Merge pull request #6103 from GPUCode/tex-cache

Start abstracting the rasterizer cache from OpenGL

.github/workflows/ci.yml

@@ -157,7 +157,7 @@ jobs:
           TRANSIFEX_API_TOKEN: ${{ secrets.TRANSIFEX_API_TOKEN }}
   release:
     runs-on: ubuntu-latest
-    needs: [build, macos, android]
+    needs: [build, android]
     if: ${{ startsWith(github.ref, 'refs/tags/') }}
     steps:
       - uses: actions/download-artifact@v3

.gitignore

@@ -1,6 +1,7 @@
 # Build directory
 [Bb]uild/
 doc-build/
+build-*/
 
 # Generated source files
 src/common/scm_rev.cpp

externals/glad/include/KHR/khrplatform.h

@@ -90,12 +90,20 @@
  *                                  int arg2) KHRONOS_APIATTRIBUTES;
  */
 
+#if defined(__SCITECH_SNAP__) && !defined(KHRONOS_STATIC)
+#   define KHRONOS_STATIC 1
+#endif
+
 /*-------------------------------------------------------------------------
  * Definition of KHRONOS_APICALL
  *-------------------------------------------------------------------------
  * This precedes the return type of the function in the function prototype.
  */
-#if defined(_WIN32) && !defined(__SCITECH_SNAP__)
+#if defined(KHRONOS_STATIC)
+    /* If the preprocessor constant KHRONOS_STATIC is defined, make the
+     * header compatible with static linking. */
+#   define KHRONOS_APICALL
+#elif defined(_WIN32)
 #   define KHRONOS_APICALL __declspec(dllimport)
 #elif defined (__SYMBIAN32__)
 #   define KHRONOS_APICALL IMPORT_C
@@ -145,6 +153,20 @@ typedef int64_t                 khronos_int64_t;
 typedef uint64_t                khronos_uint64_t;
 #define KHRONOS_SUPPORT_INT64   1
 #define KHRONOS_SUPPORT_FLOAT   1
+/*
+ * To support platform where unsigned long cannot be used interchangeably with
+ * inptr_t (e.g. CHERI-extended ISAs), we can use the stdint.h intptr_t.
+ * Ideally, we could just use (u)intptr_t everywhere, but this could result in
+ * ABI breakage if khronos_uintptr_t is changed from unsigned long to
+ * unsigned long long or similar (this results in different C++ name mangling).
+ * To avoid changes for existing platforms, we restrict usage of intptr_t to
+ * platforms where the size of a pointer is larger than the size of long.
+ */
+#if defined(__SIZEOF_LONG__) && defined(__SIZEOF_POINTER__)
+#if __SIZEOF_POINTER__ > __SIZEOF_LONG__
+#define KHRONOS_USE_INTPTR_T
+#endif
+#endif
 
 #elif defined(__VMS ) || defined(__sgi)
 
@@ -227,14 +249,21 @@ typedef unsigned short int     khronos_uint16_t;
  * pointers are 64 bits, but 'long' is still 32 bits. Win64 appears
  * to be the only LLP64 architecture in current use.
  */
-#ifdef _WIN64
+#ifdef KHRONOS_USE_INTPTR_T
+typedef intptr_t               khronos_intptr_t;
+typedef uintptr_t              khronos_uintptr_t;
+#elif defined(_WIN64)
 typedef signed   long long int khronos_intptr_t;
 typedef unsigned long long int khronos_uintptr_t;
-typedef signed   long long int khronos_ssize_t;
-typedef unsigned long long int khronos_usize_t;
 #else
 typedef signed   long  int     khronos_intptr_t;
 typedef unsigned long  int     khronos_uintptr_t;
+#endif
+
+#if defined(_WIN64)
+typedef signed   long long int khronos_ssize_t;
+typedef unsigned long long int khronos_usize_t;
+#else
 typedef signed   long  int     khronos_ssize_t;
 typedef unsigned long  int     khronos_usize_t;
 #endif

src/citra/emu_window/emu_window_sdl2.cpp

@@ -147,14 +147,16 @@ EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
 
     SDL_SetMainReady();
 
-    SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 3);
     if (Settings::values.use_gles) {
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 3);
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 2);
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_ES);
     } else {
+        SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 4);
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 3);
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_CORE);
     }
+
     SDL_GL_SetAttribute(SDL_GL_DOUBLEBUFFER, 1);
     SDL_GL_SetAttribute(SDL_GL_RED_SIZE, 8);
     SDL_GL_SetAttribute(SDL_GL_GREEN_SIZE, 8);

src/citra_qt/bootmanager.cpp

@@ -9,7 +9,7 @@
 #include <QOffscreenSurface>
 #include <QOpenGLContext>
 #include <QOpenGLFunctions>
-#include <QOpenGLFunctions_3_3_Core>
+#include <QOpenGLFunctions_4_3_Core>
 #include <fmt/format.h>
 #include "citra_qt/bootmanager.h"
 #include "citra_qt/main.h"
@@ -143,7 +143,7 @@ void OpenGLWindow::Present() {
         VideoCore::g_renderer->TryPresent(100);
     }
     context->swapBuffers(this);
-    auto f = context->versionFunctions<QOpenGLFunctions_3_3_Core>();
+    auto f = context->versionFunctions<QOpenGLFunctions_4_3_Core>();
     f->glFinish();
     QWindow::requestUpdate();
 }

src/citra_qt/main.cpp

@@ -11,7 +11,7 @@
 #include <QFutureWatcher>
 #include <QLabel>
 #include <QMessageBox>
-#include <QOpenGLFunctions_3_3_Core>
+#include <QOpenGLFunctions_4_3_Core>
 #include <QSysInfo>
 #include <QtConcurrent/QtConcurrentRun>
 #include <QtGui>
@@ -916,12 +916,12 @@ bool GMainWindow::LoadROM(const QString& filename) {
 
     Frontend::ScopeAcquireContext scope(*render_window);
 
-    const QString below_gl33_title = tr("OpenGL 3.3 Unsupported");
+    const QString below_gl43_title = tr("OpenGL 4.3 Unsupported");
-    const QString below_gl33_message = tr("Your GPU may not support OpenGL 3.3, or you do not "
+    const QString below_gl43_message = tr("Your GPU may not support OpenGL 4.3, or you do not "
                                           "have the latest graphics driver.");
 
-    if (!QOpenGLContext::globalShareContext()->versionFunctions<QOpenGLFunctions_3_3_Core>()) {
+    if (!QOpenGLContext::globalShareContext()->versionFunctions<QOpenGLFunctions_4_3_Core>()) {
-        QMessageBox::critical(this, below_gl33_title, below_gl33_message);
+        QMessageBox::critical(this, below_gl43_title, below_gl43_message);
         return false;
     }
 
@@ -991,8 +991,8 @@ bool GMainWindow::LoadROM(const QString& filename) {
                    "proper drivers for your graphics card from the manufacturer's website."));
             break;
 
-        case Core::System::ResultStatus::ErrorVideoCore_ErrorBelowGL33:
+        case Core::System::ResultStatus::ErrorVideoCore_ErrorBelowGL43:
-            QMessageBox::critical(this, below_gl33_title, below_gl33_message);
+            QMessageBox::critical(this, below_gl43_title, below_gl43_message);
             break;
 
         default:
@@ -2441,7 +2441,7 @@ int main(int argc, char* argv[]) {
     QCoreApplication::setApplicationName(QStringLiteral("Citra"));
 
     QSurfaceFormat format;
-    format.setVersion(3, 3);
+    format.setVersion(4, 3);
     format.setProfile(QSurfaceFormat::CoreProfile);
     format.setSwapInterval(0);
     // TODO: expose a setting for buffer value (ie default/single/double/triple)

src/common/vector_math.h

@@ -31,6 +31,7 @@
 #pragma once
 
 #include <cmath>
+#include <cstring>
 #include <type_traits>
 #include <boost/serialization/access.hpp>
 
@@ -60,6 +61,10 @@ public:
         return &x;
     }
 
+    const T* AsArray() const {
+        return &x;
+    }
+
     constexpr Vec2() = default;
     constexpr Vec2(const T& x_, const T& y_) : x(x_), y(y_) {}
 
@@ -123,6 +128,14 @@ public:
         return x * x + y * y;
     }
 
+    [[nodiscard]] constexpr bool operator!=(const Vec2& other) const {
+        return std::memcmp(AsArray(), other.AsArray(), sizeof(Vec2)) != 0;
+    }
+
+    [[nodiscard]] constexpr bool operator==(const Vec2& other) const {
+        return std::memcmp(AsArray(), other.AsArray(), sizeof(Vec2)) == 0;
+    }
+
     // Only implemented for T=float
     [[nodiscard]] float Length() const;
     [[nodiscard]] float Normalize(); // returns the previous length, which is often useful
@@ -184,6 +197,8 @@ template <typename T, typename V>
 }
 
 using Vec2f = Vec2<float>;
+using Vec2i = Vec2<int>;
+using Vec2u = Vec2<unsigned int>;
 
 template <>
 inline float Vec2<float>::Length() const {
@@ -216,6 +231,10 @@ public:
         return &x;
     }
 
+    const T* AsArray() const {
+        return &x;
+    }
+
     constexpr Vec3() = default;
     constexpr Vec3(const T& x_, const T& y_, const T& z_) : x(x_), y(y_), z(z_) {}
 
@@ -280,6 +299,14 @@ public:
         return *this;
     }
 
+    [[nodiscard]] constexpr bool operator!=(const Vec3& other) const {
+        return std::memcmp(AsArray(), other.AsArray(), sizeof(Vec3)) != 0;
+    }
+
+    [[nodiscard]] constexpr bool operator==(const Vec3& other) const {
+        return std::memcmp(AsArray(), other.AsArray(), sizeof(Vec3)) == 0;
+    }
+
     [[nodiscard]] constexpr T Length2() const {
         return x * x + y * y + z * z;
     }
@@ -412,6 +439,8 @@ inline float Vec3<float>::Normalize() {
 }
 
 using Vec3f = Vec3<float>;
+using Vec3i = Vec3<int>;
+using Vec3u = Vec3<unsigned int>;
 
 template <typename T>
 class Vec4 {
@@ -434,6 +463,10 @@ public:
         return &x;
     }
 
+    const T* AsArray() const {
+        return &x;
+    }
+
     constexpr Vec4() = default;
     constexpr Vec4(const T& x_, const T& y_, const T& z_, const T& w_)
         : x(x_), y(y_), z(z_), w(w_) {}
@@ -503,6 +536,14 @@ public:
         return *this;
     }
 
+    [[nodiscard]] constexpr bool operator!=(const Vec4& other) const {
+        return std::memcmp(AsArray(), other.AsArray(), sizeof(Vec4)) != 0;
+    }
+
+    [[nodiscard]] constexpr bool operator==(const Vec4& other) const {
+        return std::memcmp(AsArray(), other.AsArray(), sizeof(Vec4)) == 0;
+    }
+
     [[nodiscard]] constexpr T Length2() const {
         return x * x + y * y + z * z + w * w;
     }
@@ -623,6 +664,8 @@ template <typename T, typename V>
 }
 
 using Vec4f = Vec4<float>;
+using Vec4i = Vec4<int>;
+using Vec4u = Vec4<unsigned int>;
 
 template <typename T>
 constexpr decltype(T{} * T{} + T{} * T{}) Dot(const Vec2<T>& a, const Vec2<T>& b) {

src/core/core.cpp

@@ -422,8 +422,8 @@ System::ResultStatus System::Init(Frontend::EmuWindow& emu_window, u32 system_mo
         switch (result) {
         case VideoCore::ResultStatus::ErrorGenericDrivers:
             return ResultStatus::ErrorVideoCore_ErrorGenericDrivers;
-        case VideoCore::ResultStatus::ErrorBelowGL33:
+        case VideoCore::ResultStatus::ErrorBelowGL43:
-            return ResultStatus::ErrorVideoCore_ErrorBelowGL33;
+            return ResultStatus::ErrorVideoCore_ErrorBelowGL43;
         default:
             return ResultStatus::ErrorVideoCore;
         }

src/core/core.h

@@ -87,8 +87,8 @@ public:
         ErrorVideoCore,                     ///< Error in the video core
         ErrorVideoCore_ErrorGenericDrivers, ///< Error in the video core due to the user having
                                             /// generic drivers installed
-        ErrorVideoCore_ErrorBelowGL33,      ///< Error in the video core due to the user not having
+        ErrorVideoCore_ErrorBelowGL43,      ///< Error in the video core due to the user not having
-                                            /// OpenGL 3.3 or higher
+                                            /// OpenGL 4.3 or higher
         ErrorSavestate,                     ///< Error saving or loading
         ShutdownRequested,                  ///< Emulated program requested a system shutdown
         ErrorUnknown                        ///< Any other error

src/video_core/CMakeLists.txt

@@ -23,12 +23,23 @@ add_library(video_core STATIC
     regs_texturing.h
     renderer_base.cpp
     renderer_base.h
+    rasterizer_cache/cached_surface.cpp
+    rasterizer_cache/cached_surface.h
+    rasterizer_cache/morton_swizzle.h
+    rasterizer_cache/pixel_format.h
+    rasterizer_cache/rasterizer_cache.cpp
+    rasterizer_cache/rasterizer_cache.h
+    rasterizer_cache/rasterizer_cache_types.h
+    rasterizer_cache/rasterizer_cache_utils.cpp
+    rasterizer_cache/rasterizer_cache_utils.h
+    rasterizer_cache/surface_params.cpp
+    rasterizer_cache/surface_params.h
+    rasterizer_cache/texture_runtime.cpp
+    rasterizer_cache/texture_runtime.h
     renderer_opengl/frame_dumper_opengl.cpp
     renderer_opengl/frame_dumper_opengl.h
     renderer_opengl/gl_rasterizer.cpp
     renderer_opengl/gl_rasterizer.h
-    renderer_opengl/gl_rasterizer_cache.cpp
-    renderer_opengl/gl_rasterizer_cache.h
     renderer_opengl/gl_resource_manager.cpp
     renderer_opengl/gl_resource_manager.h
     renderer_opengl/gl_shader_decompiler.cpp
@@ -45,8 +56,6 @@ add_library(video_core STATIC
     renderer_opengl/gl_state.h
     renderer_opengl/gl_stream_buffer.cpp
     renderer_opengl/gl_stream_buffer.h
-    renderer_opengl/gl_surface_params.cpp
-    renderer_opengl/gl_surface_params.h
     renderer_opengl/gl_vars.cpp
     renderer_opengl/gl_vars.h
     renderer_opengl/pica_to_gl.h

src/video_core/rasterizer_cache/cached_surface.cpp

@@ -0,0 +1,475 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/microprofile.h"
+#include "common/scope_exit.h"
+#include "common/texture.h"
+#include "core/core.h"
+#include "video_core/rasterizer_cache/cached_surface.h"
+#include "video_core/rasterizer_cache/morton_swizzle.h"
+#include "video_core/rasterizer_cache/rasterizer_cache.h"
+#include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/texture_downloader_es.h"
+#include "video_core/renderer_opengl/texture_filters/texture_filterer.h"
+
+namespace OpenGL {
+
+static Aspect ToAspect(SurfaceType type) {
+    switch (type) {
+    case SurfaceType::Color:
+    case SurfaceType::Texture:
+    case SurfaceType::Fill:
+        return Aspect::Color;
+    case SurfaceType::Depth:
+        return Aspect::Depth;
+    case SurfaceType::DepthStencil:
+        return Aspect::DepthStencil;
+    default:
+        LOG_CRITICAL(Render_OpenGL, "Unknown SurfaceType {}", type);
+        UNREACHABLE();
+    }
+
+    return Aspect::Color;
+}
+
+CachedSurface::~CachedSurface() {
+    if (texture.handle) {
+        auto tag = is_custom ? HostTextureTag{GetFormatTuple(PixelFormat::RGBA8),
+                                              custom_tex_info.width, custom_tex_info.height}
+                             : HostTextureTag{GetFormatTuple(pixel_format), GetScaledWidth(),
+                                              GetScaledHeight()};
+
+        owner.host_texture_recycler.emplace(tag, std::move(texture));
+    }
+}
+
+MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 192, 64));
+void CachedSurface::LoadGLBuffer(PAddr load_start, PAddr load_end) {
+    ASSERT(type != SurfaceType::Fill);
+    const bool need_swap =
+        GLES && (pixel_format == PixelFormat::RGBA8 || pixel_format == PixelFormat::RGB8);
+
+    const u8* const texture_src_data = VideoCore::g_memory->GetPhysicalPointer(addr);
+    if (texture_src_data == nullptr)
+        return;
+
+    if (gl_buffer.empty()) {
+        gl_buffer.resize(width * height * GetBytesPerPixel(pixel_format));
+    }
+
+    // TODO: Should probably be done in ::Memory:: and check for other regions too
+    if (load_start < Memory::VRAM_VADDR_END && load_end > Memory::VRAM_VADDR_END)
+        load_end = Memory::VRAM_VADDR_END;
+
+    if (load_start < Memory::VRAM_VADDR && load_end > Memory::VRAM_VADDR)
+        load_start = Memory::VRAM_VADDR;
+
+    MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
+
+    ASSERT(load_start >= addr && load_end <= end);
+    const u32 start_offset = load_start - addr;
+
+    if (!is_tiled) {
+        ASSERT(type == SurfaceType::Color);
+        if (need_swap) {
+            // TODO(liushuyu): check if the byteswap here is 100% correct
+            // cannot fully test this
+            if (pixel_format == PixelFormat::RGBA8) {
+                for (std::size_t i = start_offset; i < load_end - addr; i += 4) {
+                    gl_buffer[i] = texture_src_data[i + 3];
+                    gl_buffer[i + 1] = texture_src_data[i + 2];
+                    gl_buffer[i + 2] = texture_src_data[i + 1];
+                    gl_buffer[i + 3] = texture_src_data[i];
+                }
+            } else if (pixel_format == PixelFormat::RGB8) {
+                for (std::size_t i = start_offset; i < load_end - addr; i += 3) {
+                    gl_buffer[i] = texture_src_data[i + 2];
+                    gl_buffer[i + 1] = texture_src_data[i + 1];
+                    gl_buffer[i + 2] = texture_src_data[i];
+                }
+            }
+        } else {
+            std::memcpy(&gl_buffer[start_offset], texture_src_data + start_offset,
+                        load_end - load_start);
+        }
+    } else {
+        if (type == SurfaceType::Texture) {
+            Pica::Texture::TextureInfo tex_info{};
+            tex_info.width = width;
+            tex_info.height = height;
+            tex_info.format = static_cast<Pica::TexturingRegs::TextureFormat>(pixel_format);
+            tex_info.SetDefaultStride();
+            tex_info.physical_address = addr;
+
+            const SurfaceInterval load_interval(load_start, load_end);
+            const auto rect = GetSubRect(FromInterval(load_interval));
+            ASSERT(FromInterval(load_interval).GetInterval() == load_interval);
+
+            for (unsigned y = rect.bottom; y < rect.top; ++y) {
+                for (unsigned x = rect.left; x < rect.right; ++x) {
+                    auto vec4 =
+                        Pica::Texture::LookupTexture(texture_src_data, x, height - 1 - y, tex_info);
+                    const std::size_t offset = (x + (width * y)) * 4;
+                    std::memcpy(&gl_buffer[offset], vec4.AsArray(), 4);
+                }
+            }
+        } else {
+            morton_to_gl_fns[static_cast<std::size_t>(pixel_format)](stride, height, &gl_buffer[0],
+                                                                     addr, load_start, load_end);
+        }
+    }
+}
+
+MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
+void CachedSurface::FlushGLBuffer(PAddr flush_start, PAddr flush_end) {
+    u8* const dst_buffer = VideoCore::g_memory->GetPhysicalPointer(addr);
+    if (dst_buffer == nullptr)
+        return;
+
+    ASSERT(gl_buffer.size() == width * height * GetBytesPerPixel(pixel_format));
+
+    // TODO: Should probably be done in ::Memory:: and check for other regions too
+    // same as loadglbuffer()
+    if (flush_start < Memory::VRAM_VADDR_END && flush_end > Memory::VRAM_VADDR_END)
+        flush_end = Memory::VRAM_VADDR_END;
+
+    if (flush_start < Memory::VRAM_VADDR && flush_end > Memory::VRAM_VADDR)
+        flush_start = Memory::VRAM_VADDR;
+
+    MICROPROFILE_SCOPE(OpenGL_SurfaceFlush);
+
+    ASSERT(flush_start >= addr && flush_end <= end);
+    const u32 start_offset = flush_start - addr;
+    const u32 end_offset = flush_end - addr;
+
+    if (type == SurfaceType::Fill) {
+        const u32 coarse_start_offset = start_offset - (start_offset % fill_size);
+        const u32 backup_bytes = start_offset % fill_size;
+        std::array<u8, 4> backup_data;
+        if (backup_bytes)
+            std::memcpy(&backup_data[0], &dst_buffer[coarse_start_offset], backup_bytes);
+
+        for (u32 offset = coarse_start_offset; offset < end_offset; offset += fill_size) {
+            std::memcpy(&dst_buffer[offset], &fill_data[0],
+                        std::min(fill_size, end_offset - offset));
+        }
+
+        if (backup_bytes)
+            std::memcpy(&dst_buffer[coarse_start_offset], &backup_data[0], backup_bytes);
+    } else if (!is_tiled) {
+        ASSERT(type == SurfaceType::Color);
+        if (pixel_format == PixelFormat::RGBA8 && GLES) {
+            for (std::size_t i = start_offset; i < flush_end - addr; i += 4) {
+                dst_buffer[i] = gl_buffer[i + 3];
+                dst_buffer[i + 1] = gl_buffer[i + 2];
+                dst_buffer[i + 2] = gl_buffer[i + 1];
+                dst_buffer[i + 3] = gl_buffer[i];
+            }
+        } else if (pixel_format == PixelFormat::RGB8 && GLES) {
+            for (std::size_t i = start_offset; i < flush_end - addr; i += 3) {
+                dst_buffer[i] = gl_buffer[i + 2];
+                dst_buffer[i + 1] = gl_buffer[i + 1];
+                dst_buffer[i + 2] = gl_buffer[i];
+            }
+        } else {
+            std::memcpy(dst_buffer + start_offset, &gl_buffer[start_offset],
+                        flush_end - flush_start);
+        }
+    } else {
+        gl_to_morton_fns[static_cast<std::size_t>(pixel_format)](stride, height, &gl_buffer[0],
+                                                                 addr, flush_start, flush_end);
+    }
+}
+
+bool CachedSurface::LoadCustomTexture(u64 tex_hash) {
+    auto& custom_tex_cache = Core::System::GetInstance().CustomTexCache();
+    const auto& image_interface = Core::System::GetInstance().GetImageInterface();
+
+    if (custom_tex_cache.IsTextureCached(tex_hash)) {
+        custom_tex_info = custom_tex_cache.LookupTexture(tex_hash);
+        return true;
+    }
+
+    if (!custom_tex_cache.CustomTextureExists(tex_hash)) {
+        return false;
+    }
+
+    const auto& path_info = custom_tex_cache.LookupTexturePathInfo(tex_hash);
+    if (!image_interface->DecodePNG(custom_tex_info.tex, custom_tex_info.width,
+                                    custom_tex_info.height, path_info.path)) {
+        LOG_ERROR(Render_OpenGL, "Failed to load custom texture {}", path_info.path);
+        return false;
+    }
+
+    const std::bitset<32> width_bits(custom_tex_info.width);
+    const std::bitset<32> height_bits(custom_tex_info.height);
+    if (width_bits.count() != 1 || height_bits.count() != 1) {
+        LOG_ERROR(Render_OpenGL, "Texture {} size is not a power of 2", path_info.path);
+        return false;
+    }
+
+    LOG_DEBUG(Render_OpenGL, "Loaded custom texture from {}", path_info.path);
+    Common::FlipRGBA8Texture(custom_tex_info.tex, custom_tex_info.width, custom_tex_info.height);
+    custom_tex_cache.CacheTexture(tex_hash, custom_tex_info.tex, custom_tex_info.width,
+                                  custom_tex_info.height);
+    return true;
+}
+
+void CachedSurface::DumpTexture(GLuint target_tex, u64 tex_hash) {
+    // Make sure the texture size is a power of 2
+    // If not, the surface is actually a framebuffer
+    std::bitset<32> width_bits(width);
+    std::bitset<32> height_bits(height);
+    if (width_bits.count() != 1 || height_bits.count() != 1) {
+        LOG_WARNING(Render_OpenGL, "Not dumping {:016X} because size isn't a power of 2 ({}x{})",
+                    tex_hash, width, height);
+        return;
+    }
+
+    // Dump texture to RGBA8 and encode as PNG
+    const auto& image_interface = Core::System::GetInstance().GetImageInterface();
+    auto& custom_tex_cache = Core::System::GetInstance().CustomTexCache();
+    std::string dump_path =
+        fmt::format("{}textures/{:016X}/", FileUtil::GetUserPath(FileUtil::UserPath::DumpDir),
+                    Core::System::GetInstance().Kernel().GetCurrentProcess()->codeset->program_id);
+    if (!FileUtil::CreateFullPath(dump_path)) {
+        LOG_ERROR(Render, "Unable to create {}", dump_path);
+        return;
+    }
+
+    dump_path += fmt::format("tex1_{}x{}_{:016X}_{}.png", width, height, tex_hash, pixel_format);
+    if (!custom_tex_cache.IsTextureDumped(tex_hash) && !FileUtil::Exists(dump_path)) {
+        custom_tex_cache.SetTextureDumped(tex_hash);
+
+        LOG_INFO(Render_OpenGL, "Dumping texture to {}", dump_path);
+        std::vector<u8> decoded_texture;
+        decoded_texture.resize(width * height * 4);
+        OpenGLState state = OpenGLState::GetCurState();
+        GLuint old_texture = state.texture_units[0].texture_2d;
+        state.Apply();
+        /*
+           GetTexImageOES is used even if not using OpenGL ES to work around a small issue that
+           happens if using custom textures with texture dumping at the same.
+           Let's say there's 2 textures that are both 32x32 and one of them gets replaced with a
+           higher quality 256x256 texture. If the 256x256 texture is displayed first and the
+           32x32 texture gets uploaded to the same underlying OpenGL texture, the 32x32 texture
+           will appear in the corner of the 256x256 texture. If texture dumping is enabled and
+           the 32x32 is undumped, Citra will attempt to dump it. Since the underlying OpenGL
+           texture is still 256x256, Citra crashes because it thinks the texture is only 32x32.
+           GetTexImageOES conveniently only dumps the specified region, and works on both
+           desktop and ES.
+        */
+        owner.texture_downloader_es->GetTexImage(GL_TEXTURE_2D, 0, GL_RGBA, GL_UNSIGNED_BYTE,
+                                                 height, width, &decoded_texture[0]);
+        state.texture_units[0].texture_2d = old_texture;
+        state.Apply();
+        Common::FlipRGBA8Texture(decoded_texture, width, height);
+        if (!image_interface->EncodePNG(dump_path, decoded_texture, width, height))
+            LOG_ERROR(Render_OpenGL, "Failed to save decoded texture");
+    }
+}
+
+MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 192, 64));
+void CachedSurface::UploadGLTexture(Common::Rectangle<u32> rect) {
+    if (type == SurfaceType::Fill) {
+        return;
+    }
+
+    MICROPROFILE_SCOPE(OpenGL_TextureUL);
+    ASSERT(gl_buffer.size() == width * height * GetBytesPerPixel(pixel_format));
+
+    u64 tex_hash = 0;
+
+    if (Settings::values.dump_textures || Settings::values.custom_textures) {
+        tex_hash = Common::ComputeHash64(gl_buffer.data(), gl_buffer.size());
+    }
+
+    if (Settings::values.custom_textures) {
+        is_custom = LoadCustomTexture(tex_hash);
+    }
+
+    // Load data from memory to the surface
+    GLint x0 = static_cast<GLint>(rect.left);
+    GLint y0 = static_cast<GLint>(rect.bottom);
+    std::size_t buffer_offset = (y0 * stride + x0) * GetBytesPerPixel(pixel_format);
+
+    const FormatTuple& tuple = GetFormatTuple(pixel_format);
+    GLuint target_tex = texture.handle;
+
+    // If not 1x scale, create 1x texture that we will blit from to replace texture subrect in
+    // surface
+    OGLTexture unscaled_tex;
+    if (res_scale != 1) {
+        x0 = 0;
+        y0 = 0;
+
+        if (is_custom) {
+            const auto& tuple = GetFormatTuple(PixelFormat::RGBA8);
+            unscaled_tex =
+                owner.AllocateSurfaceTexture(tuple, custom_tex_info.width, custom_tex_info.height);
+        } else {
+            unscaled_tex = owner.AllocateSurfaceTexture(tuple, rect.GetWidth(), rect.GetHeight());
+        }
+
+        target_tex = unscaled_tex.handle;
+    }
+
+    OpenGLState cur_state = OpenGLState::GetCurState();
+
+    GLuint old_tex = cur_state.texture_units[0].texture_2d;
+    cur_state.texture_units[0].texture_2d = target_tex;
+    cur_state.Apply();
+
+    // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
+    ASSERT(stride * GetBytesPerPixel(pixel_format) % 4 == 0);
+    if (is_custom) {
+        if (res_scale == 1) {
+            texture = owner.AllocateSurfaceTexture(GetFormatTuple(PixelFormat::RGBA8),
+                                                   custom_tex_info.width, custom_tex_info.height);
+            cur_state.texture_units[0].texture_2d = texture.handle;
+            cur_state.Apply();
+        }
+
+        // Always going to be using rgba8
+        glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(custom_tex_info.width));
+
+        glActiveTexture(GL_TEXTURE0);
+        glTexSubImage2D(GL_TEXTURE_2D, 0, x0, y0, custom_tex_info.width, custom_tex_info.height,
+                        GL_RGBA, GL_UNSIGNED_BYTE, custom_tex_info.tex.data());
+    } else {
+        glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(stride));
+
+        glActiveTexture(GL_TEXTURE0);
+        glTexSubImage2D(GL_TEXTURE_2D, 0, x0, y0, static_cast<GLsizei>(rect.GetWidth()),
+                        static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
+                        &gl_buffer[buffer_offset]);
+    }
+
+    glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
+    if (Settings::values.dump_textures && !is_custom) {
+        DumpTexture(target_tex, tex_hash);
+    }
+
+    cur_state.texture_units[0].texture_2d = old_tex;
+    cur_state.Apply();
+
+    if (res_scale != 1) {
+        auto scaled_rect = rect;
+        scaled_rect.left *= res_scale;
+        scaled_rect.top *= res_scale;
+        scaled_rect.right *= res_scale;
+        scaled_rect.bottom *= res_scale;
+
+        const u32 width = is_custom ? custom_tex_info.width : rect.GetWidth();
+        const u32 height = is_custom ? custom_tex_info.height : rect.GetHeight();
+        const Common::Rectangle<u32> from_rect{0, height, width, 0};
+
+        if (!owner.texture_filterer->Filter(unscaled_tex, from_rect, texture, scaled_rect, type)) {
+            const Aspect aspect = ToAspect(type);
+            runtime.BlitTextures(unscaled_tex, {aspect, from_rect}, texture, {aspect, scaled_rect});
+        }
+    }
+
+    InvalidateAllWatcher();
+}
+
+MICROPROFILE_DEFINE(OpenGL_TextureDL, "OpenGL", "Texture Download", MP_RGB(128, 192, 64));
+void CachedSurface::DownloadGLTexture(const Common::Rectangle<u32>& rect) {
+    if (type == SurfaceType::Fill) {
+        return;
+    }
+
+    MICROPROFILE_SCOPE(OpenGL_TextureDL);
+
+    if (gl_buffer.empty()) {
+        gl_buffer.resize(width * height * GetBytesPerPixel(pixel_format));
+    }
+
+    OpenGLState state = OpenGLState::GetCurState();
+    OpenGLState prev_state = state;
+    SCOPE_EXIT({ prev_state.Apply(); });
+
+    const FormatTuple& tuple = GetFormatTuple(pixel_format);
+
+    // Ensure no bad interactions with GL_PACK_ALIGNMENT
+    ASSERT(stride * GetBytesPerPixel(pixel_format) % 4 == 0);
+    glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(stride));
+    const std::size_t buffer_offset =
+        (rect.bottom * stride + rect.left) * GetBytesPerPixel(pixel_format);
+
+    // If not 1x scale, blit scaled texture to a new 1x texture and use that to flush
+    const Aspect aspect = ToAspect(type);
+    if (res_scale != 1) {
+        auto scaled_rect = rect;
+        scaled_rect.left *= res_scale;
+        scaled_rect.top *= res_scale;
+        scaled_rect.right *= res_scale;
+        scaled_rect.bottom *= res_scale;
+
+        const Common::Rectangle<u32> unscaled_tex_rect{0, rect.GetHeight(), rect.GetWidth(), 0};
+        auto unscaled_tex = owner.AllocateSurfaceTexture(tuple, rect.GetWidth(), rect.GetHeight());
+        // Blit scaled texture to the unscaled one
+        runtime.BlitTextures(texture, {aspect, scaled_rect}, unscaled_tex,
+                             {aspect, unscaled_tex_rect});
+
+        state.texture_units[0].texture_2d = unscaled_tex.handle;
+        state.Apply();
+
+        glActiveTexture(GL_TEXTURE0);
+        if (GLES) {
+            owner.texture_downloader_es->GetTexImage(GL_TEXTURE_2D, 0, tuple.format, tuple.type,
+                                                     rect.GetHeight(), rect.GetWidth(),
+                                                     &gl_buffer[buffer_offset]);
+        } else {
+            glGetTexImage(GL_TEXTURE_2D, 0, tuple.format, tuple.type, &gl_buffer[buffer_offset]);
+        }
+    } else {
+        runtime.ReadTexture(texture, {aspect, rect}, tuple, gl_buffer.data());
+    }
+
+    glPixelStorei(GL_PACK_ROW_LENGTH, 0);
+}
+
+bool CachedSurface::CanFill(const SurfaceParams& dest_surface,
+                            SurfaceInterval fill_interval) const {
+    if (type == SurfaceType::Fill && IsRegionValid(fill_interval) &&
+        boost::icl::first(fill_interval) >= addr &&
+        boost::icl::last_next(fill_interval) <= end && // dest_surface is within our fill range
+        dest_surface.FromInterval(fill_interval).GetInterval() ==
+            fill_interval) { // make sure interval is a rectangle in dest surface
+        if (fill_size * 8 != dest_surface.GetFormatBpp()) {
+            // Check if bits repeat for our fill_size
+            const u32 dest_bytes_per_pixel = std::max(dest_surface.GetFormatBpp() / 8, 1u);
+            std::vector<u8> fill_test(fill_size * dest_bytes_per_pixel);
+
+            for (u32 i = 0; i < dest_bytes_per_pixel; ++i)
+                std::memcpy(&fill_test[i * fill_size], &fill_data[0], fill_size);
+
+            for (u32 i = 0; i < fill_size; ++i)
+                if (std::memcmp(&fill_test[dest_bytes_per_pixel * i], &fill_test[0],
+                                dest_bytes_per_pixel) != 0)
+                    return false;
+
+            if (dest_surface.GetFormatBpp() == 4 && (fill_test[0] & 0xF) != (fill_test[0] >> 4))
+                return false;
+        }
+        return true;
+    }
+    return false;
+}
+
+bool CachedSurface::CanCopy(const SurfaceParams& dest_surface,
+                            SurfaceInterval copy_interval) const {
+    SurfaceParams subrect_params = dest_surface.FromInterval(copy_interval);
+    ASSERT(subrect_params.GetInterval() == copy_interval);
+    if (CanSubRect(subrect_params))
+        return true;
+
+    if (CanFill(dest_surface, copy_interval))
+        return true;
+
+    return false;
+}
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/cached_surface.h

@@ -0,0 +1,137 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include <list>
+#include "common/assert.h"
+#include "core/custom_tex_cache.h"
+#include "video_core/rasterizer_cache/surface_params.h"
+#include "video_core/rasterizer_cache/texture_runtime.h"
+
+namespace OpenGL {
+
+/**
+ * A watcher that notifies whether a cached surface has been changed. This is useful for caching
+ * surface collection objects, including texture cube and mipmap.
+ */
+class SurfaceWatcher {
+    friend class CachedSurface;
+
+public:
+    explicit SurfaceWatcher(std::weak_ptr<CachedSurface>&& surface) : surface(std::move(surface)) {}
+
+    /// Checks whether the surface has been changed.
+    bool IsValid() const {
+        return !surface.expired() && valid;
+    }
+
+    /// Marks that the content of the referencing surface has been updated to the watcher user.
+    void Validate() {
+        ASSERT(!surface.expired());
+        valid = true;
+    }
+
+    /// Gets the referencing surface. Returns null if the surface has been destroyed
+    Surface Get() const {
+        return surface.lock();
+    }
+
+private:
+    std::weak_ptr<CachedSurface> surface;
+    bool valid = false;
+};
+
+class RasterizerCacheOpenGL;
+
+class CachedSurface : public SurfaceParams, public std::enable_shared_from_this<CachedSurface> {
+public:
+    CachedSurface(SurfaceParams params, RasterizerCacheOpenGL& owner, TextureRuntime& runtime)
+        : SurfaceParams(params), owner(owner), runtime(runtime) {}
+    ~CachedSurface();
+
+    /// Read/Write data in 3DS memory to/from gl_buffer
+    void LoadGLBuffer(PAddr load_start, PAddr load_end);
+    void FlushGLBuffer(PAddr flush_start, PAddr flush_end);
+
+    /// Custom texture loading and dumping
+    bool LoadCustomTexture(u64 tex_hash);
+    void DumpTexture(GLuint target_tex, u64 tex_hash);
+
+    /// Upload/Download data in gl_buffer in/to this surface's texture
+    void UploadGLTexture(Common::Rectangle<u32> rect);
+    void DownloadGLTexture(const Common::Rectangle<u32>& rect);
+
+    bool CanFill(const SurfaceParams& dest_surface, SurfaceInterval fill_interval) const;
+    bool CanCopy(const SurfaceParams& dest_surface, SurfaceInterval copy_interval) const;
+
+    bool IsRegionValid(SurfaceInterval interval) const {
+        return (invalid_regions.find(interval) == invalid_regions.end());
+    }
+
+    bool IsSurfaceFullyInvalid() const {
+        auto interval = GetInterval();
+        return *invalid_regions.equal_range(interval).first == interval;
+    }
+
+    std::shared_ptr<SurfaceWatcher> CreateWatcher() {
+        auto watcher = std::make_shared<SurfaceWatcher>(weak_from_this());
+        watchers.push_front(watcher);
+        return watcher;
+    }
+
+    void InvalidateAllWatcher() {
+        for (const auto& watcher : watchers) {
+            if (auto locked = watcher.lock()) {
+                locked->valid = false;
+            }
+        }
+    }
+
+    void UnlinkAllWatcher() {
+        for (const auto& watcher : watchers) {
+            if (auto locked = watcher.lock()) {
+                locked->valid = false;
+                locked->surface.reset();
+            }
+        }
+
+        watchers.clear();
+    }
+
+public:
+    bool registered = false;
+    SurfaceRegions invalid_regions;
+    std::vector<u8> gl_buffer;
+
+    // Number of bytes to read from fill_data
+    u32 fill_size = 0;
+    std::array<u8, 4> fill_data;
+    OGLTexture texture;
+
+    // level_watchers[i] watches the (i+1)-th level mipmap source surface
+    std::array<std::shared_ptr<SurfaceWatcher>, 7> level_watchers;
+    u32 max_level = 0;
+
+    // Information about custom textures
+    bool is_custom = false;
+    Core::CustomTexInfo custom_tex_info;
+
+private:
+    RasterizerCacheOpenGL& owner;
+    TextureRuntime& runtime;
+    std::list<std::weak_ptr<SurfaceWatcher>> watchers;
+};
+
+struct CachedTextureCube {
+    OGLTexture texture;
+    u16 res_scale = 1;
+    std::shared_ptr<SurfaceWatcher> px;
+    std::shared_ptr<SurfaceWatcher> nx;
+    std::shared_ptr<SurfaceWatcher> py;
+    std::shared_ptr<SurfaceWatcher> ny;
+    std::shared_ptr<SurfaceWatcher> pz;
+    std::shared_ptr<SurfaceWatcher> nz;
+};
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/morton_swizzle.h

@@ -0,0 +1,171 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include "common/alignment.h"
+#include "core/memory.h"
+#include "video_core/rasterizer_cache/pixel_format.h"
+#include "video_core/renderer_opengl/gl_vars.h"
+#include "video_core/utils.h"
+#include "video_core/video_core.h"
+
+namespace OpenGL {
+
+template <bool morton_to_gl, PixelFormat format>
+static void MortonCopyTile(u32 stride, u8* tile_buffer, u8* gl_buffer) {
+    constexpr u32 bytes_per_pixel = GetFormatBpp(format) / 8;
+    constexpr u32 aligned_bytes_per_pixel = GetBytesPerPixel(format);
+    for (u32 y = 0; y < 8; ++y) {
+        for (u32 x = 0; x < 8; ++x) {
+            u8* tile_ptr = tile_buffer + VideoCore::MortonInterleave(x, y) * bytes_per_pixel;
+            u8* gl_ptr = gl_buffer + ((7 - y) * stride + x) * aligned_bytes_per_pixel;
+            if constexpr (morton_to_gl) {
+                if constexpr (format == PixelFormat::D24S8) {
+                    gl_ptr[0] = tile_ptr[3];
+                    std::memcpy(gl_ptr + 1, tile_ptr, 3);
+                } else if (format == PixelFormat::RGBA8 && GLES) {
+                    // because GLES does not have ABGR format
+                    // so we will do byteswapping here
+                    gl_ptr[0] = tile_ptr[3];
+                    gl_ptr[1] = tile_ptr[2];
+                    gl_ptr[2] = tile_ptr[1];
+                    gl_ptr[3] = tile_ptr[0];
+                } else if (format == PixelFormat::RGB8 && GLES) {
+                    gl_ptr[0] = tile_ptr[2];
+                    gl_ptr[1] = tile_ptr[1];
+                    gl_ptr[2] = tile_ptr[0];
+                } else {
+                    std::memcpy(gl_ptr, tile_ptr, bytes_per_pixel);
+                }
+            } else {
+                if constexpr (format == PixelFormat::D24S8) {
+                    std::memcpy(tile_ptr, gl_ptr + 1, 3);
+                    tile_ptr[3] = gl_ptr[0];
+                } else if (format == PixelFormat::RGBA8 && GLES) {
+                    // because GLES does not have ABGR format
+                    // so we will do byteswapping here
+                    tile_ptr[0] = gl_ptr[3];
+                    tile_ptr[1] = gl_ptr[2];
+                    tile_ptr[2] = gl_ptr[1];
+                    tile_ptr[3] = gl_ptr[0];
+                } else if (format == PixelFormat::RGB8 && GLES) {
+                    tile_ptr[0] = gl_ptr[2];
+                    tile_ptr[1] = gl_ptr[1];
+                    tile_ptr[2] = gl_ptr[0];
+                } else {
+                    std::memcpy(tile_ptr, gl_ptr, bytes_per_pixel);
+                }
+            }
+        }
+    }
+}
+
+template <bool morton_to_gl, PixelFormat format>
+static void MortonCopy(u32 stride, u32 height, u8* gl_buffer, PAddr base, PAddr start, PAddr end) {
+    constexpr u32 bytes_per_pixel = GetFormatBpp(format) / 8;
+    constexpr u32 tile_size = bytes_per_pixel * 64;
+
+    constexpr u32 aligned_bytes_per_pixel = GetBytesPerPixel(format);
+    static_assert(aligned_bytes_per_pixel >= bytes_per_pixel, "");
+    gl_buffer += aligned_bytes_per_pixel - bytes_per_pixel;
+
+    const PAddr aligned_down_start = base + Common::AlignDown(start - base, tile_size);
+    const PAddr aligned_start = base + Common::AlignUp(start - base, tile_size);
+    const PAddr aligned_end = base + Common::AlignDown(end - base, tile_size);
+
+    ASSERT(!morton_to_gl || (aligned_start == start && aligned_end == end));
+
+    const u32 begin_pixel_index = (aligned_down_start - base) / bytes_per_pixel;
+    u32 x = (begin_pixel_index % (stride * 8)) / 8;
+    u32 y = (begin_pixel_index / (stride * 8)) * 8;
+
+    gl_buffer += ((height - 8 - y) * stride + x) * aligned_bytes_per_pixel;
+
+    auto glbuf_next_tile = [&] {
+        x = (x + 8) % stride;
+        gl_buffer += 8 * aligned_bytes_per_pixel;
+        if (!x) {
+            y += 8;
+            gl_buffer -= stride * 9 * aligned_bytes_per_pixel;
+        }
+    };
+
+    u8* tile_buffer = VideoCore::g_memory->GetPhysicalPointer(start);
+
+    if (start < aligned_start && !morton_to_gl) {
+        std::array<u8, tile_size> tmp_buf;
+        MortonCopyTile<morton_to_gl, format>(stride, &tmp_buf[0], gl_buffer);
+        std::memcpy(tile_buffer, &tmp_buf[start - aligned_down_start],
+                    std::min(aligned_start, end) - start);
+
+        tile_buffer += aligned_start - start;
+        glbuf_next_tile();
+    }
+
+    const u8* const buffer_end = tile_buffer + aligned_end - aligned_start;
+    PAddr current_paddr = aligned_start;
+    while (tile_buffer < buffer_end) {
+        // Pokemon Super Mystery Dungeon will try to use textures that go beyond
+        // the end address of VRAM. Stop reading if reaches invalid address
+        if (!VideoCore::g_memory->IsValidPhysicalAddress(current_paddr) ||
+            !VideoCore::g_memory->IsValidPhysicalAddress(current_paddr + tile_size)) {
+            LOG_ERROR(Render_OpenGL, "Out of bound texture");
+            break;
+        }
+        MortonCopyTile<morton_to_gl, format>(stride, tile_buffer, gl_buffer);
+        tile_buffer += tile_size;
+        current_paddr += tile_size;
+        glbuf_next_tile();
+    }
+
+    if (end > std::max(aligned_start, aligned_end) && !morton_to_gl) {
+        std::array<u8, tile_size> tmp_buf;
+        MortonCopyTile<morton_to_gl, format>(stride, &tmp_buf[0], gl_buffer);
+        std::memcpy(tile_buffer, &tmp_buf[0], end - aligned_end);
+    }
+}
+
+static constexpr std::array<void (*)(u32, u32, u8*, PAddr, PAddr, PAddr), 18> morton_to_gl_fns = {
+    MortonCopy<true, PixelFormat::RGBA8>,  // 0
+    MortonCopy<true, PixelFormat::RGB8>,   // 1
+    MortonCopy<true, PixelFormat::RGB5A1>, // 2
+    MortonCopy<true, PixelFormat::RGB565>, // 3
+    MortonCopy<true, PixelFormat::RGBA4>,  // 4
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,                             // 5 - 13
+    MortonCopy<true, PixelFormat::D16>,  // 14
+    nullptr,                             // 15
+    MortonCopy<true, PixelFormat::D24>,  // 16
+    MortonCopy<true, PixelFormat::D24S8> // 17
+};
+
+static constexpr std::array<void (*)(u32, u32, u8*, PAddr, PAddr, PAddr), 18> gl_to_morton_fns = {
+    MortonCopy<false, PixelFormat::RGBA8>,  // 0
+    MortonCopy<false, PixelFormat::RGB8>,   // 1
+    MortonCopy<false, PixelFormat::RGB5A1>, // 2
+    MortonCopy<false, PixelFormat::RGB565>, // 3
+    MortonCopy<false, PixelFormat::RGBA4>,  // 4
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,                              // 5 - 13
+    MortonCopy<false, PixelFormat::D16>,  // 14
+    nullptr,                              // 15
+    MortonCopy<false, PixelFormat::D24>,  // 16
+    MortonCopy<false, PixelFormat::D24S8> // 17
+};
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/pixel_format.h

@@ -0,0 +1,196 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include <string_view>
+#include "core/hw/gpu.h"
+#include "video_core/regs_framebuffer.h"
+#include "video_core/regs_texturing.h"
+
+namespace OpenGL {
+
+constexpr u32 PIXEL_FORMAT_COUNT = 18;
+
+enum class PixelFormat : u8 {
+    // First 5 formats are shared between textures and color buffers
+    RGBA8 = 0,
+    RGB8 = 1,
+    RGB5A1 = 2,
+    RGB565 = 3,
+    RGBA4 = 4,
+    // Texture-only formats
+    IA8 = 5,
+    RG8 = 6,
+    I8 = 7,
+    A8 = 8,
+    IA4 = 9,
+    I4 = 10,
+    A4 = 11,
+    ETC1 = 12,
+    ETC1A4 = 13,
+    // Depth buffer-only formats
+    D16 = 14,
+    D24 = 16,
+    D24S8 = 17,
+    Invalid = 255,
+};
+
+enum class SurfaceType {
+    Color = 0,
+    Texture = 1,
+    Depth = 2,
+    DepthStencil = 3,
+    Fill = 4,
+    Invalid = 5
+};
+
+constexpr std::string_view PixelFormatAsString(PixelFormat format) {
+    switch (format) {
+    case PixelFormat::RGBA8:
+        return "RGBA8";
+    case PixelFormat::RGB8:
+        return "RGB8";
+    case PixelFormat::RGB5A1:
+        return "RGB5A1";
+    case PixelFormat::RGB565:
+        return "RGB565";
+    case PixelFormat::RGBA4:
+        return "RGBA4";
+    case PixelFormat::IA8:
+        return "IA8";
+    case PixelFormat::RG8:
+        return "RG8";
+    case PixelFormat::I8:
+        return "I8";
+    case PixelFormat::A8:
+        return "A8";
+    case PixelFormat::IA4:
+        return "IA4";
+    case PixelFormat::I4:
+        return "I4";
+    case PixelFormat::A4:
+        return "A4";
+    case PixelFormat::ETC1:
+        return "ETC1";
+    case PixelFormat::ETC1A4:
+        return "ETC1A4";
+    case PixelFormat::D16:
+        return "D16";
+    case PixelFormat::D24:
+        return "D24";
+    case PixelFormat::D24S8:
+        return "D24S8";
+    default:
+        return "NotReal";
+    }
+}
+
+constexpr PixelFormat PixelFormatFromTextureFormat(Pica::TexturingRegs::TextureFormat format) {
+    const u32 format_index = static_cast<u32>(format);
+    return (format_index < 14) ? static_cast<PixelFormat>(format) : PixelFormat::Invalid;
+}
+
+constexpr PixelFormat PixelFormatFromColorFormat(Pica::FramebufferRegs::ColorFormat format) {
+    const u32 format_index = static_cast<u32>(format);
+    return (format_index < 5) ? static_cast<PixelFormat>(format) : PixelFormat::Invalid;
+}
+
+constexpr PixelFormat PixelFormatFromDepthFormat(Pica::FramebufferRegs::DepthFormat format) {
+    const u32 format_index = static_cast<u32>(format);
+    return (format_index < 4) ? static_cast<PixelFormat>(format_index + 14) : PixelFormat::Invalid;
+}
+
+constexpr PixelFormat PixelFormatFromGPUPixelFormat(GPU::Regs::PixelFormat format) {
+    const u32 format_index = static_cast<u32>(format);
+    switch (format) {
+    // RGB565 and RGB5A1 are switched in PixelFormat compared to ColorFormat
+    case GPU::Regs::PixelFormat::RGB565:
+        return PixelFormat::RGB565;
+    case GPU::Regs::PixelFormat::RGB5A1:
+        return PixelFormat::RGB5A1;
+    default:
+        return (format_index < 5) ? static_cast<PixelFormat>(format) : PixelFormat::Invalid;
+    }
+}
+
+constexpr SurfaceType GetFormatType(PixelFormat pixel_format) {
+    const u32 format_index = static_cast<u32>(pixel_format);
+    if (format_index < 5) {
+        return SurfaceType::Color;
+    }
+
+    if (format_index < 14) {
+        return SurfaceType::Texture;
+    }
+
+    if (pixel_format == PixelFormat::D16 || pixel_format == PixelFormat::D24) {
+        return SurfaceType::Depth;
+    }
+
+    if (pixel_format == PixelFormat::D24S8) {
+        return SurfaceType::DepthStencil;
+    }
+
+    return SurfaceType::Invalid;
+}
+
+constexpr bool CheckFormatsBlittable(PixelFormat source_format, PixelFormat dest_format) {
+    SurfaceType source_type = GetFormatType(source_format);
+    SurfaceType dest_type = GetFormatType(dest_format);
+
+    if ((source_type == SurfaceType::Color || source_type == SurfaceType::Texture) &&
+        (dest_type == SurfaceType::Color || dest_type == SurfaceType::Texture)) {
+        return true;
+    }
+
+    if (source_type == SurfaceType::Depth && dest_type == SurfaceType::Depth) {
+        return true;
+    }
+
+    if (source_type == SurfaceType::DepthStencil && dest_type == SurfaceType::DepthStencil) {
+        return true;
+    }
+
+    return false;
+}
+
+constexpr u32 GetFormatBpp(PixelFormat format) {
+    switch (format) {
+    case PixelFormat::RGBA8:
+    case PixelFormat::D24S8:
+        return 32;
+    case PixelFormat::RGB8:
+    case PixelFormat::D24:
+        return 24;
+    case PixelFormat::RGB5A1:
+    case PixelFormat::RGB565:
+    case PixelFormat::RGBA4:
+    case PixelFormat::IA8:
+    case PixelFormat::RG8:
+    case PixelFormat::D16:
+        return 16;
+    case PixelFormat::I8:
+    case PixelFormat::A8:
+    case PixelFormat::IA4:
+    case PixelFormat::ETC1A4:
+        return 8;
+    case PixelFormat::I4:
+    case PixelFormat::A4:
+    case PixelFormat::ETC1:
+        return 4;
+    default:
+        return 0;
+    }
+}
+
+constexpr u32 GetBytesPerPixel(PixelFormat format) {
+    // OpenGL needs 4 bpp alignment for D24 since using GL_UNSIGNED_INT as type
+    if (format == PixelFormat::D24 || GetFormatType(format) == SurfaceType::Texture) {
+        return 4;
+    }
+
+    return GetFormatBpp(format) / 8;
+}
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/rasterizer_cache.h

@@ -0,0 +1,131 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include <unordered_map>
+#include "video_core/rasterizer_cache/cached_surface.h"
+#include "video_core/rasterizer_cache/rasterizer_cache_utils.h"
+#include "video_core/rasterizer_cache/surface_params.h"
+#include "video_core/texture/texture_decode.h"
+
+namespace OpenGL {
+
+enum class ScaleMatch {
+    Exact,   // only accept same res scale
+    Upscale, // only allow higher scale than params
+    Ignore   // accept every scaled res
+};
+
+class TextureDownloaderES;
+class TextureFilterer;
+class FormatReinterpreterOpenGL;
+
+class RasterizerCacheOpenGL : NonCopyable {
+public:
+    RasterizerCacheOpenGL();
+    ~RasterizerCacheOpenGL();
+
+    /// Blit one surface's texture to another
+    bool BlitSurfaces(const Surface& src_surface, const Common::Rectangle<u32>& src_rect,
+                      const Surface& dst_surface, const Common::Rectangle<u32>& dst_rect);
+
+    /// Copy one surface's region to another
+    void CopySurface(const Surface& src_surface, const Surface& dst_surface,
+                     SurfaceInterval copy_interval);
+
+    /// Load a texture from 3DS memory to OpenGL and cache it (if not already cached)
+    Surface GetSurface(const SurfaceParams& params, ScaleMatch match_res_scale,
+                       bool load_if_create);
+
+    /// Attempt to find a subrect (resolution scaled) of a surface, otherwise loads a texture from
+    /// 3DS memory to OpenGL and caches it (if not already cached)
+    SurfaceRect_Tuple GetSurfaceSubRect(const SurfaceParams& params, ScaleMatch match_res_scale,
+                                        bool load_if_create);
+
+    /// Get a surface based on the texture configuration
+    Surface GetTextureSurface(const Pica::TexturingRegs::FullTextureConfig& config);
+    Surface GetTextureSurface(const Pica::Texture::TextureInfo& info, u32 max_level = 0);
+
+    /// Get a texture cube based on the texture configuration
+    const CachedTextureCube& GetTextureCube(const TextureCubeConfig& config);
+
+    /// Get the color and depth surfaces based on the framebuffer configuration
+    SurfaceSurfaceRect_Tuple GetFramebufferSurfaces(bool using_color_fb, bool using_depth_fb,
+                                                    const Common::Rectangle<s32>& viewport_rect);
+
+    /// Get a surface that matches the fill config
+    Surface GetFillSurface(const GPU::Regs::MemoryFillConfig& config);
+
+    /// Get a surface that matches a "texture copy" display transfer config
+    SurfaceRect_Tuple GetTexCopySurface(const SurfaceParams& params);
+
+    /// Write any cached resources overlapping the region back to memory (if dirty)
+    void FlushRegion(PAddr addr, u32 size, Surface flush_surface = nullptr);
+
+    /// Mark region as being invalidated by region_owner (nullptr if 3DS memory)
+    void InvalidateRegion(PAddr addr, u32 size, const Surface& region_owner);
+
+    /// Flush all cached resources tracked by this cache manager
+    void FlushAll();
+
+    /// Clear all cached resources tracked by this cache manager
+    void ClearAll(bool flush);
+
+    // Textures from destroyed surfaces are stored here to be recyled to reduce allocation overhead
+    // in the driver
+    // this must be placed above the surface_cache to ensure all cached surfaces are destroyed
+    // before destroying the recycler
+    std::unordered_multimap<HostTextureTag, OGLTexture> host_texture_recycler;
+
+private:
+    void DuplicateSurface(const Surface& src_surface, const Surface& dest_surface);
+
+    /// Update surface's texture for given region when necessary
+    void ValidateSurface(const Surface& surface, PAddr addr, u32 size);
+
+    // Returns false if there is a surface in the cache at the interval with the same bit-width,
+    bool NoUnimplementedReinterpretations(const OpenGL::Surface& surface,
+                                          OpenGL::SurfaceParams& params,
+                                          const OpenGL::SurfaceInterval& interval);
+
+    // Return true if a surface with an invalid pixel format exists at the interval
+    bool IntervalHasInvalidPixelFormat(SurfaceParams& params, const SurfaceInterval& interval);
+
+    // Attempt to find a reinterpretable surface in the cache and use it to copy for validation
+    bool ValidateByReinterpretation(const Surface& surface, SurfaceParams& params,
+                                    const SurfaceInterval& interval);
+
+    /// Create a new surface
+    Surface CreateSurface(const SurfaceParams& params);
+
+    /// Register surface into the cache
+    void RegisterSurface(const Surface& surface);
+
+    /// Remove surface from the cache
+    void UnregisterSurface(const Surface& surface);
+
+    /// Increase/decrease the number of surface in pages touching the specified region
+    void UpdatePagesCachedCount(PAddr addr, u32 size, int delta);
+
+    TextureRuntime runtime;
+    SurfaceCache surface_cache;
+    PageMap cached_pages;
+    SurfaceMap dirty_regions;
+    SurfaceSet remove_surfaces;
+
+    u16 resolution_scale_factor;
+
+    std::unordered_map<TextureCubeConfig, CachedTextureCube> texture_cube_cache;
+
+    std::recursive_mutex mutex;
+
+public:
+    OGLTexture AllocateSurfaceTexture(const FormatTuple& format_tuple, u32 width, u32 height);
+
+    std::unique_ptr<TextureFilterer> texture_filterer;
+    std::unique_ptr<FormatReinterpreterOpenGL> format_reinterpreter;
+    std::unique_ptr<TextureDownloaderES> texture_downloader_es;
+};
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/rasterizer_cache_types.h

@@ -0,0 +1,38 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include <memory>
+#include <set>
+#include <tuple>
+#include <boost/icl/interval_map.hpp>
+#include <boost/icl/interval_set.hpp>
+#include "common/common_types.h"
+#include "common/math_util.h"
+
+namespace OpenGL {
+
+class CachedSurface;
+using Surface = std::shared_ptr<CachedSurface>;
+
+// Declare rasterizer interval types
+using SurfaceInterval = boost::icl::right_open_interval<PAddr>;
+using SurfaceSet = std::set<Surface>;
+using SurfaceRegions = boost::icl::interval_set<PAddr, std::less, SurfaceInterval>;
+using SurfaceMap =
+    boost::icl::interval_map<PAddr, Surface, boost::icl::partial_absorber, std::less,
+                             boost::icl::inplace_plus, boost::icl::inter_section, SurfaceInterval>;
+using SurfaceCache =
+    boost::icl::interval_map<PAddr, SurfaceSet, boost::icl::partial_absorber, std::less,
+                             boost::icl::inplace_plus, boost::icl::inter_section, SurfaceInterval>;
+
+static_assert(std::is_same<SurfaceRegions::interval_type, SurfaceCache::interval_type>() &&
+                  std::is_same<SurfaceMap::interval_type, SurfaceCache::interval_type>(),
+              "Incorrect interval types");
+
+using SurfaceRect_Tuple = std::tuple<Surface, Common::Rectangle<u32>>;
+using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, Common::Rectangle<u32>>;
+using PageMap = boost::icl::interval_map<u32, int>;
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/rasterizer_cache_utils.cpp

@@ -0,0 +1,56 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include <glad/glad.h>
+#include "video_core/rasterizer_cache/rasterizer_cache_utils.h"
+#include "video_core/renderer_opengl/gl_vars.h"
+
+namespace OpenGL {
+
+constexpr FormatTuple tex_tuple = {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE};
+
+static constexpr std::array<FormatTuple, 4> depth_format_tuples = {{
+    {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // D16
+    {},
+    {GL_DEPTH_COMPONENT24, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT},   // D24
+    {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // D24S8
+}};
+
+static constexpr std::array<FormatTuple, 5> fb_format_tuples = {{
+    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8},     // RGBA8
+    {GL_RGB8, GL_BGR, GL_UNSIGNED_BYTE},              // RGB8
+    {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_5_5_5_1}, // RGB5A1
+    {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5},     // RGB565
+    {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4},   // RGBA4
+}};
+
+// Same as above, with minor changes for OpenGL ES. Replaced
+// GL_UNSIGNED_INT_8_8_8_8 with GL_UNSIGNED_BYTE and
+// GL_BGR with GL_RGB
+static constexpr std::array<FormatTuple, 5> fb_format_tuples_oes = {{
+    {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE},            // RGBA8
+    {GL_RGB8, GL_RGB, GL_UNSIGNED_BYTE},              // RGB8
+    {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_5_5_5_1}, // RGB5A1
+    {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5},     // RGB565
+    {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4},   // RGBA4
+}};
+
+const FormatTuple& GetFormatTuple(PixelFormat pixel_format) {
+    const SurfaceType type = GetFormatType(pixel_format);
+    const std::size_t format_index = static_cast<std::size_t>(pixel_format);
+
+    if (type == SurfaceType::Color) {
+        ASSERT(format_index < fb_format_tuples.size());
+        return (GLES ? fb_format_tuples_oes : fb_format_tuples)[format_index];
+    } else if (type == SurfaceType::Depth || type == SurfaceType::DepthStencil) {
+        const std::size_t tuple_idx = format_index - 14;
+        ASSERT(tuple_idx < depth_format_tuples.size());
+        return depth_format_tuples[tuple_idx];
+    }
+
+    return tex_tuple;
+}
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/rasterizer_cache_utils.h

@@ -0,0 +1,73 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include <functional>
+#include "common/hash.h"
+#include "video_core/rasterizer_cache/pixel_format.h"
+
+namespace OpenGL {
+
+struct FormatTuple {
+    int internal_format;
+    u32 format;
+    u32 type;
+};
+
+const FormatTuple& GetFormatTuple(PixelFormat pixel_format);
+
+struct HostTextureTag {
+    FormatTuple format_tuple{};
+    u32 width = 0;
+    u32 height = 0;
+
+    bool operator==(const HostTextureTag& rhs) const noexcept {
+        return std::memcmp(this, &rhs, sizeof(HostTextureTag)) == 0;
+    };
+
+    const u64 Hash() const {
+        return Common::ComputeHash64(this, sizeof(HostTextureTag));
+    }
+};
+
+struct TextureCubeConfig {
+    PAddr px;
+    PAddr nx;
+    PAddr py;
+    PAddr ny;
+    PAddr pz;
+    PAddr nz;
+    u32 width;
+    Pica::TexturingRegs::TextureFormat format;
+
+    bool operator==(const TextureCubeConfig& rhs) const {
+        return std::memcmp(this, &rhs, sizeof(TextureCubeConfig)) == 0;
+    }
+
+    bool operator!=(const TextureCubeConfig& rhs) const {
+        return std::memcmp(this, &rhs, sizeof(TextureCubeConfig)) != 0;
+    }
+
+    const u64 Hash() const {
+        return Common::ComputeHash64(this, sizeof(TextureCubeConfig));
+    }
+};
+
+} // namespace OpenGL
+
+namespace std {
+template <>
+struct hash<OpenGL::HostTextureTag> {
+    std::size_t operator()(const OpenGL::HostTextureTag& tag) const noexcept {
+        return tag.Hash();
+    }
+};
+
+template <>
+struct hash<OpenGL::TextureCubeConfig> {
+    std::size_t operator()(const OpenGL::TextureCubeConfig& config) const noexcept {
+        return config.Hash();
+    }
+};
+} // namespace std

src/video_core/rasterizer_cache/surface_params.cpp

@@ -1,10 +1,10 @@
-// Copyright 2020 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #include "common/alignment.h"
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
+#include "video_core/rasterizer_cache/rasterizer_cache.h"
-#include "video_core/renderer_opengl/gl_surface_params.h"
+#include "video_core/rasterizer_cache/surface_params.h"
 
 namespace OpenGL {
 
@@ -12,6 +12,7 @@ SurfaceParams SurfaceParams::FromInterval(SurfaceInterval interval) const {
     SurfaceParams params = *this;
     const u32 tiled_size = is_tiled ? 8 : 1;
     const u32 stride_tiled_bytes = BytesInPixels(stride * tiled_size);
+
     PAddr aligned_start =
         addr + Common::AlignDown(boost::icl::first(interval) - addr, stride_tiled_bytes);
     PAddr aligned_end =
@@ -24,17 +25,19 @@ SurfaceParams SurfaceParams::FromInterval(SurfaceInterval interval) const {
         // 1 row
         ASSERT(aligned_end - aligned_start == stride_tiled_bytes);
         const u32 tiled_alignment = BytesInPixels(is_tiled ? 8 * 8 : 1);
+
         aligned_start =
             addr + Common::AlignDown(boost::icl::first(interval) - addr, tiled_alignment);
         aligned_end =
             addr + Common::AlignUp(boost::icl::last_next(interval) - addr, tiled_alignment);
+
         params.addr = aligned_start;
         params.width = PixelsInBytes(aligned_end - aligned_start) / tiled_size;
         params.stride = params.width;
         params.height = tiled_size;
     }
-    params.UpdateParams();
 
+    params.UpdateParams();
     return params;
 }
 
@@ -158,6 +161,7 @@ bool SurfaceParams::CanTexCopy(const SurfaceParams& texcopy_params) const {
         end < texcopy_params.end) {
         return false;
     }
+
     if (texcopy_params.width != texcopy_params.stride) {
         const u32 tile_stride = BytesInPixels(stride * (is_tiled ? 8 : 1));
         return (texcopy_params.addr - addr) % BytesInPixels(is_tiled ? 64 : 1) == 0 &&
@@ -165,6 +169,7 @@ bool SurfaceParams::CanTexCopy(const SurfaceParams& texcopy_params) const {
                (texcopy_params.height == 1 || texcopy_params.stride == tile_stride) &&
                ((texcopy_params.addr - addr) % tile_stride) + texcopy_params.width <= tile_stride;
     }
+
     return FromInterval(texcopy_params.GetInterval()).GetInterval() == texcopy_params.GetInterval();
 }
 

src/video_core/rasterizer_cache/surface_params.h

@@ -0,0 +1,91 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <climits>
+#include "video_core/rasterizer_cache/pixel_format.h"
+#include "video_core/rasterizer_cache/rasterizer_cache_types.h"
+
+namespace OpenGL {
+
+class SurfaceParams {
+public:
+    // Surface match traits
+    bool ExactMatch(const SurfaceParams& other_surface) const;
+    bool CanSubRect(const SurfaceParams& sub_surface) const;
+    bool CanExpand(const SurfaceParams& expanded_surface) const;
+    bool CanTexCopy(const SurfaceParams& texcopy_params) const;
+
+    Common::Rectangle<u32> GetSubRect(const SurfaceParams& sub_surface) const;
+    Common::Rectangle<u32> GetScaledSubRect(const SurfaceParams& sub_surface) const;
+
+    // Returns the outer rectangle containing "interval"
+    SurfaceParams FromInterval(SurfaceInterval interval) const;
+    SurfaceInterval GetSubRectInterval(Common::Rectangle<u32> unscaled_rect) const;
+
+    // Returns the region of the biggest valid rectange within interval
+    SurfaceInterval GetCopyableInterval(const Surface& src_surface) const;
+
+    /// Updates remaining members from the already set addr, width, height and pixel_format
+    void UpdateParams() {
+        if (stride == 0) {
+            stride = width;
+        }
+
+        type = GetFormatType(pixel_format);
+        size = !is_tiled ? BytesInPixels(stride * (height - 1) + width)
+                         : BytesInPixels(stride * 8 * (height / 8 - 1) + width * 8);
+        end = addr + size;
+    }
+
+    SurfaceInterval GetInterval() const {
+        return SurfaceInterval(addr, end);
+    }
+
+    u32 GetFormatBpp() const {
+        return OpenGL::GetFormatBpp(pixel_format);
+    }
+
+    u32 GetScaledWidth() const {
+        return width * res_scale;
+    }
+
+    u32 GetScaledHeight() const {
+        return height * res_scale;
+    }
+
+    Common::Rectangle<u32> GetRect() const {
+        return {0, height, width, 0};
+    }
+
+    Common::Rectangle<u32> GetScaledRect() const {
+        return {0, GetScaledHeight(), GetScaledWidth(), 0};
+    }
+
+    u32 PixelsInBytes(u32 size) const {
+        return size * 8 / GetFormatBpp();
+    }
+
+    u32 BytesInPixels(u32 pixels) const {
+        return pixels * GetFormatBpp() / 8;
+    }
+
+public:
+    PAddr addr = 0;
+    PAddr end = 0;
+    u32 size = 0;
+
+    u32 width = 0;
+    u32 height = 0;
+    u32 stride = 0;
+    u16 res_scale = 1;
+
+    bool is_tiled = false;
+    PixelFormat pixel_format = PixelFormat::Invalid;
+    SurfaceType type = SurfaceType::Invalid;
+};
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/texture_runtime.cpp

@@ -0,0 +1,195 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/scope_exit.h"
+#include "video_core/rasterizer_cache/rasterizer_cache_utils.h"
+#include "video_core/rasterizer_cache/texture_runtime.h"
+#include "video_core/renderer_opengl/gl_state.h"
+
+namespace OpenGL {
+
+GLbitfield ToBufferMask(Aspect aspect) {
+    switch (aspect) {
+    case Aspect::Color:
+        return GL_COLOR_BUFFER_BIT;
+    case Aspect::Depth:
+        return GL_DEPTH_BUFFER_BIT;
+    case Aspect::DepthStencil:
+        return GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
+    }
+}
+
+TextureRuntime::TextureRuntime() {
+    read_fbo.Create();
+    draw_fbo.Create();
+}
+
+void TextureRuntime::ReadTexture(const OGLTexture& tex, Subresource subresource,
+                                 const FormatTuple& tuple, u8* pixels) {
+
+    OpenGLState prev_state = OpenGLState::GetCurState();
+    SCOPE_EXIT({ prev_state.Apply(); });
+
+    OpenGLState state;
+    state.ResetTexture(tex.handle);
+    state.draw.read_framebuffer = read_fbo.handle;
+    state.Apply();
+
+    const u32 level = subresource.level;
+    switch (subresource.aspect) {
+    case Aspect::Color:
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, tex.handle,
+                               level);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
+                               0);
+        break;
+    case Aspect::Depth:
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, tex.handle,
+                               level);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
+        break;
+    case Aspect::DepthStencil:
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                               tex.handle, level);
+        break;
+    }
+
+    const auto& rect = subresource.region;
+    glReadPixels(rect.left, rect.bottom, rect.GetWidth(), rect.GetHeight(), tuple.format,
+                 tuple.type, pixels);
+}
+
+bool TextureRuntime::ClearTexture(const OGLTexture& tex, Subresource subresource,
+                                  ClearValue value) {
+    OpenGLState prev_state = OpenGLState::GetCurState();
+    SCOPE_EXIT({ prev_state.Apply(); });
+
+    // Setup scissor rectangle according to the clear rectangle
+    const auto& clear_rect = subresource.region;
+    OpenGLState state;
+    state.scissor.enabled = true;
+    state.scissor.x = clear_rect.left;
+    state.scissor.y = clear_rect.bottom;
+    state.scissor.width = clear_rect.GetWidth();
+    state.scissor.height = clear_rect.GetHeight();
+    state.draw.draw_framebuffer = draw_fbo.handle;
+    state.Apply();
+
+    const u32 level = subresource.level;
+    switch (subresource.aspect) {
+    case Aspect::Color:
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, tex.handle,
+                               level);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
+                               0);
+
+        state.color_mask.red_enabled = true;
+        state.color_mask.green_enabled = true;
+        state.color_mask.blue_enabled = true;
+        state.color_mask.alpha_enabled = true;
+        state.Apply();
+
+        glClearBufferfv(GL_COLOR, 0, value.color.AsArray());
+        break;
+    case Aspect::Depth:
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, tex.handle,
+                               level);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
+
+        state.depth.write_mask = GL_TRUE;
+        state.Apply();
+
+        glClearBufferfv(GL_DEPTH, 0, &value.depth);
+        break;
+    case Aspect::DepthStencil:
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                               tex.handle, level);
+
+        state.depth.write_mask = GL_TRUE;
+        state.stencil.write_mask = -1;
+        state.Apply();
+
+        glClearBufferfi(GL_DEPTH_STENCIL, 0, value.depth, value.stencil);
+        break;
+    }
+
+    return true;
+}
+
+bool TextureRuntime::CopyTextures(const OGLTexture& src_tex, Subresource src_subresource,
+                                  const OGLTexture& dst_tex, Subresource dst_subresource) {
+    return true;
+}
+
+bool TextureRuntime::BlitTextures(const OGLTexture& src_tex, Subresource src_subresource,
+                                  const OGLTexture& dst_tex, Subresource dst_subresource) {
+    OpenGLState prev_state = OpenGLState::GetCurState();
+    SCOPE_EXIT({ prev_state.Apply(); });
+
+    OpenGLState state;
+    state.draw.read_framebuffer = read_fbo.handle;
+    state.draw.draw_framebuffer = draw_fbo.handle;
+    state.Apply();
+
+    auto BindAttachment = [src_level = src_subresource.level, dst_level = dst_subresource.level](
+                              GLenum target, u32 src_tex, u32 dst_tex) -> void {
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, target, GL_TEXTURE_2D, src_tex, src_level);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, target, GL_TEXTURE_2D, dst_tex, dst_level);
+    };
+
+    // Sanity check; Can't blit a color texture to a depth buffer
+    ASSERT(src_subresource.aspect == dst_subresource.aspect);
+    switch (src_subresource.aspect) {
+    case Aspect::Color:
+        // Bind only color
+        BindAttachment(GL_COLOR_ATTACHMENT0, src_tex.handle, dst_tex.handle);
+        BindAttachment(GL_DEPTH_STENCIL_ATTACHMENT, 0, 0);
+        break;
+    case Aspect::Depth:
+        // Bind only depth
+        BindAttachment(GL_COLOR_ATTACHMENT0, 0, 0);
+        BindAttachment(GL_DEPTH_ATTACHMENT, src_tex.handle, dst_tex.handle);
+        BindAttachment(GL_STENCIL_ATTACHMENT, 0, 0);
+        break;
+    case Aspect::DepthStencil:
+        // Bind to combined depth + stencil
+        BindAttachment(GL_COLOR_ATTACHMENT0, 0, 0);
+        BindAttachment(GL_DEPTH_STENCIL_ATTACHMENT, src_tex.handle, dst_tex.handle);
+        break;
+    }
+
+    // TODO (wwylele): use GL_NEAREST for shadow map texture
+    // Note: shadow map is treated as RGBA8 format in PICA, as well as in the rasterizer cache, but
+    // doing linear intepolation componentwise would cause incorrect value. However, for a
+    // well-programmed game this code path should be rarely executed for shadow map with
+    // inconsistent scale.
+    const GLenum filter = src_subresource.aspect == Aspect::Color ? GL_LINEAR : GL_NEAREST;
+    const auto& src_rect = src_subresource.region;
+    const auto& dst_rect = dst_subresource.region;
+    glBlitFramebuffer(src_rect.left, src_rect.bottom, src_rect.right, src_rect.top, dst_rect.left,
+                      dst_rect.bottom, dst_rect.right, dst_rect.top,
+                      ToBufferMask(src_subresource.aspect), filter);
+
+    return true;
+}
+
+void TextureRuntime::GenerateMipmaps(const OGLTexture& tex, u32 max_level) {
+    OpenGLState prev_state = OpenGLState::GetCurState();
+    SCOPE_EXIT({ prev_state.Apply(); });
+
+    OpenGLState state;
+    state.texture_units[0].texture_2d = tex.handle;
+    state.Apply();
+
+    glActiveTexture(GL_TEXTURE0);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, max_level);
+
+    glGenerateMipmap(GL_TEXTURE_2D);
+}
+
+} // namespace OpenGL

src/video_core/rasterizer_cache/texture_runtime.h

@@ -0,0 +1,69 @@
+// Copyright 2022 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include "common/math_util.h"
+#include "common/vector_math.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+
+namespace OpenGL {
+
+// Describes the type of data a texture holds
+enum class Aspect { Color = 0, Depth = 1, DepthStencil = 2 };
+
+// A union for both color and depth/stencil clear values
+union ClearValue {
+    Common::Vec4f color;
+    struct {
+        float depth;
+        u8 stencil;
+    };
+};
+
+struct Subresource {
+    Subresource(Aspect aspect, Common::Rectangle<u32> region, u32 level = 0, u32 layer = 0)
+        : aspect(aspect), region(region), level(level), layer(layer) {}
+
+    Aspect aspect;
+    Common::Rectangle<u32> region;
+    u32 level = 0;
+    u32 layer = 0;
+};
+
+struct FormatTuple;
+
+/**
+ * Provides texture manipulation functions to the rasterizer cache
+ * Separating this into a class makes it easier to abstract graphics API code
+ */
+class TextureRuntime {
+public:
+    TextureRuntime();
+    ~TextureRuntime() = default;
+
+    // Copies the GPU pixel data to the provided pixels buffer
+    void ReadTexture(const OGLTexture& tex, Subresource subresource, const FormatTuple& tuple,
+                     u8* pixels);
+
+    // Fills the rectangle of the texture with the clear value provided
+    bool ClearTexture(const OGLTexture& texture, Subresource subresource, ClearValue value);
+
+    // Copies a rectangle of src_tex to another rectange of dst_rect
+    // NOTE: The width and height of the rectangles must be equal
+    bool CopyTextures(const OGLTexture& src_tex, Subresource src_subresource,
+                      const OGLTexture& dst_tex, Subresource dst_subresource);
+
+    // Copies a rectangle of src_tex to another rectange of dst_rect performing
+    // scaling and format conversions
+    bool BlitTextures(const OGLTexture& src_tex, Subresource src_subresource,
+                      const OGLTexture& dst_tex, Subresource dst_subresource);
+
+    // Generates mipmaps for all the available levels of the texture
+    void GenerateMipmaps(const OGLTexture& tex, u32 max_level);
+
+private:
+    OGLFramebuffer read_fbo, draw_fbo;
+};
+
+} // namespace OpenGL

src/video_core/regs.h

@@ -3,13 +3,6 @@
 // Refer to the license.txt file included.
 
 #pragma once
-
-#include <array>
-#include <cstddef>
-#include <string>
-
-#include "common/common_funcs.h"
-#include "common/common_types.h"
 #include "video_core/regs_framebuffer.h"
 #include "video_core/regs_lighting.h"
 #include "video_core/regs_pipeline.h"

src/video_core/regs_framebuffer.h

@@ -3,9 +3,7 @@
 // Refer to the license.txt file included.
 
 #pragma once
-
 #include <array>
-
 #include "common/assert.h"
 #include "common/bit_field.h"
 #include "common/common_funcs.h"

src/video_core/renderer_opengl/gl_format_reinterpreter.cpp

@@ -1,19 +1,13 @@
-// Copyright 2020 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "common/assert.h"
 #include "common/scope_exit.h"
 #include "video_core/renderer_opengl/gl_format_reinterpreter.h"
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_state.h"
-#include "video_core/renderer_opengl/gl_vars.h"
-#include "video_core/renderer_opengl/texture_filters/texture_filterer.h"
 
 namespace OpenGL {
 
-using PixelFormat = SurfaceParams::PixelFormat;
-
 class RGBA4toRGB5A1 final : public FormatReinterpreterBase {
 public:
     RGBA4toRGB5A1() {
@@ -66,15 +60,18 @@ void main() {
         vao.Create();
     }
 
-    void Reinterpret(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint read_fb_handle,
+    PixelFormat GetSourceFormat() const override {
-                     GLuint dst_tex, const Common::Rectangle<u32>& dst_rect,
+        return PixelFormat::RGBA4;
-                     GLuint draw_fb_handle) override {
+    }
+
+    void Reinterpret(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
+                     const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) override {
         OpenGLState prev_state = OpenGLState::GetCurState();
         SCOPE_EXIT({ prev_state.Apply(); });
 
         OpenGLState state;
-        state.texture_units[0].texture_2d = src_tex;
+        state.texture_units[0].texture_2d = src_tex.handle;
-        state.draw.draw_framebuffer = draw_fb_handle;
+        state.draw.draw_framebuffer = draw_fbo.handle;
         state.draw.shader_program = program.handle;
         state.draw.vertex_array = vao.handle;
         state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
@@ -82,8 +79,8 @@ void main() {
                           static_cast<GLsizei>(dst_rect.GetHeight())};
         state.Apply();
 
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex,
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
-                               0);
+                               dst_tex.handle, 0);
         glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
                                0);
 
@@ -99,127 +96,6 @@ private:
     OGLVertexArray vao;
 };
 
-class PixelBufferD24S8toABGR final : public FormatReinterpreterBase {
-public:
-    PixelBufferD24S8toABGR() {
-        attributeless_vao.Create();
-        d24s8_abgr_buffer.Create();
-        d24s8_abgr_buffer_size = 0;
-
-        constexpr std::string_view vs_source = R"(
-const vec2 vertices[4] = vec2[4](vec2(-1.0, -1.0), vec2(1.0, -1.0),
-                                 vec2(-1.0,  1.0), vec2(1.0,  1.0));
-void main() {
-    gl_Position = vec4(vertices[gl_VertexID], 0.0, 1.0);
-}
-)";
-
-        std::string fs_source = GLES ? fragment_shader_precision_OES : "";
-        fs_source += R"(
-uniform samplerBuffer tbo;
-uniform vec2 tbo_size;
-uniform vec4 viewport;
-
-out vec4 color;
-
-void main() {
-    vec2 tbo_coord = (gl_FragCoord.xy - viewport.xy) * tbo_size / viewport.zw;
-    int tbo_offset = int(tbo_coord.y) * int(tbo_size.x) + int(tbo_coord.x);
-    color = texelFetch(tbo, tbo_offset).rabg;
-}
-)";
-        d24s8_abgr_shader.Create(vs_source.data(), fs_source.c_str());
-
-        OpenGLState state = OpenGLState::GetCurState();
-        GLuint old_program = state.draw.shader_program;
-        state.draw.shader_program = d24s8_abgr_shader.handle;
-        state.Apply();
-
-        GLint tbo_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "tbo");
-        ASSERT(tbo_u_id != -1);
-        glUniform1i(tbo_u_id, 0);
-
-        state.draw.shader_program = old_program;
-        state.Apply();
-
-        d24s8_abgr_tbo_size_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "tbo_size");
-        ASSERT(d24s8_abgr_tbo_size_u_id != -1);
-        d24s8_abgr_viewport_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "viewport");
-        ASSERT(d24s8_abgr_viewport_u_id != -1);
-    }
-
-    ~PixelBufferD24S8toABGR() {}
-
-    void Reinterpret(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint read_fb_handle,
-                     GLuint dst_tex, const Common::Rectangle<u32>& dst_rect,
-                     GLuint draw_fb_handle) override {
-        OpenGLState prev_state = OpenGLState::GetCurState();
-        SCOPE_EXIT({ prev_state.Apply(); });
-
-        OpenGLState state;
-        state.draw.read_framebuffer = read_fb_handle;
-        state.draw.draw_framebuffer = draw_fb_handle;
-        state.Apply();
-
-        glBindBuffer(GL_PIXEL_PACK_BUFFER, d24s8_abgr_buffer.handle);
-
-        GLsizeiptr target_pbo_size =
-            static_cast<GLsizeiptr>(src_rect.GetWidth()) * src_rect.GetHeight() * 4;
-        if (target_pbo_size > d24s8_abgr_buffer_size) {
-            d24s8_abgr_buffer_size = target_pbo_size * 2;
-            glBufferData(GL_PIXEL_PACK_BUFFER, d24s8_abgr_buffer_size, nullptr, GL_STREAM_COPY);
-        }
-
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
-                               src_tex, 0);
-        glReadPixels(static_cast<GLint>(src_rect.left), static_cast<GLint>(src_rect.bottom),
-                     static_cast<GLsizei>(src_rect.GetWidth()),
-                     static_cast<GLsizei>(src_rect.GetHeight()), GL_DEPTH_STENCIL,
-                     GL_UNSIGNED_INT_24_8, 0);
-
-        glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
-
-        // PBO now contains src_tex in RABG format
-        state.draw.shader_program = d24s8_abgr_shader.handle;
-        state.draw.vertex_array = attributeless_vao.handle;
-        state.viewport.x = static_cast<GLint>(dst_rect.left);
-        state.viewport.y = static_cast<GLint>(dst_rect.bottom);
-        state.viewport.width = static_cast<GLsizei>(dst_rect.GetWidth());
-        state.viewport.height = static_cast<GLsizei>(dst_rect.GetHeight());
-        state.Apply();
-
-        OGLTexture tbo;
-        tbo.Create();
-        glActiveTexture(GL_TEXTURE0);
-        glBindTexture(GL_TEXTURE_BUFFER, tbo.handle);
-        glTexBuffer(GL_TEXTURE_BUFFER, GL_RGBA8, d24s8_abgr_buffer.handle);
-
-        glUniform2f(d24s8_abgr_tbo_size_u_id, static_cast<GLfloat>(src_rect.GetWidth()),
-                    static_cast<GLfloat>(src_rect.GetHeight()));
-        glUniform4f(d24s8_abgr_viewport_u_id, static_cast<GLfloat>(state.viewport.x),
-                    static_cast<GLfloat>(state.viewport.y),
-                    static_cast<GLfloat>(state.viewport.width),
-                    static_cast<GLfloat>(state.viewport.height));
-
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex,
-                               0);
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
-                               0);
-        glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
-
-        glBindTexture(GL_TEXTURE_BUFFER, 0);
-    }
-
-private:
-    OGLVertexArray attributeless_vao;
-    OGLBuffer d24s8_abgr_buffer;
-    GLsizeiptr d24s8_abgr_buffer_size;
-    OGLProgram d24s8_abgr_shader;
-    GLint d24s8_abgr_tbo_size_u_id;
-    GLint d24s8_abgr_viewport_u_id;
-};
-
 class ShaderD24S8toRGBA8 final : public FormatReinterpreterBase {
 public:
     ShaderD24S8toRGBA8() {
@@ -294,19 +170,23 @@ void main() {
         }
     }
 
-    void Reinterpret(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint read_fb_handle,
+    PixelFormat GetSourceFormat() const override {
-                     GLuint dst_tex, const Common::Rectangle<u32>& dst_rect,
+        return PixelFormat::D24S8;
-                     GLuint draw_fb_handle) override {
+    }
+
+    void Reinterpret(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
+                     const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) override {
         OpenGLState prev_state = OpenGLState::GetCurState();
         SCOPE_EXIT({ prev_state.Apply(); });
 
         OpenGLState state;
-        state.texture_units[0].texture_2d = src_tex;
+        state.texture_units[0].texture_2d = src_tex.handle;
 
         if (use_texture_view) {
             temp_tex.Create();
             glActiveTexture(GL_TEXTURE1);
-            glTextureView(temp_tex.handle, GL_TEXTURE_2D, src_tex, GL_DEPTH24_STENCIL8, 0, 1, 0, 1);
+            glTextureView(temp_tex.handle, GL_TEXTURE_2D, src_tex.handle, GL_DEPTH24_STENCIL8, 0, 1,
+                          0, 1);
             glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
             glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
         } else if (src_rect.top > temp_rect.top || src_rect.right > temp_rect.right) {
@@ -322,7 +202,7 @@ void main() {
         }
 
         state.texture_units[1].texture_2d = temp_tex.handle;
-        state.draw.draw_framebuffer = draw_fb_handle;
+        state.draw.draw_framebuffer = draw_fbo.handle;
         state.draw.shader_program = program.handle;
         state.draw.vertex_array = vao.handle;
         state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
@@ -332,14 +212,14 @@ void main() {
 
         glActiveTexture(GL_TEXTURE1);
         if (!use_texture_view) {
-            glCopyImageSubData(src_tex, GL_TEXTURE_2D, 0, src_rect.left, src_rect.bottom, 0,
+            glCopyImageSubData(src_tex.handle, GL_TEXTURE_2D, 0, src_rect.left, src_rect.bottom, 0,
                                temp_tex.handle, GL_TEXTURE_2D, 0, src_rect.left, src_rect.bottom, 0,
                                src_rect.GetWidth(), src_rect.GetHeight(), 1);
         }
         glTexParameteri(GL_TEXTURE_2D, GL_DEPTH_STENCIL_TEXTURE_MODE, GL_STENCIL_INDEX);
 
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex,
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
-                               0);
+                               dst_tex.handle, 0);
         glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
                                0);
 
@@ -365,32 +245,21 @@ private:
 FormatReinterpreterOpenGL::FormatReinterpreterOpenGL() {
     const std::string_view vendor{reinterpret_cast<const char*>(glGetString(GL_VENDOR))};
     const std::string_view version{reinterpret_cast<const char*>(glGetString(GL_VERSION))};
-    // Fallback to PBO path on obsolete intel drivers
-    // intel`s GL_VERSION string - `3.3.0 - Build 25.20.100.6373`
-    const bool intel_broken_drivers =
-        vendor.find("Intel") != vendor.npos && (std::atoi(version.substr(14, 2).data()) < 30);
 
-    if ((!intel_broken_drivers && GLAD_GL_ARB_stencil_texturing && GLAD_GL_ARB_texture_storage &&
+    auto Register = [this](PixelFormat dest, std::unique_ptr<FormatReinterpreterBase>&& obj) {
-         GLAD_GL_ARB_copy_image) ||
+        const u32 dst_index = static_cast<u32>(dest);
-        GLES) {
+        return reinterpreters[dst_index].push_back(std::move(obj));
-        reinterpreters.emplace(PixelFormatPair{PixelFormat::RGBA8, PixelFormat::D24S8},
+    };
-                               std::make_unique<ShaderD24S8toRGBA8>());
+
-        LOG_INFO(Render_OpenGL, "Using shader for D24S8 to RGBA8 reinterpretation");
+    Register(PixelFormat::RGBA8, std::make_unique<ShaderD24S8toRGBA8>());
-    } else {
+    LOG_INFO(Render_OpenGL, "Using shader for D24S8 to RGBA8 reinterpretation");
-        reinterpreters.emplace(PixelFormatPair{PixelFormat::RGBA8, PixelFormat::D24S8},
+
-                               std::make_unique<PixelBufferD24S8toABGR>());
+    Register(PixelFormat::RGB5A1, std::make_unique<RGBA4toRGB5A1>());
-        LOG_INFO(Render_OpenGL, "Using pbo for D24S8 to RGBA8 reinterpretation");
-    }
-    reinterpreters.emplace(PixelFormatPair{PixelFormat::RGB5A1, PixelFormat::RGBA4},
-                           std::make_unique<RGBA4toRGB5A1>());
 }
 
-FormatReinterpreterOpenGL::~FormatReinterpreterOpenGL() = default;
+auto FormatReinterpreterOpenGL::GetPossibleReinterpretations(PixelFormat dst_format)
-
+    -> const ReinterpreterList& {
-std::pair<FormatReinterpreterOpenGL::ReinterpreterMap::iterator,
+    return reinterpreters[static_cast<u32>(dst_format)];
-          FormatReinterpreterOpenGL::ReinterpreterMap::iterator>
-FormatReinterpreterOpenGL::GetPossibleReinterpretations(PixelFormat dst_format) {
-    return reinterpreters.equal_range(dst_format);
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_format_reinterpreter.h

@@ -1,62 +1,47 @@
-// Copyright 2020 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
 
-#include <map>
+#include <unordered_map>
-#include <type_traits>
-#include <glad/glad.h>
-#include "common/common_types.h"
 #include "common/math_util.h"
+#include "video_core/rasterizer_cache/pixel_format.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_surface_params.h"
 
 namespace OpenGL {
 
 class RasterizerCacheOpenGL;
 
-struct PixelFormatPair {
-    const SurfaceParams::PixelFormat dst_format, src_format;
-    struct less {
-        using is_transparent = void;
-        constexpr bool operator()(OpenGL::PixelFormatPair lhs, OpenGL::PixelFormatPair rhs) const {
-            return std::tie(lhs.dst_format, lhs.src_format) <
-                   std::tie(rhs.dst_format, rhs.src_format);
-        }
-        constexpr bool operator()(OpenGL::SurfaceParams::PixelFormat lhs,
-                                  OpenGL::PixelFormatPair rhs) const {
-            return lhs < rhs.dst_format;
-        }
-        constexpr bool operator()(OpenGL::PixelFormatPair lhs,
-                                  OpenGL::SurfaceParams::PixelFormat rhs) const {
-            return lhs.dst_format < rhs;
-        }
-    };
-};
-
 class FormatReinterpreterBase {
 public:
+    FormatReinterpreterBase() {
+        read_fbo.Create();
+        draw_fbo.Create();
+    }
+
     virtual ~FormatReinterpreterBase() = default;
 
-    virtual void Reinterpret(GLuint src_tex, const Common::Rectangle<u32>& src_rect,
+    virtual PixelFormat GetSourceFormat() const = 0;
-                             GLuint read_fb_handle, GLuint dst_tex,
+    virtual void Reinterpret(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                             const Common::Rectangle<u32>& dst_rect, GLuint draw_fb_handle) = 0;
+                             const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) = 0;
+
+protected:
+    OGLFramebuffer read_fbo;
+    OGLFramebuffer draw_fbo;
 };
 
+using ReinterpreterList = std::vector<std::unique_ptr<FormatReinterpreterBase>>;
+
 class FormatReinterpreterOpenGL : NonCopyable {
-    using ReinterpreterMap =
-        std::map<PixelFormatPair, std::unique_ptr<FormatReinterpreterBase>, PixelFormatPair::less>;
-
 public:
-    explicit FormatReinterpreterOpenGL();
+    FormatReinterpreterOpenGL();
-    ~FormatReinterpreterOpenGL();
+    ~FormatReinterpreterOpenGL() = default;
 
-    std::pair<ReinterpreterMap::iterator, ReinterpreterMap::iterator> GetPossibleReinterpretations(
+    const ReinterpreterList& GetPossibleReinterpretations(PixelFormat dst_format);
-        SurfaceParams::PixelFormat dst_format);
 
 private:
-    ReinterpreterMap reinterpreters;
+    std::array<ReinterpreterList, PIXEL_FORMAT_COUNT> reinterpreters;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -1,25 +1,15 @@
-// Copyright 2015 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <algorithm>
-#include <memory>
-#include <string>
-#include <tuple>
-#include <utility>
-#include <glad/glad.h>
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/math_util.h"
 #include "common/microprofile.h"
-#include "common/scope_exit.h"
-#include "common/vector_math.h"
-#include "core/hw/gpu.h"
 #include "video_core/pica_state.h"
 #include "video_core/regs_framebuffer.h"
 #include "video_core/regs_rasterizer.h"
-#include "video_core/regs_texturing.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
 #include "video_core/renderer_opengl/gl_vars.h"
@@ -29,9 +19,6 @@
 
 namespace OpenGL {
 
-using PixelFormat = SurfaceParams::PixelFormat;
-using SurfaceType = SurfaceParams::SurfaceType;
-
 MICROPROFILE_DEFINE(OpenGL_VAO, "OpenGL", "Vertex Array Setup", MP_RGB(255, 128, 0));
 MICROPROFILE_DEFINE(OpenGL_VS, "OpenGL", "Vertex Shader Setup", MP_RGB(192, 128, 128));
 MICROPROFILE_DEFINE(OpenGL_GS, "OpenGL", "Geometry Shader Setup", MP_RGB(128, 192, 128));
@@ -39,33 +26,19 @@ MICROPROFILE_DEFINE(OpenGL_Drawing, "OpenGL", "Drawing", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255));
 MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100));
 
-static bool IsVendorAmd() {
+#ifdef __APPLE__
-    const std::string_view gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
-    return gpu_vendor == "ATI Technologies Inc." || gpu_vendor == "Advanced Micro Devices, Inc.";
-}
 static bool IsVendorIntel() {
     std::string gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
     return gpu_vendor == "Intel Inc.";
 }
+#endif
 
 RasterizerOpenGL::RasterizerOpenGL(Frontend::EmuWindow& emu_window)
-    : is_amd(IsVendorAmd()), vertex_buffer(GL_ARRAY_BUFFER, VERTEX_BUFFER_SIZE, is_amd),
+    : vertex_buffer(GL_ARRAY_BUFFER, VERTEX_BUFFER_SIZE),
-      uniform_buffer(GL_UNIFORM_BUFFER, UNIFORM_BUFFER_SIZE, false),
+      uniform_buffer(GL_UNIFORM_BUFFER, UNIFORM_BUFFER_SIZE),
-      index_buffer(GL_ELEMENT_ARRAY_BUFFER, INDEX_BUFFER_SIZE, false),
+      index_buffer(GL_ELEMENT_ARRAY_BUFFER, INDEX_BUFFER_SIZE),
-      texture_buffer(GL_TEXTURE_BUFFER, TEXTURE_BUFFER_SIZE, false),
+      texture_buffer(GL_TEXTURE_BUFFER, TEXTURE_BUFFER_SIZE),
-      texture_lf_buffer(GL_TEXTURE_BUFFER, TEXTURE_BUFFER_SIZE, false) {
+      texture_lf_buffer(GL_TEXTURE_BUFFER, TEXTURE_BUFFER_SIZE) {
-
-    allow_shadow = GLES || (GLAD_GL_ARB_shader_image_load_store && GLAD_GL_ARB_shader_image_size &&
-                            GLAD_GL_ARB_framebuffer_no_attachments);
-    if (!allow_shadow) {
-        LOG_WARNING(Render_OpenGL,
-                    "Shadow might not be able to render because of unsupported OpenGL extensions.");
-    }
-
-    if (!GLAD_GL_ARB_copy_image && !GLES) {
-        LOG_WARNING(Render_OpenGL,
-                    "ARB_copy_image not supported. Some games might produce artifacts.");
-    }
 
     // Clipping plane 0 is always enabled for PICA fixed clip plane z <= 0
     state.clip_distance[0] = true;
@@ -171,22 +144,19 @@ RasterizerOpenGL::RasterizerOpenGL(Frontend::EmuWindow& emu_window)
 
 #ifdef __APPLE__
     if (IsVendorIntel()) {
-        shader_program_manager = std::make_unique<ShaderProgramManager>(
+        shader_program_manager = std::make_unique<ShaderProgramManager>(emu_window, false);
-            emu_window,
-            VideoCore::g_separable_shader_enabled ? GLAD_GL_ARB_separate_shader_objects : false,
-            is_amd);
     } else {
-        shader_program_manager = std::make_unique<ShaderProgramManager>(
+        shader_program_manager = std::make_unique<ShaderProgramManager>(emu_window, false);
-            emu_window, GLAD_GL_ARB_separate_shader_objects, is_amd);
     }
 #else
-    shader_program_manager = std::make_unique<ShaderProgramManager>(
+    shader_program_manager = std::make_unique<ShaderProgramManager>(emu_window, !GLES);
-        emu_window, GLAD_GL_ARB_separate_shader_objects, is_amd);
 #endif
 
     glEnable(GL_BLEND);
 
-    SyncEntireState();
+    // Explicitly call the derived version to avoid warnings about calling virtual
+    // methods in the constructor
+    RasterizerOpenGL::SyncEntireState();
 }
 
 RasterizerOpenGL::~RasterizerOpenGL() = default;
@@ -577,9 +547,10 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
     state.Apply();
 
     if (shadow_rendering) {
-        if (!allow_shadow || color_surface == nullptr) {
+        if (color_surface == nullptr) {
             return true;
         }
+
         glFramebufferParameteri(GL_DRAW_FRAMEBUFFER, GL_FRAMEBUFFER_DEFAULT_WIDTH,
                                 color_surface->width * color_surface->res_scale);
         glFramebufferParameteri(GL_DRAW_FRAMEBUFFER, GL_FRAMEBUFFER_DEFAULT_HEIGHT,
@@ -656,6 +627,18 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
         }
     };
 
+    const auto BindCubeFace = [&](GLuint& target, Pica::TexturingRegs::CubeFace face,
+                                  Pica::Texture::TextureInfo& info) {
+        info.physical_address = regs.texturing.GetCubePhysicalAddress(face);
+        Surface surface = res_cache.GetTextureSurface(info);
+
+        if (surface != nullptr) {
+            CheckBarrier(target = surface->texture.handle);
+        } else {
+            target = 0;
+        }
+    };
+
     // Sync and bind the texture surfaces
     const auto pica_textures = regs.texturing.GetTextures();
     for (unsigned texture_index = 0; texture_index < pica_textures.size(); ++texture_index) {
@@ -666,9 +649,6 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
                 using TextureType = Pica::TexturingRegs::TextureConfig::TextureType;
                 switch (texture.config.type.Value()) {
                 case TextureType::Shadow2D: {
-                    if (!allow_shadow)
-                        continue;
-
                     Surface surface = res_cache.GetTextureSurface(texture);
                     if (surface != nullptr) {
                         CheckBarrier(state.image_shadow_texture_px = surface->texture.handle);
@@ -678,67 +658,15 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
                     continue;
                 }
                 case TextureType::ShadowCube: {
-                    if (!allow_shadow)
-                        continue;
-                    Pica::Texture::TextureInfo info = Pica::Texture::TextureInfo::FromPicaRegister(
-                        texture.config, texture.format);
-                    Surface surface;
-
                     using CubeFace = Pica::TexturingRegs::CubeFace;
-                    info.physical_address =
+                    auto info = Pica::Texture::TextureInfo::FromPicaRegister(texture.config,
-                        regs.texturing.GetCubePhysicalAddress(CubeFace::PositiveX);
+                                                                             texture.format);
-                    surface = res_cache.GetTextureSurface(info);
+                    BindCubeFace(state.image_shadow_texture_px, CubeFace::PositiveX, info);
-                    if (surface != nullptr) {
+                    BindCubeFace(state.image_shadow_texture_nx, CubeFace::NegativeX, info);
-                        CheckBarrier(state.image_shadow_texture_px = surface->texture.handle);
+                    BindCubeFace(state.image_shadow_texture_py, CubeFace::PositiveY, info);
-                    } else {
+                    BindCubeFace(state.image_shadow_texture_ny, CubeFace::NegativeY, info);
-                        state.image_shadow_texture_px = 0;
+                    BindCubeFace(state.image_shadow_texture_pz, CubeFace::PositiveZ, info);
-                    }
+                    BindCubeFace(state.image_shadow_texture_nz, CubeFace::NegativeZ, info);
-
-                    info.physical_address =
-                        regs.texturing.GetCubePhysicalAddress(CubeFace::NegativeX);
-                    surface = res_cache.GetTextureSurface(info);
-                    if (surface != nullptr) {
-                        CheckBarrier(state.image_shadow_texture_nx = surface->texture.handle);
-                    } else {
-                        state.image_shadow_texture_nx = 0;
-                    }
-
-                    info.physical_address =
-                        regs.texturing.GetCubePhysicalAddress(CubeFace::PositiveY);
-                    surface = res_cache.GetTextureSurface(info);
-                    if (surface != nullptr) {
-                        CheckBarrier(state.image_shadow_texture_py = surface->texture.handle);
-                    } else {
-                        state.image_shadow_texture_py = 0;
-                    }
-
-                    info.physical_address =
-                        regs.texturing.GetCubePhysicalAddress(CubeFace::NegativeY);
-                    surface = res_cache.GetTextureSurface(info);
-                    if (surface != nullptr) {
-                        CheckBarrier(state.image_shadow_texture_ny = surface->texture.handle);
-                    } else {
-                        state.image_shadow_texture_ny = 0;
-                    }
-
-                    info.physical_address =
-                        regs.texturing.GetCubePhysicalAddress(CubeFace::PositiveZ);
-                    surface = res_cache.GetTextureSurface(info);
-                    if (surface != nullptr) {
-                        CheckBarrier(state.image_shadow_texture_pz = surface->texture.handle);
-                    } else {
-                        state.image_shadow_texture_pz = 0;
-                    }
-
-                    info.physical_address =
-                        regs.texturing.GetCubePhysicalAddress(CubeFace::NegativeZ);
-                    surface = res_cache.GetTextureSurface(info);
-                    if (surface != nullptr) {
-                        CheckBarrier(state.image_shadow_texture_nz = surface->texture.handle);
-                    } else {
-                        state.image_shadow_texture_nz = 0;
-                    }
-
                     continue;
                 }
                 case TextureType::TextureCube:
@@ -758,7 +686,10 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
                     texture_cube_sampler.SyncWithConfig(texture.config);
                     state.texture_units[texture_index].texture_2d = 0;
                     continue; // Texture unit 0 setup finished. Continue to next unit
+                default:
+                    break;
                 }
+
                 state.texture_cube_unit.texture_cube = 0;
             }
 
@@ -783,27 +714,19 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
     }
 
     OGLTexture temp_tex;
-    if (need_duplicate_texture && (GLAD_GL_ARB_copy_image || GLES)) {
+    if (need_duplicate_texture) {
+        const auto& tuple = GetFormatTuple(color_surface->pixel_format);
+        const GLsizei levels = color_surface->max_level + 1;
+
         // The game is trying to use a surface as a texture and framebuffer at the same time
         // which causes unpredictable behavior on the host.
         // Making a copy to sample from eliminates this issue and seems to be fairly cheap.
         temp_tex.Create();
-        glBindTexture(GL_TEXTURE_2D, temp_tex.handle);
+        temp_tex.Allocate(GL_TEXTURE_2D, levels, tuple.internal_format,
-        auto [internal_format, format, type] = GetFormatTuple(color_surface->pixel_format);
+                          color_surface->GetScaledWidth(), color_surface->GetScaledHeight());
-        OGLTexture::Allocate(GL_TEXTURE_2D, color_surface->max_level + 1, internal_format, format,
-                             type, color_surface->GetScaledWidth(),
-                             color_surface->GetScaledHeight());
-        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
-        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-        glBindTexture(GL_TEXTURE_2D, state.texture_units[0].texture_2d);
 
-        for (u32 level{0}; level <= color_surface->max_level; ++level) {
+        temp_tex.CopyFrom(color_surface->texture, GL_TEXTURE_2D, levels,
-            glCopyImageSubData(color_surface->texture.handle, GL_TEXTURE_2D, level, 0, 0, 0,
+                          color_surface->GetScaledWidth(), color_surface->GetScaledHeight());
-                               temp_tex.handle, GL_TEXTURE_2D, level, 0, 0, 0,
-                               color_surface->GetScaledWidth() >> level,
-                               color_surface->GetScaledHeight() >> level, 1);
-        }
 
         for (auto& unit : state.texture_units) {
             if (unit.texture_2d == color_surface->texture.handle) {
@@ -878,15 +801,13 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
         state.texture_units[texture_index].texture_2d = 0;
     }
     state.texture_cube_unit.texture_cube = 0;
-    if (allow_shadow) {
+    state.image_shadow_texture_px = 0;
-        state.image_shadow_texture_px = 0;
+    state.image_shadow_texture_nx = 0;
-        state.image_shadow_texture_nx = 0;
+    state.image_shadow_texture_py = 0;
-        state.image_shadow_texture_py = 0;
+    state.image_shadow_texture_ny = 0;
-        state.image_shadow_texture_ny = 0;
+    state.image_shadow_texture_pz = 0;
-        state.image_shadow_texture_pz = 0;
+    state.image_shadow_texture_nz = 0;
-        state.image_shadow_texture_nz = 0;
+    state.image_shadow_buffer = 0;
-        state.image_shadow_buffer = 0;
-    }
     state.Apply();
 
     if (shadow_rendering) {
@@ -1445,7 +1366,7 @@ bool RasterizerOpenGL::AccelerateDisplayTransfer(const GPU::Regs::DisplayTransfe
     src_params.stride = config.input_width;
     src_params.height = config.output_height;
     src_params.is_tiled = !config.input_linear;
-    src_params.pixel_format = SurfaceParams::PixelFormatFromGPUPixelFormat(config.input_format);
+    src_params.pixel_format = PixelFormatFromGPUPixelFormat(config.input_format);
     src_params.UpdateParams();
 
     SurfaceParams dst_params;
@@ -1455,7 +1376,7 @@ bool RasterizerOpenGL::AccelerateDisplayTransfer(const GPU::Regs::DisplayTransfe
     dst_params.height = config.scaling == config.ScaleXY ? config.output_height.Value() / 2
                                                          : config.output_height.Value();
     dst_params.is_tiled = config.input_linear != config.dont_swizzle;
-    dst_params.pixel_format = SurfaceParams::PixelFormatFromGPUPixelFormat(config.output_format);
+    dst_params.pixel_format = PixelFormatFromGPUPixelFormat(config.output_format);
     dst_params.UpdateParams();
 
     Common::Rectangle<u32> src_rect;
@@ -1595,7 +1516,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const GPU::Regs::FramebufferConfig& con
     src_params.height = config.height;
     src_params.stride = pixel_stride;
     src_params.is_tiled = false;
-    src_params.pixel_format = SurfaceParams::PixelFormatFromGPUPixelFormat(config.color_format);
+    src_params.pixel_format = PixelFormatFromGPUPixelFormat(config.color_format);
     src_params.UpdateParams();
 
     Common::Rectangle<u32> src_rect;
@@ -1674,7 +1595,7 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(
         if (border_color != config.border_color.raw) {
             border_color = config.border_color.raw;
             auto gl_color = PicaToGL::ColorRGBA8(border_color);
-            glSamplerParameterfv(s, GL_TEXTURE_BORDER_COLOR, gl_color.data());
+            glSamplerParameterfv(s, GL_TEXTURE_BORDER_COLOR, gl_color.AsArray());
         }
     }
 
@@ -1704,8 +1625,8 @@ void RasterizerOpenGL::SyncClipEnabled() {
 
 void RasterizerOpenGL::SyncClipCoef() {
     const auto raw_clip_coef = Pica::g_state.regs.rasterizer.GetClipCoef();
-    const GLvec4 new_clip_coef = {raw_clip_coef.x.ToFloat32(), raw_clip_coef.y.ToFloat32(),
+    const Common::Vec4f new_clip_coef = {raw_clip_coef.x.ToFloat32(), raw_clip_coef.y.ToFloat32(),
-                                  raw_clip_coef.z.ToFloat32(), raw_clip_coef.w.ToFloat32()};
+                                         raw_clip_coef.z.ToFloat32(), raw_clip_coef.w.ToFloat32()};
     if (new_clip_coef != uniform_block_data.data.clip_coef) {
         uniform_block_data.data.clip_coef = new_clip_coef;
         uniform_block_data.dirty = true;
@@ -1975,10 +1896,10 @@ void RasterizerOpenGL::SyncLightAmbient(int light_index) {
 }
 
 void RasterizerOpenGL::SyncLightPosition(int light_index) {
-    GLvec3 position = {
+    const auto& light = Pica::g_state.regs.lighting.light[light_index];
-        Pica::float16::FromRaw(Pica::g_state.regs.lighting.light[light_index].x).ToFloat32(),
+    const Common::Vec3f position = {Pica::float16::FromRaw(light.x).ToFloat32(),
-        Pica::float16::FromRaw(Pica::g_state.regs.lighting.light[light_index].y).ToFloat32(),
+                                    Pica::float16::FromRaw(light.y).ToFloat32(),
-        Pica::float16::FromRaw(Pica::g_state.regs.lighting.light[light_index].z).ToFloat32()};
+                                    Pica::float16::FromRaw(light.z).ToFloat32()};
 
     if (position != uniform_block_data.data.light_src[light_index].position) {
         uniform_block_data.data.light_src[light_index].position = position;
@@ -1988,8 +1909,7 @@ void RasterizerOpenGL::SyncLightPosition(int light_index) {
 
 void RasterizerOpenGL::SyncLightSpotDirection(int light_index) {
     const auto& light = Pica::g_state.regs.lighting.light[light_index];
-    GLvec3 spot_direction = {light.spot_x / 2047.0f, light.spot_y / 2047.0f,
+    const auto spot_direction = Common::Vec3u{light.spot_x, light.spot_y, light.spot_z} / 2047.0f;
-                             light.spot_z / 2047.0f};
 
     if (spot_direction != uniform_block_data.data.light_src[light_index].spot_direction) {
         uniform_block_data.data.light_src[light_index].spot_direction = spot_direction;
@@ -1998,9 +1918,8 @@ void RasterizerOpenGL::SyncLightSpotDirection(int light_index) {
 }
 
 void RasterizerOpenGL::SyncLightDistanceAttenuationBias(int light_index) {
-    GLfloat dist_atten_bias =
+    const auto& light = Pica::g_state.regs.lighting.light[light_index];
-        Pica::float20::FromRaw(Pica::g_state.regs.lighting.light[light_index].dist_atten_bias)
+    float dist_atten_bias = Pica::float20::FromRaw(light.dist_atten_bias).ToFloat32();
-            .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;
@@ -2009,9 +1928,8 @@ void RasterizerOpenGL::SyncLightDistanceAttenuationBias(int light_index) {
 }
 
 void RasterizerOpenGL::SyncLightDistanceAttenuationScale(int light_index) {
-    GLfloat dist_atten_scale =
+    const auto& light = Pica::g_state.regs.lighting.light[light_index];
-        Pica::float20::FromRaw(Pica::g_state.regs.lighting.light[light_index].dist_atten_scale)
+    float dist_atten_scale = Pica::float20::FromRaw(light.dist_atten_scale).ToFloat32();
-            .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;
@@ -2042,7 +1960,8 @@ void RasterizerOpenGL::SyncShadowTextureBias() {
 
 void RasterizerOpenGL::SyncAndUploadLUTsLF() {
     constexpr std::size_t max_size =
-        sizeof(GLvec2) * 256 * Pica::LightingRegs::NumLightingSampler + sizeof(GLvec2) * 128; // fog
+        sizeof(Common::Vec2f) * 256 * Pica::LightingRegs::NumLightingSampler +
+        sizeof(Common::Vec2f) * 128; // fog
 
     if (!uniform_block_data.lighting_lut_dirty_any && !uniform_block_data.fog_lut_dirty) {
         return;
@@ -2053,27 +1972,27 @@ void RasterizerOpenGL::SyncAndUploadLUTsLF() {
     bool invalidate;
     std::size_t bytes_used = 0;
     glBindBuffer(GL_TEXTURE_BUFFER, texture_lf_buffer.GetHandle());
-    std::tie(buffer, offset, invalidate) = texture_lf_buffer.Map(max_size, sizeof(GLvec4));
+    std::tie(buffer, offset, invalidate) = texture_lf_buffer.Map(max_size, sizeof(Common::Vec4f));
 
     // Sync the lighting luts
     if (uniform_block_data.lighting_lut_dirty_any || invalidate) {
         for (unsigned index = 0; index < uniform_block_data.lighting_lut_dirty.size(); index++) {
             if (uniform_block_data.lighting_lut_dirty[index] || invalidate) {
-                std::array<GLvec2, 256> new_data;
+                std::array<Common::Vec2f, 256> new_data;
                 const auto& source_lut = Pica::g_state.lighting.luts[index];
                 std::transform(source_lut.begin(), source_lut.end(), new_data.begin(),
                                [](const auto& entry) {
-                                   return GLvec2{entry.ToFloat(), entry.DiffToFloat()};
+                                   return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()};
                                });
 
                 if (new_data != lighting_lut_data[index] || invalidate) {
                     lighting_lut_data[index] = new_data;
                     std::memcpy(buffer + bytes_used, new_data.data(),
-                                new_data.size() * sizeof(GLvec2));
+                                new_data.size() * sizeof(Common::Vec2f));
                     uniform_block_data.data.lighting_lut_offset[index / 4][index % 4] =
-                        static_cast<GLint>((offset + bytes_used) / sizeof(GLvec2));
+                        static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec2f));
                     uniform_block_data.dirty = true;
-                    bytes_used += new_data.size() * sizeof(GLvec2);
+                    bytes_used += new_data.size() * sizeof(Common::Vec2f);
                 }
                 uniform_block_data.lighting_lut_dirty[index] = false;
             }
@@ -2083,20 +2002,21 @@ void RasterizerOpenGL::SyncAndUploadLUTsLF() {
 
     // Sync the fog lut
     if (uniform_block_data.fog_lut_dirty || invalidate) {
-        std::array<GLvec2, 128> new_data;
+        std::array<Common::Vec2f, 128> new_data;
 
         std::transform(Pica::g_state.fog.lut.begin(), Pica::g_state.fog.lut.end(), new_data.begin(),
                        [](const auto& entry) {
-                           return GLvec2{entry.ToFloat(), entry.DiffToFloat()};
+                           return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()};
                        });
 
         if (new_data != fog_lut_data || invalidate) {
             fog_lut_data = new_data;
-            std::memcpy(buffer + bytes_used, new_data.data(), new_data.size() * sizeof(GLvec2));
+            std::memcpy(buffer + bytes_used, new_data.data(),
+                        new_data.size() * sizeof(Common::Vec2f));
             uniform_block_data.data.fog_lut_offset =
-                static_cast<GLint>((offset + bytes_used) / sizeof(GLvec2));
+                static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
             uniform_block_data.dirty = true;
-            bytes_used += new_data.size() * sizeof(GLvec2);
+            bytes_used += new_data.size() * sizeof(Common::Vec2f);
         }
         uniform_block_data.fog_lut_dirty = false;
     }
@@ -2105,9 +2025,10 @@ void RasterizerOpenGL::SyncAndUploadLUTsLF() {
 }
 
 void RasterizerOpenGL::SyncAndUploadLUTs() {
-    constexpr std::size_t max_size = sizeof(GLvec2) * 128 * 3 + // proctex: noise + color + alpha
+    constexpr std::size_t max_size =
-                                     sizeof(GLvec4) * 256 +     // proctex
+        sizeof(Common::Vec2f) * 128 * 3 + // proctex: noise + color + alpha
-                                     sizeof(GLvec4) * 256;      // proctex diff
+        sizeof(Common::Vec4f) * 256 +     // proctex
+        sizeof(Common::Vec4f) * 256;      // proctex diff
 
     if (!uniform_block_data.proctex_noise_lut_dirty &&
         !uniform_block_data.proctex_color_map_dirty &&
@@ -2121,23 +2042,24 @@ void RasterizerOpenGL::SyncAndUploadLUTs() {
     bool invalidate;
     std::size_t bytes_used = 0;
     glBindBuffer(GL_TEXTURE_BUFFER, texture_buffer.GetHandle());
-    std::tie(buffer, offset, invalidate) = texture_buffer.Map(max_size, sizeof(GLvec4));
+    std::tie(buffer, offset, invalidate) = texture_buffer.Map(max_size, sizeof(Common::Vec4f));
 
     // helper function for SyncProcTexNoiseLUT/ColorMap/AlphaMap
     auto SyncProcTexValueLUT = [this, buffer, offset, invalidate, &bytes_used](
                                    const std::array<Pica::State::ProcTex::ValueEntry, 128>& lut,
-                                   std::array<GLvec2, 128>& lut_data, GLint& lut_offset) {
+                                   std::array<Common::Vec2f, 128>& lut_data, GLint& lut_offset) {
-        std::array<GLvec2, 128> new_data;
+        std::array<Common::Vec2f, 128> new_data;
         std::transform(lut.begin(), lut.end(), new_data.begin(), [](const auto& entry) {
-            return GLvec2{entry.ToFloat(), entry.DiffToFloat()};
+            return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()};
         });
 
         if (new_data != lut_data || invalidate) {
             lut_data = new_data;
-            std::memcpy(buffer + bytes_used, new_data.data(), new_data.size() * sizeof(GLvec2));
+            std::memcpy(buffer + bytes_used, new_data.data(),
-            lut_offset = static_cast<GLint>((offset + bytes_used) / sizeof(GLvec2));
+                        new_data.size() * sizeof(Common::Vec2f));
+            lut_offset = static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec2f));
             uniform_block_data.dirty = true;
-            bytes_used += new_data.size() * sizeof(GLvec2);
+            bytes_used += new_data.size() * sizeof(Common::Vec2f);
         }
     };
 
@@ -2164,44 +2086,46 @@ void RasterizerOpenGL::SyncAndUploadLUTs() {
 
     // Sync the proctex lut
     if (uniform_block_data.proctex_lut_dirty || invalidate) {
-        std::array<GLvec4, 256> new_data;
+        std::array<Common::Vec4f, 256> new_data;
 
         std::transform(Pica::g_state.proctex.color_table.begin(),
                        Pica::g_state.proctex.color_table.end(), new_data.begin(),
                        [](const auto& entry) {
                            auto rgba = entry.ToVector() / 255.0f;
-                           return GLvec4{rgba.r(), rgba.g(), rgba.b(), rgba.a()};
+                           return Common::Vec4f{rgba.r(), rgba.g(), rgba.b(), rgba.a()};
                        });
 
         if (new_data != proctex_lut_data || invalidate) {
             proctex_lut_data = new_data;
-            std::memcpy(buffer + bytes_used, new_data.data(), new_data.size() * sizeof(GLvec4));
+            std::memcpy(buffer + bytes_used, new_data.data(),
+                        new_data.size() * sizeof(Common::Vec4f));
             uniform_block_data.data.proctex_lut_offset =
-                static_cast<GLint>((offset + bytes_used) / sizeof(GLvec4));
+                static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec4f));
             uniform_block_data.dirty = true;
-            bytes_used += new_data.size() * sizeof(GLvec4);
+            bytes_used += new_data.size() * sizeof(Common::Vec4f);
         }
         uniform_block_data.proctex_lut_dirty = false;
     }
 
     // Sync the proctex difference lut
     if (uniform_block_data.proctex_diff_lut_dirty || invalidate) {
-        std::array<GLvec4, 256> new_data;
+        std::array<Common::Vec4f, 256> new_data;
 
         std::transform(Pica::g_state.proctex.color_diff_table.begin(),
                        Pica::g_state.proctex.color_diff_table.end(), new_data.begin(),
                        [](const auto& entry) {
                            auto rgba = entry.ToVector() / 255.0f;
-                           return GLvec4{rgba.r(), rgba.g(), rgba.b(), rgba.a()};
+                           return Common::Vec4f{rgba.r(), rgba.g(), rgba.b(), rgba.a()};
                        });
 
         if (new_data != proctex_diff_lut_data || invalidate) {
             proctex_diff_lut_data = new_data;
-            std::memcpy(buffer + bytes_used, new_data.data(), new_data.size() * sizeof(GLvec4));
+            std::memcpy(buffer + bytes_used, new_data.data(),
+                        new_data.size() * sizeof(Common::Vec4f));
             uniform_block_data.data.proctex_diff_lut_offset =
-                static_cast<GLint>((offset + bytes_used) / sizeof(GLvec4));
+                static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec4f));
             uniform_block_data.dirty = true;
-            bytes_used += new_data.size() * sizeof(GLvec4);
+            bytes_used += new_data.size() * sizeof(Common::Vec4f);
         }
         uniform_block_data.proctex_diff_lut_dirty = false;
     }

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -1,32 +1,18 @@
-// Copyright 2015 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
-
-#include <array>
-#include <cstddef>
-#include <cstring>
-#include <memory>
-#include <vector>
-#include <glad/glad.h>
-#include "common/bit_field.h"
-#include "common/common_types.h"
 #include "common/vector_math.h"
 #include "core/hw/gpu.h"
-#include "video_core/pica_state.h"
 #include "video_core/pica_types.h"
+#include "video_core/rasterizer_cache/rasterizer_cache.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/regs_framebuffer.h"
 #include "video_core/regs_lighting.h"
-#include "video_core/regs_rasterizer.h"
 #include "video_core/regs_texturing.h"
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
-#include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
-#include "video_core/renderer_opengl/pica_to_gl.h"
 #include "video_core/shader/shader.h"
 
 namespace Frontend {
@@ -118,20 +104,18 @@ private:
             view[2] = v.view.z.ToFloat32();
 
             if (flip_quaternion) {
-                for (float& x : normquat) {
+                normquat = -normquat;
-                    x = -x;
-                }
             }
         }
 
-        GLvec4 position;
+        Common::Vec4f position;
-        GLvec4 color;
+        Common::Vec4f color;
-        GLvec2 tex_coord0;
+        Common::Vec2f tex_coord0;
-        GLvec2 tex_coord1;
+        Common::Vec2f tex_coord1;
-        GLvec2 tex_coord2;
+        Common::Vec2f tex_coord2;
-        GLfloat tex_coord0_w;
+        float tex_coord0_w;
-        GLvec4 normquat;
+        Common::Vec4f normquat;
-        GLvec3 view;
+        Common::Vec3f view;
     };
 
     /// Syncs the clip enabled status to match the PICA register
@@ -263,8 +247,6 @@ private:
     /// Setup geometry shader for AccelerateDrawBatch
     bool SetupGeometryShader();
 
-    bool is_amd;
-
     OpenGLState state;
     GLuint default_texture;
 
@@ -316,15 +298,14 @@ private:
     OGLTexture texture_buffer_lut_rg;
     OGLTexture texture_buffer_lut_rgba;
 
-    std::array<std::array<GLvec2, 256>, Pica::LightingRegs::NumLightingSampler> lighting_lut_data{};
+    std::array<std::array<Common::Vec2f, 256>, Pica::LightingRegs::NumLightingSampler>
-    std::array<GLvec2, 128> fog_lut_data{};
+        lighting_lut_data{};
-    std::array<GLvec2, 128> proctex_noise_lut_data{};
+    std::array<Common::Vec2f, 128> fog_lut_data{};
-    std::array<GLvec2, 128> proctex_color_map_data{};
+    std::array<Common::Vec2f, 128> proctex_noise_lut_data{};
-    std::array<GLvec2, 128> proctex_alpha_map_data{};
+    std::array<Common::Vec2f, 128> proctex_color_map_data{};
-    std::array<GLvec4, 256> proctex_lut_data{};
+    std::array<Common::Vec2f, 128> proctex_alpha_map_data{};
-    std::array<GLvec4, 256> proctex_diff_lut_data{};
+    std::array<Common::Vec4f, 256> proctex_lut_data{};
-
+    std::array<Common::Vec4f, 256> proctex_diff_lut_data{};
-    bool allow_shadow;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -1,390 +0,0 @@
-// Copyright 2015 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-#include <list>
-#include <memory>
-#include <mutex>
-#include <set>
-#include <tuple>
-#ifdef __GNUC__
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wunused-local-typedefs"
-#endif
-#include <boost/icl/interval_map.hpp>
-#include <boost/icl/interval_set.hpp>
-#ifdef __GNUC__
-#pragma GCC diagnostic pop
-#endif
-#include <unordered_map>
-#include <boost/functional/hash.hpp>
-#include <glad/glad.h>
-#include "common/assert.h"
-#include "common/common_funcs.h"
-#include "common/common_types.h"
-#include "common/math_util.h"
-#include "core/custom_tex_cache.h"
-#include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_surface_params.h"
-#include "video_core/texture/texture_decode.h"
-
-namespace OpenGL {
-
-class RasterizerCacheOpenGL;
-class TextureFilterer;
-class FormatReinterpreterOpenGL;
-
-struct FormatTuple {
-    GLint internal_format;
-    GLenum format;
-    GLenum type;
-};
-
-constexpr FormatTuple tex_tuple = {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE};
-
-const FormatTuple& GetFormatTuple(SurfaceParams::PixelFormat pixel_format);
-
-struct HostTextureTag {
-    FormatTuple format_tuple;
-    u32 width;
-    u32 height;
-    bool operator==(const HostTextureTag& rhs) const noexcept {
-        return std::tie(format_tuple.format, format_tuple.internal_format, width, height) ==
-               std::tie(rhs.format_tuple.format, rhs.format_tuple.internal_format, rhs.width,
-                        rhs.height);
-    };
-};
-
-struct TextureCubeConfig {
-    PAddr px;
-    PAddr nx;
-    PAddr py;
-    PAddr ny;
-    PAddr pz;
-    PAddr nz;
-    u32 width;
-    Pica::TexturingRegs::TextureFormat format;
-
-    bool operator==(const TextureCubeConfig& rhs) const {
-        return std::tie(px, nx, py, ny, pz, nz, width, format) ==
-               std::tie(rhs.px, rhs.nx, rhs.py, rhs.ny, rhs.pz, rhs.nz, rhs.width, rhs.format);
-    }
-
-    bool operator!=(const TextureCubeConfig& rhs) const {
-        return !(*this == rhs);
-    }
-};
-
-} // namespace OpenGL
-
-namespace std {
-template <>
-struct hash<OpenGL::HostTextureTag> {
-    std::size_t operator()(const OpenGL::HostTextureTag& tag) const noexcept {
-        std::size_t hash = 0;
-        boost::hash_combine(hash, tag.format_tuple.format);
-        boost::hash_combine(hash, tag.format_tuple.internal_format);
-        boost::hash_combine(hash, tag.width);
-        boost::hash_combine(hash, tag.height);
-        return hash;
-    }
-};
-
-template <>
-struct hash<OpenGL::TextureCubeConfig> {
-    std::size_t operator()(const OpenGL::TextureCubeConfig& config) const noexcept {
-        std::size_t hash = 0;
-        boost::hash_combine(hash, config.px);
-        boost::hash_combine(hash, config.nx);
-        boost::hash_combine(hash, config.py);
-        boost::hash_combine(hash, config.ny);
-        boost::hash_combine(hash, config.pz);
-        boost::hash_combine(hash, config.nz);
-        boost::hash_combine(hash, config.width);
-        boost::hash_combine(hash, static_cast<u32>(config.format));
-        return hash;
-    }
-};
-} // namespace std
-
-namespace OpenGL {
-
-using SurfaceSet = std::set<Surface>;
-
-using SurfaceRegions = boost::icl::interval_set<PAddr, std::less, SurfaceInterval>;
-using SurfaceMap =
-    boost::icl::interval_map<PAddr, Surface, boost::icl::partial_absorber, std::less,
-                             boost::icl::inplace_plus, boost::icl::inter_section, SurfaceInterval>;
-using SurfaceCache =
-    boost::icl::interval_map<PAddr, SurfaceSet, boost::icl::partial_absorber, std::less,
-                             boost::icl::inplace_plus, boost::icl::inter_section, SurfaceInterval>;
-
-static_assert(std::is_same<SurfaceRegions::interval_type, SurfaceCache::interval_type>() &&
-                  std::is_same<SurfaceMap::interval_type, SurfaceCache::interval_type>(),
-              "incorrect interval types");
-
-using SurfaceRect_Tuple = std::tuple<Surface, Common::Rectangle<u32>>;
-using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, Common::Rectangle<u32>>;
-
-using PageMap = boost::icl::interval_map<u32, int>;
-
-enum class ScaleMatch {
-    Exact,   // only accept same res scale
-    Upscale, // only allow higher scale than params
-    Ignore   // accept every scaled res
-};
-
-/**
- * A watcher that notifies whether a cached surface has been changed. This is useful for caching
- * surface collection objects, including texture cube and mipmap.
- */
-struct SurfaceWatcher {
-public:
-    explicit SurfaceWatcher(std::weak_ptr<CachedSurface>&& surface) : surface(std::move(surface)) {}
-
-    /**
-     * Checks whether the surface has been changed.
-     * @return false if the surface content has been changed since last Validate() call or has been
-     * destroyed; otherwise true
-     */
-    bool IsValid() const {
-        return !surface.expired() && valid;
-    }
-
-    /// Marks that the content of the referencing surface has been updated to the watcher user.
-    void Validate() {
-        ASSERT(!surface.expired());
-        valid = true;
-    }
-
-    /// Gets the referencing surface. Returns null if the surface has been destroyed
-    Surface Get() const {
-        return surface.lock();
-    }
-
-private:
-    friend struct CachedSurface;
-    std::weak_ptr<CachedSurface> surface;
-    bool valid = false;
-};
-
-class RasterizerCacheOpenGL;
-
-struct CachedSurface : SurfaceParams, std::enable_shared_from_this<CachedSurface> {
-    CachedSurface(RasterizerCacheOpenGL& owner) : owner{owner} {}
-    ~CachedSurface();
-
-    bool CanFill(const SurfaceParams& dest_surface, SurfaceInterval fill_interval) const;
-    bool CanCopy(const SurfaceParams& dest_surface, SurfaceInterval copy_interval) const;
-
-    bool IsRegionValid(SurfaceInterval interval) const {
-        return (invalid_regions.find(interval) == invalid_regions.end());
-    }
-
-    bool IsSurfaceFullyInvalid() const {
-        auto interval = GetInterval();
-        return *invalid_regions.equal_range(interval).first == interval;
-    }
-
-    bool registered = false;
-    SurfaceRegions invalid_regions;
-
-    u32 fill_size = 0; /// Number of bytes to read from fill_data
-    std::array<u8, 4> fill_data;
-
-    OGLTexture texture;
-
-    /// max mipmap level that has been attached to the texture
-    u32 max_level = 0;
-    /// level_watchers[i] watches the (i+1)-th level mipmap source surface
-    std::array<std::shared_ptr<SurfaceWatcher>, 7> level_watchers;
-
-    bool is_custom = false;
-    Core::CustomTexInfo custom_tex_info;
-
-    static constexpr unsigned int GetGLBytesPerPixel(PixelFormat format) {
-        // OpenGL needs 4 bpp alignment for D24 since using GL_UNSIGNED_INT as type
-        return format == PixelFormat::Invalid
-                   ? 0
-                   : (format == PixelFormat::D24 || GetFormatType(format) == SurfaceType::Texture)
-                         ? 4
-                         : SurfaceParams::GetFormatBpp(format) / 8;
-    }
-
-    std::vector<u8> gl_buffer;
-
-    // Read/Write data in 3DS memory to/from gl_buffer
-    void LoadGLBuffer(PAddr load_start, PAddr load_end);
-    void FlushGLBuffer(PAddr flush_start, PAddr flush_end);
-
-    // Custom texture loading and dumping
-    bool LoadCustomTexture(u64 tex_hash);
-    void DumpTexture(GLuint target_tex, u64 tex_hash);
-
-    // Upload/Download data in gl_buffer in/to this surface's texture
-    void UploadGLTexture(Common::Rectangle<u32> rect, GLuint read_fb_handle, GLuint draw_fb_handle);
-    void DownloadGLTexture(const Common::Rectangle<u32>& rect, GLuint read_fb_handle,
-                           GLuint draw_fb_handle);
-
-    std::shared_ptr<SurfaceWatcher> CreateWatcher() {
-        auto watcher = std::make_shared<SurfaceWatcher>(weak_from_this());
-        watchers.push_front(watcher);
-        return watcher;
-    }
-
-    void InvalidateAllWatcher() {
-        for (const auto& watcher : watchers) {
-            if (auto locked = watcher.lock()) {
-                locked->valid = false;
-            }
-        }
-    }
-
-    void UnlinkAllWatcher() {
-        for (const auto& watcher : watchers) {
-            if (auto locked = watcher.lock()) {
-                locked->valid = false;
-                locked->surface.reset();
-            }
-        }
-        watchers.clear();
-    }
-
-private:
-    RasterizerCacheOpenGL& owner;
-    std::list<std::weak_ptr<SurfaceWatcher>> watchers;
-};
-
-struct CachedTextureCube {
-    OGLTexture texture;
-    u16 res_scale = 1;
-    std::shared_ptr<SurfaceWatcher> px;
-    std::shared_ptr<SurfaceWatcher> nx;
-    std::shared_ptr<SurfaceWatcher> py;
-    std::shared_ptr<SurfaceWatcher> ny;
-    std::shared_ptr<SurfaceWatcher> pz;
-    std::shared_ptr<SurfaceWatcher> nz;
-};
-
-static constexpr std::array<FormatTuple, 4> depth_format_tuples = {{
-    {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // D16
-    {},
-    {GL_DEPTH_COMPONENT24, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT},   // D24
-    {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // D24S8
-}};
-
-class TextureDownloaderES;
-
-class RasterizerCacheOpenGL : NonCopyable {
-public:
-    RasterizerCacheOpenGL();
-    ~RasterizerCacheOpenGL();
-
-    /// Blit one surface's texture to another
-    bool BlitSurfaces(const Surface& src_surface, const Common::Rectangle<u32>& src_rect,
-                      const Surface& dst_surface, const Common::Rectangle<u32>& dst_rect);
-
-    /// Copy one surface's region to another
-    void CopySurface(const Surface& src_surface, const Surface& dst_surface,
-                     SurfaceInterval copy_interval);
-
-    /// Load a texture from 3DS memory to OpenGL and cache it (if not already cached)
-    Surface GetSurface(const SurfaceParams& params, ScaleMatch match_res_scale,
-                       bool load_if_create);
-
-    /// Attempt to find a subrect (resolution scaled) of a surface, otherwise loads a texture from
-    /// 3DS memory to OpenGL and caches it (if not already cached)
-    SurfaceRect_Tuple GetSurfaceSubRect(const SurfaceParams& params, ScaleMatch match_res_scale,
-                                        bool load_if_create);
-
-    /// Get a surface based on the texture configuration
-    Surface GetTextureSurface(const Pica::TexturingRegs::FullTextureConfig& config);
-    Surface GetTextureSurface(const Pica::Texture::TextureInfo& info, u32 max_level = 0);
-
-    /// Get a texture cube based on the texture configuration
-    const CachedTextureCube& GetTextureCube(const TextureCubeConfig& config);
-
-    /// Get the color and depth surfaces based on the framebuffer configuration
-    SurfaceSurfaceRect_Tuple GetFramebufferSurfaces(bool using_color_fb, bool using_depth_fb,
-                                                    const Common::Rectangle<s32>& viewport_rect);
-
-    /// Get a surface that matches the fill config
-    Surface GetFillSurface(const GPU::Regs::MemoryFillConfig& config);
-
-    /// Get a surface that matches a "texture copy" display transfer config
-    SurfaceRect_Tuple GetTexCopySurface(const SurfaceParams& params);
-
-    /// Write any cached resources overlapping the region back to memory (if dirty)
-    void FlushRegion(PAddr addr, u32 size, Surface flush_surface = nullptr);
-
-    /// Mark region as being invalidated by region_owner (nullptr if 3DS memory)
-    void InvalidateRegion(PAddr addr, u32 size, const Surface& region_owner);
-
-    /// Flush all cached resources tracked by this cache manager
-    void FlushAll();
-
-    /// Clear all cached resources tracked by this cache manager
-    void ClearAll(bool flush);
-
-    // Textures from destroyed surfaces are stored here to be recyled to reduce allocation overhead
-    // in the driver
-    // this must be placed above the surface_cache to ensure all cached surfaces are destroyed
-    // before destroying the recycler
-    std::unordered_multimap<HostTextureTag, OGLTexture> host_texture_recycler;
-
-private:
-    void DuplicateSurface(const Surface& src_surface, const Surface& dest_surface);
-
-    /// Update surface's texture for given region when necessary
-    void ValidateSurface(const Surface& surface, PAddr addr, u32 size);
-
-    // Returns false if there is a surface in the cache at the interval with the same bit-width,
-    bool NoUnimplementedReinterpretations(const OpenGL::Surface& surface,
-                                          OpenGL::SurfaceParams& params,
-                                          const OpenGL::SurfaceInterval& interval);
-
-    // Return true if a surface with an invalid pixel format exists at the interval
-    bool IntervalHasInvalidPixelFormat(SurfaceParams& params, const SurfaceInterval& interval);
-
-    // Attempt to find a reinterpretable surface in the cache and use it to copy for validation
-    bool ValidateByReinterpretation(const Surface& surface, SurfaceParams& params,
-                                    const SurfaceInterval& interval);
-
-    /// Create a new surface
-    Surface CreateSurface(const SurfaceParams& params);
-
-    /// Register surface into the cache
-    void RegisterSurface(const Surface& surface);
-
-    /// Remove surface from the cache
-    void UnregisterSurface(const Surface& surface);
-
-    /// Increase/decrease the number of surface in pages touching the specified region
-    void UpdatePagesCachedCount(PAddr addr, u32 size, int delta);
-
-    SurfaceCache surface_cache;
-    PageMap cached_pages;
-    SurfaceMap dirty_regions;
-    SurfaceSet remove_surfaces;
-
-    OGLFramebuffer read_framebuffer;
-    OGLFramebuffer draw_framebuffer;
-
-    u16 resolution_scale_factor;
-
-    std::unordered_map<TextureCubeConfig, CachedTextureCube> texture_cube_cache;
-
-    std::recursive_mutex mutex;
-
-public:
-    OGLTexture AllocateSurfaceTexture(const FormatTuple& format_tuple, u32 width, u32 height);
-
-    std::unique_ptr<TextureFilterer> texture_filterer;
-    std::unique_ptr<FormatReinterpreterOpenGL> format_reinterpreter;
-    std::unique_ptr<TextureDownloaderES> texture_downloader_es;
-};
-
-} // namespace OpenGL

src/video_core/renderer_opengl/gl_resource_manager.cpp

@@ -1,15 +1,11 @@
-// Copyright 2015 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <utility>
-#include <glad/glad.h>
-#include "common/common_types.h"
 #include "common/microprofile.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
 #include "video_core/renderer_opengl/gl_state.h"
-#include "video_core/renderer_opengl/gl_vars.h"
 
 MICROPROFILE_DEFINE(OpenGL_ResourceCreation, "OpenGL", "Resource Creation", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_RGB(128, 128, 192));
@@ -17,16 +13,18 @@ MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_R
 namespace OpenGL {
 
 void OGLRenderbuffer::Create() {
-    if (handle != 0)
+    if (handle != 0) {
         return;
+    }
 
     MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenRenderbuffers(1, &handle);
 }
 
 void OGLRenderbuffer::Release() {
-    if (handle == 0)
+    if (handle == 0) {
         return;
+    }
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteRenderbuffers(1, &handle);
@@ -35,16 +33,18 @@ void OGLRenderbuffer::Release() {
 }
 
 void OGLTexture::Create() {
-    if (handle != 0)
+    if (handle != 0) {
         return;
+    }
 
     MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenTextures(1, &handle);
 }
 
 void OGLTexture::Release() {
-    if (handle == 0)
+    if (handle == 0) {
         return;
+    }
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteTextures(1, &handle);
@@ -52,72 +52,64 @@ void OGLTexture::Release() {
     handle = 0;
 }
 
-void OGLTexture::Allocate(GLenum target, GLsizei levels, GLenum internalformat, GLenum format,
+void OGLTexture::Allocate(GLenum target, GLsizei levels, GLenum internalformat, GLsizei width,
-                          GLenum type, GLsizei width, GLsizei height, GLsizei depth) {
+                          GLsizei height, GLsizei depth) {
-    const bool tex_storage = GLAD_GL_ARB_texture_storage || GLES;
+    GLuint old_tex = OpenGLState::GetCurState().texture_units[0].texture_2d;
+    glActiveTexture(GL_TEXTURE0);
+    glBindTexture(GL_TEXTURE_2D, handle);
 
     switch (target) {
     case GL_TEXTURE_1D:
     case GL_TEXTURE:
-        if (tex_storage) {
+        glTexStorage1D(target, levels, internalformat, width);
-            glTexStorage1D(target, levels, internalformat, width);
-        } else {
-            for (GLsizei level{0}; level < levels; ++level) {
-                glTexImage1D(target, level, internalformat, width, 0, format, type, nullptr);
-                width >>= 1;
-            }
-        }
         break;
     case GL_TEXTURE_2D:
     case GL_TEXTURE_1D_ARRAY:
     case GL_TEXTURE_RECTANGLE:
     case GL_TEXTURE_CUBE_MAP:
-        if (tex_storage) {
+        glTexStorage2D(target, levels, internalformat, width, height);
-            glTexStorage2D(target, levels, internalformat, width, height);
-        } else {
-            for (GLsizei level{0}; level < levels; ++level) {
-                glTexImage2D(target, level, internalformat, width, height, 0, format, type,
-                             nullptr);
-                width >>= 1;
-                if (target != GL_TEXTURE_1D_ARRAY)
-                    height >>= 1;
-            }
-        }
         break;
     case GL_TEXTURE_3D:
     case GL_TEXTURE_2D_ARRAY:
     case GL_TEXTURE_CUBE_MAP_ARRAY:
-        if (tex_storage) {
+        glTexStorage3D(target, levels, internalformat, width, height, depth);
-            glTexStorage3D(target, levels, internalformat, width, height, depth);
-        } else {
-            for (GLsizei level{0}; level < levels; ++level) {
-                glTexImage3D(target, level, internalformat, width, height, depth, 0, format, type,
-                             nullptr);
-            }
-            width >>= 1;
-            height >>= 1;
-            if (target == GL_TEXTURE_3D)
-                depth >>= 1;
-        }
         break;
     }
 
-    if (!tex_storage) {
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-        glTexParameteri(target, GL_TEXTURE_MAX_LEVEL, levels - 1);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+
+    glBindTexture(GL_TEXTURE_2D, old_tex);
+}
+
+void OGLTexture::CopyFrom(const OGLTexture& other, GLenum target, GLsizei levels, GLsizei width,
+                          GLsizei height) {
+    GLuint old_tex = OpenGLState::GetCurState().texture_units[0].texture_2d;
+    glActiveTexture(GL_TEXTURE0);
+    glBindTexture(GL_TEXTURE_2D, handle);
+
+    for (u32 level = 0; level < levels; level++) {
+        glCopyImageSubData(other.handle, target, level, 0, 0, 0, handle, target, level, 0, 0, 0,
+                           width >> level, height >> level, 1);
     }
+
+    glBindTexture(GL_TEXTURE_2D, old_tex);
 }
 
 void OGLSampler::Create() {
-    if (handle != 0)
+    if (handle != 0) {
         return;
+    }
 
     MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
     glGenSamplers(1, &handle);
 }
 
 void OGLSampler::Release() {
-    if (handle == 0)
+    if (handle == 0) {
         return;
+    }
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteSamplers(1, &handle);

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -1,4 +1,4 @@
-// Copyright 2015 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
@@ -8,7 +8,6 @@
 #include <vector>
 #include <glad/glad.h>
 #include "common/common_types.h"
-#include "video_core/renderer_opengl/gl_shader_util.h"
 
 namespace OpenGL {
 
@@ -59,8 +58,11 @@ public:
     /// Deletes the internal OpenGL resource
     void Release();
 
-    static void Allocate(GLenum target, GLsizei levels, GLenum internalformat, GLenum format,
+    void Allocate(GLenum target, GLsizei levels, GLenum internalformat, GLsizei width,
-                         GLenum type, GLsizei width, GLsizei height = 1, GLsizei depth = 1);
+                  GLsizei height = 1, GLsizei depth = 1);
+
+    void CopyFrom(const OGLTexture& other, GLenum target, GLsizei levels, GLsizei width,
+                  GLsizei height);
 
     GLuint handle = 0;
 };

src/video_core/renderer_opengl/gl_shader_gen.cpp

@@ -1,21 +1,13 @@
-// Copyright 2015 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <array>
-#include <cstddef>
 #include <string_view>
 #include <fmt/format.h>
-#include "common/assert.h"
-#include "common/bit_field.h"
 #include "common/bit_set.h"
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "video_core/regs_framebuffer.h"
+#include "video_core/pica_state.h"
-#include "video_core/regs_lighting.h"
-#include "video_core/regs_rasterizer.h"
-#include "video_core/regs_texturing.h"
-#include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
@@ -1233,18 +1225,6 @@ ShaderDecompiler::ProgramResult GenerateFragmentShader(const PicaFSConfig& confi
     const auto& state = config.state;
     std::string out;
 
-    if (GLES) {
-        out += R"(
-#define ALLOW_SHADOW (defined(CITRA_GLES))
-)";
-    } else {
-        out += R"(
-#extension GL_ARB_shader_image_load_store : enable
-#extension GL_ARB_shader_image_size : enable
-#define ALLOW_SHADOW (defined(GL_ARB_shader_image_load_store) && defined(GL_ARB_shader_image_size))
-)";
-    }
-
     if (separable_shader && !GLES) {
         out += "#extension GL_ARB_separate_shader_objects : enable\n";
     }
@@ -1270,7 +1250,6 @@ uniform samplerBuffer texture_buffer_lut_lf;
 uniform samplerBuffer texture_buffer_lut_rg;
 uniform samplerBuffer texture_buffer_lut_rgba;
 
-#if ALLOW_SHADOW
 layout(r32ui) uniform readonly uimage2D shadow_texture_px;
 layout(r32ui) uniform readonly uimage2D shadow_texture_nx;
 layout(r32ui) uniform readonly uimage2D shadow_texture_py;
@@ -1278,7 +1257,6 @@ layout(r32ui) uniform readonly uimage2D shadow_texture_ny;
 layout(r32ui) uniform readonly uimage2D shadow_texture_pz;
 layout(r32ui) uniform readonly uimage2D shadow_texture_nz;
 layout(r32ui) uniform uimage2D shadow_buffer;
-#endif
 )";
 
     out += UniformBlockDef;
@@ -1332,8 +1310,6 @@ float getLod(vec2 coord) {
     return log2(max(d.x, d.y));
 }
 
-#if ALLOW_SHADOW
-
 uvec2 DecodeShadow(uint pixel) {
     return uvec2(pixel >> 8, pixel & 0xFFu);
 }
@@ -1459,18 +1435,6 @@ vec4 shadowTextureCube(vec2 uv, float w) {
         CompareShadow(pixels.w, z));
     return vec4(mix2(s, f));
 }
-
-#else
-
-vec4 shadowTexture(vec2 uv, float w) {
-    return vec4(1.0);
-}
-
-vec4 shadowTextureCube(vec2 uv, float w) {
-    return vec4(1.0);
-}
-
-#endif
 )";
 
     if (config.state.proctex.enable)
@@ -1559,7 +1523,6 @@ vec4 secondary_fragment_color = vec4(0.0);
 
     if (state.shadow_rendering) {
         out += R"(
-#if ALLOW_SHADOW
 uint d = uint(clamp(depth, 0.0, 1.0) * float(0xFFFFFF));
 uint s = uint(last_tex_env_out.g * float(0xFF));
 ivec2 image_coord = ivec2(gl_FragCoord.xy);
@@ -1582,7 +1545,6 @@ do {
     new = EncodeShadow(ref);
 
 } while ((old = imageAtomicCompSwap(shadow_buffer, image_coord, old, new)) != old2);
-#endif // ALLOW_SHADOW
 )";
     } else {
         out += "gl_FragDepth = depth;\n";

src/video_core/renderer_opengl/gl_shader_gen.h

@@ -3,13 +3,8 @@
 // Refer to the license.txt file included.
 
 #pragma once
-
-#include <array>
-#include <cstring>
 #include <functional>
 #include <optional>
-#include <string>
-#include <type_traits>
 #include "common/hash.h"
 #include "video_core/regs.h"
 #include "video_core/shader/shader.h"

src/video_core/renderer_opengl/gl_shader_manager.cpp

@@ -1,16 +1,18 @@
-// Copyright 2018 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #include <algorithm>
+#include <set>
 #include <thread>
 #include <unordered_map>
 #include <boost/functional/hash.hpp>
 #include <boost/variant.hpp>
-#include "core/core.h"
 #include "core/frontend/scope_acquire_context.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_disk_cache.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
+#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_vars.h"
 #include "video_core/video_core.h"
 
@@ -149,11 +151,11 @@ void PicaUniformsData::SetFromRegs(const Pica::ShaderRegs& regs,
     std::transform(std::begin(setup.uniforms.b), std::end(setup.uniforms.b), std::begin(bools),
                    [](bool value) -> BoolAligned { return {value ? GL_TRUE : GL_FALSE}; });
     std::transform(std::begin(regs.int_uniforms), std::end(regs.int_uniforms), std::begin(i),
-                   [](const auto& value) -> GLuvec4 {
+                   [](const auto& value) -> Common::Vec4u {
                        return {value.x.Value(), value.y.Value(), value.z.Value(), value.w.Value()};
                    });
     std::transform(std::begin(setup.uniforms.f), std::end(setup.uniforms.f), std::begin(f),
-                   [](const auto& value) -> GLvec4 {
+                   [](const auto& value) -> Common::Vec4f {
                        return {value.x.ToFloat32(), value.y.ToFloat32(), value.z.ToFloat32(),
                                value.w.ToFloat32()};
                    });
@@ -325,8 +327,8 @@ using FragmentShaders = ShaderCache<PicaFSConfig, &GenerateFragmentShader, GL_FR
 
 class ShaderProgramManager::Impl {
 public:
-    explicit Impl(bool separable, bool is_amd)
+    explicit Impl(bool separable)
-        : is_amd(is_amd), separable(separable), programmable_vertex_shaders(separable),
+        : separable(separable), programmable_vertex_shaders(separable),
           trivial_vertex_shader(separable), fixed_geometry_shaders(separable),
           fragment_shaders(separable), disk_cache(separable) {
         if (separable)
@@ -359,7 +361,6 @@ public:
         }
     };
 
-    bool is_amd;
     bool separable;
 
     ShaderTuple current;
@@ -375,9 +376,8 @@ public:
     ShaderDiskCache disk_cache;
 };
 
-ShaderProgramManager::ShaderProgramManager(Frontend::EmuWindow& emu_window_, bool separable,
+ShaderProgramManager::ShaderProgramManager(Frontend::EmuWindow& emu_window_, bool separable)
-                                           bool is_amd)
+    : impl(std::make_unique<Impl>(separable)), emu_window{emu_window_} {}
-    : impl(std::make_unique<Impl>(separable, is_amd)), emu_window{emu_window_} {}
 
 ShaderProgramManager::~ShaderProgramManager() = default;
 
@@ -439,15 +439,6 @@ void ShaderProgramManager::UseFragmentShader(const Pica::Regs& regs) {
 
 void ShaderProgramManager::ApplyTo(OpenGLState& state) {
     if (impl->separable) {
-        if (impl->is_amd) {
-            // Without this reseting, AMD sometimes freezes when one stage is changed but not
-            // for the others. On the other hand, including this reset seems to introduce memory
-            // leak in Intel Graphics.
-            glUseProgramStages(
-                impl->pipeline.handle,
-                GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT | GL_FRAGMENT_SHADER_BIT, 0);
-        }
-
         glUseProgramStages(impl->pipeline.handle, GL_VERTEX_SHADER_BIT, impl->current.vs);
         glUseProgramStages(impl->pipeline.handle, GL_GEOMETRY_SHADER_BIT, impl->current.gs);
         glUseProgramStages(impl->pipeline.handle, GL_FRAGMENT_SHADER_BIT, impl->current.fs);
@@ -471,12 +462,6 @@ void ShaderProgramManager::ApplyTo(OpenGLState& state) {
 
 void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
                                          const VideoCore::DiskResourceLoadCallback& callback) {
-    if (!GLAD_GL_ARB_get_program_binary && !GLES) {
-        LOG_ERROR(Render_OpenGL,
-                  "Cannot load disk cache as ARB_get_program_binary is not supported!");
-        return;
-    }
-
     auto& disk_cache = impl->disk_cache;
     const auto transferable = disk_cache.LoadTransferable();
     if (!transferable) {

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -1,35 +1,44 @@
-// Copyright 2018 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
 
 #include <memory>
-#include <glad/glad.h>
+#include "common/vector_math.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/regs_lighting.h"
-#include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_shader_gen.h"
-#include "video_core/renderer_opengl/gl_state.h"
-#include "video_core/renderer_opengl/pica_to_gl.h"
 
 namespace Core {
 class System;
 }
 
+namespace Frontend {
+class EmuWindow;
+}
+
+namespace Pica {
+struct Regs;
+struct ShaderRegs;
+} // namespace Pica
+
+namespace Pica::Shader {
+struct ShaderSetup;
+}
+
 namespace OpenGL {
 
-enum class UniformBindings : GLuint { Common, VS, GS };
+enum class UniformBindings : u32 { Common, VS, GS };
 
 struct LightSrc {
-    alignas(16) GLvec3 specular_0;
+    alignas(16) Common::Vec3f specular_0;
-    alignas(16) GLvec3 specular_1;
+    alignas(16) Common::Vec3f specular_1;
-    alignas(16) GLvec3 diffuse;
+    alignas(16) Common::Vec3f diffuse;
-    alignas(16) GLvec3 ambient;
+    alignas(16) Common::Vec3f ambient;
-    alignas(16) GLvec3 position;
+    alignas(16) Common::Vec3f position;
-    alignas(16) GLvec3 spot_direction; // negated
+    alignas(16) Common::Vec3f spot_direction; // negated
-    GLfloat dist_atten_bias;
+    float dist_atten_bias;
-    GLfloat dist_atten_scale;
+    float dist_atten_scale;
 };
 
 /// Uniform structure for the Uniform Buffer Object, all vectors must be 16-byte aligned
@@ -37,39 +46,38 @@ struct LightSrc {
 //       the end of a uniform block is included in UNIFORM_BLOCK_DATA_SIZE or not.
 //       Not following that rule will cause problems on some AMD drivers.
 struct UniformData {
-    GLint framebuffer_scale;
+    int framebuffer_scale;
-    GLint alphatest_ref;
+    int alphatest_ref;
-    GLfloat depth_scale;
+    float depth_scale;
-    GLfloat depth_offset;
+    float depth_offset;
-    GLfloat shadow_bias_constant;
+    float shadow_bias_constant;
-    GLfloat shadow_bias_linear;
+    float shadow_bias_linear;
-    GLint scissor_x1;
+    int scissor_x1;
-    GLint scissor_y1;
+    int scissor_y1;
-    GLint scissor_x2;
+    int scissor_x2;
-    GLint scissor_y2;
+    int scissor_y2;
-    GLint fog_lut_offset;
+    int fog_lut_offset;
-    GLint proctex_noise_lut_offset;
+    int proctex_noise_lut_offset;
-    GLint proctex_color_map_offset;
+    int proctex_color_map_offset;
-    GLint proctex_alpha_map_offset;
+    int proctex_alpha_map_offset;
-    GLint proctex_lut_offset;
+    int proctex_lut_offset;
-    GLint proctex_diff_lut_offset;
+    int proctex_diff_lut_offset;
-    GLfloat proctex_bias;
+    float proctex_bias;
-    GLint shadow_texture_bias;
+    int shadow_texture_bias;
-    alignas(16) GLivec4 lighting_lut_offset[Pica::LightingRegs::NumLightingSampler / 4];
+    alignas(16) Common::Vec4i lighting_lut_offset[Pica::LightingRegs::NumLightingSampler / 4];
-    alignas(16) GLvec3 fog_color;
+    alignas(16) Common::Vec3f fog_color;
-    alignas(8) GLvec2 proctex_noise_f;
+    alignas(8) Common::Vec2f proctex_noise_f;
-    alignas(8) GLvec2 proctex_noise_a;
+    alignas(8) Common::Vec2f proctex_noise_a;
-    alignas(8) GLvec2 proctex_noise_p;
+    alignas(8) Common::Vec2f proctex_noise_p;
-    alignas(16) GLvec3 lighting_global_ambient;
+    alignas(16) Common::Vec3f lighting_global_ambient;
     LightSrc light_src[8];
-    alignas(16) GLvec4 const_color[6]; // A vec4 color for each of the six tev stages
+    alignas(16) Common::Vec4f const_color[6]; // A vec4 color for each of the six tev stages
-    alignas(16) GLvec4 tev_combiner_buffer_color;
+    alignas(16) Common::Vec4f tev_combiner_buffer_color;
-    alignas(16) GLvec4 clip_coef;
+    alignas(16) Common::Vec4f clip_coef;
 };
 
-static_assert(
+static_assert(sizeof(UniformData) == 0x4F0,
-    sizeof(UniformData) == 0x4F0,
+              "The size of the UniformData does not match the structure in the shader");
-    "The size of the UniformData structure has changed, update the structure in the shader");
 static_assert(sizeof(UniformData) < 16384,
               "UniformData structure must be less than 16kb as per the OpenGL spec");
 
@@ -79,27 +87,28 @@ struct PicaUniformsData {
     void SetFromRegs(const Pica::ShaderRegs& regs, const Pica::Shader::ShaderSetup& setup);
 
     struct BoolAligned {
-        alignas(16) GLint b;
+        alignas(16) int b;
     };
 
     std::array<BoolAligned, 16> bools;
-    alignas(16) std::array<GLuvec4, 4> i;
+    alignas(16) std::array<Common::Vec4u, 4> i;
-    alignas(16) std::array<GLvec4, 96> f;
+    alignas(16) std::array<Common::Vec4f, 96> f;
 };
 
 struct VSUniformData {
     PicaUniformsData uniforms;
 };
-static_assert(
+static_assert(sizeof(VSUniformData) == 1856,
-    sizeof(VSUniformData) == 1856,
+              "The size of the VSUniformData does not match the structure in the shader");
-    "The size of the VSUniformData structure has changed, update the structure in the shader");
 static_assert(sizeof(VSUniformData) < 16384,
               "VSUniformData structure must be less than 16kb as per the OpenGL spec");
 
+class OpenGLState;
+
 /// A class that manage different shader stages and configures them with given config data.
 class ShaderProgramManager {
 public:
-    ShaderProgramManager(Frontend::EmuWindow& emu_window_, bool separable, bool is_amd);
+    ShaderProgramManager(Frontend::EmuWindow& emu_window_, bool separable);
     ~ShaderProgramManager();
 
     void LoadDiskCache(const std::atomic_bool& stop_loading,

src/video_core/renderer_opengl/gl_shader_util.cpp

@@ -26,7 +26,7 @@ GLuint LoadShader(const char* source, GLenum type) {
 #extension GL_EXT_clip_cull_distance : enable
 #endif // defined(GL_EXT_clip_cull_distance)
 )"
-                                     : "#version 330\n";
+                                     : "#version 430 core\n";
 
     const char* debug_type;
     switch (type) {

src/video_core/renderer_opengl/gl_stream_buffer.cpp

@@ -1,13 +1,10 @@
-// Copyright 2018 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <deque>
-#include <vector>
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/microprofile.h"
-#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
 
 MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
@@ -15,21 +12,12 @@ MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
 
 namespace OpenGL {
 
-OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool array_buffer_for_amd,
+OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent)
-                                 bool prefer_coherent)
     : gl_target(target), buffer_size(size) {
     gl_buffer.Create();
     glBindBuffer(gl_target, gl_buffer.handle);
 
     GLsizeiptr allocate_size = size;
-    if (array_buffer_for_amd) {
-        // On AMD GPU there is a strange crash in indexed drawing. The crash happens when the buffer
-        // read position is near the end and is an out-of-bound access to the vertex buffer. This is
-        // probably a bug in the driver and is related to the usage of vec3<byte> attributes in the
-        // vertex array. Doubling the allocation size for the vertex buffer seems to avoid the
-        // crash.
-        allocate_size *= 2;
-    }
 
     if (GLAD_GL_ARB_buffer_storage) {
         persistent = true;

src/video_core/renderer_opengl/gl_stream_buffer.h

@@ -1,20 +1,16 @@
-// Copyright 2018 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
-
 #include <tuple>
-#include <glad/glad.h>
-#include "common/common_types.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 
 namespace OpenGL {
 
 class OGLStreamBuffer : private NonCopyable {
 public:
-    explicit OGLStreamBuffer(GLenum target, GLsizeiptr size, bool array_buffer_for_amd,
+    explicit OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent = false);
-                             bool prefer_coherent = false);
     ~OGLStreamBuffer();
 
     GLuint GetHandle() const;

src/video_core/renderer_opengl/gl_surface_params.h

@@ -1,270 +0,0 @@
-// Copyright 2020 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-#include <climits>
-#include <boost/icl/interval.hpp>
-#include "common/assert.h"
-#include "common/math_util.h"
-#include "core/hw/gpu.h"
-#include "video_core/regs_framebuffer.h"
-#include "video_core/regs_texturing.h"
-
-namespace OpenGL {
-
-struct CachedSurface;
-using Surface = std::shared_ptr<CachedSurface>;
-
-using SurfaceInterval = boost::icl::right_open_interval<PAddr>;
-
-struct SurfaceParams {
-private:
-    static constexpr std::array<unsigned int, 18> BPP_TABLE = {
-        32, // RGBA8
-        24, // RGB8
-        16, // RGB5A1
-        16, // RGB565
-        16, // RGBA4
-        16, // IA8
-        16, // RG8
-        8,  // I8
-        8,  // A8
-        8,  // IA4
-        4,  // I4
-        4,  // A4
-        4,  // ETC1
-        8,  // ETC1A4
-        16, // D16
-        0,
-        24, // D24
-        32, // D24S8
-    };
-
-public:
-    enum class PixelFormat {
-        // First 5 formats are shared between textures and color buffers
-        RGBA8 = 0,
-        RGB8 = 1,
-        RGB5A1 = 2,
-        RGB565 = 3,
-        RGBA4 = 4,
-
-        // Texture-only formats
-        IA8 = 5,
-        RG8 = 6,
-        I8 = 7,
-        A8 = 8,
-        IA4 = 9,
-        I4 = 10,
-        A4 = 11,
-        ETC1 = 12,
-        ETC1A4 = 13,
-
-        // Depth buffer-only formats
-        D16 = 14,
-        // gap
-        D24 = 16,
-        D24S8 = 17,
-
-        Invalid = 255,
-    };
-
-    enum class SurfaceType {
-        Color = 0,
-        Texture = 1,
-        Depth = 2,
-        DepthStencil = 3,
-        Fill = 4,
-        Invalid = 5
-    };
-
-    static constexpr unsigned int GetFormatBpp(PixelFormat format) {
-        const auto format_idx = static_cast<std::size_t>(format);
-        DEBUG_ASSERT_MSG(format_idx < BPP_TABLE.size(), "Invalid pixel format {}", format_idx);
-        return BPP_TABLE[format_idx];
-    }
-
-    unsigned int GetFormatBpp() const {
-        return GetFormatBpp(pixel_format);
-    }
-
-    static std::string_view PixelFormatAsString(PixelFormat format) {
-        switch (format) {
-        case PixelFormat::RGBA8:
-            return "RGBA8";
-        case PixelFormat::RGB8:
-            return "RGB8";
-        case PixelFormat::RGB5A1:
-            return "RGB5A1";
-        case PixelFormat::RGB565:
-            return "RGB565";
-        case PixelFormat::RGBA4:
-            return "RGBA4";
-        case PixelFormat::IA8:
-            return "IA8";
-        case PixelFormat::RG8:
-            return "RG8";
-        case PixelFormat::I8:
-            return "I8";
-        case PixelFormat::A8:
-            return "A8";
-        case PixelFormat::IA4:
-            return "IA4";
-        case PixelFormat::I4:
-            return "I4";
-        case PixelFormat::A4:
-            return "A4";
-        case PixelFormat::ETC1:
-            return "ETC1";
-        case PixelFormat::ETC1A4:
-            return "ETC1A4";
-        case PixelFormat::D16:
-            return "D16";
-        case PixelFormat::D24:
-            return "D24";
-        case PixelFormat::D24S8:
-            return "D24S8";
-        default:
-            return "Not a real pixel format";
-        }
-    }
-
-    static PixelFormat PixelFormatFromTextureFormat(Pica::TexturingRegs::TextureFormat format) {
-        return ((unsigned int)format < 14) ? (PixelFormat)format : PixelFormat::Invalid;
-    }
-
-    static PixelFormat PixelFormatFromColorFormat(Pica::FramebufferRegs::ColorFormat format) {
-        return ((unsigned int)format < 5) ? (PixelFormat)format : PixelFormat::Invalid;
-    }
-
-    static PixelFormat PixelFormatFromDepthFormat(Pica::FramebufferRegs::DepthFormat format) {
-        return ((unsigned int)format < 4) ? (PixelFormat)((unsigned int)format + 14)
-                                          : PixelFormat::Invalid;
-    }
-
-    static PixelFormat PixelFormatFromGPUPixelFormat(GPU::Regs::PixelFormat format) {
-        switch (format) {
-        // RGB565 and RGB5A1 are switched in PixelFormat compared to ColorFormat
-        case GPU::Regs::PixelFormat::RGB565:
-            return PixelFormat::RGB565;
-        case GPU::Regs::PixelFormat::RGB5A1:
-            return PixelFormat::RGB5A1;
-        default:
-            return ((unsigned int)format < 5) ? (PixelFormat)format : PixelFormat::Invalid;
-        }
-    }
-
-    static bool CheckFormatsBlittable(PixelFormat pixel_format_a, PixelFormat pixel_format_b) {
-        SurfaceType a_type = GetFormatType(pixel_format_a);
-        SurfaceType b_type = GetFormatType(pixel_format_b);
-
-        if ((a_type == SurfaceType::Color || a_type == SurfaceType::Texture) &&
-            (b_type == SurfaceType::Color || b_type == SurfaceType::Texture)) {
-            return true;
-        }
-
-        if (a_type == SurfaceType::Depth && b_type == SurfaceType::Depth) {
-            return true;
-        }
-
-        if (a_type == SurfaceType::DepthStencil && b_type == SurfaceType::DepthStencil) {
-            return true;
-        }
-
-        return false;
-    }
-
-    static constexpr SurfaceType GetFormatType(PixelFormat pixel_format) {
-        if ((unsigned int)pixel_format < 5) {
-            return SurfaceType::Color;
-        }
-
-        if ((unsigned int)pixel_format < 14) {
-            return SurfaceType::Texture;
-        }
-
-        if (pixel_format == PixelFormat::D16 || pixel_format == PixelFormat::D24) {
-            return SurfaceType::Depth;
-        }
-
-        if (pixel_format == PixelFormat::D24S8) {
-            return SurfaceType::DepthStencil;
-        }
-
-        return SurfaceType::Invalid;
-    }
-
-    /// Update the params "size", "end" and "type" from the already set "addr", "width", "height"
-    /// and "pixel_format"
-    void UpdateParams() {
-        if (stride == 0) {
-            stride = width;
-        }
-        type = GetFormatType(pixel_format);
-        size = !is_tiled ? BytesInPixels(stride * (height - 1) + width)
-                         : BytesInPixels(stride * 8 * (height / 8 - 1) + width * 8);
-        end = addr + size;
-    }
-
-    SurfaceInterval GetInterval() const {
-        return SurfaceInterval(addr, end);
-    }
-
-    // Returns the outer rectangle containing "interval"
-    SurfaceParams FromInterval(SurfaceInterval interval) const;
-
-    SurfaceInterval GetSubRectInterval(Common::Rectangle<u32> unscaled_rect) const;
-
-    // Returns the region of the biggest valid rectange within interval
-    SurfaceInterval GetCopyableInterval(const Surface& src_surface) const;
-
-    u32 GetScaledWidth() const {
-        return width * res_scale;
-    }
-
-    u32 GetScaledHeight() const {
-        return height * res_scale;
-    }
-
-    Common::Rectangle<u32> GetRect() const {
-        return {0, height, width, 0};
-    }
-
-    Common::Rectangle<u32> GetScaledRect() const {
-        return {0, GetScaledHeight(), GetScaledWidth(), 0};
-    }
-
-    u32 PixelsInBytes(u32 size) const {
-        return size * CHAR_BIT / GetFormatBpp(pixel_format);
-    }
-
-    u32 BytesInPixels(u32 pixels) const {
-        return pixels * GetFormatBpp(pixel_format) / CHAR_BIT;
-    }
-
-    bool ExactMatch(const SurfaceParams& other_surface) const;
-    bool CanSubRect(const SurfaceParams& sub_surface) const;
-    bool CanExpand(const SurfaceParams& expanded_surface) const;
-    bool CanTexCopy(const SurfaceParams& texcopy_params) const;
-
-    Common::Rectangle<u32> GetSubRect(const SurfaceParams& sub_surface) const;
-    Common::Rectangle<u32> GetScaledSubRect(const SurfaceParams& sub_surface) const;
-
-    PAddr addr = 0;
-    PAddr end = 0;
-    u32 size = 0;
-
-    u32 width = 0;
-    u32 height = 0;
-    u32 stride = 0;
-    u16 res_scale = 1;
-
-    bool is_tiled = false;
-    PixelFormat pixel_format = PixelFormat::Invalid;
-    SurfaceType type = SurfaceType::Invalid;
-};
-
-} // namespace OpenGL

src/video_core/renderer_opengl/pica_to_gl.h

@@ -1,32 +1,18 @@
-// Copyright 2015 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
 
 #include <array>
-#include <cstddef>
 #include <glad/glad.h>
 #include "common/assert.h"
-#include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "video_core/regs_framebuffer.h"
 #include "video_core/regs_lighting.h"
 #include "video_core/regs_texturing.h"
 
-using GLvec2 = std::array<GLfloat, 2>;
-using GLvec3 = std::array<GLfloat, 3>;
-using GLvec4 = std::array<GLfloat, 4>;
-
-using GLuvec2 = std::array<GLuint, 2>;
-using GLuvec3 = std::array<GLuint, 3>;
-using GLuvec4 = std::array<GLuint, 4>;
-
-using GLivec2 = std::array<GLint, 2>;
-using GLivec3 = std::array<GLint, 3>;
-using GLivec4 = std::array<GLint, 4>;
-
 namespace PicaToGL {
 
 using TextureFilter = Pica::TexturingRegs::TextureConfig::TextureFilter;
@@ -245,21 +231,14 @@ inline GLenum StencilOp(Pica::FramebufferRegs::StencilAction action) {
     return stencil_op_table[index];
 }
 
-inline GLvec4 ColorRGBA8(const u32 color) {
+inline Common::Vec4f ColorRGBA8(const u32 color) {
-    return {{
+    const auto rgba =
-        (color >> 0 & 0xFF) / 255.0f,
+        Common::Vec4u{color >> 0 & 0xFF, color >> 8 & 0xFF, color >> 16 & 0xFF, color >> 24 & 0xFF};
-        (color >> 8 & 0xFF) / 255.0f,
+    return rgba / 255.0f;
-        (color >> 16 & 0xFF) / 255.0f,
-        (color >> 24 & 0xFF) / 255.0f,
-    }};
 }
 
-inline std::array<GLfloat, 3> LightColor(const Pica::LightingRegs::LightColor& color) {
+inline Common::Vec3f LightColor(const Pica::LightingRegs::LightColor& color) {
-    return {{
+    return Common::Vec3u{color.r, color.g, color.b} / 255.0f;
-        color.r / 255.0f,
-        color.g / 255.0f,
-        color.b / 255.0f,
-    }};
 }
 
 } // namespace PicaToGL

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -1,27 +1,14 @@
-// Copyright 2014 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <algorithm>
-#include <array>
-#include <condition_variable>
-#include <cstddef>
-#include <cstdlib>
-#include <deque>
-#include <memory>
-#include <mutex>
-#include <glad/glad.h>
 #include <queue>
-#include "common/assert.h"
-#include "common/bit_field.h"
 #include "common/logging/log.h"
 #include "common/microprofile.h"
 #include "core/core.h"
-#include "core/core_timing.h"
 #include "core/dumping/backend.h"
 #include "core/frontend/emu_window.h"
 #include "core/frontend/framebuffer_layout.h"
-#include "core/hw/gpu.h"
 #include "core/hw/hw.h"
 #include "core/hw/lcd.h"
 #include "core/memory.h"
@@ -29,6 +16,7 @@
 #include "core/tracer/recorder.h"
 #include "video_core/debug_utils/debug_utils.h"
 #include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_opengl/gl_shader_util.h"
 #include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_vars.h"
 #include "video_core/renderer_opengl/post_processing_opengl.h"
@@ -526,7 +514,6 @@ void RendererOpenGL::RenderToMailbox(const Layout::FramebufferLayout& layout,
             mailbox->ReloadRenderFrame(frame, layout.width, layout.height);
         }
 
-        GLuint render_texture = frame->color.handle;
         state.draw.draw_framebuffer = frame->render.handle;
         state.Apply();
         DrawScreens(layout, flipped);
@@ -1200,6 +1187,8 @@ static const char* GetSource(GLenum source) {
         UNREACHABLE();
     }
 #undef RET
+
+    return "";
 }
 
 static const char* GetType(GLenum type) {
@@ -1218,6 +1207,8 @@ static const char* GetType(GLenum type) {
         UNREACHABLE();
     }
 #undef RET
+
+    return "";
 }
 
 static void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severity,
@@ -1243,7 +1234,7 @@ static void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum
 VideoCore::ResultStatus RendererOpenGL::Init() {
 #ifndef ANDROID
     if (!gladLoadGL()) {
-        return VideoCore::ResultStatus::ErrorBelowGL33;
+        return VideoCore::ResultStatus::ErrorBelowGL43;
     }
 
     // Qualcomm has some spammy info messages that are marked as errors but not important
@@ -1254,9 +1245,9 @@ VideoCore::ResultStatus RendererOpenGL::Init() {
     }
 #endif
 
-    const char* gl_version{reinterpret_cast<char const*>(glGetString(GL_VERSION))};
+    const std::string_view gl_version{reinterpret_cast<char const*>(glGetString(GL_VERSION))};
-    const char* gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
+    const std::string_view gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
-    const char* gpu_model{reinterpret_cast<char const*>(glGetString(GL_RENDERER))};
+    const std::string_view gpu_model{reinterpret_cast<char const*>(glGetString(GL_RENDERER))};
 
     LOG_INFO(Render_OpenGL, "GL_VERSION: {}", gl_version);
     LOG_INFO(Render_OpenGL, "GL_VENDOR: {}", gpu_vendor);
@@ -1268,12 +1259,12 @@ VideoCore::ResultStatus RendererOpenGL::Init() {
     telemetry_session.AddField(user_system, "GPU_Model", std::string(gpu_model));
     telemetry_session.AddField(user_system, "GPU_OpenGL_Version", std::string(gl_version));
 
-    if (!strcmp(gpu_vendor, "GDI Generic")) {
+    if (gpu_vendor == "GDI Generic") {
         return VideoCore::ResultStatus::ErrorGenericDrivers;
     }
 
-    if (!(GLAD_GL_VERSION_3_3 || GLAD_GL_ES_VERSION_3_1)) {
+    if (!(GLAD_GL_VERSION_4_3 || GLAD_GL_ES_VERSION_3_1)) {
-        return VideoCore::ResultStatus::ErrorBelowGL33;
+        return VideoCore::ResultStatus::ErrorBelowGL43;
     }
 
     InitOpenGLObjects();

src/video_core/renderer_opengl/renderer_opengl.h

@@ -1,13 +1,10 @@
-// Copyright 2014 Citra Emulator Project
+// Copyright 2022 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #pragma once
 
 #include <array>
-#include <glad/glad.h>
-#include "common/common_types.h"
-#include "common/math_util.h"
 #include "core/hw/gpu.h"
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_opengl/frame_dumper_opengl.h"

src/video_core/renderer_opengl/texture_downloader_es.cpp

@@ -4,13 +4,10 @@
 
 #include <chrono>
 #include <vector>
-
 #include <fmt/chrono.h>
-
 #include "common/logging/log.h"
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
+#include "video_core/rasterizer_cache/rasterizer_cache_utils.h"
 #include "video_core/renderer_opengl/gl_state.h"
-#include "video_core/renderer_opengl/gl_vars.h"
 #include "video_core/renderer_opengl/texture_downloader_es.h"
 
 #include "shaders/depth_to_color.frag"
@@ -78,13 +75,13 @@ void TextureDownloaderES::Test() {
             }
     };
     LOG_INFO(Render_OpenGL, "GL_DEPTH24_STENCIL8 download test starting");
-    test(depth_format_tuples[3], std::vector<u32>{}, 4096,
+    test(GetFormatTuple(PixelFormat::D24S8), std::vector<u32>{}, 4096,
          [](std::size_t idx) { return static_cast<u32>((idx << 8) | (idx & 0xFF)); });
     LOG_INFO(Render_OpenGL, "GL_DEPTH_COMPONENT24 download test starting");
-    test(depth_format_tuples[2], std::vector<u32>{}, 4096,
+    test(GetFormatTuple(PixelFormat::D24), std::vector<u32>{}, 4096,
          [](std::size_t idx) { return static_cast<u32>(idx << 8); });
     LOG_INFO(Render_OpenGL, "GL_DEPTH_COMPONENT16 download test starting");
-    test(depth_format_tuples[0], std::vector<u16>{}, 256,
+    test(GetFormatTuple(PixelFormat::D16), std::vector<u16>{}, 256,
          [](std::size_t idx) { return static_cast<u16>(idx); });
 
     cur_state.Apply();

src/video_core/renderer_opengl/texture_filters/anime4k/anime4k_ultrafast.cpp

@@ -30,7 +30,6 @@
 // DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 // OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/texture_filters/anime4k/anime4k_ultrafast.h"
 
 #include "shaders/refine.frag"
@@ -72,9 +71,8 @@ Anime4kUltrafast::Anime4kUltrafast(u16 scale_factor) : TextureFilterBase(scale_f
     cur_state.Apply();
 }
 
-void Anime4kUltrafast::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect,
+void Anime4kUltrafast::Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                              GLuint dst_tex, const Common::Rectangle<u32>& dst_rect,
+                              const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) {
-                              GLuint read_fb_handle, GLuint draw_fb_handle) {
     const OpenGLState cur_state = OpenGLState::GetCurState();
 
     // These will have handles from the previous texture that was filtered, reset them to avoid
@@ -92,15 +90,9 @@ void Anime4kUltrafast::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_
         state.Apply();
         glActiveTexture(GL_TEXTURE0);
         glBindTexture(GL_TEXTURE_2D, texture.tex.handle);
-        if (GL_ARB_texture_storage) {
+        glTexStorage2D(GL_TEXTURE_2D, 1, internal_format,
-            glTexStorage2D(GL_TEXTURE_2D, 1, internal_format,
+                       src_rect.GetWidth() * internal_scale_factor,
-                           src_rect.GetWidth() * internal_scale_factor,
+                       src_rect.GetHeight() * internal_scale_factor);
-                           src_rect.GetHeight() * internal_scale_factor);
-        } else {
-            glTexImage2D(
-                GL_TEXTURE_2D, 0, internal_format, src_rect.GetWidth() * internal_scale_factor,
-                src_rect.GetHeight() * internal_scale_factor, 0, format, GL_HALF_FLOAT, nullptr);
-        }
         glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
                                texture.tex.handle, 0);
         return texture;
@@ -112,7 +104,7 @@ void Anime4kUltrafast::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_
                       static_cast<GLint>(src_rect.bottom * internal_scale_factor),
                       static_cast<GLsizei>(src_rect.GetWidth() * internal_scale_factor),
                       static_cast<GLsizei>(src_rect.GetHeight() * internal_scale_factor)};
-    state.texture_units[0].texture_2d = src_tex;
+    state.texture_units[0].texture_2d = src_tex.handle;
     state.texture_units[1].texture_2d = LUMAD.tex.handle;
     state.texture_units[2].texture_2d = XY.tex.handle;
     state.draw.draw_framebuffer = XY.fbo.handle;
@@ -131,11 +123,12 @@ void Anime4kUltrafast::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_
     state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
                       static_cast<GLsizei>(dst_rect.GetWidth()),
                       static_cast<GLsizei>(dst_rect.GetHeight())};
-    state.draw.draw_framebuffer = draw_fb_handle;
+    state.draw.draw_framebuffer = draw_fbo.handle;
     state.draw.shader_program = refine_program.handle;
     state.Apply();
 
-    glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex, 0);
+    glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex.handle,
+                           0);
     glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
 

src/video_core/renderer_opengl/texture_filters/anime4k/anime4k_ultrafast.h

@@ -15,9 +15,8 @@ public:
     static constexpr std::string_view NAME = "Anime4K Ultrafast";
 
     explicit Anime4kUltrafast(u16 scale_factor);
-    void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+    void Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
+                const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) override;
-                GLuint draw_fb_handle) override;
 
 private:
     static constexpr u8 internal_scale_factor = 2;

src/video_core/renderer_opengl/texture_filters/bicubic/bicubic.cpp

@@ -2,7 +2,6 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/texture_filters/bicubic/bicubic.h"
 
 #include "shaders/bicubic.frag"
@@ -26,18 +25,18 @@ Bicubic::Bicubic(u16 scale_factor) : TextureFilterBase(scale_factor) {
     glSamplerParameteri(src_sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
 } // namespace OpenGL
 
-void Bicubic::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+void Bicubic::Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                     const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
+                     const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) {
-                     GLuint draw_fb_handle) {
     const OpenGLState cur_state = OpenGLState::GetCurState();
-    state.texture_units[0].texture_2d = src_tex;
+    state.texture_units[0].texture_2d = src_tex.handle;
-    state.draw.draw_framebuffer = draw_fb_handle;
+    state.draw.draw_framebuffer = draw_fbo.handle;
     state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
                       static_cast<GLsizei>(dst_rect.GetWidth()),
                       static_cast<GLsizei>(dst_rect.GetHeight())};
     state.Apply();
 
-    glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex, 0);
+    glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex.handle,
+                           0);
     glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
 

src/video_core/renderer_opengl/texture_filters/bicubic/bicubic.h

@@ -15,9 +15,8 @@ public:
     static constexpr std::string_view NAME = "Bicubic";
 
     explicit Bicubic(u16 scale_factor);
-    void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+    void Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
+                const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) override;
-                GLuint draw_fb_handle) override;
 
 private:
     OpenGLState state{};

src/video_core/renderer_opengl/texture_filters/scale_force/scale_force.cpp

@@ -2,7 +2,6 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/texture_filters/scale_force/scale_force.h"
 
 #include "shaders/scale_force.frag"
@@ -26,18 +25,18 @@ ScaleForce::ScaleForce(u16 scale_factor) : TextureFilterBase(scale_factor) {
     glSamplerParameteri(src_sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
 }
 
-void ScaleForce::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+void ScaleForce::Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                        const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
+                        const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) {
-                        GLuint draw_fb_handle) {
     const OpenGLState cur_state = OpenGLState::GetCurState();
-    state.texture_units[0].texture_2d = src_tex;
+    state.texture_units[0].texture_2d = src_tex.handle;
-    state.draw.draw_framebuffer = draw_fb_handle;
+    state.draw.draw_framebuffer = draw_fbo.handle;
     state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
                       static_cast<GLsizei>(dst_rect.GetWidth()),
                       static_cast<GLsizei>(dst_rect.GetHeight())};
     state.Apply();
 
-    glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex, 0);
+    glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex.handle,
+                           0);
     glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
 

src/video_core/renderer_opengl/texture_filters/scale_force/scale_force.h

@@ -15,9 +15,8 @@ public:
     static constexpr std::string_view NAME = "ScaleForce";
 
     explicit ScaleForce(u16 scale_factor);
-    void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+    void Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
+                const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) override;
-                GLuint draw_fb_handle) override;
 
 private:
     OpenGLState state{};

src/video_core/renderer_opengl/texture_filters/texture_filter_base.h

@@ -3,23 +3,32 @@
 // Refer to the license.txt file included.
 
 #pragma once
-
+#include <string_view>
 #include "common/common_types.h"
 #include "common/math_util.h"
-#include "video_core/renderer_opengl/gl_surface_params.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
 
 namespace OpenGL {
 
+class TextureRuntime;
+class OGLTexture;
+
 class TextureFilterBase {
     friend class TextureFilterer;
-    virtual void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
-                        const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
-                        GLuint draw_fb_handle) = 0;
 
 public:
-    explicit TextureFilterBase(u16 scale_factor) : scale_factor{scale_factor} {};
+    explicit TextureFilterBase(u16 scale_factor) : scale_factor(scale_factor) {
+        draw_fbo.Create();
+    };
+
     virtual ~TextureFilterBase() = default;
 
+private:
+    virtual void Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
+                        const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) = 0;
+
+protected:
+    OGLFramebuffer draw_fbo;
     const u16 scale_factor{};
 };
 

src/video_core/renderer_opengl/texture_filters/texture_filterer.cpp

@@ -58,15 +58,16 @@ bool TextureFilterer::IsNull() const {
     return !filter;
 }
 
-bool TextureFilterer::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+bool TextureFilterer::Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                             const Common::Rectangle<u32>& dst_rect,
+                             const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect,
-                             SurfaceParams::SurfaceType type, GLuint read_fb_handle,
+                             SurfaceType type) {
-                             GLuint draw_fb_handle) {
+
-    // depth / stencil texture filtering is not supported for now
+    // Depth/Stencil texture filtering is not supported for now
-    if (IsNull() ||
+    if (IsNull() || (type != SurfaceType::Color && type != SurfaceType::Texture)) {
-        (type != SurfaceParams::SurfaceType::Color && type != SurfaceParams::SurfaceType::Texture))
         return false;
-    filter->Filter(src_tex, src_rect, dst_tex, dst_rect, read_fb_handle, draw_fb_handle);
+    }
+
+    filter->Filter(src_tex, src_rect, dst_tex, dst_rect);
     return true;
 }
 
@@ -82,6 +83,7 @@ std::vector<std::string_view> TextureFilterer::GetFilterNames() {
             return lhs_is_none && !rhs_is_none;
         return lhs < rhs;
     });
+
     return ret;
 }
 

src/video_core/renderer_opengl/texture_filters/texture_filterer.h

@@ -7,10 +7,7 @@
 #include <memory>
 #include <string_view>
 #include <vector>
-#include <glad/glad.h>
+#include "video_core/rasterizer_cache/pixel_format.h"
-#include "common/common_types.h"
-#include "common/math_util.h"
-#include "video_core/renderer_opengl/gl_surface_params.h"
 #include "video_core/renderer_opengl/texture_filters/texture_filter_base.h"
 
 namespace OpenGL {
@@ -19,15 +16,18 @@ class TextureFilterer {
 public:
     static constexpr std::string_view NONE = "none";
 
+public:
     explicit TextureFilterer(std::string_view filter_name, u16 scale_factor);
-    // returns true if the filter actually changed
+
+    // Returns true if the filter actually changed
     bool Reset(std::string_view new_filter_name, u16 new_scale_factor);
-    // returns true if there is no active filter
+
+    // Returns true if there is no active filter
     bool IsNull() const;
-    // returns true if the texture was able to be filtered
+
-    bool Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+    // Returns true if the texture was able to be filtered
-                const Common::Rectangle<u32>& dst_rect, SurfaceParams::SurfaceType type,
+    bool Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                GLuint read_fb_handle, GLuint draw_fb_handle);
+                const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect, SurfaceType type);
 
     static std::vector<std::string_view> GetFilterNames();
 

src/video_core/renderer_opengl/texture_filters/xbrz/xbrz_freescale.cpp

@@ -40,7 +40,6 @@
 // OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 // THE SOFTWARE.
 
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/texture_filters/xbrz/xbrz_freescale.h"
 
 #include "shaders/xbrz_freescale.frag"
@@ -49,6 +48,7 @@
 namespace OpenGL {
 
 XbrzFreescale::XbrzFreescale(u16 scale_factor) : TextureFilterBase(scale_factor) {
+
     const OpenGLState cur_state = OpenGLState::GetCurState();
 
     program.Create(xbrz_freescale_vert.data(), xbrz_freescale_frag.data());
@@ -62,7 +62,9 @@ XbrzFreescale::XbrzFreescale(u16 scale_factor) : TextureFilterBase(scale_factor)
     glSamplerParameteri(src_sampler.handle, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
     glSamplerParameteri(src_sampler.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
     glSamplerParameteri(src_sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-    glUniform1f(glGetUniformLocation(program.handle, "scale"), static_cast<GLfloat>(scale_factor));
+
+    const GLint scale_loc = glGetUniformLocation(program.handle, "scale");
+    glUniform1f(scale_loc, static_cast<GLfloat>(scale_factor));
 
     cur_state.Apply();
     state.draw.vertex_array = vao.handle;
@@ -70,19 +72,19 @@ XbrzFreescale::XbrzFreescale(u16 scale_factor) : TextureFilterBase(scale_factor)
     state.texture_units[0].sampler = src_sampler.handle;
 }
 
-void XbrzFreescale::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+void XbrzFreescale::Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                           const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
+                           const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) {
-                           GLuint draw_fb_handle) {
     const OpenGLState cur_state = OpenGLState::GetCurState();
 
-    state.texture_units[0].texture_2d = src_tex;
+    state.texture_units[0].texture_2d = src_tex.handle;
-    state.draw.draw_framebuffer = draw_fb_handle;
+    state.draw.draw_framebuffer = draw_fbo.handle;
     state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
                       static_cast<GLsizei>(dst_rect.GetWidth()),
                       static_cast<GLsizei>(dst_rect.GetHeight())};
     state.Apply();
 
-    glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex, 0);
+    glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex.handle,
+                           0);
     glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
 

src/video_core/renderer_opengl/texture_filters/xbrz/xbrz_freescale.h

@@ -3,7 +3,6 @@
 // Refer to the license.txt file included.
 
 #pragma once
-
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/texture_filters/texture_filter_base.h"
@@ -15,9 +14,8 @@ public:
     static constexpr std::string_view NAME = "xBRZ freescale";
 
     explicit XbrzFreescale(u16 scale_factor);
-    void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
+    void Filter(const OGLTexture& src_tex, Common::Rectangle<u32> src_rect,
-                const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
+                const OGLTexture& dst_tex, Common::Rectangle<u32> dst_rect) override;
-                GLuint draw_fb_handle) override;
 
 private:
     OpenGLState state{};

src/video_core/video_core.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <atomic>
+#include <functional>
 #include <iostream>
 #include <memory>
 #include "core/frontend/emu_window.h"
@@ -49,7 +50,7 @@ extern Memory::MemorySystem* g_memory;
 enum class ResultStatus {
     Success,
     ErrorGenericDrivers,
-    ErrorBelowGL33,
+    ErrorBelowGL43,
 };
 
 /// Initialize the video core