Merge pull request #254 from bunnei/port-citra-renderer

Port Citra OpenGL rasterizer code

externals/glad/Readme.md

@@ -1,5 +1,5 @@
 These files were generated by the [glad](https://github.com/Dav1dde/glad) OpenGL loader generator and have been checked in as-is. You can re-generate them using glad with the following command:
 
 ```
-python -m glad --profile core --out-path glad/ --api gl=3.3,gles=3.0
+python -m glad --profile core --out-path glad/ --api gl=3.3 --generator=c
 ```

externals/glad/include/KHR/khrplatform.h

@@ -26,7 +26,7 @@
 
 /* Khronos platform-specific types and definitions.
  *
- * $Revision: 23298 $ on $Date: 2013-09-30 17:07:13 -0700 (Mon, 30 Sep 2013) $
+ * $Revision: 32517 $ on $Date: 2016-03-11 02:41:19 -0800 (Fri, 11 Mar 2016) $
  *
  * Adopters may modify this file to suit their platform. Adopters are
  * encouraged to submit platform specific modifications to the Khronos
@@ -101,6 +101,8 @@
 #   define KHRONOS_APICALL __declspec(dllimport)
 #elif defined (__SYMBIAN32__)
 #   define KHRONOS_APICALL IMPORT_C
+#elif defined(__ANDROID__)
+#   define KHRONOS_APICALL __attribute__((visibility("default")))
 #else
 #   define KHRONOS_APICALL
 #endif

src/video_core/CMakeLists.txt

@@ -11,13 +11,24 @@ add_library(video_core STATIC
     gpu.h
     memory_manager.cpp
     memory_manager.h
+    rasterizer_interface.h
     renderer_base.cpp
     renderer_base.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.h
+    renderer_opengl/gl_shader_decompiler.cpp
+    renderer_opengl/gl_shader_decompiler.h
+    renderer_opengl/gl_shader_gen.cpp
+    renderer_opengl/gl_shader_gen.h
     renderer_opengl/gl_shader_util.cpp
     renderer_opengl/gl_shader_util.h
     renderer_opengl/gl_state.cpp
     renderer_opengl/gl_state.h
+    renderer_opengl/gl_stream_buffer.cpp
+    renderer_opengl/gl_stream_buffer.h
     renderer_opengl/renderer_opengl.cpp
     renderer_opengl/renderer_opengl.h
     utils.h

src/video_core/rasterizer_interface.h

@@ -0,0 +1,61 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+struct ScreenInfo;
+
+namespace VideoCore {
+
+class RasterizerInterface {
+public:
+    virtual ~RasterizerInterface() {}
+
+    /// Draw the current batch of triangles
+    virtual void DrawTriangles() = 0;
+
+    /// Notify rasterizer that the specified Maxwell register has been changed
+    virtual void NotifyMaxwellRegisterChanged(u32 id) = 0;
+
+    /// Notify rasterizer that all caches should be flushed to 3DS memory
+    virtual void FlushAll() = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be flushed to 3DS memory
+    virtual void FlushRegion(PAddr addr, u32 size) = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be invalidated
+    virtual void InvalidateRegion(PAddr addr, u32 size) = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be flushed to 3DS memory
+    /// and invalidated
+    virtual void FlushAndInvalidateRegion(PAddr addr, u32 size) = 0;
+
+    /// Attempt to use a faster method to perform a display transfer with is_texture_copy = 0
+    virtual bool AccelerateDisplayTransfer(const void* config) {
+        return false;
+    }
+
+    /// Attempt to use a faster method to perform a display transfer with is_texture_copy = 1
+    virtual bool AccelerateTextureCopy(const void* config) {
+        return false;
+    }
+
+    /// Attempt to use a faster method to fill a region
+    virtual bool AccelerateFill(const void* config) {
+        return false;
+    }
+
+    /// Attempt to use a faster method to display the framebuffer to screen
+    virtual bool AccelerateDisplay(const void* config, PAddr framebuffer_addr, u32 pixel_stride,
+                                   ScreenInfo& screen_info) {
+        return false;
+    }
+
+    virtual bool AccelerateDrawBatch(bool is_indexed) {
+        return false;
+    }
+};
+} // namespace VideoCore

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -0,0 +1,269 @@
+// Copyright 2015 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#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/settings.h"
+#include "video_core/renderer_opengl/gl_rasterizer.h"
+#include "video_core/renderer_opengl/gl_shader_gen.h"
+#include "video_core/renderer_opengl/renderer_opengl.h"
+
+using PixelFormat = SurfaceParams::PixelFormat;
+using SurfaceType = SurfaceParams::SurfaceType;
+
+MICROPROFILE_DEFINE(OpenGL_VAO, "OpenGL", "Vertex Array Setup", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_VS, "OpenGL", "Vertex Shader Setup", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_FS, "OpenGL", "Fragment Shader Setup", MP_RGB(128, 128, 192));
+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));
+
+enum class UniformBindings : GLuint { Common, VS, FS };
+
+static void SetShaderUniformBlockBinding(GLuint shader, const char* name, UniformBindings binding,
+                                         size_t expected_size) {
+    GLuint ub_index = glGetUniformBlockIndex(shader, name);
+    if (ub_index != GL_INVALID_INDEX) {
+        GLint ub_size = 0;
+        glGetActiveUniformBlockiv(shader, ub_index, GL_UNIFORM_BLOCK_DATA_SIZE, &ub_size);
+        ASSERT_MSG(ub_size == expected_size,
+                   "Uniform block size did not match! Got %d, expected %zu",
+                   static_cast<int>(ub_size), expected_size);
+        glUniformBlockBinding(shader, ub_index, static_cast<GLuint>(binding));
+    }
+}
+
+static void SetShaderUniformBlockBindings(GLuint shader) {
+    SetShaderUniformBlockBinding(shader, "shader_data", UniformBindings::Common,
+                                 sizeof(RasterizerOpenGL::UniformData));
+    SetShaderUniformBlockBinding(shader, "vs_config", UniformBindings::VS,
+                                 sizeof(RasterizerOpenGL::VSUniformData));
+    SetShaderUniformBlockBinding(shader, "fs_config", UniformBindings::FS,
+                                 sizeof(RasterizerOpenGL::FSUniformData));
+}
+
+RasterizerOpenGL::RasterizerOpenGL() {
+    has_ARB_buffer_storage = false;
+    has_ARB_direct_state_access = false;
+    has_ARB_separate_shader_objects = false;
+    has_ARB_vertex_attrib_binding = false;
+
+    GLint ext_num;
+    glGetIntegerv(GL_NUM_EXTENSIONS, &ext_num);
+    for (GLint i = 0; i < ext_num; i++) {
+        std::string extension{reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, i))};
+
+        if (extension == "GL_ARB_buffer_storage") {
+            has_ARB_buffer_storage = true;
+        } else if (extension == "GL_ARB_direct_state_access") {
+            has_ARB_direct_state_access = true;
+        } else if (extension == "GL_ARB_separate_shader_objects") {
+            has_ARB_separate_shader_objects = true;
+        } else if (extension == "GL_ARB_vertex_attrib_binding") {
+            has_ARB_vertex_attrib_binding = true;
+        }
+    }
+
+    // Clipping plane 0 is always enabled for PICA fixed clip plane z <= 0
+    state.clip_distance[0] = true;
+
+    // Generate VBO, VAO and UBO
+    vertex_buffer = OGLStreamBuffer::MakeBuffer(GLAD_GL_ARB_buffer_storage, GL_ARRAY_BUFFER);
+    vertex_buffer->Create(VERTEX_BUFFER_SIZE, VERTEX_BUFFER_SIZE / 2);
+    sw_vao.Create();
+    uniform_buffer.Create();
+
+    state.draw.vertex_array = sw_vao.handle;
+    state.draw.vertex_buffer = vertex_buffer->GetHandle();
+    state.draw.uniform_buffer = uniform_buffer.handle;
+    state.Apply();
+
+    glBufferData(GL_UNIFORM_BUFFER, sizeof(UniformData), nullptr, GL_STATIC_DRAW);
+    glBindBufferBase(GL_UNIFORM_BUFFER, 0, uniform_buffer.handle);
+
+    uniform_block_data.dirty = true;
+
+    // Create render framebuffer
+    framebuffer.Create();
+
+    if (has_ARB_separate_shader_objects) {
+        hw_vao.Create();
+        hw_vao_enabled_attributes.fill(false);
+
+        stream_buffer = OGLStreamBuffer::MakeBuffer(has_ARB_buffer_storage, GL_ARRAY_BUFFER);
+        stream_buffer->Create(STREAM_BUFFER_SIZE, STREAM_BUFFER_SIZE / 2);
+        state.draw.vertex_buffer = stream_buffer->GetHandle();
+
+        pipeline.Create();
+        vs_input_index_min = 0;
+        vs_input_index_max = 0;
+        state.draw.program_pipeline = pipeline.handle;
+        state.draw.shader_program = 0;
+        state.draw.vertex_array = hw_vao.handle;
+        state.Apply();
+
+        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, stream_buffer->GetHandle());
+
+        vs_uniform_buffer.Create();
+        glBindBuffer(GL_UNIFORM_BUFFER, vs_uniform_buffer.handle);
+        glBufferData(GL_UNIFORM_BUFFER, sizeof(VSUniformData), nullptr, GL_STREAM_COPY);
+        glBindBufferBase(GL_UNIFORM_BUFFER, 1, vs_uniform_buffer.handle);
+    } else {
+        UNIMPLEMENTED();
+    }
+
+    accelerate_draw = AccelDraw::Disabled;
+
+    glEnable(GL_BLEND);
+
+    // Sync fixed function OpenGL state
+    SyncClipEnabled();
+    SyncClipCoef();
+    SyncCullMode();
+    SyncBlendEnabled();
+    SyncBlendFuncs();
+    SyncBlendColor();
+}
+
+RasterizerOpenGL::~RasterizerOpenGL() {
+    if (stream_buffer != nullptr) {
+        state.draw.vertex_buffer = stream_buffer->GetHandle();
+        state.Apply();
+        stream_buffer->Release();
+    }
+}
+
+static constexpr std::array<GLenum, 4> vs_attrib_types{
+    GL_BYTE,          // VertexAttributeFormat::BYTE
+    GL_UNSIGNED_BYTE, // VertexAttributeFormat::UBYTE
+    GL_SHORT,         // VertexAttributeFormat::SHORT
+    GL_FLOAT          // VertexAttributeFormat::FLOAT
+};
+
+void RasterizerOpenGL::AnalyzeVertexArray(bool is_indexed) {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SetupVertexArray(u8* array_ptr, GLintptr buffer_offset) {
+    MICROPROFILE_SCOPE(OpenGL_VAO);
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SetupVertexShader(VSUniformData* ub_ptr, GLintptr buffer_offset) {
+    MICROPROFILE_SCOPE(OpenGL_VS);
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SetupFragmentShader(FSUniformData* ub_ptr, GLintptr buffer_offset) {
+    MICROPROFILE_SCOPE(OpenGL_FS);
+    UNIMPLEMENTED();
+}
+
+bool RasterizerOpenGL::AccelerateDrawBatch(bool is_indexed) {
+    if (!has_ARB_separate_shader_objects) {
+        UNIMPLEMENTED();
+        return false;
+    }
+
+    accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays;
+    DrawTriangles();
+
+    return true;
+}
+
+void RasterizerOpenGL::DrawTriangles() {
+    MICROPROFILE_SCOPE(OpenGL_Drawing);
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::NotifyMaxwellRegisterChanged(u32 id) {}
+
+void RasterizerOpenGL::FlushAll() {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    res_cache.FlushAll();
+}
+
+void RasterizerOpenGL::FlushRegion(PAddr addr, u32 size) {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    res_cache.FlushRegion(addr, size);
+}
+
+void RasterizerOpenGL::InvalidateRegion(PAddr addr, u32 size) {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    res_cache.InvalidateRegion(addr, size, nullptr);
+}
+
+void RasterizerOpenGL::FlushAndInvalidateRegion(PAddr addr, u32 size) {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    res_cache.FlushRegion(addr, size);
+    res_cache.InvalidateRegion(addr, size, nullptr);
+}
+
+bool RasterizerOpenGL::AccelerateDisplayTransfer(const void* config) {
+    MICROPROFILE_SCOPE(OpenGL_Blits);
+    UNIMPLEMENTED();
+    return true;
+}
+
+bool RasterizerOpenGL::AccelerateTextureCopy(const void* config) {
+    UNIMPLEMENTED();
+    return true;
+}
+
+bool RasterizerOpenGL::AccelerateFill(const void* config) {
+    UNIMPLEMENTED();
+    return true;
+}
+
+bool RasterizerOpenGL::AccelerateDisplay(const void* config, PAddr framebuffer_addr,
+                                         u32 pixel_stride, ScreenInfo& screen_info) {
+    UNIMPLEMENTED();
+    return true;
+}
+
+void RasterizerOpenGL::SetShader() {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SyncClipEnabled() {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SyncClipCoef() {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SyncCullMode() {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SyncDepthScale() {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SyncDepthOffset() {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SyncBlendEnabled() {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SyncBlendFuncs() {
+    UNIMPLEMENTED();
+}
+
+void RasterizerOpenGL::SyncBlendColor() {
+    UNIMPLEMENTED();
+}

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -0,0 +1,162 @@
+// Copyright 2015 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 <unordered_map>
+#include <vector>
+#include <glad/glad.h>
+#include "common/bit_field.h"
+#include "common/common_types.h"
+#include "common/hash.h"
+#include "common/vector_math.h"
+#include "video_core/rasterizer_interface.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_gen.h"
+#include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/gl_stream_buffer.h"
+
+struct ScreenInfo;
+
+class RasterizerOpenGL : public VideoCore::RasterizerInterface {
+public:
+    RasterizerOpenGL();
+    ~RasterizerOpenGL() override;
+
+    void DrawTriangles() override;
+    void NotifyMaxwellRegisterChanged(u32 id) override;
+    void FlushAll() override;
+    void FlushRegion(PAddr addr, u32 size) override;
+    void InvalidateRegion(PAddr addr, u32 size) override;
+    void FlushAndInvalidateRegion(PAddr addr, u32 size) override;
+    bool AccelerateDisplayTransfer(const void* config) override;
+    bool AccelerateTextureCopy(const void* config) override;
+    bool AccelerateFill(const void* config) override;
+    bool AccelerateDisplay(const void* config, PAddr framebuffer_addr, u32 pixel_stride,
+                           ScreenInfo& screen_info) override;
+    bool AccelerateDrawBatch(bool is_indexed) override;
+
+    struct VertexShader {
+        OGLShader shader;
+    };
+
+    struct FragmentShader {
+        OGLShader shader;
+    };
+
+    /// Uniform structure for the Uniform Buffer Object, all vectors must be 16-byte aligned
+    // NOTE: Always keep a vec4 at the end. The GL spec is not clear wether the alignment at
+    //       the end of a uniform block is included in UNIFORM_BLOCK_DATA_SIZE or not.
+    //       Not following that rule will cause problems on some AMD drivers.
+    struct UniformData {};
+
+    // static_assert(
+    //    sizeof(UniformData) == 0x460,
+    //    "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");
+
+    struct VSUniformData {};
+    // static_assert(
+    //    sizeof(VSUniformData) == 1856,
+    //    "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");
+
+    struct FSUniformData {};
+    // static_assert(
+    //    sizeof(FSUniformData) == 1856,
+    //    "The size of the FSUniformData structure has changed, update the structure in the
+    //    shader");
+    static_assert(sizeof(FSUniformData) < 16384,
+                  "FSUniformData structure must be less than 16kb as per the OpenGL spec");
+
+private:
+    struct SamplerInfo {};
+
+    /// Syncs the clip enabled status to match the guest state
+    void SyncClipEnabled();
+
+    /// Syncs the clip coefficients to match the guest state
+    void SyncClipCoef();
+
+    /// Sets the OpenGL shader in accordance with the current guest state
+    void SetShader();
+
+    /// Syncs the cull mode to match the guest state
+    void SyncCullMode();
+
+    /// Syncs the depth scale to match the guest state
+    void SyncDepthScale();
+
+    /// Syncs the depth offset to match the guest state
+    void SyncDepthOffset();
+
+    /// Syncs the blend enabled status to match the guest state
+    void SyncBlendEnabled();
+
+    /// Syncs the blend functions to match the guest state
+    void SyncBlendFuncs();
+
+    /// Syncs the blend color to match the guest state
+    void SyncBlendColor();
+
+    bool has_ARB_buffer_storage;
+    bool has_ARB_direct_state_access;
+    bool has_ARB_separate_shader_objects;
+    bool has_ARB_vertex_attrib_binding;
+
+    OpenGLState state;
+
+    RasterizerCacheOpenGL res_cache;
+
+    struct {
+        UniformData data;
+        bool dirty;
+    } uniform_block_data = {};
+
+    OGLPipeline pipeline;
+    OGLVertexArray sw_vao;
+    OGLVertexArray hw_vao;
+    std::array<bool, 16> hw_vao_enabled_attributes;
+
+    std::array<SamplerInfo, 3> texture_samplers;
+    static constexpr size_t VERTEX_BUFFER_SIZE = 128 * 1024 * 1024;
+    std::unique_ptr<OGLStreamBuffer> vertex_buffer;
+    OGLBuffer uniform_buffer;
+    OGLFramebuffer framebuffer;
+
+    static constexpr size_t STREAM_BUFFER_SIZE = 4 * 1024 * 1024;
+    std::unique_ptr<OGLStreamBuffer> stream_buffer;
+
+    GLint vs_input_index_min;
+    GLint vs_input_index_max;
+    GLsizeiptr vs_input_size;
+
+    void AnalyzeVertexArray(bool is_indexed);
+    void SetupVertexArray(u8* array_ptr, GLintptr buffer_offset);
+
+    OGLBuffer vs_uniform_buffer;
+    std::unordered_map<GLShader::MaxwellVSConfig, VertexShader*> vs_shader_map;
+    std::unordered_map<std::string, VertexShader> vs_shader_cache;
+    OGLShader vs_default_shader;
+
+    void SetupVertexShader(VSUniformData* ub_ptr, GLintptr buffer_offset);
+
+    OGLBuffer fs_uniform_buffer;
+    std::unordered_map<GLShader::MaxwellFSConfig, FragmentShader*> fs_shader_map;
+    std::unordered_map<std::string, FragmentShader> fs_shader_cache;
+    OGLShader fs_default_shader;
+
+    void SetupFragmentShader(FSUniformData* ub_ptr, GLintptr buffer_offset);
+
+    enum class AccelDraw { Disabled, Arrays, Indexed };
+    AccelDraw accelerate_draw;
+};

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -0,0 +1,350 @@
+// Copyright 2015 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+#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 <glad/glad.h>
+#include "common/assert.h"
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "common/math_util.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+
+struct CachedSurface;
+using Surface = std::shared_ptr<CachedSurface>;
+using SurfaceSet = std::set<Surface>;
+
+using SurfaceRegions = boost::icl::interval_set<PAddr>;
+using SurfaceMap = boost::icl::interval_map<PAddr, Surface>;
+using SurfaceCache = boost::icl::interval_map<PAddr, SurfaceSet>;
+
+using SurfaceInterval = SurfaceCache::interval_type;
+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, MathUtil::Rectangle<u32>>;
+using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, MathUtil::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
+};
+
+struct SurfaceParams {
+    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) {
+        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
+        };
+
+        assert(static_cast<size_t>(format) < bpp_table.size());
+        return bpp_table[static_cast<size_t>(format)];
+    }
+    unsigned int GetFormatBpp() const {
+        return GetFormatBpp(pixel_format);
+    }
+
+    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::right_open(addr, end);
+    }
+
+    // Returns the outer rectangle containing "interval"
+    SurfaceParams FromInterval(SurfaceInterval interval) const;
+
+    SurfaceInterval GetSubRectInterval(MathUtil::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;
+    }
+
+    MathUtil::Rectangle<u32> GetRect() const {
+        return {0, height, width, 0};
+    }
+
+    MathUtil::Rectangle<u32> GetScaledRect() const {
+        return {0, GetScaledHeight(), GetScaledWidth(), 0};
+    }
+
+    u64 PixelsInBytes(u64 size) const {
+        return size * CHAR_BIT / GetFormatBpp(pixel_format);
+    }
+
+    u64 BytesInPixels(u64 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;
+
+    MathUtil::Rectangle<u32> GetSubRect(const SurfaceParams& sub_surface) const;
+    MathUtil::Rectangle<u32> GetScaledSubRect(const SurfaceParams& sub_surface) const;
+
+    PAddr addr = 0;
+    PAddr end = 0;
+    u64 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;
+};
+
+struct CachedSurface : SurfaceParams {
+    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 {
+        return (invalid_regions & GetInterval()) == SurfaceRegions(GetInterval());
+    }
+
+    bool registered = false;
+    SurfaceRegions invalid_regions;
+
+    u64 fill_size = 0; /// Number of bytes to read from fill_data
+    std::array<u8, 4> fill_data;
+
+    OGLTexture texture;
+
+    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::unique_ptr<u8[]> gl_buffer;
+    size_t gl_buffer_size = 0;
+
+    // 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);
+
+    // Upload/Download data in gl_buffer in/to this surface's texture
+    void UploadGLTexture(const MathUtil::Rectangle<u32>& rect, GLuint read_fb_handle,
+                         GLuint draw_fb_handle);
+    void DownloadGLTexture(const MathUtil::Rectangle<u32>& rect, GLuint read_fb_handle,
+                           GLuint draw_fb_handle);
+};
+
+class RasterizerCacheOpenGL : NonCopyable {
+public:
+    RasterizerCacheOpenGL();
+    ~RasterizerCacheOpenGL();
+
+    /// Blit one surface's texture to another
+    bool BlitSurfaces(const Surface& src_surface, const MathUtil::Rectangle<u32>& src_rect,
+                      const Surface& dst_surface, const MathUtil::Rectangle<u32>& dst_rect);
+
+    void ConvertD24S8toABGR(GLuint src_tex, const MathUtil::Rectangle<u32>& src_rect,
+                            GLuint dst_tex, const MathUtil::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 void* config);
+
+    /// Get the color and depth surfaces based on the framebuffer configuration
+    SurfaceSurfaceRect_Tuple GetFramebufferSurfaces(bool using_color_fb, bool using_depth_fb,
+                                                    const MathUtil::Rectangle<s32>& viewport_rect);
+
+    /// Get a surface that matches the fill config
+    Surface GetFillSurface(const void* 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, u64 size, Surface flush_surface = nullptr);
+
+    /// Mark region as being invalidated by region_owner (nullptr if 3DS memory)
+    void InvalidateRegion(PAddr addr, u64 size, const Surface& region_owner);
+
+    /// Flush all cached resources tracked by this cache manager
+    void FlushAll();
+
+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, u64 size);
+
+    /// 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, u64 size, int delta);
+
+    SurfaceCache surface_cache;
+    PageMap cached_pages;
+    SurfaceMap dirty_regions;
+    SurfaceSet remove_surfaces;
+
+    OGLFramebuffer read_framebuffer;
+    OGLFramebuffer draw_framebuffer;
+
+    OGLVertexArray attributeless_vao;
+    OGLBuffer d24s8_abgr_buffer;
+    GLsizeiptr d24s8_abgr_buffer_size;
+    OGLShader d24s8_abgr_shader;
+    GLint d24s8_abgr_tbo_size_u_id;
+    GLint d24s8_abgr_viewport_u_id;
+};

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -36,7 +36,7 @@ public:
         if (handle == 0)
             return;
         glDeleteTextures(1, &handle);
-        OpenGLState::ResetTexture(handle);
+        OpenGLState::GetCurState().ResetTexture(handle).Apply();
         handle = 0;
     }
 
@@ -69,7 +69,7 @@ public:
         if (handle == 0)
             return;
         glDeleteSamplers(1, &handle);
-        OpenGLState::ResetSampler(handle);
+        OpenGLState::GetCurState().ResetSampler(handle).Apply();
         handle = 0;
     }
 
@@ -91,10 +91,13 @@ public:
     }
 
     /// Creates a new internal OpenGL resource and stores the handle
-    void Create(const char* vert_shader, const char* frag_shader) {
+    void Create(const char* vert_shader, const char* geo_shader, const char* frag_shader,
+                const std::vector<const char*>& feedback_vars = {},
+                bool separable_program = false) {
         if (handle != 0)
             return;
-        handle = GLShader::LoadProgram(vert_shader, frag_shader);
+        handle = GLShader::LoadProgram(vert_shader, geo_shader, frag_shader, feedback_vars,
+                                       separable_program);
     }
 
     /// Deletes the internal OpenGL resource
@@ -102,7 +105,40 @@ public:
         if (handle == 0)
             return;
         glDeleteProgram(handle);
-        OpenGLState::ResetProgram(handle);
+        OpenGLState::GetCurState().ResetProgram(handle).Apply();
+        handle = 0;
+    }
+
+    GLuint handle = 0;
+};
+
+class OGLPipeline : private NonCopyable {
+public:
+    OGLPipeline() = default;
+    OGLPipeline(OGLPipeline&& o) {
+        handle = std::exchange<GLuint>(o.handle, 0);
+    }
+    ~OGLPipeline() {
+        Release();
+    }
+    OGLPipeline& operator=(OGLPipeline&& o) {
+        handle = std::exchange<GLuint>(o.handle, 0);
+        return *this;
+    }
+
+    /// Creates a new internal OpenGL resource and stores the handle
+    void Create() {
+        if (handle != 0)
+            return;
+        glGenProgramPipelines(1, &handle);
+    }
+
+    /// Deletes the internal OpenGL resource
+    void Release() {
+        if (handle == 0)
+            return;
+        glDeleteProgramPipelines(1, &handle);
+        OpenGLState::GetCurState().ResetPipeline(handle).Apply();
         handle = 0;
     }
 
@@ -135,13 +171,46 @@ public:
         if (handle == 0)
             return;
         glDeleteBuffers(1, &handle);
-        OpenGLState::ResetBuffer(handle);
+        OpenGLState::GetCurState().ResetBuffer(handle).Apply();
         handle = 0;
     }
 
     GLuint handle = 0;
 };
 
+class OGLSync : private NonCopyable {
+public:
+    OGLSync() = default;
+
+    OGLSync(OGLSync&& o) : handle(std::exchange(o.handle, nullptr)) {}
+
+    ~OGLSync() {
+        Release();
+    }
+    OGLSync& operator=(OGLSync&& o) {
+        Release();
+        handle = std::exchange(o.handle, nullptr);
+        return *this;
+    }
+
+    /// Creates a new internal OpenGL resource and stores the handle
+    void Create() {
+        if (handle != 0)
+            return;
+        handle = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+    }
+
+    /// Deletes the internal OpenGL resource
+    void Release() {
+        if (handle == 0)
+            return;
+        glDeleteSync(handle);
+        handle = 0;
+    }
+
+    GLsync handle = 0;
+};
+
 class OGLVertexArray : private NonCopyable {
 public:
     OGLVertexArray() = default;
@@ -168,7 +237,7 @@ public:
         if (handle == 0)
             return;
         glDeleteVertexArrays(1, &handle);
-        OpenGLState::ResetVertexArray(handle);
+        OpenGLState::GetCurState().ResetVertexArray(handle).Apply();
         handle = 0;
     }
 
@@ -201,7 +270,7 @@ public:
         if (handle == 0)
             return;
         glDeleteFramebuffers(1, &handle);
-        OpenGLState::ResetFramebuffer(handle);
+        OpenGLState::GetCurState().ResetFramebuffer(handle).Apply();
         handle = 0;
     }
 

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -0,0 +1,58 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <string>
+#include <queue>
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/renderer_opengl/gl_shader_decompiler.h"
+
+namespace Maxwell3D {
+namespace Shader {
+namespace Decompiler {
+
+constexpr u32 PROGRAM_END = MAX_PROGRAM_CODE_LENGTH;
+
+class Impl {
+public:
+    Impl(const std::array<u32, MAX_PROGRAM_CODE_LENGTH>& program_code,
+         const std::array<u32, MAX_SWIZZLE_DATA_LENGTH>& swizzle_data, u32 main_offset,
+         const std::function<std::string(u32)>& inputreg_getter,
+         const std::function<std::string(u32)>& outputreg_getter, bool sanitize_mul,
+         const std::string& emit_cb, const std::string& setemit_cb)
+        : program_code(program_code), swizzle_data(swizzle_data), main_offset(main_offset),
+          inputreg_getter(inputreg_getter), outputreg_getter(outputreg_getter),
+          sanitize_mul(sanitize_mul), emit_cb(emit_cb), setemit_cb(setemit_cb) {}
+
+    std::string Decompile() {
+        UNIMPLEMENTED();
+        return {};
+    }
+
+private:
+    const std::array<u32, MAX_PROGRAM_CODE_LENGTH>& program_code;
+    const std::array<u32, MAX_SWIZZLE_DATA_LENGTH>& swizzle_data;
+    u32 main_offset;
+    const std::function<std::string(u32)>& inputreg_getter;
+    const std::function<std::string(u32)>& outputreg_getter;
+    bool sanitize_mul;
+    const std::string& emit_cb;
+    const std::string& setemit_cb;
+};
+
+std::string DecompileProgram(const std::array<u32, MAX_PROGRAM_CODE_LENGTH>& program_code,
+                             const std::array<u32, MAX_SWIZZLE_DATA_LENGTH>& swizzle_data,
+                             u32 main_offset,
+                             const std::function<std::string(u32)>& inputreg_getter,
+                             const std::function<std::string(u32)>& outputreg_getter,
+                             bool sanitize_mul, const std::string& emit_cb,
+                             const std::string& setemit_cb) {
+    Impl impl(program_code, swizzle_data, main_offset, inputreg_getter, outputreg_getter,
+              sanitize_mul, emit_cb, setemit_cb);
+    return impl.Decompile();
+}
+
+} // namespace Decompiler
+} // namespace Shader
+} // namespace Maxwell3D

src/video_core/renderer_opengl/gl_shader_decompiler.h

@@ -0,0 +1,27 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <functional>
+#include <string>
+#include "common/common_types.h"
+
+namespace Maxwell3D {
+namespace Shader {
+namespace Decompiler {
+
+constexpr size_t MAX_PROGRAM_CODE_LENGTH{0x100000};
+constexpr size_t MAX_SWIZZLE_DATA_LENGTH{0x100000};
+
+std::string DecompileProgram(const std::array<u32, MAX_PROGRAM_CODE_LENGTH>& program_code,
+                             const std::array<u32, MAX_SWIZZLE_DATA_LENGTH>& swizzle_data,
+                             u32 main_offset,
+                             const std::function<std::string(u32)>& inputreg_getter,
+                             const std::function<std::string(u32)>& outputreg_getter,
+                             bool sanitize_mul, const std::string& emit_cb = "",
+                             const std::string& setemit_cb = "");
+
+} // namespace Decompiler
+} // namespace Shader
+} // namespace Maxwell3D

src/video_core/renderer_opengl/gl_shader_gen.cpp

@@ -0,0 +1,20 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "video_core/renderer_opengl/gl_shader_gen.h"
+
+namespace GLShader {
+
+std::string GenerateVertexShader(const MaxwellVSConfig& config) {
+    UNIMPLEMENTED();
+    return {};
+}
+
+std::string GenerateFragmentShader(const MaxwellFSConfig& config) {
+    UNIMPLEMENTED();
+    return {};
+}
+
+} // namespace GLShader

src/video_core/renderer_opengl/gl_shader_gen.h

@@ -0,0 +1,66 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstring>
+#include <string>
+#include <type_traits>
+#include "common/hash.h"
+
+namespace GLShader {
+
+enum Attributes {
+    ATTRIBUTE_POSITION,
+    ATTRIBUTE_COLOR,
+    ATTRIBUTE_TEXCOORD0,
+    ATTRIBUTE_TEXCOORD1,
+    ATTRIBUTE_TEXCOORD2,
+    ATTRIBUTE_TEXCOORD0_W,
+    ATTRIBUTE_NORMQUAT,
+    ATTRIBUTE_VIEW,
+};
+
+struct MaxwellShaderConfigCommon {
+    explicit MaxwellShaderConfigCommon(){};
+};
+
+struct MaxwellVSConfig : MaxwellShaderConfigCommon {
+    explicit MaxwellVSConfig() : MaxwellShaderConfigCommon() {}
+
+    bool operator==(const MaxwellVSConfig& o) const {
+        return std::memcmp(this, &o, sizeof(MaxwellVSConfig)) == 0;
+    };
+};
+
+struct MaxwellFSConfig : MaxwellShaderConfigCommon {
+    explicit MaxwellFSConfig() : MaxwellShaderConfigCommon() {}
+
+    bool operator==(const MaxwellFSConfig& o) const {
+        return std::memcmp(this, &o, sizeof(MaxwellFSConfig)) == 0;
+    };
+};
+
+std::string GenerateVertexShader(const MaxwellVSConfig& config);
+std::string GenerateFragmentShader(const MaxwellFSConfig& config);
+
+} // namespace GLShader
+
+namespace std {
+
+template <>
+struct hash<GLShader::MaxwellVSConfig> {
+    size_t operator()(const GLShader::MaxwellVSConfig& k) const {
+        return Common::ComputeHash64(&k, sizeof(GLShader::MaxwellVSConfig));
+    }
+};
+
+template <>
+struct hash<GLShader::MaxwellFSConfig> {
+    size_t operator()(const GLShader::MaxwellFSConfig& k) const {
+        return Common::ComputeHash64(&k, sizeof(GLShader::MaxwellFSConfig));
+    }
+};
+
+} // namespace std

src/video_core/renderer_opengl/gl_shader_util.cpp

@@ -10,15 +10,18 @@
 
 namespace GLShader {
 
-GLuint LoadProgram(const char* vertex_shader, const char* fragment_shader) {
-
+GLuint LoadProgram(const char* vertex_shader, const char* geometry_shader,
+                   const char* fragment_shader, const std::vector<const char*>& feedback_vars,
+                   bool separable_program) {
     // Create the shaders
-    GLuint vertex_shader_id = glCreateShader(GL_VERTEX_SHADER);
-    GLuint fragment_shader_id = glCreateShader(GL_FRAGMENT_SHADER);
+    GLuint vertex_shader_id = vertex_shader ? glCreateShader(GL_VERTEX_SHADER) : 0;
+    GLuint geometry_shader_id = geometry_shader ? glCreateShader(GL_GEOMETRY_SHADER) : 0;
+    GLuint fragment_shader_id = fragment_shader ? glCreateShader(GL_FRAGMENT_SHADER) : 0;
 
     GLint result = GL_FALSE;
     int info_log_length;
 
+    if (vertex_shader) {
         // Compile Vertex Shader
         LOG_DEBUG(Render_OpenGL, "Compiling vertex shader...");
 
@@ -35,10 +38,37 @@ GLuint LoadProgram(const char* vertex_shader, const char* fragment_shader) {
             if (result == GL_TRUE) {
                 LOG_DEBUG(Render_OpenGL, "%s", &vertex_shader_error[0]);
             } else {
-            LOG_ERROR(Render_OpenGL, "Error compiling vertex shader:\n%s", &vertex_shader_error[0]);
+                LOG_ERROR(Render_OpenGL, "Error compiling vertex shader:\n%s",
+                          &vertex_shader_error[0]);
+            }
         }
     }
 
+    if (geometry_shader) {
+        // Compile Geometry Shader
+        LOG_DEBUG(Render_OpenGL, "Compiling geometry shader...");
+
+        glShaderSource(geometry_shader_id, 1, &geometry_shader, nullptr);
+        glCompileShader(geometry_shader_id);
+
+        // Check Geometry Shader
+        glGetShaderiv(geometry_shader_id, GL_COMPILE_STATUS, &result);
+        glGetShaderiv(geometry_shader_id, GL_INFO_LOG_LENGTH, &info_log_length);
+
+        if (info_log_length > 1) {
+            std::vector<char> geometry_shader_error(info_log_length);
+            glGetShaderInfoLog(geometry_shader_id, info_log_length, nullptr,
+                               &geometry_shader_error[0]);
+            if (result == GL_TRUE) {
+                LOG_DEBUG(Render_OpenGL, "%s", &geometry_shader_error[0]);
+            } else {
+                LOG_ERROR(Render_OpenGL, "Error compiling geometry shader:\n%s",
+                          &geometry_shader_error[0]);
+            }
+        }
+    }
+
+    if (fragment_shader) {
         // Compile Fragment Shader
         LOG_DEBUG(Render_OpenGL, "Compiling fragment shader...");
 
@@ -51,7 +81,8 @@ GLuint LoadProgram(const char* vertex_shader, const char* fragment_shader) {
 
         if (info_log_length > 1) {
             std::vector<char> fragment_shader_error(info_log_length);
-        glGetShaderInfoLog(fragment_shader_id, info_log_length, nullptr, &fragment_shader_error[0]);
+            glGetShaderInfoLog(fragment_shader_id, info_log_length, nullptr,
+                               &fragment_shader_error[0]);
             if (result == GL_TRUE) {
                 LOG_DEBUG(Render_OpenGL, "%s", &fragment_shader_error[0]);
             } else {
@@ -59,13 +90,31 @@ GLuint LoadProgram(const char* vertex_shader, const char* fragment_shader) {
                           &fragment_shader_error[0]);
             }
         }
+    }
 
     // Link the program
     LOG_DEBUG(Render_OpenGL, "Linking program...");
 
     GLuint program_id = glCreateProgram();
+    if (vertex_shader) {
         glAttachShader(program_id, vertex_shader_id);
+    }
+    if (geometry_shader) {
+        glAttachShader(program_id, geometry_shader_id);
+    }
+    if (fragment_shader) {
         glAttachShader(program_id, fragment_shader_id);
+    }
+
+    if (!feedback_vars.empty()) {
+        auto varyings = feedback_vars;
+        glTransformFeedbackVaryings(program_id, static_cast<GLsizei>(feedback_vars.size()),
+                                    &varyings[0], GL_INTERLEAVED_ATTRIBS);
+    }
+
+    if (separable_program) {
+        glProgramParameteri(program_id, GL_PROGRAM_SEPARABLE, GL_TRUE);
+    }
 
     glLinkProgram(program_id);
 
@@ -85,13 +134,30 @@ GLuint LoadProgram(const char* vertex_shader, const char* fragment_shader) {
 
     // If the program linking failed at least one of the shaders was probably bad
     if (result == GL_FALSE) {
+        if (vertex_shader) {
             LOG_ERROR(Render_OpenGL, "Vertex shader:\n%s", vertex_shader);
+        }
+        if (geometry_shader) {
+            LOG_ERROR(Render_OpenGL, "Geometry shader:\n%s", geometry_shader);
+        }
+        if (fragment_shader) {
             LOG_ERROR(Render_OpenGL, "Fragment shader:\n%s", fragment_shader);
         }
+    }
     ASSERT_MSG(result == GL_TRUE, "Shader not linked");
 
+    if (vertex_shader) {
+        glDetachShader(program_id, vertex_shader_id);
         glDeleteShader(vertex_shader_id);
+    }
+    if (geometry_shader) {
+        glDetachShader(program_id, geometry_shader_id);
+        glDeleteShader(geometry_shader_id);
+    }
+    if (fragment_shader) {
+        glDetachShader(program_id, fragment_shader_id);
         glDeleteShader(fragment_shader_id);
+    }
 
     return program_id;
 }

src/video_core/renderer_opengl/gl_shader_util.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <vector>
 #include <glad/glad.h>
 
 namespace GLShader {
@@ -11,9 +12,12 @@ namespace GLShader {
 /**
  * Utility function to create and compile an OpenGL GLSL shader program (vertex + fragment shader)
  * @param vertex_shader String of the GLSL vertex shader program
+ * @param geometry_shader String of the GLSL geometry shader program
  * @param fragment_shader String of the GLSL fragment shader program
  * @returns Handle of the newly created OpenGL shader object
  */
-GLuint LoadProgram(const char* vertex_shader, const char* fragment_shader);
+GLuint LoadProgram(const char* vertex_shader, const char* geometry_shader,
+                   const char* fragment_shader, const std::vector<const char*>& feedback_vars = {},
+                   bool separable_program = false);
 
 } // namespace GLShader

src/video_core/renderer_opengl/gl_state.cpp

@@ -33,7 +33,7 @@ OpenGLState::OpenGLState() {
     stencil.action_depth_pass = GL_KEEP;
     stencil.action_stencil_fail = GL_KEEP;
 
-    blend.enabled = false;
+    blend.enabled = true;
     blend.rgb_equation = GL_FUNC_ADD;
     blend.a_equation = GL_FUNC_ADD;
     blend.src_rgb_func = GL_ONE;
@@ -68,6 +68,18 @@ OpenGLState::OpenGLState() {
     draw.vertex_buffer = 0;
     draw.uniform_buffer = 0;
     draw.shader_program = 0;
+    draw.program_pipeline = 0;
+
+    scissor.enabled = false;
+    scissor.x = 0;
+    scissor.y = 0;
+    scissor.width = 0;
+    scissor.height = 0;
+
+    viewport.x = 0;
+    viewport.y = 0;
+    viewport.width = 0;
+    viewport.height = 0;
 
     clip_distance = {};
 }
@@ -148,9 +160,6 @@ void OpenGLState::Apply() const {
     if (blend.enabled != cur_state.blend.enabled) {
         if (blend.enabled) {
             glEnable(GL_BLEND);
-
-            cur_state.logic_op = GL_COPY;
-            glLogicOp(cur_state.logic_op);
             glDisable(GL_COLOR_LOGIC_OP);
         } else {
             glDisable(GL_BLEND);
@@ -196,7 +205,7 @@ void OpenGLState::Apply() const {
     // Lighting LUTs
     if (lighting_lut.texture_buffer != cur_state.lighting_lut.texture_buffer) {
         glActiveTexture(TextureUnits::LightingLUT.Enum());
-        glBindTexture(GL_TEXTURE_BUFFER, cur_state.lighting_lut.texture_buffer);
+        glBindTexture(GL_TEXTURE_BUFFER, lighting_lut.texture_buffer);
     }
 
     // Fog LUT
@@ -263,6 +272,31 @@ void OpenGLState::Apply() const {
         glUseProgram(draw.shader_program);
     }
 
+    // Program pipeline
+    if (draw.program_pipeline != cur_state.draw.program_pipeline) {
+        glBindProgramPipeline(draw.program_pipeline);
+    }
+
+    // Scissor test
+    if (scissor.enabled != cur_state.scissor.enabled) {
+        if (scissor.enabled) {
+            glEnable(GL_SCISSOR_TEST);
+        } else {
+            glDisable(GL_SCISSOR_TEST);
+        }
+    }
+
+    if (scissor.x != cur_state.scissor.x || scissor.y != cur_state.scissor.y ||
+        scissor.width != cur_state.scissor.width || scissor.height != cur_state.scissor.height) {
+        glScissor(scissor.x, scissor.y, scissor.width, scissor.height);
+    }
+
+    if (viewport.x != cur_state.viewport.x || viewport.y != cur_state.viewport.y ||
+        viewport.width != cur_state.viewport.width ||
+        viewport.height != cur_state.viewport.height) {
+        glViewport(viewport.x, viewport.y, viewport.width, viewport.height);
+    }
+
     // Clip distance
     for (size_t i = 0; i < clip_distance.size(); ++i) {
         if (clip_distance[i] != cur_state.clip_distance[i]) {
@@ -277,62 +311,75 @@ void OpenGLState::Apply() const {
     cur_state = *this;
 }
 
-void OpenGLState::ResetTexture(GLuint handle) {
-    for (auto& unit : cur_state.texture_units) {
+OpenGLState& OpenGLState::ResetTexture(GLuint handle) {
+    for (auto& unit : texture_units) {
         if (unit.texture_2d == handle) {
             unit.texture_2d = 0;
         }
     }
-    if (cur_state.lighting_lut.texture_buffer == handle)
-        cur_state.lighting_lut.texture_buffer = 0;
-    if (cur_state.fog_lut.texture_buffer == handle)
-        cur_state.fog_lut.texture_buffer = 0;
-    if (cur_state.proctex_noise_lut.texture_buffer == handle)
-        cur_state.proctex_noise_lut.texture_buffer = 0;
-    if (cur_state.proctex_color_map.texture_buffer == handle)
-        cur_state.proctex_color_map.texture_buffer = 0;
-    if (cur_state.proctex_alpha_map.texture_buffer == handle)
-        cur_state.proctex_alpha_map.texture_buffer = 0;
-    if (cur_state.proctex_lut.texture_buffer == handle)
-        cur_state.proctex_lut.texture_buffer = 0;
-    if (cur_state.proctex_diff_lut.texture_buffer == handle)
-        cur_state.proctex_diff_lut.texture_buffer = 0;
+    if (lighting_lut.texture_buffer == handle)
+        lighting_lut.texture_buffer = 0;
+    if (fog_lut.texture_buffer == handle)
+        fog_lut.texture_buffer = 0;
+    if (proctex_noise_lut.texture_buffer == handle)
+        proctex_noise_lut.texture_buffer = 0;
+    if (proctex_color_map.texture_buffer == handle)
+        proctex_color_map.texture_buffer = 0;
+    if (proctex_alpha_map.texture_buffer == handle)
+        proctex_alpha_map.texture_buffer = 0;
+    if (proctex_lut.texture_buffer == handle)
+        proctex_lut.texture_buffer = 0;
+    if (proctex_diff_lut.texture_buffer == handle)
+        proctex_diff_lut.texture_buffer = 0;
+    return *this;
 }
 
-void OpenGLState::ResetSampler(GLuint handle) {
-    for (auto& unit : cur_state.texture_units) {
+OpenGLState& OpenGLState::ResetSampler(GLuint handle) {
+    for (auto& unit : texture_units) {
         if (unit.sampler == handle) {
             unit.sampler = 0;
         }
     }
+    return *this;
 }
 
-void OpenGLState::ResetProgram(GLuint handle) {
-    if (cur_state.draw.shader_program == handle) {
-        cur_state.draw.shader_program = 0;
+OpenGLState& OpenGLState::ResetProgram(GLuint handle) {
+    if (draw.shader_program == handle) {
+        draw.shader_program = 0;
     }
+    return *this;
 }
 
-void OpenGLState::ResetBuffer(GLuint handle) {
-    if (cur_state.draw.vertex_buffer == handle) {
-        cur_state.draw.vertex_buffer = 0;
-    }
-    if (cur_state.draw.uniform_buffer == handle) {
-        cur_state.draw.uniform_buffer = 0;
+OpenGLState& OpenGLState::ResetPipeline(GLuint handle) {
+    if (draw.program_pipeline == handle) {
+        draw.program_pipeline = 0;
     }
+    return *this;
 }
 
-void OpenGLState::ResetVertexArray(GLuint handle) {
-    if (cur_state.draw.vertex_array == handle) {
-        cur_state.draw.vertex_array = 0;
+OpenGLState& OpenGLState::ResetBuffer(GLuint handle) {
+    if (draw.vertex_buffer == handle) {
+        draw.vertex_buffer = 0;
     }
+    if (draw.uniform_buffer == handle) {
+        draw.uniform_buffer = 0;
+    }
+    return *this;
 }
 
-void OpenGLState::ResetFramebuffer(GLuint handle) {
-    if (cur_state.draw.read_framebuffer == handle) {
-        cur_state.draw.read_framebuffer = 0;
-    }
-    if (cur_state.draw.draw_framebuffer == handle) {
-        cur_state.draw.draw_framebuffer = 0;
+OpenGLState& OpenGLState::ResetVertexArray(GLuint handle) {
+    if (draw.vertex_array == handle) {
+        draw.vertex_array = 0;
     }
+    return *this;
+}
+
+OpenGLState& OpenGLState::ResetFramebuffer(GLuint handle) {
+    if (draw.read_framebuffer == handle) {
+        draw.read_framebuffer = 0;
+    }
+    if (draw.draw_framebuffer == handle) {
+        draw.draw_framebuffer = 0;
+    }
+    return *this;
 }

src/video_core/renderer_opengl/gl_state.h

@@ -122,27 +122,44 @@ public:
         GLuint vertex_buffer;    // GL_ARRAY_BUFFER_BINDING
         GLuint uniform_buffer;   // GL_UNIFORM_BUFFER_BINDING
         GLuint shader_program;   // GL_CURRENT_PROGRAM
+        GLuint program_pipeline; // GL_PROGRAM_PIPELINE_BINDING
     } draw;
 
+    struct {
+        bool enabled; // GL_SCISSOR_TEST
+        GLint x;
+        GLint y;
+        GLsizei width;
+        GLsizei height;
+    } scissor;
+
+    struct {
+        GLint x;
+        GLint y;
+        GLsizei width;
+        GLsizei height;
+    } viewport;
+
     std::array<bool, 2> clip_distance; // GL_CLIP_DISTANCE
 
     OpenGLState();
 
     /// Get the currently active OpenGL state
-    static const OpenGLState& GetCurState() {
+    static OpenGLState GetCurState() {
         return cur_state;
     }
 
     /// Apply this state as the current OpenGL state
     void Apply() const;
 
-    /// Resets and unbinds any references to the given resource in the current OpenGL state
-    static void ResetTexture(GLuint handle);
-    static void ResetSampler(GLuint handle);
-    static void ResetProgram(GLuint handle);
-    static void ResetBuffer(GLuint handle);
-    static void ResetVertexArray(GLuint handle);
-    static void ResetFramebuffer(GLuint handle);
+    /// Resets any references to the given resource
+    OpenGLState& ResetTexture(GLuint handle);
+    OpenGLState& ResetSampler(GLuint handle);
+    OpenGLState& ResetProgram(GLuint handle);
+    OpenGLState& ResetPipeline(GLuint handle);
+    OpenGLState& ResetBuffer(GLuint handle);
+    OpenGLState& ResetVertexArray(GLuint handle);
+    OpenGLState& ResetFramebuffer(GLuint handle);
 
 private:
     static OpenGLState cur_state;

src/video_core/renderer_opengl/gl_stream_buffer.cpp

@@ -0,0 +1,182 @@
+// Copyright 2018 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 "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/gl_stream_buffer.h"
+
+class OrphanBuffer : public OGLStreamBuffer {
+public:
+    explicit OrphanBuffer(GLenum target) : OGLStreamBuffer(target) {}
+    ~OrphanBuffer() override;
+
+private:
+    void Create(size_t size, size_t sync_subdivide) override;
+    void Release() override;
+
+    std::pair<u8*, GLintptr> Map(size_t size, size_t alignment) override;
+    void Unmap() override;
+
+    std::vector<u8> data;
+};
+
+class StorageBuffer : public OGLStreamBuffer {
+public:
+    explicit StorageBuffer(GLenum target) : OGLStreamBuffer(target) {}
+    ~StorageBuffer() override;
+
+private:
+    void Create(size_t size, size_t sync_subdivide) override;
+    void Release() override;
+
+    std::pair<u8*, GLintptr> Map(size_t size, size_t alignment) override;
+    void Unmap() override;
+
+    struct Fence {
+        OGLSync sync;
+        size_t offset;
+    };
+    std::deque<Fence> head;
+    std::deque<Fence> tail;
+
+    u8* mapped_ptr;
+};
+
+OGLStreamBuffer::OGLStreamBuffer(GLenum target) {
+    gl_target = target;
+}
+
+GLuint OGLStreamBuffer::GetHandle() const {
+    return gl_buffer.handle;
+}
+
+std::unique_ptr<OGLStreamBuffer> OGLStreamBuffer::MakeBuffer(bool storage_buffer, GLenum target) {
+    if (storage_buffer) {
+        return std::make_unique<StorageBuffer>(target);
+    }
+    return std::make_unique<OrphanBuffer>(target);
+}
+
+OrphanBuffer::~OrphanBuffer() {
+    Release();
+}
+
+void OrphanBuffer::Create(size_t size, size_t /*sync_subdivide*/) {
+    buffer_pos = 0;
+    buffer_size = size;
+    data.resize(buffer_size);
+
+    if (gl_buffer.handle == 0) {
+        gl_buffer.Create();
+        glBindBuffer(gl_target, gl_buffer.handle);
+    }
+
+    glBufferData(gl_target, static_cast<GLsizeiptr>(buffer_size), nullptr, GL_STREAM_DRAW);
+}
+
+void OrphanBuffer::Release() {
+    gl_buffer.Release();
+}
+
+std::pair<u8*, GLintptr> OrphanBuffer::Map(size_t size, size_t alignment) {
+    buffer_pos = Common::AlignUp(buffer_pos, alignment);
+
+    if (buffer_pos + size > buffer_size) {
+        Create(std::max(buffer_size, size), 0);
+    }
+
+    mapped_size = size;
+    return std::make_pair(&data[buffer_pos], static_cast<GLintptr>(buffer_pos));
+}
+
+void OrphanBuffer::Unmap() {
+    glBufferSubData(gl_target, static_cast<GLintptr>(buffer_pos),
+                    static_cast<GLsizeiptr>(mapped_size), &data[buffer_pos]);
+    buffer_pos += mapped_size;
+}
+
+StorageBuffer::~StorageBuffer() {
+    Release();
+}
+
+void StorageBuffer::Create(size_t size, size_t sync_subdivide) {
+    if (gl_buffer.handle != 0)
+        return;
+
+    buffer_pos = 0;
+    buffer_size = size;
+    buffer_sync_subdivide = std::max<size_t>(sync_subdivide, 1);
+
+    gl_buffer.Create();
+    glBindBuffer(gl_target, gl_buffer.handle);
+
+    glBufferStorage(gl_target, static_cast<GLsizeiptr>(buffer_size), nullptr,
+                    GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT);
+    mapped_ptr = reinterpret_cast<u8*>(
+        glMapBufferRange(gl_target, 0, static_cast<GLsizeiptr>(buffer_size),
+                         GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | GL_MAP_FLUSH_EXPLICIT_BIT));
+}
+
+void StorageBuffer::Release() {
+    if (gl_buffer.handle == 0)
+        return;
+
+    glUnmapBuffer(gl_target);
+
+    gl_buffer.Release();
+    head.clear();
+    tail.clear();
+}
+
+std::pair<u8*, GLintptr> StorageBuffer::Map(size_t size, size_t alignment) {
+    ASSERT(size <= buffer_size);
+
+    OGLSync sync;
+
+    buffer_pos = Common::AlignUp(buffer_pos, alignment);
+    size_t effective_offset = Common::AlignDown(buffer_pos, buffer_sync_subdivide);
+
+    if (!head.empty() &&
+        (effective_offset > head.back().offset || buffer_pos + size > buffer_size)) {
+        ASSERT(head.back().sync.handle == 0);
+        head.back().sync.Create();
+    }
+
+    if (buffer_pos + size > buffer_size) {
+        if (!tail.empty()) {
+            std::swap(sync, tail.back().sync);
+            tail.clear();
+        }
+        std::swap(tail, head);
+        buffer_pos = 0;
+        effective_offset = 0;
+    }
+
+    while (!tail.empty() && buffer_pos + size > tail.front().offset) {
+        std::swap(sync, tail.front().sync);
+        tail.pop_front();
+    }
+
+    if (sync.handle != 0) {
+        glClientWaitSync(sync.handle, GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED);
+        sync.Release();
+    }
+
+    if (head.empty() || effective_offset > head.back().offset) {
+        head.emplace_back();
+        head.back().offset = effective_offset;
+    }
+
+    mapped_size = size;
+    return std::make_pair(&mapped_ptr[buffer_pos], static_cast<GLintptr>(buffer_pos));
+}
+
+void StorageBuffer::Unmap() {
+    glFlushMappedBufferRange(gl_target, static_cast<GLintptr>(buffer_pos),
+                             static_cast<GLsizeiptr>(mapped_size));
+    buffer_pos += mapped_size;
+}

src/video_core/renderer_opengl/gl_stream_buffer.h

@@ -0,0 +1,34 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <memory>
+#include <glad/glad.h>
+#include "common/common_types.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+
+class OGLStreamBuffer : private NonCopyable {
+public:
+    explicit OGLStreamBuffer(GLenum target);
+    virtual ~OGLStreamBuffer() = default;
+
+public:
+    static std::unique_ptr<OGLStreamBuffer> MakeBuffer(bool storage_buffer, GLenum target);
+
+    virtual void Create(size_t size, size_t sync_subdivide) = 0;
+    virtual void Release() {}
+
+    GLuint GetHandle() const;
+
+    virtual std::pair<u8*, GLintptr> Map(size_t size, size_t alignment) = 0;
+    virtual void Unmap() = 0;
+
+protected:
+    OGLBuffer gl_buffer;
+    GLenum gl_target;
+
+    size_t buffer_pos = 0;
+    size_t buffer_size = 0;
+    size_t buffer_sync_subdivide = 0;
+    size_t mapped_size = 0;
+};

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -318,7 +318,7 @@ void RendererOpenGL::InitOpenGLObjects() {
                  0.0f);
 
     // Link shaders and get variable locations
-    shader.Create(vertex_shader, fragment_shader);
+    shader.Create(vertex_shader, nullptr, fragment_shader);
     state.draw.shader_program = shader.handle;
     state.Apply();
     uniform_modelview_matrix = glGetUniformLocation(shader.handle, "modelview_matrix");