Refactor software renderer (#6621)

src/citra/citra.cpp

@@ -344,11 +344,14 @@ int main(int argc, char** argv) {
         return -1;
     }
 
+    auto& system = Core::System::GetInstance();
+    auto& movie = Core::Movie::GetInstance();
+
     if (!movie_record.empty()) {
-        Core::Movie::GetInstance().PrepareForRecording();
+        movie.PrepareForRecording();
     }
     if (!movie_play.empty()) {
-        Core::Movie::GetInstance().PrepareForPlayback(movie_play);
+        movie.PrepareForPlayback(movie_play);
     }
 
     // Apply the command line arguments
@@ -361,13 +364,13 @@ int main(int argc, char** argv) {
 
     EmuWindow_SDL2::InitializeSDL2();
 
-    const auto create_emu_window = [](bool fullscreen,
+    const auto create_emu_window = [&](bool fullscreen,
-                                      bool is_secondary) -> std::unique_ptr<EmuWindow_SDL2> {
+                                       bool is_secondary) -> std::unique_ptr<EmuWindow_SDL2> {
         switch (Settings::values.graphics_api.GetValue()) {
         case Settings::GraphicsAPI::OpenGL:
             return std::make_unique<EmuWindow_SDL2_GL>(fullscreen, is_secondary);
         case Settings::GraphicsAPI::Software:
-            return std::make_unique<EmuWindow_SDL2_SW>(fullscreen, is_secondary);
+            return std::make_unique<EmuWindow_SDL2_SW>(system, fullscreen, is_secondary);
         }
         LOG_ERROR(Frontend, "Invalid Graphics API, using OpenGL");
         return std::make_unique<EmuWindow_SDL2_GL>(fullscreen, is_secondary);
@@ -385,7 +388,6 @@ int main(int argc, char** argv) {
              Common::g_scm_desc);
     Settings::LogSettings();
 
-    Core::System& system = Core::System::GetInstance();
     const Core::System::ResultStatus load_result{
         system.Load(*emu_window, filepath, secondary_window.get())};
 
@@ -437,21 +439,21 @@ int main(int argc, char** argv) {
     }
 
     if (!movie_play.empty()) {
-        auto metadata = Core::Movie::GetInstance().GetMovieMetadata(movie_play);
+        auto metadata = movie.GetMovieMetadata(movie_play);
         LOG_INFO(Movie, "Author: {}", metadata.author);
         LOG_INFO(Movie, "Rerecord count: {}", metadata.rerecord_count);
         LOG_INFO(Movie, "Input count: {}", metadata.input_count);
-        Core::Movie::GetInstance().StartPlayback(movie_play);
+        movie.StartPlayback(movie_play);
     }
     if (!movie_record.empty()) {
-        Core::Movie::GetInstance().StartRecording(movie_record, movie_record_author);
+        movie.StartRecording(movie_record, movie_record_author);
     }
     if (!dump_video.empty() && DynamicLibrary::FFmpeg::LoadFFmpeg()) {
-        Layout::FramebufferLayout layout{Layout::FrameLayoutFromResolutionScale(
+        const auto layout{
-            VideoCore::g_renderer->GetResolutionScaleFactor())};
+            Layout::FrameLayoutFromResolutionScale(system.Renderer().GetResolutionScaleFactor())};
         auto dumper = std::make_shared<VideoDumper::FFmpegBackend>();
         if (dumper->StartDumping(dump_video, layout)) {
-            Core::System::GetInstance().RegisterVideoDumper(dumper);
+            system.RegisterVideoDumper(dumper);
         }
     }
 
@@ -494,7 +496,7 @@ int main(int argc, char** argv) {
     main_render_thread.join();
     secondary_render_thread.join();
 
-    Core::Movie::GetInstance().Shutdown();
+    movie.Shutdown();
 
     auto video_dumper = system.GetVideoDumper();
     if (video_dumper && video_dumper->IsDumping()) {

src/citra/emu_window/emu_window_sdl2_sw.cpp

@@ -9,18 +9,16 @@
 #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 "common/settings.h"
+#include "core/core.h"
 #include "core/frontend/emu_window.h"
-#include "core/hw/gpu.h"
+#include "video_core/renderer_software/renderer_software.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_SW::EmuWindow_SDL2_SW(Core::System& system_, bool fullscreen, bool is_secondary)
-    : EmuWindow_SDL2{is_secondary} {
+    : EmuWindow_SDL2{is_secondary}, system{system_} {
     std::string window_title = fmt::format("Citra {} | {}-{}", Common::g_build_fullname,
                                            Common::g_scm_branch, Common::g_scm_desc);
     render_window =
@@ -67,6 +65,8 @@ void EmuWindow_SDL2_SW::Present() {
     const auto layout{Layout::DefaultFrameLayout(
         Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight, false, false)};
 
+    using VideoCore::ScreenId;
+
     while (IsOpen()) {
         SDL_SetRenderDrawColor(renderer,
                                static_cast<Uint8>(Settings::values.bg_red.GetValue() * 255),
@@ -74,62 +74,34 @@ void EmuWindow_SDL2_SW::Present() {
                                static_cast<Uint8>(Settings::values.bg_blue.GetValue() * 255), 0xFF);
         SDL_RenderClear(renderer);
 
-        const auto draw_screen = [&](int fb_id) {
+        const auto draw_screen = [&](ScreenId screen_id) {
-            const auto dst_rect = fb_id == 0 ? layout.top_screen : layout.bottom_screen;
+            const auto dst_rect =
+                screen_id == ScreenId::TopLeft ? 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_Surface* screen = LoadFramebuffer(screen_id);
             SDL_BlitSurface(screen, nullptr, window_surface, &sdl_rect);
             SDL_FreeSurface(screen);
         };
 
-        draw_screen(0);
+        draw_screen(ScreenId::TopLeft);
-        draw_screen(1);
+        draw_screen(ScreenId::Bottom);
 
         SDL_RenderPresent(renderer);
         SDL_UpdateWindowSurface(render_window);
     }
 }
 
-SDL_Surface* EmuWindow_SDL2_SW::LoadFramebuffer(int fb_id) {
+SDL_Surface* EmuWindow_SDL2_SW::LoadFramebuffer(VideoCore::ScreenId screen_id) {
-    const auto& framebuffer = GPU::g_regs.framebuffer_config[fb_id];
+    const auto& renderer = static_cast<SwRenderer::RendererSoftware&>(system.Renderer());
-    const PAddr framebuffer_addr =
+    const auto& info = renderer.Screen(screen_id);
-        framebuffer.active_fb == 0 ? framebuffer.address_left1 : framebuffer.address_left2;
+    const int width = static_cast<int>(info.width);
-
+    const int height = static_cast<int>(info.height);
-    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++) {
+    std::memcpy(surface->pixels, info.pixels.data(), info.pixels.size());
-        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);
-                }
-                UNREACHABLE();
-            }();
-
-            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

@@ -10,9 +10,17 @@
 struct SDL_Renderer;
 struct SDL_Surface;
 
+namespace VideoCore {
+enum class ScreenId : u32;
+}
+
+namespace Core {
+class System;
+}
+
 class EmuWindow_SDL2_SW : public EmuWindow_SDL2 {
 public:
-    explicit EmuWindow_SDL2_SW(bool fullscreen, bool is_secondary);
+    explicit EmuWindow_SDL2_SW(Core::System& system, bool fullscreen, bool is_secondary);
     ~EmuWindow_SDL2_SW();
 
     void Present() override;
@@ -22,7 +30,10 @@ public:
 
 private:
     /// Loads a framebuffer to an SDL surface
-    SDL_Surface* LoadFramebuffer(int fb_id);
+    SDL_Surface* LoadFramebuffer(VideoCore::ScreenId screen_id);
+
+    /// The system class.
+    Core::System& system;
 
     /// The SDL software renderer
     SDL_Renderer* renderer;

src/citra_qt/bootmanager.cpp

@@ -25,6 +25,7 @@
 #include "input_common/motion_emu.h"
 #include "video_core/custom_textures/custom_tex_manager.h"
 #include "video_core/renderer_base.h"
+#include "video_core/renderer_software/renderer_software.h"
 #include "video_core/video_core.h"
 
 #ifdef HAS_OPENGL
@@ -288,7 +289,8 @@ private:
 #endif
 
 struct SoftwareRenderWidget : public RenderWidget {
-    explicit SoftwareRenderWidget(GRenderWindow* parent) : RenderWidget(parent) {}
+    explicit SoftwareRenderWidget(GRenderWindow* parent, Core::System& system_)
+        : RenderWidget(parent), system(system_) {}
 
     void Present() override {
         if (!isVisible()) {
@@ -298,61 +300,40 @@ struct SoftwareRenderWidget : public RenderWidget {
             return;
         }
 
+        using VideoCore::ScreenId;
+
         const auto layout{Layout::DefaultFrameLayout(width(), height(), false, false)};
         QPainter painter(this);
 
-        const auto draw_screen = [&](int fb_id) {
+        const auto draw_screen = [&](ScreenId screen_id) {
-            const auto rect = fb_id == 0 ? layout.top_screen : layout.bottom_screen;
+            const auto rect =
-            const QImage screen = LoadFramebuffer(fb_id).scaled(rect.GetWidth(), rect.GetHeight());
+                screen_id == ScreenId::TopLeft ? layout.top_screen : layout.bottom_screen;
+            const QImage screen =
+                LoadFramebuffer(screen_id).scaled(rect.GetWidth(), rect.GetHeight());
             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(ScreenId::TopLeft);
-        draw_screen(1);
+        draw_screen(ScreenId::Bottom);
 
         painter.end();
     }
 
-    QImage LoadFramebuffer(int fb_id) {
+    QImage LoadFramebuffer(VideoCore::ScreenId screen_id) {
-        const auto& framebuffer = GPU::g_regs.framebuffer_config[fb_id];
+        const auto& renderer = static_cast<SwRenderer::RendererSoftware&>(system.Renderer());
-        const PAddr framebuffer_addr =
+        const auto& info = renderer.Screen(screen_id);
-            framebuffer.active_fb == 0 ? framebuffer.address_left1 : framebuffer.address_left2;
+        const int width = static_cast<int>(info.width);
-
+        const int height = static_cast<int>(info.height);
-        Memory::RasterizerFlushRegion(framebuffer_addr, framebuffer.stride * framebuffer.height);
+        QImage image{height, width, QImage::Format_RGBA8888};
-        const u8* framebuffer_data = VideoCore::g_memory->GetPhysicalPointer(framebuffer_addr);
+        std::memcpy(image.bits(), info.pixels.data(), info.pixels.size());
-
-        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);
-                    }
-                    UNREACHABLE();
-                }();
-
-                image.setPixel(x, y, qRgba(color.r(), color.g(), color.b(), color.a()));
-            }
-        }
         return image;
     }
+
+private:
+    Core::System& system;
 };
 
 static Frontend::WindowSystemType GetWindowSystemType() {
@@ -401,8 +382,9 @@ static Frontend::EmuWindow::WindowSystemInfo GetWindowSystemInfo(QWindow* window
 
 std::unique_ptr<Frontend::GraphicsContext> GRenderWindow::main_context;
 
-GRenderWindow::GRenderWindow(QWidget* parent_, EmuThread* emu_thread, bool is_secondary_)
+GRenderWindow::GRenderWindow(QWidget* parent_, EmuThread* emu_thread_, Core::System& system_,
-    : QWidget(parent_), EmuWindow(is_secondary_), emu_thread(emu_thread) {
+                             bool is_secondary_)
+    : QWidget(parent_), EmuWindow(is_secondary_), emu_thread(emu_thread_), system{system_} {
 
     setWindowTitle(QStringLiteral("Citra %1 | %2-%3")
                        .arg(QString::fromUtf8(Common::g_build_name),
@@ -652,12 +634,12 @@ void GRenderWindow::ReleaseRenderTarget() {
 
 void GRenderWindow::CaptureScreenshot(u32 res_scale, const QString& screenshot_path) {
     if (res_scale == 0) {
-        res_scale = VideoCore::g_renderer->GetResolutionScaleFactor();
+        res_scale = system.Renderer().GetResolutionScaleFactor();
     }
 
     const auto layout{Layout::FrameLayoutFromResolutionScale(res_scale, is_secondary)};
     screenshot_image = QImage(QSize(layout.width, layout.height), QImage::Format_RGB32);
-    VideoCore::g_renderer->RequestScreenshot(
+    system.Renderer().RequestScreenshot(
         screenshot_image.bits(),
         [this, screenshot_path] {
             const std::string std_screenshot_path = screenshot_path.toStdString();
@@ -708,7 +690,7 @@ bool GRenderWindow::InitializeOpenGL() {
 }
 
 void GRenderWindow::InitializeSoftware() {
-    child_widget = new SoftwareRenderWidget(this);
+    child_widget = new SoftwareRenderWidget(this, system);
     main_context = std::make_unique<DummyContext>();
 }
 

src/citra_qt/bootmanager.h

@@ -112,7 +112,7 @@ class GRenderWindow : public QWidget, public Frontend::EmuWindow {
     Q_OBJECT
 
 public:
-    GRenderWindow(QWidget* parent, EmuThread* emu_thread, bool is_secondary);
+    GRenderWindow(QWidget* parent, EmuThread* emu_thread, Core::System& system, bool is_secondary);
     ~GRenderWindow() override;
 
     // EmuWindow implementation.
@@ -188,6 +188,7 @@ private:
     QWidget* child_widget = nullptr;
 
     EmuThread* emu_thread;
+    Core::System& system;
 
     /// 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

src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp

@@ -550,8 +550,8 @@ void GraphicsVertexShaderWidget::OnResumed() {
 }
 
 void GraphicsVertexShaderWidget::OnInputAttributeChanged(int index) {
-    float value = input_data[index]->text().toFloat();
+    const f32 value = input_data[index]->text().toFloat();
-    input_vertex.attr[index / 4][index % 4] = Pica::float24::FromFloat32(value);
+    input_vertex.attr[index / 4][index % 4] = Pica::f24::FromFloat32(value);
     // Re-execute shader with updated value
     Reload();
 }

src/citra_qt/main.cpp

@@ -297,8 +297,8 @@ void GMainWindow::InitializeWidgets() {
 #ifdef CITRA_ENABLE_COMPATIBILITY_REPORTING
     ui->action_Report_Compatibility->setVisible(true);
 #endif
-    render_window = new GRenderWindow(this, emu_thread.get(), false);
+    render_window = new GRenderWindow(this, emu_thread.get(), system, false);
-    secondary_window = new GRenderWindow(this, emu_thread.get(), true);
+    secondary_window = new GRenderWindow(this, emu_thread.get(), system, true);
     render_window->hide();
     secondary_window->hide();
     secondary_window->setParent(nullptr);

src/tests/video_core/shader/shader_jit_x64_compiler.cpp

@@ -14,7 +14,6 @@
 #include "video_core/shader/shader_interpreter.h"
 #include "video_core/shader/shader_jit_x64_compiler.h"
 
-using float24 = Pica::float24;
 using JitShader = Pica::Shader::JitShader;
 using ShaderInterpreter = Pica::Shader::InterpreterEngine;
 
@@ -51,14 +50,14 @@ public:
     }
 
     void RunJit(Pica::Shader::UnitState& shader_unit, float input) {
-        shader_unit.registers.input[0].x = float24::FromFloat32(input);
+        shader_unit.registers.input[0].x = Pica::f24::FromFloat32(input);
-        shader_unit.registers.temporary[0].x = float24::FromFloat32(0);
+        shader_unit.registers.temporary[0].x = Pica::f24::Zero();
         shader_jit.Run(*shader_setup, shader_unit, 0);
     }
 
     void RunInterpreter(Pica::Shader::UnitState& shader_unit, float input) {
-        shader_unit.registers.input[0].x = float24::FromFloat32(input);
+        shader_unit.registers.input[0].x = Pica::f24::FromFloat32(input);
-        shader_unit.registers.temporary[0].x = float24::FromFloat32(0);
+        shader_unit.registers.temporary[0].x = Pica::f24::Zero();
         shader_interpreter.Run(*shader_setup, shader_unit);
     }
 

src/video_core/CMakeLists.txt

@@ -83,8 +83,6 @@ add_library(video_core STATIC
     renderer_opengl/post_processing_opengl.h
     renderer_opengl/renderer_opengl.cpp
     renderer_opengl/renderer_opengl.h
-    renderer_software/rasterizer.cpp
-    renderer_software/rasterizer.h
     renderer_software/renderer_software.cpp
     renderer_software/renderer_software.h
     renderer_software/sw_clipper.cpp

src/video_core/command_processor.cpp

@@ -90,16 +90,16 @@ static void WriteUniformFloatReg(ShaderRegs& config, Shader::ShaderSetup& setup,
                 for (auto i : {0, 1, 2, 3}) {
                     float buffer_value;
                     std::memcpy(&buffer_value, &uniform_write_buffer[i], sizeof(float));
-                    uniform[3 - i] = float24::FromFloat32(buffer_value);
+                    uniform[3 - i] = f24::FromFloat32(buffer_value);
                 }
             } else {
                 // TODO: Untested
-                uniform.w = float24::FromRaw(uniform_write_buffer[0] >> 8);
+                uniform.w = f24::FromRaw(uniform_write_buffer[0] >> 8);
-                uniform.z = float24::FromRaw(((uniform_write_buffer[0] & 0xFF) << 16) |
+                uniform.z = f24::FromRaw(((uniform_write_buffer[0] & 0xFF) << 16) |
-                                             ((uniform_write_buffer[1] >> 16) & 0xFFFF));
+                                         ((uniform_write_buffer[1] >> 16) & 0xFFFF));
-                uniform.y = float24::FromRaw(((uniform_write_buffer[1] & 0xFFFF) << 8) |
+                uniform.y = f24::FromRaw(((uniform_write_buffer[1] & 0xFFFF) << 8) |
-                                             ((uniform_write_buffer[2] >> 24) & 0xFF));
+                                         ((uniform_write_buffer[2] >> 24) & 0xFF));
-                uniform.x = float24::FromRaw(uniform_write_buffer[2] & 0xFFFFFF);
+                uniform.x = f24::FromRaw(uniform_write_buffer[2] & 0xFFFFFF);
             }
 
             LOG_TRACE(HW_GPU, "Set {} float uniform {:x} to ({} {} {} {})",
@@ -182,15 +182,15 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
                 break;
             }
 
-            Common::Vec4<float24> attribute;
+            Common::Vec4<f24> attribute;
 
             // NOTE: The destination component order indeed is "backwards"
-            attribute.w = float24::FromRaw(g_state.default_attr_write_buffer[0] >> 8);
+            attribute.w = f24::FromRaw(g_state.default_attr_write_buffer[0] >> 8);
-            attribute.z = float24::FromRaw(((g_state.default_attr_write_buffer[0] & 0xFF) << 16) |
+            attribute.z = f24::FromRaw(((g_state.default_attr_write_buffer[0] & 0xFF) << 16) |
-                                           ((g_state.default_attr_write_buffer[1] >> 16) & 0xFFFF));
+                                       ((g_state.default_attr_write_buffer[1] >> 16) & 0xFFFF));
-            attribute.y = float24::FromRaw(((g_state.default_attr_write_buffer[1] & 0xFFFF) << 8) |
+            attribute.y = f24::FromRaw(((g_state.default_attr_write_buffer[1] & 0xFFFF) << 8) |
-                                           ((g_state.default_attr_write_buffer[2] >> 24) & 0xFF));
+                                       ((g_state.default_attr_write_buffer[2] >> 24) & 0xFF));
-            attribute.x = float24::FromRaw(g_state.default_attr_write_buffer[2] & 0xFFFFFF);
+            attribute.x = f24::FromRaw(g_state.default_attr_write_buffer[2] & 0xFFFFFF);
 
             LOG_TRACE(HW_GPU, "Set default VS attribute {:x} to ({} {} {} {})", (int)setup.index,
                       attribute.x.ToFloat32(), attribute.y.ToFloat32(), attribute.z.ToFloat32(),

src/video_core/geometry_pipeline.cpp

@@ -85,8 +85,8 @@ private:
     const Regs& regs;
     Shader::GSUnitState& unit;
     Shader::AttributeBuffer attribute_buffer;
-    Common::Vec4<float24>* buffer_cur;
+    Common::Vec4<f24>* buffer_cur;
-    Common::Vec4<float24>* buffer_end;
+    Common::Vec4<f24>* buffer_end;
     unsigned int vs_output_num;
 
     GeometryPipeline_Point() : regs(g_state.regs), unit(g_state.gs_unit) {}
@@ -146,7 +146,7 @@ public:
         DEBUG_ASSERT(need_index);
 
         // The number of vertex input is put to the uniform register
-        float24 vertex_num = float24::FromFloat32(static_cast<float>(val));
+        f24 vertex_num = f24::FromFloat32(static_cast<float>(val));
         setup.uniforms.f[0] = Common::MakeVec(vertex_num, vertex_num, vertex_num, vertex_num);
 
         // The second uniform register and so on are used for receiving input vertices
@@ -183,7 +183,7 @@ private:
     Shader::ShaderSetup& setup;
     unsigned int main_vertex_num;
     unsigned int total_vertex_num;
-    Common::Vec4<float24>* buffer_cur;
+    Common::Vec4<f24>* buffer_cur;
     unsigned int vs_output_num;
 
     GeometryPipeline_VariablePrimitive() : regs(g_state.regs), setup(g_state.gs) {}
@@ -257,9 +257,9 @@ public:
 private:
     [[maybe_unused]] const Regs& regs;
     Shader::ShaderSetup& setup;
-    Common::Vec4<float24>* buffer_begin;
+    Common::Vec4<f24>* buffer_begin;
-    Common::Vec4<float24>* buffer_cur;
+    Common::Vec4<f24>* buffer_cur;
-    Common::Vec4<float24>* buffer_end;
+    Common::Vec4<f24>* buffer_end;
     unsigned int vs_output_num;
 
     GeometryPipeline_FixedPrimitive() : regs(g_state.regs), setup(g_state.gs) {}

src/video_core/pica_types.h

@@ -25,20 +25,20 @@ namespace Pica {
 template <unsigned M, unsigned E>
 struct Float {
 public:
-    static Float<M, E> FromFloat32(float val) {
+    static constexpr Float<M, E> FromFloat32(float val) {
         Float<M, E> ret;
         ret.value = val;
         return ret;
     }
 
-    static Float<M, E> FromRaw(u32 hex) {
+    static constexpr Float<M, E> FromRaw(u32 hex) {
         Float<M, E> res;
 
-        const int width = M + E + 1;
+        const s32 width = M + E + 1;
-        const int bias = 128 - (1 << (E - 1));
+        const s32 bias = 128 - (1 << (E - 1));
-        int exponent = (hex >> M) & ((1 << E) - 1);
+        s32 exponent = (hex >> M) & ((1 << E) - 1);
-        const unsigned mantissa = hex & ((1 << M) - 1);
+        const u32 mantissa = hex & ((1 << M) - 1);
-        const unsigned sign = (hex >> (E + M)) << 31;
+        const u32 sign = (hex >> (E + M)) << 31;
 
         if (hex & ((1 << (width - 1)) - 1)) {
             if (exponent == (1 << E) - 1)
@@ -55,16 +55,20 @@ public:
         return res;
     }
 
-    static Float<M, E> Zero() {
+    static constexpr Float<M, E> Zero() {
         return FromFloat32(0.f);
     }
 
+    static constexpr Float<M, E> One() {
+        return FromFloat32(1.f);
+    }
+
     // Not recommended for anything but logging
-    float ToFloat32() const {
+    constexpr float ToFloat32() const {
         return value;
     }
 
-    Float<M, E> operator*(const Float<M, E>& flt) const {
+    constexpr Float<M, E> operator*(const Float<M, E>& flt) const {
         float result = value * flt.ToFloat32();
         // PICA gives 0 instead of NaN when multiplying by inf
         if (std::isnan(result))
@@ -73,70 +77,70 @@ public:
         return Float<M, E>::FromFloat32(result);
     }
 
-    Float<M, E> operator/(const Float<M, E>& flt) const {
+    constexpr Float<M, E> operator/(const Float<M, E>& flt) const {
         return Float<M, E>::FromFloat32(ToFloat32() / flt.ToFloat32());
     }
 
-    Float<M, E> operator+(const Float<M, E>& flt) const {
+    constexpr Float<M, E> operator+(const Float<M, E>& flt) const {
         return Float<M, E>::FromFloat32(ToFloat32() + flt.ToFloat32());
     }
 
-    Float<M, E> operator-(const Float<M, E>& flt) const {
+    constexpr Float<M, E> operator-(const Float<M, E>& flt) const {
         return Float<M, E>::FromFloat32(ToFloat32() - flt.ToFloat32());
     }
 
-    Float<M, E>& operator*=(const Float<M, E>& flt) {
+    constexpr Float<M, E>& operator*=(const Float<M, E>& flt) {
         value = operator*(flt).value;
         return *this;
     }
 
-    Float<M, E>& operator/=(const Float<M, E>& flt) {
+    constexpr Float<M, E>& operator/=(const Float<M, E>& flt) {
         value /= flt.ToFloat32();
         return *this;
     }
 
-    Float<M, E>& operator+=(const Float<M, E>& flt) {
+    constexpr Float<M, E>& operator+=(const Float<M, E>& flt) {
         value += flt.ToFloat32();
         return *this;
     }
 
-    Float<M, E>& operator-=(const Float<M, E>& flt) {
+    constexpr Float<M, E>& operator-=(const Float<M, E>& flt) {
         value -= flt.ToFloat32();
         return *this;
     }
 
-    Float<M, E> operator-() const {
+    constexpr Float<M, E> operator-() const {
         return Float<M, E>::FromFloat32(-ToFloat32());
     }
 
-    bool operator<(const Float<M, E>& flt) const {
+    constexpr bool operator<(const Float<M, E>& flt) const {
         return ToFloat32() < flt.ToFloat32();
     }
 
-    bool operator>(const Float<M, E>& flt) const {
+    constexpr bool operator>(const Float<M, E>& flt) const {
         return ToFloat32() > flt.ToFloat32();
     }
 
-    bool operator>=(const Float<M, E>& flt) const {
+    constexpr bool operator>=(const Float<M, E>& flt) const {
         return ToFloat32() >= flt.ToFloat32();
     }
 
-    bool operator<=(const Float<M, E>& flt) const {
+    constexpr bool operator<=(const Float<M, E>& flt) const {
         return ToFloat32() <= flt.ToFloat32();
     }
 
-    bool operator==(const Float<M, E>& flt) const {
+    constexpr bool operator==(const Float<M, E>& flt) const {
         return ToFloat32() == flt.ToFloat32();
     }
 
-    bool operator!=(const Float<M, E>& flt) const {
+    constexpr bool operator!=(const Float<M, E>& flt) const {
         return ToFloat32() != flt.ToFloat32();
     }
 
 private:
-    static const unsigned MASK = (1 << (M + E + 1)) - 1;
+    static constexpr u32 MASK = (1 << (M + E + 1)) - 1;
-    static const unsigned MANTISSA_MASK = (1 << M) - 1;
+    static constexpr u32 MANTISSA_MASK = (1 << M) - 1;
-    static const unsigned EXPONENT_MASK = (1 << E) - 1;
+    static constexpr u32 EXPONENT_MASK = (1 << E) - 1;
 
     // Stored as a regular float, merely for convenience
     // TODO: Perform proper arithmetic on this!
@@ -149,8 +153,8 @@ private:
     }
 };
 
-using float24 = Float<16, 7>;
+using f24 = Pica::Float<16, 7>;
-using float20 = Float<12, 7>;
+using f20 = Pica::Float<12, 7>;
-using float16 = Float<10, 5>;
+using f16 = Pica::Float<10, 5>;
 
 } // namespace Pica

src/video_core/rasterizer_accelerated.cpp

@@ -10,6 +10,8 @@
 
 namespace VideoCore {
 
+using Pica::f24;
+
 static Common::Vec4f ColorRGBA8(const u32 color) {
     const auto rgba =
         Common::Vec4u{color >> 0 & 0xFF, color >> 8 & 0xFF, color >> 16 & 0xFF, color >> 24 & 0xFF};
@@ -73,7 +75,7 @@ RasterizerAccelerated::RasterizerAccelerated(Memory::MemorySystem& memory_)
  * 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) {
+static bool AreQuaternionsOpposite(Common::Vec4<f24> qa, Common::Vec4<f24> 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()};
 
@@ -612,7 +614,7 @@ void RasterizerAccelerated::NotifyPicaRegisterChanged(u32 id) {
 }
 
 void RasterizerAccelerated::SyncDepthScale() {
-    float depth_scale = Pica::float24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32();
+    const f32 depth_scale = f24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32();
 
     if (depth_scale != uniform_block_data.data.depth_scale) {
         uniform_block_data.data.depth_scale = depth_scale;
@@ -621,8 +623,7 @@ void RasterizerAccelerated::SyncDepthScale() {
 }
 
 void RasterizerAccelerated::SyncDepthOffset() {
-    float depth_offset =
+    const f32 depth_offset = f24::FromRaw(regs.rasterizer.viewport_depth_near_plane).ToFloat32();
-        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;
@@ -646,16 +647,16 @@ void RasterizerAccelerated::SyncFogColor() {
 
 void RasterizerAccelerated::SyncProcTexNoise() {
     const Common::Vec2f proctex_noise_f = {
-        Pica::float16::FromRaw(regs.texturing.proctex_noise_frequency.u).ToFloat32(),
+        Pica::f16::FromRaw(regs.texturing.proctex_noise_frequency.u).ToFloat32(),
-        Pica::float16::FromRaw(regs.texturing.proctex_noise_frequency.v).ToFloat32(),
+        Pica::f16::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::f16::FromRaw(regs.texturing.proctex_noise_u.phase).ToFloat32(),
-        Pica::float16::FromRaw(regs.texturing.proctex_noise_v.phase).ToFloat32(),
+        Pica::f16::FromRaw(regs.texturing.proctex_noise_v.phase).ToFloat32(),
     };
 
     if (proctex_noise_f != uniform_block_data.data.proctex_noise_f ||
@@ -669,8 +670,8 @@ void RasterizerAccelerated::SyncProcTexNoise() {
 }
 
 void RasterizerAccelerated::SyncProcTexBias() {
-    const auto proctex_bias = Pica::float16::FromRaw(regs.texturing.proctex.bias_low |
+    const auto proctex_bias = Pica::f16::FromRaw(regs.texturing.proctex.bias_low |
-                                                     (regs.texturing.proctex_lut.bias_high << 8))
+                                                 (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;
@@ -687,7 +688,7 @@ void RasterizerAccelerated::SyncAlphaTest() {
 }
 
 void RasterizerAccelerated::SyncCombinerColor() {
-    auto combiner_color = ColorRGBA8(regs.texturing.tev_combiner_buffer_color.raw);
+    const 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;
@@ -695,7 +696,7 @@ void RasterizerAccelerated::SyncCombinerColor() {
 }
 
 void RasterizerAccelerated::SyncTevConstColor(
-    std::size_t stage_index, const Pica::TexturingRegs::TevStageConfig& tev_stage) {
+    const 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]) {
@@ -707,7 +708,7 @@ void RasterizerAccelerated::SyncTevConstColor(
 }
 
 void RasterizerAccelerated::SyncGlobalAmbient() {
-    auto color = LightColor(regs.lighting.global_ambient);
+    const 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;
@@ -715,7 +716,7 @@ void RasterizerAccelerated::SyncGlobalAmbient() {
 }
 
 void RasterizerAccelerated::SyncLightSpecular0(int light_index) {
-    auto color = LightColor(regs.lighting.light[light_index].specular_0);
+    const 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;
@@ -723,7 +724,7 @@ void RasterizerAccelerated::SyncLightSpecular0(int light_index) {
 }
 
 void RasterizerAccelerated::SyncLightSpecular1(int light_index) {
-    auto color = LightColor(regs.lighting.light[light_index].specular_1);
+    const 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;
@@ -731,7 +732,7 @@ void RasterizerAccelerated::SyncLightSpecular1(int light_index) {
 }
 
 void RasterizerAccelerated::SyncLightDiffuse(int light_index) {
-    auto color = LightColor(regs.lighting.light[light_index].diffuse);
+    const 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;
@@ -739,7 +740,7 @@ void RasterizerAccelerated::SyncLightDiffuse(int light_index) {
 }
 
 void RasterizerAccelerated::SyncLightAmbient(int light_index) {
-    auto color = LightColor(regs.lighting.light[light_index].ambient);
+    const 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;
@@ -748,9 +749,9 @@ void RasterizerAccelerated::SyncLightAmbient(int light_index) {
 
 void RasterizerAccelerated::SyncLightPosition(int light_index) {
     const Common::Vec3f position = {
-        Pica::float16::FromRaw(regs.lighting.light[light_index].x).ToFloat32(),
+        Pica::f16::FromRaw(regs.lighting.light[light_index].x).ToFloat32(),
-        Pica::float16::FromRaw(regs.lighting.light[light_index].y).ToFloat32(),
+        Pica::f16::FromRaw(regs.lighting.light[light_index].y).ToFloat32(),
-        Pica::float16::FromRaw(regs.lighting.light[light_index].z).ToFloat32(),
+        Pica::f16::FromRaw(regs.lighting.light[light_index].z).ToFloat32(),
     };
 
     if (position != uniform_block_data.data.light_src[light_index].position) {
@@ -771,8 +772,8 @@ void RasterizerAccelerated::SyncLightSpotDirection(int light_index) {
 }
 
 void RasterizerAccelerated::SyncLightDistanceAttenuationBias(int light_index) {
-    float dist_atten_bias =
+    const f32 dist_atten_bias =
-        Pica::float20::FromRaw(regs.lighting.light[light_index].dist_atten_bias).ToFloat32();
+        Pica::f20::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;
@@ -781,8 +782,8 @@ void RasterizerAccelerated::SyncLightDistanceAttenuationBias(int light_index) {
 }
 
 void RasterizerAccelerated::SyncLightDistanceAttenuationScale(int light_index) {
-    float dist_atten_scale =
+    const f32 dist_atten_scale =
-        Pica::float20::FromRaw(regs.lighting.light[light_index].dist_atten_scale).ToFloat32();
+        Pica::f20::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;
@@ -792,8 +793,8 @@ void RasterizerAccelerated::SyncLightDistanceAttenuationScale(int light_index) {
 
 void RasterizerAccelerated::SyncShadowBias() {
     const auto& shadow = regs.framebuffer.shadow;
-    float constant = Pica::float16::FromRaw(shadow.constant).ToFloat32();
+    const f32 constant = Pica::f16::FromRaw(shadow.constant).ToFloat32();
-    float linear = Pica::float16::FromRaw(shadow.linear).ToFloat32();
+    const f32 linear = Pica::f16::FromRaw(shadow.linear).ToFloat32();
 
     if (constant != uniform_block_data.data.shadow_bias_constant ||
         linear != uniform_block_data.data.shadow_bias_linear) {
@@ -804,7 +805,7 @@ void RasterizerAccelerated::SyncShadowBias() {
 }
 
 void RasterizerAccelerated::SyncShadowTextureBias() {
-    int bias = regs.texturing.shadow.bias << 1;
+    const s32 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;
@@ -813,7 +814,7 @@ void RasterizerAccelerated::SyncShadowTextureBias() {
 
 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;
+    const f32 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;

src/video_core/regs_rasterizer.h

@@ -37,9 +37,9 @@ struct RasterizerRegs {
     BitField<0, 1, u32> clip_enable;
     BitField<0, 24, u32> clip_coef[4]; // float24
 
-    Common::Vec4<float24> GetClipCoef() const {
+    Common::Vec4<f24> GetClipCoef() const {
-        return {float24::FromRaw(clip_coef[0]), float24::FromRaw(clip_coef[1]),
+        return {f24::FromRaw(clip_coef[0]), f24::FromRaw(clip_coef[1]), f24::FromRaw(clip_coef[2]),
-                float24::FromRaw(clip_coef[2]), float24::FromRaw(clip_coef[3])};
+                f24::FromRaw(clip_coef[3])};
     }
 
     Common::Rectangle<s32> GetViewportRect() const {
@@ -47,9 +47,9 @@ struct RasterizerRegs {
             // These registers hold half-width and half-height, so must be multiplied by 2
             viewport_corner.x,  // left
             viewport_corner.y + // top
-                static_cast<s32>(float24::FromRaw(viewport_size_y).ToFloat32() * 2),
+                static_cast<s32>(f24::FromRaw(viewport_size_y).ToFloat32() * 2),
             viewport_corner.x + // right
-                static_cast<s32>(float24::FromRaw(viewport_size_x).ToFloat32() * 2),
+                static_cast<s32>(f24::FromRaw(viewport_size_x).ToFloat32() * 2),
             viewport_corner.y // bottom
         };
     }

src/video_core/renderer_base.h

@@ -18,6 +18,12 @@ class System;
 
 namespace VideoCore {
 
+enum class ScreenId : u32 {
+    TopLeft,
+    TopRight,
+    Bottom,
+};
+
 struct RendererSettings {
     // Screenshot
     std::atomic_bool screenshot_requested{false};
@@ -75,7 +81,7 @@ public:
         return current_fps;
     }
 
-    int GetCurrentFrame() const {
+    s32 GetCurrentFrame() const {
         return current_frame;
     }
 
@@ -108,7 +114,7 @@ protected:
     Frontend::EmuWindow& render_window;    ///< Reference to the render window handle.
     Frontend::EmuWindow* secondary_window; ///< Reference to the secondary render window handle.
     f32 current_fps = 0.0f;                ///< Current framerate, should be set by the renderer
-    int current_frame = 0;                 ///< Current frame, should be set by the renderer
+    s32 current_frame = 0;                 ///< Current frame, should be set by the renderer
 };
 
 } // namespace VideoCore

src/video_core/renderer_software/rasterizer.cpp

@@ -1,901 +0,0 @@
-// Copyright 2014 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <array>
-#include <cmath>
-#include <tuple>
-#include "common/assert.h"
-#include "common/bit_field.h"
-#include "common/color.h"
-#include "common/common_types.h"
-#include "common/logging/log.h"
-#include "common/microprofile.h"
-#include "common/quaternion.h"
-#include "common/vector_math.h"
-#include "core/hw/gpu.h"
-#include "core/memory.h"
-#include "video_core/debug_utils/debug_utils.h"
-#include "video_core/pica_state.h"
-#include "video_core/pica_types.h"
-#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/texture/texture_decode.h"
-#include "video_core/utils.h"
-#include "video_core/video_core.h"
-
-namespace Pica::Rasterizer {
-
-// NOTE: Assuming that rasterizer coordinates are 12.4 fixed-point values
-struct Fix12P4 {
-    Fix12P4() {}
-    Fix12P4(u16 val) : val(val) {}
-
-    static u16 FracMask() {
-        return 0xF;
-    }
-    static u16 IntMask() {
-        return (u16)~0xF;
-    }
-
-    operator u16() const {
-        return val;
-    }
-
-    bool operator<(const Fix12P4& oth) const {
-        return (u16) * this < (u16)oth;
-    }
-
-private:
-    u16 val;
-};
-
-/**
- * Calculate signed area of the triangle spanned by the three argument vertices.
- * The sign denotes an orientation.
- *
- * @todo define orientation concretely.
- */
-static int SignedArea(const Common::Vec2<Fix12P4>& vtx1, const Common::Vec2<Fix12P4>& vtx2,
-                      const Common::Vec2<Fix12P4>& vtx3) {
-    const auto vec1 = Common::MakeVec(vtx2 - vtx1, 0);
-    const auto vec2 = Common::MakeVec(vtx3 - vtx1, 0);
-    // TODO: There is a very small chance this will overflow for sizeof(int) == 4
-    return Common::Cross(vec1, vec2).z;
-};
-
-/// Convert a 3D vector for cube map coordinates to 2D texture coordinates along with the face name
-static std::tuple<float24, float24, float24, PAddr> ConvertCubeCoord(float24 u, float24 v,
-                                                                     float24 w,
-                                                                     const TexturingRegs& regs) {
-    const float abs_u = std::abs(u.ToFloat32());
-    const float abs_v = std::abs(v.ToFloat32());
-    const float abs_w = std::abs(w.ToFloat32());
-    float24 x, y, z;
-    PAddr addr;
-    if (abs_u > abs_v && abs_u > abs_w) {
-        if (u > float24::FromFloat32(0)) {
-            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::PositiveX);
-            y = -v;
-        } else {
-            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::NegativeX);
-            y = v;
-        }
-        x = -w;
-        z = u;
-    } else if (abs_v > abs_w) {
-        if (v > float24::FromFloat32(0)) {
-            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::PositiveY);
-            x = u;
-        } else {
-            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::NegativeY);
-            x = -u;
-        }
-        y = w;
-        z = v;
-    } else {
-        if (w > float24::FromFloat32(0)) {
-            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::PositiveZ);
-            y = -v;
-        } else {
-            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::NegativeZ);
-            y = v;
-        }
-        x = u;
-        z = w;
-    }
-    float24 z_abs = float24::FromFloat32(std::abs(z.ToFloat32()));
-    const float24 half = float24::FromFloat32(0.5f);
-    return std::make_tuple(x / z * half + half, y / z * half + half, z_abs, addr);
-}
-
-MICROPROFILE_DEFINE(GPU_Rasterization, "GPU", "Rasterization", MP_RGB(50, 50, 240));
-
-/**
- * Helper function for ProcessTriangle with the "reversed" flag to allow for implementing
- * culling via recursion.
- */
-static void ProcessTriangleInternal(const Vertex& v0, const Vertex& v1, const Vertex& v2,
-                                    bool reversed = false) {
-    const auto& regs = g_state.regs;
-    MICROPROFILE_SCOPE(GPU_Rasterization);
-
-    // vertex positions in rasterizer coordinates
-    static auto FloatToFix = [](float24 flt) {
-        // TODO: Rounding here is necessary to prevent garbage pixels at
-        //       triangle borders. Is it that the correct solution, though?
-        return Fix12P4(static_cast<unsigned short>(round(flt.ToFloat32() * 16.0f)));
-    };
-    static auto ScreenToRasterizerCoordinates = [](const Common::Vec3<float24>& vec) {
-        return Common::Vec3<Fix12P4>{FloatToFix(vec.x), FloatToFix(vec.y), FloatToFix(vec.z)};
-    };
-
-    Common::Vec3<Fix12P4> vtxpos[3]{ScreenToRasterizerCoordinates(v0.screenpos),
-                                    ScreenToRasterizerCoordinates(v1.screenpos),
-                                    ScreenToRasterizerCoordinates(v2.screenpos)};
-
-    if (regs.rasterizer.cull_mode == RasterizerRegs::CullMode::KeepAll) {
-        // Make sure we always end up with a triangle wound counter-clockwise
-        if (!reversed && SignedArea(vtxpos[0].xy(), vtxpos[1].xy(), vtxpos[2].xy()) <= 0) {
-            ProcessTriangleInternal(v0, v2, v1, true);
-            return;
-        }
-    } else {
-        if (!reversed && regs.rasterizer.cull_mode == RasterizerRegs::CullMode::KeepClockWise) {
-            // Reverse vertex order and use the CCW code path.
-            ProcessTriangleInternal(v0, v2, v1, true);
-            return;
-        }
-
-        // Cull away triangles which are wound clockwise.
-        if (SignedArea(vtxpos[0].xy(), vtxpos[1].xy(), vtxpos[2].xy()) <= 0)
-            return;
-    }
-
-    u16 min_x = std::min({vtxpos[0].x, vtxpos[1].x, vtxpos[2].x});
-    u16 min_y = std::min({vtxpos[0].y, vtxpos[1].y, vtxpos[2].y});
-    u16 max_x = std::max({vtxpos[0].x, vtxpos[1].x, vtxpos[2].x});
-    u16 max_y = std::max({vtxpos[0].y, vtxpos[1].y, vtxpos[2].y});
-
-    // Convert the scissor box coordinates to 12.4 fixed point
-    u16 scissor_x1 = (u16)(regs.rasterizer.scissor_test.x1 << 4);
-    u16 scissor_y1 = (u16)(regs.rasterizer.scissor_test.y1 << 4);
-    // x2,y2 have +1 added to cover the entire sub-pixel area
-    u16 scissor_x2 = (u16)((regs.rasterizer.scissor_test.x2 + 1) << 4);
-    u16 scissor_y2 = (u16)((regs.rasterizer.scissor_test.y2 + 1) << 4);
-
-    if (regs.rasterizer.scissor_test.mode == RasterizerRegs::ScissorMode::Include) {
-        // Calculate the new bounds
-        min_x = std::max(min_x, scissor_x1);
-        min_y = std::max(min_y, scissor_y1);
-        max_x = std::min(max_x, scissor_x2);
-        max_y = std::min(max_y, scissor_y2);
-    }
-
-    min_x &= Fix12P4::IntMask();
-    min_y &= Fix12P4::IntMask();
-    max_x = ((max_x + Fix12P4::FracMask()) & Fix12P4::IntMask());
-    max_y = ((max_y + Fix12P4::FracMask()) & Fix12P4::IntMask());
-
-    // Triangle filling rules: Pixels on the right-sided edge or on flat bottom edges are not
-    // drawn. Pixels on any other triangle border are drawn. This is implemented with three bias
-    // values which are added to the barycentric coordinates w0, w1 and w2, respectively.
-    // NOTE: These are the PSP filling rules. Not sure if the 3DS uses the same ones...
-    auto IsRightSideOrFlatBottomEdge = [](const Common::Vec2<Fix12P4>& vtx,
-                                          const Common::Vec2<Fix12P4>& line1,
-                                          const Common::Vec2<Fix12P4>& line2) {
-        if (line1.y == line2.y) {
-            // just check if vertex is above us => bottom line parallel to x-axis
-            return vtx.y < line1.y;
-        } else {
-            // check if vertex is on our left => right side
-            // TODO: Not sure how likely this is to overflow
-            return (int)vtx.x < (int)line1.x + ((int)line2.x - (int)line1.x) *
-                                                   ((int)vtx.y - (int)line1.y) /
-                                                   ((int)line2.y - (int)line1.y);
-        }
-    };
-    int bias0 =
-        IsRightSideOrFlatBottomEdge(vtxpos[0].xy(), vtxpos[1].xy(), vtxpos[2].xy()) ? -1 : 0;
-    int bias1 =
-        IsRightSideOrFlatBottomEdge(vtxpos[1].xy(), vtxpos[2].xy(), vtxpos[0].xy()) ? -1 : 0;
-    int bias2 =
-        IsRightSideOrFlatBottomEdge(vtxpos[2].xy(), vtxpos[0].xy(), vtxpos[1].xy()) ? -1 : 0;
-
-    auto w_inverse = Common::MakeVec(v0.pos.w, v1.pos.w, v2.pos.w);
-
-    auto textures = regs.texturing.GetTextures();
-    auto tev_stages = regs.texturing.GetTevStages();
-
-    bool stencil_action_enable =
-        g_state.regs.framebuffer.output_merger.stencil_test.enable &&
-        g_state.regs.framebuffer.framebuffer.depth_format == FramebufferRegs::DepthFormat::D24S8;
-    const auto stencil_test = g_state.regs.framebuffer.output_merger.stencil_test;
-
-    // Enter rasterization loop, starting at the center of the topleft bounding box corner.
-    // TODO: Not sure if looping through x first might be faster
-    for (u16 y = min_y + 8; y < max_y; y += 0x10) {
-        for (u16 x = min_x + 8; x < max_x; x += 0x10) {
-
-            // Do not process the pixel if it's inside the scissor box and the scissor mode is set
-            // to Exclude
-            if (regs.rasterizer.scissor_test.mode == RasterizerRegs::ScissorMode::Exclude) {
-                if (x >= scissor_x1 && x < scissor_x2 && y >= scissor_y1 && y < scissor_y2)
-                    continue;
-            }
-
-            // Calculate the barycentric coordinates w0, w1 and w2
-            int w0 = bias0 + SignedArea(vtxpos[1].xy(), vtxpos[2].xy(), {x, y});
-            int w1 = bias1 + SignedArea(vtxpos[2].xy(), vtxpos[0].xy(), {x, y});
-            int w2 = bias2 + SignedArea(vtxpos[0].xy(), vtxpos[1].xy(), {x, y});
-            int wsum = w0 + w1 + w2;
-
-            // If current pixel is not covered by the current primitive
-            if (w0 < 0 || w1 < 0 || w2 < 0)
-                continue;
-
-            auto baricentric_coordinates =
-                Common::MakeVec(float24::FromFloat32(static_cast<float>(w0)),
-                                float24::FromFloat32(static_cast<float>(w1)),
-                                float24::FromFloat32(static_cast<float>(w2)));
-            float24 interpolated_w_inverse =
-                float24::FromFloat32(1.0f) / Common::Dot(w_inverse, baricentric_coordinates);
-
-            // interpolated_z = z / w
-            float interpolated_z_over_w =
-                (v0.screenpos[2].ToFloat32() * w0 + v1.screenpos[2].ToFloat32() * w1 +
-                 v2.screenpos[2].ToFloat32() * w2) /
-                wsum;
-
-            // Not fully accurate. About 3 bits in precision are missing.
-            // Z-Buffer (z / w * scale + offset)
-            float depth_scale = float24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32();
-            float depth_offset =
-                float24::FromRaw(regs.rasterizer.viewport_depth_near_plane).ToFloat32();
-            float depth = interpolated_z_over_w * depth_scale + depth_offset;
-
-            // Potentially switch to W-Buffer
-            if (regs.rasterizer.depthmap_enable ==
-                Pica::RasterizerRegs::DepthBuffering::WBuffering) {
-                // W-Buffer (z * scale + w * offset = (z / w * scale + offset) * w)
-                depth *= interpolated_w_inverse.ToFloat32() * wsum;
-            }
-
-            // Clamp the result
-            depth = std::clamp(depth, 0.0f, 1.0f);
-
-            // Perspective correct attribute interpolation:
-            // Attribute values cannot be calculated by simple linear interpolation since
-            // they are not linear in screen space. For example, when interpolating a
-            // texture coordinate across two vertices, something simple like
-            //     u = (u0*w0 + u1*w1)/(w0+w1)
-            // will not work. However, the attribute value divided by the
-            // clipspace w-coordinate (u/w) and and the inverse w-coordinate (1/w) are linear
-            // in screenspace. Hence, we can linearly interpolate these two independently and
-            // calculate the interpolated attribute by dividing the results.
-            // I.e.
-            //     u_over_w   = ((u0/v0.pos.w)*w0 + (u1/v1.pos.w)*w1)/(w0+w1)
-            //     one_over_w = (( 1/v0.pos.w)*w0 + ( 1/v1.pos.w)*w1)/(w0+w1)
-            //     u = u_over_w / one_over_w
-            //
-            // The generalization to three vertices is straightforward in baricentric coordinates.
-            auto GetInterpolatedAttribute = [&](float24 attr0, float24 attr1, float24 attr2) {
-                auto attr_over_w = Common::MakeVec(attr0, attr1, attr2);
-                float24 interpolated_attr_over_w =
-                    Common::Dot(attr_over_w, baricentric_coordinates);
-                return interpolated_attr_over_w * interpolated_w_inverse;
-            };
-
-            Common::Vec4<u8> primary_color{
-                static_cast<u8>(round(
-                    GetInterpolatedAttribute(v0.color.r(), v1.color.r(), v2.color.r()).ToFloat32() *
-                    255)),
-                static_cast<u8>(round(
-                    GetInterpolatedAttribute(v0.color.g(), v1.color.g(), v2.color.g()).ToFloat32() *
-                    255)),
-                static_cast<u8>(round(
-                    GetInterpolatedAttribute(v0.color.b(), v1.color.b(), v2.color.b()).ToFloat32() *
-                    255)),
-                static_cast<u8>(round(
-                    GetInterpolatedAttribute(v0.color.a(), v1.color.a(), v2.color.a()).ToFloat32() *
-                    255)),
-            };
-
-            Common::Vec2<float24> uv[3];
-            uv[0].u() = GetInterpolatedAttribute(v0.tc0.u(), v1.tc0.u(), v2.tc0.u());
-            uv[0].v() = GetInterpolatedAttribute(v0.tc0.v(), v1.tc0.v(), v2.tc0.v());
-            uv[1].u() = GetInterpolatedAttribute(v0.tc1.u(), v1.tc1.u(), v2.tc1.u());
-            uv[1].v() = GetInterpolatedAttribute(v0.tc1.v(), v1.tc1.v(), v2.tc1.v());
-            uv[2].u() = GetInterpolatedAttribute(v0.tc2.u(), v1.tc2.u(), v2.tc2.u());
-            uv[2].v() = GetInterpolatedAttribute(v0.tc2.v(), v1.tc2.v(), v2.tc2.v());
-
-            Common::Vec4<u8> texture_color[4]{};
-            for (int i = 0; i < 3; ++i) {
-                const auto& texture = textures[i];
-                if (!texture.enabled)
-                    continue;
-
-                if (texture.config.address == 0) {
-                    texture_color[i] = {0, 0, 0, 255};
-                    continue;
-                }
-
-                int coordinate_i =
-                    (i == 2 && regs.texturing.main_config.texture2_use_coord1) ? 1 : i;
-                float24 u = uv[coordinate_i].u();
-                float24 v = uv[coordinate_i].v();
-
-                // Only unit 0 respects the texturing type (according to 3DBrew)
-                // TODO: Refactor so cubemaps and shadowmaps can be handled
-                PAddr texture_address = texture.config.GetPhysicalAddress();
-                float24 shadow_z;
-                if (i == 0) {
-                    switch (texture.config.type) {
-                    case TexturingRegs::TextureConfig::Texture2D:
-                        break;
-                    case TexturingRegs::TextureConfig::ShadowCube:
-                    case TexturingRegs::TextureConfig::TextureCube: {
-                        auto w = GetInterpolatedAttribute(v0.tc0_w, v1.tc0_w, v2.tc0_w);
-                        std::tie(u, v, shadow_z, texture_address) =
-                            ConvertCubeCoord(u, v, w, regs.texturing);
-                        break;
-                    }
-                    case TexturingRegs::TextureConfig::Projection2D: {
-                        auto tc0_w = GetInterpolatedAttribute(v0.tc0_w, v1.tc0_w, v2.tc0_w);
-                        u /= tc0_w;
-                        v /= tc0_w;
-                        break;
-                    }
-                    case TexturingRegs::TextureConfig::Shadow2D: {
-                        auto tc0_w = GetInterpolatedAttribute(v0.tc0_w, v1.tc0_w, v2.tc0_w);
-                        if (!regs.texturing.shadow.orthographic) {
-                            u /= tc0_w;
-                            v /= tc0_w;
-                        }
-
-                        shadow_z = float24::FromFloat32(std::abs(tc0_w.ToFloat32()));
-                        break;
-                    }
-                    case TexturingRegs::TextureConfig::Disabled:
-                        continue; // skip this unit and continue to the next unit
-                    default:
-                        LOG_ERROR(HW_GPU, "Unhandled texture type {:x}", (int)texture.config.type);
-                        UNIMPLEMENTED();
-                        break;
-                    }
-                }
-
-                int s = (int)(u * float24::FromFloat32(static_cast<float>(texture.config.width)))
-                            .ToFloat32();
-                int t = (int)(v * float24::FromFloat32(static_cast<float>(texture.config.height)))
-                            .ToFloat32();
-
-                bool use_border_s = false;
-                bool use_border_t = false;
-
-                if (texture.config.wrap_s == TexturingRegs::TextureConfig::ClampToBorder) {
-                    use_border_s = s < 0 || s >= static_cast<int>(texture.config.width);
-                } else if (texture.config.wrap_s == TexturingRegs::TextureConfig::ClampToBorder2) {
-                    use_border_s = s >= static_cast<int>(texture.config.width);
-                }
-
-                if (texture.config.wrap_t == TexturingRegs::TextureConfig::ClampToBorder) {
-                    use_border_t = t < 0 || t >= static_cast<int>(texture.config.height);
-                } else if (texture.config.wrap_t == TexturingRegs::TextureConfig::ClampToBorder2) {
-                    use_border_t = t >= static_cast<int>(texture.config.height);
-                }
-
-                if (use_border_s || use_border_t) {
-                    auto border_color = texture.config.border_color;
-                    texture_color[i] =
-                        Common::MakeVec(border_color.r.Value(), border_color.g.Value(),
-                                        border_color.b.Value(), border_color.a.Value())
-                            .Cast<u8>();
-                } else {
-                    // Textures are laid out from bottom to top, hence we invert the t coordinate.
-                    // NOTE: This may not be the right place for the inversion.
-                    // TODO: Check if this applies to ETC textures, too.
-                    s = GetWrappedTexCoord(texture.config.wrap_s, s, texture.config.width);
-                    t = texture.config.height - 1 -
-                        GetWrappedTexCoord(texture.config.wrap_t, t, texture.config.height);
-
-                    const u8* texture_data =
-                        VideoCore::g_memory->GetPhysicalPointer(texture_address);
-                    auto info =
-                        Texture::TextureInfo::FromPicaRegister(texture.config, texture.format);
-
-                    // TODO: Apply the min and mag filters to the texture
-                    texture_color[i] = Texture::LookupTexture(texture_data, s, t, info);
-                }
-
-                if (i == 0 && (texture.config.type == TexturingRegs::TextureConfig::Shadow2D ||
-                               texture.config.type == TexturingRegs::TextureConfig::ShadowCube)) {
-
-                    s32 z_int = static_cast<s32>(std::min(shadow_z.ToFloat32(), 1.0f) * 0xFFFFFF);
-                    z_int -= regs.texturing.shadow.bias << 1;
-                    auto& color = texture_color[i];
-                    s32 z_ref = (color.w << 16) | (color.z << 8) | color.y;
-                    u8 density;
-                    if (z_ref >= z_int) {
-                        density = color.x;
-                    } else {
-                        density = 0;
-                    }
-                    texture_color[i] = {density, density, density, density};
-                }
-            }
-
-            // sample procedural texture
-            if (regs.texturing.main_config.texture3_enable) {
-                const auto& proctex_uv = uv[regs.texturing.main_config.texture3_coordinates];
-                texture_color[3] = ProcTex(proctex_uv.u().ToFloat32(), proctex_uv.v().ToFloat32(),
-                                           g_state.regs.texturing, g_state.proctex);
-            }
-
-            // Texture environment - consists of 6 stages of color and alpha combining.
-            //
-            // Color combiners take three input color values from some source (e.g. interpolated
-            // vertex color, texture color, previous stage, etc), perform some very simple
-            // operations on each of them (e.g. inversion) and then calculate the output color
-            // with some basic arithmetic. Alpha combiners can be configured separately but work
-            // analogously.
-            Common::Vec4<u8> combiner_output;
-            Common::Vec4<u8> combiner_buffer = {0, 0, 0, 0};
-            Common::Vec4<u8> next_combiner_buffer =
-                Common::MakeVec(regs.texturing.tev_combiner_buffer_color.r.Value(),
-                                regs.texturing.tev_combiner_buffer_color.g.Value(),
-                                regs.texturing.tev_combiner_buffer_color.b.Value(),
-                                regs.texturing.tev_combiner_buffer_color.a.Value())
-                    .Cast<u8>();
-
-            Common::Vec4<u8> primary_fragment_color = {0, 0, 0, 0};
-            Common::Vec4<u8> secondary_fragment_color = {0, 0, 0, 0};
-
-            if (!g_state.regs.lighting.disable) {
-                Common::Quaternion<float> normquat =
-                    Common::Quaternion<float>{
-                        {GetInterpolatedAttribute(v0.quat.x, v1.quat.x, v2.quat.x).ToFloat32(),
-                         GetInterpolatedAttribute(v0.quat.y, v1.quat.y, v2.quat.y).ToFloat32(),
-                         GetInterpolatedAttribute(v0.quat.z, v1.quat.z, v2.quat.z).ToFloat32()},
-                        GetInterpolatedAttribute(v0.quat.w, v1.quat.w, v2.quat.w).ToFloat32(),
-                    }
-                        .Normalized();
-
-                Common::Vec3<float> view{
-                    GetInterpolatedAttribute(v0.view.x, v1.view.x, v2.view.x).ToFloat32(),
-                    GetInterpolatedAttribute(v0.view.y, v1.view.y, v2.view.y).ToFloat32(),
-                    GetInterpolatedAttribute(v0.view.z, v1.view.z, v2.view.z).ToFloat32(),
-                };
-                std::tie(primary_fragment_color, secondary_fragment_color) = ComputeFragmentsColors(
-                    g_state.regs.lighting, g_state.lighting, normquat, view, texture_color);
-            }
-
-            for (unsigned tev_stage_index = 0; tev_stage_index < tev_stages.size();
-                 ++tev_stage_index) {
-                const auto& tev_stage = tev_stages[tev_stage_index];
-                using Source = TexturingRegs::TevStageConfig::Source;
-
-                auto GetSource = [&](Source source) -> Common::Vec4<u8> {
-                    switch (source) {
-                    case Source::PrimaryColor:
-                        return primary_color;
-
-                    case Source::PrimaryFragmentColor:
-                        return primary_fragment_color;
-
-                    case Source::SecondaryFragmentColor:
-                        return secondary_fragment_color;
-
-                    case Source::Texture0:
-                        return texture_color[0];
-
-                    case Source::Texture1:
-                        return texture_color[1];
-
-                    case Source::Texture2:
-                        return texture_color[2];
-
-                    case Source::Texture3:
-                        return texture_color[3];
-
-                    case Source::PreviousBuffer:
-                        return combiner_buffer;
-
-                    case Source::Constant:
-                        return Common::MakeVec(tev_stage.const_r.Value(), tev_stage.const_g.Value(),
-                                               tev_stage.const_b.Value(), tev_stage.const_a.Value())
-                            .Cast<u8>();
-
-                    case Source::Previous:
-                        return combiner_output;
-
-                    default:
-                        LOG_ERROR(HW_GPU, "Unknown color combiner source {}", (int)source);
-                        UNIMPLEMENTED();
-                        return {0, 0, 0, 0};
-                    }
-                };
-
-                // color combiner
-                // NOTE: Not sure if the alpha combiner might use the color output of the previous
-                //       stage as input. Hence, we currently don't directly write the result to
-                //       combiner_output.rgb(), but instead store it in a temporary variable until
-                //       alpha combining has been done.
-                Common::Vec3<u8> color_result[3] = {
-                    GetColorModifier(tev_stage.color_modifier1, GetSource(tev_stage.color_source1)),
-                    GetColorModifier(tev_stage.color_modifier2, GetSource(tev_stage.color_source2)),
-                    GetColorModifier(tev_stage.color_modifier3, GetSource(tev_stage.color_source3)),
-                };
-                auto color_output = ColorCombine(tev_stage.color_op, color_result);
-
-                u8 alpha_output;
-                if (tev_stage.color_op == TexturingRegs::TevStageConfig::Operation::Dot3_RGBA) {
-                    // result of Dot3_RGBA operation is also placed to the alpha component
-                    alpha_output = color_output.x;
-                } else {
-                    // alpha combiner
-                    std::array<u8, 3> alpha_result = {{
-                        GetAlphaModifier(tev_stage.alpha_modifier1,
-                                         GetSource(tev_stage.alpha_source1)),
-                        GetAlphaModifier(tev_stage.alpha_modifier2,
-                                         GetSource(tev_stage.alpha_source2)),
-                        GetAlphaModifier(tev_stage.alpha_modifier3,
-                                         GetSource(tev_stage.alpha_source3)),
-                    }};
-                    alpha_output = AlphaCombine(tev_stage.alpha_op, alpha_result);
-                }
-
-                combiner_output[0] =
-                    std::min((unsigned)255, color_output.r() * tev_stage.GetColorMultiplier());
-                combiner_output[1] =
-                    std::min((unsigned)255, color_output.g() * tev_stage.GetColorMultiplier());
-                combiner_output[2] =
-                    std::min((unsigned)255, color_output.b() * tev_stage.GetColorMultiplier());
-                combiner_output[3] =
-                    std::min((unsigned)255, alpha_output * tev_stage.GetAlphaMultiplier());
-
-                combiner_buffer = next_combiner_buffer;
-
-                if (regs.texturing.tev_combiner_buffer_input.TevStageUpdatesCombinerBufferColor(
-                        tev_stage_index)) {
-                    next_combiner_buffer.r() = combiner_output.r();
-                    next_combiner_buffer.g() = combiner_output.g();
-                    next_combiner_buffer.b() = combiner_output.b();
-                }
-
-                if (regs.texturing.tev_combiner_buffer_input.TevStageUpdatesCombinerBufferAlpha(
-                        tev_stage_index)) {
-                    next_combiner_buffer.a() = combiner_output.a();
-                }
-            }
-
-            const auto& output_merger = regs.framebuffer.output_merger;
-
-            if (output_merger.fragment_operation_mode ==
-                FramebufferRegs::FragmentOperationMode::Shadow) {
-                u32 depth_int = static_cast<u32>(depth * 0xFFFFFF);
-                // use green color as the shadow intensity
-                u8 stencil = combiner_output.y;
-                DrawShadowMapPixel(x >> 4, y >> 4, depth_int, stencil);
-                // skip the normal output merger pipeline if it is in shadow mode
-                continue;
-            }
-
-            // TODO: Does alpha testing happen before or after stencil?
-            if (output_merger.alpha_test.enable) {
-                bool pass = false;
-
-                switch (output_merger.alpha_test.func) {
-                case FramebufferRegs::CompareFunc::Never:
-                    pass = false;
-                    break;
-
-                case FramebufferRegs::CompareFunc::Always:
-                    pass = true;
-                    break;
-
-                case FramebufferRegs::CompareFunc::Equal:
-                    pass = combiner_output.a() == output_merger.alpha_test.ref;
-                    break;
-
-                case FramebufferRegs::CompareFunc::NotEqual:
-                    pass = combiner_output.a() != output_merger.alpha_test.ref;
-                    break;
-
-                case FramebufferRegs::CompareFunc::LessThan:
-                    pass = combiner_output.a() < output_merger.alpha_test.ref;
-                    break;
-
-                case FramebufferRegs::CompareFunc::LessThanOrEqual:
-                    pass = combiner_output.a() <= output_merger.alpha_test.ref;
-                    break;
-
-                case FramebufferRegs::CompareFunc::GreaterThan:
-                    pass = combiner_output.a() > output_merger.alpha_test.ref;
-                    break;
-
-                case FramebufferRegs::CompareFunc::GreaterThanOrEqual:
-                    pass = combiner_output.a() >= output_merger.alpha_test.ref;
-                    break;
-                }
-
-                if (!pass)
-                    continue;
-            }
-
-            // Apply fog combiner
-            // Not fully accurate. We'd have to know what data type is used to
-            // store the depth etc. Using float for now until we know more
-            // about Pica datatypes
-            if (regs.texturing.fog_mode == TexturingRegs::FogMode::Fog) {
-                const Common::Vec3<u8> fog_color =
-                    Common::MakeVec(regs.texturing.fog_color.r.Value(),
-                                    regs.texturing.fog_color.g.Value(),
-                                    regs.texturing.fog_color.b.Value())
-                        .Cast<u8>();
-
-                // Get index into fog LUT
-                float fog_index;
-                if (g_state.regs.texturing.fog_flip) {
-                    fog_index = (1.0f - depth) * 128.0f;
-                } else {
-                    fog_index = depth * 128.0f;
-                }
-
-                // Generate clamped fog factor from LUT for given fog index
-                float fog_i = std::clamp(floorf(fog_index), 0.0f, 127.0f);
-                float fog_f = fog_index - fog_i;
-                const auto& fog_lut_entry = g_state.fog.lut[static_cast<unsigned int>(fog_i)];
-                float fog_factor = fog_lut_entry.ToFloat() + fog_lut_entry.DiffToFloat() * fog_f;
-                fog_factor = std::clamp(fog_factor, 0.0f, 1.0f);
-
-                // Blend the fog
-                for (unsigned i = 0; i < 3; i++) {
-                    combiner_output[i] = static_cast<u8>(fog_factor * combiner_output[i] +
-                                                         (1.0f - fog_factor) * fog_color[i]);
-                }
-            }
-
-            u8 old_stencil = 0;
-
-            auto UpdateStencil = [stencil_test, x, y,
-                                  &old_stencil](Pica::FramebufferRegs::StencilAction action) {
-                u8 new_stencil =
-                    PerformStencilAction(action, old_stencil, stencil_test.reference_value);
-                if (g_state.regs.framebuffer.framebuffer.allow_depth_stencil_write != 0)
-                    SetStencil(x >> 4, y >> 4,
-                               (new_stencil & stencil_test.write_mask) |
-                                   (old_stencil & ~stencil_test.write_mask));
-            };
-
-            if (stencil_action_enable) {
-                old_stencil = GetStencil(x >> 4, y >> 4);
-                u8 dest = old_stencil & stencil_test.input_mask;
-                u8 ref = stencil_test.reference_value & stencil_test.input_mask;
-
-                bool pass = false;
-                switch (stencil_test.func) {
-                case FramebufferRegs::CompareFunc::Never:
-                    pass = false;
-                    break;
-
-                case FramebufferRegs::CompareFunc::Always:
-                    pass = true;
-                    break;
-
-                case FramebufferRegs::CompareFunc::Equal:
-                    pass = (ref == dest);
-                    break;
-
-                case FramebufferRegs::CompareFunc::NotEqual:
-                    pass = (ref != dest);
-                    break;
-
-                case FramebufferRegs::CompareFunc::LessThan:
-                    pass = (ref < dest);
-                    break;
-
-                case FramebufferRegs::CompareFunc::LessThanOrEqual:
-                    pass = (ref <= dest);
-                    break;
-
-                case FramebufferRegs::CompareFunc::GreaterThan:
-                    pass = (ref > dest);
-                    break;
-
-                case FramebufferRegs::CompareFunc::GreaterThanOrEqual:
-                    pass = (ref >= dest);
-                    break;
-                }
-
-                if (!pass) {
-                    UpdateStencil(stencil_test.action_stencil_fail);
-                    continue;
-                }
-            }
-
-            // Convert float to integer
-            unsigned num_bits =
-                FramebufferRegs::DepthBitsPerPixel(regs.framebuffer.framebuffer.depth_format);
-            u32 z = (u32)(depth * ((1 << num_bits) - 1));
-
-            if (output_merger.depth_test_enable) {
-                u32 ref_z = GetDepth(x >> 4, y >> 4);
-
-                bool pass = false;
-
-                switch (output_merger.depth_test_func) {
-                case FramebufferRegs::CompareFunc::Never:
-                    pass = false;
-                    break;
-
-                case FramebufferRegs::CompareFunc::Always:
-                    pass = true;
-                    break;
-
-                case FramebufferRegs::CompareFunc::Equal:
-                    pass = z == ref_z;
-                    break;
-
-                case FramebufferRegs::CompareFunc::NotEqual:
-                    pass = z != ref_z;
-                    break;
-
-                case FramebufferRegs::CompareFunc::LessThan:
-                    pass = z < ref_z;
-                    break;
-
-                case FramebufferRegs::CompareFunc::LessThanOrEqual:
-                    pass = z <= ref_z;
-                    break;
-
-                case FramebufferRegs::CompareFunc::GreaterThan:
-                    pass = z > ref_z;
-                    break;
-
-                case FramebufferRegs::CompareFunc::GreaterThanOrEqual:
-                    pass = z >= ref_z;
-                    break;
-                }
-
-                if (!pass) {
-                    if (stencil_action_enable)
-                        UpdateStencil(stencil_test.action_depth_fail);
-                    continue;
-                }
-            }
-
-            if (regs.framebuffer.framebuffer.allow_depth_stencil_write != 0 &&
-                output_merger.depth_write_enable) {
-
-                SetDepth(x >> 4, y >> 4, z);
-            }
-
-            // The stencil depth_pass action is executed even if depth testing is disabled
-            if (stencil_action_enable)
-                UpdateStencil(stencil_test.action_depth_pass);
-
-            auto dest = GetPixel(x >> 4, y >> 4);
-            Common::Vec4<u8> blend_output = combiner_output;
-
-            if (output_merger.alphablend_enable) {
-                auto params = output_merger.alpha_blending;
-
-                auto LookupFactor = [&](unsigned channel,
-                                        FramebufferRegs::BlendFactor factor) -> u8 {
-                    DEBUG_ASSERT(channel < 4);
-
-                    const Common::Vec4<u8> blend_const =
-                        Common::MakeVec(output_merger.blend_const.r.Value(),
-                                        output_merger.blend_const.g.Value(),
-                                        output_merger.blend_const.b.Value(),
-                                        output_merger.blend_const.a.Value())
-                            .Cast<u8>();
-
-                    switch (factor) {
-                    case FramebufferRegs::BlendFactor::Zero:
-                        return 0;
-
-                    case FramebufferRegs::BlendFactor::One:
-                        return 255;
-
-                    case FramebufferRegs::BlendFactor::SourceColor:
-                        return combiner_output[channel];
-
-                    case FramebufferRegs::BlendFactor::OneMinusSourceColor:
-                        return 255 - combiner_output[channel];
-
-                    case FramebufferRegs::BlendFactor::DestColor:
-                        return dest[channel];
-
-                    case FramebufferRegs::BlendFactor::OneMinusDestColor:
-                        return 255 - dest[channel];
-
-                    case FramebufferRegs::BlendFactor::SourceAlpha:
-                        return combiner_output.a();
-
-                    case FramebufferRegs::BlendFactor::OneMinusSourceAlpha:
-                        return 255 - combiner_output.a();
-
-                    case FramebufferRegs::BlendFactor::DestAlpha:
-                        return dest.a();
-
-                    case FramebufferRegs::BlendFactor::OneMinusDestAlpha:
-                        return 255 - dest.a();
-
-                    case FramebufferRegs::BlendFactor::ConstantColor:
-                        return blend_const[channel];
-
-                    case FramebufferRegs::BlendFactor::OneMinusConstantColor:
-                        return 255 - blend_const[channel];
-
-                    case FramebufferRegs::BlendFactor::ConstantAlpha:
-                        return blend_const.a();
-
-                    case FramebufferRegs::BlendFactor::OneMinusConstantAlpha:
-                        return 255 - blend_const.a();
-
-                    case FramebufferRegs::BlendFactor::SourceAlphaSaturate:
-                        // Returns 1.0 for the alpha channel
-                        if (channel == 3)
-                            return 255;
-                        return std::min(combiner_output.a(), static_cast<u8>(255 - dest.a()));
-
-                    default:
-                        LOG_CRITICAL(HW_GPU, "Unknown blend factor {:x}", factor);
-                        UNIMPLEMENTED();
-                        break;
-                    }
-
-                    return combiner_output[channel];
-                };
-
-                auto srcfactor = Common::MakeVec(LookupFactor(0, params.factor_source_rgb),
-                                                 LookupFactor(1, params.factor_source_rgb),
-                                                 LookupFactor(2, params.factor_source_rgb),
-                                                 LookupFactor(3, params.factor_source_a));
-
-                auto dstfactor = Common::MakeVec(LookupFactor(0, params.factor_dest_rgb),
-                                                 LookupFactor(1, params.factor_dest_rgb),
-                                                 LookupFactor(2, params.factor_dest_rgb),
-                                                 LookupFactor(3, params.factor_dest_a));
-
-                blend_output = EvaluateBlendEquation(combiner_output, srcfactor, dest, dstfactor,
-                                                     params.blend_equation_rgb);
-                blend_output.a() = EvaluateBlendEquation(combiner_output, srcfactor, dest,
-                                                         dstfactor, params.blend_equation_a)
-                                       .a();
-            } else {
-                blend_output =
-                    Common::MakeVec(LogicOp(combiner_output.r(), dest.r(), output_merger.logic_op),
-                                    LogicOp(combiner_output.g(), dest.g(), output_merger.logic_op),
-                                    LogicOp(combiner_output.b(), dest.b(), output_merger.logic_op),
-                                    LogicOp(combiner_output.a(), dest.a(), output_merger.logic_op));
-            }
-
-            const Common::Vec4<u8> result = {
-                output_merger.red_enable ? blend_output.r() : dest.r(),
-                output_merger.green_enable ? blend_output.g() : dest.g(),
-                output_merger.blue_enable ? blend_output.b() : dest.b(),
-                output_merger.alpha_enable ? blend_output.a() : dest.a(),
-            };
-
-            if (regs.framebuffer.framebuffer.allow_color_write != 0)
-                DrawPixel(x >> 4, y >> 4, result);
-        }
-    }
-}
-
-void ProcessTriangle(const Vertex& v0, const Vertex& v1, const Vertex& v2) {
-    ProcessTriangleInternal(v0, v1, v2);
-}
-
-} // namespace Pica::Rasterizer

src/video_core/renderer_software/rasterizer.h

@@ -1,44 +0,0 @@
-// Copyright 2014 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include "video_core/shader/shader.h"
-
-namespace Pica::Rasterizer {
-
-struct Vertex : Shader::OutputVertex {
-    Vertex(const OutputVertex& v) : OutputVertex(v) {}
-
-    // Attributes used to store intermediate results
-    // position after perspective divide
-    Common::Vec3<float24> screenpos;
-
-    // Linear interpolation
-    // factor: 0=this, 1=vtx
-    // Note: This function cannot be called after perspective divide
-    void Lerp(float24 factor, const Vertex& vtx) {
-        pos = pos * factor + vtx.pos * (float24::FromFloat32(1) - factor);
-        quat = quat * factor + vtx.quat * (float24::FromFloat32(1) - factor);
-        color = color * factor + vtx.color * (float24::FromFloat32(1) - factor);
-        tc0 = tc0 * factor + vtx.tc0 * (float24::FromFloat32(1) - factor);
-        tc1 = tc1 * factor + vtx.tc1 * (float24::FromFloat32(1) - factor);
-        tc0_w = tc0_w * factor + vtx.tc0_w * (float24::FromFloat32(1) - factor);
-        view = view * factor + vtx.view * (float24::FromFloat32(1) - factor);
-        tc2 = tc2 * factor + vtx.tc2 * (float24::FromFloat32(1) - factor);
-    }
-
-    // Linear interpolation
-    // factor: 0=v0, 1=v1
-    // Note: This function cannot be called after perspective divide
-    static Vertex Lerp(float24 factor, const Vertex& v0, const Vertex& v1) {
-        Vertex ret = v0;
-        ret.Lerp(factor, v1);
-        return ret;
-    }
-};
-
-void ProcessTriangle(const Vertex& v0, const Vertex& v1, const Vertex& v2);
-
-} // namespace Pica::Rasterizer

src/video_core/renderer_software/renderer_software.cpp

@@ -2,18 +2,86 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/color.h"
+#include "core/core.h"
+#include "core/hw/gpu.h"
+#include "core/hw/hw.h"
+#include "core/hw/lcd.h"
 #include "video_core/renderer_software/renderer_software.h"
 
-namespace VideoCore {
+namespace SwRenderer {
 
 RendererSoftware::RendererSoftware(Core::System& system, Frontend::EmuWindow& window)
-    : VideoCore::RendererBase{system, window, nullptr},
+    : VideoCore::RendererBase{system, window, nullptr}, memory{system.Memory()},
-      rasterizer{std::make_unique<RasterizerSoftware>()} {}
+      rasterizer{std::make_unique<RasterizerSoftware>(system.Memory())} {}
 
 RendererSoftware::~RendererSoftware() = default;
 
 void RendererSoftware::SwapBuffers() {
+    PrepareRenderTarget();
     EndFrame();
 }
 
-} // namespace VideoCore
+void RendererSoftware::PrepareRenderTarget() {
+    for (int i : {0, 1, 2}) {
+        const int fb_id = i == 2 ? 1 : 0;
+        const auto& framebuffer = GPU::g_regs.framebuffer_config[fb_id];
+        auto& info = screen_infos[i];
+
+        u32 lcd_color_addr =
+            (fb_id == 0) ? LCD_REG_INDEX(color_fill_top) : LCD_REG_INDEX(color_fill_bottom);
+        lcd_color_addr = HW::VADDR_LCD + 4 * lcd_color_addr;
+        LCD::Regs::ColorFill color_fill = {0};
+        LCD::Read(color_fill.raw, lcd_color_addr);
+
+        if (!color_fill.is_enabled) {
+            const u32 old_width = std::exchange(info.width, framebuffer.width);
+            const u32 old_height = std::exchange(info.height, framebuffer.height);
+            if (framebuffer.width != old_width || framebuffer.height != old_height) [[unlikely]] {
+                info.pixels.resize(framebuffer.width * framebuffer.height * 4);
+            }
+            CopyPixels(i);
+        }
+    }
+}
+
+void RendererSoftware::CopyPixels(int i) {
+    const u32 fb_id = i == 2 ? 1 : 0;
+    const auto& framebuffer = GPU::g_regs.framebuffer_config[fb_id];
+
+    const PAddr framebuffer_addr =
+        framebuffer.active_fb == 0 ? framebuffer.address_left1 : framebuffer.address_left2;
+    const s32 bpp = GPU::Regs::BytesPerPixel(framebuffer.color_format);
+    const u8* framebuffer_data = memory.GetPhysicalPointer(framebuffer_addr);
+
+    const s32 stride = framebuffer.stride;
+    const s32 height = framebuffer.height;
+    ASSERT(stride * height != 0);
+
+    u32 output_offset = 0;
+    for (u32 y = 0; y < framebuffer.height; y++) {
+        for (u32 x = 0; x < framebuffer.width; x++) {
+            const u8* pixel = framebuffer_data + (y * stride + x) * 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);
+                }
+                UNREACHABLE();
+            }();
+            u8* dest = screen_infos[i].pixels.data() + output_offset;
+            std::memcpy(dest, color.AsArray(), sizeof(color));
+            output_offset += sizeof(color);
+        }
+    }
+}
+
+} // namespace SwRenderer

src/video_core/renderer_software/renderer_software.h

@@ -11,7 +11,13 @@ namespace Core {
 class System;
 }
 
-namespace VideoCore {
+namespace SwRenderer {
+
+struct ScreenInfo {
+    u32 width;
+    u32 height;
+    std::vector<u8> pixels;
+};
 
 class RendererSoftware : public VideoCore::RendererBase {
 public:
@@ -22,12 +28,22 @@ public:
         return rasterizer.get();
     }
 
+    [[nodiscard]] const ScreenInfo& Screen(VideoCore::ScreenId id) const noexcept {
+        return screen_infos[static_cast<u32>(id)];
+    }
+
     void SwapBuffers() override;
     void TryPresent(int timeout_ms, bool is_secondary) override {}
     void Sync() override {}
 
 private:
+    void PrepareRenderTarget();
+    void CopyPixels(int i);
+
+private:
+    Memory::MemorySystem& memory;
     std::unique_ptr<RasterizerSoftware> rasterizer;
+    std::array<ScreenInfo, 3> screen_infos{};
 };
 
-} // namespace VideoCore
+} // namespace SwRenderer

src/video_core/renderer_software/sw_clipper.cpp

@@ -1,196 +1,88 @@
-// Copyright 2014 Citra Emulator Project
+// Copyright 2023 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <algorithm>
 #include <array>
 #include <cstddef>
-#include <boost/container/static_vector.hpp>
+#include "video_core/regs_texturing.h"
-#include "common/bit_field.h"
-#include "common/common_types.h"
-#include "common/logging/log.h"
-#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"
 
-using Pica::Rasterizer::Vertex;
+namespace SwRenderer {
 
-namespace Pica::Clipper {
+using Pica::TexturingRegs;
 
-struct ClippingEdge {
+void FlipQuaternionIfOpposite(Common::Vec4<f24>& a, const Common::Vec4<f24>& b) {
-public:
+    if (Common::Dot(a, b) < f24::Zero()) {
-    ClippingEdge(Common::Vec4<float24> coeffs,
+        a *= f24::FromFloat32(-1.0f);
-                 Common::Vec4<float24> bias = Common::Vec4<float24>(float24::FromFloat32(0),
-                                                                    float24::FromFloat32(0),
-                                                                    float24::FromFloat32(0),
-                                                                    float24::FromFloat32(0)))
-        : coeffs(coeffs), bias(bias) {}
-
-    bool IsInside(const Vertex& vertex) const {
-        return Common::Dot(vertex.pos + bias, coeffs) >= float24::FromFloat32(0);
     }
-
-    bool IsOutSide(const Vertex& vertex) const {
-        return !IsInside(vertex);
-    }
-
-    Vertex GetIntersection(const Vertex& v0, const Vertex& v1) const {
-        float24 dp = Common::Dot(v0.pos + bias, coeffs);
-        float24 dp_prev = Common::Dot(v1.pos + bias, coeffs);
-        float24 factor = dp_prev / (dp_prev - dp);
-
-        return Vertex::Lerp(factor, v0, v1);
-    }
-
-private:
-    [[maybe_unused]] float24 pos;
-    Common::Vec4<float24> coeffs;
-    Common::Vec4<float24> bias;
 };
 
-static void InitScreenCoordinates(Vertex& vtx) {
+int SignedArea(const Common::Vec2<Fix12P4>& vtx1, const Common::Vec2<Fix12P4>& vtx2,
-    struct {
+               const Common::Vec2<Fix12P4>& vtx3) {
-        float24 halfsize_x;
+    const auto vec1 = Common::MakeVec(vtx2 - vtx1, 0);
-        float24 offset_x;
+    const auto vec2 = Common::MakeVec(vtx3 - vtx1, 0);
-        float24 halfsize_y;
+    // TODO: There is a very small chance this will overflow for sizeof(int) == 4
-        float24 offset_y;
+    return Common::Cross(vec1, vec2).z;
-        float24 zscale;
+};
-        float24 offset_z;
-    } viewport;
 
-    const auto& regs = g_state.regs;
+std::tuple<f24, f24, f24, PAddr> ConvertCubeCoord(f24 u, f24 v, f24 w,
-    viewport.halfsize_x = float24::FromRaw(regs.rasterizer.viewport_size_x);
+                                                  const Pica::TexturingRegs& regs) {
-    viewport.halfsize_y = float24::FromRaw(regs.rasterizer.viewport_size_y);
+    const float abs_u = std::abs(u.ToFloat32());
-    viewport.offset_x = float24::FromFloat32(static_cast<float>(regs.rasterizer.viewport_corner.x));
+    const float abs_v = std::abs(v.ToFloat32());
-    viewport.offset_y = float24::FromFloat32(static_cast<float>(regs.rasterizer.viewport_corner.y));
+    const float abs_w = std::abs(w.ToFloat32());
-
+    f24 x, y, z;
-    float24 inv_w = float24::FromFloat32(1.f) / vtx.pos.w;
+    PAddr addr;
-    vtx.pos.w = inv_w;
+    if (abs_u > abs_v && abs_u > abs_w) {
-    vtx.quat *= inv_w;
+        if (u > f24::Zero()) {
-    vtx.color *= inv_w;
+            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::PositiveX);
-    vtx.tc0 *= inv_w;
+            y = -v;
-    vtx.tc1 *= inv_w;
+        } else {
-    vtx.tc0_w *= inv_w;
+            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::NegativeX);
-    vtx.view *= inv_w;
+            y = v;
-    vtx.tc2 *= inv_w;
-
-    vtx.screenpos[0] =
-        (vtx.pos.x * inv_w + float24::FromFloat32(1.0)) * viewport.halfsize_x + viewport.offset_x;
-    vtx.screenpos[1] =
-        (vtx.pos.y * inv_w + float24::FromFloat32(1.0)) * viewport.halfsize_y + viewport.offset_y;
-    vtx.screenpos[2] = vtx.pos.z * inv_w;
-}
-
-void ProcessTriangle(const OutputVertex& v0, const OutputVertex& v1, const OutputVertex& v2) {
-    using boost::container::static_vector;
-
-    // Clipping a planar n-gon against a plane will remove at least 1 vertex and introduces 2 at
-    // the new edge (or less in degenerate cases). As such, we can say that each clipping plane
-    // introduces at most 1 new vertex to the polygon. Since we start with a triangle and have a
-    // fixed 6 clipping planes, the maximum number of vertices of the clipped polygon is 3 + 6 = 9.
-    static const std::size_t MAX_VERTICES = 9;
-    static_vector<Vertex, MAX_VERTICES> buffer_a = {v0, v1, v2};
-    static_vector<Vertex, MAX_VERTICES> buffer_b;
-
-    auto FlipQuaternionIfOpposite = [](auto& a, const auto& b) {
-        if (Common::Dot(a, b) < float24::Zero())
-            a = a * float24::FromFloat32(-1.0f);
-    };
-
-    // Flip the quaternions if they are opposite to prevent interpolating them over the wrong
-    // direction.
-    FlipQuaternionIfOpposite(buffer_a[1].quat, buffer_a[0].quat);
-    FlipQuaternionIfOpposite(buffer_a[2].quat, buffer_a[0].quat);
-
-    auto* output_list = &buffer_a;
-    auto* input_list = &buffer_b;
-
-    // NOTE: We clip against a w=epsilon plane to guarantee that the output has a positive w value.
-    // TODO: Not sure if this is a valid approach. Also should probably instead use the smallest
-    //       epsilon possible within float24 accuracy.
-    static const float24 EPSILON = float24::FromFloat32(0.00001f);
-    static const float24 f0 = float24::FromFloat32(0.0);
-    static const float24 f1 = float24::FromFloat32(1.0);
-    static const std::array<ClippingEdge, 7> clipping_edges = {{
-        {Common::MakeVec(-f1, f0, f0, f1)}, // x = +w
-        {Common::MakeVec(f1, f0, f0, f1)},  // x = -w
-        {Common::MakeVec(f0, -f1, f0, f1)}, // y = +w
-        {Common::MakeVec(f0, f1, f0, f1)},  // y = -w
-        {Common::MakeVec(f0, f0, -f1, f0)}, // z =  0
-        {Common::MakeVec(f0, f0, f1, f1)},  // z = -w
-        {Common::MakeVec(f0, f0, f0, f1),
-         Common::Vec4<float24>(f0, f0, f0, EPSILON)}, // w = EPSILON
-    }};
-
-    // Simple implementation of the Sutherland-Hodgman clipping algorithm.
-    // TODO: Make this less inefficient (currently lots of useless buffering overhead happens here)
-    auto Clip = [&](const ClippingEdge& edge) {
-        std::swap(input_list, output_list);
-        output_list->clear();
-
-        const Vertex* reference_vertex = &input_list->back();
-
-        for (const auto& vertex : *input_list) {
-            // NOTE: This algorithm changes vertex order in some cases!
-            if (edge.IsInside(vertex)) {
-                if (edge.IsOutSide(*reference_vertex)) {
-                    output_list->push_back(edge.GetIntersection(vertex, *reference_vertex));
-                }
-
-                output_list->push_back(vertex);
-            } else if (edge.IsInside(*reference_vertex)) {
-                output_list->push_back(edge.GetIntersection(vertex, *reference_vertex));
-            }
-            reference_vertex = &vertex;
         }
-    };
+        x = -w;
-
+        z = u;
-    for (auto edge : clipping_edges) {
+    } else if (abs_v > abs_w) {
-        Clip(edge);
+        if (v > f24::Zero()) {
-
+            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::PositiveY);
-        // Need to have at least a full triangle to continue...
+            x = u;
-        if (output_list->size() < 3)
+        } else {
-            return;
+            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::NegativeY);
+            x = -u;
+        }
+        y = w;
+        z = v;
+    } else {
+        if (w > f24::Zero()) {
+            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::PositiveZ);
+            y = -v;
+        } else {
+            addr = regs.GetCubePhysicalAddress(TexturingRegs::CubeFace::NegativeZ);
+            y = v;
+        }
+        x = u;
+        z = w;
     }
+    const f24 z_abs = f24::FromFloat32(std::abs(z.ToFloat32()));
+    const f24 half = f24::FromFloat32(0.5f);
+    return std::make_tuple(x / z * half + half, y / z * half + half, z_abs, addr);
+}
 
-    if (g_state.regs.rasterizer.clip_enable) {
+bool IsRightSideOrFlatBottomEdge(const Common::Vec2<Fix12P4>& vtx,
-        ClippingEdge custom_edge{g_state.regs.rasterizer.GetClipCoef()};
+                                 const Common::Vec2<Fix12P4>& line1,
-        Clip(custom_edge);
+                                 const Common::Vec2<Fix12P4>& line2) {
-
+    if (line1.y == line2.y) {
-        if (output_list->size() < 3)
+        // Just check if vertex is above us => bottom line parallel to x-axis
-            return;
+        return vtx.y < line1.y;
-    }
+    } else {
-
+        // Check if vertex is on our left => right side
-    InitScreenCoordinates((*output_list)[0]);
+        // TODO: Not sure how likely this is to overflow
-    InitScreenCoordinates((*output_list)[1]);
+        const auto svtx = vtx.Cast<s32>();
-
+        const auto sline1 = line1.Cast<s32>();
-    for (std::size_t i = 0; i < output_list->size() - 2; i++) {
+        const auto sline2 = line2.Cast<s32>();
-        Vertex& vtx0 = (*output_list)[0];
+        return svtx.x <
-        Vertex& vtx1 = (*output_list)[i + 1];
+               sline1.x + (sline2.x - sline1.x) * (svtx.y - sline1.y) / (sline2.y - sline1.y);
-        Vertex& vtx2 = (*output_list)[i + 2];
-
-        InitScreenCoordinates(vtx2);
-
-        LOG_TRACE(
-            Render_Software,
-            "Triangle {}/{} at position ({:.3}, {:.3}, {:.3}, {:.3f}), "
-            "({:.3}, {:.3}, {:.3}, {:.3}), ({:.3}, {:.3}, {:.3}, {:.3}) and "
-            "screen position ({:.2}, {:.2}, {:.2}), ({:.2}, {:.2}, {:.2}), ({:.2}, {:.2}, {:.2})",
-            i + 1, output_list->size() - 2, vtx0.pos.x.ToFloat32(), vtx0.pos.y.ToFloat32(),
-            vtx0.pos.z.ToFloat32(), vtx0.pos.w.ToFloat32(), vtx1.pos.x.ToFloat32(),
-            vtx1.pos.y.ToFloat32(), vtx1.pos.z.ToFloat32(), vtx1.pos.w.ToFloat32(),
-            vtx2.pos.x.ToFloat32(), vtx2.pos.y.ToFloat32(), vtx2.pos.z.ToFloat32(),
-            vtx2.pos.w.ToFloat32(), vtx0.screenpos.x.ToFloat32(), vtx0.screenpos.y.ToFloat32(),
-            vtx0.screenpos.z.ToFloat32(), vtx1.screenpos.x.ToFloat32(),
-            vtx1.screenpos.y.ToFloat32(), vtx1.screenpos.z.ToFloat32(),
-            vtx2.screenpos.x.ToFloat32(), vtx2.screenpos.y.ToFloat32(),
-            vtx2.screenpos.z.ToFloat32());
-
-        Rasterizer::ProcessTriangle(vtx0, vtx1, vtx2);
     }
 }
 
-} // namespace Pica::Clipper
+} // namespace SwRenderer

src/video_core/renderer_software/sw_clipper.h

@@ -1,19 +1,87 @@
-// Copyright 2014 Citra Emulator Project
+// Copyright 2023 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
 
+#include "common/common_types.h"
+#include "common/vector_math.h"
+#include "video_core/pica_types.h"
+
 namespace Pica {
-namespace Shader {
+struct TexturingRegs;
-struct OutputVertex;
 }
 
-namespace Clipper {
+namespace SwRenderer {
 
-using Shader::OutputVertex;
+using Pica::f24;
 
-void ProcessTriangle(const OutputVertex& v0, const OutputVertex& v1, const OutputVertex& v2);
+// NOTE: Assuming that rasterizer coordinates are 12.4 fixed-point values
+struct Fix12P4 {
+    Fix12P4() {}
+    Fix12P4(u16 val) : val(val) {}
 
-} // namespace Clipper
+    static Fix12P4 FromFloat24(f24 flt) {
-} // namespace Pica
+        // TODO: Rounding here is necessary to prevent garbage pixels at
+        //       triangle borders. Is it that the correct solution, though?
+        return Fix12P4(static_cast<u16>(round(flt.ToFloat32() * 16.0f)));
+    }
+
+    static u16 FracMask() {
+        return 0xF;
+    }
+    static u16 IntMask() {
+        return static_cast<u16>(~0xF);
+    }
+
+    operator u16() const {
+        return val;
+    }
+
+    bool operator<(const Fix12P4& oth) const {
+        return (u16) * this < (u16)oth;
+    }
+
+private:
+    u16 val;
+};
+
+struct Viewport {
+    f24 halfsize_x;
+    f24 offset_x;
+    f24 halfsize_y;
+    f24 offset_y;
+    f24 zscale;
+    f24 offset_z;
+};
+
+/**
+ * Flips the quaternions if they are opposite to prevent
+ * interpolating them over the wrong direction.
+ */
+void FlipQuaternionIfOpposite(Common::Vec4<f24>& a, const Common::Vec4<f24>& b);
+
+/**
+ * Calculate signed area of the triangle spanned by the three argument vertices.
+ * The sign denotes an orientation.
+ **/
+int SignedArea(const Common::Vec2<Fix12P4>& vtx1, const Common::Vec2<Fix12P4>& vtx2,
+               const Common::Vec2<Fix12P4>& vtx3);
+
+/**
+ * Convert a 3D vector for cube map coordinates to 2D texture coordinates along with the face name.
+ **/
+std::tuple<f24, f24, f24, PAddr> ConvertCubeCoord(f24 u, f24 v, f24 w,
+                                                  const Pica::TexturingRegs& regs);
+
+/**
+ * Triangle filling rules: Pixels on the right-sided edge or on flat bottom edges are not
+ * drawn. Pixels on any other triangle border are drawn. This is implemented with three bias
+ * values which are added to the barycentric coordinates w0, w1 and w2, respectively.
+ * NOTE: These are the PSP filling rules. Not sure if the 3DS uses the same ones...
+ **/
+bool IsRightSideOrFlatBottomEdge(const Common::Vec2<Fix12P4>& vtx,
+                                 const Common::Vec2<Fix12P4>& line1,
+                                 const Common::Vec2<Fix12P4>& line2);
+
+} // namespace SwRenderer

src/video_core/renderer_software/sw_framebuffer.cpp

@@ -3,23 +3,46 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
-#include "common/assert.h"
 #include "common/color.h"
-#include "common/common_types.h"
 #include "common/logging/log.h"
-#include "common/vector_math.h"
 #include "core/hw/gpu.h"
 #include "core/memory.h"
-#include "video_core/pica_state.h"
+#include "video_core/pica_types.h"
 #include "video_core/regs_framebuffer.h"
 #include "video_core/renderer_software/sw_framebuffer.h"
 #include "video_core/utils.h"
-#include "video_core/video_core.h"
 
-namespace Pica::Rasterizer {
+namespace SwRenderer {
 
-void DrawPixel(int x, int y, const Common::Vec4<u8>& color) {
+using Pica::f16;
-    const auto& framebuffer = g_state.regs.framebuffer.framebuffer;
+using Pica::FramebufferRegs;
+
+namespace {
+
+/// Decode/Encode for shadow map format. It is similar to D24S8 format,
+/// but the depth field is in big-endian.
+const Common::Vec2<u32> DecodeD24S8Shadow(const u8* bytes) {
+    return {static_cast<u32>((bytes[0] << 16) | (bytes[1] << 8) | bytes[2]), bytes[3]};
+}
+
+void EncodeD24X8Shadow(u32 depth, u8* bytes) {
+    bytes[2] = depth & 0xFF;
+    bytes[1] = (depth >> 8) & 0xFF;
+    bytes[0] = (depth >> 16) & 0xFF;
+}
+
+void EncodeX24S8Shadow(u8 stencil, u8* bytes) {
+    bytes[3] = stencil;
+}
+} // Anonymous namespace
+
+Framebuffer::Framebuffer(Memory::MemorySystem& memory_, const Pica::FramebufferRegs& regs_)
+    : memory{memory_}, regs{regs_} {}
+
+Framebuffer::~Framebuffer() = default;
+
+void Framebuffer::DrawPixel(int x, int y, const Common::Vec4<u8>& color) const {
+    const auto& framebuffer = regs.framebuffer;
     const PAddr addr = framebuffer.GetColorBufferPhysicalAddress();
 
     // Similarly to textures, the render framebuffer is laid out from bottom to top, too.
@@ -27,33 +50,29 @@ void DrawPixel(int x, int y, const Common::Vec4<u8>& color) {
     y = framebuffer.height - y;
 
     const u32 coarse_y = y & ~7;
-    u32 bytes_per_pixel =
+    const u32 bytes_per_pixel =
         GPU::Regs::BytesPerPixel(GPU::Regs::PixelFormat(framebuffer.color_format.Value()));
-    u32 dst_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) +
+    const u32 dst_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) +
-                     coarse_y * framebuffer.width * bytes_per_pixel;
+                           coarse_y * framebuffer.width * bytes_per_pixel;
-    u8* dst_pixel = VideoCore::g_memory->GetPhysicalPointer(addr) + dst_offset;
+    u8* depth_buffer = memory.GetPhysicalPointer(addr);
+    u8* dst_pixel = depth_buffer + dst_offset;
 
     switch (framebuffer.color_format) {
     case FramebufferRegs::ColorFormat::RGBA8:
         Common::Color::EncodeRGBA8(color, dst_pixel);
         break;
-
     case FramebufferRegs::ColorFormat::RGB8:
         Common::Color::EncodeRGB8(color, dst_pixel);
         break;
-
     case FramebufferRegs::ColorFormat::RGB5A1:
         Common::Color::EncodeRGB5A1(color, dst_pixel);
         break;
-
     case FramebufferRegs::ColorFormat::RGB565:
         Common::Color::EncodeRGB565(color, dst_pixel);
         break;
-
     case FramebufferRegs::ColorFormat::RGBA4:
         Common::Color::EncodeRGBA4(color, dst_pixel);
         break;
-
     default:
         LOG_CRITICAL(Render_Software, "Unknown framebuffer color format {:x}",
                      static_cast<u32>(framebuffer.color_format.Value()));
@@ -61,35 +80,31 @@ void DrawPixel(int x, int y, const Common::Vec4<u8>& color) {
     }
 }
 
-const Common::Vec4<u8> GetPixel(int x, int y) {
+const Common::Vec4<u8> Framebuffer::GetPixel(int x, int y) const {
-    const auto& framebuffer = g_state.regs.framebuffer.framebuffer;
+    const auto& framebuffer = regs.framebuffer;
     const PAddr addr = framebuffer.GetColorBufferPhysicalAddress();
 
     y = framebuffer.height - y;
 
     const u32 coarse_y = y & ~7;
-    u32 bytes_per_pixel =
+    const u32 bytes_per_pixel =
         GPU::Regs::BytesPerPixel(GPU::Regs::PixelFormat(framebuffer.color_format.Value()));
-    u32 src_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) +
+    const u32 src_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) +
-                     coarse_y * framebuffer.width * bytes_per_pixel;
+                           coarse_y * framebuffer.width * bytes_per_pixel;
-    u8* src_pixel = VideoCore::g_memory->GetPhysicalPointer(addr) + src_offset;
+    const u8* color_buffer = memory.GetPhysicalPointer(addr);
+    const u8* src_pixel = color_buffer + src_offset;
 
     switch (framebuffer.color_format) {
     case FramebufferRegs::ColorFormat::RGBA8:
         return Common::Color::DecodeRGBA8(src_pixel);
-
     case FramebufferRegs::ColorFormat::RGB8:
         return Common::Color::DecodeRGB8(src_pixel);
-
     case FramebufferRegs::ColorFormat::RGB5A1:
         return Common::Color::DecodeRGB5A1(src_pixel);
-
     case FramebufferRegs::ColorFormat::RGB565:
         return Common::Color::DecodeRGB565(src_pixel);
-
     case FramebufferRegs::ColorFormat::RGBA4:
         return Common::Color::DecodeRGBA4(src_pixel);
-
     default:
         LOG_CRITICAL(Render_Software, "Unknown framebuffer color format {:x}",
                      static_cast<u32>(framebuffer.color_format.Value()));
@@ -99,19 +114,19 @@ const Common::Vec4<u8> GetPixel(int x, int y) {
     return {0, 0, 0, 0};
 }
 
-u32 GetDepth(int x, int y) {
+u32 Framebuffer::GetDepth(int x, int y) const {
-    const auto& framebuffer = g_state.regs.framebuffer.framebuffer;
+    const auto& framebuffer = regs.framebuffer;
     const PAddr addr = framebuffer.GetDepthBufferPhysicalAddress();
-    u8* depth_buffer = VideoCore::g_memory->GetPhysicalPointer(addr);
 
     y = framebuffer.height - y;
 
     const u32 coarse_y = y & ~7;
-    u32 bytes_per_pixel = FramebufferRegs::BytesPerDepthPixel(framebuffer.depth_format);
+    const u32 bytes_per_pixel = FramebufferRegs::BytesPerDepthPixel(framebuffer.depth_format);
-    u32 stride = framebuffer.width * bytes_per_pixel;
+    const u32 stride = framebuffer.width * bytes_per_pixel;
 
-    u32 src_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) + coarse_y * stride;
+    const u32 src_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) + coarse_y * stride;
-    u8* src_pixel = depth_buffer + src_offset;
+    const u8* depth_buffer = memory.GetPhysicalPointer(addr);
+    const u8* src_pixel = depth_buffer + src_offset;
 
     switch (framebuffer.depth_format) {
     case FramebufferRegs::DepthFormat::D16:
@@ -128,24 +143,23 @@ u32 GetDepth(int x, int y) {
     }
 }
 
-u8 GetStencil(int x, int y) {
+u8 Framebuffer::GetStencil(int x, int y) const {
-    const auto& framebuffer = g_state.regs.framebuffer.framebuffer;
+    const auto& framebuffer = regs.framebuffer;
     const PAddr addr = framebuffer.GetDepthBufferPhysicalAddress();
-    u8* depth_buffer = VideoCore::g_memory->GetPhysicalPointer(addr);
 
     y = framebuffer.height - y;
 
     const u32 coarse_y = y & ~7;
-    u32 bytes_per_pixel = Pica::FramebufferRegs::BytesPerDepthPixel(framebuffer.depth_format);
+    const u32 bytes_per_pixel = Pica::FramebufferRegs::BytesPerDepthPixel(framebuffer.depth_format);
-    u32 stride = framebuffer.width * bytes_per_pixel;
+    const u32 stride = framebuffer.width * bytes_per_pixel;
 
-    u32 src_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) + coarse_y * stride;
+    const u32 src_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) + coarse_y * stride;
-    u8* src_pixel = depth_buffer + src_offset;
+    const u8* depth_buffer = memory.GetPhysicalPointer(addr);
+    const u8* src_pixel = depth_buffer + src_offset;
 
     switch (framebuffer.depth_format) {
     case FramebufferRegs::DepthFormat::D24S8:
         return Common::Color::DecodeD24S8(src_pixel).y;
-
     default:
         LOG_WARNING(
             HW_GPU,
@@ -155,33 +169,30 @@ u8 GetStencil(int x, int y) {
     }
 }
 
-void SetDepth(int x, int y, u32 value) {
+void Framebuffer::SetDepth(int x, int y, u32 value) const {
-    const auto& framebuffer = g_state.regs.framebuffer.framebuffer;
+    const auto& framebuffer = regs.framebuffer;
     const PAddr addr = framebuffer.GetDepthBufferPhysicalAddress();
-    u8* depth_buffer = VideoCore::g_memory->GetPhysicalPointer(addr);
 
     y = framebuffer.height - y;
 
     const u32 coarse_y = y & ~7;
-    u32 bytes_per_pixel = FramebufferRegs::BytesPerDepthPixel(framebuffer.depth_format);
+    const u32 bytes_per_pixel = FramebufferRegs::BytesPerDepthPixel(framebuffer.depth_format);
-    u32 stride = framebuffer.width * bytes_per_pixel;
+    const u32 stride = framebuffer.width * bytes_per_pixel;
 
-    u32 dst_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) + coarse_y * stride;
+    const u32 dst_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) + coarse_y * stride;
+    u8* depth_buffer = memory.GetPhysicalPointer(addr);
     u8* dst_pixel = depth_buffer + dst_offset;
 
     switch (framebuffer.depth_format) {
     case FramebufferRegs::DepthFormat::D16:
         Common::Color::EncodeD16(value, dst_pixel);
         break;
-
     case FramebufferRegs::DepthFormat::D24:
         Common::Color::EncodeD24(value, dst_pixel);
         break;
-
     case FramebufferRegs::DepthFormat::D24S8:
         Common::Color::EncodeD24X8(value, dst_pixel);
         break;
-
     default:
         LOG_CRITICAL(HW_GPU, "Unimplemented depth format {}",
                      static_cast<u32>(framebuffer.depth_format.Value()));
@@ -190,18 +201,18 @@ void SetDepth(int x, int y, u32 value) {
     }
 }
 
-void SetStencil(int x, int y, u8 value) {
+void Framebuffer::SetStencil(int x, int y, u8 value) const {
-    const auto& framebuffer = g_state.regs.framebuffer.framebuffer;
+    const auto& framebuffer = regs.framebuffer;
     const PAddr addr = framebuffer.GetDepthBufferPhysicalAddress();
-    u8* depth_buffer = VideoCore::g_memory->GetPhysicalPointer(addr);
 
     y = framebuffer.height - y;
 
     const u32 coarse_y = y & ~7;
-    u32 bytes_per_pixel = Pica::FramebufferRegs::BytesPerDepthPixel(framebuffer.depth_format);
+    const u32 bytes_per_pixel = Pica::FramebufferRegs::BytesPerDepthPixel(framebuffer.depth_format);
-    u32 stride = framebuffer.width * bytes_per_pixel;
+    const u32 stride = framebuffer.width * bytes_per_pixel;
 
-    u32 dst_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) + coarse_y * stride;
+    const u32 dst_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) + coarse_y * stride;
+    u8* depth_buffer = memory.GetPhysicalPointer(addr);
     u8* dst_pixel = depth_buffer + dst_offset;
 
     switch (framebuffer.depth_format) {
@@ -209,11 +220,9 @@ void SetStencil(int x, int y, u8 value) {
     case Pica::FramebufferRegs::DepthFormat::D24:
         // Nothing to do
         break;
-
     case Pica::FramebufferRegs::DepthFormat::D24S8:
         Common::Color::EncodeX24S8(value, dst_pixel);
         break;
-
     default:
         LOG_CRITICAL(HW_GPU, "Unimplemented depth format {}",
                      static_cast<u32>(framebuffer.depth_format.Value()));
@@ -222,36 +231,65 @@ void SetStencil(int x, int y, u8 value) {
     }
 }
 
+void Framebuffer::DrawShadowMapPixel(int x, int y, u32 depth, u8 stencil) const {
+    const auto& framebuffer = regs.framebuffer;
+    const auto& shadow = regs.shadow;
+    const PAddr addr = framebuffer.GetColorBufferPhysicalAddress();
+
+    y = framebuffer.height - y;
+
+    const u32 coarse_y = y & ~7;
+    u32 bytes_per_pixel = 4;
+    u32 dst_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) +
+                     coarse_y * framebuffer.width * bytes_per_pixel;
+    u8* shadow_buffer = memory.GetPhysicalPointer(addr);
+    u8* dst_pixel = shadow_buffer + dst_offset;
+
+    const auto ref = DecodeD24S8Shadow(dst_pixel);
+    const u32 ref_z = ref.x;
+    const u32 ref_s = ref.y;
+
+    if (depth >= ref_z) {
+        return;
+    }
+
+    if (stencil == 0) {
+        EncodeD24X8Shadow(depth, dst_pixel);
+    } else {
+        const f16 constant = f16::FromRaw(shadow.constant);
+        const f16 linear = f16::FromRaw(shadow.linear);
+        const f16 x_ = f16::FromFloat32(static_cast<float>(depth) / ref_z);
+        const f16 stencil_new = f16::FromFloat32(stencil) / (constant + linear * x_);
+        stencil = static_cast<u8>(std::clamp(stencil_new.ToFloat32(), 0.0f, 255.0f));
+
+        if (stencil < ref_s) {
+            EncodeX24S8Shadow(stencil, dst_pixel);
+        }
+    }
+}
+
 u8 PerformStencilAction(FramebufferRegs::StencilAction action, u8 old_stencil, u8 ref) {
     switch (action) {
     case FramebufferRegs::StencilAction::Keep:
         return old_stencil;
-
     case FramebufferRegs::StencilAction::Zero:
         return 0;
-
     case FramebufferRegs::StencilAction::Replace:
         return ref;
-
     case FramebufferRegs::StencilAction::Increment:
         // Saturated increment
         return std::min<u8>(old_stencil, 254) + 1;
-
     case FramebufferRegs::StencilAction::Decrement:
         // Saturated decrement
         return std::max<u8>(old_stencil, 1) - 1;
-
     case FramebufferRegs::StencilAction::Invert:
         return ~old_stencil;
-
     case FramebufferRegs::StencilAction::IncrementWrap:
         return old_stencil + 1;
-
     case FramebufferRegs::StencilAction::DecrementWrap:
         return old_stencil - 1;
-
     default:
-        LOG_CRITICAL(HW_GPU, "Unknown stencil action {:x}", (int)action);
+        LOG_CRITICAL(HW_GPU, "Unknown stencil action {:x}", static_cast<int>(action));
         UNIMPLEMENTED();
         return 0;
     }
@@ -262,24 +300,21 @@ Common::Vec4<u8> EvaluateBlendEquation(const Common::Vec4<u8>& src,
                                        const Common::Vec4<u8>& dest,
                                        const Common::Vec4<u8>& destfactor,
                                        FramebufferRegs::BlendEquation equation) {
-    Common::Vec4<int> result;
+    Common::Vec4i result;
 
-    auto src_result = (src * srcfactor).Cast<int>();
+    const auto src_result = (src * srcfactor).Cast<s32>();
-    auto dst_result = (dest * destfactor).Cast<int>();
+    const auto dst_result = (dest * destfactor).Cast<s32>();
 
     switch (equation) {
     case FramebufferRegs::BlendEquation::Add:
         result = (src_result + dst_result) / 255;
         break;
-
     case FramebufferRegs::BlendEquation::Subtract:
         result = (src_result - dst_result) / 255;
         break;
-
     case FramebufferRegs::BlendEquation::ReverseSubtract:
         result = (dst_result - src_result) / 255;
         break;
-
     // TODO: How do these two actually work?  OpenGL doesn't include the blend factors in the
     //       min/max computations, but is this what the 3DS actually does?
     case FramebufferRegs::BlendEquation::Min:
@@ -288,14 +323,12 @@ Common::Vec4<u8> EvaluateBlendEquation(const Common::Vec4<u8>& src,
         result.b() = std::min(src.b(), dest.b());
         result.a() = std::min(src.a(), dest.a());
         break;
-
     case FramebufferRegs::BlendEquation::Max:
         result.r() = std::max(src.r(), dest.r());
         result.g() = std::max(src.g(), dest.g());
         result.b() = std::max(src.b(), dest.b());
         result.a() = std::max(src.a(), dest.a());
         break;
-
     default:
         LOG_CRITICAL(HW_GPU, "Unknown RGB blend equation 0x{:x}", equation);
         UNIMPLEMENTED();
@@ -309,103 +342,38 @@ u8 LogicOp(u8 src, u8 dest, FramebufferRegs::LogicOp op) {
     switch (op) {
     case FramebufferRegs::LogicOp::Clear:
         return 0;
-
     case FramebufferRegs::LogicOp::And:
         return src & dest;
-
     case FramebufferRegs::LogicOp::AndReverse:
         return src & ~dest;
-
     case FramebufferRegs::LogicOp::Copy:
         return src;
-
     case FramebufferRegs::LogicOp::Set:
         return 255;
-
     case FramebufferRegs::LogicOp::CopyInverted:
         return ~src;
-
     case FramebufferRegs::LogicOp::NoOp:
         return dest;
-
     case FramebufferRegs::LogicOp::Invert:
         return ~dest;
-
     case FramebufferRegs::LogicOp::Nand:
         return ~(src & dest);
-
     case FramebufferRegs::LogicOp::Or:
         return src | dest;
-
     case FramebufferRegs::LogicOp::Nor:
         return ~(src | dest);
-
     case FramebufferRegs::LogicOp::Xor:
         return src ^ dest;
-
     case FramebufferRegs::LogicOp::Equiv:
         return ~(src ^ dest);
-
     case FramebufferRegs::LogicOp::AndInverted:
         return ~src & dest;
-
     case FramebufferRegs::LogicOp::OrReverse:
         return src | ~dest;
-
     case FramebufferRegs::LogicOp::OrInverted:
         return ~src | dest;
     }
-
     UNREACHABLE();
 };
 
-// Decode/Encode for shadow map format. It is similar to D24S8 format, but the depth field is in
+} // namespace SwRenderer
-// big-endian
-static const Common::Vec2<u32> DecodeD24S8Shadow(const u8* bytes) {
-    return {static_cast<u32>((bytes[0] << 16) | (bytes[1] << 8) | bytes[2]), bytes[3]};
-}
-
-static void EncodeD24X8Shadow(u32 depth, u8* bytes) {
-    bytes[2] = depth & 0xFF;
-    bytes[1] = (depth >> 8) & 0xFF;
-    bytes[0] = (depth >> 16) & 0xFF;
-}
-
-static void EncodeX24S8Shadow(u8 stencil, u8* bytes) {
-    bytes[3] = stencil;
-}
-
-void DrawShadowMapPixel(int x, int y, u32 depth, u8 stencil) {
-    const auto& framebuffer = g_state.regs.framebuffer.framebuffer;
-    const auto& shadow = g_state.regs.framebuffer.shadow;
-    const PAddr addr = framebuffer.GetColorBufferPhysicalAddress();
-
-    y = framebuffer.height - y;
-
-    const u32 coarse_y = y & ~7;
-    u32 bytes_per_pixel = 4;
-    u32 dst_offset = VideoCore::GetMortonOffset(x, y, bytes_per_pixel) +
-                     coarse_y * framebuffer.width * bytes_per_pixel;
-    u8* dst_pixel = VideoCore::g_memory->GetPhysicalPointer(addr) + dst_offset;
-
-    auto ref = DecodeD24S8Shadow(dst_pixel);
-    u32 ref_z = ref.x;
-    u32 ref_s = ref.y;
-
-    if (depth < ref_z) {
-        if (stencil == 0) {
-            EncodeD24X8Shadow(depth, dst_pixel);
-        } else {
-            float16 constant = float16::FromRaw(shadow.constant);
-            float16 linear = float16::FromRaw(shadow.linear);
-            float16 x_ = float16::FromFloat32(static_cast<float>(depth) / ref_z);
-            float16 stencil_new = float16::FromFloat32(stencil) / (constant + linear * x_);
-            stencil = static_cast<u8>(std::clamp(stencil_new.ToFloat32(), 0.0f, 255.0f));
-
-            if (stencil < ref_s)
-                EncodeX24S8Shadow(stencil, dst_pixel);
-        }
-    }
-}
-
-} // namespace Pica::Rasterizer

src/video_core/renderer_software/sw_framebuffer.h

@@ -8,24 +8,55 @@
 #include "common/vector_math.h"
 #include "video_core/regs_framebuffer.h"
 
-namespace Pica::Rasterizer {
+namespace Memory {
+class MemorySystem;
+}
 
-void DrawPixel(int x, int y, const Common::Vec4<u8>& color);
+namespace Pica {
-const Common::Vec4<u8> GetPixel(int x, int y);
+struct FramebufferRegs;
-u32 GetDepth(int x, int y);
+}
-u8 GetStencil(int x, int y);
+
-void SetDepth(int x, int y, u32 value);
+namespace SwRenderer {
-void SetStencil(int x, int y, u8 value);
+
-u8 PerformStencilAction(FramebufferRegs::StencilAction action, u8 old_stencil, u8 ref);
+class Framebuffer {
+public:
+    explicit Framebuffer(Memory::MemorySystem& memory, const Pica::FramebufferRegs& framebuffer);
+    ~Framebuffer();
+
+    /// Draws a pixel at the specified coordinates.
+    void DrawPixel(int x, int y, const Common::Vec4<u8>& color) const;
+
+    /// Returns the current color at the specified coordinates.
+    [[nodiscard]] const Common::Vec4<u8> GetPixel(int x, int y) const;
+
+    /// Returns the depth value at the specified coordinates.
+    [[nodiscard]] u32 GetDepth(int x, int y) const;
+
+    /// Returns the stencil value at the specified coordinates.
+    [[nodiscard]] u8 GetStencil(int x, int y) const;
+
+    /// Stores the provided depth value at the specified coordinates.
+    void SetDepth(int x, int y, u32 value) const;
+
+    /// Stores the provided stencil value at the specified coordinates.
+    void SetStencil(int x, int y, u8 value) const;
+
+    /// Draws a pixel to the shadow buffer.
+    void DrawShadowMapPixel(int x, int y, u32 depth, u8 stencil) const;
+
+private:
+    Memory::MemorySystem& memory;
+    const Pica::FramebufferRegs& regs;
+};
+
+u8 PerformStencilAction(Pica::FramebufferRegs::StencilAction action, u8 old_stencil, u8 ref);
 
 Common::Vec4<u8> EvaluateBlendEquation(const Common::Vec4<u8>& src,
                                        const Common::Vec4<u8>& srcfactor,
                                        const Common::Vec4<u8>& dest,
                                        const Common::Vec4<u8>& destfactor,
-                                       FramebufferRegs::BlendEquation equation);
+                                       Pica::FramebufferRegs::BlendEquation equation);
 
-u8 LogicOp(u8 src, u8 dest, FramebufferRegs::LogicOp op);
+u8 LogicOp(u8 src, u8 dest, Pica::FramebufferRegs::LogicOp op);
 
-void DrawShadowMapPixel(int x, int y, u32 depth, u8 stencil);
+} // namespace SwRenderer
-
-} // namespace Pica::Rasterizer

src/video_core/renderer_software/sw_lighting.cpp

@@ -5,7 +5,10 @@
 #include <algorithm>
 #include "video_core/renderer_software/sw_lighting.h"
 
-namespace Pica {
+namespace SwRenderer {
+
+using Pica::f16;
+using Pica::LightingRegs;
 
 static float LookupLightingLut(const Pica::State::Lighting& lighting, std::size_t lut_index,
                                u8 index, float delta) {
@@ -14,18 +17,18 @@ static float LookupLightingLut(const Pica::State::Lighting& lighting, std::size_
 
     const auto& lut = lighting.luts[lut_index][index];
 
-    float lut_value = lut.ToFloat();
+    const float lut_value = lut.ToFloat();
-    float lut_diff = lut.DiffToFloat();
+    const float lut_diff = lut.DiffToFloat();
 
     return lut_value + lut_diff * delta;
 }
 
-std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
+std::pair<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
     const Pica::LightingRegs& lighting, const Pica::State::Lighting& lighting_state,
-    const Common::Quaternion<float>& normquat, const Common::Vec3<float>& view,
+    const Common::Quaternion<f32>& normquat, const Common::Vec3f& view,
-    const Common::Vec4<u8> (&texture_color)[4]) {
+    std::span<const Common::Vec4<u8>, 4> texture_color) {
 
-    Common::Vec4<float> shadow;
+    Common::Vec4f shadow;
     if (lighting.config0.enable_shadow) {
         shadow = texture_color[lighting.config0.shadow_selector].Cast<float>() / 255.0f;
         if (lighting.config0.shadow_invert) {
@@ -35,16 +38,16 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
         shadow = Common::MakeVec(1.0f, 1.0f, 1.0f, 1.0f);
     }
 
-    Common::Vec3<float> surface_normal{};
+    Common::Vec3f surface_normal{};
-    Common::Vec3<float> surface_tangent{};
+    Common::Vec3f surface_tangent{};
 
     if (lighting.config0.bump_mode != LightingRegs::LightingBumpMode::None) {
-        Common::Vec3<float> perturbation =
+        Common::Vec3f perturbation =
             texture_color[lighting.config0.bump_selector].xyz().Cast<float>() / 127.5f -
             Common::MakeVec(1.0f, 1.0f, 1.0f);
         if (lighting.config0.bump_mode == LightingRegs::LightingBumpMode::NormalMap) {
             if (!lighting.config0.disable_bump_renorm) {
-                const float z_square = 1 - perturbation.xy().Length2();
+                const f32 z_square = 1 - perturbation.xy().Length2();
                 perturbation.z = std::sqrt(std::max(z_square, 0.0f));
             }
             surface_normal = perturbation;
@@ -65,66 +68,64 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
     auto normal = Common::QuaternionRotate(normquat, surface_normal);
     auto tangent = Common::QuaternionRotate(normquat, surface_tangent);
 
-    Common::Vec4<float> diffuse_sum = {0.0f, 0.0f, 0.0f, 1.0f};
+    Common::Vec4f diffuse_sum = {0.0f, 0.0f, 0.0f, 1.0f};
-    Common::Vec4<float> specular_sum = {0.0f, 0.0f, 0.0f, 1.0f};
+    Common::Vec4f specular_sum = {0.0f, 0.0f, 0.0f, 1.0f};
 
-    for (unsigned light_index = 0; light_index <= lighting.max_light_index; ++light_index) {
+    for (u32 light_index = 0; light_index <= lighting.max_light_index; ++light_index) {
-        unsigned num = lighting.light_enable.GetNum(light_index);
+        u32 num = lighting.light_enable.GetNum(light_index);
         const auto& light_config = lighting.light[num];
 
-        Common::Vec3<float> refl_value = {};
+        const Common::Vec3f position = {f16::FromRaw(light_config.x).ToFloat32(),
-        Common::Vec3<float> position = {float16::FromRaw(light_config.x).ToFloat32(),
+                                        f16::FromRaw(light_config.y).ToFloat32(),
-                                        float16::FromRaw(light_config.y).ToFloat32(),
+                                        f16::FromRaw(light_config.z).ToFloat32()};
-                                        float16::FromRaw(light_config.z).ToFloat32()};
+        Common::Vec3f refl_value{};
-        Common::Vec3<float> light_vector;
+        Common::Vec3f light_vector{};
 
-        if (light_config.config.directional)
+        if (light_config.config.directional) {
             light_vector = position;
-        else
+        } else {
             light_vector = position + view;
+        }
 
-        [[maybe_unused]] float length = light_vector.Normalize();
+        [[maybe_unused]] const f32 length = light_vector.Normalize();
 
-        Common::Vec3<float> norm_view = view.Normalized();
+        Common::Vec3f norm_view = view.Normalized();
-        Common::Vec3<float> half_vector = norm_view + light_vector;
+        Common::Vec3f half_vector = norm_view + light_vector;
 
-        float dist_atten = 1.0f;
+        f32 dist_atten = 1.0f;
         if (!lighting.IsDistAttenDisabled(num)) {
-            float scale = Pica::float20::FromRaw(light_config.dist_atten_scale).ToFloat32();
+            const f32 scale = Pica::f20::FromRaw(light_config.dist_atten_scale).ToFloat32();
-            float bias = Pica::float20::FromRaw(light_config.dist_atten_bias).ToFloat32();
+            const f32 bias = Pica::f20::FromRaw(light_config.dist_atten_bias).ToFloat32();
-            std::size_t lut =
+            const std::size_t lut =
                 static_cast<std::size_t>(LightingRegs::LightingSampler::DistanceAttenuation) + num;
 
-            float sample_loc = std::clamp(scale * length + bias, 0.0f, 1.0f);
+            const f32 sample_loc = std::clamp(scale * length + bias, 0.0f, 1.0f);
 
-            u8 lutindex =
+            const u8 lutindex =
                 static_cast<u8>(std::clamp(std::floor(sample_loc * 256.0f), 0.0f, 255.0f));
-            float delta = sample_loc * 256 - lutindex;
+            const f32 delta = sample_loc * 256 - lutindex;
+
             dist_atten = LookupLightingLut(lighting_state, lut, lutindex, delta);
         }
 
-        auto GetLutValue = [&](LightingRegs::LightingLutInput input, bool abs,
+        auto get_lut_value = [&](LightingRegs::LightingLutInput input, bool abs,
-                               LightingRegs::LightingScale scale_enum,
+                                 LightingRegs::LightingScale scale_enum,
-                               LightingRegs::LightingSampler sampler) {
+                                 LightingRegs::LightingSampler sampler) {
-            float result = 0.0f;
+            f32 result = 0.0f;
 
             switch (input) {
             case LightingRegs::LightingLutInput::NH:
                 result = Common::Dot(normal, half_vector.Normalized());
                 break;
-
             case LightingRegs::LightingLutInput::VH:
                 result = Common::Dot(norm_view, half_vector.Normalized());
                 break;
-
             case LightingRegs::LightingLutInput::NV:
                 result = Common::Dot(normal, norm_view);
                 break;
-
             case LightingRegs::LightingLutInput::LN:
                 result = Common::Dot(light_vector, normal);
                 break;
-
             case LightingRegs::LightingLutInput::SP: {
                 Common::Vec3<s32> spot_dir{light_config.spot_x.Value(), light_config.spot_y.Value(),
                                            light_config.spot_z.Value()};
@@ -133,8 +134,8 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
             }
             case LightingRegs::LightingLutInput::CP:
                 if (lighting.config0.config == LightingRegs::LightingConfig::Config7) {
-                    const Common::Vec3<float> norm_half_vector = half_vector.Normalized();
+                    const Common::Vec3f norm_half_vector = half_vector.Normalized();
-                    const Common::Vec3<float> half_vector_proj =
+                    const Common::Vec3f half_vector_proj =
                         norm_half_vector - normal * Common::Dot(normal, norm_half_vector);
                     result = Common::Dot(half_vector_proj, tangent);
                 } else {
@@ -148,58 +149,60 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
             }
 
             u8 index;
-            float delta;
+            f32 delta;
 
             if (abs) {
-                if (light_config.config.two_sided_diffuse)
+                if (light_config.config.two_sided_diffuse) {
                     result = std::abs(result);
-                else
+                } else {
                     result = std::max(result, 0.0f);
+                }
 
-                float flr = std::floor(result * 256.0f);
+                const f32 flr = std::floor(result * 256.0f);
                 index = static_cast<u8>(std::clamp(flr, 0.0f, 255.0f));
                 delta = result * 256 - index;
             } else {
-                float flr = std::floor(result * 128.0f);
+                const f32 flr = std::floor(result * 128.0f);
-                s8 signed_index = static_cast<s8>(std::clamp(flr, -128.0f, 127.0f));
+                const s8 signed_index = static_cast<s8>(std::clamp(flr, -128.0f, 127.0f));
                 delta = result * 128.0f - signed_index;
                 index = static_cast<u8>(signed_index);
             }
 
-            float scale = lighting.lut_scale.GetScale(scale_enum);
+            const f32 scale = lighting.lut_scale.GetScale(scale_enum);
             return scale * LookupLightingLut(lighting_state, static_cast<std::size_t>(sampler),
                                              index, delta);
         };
 
         // If enabled, compute spot light attenuation value
-        float spot_atten = 1.0f;
+        f32 spot_atten = 1.0f;
         if (!lighting.IsSpotAttenDisabled(num) &&
             LightingRegs::IsLightingSamplerSupported(
                 lighting.config0.config, LightingRegs::LightingSampler::SpotlightAttenuation)) {
             auto lut = LightingRegs::SpotlightAttenuationSampler(num);
-            spot_atten = GetLutValue(lighting.lut_input.sp, lighting.abs_lut_input.disable_sp == 0,
+            spot_atten =
-                                     lighting.lut_scale.sp, lut);
+                get_lut_value(lighting.lut_input.sp, lighting.abs_lut_input.disable_sp == 0,
+                              lighting.lut_scale.sp, lut);
         }
 
         // Specular 0 component
-        float d0_lut_value = 1.0f;
+        f32 d0_lut_value = 1.0f;
         if (lighting.config1.disable_lut_d0 == 0 &&
             LightingRegs::IsLightingSamplerSupported(
                 lighting.config0.config, LightingRegs::LightingSampler::Distribution0)) {
             d0_lut_value =
-                GetLutValue(lighting.lut_input.d0, lighting.abs_lut_input.disable_d0 == 0,
+                get_lut_value(lighting.lut_input.d0, lighting.abs_lut_input.disable_d0 == 0,
-                            lighting.lut_scale.d0, LightingRegs::LightingSampler::Distribution0);
+                              lighting.lut_scale.d0, LightingRegs::LightingSampler::Distribution0);
         }
 
-        Common::Vec3<float> specular_0 = d0_lut_value * light_config.specular_0.ToVec3f();
+        Common::Vec3f specular_0 = d0_lut_value * light_config.specular_0.ToVec3f();
 
         // If enabled, lookup ReflectRed value, otherwise, 1.0 is used
         if (lighting.config1.disable_lut_rr == 0 &&
             LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
                                                      LightingRegs::LightingSampler::ReflectRed)) {
             refl_value.x =
-                GetLutValue(lighting.lut_input.rr, lighting.abs_lut_input.disable_rr == 0,
+                get_lut_value(lighting.lut_input.rr, lighting.abs_lut_input.disable_rr == 0,
-                            lighting.lut_scale.rr, LightingRegs::LightingSampler::ReflectRed);
+                              lighting.lut_scale.rr, LightingRegs::LightingSampler::ReflectRed);
         } else {
             refl_value.x = 1.0f;
         }
@@ -209,8 +212,8 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
             LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
                                                      LightingRegs::LightingSampler::ReflectGreen)) {
             refl_value.y =
-                GetLutValue(lighting.lut_input.rg, lighting.abs_lut_input.disable_rg == 0,
+                get_lut_value(lighting.lut_input.rg, lighting.abs_lut_input.disable_rg == 0,
-                            lighting.lut_scale.rg, LightingRegs::LightingSampler::ReflectGreen);
+                              lighting.lut_scale.rg, LightingRegs::LightingSampler::ReflectGreen);
         } else {
             refl_value.y = refl_value.x;
         }
@@ -220,24 +223,23 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
             LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
                                                      LightingRegs::LightingSampler::ReflectBlue)) {
             refl_value.z =
-                GetLutValue(lighting.lut_input.rb, lighting.abs_lut_input.disable_rb == 0,
+                get_lut_value(lighting.lut_input.rb, lighting.abs_lut_input.disable_rb == 0,
-                            lighting.lut_scale.rb, LightingRegs::LightingSampler::ReflectBlue);
+                              lighting.lut_scale.rb, LightingRegs::LightingSampler::ReflectBlue);
         } else {
             refl_value.z = refl_value.x;
         }
 
         // Specular 1 component
-        float d1_lut_value = 1.0f;
+        f32 d1_lut_value = 1.0f;
         if (lighting.config1.disable_lut_d1 == 0 &&
             LightingRegs::IsLightingSamplerSupported(
                 lighting.config0.config, LightingRegs::LightingSampler::Distribution1)) {
             d1_lut_value =
-                GetLutValue(lighting.lut_input.d1, lighting.abs_lut_input.disable_d1 == 0,
+                get_lut_value(lighting.lut_input.d1, lighting.abs_lut_input.disable_d1 == 0,
-                            lighting.lut_scale.d1, LightingRegs::LightingSampler::Distribution1);
+                              lighting.lut_scale.d1, LightingRegs::LightingSampler::Distribution1);
         }
 
-        Common::Vec3<float> specular_1 =
+        Common::Vec3f specular_1 = d1_lut_value * refl_value * light_config.specular_1.ToVec3f();
-            d1_lut_value * refl_value * light_config.specular_1.ToVec3f();
 
         // Fresnel
         // Note: only the last entry in the light slots applies the Fresnel factor
@@ -245,9 +247,9 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
             LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
                                                      LightingRegs::LightingSampler::Fresnel)) {
 
-            float lut_value =
+            const f32 lut_value =
-                GetLutValue(lighting.lut_input.fr, lighting.abs_lut_input.disable_fr == 0,
+                get_lut_value(lighting.lut_input.fr, lighting.abs_lut_input.disable_fr == 0,
-                            lighting.lut_scale.fr, LightingRegs::LightingSampler::Fresnel);
+                              lighting.lut_scale.fr, LightingRegs::LightingSampler::Fresnel);
 
             // Enabled for diffuse lighting alpha component
             if (lighting.config0.enable_primary_alpha) {
@@ -261,18 +263,19 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
         }
 
         auto dot_product = Common::Dot(light_vector, normal);
-        if (light_config.config.two_sided_diffuse)
+        if (light_config.config.two_sided_diffuse) {
             dot_product = std::abs(dot_product);
-        else
+        } else {
             dot_product = std::max(dot_product, 0.0f);
+        }
 
-        float clamp_highlights = 1.0f;
+        f32 clamp_highlights = 1.0f;
         if (lighting.config0.clamp_highlights) {
             clamp_highlights = dot_product == 0.0f ? 0.0f : 1.0f;
         }
 
         if (light_config.config.geometric_factor_0 || light_config.config.geometric_factor_1) {
-            float geo_factor = half_vector.Length2();
+            f32 geo_factor = half_vector.Length2();
             geo_factor = geo_factor == 0.0f ? 0.0f : std::min(dot_product / geo_factor, 1.0f);
             if (light_config.config.geometric_factor_0) {
                 specular_0 *= geo_factor;
@@ -315,17 +318,17 @@ std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
 
     diffuse_sum += Common::MakeVec(lighting.global_ambient.ToVec3f(), 0.0f);
 
-    auto diffuse = Common::MakeVec<float>(std::clamp(diffuse_sum.x, 0.0f, 1.0f) * 255,
+    const auto diffuse = Common::MakeVec(std::clamp(diffuse_sum.x, 0.0f, 1.0f) * 255,
-                                          std::clamp(diffuse_sum.y, 0.0f, 1.0f) * 255,
+                                         std::clamp(diffuse_sum.y, 0.0f, 1.0f) * 255,
-                                          std::clamp(diffuse_sum.z, 0.0f, 1.0f) * 255,
+                                         std::clamp(diffuse_sum.z, 0.0f, 1.0f) * 255,
-                                          std::clamp(diffuse_sum.w, 0.0f, 1.0f) * 255)
+                                         std::clamp(diffuse_sum.w, 0.0f, 1.0f) * 255)
-                       .Cast<u8>();
+                             .Cast<u8>();
-    auto specular = Common::MakeVec<float>(std::clamp(specular_sum.x, 0.0f, 1.0f) * 255,
+    const auto specular = Common::MakeVec(std::clamp(specular_sum.x, 0.0f, 1.0f) * 255,
-                                           std::clamp(specular_sum.y, 0.0f, 1.0f) * 255,
+                                          std::clamp(specular_sum.y, 0.0f, 1.0f) * 255,
-                                           std::clamp(specular_sum.z, 0.0f, 1.0f) * 255,
+                                          std::clamp(specular_sum.z, 0.0f, 1.0f) * 255,
-                                           std::clamp(specular_sum.w, 0.0f, 1.0f) * 255)
+                                          std::clamp(specular_sum.w, 0.0f, 1.0f) * 255)
-                        .Cast<u8>();
+                              .Cast<u8>();
-    return std::make_tuple(diffuse, specular);
+    return std::make_pair(diffuse, specular);
 }
 
-} // namespace Pica
+} // namespace SwRenderer

src/video_core/renderer_software/sw_lighting.h

@@ -4,16 +4,18 @@
 
 #pragma once
 
-#include <tuple>
+#include <span>
+#include <utility>
+
 #include "common/quaternion.h"
 #include "common/vector_math.h"
 #include "video_core/pica_state.h"
 
-namespace Pica {
+namespace SwRenderer {
 
-std::tuple<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
+std::pair<Common::Vec4<u8>, Common::Vec4<u8>> ComputeFragmentsColors(
     const Pica::LightingRegs& lighting, const Pica::State::Lighting& lighting_state,
-    const Common::Quaternion<float>& normquat, const Common::Vec3<float>& view,
+    const Common::Quaternion<f32>& normquat, const Common::Vec3f& view,
-    const Common::Vec4<u8> (&texture_color)[4]);
+    std::span<const Common::Vec4<u8>, 4> texture_color);
 
-} // namespace Pica
+} // namespace SwRenderer

src/video_core/renderer_software/sw_proctex.cpp

@@ -4,17 +4,18 @@
 
 #include <array>
 #include <cmath>
-#include "common/math_util.h"
 #include "video_core/renderer_software/sw_proctex.h"
 
-namespace Pica::Rasterizer {
+namespace SwRenderer {
 
-using ProcTexClamp = TexturingRegs::ProcTexClamp;
+namespace {
-using ProcTexShift = TexturingRegs::ProcTexShift;
+using ProcTexClamp = Pica::TexturingRegs::ProcTexClamp;
-using ProcTexCombiner = TexturingRegs::ProcTexCombiner;
+using ProcTexShift = Pica::TexturingRegs::ProcTexShift;
-using ProcTexFilter = TexturingRegs::ProcTexFilter;
+using ProcTexCombiner = Pica::TexturingRegs::ProcTexCombiner;
+using ProcTexFilter = Pica::TexturingRegs::ProcTexFilter;
+using Pica::f16;
 
-static float LookupLUT(const std::array<State::ProcTex::ValueEntry, 128>& lut, float coord) {
+float LookupLUT(const std::array<Pica::State::ProcTex::ValueEntry, 128>& lut, float coord) {
     // For NoiseLUT/ColorMap/AlphaMap, coord=0.0 is lut[0], coord=127.0/128.0 is lut[127] and
     // coord=1.0 is lut[127]+lut_diff[127]. For other indices, the result is interpolated using
     // value entries and difference entries.
@@ -26,13 +27,13 @@ static float LookupLUT(const std::array<State::ProcTex::ValueEntry, 128>& lut, f
 
 // These function are used to generate random noise for procedural texture. Their results are
 // verified against real hardware, but it's not known if the algorithm is the same as hardware.
-static unsigned int NoiseRand1D(unsigned int v) {
+unsigned int NoiseRand1D(unsigned int v) {
     static constexpr std::array<unsigned int, 16> table{
         {0, 4, 10, 8, 4, 9, 7, 12, 5, 15, 13, 14, 11, 15, 2, 11}};
     return ((v % 9 + 2) * 3 & 0xF) ^ table[(v / 9) & 0xF];
 }
 
-static float NoiseRand2D(unsigned int x, unsigned int y) {
+float NoiseRand2D(unsigned int x, unsigned int y) {
     static constexpr std::array<unsigned int, 16> table{
         {10, 2, 15, 8, 0, 7, 4, 5, 5, 13, 2, 6, 13, 9, 3, 14}};
     unsigned int u2 = NoiseRand1D(x);
@@ -45,11 +46,12 @@ static float NoiseRand2D(unsigned int x, unsigned int y) {
     return -1.0f + v2 * 2.0f / 15.0f;
 }
 
-static float NoiseCoef(float u, float v, const TexturingRegs& regs, const State::ProcTex& state) {
+float NoiseCoef(float u, float v, const Pica::TexturingRegs& regs,
-    const float freq_u = float16::FromRaw(regs.proctex_noise_frequency.u).ToFloat32();
+                const Pica::State::ProcTex& state) {
-    const float freq_v = float16::FromRaw(regs.proctex_noise_frequency.v).ToFloat32();
+    const float freq_u = f16::FromRaw(regs.proctex_noise_frequency.u).ToFloat32();
-    const float phase_u = float16::FromRaw(regs.proctex_noise_u.phase).ToFloat32();
+    const float freq_v = f16::FromRaw(regs.proctex_noise_frequency.v).ToFloat32();
-    const float phase_v = float16::FromRaw(regs.proctex_noise_v.phase).ToFloat32();
+    const float phase_u = f16::FromRaw(regs.proctex_noise_u.phase).ToFloat32();
+    const float phase_v = f16::FromRaw(regs.proctex_noise_v.phase).ToFloat32();
     const float x = 9 * freq_u * std::abs(u + phase_u);
     const float y = 9 * freq_v * std::abs(v + phase_v);
     const int x_int = static_cast<int>(x);
@@ -66,7 +68,7 @@ static float NoiseCoef(float u, float v, const TexturingRegs& regs, const State:
     return Common::BilinearInterp(g0, g1, g2, g3, x_noise, y_noise);
 }
 
-static float GetShiftOffset(float v, ProcTexShift mode, ProcTexClamp clamp_mode) {
+float GetShiftOffset(float v, ProcTexShift mode, ProcTexClamp clamp_mode) {
     const float offset = (clamp_mode == ProcTexClamp::MirroredRepeat) ? 1 : 0.5f;
     switch (mode) {
     case ProcTexShift::None:
@@ -81,7 +83,7 @@ static float GetShiftOffset(float v, ProcTexShift mode, ProcTexClamp clamp_mode)
     }
 };
 
-static void ClampCoord(float& coord, ProcTexClamp mode) {
+void ClampCoord(float& coord, ProcTexClamp mode) {
     switch (mode) {
     case ProcTexClamp::ToZero:
         if (coord > 1.0f)
@@ -112,8 +114,8 @@ static void ClampCoord(float& coord, ProcTexClamp mode) {
     }
 }
 
-static float CombineAndMap(float u, float v, ProcTexCombiner combiner,
+float CombineAndMap(float u, float v, ProcTexCombiner combiner,
-                           const std::array<State::ProcTex::ValueEntry, 128>& map_table) {
+                    const std::array<Pica::State::ProcTex::ValueEntry, 128>& map_table) {
     float f;
     switch (combiner) {
     case ProcTexCombiner::U:
@@ -122,28 +124,28 @@ static float CombineAndMap(float u, float v, ProcTexCombiner combiner,
     case ProcTexCombiner::U2:
         f = u * u;
         break;
-    case TexturingRegs::ProcTexCombiner::V:
+    case ProcTexCombiner::V:
         f = v;
         break;
-    case TexturingRegs::ProcTexCombiner::V2:
+    case ProcTexCombiner::V2:
         f = v * v;
         break;
-    case TexturingRegs::ProcTexCombiner::Add:
+    case ProcTexCombiner::Add:
         f = (u + v) * 0.5f;
         break;
-    case TexturingRegs::ProcTexCombiner::Add2:
+    case ProcTexCombiner::Add2:
         f = (u * u + v * v) * 0.5f;
         break;
-    case TexturingRegs::ProcTexCombiner::SqrtAdd2:
+    case ProcTexCombiner::SqrtAdd2:
         f = std::min(std::sqrt(u * u + v * v), 1.0f);
         break;
-    case TexturingRegs::ProcTexCombiner::Min:
+    case ProcTexCombiner::Min:
         f = std::min(u, v);
         break;
-    case TexturingRegs::ProcTexCombiner::Max:
+    case ProcTexCombiner::Max:
         f = std::max(u, v);
         break;
-    case TexturingRegs::ProcTexCombiner::RMax:
+    case ProcTexCombiner::RMax:
         f = std::min(((u + v) * 0.5f + std::sqrt(u * u + v * v)) * 0.5f, 1.0f);
         break;
     default:
@@ -153,8 +155,10 @@ static float CombineAndMap(float u, float v, ProcTexCombiner combiner,
     }
     return LookupLUT(map_table, f);
 }
+} // Anonymous namespace
 
-Common::Vec4<u8> ProcTex(float u, float v, const TexturingRegs& regs, const State::ProcTex& state) {
+Common::Vec4<u8> ProcTex(float u, float v, const Pica::TexturingRegs& regs,
+                         const Pica::State::ProcTex& state) {
     u = std::abs(u);
     v = std::abs(v);
 
@@ -218,4 +222,4 @@ Common::Vec4<u8> ProcTex(float u, float v, const TexturingRegs& regs, const Stat
     }
 }
 
-} // namespace Pica::Rasterizer
+} // namespace SwRenderer

src/video_core/renderer_software/sw_proctex.h

@@ -8,9 +8,10 @@
 #include "common/vector_math.h"
 #include "video_core/pica_state.h"
 
-namespace Pica::Rasterizer {
+namespace SwRenderer {
 
 /// Generates procedural texture color for the given coordinates
-Common::Vec4<u8> ProcTex(float u, float v, const TexturingRegs& regs, const State::ProcTex& state);
+Common::Vec4<u8> ProcTex(float u, float v, const Pica::TexturingRegs& regs,
+                         const Pica::State::ProcTex& state);
 
-} // namespace Pica::Rasterizer
+} // namespace SwRenderer

src/video_core/renderer_software/sw_rasterizer.cpp

@@ -2,15 +2,937 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "video_core/renderer_software/sw_clipper.h"
+#include <boost/container/static_vector.hpp>
+#include "common/logging/log.h"
+#include "common/microprofile.h"
+#include "common/quaternion.h"
+#include "common/vector_math.h"
+#include "core/memory.h"
+#include "video_core/pica_state.h"
+#include "video_core/pica_types.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_rasterizer.h"
+#include "video_core/renderer_software/sw_texturing.h"
+#include "video_core/shader/shader.h"
+#include "video_core/texture/texture_decode.h"
 
-namespace VideoCore {
+namespace SwRenderer {
+
+using Pica::f24;
+using Pica::FramebufferRegs;
+using Pica::RasterizerRegs;
+using Pica::TexturingRegs;
+using Pica::Texture::LookupTexture;
+using Pica::Texture::TextureInfo;
+
+struct Vertex : Pica::Shader::OutputVertex {
+    Vertex(const OutputVertex& v) : OutputVertex(v) {}
+
+    /// Attributes used to store intermediate results position after perspective divide.
+    Common::Vec3<f24> screenpos;
+
+    /**
+     * Linear interpolation
+     * factor: 0=this, 1=vtx
+     * Note: This function cannot be called after perspective divide.
+     **/
+    void Lerp(f24 factor, const Vertex& vtx) {
+        pos = pos * factor + vtx.pos * (f24::One() - factor);
+        quat = quat * factor + vtx.quat * (f24::One() - factor);
+        color = color * factor + vtx.color * (f24::One() - factor);
+        tc0 = tc0 * factor + vtx.tc0 * (f24::One() - factor);
+        tc1 = tc1 * factor + vtx.tc1 * (f24::One() - factor);
+        tc0_w = tc0_w * factor + vtx.tc0_w * (f24::One() - factor);
+        view = view * factor + vtx.view * (f24::One() - factor);
+        tc2 = tc2 * factor + vtx.tc2 * (f24::One() - factor);
+    }
+
+    /**
+     * Linear interpolation
+     * factor: 0=v0, 1=v1
+     * Note: This function cannot be called after perspective divide.
+     **/
+    static Vertex Lerp(f24 factor, const Vertex& v0, const Vertex& v1) {
+        Vertex ret = v0;
+        ret.Lerp(factor, v1);
+        return ret;
+    }
+};
+
+namespace {
+
+MICROPROFILE_DEFINE(GPU_Rasterization, "GPU", "Rasterization", MP_RGB(50, 50, 240));
+
+struct ClippingEdge {
+public:
+    constexpr ClippingEdge(Common::Vec4<f24> coeffs,
+                           Common::Vec4<f24> bias = Common::Vec4<f24>(f24::Zero(), f24::Zero(),
+                                                                      f24::Zero(), f24::Zero()))
+        : pos(f24::Zero()), coeffs(coeffs), bias(bias) {}
+
+    bool IsInside(const Vertex& vertex) const {
+        return Common::Dot(vertex.pos + bias, coeffs) >= f24::Zero();
+    }
+
+    bool IsOutSide(const Vertex& vertex) const {
+        return !IsInside(vertex);
+    }
+
+    Vertex GetIntersection(const Vertex& v0, const Vertex& v1) const {
+        const f24 dp = Common::Dot(v0.pos + bias, coeffs);
+        const f24 dp_prev = Common::Dot(v1.pos + bias, coeffs);
+        const f24 factor = dp_prev / (dp_prev - dp);
+        return Vertex::Lerp(factor, v0, v1);
+    }
+
+private:
+    [[maybe_unused]] f24 pos;
+    Common::Vec4<f24> coeffs;
+    Common::Vec4<f24> bias;
+};
+
+} // Anonymous namespace
+
+RasterizerSoftware::RasterizerSoftware(Memory::MemorySystem& memory_)
+    : memory{memory_}, state{Pica::g_state}, regs{state.regs}, fb{memory, regs.framebuffer} {}
 
 void RasterizerSoftware::AddTriangle(const Pica::Shader::OutputVertex& v0,
                                      const Pica::Shader::OutputVertex& v1,
                                      const Pica::Shader::OutputVertex& v2) {
-    Pica::Clipper::ProcessTriangle(v0, v1, v2);
+    /**
+     * Clipping a planar n-gon against a plane will remove at least 1 vertex and introduces 2 at
+     * the new edge (or less in degenerate cases). As such, we can say that each clipping plane
+     * introduces at most 1 new vertex to the polygon. Since we start with a triangle and have a
+     * fixed 6 clipping planes, the maximum number of vertices of the clipped polygon is 3 + 6 = 9.
+     **/
+    static constexpr std::size_t MAX_VERTICES = 9;
+
+    boost::container::static_vector<Vertex, MAX_VERTICES> buffer_a = {v0, v1, v2};
+    boost::container::static_vector<Vertex, MAX_VERTICES> buffer_b;
+
+    FlipQuaternionIfOpposite(buffer_a[1].quat, buffer_a[0].quat);
+    FlipQuaternionIfOpposite(buffer_a[2].quat, buffer_a[0].quat);
+
+    auto* output_list = &buffer_a;
+    auto* input_list = &buffer_b;
+
+    // NOTE: We clip against a w=epsilon plane to guarantee that the output has a positive w value.
+    // TODO: Not sure if this is a valid approach. Also should probably instead use the smallest
+    //       epsilon possible within f24 accuracy.
+    static constexpr f24 EPSILON = f24::FromFloat32(0.00001f);
+    static constexpr f24 f0 = f24::Zero();
+    static constexpr f24 f1 = f24::One();
+    static constexpr std::array<ClippingEdge, 7> clipping_edges = {{
+        {Common::MakeVec(-f1, f0, f0, f1)},                                        // x = +w
+        {Common::MakeVec(f1, f0, f0, f1)},                                         // x = -w
+        {Common::MakeVec(f0, -f1, f0, f1)},                                        // y = +w
+        {Common::MakeVec(f0, f1, f0, f1)},                                         // y = -w
+        {Common::MakeVec(f0, f0, -f1, f0)},                                        // z =  0
+        {Common::MakeVec(f0, f0, f1, f1)},                                         // z = -w
+        {Common::MakeVec(f0, f0, f0, f1), Common::Vec4<f24>(f0, f0, f0, EPSILON)}, // w = EPSILON
+    }};
+
+    // Simple implementation of the Sutherland-Hodgman clipping algorithm.
+    // TODO: Make this less inefficient (currently lots of useless buffering overhead happens here)
+    const auto clip = [&](const ClippingEdge& edge) {
+        std::swap(input_list, output_list);
+        output_list->clear();
+
+        const Vertex* reference_vertex = &input_list->back();
+        for (const auto& vertex : *input_list) {
+            // NOTE: This algorithm changes vertex order in some cases!
+            if (edge.IsInside(vertex)) {
+                if (edge.IsOutSide(*reference_vertex)) {
+                    output_list->push_back(edge.GetIntersection(vertex, *reference_vertex));
+                }
+                output_list->push_back(vertex);
+            } else if (edge.IsInside(*reference_vertex)) {
+                output_list->push_back(edge.GetIntersection(vertex, *reference_vertex));
+            }
+            reference_vertex = &vertex;
+        }
+    };
+
+    for (const ClippingEdge& edge : clipping_edges) {
+        clip(edge);
+        if (output_list->size() < 3) {
+            return;
+        }
+    }
+
+    if (state.regs.rasterizer.clip_enable) {
+        const ClippingEdge custom_edge{state.regs.rasterizer.GetClipCoef()};
+        clip(custom_edge);
+        if (output_list->size() < 3) {
+            return;
+        }
+    }
+
+    MakeScreenCoords((*output_list)[0]);
+    MakeScreenCoords((*output_list)[1]);
+
+    for (std::size_t i = 0; i < output_list->size() - 2; i++) {
+        Vertex& vtx0 = (*output_list)[0];
+        Vertex& vtx1 = (*output_list)[i + 1];
+        Vertex& vtx2 = (*output_list)[i + 2];
+
+        MakeScreenCoords(vtx2);
+
+        LOG_TRACE(
+            Render_Software,
+            "Triangle {}/{} at position ({:.3}, {:.3}, {:.3}, {:.3f}), "
+            "({:.3}, {:.3}, {:.3}, {:.3}), ({:.3}, {:.3}, {:.3}, {:.3}) and "
+            "screen position ({:.2}, {:.2}, {:.2}), ({:.2}, {:.2}, {:.2}), ({:.2}, {:.2}, {:.2})",
+            i + 1, output_list->size() - 2, vtx0.pos.x.ToFloat32(), vtx0.pos.y.ToFloat32(),
+            vtx0.pos.z.ToFloat32(), vtx0.pos.w.ToFloat32(), vtx1.pos.x.ToFloat32(),
+            vtx1.pos.y.ToFloat32(), vtx1.pos.z.ToFloat32(), vtx1.pos.w.ToFloat32(),
+            vtx2.pos.x.ToFloat32(), vtx2.pos.y.ToFloat32(), vtx2.pos.z.ToFloat32(),
+            vtx2.pos.w.ToFloat32(), vtx0.screenpos.x.ToFloat32(), vtx0.screenpos.y.ToFloat32(),
+            vtx0.screenpos.z.ToFloat32(), vtx1.screenpos.x.ToFloat32(),
+            vtx1.screenpos.y.ToFloat32(), vtx1.screenpos.z.ToFloat32(),
+            vtx2.screenpos.x.ToFloat32(), vtx2.screenpos.y.ToFloat32(),
+            vtx2.screenpos.z.ToFloat32());
+
+        ProcessTriangle(vtx0, vtx1, vtx2);
+    }
 }
 
-} // namespace VideoCore
+void RasterizerSoftware::MakeScreenCoords(Vertex& vtx) {
+    Viewport viewport{};
+    viewport.halfsize_x = f24::FromRaw(regs.rasterizer.viewport_size_x);
+    viewport.halfsize_y = f24::FromRaw(regs.rasterizer.viewport_size_y);
+    viewport.offset_x = f24::FromFloat32(static_cast<f32>(regs.rasterizer.viewport_corner.x));
+    viewport.offset_y = f24::FromFloat32(static_cast<f32>(regs.rasterizer.viewport_corner.y));
+
+    f24 inv_w = f24::One() / vtx.pos.w;
+    vtx.pos.w = inv_w;
+    vtx.quat *= inv_w;
+    vtx.color *= inv_w;
+    vtx.tc0 *= inv_w;
+    vtx.tc1 *= inv_w;
+    vtx.tc0_w *= inv_w;
+    vtx.view *= inv_w;
+    vtx.tc2 *= inv_w;
+
+    vtx.screenpos[0] = (vtx.pos.x * inv_w + f24::One()) * viewport.halfsize_x + viewport.offset_x;
+    vtx.screenpos[1] = (vtx.pos.y * inv_w + f24::One()) * viewport.halfsize_y + viewport.offset_y;
+    vtx.screenpos[2] = vtx.pos.z * inv_w;
+}
+
+void RasterizerSoftware::ProcessTriangle(const Vertex& v0, const Vertex& v1, const Vertex& v2,
+                                         bool reversed) {
+    MICROPROFILE_SCOPE(GPU_Rasterization);
+
+    // Vertex positions in rasterizer coordinates
+    static auto screen_to_rasterizer_coords = [](const Common::Vec3<f24>& vec) {
+        return Common::Vec3{Fix12P4::FromFloat24(vec.x), Fix12P4::FromFloat24(vec.y),
+                            Fix12P4::FromFloat24(vec.z)};
+    };
+
+    const std::array<Common::Vec3<Fix12P4>, 3> vtxpos = {
+        screen_to_rasterizer_coords(v0.screenpos),
+        screen_to_rasterizer_coords(v1.screenpos),
+        screen_to_rasterizer_coords(v2.screenpos),
+    };
+
+    if (regs.rasterizer.cull_mode == RasterizerRegs::CullMode::KeepAll) {
+        // Make sure we always end up with a triangle wound counter-clockwise
+        if (!reversed && SignedArea(vtxpos[0].xy(), vtxpos[1].xy(), vtxpos[2].xy()) <= 0) {
+            ProcessTriangle(v0, v2, v1, true);
+            return;
+        }
+    } else {
+        if (!reversed && regs.rasterizer.cull_mode == RasterizerRegs::CullMode::KeepClockWise) {
+            // Reverse vertex order and use the CCW code path.
+            ProcessTriangle(v0, v2, v1, true);
+            return;
+        }
+        // Cull away triangles which are wound clockwise.
+        if (SignedArea(vtxpos[0].xy(), vtxpos[1].xy(), vtxpos[2].xy()) <= 0) {
+            return;
+        }
+    }
+
+    u16 min_x = std::min({vtxpos[0].x, vtxpos[1].x, vtxpos[2].x});
+    u16 min_y = std::min({vtxpos[0].y, vtxpos[1].y, vtxpos[2].y});
+    u16 max_x = std::max({vtxpos[0].x, vtxpos[1].x, vtxpos[2].x});
+    u16 max_y = std::max({vtxpos[0].y, vtxpos[1].y, vtxpos[2].y});
+
+    // Convert the scissor box coordinates to 12.4 fixed point
+    const u16 scissor_x1 = static_cast<u16>(regs.rasterizer.scissor_test.x1 << 4);
+    const u16 scissor_y1 = static_cast<u16>(regs.rasterizer.scissor_test.y1 << 4);
+    // x2,y2 have +1 added to cover the entire sub-pixel area
+    const u16 scissor_x2 = static_cast<u16>((regs.rasterizer.scissor_test.x2 + 1) << 4);
+    const u16 scissor_y2 = static_cast<u16>((regs.rasterizer.scissor_test.y2 + 1) << 4);
+
+    if (regs.rasterizer.scissor_test.mode == RasterizerRegs::ScissorMode::Include) {
+        // Calculate the new bounds
+        min_x = std::max(min_x, scissor_x1);
+        min_y = std::max(min_y, scissor_y1);
+        max_x = std::min(max_x, scissor_x2);
+        max_y = std::min(max_y, scissor_y2);
+    }
+
+    min_x &= Fix12P4::IntMask();
+    min_y &= Fix12P4::IntMask();
+    max_x = ((max_x + Fix12P4::FracMask()) & Fix12P4::IntMask());
+    max_y = ((max_y + Fix12P4::FracMask()) & Fix12P4::IntMask());
+
+    const int bias0 =
+        IsRightSideOrFlatBottomEdge(vtxpos[0].xy(), vtxpos[1].xy(), vtxpos[2].xy()) ? -1 : 0;
+    const int bias1 =
+        IsRightSideOrFlatBottomEdge(vtxpos[1].xy(), vtxpos[2].xy(), vtxpos[0].xy()) ? -1 : 0;
+    const int bias2 =
+        IsRightSideOrFlatBottomEdge(vtxpos[2].xy(), vtxpos[0].xy(), vtxpos[1].xy()) ? -1 : 0;
+
+    const auto w_inverse = Common::MakeVec(v0.pos.w, v1.pos.w, v2.pos.w);
+
+    auto textures = regs.texturing.GetTextures();
+    const auto tev_stages = regs.texturing.GetTevStages();
+
+    const bool stencil_action_enable =
+        regs.framebuffer.output_merger.stencil_test.enable &&
+        regs.framebuffer.framebuffer.depth_format == FramebufferRegs::DepthFormat::D24S8;
+    const auto stencil_test = regs.framebuffer.output_merger.stencil_test;
+
+    // Enter rasterization loop, starting at the center of the topleft bounding box corner.
+    // TODO: Not sure if looping through x first might be faster
+    for (u16 y = min_y + 8; y < max_y; y += 0x10) {
+        for (u16 x = min_x + 8; x < max_x; x += 0x10) {
+            // Do not process the pixel if it's inside the scissor box and the scissor mode is set
+            // to Exclude.
+            if (regs.rasterizer.scissor_test.mode == RasterizerRegs::ScissorMode::Exclude) {
+                if (x >= scissor_x1 && x < scissor_x2 && y >= scissor_y1 && y < scissor_y2) {
+                    continue;
+                }
+            }
+
+            // Calculate the barycentric coordinates w0, w1 and w2
+            const s32 w0 = bias0 + SignedArea(vtxpos[1].xy(), vtxpos[2].xy(), {x, y});
+            const s32 w1 = bias1 + SignedArea(vtxpos[2].xy(), vtxpos[0].xy(), {x, y});
+            const s32 w2 = bias2 + SignedArea(vtxpos[0].xy(), vtxpos[1].xy(), {x, y});
+            const s32 wsum = w0 + w1 + w2;
+
+            // If current pixel is not covered by the current primitive
+            if (w0 < 0 || w1 < 0 || w2 < 0) {
+                continue;
+            }
+
+            const auto baricentric_coordinates = Common::MakeVec(
+                f24::FromFloat32(static_cast<f32>(w0)), f24::FromFloat32(static_cast<f32>(w1)),
+                f24::FromFloat32(static_cast<f32>(w2)));
+            const f24 interpolated_w_inverse =
+                f24::One() / Common::Dot(w_inverse, baricentric_coordinates);
+
+            // interpolated_z = z / w
+            const float interpolated_z_over_w =
+                (v0.screenpos[2].ToFloat32() * w0 + v1.screenpos[2].ToFloat32() * w1 +
+                 v2.screenpos[2].ToFloat32() * w2) /
+                wsum;
+
+            // Not fully accurate. About 3 bits in precision are missing.
+            // Z-Buffer (z / w * scale + offset)
+            const float depth_scale =
+                f24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32();
+            const float depth_offset =
+                f24::FromRaw(regs.rasterizer.viewport_depth_near_plane).ToFloat32();
+            float depth = interpolated_z_over_w * depth_scale + depth_offset;
+
+            // Potentially switch to W-Buffer
+            if (regs.rasterizer.depthmap_enable ==
+                Pica::RasterizerRegs::DepthBuffering::WBuffering) {
+                // W-Buffer (z * scale + w * offset = (z / w * scale + offset) * w)
+                depth *= interpolated_w_inverse.ToFloat32() * wsum;
+            }
+
+            // Clamp the result
+            depth = std::clamp(depth, 0.0f, 1.0f);
+
+            /**
+             * Perspective correct attribute interpolation:
+             * Attribute values cannot be calculated by simple linear interpolation since
+             * they are not linear in screen space. For example, when interpolating a
+             * texture coordinate across two vertices, something simple like
+             *     u = (u0*w0 + u1*w1)/(w0+w1)
+             * will not work. However, the attribute value divided by the
+             * clipspace w-coordinate (u/w) and and the inverse w-coordinate (1/w) are linear
+             * in screenspace. Hence, we can linearly interpolate these two independently and
+             * calculate the interpolated attribute by dividing the results.
+             * I.e.
+             *     u_over_w   = ((u0/v0.pos.w)*w0 + (u1/v1.pos.w)*w1)/(w0+w1)
+             *     one_over_w = (( 1/v0.pos.w)*w0 + ( 1/v1.pos.w)*w1)/(w0+w1)
+             *     u = u_over_w / one_over_w
+             *
+             * The generalization to three vertices is straightforward in baricentric coordinates.
+             **/
+            const auto get_interpolated_attribute = [&](f24 attr0, f24 attr1, f24 attr2) {
+                auto attr_over_w = Common::MakeVec(attr0, attr1, attr2);
+                f24 interpolated_attr_over_w = Common::Dot(attr_over_w, baricentric_coordinates);
+                return interpolated_attr_over_w * interpolated_w_inverse;
+            };
+
+            const Common::Vec4<u8> primary_color{
+                static_cast<u8>(
+                    round(get_interpolated_attribute(v0.color.r(), v1.color.r(), v2.color.r())
+                              .ToFloat32() *
+                          255)),
+                static_cast<u8>(
+                    round(get_interpolated_attribute(v0.color.g(), v1.color.g(), v2.color.g())
+                              .ToFloat32() *
+                          255)),
+                static_cast<u8>(
+                    round(get_interpolated_attribute(v0.color.b(), v1.color.b(), v2.color.b())
+                              .ToFloat32() *
+                          255)),
+                static_cast<u8>(
+                    round(get_interpolated_attribute(v0.color.a(), v1.color.a(), v2.color.a())
+                              .ToFloat32() *
+                          255)),
+            };
+
+            std::array<Common::Vec2<f24>, 3> uv;
+            uv[0].u() = get_interpolated_attribute(v0.tc0.u(), v1.tc0.u(), v2.tc0.u());
+            uv[0].v() = get_interpolated_attribute(v0.tc0.v(), v1.tc0.v(), v2.tc0.v());
+            uv[1].u() = get_interpolated_attribute(v0.tc1.u(), v1.tc1.u(), v2.tc1.u());
+            uv[1].v() = get_interpolated_attribute(v0.tc1.v(), v1.tc1.v(), v2.tc1.v());
+            uv[2].u() = get_interpolated_attribute(v0.tc2.u(), v1.tc2.u(), v2.tc2.u());
+            uv[2].v() = get_interpolated_attribute(v0.tc2.v(), v1.tc2.v(), v2.tc2.v());
+
+            // Sample bound texture units.
+            const f24 tc0_w = get_interpolated_attribute(v0.tc0_w, v1.tc0_w, v2.tc0_w);
+            const auto texture_color = TextureColor(uv, textures, tc0_w);
+
+            Common::Vec4<u8> primary_fragment_color{0, 0, 0, 0};
+            Common::Vec4<u8> secondary_fragment_color{0, 0, 0, 0};
+            if (!regs.lighting.disable) {
+                const auto normquat =
+                    Common::Quaternion<f32>{
+                        {get_interpolated_attribute(v0.quat.x, v1.quat.x, v2.quat.x).ToFloat32(),
+                         get_interpolated_attribute(v0.quat.y, v1.quat.y, v2.quat.y).ToFloat32(),
+                         get_interpolated_attribute(v0.quat.z, v1.quat.z, v2.quat.z).ToFloat32()},
+                        get_interpolated_attribute(v0.quat.w, v1.quat.w, v2.quat.w).ToFloat32(),
+                    }
+                        .Normalized();
+
+                const Common::Vec3f view{
+                    get_interpolated_attribute(v0.view.x, v1.view.x, v2.view.x).ToFloat32(),
+                    get_interpolated_attribute(v0.view.y, v1.view.y, v2.view.y).ToFloat32(),
+                    get_interpolated_attribute(v0.view.z, v1.view.z, v2.view.z).ToFloat32(),
+                };
+                std::tie(primary_fragment_color, secondary_fragment_color) = ComputeFragmentsColors(
+                    regs.lighting, state.lighting, normquat, view, texture_color);
+            }
+
+            // Write the TEV stages.
+            Common::Vec4<u8> combiner_output =
+                WriteTevConfig(texture_color, tev_stages, primary_color, primary_fragment_color,
+                               secondary_fragment_color);
+
+            const auto& output_merger = regs.framebuffer.output_merger;
+            if (output_merger.fragment_operation_mode ==
+                FramebufferRegs::FragmentOperationMode::Shadow) {
+                u32 depth_int = static_cast<u32>(depth * 0xFFFFFF);
+                // Use green color as the shadow intensity
+                u8 stencil = combiner_output.y;
+                fb.DrawShadowMapPixel(x >> 4, y >> 4, depth_int, stencil);
+                // Skip the normal output merger pipeline if it is in shadow mode
+                continue;
+            }
+
+            // Does alpha testing happen before or after stencil?
+            if (!DoAlphaTest(combiner_output.a())) {
+                continue;
+            }
+            WriteFog(combiner_output, depth);
+            if (!DoDepthStencilTest(x, y, depth, stencil_action_enable)) {
+                continue;
+            }
+            const auto result = PixelColor(x, y, combiner_output);
+            if (regs.framebuffer.framebuffer.allow_color_write != 0) {
+                fb.DrawPixel(x >> 4, y >> 4, result);
+            }
+        }
+    }
+}
+
+std::array<Common::Vec4<u8>, 4> RasterizerSoftware::TextureColor(
+    std::span<const Common::Vec2<f24>, 3> uv,
+    std::span<const Pica::TexturingRegs::FullTextureConfig, 3> textures, f24 tc0_w) const {
+    std::array<Common::Vec4<u8>, 4> texture_color{};
+    for (u32 i = 0; i < 3; ++i) {
+        const auto& texture = textures[i];
+        if (!texture.enabled) [[unlikely]] {
+            continue;
+        }
+        if (texture.config.address == 0) [[unlikely]] {
+            texture_color[i] = {0, 0, 0, 255};
+            continue;
+        }
+
+        const s32 coordinate_i = (i == 2 && regs.texturing.main_config.texture2_use_coord1) ? 1 : i;
+        f24 u = uv[coordinate_i].u();
+        f24 v = uv[coordinate_i].v();
+
+        // Only unit 0 respects the texturing type (according to 3DBrew)
+        PAddr texture_address = texture.config.GetPhysicalAddress();
+        f24 shadow_z;
+        if (i == 0) {
+            switch (texture.config.type) {
+            case TexturingRegs::TextureConfig::Texture2D:
+                break;
+            case TexturingRegs::TextureConfig::ShadowCube:
+            case TexturingRegs::TextureConfig::TextureCube: {
+                std::tie(u, v, shadow_z, texture_address) =
+                    ConvertCubeCoord(u, v, tc0_w, regs.texturing);
+                break;
+            }
+            case TexturingRegs::TextureConfig::Projection2D: {
+                u /= tc0_w;
+                v /= tc0_w;
+                break;
+            }
+            case TexturingRegs::TextureConfig::Shadow2D: {
+                if (!regs.texturing.shadow.orthographic) {
+                    u /= tc0_w;
+                    v /= tc0_w;
+                }
+                shadow_z = f24::FromFloat32(std::abs(tc0_w.ToFloat32()));
+                break;
+            }
+            case TexturingRegs::TextureConfig::Disabled:
+                continue; // skip this unit and continue to the next unit
+            default:
+                LOG_ERROR(HW_GPU, "Unhandled texture type {:x}", (int)texture.config.type);
+                UNIMPLEMENTED();
+                break;
+            }
+        }
+
+        const f24 width = f24::FromFloat32(static_cast<f32>(texture.config.width));
+        const f24 height = f24::FromFloat32(static_cast<f32>(texture.config.height));
+        s32 s = static_cast<s32>((u * width).ToFloat32());
+        s32 t = static_cast<s32>((v * height).ToFloat32());
+
+        bool use_border_s = false;
+        bool use_border_t = false;
+
+        if (texture.config.wrap_s == TexturingRegs::TextureConfig::ClampToBorder) {
+            use_border_s = s < 0 || s >= static_cast<s32>(texture.config.width);
+        } else if (texture.config.wrap_s == TexturingRegs::TextureConfig::ClampToBorder2) {
+            use_border_s = s >= static_cast<s32>(texture.config.width);
+        }
+
+        if (texture.config.wrap_t == TexturingRegs::TextureConfig::ClampToBorder) {
+            use_border_t = t < 0 || t >= static_cast<s32>(texture.config.height);
+        } else if (texture.config.wrap_t == TexturingRegs::TextureConfig::ClampToBorder2) {
+            use_border_t = t >= static_cast<s32>(texture.config.height);
+        }
+
+        if (use_border_s || use_border_t) {
+            const auto border_color = texture.config.border_color;
+            texture_color[i] = Common::MakeVec(border_color.r.Value(), border_color.g.Value(),
+                                               border_color.b.Value(), border_color.a.Value())
+                                   .Cast<u8>();
+        } else {
+            // Textures are laid out from bottom to top, hence we invert the t coordinate.
+            // NOTE: This may not be the right place for the inversion.
+            // TODO: Check if this applies to ETC textures, too.
+            s = GetWrappedTexCoord(texture.config.wrap_s, s, texture.config.width);
+            t = texture.config.height - 1 -
+                GetWrappedTexCoord(texture.config.wrap_t, t, texture.config.height);
+
+            const u8* texture_data = memory.GetPhysicalPointer(texture_address);
+            const auto info = TextureInfo::FromPicaRegister(texture.config, texture.format);
+
+            // TODO: Apply the min and mag filters to the texture
+            texture_color[i] = LookupTexture(texture_data, s, t, info);
+        }
+
+        if (i == 0 && (texture.config.type == TexturingRegs::TextureConfig::Shadow2D ||
+                       texture.config.type == TexturingRegs::TextureConfig::ShadowCube)) {
+
+            s32 z_int = static_cast<s32>(std::min(shadow_z.ToFloat32(), 1.0f) * 0xFFFFFF);
+            z_int -= regs.texturing.shadow.bias << 1;
+            const auto& color = texture_color[i];
+            const s32 z_ref = (color.w << 16) | (color.z << 8) | color.y;
+            u8 density;
+            if (z_ref >= z_int) {
+                density = color.x;
+            } else {
+                density = 0;
+            }
+            texture_color[i] = {density, density, density, density};
+        }
+    }
+
+    // Sample procedural texture
+    if (regs.texturing.main_config.texture3_enable) {
+        const auto& proctex_uv = uv[regs.texturing.main_config.texture3_coordinates];
+        texture_color[3] = ProcTex(proctex_uv.u().ToFloat32(), proctex_uv.v().ToFloat32(),
+                                   regs.texturing, state.proctex);
+    }
+
+    return texture_color;
+}
+
+Common::Vec4<u8> RasterizerSoftware::PixelColor(u16 x, u16 y,
+                                                Common::Vec4<u8>& combiner_output) const {
+    const auto dest = fb.GetPixel(x >> 4, y >> 4);
+    Common::Vec4<u8> blend_output = combiner_output;
+
+    const auto& output_merger = regs.framebuffer.output_merger;
+    if (output_merger.alphablend_enable) {
+        const auto params = output_merger.alpha_blending;
+        const auto lookup_factor = [&](u32 channel, FramebufferRegs::BlendFactor factor) -> u8 {
+            DEBUG_ASSERT(channel < 4);
+
+            const Common::Vec4<u8> blend_const =
+                Common::MakeVec(
+                    output_merger.blend_const.r.Value(), output_merger.blend_const.g.Value(),
+                    output_merger.blend_const.b.Value(), output_merger.blend_const.a.Value())
+                    .Cast<u8>();
+
+            switch (factor) {
+            case FramebufferRegs::BlendFactor::Zero:
+                return 0;
+            case FramebufferRegs::BlendFactor::One:
+                return 255;
+            case FramebufferRegs::BlendFactor::SourceColor:
+                return combiner_output[channel];
+            case FramebufferRegs::BlendFactor::OneMinusSourceColor:
+                return 255 - combiner_output[channel];
+            case FramebufferRegs::BlendFactor::DestColor:
+                return dest[channel];
+            case FramebufferRegs::BlendFactor::OneMinusDestColor:
+                return 255 - dest[channel];
+            case FramebufferRegs::BlendFactor::SourceAlpha:
+                return combiner_output.a();
+            case FramebufferRegs::BlendFactor::OneMinusSourceAlpha:
+                return 255 - combiner_output.a();
+            case FramebufferRegs::BlendFactor::DestAlpha:
+                return dest.a();
+            case FramebufferRegs::BlendFactor::OneMinusDestAlpha:
+                return 255 - dest.a();
+            case FramebufferRegs::BlendFactor::ConstantColor:
+                return blend_const[channel];
+            case FramebufferRegs::BlendFactor::OneMinusConstantColor:
+                return 255 - blend_const[channel];
+            case FramebufferRegs::BlendFactor::ConstantAlpha:
+                return blend_const.a();
+            case FramebufferRegs::BlendFactor::OneMinusConstantAlpha:
+                return 255 - blend_const.a();
+            case FramebufferRegs::BlendFactor::SourceAlphaSaturate:
+                // Returns 1.0 for the alpha channel
+                if (channel == 3) {
+                    return 255;
+                }
+                return std::min(combiner_output.a(), static_cast<u8>(255 - dest.a()));
+            default:
+                LOG_CRITICAL(HW_GPU, "Unknown blend factor {:x}", factor);
+                UNIMPLEMENTED();
+                break;
+            }
+            return combiner_output[channel];
+        };
+
+        const auto srcfactor = Common::MakeVec(
+            lookup_factor(0, params.factor_source_rgb), lookup_factor(1, params.factor_source_rgb),
+            lookup_factor(2, params.factor_source_rgb), lookup_factor(3, params.factor_source_a));
+
+        const auto dstfactor = Common::MakeVec(
+            lookup_factor(0, params.factor_dest_rgb), lookup_factor(1, params.factor_dest_rgb),
+            lookup_factor(2, params.factor_dest_rgb), lookup_factor(3, params.factor_dest_a));
+
+        blend_output = EvaluateBlendEquation(combiner_output, srcfactor, dest, dstfactor,
+                                             params.blend_equation_rgb);
+        blend_output.a() = EvaluateBlendEquation(combiner_output, srcfactor, dest, dstfactor,
+                                                 params.blend_equation_a)
+                               .a();
+    } else {
+        blend_output =
+            Common::MakeVec(LogicOp(combiner_output.r(), dest.r(), output_merger.logic_op),
+                            LogicOp(combiner_output.g(), dest.g(), output_merger.logic_op),
+                            LogicOp(combiner_output.b(), dest.b(), output_merger.logic_op),
+                            LogicOp(combiner_output.a(), dest.a(), output_merger.logic_op));
+    }
+
+    const Common::Vec4<u8> result = {
+        output_merger.red_enable ? blend_output.r() : dest.r(),
+        output_merger.green_enable ? blend_output.g() : dest.g(),
+        output_merger.blue_enable ? blend_output.b() : dest.b(),
+        output_merger.alpha_enable ? blend_output.a() : dest.a(),
+    };
+
+    return result;
+}
+
+Common::Vec4<u8> RasterizerSoftware::WriteTevConfig(
+    std::span<const Common::Vec4<u8>, 4> texture_color,
+    std::span<const Pica::TexturingRegs::TevStageConfig, 6> tev_stages,
+    Common::Vec4<u8> primary_color, Common::Vec4<u8> primary_fragment_color,
+    Common::Vec4<u8> secondary_fragment_color) const {
+    /**
+     * Texture environment - consists of 6 stages of color and alpha combining.
+     * Color combiners take three input color values from some source (e.g. interpolated
+     * vertex color, texture color, previous stage, etc), perform some very simple
+     * operations on each of them (e.g. inversion) and then calculate the output color
+     * with some basic arithmetic. Alpha combiners can be configured separately but work
+     * analogously.
+     **/
+    Common::Vec4<u8> combiner_output;
+    Common::Vec4<u8> combiner_buffer = {0, 0, 0, 0};
+    Common::Vec4<u8> next_combiner_buffer =
+        Common::MakeVec(regs.texturing.tev_combiner_buffer_color.r.Value(),
+                        regs.texturing.tev_combiner_buffer_color.g.Value(),
+                        regs.texturing.tev_combiner_buffer_color.b.Value(),
+                        regs.texturing.tev_combiner_buffer_color.a.Value())
+            .Cast<u8>();
+
+    for (u32 tev_stage_index = 0; tev_stage_index < tev_stages.size(); ++tev_stage_index) {
+        const auto& tev_stage = tev_stages[tev_stage_index];
+        using Source = TexturingRegs::TevStageConfig::Source;
+
+        auto get_source = [&](Source source) -> Common::Vec4<u8> {
+            switch (source) {
+            case Source::PrimaryColor:
+                return primary_color;
+            case Source::PrimaryFragmentColor:
+                return primary_fragment_color;
+            case Source::SecondaryFragmentColor:
+                return secondary_fragment_color;
+            case Source::Texture0:
+                return texture_color[0];
+            case Source::Texture1:
+                return texture_color[1];
+            case Source::Texture2:
+                return texture_color[2];
+            case Source::Texture3:
+                return texture_color[3];
+            case Source::PreviousBuffer:
+                return combiner_buffer;
+            case Source::Constant:
+                return Common::MakeVec(tev_stage.const_r.Value(), tev_stage.const_g.Value(),
+                                       tev_stage.const_b.Value(), tev_stage.const_a.Value())
+                    .Cast<u8>();
+            case Source::Previous:
+                return combiner_output;
+            default:
+                LOG_ERROR(HW_GPU, "Unknown color combiner source {}", (int)source);
+                UNIMPLEMENTED();
+                return {0, 0, 0, 0};
+            }
+        };
+
+        /**
+         * Color combiner
+         * NOTE: Not sure if the alpha combiner might use the color output of the previous
+         *       stage as input. Hence, we currently don't directly write the result to
+         *       combiner_output.rgb(), but instead store it in a temporary variable until
+         *       alpha combining has been done.
+         **/
+        const std::array<Common::Vec3<u8>, 3> color_result = {
+            GetColorModifier(tev_stage.color_modifier1, get_source(tev_stage.color_source1)),
+            GetColorModifier(tev_stage.color_modifier2, get_source(tev_stage.color_source2)),
+            GetColorModifier(tev_stage.color_modifier3, get_source(tev_stage.color_source3)),
+        };
+        const Common::Vec3<u8> color_output = ColorCombine(tev_stage.color_op, color_result);
+
+        u8 alpha_output;
+        if (tev_stage.color_op == TexturingRegs::TevStageConfig::Operation::Dot3_RGBA) {
+            // result of Dot3_RGBA operation is also placed to the alpha component
+            alpha_output = color_output.x;
+        } else {
+            // alpha combiner
+            const std::array<u8, 3> alpha_result = {{
+                GetAlphaModifier(tev_stage.alpha_modifier1, get_source(tev_stage.alpha_source1)),
+                GetAlphaModifier(tev_stage.alpha_modifier2, get_source(tev_stage.alpha_source2)),
+                GetAlphaModifier(tev_stage.alpha_modifier3, get_source(tev_stage.alpha_source3)),
+            }};
+            alpha_output = AlphaCombine(tev_stage.alpha_op, alpha_result);
+        }
+
+        combiner_output[0] = std::min(255U, color_output.r() * tev_stage.GetColorMultiplier());
+        combiner_output[1] = std::min(255U, color_output.g() * tev_stage.GetColorMultiplier());
+        combiner_output[2] = std::min(255U, color_output.b() * tev_stage.GetColorMultiplier());
+        combiner_output[3] = std::min(255U, alpha_output * tev_stage.GetAlphaMultiplier());
+
+        combiner_buffer = next_combiner_buffer;
+
+        if (regs.texturing.tev_combiner_buffer_input.TevStageUpdatesCombinerBufferColor(
+                tev_stage_index)) {
+            next_combiner_buffer.r() = combiner_output.r();
+            next_combiner_buffer.g() = combiner_output.g();
+            next_combiner_buffer.b() = combiner_output.b();
+        }
+
+        if (regs.texturing.tev_combiner_buffer_input.TevStageUpdatesCombinerBufferAlpha(
+                tev_stage_index)) {
+            next_combiner_buffer.a() = combiner_output.a();
+        }
+    }
+    return combiner_output;
+}
+
+void RasterizerSoftware::WriteFog(Common::Vec4<u8>& combiner_output, float depth) const {
+    /**
+     * Apply fog combiner. Not fully accurate. We'd have to know what data type is used to
+     * store the depth etc. Using float for now until we know more about Pica datatypes.
+     **/
+    if (regs.texturing.fog_mode == TexturingRegs::FogMode::Fog) {
+        const Common::Vec3<u8> fog_color =
+            Common::MakeVec(regs.texturing.fog_color.r.Value(), regs.texturing.fog_color.g.Value(),
+                            regs.texturing.fog_color.b.Value())
+                .Cast<u8>();
+
+        float fog_index;
+        if (regs.texturing.fog_flip) {
+            fog_index = (1.0f - depth) * 128.0f;
+        } else {
+            fog_index = depth * 128.0f;
+        }
+
+        // Generate clamped fog factor from LUT for given fog index
+        const f32 fog_i = std::clamp(floorf(fog_index), 0.0f, 127.0f);
+        const f32 fog_f = fog_index - fog_i;
+        const auto& fog_lut_entry = state.fog.lut[static_cast<u32>(fog_i)];
+        f32 fog_factor = fog_lut_entry.ToFloat() + fog_lut_entry.DiffToFloat() * fog_f;
+        fog_factor = std::clamp(fog_factor, 0.0f, 1.0f);
+        for (u32 i = 0; i < 3; i++) {
+            combiner_output[i] = static_cast<u8>(fog_factor * combiner_output[i] +
+                                                 (1.0f - fog_factor) * fog_color[i]);
+        }
+    }
+}
+
+bool RasterizerSoftware::DoAlphaTest(u8 alpha) const {
+    const auto& output_merger = regs.framebuffer.output_merger;
+    if (!output_merger.alpha_test.enable) {
+        return true;
+    }
+    switch (output_merger.alpha_test.func) {
+    case FramebufferRegs::CompareFunc::Never:
+        return false;
+    case FramebufferRegs::CompareFunc::Always:
+        return true;
+    case FramebufferRegs::CompareFunc::Equal:
+        return alpha == output_merger.alpha_test.ref;
+    case FramebufferRegs::CompareFunc::NotEqual:
+        return alpha != output_merger.alpha_test.ref;
+    case FramebufferRegs::CompareFunc::LessThan:
+        return alpha < output_merger.alpha_test.ref;
+    case FramebufferRegs::CompareFunc::LessThanOrEqual:
+        return alpha <= output_merger.alpha_test.ref;
+    case FramebufferRegs::CompareFunc::GreaterThan:
+        return alpha > output_merger.alpha_test.ref;
+    case FramebufferRegs::CompareFunc::GreaterThanOrEqual:
+        return alpha >= output_merger.alpha_test.ref;
+    }
+}
+
+bool RasterizerSoftware::DoDepthStencilTest(u16 x, u16 y, float depth,
+                                            bool stencil_action_enable) const {
+    const auto& framebuffer = regs.framebuffer.framebuffer;
+    const auto stencil_test = regs.framebuffer.output_merger.stencil_test;
+    u8 old_stencil = 0;
+
+    const auto update_stencil = [&](Pica::FramebufferRegs::StencilAction action) {
+        const u8 new_stencil =
+            PerformStencilAction(action, old_stencil, stencil_test.reference_value);
+        if (framebuffer.allow_depth_stencil_write != 0) {
+            const u8 stencil =
+                (new_stencil & stencil_test.write_mask) | (old_stencil & ~stencil_test.write_mask);
+            fb.SetStencil(x >> 4, y >> 4, stencil);
+        }
+    };
+
+    if (stencil_action_enable) {
+        old_stencil = fb.GetStencil(x >> 4, y >> 4);
+        const u8 dest = old_stencil & stencil_test.input_mask;
+        const u8 ref = stencil_test.reference_value & stencil_test.input_mask;
+        bool pass = false;
+        switch (stencil_test.func) {
+        case FramebufferRegs::CompareFunc::Never:
+            pass = false;
+            break;
+        case FramebufferRegs::CompareFunc::Always:
+            pass = true;
+            break;
+        case FramebufferRegs::CompareFunc::Equal:
+            pass = (ref == dest);
+            break;
+        case FramebufferRegs::CompareFunc::NotEqual:
+            pass = (ref != dest);
+            break;
+        case FramebufferRegs::CompareFunc::LessThan:
+            pass = (ref < dest);
+            break;
+        case FramebufferRegs::CompareFunc::LessThanOrEqual:
+            pass = (ref <= dest);
+            break;
+        case FramebufferRegs::CompareFunc::GreaterThan:
+            pass = (ref > dest);
+            break;
+        case FramebufferRegs::CompareFunc::GreaterThanOrEqual:
+            pass = (ref >= dest);
+            break;
+        }
+        if (!pass) {
+            update_stencil(stencil_test.action_stencil_fail);
+            return false;
+        }
+    }
+
+    const u32 num_bits = FramebufferRegs::DepthBitsPerPixel(framebuffer.depth_format);
+    const u32 z = static_cast<u32>(depth * ((1 << num_bits) - 1));
+
+    const auto& output_merger = regs.framebuffer.output_merger;
+    if (output_merger.depth_test_enable) {
+        const u32 ref_z = fb.GetDepth(x >> 4, y >> 4);
+        bool pass = false;
+        switch (output_merger.depth_test_func) {
+        case FramebufferRegs::CompareFunc::Never:
+            pass = false;
+            break;
+        case FramebufferRegs::CompareFunc::Always:
+            pass = true;
+            break;
+        case FramebufferRegs::CompareFunc::Equal:
+            pass = z == ref_z;
+            break;
+        case FramebufferRegs::CompareFunc::NotEqual:
+            pass = z != ref_z;
+            break;
+        case FramebufferRegs::CompareFunc::LessThan:
+            pass = z < ref_z;
+            break;
+        case FramebufferRegs::CompareFunc::LessThanOrEqual:
+            pass = z <= ref_z;
+            break;
+        case FramebufferRegs::CompareFunc::GreaterThan:
+            pass = z > ref_z;
+            break;
+        case FramebufferRegs::CompareFunc::GreaterThanOrEqual:
+            pass = z >= ref_z;
+            break;
+        }
+        if (!pass) {
+            if (stencil_action_enable) {
+                update_stencil(stencil_test.action_depth_fail);
+            }
+            return false;
+        }
+    }
+    if (framebuffer.allow_depth_stencil_write != 0 && output_merger.depth_write_enable) {
+        fb.SetDepth(x >> 4, y >> 4, z);
+    }
+    // The stencil depth_pass action is executed even if depth testing is disabled
+    if (stencil_action_enable) {
+        update_stencil(stencil_test.action_depth_pass);
+    }
+
+    return true;
+}
+
+} // namespace SwRenderer

src/video_core/renderer_software/sw_rasterizer.h

@@ -4,16 +4,30 @@
 
 #pragma once
 
-#include "common/common_types.h"
+#include <span>
+
 #include "video_core/rasterizer_interface.h"
+#include "video_core/regs_texturing.h"
+#include "video_core/renderer_software/sw_clipper.h"
+#include "video_core/renderer_software/sw_framebuffer.h"
 
 namespace Pica::Shader {
 struct OutputVertex;
-} // namespace Pica::Shader
+}
 
-namespace VideoCore {
+namespace Pica {
+struct State;
+struct Regs;
+} // namespace Pica
+
+namespace SwRenderer {
+
+struct Vertex;
+
+class RasterizerSoftware : public VideoCore::RasterizerInterface {
+public:
+    explicit RasterizerSoftware(Memory::MemorySystem& memory);
 
-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 {}
@@ -23,6 +37,44 @@ class RasterizerSoftware : public RasterizerInterface {
     void InvalidateRegion(PAddr addr, u32 size) override {}
     void FlushAndInvalidateRegion(PAddr addr, u32 size) override {}
     void ClearAll(bool flush) override {}
+
+private:
+    /// Computes the screen coordinates of the provided vertex.
+    void MakeScreenCoords(Vertex& vtx);
+
+    /// Processes the triangle defined by the provided vertices.
+    void ProcessTriangle(const Vertex& v0, const Vertex& v1, const Vertex& v2,
+                         bool reversed = false);
+
+    /// Returns the texture color of the currently processed pixel.
+    std::array<Common::Vec4<u8>, 4> TextureColor(
+        std::span<const Common::Vec2<f24>, 3> uv,
+        std::span<const Pica::TexturingRegs::FullTextureConfig, 3> textures, f24 tc0_w) const;
+
+    /// Returns the final pixel color with blending or logic ops applied.
+    Common::Vec4<u8> PixelColor(u16 x, u16 y, Common::Vec4<u8>& combiner_output) const;
+
+    /// Emulates the TEV configuration and returns the combiner output.
+    Common::Vec4<u8> WriteTevConfig(
+        std::span<const Common::Vec4<u8>, 4> texture_color,
+        std::span<const Pica::TexturingRegs::TevStageConfig, 6> tev_stages,
+        Common::Vec4<u8> primary_color, Common::Vec4<u8> primary_fragment_color,
+        Common::Vec4<u8> secondary_fragment_color) const;
+
+    /// Blends fog to the combiner output if enabled.
+    void WriteFog(Common::Vec4<u8>& combiner_output, float depth) const;
+
+    /// Performs the alpha test. Returns false if the test failed.
+    bool DoAlphaTest(u8 alpha) const;
+
+    /// Performs the depth stencil test. Returns false if the test failed.
+    bool DoDepthStencilTest(u16 x, u16 y, float depth, bool stencil_action_enable) const;
+
+private:
+    Memory::MemorySystem& memory;
+    Pica::State& state;
+    const Pica::Regs& regs;
+    Framebuffer fb;
 };
 
-} // namespace VideoCore
+} // namespace SwRenderer

src/video_core/renderer_software/sw_texturing.cpp

@@ -9,41 +9,40 @@
 #include "video_core/regs_texturing.h"
 #include "video_core/renderer_software/sw_texturing.h"
 
-namespace Pica::Rasterizer {
+namespace SwRenderer {
 
-using TevStageConfig = TexturingRegs::TevStageConfig;
+using TevStageConfig = Pica::TexturingRegs::TevStageConfig;
+
+int GetWrappedTexCoord(Pica::TexturingRegs::TextureConfig::WrapMode mode, s32 val, u32 size) {
+    using TextureConfig = Pica::TexturingRegs::TextureConfig;
 
-int GetWrappedTexCoord(TexturingRegs::TextureConfig::WrapMode mode, int val, unsigned size) {
     switch (mode) {
-    case TexturingRegs::TextureConfig::ClampToEdge2:
+    case TextureConfig::ClampToEdge2:
         // For negative coordinate, ClampToEdge2 behaves the same as Repeat
         if (val < 0) {
-            return static_cast<int>(static_cast<unsigned>(val) % size);
+            return static_cast<s32>(static_cast<u32>(val) % size);
         }
-    // [[fallthrough]]
+        [[fallthrough]];
-    case TexturingRegs::TextureConfig::ClampToEdge:
+    case TextureConfig::ClampToEdge:
         val = std::max(val, 0);
-        val = std::min(val, static_cast<int>(size) - 1);
+        val = std::min(val, static_cast<s32>(size) - 1);
         return val;
-
+    case TextureConfig::ClampToBorder:
-    case TexturingRegs::TextureConfig::ClampToBorder:
         return val;
-
+    case TextureConfig::ClampToBorder2:
-    case TexturingRegs::TextureConfig::ClampToBorder2:
     // For ClampToBorder2, the case of positive coordinate beyond the texture size is already
     // handled outside. Here we only handle the negative coordinate in the same way as Repeat.
-    case TexturingRegs::TextureConfig::Repeat2:
+    case TextureConfig::Repeat2:
-    case TexturingRegs::TextureConfig::Repeat3:
+    case TextureConfig::Repeat3:
-    case TexturingRegs::TextureConfig::Repeat:
+    case TextureConfig::Repeat:
-        return static_cast<int>(static_cast<unsigned>(val) % size);
+        return static_cast<s32>(static_cast<u32>(val) % size);
-
+    case TextureConfig::MirroredRepeat: {
-    case TexturingRegs::TextureConfig::MirroredRepeat: {
+        u32 coord = (static_cast<u32>(val) % (2 * size));
-        unsigned int coord = (static_cast<unsigned>(val) % (2 * size));
+        if (coord >= size) {
-        if (coord >= size)
             coord = 2 * size - 1 - coord;
-        return static_cast<int>(coord);
+        }
+        return static_cast<s32>(coord);
     }
-
     default:
         LOG_ERROR(HW_GPU, "Unknown texture coordinate wrapping mode {:x}", (int)mode);
         UNIMPLEMENTED();
@@ -58,35 +57,25 @@ Common::Vec3<u8> GetColorModifier(TevStageConfig::ColorModifier factor,
     switch (factor) {
     case ColorModifier::SourceColor:
         return values.rgb();
-
     case ColorModifier::OneMinusSourceColor:
         return (Common::Vec3<u8>(255, 255, 255) - values.rgb()).Cast<u8>();
-
     case ColorModifier::SourceAlpha:
         return values.aaa();
-
     case ColorModifier::OneMinusSourceAlpha:
         return (Common::Vec3<u8>(255, 255, 255) - values.aaa()).Cast<u8>();
-
     case ColorModifier::SourceRed:
         return values.rrr();
-
     case ColorModifier::OneMinusSourceRed:
         return (Common::Vec3<u8>(255, 255, 255) - values.rrr()).Cast<u8>();
-
     case ColorModifier::SourceGreen:
         return values.ggg();
-
     case ColorModifier::OneMinusSourceGreen:
         return (Common::Vec3<u8>(255, 255, 255) - values.ggg()).Cast<u8>();
-
     case ColorModifier::SourceBlue:
         return values.bbb();
-
     case ColorModifier::OneMinusSourceBlue:
         return (Common::Vec3<u8>(255, 255, 255) - values.bbb()).Cast<u8>();
     }
-
     UNREACHABLE();
 };
 
@@ -96,42 +85,33 @@ u8 GetAlphaModifier(TevStageConfig::AlphaModifier factor, const Common::Vec4<u8>
     switch (factor) {
     case AlphaModifier::SourceAlpha:
         return values.a();
-
     case AlphaModifier::OneMinusSourceAlpha:
         return 255 - values.a();
-
     case AlphaModifier::SourceRed:
         return values.r();
-
     case AlphaModifier::OneMinusSourceRed:
         return 255 - values.r();
-
     case AlphaModifier::SourceGreen:
         return values.g();
-
     case AlphaModifier::OneMinusSourceGreen:
         return 255 - values.g();
-
     case AlphaModifier::SourceBlue:
         return values.b();
-
     case AlphaModifier::OneMinusSourceBlue:
         return 255 - values.b();
     }
-
     UNREACHABLE();
 };
 
-Common::Vec3<u8> ColorCombine(TevStageConfig::Operation op, const Common::Vec3<u8> input[3]) {
+Common::Vec3<u8> ColorCombine(TevStageConfig::Operation op,
+                              std::span<const Common::Vec3<u8>, 3> input) {
     using Operation = TevStageConfig::Operation;
 
     switch (op) {
     case Operation::Replace:
         return input[0];
-
     case Operation::Modulate:
         return ((input[0] * input[1]) / 255).Cast<u8>();
-
     case Operation::Add: {
         auto result = input[0] + input[1];
         result.r() = std::min(255, result.r());
@@ -139,46 +119,41 @@ Common::Vec3<u8> ColorCombine(TevStageConfig::Operation op, const Common::Vec3<u
         result.b() = std::min(255, result.b());
         return result.Cast<u8>();
     }
-
     case Operation::AddSigned: {
         // TODO(bunnei): Verify that the color conversion from (float) 0.5f to
         // (byte) 128 is correct
-        auto result =
+        Common::Vec3i result =
-            input[0].Cast<int>() + input[1].Cast<int>() - Common::MakeVec<int>(128, 128, 128);
+            input[0].Cast<s32>() + input[1].Cast<s32>() - Common::MakeVec<s32>(128, 128, 128);
-        result.r() = std::clamp<int>(result.r(), 0, 255);
+        result.r() = std::clamp<s32>(result.r(), 0, 255);
-        result.g() = std::clamp<int>(result.g(), 0, 255);
+        result.g() = std::clamp<s32>(result.g(), 0, 255);
-        result.b() = std::clamp<int>(result.b(), 0, 255);
+        result.b() = std::clamp<s32>(result.b(), 0, 255);
         return result.Cast<u8>();
     }
-
     case Operation::Lerp:
         return ((input[0] * input[2] +
                  input[1] * (Common::MakeVec<u8>(255, 255, 255) - input[2]).Cast<u8>()) /
                 255)
             .Cast<u8>();
-
     case Operation::Subtract: {
-        auto result = input[0].Cast<int>() - input[1].Cast<int>();
+        auto result = input[0].Cast<s32>() - input[1].Cast<s32>();
         result.r() = std::max(0, result.r());
         result.g() = std::max(0, result.g());
         result.b() = std::max(0, result.b());
         return result.Cast<u8>();
     }
-
     case Operation::MultiplyThenAdd: {
-        auto result = (input[0] * input[1] + 255 * input[2].Cast<int>()) / 255;
+        auto result = (input[0] * input[1] + 255 * input[2].Cast<s32>()) / 255;
         result.r() = std::min(255, result.r());
         result.g() = std::min(255, result.g());
         result.b() = std::min(255, result.b());
         return result.Cast<u8>();
     }
-
     case Operation::AddThenMultiply: {
         auto result = input[0] + input[1];
         result.r() = std::min(255, result.r());
         result.g() = std::min(255, result.g());
         result.b() = std::min(255, result.b());
-        result = (result * input[2].Cast<int>()) / 255;
+        result = (result * input[2].Cast<s32>()) / 255;
         return result.Cast<u8>();
     }
     case Operation::Dot3_RGB:
@@ -187,11 +162,11 @@ Common::Vec3<u8> ColorCombine(TevStageConfig::Operation op, const Common::Vec3<u
         // indicate that the per-component computation can't have a higher precision than 1/256,
         // while dot3_rgb((0x80,g0,b0), (0x7F,g1,b1)) and dot3_rgb((0x80,g0,b0), (0x80,g1,b1)) give
         // different results.
-        int result = ((input[0].r() * 2 - 255) * (input[1].r() * 2 - 255) + 128) / 256 +
+        s32 result = ((input[0].r() * 2 - 255) * (input[1].r() * 2 - 255) + 128) / 256 +
                      ((input[0].g() * 2 - 255) * (input[1].g() * 2 - 255) + 128) / 256 +
                      ((input[0].b() * 2 - 255) * (input[1].b() * 2 - 255) + 128) / 256;
-        result = std::max(0, std::min(255, result));
+        result = std::clamp(result, 0, 255);
-        return {(u8)result, (u8)result, (u8)result};
+        return Common::Vec3{result, result, result}.Cast<u8>();
     }
     default:
         LOG_ERROR(HW_GPU, "Unknown color combiner operation {}", (int)op);
@@ -205,31 +180,23 @@ u8 AlphaCombine(TevStageConfig::Operation op, const std::array<u8, 3>& input) {
         using Operation = TevStageConfig::Operation;
     case Operation::Replace:
         return input[0];
-
     case Operation::Modulate:
         return input[0] * input[1] / 255;
-
     case Operation::Add:
         return std::min(255, input[0] + input[1]);
-
     case Operation::AddSigned: {
         // TODO(bunnei): Verify that the color conversion from (float) 0.5f to (byte) 128 is correct
-        auto result = static_cast<int>(input[0]) + static_cast<int>(input[1]) - 128;
+        auto result = static_cast<s32>(input[0]) + static_cast<s32>(input[1]) - 128;
-        return static_cast<u8>(std::clamp<int>(result, 0, 255));
+        return static_cast<u8>(std::clamp<s32>(result, 0, 255));
     }
-
     case Operation::Lerp:
         return (input[0] * input[2] + input[1] * (255 - input[2])) / 255;
-
     case Operation::Subtract:
-        return std::max(0, (int)input[0] - (int)input[1]);
+        return std::max(0, static_cast<s32>(input[0]) - static_cast<s32>(input[1]));
-
     case Operation::MultiplyThenAdd:
         return std::min(255, (input[0] * input[1] + 255 * input[2]) / 255);
-
     case Operation::AddThenMultiply:
         return (std::min(255, (input[0] + input[1])) * input[2]) / 255;
-
     default:
         LOG_ERROR(HW_GPU, "Unknown alpha combiner operation {}", (int)op);
         UNIMPLEMENTED();
@@ -237,4 +204,4 @@ u8 AlphaCombine(TevStageConfig::Operation op, const std::array<u8, 3>& input) {
     }
 };
 
-} // namespace Pica::Rasterizer
+} // namespace SwRenderer

src/video_core/renderer_software/sw_texturing.h

@@ -4,23 +4,25 @@
 
 #pragma once
 
+#include <span>
+
 #include "common/common_types.h"
 #include "common/vector_math.h"
 #include "video_core/regs_texturing.h"
 
-namespace Pica::Rasterizer {
+namespace SwRenderer {
 
-int GetWrappedTexCoord(TexturingRegs::TextureConfig::WrapMode mode, int val, unsigned size);
+int GetWrappedTexCoord(Pica::TexturingRegs::TextureConfig::WrapMode mode, s32 val, u32 size);
 
-Common::Vec3<u8> GetColorModifier(TexturingRegs::TevStageConfig::ColorModifier factor,
+Common::Vec3<u8> GetColorModifier(Pica::TexturingRegs::TevStageConfig::ColorModifier factor,
                                   const Common::Vec4<u8>& values);
 
-u8 GetAlphaModifier(TexturingRegs::TevStageConfig::AlphaModifier factor,
+u8 GetAlphaModifier(Pica::TexturingRegs::TevStageConfig::AlphaModifier factor,
                     const Common::Vec4<u8>& values);
 
-Common::Vec3<u8> ColorCombine(TexturingRegs::TevStageConfig::Operation op,
+Common::Vec3<u8> ColorCombine(Pica::TexturingRegs::TevStageConfig::Operation op,
-                              const Common::Vec3<u8> input[3]);
+                              std::span<const Common::Vec3<u8>, 3> input);
 
-u8 AlphaCombine(TexturingRegs::TevStageConfig::Operation op, const std::array<u8, 3>& input);
+u8 AlphaCombine(Pica::TexturingRegs::TevStageConfig::Operation op, const std::array<u8, 3>& input);
 
-} // namespace Pica::Rasterizer
+} // namespace SwRenderer

src/video_core/shader/debug_data.h

@@ -54,12 +54,12 @@ struct DebugData<true> {
             LOOP_INT_IN = 0x800,
         };
 
-        Common::Vec4<float24> src1;
+        Common::Vec4<f24> src1;
-        Common::Vec4<float24> src2;
+        Common::Vec4<f24> src2;
-        Common::Vec4<float24> src3;
+        Common::Vec4<f24> src3;
 
-        Common::Vec4<float24> dest_in;
+        Common::Vec4<f24> dest_in;
-        Common::Vec4<float24> dest_out;
+        Common::Vec4<f24> dest_out;
 
         s32 address_registers[2];
         bool conditional_code[2];
@@ -89,7 +89,7 @@ template <DebugDataRecord::Type type, typename ValueType>
 inline void SetField(DebugDataRecord& record, ValueType value);
 
 template <>
-inline void SetField<DebugDataRecord::SRC1>(DebugDataRecord& record, float24* value) {
+inline void SetField<DebugDataRecord::SRC1>(DebugDataRecord& record, f24* value) {
     record.src1.x = value[0];
     record.src1.y = value[1];
     record.src1.z = value[2];
@@ -97,7 +97,7 @@ inline void SetField<DebugDataRecord::SRC1>(DebugDataRecord& record, float24* va
 }
 
 template <>
-inline void SetField<DebugDataRecord::SRC2>(DebugDataRecord& record, float24* value) {
+inline void SetField<DebugDataRecord::SRC2>(DebugDataRecord& record, f24* value) {
     record.src2.x = value[0];
     record.src2.y = value[1];
     record.src2.z = value[2];
@@ -105,7 +105,7 @@ inline void SetField<DebugDataRecord::SRC2>(DebugDataRecord& record, float24* va
 }
 
 template <>
-inline void SetField<DebugDataRecord::SRC3>(DebugDataRecord& record, float24* value) {
+inline void SetField<DebugDataRecord::SRC3>(DebugDataRecord& record, f24* value) {
     record.src3.x = value[0];
     record.src3.y = value[1];
     record.src3.z = value[2];
@@ -113,7 +113,7 @@ inline void SetField<DebugDataRecord::SRC3>(DebugDataRecord& record, float24* va
 }
 
 template <>
-inline void SetField<DebugDataRecord::DEST_IN>(DebugDataRecord& record, float24* value) {
+inline void SetField<DebugDataRecord::DEST_IN>(DebugDataRecord& record, f24* value) {
     record.dest_in.x = value[0];
     record.dest_in.y = value[1];
     record.dest_in.z = value[2];
@@ -121,7 +121,7 @@ inline void SetField<DebugDataRecord::DEST_IN>(DebugDataRecord& record, float24*
 }
 
 template <>
-inline void SetField<DebugDataRecord::DEST_OUT>(DebugDataRecord& record, float24* value) {
+inline void SetField<DebugDataRecord::DEST_OUT>(DebugDataRecord& record, f24* value) {
     record.dest_out.x = value[0];
     record.dest_out.y = value[1];
     record.dest_out.z = value[2];

src/video_core/shader/shader.cpp

@@ -5,10 +5,10 @@
 #include <cmath>
 #include <cstring>
 #include "common/arch.h"
+#include "common/assert.h"
 #include "common/bit_set.h"
 #include "common/logging/log.h"
 #include "common/microprofile.h"
-#include "video_core/pica_state.h"
 #include "video_core/regs_rasterizer.h"
 #include "video_core/regs_shader.h"
 #include "video_core/shader/shader.h"
@@ -41,11 +41,11 @@ OutputVertex OutputVertex::FromAttributeBuffer(const RasterizerRegs& regs,
         // Allow us to overflow OutputVertex to avoid branches, since
         // RasterizerRegs::VSOutputAttributes::INVALID would write to slot 31, which
         // would be out of bounds otherwise.
-        std::array<float24, 32> vertex_slots_overflow;
+        std::array<f24, 32> vertex_slots_overflow;
     };
 
     // Assert that OutputVertex has enough space for 24 semantic registers
-    static_assert(sizeof(std::array<float24, 24>) == sizeof(ret),
+    static_assert(sizeof(std::array<f24, 24>) == sizeof(ret),
                   "Struct and array have different sizes.");
 
     unsigned int num_attributes = regs.vs_output_total & 7;
@@ -61,7 +61,7 @@ OutputVertex OutputVertex::FromAttributeBuffer(const RasterizerRegs& regs,
     // interpolation
     for (unsigned i = 0; i < 4; ++i) {
         float c = std::fabs(ret.color[i].ToFloat32());
-        ret.color[i] = float24::FromFloat32(c < 1.0f ? c : 1.0f);
+        ret.color[i] = f24::FromFloat32(c < 1.0f ? c : 1.0f);
     }
 
     LOG_TRACE(HW_GPU,
@@ -86,7 +86,7 @@ void UnitState::LoadInput(const ShaderRegs& config, const AttributeBuffer& input
     }
 }
 
-static void CopyRegistersToOutput(std::span<Common::Vec4<float24>, 16> regs, u32 mask,
+static void CopyRegistersToOutput(std::span<Common::Vec4<f24>, 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(std::span<Common::Vec4<float24>, 16> output_regs) {
+void GSEmitter::Emit(std::span<Common::Vec4<f24>, 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

@@ -12,7 +12,6 @@
 #include <boost/serialization/access.hpp>
 #include <boost/serialization/array.hpp>
 #include <boost/serialization/base_object.hpp>
-#include "common/assert.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/hash.h"
@@ -29,7 +28,7 @@ using ProgramCode = std::array<u32, MAX_PROGRAM_CODE_LENGTH>;
 using SwizzleData = std::array<u32, MAX_SWIZZLE_DATA_LENGTH>;
 
 struct AttributeBuffer {
-    alignas(16) Common::Vec4<float24> attr[16];
+    alignas(16) Common::Vec4<f24> attr[16];
 
 private:
     friend class boost::serialization::access;
@@ -46,16 +45,16 @@ using VertexHandler = std::function<void(const AttributeBuffer&)>;
 using WindingSetter = std::function<void()>;
 
 struct OutputVertex {
-    Common::Vec4<float24> pos;
+    Common::Vec4<f24> pos;
-    Common::Vec4<float24> quat;
+    Common::Vec4<f24> quat;
-    Common::Vec4<float24> color;
+    Common::Vec4<f24> color;
-    Common::Vec2<float24> tc0;
+    Common::Vec2<f24> tc0;
-    Common::Vec2<float24> tc1;
+    Common::Vec2<f24> tc1;
-    float24 tc0_w;
+    f24 tc0_w;
     INSERT_PADDING_WORDS(1);
-    Common::Vec3<float24> view;
+    Common::Vec3<f24> view;
     INSERT_PADDING_WORDS(1);
-    Common::Vec2<float24> tc2;
+    Common::Vec2<f24> tc2;
 
     static void ValidateSemantics(const RasterizerRegs& regs);
     static OutputVertex FromAttributeBuffer(const RasterizerRegs& regs,
@@ -76,8 +75,8 @@ private:
     friend class boost::serialization::access;
 };
 #define ASSERT_POS(var, pos)                                                                       \
-    static_assert(offsetof(OutputVertex, var) == pos * sizeof(float24), "Semantic at wrong "       \
+    static_assert(offsetof(OutputVertex, var) == pos * sizeof(f24), "Semantic at wrong "           \
-                                                                        "offset.")
+                                                                    "offset.")
 ASSERT_POS(pos, RasterizerRegs::VSOutputAttributes::POSITION_X);
 ASSERT_POS(quat, RasterizerRegs::VSOutputAttributes::QUATERNION_X);
 ASSERT_POS(color, RasterizerRegs::VSOutputAttributes::COLOR_R);
@@ -109,7 +108,7 @@ struct GSEmitter {
 
     GSEmitter();
     ~GSEmitter();
-    void Emit(std::span<Common::Vec4<float24>, 16> output_regs);
+    void Emit(std::span<Common::Vec4<f24>, 16> output_regs);
 
 private:
     friend class boost::serialization::access;
@@ -136,9 +135,9 @@ 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) std::array<Common::Vec4<float24>, 16> input;
+        alignas(16) std::array<Common::Vec4<f24>, 16> input;
-        alignas(16) std::array<Common::Vec4<float24>, 16> temporary;
+        alignas(16) std::array<Common::Vec4<f24>, 16> temporary;
-        alignas(16) std::array<Common::Vec4<float24>, 16> output;
+        alignas(16) std::array<Common::Vec4<f24>, 16> output;
 
     private:
         friend class boost::serialization::access;
@@ -160,18 +159,16 @@ struct UnitState {
     GSEmitter* emitter_ptr;
 
     static std::size_t InputOffset(int register_index) {
-        return offsetof(UnitState, registers.input) +
+        return offsetof(UnitState, registers.input) + register_index * sizeof(Common::Vec4<f24>);
-               register_index * sizeof(Common::Vec4<float24>);
     }
 
     static std::size_t OutputOffset(int register_index) {
-        return offsetof(UnitState, registers.output) +
+        return offsetof(UnitState, registers.output) + register_index * sizeof(Common::Vec4<f24>);
-               register_index * sizeof(Common::Vec4<float24>);
     }
 
     static std::size_t TemporaryOffset(int register_index) {
         return offsetof(UnitState, registers.temporary) +
-               register_index * sizeof(Common::Vec4<float24>);
+               register_index * sizeof(Common::Vec4<f24>);
     }
 
     /**
@@ -219,13 +216,13 @@ private:
 struct Uniforms {
     // The float uniforms are accessed by the shader JIT using SSE instructions, and are
     // therefore required to be 16-byte aligned.
-    alignas(16) std::array<Common::Vec4<float24>, 96> f;
+    alignas(16) std::array<Common::Vec4<f24>, 96> f;
 
     std::array<bool, 16> b;
     std::array<Common::Vec4<u8>, 4> i;
 
     static std::size_t GetFloatUniformOffset(unsigned index) {
-        return offsetof(Uniforms, f) + index * sizeof(Common::Vec4<float24>);
+        return offsetof(Uniforms, f) + index * sizeof(Common::Vec4<f24>);
     }
 
     static std::size_t GetBoolUniformOffset(unsigned index) {

src/video_core/shader/shader_interpreter.cpp

@@ -80,7 +80,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
     const auto& program_code = setup.program_code;
 
     // Placeholder for invalid inputs
-    static float24 dummy_vec4_float24[4];
+    static f24 dummy_vec4_float24[4];
 
     unsigned iteration = 0;
     bool exit_loop = false;
@@ -111,7 +111,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
 
         debug_data.max_offset = std::max<u32>(debug_data.max_offset, 1 + program_counter);
 
-        auto LookupSourceRegister = [&](const SourceRegister& source_reg) -> const float24* {
+        auto LookupSourceRegister = [&](const SourceRegister& source_reg) -> const f24* {
             switch (source_reg.GetRegisterType()) {
             case RegisterType::Input:
                 return &state.registers.input[source_reg.GetIndex()].x;
@@ -137,15 +137,15 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                     ? 0
                     : state.address_registers[instr.common.address_register_index - 1];
 
-            const float24* src1_ = LookupSourceRegister(instr.common.GetSrc1(is_inverted) +
+            const f24* src1_ = LookupSourceRegister(instr.common.GetSrc1(is_inverted) +
-                                                        (is_inverted ? 0 : address_offset));
+                                                    (is_inverted ? 0 : address_offset));
-            const float24* src2_ = LookupSourceRegister(instr.common.GetSrc2(is_inverted) +
+            const f24* src2_ = LookupSourceRegister(instr.common.GetSrc2(is_inverted) +
-                                                        (is_inverted ? address_offset : 0));
+                                                    (is_inverted ? address_offset : 0));
 
             const bool negate_src1 = ((bool)swizzle.negate_src1 != false);
             const bool negate_src2 = ((bool)swizzle.negate_src2 != false);
 
-            float24 src1[4] = {
+            f24 src1[4] = {
                 src1_[(int)swizzle.src1_selector_0.Value()],
                 src1_[(int)swizzle.src1_selector_1.Value()],
                 src1_[(int)swizzle.src1_selector_2.Value()],
@@ -157,7 +157,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                 src1[2] = -src1[2];
                 src1[3] = -src1[3];
             }
-            float24 src2[4] = {
+            f24 src2[4] = {
                 src2_[(int)swizzle.src2_selector_0.Value()],
                 src2_[(int)swizzle.src2_selector_1.Value()],
                 src2_[(int)swizzle.src2_selector_2.Value()],
@@ -170,12 +170,11 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                 src2[3] = -src2[3];
             }
 
-            float24* dest =
+            f24* dest = (instr.common.dest.Value() < 0x10)
-                (instr.common.dest.Value() < 0x10)
+                            ? &state.registers.output[instr.common.dest.Value().GetIndex()][0]
-                    ? &state.registers.output[instr.common.dest.Value().GetIndex()][0]
+                        : (instr.common.dest.Value() < 0x20)
-                : (instr.common.dest.Value() < 0x20)
+                            ? &state.registers.temporary[instr.common.dest.Value().GetIndex()][0]
-                    ? &state.registers.temporary[instr.common.dest.Value().GetIndex()][0]
+                            : dummy_vec4_float24;
-                    : dummy_vec4_float24;
 
             debug_data.max_opdesc_id =
                 std::max<u32>(debug_data.max_opdesc_id, 1 + instr.common.operand_desc_id);
@@ -216,7 +215,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
-                    dest[i] = float24::FromFloat32(std::floor(src1[i].ToFloat32()));
+                    dest[i] = f24::FromFloat32(std::floor(src1[i].ToFloat32()));
                 }
                 Record<DebugDataRecord::DEST_OUT>(debug_data, iteration, dest);
                 break;
@@ -263,11 +262,10 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
 
                 OpCode::Id opcode = instr.opcode.Value().EffectiveOpCode();
                 if (opcode == OpCode::Id::DPH || opcode == OpCode::Id::DPHI)
-                    src1[3] = float24::FromFloat32(1.0f);
+                    src1[3] = f24::One();
 
                 int num_components = (opcode == OpCode::Id::DP3) ? 3 : 4;
-                float24 dot = std::inner_product(src1, src1 + num_components, src2,
+                f24 dot = std::inner_product(src1, src1 + num_components, src2, f24::Zero());
-                                                 float24::FromFloat32(0.f));
 
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
@@ -283,7 +281,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
             case OpCode::Id::RCP: {
                 Record<DebugDataRecord::SRC1>(debug_data, iteration, src1);
                 Record<DebugDataRecord::DEST_IN>(debug_data, iteration, dest);
-                float24 rcp_res = float24::FromFloat32(1.0f / src1[0].ToFloat32());
+                f24 rcp_res = f24::FromFloat32(1.0f / src1[0].ToFloat32());
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
@@ -298,7 +296,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
             case OpCode::Id::RSQ: {
                 Record<DebugDataRecord::SRC1>(debug_data, iteration, src1);
                 Record<DebugDataRecord::DEST_IN>(debug_data, iteration, dest);
-                float24 rsq_res = float24::FromFloat32(1.0f / std::sqrt(src1[0].ToFloat32()));
+                f24 rsq_res = f24::FromFloat32(1.0f / std::sqrt(src1[0].ToFloat32()));
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
@@ -345,8 +343,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
-                    dest[i] = (src1[i] >= src2[i]) ? float24::FromFloat32(1.0f)
+                    dest[i] = (src1[i] >= src2[i]) ? f24::One() : f24::Zero();
-                                                   : float24::FromFloat32(0.0f);
                 }
                 Record<DebugDataRecord::DEST_OUT>(debug_data, iteration, dest);
                 break;
@@ -360,8 +357,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
-                    dest[i] = (src1[i] < src2[i]) ? float24::FromFloat32(1.0f)
+                    dest[i] = (src1[i] < src2[i]) ? f24::One() : f24::Zero();
-                                                  : float24::FromFloat32(0.0f);
                 }
                 Record<DebugDataRecord::DEST_OUT>(debug_data, iteration, dest);
                 break;
@@ -413,7 +409,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                 Record<DebugDataRecord::DEST_IN>(debug_data, iteration, dest);
 
                 // EX2 only takes first component exp2 and writes it to all dest components
-                float24 ex2_res = float24::FromFloat32(std::exp2(src1[0].ToFloat32()));
+                f24 ex2_res = f24::FromFloat32(std::exp2(src1[0].ToFloat32()));
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
@@ -430,7 +426,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                 Record<DebugDataRecord::DEST_IN>(debug_data, iteration, dest);
 
                 // LG2 only takes the first component log2 and writes it to all dest components
-                float24 lg2_res = float24::FromFloat32(std::log2(src1[0].ToFloat32()));
+                f24 lg2_res = f24::FromFloat32(std::log2(src1[0].ToFloat32()));
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
@@ -466,17 +462,17 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                         ? 0
                         : state.address_registers[instr.mad.address_register_index - 1];
 
-                const float24* src1_ = LookupSourceRegister(instr.mad.GetSrc1(is_inverted));
+                const f24* src1_ = LookupSourceRegister(instr.mad.GetSrc1(is_inverted));
-                const float24* src2_ = LookupSourceRegister(instr.mad.GetSrc2(is_inverted) +
+                const f24* src2_ = LookupSourceRegister(instr.mad.GetSrc2(is_inverted) +
-                                                            (!is_inverted * address_offset));
+                                                        (!is_inverted * address_offset));
-                const float24* src3_ = LookupSourceRegister(instr.mad.GetSrc3(is_inverted) +
+                const f24* src3_ = LookupSourceRegister(instr.mad.GetSrc3(is_inverted) +
-                                                            (is_inverted * address_offset));
+                                                        (is_inverted * address_offset));
 
                 const bool negate_src1 = ((bool)mad_swizzle.negate_src1 != false);
                 const bool negate_src2 = ((bool)mad_swizzle.negate_src2 != false);
                 const bool negate_src3 = ((bool)mad_swizzle.negate_src3 != false);
 
-                float24 src1[4] = {
+                f24 src1[4] = {
                     src1_[(int)mad_swizzle.src1_selector_0.Value()],
                     src1_[(int)mad_swizzle.src1_selector_1.Value()],
                     src1_[(int)mad_swizzle.src1_selector_2.Value()],
@@ -488,7 +484,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                     src1[2] = -src1[2];
                     src1[3] = -src1[3];
                 }
-                float24 src2[4] = {
+                f24 src2[4] = {
                     src2_[(int)mad_swizzle.src2_selector_0.Value()],
                     src2_[(int)mad_swizzle.src2_selector_1.Value()],
                     src2_[(int)mad_swizzle.src2_selector_2.Value()],
@@ -500,7 +496,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                     src2[2] = -src2[2];
                     src2[3] = -src2[3];
                 }
-                float24 src3[4] = {
+                f24 src3[4] = {
                     src3_[(int)mad_swizzle.src3_selector_0.Value()],
                     src3_[(int)mad_swizzle.src3_selector_1.Value()],
                     src3_[(int)mad_swizzle.src3_selector_2.Value()],
@@ -513,12 +509,11 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData
                     src3[3] = -src3[3];
                 }
 
-                float24* dest =
+                f24* dest = (instr.mad.dest.Value() < 0x10)
-                    (instr.mad.dest.Value() < 0x10)
+                                ? &state.registers.output[instr.mad.dest.Value().GetIndex()][0]
-                        ? &state.registers.output[instr.mad.dest.Value().GetIndex()][0]
+                            : (instr.mad.dest.Value() < 0x20)
-                    : (instr.mad.dest.Value() < 0x20)
+                                ? &state.registers.temporary[instr.mad.dest.Value().GetIndex()][0]
-                        ? &state.registers.temporary[instr.mad.dest.Value().GetIndex()][0]
+                                : dummy_vec4_float24;
-                        : dummy_vec4_float24;
 
                 Record<DebugDataRecord::SRC1>(debug_data, iteration, src1);
                 Record<DebugDataRecord::SRC2>(debug_data, iteration, src2);
@@ -687,7 +682,7 @@ DebugData<true> InterpreterEngine::ProduceDebugInfo(const ShaderSetup& setup,
     DebugData<true> debug_data;
 
     // Setup input register table
-    state.registers.input.fill(Common::Vec4<float24>::AssignToAll(float24::Zero()));
+    state.registers.input.fill(Common::Vec4<f24>::AssignToAll(f24::Zero()));
     state.LoadInput(config, input);
     RunInterpreter(setup, state, debug_data, setup.engine_data.entry_point);
     return debug_data;

src/video_core/shader/shader_jit_x64.cpp

@@ -5,6 +5,7 @@
 #include "common/arch.h"
 #if CITRA_ARCH(x86_64)
 
+#include "common/assert.h"
 #include "common/microprofile.h"
 #include "video_core/shader/shader.h"
 #include "video_core/shader/shader_jit_x64.h"

src/video_core/shader/shader_jit_x64_compiler.cpp

@@ -813,7 +813,7 @@ void JitShader::Compile_JMP(Instruction instr) {
     }
 }
 
-static void Emit(GSEmitter* emitter, Common::Vec4<float24> (*output)[16]) {
+static void Emit(GSEmitter* emitter, Common::Vec4<f24> (*output)[16]) {
     emitter->Emit(*output);
 }
 

src/video_core/vertex_loader.cpp

@@ -98,7 +98,7 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
                 const s8* srcdata = reinterpret_cast<const s8*>(
                     VideoCore::g_memory->GetPhysicalPointer(source_addr));
                 for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
-                    input.attr[i][comp] = float24::FromFloat32(srcdata[comp]);
+                    input.attr[i][comp] = f24::FromFloat32(srcdata[comp]);
                 }
                 break;
             }
@@ -106,7 +106,7 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
                 const u8* srcdata = reinterpret_cast<const u8*>(
                     VideoCore::g_memory->GetPhysicalPointer(source_addr));
                 for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
-                    input.attr[i][comp] = float24::FromFloat32(srcdata[comp]);
+                    input.attr[i][comp] = f24::FromFloat32(srcdata[comp]);
                 }
                 break;
             }
@@ -114,7 +114,7 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
                 const s16* srcdata = reinterpret_cast<const s16*>(
                     VideoCore::g_memory->GetPhysicalPointer(source_addr));
                 for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
-                    input.attr[i][comp] = float24::FromFloat32(srcdata[comp]);
+                    input.attr[i][comp] = f24::FromFloat32(srcdata[comp]);
                 }
                 break;
             }
@@ -122,7 +122,7 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
                 const float* srcdata = reinterpret_cast<const float*>(
                     VideoCore::g_memory->GetPhysicalPointer(source_addr));
                 for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
-                    input.attr[i][comp] = float24::FromFloat32(srcdata[comp]);
+                    input.attr[i][comp] = f24::FromFloat32(srcdata[comp]);
                 }
                 break;
             }
@@ -132,8 +132,7 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
             // is *not* carried over from the default attribute settings even if they're
             // enabled for this attribute.
             for (unsigned int comp = vertex_attribute_elements[i]; comp < 4; ++comp) {
-                input.attr[i][comp] =
+                input.attr[i][comp] = comp == 3 ? f24::One() : f24::Zero();
-                    comp == 3 ? float24::FromFloat32(1.0f) : float24::FromFloat32(0.0f);
             }
 
             LOG_TRACE(HW_GPU,

src/video_core/video_core.cpp

@@ -40,7 +40,7 @@ void Init(Frontend::EmuWindow& emu_window, Frontend::EmuWindow* secondary_window
 
     switch (graphics_api) {
     case Settings::GraphicsAPI::Software:
-        g_renderer = std::make_unique<VideoCore::RendererSoftware>(system, emu_window);
+        g_renderer = std::make_unique<SwRenderer::RendererSoftware>(system, emu_window);
         break;
     case Settings::GraphicsAPI::OpenGL:
         g_renderer = std::make_unique<OpenGL::RendererOpenGL>(system, emu_window, secondary_window);