) {
+    std::string setting_value = sdl2_config->Get(group, setting.GetLabel(), setting.GetDefault());
+    if (setting_value.empty()) {
+        setting_value = setting.GetDefault();
+    }
+    setting = std::move(setting_value);
+}
+
+template <>
+void Config::ReadSetting(const std::string& group, Settings::Setting<bool>& setting) {
+    setting = sdl2_config->GetBoolean(group, setting.GetLabel(), setting.GetDefault());
+}
+
+template <typename Type, bool ranged>
+void Config::ReadSetting(const std::string& group, Settings::Setting<Type, ranged>& setting) {
+    if constexpr (std::is_floating_point_v<Type>) {
+        setting = sdl2_config->GetReal(group, setting.GetLabel(), setting.GetDefault());
+    } else {
+        setting = static_cast<Type>(sdl2_config->GetInteger(
+            group, setting.GetLabel(), static_cast<long>(setting.GetDefault())));
+    }
+}
+
 void Config::ReadValues() {
     // Controls
     for (int i = 0; i < Settings::NativeButton::NumButtons; ++i) {
@@ -112,39 +136,32 @@ void Config::ReadValues() {
                                                  InputCommon::CemuhookUDP::DEFAULT_PORT));
 
     // Core
-    Settings::values.use_cpu_jit = sdl2_config->GetBoolean("Core", "use_cpu_jit", true);
+    ReadSetting("Core", Settings::values.use_cpu_jit);
-    Settings::values.cpu_clock_percentage =
+    ReadSetting("Core", Settings::values.cpu_clock_percentage);
-        static_cast<int>(sdl2_config->GetInteger("Core", "cpu_clock_percentage", 100));
 
     // Premium
-    Settings::values.texture_filter_name =
+    ReadSetting("Premium", Settings::values.texture_filter_name);
-        sdl2_config->GetString("Premium", "texture_filter_name", "none");
 
     // Renderer
     Settings::values.use_gles = sdl2_config->GetBoolean("Renderer", "use_gles", true);
-    Settings::values.use_hw_renderer = sdl2_config->GetBoolean("Renderer", "use_hw_renderer", true);
-    Settings::values.use_hw_shader = sdl2_config->GetBoolean("Renderer", "use_hw_shader", true);
     Settings::values.shaders_accurate_mul =
         sdl2_config->GetBoolean("Renderer", "shaders_accurate_mul", false);
-    Settings::values.use_shader_jit = sdl2_config->GetBoolean("Renderer", "use_shader_jit", true);
+    ReadSetting("Renderer", Settings::values.graphics_api);
-    Settings::values.resolution_factor =
+    ReadSetting("Renderer", Settings::values.use_hw_shader);
-        static_cast<u16>(sdl2_config->GetInteger("Renderer", "resolution_factor", 1));
+    ReadSetting("Renderer", Settings::values.use_shader_jit);
-    Settings::values.use_disk_shader_cache =
+    ReadSetting("Renderer", Settings::values.resolution_factor);
-        sdl2_config->GetBoolean("Renderer", "use_disk_shader_cache", true);
+    ReadSetting("Renderer", Settings::values.use_disk_shader_cache);
-    Settings::values.use_vsync_new = sdl2_config->GetBoolean("Renderer", "use_vsync_new", true);
+    ReadSetting("Renderer", Settings::values.use_vsync_new);
 
     // Work around to map Android setting for enabling the frame limiter to the format Citra expects
     if (sdl2_config->GetBoolean("Renderer", "use_frame_limit", true)) {
-        Settings::values.frame_limit =
+        ReadSetting("Renderer", Settings::values.frame_limit);
-            static_cast<u16>(sdl2_config->GetInteger("Renderer", "frame_limit", 100));
     } else {
         Settings::values.frame_limit = 0;
     }
 
-    Settings::values.render_3d = static_cast<Settings::StereoRenderOption>(
+    ReadSetting("Renderer", Settings::values.render_3d);
-        sdl2_config->GetInteger("Renderer", "render_3d", 0));
+    ReadSetting("Renderer", Settings::values.factor_3d);
-    Settings::values.factor_3d =
-        static_cast<u8>(sdl2_config->GetInteger("Renderer", "factor_3d", 0));
     std::string default_shader = "none (builtin)";
     if (Settings::values.render_3d.GetValue() == Settings::StereoRenderOption::Anaglyph)
         default_shader = "dubois (builtin)";
@@ -152,70 +169,49 @@ void Config::ReadValues() {
         default_shader = "horizontal (builtin)";
     Settings::values.pp_shader_name =
         sdl2_config->GetString("Renderer", "pp_shader_name", default_shader);
-    Settings::values.filter_mode = sdl2_config->GetBoolean("Renderer", "filter_mode", true);
+    ReadSetting("Renderer", Settings::values.filter_mode);
 
-    Settings::values.bg_red = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_red", 0.0));
+    ReadSetting("Renderer", Settings::values.bg_red);
-    Settings::values.bg_green =
+    ReadSetting("Renderer", Settings::values.bg_green);
-        static_cast<float>(sdl2_config->GetReal("Renderer", "bg_green", 0.0));
+    ReadSetting("Renderer", Settings::values.bg_blue);
-    Settings::values.bg_blue = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_blue", 0.0));
 
     // Layout
     Settings::values.layout_option = static_cast<Settings::LayoutOption>(sdl2_config->GetInteger(
         "Layout", "layout_option", static_cast<int>(Settings::LayoutOption::MobileLandscape)));
-    Settings::values.custom_layout = sdl2_config->GetBoolean("Layout", "custom_layout", false);
+    ReadSetting("Layout", Settings::values.custom_layout);
-    Settings::values.custom_top_left =
+    ReadSetting("Layout", Settings::values.custom_top_left);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_top_left", 0));
+    ReadSetting("Layout", Settings::values.custom_top_top);
-    Settings::values.custom_top_top =
+    ReadSetting("Layout", Settings::values.custom_top_right);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_top_top", 0));
+    ReadSetting("Layout", Settings::values.custom_top_bottom);
-    Settings::values.custom_top_right =
+    ReadSetting("Layout", Settings::values.custom_bottom_left);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_top_right", 400));
+    ReadSetting("Layout", Settings::values.custom_bottom_top);
-    Settings::values.custom_top_bottom =
+    ReadSetting("Layout", Settings::values.custom_bottom_right);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_top_bottom", 240));
+    ReadSetting("Layout", Settings::values.custom_bottom_bottom);
-    Settings::values.custom_bottom_left =
+    ReadSetting("Layout", Settings::values.cardboard_screen_size);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_bottom_left", 40));
+    ReadSetting("Layout", Settings::values.cardboard_x_shift);
-    Settings::values.custom_bottom_top =
+    ReadSetting("Layout", Settings::values.cardboard_y_shift);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_bottom_top", 240));
-    Settings::values.custom_bottom_right =
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_bottom_right", 360));
-    Settings::values.custom_bottom_bottom =
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_bottom_bottom", 480));
-    Settings::values.cardboard_screen_size =
-        static_cast<int>(sdl2_config->GetInteger("Layout", "cardboard_screen_size", 85));
-    Settings::values.cardboard_x_shift =
-        static_cast<int>(sdl2_config->GetInteger("Layout", "cardboard_x_shift", 0));
-    Settings::values.cardboard_y_shift =
-        static_cast<int>(sdl2_config->GetInteger("Layout", "cardboard_y_shift", 0));
 
     // Utility
-    Settings::values.dump_textures = sdl2_config->GetBoolean("Utility", "dump_textures", false);
+    ReadSetting("Utility", Settings::values.dump_textures);
-    Settings::values.custom_textures = sdl2_config->GetBoolean("Utility", "custom_textures", false);
+    ReadSetting("Utility", Settings::values.custom_textures);
-    Settings::values.preload_textures =
+    ReadSetting("Utility", Settings::values.preload_textures);
-        sdl2_config->GetBoolean("Utility", "preload_textures", false);
 
     // Audio
-    Settings::values.audio_emulation =
+    ReadSetting("Audio", Settings::values.audio_emulation);
-        static_cast<Settings::AudioEmulation>(sdl2_config->GetInteger(
+    ReadSetting("Audio", Settings::values.sink_id);
-            "Audio", "audio_emulation", static_cast<int>(Settings::AudioEmulation::HLE)));
+    ReadSetting("Audio", Settings::values.enable_audio_stretching);
-    Settings::values.sink_id = sdl2_config->GetString("Audio", "output_engine", "auto");
+    ReadSetting("Audio", Settings::values.audio_device_id);
-    Settings::values.enable_audio_stretching =
+    ReadSetting("Audio", Settings::values.volume);
-        sdl2_config->GetBoolean("Audio", "enable_audio_stretching", true);
+    ReadSetting("Audio", Settings::values.mic_input_device);
-    Settings::values.audio_device_id = sdl2_config->GetString("Audio", "output_device", "auto");
+    ReadSetting("Audio", Settings::values.mic_input_type);
-    Settings::values.volume = static_cast<float>(sdl2_config->GetReal("Audio", "volume", 1));
-    Settings::values.mic_input_device =
-        sdl2_config->GetString("Audio", "mic_input_device", "Default");
-    Settings::values.mic_input_type =
-        static_cast<Settings::MicInputType>(sdl2_config->GetInteger("Audio", "mic_input_type", 1));
 
     // Data Storage
-    Settings::values.use_virtual_sd =
+    ReadSetting("Data Storage", Settings::values.use_virtual_sd);
-        sdl2_config->GetBoolean("Data Storage", "use_virtual_sd", true);
 
     // System
-    Settings::values.is_new_3ds = sdl2_config->GetBoolean("System", "is_new_3ds", true);
+    ReadSetting("System", Settings::values.is_new_3ds);
-    Settings::values.region_value =
+    ReadSetting("System", Settings::values.region_value);
-        sdl2_config->GetInteger("System", "region_value", Settings::REGION_VALUE_AUTO_SELECT);
+    ReadSetting("System", Settings::values.init_clock);
-    Settings::values.init_clock =
-        static_cast<Settings::InitClock>(sdl2_config->GetInteger("System", "init_clock", 0));
     {
         std::tm t;
         t.tm_sec = 1;
@@ -236,10 +232,8 @@ void Config::ReadValues() {
                 std::chrono::system_clock::from_time_t(std::mktime(&t)).time_since_epoch())
                 .count();
     }
-    Settings::values.plugin_loader_enabled =
+    ReadSetting("System", Settings::values.plugin_loader_enabled);
-        sdl2_config->GetBoolean("System", "plugin_loader", false);
+    ReadSetting("System", Settings::values.allow_plugin_loader);
-    Settings::values.allow_plugin_loader =
-        sdl2_config->GetBoolean("System", "allow_plugin_loader", true);
 
     // Camera
     using namespace Service::CAM;
@@ -263,14 +257,14 @@ void Config::ReadValues() {
         sdl2_config->GetInteger("Camera", "camera_outer_left_flip", 0);
 
     // Miscellaneous
-    Settings::values.log_filter = sdl2_config->GetString("Miscellaneous", "log_filter", "*:Info");
+    ReadSetting("Miscellaneous", Settings::values.log_filter);
 
     // Debugging
     Settings::values.record_frame_times =
         sdl2_config->GetBoolean("Debugging", "record_frame_times", false);
-    Settings::values.use_gdbstub = sdl2_config->GetBoolean("Debugging", "use_gdbstub", false);
+    ReadSetting("Debugging", Settings::values.renderer_debug);
-    Settings::values.gdbstub_port =
+    ReadSetting("Debugging", Settings::values.use_gdbstub);
-        static_cast<u16>(sdl2_config->GetInteger("Debugging", "gdbstub_port", 24689));
+    ReadSetting("Debugging", Settings::values.gdbstub_port);
 
     for (const auto& service_module : Service::service_module_map) {
         bool use_lle = sdl2_config->GetBoolean("Debugging", "LLE\\" + service_module.name, false);

src/android/app/src/main/jni/config.h

@@ -6,6 +6,7 @@
 
 #include <memory>
 #include <string>
+#include "common/settings.h"
 
 class INIReader;
 
@@ -23,4 +24,14 @@ public:
     ~Config();
 
     void Reload();
+
+private:
+    /**
+     * Applies a value read from the sdl2_config to a Setting.
+     *
+     * @param group The name of the INI group
+     * @param setting The yuzu setting to modify
+     */
+    template <typename Type, bool ranged>
+    void ReadSetting(const std::string& group, Settings::Setting<Type, ranged>& setting);
 };

src/android/app/src/main/jni/default_ini.h

@@ -98,13 +98,9 @@ use_cpu_jit =
 cpu_clock_percentage =
 
 [Renderer]
-# Whether to render using GLES or OpenGL
+# Whether to render using OpenGL
-# 0: OpenGL, 1 (default): GLES
+# 1: OpenGLES (default)
-use_gles =
+graphics_api =
-
-# Whether to use software or hardware rendering.
-# 0: Software, 1 (default): Hardware
-use_hw_renderer =
 
 # Whether to use hardware shaders to emulate 3DS shaders
 # 0: Software, 1 (default): Hardware
@@ -118,10 +114,6 @@ separable_shader =
 # 0: Off (Default. Faster, but causes issues in some games) 1: On (Slower, but correct)
 shaders_accurate_mul =
 
-# Enable asynchronous GPU emulation
-# 0: Off (Slower, but more accurate) 1: On (Default. Faster, but may cause issues in some games)
-use_asynchronous_gpu_emulation =
-
 # Whether to use the Just-In-Time (JIT) compiler for shader emulation
 # 0: Interpreter (slow), 1 (default): JIT (fast)
 use_shader_jit =
@@ -325,9 +317,15 @@ log_filter = *:Info
 [Debugging]
 # Record frame time data, can be found in the log directory. Boolean value
 record_frame_times =
+
+# Whether to enable additional debugging information during emulation
+# 0 (default): Off, 1: On
+renderer_debug =
+
 # Port for listening to GDB connections.
 use_gdbstub=false
 gdbstub_port=24689
+
 # To LLE a service module add "LLE\<module name>=true"
 
 [WebService]

src/android/app/src/main/jni/native.cpp

@@ -25,7 +25,6 @@
 #include "core/frontend/applets/default_applets.h"
 #include "core/frontend/camera/factory.h"
 #include "core/frontend/mic.h"
-#include "core/frontend/scope_acquire_context.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/nfc/nfc.h"
 #include "core/savestate.h"
@@ -46,6 +45,7 @@
 #include "jni/ndk_motion.h"
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_opengl/texture_filters/texture_filterer.h"
+#include "video_core/video_core.h"
 
 namespace {
 
@@ -149,7 +149,15 @@ static Core::System::ResultStatus RunCitra(const std::string& filepath) {
         return Core::System::ResultStatus::ErrorLoader;
     }
 
+    const auto graphics_api = Settings::values.graphics_api.GetValue();
+    switch (graphics_api) {
+    case Settings::GraphicsAPI::OpenGL:
         window = std::make_unique<EmuWindow_Android>(s_surf);
+        break;
+    default:
+        LOG_CRITICAL(Frontend, "Unknown graphics API {}, using OpenGL", graphics_api);
+        window = std::make_unique<EmuWindow_Android>(s_surf);
+    }
 
     Core::System& system{Core::System::GetInstance()};
 

src/android/app/src/main/res/values-de/strings.xml

@@ -58,8 +58,6 @@
     <string name="linear_filtering_description">Aktiviert lineare Filterung, welche die Spieletexturen glättet.</string>
     <string name="texture_filter_name">Texturfilter</string>
     <string name="texture_filter_description">Verbessert die Grafik von Spielen durch das Anwendung eines Texturfilters. Die unterstützten Filter sind Anime4K Ultrafast, Bicubic, ScaleForce und xBRZ freescale.</string>
-    <string name="hw_renderer">Aktiviere Hardware Renderer</string>
-    <string name="hw_renderer_description">Benutzt Hardware, um die 3DS-Grafik zu emulieren. Wenn aktiviert, wird die Spieleleistung stark verbessert.</string>
     <string name="hw_shaders">Aktiviere Hardware Shader</string>
     <string name="hw_shaders_description">Benutzt Hardware, um die 3DS-Shader zu emulieren. Wenn aktiviert, wird die Spieleleistung stark verbessert.</string>
     <string name="shaders_accurate_mul">Aktiviere genaue Shader-Multiplikation</string>

src/android/app/src/main/res/values-es/strings.xml

@@ -62,8 +62,6 @@
     <string name="linear_filtering_description">Activa el filtro linear, que hace que los gráficos del juego se vean más suaves.</string>
     <string name="texture_filter_name">Filtro de Texturas</string>
     <string name="texture_filter_description">Mejora los gráficos visuales de los juegos al aplicar un filtro a las texturas. Los filtros soportados son Anime4K Ultrafast, Bicubic, ScaleForce, y xBRZ freescale.</string>
-    <string name="hw_renderer">Activar renderizador de hardware</string>
-    <string name="hw_renderer_description">Usa el hardware para emular los gráficos de 3DS. Cuando se active, el rendimiento mejorará notablemente.</string>
     <string name="hw_shaders">Activar sombreador de hardware</string>
     <string name="hw_shaders_description">Usa el hardware para emular los sombreadores de 3DS. Cuando se active, el rendimiento mejorará notablemente.</string>
     <string name="shaders_accurate_mul">Activar multiplicación precisa de sombreado</string>

src/android/app/src/main/res/values-fi/strings.xml

@@ -46,8 +46,6 @@
     <!-- Graphics settings strings -->
     <string name="vsync">Aktivoi V-Sync</string>
     <string name="vsync_description">Synkronoi pelin virkistystaajuus laitteesi virkistystaajuuteen.</string>
-    <string name="hw_renderer">Aktivoi Laitteistorenderöinti</string>
-    <string name="hw_renderer_description">Käyttää laitteistoa emuloidakseen 3DS-grafiikoita. Kun tämä on päällä, pelien suorituskyky on huomattavasti parempi.</string>
     <string name="hw_shaders">Aktivoi Laitteistovarjostin</string>
     <string name="hw_shaders_description">Käyttää laitteistoa emuloidakseen 3DS:n varjostimia. Kun tämä on päällä, pelien suorituskyky on huomattavasti parempi.</string>
     <string name="frame_limit_enable">Aktivoi nopeuden rajoitus</string>

src/android/app/src/main/res/values-fr/strings.xml

@@ -58,8 +58,6 @@
     <string name="linear_filtering_description">Active le filtrage linéaire, qui améliorera le lissage graphique du jeu.</string>
     <string name="texture_filter_name">Filtrage des textures</string>
     <string name="texture_filter_description">Améliore l\'aspect graphique des jeux en appliquant un filtre aux textures. Les filtres supportés sont Anime4K Ultrafast, Bicubic, ScaleForce, et xBRZ freescale.</string>
-    <string name="hw_renderer">Activer le rendu matériel</string>
-    <string name="hw_renderer_description">Utilise le matériel pour émuler les graphismes de la 3DS. Lorsqu\'il est activé, la performance des jeux sera améliorée de manière significative.</string>
     <string name="hw_shaders">Activer le shader (nuanceur) matériel </string>
     <string name="hw_shaders_description">Utilise le matériel pour émuler les shaders de la 3DS. Lorsqu\'il est activé, la performance des jeux sera améliorée de manière significative.</string>
     <string name="shaders_accurate_mul">Activer la multiplication précise dans les shaders</string>

src/android/app/src/main/res/values-it/strings.xml

@@ -58,8 +58,6 @@
     <string name="linear_filtering_description">Abilita il filtro lineare, che fa sembrare più smussata la grafica dei giochi.</string>
     <string name="texture_filter_name">Filtro Texture</string>
     <string name="texture_filter_description">Migliora la grafica dei giochi applicando un filtro alle texture. I filtri supportati sono Anime4k Ultrafast, Bicubic, ScaleForce, e xBRZ freescale.</string>
-    <string name="hw_renderer">Abilita renderer hardware</string>
-    <string name="hw_renderer_description">Utilizza l\'hardware per emulare la grafica del 3DS. Se abilitato, le prestazioni dei giochi miglioreranno significativamente.</string>
     <string name="hw_shaders">Abilita shader hardware</string>
     <string name="hw_shaders_description">Utilizza l\'hardware per emulare gli shader del 3DS. Se abilitato, le prestazioni dei giochi miglioreranno significativamente.</string>
     <string name="shaders_accurate_mul">Abilita moltiplicazione shader accurata</string>

src/android/app/src/main/res/values-ja/strings.xml

@@ -40,8 +40,6 @@
     <string name="linear_filtering">リニアフィルタリングを有効化</string>
     <string name="linear_filtering_description">有効にすると、よりなめらかな画質が期待できます。</string>
     <string name="texture_filter_name">テクスチャフィルタ</string>
-    <string name="hw_renderer">ハードウェアレンダラを有効にする</string>
-    <string name="hw_renderer_description">グラフィックエミュレーションにハードウェアを使用します。有効にすると、パフォーマンスが大幅に向上します。</string>
     <string name="hw_shaders">ハードウェアシェーダを有効にする</string>
     <string name="hw_shaders_description">シェーダエミュレーションにハードウェアを使用します。有効にすると、パフォーマンスが大幅に向上します。</string>
     <string name="shaders_accurate_mul">正確なシェーダ乗算を有効にする</string>

src/android/app/src/main/res/values-ko/strings.xml

@@ -58,8 +58,6 @@
     <string name="linear_filtering_description">게임 필터링이 매끄럽게 보이도록 선형 필터링을 활성화합니다.</string>
     <string name="texture_filter_name">텍스처 필터</string>
     <string name="texture_filter_description">텍스처에 필터를 적용하여 게임의 시각적 효과를 향상시킵니다. 지원되는 필터는 Anime4K Ultrafast, Bicubic, ScaleForce 및 xBRZ 프리스케일입니다.</string>
-    <string name="hw_renderer">하드웨어 렌더러 사용</string>
-    <string name="hw_renderer_description">하드웨어를 사용하여 3DS 그래픽을 에뮬레이션합니다. 활성화하면 게임 성능이 크게 향상됩니다.</string>
     <string name="hw_shaders">하드웨어 쉐이더 사용</string>
     <string name="hw_shaders_description">하드웨어를 사용하여 3DS 쉐이더를 에뮬레이션합니다. 활성화하면 게임 성능이 크게 향상됩니다.</string>
     <string name="shaders_accurate_mul">정확한 쉐이더 곱셉 사용</string>

src/android/app/src/main/res/values-nb/strings.xml

@@ -58,8 +58,6 @@
     <string name="linear_filtering_description">Aktiverer lineær filtrering, noe som får spillvisualer til å vises jevnere.</string>
     <string name="texture_filter_name">Tekstur Filter</string>
     <string name="texture_filter_description">Forbedrer det visuelle i spill ved å bruke et filter på teksturer. De støttede filtrene er Anime4K Ultrafast, Bicubic, ScaleForce og xBRZ freescale.</string>
-    <string name="hw_renderer">Aktiver maskinvaregjengivelse</string>
-    <string name="hw_renderer_description">Bruker maskinvare til å emulere 3DS grafikk. Når dette er aktivert, vil spillytelsen bli betydelig forbedret.</string>
     <string name="hw_shaders">Aktiver maskinvare shader</string>
     <string name="hw_shaders_description">Bruker maskinvare for å etterligne 3DS shaders. Når dette er aktivert, vil spillytelsen bli betydelig forbedret.</string>
     <string name="shaders_accurate_mul">Aktiver nøyaktig shader-multiplikasjon</string>

src/android/app/src/main/res/values-pt/strings.xml

@@ -58,8 +58,6 @@
     <string name="linear_filtering_description">Ativa a filtragem linear, que suaviza o visual do jogo.</string>
     <string name="texture_filter_name">Filtro de texturas</string>
     <string name="texture_filter_description">Aprimora o visual dos jogos ao aplicar filtros às texturas. Os filtros compatíveis são: Anime4K Ultrafast, Bicúbico, ScaleForce e xBRZ Freescale.</string>
-    <string name="hw_renderer">Ativar renderizador por hardware</string>
-    <string name="hw_renderer_description">Utiliza o hardware para emular os gráficos do 3DS. Quando ativado, o desempenho de jogo será consideravelmente melhorado.</string>
     <string name="hw_shaders">Ativar shaders via hardware</string>
     <string name="hw_shaders_description">Utiliza o hardware para emular os shaders do 3DS. Quando ativado, o desempenho do jogo será consideravelmente melhorado.</string>
     <string name="shaders_accurate_mul">Ativar multiplicação precisa de shaders</string>

src/android/app/src/main/res/values-zh/strings.xml

@@ -58,8 +58,6 @@
     <string name="linear_filtering_description">开启后，游戏视觉效果会更加平滑。</string>
     <string name="texture_filter_name">纹理滤镜</string>
     <string name="texture_filter_description">通过对纹理使用滤镜来增强游戏的视觉效果。支持的滤镜有 Anime4K Ultrafast, Bicubic, ScaleForce 和 xBRZ freescale。</string>
-    <string name="hw_renderer">启用硬件渲染器</string>
-    <string name="hw_renderer_description">使用硬件模拟 3DS 图形。启用后，游戏性能将显著提高。</string>
     <string name="hw_shaders">启用硬件着色器</string>
     <string name="hw_shaders_description">使用硬件模拟 3DS 着色器。启用后，游戏性能将显著提高。</string>
     <string name="shaders_accurate_mul">启用精确乘法运算</string>

src/android/app/src/main/res/values/arrays.xml

@@ -171,4 +171,12 @@
         <item>4</item>
         <item>5</item>
     </integer-array>
+
+    <string-array name="graphicsApiNames">
+        <item>OpenGLES</item>
+    </string-array>
+
+    <integer-array name="graphicsApiValues">
+        <item>1</item>
+    </integer-array>
 </resources>

src/android/app/src/main/res/values/strings.xml

@@ -72,14 +72,15 @@
 
     <!-- Graphics settings strings -->
     <string name="renderer">Renderer</string>
+    <string name="graphics_api">Graphics API</string>
+    <string name="renderer_debug">Enable debug renderer</string>
+    <string name="renderer_debug_description">Log additional graphics related debug information. When enabled, game performance will be significantly reduced</string>
     <string name="vsync">Enable V-Sync</string>
     <string name="vsync_description">Synchronizes the game frame rate to the refresh rate of your device.</string>
     <string name="linear_filtering">Enable linear filtering</string>
     <string name="linear_filtering_description">Enables linear filtering, which causes game visuals to appear smoother.</string>
     <string name="texture_filter_name">Texture Filter</string>
     <string name="texture_filter_description">Enhances the visuals of games by applying a filter to textures. The supported filters are Anime4K Ultrafast, Bicubic, ScaleForce, and xBRZ freescale.</string>
-    <string name="hw_renderer">Enable hardware renderer</string>
-    <string name="hw_renderer_description">Uses hardware to emulate 3DS graphics. When enabled, game performance will be significantly improved.</string>
     <string name="hw_shaders">Enable hardware shader</string>
     <string name="hw_shaders_description">Uses hardware to emulate 3DS shaders. When enabled, game performance will be significantly improved.</string>
     <string name="shaders_accurate_mul">Enable accurate shader multiplication</string>

src/citra/CMakeLists.txt

@@ -8,6 +8,10 @@ add_executable(citra
     default_ini.h
     emu_window/emu_window_sdl2.cpp
     emu_window/emu_window_sdl2.h
+    emu_window/emu_window_sdl2_gl.cpp
+    emu_window/emu_window_sdl2_gl.h
+    emu_window/emu_window_sdl2_sw.cpp
+    emu_window/emu_window_sdl2_sw.h
     lodepng_image_interface.cpp
     lodepng_image_interface.h
     precompiled_headers.h

src/citra/citra.cpp

@@ -13,6 +13,8 @@
 
 #include "citra/config.h"
 #include "citra/emu_window/emu_window_sdl2.h"
+#include "citra/emu_window/emu_window_sdl2_gl.h"
+#include "citra/emu_window/emu_window_sdl2_sw.h"
 #include "citra/lodepng_image_interface.h"
 #include "common/common_paths.h"
 #include "common/detached_tasks.h"
@@ -29,7 +31,6 @@
 #include "core/file_sys/cia_container.h"
 #include "core/frontend/applets/default_applets.h"
 #include "core/frontend/framebuffer_layout.h"
-#include "core/frontend/scope_acquire_context.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/cfg/cfg.h"
@@ -38,6 +39,7 @@
 #include "input_common/main.h"
 #include "network/network.h"
 #include "video_core/renderer_base.h"
+#include "video_core/video_core.h"
 
 #undef _UNICODE
 #include <getopt.h>
@@ -362,13 +364,23 @@ int main(int argc, char** argv) {
 
     EmuWindow_SDL2::InitializeSDL2();
 
-    const auto emu_window{std::make_unique<EmuWindow_SDL2>(fullscreen, false)};
+    const auto CreateEmuWindow = [](bool fullscreen,
-    const bool use_secondary_window{Settings::values.layout_option.GetValue() ==
+                                    bool is_secondary) -> std::unique_ptr<EmuWindow_SDL2> {
-                                    Settings::LayoutOption::SeparateWindows};
+        switch (Settings::values.graphics_api.GetValue()) {
-    const auto secondary_window =
+        case Settings::GraphicsAPI::OpenGL:
-        use_secondary_window ? std::make_unique<EmuWindow_SDL2>(false, true) : nullptr;
+            return std::make_unique<EmuWindow_SDL2_GL>(fullscreen, is_secondary);
+        case Settings::GraphicsAPI::Software:
+            return std::make_unique<EmuWindow_SDL2_SW>(fullscreen, is_secondary);
+        }
+    };
 
-    Frontend::ScopeAcquireContext scope(*emu_window);
+    const auto emu_window{CreateEmuWindow(fullscreen, false)};
+    const bool use_secondary_window{
+        Settings::values.layout_option.GetValue() == Settings::LayoutOption::SeparateWindows &&
+        Settings::values.graphics_api.GetValue() != Settings::GraphicsAPI::Software};
+    const auto secondary_window = use_secondary_window ? CreateEmuWindow(false, true) : nullptr;
+
+    const auto scope = emu_window->Acquire();
 
     LOG_INFO(Frontend, "Citra Version: {} | {}-{}", Common::g_build_fullname, Common::g_scm_branch,
              Common::g_scm_desc);
@@ -400,9 +412,6 @@ int main(int argc, char** argv) {
     case Core::System::ResultStatus::ErrorSystemMode:
         LOG_CRITICAL(Frontend, "Failed to determine system mode!");
         return -1;
-    case Core::System::ResultStatus::ErrorVideoCore:
-        LOG_CRITICAL(Frontend, "VideoCore not initialized");
-        return -1;
     case Core::System::ResultStatus::Success:
         break; // Expected case
     default:

src/citra/config.cpp

@@ -5,6 +5,7 @@
 #include <iomanip>
 #include <memory>
 #include <sstream>
+#include <type_traits>
 #include <unordered_map>
 #include <SDL.h>
 #include <inih/cpp/INIReader.h>
@@ -71,6 +72,30 @@ static const std::array<std::array<int, 5>, Settings::NativeAnalog::NumAnalogs>
     },
 }};
 
+template <>
+void Config::ReadSetting(const std::string& group, Settings::Setting<std::string>& setting) {
+    std::string setting_value = sdl2_config->Get(group, setting.GetLabel(), setting.GetDefault());
+    if (setting_value.empty()) {
+        setting_value = setting.GetDefault();
+    }
+    setting = std::move(setting_value);
+}
+
+template <>
+void Config::ReadSetting(const std::string& group, Settings::Setting<bool>& setting) {
+    setting = sdl2_config->GetBoolean(group, setting.GetLabel(), setting.GetDefault());
+}
+
+template <typename Type, bool ranged>
+void Config::ReadSetting(const std::string& group, Settings::Setting<Type, ranged>& setting) {
+    if constexpr (std::is_floating_point_v<Type>) {
+        setting = sdl2_config->GetReal(group, setting.GetLabel(), setting.GetDefault());
+    } else {
+        setting = static_cast<Type>(sdl2_config->GetInteger(
+            group, setting.GetLabel(), static_cast<long>(setting.GetDefault())));
+    }
+}
+
 void Config::ReadValues() {
     // Controls
     // TODO: add multiple input profile support
@@ -104,104 +129,71 @@ void Config::ReadValues() {
                                                  InputCommon::CemuhookUDP::DEFAULT_PORT));
 
     // Core
-    Settings::values.use_cpu_jit = sdl2_config->GetBoolean("Core", "use_cpu_jit", true);
+    ReadSetting("Core", Settings::values.use_cpu_jit);
-    Settings::values.cpu_clock_percentage =
+    ReadSetting("Core", Settings::values.cpu_clock_percentage);
-        sdl2_config->GetInteger("Core", "cpu_clock_percentage", 100);
 
     // Renderer
-    Settings::values.use_gles = sdl2_config->GetBoolean("Renderer", "use_gles", false);
+    ReadSetting("Renderer", Settings::values.graphics_api);
-    Settings::values.use_hw_renderer = sdl2_config->GetBoolean("Renderer", "use_hw_renderer", true);
+    ReadSetting("Renderer", Settings::values.use_gles);
-    Settings::values.use_hw_shader = sdl2_config->GetBoolean("Renderer", "use_hw_shader", true);
+    ReadSetting("Renderer", Settings::values.use_hw_shader);
 #ifdef __APPLE__
     // Separable shader is broken on macos with Intel GPU thanks to poor drivers.
     // We still want to provide this option for test/development purposes, but disable it by
     // default.
-    Settings::values.separable_shader =
+    ReadSetting("Renderer", Settings::values.separable_shader);
-        sdl2_config->GetBoolean("Renderer", "separable_shader", false);
 #endif
-    Settings::values.shaders_accurate_mul =
+    ReadSetting("Renderer", Settings::values.shaders_accurate_mul);
-        sdl2_config->GetBoolean("Renderer", "shaders_accurate_mul", true);
+    ReadSetting("Renderer", Settings::values.use_shader_jit);
-    Settings::values.use_shader_jit = sdl2_config->GetBoolean("Renderer", "use_shader_jit", true);
+    ReadSetting("Renderer", Settings::values.resolution_factor);
-    Settings::values.resolution_factor =
+    ReadSetting("Renderer", Settings::values.use_disk_shader_cache);
-        static_cast<u16>(sdl2_config->GetInteger("Renderer", "resolution_factor", 1));
+    ReadSetting("Renderer", Settings::values.frame_limit);
-    Settings::values.use_disk_shader_cache =
+    ReadSetting("Renderer", Settings::values.use_vsync_new);
-        sdl2_config->GetBoolean("Renderer", "use_disk_shader_cache", true);
+    ReadSetting("Renderer", Settings::values.texture_filter_name);
-    Settings::values.frame_limit =
-        static_cast<u16>(sdl2_config->GetInteger("Renderer", "frame_limit", 100));
-    Settings::values.use_vsync_new =
-        static_cast<u16>(sdl2_config->GetInteger("Renderer", "use_vsync_new", 1));
-    Settings::values.texture_filter_name =
-        sdl2_config->GetString("Renderer", "texture_filter_name", "none");
 
-    Settings::values.mono_render_option = static_cast<Settings::MonoRenderOption>(
+    ReadSetting("Renderer", Settings::values.mono_render_option);
-        sdl2_config->GetInteger("Renderer", "mono_render_option", 0));
+    ReadSetting("Renderer", Settings::values.render_3d);
-    Settings::values.render_3d = static_cast<Settings::StereoRenderOption>(
+    ReadSetting("Renderer", Settings::values.factor_3d);
-        sdl2_config->GetInteger("Renderer", "render_3d", 0));
+    ReadSetting("Renderer", Settings::values.pp_shader_name);
-    Settings::values.factor_3d =
+    ReadSetting("Renderer", Settings::values.anaglyph_shader_name);
-        static_cast<u8>(sdl2_config->GetInteger("Renderer", "factor_3d", 0));
+    ReadSetting("Renderer", Settings::values.filter_mode);
-    Settings::values.pp_shader_name =
-        sdl2_config->GetString("Renderer", "pp_shader_name", "none (builtin)");
-    Settings::values.anaglyph_shader_name =
-        sdl2_config->GetString("Renderer", "anaglyph_shader_name", "dubois (builtin)");
-    Settings::values.filter_mode = sdl2_config->GetBoolean("Renderer", "filter_mode", true);
 
-    Settings::values.bg_red = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_red", 0.0));
+    ReadSetting("Renderer", Settings::values.bg_red);
-    Settings::values.bg_green =
+    ReadSetting("Renderer", Settings::values.bg_green);
-        static_cast<float>(sdl2_config->GetReal("Renderer", "bg_green", 0.0));
+    ReadSetting("Renderer", Settings::values.bg_blue);
-    Settings::values.bg_blue = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_blue", 0.0));
 
     // Layout
-    Settings::values.layout_option =
+    ReadSetting("Layout", Settings::values.layout_option);
-        static_cast<Settings::LayoutOption>(sdl2_config->GetInteger("Layout", "layout_option", 0));
+    ReadSetting("Layout", Settings::values.swap_screen);
-    Settings::values.swap_screen = sdl2_config->GetBoolean("Layout", "swap_screen", false);
+    ReadSetting("Layout", Settings::values.upright_screen);
-    Settings::values.upright_screen = sdl2_config->GetBoolean("Layout", "upright_screen", false);
+    ReadSetting("Layout", Settings::values.large_screen_proportion);
-    Settings::values.large_screen_proportion =
+    ReadSetting("Layout", Settings::values.custom_layout);
-        sdl2_config->GetReal("Layout", "large_screen_proportion", 4.0);
+    ReadSetting("Layout", Settings::values.custom_top_left);
-    Settings::values.custom_layout = sdl2_config->GetBoolean("Layout", "custom_layout", false);
+    ReadSetting("Layout", Settings::values.custom_top_top);
-    Settings::values.custom_top_left =
+    ReadSetting("Layout", Settings::values.custom_top_right);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_top_left", 0));
+    ReadSetting("Layout", Settings::values.custom_top_bottom);
-    Settings::values.custom_top_top =
+    ReadSetting("Layout", Settings::values.custom_bottom_left);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_top_top", 0));
+    ReadSetting("Layout", Settings::values.custom_bottom_top);
-    Settings::values.custom_top_right =
+    ReadSetting("Layout", Settings::values.custom_bottom_right);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_top_right", 400));
+    ReadSetting("Layout", Settings::values.custom_bottom_bottom);
-    Settings::values.custom_top_bottom =
+    ReadSetting("Layout", Settings::values.custom_second_layer_opacity);
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_top_bottom", 240));
-    Settings::values.custom_bottom_left =
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_bottom_left", 40));
-    Settings::values.custom_bottom_top =
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_bottom_top", 240));
-    Settings::values.custom_bottom_right =
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_bottom_right", 360));
-    Settings::values.custom_bottom_bottom =
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_bottom_bottom", 480));
-    Settings::values.custom_second_layer_opacity =
-        static_cast<u16>(sdl2_config->GetInteger("Layout", "custom_second_layer_opacity", 100));
 
     // Utility
-    Settings::values.dump_textures = sdl2_config->GetBoolean("Utility", "dump_textures", false);
+    ReadSetting("Utility", Settings::values.dump_textures);
-    Settings::values.custom_textures = sdl2_config->GetBoolean("Utility", "custom_textures", false);
+    ReadSetting("Utility", Settings::values.custom_textures);
-    Settings::values.preload_textures =
+    ReadSetting("Utility", Settings::values.preload_textures);
-        sdl2_config->GetBoolean("Utility", "preload_textures", false);
 
     // Audio
-    Settings::values.audio_emulation = static_cast<Settings::AudioEmulation>(
+    ReadSetting("Audio", Settings::values.audio_emulation);
-        sdl2_config->GetInteger("Audio", "audio_emulation", 0));
+    ReadSetting("Audio", Settings::values.sink_id);
-    Settings::values.sink_id = sdl2_config->GetString("Audio", "output_engine", "auto");
+    ReadSetting("Audio", Settings::values.enable_audio_stretching);
-    Settings::values.enable_audio_stretching =
+    ReadSetting("Audio", Settings::values.audio_device_id);
-        sdl2_config->GetBoolean("Audio", "enable_audio_stretching", true);
+    ReadSetting("Audio", Settings::values.volume);
-    Settings::values.audio_device_id = sdl2_config->GetString("Audio", "output_device", "auto");
+    ReadSetting("Audio", Settings::values.mic_input_device);
-    Settings::values.volume = static_cast<float>(sdl2_config->GetReal("Audio", "volume", 1));
+    ReadSetting("Audio", Settings::values.mic_input_type);
-    Settings::values.mic_input_device =
-        sdl2_config->GetString("Audio", "mic_input_device", Frontend::Mic::default_device_name);
-    Settings::values.mic_input_type =
-        static_cast<Settings::MicInputType>(sdl2_config->GetInteger("Audio", "mic_input_type", 0));
 
     // Data Storage
-    Settings::values.use_virtual_sd =
+    ReadSetting("Data Storage", Settings::values.use_virtual_sd);
-        sdl2_config->GetBoolean("Data Storage", "use_virtual_sd", true);
+    ReadSetting("Data Storage", Settings::values.use_custom_storage);
-
-    Settings::values.use_custom_storage =
-        sdl2_config->GetBoolean("Data Storage", "use_custom_storage", false);
 
     if (Settings::values.use_custom_storage) {
         FileUtil::UpdateUserPath(FileUtil::UserPath::NANDDir,
@@ -211,11 +203,9 @@ void Config::ReadValues() {
     }
 
     // System
-    Settings::values.is_new_3ds = sdl2_config->GetBoolean("System", "is_new_3ds", true);
+    ReadSetting("System", Settings::values.is_new_3ds);
-    Settings::values.region_value =
+    ReadSetting("System", Settings::values.region_value);
-        sdl2_config->GetInteger("System", "region_value", Settings::REGION_VALUE_AUTO_SELECT);
+    ReadSetting("System", Settings::values.init_clock);
-    Settings::values.init_clock =
-        static_cast<Settings::InitClock>(sdl2_config->GetInteger("System", "init_clock", 1));
     {
         std::tm t;
         t.tm_sec = 1;
@@ -236,6 +226,8 @@ void Config::ReadValues() {
                 std::chrono::system_clock::from_time_t(std::mktime(&t)).time_since_epoch())
                 .count();
     }
+    ReadSetting("System", Settings::values.plugin_loader_enabled);
+    ReadSetting("System", Settings::values.allow_plugin_loader);
 
     {
         constexpr const char* default_init_time_offset = "0 00:00:00";
@@ -311,14 +303,14 @@ void Config::ReadValues() {
         sdl2_config->GetInteger("Camera", "camera_outer_left_flip", 0);
 
     // Miscellaneous
-    Settings::values.log_filter = sdl2_config->GetString("Miscellaneous", "log_filter", "*:Info");
+    ReadSetting("Miscellaneous", Settings::values.log_filter);
 
     // Debugging
     Settings::values.record_frame_times =
         sdl2_config->GetBoolean("Debugging", "record_frame_times", false);
-    Settings::values.use_gdbstub = sdl2_config->GetBoolean("Debugging", "use_gdbstub", false);
+    ReadSetting("Debugging", Settings::values.renderer_debug);
-    Settings::values.gdbstub_port =
+    ReadSetting("Debugging", Settings::values.use_gdbstub);
-        static_cast<u16>(sdl2_config->GetInteger("Debugging", "gdbstub_port", 24689));
+    ReadSetting("Debugging", Settings::values.gdbstub_port);
 
     for (const auto& service_module : Service::service_module_map) {
         bool use_lle = sdl2_config->GetBoolean("Debugging", "LLE\\" + service_module.name, false);

src/citra/config.h

@@ -6,6 +6,7 @@
 
 #include <memory>
 #include <string>
+#include "common/settings.h"
 
 class INIReader;
 
@@ -21,4 +22,14 @@ public:
     ~Config();
 
     void Reload();
+
+private:
+    /**
+     * Applies a value read from the sdl2_config to a Setting.
+     *
+     * @param group The name of the INI group
+     * @param setting The yuzu setting to modify
+     */
+    template <typename Type, bool ranged>
+    void ReadSetting(const std::string& group, Settings::Setting<Type, ranged>& setting);
 };

src/citra/default_ini.h

@@ -98,14 +98,14 @@ use_cpu_jit =
 cpu_clock_percentage =
 
 [Renderer]
+# Whether to render using OpenGL or Software
+# 0: Software, 1: OpenGL (default)
+graphics_api =
+
 # Whether to render using GLES or OpenGL
 # 0 (default): OpenGL, 1: GLES
 use_gles =
 
-# Whether to use software or hardware rendering.
-# 0: Software, 1 (default): Hardware
-use_hw_renderer =
-
 # Whether to use hardware shaders to emulate 3DS shaders
 # 0: Software, 1 (default): Hardware
 use_hw_shader =
@@ -328,9 +328,15 @@ log_filter = *:Info
 [Debugging]
 # Record frame time data, can be found in the log directory. Boolean value
 record_frame_times =
+
 # Port for listening to GDB connections.
 use_gdbstub=false
 gdbstub_port=24689
+
+# Whether to enable additional debugging information during emulation
+# 0 (default): Off, 1: On
+renderer_debug =
+
 # To LLE a service module add "LLE\<module name>=true"
 
 [WebService]

src/citra/emu_window/emu_window_sdl2.cpp

@@ -7,40 +7,14 @@
 #include <string>
 #define SDL_MAIN_HANDLED
 #include <SDL.h>
-#include <fmt/format.h>
-#include <glad/glad.h>
 #include "citra/emu_window/emu_window_sdl2.h"
 #include "common/logging/log.h"
 #include "common/scm_rev.h"
-#include "common/settings.h"
-#include "core/3ds.h"
 #include "core/core.h"
 #include "input_common/keyboard.h"
 #include "input_common/main.h"
 #include "input_common/motion_emu.h"
-#include "input_common/sdl/sdl.h"
 #include "network/network.h"
-#include "video_core/renderer_base.h"
-#include "video_core/video_core.h"
-
-SharedContext_SDL2::SharedContext_SDL2() {
-    window = SDL_CreateWindow(NULL, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED, 0, 0,
-                              SDL_WINDOW_HIDDEN | SDL_WINDOW_OPENGL);
-    context = SDL_GL_CreateContext(window);
-}
-
-SharedContext_SDL2::~SharedContext_SDL2() {
-    SDL_GL_DeleteContext(context);
-    SDL_DestroyWindow(window);
-}
-
-void SharedContext_SDL2::MakeCurrent() {
-    SDL_GL_MakeCurrent(window, context);
-}
-
-void SharedContext_SDL2::DoneCurrent() {
-    SDL_GL_MakeCurrent(window, nullptr);
-}
 
 void EmuWindow_SDL2::OnMouseMotion(s32 x, s32 y) {
     TouchMoved((unsigned)std::max(x, 0), (unsigned)std::max(y, 0));
@@ -135,78 +109,9 @@ void EmuWindow_SDL2::Fullscreen() {
     SDL_MaximizeWindow(render_window);
 }
 
-EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen, bool is_secondary) : EmuWindow(is_secondary) {
+EmuWindow_SDL2::EmuWindow_SDL2(bool is_secondary) : EmuWindow(is_secondary) {}
-    // Initialize the window
-    if (Settings::values.use_gles) {
-        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 3);
-        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 2);
-        SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_ES);
-    } else {
-        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 4);
-        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 3);
-        SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_CORE);
-    }
-
-    SDL_GL_SetAttribute(SDL_GL_DOUBLEBUFFER, 1);
-    SDL_GL_SetAttribute(SDL_GL_RED_SIZE, 8);
-    SDL_GL_SetAttribute(SDL_GL_GREEN_SIZE, 8);
-    SDL_GL_SetAttribute(SDL_GL_BLUE_SIZE, 8);
-    SDL_GL_SetAttribute(SDL_GL_ALPHA_SIZE, 0);
-    // Enable context sharing for the shared context
-    SDL_GL_SetAttribute(SDL_GL_SHARE_WITH_CURRENT_CONTEXT, 1);
-    // Enable vsync
-    SDL_GL_SetSwapInterval(1);
-
-    std::string window_title = fmt::format("Citra {} | {}-{}", Common::g_build_fullname,
-                                           Common::g_scm_branch, Common::g_scm_desc);
-    render_window =
-        SDL_CreateWindow(window_title.c_str(),
-                         SDL_WINDOWPOS_UNDEFINED, // x position
-                         SDL_WINDOWPOS_UNDEFINED, // y position
-                         Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight,
-                         SDL_WINDOW_OPENGL | SDL_WINDOW_RESIZABLE | SDL_WINDOW_ALLOW_HIGHDPI);
-
-    if (render_window == nullptr) {
-        LOG_CRITICAL(Frontend, "Failed to create SDL2 window: {}", SDL_GetError());
-        exit(1);
-    }
-
-    dummy_window = SDL_CreateWindow(NULL, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED, 0, 0,
-                                    SDL_WINDOW_HIDDEN | SDL_WINDOW_OPENGL);
-
-    if (fullscreen) {
-        Fullscreen();
-    }
-
-    window_context = SDL_GL_CreateContext(render_window);
-    core_context = CreateSharedContext();
-    last_saved_context = nullptr;
-
-    if (window_context == nullptr) {
-        LOG_CRITICAL(Frontend, "Failed to create SDL2 GL context: {}", SDL_GetError());
-        exit(1);
-    }
-    if (core_context == nullptr) {
-        LOG_CRITICAL(Frontend, "Failed to create shared SDL2 GL context: {}", SDL_GetError());
-        exit(1);
-    }
-
-    render_window_id = SDL_GetWindowID(render_window);
-    auto gl_load_func = Settings::values.use_gles ? gladLoadGLES2Loader : gladLoadGLLoader;
-
-    if (!gl_load_func(static_cast<GLADloadproc>(SDL_GL_GetProcAddress))) {
-        LOG_CRITICAL(Frontend, "Failed to initialize GL functions: {}", SDL_GetError());
-        exit(1);
-    }
-
-    OnResize();
-    OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
-    SDL_PumpEvents();
-}
 
 EmuWindow_SDL2::~EmuWindow_SDL2() {
-    core_context.reset();
-    SDL_GL_DeleteContext(window_context);
     SDL_Quit();
 }
 
@@ -222,28 +127,6 @@ void EmuWindow_SDL2::InitializeSDL2() {
     SDL_SetMainReady();
 }
 
-std::unique_ptr<Frontend::GraphicsContext> EmuWindow_SDL2::CreateSharedContext() const {
-    return std::make_unique<SharedContext_SDL2>();
-}
-
-void EmuWindow_SDL2::SaveContext() {
-    last_saved_context = SDL_GL_GetCurrentContext();
-}
-
-void EmuWindow_SDL2::RestoreContext() {
-    SDL_GL_MakeCurrent(render_window, last_saved_context);
-}
-
-void EmuWindow_SDL2::Present() {
-    SDL_GL_MakeCurrent(render_window, window_context);
-    SDL_GL_SetSwapInterval(1);
-    while (IsOpen()) {
-        VideoCore::g_renderer->TryPresent(100, is_secondary);
-        SDL_GL_SwapWindow(render_window);
-    }
-    SDL_GL_MakeCurrent(render_window, nullptr);
-}
-
 void EmuWindow_SDL2::PollEvents() {
     SDL_Event event;
     std::vector<SDL_Event> other_window_events;
@@ -312,14 +195,6 @@ void EmuWindow_SDL2::PollEvents() {
     }
 }
 
-void EmuWindow_SDL2::MakeCurrent() {
-    core_context->MakeCurrent();
-}
-
-void EmuWindow_SDL2::DoneCurrent() {
-    core_context->DoneCurrent();
-}
-
 void EmuWindow_SDL2::OnMinimalClientAreaChangeRequest(std::pair<u32, u32> minimal_size) {
     SDL_SetWindowMinimumSize(render_window, minimal_size.first, minimal_size.second);
 }

src/citra/emu_window/emu_window_sdl2.h

@@ -10,56 +10,27 @@
 
 struct SDL_Window;
 
-class SharedContext_SDL2 : public Frontend::GraphicsContext {
-public:
-    using SDL_GLContext = void*;
-
-    SharedContext_SDL2();
-
-    ~SharedContext_SDL2() override;
-
-    void MakeCurrent() override;
-
-    void DoneCurrent() override;
-
-private:
-    SDL_GLContext context;
-    SDL_Window* window;
-};
-
 class EmuWindow_SDL2 : public Frontend::EmuWindow {
 public:
-    explicit EmuWindow_SDL2(bool fullscreen, bool is_secondary);
+    explicit EmuWindow_SDL2(bool is_secondary);
     ~EmuWindow_SDL2();
 
+    /// Initializes SDL2
     static void InitializeSDL2();
 
-    void Present();
+    /// Presents the most recent frame from the video backend
+    virtual void Present() {}
 
     /// Polls window events
     void PollEvents() override;
 
-    /// Makes the graphics context current for the caller thread
-    void MakeCurrent() override;
-
-    /// Releases the GL context from the caller thread
-    void DoneCurrent() override;
-
     /// Whether the window is still open, and a close request hasn't yet been sent
     bool IsOpen() const;
 
     /// Close the window.
     void RequestClose();
 
-    /// Creates a new context that is shared with the current context
+protected:
-    std::unique_ptr<GraphicsContext> CreateSharedContext() const override;
-
-    /// Saves the current context, for the purpose of e.g. creating new shared contexts
-    void SaveContext() override;
-    /// Restores the context previously saved
-    void RestoreContext() override;
-
-private:
     /// Called by PollEvents when a key is pressed or released.
     void OnKeyEvent(int key, u8 state);
 
@@ -105,17 +76,6 @@ private:
     /// Fake hidden window for the core context
     SDL_Window* dummy_window;
 
-    using SDL_GLContext = void*;
-
-    /// The OpenGL context associated with the window
-    SDL_GLContext window_context;
-
-    /// Used by SaveContext and RestoreContext
-    SDL_GLContext last_saved_context;
-
-    /// The OpenGL context associated with the core
-    std::unique_ptr<Frontend::GraphicsContext> core_context;
-
     /// Keeps track of how often to update the title bar during gameplay
     u32 last_time = 0;
 };

src/citra/emu_window/emu_window_sdl2_gl.cpp

@@ -0,0 +1,152 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstdlib>
+#include <string>
+#define SDL_MAIN_HANDLED
+#include <SDL.h>
+#include <glad/glad.h>
+#include "citra/emu_window/emu_window_sdl2_gl.h"
+#include "common/scm_rev.h"
+#include "video_core/renderer_base.h"
+#include "video_core/video_core.h"
+
+class SDLGLContext : public Frontend::GraphicsContext {
+public:
+    using SDL_GLContext = void*;
+
+    SDLGLContext() {
+        window = SDL_CreateWindow(NULL, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED, 0, 0,
+                                  SDL_WINDOW_HIDDEN | SDL_WINDOW_OPENGL);
+        context = SDL_GL_CreateContext(window);
+    }
+
+    ~SDLGLContext() override {
+        SDL_GL_DeleteContext(context);
+        SDL_DestroyWindow(window);
+    }
+
+    void MakeCurrent() override {
+        SDL_GL_MakeCurrent(window, context);
+    }
+
+    void DoneCurrent() override {
+        SDL_GL_MakeCurrent(window, nullptr);
+    }
+
+private:
+    SDL_Window* window;
+    SDL_GLContext context;
+};
+
+EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen, bool is_secondary)
+    : EmuWindow_SDL2{is_secondary} {
+    // Initialize the window
+    if (Settings::values.use_gles) {
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 3);
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 2);
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_ES);
+    } else {
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 4);
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 3);
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_CORE);
+    }
+
+    SDL_GL_SetAttribute(SDL_GL_DOUBLEBUFFER, 1);
+    SDL_GL_SetAttribute(SDL_GL_RED_SIZE, 8);
+    SDL_GL_SetAttribute(SDL_GL_GREEN_SIZE, 8);
+    SDL_GL_SetAttribute(SDL_GL_BLUE_SIZE, 8);
+    SDL_GL_SetAttribute(SDL_GL_ALPHA_SIZE, 0);
+    // Enable context sharing for the shared context
+    SDL_GL_SetAttribute(SDL_GL_SHARE_WITH_CURRENT_CONTEXT, 1);
+    // Enable vsync
+    SDL_GL_SetSwapInterval(1);
+    // Enable debug context
+    if (Settings::values.renderer_debug) {
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_FLAGS, SDL_GL_CONTEXT_DEBUG_FLAG);
+    }
+
+    std::string window_title = fmt::format("Citra {} | {}-{}", Common::g_build_fullname,
+                                           Common::g_scm_branch, Common::g_scm_desc);
+    render_window =
+        SDL_CreateWindow(window_title.c_str(),
+                         SDL_WINDOWPOS_UNDEFINED, // x position
+                         SDL_WINDOWPOS_UNDEFINED, // y position
+                         Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight,
+                         SDL_WINDOW_OPENGL | SDL_WINDOW_RESIZABLE | SDL_WINDOW_ALLOW_HIGHDPI);
+
+    if (render_window == nullptr) {
+        LOG_CRITICAL(Frontend, "Failed to create SDL2 window: {}", SDL_GetError());
+        exit(1);
+    }
+
+    dummy_window = SDL_CreateWindow(NULL, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED, 0, 0,
+                                    SDL_WINDOW_HIDDEN | SDL_WINDOW_OPENGL);
+
+    if (fullscreen) {
+        Fullscreen();
+    }
+
+    window_context = SDL_GL_CreateContext(render_window);
+    core_context = CreateSharedContext();
+    last_saved_context = nullptr;
+
+    if (window_context == nullptr) {
+        LOG_CRITICAL(Frontend, "Failed to create SDL2 GL context: {}", SDL_GetError());
+        exit(1);
+    }
+    if (core_context == nullptr) {
+        LOG_CRITICAL(Frontend, "Failed to create shared SDL2 GL context: {}", SDL_GetError());
+        exit(1);
+    }
+
+    render_window_id = SDL_GetWindowID(render_window);
+    auto gl_load_func = Settings::values.use_gles ? gladLoadGLES2Loader : gladLoadGLLoader;
+
+    if (!gl_load_func(static_cast<GLADloadproc>(SDL_GL_GetProcAddress))) {
+        LOG_CRITICAL(Frontend, "Failed to initialize GL functions: {}", SDL_GetError());
+        exit(1);
+    }
+
+    OnResize();
+    OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
+    SDL_PumpEvents();
+}
+
+EmuWindow_SDL2_GL::~EmuWindow_SDL2_GL() {
+    core_context.reset();
+    SDL_DestroyWindow(render_window);
+    SDL_GL_DeleteContext(window_context);
+}
+
+std::unique_ptr<Frontend::GraphicsContext> EmuWindow_SDL2_GL::CreateSharedContext() const {
+    return std::make_unique<SDLGLContext>();
+}
+
+void EmuWindow_SDL2_GL::MakeCurrent() {
+    core_context->MakeCurrent();
+}
+
+void EmuWindow_SDL2_GL::DoneCurrent() {
+    core_context->DoneCurrent();
+}
+
+void EmuWindow_SDL2_GL::SaveContext() {
+    last_saved_context = SDL_GL_GetCurrentContext();
+}
+
+void EmuWindow_SDL2_GL::RestoreContext() {
+    SDL_GL_MakeCurrent(render_window, last_saved_context);
+}
+
+void EmuWindow_SDL2_GL::Present() {
+    SDL_GL_MakeCurrent(render_window, window_context);
+    SDL_GL_SetSwapInterval(1);
+    while (IsOpen()) {
+        VideoCore::g_renderer->TryPresent(100, is_secondary);
+        SDL_GL_SwapWindow(render_window);
+    }
+    SDL_GL_MakeCurrent(render_window, nullptr);
+}

src/citra/emu_window/emu_window_sdl2_gl.h

@@ -0,0 +1,35 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include "citra/emu_window/emu_window_sdl2.h"
+
+struct SDL_Window;
+
+class EmuWindow_SDL2_GL : public EmuWindow_SDL2 {
+public:
+    explicit EmuWindow_SDL2_GL(bool fullscreen, bool is_secondary);
+    ~EmuWindow_SDL2_GL();
+
+    void Present() override;
+    std::unique_ptr<GraphicsContext> CreateSharedContext() const override;
+    void MakeCurrent() override;
+    void DoneCurrent() override;
+    void SaveContext() override;
+    void RestoreContext() override;
+
+private:
+    using SDL_GLContext = void*;
+
+    /// The OpenGL context associated with the window
+    SDL_GLContext window_context;
+
+    /// Used by SaveContext and RestoreContext
+    SDL_GLContext last_saved_context;
+
+    /// The OpenGL context associated with the core
+    std::unique_ptr<Frontend::GraphicsContext> core_context;
+};

src/citra/emu_window/emu_window_sdl2_sw.cpp

@@ -0,0 +1,127 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstdlib>
+#include <string>
+#define SDL_MAIN_HANDLED
+#include <SDL.h>
+#include <SDL_rect.h>
+#include "citra/emu_window/emu_window_sdl2_sw.h"
+#include "common/color.h"
+#include "common/scm_rev.h"
+#include "core/frontend/emu_window.h"
+#include "core/hw/gpu.h"
+#include "core/memory.h"
+#include "video_core/video_core.h"
+
+class DummyContext : public Frontend::GraphicsContext {};
+
+EmuWindow_SDL2_SW::EmuWindow_SDL2_SW(bool fullscreen, bool is_secondary)
+    : EmuWindow_SDL2{is_secondary} {
+    std::string window_title = fmt::format("Citra {} | {}-{}", Common::g_build_fullname,
+                                           Common::g_scm_branch, Common::g_scm_desc);
+    render_window =
+        SDL_CreateWindow(window_title.c_str(),
+                         SDL_WINDOWPOS_UNDEFINED, // x position
+                         SDL_WINDOWPOS_UNDEFINED, // y position
+                         Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight,
+                         SDL_WINDOW_SHOWN);
+
+    if (render_window == nullptr) {
+        LOG_CRITICAL(Frontend, "Failed to create SDL2 window: {}", SDL_GetError());
+        exit(1);
+    }
+
+    window_surface = SDL_GetWindowSurface(render_window);
+    renderer = SDL_CreateSoftwareRenderer(window_surface);
+
+    if (fullscreen) {
+        Fullscreen();
+    }
+
+    render_window_id = SDL_GetWindowID(render_window);
+
+    OnResize();
+    OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
+    SDL_PumpEvents();
+}
+
+EmuWindow_SDL2_SW::~EmuWindow_SDL2_SW() {
+    SDL_DestroyRenderer(renderer);
+    SDL_DestroyWindow(render_window);
+}
+
+std::unique_ptr<Frontend::GraphicsContext> EmuWindow_SDL2_SW::CreateSharedContext() const {
+    return std::make_unique<DummyContext>();
+}
+
+void EmuWindow_SDL2_SW::Present() {
+    const auto layout{Layout::DefaultFrameLayout(
+        Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight, false, false)};
+
+    while (IsOpen()) {
+        SDL_SetRenderDrawColor(renderer, Settings::values.bg_red.GetValue() * 255,
+                               Settings::values.bg_green.GetValue() * 255,
+                               Settings::values.bg_blue.GetValue() * 255, 0xFF);
+        SDL_RenderClear(renderer);
+
+        const auto draw_screen = [&](int fb_id) {
+            const auto dst_rect = fb_id == 0 ? layout.top_screen : layout.bottom_screen;
+            SDL_Rect sdl_rect{static_cast<int>(dst_rect.left), static_cast<int>(dst_rect.top),
+                              static_cast<int>(dst_rect.GetWidth()),
+                              static_cast<int>(dst_rect.GetHeight())};
+            SDL_Surface* screen = LoadFramebuffer(fb_id);
+            SDL_BlitSurface(screen, nullptr, window_surface, &sdl_rect);
+            SDL_FreeSurface(screen);
+        };
+
+        draw_screen(0);
+        draw_screen(1);
+
+        SDL_RenderPresent(renderer);
+        SDL_UpdateWindowSurface(render_window);
+    }
+}
+
+SDL_Surface* EmuWindow_SDL2_SW::LoadFramebuffer(int fb_id) {
+    const auto& framebuffer = GPU::g_regs.framebuffer_config[fb_id];
+    const PAddr framebuffer_addr =
+        framebuffer.active_fb == 0 ? framebuffer.address_left1 : framebuffer.address_left2;
+
+    Memory::RasterizerFlushRegion(framebuffer_addr, framebuffer.stride * framebuffer.height);
+    const u8* framebuffer_data = VideoCore::g_memory->GetPhysicalPointer(framebuffer_addr);
+
+    const int width = framebuffer.height;
+    const int height = framebuffer.width;
+    const int bpp = GPU::Regs::BytesPerPixel(framebuffer.color_format);
+
+    SDL_Surface* surface =
+        SDL_CreateRGBSurfaceWithFormat(0, width, height, 0, SDL_PIXELFORMAT_ABGR8888);
+    SDL_LockSurface(surface);
+    for (int y = 0; y < height; y++) {
+        for (int x = 0; x < width; x++) {
+            const u8* pixel = framebuffer_data + (x * height + height - y) * bpp;
+            const Common::Vec4 color = [&] {
+                switch (framebuffer.color_format) {
+                case GPU::Regs::PixelFormat::RGBA8:
+                    return Common::Color::DecodeRGBA8(pixel);
+                case GPU::Regs::PixelFormat::RGB8:
+                    return Common::Color::DecodeRGB8(pixel);
+                case GPU::Regs::PixelFormat::RGB565:
+                    return Common::Color::DecodeRGB565(pixel);
+                case GPU::Regs::PixelFormat::RGB5A1:
+                    return Common::Color::DecodeRGB5A1(pixel);
+                case GPU::Regs::PixelFormat::RGBA4:
+                    return Common::Color::DecodeRGBA4(pixel);
+                }
+            }();
+
+            u8* dst_pixel = reinterpret_cast<u8*>(surface->pixels) + (y * width + x) * 4;
+            std::memcpy(dst_pixel, color.AsArray(), sizeof(color));
+        }
+    }
+    SDL_UnlockSurface(surface);
+    return surface;
+}

src/citra/emu_window/emu_window_sdl2_sw.h

@@ -0,0 +1,32 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include "citra/emu_window/emu_window_sdl2.h"
+
+struct SDL_Renderer;
+struct SDL_Surface;
+
+class EmuWindow_SDL2_SW : public EmuWindow_SDL2 {
+public:
+    explicit EmuWindow_SDL2_SW(bool fullscreen, bool is_secondary);
+    ~EmuWindow_SDL2_SW();
+
+    void Present() override;
+    std::unique_ptr<GraphicsContext> CreateSharedContext() const override;
+    void MakeCurrent() override {}
+    void DoneCurrent() override {}
+
+private:
+    /// Loads a framebuffer to an SDL surface
+    SDL_Surface* LoadFramebuffer(int fb_id);
+
+    /// The SDL software renderer
+    SDL_Renderer* renderer;
+
+    /// The window surface
+    SDL_Surface* window_surface;
+};

src/citra_qt/CMakeLists.txt

@@ -269,6 +269,10 @@ target_link_libraries(citra-qt PRIVATE audio_core common core input_common netwo
 target_link_libraries(citra-qt PRIVATE Boost::boost glad nihstro-headers Qt5::Widgets Qt5::Multimedia Qt5::Concurrent)
 target_link_libraries(citra-qt PRIVATE ${PLATFORM_LIBRARIES} Threads::Threads)
 
+if (NOT WIN32)
+    target_include_directories(citra-qt PRIVATE ${Qt5Gui_PRIVATE_INCLUDE_DIRS})
+endif()
+
 if (UNIX AND NOT APPLE)
     target_link_libraries(citra-qt PRIVATE Qt5::DBus)
 endif()
@@ -325,6 +329,10 @@ if (MSVC)
     endif()
 endif()
 
+if (NOT APPLE)
+    target_compile_definitions(citra-qt PRIVATE HAS_OPENGL)
+endif()
+
 if (CITRA_USE_PRECOMPILED_HEADERS)
     target_precompile_headers(citra-qt PRIVATE precompiled_headers.h)
 endif()

src/citra_qt/bootmanager.cpp

@@ -2,31 +2,45 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <glad/glad.h>
+
 #include <QApplication>
 #include <QDragEnterEvent>
 #include <QHBoxLayout>
 #include <QKeyEvent>
-#include <QOffscreenSurface>
+#include <QMessageBox>
-#include <QOpenGLContext>
+#include <QPainter>
-#include <QOpenGLFunctions>
-#include <QOpenGLFunctions_4_3_Core>
 #include <fmt/format.h>
 #include "citra_qt/bootmanager.h"
 #include "citra_qt/main.h"
+#include "common/color.h"
 #include "common/microprofile.h"
 #include "common/scm_rev.h"
 #include "common/settings.h"
 #include "core/3ds.h"
 #include "core/core.h"
-#include "core/frontend/scope_acquire_context.h"
+#include "core/frontend/framebuffer_layout.h"
 #include "core/perf_stats.h"
 #include "input_common/keyboard.h"
 #include "input_common/main.h"
 #include "input_common/motion_emu.h"
-#include "network/network.h"
 #include "video_core/renderer_base.h"
 #include "video_core/video_core.h"
 
+#ifdef HAS_OPENGL
+#include <QOffscreenSurface>
+#include <QOpenGLContext>
+#endif
+
+#if defined(__APPLE__)
+#include <objc/message.h>
+#include <objc/objc.h>
+#endif
+
+#if !defined(WIN32)
+#include <qpa/qplatformnativeinterface.h>
+#endif
+
 EmuThread::EmuThread(Frontend::GraphicsContext& core_context) : core_context(core_context) {}
 
 EmuThread::~EmuThread() = default;
@@ -44,7 +58,7 @@ static GMainWindow* GetMainWindow() {
 
 void EmuThread::run() {
     MicroProfileOnThreadCreate("EmuThread");
-    Frontend::ScopeAcquireContext scope(core_context);
+    const auto scope = core_context.Acquire();
 
     emit LoadProgress(VideoCore::LoadCallbackStage::Prepare, 0, 0);
 
@@ -55,6 +69,7 @@ void EmuThread::run() {
         });
 
     emit LoadProgress(VideoCore::LoadCallbackStage::Complete, 0, 0);
+    emit HideLoadingScreen();
 
     core_context.MakeCurrent();
 
@@ -113,90 +128,263 @@ void EmuThread::run() {
 #endif
 }
 
-OpenGLWindow::OpenGLWindow(QWindow* parent, QWidget* event_handler, QOpenGLContext* shared_context,
+#ifdef HAS_OPENGL
-                           bool is_secondary)
+class OpenGLSharedContext : public Frontend::GraphicsContext {
-    : QWindow(parent), context(std::make_unique<QOpenGLContext>(shared_context->parent())),
+public:
-      event_handler(event_handler), is_secondary{is_secondary} {
+    /// Create the original context that should be shared from
+    explicit OpenGLSharedContext() {
+        QSurfaceFormat format;
+
+        format.setVersion(4, 3);
+        format.setProfile(QSurfaceFormat::CoreProfile);
+
+        if (Settings::values.renderer_debug) {
+            format.setOption(QSurfaceFormat::FormatOption::DebugContext);
+        }
+
+        // TODO: expose a setting for buffer value (ie default/single/double/triple)
+        format.setSwapBehavior(QSurfaceFormat::DefaultSwapBehavior);
+        format.setSwapInterval(0);
+
+        context = std::make_unique<QOpenGLContext>();
+        context->setFormat(format);
+        if (!context->create()) {
+            LOG_ERROR(Frontend, "Unable to create main openGL context");
+        }
+
+        offscreen_surface = std::make_unique<QOffscreenSurface>(nullptr);
+        offscreen_surface->setFormat(format);
+        offscreen_surface->create();
+        surface = offscreen_surface.get();
+    }
+
+    /// Create the shared contexts for rendering and presentation
+    explicit OpenGLSharedContext(QOpenGLContext* share_context, QSurface* main_surface) {
 
         // disable vsync for any shared contexts
-    auto format = shared_context->format();
+        auto format = share_context->format();
-    format.setSwapInterval(Settings::values.use_vsync_new ? 1 : 0);
+        format.setSwapInterval(main_surface ? Settings::values.use_vsync_new.GetValue() : 0);
-    this->setFormat(format);
 
-    context->setShareContext(shared_context);
+        context = std::make_unique<QOpenGLContext>();
-    context->setScreen(this->screen());
+        context->setShareContext(share_context);
         context->setFormat(format);
-    context->create();
+        if (!context->create()) {
+            LOG_ERROR(Frontend, "Unable to create shared openGL context");
+        }
 
-    setSurfaceType(QWindow::OpenGLSurface);
+        surface = main_surface;
+    }
 
-    // TODO: One of these flags might be interesting: WA_OpaquePaintEvent, WA_NoBackground,
+    ~OpenGLSharedContext() {
-    // WA_DontShowOnScreen, WA_DeleteOnClose
-}
-
-OpenGLWindow::~OpenGLWindow() {
         context->doneCurrent();
-}
-
-void OpenGLWindow::Present() {
-    if (!isExposed())
-        return;
-
-    context->makeCurrent(this);
-    if (VideoCore::g_renderer) {
-        VideoCore::g_renderer->TryPresent(100, is_secondary);
     }
-    context->swapBuffers(this);
-    auto f = context->versionFunctions<QOpenGLFunctions_4_3_Core>();
-    f->glFinish();
-    QWindow::requestUpdate();
-}
 
-bool OpenGLWindow::event(QEvent* event) {
+    void SwapBuffers() override {
-    switch (event->type()) {
+        context->swapBuffers(surface);
-    case QEvent::UpdateRequest:
+    }
+
+    void MakeCurrent() override {
+        // We can't track the current state of the underlying context in this wrapper class because
+        // Qt may make the underlying context not current for one reason or another. In particular,
+        // the WebBrowser uses GL, so it seems to conflict if we aren't careful.
+        // Instead of always just making the context current (which does not have any caching to
+        // check if the underlying context is already current) we can check for the current context
+        // in the thread local data by calling `currentContext()` and checking if its ours.
+        if (QOpenGLContext::currentContext() != context.get()) {
+            context->makeCurrent(surface);
+        }
+    }
+
+    void DoneCurrent() override {
+        context->doneCurrent();
+    }
+
+    QOpenGLContext* GetShareContext() const {
+        return context.get();
+    }
+
+private:
+    // Avoid using Qt parent system here since we might move the QObjects to new threads
+    // As a note, this means we should avoid using slots/signals with the objects too
+    std::unique_ptr<QOpenGLContext> context;
+    std::unique_ptr<QOffscreenSurface> offscreen_surface{};
+    QSurface* surface;
+};
+#endif
+
+class DummyContext : public Frontend::GraphicsContext {};
+
+class RenderWidget : public QWidget {
+public:
+    RenderWidget(GRenderWindow* parent) : QWidget(parent) {
+        setMouseTracking(true);
+    }
+
+    virtual ~RenderWidget() = default;
+
+    virtual void Present() {}
+
+    void paintEvent(QPaintEvent* event) override {
         Present();
-        return true;
+        update();
-    case QEvent::MouseButtonPress:
-    case QEvent::MouseButtonRelease:
-    case QEvent::MouseButtonDblClick:
-    case QEvent::MouseMove:
-    case QEvent::KeyPress:
-    case QEvent::KeyRelease:
-    case QEvent::FocusIn:
-    case QEvent::FocusOut:
-    case QEvent::FocusAboutToChange:
-    case QEvent::Enter:
-    case QEvent::Leave:
-    case QEvent::Wheel:
-    case QEvent::TabletMove:
-    case QEvent::TabletPress:
-    case QEvent::TabletRelease:
-    case QEvent::TabletEnterProximity:
-    case QEvent::TabletLeaveProximity:
-    case QEvent::TouchBegin:
-    case QEvent::TouchUpdate:
-    case QEvent::TouchEnd:
-    case QEvent::InputMethodQuery:
-    case QEvent::TouchCancel:
-        return QCoreApplication::sendEvent(event_handler, event);
-    case QEvent::Drop:
-        GetMainWindow()->DropAction(static_cast<QDropEvent*>(event));
-        return true;
-    case QEvent::DragEnter:
-    case QEvent::DragMove:
-        GetMainWindow()->AcceptDropEvent(static_cast<QDropEvent*>(event));
-        return true;
-    default:
-        return QWindow::event(event);
     }
+
+    std::pair<unsigned, unsigned> GetSize() const {
+        return std::make_pair(width(), height());
+    }
+};
+
+#ifdef HAS_OPENGL
+class OpenGLRenderWidget : public RenderWidget {
+public:
+    explicit OpenGLRenderWidget(GRenderWindow* parent, bool is_secondary)
+        : RenderWidget(parent), is_secondary(is_secondary) {
+        setAttribute(Qt::WA_NativeWindow);
+        setAttribute(Qt::WA_PaintOnScreen);
+        windowHandle()->setSurfaceType(QWindow::OpenGLSurface);
+    }
+
+    void SetContext(std::unique_ptr<Frontend::GraphicsContext>&& context_) {
+        context = std::move(context_);
+    }
+
+    void Present() override {
+        if (!isVisible()) {
+            return;
+        }
+        if (!Core::System::GetInstance().IsPoweredOn()) {
+            return;
+        }
+        context->MakeCurrent();
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0);
+        VideoCore::g_renderer->TryPresent(100, is_secondary);
+        context->SwapBuffers();
+        glFinish();
+    }
+
+    QPaintEngine* paintEngine() const override {
+        return nullptr;
+    }
+
+private:
+    std::unique_ptr<Frontend::GraphicsContext> context{};
+    bool is_secondary;
+};
+#endif
+
+struct SoftwareRenderWidget : public RenderWidget {
+    explicit SoftwareRenderWidget(GRenderWindow* parent) : RenderWidget(parent) {}
+
+    void Present() override {
+        if (!isVisible()) {
+            return;
+        }
+        if (!Core::System::GetInstance().IsPoweredOn()) {
+            return;
+        }
+
+        const auto layout{Layout::DefaultFrameLayout(width(), height(), false, false)};
+        QPainter painter(this);
+
+        const auto draw_screen = [&](int fb_id) {
+            const auto rect = fb_id == 0 ? layout.top_screen : layout.bottom_screen;
+            const QImage screen = LoadFramebuffer(fb_id);
+            painter.drawImage(rect.left, rect.top, screen);
+        };
+
+        painter.fillRect(rect(), qRgb(Settings::values.bg_red.GetValue() * 255,
+                                      Settings::values.bg_green.GetValue() * 255,
+                                      Settings::values.bg_blue.GetValue() * 255));
+        draw_screen(0);
+        draw_screen(1);
+
+        painter.end();
+    }
+
+    QImage LoadFramebuffer(int fb_id) {
+        const auto& framebuffer = GPU::g_regs.framebuffer_config[fb_id];
+        const PAddr framebuffer_addr =
+            framebuffer.active_fb == 0 ? framebuffer.address_left1 : framebuffer.address_left2;
+
+        Memory::RasterizerFlushRegion(framebuffer_addr, framebuffer.stride * framebuffer.height);
+        const u8* framebuffer_data = VideoCore::g_memory->GetPhysicalPointer(framebuffer_addr);
+
+        const int width = framebuffer.height;
+        const int height = framebuffer.width;
+        const int bpp = GPU::Regs::BytesPerPixel(framebuffer.color_format);
+
+        QImage image{width, height, QImage::Format_RGBA8888};
+        for (int y = 0; y < height; y++) {
+            for (int x = 0; x < width; x++) {
+                const u8* pixel = framebuffer_data + (x * height + height - y) * bpp;
+                const Common::Vec4 color = [&] {
+                    switch (framebuffer.color_format) {
+                    case GPU::Regs::PixelFormat::RGBA8:
+                        return Common::Color::DecodeRGBA8(pixel);
+                    case GPU::Regs::PixelFormat::RGB8:
+                        return Common::Color::DecodeRGB8(pixel);
+                    case GPU::Regs::PixelFormat::RGB565:
+                        return Common::Color::DecodeRGB565(pixel);
+                    case GPU::Regs::PixelFormat::RGB5A1:
+                        return Common::Color::DecodeRGB5A1(pixel);
+                    case GPU::Regs::PixelFormat::RGBA4:
+                        return Common::Color::DecodeRGBA4(pixel);
+                    }
+                }();
+
+                image.setPixel(x, y, qRgba(color.r(), color.g(), color.b(), color.a()));
+            }
+        }
+        return image;
+    }
+};
+
+static Frontend::WindowSystemType GetWindowSystemType() {
+    // Determine WSI type based on Qt platform.
+    const QString platform_name = QGuiApplication::platformName();
+    if (platform_name == QStringLiteral("windows"))
+        return Frontend::WindowSystemType::Windows;
+    else if (platform_name == QStringLiteral("xcb"))
+        return Frontend::WindowSystemType::X11;
+    else if (platform_name == QStringLiteral("wayland"))
+        return Frontend::WindowSystemType::Wayland;
+    else if (platform_name == QStringLiteral("cocoa"))
+        return Frontend::WindowSystemType::MacOS;
+
+    LOG_CRITICAL(Frontend, "Unknown Qt platform!");
+    return Frontend::WindowSystemType::Windows;
 }
 
-void OpenGLWindow::exposeEvent(QExposeEvent* event) {
+static Frontend::EmuWindow::WindowSystemInfo GetWindowSystemInfo(QWindow* window) {
-    QWindow::requestUpdate();
+    Frontend::EmuWindow::WindowSystemInfo wsi;
-    QWindow::exposeEvent(event);
+    wsi.type = GetWindowSystemType();
+
+    if (window) {
+#if defined(WIN32)
+        // Our Win32 Qt external doesn't have the private API.
+        wsi.render_surface = reinterpret_cast<void*>(window->winId());
+#elif defined(__APPLE__)
+        wsi.render_surface = reinterpret_cast<void* (*)(id, SEL)>(objc_msgSend)(
+            reinterpret_cast<id>(window->winId()), sel_registerName("layer"));
+#else
+        QPlatformNativeInterface* pni = QGuiApplication::platformNativeInterface();
+        wsi.display_connection = pni->nativeResourceForWindow("display", window);
+        if (wsi.type == Frontend::WindowSystemType::Wayland)
+            wsi.render_surface = pni->nativeResourceForWindow("surface", window);
+        else
+            wsi.render_surface = reinterpret_cast<void*>(window->winId());
+#endif
+        wsi.render_surface_scale = static_cast<float>(window->devicePixelRatio());
+    } else {
+        wsi.render_surface = nullptr;
+        wsi.render_surface_scale = 1.0f;
+    }
+
+    return wsi;
 }
 
+std::shared_ptr<Frontend::GraphicsContext> GRenderWindow::main_context;
+
 GRenderWindow::GRenderWindow(QWidget* parent_, EmuThread* emu_thread, bool is_secondary_)
     : QWidget(parent_), EmuWindow(is_secondary_), emu_thread(emu_thread) {
 
@@ -218,11 +406,11 @@ GRenderWindow::GRenderWindow(QWidget* parent_, EmuThread* emu_thread, bool is_se
 GRenderWindow::~GRenderWindow() = default;
 
 void GRenderWindow::MakeCurrent() {
-    core_context->MakeCurrent();
+    main_context->MakeCurrent();
 }
 
 void GRenderWindow::DoneCurrent() {
-    core_context->DoneCurrent();
+    main_context->DoneCurrent();
 }
 
 void GRenderWindow::PollEvents() {
@@ -295,8 +483,9 @@ void GRenderWindow::keyReleaseEvent(QKeyEvent* event) {
 }
 
 void GRenderWindow::mousePressEvent(QMouseEvent* event) {
-    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+    if (event->source() == Qt::MouseEventSynthesizedBySystem) {
         return; // touch input is handled in TouchBeginEvent
+    }
 
     auto pos = event->pos();
     if (event->button() == Qt::LeftButton) {
@@ -309,8 +498,9 @@ void GRenderWindow::mousePressEvent(QMouseEvent* event) {
 }
 
 void GRenderWindow::mouseMoveEvent(QMouseEvent* event) {
-    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+    if (event->source() == Qt::MouseEventSynthesizedBySystem) {
         return; // touch input is handled in TouchUpdateEvent
+    }
 
     auto pos = event->pos();
     const auto [x, y] = ScaleTouch(pos);
@@ -320,8 +510,9 @@ void GRenderWindow::mouseMoveEvent(QMouseEvent* event) {
 }
 
 void GRenderWindow::mouseReleaseEvent(QMouseEvent* event) {
-    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+    if (event->source() == Qt::MouseEventSynthesizedBySystem) {
         return; // touch input is handled in TouchEndEvent
+    }
 
     if (event->button() == Qt::LeftButton)
         this->TouchReleased();
@@ -393,42 +584,61 @@ void GRenderWindow::resizeEvent(QResizeEvent* event) {
     OnFramebufferSizeChanged();
 }
 
-void GRenderWindow::InitRenderTarget() {
+bool GRenderWindow::InitRenderTarget() {
-    ReleaseRenderTarget();
+    {
+        // Create a dummy render widget so that Qt
+        // places the render window at the correct position.
+        const RenderWidget dummy_widget{this};
+    }
 
     first_frame = false;
 
-    GMainWindow* parent = GetMainWindow();
+    const auto graphics_api = Settings::values.graphics_api.GetValue();
-    QWindow* parent_win_handle = parent ? parent->windowHandle() : nullptr;
+    switch (graphics_api) {
-    child_window = new OpenGLWindow(parent_win_handle, this, QOpenGLContext::globalShareContext(),
+    case Settings::GraphicsAPI::Software:
-                                    is_secondary);
+        InitializeSoftware();
-    child_window->create();
+        break;
-    child_widget = createWindowContainer(child_window, this);
+    case Settings::GraphicsAPI::OpenGL:
+        if (!InitializeOpenGL() || !LoadOpenGL()) {
+            return false;
+        }
+        break;
+    }
+
+    // Update the Window System information with the new render target
+    window_info = GetWindowSystemInfo(child_widget->windowHandle());
+
     child_widget->resize(Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight);
 
     layout()->addWidget(child_widget);
+    // Reset minimum required size to avoid resizing issues on the main window after restarting.
+    setMinimumSize(1, 1);
 
-    core_context = CreateSharedContext();
     resize(Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight);
     OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
     OnFramebufferSizeChanged();
     BackupGeometry();
+
+    return true;
 }
 
 void GRenderWindow::ReleaseRenderTarget() {
     if (child_widget) {
         layout()->removeWidget(child_widget);
-        delete child_widget;
+        child_widget->deleteLater();
         child_widget = nullptr;
     }
+    main_context.reset();
 }
 
 void GRenderWindow::CaptureScreenshot(u32 res_scale, const QString& screenshot_path) {
-    if (res_scale == 0)
+    if (res_scale == 0) {
         res_scale = VideoCore::GetResolutionScaleFactor();
+    }
+
     const auto layout{Layout::FrameLayoutFromResolutionScale(res_scale, is_secondary)};
     screenshot_image = QImage(QSize(layout.width, layout.height), QImage::Format_RGB32);
-    VideoCore::RequestScreenshot(
+    VideoCore::g_renderer->RequestScreenshot(
         screenshot_image.bits(),
         [this, screenshot_path] {
             const std::string std_screenshot_path = screenshot_path.toStdString();
@@ -445,6 +655,59 @@ void GRenderWindow::OnMinimalClientAreaChangeRequest(std::pair<u32, u32> minimal
     setMinimumSize(minimal_size.first, minimal_size.second);
 }
 
+bool GRenderWindow::InitializeOpenGL() {
+#ifdef HAS_OPENGL
+    // TODO: One of these flags might be interesting: WA_OpaquePaintEvent, WA_NoBackground,
+    // WA_DontShowOnScreen, WA_DeleteOnClose
+    auto child = new OpenGLRenderWidget(this, is_secondary);
+    child_widget = child;
+    child_widget->windowHandle()->create();
+
+    if (!main_context) {
+        main_context = std::make_shared<OpenGLSharedContext>();
+    }
+
+    auto child_context = CreateSharedContext();
+    child->SetContext(std::move(child_context));
+
+    return true;
+#else
+    QMessageBox::warning(this, tr("OpenGL not available!"),
+                         tr("Citra has not been compiled with OpenGL support."));
+    return false;
+#endif
+}
+
+void GRenderWindow::InitializeSoftware() {
+    child_widget = new SoftwareRenderWidget(this);
+    main_context = std::make_unique<DummyContext>();
+}
+
+bool GRenderWindow::LoadOpenGL() {
+    auto context = CreateSharedContext();
+    auto scope = context->Acquire();
+    if (!gladLoadGL()) {
+        QMessageBox::warning(
+            this, tr("Error while initializing OpenGL!"),
+            tr("Your GPU may not support OpenGL, or you do not have the latest graphics driver."));
+        return false;
+    }
+
+    const QString renderer =
+        QString::fromUtf8(reinterpret_cast<const char*>(glGetString(GL_RENDERER)));
+
+    if (!GLAD_GL_VERSION_4_3) {
+        LOG_ERROR(Frontend, "GPU does not support OpenGL 4.3: {}", renderer.toStdString());
+        QMessageBox::warning(this, tr("Error while initializing OpenGL 4.3!"),
+                             tr("Your GPU may not support OpenGL 4.3, or you do not have the "
+                                "latest graphics driver.<br><br>GL Renderer:<br>%1")
+                                 .arg(renderer));
+        return false;
+    }
+
+    return true;
+}
+
 void GRenderWindow::OnEmulationStarting(EmuThread* emu_thread) {
     this->emu_thread = emu_thread;
 }
@@ -458,29 +721,15 @@ void GRenderWindow::showEvent(QShowEvent* event) {
 }
 
 std::unique_ptr<Frontend::GraphicsContext> GRenderWindow::CreateSharedContext() const {
-    return std::make_unique<GLContext>(QOpenGLContext::globalShareContext());
+#ifdef HAS_OPENGL
-}
+    const auto graphics_api = Settings::values.graphics_api.GetValue();
-
+    if (graphics_api == Settings::GraphicsAPI::OpenGL) {
-GLContext::GLContext(QOpenGLContext* shared_context)
+        auto gl_context = static_cast<OpenGLSharedContext*>(main_context.get());
-    : context(std::make_unique<QOpenGLContext>(shared_context->parent())),
+        // Bind the shared contexts to the main surface in case the backend wants to take over
-      surface(std::make_unique<QOffscreenSurface>(nullptr)) {
+        // presentation
-
+        return std::make_unique<OpenGLSharedContext>(gl_context->GetShareContext(),
-    // disable vsync for any shared contexts
+                                                     child_widget->windowHandle());
-    auto format = shared_context->format();
+    }
-    format.setSwapInterval(0);
+#endif
-
+    return std::make_unique<DummyContext>();
-    context->setShareContext(shared_context);
-    context->setFormat(format);
-    context->create();
-    surface->setParent(shared_context->parent());
-    surface->setFormat(format);
-    surface->create();
-}
-
-void GLContext::MakeCurrent() {
-    context->makeCurrent(surface.get());
-}
-
-void GLContext::DoneCurrent() {
-    context->doneCurrent();
 }

src/citra_qt/bootmanager.h

@@ -27,19 +27,6 @@ namespace VideoCore {
 enum class LoadCallbackStage;
 }
 
-class GLContext : public Frontend::GraphicsContext {
-public:
-    explicit GLContext(QOpenGLContext* shared_context);
-
-    void MakeCurrent() override;
-
-    void DoneCurrent() override;
-
-private:
-    std::unique_ptr<QOpenGLContext> context;
-    std::unique_ptr<QOffscreenSurface> surface;
-};
-
 class EmuThread final : public QThread {
     Q_OBJECT
 
@@ -126,26 +113,6 @@ signals:
     void HideLoadingScreen();
 };
 
-class OpenGLWindow : public QWindow {
-    Q_OBJECT
-public:
-    explicit OpenGLWindow(QWindow* parent, QWidget* event_handler, QOpenGLContext* shared_context,
-                          bool is_secondary = false);
-
-    ~OpenGLWindow();
-
-    void Present();
-
-protected:
-    bool event(QEvent* event) override;
-    void exposeEvent(QExposeEvent* event) override;
-
-private:
-    std::unique_ptr<QOpenGLContext> context;
-    QWidget* event_handler;
-    bool is_secondary;
-};
-
 class GRenderWindow : public QWidget, public Frontend::EmuWindow {
     Q_OBJECT
 
@@ -185,13 +152,15 @@ public:
         return has_focus;
     }
 
-    void InitRenderTarget();
+    bool InitRenderTarget();
 
     /// Destroy the previous run's child_widget which should also destroy the child_window
     void ReleaseRenderTarget();
 
     void CaptureScreenshot(u32 res_scale, const QString& screenshot_path);
 
+    std::pair<u32, u32> ScaleTouch(const QPointF pos) const;
+
 public slots:
 
     void OnEmulationStarting(EmuThread* emu_thread);
@@ -211,29 +180,28 @@ signals:
     void MouseActivity();
 
 private:
-    std::pair<u32, u32> ScaleTouch(QPointF pos) const;
     void TouchBeginEvent(const QTouchEvent* event);
     void TouchUpdateEvent(const QTouchEvent* event);
     void TouchEndEvent();
 
     void OnMinimalClientAreaChangeRequest(std::pair<u32, u32> minimal_size) override;
 
-    std::unique_ptr<GraphicsContext> core_context;
+    bool InitializeOpenGL();
+    void InitializeSoftware();
+    bool LoadOpenGL();
 
-    QByteArray geometry;
-
-    /// Native window handle that backs this presentation widget
-    QWindow* child_window = nullptr;
-
-    /// In order to embed the window into GRenderWindow, you need to use createWindowContainer to
-    /// put the child_window into a widget then add it to the layout. This child_widget can be
-    /// parented to GRenderWindow and use Qt's lifetime system
     QWidget* child_widget = nullptr;
 
     EmuThread* emu_thread;
 
+    /// Main context that will be shared with all other contexts that are requested.
+    /// If this is used in a shared context setting, then this should not be used directly, but
+    /// should instead be shared from
+    static std::shared_ptr<Frontend::GraphicsContext> main_context;
+
     /// Temporary storage of the screenshot taken
     QImage screenshot_image;
+    QByteArray geometry;
     bool first_frame = false;
     bool has_focus = false;
 

src/citra_qt/configuration/config.cpp

@@ -482,6 +482,7 @@ void Config::ReadDebuggingValues() {
         qt_config->value(QStringLiteral("record_frame_times"), false).toBool();
     ReadBasicSetting(Settings::values.use_gdbstub);
     ReadBasicSetting(Settings::values.gdbstub_port);
+    ReadBasicSetting(Settings::values.renderer_debug);
 
     qt_config->beginGroup(QStringLiteral("LLE"));
     for (const auto& service_module : Service::service_module_map) {
@@ -625,7 +626,7 @@ void Config::ReadPathValues() {
 void Config::ReadRendererValues() {
     qt_config->beginGroup(QStringLiteral("Renderer"));
 
-    ReadGlobalSetting(Settings::values.use_hw_renderer);
+    ReadGlobalSetting(Settings::values.graphics_api);
     ReadGlobalSetting(Settings::values.use_hw_shader);
 #ifdef __APPLE__
     // Hardware shader is broken on macos with Intel GPUs thanks to poor drivers.
@@ -992,6 +993,7 @@ void Config::SaveDebuggingValues() {
     qt_config->setValue(QStringLiteral("record_frame_times"), Settings::values.record_frame_times);
     WriteBasicSetting(Settings::values.use_gdbstub);
     WriteBasicSetting(Settings::values.gdbstub_port);
+    WriteBasicSetting(Settings::values.renderer_debug);
 
     qt_config->beginGroup(QStringLiteral("LLE"));
     for (const auto& service_module : Settings::values.lle_modules) {
@@ -1103,7 +1105,7 @@ void Config::SavePathValues() {
 void Config::SaveRendererValues() {
     qt_config->beginGroup(QStringLiteral("Renderer"));
 
-    WriteGlobalSetting(Settings::values.use_hw_renderer);
+    WriteGlobalSetting(Settings::values.graphics_api);
     WriteGlobalSetting(Settings::values.use_hw_shader);
 #ifdef __APPLE__
     // Hardware shader is broken on macos thanks to poor drivers.

src/citra_qt/configuration/configuration_shared.h

@@ -83,6 +83,21 @@ template <>
 void SetPerGameSetting(QComboBox* combobox,
                        const Settings::SwitchableSetting<std::string>* setting);
 
+/// Given an index of a combobox setting extracts the setting taking into
+/// account per-game status
+template <typename Type, bool ranged>
+Type GetComboboxSetting(int index, const Settings::SwitchableSetting<Type, ranged>* setting) {
+    if (Settings::IsConfiguringGlobal() && setting->UsingGlobal()) {
+        return static_cast<Type>(index);
+    } else if (!Settings::IsConfiguringGlobal()) {
+        if (index == 0) {
+            return setting->GetValue();
+        } else {
+            return static_cast<Type>(index - ConfigurationShared::USE_GLOBAL_OFFSET);
+        }
+    }
+}
+
 /// Given a Qt widget sets the background color to indicate whether the setting
 /// is per-game overriden (highlighted) or global (non-highlighted)
 void SetHighlight(QWidget* widget, bool highlighted);

src/citra_qt/configuration/configure_debug.cpp

@@ -37,6 +37,7 @@ ConfigureDebug::ConfigureDebug(QWidget* parent)
 
     const bool is_powered_on = Core::System::GetInstance().IsPoweredOn();
     ui->toggle_cpu_jit->setEnabled(!is_powered_on);
+    ui->toggle_renderer_debug->setEnabled(!is_powered_on);
 
     // Set a minimum width for the label to prevent the slider from changing size.
     // This scales across DPIs. (This value should be enough for "xxx%")
@@ -62,6 +63,7 @@ void ConfigureDebug::SetConfiguration() {
     ui->toggle_console->setChecked(UISettings::values.show_console.GetValue());
     ui->log_filter_edit->setText(QString::fromStdString(Settings::values.log_filter.GetValue()));
     ui->toggle_cpu_jit->setChecked(Settings::values.use_cpu_jit.GetValue());
+    ui->toggle_renderer_debug->setChecked(Settings::values.renderer_debug.GetValue());
 
     if (!Settings::IsConfiguringGlobal()) {
         if (Settings::values.cpu_clock_percentage.UsingGlobal()) {
@@ -91,6 +93,7 @@ void ConfigureDebug::ApplyConfiguration() {
     filter.ParseFilterString(Settings::values.log_filter.GetValue());
     Log::SetGlobalFilter(filter);
     Settings::values.use_cpu_jit = ui->toggle_cpu_jit->isChecked();
+    Settings::values.renderer_debug = ui->toggle_renderer_debug->isChecked();
 
     ConfigurationShared::ApplyPerGameSetting(
         &Settings::values.cpu_clock_percentage, ui->clock_speed_combo,

src/citra_qt/configuration/configure_debug.h

@@ -23,5 +23,6 @@ public:
     void SetConfiguration();
     void SetupPerGameUI();
 
+private:
     std::unique_ptr<Ui::ConfigureDebug> ui;
 };

src/citra_qt/configuration/configure_debug.ui

@@ -6,8 +6,8 @@
    <rect>
     <x>0</x>
     <y>0</y>
-    <width>443</width>
+    <width>523</width>
-    <height>358</height>
+    <height>447</height>
    </rect>
   </property>
   <property name="windowTitle">
@@ -112,12 +112,34 @@
       <string>CPU</string>
      </property>
      <layout class="QGridLayout" name="gridLayout_2">
+      <item row="2" column="0">
+       <widget class="QCheckBox" name="toggle_cpu_jit">
+        <property name="toolTip">
+         <string>&lt;html&gt;&lt;head/&gt;&lt;body&gt;&lt;p&gt;Enables the use of the ARM JIT compiler for emulating the 3DS CPUs. Don't disable unless for debugging purposes&lt;/p&gt;&lt;/body&gt;&lt;/html&gt;</string>
+        </property>
+        <property name="text">
+         <string>Enable CPU JIT</string>
+        </property>
+       </widget>
+      </item>
       <item row="1" column="0">
        <widget class="QWidget" name="clock_speed_widget" native="true">
         <layout class="QHBoxLayout" name="clock_speed_layout">
          <property name="spacing">
           <number>7</number>
          </property>
+         <property name="leftMargin">
+          <number>0</number>
+         </property>
+         <property name="topMargin">
+          <number>0</number>
+         </property>
+         <property name="rightMargin">
+          <number>0</number>
+         </property>
+         <property name="bottomMargin">
+          <number>0</number>
+         </property>
          <item>
           <widget class="QComboBox" name="clock_speed_combo">
            <item>
@@ -180,13 +202,10 @@
         </layout>
        </widget>
       </item>
-      <item row="2" column="0">
+      <item row="3" column="0">
-       <widget class="QCheckBox" name="toggle_cpu_jit">
+       <widget class="QCheckBox" name="toggle_renderer_debug">
-        <property name="toolTip">
-         <string>&lt;html&gt;&lt;head/&gt;&lt;body&gt;&lt;p&gt;Enables the use of the ARM JIT compiler for emulating the 3DS CPUs. Don't disable unless for debugging purposes&lt;/p&gt;&lt;/body&gt;&lt;/html&gt;</string>
-        </property>
         <property name="text">
-         <string>Enable CPU JIT</string>
+         <string>Enable debug renderer</string>
         </property>
        </widget>
       </item>

src/citra_qt/configuration/configure_enhancements.cpp

@@ -22,7 +22,9 @@ ConfigureEnhancements::ConfigureEnhancements(QWidget* parent)
 
     ui->layout_group->setEnabled(!Settings::values.custom_layout);
 
-    ui->resolution_factor_combobox->setEnabled(Settings::values.use_hw_renderer.GetValue());
+    const auto graphics_api = Settings::values.graphics_api.GetValue();
+    const bool res_scale_enabled = graphics_api != Settings::GraphicsAPI::Software;
+    ui->resolution_factor_combobox->setEnabled(res_scale_enabled);
 
     connect(ui->render_3d_combobox,
             static_cast<void (QComboBox::*)(int)>(&QComboBox::currentIndexChanged), this,

src/citra_qt/configuration/configure_graphics.cpp

@@ -16,22 +16,20 @@ ConfigureGraphics::ConfigureGraphics(QWidget* parent)
     : QWidget(parent), ui(std::make_unique<Ui::ConfigureGraphics>()) {
     ui->setupUi(this);
 
-    SetupPerGameUI();
-    SetConfiguration();
-
-    ui->hw_renderer_group->setEnabled(ui->hw_renderer_group->isEnabled() &&
-                                      ui->toggle_hw_renderer->isChecked());
     ui->toggle_vsync_new->setEnabled(!Core::System::GetInstance().IsPoweredOn());
+    // Set the index to -1 to ensure the below lambda is called with setCurrentIndex
+    ui->graphics_api_combo->setCurrentIndex(-1);
 
-    connect(ui->toggle_hw_renderer, &QCheckBox::toggled, this, [this] {
+    connect(ui->graphics_api_combo, qOverload<int>(&QComboBox::currentIndexChanged), this,
-        const bool checked = ui->toggle_hw_renderer->isChecked();
+            [this](int index) {
-        ui->hw_renderer_group->setEnabled(checked);
+                const auto graphics_api =
-        ui->toggle_disk_shader_cache->setEnabled(checked && ui->toggle_hw_shader->isChecked());
+                    ConfigurationShared::GetComboboxSetting(index, &Settings::values.graphics_api);
-    });
+                const bool is_software = graphics_api == Settings::GraphicsAPI::Software;
 
-    ui->hw_shader_group->setEnabled(ui->toggle_hw_shader->isChecked());
+                ui->hw_renderer_group->setEnabled(!is_software);
-    ui->toggle_disk_shader_cache->setEnabled(ui->toggle_hw_renderer->isChecked() &&
+                ui->toggle_disk_shader_cache->setEnabled(!is_software &&
                                                          ui->toggle_hw_shader->isChecked());
+            });
 
     connect(ui->toggle_hw_shader, &QCheckBox::toggled, this, [this] {
         const bool checked = ui->toggle_hw_shader->isChecked();
@@ -60,12 +58,24 @@ ConfigureGraphics::ConfigureGraphics(QWidget* parent)
     // TODO(B3N30): Hide this for macs with none Intel GPUs, too.
     ui->toggle_separable_shader->setVisible(false);
 #endif
+
+    SetupPerGameUI();
+    SetConfiguration();
 }
 
 ConfigureGraphics::~ConfigureGraphics() = default;
 
 void ConfigureGraphics::SetConfiguration() {
-    ui->toggle_hw_renderer->setChecked(Settings::values.use_hw_renderer.GetValue());
+    if (!Settings::IsConfiguringGlobal()) {
+        ConfigurationShared::SetHighlight(ui->graphics_api_group,
+                                          !Settings::values.graphics_api.UsingGlobal());
+        ConfigurationShared::SetPerGameSetting(ui->graphics_api_combo,
+                                               &Settings::values.graphics_api);
+    } else {
+        ui->graphics_api_combo->setCurrentIndex(
+            static_cast<int>(Settings::values.graphics_api.GetValue()));
+    }
+
     ui->toggle_hw_shader->setChecked(Settings::values.use_hw_shader.GetValue());
     ui->toggle_separable_shader->setChecked(Settings::values.separable_shader.GetValue());
     ui->toggle_accurate_mul->setChecked(Settings::values.shaders_accurate_mul.GetValue());
@@ -78,8 +88,8 @@ void ConfigureGraphics::SetConfiguration() {
 }
 
 void ConfigureGraphics::ApplyConfiguration() {
-    ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_hw_renderer,
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.graphics_api,
-                                             ui->toggle_hw_renderer, use_hw_renderer);
+                                             ui->graphics_api_combo);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_hw_shader, ui->toggle_hw_shader,
                                              use_hw_shader);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.separable_shader,
@@ -103,7 +113,7 @@ void ConfigureGraphics::RetranslateUI() {
 void ConfigureGraphics::SetupPerGameUI() {
     // Block the global settings if a game is currently running that overrides them
     if (Settings::IsConfiguringGlobal()) {
-        ui->toggle_hw_renderer->setEnabled(Settings::values.use_hw_renderer.UsingGlobal());
+        ui->graphics_api_group->setEnabled(Settings::values.graphics_api.UsingGlobal());
         ui->toggle_hw_shader->setEnabled(Settings::values.use_hw_shader.UsingGlobal());
         ui->toggle_separable_shader->setEnabled(Settings::values.separable_shader.UsingGlobal());
         ui->toggle_accurate_mul->setEnabled(Settings::values.shaders_accurate_mul.UsingGlobal());
@@ -115,8 +125,10 @@ void ConfigureGraphics::SetupPerGameUI() {
 
     ui->toggle_shader_jit->setVisible(false);
 
-    ConfigurationShared::SetColoredTristate(ui->toggle_hw_renderer,
+    ConfigurationShared::SetColoredComboBox(
-                                            Settings::values.use_hw_renderer, use_hw_renderer);
+        ui->graphics_api_combo, ui->graphics_api_group,
+        static_cast<u32>(Settings::values.graphics_api.GetValue(true)));
+
     ConfigurationShared::SetColoredTristate(ui->toggle_hw_shader, Settings::values.use_hw_shader,
                                             use_hw_shader);
     ConfigurationShared::SetColoredTristate(ui->toggle_separable_shader,

src/citra_qt/configuration/configure_graphics.h

@@ -28,9 +28,9 @@ public:
 
     void UpdateBackgroundColorButton(const QColor& color);
 
+private:
     void SetupPerGameUI();
 
-    ConfigurationShared::CheckState use_hw_renderer;
     ConfigurationShared::CheckState use_hw_shader;
     ConfigurationShared::CheckState separable_shader;
     ConfigurationShared::CheckState shaders_accurate_mul;

src/citra_qt/configuration/configure_graphics.ui

@@ -7,7 +7,7 @@
     <x>0</x>
     <y>0</y>
     <width>400</width>
-    <height>430</height>
+    <height>443</height>
    </rect>
   </property>
   <property name="minimumSize">
@@ -20,27 +20,65 @@
    <string>Form</string>
   </property>
   <layout class="QVBoxLayout" name="verticalLayout">
+   <item>
+    <widget class="QGroupBox" name="apiBox">
+     <property name="title">
+      <string>API Settings</string>
+     </property>
+     <layout class="QVBoxLayout" name="verticalLayout_3">
+      <item>
+       <widget class="QWidget" name="graphics_api_group" native="true">
+        <layout class="QHBoxLayout" name="graphics_api_group_2">
+         <property name="leftMargin">
+          <number>0</number>
+         </property>
+         <property name="topMargin">
+          <number>0</number>
+         </property>
+         <property name="rightMargin">
+          <number>0</number>
+         </property>
+         <property name="bottomMargin">
+          <number>0</number>
+         </property>
+         <item>
+          <widget class="QLabel" name="graphics_api_label">
+           <property name="text">
+            <string>Graphics API</string>
+           </property>
+          </widget>
+         </item>
+         <item>
+          <widget class="QComboBox" name="graphics_api_combo">
+           <item>
+            <property name="text">
+             <string>Software</string>
+            </property>
+           </item>
+           <item>
+            <property name="text">
+             <string>OpenGL</string>
+            </property>
+           </item>
+          </widget>
+         </item>
+        </layout>
+       </widget>
+      </item>
+     </layout>
+    </widget>
+   </item>
    <item>
     <widget class="QGroupBox" name="rendererBox">
      <property name="title">
       <string>Renderer</string>
      </property>
      <layout class="QVBoxLayout" name="verticalLayout_6">
-      <item>
-       <widget class="QCheckBox" name="toggle_hw_renderer">
-        <property name="toolTip">
-         <string>&lt;html&gt;&lt;head/&gt;&lt;body&gt;&lt;p&gt;Use OpenGL to accelerate rendering.&lt;/p&gt;&lt;p&gt;Disable to debug graphics-related problem.&lt;/p&gt;&lt;/body&gt;&lt;/html&gt;</string>
-        </property>
-        <property name="text">
-         <string>Enable Hardware Renderer</string>
-        </property>
-       </widget>
-      </item>
       <item>
        <widget class="QWidget" name="hw_renderer_group" native="true">
         <layout class="QVBoxLayout" name="verticalLayout_5">
          <property name="leftMargin">
-          <number>16</number>
+          <number>0</number>
          </property>
          <property name="topMargin">
           <number>0</number>
@@ -157,8 +195,6 @@
   </layout>
  </widget>
  <tabstops>
-  <tabstop>toggle_hw_renderer</tabstop>
-  <tabstop>toggle_hw_shader</tabstop>
   <tabstop>toggle_separable_shader</tabstop>
   <tabstop>toggle_accurate_mul</tabstop>
   <tabstop>toggle_shader_jit</tabstop>

src/citra_qt/main.cpp

@@ -12,7 +12,6 @@
 #include <QFutureWatcher>
 #include <QLabel>
 #include <QMessageBox>
-#include <QOpenGLFunctions_4_3_Core>
 #include <QSysInfo>
 #include <QtConcurrent/QtConcurrentRun>
 #include <QtGui>
@@ -88,7 +87,6 @@
 #include "core/file_sys/archive_extsavedata.h"
 #include "core/file_sys/archive_source_sd_savedata.h"
 #include "core/frontend/applets/default_applets.h"
-#include "core/frontend/scope_acquire_context.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hle/service/cfg/cfg.h"
 #include "core/hle/service/fs/archive.h"
@@ -1026,16 +1024,7 @@ bool GMainWindow::LoadROM(const QString& filename) {
     render_window->InitRenderTarget();
     secondary_window->InitRenderTarget();
 
-    Frontend::ScopeAcquireContext scope(*render_window);
+    const auto scope = render_window->Acquire();
-
-    const QString below_gl43_title = tr("OpenGL 4.3 Unsupported");
-    const QString below_gl43_message = tr("Your GPU may not support OpenGL 4.3, or you do not "
-                                          "have the latest graphics driver.");
-
-    if (!QOpenGLContext::globalShareContext()->versionFunctions<QOpenGLFunctions_4_3_Core>()) {
-        QMessageBox::critical(this, below_gl43_title, below_gl43_message);
-        return false;
-    }
 
     Core::System& system{Core::System::GetInstance()};
 
@@ -1091,28 +1080,6 @@ bool GMainWindow::LoadROM(const QString& filename) {
                                   tr("GBA Virtual Console ROMs are not supported by Citra."));
             break;
 
-        case Core::System::ResultStatus::ErrorVideoCore:
-            QMessageBox::critical(
-                this, tr("Video Core Error"),
-                tr("An error has occurred. Please <a "
-                   "href='https://community.citra-emu.org/t/how-to-upload-the-log-file/296'>see "
-                   "the "
-                   "log</a> for more details. "
-                   "Ensure that you have the latest graphics drivers for your GPU."));
-            break;
-
-        case Core::System::ResultStatus::ErrorVideoCore_ErrorGenericDrivers:
-            QMessageBox::critical(
-                this, tr("Video Core Error"),
-                tr("You are running default Windows drivers "
-                   "for your GPU. You need to install the "
-                   "proper drivers for your graphics card from the manufacturer's website."));
-            break;
-
-        case Core::System::ResultStatus::ErrorVideoCore_ErrorBelowGL43:
-            QMessageBox::critical(this, below_gl43_title, below_gl43_message);
-            break;
-
         default:
             QMessageBox::critical(
                 this, tr("Error while loading ROM!"),
@@ -2786,14 +2753,6 @@ int main(int argc, char* argv[]) {
     QCoreApplication::setOrganizationName(QStringLiteral("Citra team"));
     QCoreApplication::setApplicationName(QStringLiteral("Citra"));
 
-    QSurfaceFormat format;
-    format.setVersion(4, 3);
-    format.setProfile(QSurfaceFormat::CoreProfile);
-    format.setSwapInterval(0);
-    // TODO: expose a setting for buffer value (ie default/single/double/triple)
-    format.setSwapBehavior(QSurfaceFormat::DefaultSwapBehavior);
-    QSurfaceFormat::setDefaultFormat(format);
-
     SetHighDPIAttributes();
 
 #ifdef __APPLE__

src/common/common_funcs.h

@@ -55,6 +55,60 @@ __declspec(dllimport) void __stdcall DebugBreak(void);
 
 #endif // _MSC_VER
 
+#define DECLARE_ENUM_FLAG_OPERATORS(type)                                                          \
+    [[nodiscard]] constexpr type operator|(type a, type b) noexcept {                              \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<type>(static_cast<T>(a) | static_cast<T>(b));                           \
+    }                                                                                              \
+    [[nodiscard]] constexpr type operator&(type a, type b) noexcept {                              \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<type>(static_cast<T>(a) & static_cast<T>(b));                           \
+    }                                                                                              \
+    [[nodiscard]] constexpr type operator^(type a, type b) noexcept {                              \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<type>(static_cast<T>(a) ^ static_cast<T>(b));                           \
+    }                                                                                              \
+    [[nodiscard]] constexpr type operator<<(type a, type b) noexcept {                             \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<type>(static_cast<T>(a) << static_cast<T>(b));                          \
+    }                                                                                              \
+    [[nodiscard]] constexpr type operator>>(type a, type b) noexcept {                             \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<type>(static_cast<T>(a) >> static_cast<T>(b));                          \
+    }                                                                                              \
+    constexpr type& operator|=(type& a, type b) noexcept {                                         \
+        a = a | b;                                                                                 \
+        return a;                                                                                  \
+    }                                                                                              \
+    constexpr type& operator&=(type& a, type b) noexcept {                                         \
+        a = a & b;                                                                                 \
+        return a;                                                                                  \
+    }                                                                                              \
+    constexpr type& operator^=(type& a, type b) noexcept {                                         \
+        a = a ^ b;                                                                                 \
+        return a;                                                                                  \
+    }                                                                                              \
+    constexpr type& operator<<=(type& a, type b) noexcept {                                        \
+        a = a << b;                                                                                \
+        return a;                                                                                  \
+    }                                                                                              \
+    constexpr type& operator>>=(type& a, type b) noexcept {                                        \
+        a = a >> b;                                                                                \
+        return a;                                                                                  \
+    }                                                                                              \
+    [[nodiscard]] constexpr type operator~(type key) noexcept {                                    \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<type>(~static_cast<T>(key));                                            \
+    }                                                                                              \
+    [[nodiscard]] constexpr bool True(type key) noexcept {                                         \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<T>(key) != 0;                                                           \
+    }                                                                                              \
+    [[nodiscard]] constexpr bool False(type key) noexcept {                                        \
+        using T = std::underlying_type_t<type>;                                                    \
+        return static_cast<T>(key) == 0;                                                           \
+    }
+
 // Generic function to get last error message.
 // Call directly after the command or use the error num.
 // This function might change the error code.

src/common/settings.cpp

@@ -20,6 +20,8 @@
 
 namespace Settings {
 
+namespace {
+
 std::string_view GetAudioEmulationName(AudioEmulation emulation) {
     switch (emulation) {
     case AudioEmulation::HLE:
@@ -31,6 +33,17 @@ std::string_view GetAudioEmulationName(AudioEmulation emulation) {
     }
 };
 
+std::string_view GetGraphicsAPIName(GraphicsAPI api) {
+    switch (api) {
+    case GraphicsAPI::Software:
+        return "Software";
+    case GraphicsAPI::OpenGL:
+        return "OpenGL";
+    }
+}
+
+} // Anonymous namespace
+
 Values values = {};
 static bool configuring_global = true;
 
@@ -38,7 +51,6 @@ void Apply() {
     GDBStub::SetServerPort(values.gdbstub_port.GetValue());
     GDBStub::ToggleServer(values.use_gdbstub.GetValue());
 
-    VideoCore::g_hw_renderer_enabled = values.use_hw_renderer.GetValue();
     VideoCore::g_shader_jit_enabled = values.use_shader_jit.GetValue();
     VideoCore::g_hw_shader_enabled = values.use_hw_shader.GetValue();
     VideoCore::g_separable_shader_enabled = values.separable_shader.GetValue();
@@ -101,7 +113,7 @@ void LogSettings() {
     log_setting("Core_UseCpuJit", values.use_cpu_jit.GetValue());
     log_setting("Core_CPUClockPercentage", values.cpu_clock_percentage.GetValue());
     log_setting("Renderer_UseGLES", values.use_gles.GetValue());
-    log_setting("Renderer_UseHwRenderer", values.use_hw_renderer.GetValue());
+    log_setting("Renderer_GraphicsAPI", GetGraphicsAPIName(values.graphics_api.GetValue()));
     log_setting("Renderer_UseHwShader", values.use_hw_shader.GetValue());
     log_setting("Renderer_SeparableShader", values.separable_shader.GetValue());
     log_setting("Renderer_ShadersAccurateMul", values.shaders_accurate_mul.GetValue());
@@ -186,7 +198,7 @@ void RestoreGlobalState(bool is_powered_on) {
     values.is_new_3ds.SetGlobal(true);
 
     // Renderer
-    values.use_hw_renderer.SetGlobal(true);
+    values.graphics_api.SetGlobal(true);
     values.use_hw_shader.SetGlobal(true);
     values.separable_shader.SetGlobal(true);
     values.use_disk_shader_cache.SetGlobal(true);

src/common/settings.h

@@ -15,6 +15,11 @@
 
 namespace Settings {
 
+enum class GraphicsAPI {
+    Software = 0,
+    OpenGL = 1,
+};
+
 enum class InitClock : u32 {
     SystemTime = 0,
     FixedTime = 1,
@@ -415,8 +420,9 @@ struct Values {
     Setting<bool> allow_plugin_loader{true, "allow_plugin_loader"};
 
     // Renderer
+    SwitchableSetting<GraphicsAPI> graphics_api{GraphicsAPI::OpenGL, "graphics_api"};
     Setting<bool> use_gles{false, "use_gles"};
-    SwitchableSetting<bool> use_hw_renderer{true, "use_hw_renderer"};
+    Setting<bool> renderer_debug{false, "renderer_debug"};
     SwitchableSetting<bool> use_hw_shader{true, "use_hw_shader"};
     SwitchableSetting<bool> separable_shader{false, "use_separable_shader"};
     SwitchableSetting<bool> use_disk_shader_cache{true, "use_disk_shader_cache"};

src/core/CMakeLists.txt

@@ -113,8 +113,6 @@ add_library(core STATIC
     frontend/input.h
     frontend/mic.cpp
     frontend/mic.h
-    frontend/scope_acquire_context.cpp
-    frontend/scope_acquire_context.h
     gdbstub/gdbstub.cpp
     gdbstub/gdbstub.h
     hle/applets/applet.cpp

src/core/core.cpp

@@ -431,17 +431,7 @@ System::ResultStatus System::Init(Frontend::EmuWindow& emu_window,
     video_dumper = std::make_unique<VideoDumper::NullBackend>();
 #endif
 
-    VideoCore::ResultStatus result = VideoCore::Init(emu_window, secondary_window, *memory);
+    VideoCore::Init(emu_window, secondary_window, *this);
-    if (result != VideoCore::ResultStatus::Success) {
-        switch (result) {
-        case VideoCore::ResultStatus::ErrorGenericDrivers:
-            return ResultStatus::ErrorVideoCore_ErrorGenericDrivers;
-        case VideoCore::ResultStatus::ErrorBelowGL43:
-            return ResultStatus::ErrorVideoCore_ErrorBelowGL43;
-        default:
-            return ResultStatus::ErrorVideoCore;
-        }
-    }
 
     LOG_DEBUG(Core, "Initialized OK");
 
@@ -450,7 +440,7 @@ System::ResultStatus System::Init(Frontend::EmuWindow& emu_window,
     return ResultStatus::Success;
 }
 
-RendererBase& System::Renderer() {
+VideoCore::RendererBase& System::Renderer() {
     return *VideoCore::g_renderer;
 }
 

src/core/core.h

@@ -57,7 +57,9 @@ namespace VideoDumper {
 class Backend;
 }
 
+namespace VideoCore {
 class RendererBase;
+}
 
 namespace Core {
 
@@ -87,11 +89,6 @@ public:
         ErrorLoader_ErrorGbaTitle, ///< Error loading the specified application as it is GBA Virtual
                                    ///< Console
         ErrorSystemFiles,          ///< Error in finding system files
-        ErrorVideoCore,            ///< Error in the video core
-        ErrorVideoCore_ErrorGenericDrivers, ///< Error in the video core due to the user having
-                                            /// generic drivers installed
-        ErrorVideoCore_ErrorBelowGL43,      ///< Error in the video core due to the user not having
-                                            /// OpenGL 4.3 or higher
         ErrorSavestate,            ///< Error saving or loading
         ShutdownRequested,         ///< Emulated program requested a system shutdown
         ErrorUnknown               ///< Any other error
@@ -210,7 +207,7 @@ public:
         return *dsp_core;
     }
 
-    [[nodiscard]] RendererBase& Renderer();
+    [[nodiscard]] VideoCore::RendererBase& Renderer();
 
     /**
      * Gets a reference to the service manager.

src/core/frontend/emu_window.h

@@ -14,6 +14,17 @@
 
 namespace Frontend {
 
+/// Information for the Graphics Backends signifying what type of screen pointer is in
+/// WindowInformation
+enum class WindowSystemType : u8 {
+    Headless,
+    Android,
+    Windows,
+    MacOS,
+    X11,
+    Wayland,
+};
+
 struct Frame;
 /**
  * For smooth Vsync rendering, we want to always present the latest frame that the core generates,
@@ -62,11 +73,33 @@ 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
+    [[nodiscard]] Scoped Acquire() {
+        return Scoped{*this};
+    }
 };
 
 /**
@@ -100,6 +133,23 @@ public:
             Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight};
     };
 
+    /// Data describing host window system information
+    struct WindowSystemInfo {
+        // Window system type. Determines which GL context or Vulkan WSI is used.
+        WindowSystemType type = WindowSystemType::Headless;
+
+        // Connection to a display server. This is used on X11 and Wayland platforms.
+        void* display_connection = nullptr;
+
+        // Render surface. This is a pointer to the native window handle, which depends
+        // on the platform. e.g. HWND for Windows, Window for X11. If the surface is
+        // set to nullptr, the video backend will run in headless mode.
+        void* render_surface = nullptr;
+
+        // Scale of the render surface. For hidpi systems, this will be >1.
+        float render_surface_scale = 1.0f;
+    };
+
     /// Polls window events
     virtual void PollEvents() = 0;
 
@@ -163,6 +213,13 @@ public:
         config = val;
     }
 
+    /**
+     * Returns system information about the drawing area.
+     */
+    const WindowSystemInfo& GetWindowInfo() const {
+        return window_info;
+    }
+
     /**
      * Gets the framebuffer layout (width, height, and screen regions)
      * @note This method is thread-safe
@@ -211,6 +268,7 @@ protected:
     }
 
     bool is_secondary{};
+    WindowSystemInfo window_info;
 
 private:
     /**

src/core/frontend/scope_acquire_context.cpp

@@ -1,17 +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 Frontend {
-
-ScopeAcquireContext::ScopeAcquireContext(Frontend::GraphicsContext& context) : context{context} {
-    context.MakeCurrent();
-}
-ScopeAcquireContext::~ScopeAcquireContext() {
-    context.DoneCurrent();
-}
-
-} // namespace 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 Frontend {
-
-class GraphicsContext;
-
-/// Helper class to acquire/release window context within a given scope
-class ScopeAcquireContext : NonCopyable {
-public:
-    explicit ScopeAcquireContext(Frontend::GraphicsContext& context);
-    ~ScopeAcquireContext();
-
-private:
-    Frontend::GraphicsContext& context;
-};
-
-} // namespace Frontend

src/core/telemetry_session.cpp

@@ -133,8 +133,8 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
              Settings::values.resolution_factor.GetValue());
     AddField(Telemetry::FieldType::UserConfig, "Renderer_FrameLimit",
              Settings::values.frame_limit.GetValue());
-    AddField(Telemetry::FieldType::UserConfig, "Renderer_UseHwRenderer",
+    AddField(Telemetry::FieldType::UserConfig, "Renderer_Backend",
-             Settings::values.use_hw_renderer.GetValue());
+             static_cast<int>(Settings::values.graphics_api.GetValue()));
     AddField(Telemetry::FieldType::UserConfig, "Renderer_UseHwShader",
              Settings::values.use_hw_shader.GetValue());
     AddField(Telemetry::FieldType::UserConfig, "Renderer_ShadersAccurateMul",

src/video_core/CMakeLists.txt

@@ -13,6 +13,8 @@ add_library(video_core STATIC
     precompiled_headers.h
     primitive_assembly.cpp
     primitive_assembly.h
+    rasterizer_accelerated.cpp
+    rasterizer_accelerated.h
     rasterizer_interface.h
     regs.cpp
     regs.h
@@ -39,6 +41,8 @@ add_library(video_core STATIC
     rasterizer_cache/texture_runtime.h
     renderer_opengl/frame_dumper_opengl.cpp
     renderer_opengl/frame_dumper_opengl.h
+    renderer_opengl/gl_driver.cpp
+    renderer_opengl/gl_driver.h
     renderer_opengl/gl_rasterizer.cpp
     renderer_opengl/gl_rasterizer.h
     renderer_opengl/gl_resource_manager.cpp
@@ -82,6 +86,22 @@ add_library(video_core STATIC
     #temporary, move these back in alphabetical order before merging
     renderer_opengl/gl_format_reinterpreter.cpp
     renderer_opengl/gl_format_reinterpreter.h
+    renderer_software/rasterizer.cpp
+    renderer_software/rasterizer.h
+    renderer_software/renderer_software.cpp
+    renderer_software/renderer_software.h
+    renderer_software/sw_clipper.cpp
+    renderer_software/sw_clipper.h
+    renderer_software/sw_framebuffer.cpp
+    renderer_software/sw_framebuffer.h
+    renderer_software/sw_lighting.cpp
+    renderer_software/sw_lighting.h
+    renderer_software/sw_proctex.cpp
+    renderer_software/sw_proctex.h
+    renderer_software/sw_rasterizer.cpp
+    renderer_software/sw_rasterizer.h
+    renderer_software/sw_texturing.cpp
+    renderer_software/sw_texturing.h
     shader/debug_data.h
     shader/shader.cpp
     shader/shader.h
@@ -91,20 +111,8 @@ add_library(video_core STATIC
     shader/shader_jit_x64_compiler.cpp
     shader/shader_jit_x64.h
     shader/shader_jit_x64_compiler.h
-    swrasterizer/clipper.cpp
+    shader/shader_uniforms.cpp
-    swrasterizer/clipper.h
+    shader/shader_uniforms.h
-    swrasterizer/framebuffer.cpp
-    swrasterizer/framebuffer.h
-    swrasterizer/lighting.cpp
-    swrasterizer/lighting.h
-    swrasterizer/proctex.cpp
-    swrasterizer/proctex.h
-    swrasterizer/rasterizer.cpp
-    swrasterizer/rasterizer.h
-    swrasterizer/swrasterizer.cpp
-    swrasterizer/swrasterizer.h
-    swrasterizer/texturing.cpp
-    swrasterizer/texturing.h
     texture/etc1.cpp
     texture/etc1.h
     texture/texture_decode.cpp

src/video_core/rasterizer_accelerated.cpp

@@ -0,0 +1,832 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <limits>
+#include "common/alignment.h"
+#include "core/memory.h"
+#include "video_core/pica_state.h"
+#include "video_core/rasterizer_accelerated.h"
+
+namespace VideoCore {
+
+static Common::Vec4f ColorRGBA8(const u32 color) {
+    const auto rgba =
+        Common::Vec4u{color >> 0 & 0xFF, color >> 8 & 0xFF, color >> 16 & 0xFF, color >> 24 & 0xFF};
+    return rgba / 255.0f;
+}
+
+static Common::Vec3f LightColor(const Pica::LightingRegs::LightColor& color) {
+    return Common::Vec3u{color.r, color.g, color.b} / 255.0f;
+}
+
+RasterizerAccelerated::HardwareVertex::HardwareVertex(const Pica::Shader::OutputVertex& v,
+                                                      bool flip_quaternion) {
+    position[0] = v.pos.x.ToFloat32();
+    position[1] = v.pos.y.ToFloat32();
+    position[2] = v.pos.z.ToFloat32();
+    position[3] = v.pos.w.ToFloat32();
+    color[0] = v.color.x.ToFloat32();
+    color[1] = v.color.y.ToFloat32();
+    color[2] = v.color.z.ToFloat32();
+    color[3] = v.color.w.ToFloat32();
+    tex_coord0[0] = v.tc0.x.ToFloat32();
+    tex_coord0[1] = v.tc0.y.ToFloat32();
+    tex_coord1[0] = v.tc1.x.ToFloat32();
+    tex_coord1[1] = v.tc1.y.ToFloat32();
+    tex_coord2[0] = v.tc2.x.ToFloat32();
+    tex_coord2[1] = v.tc2.y.ToFloat32();
+    tex_coord0_w = v.tc0_w.ToFloat32();
+    normquat[0] = v.quat.x.ToFloat32();
+    normquat[1] = v.quat.y.ToFloat32();
+    normquat[2] = v.quat.z.ToFloat32();
+    normquat[3] = v.quat.w.ToFloat32();
+    view[0] = v.view.x.ToFloat32();
+    view[1] = v.view.y.ToFloat32();
+    view[2] = v.view.z.ToFloat32();
+
+    if (flip_quaternion) {
+        normquat = -normquat;
+    }
+}
+
+RasterizerAccelerated::RasterizerAccelerated(Memory::MemorySystem& memory_)
+    : memory{memory_}, regs{Pica::g_state.regs} {
+    uniform_block_data.lighting_lut_dirty.fill(true);
+}
+
+/**
+ * This is a helper function to resolve an issue when interpolating opposite quaternions. See below
+ * for a detailed description of this issue (yuriks):
+ *
+ * For any rotation, there are two quaternions Q, and -Q, that represent the same rotation. If you
+ * interpolate two quaternions that are opposite, instead of going from one rotation to another
+ * using the shortest path, you'll go around the longest path. You can test if two quaternions are
+ * opposite by checking if Dot(Q1, Q2) < 0. In that case, you can flip either of them, therefore
+ * making Dot(Q1, -Q2) positive.
+ *
+ * This solution corrects this issue per-vertex before passing the quaternions to OpenGL. This is
+ * correct for most cases but can still rotate around the long way sometimes. An implementation
+ * which did `lerp(lerp(Q1, Q2), Q3)` (with proper weighting), applying the dot product check
+ * between each step would work for those cases at the cost of being more complex to implement.
+ *
+ * Fortunately however, the 3DS hardware happens to also use this exact same logic to work around
+ * these issues, making this basic implementation actually more accurate to the hardware.
+ */
+static bool AreQuaternionsOpposite(Common::Vec4<Pica::float24> qa, Common::Vec4<Pica::float24> qb) {
+    Common::Vec4f a{qa.x.ToFloat32(), qa.y.ToFloat32(), qa.z.ToFloat32(), qa.w.ToFloat32()};
+    Common::Vec4f b{qb.x.ToFloat32(), qb.y.ToFloat32(), qb.z.ToFloat32(), qb.w.ToFloat32()};
+
+    return (Common::Dot(a, b) < 0.f);
+}
+
+void RasterizerAccelerated::AddTriangle(const Pica::Shader::OutputVertex& v0,
+                                        const Pica::Shader::OutputVertex& v1,
+                                        const Pica::Shader::OutputVertex& v2) {
+    vertex_batch.emplace_back(v0, false);
+    vertex_batch.emplace_back(v1, AreQuaternionsOpposite(v0.quat, v1.quat));
+    vertex_batch.emplace_back(v2, AreQuaternionsOpposite(v0.quat, v2.quat));
+}
+
+RasterizerAccelerated::VertexArrayInfo RasterizerAccelerated::AnalyzeVertexArray(
+    bool is_indexed, u32 stride_alignment) {
+    const auto& vertex_attributes = regs.pipeline.vertex_attributes;
+
+    u32 vertex_min;
+    u32 vertex_max;
+    if (is_indexed) {
+        const auto& index_info = regs.pipeline.index_array;
+        const PAddr address = vertex_attributes.GetPhysicalBaseAddress() + index_info.offset;
+        const u8* index_address_8 = memory.GetPhysicalPointer(address);
+        const u16* index_address_16 = reinterpret_cast<const u16*>(index_address_8);
+        const bool index_u16 = index_info.format != 0;
+
+        vertex_min = 0xFFFF;
+        vertex_max = 0;
+        const u32 size = regs.pipeline.num_vertices * (index_u16 ? 2 : 1);
+        FlushRegion(address, size);
+        for (u32 index = 0; index < regs.pipeline.num_vertices; ++index) {
+            const u32 vertex = index_u16 ? index_address_16[index] : index_address_8[index];
+            vertex_min = std::min(vertex_min, vertex);
+            vertex_max = std::max(vertex_max, vertex);
+        }
+    } else {
+        vertex_min = regs.pipeline.vertex_offset;
+        vertex_max = regs.pipeline.vertex_offset + regs.pipeline.num_vertices - 1;
+    }
+
+    const u32 vertex_num = vertex_max - vertex_min + 1;
+    u32 vs_input_size = 0;
+    for (const auto& loader : vertex_attributes.attribute_loaders) {
+        if (loader.component_count != 0) {
+            const u32 aligned_stride =
+                Common::AlignUp(static_cast<u32>(loader.byte_count), stride_alignment);
+            vs_input_size += Common::AlignUp(aligned_stride * vertex_num, 4);
+        }
+    }
+
+    return {vertex_min, vertex_max, vs_input_size};
+}
+
+void RasterizerAccelerated::SyncEntireState() {
+    // Sync renderer-specific fixed-function state
+    SyncFixedState();
+
+    // Sync uniforms
+    SyncClipCoef();
+    SyncDepthScale();
+    SyncDepthOffset();
+    SyncAlphaTest();
+    SyncCombinerColor();
+    auto& tev_stages = regs.texturing.GetTevStages();
+    for (std::size_t index = 0; index < tev_stages.size(); ++index) {
+        SyncTevConstColor(index, tev_stages[index]);
+    }
+
+    SyncGlobalAmbient();
+    for (unsigned light_index = 0; light_index < 8; light_index++) {
+        SyncLightSpecular0(light_index);
+        SyncLightSpecular1(light_index);
+        SyncLightDiffuse(light_index);
+        SyncLightAmbient(light_index);
+        SyncLightPosition(light_index);
+        SyncLightDistanceAttenuationBias(light_index);
+        SyncLightDistanceAttenuationScale(light_index);
+    }
+
+    SyncFogColor();
+    SyncProcTexNoise();
+    SyncProcTexBias();
+    SyncShadowBias();
+    SyncShadowTextureBias();
+
+    for (unsigned tex_index = 0; tex_index < 3; tex_index++) {
+        SyncTextureLodBias(tex_index);
+    }
+}
+
+void RasterizerAccelerated::NotifyPicaRegisterChanged(u32 id) {
+    switch (id) {
+    // Depth modifiers
+    case PICA_REG_INDEX(rasterizer.viewport_depth_range):
+        SyncDepthScale();
+        break;
+    case PICA_REG_INDEX(rasterizer.viewport_depth_near_plane):
+        SyncDepthOffset();
+        break;
+
+    // Depth buffering
+    case PICA_REG_INDEX(rasterizer.depthmap_enable):
+        shader_dirty = true;
+        break;
+
+    // Shadow texture
+    case PICA_REG_INDEX(texturing.shadow):
+        SyncShadowTextureBias();
+        break;
+
+    // Fog state
+    case PICA_REG_INDEX(texturing.fog_color):
+        SyncFogColor();
+        break;
+    case PICA_REG_INDEX(texturing.fog_lut_data[0]):
+    case PICA_REG_INDEX(texturing.fog_lut_data[1]):
+    case PICA_REG_INDEX(texturing.fog_lut_data[2]):
+    case PICA_REG_INDEX(texturing.fog_lut_data[3]):
+    case PICA_REG_INDEX(texturing.fog_lut_data[4]):
+    case PICA_REG_INDEX(texturing.fog_lut_data[5]):
+    case PICA_REG_INDEX(texturing.fog_lut_data[6]):
+    case PICA_REG_INDEX(texturing.fog_lut_data[7]):
+        uniform_block_data.fog_lut_dirty = true;
+        break;
+
+    // ProcTex state
+    case PICA_REG_INDEX(texturing.proctex):
+    case PICA_REG_INDEX(texturing.proctex_lut):
+    case PICA_REG_INDEX(texturing.proctex_lut_offset):
+        SyncProcTexBias();
+        shader_dirty = true;
+        break;
+
+    case PICA_REG_INDEX(texturing.proctex_noise_u):
+    case PICA_REG_INDEX(texturing.proctex_noise_v):
+    case PICA_REG_INDEX(texturing.proctex_noise_frequency):
+        SyncProcTexNoise();
+        break;
+
+    case PICA_REG_INDEX(texturing.proctex_lut_data[0]):
+    case PICA_REG_INDEX(texturing.proctex_lut_data[1]):
+    case PICA_REG_INDEX(texturing.proctex_lut_data[2]):
+    case PICA_REG_INDEX(texturing.proctex_lut_data[3]):
+    case PICA_REG_INDEX(texturing.proctex_lut_data[4]):
+    case PICA_REG_INDEX(texturing.proctex_lut_data[5]):
+    case PICA_REG_INDEX(texturing.proctex_lut_data[6]):
+    case PICA_REG_INDEX(texturing.proctex_lut_data[7]):
+        using Pica::TexturingRegs;
+        switch (regs.texturing.proctex_lut_config.ref_table.Value()) {
+        case TexturingRegs::ProcTexLutTable::Noise:
+            uniform_block_data.proctex_noise_lut_dirty = true;
+            break;
+        case TexturingRegs::ProcTexLutTable::ColorMap:
+            uniform_block_data.proctex_color_map_dirty = true;
+            break;
+        case TexturingRegs::ProcTexLutTable::AlphaMap:
+            uniform_block_data.proctex_alpha_map_dirty = true;
+            break;
+        case TexturingRegs::ProcTexLutTable::Color:
+            uniform_block_data.proctex_lut_dirty = true;
+            break;
+        case TexturingRegs::ProcTexLutTable::ColorDiff:
+            uniform_block_data.proctex_diff_lut_dirty = true;
+            break;
+        }
+        break;
+
+    // Alpha test
+    case PICA_REG_INDEX(framebuffer.output_merger.alpha_test):
+        SyncAlphaTest();
+        shader_dirty = true;
+        break;
+
+    case PICA_REG_INDEX(framebuffer.shadow):
+        SyncShadowBias();
+        break;
+
+    // Scissor test
+    case PICA_REG_INDEX(rasterizer.scissor_test.mode):
+        shader_dirty = true;
+        break;
+
+    case PICA_REG_INDEX(texturing.main_config):
+        shader_dirty = true;
+        break;
+
+    // Texture 0 type
+    case PICA_REG_INDEX(texturing.texture0.type):
+        shader_dirty = true;
+        break;
+
+    // TEV stages
+    // (This also syncs fog_mode and fog_flip which are part of tev_combiner_buffer_input)
+    case PICA_REG_INDEX(texturing.tev_stage0.color_source1):
+    case PICA_REG_INDEX(texturing.tev_stage0.color_modifier1):
+    case PICA_REG_INDEX(texturing.tev_stage0.color_op):
+    case PICA_REG_INDEX(texturing.tev_stage0.color_scale):
+    case PICA_REG_INDEX(texturing.tev_stage1.color_source1):
+    case PICA_REG_INDEX(texturing.tev_stage1.color_modifier1):
+    case PICA_REG_INDEX(texturing.tev_stage1.color_op):
+    case PICA_REG_INDEX(texturing.tev_stage1.color_scale):
+    case PICA_REG_INDEX(texturing.tev_stage2.color_source1):
+    case PICA_REG_INDEX(texturing.tev_stage2.color_modifier1):
+    case PICA_REG_INDEX(texturing.tev_stage2.color_op):
+    case PICA_REG_INDEX(texturing.tev_stage2.color_scale):
+    case PICA_REG_INDEX(texturing.tev_stage3.color_source1):
+    case PICA_REG_INDEX(texturing.tev_stage3.color_modifier1):
+    case PICA_REG_INDEX(texturing.tev_stage3.color_op):
+    case PICA_REG_INDEX(texturing.tev_stage3.color_scale):
+    case PICA_REG_INDEX(texturing.tev_stage4.color_source1):
+    case PICA_REG_INDEX(texturing.tev_stage4.color_modifier1):
+    case PICA_REG_INDEX(texturing.tev_stage4.color_op):
+    case PICA_REG_INDEX(texturing.tev_stage4.color_scale):
+    case PICA_REG_INDEX(texturing.tev_stage5.color_source1):
+    case PICA_REG_INDEX(texturing.tev_stage5.color_modifier1):
+    case PICA_REG_INDEX(texturing.tev_stage5.color_op):
+    case PICA_REG_INDEX(texturing.tev_stage5.color_scale):
+    case PICA_REG_INDEX(texturing.tev_combiner_buffer_input):
+        shader_dirty = true;
+        break;
+    case PICA_REG_INDEX(texturing.tev_stage0.const_r):
+        SyncTevConstColor(0, regs.texturing.tev_stage0);
+        break;
+    case PICA_REG_INDEX(texturing.tev_stage1.const_r):
+        SyncTevConstColor(1, regs.texturing.tev_stage1);
+        break;
+    case PICA_REG_INDEX(texturing.tev_stage2.const_r):
+        SyncTevConstColor(2, regs.texturing.tev_stage2);
+        break;
+    case PICA_REG_INDEX(texturing.tev_stage3.const_r):
+        SyncTevConstColor(3, regs.texturing.tev_stage3);
+        break;
+    case PICA_REG_INDEX(texturing.tev_stage4.const_r):
+        SyncTevConstColor(4, regs.texturing.tev_stage4);
+        break;
+    case PICA_REG_INDEX(texturing.tev_stage5.const_r):
+        SyncTevConstColor(5, regs.texturing.tev_stage5);
+        break;
+
+    // TEV combiner buffer color
+    case PICA_REG_INDEX(texturing.tev_combiner_buffer_color):
+        SyncCombinerColor();
+        break;
+
+    // Fragment lighting switches
+    case PICA_REG_INDEX(lighting.disable):
+    case PICA_REG_INDEX(lighting.max_light_index):
+    case PICA_REG_INDEX(lighting.config0):
+    case PICA_REG_INDEX(lighting.config1):
+    case PICA_REG_INDEX(lighting.abs_lut_input):
+    case PICA_REG_INDEX(lighting.lut_input):
+    case PICA_REG_INDEX(lighting.lut_scale):
+    case PICA_REG_INDEX(lighting.light_enable):
+        break;
+
+    // Fragment lighting specular 0 color
+    case PICA_REG_INDEX(lighting.light[0].specular_0):
+        SyncLightSpecular0(0);
+        break;
+    case PICA_REG_INDEX(lighting.light[1].specular_0):
+        SyncLightSpecular0(1);
+        break;
+    case PICA_REG_INDEX(lighting.light[2].specular_0):
+        SyncLightSpecular0(2);
+        break;
+    case PICA_REG_INDEX(lighting.light[3].specular_0):
+        SyncLightSpecular0(3);
+        break;
+    case PICA_REG_INDEX(lighting.light[4].specular_0):
+        SyncLightSpecular0(4);
+        break;
+    case PICA_REG_INDEX(lighting.light[5].specular_0):
+        SyncLightSpecular0(5);
+        break;
+    case PICA_REG_INDEX(lighting.light[6].specular_0):
+        SyncLightSpecular0(6);
+        break;
+    case PICA_REG_INDEX(lighting.light[7].specular_0):
+        SyncLightSpecular0(7);
+        break;
+
+    // Fragment lighting specular 1 color
+    case PICA_REG_INDEX(lighting.light[0].specular_1):
+        SyncLightSpecular1(0);
+        break;
+    case PICA_REG_INDEX(lighting.light[1].specular_1):
+        SyncLightSpecular1(1);
+        break;
+    case PICA_REG_INDEX(lighting.light[2].specular_1):
+        SyncLightSpecular1(2);
+        break;
+    case PICA_REG_INDEX(lighting.light[3].specular_1):
+        SyncLightSpecular1(3);
+        break;
+    case PICA_REG_INDEX(lighting.light[4].specular_1):
+        SyncLightSpecular1(4);
+        break;
+    case PICA_REG_INDEX(lighting.light[5].specular_1):
+        SyncLightSpecular1(5);
+        break;
+    case PICA_REG_INDEX(lighting.light[6].specular_1):
+        SyncLightSpecular1(6);
+        break;
+    case PICA_REG_INDEX(lighting.light[7].specular_1):
+        SyncLightSpecular1(7);
+        break;
+
+    // Fragment lighting diffuse color
+    case PICA_REG_INDEX(lighting.light[0].diffuse):
+        SyncLightDiffuse(0);
+        break;
+    case PICA_REG_INDEX(lighting.light[1].diffuse):
+        SyncLightDiffuse(1);
+        break;
+    case PICA_REG_INDEX(lighting.light[2].diffuse):
+        SyncLightDiffuse(2);
+        break;
+    case PICA_REG_INDEX(lighting.light[3].diffuse):
+        SyncLightDiffuse(3);
+        break;
+    case PICA_REG_INDEX(lighting.light[4].diffuse):
+        SyncLightDiffuse(4);
+        break;
+    case PICA_REG_INDEX(lighting.light[5].diffuse):
+        SyncLightDiffuse(5);
+        break;
+    case PICA_REG_INDEX(lighting.light[6].diffuse):
+        SyncLightDiffuse(6);
+        break;
+    case PICA_REG_INDEX(lighting.light[7].diffuse):
+        SyncLightDiffuse(7);
+        break;
+
+    // Fragment lighting ambient color
+    case PICA_REG_INDEX(lighting.light[0].ambient):
+        SyncLightAmbient(0);
+        break;
+    case PICA_REG_INDEX(lighting.light[1].ambient):
+        SyncLightAmbient(1);
+        break;
+    case PICA_REG_INDEX(lighting.light[2].ambient):
+        SyncLightAmbient(2);
+        break;
+    case PICA_REG_INDEX(lighting.light[3].ambient):
+        SyncLightAmbient(3);
+        break;
+    case PICA_REG_INDEX(lighting.light[4].ambient):
+        SyncLightAmbient(4);
+        break;
+    case PICA_REG_INDEX(lighting.light[5].ambient):
+        SyncLightAmbient(5);
+        break;
+    case PICA_REG_INDEX(lighting.light[6].ambient):
+        SyncLightAmbient(6);
+        break;
+    case PICA_REG_INDEX(lighting.light[7].ambient):
+        SyncLightAmbient(7);
+        break;
+
+    // Fragment lighting position
+    case PICA_REG_INDEX(lighting.light[0].x):
+    case PICA_REG_INDEX(lighting.light[0].z):
+        SyncLightPosition(0);
+        break;
+    case PICA_REG_INDEX(lighting.light[1].x):
+    case PICA_REG_INDEX(lighting.light[1].z):
+        SyncLightPosition(1);
+        break;
+    case PICA_REG_INDEX(lighting.light[2].x):
+    case PICA_REG_INDEX(lighting.light[2].z):
+        SyncLightPosition(2);
+        break;
+    case PICA_REG_INDEX(lighting.light[3].x):
+    case PICA_REG_INDEX(lighting.light[3].z):
+        SyncLightPosition(3);
+        break;
+    case PICA_REG_INDEX(lighting.light[4].x):
+    case PICA_REG_INDEX(lighting.light[4].z):
+        SyncLightPosition(4);
+        break;
+    case PICA_REG_INDEX(lighting.light[5].x):
+    case PICA_REG_INDEX(lighting.light[5].z):
+        SyncLightPosition(5);
+        break;
+    case PICA_REG_INDEX(lighting.light[6].x):
+    case PICA_REG_INDEX(lighting.light[6].z):
+        SyncLightPosition(6);
+        break;
+    case PICA_REG_INDEX(lighting.light[7].x):
+    case PICA_REG_INDEX(lighting.light[7].z):
+        SyncLightPosition(7);
+        break;
+
+    // Fragment spot lighting direction
+    case PICA_REG_INDEX(lighting.light[0].spot_x):
+    case PICA_REG_INDEX(lighting.light[0].spot_z):
+        SyncLightSpotDirection(0);
+        break;
+    case PICA_REG_INDEX(lighting.light[1].spot_x):
+    case PICA_REG_INDEX(lighting.light[1].spot_z):
+        SyncLightSpotDirection(1);
+        break;
+    case PICA_REG_INDEX(lighting.light[2].spot_x):
+    case PICA_REG_INDEX(lighting.light[2].spot_z):
+        SyncLightSpotDirection(2);
+        break;
+    case PICA_REG_INDEX(lighting.light[3].spot_x):
+    case PICA_REG_INDEX(lighting.light[3].spot_z):
+        SyncLightSpotDirection(3);
+        break;
+    case PICA_REG_INDEX(lighting.light[4].spot_x):
+    case PICA_REG_INDEX(lighting.light[4].spot_z):
+        SyncLightSpotDirection(4);
+        break;
+    case PICA_REG_INDEX(lighting.light[5].spot_x):
+    case PICA_REG_INDEX(lighting.light[5].spot_z):
+        SyncLightSpotDirection(5);
+        break;
+    case PICA_REG_INDEX(lighting.light[6].spot_x):
+    case PICA_REG_INDEX(lighting.light[6].spot_z):
+        SyncLightSpotDirection(6);
+        break;
+    case PICA_REG_INDEX(lighting.light[7].spot_x):
+    case PICA_REG_INDEX(lighting.light[7].spot_z):
+        SyncLightSpotDirection(7);
+        break;
+
+    // Fragment lighting light source config
+    case PICA_REG_INDEX(lighting.light[0].config):
+    case PICA_REG_INDEX(lighting.light[1].config):
+    case PICA_REG_INDEX(lighting.light[2].config):
+    case PICA_REG_INDEX(lighting.light[3].config):
+    case PICA_REG_INDEX(lighting.light[4].config):
+    case PICA_REG_INDEX(lighting.light[5].config):
+    case PICA_REG_INDEX(lighting.light[6].config):
+    case PICA_REG_INDEX(lighting.light[7].config):
+        shader_dirty = true;
+        break;
+
+    // Fragment lighting distance attenuation bias
+    case PICA_REG_INDEX(lighting.light[0].dist_atten_bias):
+        SyncLightDistanceAttenuationBias(0);
+        break;
+    case PICA_REG_INDEX(lighting.light[1].dist_atten_bias):
+        SyncLightDistanceAttenuationBias(1);
+        break;
+    case PICA_REG_INDEX(lighting.light[2].dist_atten_bias):
+        SyncLightDistanceAttenuationBias(2);
+        break;
+    case PICA_REG_INDEX(lighting.light[3].dist_atten_bias):
+        SyncLightDistanceAttenuationBias(3);
+        break;
+    case PICA_REG_INDEX(lighting.light[4].dist_atten_bias):
+        SyncLightDistanceAttenuationBias(4);
+        break;
+    case PICA_REG_INDEX(lighting.light[5].dist_atten_bias):
+        SyncLightDistanceAttenuationBias(5);
+        break;
+    case PICA_REG_INDEX(lighting.light[6].dist_atten_bias):
+        SyncLightDistanceAttenuationBias(6);
+        break;
+    case PICA_REG_INDEX(lighting.light[7].dist_atten_bias):
+        SyncLightDistanceAttenuationBias(7);
+        break;
+
+    // Fragment lighting distance attenuation scale
+    case PICA_REG_INDEX(lighting.light[0].dist_atten_scale):
+        SyncLightDistanceAttenuationScale(0);
+        break;
+    case PICA_REG_INDEX(lighting.light[1].dist_atten_scale):
+        SyncLightDistanceAttenuationScale(1);
+        break;
+    case PICA_REG_INDEX(lighting.light[2].dist_atten_scale):
+        SyncLightDistanceAttenuationScale(2);
+        break;
+    case PICA_REG_INDEX(lighting.light[3].dist_atten_scale):
+        SyncLightDistanceAttenuationScale(3);
+        break;
+    case PICA_REG_INDEX(lighting.light[4].dist_atten_scale):
+        SyncLightDistanceAttenuationScale(4);
+        break;
+    case PICA_REG_INDEX(lighting.light[5].dist_atten_scale):
+        SyncLightDistanceAttenuationScale(5);
+        break;
+    case PICA_REG_INDEX(lighting.light[6].dist_atten_scale):
+        SyncLightDistanceAttenuationScale(6);
+        break;
+    case PICA_REG_INDEX(lighting.light[7].dist_atten_scale):
+        SyncLightDistanceAttenuationScale(7);
+        break;
+
+    // Fragment lighting global ambient color (emission + ambient * ambient)
+    case PICA_REG_INDEX(lighting.global_ambient):
+        SyncGlobalAmbient();
+        break;
+
+    // Fragment lighting lookup tables
+    case PICA_REG_INDEX(lighting.lut_data[0]):
+    case PICA_REG_INDEX(lighting.lut_data[1]):
+    case PICA_REG_INDEX(lighting.lut_data[2]):
+    case PICA_REG_INDEX(lighting.lut_data[3]):
+    case PICA_REG_INDEX(lighting.lut_data[4]):
+    case PICA_REG_INDEX(lighting.lut_data[5]):
+    case PICA_REG_INDEX(lighting.lut_data[6]):
+    case PICA_REG_INDEX(lighting.lut_data[7]): {
+        const auto& lut_config = regs.lighting.lut_config;
+        uniform_block_data.lighting_lut_dirty[lut_config.type] = true;
+        uniform_block_data.lighting_lut_dirty_any = true;
+        break;
+    }
+
+    // Texture LOD biases
+    case PICA_REG_INDEX(texturing.texture0.lod.bias):
+        SyncTextureLodBias(0);
+        break;
+    case PICA_REG_INDEX(texturing.texture1.lod.bias):
+        SyncTextureLodBias(1);
+        break;
+    case PICA_REG_INDEX(texturing.texture2.lod.bias):
+        SyncTextureLodBias(2);
+        break;
+
+    // Clipping plane
+    case PICA_REG_INDEX(rasterizer.clip_coef[0]):
+    case PICA_REG_INDEX(rasterizer.clip_coef[1]):
+    case PICA_REG_INDEX(rasterizer.clip_coef[2]):
+    case PICA_REG_INDEX(rasterizer.clip_coef[3]):
+        SyncClipCoef();
+        break;
+
+    default:
+        // Forward registers that map to fixed function API features to the video backend
+        NotifyFixedFunctionPicaRegisterChanged(id);
+    }
+}
+
+void RasterizerAccelerated::SyncDepthScale() {
+    float depth_scale = Pica::float24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32();
+
+    if (depth_scale != uniform_block_data.data.depth_scale) {
+        uniform_block_data.data.depth_scale = depth_scale;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncDepthOffset() {
+    float depth_offset =
+        Pica::float24::FromRaw(regs.rasterizer.viewport_depth_near_plane).ToFloat32();
+
+    if (depth_offset != uniform_block_data.data.depth_offset) {
+        uniform_block_data.data.depth_offset = depth_offset;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncFogColor() {
+    const auto& fog_color_regs = regs.texturing.fog_color;
+    const Common::Vec3f fog_color = {
+        fog_color_regs.r.Value() / 255.0f,
+        fog_color_regs.g.Value() / 255.0f,
+        fog_color_regs.b.Value() / 255.0f,
+    };
+
+    if (fog_color != uniform_block_data.data.fog_color) {
+        uniform_block_data.data.fog_color = fog_color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncProcTexNoise() {
+    const Common::Vec2f proctex_noise_f = {
+        Pica::float16::FromRaw(regs.texturing.proctex_noise_frequency.u).ToFloat32(),
+        Pica::float16::FromRaw(regs.texturing.proctex_noise_frequency.v).ToFloat32(),
+    };
+    const Common::Vec2f proctex_noise_a = {
+        regs.texturing.proctex_noise_u.amplitude / 4095.0f,
+        regs.texturing.proctex_noise_v.amplitude / 4095.0f,
+    };
+    const Common::Vec2f proctex_noise_p = {
+        Pica::float16::FromRaw(regs.texturing.proctex_noise_u.phase).ToFloat32(),
+        Pica::float16::FromRaw(regs.texturing.proctex_noise_v.phase).ToFloat32(),
+    };
+
+    if (proctex_noise_f != uniform_block_data.data.proctex_noise_f ||
+        proctex_noise_a != uniform_block_data.data.proctex_noise_a ||
+        proctex_noise_p != uniform_block_data.data.proctex_noise_p) {
+        uniform_block_data.data.proctex_noise_f = proctex_noise_f;
+        uniform_block_data.data.proctex_noise_a = proctex_noise_a;
+        uniform_block_data.data.proctex_noise_p = proctex_noise_p;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncProcTexBias() {
+    const auto proctex_bias = Pica::float16::FromRaw(regs.texturing.proctex.bias_low |
+                                                     (regs.texturing.proctex_lut.bias_high << 8))
+                                  .ToFloat32();
+    if (proctex_bias != uniform_block_data.data.proctex_bias) {
+        uniform_block_data.data.proctex_bias = proctex_bias;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncAlphaTest() {
+    if (regs.framebuffer.output_merger.alpha_test.ref != uniform_block_data.data.alphatest_ref) {
+        uniform_block_data.data.alphatest_ref = regs.framebuffer.output_merger.alpha_test.ref;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncCombinerColor() {
+    auto combiner_color = ColorRGBA8(regs.texturing.tev_combiner_buffer_color.raw);
+    if (combiner_color != uniform_block_data.data.tev_combiner_buffer_color) {
+        uniform_block_data.data.tev_combiner_buffer_color = combiner_color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncTevConstColor(
+    std::size_t stage_index, const Pica::TexturingRegs::TevStageConfig& tev_stage) {
+    const auto const_color = ColorRGBA8(tev_stage.const_color);
+
+    if (const_color == uniform_block_data.data.const_color[stage_index]) {
+        return;
+    }
+
+    uniform_block_data.data.const_color[stage_index] = const_color;
+    uniform_block_data.dirty = true;
+}
+
+void RasterizerAccelerated::SyncGlobalAmbient() {
+    auto color = LightColor(regs.lighting.global_ambient);
+    if (color != uniform_block_data.data.lighting_global_ambient) {
+        uniform_block_data.data.lighting_global_ambient = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncLightSpecular0(int light_index) {
+    auto color = LightColor(regs.lighting.light[light_index].specular_0);
+    if (color != uniform_block_data.data.light_src[light_index].specular_0) {
+        uniform_block_data.data.light_src[light_index].specular_0 = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncLightSpecular1(int light_index) {
+    auto color = LightColor(regs.lighting.light[light_index].specular_1);
+    if (color != uniform_block_data.data.light_src[light_index].specular_1) {
+        uniform_block_data.data.light_src[light_index].specular_1 = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncLightDiffuse(int light_index) {
+    auto color = LightColor(regs.lighting.light[light_index].diffuse);
+    if (color != uniform_block_data.data.light_src[light_index].diffuse) {
+        uniform_block_data.data.light_src[light_index].diffuse = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncLightAmbient(int light_index) {
+    auto color = LightColor(regs.lighting.light[light_index].ambient);
+    if (color != uniform_block_data.data.light_src[light_index].ambient) {
+        uniform_block_data.data.light_src[light_index].ambient = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncLightPosition(int light_index) {
+    const Common::Vec3f position = {
+        Pica::float16::FromRaw(regs.lighting.light[light_index].x).ToFloat32(),
+        Pica::float16::FromRaw(regs.lighting.light[light_index].y).ToFloat32(),
+        Pica::float16::FromRaw(regs.lighting.light[light_index].z).ToFloat32(),
+    };
+
+    if (position != uniform_block_data.data.light_src[light_index].position) {
+        uniform_block_data.data.light_src[light_index].position = position;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncLightSpotDirection(int light_index) {
+    const auto& light = regs.lighting.light[light_index];
+    const auto spot_direction =
+        Common::Vec3f{light.spot_x / 2047.0f, light.spot_y / 2047.0f, light.spot_z / 2047.0f};
+
+    if (spot_direction != uniform_block_data.data.light_src[light_index].spot_direction) {
+        uniform_block_data.data.light_src[light_index].spot_direction = spot_direction;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncLightDistanceAttenuationBias(int light_index) {
+    float dist_atten_bias =
+        Pica::float20::FromRaw(regs.lighting.light[light_index].dist_atten_bias).ToFloat32();
+
+    if (dist_atten_bias != uniform_block_data.data.light_src[light_index].dist_atten_bias) {
+        uniform_block_data.data.light_src[light_index].dist_atten_bias = dist_atten_bias;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncLightDistanceAttenuationScale(int light_index) {
+    float dist_atten_scale =
+        Pica::float20::FromRaw(regs.lighting.light[light_index].dist_atten_scale).ToFloat32();
+
+    if (dist_atten_scale != uniform_block_data.data.light_src[light_index].dist_atten_scale) {
+        uniform_block_data.data.light_src[light_index].dist_atten_scale = dist_atten_scale;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncShadowBias() {
+    const auto& shadow = regs.framebuffer.shadow;
+    float constant = Pica::float16::FromRaw(shadow.constant).ToFloat32();
+    float linear = Pica::float16::FromRaw(shadow.linear).ToFloat32();
+
+    if (constant != uniform_block_data.data.shadow_bias_constant ||
+        linear != uniform_block_data.data.shadow_bias_linear) {
+        uniform_block_data.data.shadow_bias_constant = constant;
+        uniform_block_data.data.shadow_bias_linear = linear;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncShadowTextureBias() {
+    int bias = regs.texturing.shadow.bias << 1;
+    if (bias != uniform_block_data.data.shadow_texture_bias) {
+        uniform_block_data.data.shadow_texture_bias = bias;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncTextureLodBias(int tex_index) {
+    const auto pica_textures = regs.texturing.GetTextures();
+    const float bias = pica_textures[tex_index].config.lod.bias / 256.0f;
+    if (bias != uniform_block_data.data.tex_lod_bias[tex_index]) {
+        uniform_block_data.data.tex_lod_bias[tex_index] = bias;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerAccelerated::SyncClipCoef() {
+    const auto raw_clip_coef = regs.rasterizer.GetClipCoef();
+    const Common::Vec4f new_clip_coef = {raw_clip_coef.x.ToFloat32(), raw_clip_coef.y.ToFloat32(),
+                                         raw_clip_coef.z.ToFloat32(), raw_clip_coef.w.ToFloat32()};
+    if (new_clip_coef != uniform_block_data.data.clip_coef) {
+        uniform_block_data.data.clip_coef = new_clip_coef;
+        uniform_block_data.dirty = true;
+    }
+}
+
+} // namespace VideoCore

src/video_core/rasterizer_accelerated.h

@@ -0,0 +1,159 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/vector_math.h"
+#include "video_core/rasterizer_interface.h"
+#include "video_core/regs_texturing.h"
+#include "video_core/shader/shader_uniforms.h"
+
+namespace Memory {
+class MemorySystem;
+}
+
+namespace Pica {
+struct Regs;
+}
+
+namespace VideoCore {
+
+class RasterizerAccelerated : public RasterizerInterface {
+public:
+    RasterizerAccelerated(Memory::MemorySystem& memory);
+    virtual ~RasterizerAccelerated() = default;
+
+    void AddTriangle(const Pica::Shader::OutputVertex& v0, const Pica::Shader::OutputVertex& v1,
+                     const Pica::Shader::OutputVertex& v2) override;
+    void NotifyPicaRegisterChanged(u32 id) override;
+    void SyncEntireState() override;
+
+protected:
+    /// Sync fixed-function pipeline state
+    virtual void SyncFixedState() = 0;
+
+    /// Notifies that a fixed function PICA register changed to the video backend
+    virtual void NotifyFixedFunctionPicaRegisterChanged(u32 id) = 0;
+
+    /// Syncs the depth scale to match the PICA register
+    void SyncDepthScale();
+
+    /// Syncs the depth offset to match the PICA register
+    void SyncDepthOffset();
+
+    /// Syncs the fog states to match the PICA register
+    void SyncFogColor();
+
+    /// Sync the procedural texture noise configuration to match the PICA register
+    void SyncProcTexNoise();
+
+    /// Sync the procedural texture bias configuration to match the PICA register
+    void SyncProcTexBias();
+
+    /// Syncs the alpha test states to match the PICA register
+    void SyncAlphaTest();
+
+    /// Syncs the TEV combiner color buffer to match the PICA register
+    void SyncCombinerColor();
+
+    /// Syncs the TEV constant color to match the PICA register
+    void SyncTevConstColor(std::size_t tev_index,
+                           const Pica::TexturingRegs::TevStageConfig& tev_stage);
+
+    /// Syncs the lighting global ambient color to match the PICA register
+    void SyncGlobalAmbient();
+
+    /// Syncs the specified light's specular 0 color to match the PICA register
+    void SyncLightSpecular0(int light_index);
+
+    /// Syncs the specified light's specular 1 color to match the PICA register
+    void SyncLightSpecular1(int light_index);
+
+    /// Syncs the specified light's diffuse color to match the PICA register
+    void SyncLightDiffuse(int light_index);
+
+    /// Syncs the specified light's ambient color to match the PICA register
+    void SyncLightAmbient(int light_index);
+
+    /// Syncs the specified light's position to match the PICA register
+    void SyncLightPosition(int light_index);
+
+    /// Syncs the specified spot light direcition to match the PICA register
+    void SyncLightSpotDirection(int light_index);
+
+    /// Syncs the specified light's distance attenuation bias to match the PICA register
+    void SyncLightDistanceAttenuationBias(int light_index);
+
+    /// Syncs the specified light's distance attenuation scale to match the PICA register
+    void SyncLightDistanceAttenuationScale(int light_index);
+
+    /// Syncs the shadow rendering bias to match the PICA register
+    void SyncShadowBias();
+
+    /// Syncs the shadow texture bias to match the PICA register
+    void SyncShadowTextureBias();
+
+    /// Syncs the texture LOD bias to match the PICA register
+    void SyncTextureLodBias(int tex_index);
+
+    /// Syncs the clip coefficients to match the PICA register
+    void SyncClipCoef();
+
+protected:
+    /// Structure that keeps tracks of the uniform state
+    struct UniformBlockData {
+        Pica::Shader::UniformData data{};
+        std::array<bool, Pica::LightingRegs::NumLightingSampler> lighting_lut_dirty{};
+        bool lighting_lut_dirty_any = true;
+        bool fog_lut_dirty = true;
+        bool proctex_noise_lut_dirty = true;
+        bool proctex_color_map_dirty = true;
+        bool proctex_alpha_map_dirty = true;
+        bool proctex_lut_dirty = true;
+        bool proctex_diff_lut_dirty = true;
+        bool dirty = true;
+    };
+
+    /// Structure that the hardware rendered vertices are composed of
+    struct HardwareVertex {
+        HardwareVertex() = default;
+        HardwareVertex(const Pica::Shader::OutputVertex& v, bool flip_quaternion);
+
+        Common::Vec4f position;
+        Common::Vec4f color;
+        Common::Vec2f tex_coord0;
+        Common::Vec2f tex_coord1;
+        Common::Vec2f tex_coord2;
+        float tex_coord0_w;
+        Common::Vec4f normquat;
+        Common::Vec3f view;
+    };
+
+    struct VertexArrayInfo {
+        u32 vs_input_index_min;
+        u32 vs_input_index_max;
+        u32 vs_input_size;
+    };
+
+    /// Retrieve the range and the size of the input vertex
+    VertexArrayInfo AnalyzeVertexArray(bool is_indexed, u32 stride_alignment = 1);
+
+protected:
+    Memory::MemorySystem& memory;
+    Pica::Regs& regs;
+
+    std::vector<HardwareVertex> vertex_batch;
+    bool shader_dirty = true;
+
+    UniformBlockData uniform_block_data{};
+    std::array<std::array<Common::Vec2f, 256>, Pica::LightingRegs::NumLightingSampler>
+        lighting_lut_data{};
+    std::array<Common::Vec2f, 128> fog_lut_data{};
+    std::array<Common::Vec2f, 128> proctex_noise_lut_data{};
+    std::array<Common::Vec2f, 128> proctex_color_map_data{};
+    std::array<Common::Vec2f, 128> proctex_alpha_map_data{};
+    std::array<Common::Vec4f, 256> proctex_lut_data{};
+    std::array<Common::Vec4f, 256> proctex_diff_lut_data{};
+};
+} // namespace VideoCore

src/video_core/renderer_base.cpp

@@ -2,15 +2,17 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <memory>
+#include "core/core.h"
 #include "core/frontend/emu_window.h"
+#include "core/tracer/recorder.h"
+#include "video_core/debug_utils/debug_utils.h"
 #include "video_core/renderer_base.h"
-#include "video_core/renderer_opengl/gl_rasterizer.h"
-#include "video_core/swrasterizer/swrasterizer.h"
-#include "video_core/video_core.h"
 
-RendererBase::RendererBase(Frontend::EmuWindow& window, Frontend::EmuWindow* secondary_window_)
+namespace VideoCore {
-    : render_window{window}, secondary_window{secondary_window_} {}
+
+RendererBase::RendererBase(Core::System& system_, Frontend::EmuWindow& window,
+                           Frontend::EmuWindow* secondary_window_)
+    : system{system_}, render_window{window}, secondary_window{secondary_window_} {}
 
 RendererBase::~RendererBase() = default;
 
@@ -25,19 +27,35 @@ void RendererBase::UpdateCurrentFramebufferLayout(bool is_portrait_mode) {
     }
 }
 
-void RendererBase::RefreshRasterizerSetting() {
+void RendererBase::EndFrame() {
-    bool hw_renderer_enabled = VideoCore::g_hw_renderer_enabled;
+    current_frame++;
-    if (rasterizer == nullptr || opengl_rasterizer_active != hw_renderer_enabled) {
-        opengl_rasterizer_active = hw_renderer_enabled;
 
-        if (hw_renderer_enabled) {
+    system.perf_stats->EndSystemFrame();
-            rasterizer = std::make_unique<OpenGL::RasterizerOpenGL>(render_window);
+
-        } else {
+    render_window.PollEvents();
-            rasterizer = std::make_unique<VideoCore::SWRasterizer>();
+
-        }
+    system.frame_limiter.DoFrameLimiting(system.CoreTiming().GetGlobalTimeUs());
+    system.perf_stats->BeginSystemFrame();
+
+    if (Pica::g_debug_context && Pica::g_debug_context->recorder) {
+        Pica::g_debug_context->recorder->FrameFinished();
     }
 }
 
-void RendererBase::Sync() {
+bool RendererBase::IsScreenshotPending() const {
-    rasterizer->SyncEntireState();
+    return renderer_settings.screenshot_requested;
 }
+
+void RendererBase::RequestScreenshot(void* data, std::function<void()> callback,
+                                     const Layout::FramebufferLayout& layout) {
+    if (renderer_settings.screenshot_requested) {
+        LOG_ERROR(Render, "A screenshot is already requested or in progress, ignoring the request");
+        return;
+    }
+    renderer_settings.screenshot_bits = data;
+    renderer_settings.screenshot_complete_callback = callback;
+    renderer_settings.screenshot_framebuffer_layout = layout;
+    renderer_settings.screenshot_requested = true;
+}
+
+} // namespace VideoCore

src/video_core/renderer_base.h

@@ -4,25 +4,36 @@
 
 #pragma once
 
-#include <memory>
 #include "common/common_types.h"
+#include "core/frontend/framebuffer_layout.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/video_core.h"
 
 namespace Frontend {
 class EmuWindow;
 }
 
+namespace Core {
+class System;
+}
+
+namespace VideoCore {
+
+struct RendererSettings {
+    // Screenshot
+    std::atomic_bool screenshot_requested{false};
+    void* screenshot_bits{};
+    std::function<void()> screenshot_complete_callback;
+    Layout::FramebufferLayout screenshot_framebuffer_layout;
+};
+
 class RendererBase : NonCopyable {
 public:
-    explicit RendererBase(Frontend::EmuWindow& window, Frontend::EmuWindow* secondary_window);
+    explicit RendererBase(Core::System& system, Frontend::EmuWindow& window,
+                          Frontend::EmuWindow* secondary_window);
     virtual ~RendererBase();
 
-    /// Initialize the renderer
+    /// Returns the rasterizer owned by the renderer
-    virtual VideoCore::ResultStatus Init() = 0;
+    virtual VideoCore::RasterizerInterface* Rasterizer() const = 0;
-
-    /// Shutdown the renderer
-    virtual void ShutDown() = 0;
 
     /// Finalize rendering the guest frame and draw into the presentation texture
     virtual void SwapBuffers() = 0;
@@ -35,27 +46,29 @@ public:
     }
 
     /// Prepares for video dumping (e.g. create necessary buffers, etc)
-    virtual void PrepareVideoDumping() = 0;
+    virtual void PrepareVideoDumping() {}
 
     /// Cleans up after video dumping is ended
-    virtual void CleanupVideoDumping() = 0;
+    virtual void CleanupVideoDumping() {}
+
+    /// Synchronizes fixed function renderer state
+    virtual void Sync() {}
 
     /// Updates the framebuffer layout of the contained render window handle.
     void UpdateCurrentFramebufferLayout(bool is_portrait_mode = {});
 
+    /// Ends the current frame
+    void EndFrame();
+
     // Getter/setter functions:
     // ------------------------
 
     f32 GetCurrentFPS() const {
-        return m_current_fps;
+        return current_fps;
     }
 
     int GetCurrentFrame() const {
-        return m_current_frame;
+        return current_frame;
-    }
-
-    VideoCore::RasterizerInterface* Rasterizer() const {
-        return rasterizer.get();
     }
 
     Frontend::EmuWindow& GetRenderWindow() {
@@ -66,16 +79,28 @@ public:
         return render_window;
     }
 
-    void RefreshRasterizerSetting();
+    [[nodiscard]] RendererSettings& Settings() {
-    void Sync();
+        return renderer_settings;
+    }
+
+    [[nodiscard]] const RendererSettings& Settings() const {
+        return renderer_settings;
+    }
+
+    /// Returns true if a screenshot is being processed
+    [[nodiscard]] bool IsScreenshotPending() const;
+
+    /// Request a screenshot of the next frame
+    void RequestScreenshot(void* data, std::function<void()> callback,
+                           const Layout::FramebufferLayout& layout);
 
 protected:
+    Core::System& system;
+    RendererSettings renderer_settings;
     Frontend::EmuWindow& render_window;    ///< Reference to the render window handle.
     Frontend::EmuWindow* secondary_window; ///< Reference to the secondary render window handle.
-    std::unique_ptr<VideoCore::RasterizerInterface> rasterizer;
+    f32 current_fps = 0.0f;                ///< Current framerate, should be set by the renderer
-    f32 m_current_fps = 0.0f; ///< Current framerate, should be set by the renderer
+    int current_frame = 0;                 ///< Current frame, should be set by the renderer
-    int m_current_frame = 0;  ///< Current frame, should be set by the renderer
-
-private:
-    bool opengl_rasterizer_active = false;
 };
+
+} // namespace VideoCore

src/video_core/renderer_opengl/frame_dumper_opengl.cpp

@@ -4,7 +4,6 @@
 
 #include <glad/glad.h>
 #include "core/frontend/emu_window.h"
-#include "core/frontend/scope_acquire_context.h"
 #include "video_core/renderer_opengl/frame_dumper_opengl.h"
 #include "video_core/renderer_opengl/renderer_opengl.h"
 
@@ -39,7 +38,7 @@ void FrameDumperOpenGL::StopDumping() {
 }
 
 void FrameDumperOpenGL::PresentLoop() {
-    Frontend::ScopeAcquireContext scope{*context};
+    const auto scope = context->Acquire();
     InitializeOpenGLObjects();
 
     const auto& layout = GetLayout();

src/video_core/renderer_opengl/gl_driver.cpp

@@ -0,0 +1,156 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <glad/glad.h>
+#include "common/assert.h"
+#include "common/settings.h"
+#include "core/telemetry_session.h"
+#include "video_core/renderer_opengl/gl_driver.h"
+
+namespace OpenGL {
+
+DECLARE_ENUM_FLAG_OPERATORS(DriverBug);
+
+inline std::string_view GetSource(GLenum source) {
+#define RET(s)                                                                                     \
+    case GL_DEBUG_SOURCE_##s:                                                                      \
+        return #s
+    switch (source) {
+        RET(API);
+        RET(WINDOW_SYSTEM);
+        RET(SHADER_COMPILER);
+        RET(THIRD_PARTY);
+        RET(APPLICATION);
+        RET(OTHER);
+    default:
+        UNREACHABLE();
+    }
+#undef RET
+
+    return std::string_view{};
+}
+
+inline std::string_view GetType(GLenum type) {
+#define RET(t)                                                                                     \
+    case GL_DEBUG_TYPE_##t:                                                                        \
+        return #t
+    switch (type) {
+        RET(ERROR);
+        RET(DEPRECATED_BEHAVIOR);
+        RET(UNDEFINED_BEHAVIOR);
+        RET(PORTABILITY);
+        RET(PERFORMANCE);
+        RET(OTHER);
+        RET(MARKER);
+    default:
+        UNREACHABLE();
+    }
+#undef RET
+
+    return std::string_view{};
+}
+
+static void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severity,
+                                  GLsizei length, const GLchar* message, const void* user_param) {
+    Log::Level level = Log::Level::Info;
+    switch (severity) {
+    case GL_DEBUG_SEVERITY_HIGH:
+        level = Log::Level::Critical;
+        break;
+    case GL_DEBUG_SEVERITY_MEDIUM:
+        level = Log::Level::Warning;
+        break;
+    case GL_DEBUG_SEVERITY_NOTIFICATION:
+    case GL_DEBUG_SEVERITY_LOW:
+        level = Log::Level::Debug;
+        break;
+    }
+
+    LOG_GENERIC(Log::Class::Render_OpenGL, level, "{} {} {}: {}", GetSource(source), GetType(type),
+                id, message);
+}
+
+Driver::Driver(Core::TelemetrySession& telemetry_session_) : telemetry_session{telemetry_session_} {
+    const bool enable_debug = Settings::values.renderer_debug.GetValue();
+    if (enable_debug) {
+        glEnable(GL_DEBUG_OUTPUT);
+        glDebugMessageCallback(DebugHandler, nullptr);
+    }
+
+    ReportDriverInfo();
+    DeduceVendor();
+    CheckExtensionSupport();
+    FindBugs();
+}
+
+Driver::~Driver() = default;
+
+bool Driver::HasBug(DriverBug bug) const {
+    return True(bugs & bug);
+}
+
+void Driver::ReportDriverInfo() {
+    // Report the context version and the vendor string
+    gl_version = std::string_view{reinterpret_cast<const char*>(glGetString(GL_VERSION))};
+    gpu_vendor = std::string_view{reinterpret_cast<const char*>(glGetString(GL_VENDOR))};
+    gpu_model = std::string_view{reinterpret_cast<const char*>(glGetString(GL_RENDERER))};
+
+    LOG_INFO(Render_OpenGL, "GL_VERSION: {}", gl_version);
+    LOG_INFO(Render_OpenGL, "GL_VENDOR: {}", gpu_vendor);
+    LOG_INFO(Render_OpenGL, "GL_RENDERER: {}", gpu_model);
+
+    // Add the information to the telemetry system
+    constexpr auto user_system = Common::Telemetry::FieldType::UserSystem;
+    telemetry_session.AddField(user_system, "GPU_Vendor", std::string{gpu_vendor});
+    telemetry_session.AddField(user_system, "GPU_Model", std::string{gpu_model});
+    telemetry_session.AddField(user_system, "GPU_OpenGL_Version", std::string{gl_version});
+}
+
+void Driver::DeduceVendor() {
+    if (gpu_vendor.find("NVIDIA") != gpu_vendor.npos) {
+        vendor = Vendor::Nvidia;
+    } else if ((gpu_vendor.find("ATI") != gpu_vendor.npos) ||
+               (gpu_vendor.find("AMD") != gpu_vendor.npos) ||
+               (gpu_vendor.find("Advanced Micro Devices") != gpu_vendor.npos)) {
+        vendor = Vendor::AMD;
+    } else if (gpu_vendor.find("Intel") != gpu_vendor.npos) {
+        vendor = Vendor::Intel;
+    } else if (gpu_vendor.find("ARM") != gpu_vendor.npos) {
+        vendor = Vendor::ARM;
+    } else if (gpu_vendor.find("Qualcomm") != gpu_vendor.npos) {
+        vendor = Vendor::Qualcomm;
+    } else if (gpu_vendor.find("Samsung") != gpu_vendor.npos) {
+        vendor = Vendor::Samsung;
+    } else if (gpu_vendor.find("GDI Generic") != gpu_vendor.npos) {
+        vendor = Vendor::Generic;
+    }
+}
+
+void Driver::CheckExtensionSupport() {
+    ext_buffer_storage = GLAD_GL_EXT_buffer_storage;
+    arb_buffer_storage = GLAD_GL_ARB_buffer_storage;
+    arb_clear_texture = GLAD_GL_ARB_clear_texture;
+    arb_get_texture_sub_image = GLAD_GL_ARB_get_texture_sub_image;
+    ext_clip_cull_distance = GLAD_GL_EXT_clip_cull_distance;
+    is_suitable = GLAD_GL_VERSION_4_3 || GLAD_GL_ES_VERSION_3_1;
+}
+
+void Driver::FindBugs() {
+#ifdef __unix__
+    const bool is_linux = true;
+#else
+    const bool is_linux = false;
+#endif
+
+    // TODO: Check if these have been fixed in the newer driver
+    if (vendor == Vendor::AMD) {
+        bugs |= DriverBug::ShaderStageChangeFreeze | DriverBug::VertexArrayOutOfBound;
+    }
+
+    if (vendor == Vendor::AMD || (vendor == Vendor::Intel && !is_linux)) {
+        bugs |= DriverBug::BrokenTextureView;
+    }
+}
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_driver.h

@@ -0,0 +1,114 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <string_view>
+#include "common/common_types.h"
+
+namespace Core {
+class TelemetrySession;
+}
+
+namespace OpenGL {
+
+enum class Vendor {
+    Unknown = 0,
+    AMD = 1,
+    Nvidia = 2,
+    Intel = 3,
+    ARM = 4,
+    Qualcomm = 5,
+    Samsung = 6,
+    Generic = 7,
+};
+
+enum class DriverBug {
+    // AMD drivers sometimes freezes when one shader stage is changed but not the others.
+    ShaderStageChangeFreeze = 1 << 0,
+    // On AMD drivers there is a strange crash in indexed drawing. The crash happens when the buffer
+    // read position is near the end and is an out-of-bound access to the vertex buffer. This is
+    // probably a bug in the driver and is related to the usage of vec3<byte> attributes in the
+    // vertex array. Doubling the allocation size for the vertex buffer seems to avoid the crash.
+    VertexArrayOutOfBound = 1 << 1,
+    // On AMD and Intel drivers on Windows glTextureView produces incorrect results
+    BrokenTextureView = 1 << 2,
+};
+
+/**
+ * Utility class that loads the OpenGL function pointers and reports
+ * information about the graphics device and driver used
+ */
+class Driver {
+public:
+    Driver(Core::TelemetrySession& telemetry_session);
+    ~Driver();
+
+    /// Returns true of the driver has a particular bug stated in the DriverBug enum
+    bool HasBug(DriverBug bug) const;
+
+    /// Returns the vendor of the currently selected physical device
+    Vendor GetVendor() const {
+        return vendor;
+    }
+
+    /// Returns the gpu vendor string returned by the driver
+    std::string_view GetVendorString() const {
+        return gpu_vendor;
+    }
+
+    /// Returns true if the implementation is suitable for emulation
+    bool IsSuitable() const {
+        return is_suitable;
+    }
+
+    /// Returns true if the implementation supports ARB_buffer_storage
+    bool HasArbBufferStorage() const {
+        return arb_buffer_storage;
+    }
+
+    /// Returns true if the implementation supports EXT_buffer_storage
+    bool HasExtBufferStorage() const {
+        return ext_buffer_storage;
+    }
+
+    /// Returns true if the implementation supports ARB_clear_texture
+    bool HasArbClearTexture() const {
+        return arb_clear_texture;
+    }
+
+    /// Returns true if the implementation supports ARB_get_texture_sub_image
+    bool HasArbGetTextureSubImage() const {
+        return arb_get_texture_sub_image;
+    }
+
+    /// Returns true if the implementation supports EXT_clip_cull_distance
+    bool HasExtClipCullDistance() const {
+        return ext_clip_cull_distance;
+    }
+
+private:
+    void ReportDriverInfo();
+    void DeduceVendor();
+    void CheckExtensionSupport();
+    void FindBugs();
+
+private:
+    Core::TelemetrySession& telemetry_session;
+    Vendor vendor = Vendor::Unknown;
+    DriverBug bugs{};
+    bool is_suitable{};
+
+    bool ext_buffer_storage{};
+    bool arb_buffer_storage{};
+    bool arb_clear_texture{};
+    bool arb_get_texture_sub_image{};
+    bool ext_clip_cull_distance{};
+
+    std::string_view gl_version{};
+    std::string_view gpu_vendor{};
+    std::string_view gpu_model{};
+};
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -3,12 +3,11 @@
 // Refer to the license.txt file included.
 
 #pragma once
-#include "common/vector_math.h"
+
 #include "core/hw/gpu.h"
-#include "video_core/pica_types.h"
+#include "video_core/rasterizer_accelerated.h"
 #include "video_core/rasterizer_cache/rasterizer_cache.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/regs_lighting.h"
 #include "video_core/regs_texturing.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
@@ -20,20 +19,20 @@ class EmuWindow;
 }
 
 namespace OpenGL {
+
+class Driver;
 class ShaderProgramManager;
 
-class RasterizerOpenGL : public VideoCore::RasterizerInterface {
+class RasterizerOpenGL : public VideoCore::RasterizerAccelerated {
 public:
-    explicit RasterizerOpenGL(Frontend::EmuWindow& emu_window);
+    explicit RasterizerOpenGL(Memory::MemorySystem& memory, Frontend::EmuWindow& emu_window,
+                              Driver& driver);
     ~RasterizerOpenGL() override;
 
     void LoadDiskResources(const std::atomic_bool& stop_loading,
                            const VideoCore::DiskResourceLoadCallback& callback) override;
 
-    void AddTriangle(const Pica::Shader::OutputVertex& v0, const Pica::Shader::OutputVertex& v1,
-                     const Pica::Shader::OutputVertex& v2) override;
     void DrawTriangles() override;
-    void NotifyPicaRegisterChanged(u32 id) override;
     void FlushAll() override;
     void FlushRegion(PAddr addr, u32 size) override;
     void InvalidateRegion(PAddr addr, u32 size) override;
@@ -46,10 +45,10 @@ public:
                            u32 pixel_stride, ScreenInfo& screen_info) override;
     bool AccelerateDrawBatch(bool is_indexed) override;
 
-    /// Syncs entire status to match PICA registers
-    void SyncEntireState() override;
-
 private:
+    void SyncFixedState() override;
+    void NotifyFixedFunctionPicaRegisterChanged(u32 id) override;
+
     struct SamplerInfo {
         using TextureConfig = Pica::TexturingRegs::TextureConfig;
 
@@ -76,66 +75,15 @@ private:
         bool supress_mipmap_for_cube = false;
     };
 
-    /// Structure that the hardware rendered vertices are composed of
-    struct HardwareVertex {
-        HardwareVertex() = default;
-        HardwareVertex(const Pica::Shader::OutputVertex& v, bool flip_quaternion) {
-            position[0] = v.pos.x.ToFloat32();
-            position[1] = v.pos.y.ToFloat32();
-            position[2] = v.pos.z.ToFloat32();
-            position[3] = v.pos.w.ToFloat32();
-            color[0] = v.color.x.ToFloat32();
-            color[1] = v.color.y.ToFloat32();
-            color[2] = v.color.z.ToFloat32();
-            color[3] = v.color.w.ToFloat32();
-            tex_coord0[0] = v.tc0.x.ToFloat32();
-            tex_coord0[1] = v.tc0.y.ToFloat32();
-            tex_coord1[0] = v.tc1.x.ToFloat32();
-            tex_coord1[1] = v.tc1.y.ToFloat32();
-            tex_coord2[0] = v.tc2.x.ToFloat32();
-            tex_coord2[1] = v.tc2.y.ToFloat32();
-            tex_coord0_w = v.tc0_w.ToFloat32();
-            normquat[0] = v.quat.x.ToFloat32();
-            normquat[1] = v.quat.y.ToFloat32();
-            normquat[2] = v.quat.z.ToFloat32();
-            normquat[3] = v.quat.w.ToFloat32();
-            view[0] = v.view.x.ToFloat32();
-            view[1] = v.view.y.ToFloat32();
-            view[2] = v.view.z.ToFloat32();
-
-            if (flip_quaternion) {
-                normquat = -normquat;
-            }
-        }
-
-        Common::Vec4f position;
-        Common::Vec4f color;
-        Common::Vec2f tex_coord0;
-        Common::Vec2f tex_coord1;
-        Common::Vec2f tex_coord2;
-        float tex_coord0_w;
-        Common::Vec4f normquat;
-        Common::Vec3f view;
-    };
-
     /// Syncs the clip enabled status to match the PICA register
     void SyncClipEnabled();
 
-    /// Syncs the clip coefficients to match the PICA register
-    void SyncClipCoef();
-
     /// Sets the OpenGL shader in accordance with the current PICA register state
     void SetShader();
 
     /// Syncs the cull mode to match the PICA register
     void SyncCullMode();
 
-    /// Syncs the depth scale to match the PICA register
-    void SyncDepthScale();
-
-    /// Syncs the depth offset to match the PICA register
-    void SyncDepthOffset();
-
     /// Syncs the blend enabled status to match the PICA register
     void SyncBlendEnabled();
 
@@ -145,18 +93,6 @@ private:
     /// Syncs the blend color to match the PICA register
     void SyncBlendColor();
 
-    /// Syncs the fog states to match the PICA register
-    void SyncFogColor();
-
-    /// Sync the procedural texture noise configuration to match the PICA register
-    void SyncProcTexNoise();
-
-    /// Sync the procedural texture bias configuration to match the PICA register
-    void SyncProcTexBias();
-
-    /// Syncs the alpha test states to match the PICA register
-    void SyncAlphaTest();
-
     /// Syncs the logic op states to match the PICA register
     void SyncLogicOp();
 
@@ -175,46 +111,6 @@ private:
     /// Syncs the depth test states to match the PICA register
     void SyncDepthTest();
 
-    /// Syncs the TEV combiner color buffer to match the PICA register
-    void SyncCombinerColor();
-
-    /// Syncs the TEV constant color to match the PICA register
-    void SyncTevConstColor(std::size_t tev_index,
-                           const Pica::TexturingRegs::TevStageConfig& tev_stage);
-
-    /// Syncs the lighting global ambient color to match the PICA register
-    void SyncGlobalAmbient();
-
-    /// Syncs the specified light's specular 0 color to match the PICA register
-    void SyncLightSpecular0(int light_index);
-
-    /// Syncs the specified light's specular 1 color to match the PICA register
-    void SyncLightSpecular1(int light_index);
-
-    /// Syncs the specified light's diffuse color to match the PICA register
-    void SyncLightDiffuse(int light_index);
-
-    /// Syncs the specified light's ambient color to match the PICA register
-    void SyncLightAmbient(int light_index);
-
-    /// Syncs the specified light's position to match the PICA register
-    void SyncLightPosition(int light_index);
-
-    /// Syncs the specified spot light direcition to match the PICA register
-    void SyncLightSpotDirection(int light_index);
-
-    /// Syncs the specified light's distance attenuation bias to match the PICA register
-    void SyncLightDistanceAttenuationBias(int light_index);
-
-    /// Syncs the specified light's distance attenuation scale to match the PICA register
-    void SyncLightDistanceAttenuationScale(int light_index);
-
-    /// Syncs the shadow rendering bias to match the PICA register
-    void SyncShadowBias();
-
-    /// Syncs the shadow texture bias to match the PICA register
-    void SyncShadowTextureBias();
-
     /// Syncs and uploads the lighting, fog and proctex LUTs
     void SyncAndUploadLUTs();
     void SyncAndUploadLUTsLF();
@@ -228,15 +124,6 @@ private:
     /// Internal implementation for AccelerateDrawBatch
     bool AccelerateDrawBatchInternal(bool is_indexed);
 
-    struct VertexArrayInfo {
-        u32 vs_input_index_min;
-        u32 vs_input_index_max;
-        u32 vs_input_size;
-    };
-
-    /// Retrieve the range and the size of the input vertex
-    VertexArrayInfo AnalyzeVertexArray(bool is_indexed);
-
     /// Setup vertex array for AccelerateDrawBatch
     void SetupVertexArray(u8* array_ptr, GLintptr buffer_offset, GLuint vs_input_index_min,
                           GLuint vs_input_index_max);
@@ -247,38 +134,13 @@ private:
     /// Setup geometry shader for AccelerateDrawBatch
     bool SetupGeometryShader();
 
-    bool is_amd;
+private:
-
+    Driver& driver;
     OpenGLState state;
     GLuint default_texture;
-
     RasterizerCacheOpenGL res_cache;
-
-    std::vector<HardwareVertex> vertex_batch;
-
-    bool shader_dirty = true;
-
-    struct {
-        UniformData data;
-        std::array<bool, Pica::LightingRegs::NumLightingSampler> lighting_lut_dirty;
-        bool lighting_lut_dirty_any;
-        bool fog_lut_dirty;
-        bool proctex_noise_lut_dirty;
-        bool proctex_color_map_dirty;
-        bool proctex_alpha_map_dirty;
-        bool proctex_lut_dirty;
-        bool proctex_diff_lut_dirty;
-        bool dirty;
-    } uniform_block_data = {};
-
     std::unique_ptr<ShaderProgramManager> shader_program_manager;
 
-    // They shall be big enough for about one frame.
-    static constexpr std::size_t VERTEX_BUFFER_SIZE = 16 * 1024 * 1024;
-    static constexpr std::size_t INDEX_BUFFER_SIZE = 1 * 1024 * 1024;
-    static constexpr std::size_t UNIFORM_BUFFER_SIZE = 2 * 1024 * 1024;
-    static constexpr std::size_t TEXTURE_BUFFER_SIZE = 1 * 1024 * 1024;
-
     OGLVertexArray sw_vao; // VAO for software shader draw
     OGLVertexArray hw_vao; // VAO for hardware shader / accelerate draw
     std::array<bool, 16> hw_vao_enabled_attributes{};
@@ -299,15 +161,6 @@ private:
     OGLTexture texture_buffer_lut_lf;
     OGLTexture texture_buffer_lut_rg;
     OGLTexture texture_buffer_lut_rgba;
-
-    std::array<std::array<Common::Vec2f, 256>, Pica::LightingRegs::NumLightingSampler>
-        lighting_lut_data{};
-    std::array<Common::Vec2f, 128> fog_lut_data{};
-    std::array<Common::Vec2f, 128> proctex_noise_lut_data{};
-    std::array<Common::Vec2f, 128> proctex_color_map_data{};
-    std::array<Common::Vec2f, 128> proctex_alpha_map_data{};
-    std::array<Common::Vec4f, 256> proctex_lut_data{};
-    std::array<Common::Vec4f, 256> proctex_diff_lut_data{};
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_shader_gen.cpp

@@ -12,6 +12,7 @@
 #include "video_core/renderer_opengl/gl_shader_gen.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
 #include "video_core/renderer_opengl/gl_vars.h"
+#include "video_core/shader/shader_uniforms.h"
 #include "video_core/video_core.h"
 
 using Pica::FramebufferRegs;
@@ -23,53 +24,7 @@ using VSOutputAttributes = RasterizerRegs::VSOutputAttributes;
 
 namespace OpenGL {
 
-constexpr std::string_view UniformBlockDef = R"(
+const std::string UniformBlockDef = Pica::Shader::BuildShaderUniformDefinitions();
-#define NUM_TEV_STAGES 6
-#define NUM_LIGHTS 8
-#define NUM_LIGHTING_SAMPLERS 24
-
-struct LightSrc {
-    vec3 specular_0;
-    vec3 specular_1;
-    vec3 diffuse;
-    vec3 ambient;
-    vec3 position;
-    vec3 spot_direction;
-    float dist_atten_bias;
-    float dist_atten_scale;
-};
-
-layout (std140) uniform shader_data {
-    int framebuffer_scale;
-    int alphatest_ref;
-    float depth_scale;
-    float depth_offset;
-    float shadow_bias_constant;
-    float shadow_bias_linear;
-    int scissor_x1;
-    int scissor_y1;
-    int scissor_x2;
-    int scissor_y2;
-    int fog_lut_offset;
-    int proctex_noise_lut_offset;
-    int proctex_color_map_offset;
-    int proctex_alpha_map_offset;
-    int proctex_lut_offset;
-    int proctex_diff_lut_offset;
-    float proctex_bias;
-    int shadow_texture_bias;
-    ivec4 lighting_lut_offset[NUM_LIGHTING_SAMPLERS / 4];
-    vec3 fog_color;
-    vec2 proctex_noise_f;
-    vec2 proctex_noise_a;
-    vec2 proctex_noise_p;
-    vec3 lighting_global_ambient;
-    LightSrc light_src[NUM_LIGHTS];
-    vec4 const_color[NUM_TEV_STAGES];
-    vec4 tev_combiner_buffer_color;
-    vec4 clip_coef;
-};
-)";
 
 static std::string GetVertexInterfaceDeclaration(bool is_output, bool separable_shader) {
     std::string out;

src/video_core/renderer_opengl/gl_shader_manager.cpp

@@ -6,13 +6,13 @@
 #include <set>
 #include <thread>
 #include <unordered_map>
-#include <boost/variant.hpp>
+#include <variant>
-#include "core/frontend/scope_acquire_context.h"
+#include "video_core/renderer_opengl/gl_driver.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_disk_cache.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
-#include "video_core/renderer_opengl/gl_vars.h"
+#include "video_core/shader/shader_uniforms.h"
 #include "video_core/video_core.h"
 
 namespace OpenGL {
@@ -85,7 +85,8 @@ static std::tuple<PicaVSConfig, Pica::Shader::ShaderSetup> BuildVSConfigFromRaw(
     return {PicaVSConfig{raw.GetRawShaderConfig().vs, setup}, setup};
 }
 
-static void SetShaderUniformBlockBinding(GLuint shader, const char* name, UniformBindings binding,
+static void SetShaderUniformBlockBinding(GLuint shader, const char* name,
+                                         Pica::Shader::UniformBindings binding,
                                          std::size_t expected_size) {
     const GLuint ub_index = glGetUniformBlockIndex(shader, name);
     if (ub_index == GL_INVALID_INDEX) {
@@ -100,9 +101,10 @@ static void SetShaderUniformBlockBinding(GLuint shader, const char* name, Unifor
 }
 
 static void SetShaderUniformBlockBindings(GLuint shader) {
-    SetShaderUniformBlockBinding(shader, "shader_data", UniformBindings::Common,
+    SetShaderUniformBlockBinding(shader, "shader_data", Pica::Shader::UniformBindings::Common,
-                                 sizeof(UniformData));
+                                 sizeof(Pica::Shader::UniformData));
-    SetShaderUniformBlockBinding(shader, "vs_config", UniformBindings::VS, sizeof(VSUniformData));
+    SetShaderUniformBlockBinding(shader, "vs_config", Pica::Shader::UniformBindings::VS,
+                                 sizeof(Pica::Shader::VSUniformData));
 }
 
 static void SetShaderSamplerBinding(GLuint shader, const char* name,
@@ -148,21 +150,6 @@ static void SetShaderSamplerBindings(GLuint shader) {
     cur_state.Apply();
 }
 
-void PicaUniformsData::SetFromRegs(const Pica::ShaderRegs& regs,
-                                   const Pica::Shader::ShaderSetup& setup) {
-    std::transform(std::begin(setup.uniforms.b), std::end(setup.uniforms.b), std::begin(bools),
-                   [](bool value) -> BoolAligned { return {value ? GL_TRUE : GL_FALSE}; });
-    std::transform(std::begin(regs.int_uniforms), std::end(regs.int_uniforms), std::begin(i),
-                   [](const auto& value) -> Common::Vec4u {
-                       return {value.x.Value(), value.y.Value(), value.z.Value(), value.w.Value()};
-                   });
-    std::transform(std::begin(setup.uniforms.f), std::end(setup.uniforms.f), std::begin(f),
-                   [](const auto& value) -> Common::Vec4f {
-                       return {value.x.ToFloat32(), value.y.ToFloat32(), value.z.ToFloat32(),
-                               value.w.ToFloat32()};
-                   });
-}
-
 /**
  * An object representing a shader program staging. It can be either a shader object or a program
  * object, depending on whether separable program is used.
@@ -178,12 +165,12 @@ public:
     }
 
     void Create(const char* source, GLenum type) {
-        if (shader_or_program.which() == 0) {
+        if (shader_or_program.index() == 0) {
-            boost::get<OGLShader>(shader_or_program).Create(source, type);
+            std::get<OGLShader>(shader_or_program).Create(source, type);
         } else {
             OGLShader shader;
             shader.Create(source, type);
-            OGLProgram& program = boost::get<OGLProgram>(shader_or_program);
+            OGLProgram& program = std::get<OGLProgram>(shader_or_program);
             program.Create(true, {shader.handle});
             SetShaderUniformBlockBindings(program.handle);
 
@@ -194,10 +181,10 @@ public:
     }
 
     GLuint GetHandle() const {
-        if (shader_or_program.which() == 0) {
+        if (shader_or_program.index() == 0) {
-            return boost::get<OGLShader>(shader_or_program).handle;
+            return std::get<OGLShader>(shader_or_program).handle;
         } else {
-            return boost::get<OGLProgram>(shader_or_program).handle;
+            return std::get<OGLProgram>(shader_or_program).handle;
         }
     }
 
@@ -208,7 +195,7 @@ public:
     }
 
 private:
-    boost::variant<OGLShader, OGLProgram> shader_or_program;
+    std::variant<OGLShader, OGLProgram> shader_or_program;
 };
 
 class TrivialVertexShader {
@@ -329,8 +316,8 @@ using FragmentShaders = ShaderCache<PicaFSConfig, &GenerateFragmentShader, GL_FR
 
 class ShaderProgramManager::Impl {
 public:
-    explicit Impl(bool separable, bool is_amd)
+    explicit Impl(bool separable)
-        : is_amd(is_amd), separable(separable), programmable_vertex_shaders(separable),
+        : separable(separable), programmable_vertex_shaders(separable),
           trivial_vertex_shader(separable), fixed_geometry_shaders(separable),
           fragment_shaders(separable), disk_cache(separable) {
         if (separable)
@@ -363,7 +350,6 @@ public:
     static_assert(offsetof(ShaderTuple, fs_hash) == sizeof(std::size_t) * 2,
                   "ShaderTuple layout changed!");
 
-    bool is_amd;
     bool separable;
 
     ShaderTuple current;
@@ -379,9 +365,9 @@ public:
     ShaderDiskCache disk_cache;
 };
 
-ShaderProgramManager::ShaderProgramManager(Frontend::EmuWindow& emu_window_, bool separable,
+ShaderProgramManager::ShaderProgramManager(Frontend::EmuWindow& emu_window_, const Driver& driver_,
-                                           bool is_amd)
+                                           bool separable)
-    : impl(std::make_unique<Impl>(separable, is_amd)), emu_window{emu_window_} {}
+    : impl(std::make_unique<Impl>(separable)), emu_window{emu_window_}, driver{driver_} {}
 
 ShaderProgramManager::~ShaderProgramManager() = default;
 
@@ -443,10 +429,7 @@ void ShaderProgramManager::UseFragmentShader(const Pica::Regs& regs) {
 
 void ShaderProgramManager::ApplyTo(OpenGLState& state) {
     if (impl->separable) {
-        if (impl->is_amd) {
+        if (driver.HasBug(DriverBug::ShaderStageChangeFreeze)) {
-            // Without this reseting, AMD sometimes freezes when one stage is changed but not
-            // for the others. On the other hand, including this reset seems to introduce memory
-            // leak in Intel Graphics.
             glUseProgramStages(
                 impl->pipeline.handle,
                 GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT | GL_FRAGMENT_SHADER_BIT, 0);
@@ -641,7 +624,7 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
     std::size_t built_shaders = 0; // It doesn't have be atomic since it's used behind a mutex
     const auto LoadRawSepareble = [&](Frontend::GraphicsContext* context, std::size_t begin,
                                       std::size_t end) {
-        Frontend::ScopeAcquireContext scope(*context);
+        const auto scope = context->Acquire();
         for (std::size_t i = begin; i < end; ++i) {
             if (stop_loading || compilation_failed) {
                 return;

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -5,13 +5,7 @@
 #pragma once
 
 #include <memory>
-#include "common/vector_math.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/regs_lighting.h"
-
-namespace Core {
-class System;
-}
 
 namespace Frontend {
 class EmuWindow;
@@ -19,8 +13,7 @@ class EmuWindow;
 
 namespace Pica {
 struct Regs;
-struct ShaderRegs;
+}
-} // namespace Pica
 
 namespace Pica::Shader {
 struct ShaderSetup;
@@ -28,87 +21,13 @@ struct ShaderSetup;
 
 namespace OpenGL {
 
-enum class UniformBindings : u32 { Common, VS, GS };
+class Driver;
-
-struct LightSrc {
-    alignas(16) Common::Vec3f specular_0;
-    alignas(16) Common::Vec3f specular_1;
-    alignas(16) Common::Vec3f diffuse;
-    alignas(16) Common::Vec3f ambient;
-    alignas(16) Common::Vec3f position;
-    alignas(16) Common::Vec3f spot_direction; // negated
-    float dist_atten_bias;
-    float dist_atten_scale;
-};
-
-/// Uniform structure for the Uniform Buffer Object, all vectors must be 16-byte aligned
-// NOTE: Always keep a vec4 at the end. The GL spec is not clear wether the alignment at
-//       the end of a uniform block is included in UNIFORM_BLOCK_DATA_SIZE or not.
-//       Not following that rule will cause problems on some AMD drivers.
-struct UniformData {
-    int framebuffer_scale;
-    int alphatest_ref;
-    float depth_scale;
-    float depth_offset;
-    float shadow_bias_constant;
-    float shadow_bias_linear;
-    int scissor_x1;
-    int scissor_y1;
-    int scissor_x2;
-    int scissor_y2;
-    int fog_lut_offset;
-    int proctex_noise_lut_offset;
-    int proctex_color_map_offset;
-    int proctex_alpha_map_offset;
-    int proctex_lut_offset;
-    int proctex_diff_lut_offset;
-    float proctex_bias;
-    int shadow_texture_bias;
-    alignas(16) Common::Vec4i lighting_lut_offset[Pica::LightingRegs::NumLightingSampler / 4];
-    alignas(16) Common::Vec3f fog_color;
-    alignas(8) Common::Vec2f proctex_noise_f;
-    alignas(8) Common::Vec2f proctex_noise_a;
-    alignas(8) Common::Vec2f proctex_noise_p;
-    alignas(16) Common::Vec3f lighting_global_ambient;
-    LightSrc light_src[8];
-    alignas(16) Common::Vec4f const_color[6]; // A vec4 color for each of the six tev stages
-    alignas(16) Common::Vec4f tev_combiner_buffer_color;
-    alignas(16) Common::Vec4f clip_coef;
-};
-
-static_assert(sizeof(UniformData) == 0x4F0,
-              "The size of the UniformData does not match the structure in the shader");
-static_assert(sizeof(UniformData) < 16384,
-              "UniformData structure must be less than 16kb as per the OpenGL spec");
-
-/// Uniform struct for the Uniform Buffer Object that contains PICA vertex/geometry shader uniforms.
-// NOTE: the same rule from UniformData also applies here.
-struct PicaUniformsData {
-    void SetFromRegs(const Pica::ShaderRegs& regs, const Pica::Shader::ShaderSetup& setup);
-
-    struct BoolAligned {
-        alignas(16) int b;
-    };
-
-    std::array<BoolAligned, 16> bools;
-    alignas(16) std::array<Common::Vec4u, 4> i;
-    alignas(16) std::array<Common::Vec4f, 96> f;
-};
-
-struct VSUniformData {
-    PicaUniformsData uniforms;
-};
-static_assert(sizeof(VSUniformData) == 1856,
-              "The size of the VSUniformData does not match the structure in the shader");
-static_assert(sizeof(VSUniformData) < 16384,
-              "VSUniformData structure must be less than 16kb as per the OpenGL spec");
-
 class OpenGLState;
 
 /// A class that manage different shader stages and configures them with given config data.
 class ShaderProgramManager {
 public:
-    ShaderProgramManager(Frontend::EmuWindow& emu_window_, bool separable, bool is_amd);
+    ShaderProgramManager(Frontend::EmuWindow& emu_window, const Driver& driver, bool separable);
     ~ShaderProgramManager();
 
     void LoadDiskCache(const std::atomic_bool& stop_loading,
@@ -131,5 +50,6 @@ private:
     std::unique_ptr<Impl> impl;
 
     Frontend::EmuWindow& emu_window;
+    const Driver& driver;
 };
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_stream_buffer.cpp

@@ -5,6 +5,7 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/microprofile.h"
+#include "video_core/renderer_opengl/gl_driver.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
 
 MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
@@ -12,19 +13,14 @@ MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
 
 namespace OpenGL {
 
-OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool array_buffer_for_amd,
+OGLStreamBuffer::OGLStreamBuffer(Driver& driver, GLenum target, GLsizeiptr size,
                                  bool prefer_coherent)
     : gl_target(target), buffer_size(size) {
     gl_buffer.Create();
     glBindBuffer(gl_target, gl_buffer.handle);
 
     GLsizeiptr allocate_size = size;
-    if (array_buffer_for_amd) {
+    if (driver.HasBug(DriverBug::VertexArrayOutOfBound) && target == GL_ARRAY_BUFFER) {
-        // On AMD GPU there is a strange crash in indexed drawing. The crash happens when the buffer
-        // read position is near the end and is an out-of-bound access to the vertex buffer. This is
-        // probably a bug in the driver and is related to the usage of vec3<byte> attributes in the
-        // vertex array. Doubling the allocation size for the vertex buffer seems to avoid the
-        // crash.
         allocate_size *= 2;
     }
 

src/video_core/renderer_opengl/gl_stream_buffer.h

@@ -3,14 +3,17 @@
 // Refer to the license.txt file included.
 
 #pragma once
+
 #include <tuple>
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 
 namespace OpenGL {
 
+class Driver;
+
 class OGLStreamBuffer : private NonCopyable {
 public:
-    explicit OGLStreamBuffer(GLenum target, GLsizeiptr size, bool array_buffer_for_amd,
+    explicit OGLStreamBuffer(Driver& driver, GLenum target, GLsizeiptr size,
                              bool prefer_coherent = false);
     ~OGLStreamBuffer();
 

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -13,8 +13,6 @@
 #include "core/hw/hw.h"
 #include "core/hw/lcd.h"
 #include "core/memory.h"
-#include "core/tracer/recorder.h"
-#include "video_core/debug_utils/debug_utils.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
 #include "video_core/renderer_opengl/gl_state.h"
@@ -352,14 +350,17 @@ static std::array<GLfloat, 3 * 2> MakeOrthographicMatrix(const float width, cons
     return matrix;
 }
 
-RendererOpenGL::RendererOpenGL(Frontend::EmuWindow& window, Frontend::EmuWindow* secondary_window)
+RendererOpenGL::RendererOpenGL(Core::System& system, Frontend::EmuWindow& window,
-    : RendererBase{window, secondary_window},
+                               Frontend::EmuWindow* secondary_window)
-      frame_dumper(Core::System::GetInstance().VideoDumper(), window) {
+    : VideoCore::RendererBase{system, window, secondary_window}, driver{system.TelemetrySession()},
+      frame_dumper{system.VideoDumper(), window} {
     window.mailbox = std::make_unique<OGLTextureMailbox>();
     if (secondary_window) {
         secondary_window->mailbox = std::make_unique<OGLTextureMailbox>();
     }
     frame_dumper.mailbox = std::make_unique<OGLVideoDumpingMailbox>();
+    InitOpenGLObjects();
+    rasterizer = std::make_unique<RasterizerOpenGL>(system.Memory(), render_window, driver);
 }
 
 RendererOpenGL::~RendererOpenGL() = default;
@@ -374,7 +375,6 @@ void RendererOpenGL::SwapBuffers() {
     state.Apply();
 
     PrepareRendertarget();
-
     RenderScreenshot();
 
     const auto& main_layout = render_window.GetFramebufferLayout();
@@ -396,26 +396,12 @@ void RendererOpenGL::SwapBuffers() {
         }
     }
 
-    m_current_frame++;
+    EndFrame();
-
-    Core::System::GetInstance().perf_stats->EndSystemFrame();
-
-    render_window.PollEvents();
-
-    Core::System::GetInstance().frame_limiter.DoFrameLimiting(
-        Core::System::GetInstance().CoreTiming().GetGlobalTimeUs());
-    Core::System::GetInstance().perf_stats->BeginSystemFrame();
-
     prev_state.Apply();
-    RefreshRasterizerSetting();
-
-    if (Pica::g_debug_context && Pica::g_debug_context->recorder) {
-        Pica::g_debug_context->recorder->FrameFinished();
-    }
 }
 
 void RendererOpenGL::RenderScreenshot() {
-    if (VideoCore::g_renderer_screenshot_requested) {
+    if (renderer_settings.screenshot_requested.exchange(false)) {
         // Draw this frame to the screenshot framebuffer
         screenshot_framebuffer.Create();
         GLuint old_read_fb = state.draw.read_framebuffer;
@@ -423,7 +409,7 @@ void RendererOpenGL::RenderScreenshot() {
         state.draw.read_framebuffer = state.draw.draw_framebuffer = screenshot_framebuffer.handle;
         state.Apply();
 
-        Layout::FramebufferLayout layout{VideoCore::g_screenshot_framebuffer_layout};
+        const auto layout{renderer_settings.screenshot_framebuffer_layout};
 
         GLuint renderbuffer;
         glGenRenderbuffers(1, &renderbuffer);
@@ -435,7 +421,7 @@ void RendererOpenGL::RenderScreenshot() {
         DrawScreens(layout, false);
 
         glReadPixels(0, 0, layout.width, layout.height, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV,
-                     VideoCore::g_screenshot_bits);
+                     renderer_settings.screenshot_bits);
 
         screenshot_framebuffer.Release();
         state.draw.read_framebuffer = old_read_fb;
@@ -443,8 +429,7 @@ void RendererOpenGL::RenderScreenshot() {
         state.Apply();
         glDeleteRenderbuffers(1, &renderbuffer);
 
-        VideoCore::g_screenshot_complete_callback();
+        renderer_settings.screenshot_complete_callback();
-        VideoCore::g_renderer_screenshot_requested = false;
     }
 }
 
@@ -1226,109 +1211,8 @@ void RendererOpenGL::CleanupVideoDumping() {
     mailbox->free_cv.notify_one();
 }
 
-static const char* GetSource(GLenum source) {
+void RendererOpenGL::Sync() {
-#define RET(s)                                                                                     \
+    rasterizer->SyncEntireState();
-    case GL_DEBUG_SOURCE_##s:                                                                      \
-        return #s
-    switch (source) {
-        RET(API);
-        RET(WINDOW_SYSTEM);
-        RET(SHADER_COMPILER);
-        RET(THIRD_PARTY);
-        RET(APPLICATION);
-        RET(OTHER);
-    default:
-        UNREACHABLE();
-    }
-#undef RET
-
-    return "";
 }
 
-static const char* GetType(GLenum type) {
-#define RET(t)                                                                                     \
-    case GL_DEBUG_TYPE_##t:                                                                        \
-        return #t
-    switch (type) {
-        RET(ERROR);
-        RET(DEPRECATED_BEHAVIOR);
-        RET(UNDEFINED_BEHAVIOR);
-        RET(PORTABILITY);
-        RET(PERFORMANCE);
-        RET(OTHER);
-        RET(MARKER);
-    default:
-        UNREACHABLE();
-    }
-#undef RET
-
-    return "";
-}
-
-static void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severity,
-                                  GLsizei length, const GLchar* message, const void* user_param) {
-    Log::Level level;
-    switch (severity) {
-    case GL_DEBUG_SEVERITY_HIGH:
-        level = Log::Level::Critical;
-        break;
-    case GL_DEBUG_SEVERITY_MEDIUM:
-        level = Log::Level::Warning;
-        break;
-    case GL_DEBUG_SEVERITY_NOTIFICATION:
-    case GL_DEBUG_SEVERITY_LOW:
-        level = Log::Level::Debug;
-        break;
-    }
-    LOG_GENERIC(Log::Class::Render_OpenGL, level, "{} {} {}: {}", GetSource(source), GetType(type),
-                id, message);
-}
-
-/// Initialize the renderer
-VideoCore::ResultStatus RendererOpenGL::Init() {
-#ifndef ANDROID
-    if (!gladLoadGL()) {
-        return VideoCore::ResultStatus::ErrorBelowGL43;
-    }
-
-    // Qualcomm has some spammy info messages that are marked as errors but not important
-    // https://developer.qualcomm.com/comment/11845
-    if (GLAD_GL_KHR_debug) {
-        glEnable(GL_DEBUG_OUTPUT);
-        glDebugMessageCallback(DebugHandler, nullptr);
-    }
-#endif
-
-    const std::string_view gl_version{reinterpret_cast<char const*>(glGetString(GL_VERSION))};
-    const std::string_view gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
-    const std::string_view gpu_model{reinterpret_cast<char const*>(glGetString(GL_RENDERER))};
-
-    LOG_INFO(Render_OpenGL, "GL_VERSION: {}", gl_version);
-    LOG_INFO(Render_OpenGL, "GL_VENDOR: {}", gpu_vendor);
-    LOG_INFO(Render_OpenGL, "GL_RENDERER: {}", gpu_model);
-
-    auto& telemetry_session = Core::System::GetInstance().TelemetrySession();
-    constexpr auto user_system = Common::Telemetry::FieldType::UserSystem;
-    telemetry_session.AddField(user_system, "GPU_Vendor", std::string(gpu_vendor));
-    telemetry_session.AddField(user_system, "GPU_Model", std::string(gpu_model));
-    telemetry_session.AddField(user_system, "GPU_OpenGL_Version", std::string(gl_version));
-
-    if (gpu_vendor == "GDI Generic") {
-        return VideoCore::ResultStatus::ErrorGenericDrivers;
-    }
-
-    if (!(GLAD_GL_VERSION_4_3 || GLAD_GL_ES_VERSION_3_1)) {
-        return VideoCore::ResultStatus::ErrorBelowGL43;
-    }
-
-    InitOpenGLObjects();
-
-    RefreshRasterizerSetting();
-
-    return VideoCore::ResultStatus::Success;
-}
-
-/// Shutdown the renderer
-void RendererOpenGL::ShutDown() {}
-
 } // namespace OpenGL

src/video_core/renderer_opengl/renderer_opengl.h

@@ -8,6 +8,8 @@
 #include "core/hw/gpu.h"
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_opengl/frame_dumper_opengl.h"
+#include "video_core/renderer_opengl/gl_driver.h"
+#include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
 
@@ -15,6 +17,10 @@ namespace Layout {
 struct FramebufferLayout;
 }
 
+namespace Core {
+class System;
+}
+
 namespace Frontend {
 
 struct Frame {
@@ -48,35 +54,21 @@ struct ScreenInfo {
     TextureInfo texture;
 };
 
-struct PresentationTexture {
+class RendererOpenGL : public VideoCore::RendererBase {
-    u32 width = 0;
-    u32 height = 0;
-    OGLTexture texture;
-};
-
-class RendererOpenGL : public RendererBase {
 public:
-    explicit RendererOpenGL(Frontend::EmuWindow& window, Frontend::EmuWindow* secondary_window);
+    explicit RendererOpenGL(Core::System& system, Frontend::EmuWindow& window,
+                            Frontend::EmuWindow* secondary_window);
     ~RendererOpenGL() override;
 
-    /// Initialize the renderer
+    [[nodiscard]] VideoCore::RasterizerInterface* Rasterizer() const override {
-    VideoCore::ResultStatus Init() override;
+        return rasterizer.get();
+    }
 
-    /// Shutdown the renderer
-    void ShutDown() override;
-
-    /// Finalizes rendering the guest frame
     void SwapBuffers() override;
-
-    /// Draws the latest frame from texture mailbox to the currently bound draw framebuffer in this
-    /// context
     void TryPresent(int timeout_ms, bool is_secondary) override;
-
-    /// Prepares for video dumping (e.g. create necessary buffers, etc)
     void PrepareVideoDumping() override;
-
-    /// Cleans up after video dumping is ended
     void CleanupVideoDumping() override;
+    void Sync() override;
 
 private:
     void InitOpenGLObjects();
@@ -111,7 +103,10 @@ private:
     // Fills active OpenGL texture with the given RGB color.
     void LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, const TextureInfo& texture);
 
+private:
+    Driver driver;
     OpenGLState state;
+    std::unique_ptr<RasterizerOpenGL> rasterizer;
 
     // OpenGL object IDs
     OGLVertexArray vertex_array;

src/video_core/renderer_software/rasterizer.cpp

@@ -22,12 +22,12 @@
 #include "video_core/regs_framebuffer.h"
 #include "video_core/regs_rasterizer.h"
 #include "video_core/regs_texturing.h"
+#include "video_core/renderer_software/rasterizer.h"
+#include "video_core/renderer_software/sw_framebuffer.h"
+#include "video_core/renderer_software/sw_lighting.h"
+#include "video_core/renderer_software/sw_proctex.h"
+#include "video_core/renderer_software/sw_texturing.h"
 #include "video_core/shader/shader.h"
-#include "video_core/swrasterizer/framebuffer.h"
-#include "video_core/swrasterizer/lighting.h"
-#include "video_core/swrasterizer/proctex.h"
-#include "video_core/swrasterizer/rasterizer.h"
-#include "video_core/swrasterizer/texturing.h"
 #include "video_core/texture/texture_decode.h"
 #include "video_core/utils.h"
 #include "video_core/video_core.h"

src/video_core/renderer_software/renderer_software.cpp

@@ -0,0 +1,19 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/renderer_software/renderer_software.h"
+
+namespace VideoCore {
+
+RendererSoftware::RendererSoftware(Core::System& system, Frontend::EmuWindow& window)
+    : VideoCore::RendererBase{system, window, nullptr},
+      rasterizer{std::make_unique<RasterizerSoftware>()} {}
+
+RendererSoftware::~RendererSoftware() = default;
+
+void RendererSoftware::SwapBuffers() {
+    EndFrame();
+}
+
+} // namespace VideoCore

src/video_core/renderer_software/renderer_software.h

@@ -0,0 +1,33 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "video_core/renderer_base.h"
+#include "video_core/renderer_software/sw_rasterizer.h"
+
+namespace Core {
+class System;
+}
+
+namespace VideoCore {
+
+class RendererSoftware : public VideoCore::RendererBase {
+public:
+    explicit RendererSoftware(Core::System& system, Frontend::EmuWindow& window);
+    ~RendererSoftware() override;
+
+    [[nodiscard]] VideoCore::RasterizerInterface* Rasterizer() const override {
+        return rasterizer.get();
+    }
+
+    void SwapBuffers() override;
+    void TryPresent(int timeout_ms, bool is_secondary) override {}
+    void Sync() override {}
+
+private:
+    std::unique_ptr<RasterizerSoftware> rasterizer;
+};
+
+} // namespace VideoCore

src/video_core/renderer_software/sw_clipper.cpp

@@ -12,9 +12,9 @@
 #include "common/vector_math.h"
 #include "video_core/pica_state.h"
 #include "video_core/pica_types.h"
+#include "video_core/renderer_software/rasterizer.h"
+#include "video_core/renderer_software/sw_clipper.h"
 #include "video_core/shader/shader.h"
-#include "video_core/swrasterizer/clipper.h"
-#include "video_core/swrasterizer/rasterizer.h"
 
 using Pica::Rasterizer::Vertex;
 

src/video_core/renderer_software/sw_framebuffer.cpp

@@ -12,7 +12,7 @@
 #include "core/memory.h"
 #include "video_core/pica_state.h"
 #include "video_core/regs_framebuffer.h"
-#include "video_core/swrasterizer/framebuffer.h"
+#include "video_core/renderer_software/sw_framebuffer.h"
 #include "video_core/utils.h"
 #include "video_core/video_core.h"
 

src/video_core/renderer_software/sw_lighting.cpp

@@ -3,7 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
-#include "video_core/swrasterizer/lighting.h"
+#include "video_core/renderer_software/sw_lighting.h"
 
 namespace Pica {
 

src/video_core/renderer_software/sw_proctex.cpp

@@ -5,7 +5,7 @@
 #include <array>
 #include <cmath>
 #include "common/math_util.h"
-#include "video_core/swrasterizer/proctex.h"
+#include "video_core/renderer_software/sw_proctex.h"
 
 namespace Pica::Rasterizer {
 

src/video_core/renderer_software/sw_rasterizer.cpp

@@ -0,0 +1,16 @@
+// Copyright 2015 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/renderer_software/sw_clipper.h"
+#include "video_core/renderer_software/sw_rasterizer.h"
+
+namespace VideoCore {
+
+void RasterizerSoftware::AddTriangle(const Pica::Shader::OutputVertex& v0,
+                                     const Pica::Shader::OutputVertex& v1,
+                                     const Pica::Shader::OutputVertex& v2) {
+    Pica::Clipper::ProcessTriangle(v0, v1, v2);
+}
+
+} // namespace VideoCore

src/video_core/renderer_software/sw_rasterizer.h

@@ -13,7 +13,7 @@ struct OutputVertex;
 
 namespace VideoCore {
 
-class SWRasterizer : public RasterizerInterface {
+class RasterizerSoftware : public RasterizerInterface {
     void AddTriangle(const Pica::Shader::OutputVertex& v0, const Pica::Shader::OutputVertex& v1,
                      const Pica::Shader::OutputVertex& v2) override;
     void DrawTriangles() override {}

src/video_core/renderer_software/sw_texturing.cpp

@@ -7,7 +7,7 @@
 #include "common/common_types.h"
 #include "common/vector_math.h"
 #include "video_core/regs_texturing.h"
-#include "video_core/swrasterizer/texturing.h"
+#include "video_core/renderer_software/sw_texturing.h"
 
 namespace Pica::Rasterizer {
 

src/video_core/shader/shader.cpp

@@ -86,7 +86,7 @@ void UnitState::LoadInput(const ShaderRegs& config, const AttributeBuffer& input
     }
 }
 
-static void CopyRegistersToOutput(const Common::Vec4<float24>* regs, u32 mask,
+static void CopyRegistersToOutput(std::span<Common::Vec4<float24>, 16> regs, u32 mask,
                                   AttributeBuffer& buffer) {
     int output_i = 0;
     for (int reg : Common::BitSet<u32>(mask)) {
@@ -108,7 +108,7 @@ GSEmitter::~GSEmitter() {
     delete handlers;
 }
 
-void GSEmitter::Emit(Common::Vec4<float24> (&output_regs)[16]) {
+void GSEmitter::Emit(std::span<Common::Vec4<float24>, 16> output_regs) {
     ASSERT(vertex_id < 3);
     // TODO: This should be merged with UnitState::WriteOutput somehow
     CopyRegistersToOutput(output_regs, output_mask, buffer[vertex_id]);

src/video_core/shader/shader.h

@@ -7,6 +7,7 @@
 #include <array>
 #include <cstddef>
 #include <functional>
+#include <span>
 #include <type_traits>
 #include <boost/serialization/access.hpp>
 #include <boost/serialization/array.hpp>
@@ -113,7 +114,7 @@ struct GSEmitter {
 
     GSEmitter();
     ~GSEmitter();
-    void Emit(Common::Vec4<float24> (&output_regs)[16]);
+    void Emit(std::span<Common::Vec4<float24>, 16> output_regs);
 
 private:
     friend class boost::serialization::access;
@@ -140,9 +141,9 @@ struct UnitState {
     struct Registers {
         // The registers are accessed by the shader JIT using SSE instructions, and are therefore
         // required to be 16-byte aligned.
-        alignas(16) Common::Vec4<float24> input[16];
+        alignas(16) std::array<Common::Vec4<float24>, 16> input;
-        alignas(16) Common::Vec4<float24> temporary[16];
+        alignas(16) std::array<Common::Vec4<float24>, 16> temporary;
-        alignas(16) Common::Vec4<float24> output[16];
+        alignas(16) std::array<Common::Vec4<float24>, 16> output;
 
     private:
         friend class boost::serialization::access;

src/video_core/shader/shader_interpreter.cpp

@@ -7,7 +7,6 @@
 #include <cmath>
 #include <numeric>
 #include <boost/container/static_vector.hpp>
-#include <boost/range/algorithm/fill.hpp>
 #include <nihstro/shader_bytecode.h>
 #include "common/assert.h"
 #include "common/common_types.h"
@@ -688,7 +687,7 @@ DebugData<true> InterpreterEngine::ProduceDebugInfo(const ShaderSetup& setup,
     DebugData<true> debug_data;
 
     // Setup input register table
-    boost::fill(state.registers.input, Common::Vec4<float24>::AssignToAll(float24::Zero()));
+    state.registers.input.fill(Common::Vec4<float24>::AssignToAll(float24::Zero()));
     state.LoadInput(config, input);
     RunInterpreter(setup, state, debug_data, setup.engine_data.entry_point);
     return debug_data;

src/video_core/shader/shader_uniforms.cpp

@@ -0,0 +1,78 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include "video_core/shader/shader.h"
+#include "video_core/shader/shader_uniforms.h"
+
+namespace Pica::Shader {
+
+void PicaUniformsData::SetFromRegs(const Pica::ShaderRegs& regs,
+                                   const Pica::Shader::ShaderSetup& setup) {
+    std::transform(std::begin(setup.uniforms.b), std::end(setup.uniforms.b), std::begin(bools),
+                   [](bool value) -> BoolAligned { return {value ? 1 : 0}; });
+    std::transform(std::begin(regs.int_uniforms), std::end(regs.int_uniforms), std::begin(i),
+                   [](const auto& value) -> Common::Vec4u {
+                       return {value.x.Value(), value.y.Value(), value.z.Value(), value.w.Value()};
+                   });
+    std::transform(std::begin(setup.uniforms.f), std::end(setup.uniforms.f), std::begin(f),
+                   [](const auto& value) -> Common::Vec4f {
+                       return {value.x.ToFloat32(), value.y.ToFloat32(), value.z.ToFloat32(),
+                               value.w.ToFloat32()};
+                   });
+}
+
+constexpr std::string_view UniformBlockDefFormat = R"(
+#define NUM_TEV_STAGES 6
+#define NUM_LIGHTS 8
+#define NUM_LIGHTING_SAMPLERS 24
+struct LightSrc {{
+    vec3 specular_0;
+    vec3 specular_1;
+    vec3 diffuse;
+    vec3 ambient;
+    vec3 position;
+    vec3 spot_direction;
+    float dist_atten_bias;
+    float dist_atten_scale;
+}};
+layout ({}std140) uniform shader_data {{
+    int framebuffer_scale;
+    int alphatest_ref;
+    float depth_scale;
+    float depth_offset;
+    float shadow_bias_constant;
+    float shadow_bias_linear;
+    int scissor_x1;
+    int scissor_y1;
+    int scissor_x2;
+    int scissor_y2;
+    int fog_lut_offset;
+    int proctex_noise_lut_offset;
+    int proctex_color_map_offset;
+    int proctex_alpha_map_offset;
+    int proctex_lut_offset;
+    int proctex_diff_lut_offset;
+    float proctex_bias;
+    int shadow_texture_bias;
+    bool enable_clip1;
+    ivec4 lighting_lut_offset[NUM_LIGHTING_SAMPLERS / 4];
+    vec3 fog_color;
+    vec2 proctex_noise_f;
+    vec2 proctex_noise_a;
+    vec2 proctex_noise_p;
+    vec3 lighting_global_ambient;
+    LightSrc light_src[NUM_LIGHTS];
+    vec4 const_color[NUM_TEV_STAGES];
+    vec4 tev_combiner_buffer_color;
+    vec3 tex_lod_bias;
+    vec4 clip_coef;
+}};
+)";
+
+std::string BuildShaderUniformDefinitions(const std::string& extra_layout_parameters) {
+    return fmt::format(UniformBlockDefFormat, extra_layout_parameters);
+}
+
+} // namespace Pica::Shader

src/video_core/shader/shader_uniforms.h

@@ -0,0 +1,101 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/vector_math.h"
+#include "video_core/regs_lighting.h"
+
+namespace Pica {
+struct ShaderRegs;
+}
+
+namespace Pica::Shader {
+
+struct ShaderSetup;
+
+enum class UniformBindings : u32 { Common, VS, GS };
+
+struct LightSrc {
+    alignas(16) Common::Vec3f specular_0;
+    alignas(16) Common::Vec3f specular_1;
+    alignas(16) Common::Vec3f diffuse;
+    alignas(16) Common::Vec3f ambient;
+    alignas(16) Common::Vec3f position;
+    alignas(16) Common::Vec3f spot_direction; // negated
+    float dist_atten_bias;
+    float dist_atten_scale;
+};
+
+/**
+ * Uniform structure for the Uniform Buffer Object, all vectors must be 16-byte aligned
+ * NOTE: Always keep a vec4 at the end. The GL spec is not clear wether the alignment at
+ *       the end of a uniform block is included in UNIFORM_BLOCK_DATA_SIZE or not.
+ *       Not following that rule will cause problems on some AMD drivers.
+ */
+struct UniformData {
+    int framebuffer_scale;
+    int alphatest_ref;
+    float depth_scale;
+    float depth_offset;
+    float shadow_bias_constant;
+    float shadow_bias_linear;
+    int scissor_x1;
+    int scissor_y1;
+    int scissor_x2;
+    int scissor_y2;
+    int fog_lut_offset;
+    int proctex_noise_lut_offset;
+    int proctex_color_map_offset;
+    int proctex_alpha_map_offset;
+    int proctex_lut_offset;
+    int proctex_diff_lut_offset;
+    float proctex_bias;
+    int shadow_texture_bias;
+    alignas(4) bool enable_clip1;
+    alignas(16) Common::Vec4i lighting_lut_offset[LightingRegs::NumLightingSampler / 4];
+    alignas(16) Common::Vec3f fog_color;
+    alignas(8) Common::Vec2f proctex_noise_f;
+    alignas(8) Common::Vec2f proctex_noise_a;
+    alignas(8) Common::Vec2f proctex_noise_p;
+    alignas(16) Common::Vec3f lighting_global_ambient;
+    LightSrc light_src[8];
+    alignas(16) Common::Vec4f const_color[6]; // A vec4 color for each of the six tev stages
+    alignas(16) Common::Vec4f tev_combiner_buffer_color;
+    alignas(16) Common::Vec3f tex_lod_bias;
+    alignas(16) Common::Vec4f clip_coef;
+};
+
+static_assert(sizeof(UniformData) == 0x500,
+              "The size of the UniformData does not match the structure in the shader");
+static_assert(sizeof(UniformData) < 16384,
+              "UniformData structure must be less than 16kb as per the OpenGL spec");
+
+/**
+ * Uniform struct for the Uniform Buffer Object that contains PICA vertex/geometry shader uniforms.
+ * NOTE: the same rule from UniformData also applies here.
+ */
+struct PicaUniformsData {
+    void SetFromRegs(const ShaderRegs& regs, const ShaderSetup& setup);
+
+    struct BoolAligned {
+        alignas(16) int b;
+    };
+
+    std::array<BoolAligned, 16> bools;
+    alignas(16) std::array<Common::Vec4u, 4> i;
+    alignas(16) std::array<Common::Vec4f, 96> f;
+};
+
+struct VSUniformData {
+    PicaUniformsData uniforms;
+};
+static_assert(sizeof(VSUniformData) == 1856,
+              "The size of the VSUniformData does not match the structure in the shader");
+static_assert(sizeof(VSUniformData) < 16384,
+              "VSUniformData structure must be less than 16kb as per the OpenGL spec");
+
+std::string BuildShaderUniformDefinitions(const std::string& extra_layout_parameters = "");
+
+} // namespace Pica::Shader

src/video_core/swrasterizer/swrasterizer.cpp

@@ -1,16 +0,0 @@
-// Copyright 2015 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "video_core/swrasterizer/clipper.h"
-#include "video_core/swrasterizer/swrasterizer.h"
-
-namespace VideoCore {
-
-void SWRasterizer::AddTriangle(const Pica::Shader::OutputVertex& v0,
-                               const Pica::Shader::OutputVertex& v1,
-                               const Pica::Shader::OutputVertex& v2) {
-    Pica::Clipper::ProcessTriangle(v0, v1, v2);
-}
-
-} // namespace VideoCore

src/video_core/vertex_loader.cpp

@@ -1,5 +1,4 @@
 #include <memory>
-#include <boost/range/algorithm/fill.hpp>
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/bit_field.h"
@@ -23,7 +22,7 @@ void VertexLoader::Setup(const PipelineRegs& regs) {
     const auto& attribute_config = regs.vertex_attributes;
     num_total_attributes = attribute_config.GetNumTotalAttributes();
 
-    boost::fill(vertex_attribute_sources, 0xdeadbeef);
+    vertex_attribute_sources.fill(0xdeadbeef);
 
     for (int i = 0; i < 16; i++) {
         vertex_attribute_is_default[i] = attribute_config.IsDefaultAttribute(i);

src/video_core/video_core.cpp

@@ -6,11 +6,13 @@
 #include "common/archives.h"
 #include "common/logging/log.h"
 #include "common/settings.h"
+#include "core/core.h"
 #include "video_core/pica.h"
 #include "video_core/pica_state.h"
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_opengl/gl_vars.h"
 #include "video_core/renderer_opengl/renderer_opengl.h"
+#include "video_core/renderer_software/renderer_software.h"
 #include "video_core/video_core.h"
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -18,9 +20,8 @@
 
 namespace VideoCore {
 
-std::unique_ptr<RendererBase> g_renderer; ///< Renderer plugin
+std::unique_ptr<RendererBase> g_renderer{}; ///< Renderer plugin
 
-std::atomic<bool> g_hw_renderer_enabled;
 std::atomic<bool> g_shader_jit_enabled;
 std::atomic<bool> g_hw_shader_enabled;
 std::atomic<bool> g_separable_shader_enabled;
@@ -30,65 +31,49 @@ std::atomic<bool> g_renderer_bg_color_update_requested;
 std::atomic<bool> g_renderer_sampler_update_requested;
 std::atomic<bool> g_renderer_shader_update_requested;
 std::atomic<bool> g_texture_filter_update_requested;
-// Screenshot
-std::atomic<bool> g_renderer_screenshot_requested;
-void* g_screenshot_bits;
-std::function<void()> g_screenshot_complete_callback;
-Layout::FramebufferLayout g_screenshot_framebuffer_layout;
 
 Memory::MemorySystem* g_memory;
 
 /// Initialize the video core
-ResultStatus Init(Frontend::EmuWindow& emu_window, Frontend::EmuWindow* secondary_window,
+void Init(Frontend::EmuWindow& emu_window, Frontend::EmuWindow* secondary_window,
-                  Memory::MemorySystem& memory) {
+          Core::System& system) {
-    g_memory = &memory;
+    g_memory = &system.Memory();
     Pica::Init();
 
+    const Settings::GraphicsAPI graphics_api = Settings::values.graphics_api.GetValue();
     OpenGL::GLES = Settings::values.use_gles.GetValue();
 
-    g_renderer = std::make_unique<OpenGL::RendererOpenGL>(emu_window, secondary_window);
+    switch (graphics_api) {
-    ResultStatus result = g_renderer->Init();
+    case Settings::GraphicsAPI::Software:
-
+        g_renderer = std::make_unique<VideoCore::RendererSoftware>(system, emu_window);
-    if (result != ResultStatus::Success) {
+        break;
-        LOG_ERROR(Render, "initialization failed !");
+    case Settings::GraphicsAPI::OpenGL:
-    } else {
+        g_renderer = std::make_unique<OpenGL::RendererOpenGL>(system, emu_window, secondary_window);
-        LOG_DEBUG(Render, "initialized OK");
+        break;
+    default:
+        LOG_CRITICAL(Render, "Unknown graphics API {}, using OpenGL", graphics_api);
+        g_renderer = std::make_unique<OpenGL::RendererOpenGL>(system, emu_window, secondary_window);
     }
-
-    return result;
 }
 
 /// Shutdown the video core
 void Shutdown() {
     Pica::Shutdown();
-
-    g_renderer->ShutDown();
     g_renderer.reset();
 
     LOG_DEBUG(Render, "shutdown OK");
 }
 
-void RequestScreenshot(void* data, std::function<void()> callback,
-                       const Layout::FramebufferLayout& layout) {
-    if (g_renderer_screenshot_requested) {
-        LOG_ERROR(Render, "A screenshot is already requested or in progress, ignoring the request");
-        return;
-    }
-    g_screenshot_bits = data;
-    g_screenshot_complete_callback = std::move(callback);
-    g_screenshot_framebuffer_layout = layout;
-    g_renderer_screenshot_requested = true;
-}
-
 u16 GetResolutionScaleFactor() {
-    if (g_hw_renderer_enabled) {
+    const auto graphics_api = Settings::values.graphics_api.GetValue();
-        return Settings::values.resolution_factor.GetValue()
+    if (graphics_api == Settings::GraphicsAPI::Software) {
-                   ? Settings::values.resolution_factor.GetValue()
-                   : g_renderer->GetRenderWindow().GetFramebufferLayout().GetScalingRatio();
-    } else {
         // Software renderer always render at native resolution
         return 1;
     }
+
+    return Settings::values.resolution_factor.GetValue()
+               ? Settings::values.resolution_factor.GetValue()
+               : g_renderer->GetRenderWindow().GetFramebufferLayout().GetScalingRatio();
 }
 
 template <class Archive>

src/video_core/video_core.h

@@ -14,7 +14,9 @@ namespace Frontend {
 class EmuWindow;
 }
 
-class RendererBase;
+namespace Core {
+class System;
+}
 
 namespace Memory {
 class MemorySystem;
@@ -25,11 +27,12 @@ class MemorySystem;
 
 namespace VideoCore {
 
+class RendererBase;
+
 extern std::unique_ptr<RendererBase> g_renderer; ///< Renderer plugin
 
 // TODO: Wrap these in a user settings struct along with any other graphics settings (often set from
 // qt ui)
-extern std::atomic<bool> g_hw_renderer_enabled;
 extern std::atomic<bool> g_shader_jit_enabled;
 extern std::atomic<bool> g_hw_shader_enabled;
 extern std::atomic<bool> g_separable_shader_enabled;
@@ -39,31 +42,16 @@ extern std::atomic<bool> g_renderer_bg_color_update_requested;
 extern std::atomic<bool> g_renderer_sampler_update_requested;
 extern std::atomic<bool> g_renderer_shader_update_requested;
 extern std::atomic<bool> g_texture_filter_update_requested;
-// Screenshot
-extern std::atomic<bool> g_renderer_screenshot_requested;
-extern void* g_screenshot_bits;
-extern std::function<void()> g_screenshot_complete_callback;
-extern Layout::FramebufferLayout g_screenshot_framebuffer_layout;
 
 extern Memory::MemorySystem* g_memory;
 
-enum class ResultStatus {
-    Success,
-    ErrorGenericDrivers,
-    ErrorBelowGL43,
-};
-
 /// Initialize the video core
-ResultStatus Init(Frontend::EmuWindow& emu_window, Frontend::EmuWindow* secondary_window,
+void Init(Frontend::EmuWindow& emu_window, Frontend::EmuWindow* secondary_window,
-                  Memory::MemorySystem& memory);
+          Core::System& system);
 
 /// Shutdown the video core
 void Shutdown();
 
-/// Request a screenshot of the next frame
-void RequestScreenshot(void* data, std::function<void()> callback,
-                       const Layout::FramebufferLayout& layout);
-
 u16 GetResolutionScaleFactor();
 
 template <class Archive>