6 files changed, 137 insertions, 70 deletions

diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index 1d1581f49..65a88b06c 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -2,11 +2,14 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <array>
 #include <cstddef>
 #include <glad/glad.h>
 
 #include "common/logging/log.h"
+#include "common/scope_exit.h"
 #include "video_core/renderer_opengl/gl_device.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
 
 namespace OpenGL {
 
@@ -24,6 +27,7 @@ Device::Device() {
     max_vertex_attributes = GetInteger<u32>(GL_MAX_VERTEX_ATTRIBS);
     max_varyings = GetInteger<u32>(GL_MAX_VARYING_VECTORS);
     has_variable_aoffi = TestVariableAoffi();
+    has_component_indexing_bug = TestComponentIndexingBug();
 }
 
 Device::Device(std::nullptr_t) {
@@ -31,6 +35,7 @@ Device::Device(std::nullptr_t) {
     max_vertex_attributes = 16;
     max_varyings = 15;
     has_variable_aoffi = true;
+    has_component_indexing_bug = false;
 }
 
 bool Device::TestVariableAoffi() {
@@ -52,4 +57,53 @@ void main() {
     return supported;
 }
 
+bool Device::TestComponentIndexingBug() {
+    constexpr char log_message[] = "Renderer_ComponentIndexingBug: {}";
+    const GLchar* COMPONENT_TEST = R"(#version 430 core
+layout (std430, binding = 0) buffer OutputBuffer {
+    uint output_value;
+};
+layout (std140, binding = 0) uniform InputBuffer {
+    uvec4 input_value[4096];
+};
+layout (location = 0) uniform uint idx;
+void main() {
+    output_value = input_value[idx >> 2][idx & 3];
+})";
+    const GLuint shader{glCreateShaderProgramv(GL_VERTEX_SHADER, 1, &COMPONENT_TEST)};
+    SCOPE_EXIT({ glDeleteProgram(shader); });
+    glUseProgram(shader);
+
+    OGLVertexArray vao;
+    vao.Create();
+    glBindVertexArray(vao.handle);
+
+    constexpr std::array<GLuint, 8> values{0, 0, 0, 0, 0x1236327, 0x985482, 0x872753, 0x2378432};
+    OGLBuffer ubo;
+    ubo.Create();
+    glNamedBufferData(ubo.handle, sizeof(values), values.data(), GL_STATIC_DRAW);
+    glBindBufferBase(GL_UNIFORM_BUFFER, 0, ubo.handle);
+
+    OGLBuffer ssbo;
+    ssbo.Create();
+    glNamedBufferStorage(ssbo.handle, sizeof(GLuint), nullptr, GL_CLIENT_STORAGE_BIT);
+
+    for (GLuint index = 4; index < 8; ++index) {
+        glInvalidateBufferData(ssbo.handle);
+        glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 0, ssbo.handle);
+
+        glProgramUniform1ui(shader, 0, index);
+        glDrawArrays(GL_POINTS, 0, 1);
+
+        GLuint result;
+        glGetNamedBufferSubData(ssbo.handle, 0, sizeof(result), &result);
+        if (result != values.at(index)) {
+            LOG_INFO(Render_OpenGL, log_message, true);
+            return true;
+        }
+    }
+    LOG_INFO(Render_OpenGL, log_message, false);
+    return false;
+}
+
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index de8490682..8c8c93760 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/src/video_core/renderer_opengl/gl_device.h
@@ -30,13 +30,19 @@ public:
         return has_variable_aoffi;
     }
 
+    bool HasComponentIndexingBug() const {
+        return has_component_indexing_bug;
+    }
+
 private:
     static bool TestVariableAoffi();
+    static bool TestComponentIndexingBug();
 
     std::size_t uniform_buffer_alignment{};
     u32 max_vertex_attributes{};
     u32 max_varyings{};
     bool has_variable_aoffi{};
+    bool has_component_indexing_bug{};
 };
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index d66252224..ac8a9e6b7 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -35,8 +35,8 @@ struct UnspecializedShader {
 namespace {
 
 /// Gets the address for the specified shader stage program
-GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) {
-    const auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
+GPUVAddr GetShaderAddress(Core::System& system, Maxwell::ShaderProgram program) {
+    const auto& gpu{system.GPU().Maxwell3D()};
     const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]};
     return gpu.regs.code_address.CodeAddress() + shader_config.offset;
 }
@@ -350,7 +350,8 @@ ShaderDiskCacheUsage CachedShader::GetUsage(GLenum primitive_mode,
 
 ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
                                      Core::Frontend::EmuWindow& emu_window, const Device& device)
-    : RasterizerCache{rasterizer}, emu_window{emu_window}, device{device}, disk_cache{system} {}
+    : RasterizerCache{rasterizer}, system{system}, emu_window{emu_window}, device{device},
+      disk_cache{system} {}
 
 void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
                                       const VideoCore::DiskResourceLoadCallback& callback) {
@@ -546,42 +547,45 @@ std::unordered_map<u64, UnspecializedShader> ShaderCacheOpenGL::GenerateUnspecia
 }
 
 Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
-    if (!Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.shaders) {
-        return last_shaders[static_cast<u32>(program)];
+    if (!system.GPU().Maxwell3D().dirty_flags.shaders) {
+        return last_shaders[static_cast<std::size_t>(program)];
     }
 
-    auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
-    const GPUVAddr program_addr{GetShaderAddress(program)};
+    auto& memory_manager{system.GPU().MemoryManager()};
+    const GPUVAddr program_addr{GetShaderAddress(system, program)};
 
     // Look up shader in the cache based on address
-    const auto& host_ptr{memory_manager.GetPointer(program_addr)};
+    const auto host_ptr{memory_manager.GetPointer(program_addr)};
     Shader shader{TryGet(host_ptr)};
+    if (shader) {
+        return last_shaders[static_cast<std::size_t>(program)] = shader;
+    }
 
-    if (!shader) {
-        // No shader found - create a new one
-        ProgramCode program_code{GetShaderCode(memory_manager, program_addr, host_ptr)};
-        ProgramCode program_code_b;
-        if (program == Maxwell::ShaderProgram::VertexA) {
-            const GPUVAddr program_addr_b{GetShaderAddress(Maxwell::ShaderProgram::VertexB)};
-            program_code_b = GetShaderCode(memory_manager, program_addr_b,
-                                           memory_manager.GetPointer(program_addr_b));
-        }
-        const u64 unique_identifier = GetUniqueIdentifier(program, program_code, program_code_b);
-        const VAddr cpu_addr{*memory_manager.GpuToCpuAddress(program_addr)};
-        const auto found = precompiled_shaders.find(unique_identifier);
-        if (found != precompiled_shaders.end()) {
-            shader =
-                std::make_shared<CachedShader>(cpu_addr, unique_identifier, program, disk_cache,
-                                               precompiled_programs, found->second, host_ptr);
-        } else {
-            shader = std::make_shared<CachedShader>(
-                device, cpu_addr, unique_identifier, program, disk_cache, precompiled_programs,
-                std::move(program_code), std::move(program_code_b), host_ptr);
-        }
-        Register(shader);
+    // No shader found - create a new one
+    ProgramCode program_code{GetShaderCode(memory_manager, program_addr, host_ptr)};
+    ProgramCode program_code_b;
+    if (program == Maxwell::ShaderProgram::VertexA) {
+        const GPUVAddr program_addr_b{GetShaderAddress(system, Maxwell::ShaderProgram::VertexB)};
+        program_code_b = GetShaderCode(memory_manager, program_addr_b,
+                                       memory_manager.GetPointer(program_addr_b));
+    }
+
+    const u64 unique_identifier = GetUniqueIdentifier(program, program_code, program_code_b);
+    const VAddr cpu_addr{*memory_manager.GpuToCpuAddress(program_addr)};
+    const auto found = precompiled_shaders.find(unique_identifier);
+    if (found != precompiled_shaders.end()) {
+        // Create a shader from the cache
+        shader = std::make_shared<CachedShader>(cpu_addr, unique_identifier, program, disk_cache,
+                                                precompiled_programs, found->second, host_ptr);
+    } else {
+        // Create a shader from guest memory
+        shader = std::make_shared<CachedShader>(
+            device, cpu_addr, unique_identifier, program, disk_cache, precompiled_programs,
+            std::move(program_code), std::move(program_code_b), host_ptr);
     }
+    Register(shader);
 
-    return last_shaders[static_cast<u32>(program)] = shader;
+    return last_shaders[static_cast<std::size_t>(program)] = shader;
 }
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index 64e5a5594..09bd0761d 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -137,6 +137,7 @@ private:
     CachedProgram GeneratePrecompiledProgram(const ShaderDiskCacheDump& dump,
                                              const std::set<GLenum>& supported_formats);
 
+    Core::System& system;
     Core::Frontend::EmuWindow& emu_window;
     const Device& device;
     ShaderDiskCacheOpenGL disk_cache;
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index e9f8d40db..29de5c9db 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -45,7 +45,6 @@ struct TextureAoffi {};
 using TextureArgument = std::pair<Type, Node>;
 using TextureIR = std::variant<TextureAoffi, TextureArgument>;
 
-enum : u32 { POSITION_VARYING_LOCATION = 0, GENERIC_VARYING_START_LOCATION = 1 };
 constexpr u32 MAX_CONSTBUFFER_ELEMENTS =
     static_cast<u32>(RasterizerOpenGL::MaxConstbufferSize) / (4 * sizeof(float));
 
@@ -247,6 +246,12 @@ private:
         code.AddLine("layout ({}, max_vertices = {}) out;", topology, max_vertices);
         code.AddNewLine();
 
+        code.AddLine("in gl_PerVertex {{");
+        ++code.scope;
+        code.AddLine("vec4 gl_Position;");
+        --code.scope;
+        code.AddLine("}} gl_in[];");
+
         DeclareVertexRedeclarations();
     }
 
@@ -349,7 +354,7 @@ private:
     }
 
     void DeclareInputAttribute(Attribute::Index index, bool skip_unused) {
-        const u32 generic_index{GetGenericAttributeIndex(index)};
+        const u32 location{GetGenericAttributeIndex(index)};
 
         std::string name{GetInputAttribute(index)};
         if (stage == ShaderStage::Geometry) {
@@ -358,19 +363,13 @@ private:
 
         std::string suffix;
         if (stage == ShaderStage::Fragment) {
-            const auto input_mode{header.ps.GetAttributeUse(generic_index)};
+            const auto input_mode{header.ps.GetAttributeUse(location)};
             if (skip_unused && input_mode == AttributeUse::Unused) {
                 return;
             }
             suffix = GetInputFlags(input_mode);
         }
 
-        u32 location = generic_index;
-        if (stage != ShaderStage::Vertex) {
-            // If inputs are varyings, add an offset
-            location += GENERIC_VARYING_START_LOCATION;
-        }
-
         code.AddLine("layout (location = {}) {} in vec4 {};", location, suffix, name);
     }
 
@@ -395,7 +394,7 @@ private:
     }
 
     void DeclareOutputAttribute(Attribute::Index index) {
-        const u32 location{GetGenericAttributeIndex(index) + GENERIC_VARYING_START_LOCATION};
+        const u32 location{GetGenericAttributeIndex(index)};
         code.AddLine("layout (location = {}) out vec4 {};", location, GetOutputAttribute(index));
     }
 
@@ -577,9 +576,26 @@ private:
             if (std::holds_alternative<OperationNode>(*offset)) {
                 // Indirect access
                 const std::string final_offset = code.GenerateTemporary();
-                code.AddLine("uint {} = (ftou({}) / 4);", final_offset, Visit(offset));
-                return fmt::format("{}[{} / 4][{} % 4]", GetConstBuffer(cbuf->GetIndex()),
-                                   final_offset, final_offset);
+                code.AddLine("uint {} = ftou({}) >> 2;", final_offset, Visit(offset));
+
+                if (!device.HasComponentIndexingBug()) {
+                    return fmt::format("{}[{} >> 2][{} & 3]", GetConstBuffer(cbuf->GetIndex()),
+                                       final_offset, final_offset);
+                }
+
+                // AMD's proprietary GLSL compiler emits ill code for variable component access.
+                // To bypass this driver bug generate 4 ifs, one per each component.
+                const std::string pack = code.GenerateTemporary();
+                code.AddLine("vec4 {} = {}[{} >> 2];", pack, GetConstBuffer(cbuf->GetIndex()),
+                             final_offset);
+
+                const std::string result = code.GenerateTemporary();
+                code.AddLine("float {};", result);
+                for (u32 swizzle = 0; swizzle < 4; ++swizzle) {
+                    code.AddLine("if (({} & 3) == {}) {} = {}{};", final_offset, swizzle, result,
+                                 pack, GetSwizzle(swizzle));
+                }
+                return result;
             }
 
             UNREACHABLE_MSG("Unmanaged offset node type");
@@ -633,10 +649,14 @@ private:
 
         switch (attribute) {
         case Attribute::Index::Position:
-            if (stage != ShaderStage::Fragment) {
-                return GeometryPass("position") + GetSwizzle(element);
-            } else {
+            switch (stage) {
+            case ShaderStage::Geometry:
+                return fmt::format("gl_in[ftou({})].gl_Position{}", Visit(buffer),
+                                   GetSwizzle(element));
+            case ShaderStage::Fragment:
                 return element == 3 ? "1.0f" : ("gl_FragCoord"s + GetSwizzle(element));
+            default:
+                UNREACHABLE();
             }
         case Attribute::Index::PointCoord:
             switch (element) {
@@ -921,7 +941,7 @@ private:
             target = [&]() -> std::string {
                 switch (const auto attribute = abuf->GetIndex(); abuf->GetIndex()) {
                 case Attribute::Index::Position:
-                    return "position"s + GetSwizzle(abuf->GetElement());
+                    return "gl_Position"s + GetSwizzle(abuf->GetElement());
                 case Attribute::Index::PointSize:
                     return "gl_PointSize";
                 case Attribute::Index::ClipDistances0123:
@@ -1506,9 +1526,7 @@ private:
 
         // If a geometry shader is attached, it will always flip (it's the last stage before
         // fragment). For more info about flipping, refer to gl_shader_gen.cpp.
-        code.AddLine("position.xy *= viewport_flip.xy;");
-        code.AddLine("gl_Position = position;");
-        code.AddLine("position.w = 1.0;");
+        code.AddLine("gl_Position.xy *= viewport_flip.xy;");
         code.AddLine("EmitVertex();");
         return {};
     }
@@ -1746,8 +1764,7 @@ private:
     }
 
     u32 GetNumPhysicalVaryings() const {
-        return std::min<u32>(device.GetMaxVaryings() - GENERIC_VARYING_START_LOCATION,
-                             Maxwell::NumVaryings);
+        return std::min<u32>(device.GetMaxVaryings(), Maxwell::NumVaryings);
     }
 
     const Device& device;
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp
index d2bb705a9..c845b29aa 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_gen.cpp
@@ -23,8 +23,6 @@ ProgramResult GenerateVertexShader(const Device& device, const ShaderSetup& setu
     out += GetCommonDeclarations();
 
     out += R"(
-layout (location = 0) out vec4 position;
-
 layout (std140, binding = EMULATION_UBO_BINDING) uniform vs_config {
     vec4 viewport_flip;
     uvec4 config_pack; // instance_id, flip_stage, y_direction, padding
@@ -48,7 +46,6 @@ layout (std140, binding = EMULATION_UBO_BINDING) uniform vs_config {
 
     out += R"(
 void main() {
-    position = vec4(0.0, 0.0, 0.0, 0.0);
     execute_vertex();
 )";
 
@@ -59,19 +56,12 @@ void main() {
     out += R"(
 
     // Set Position Y direction
-    position.y *= utof(config_pack[2]);
+    gl_Position.y *= utof(config_pack[2]);
     // Check if the flip stage is VertexB
     // Config pack's second value is flip_stage
     if (config_pack[1] == 1) {
         // Viewport can be flipped, which is unsupported by glViewport
-        position.xy *= viewport_flip.xy;
-    }
-    gl_Position = position;
-
-    // TODO(bunnei): This is likely a hack, position.w should be interpolated as 1.0
-    // For now, this is here to bring order in lieu of proper emulation
-    if (config_pack[1] == 1) {
-        position.w = 1.0;
+        gl_Position.xy *= viewport_flip.xy;
     }
 })";
 
@@ -85,9 +75,6 @@ ProgramResult GenerateGeometryShader(const Device& device, const ShaderSetup& se
     out += GetCommonDeclarations();
 
     out += R"(
-layout (location = 0) in vec4 gs_position[];
-layout (location = 0) out vec4 position;
-
 layout (std140, binding = EMULATION_UBO_BINDING) uniform gs_config {
     vec4 viewport_flip;
     uvec4 config_pack; // instance_id, flip_stage, y_direction, padding
@@ -124,8 +111,6 @@ layout (location = 5) out vec4 FragColor5;
 layout (location = 6) out vec4 FragColor6;
 layout (location = 7) out vec4 FragColor7;
 
-layout (location = 0) in noperspective vec4 position;
-
 layout (std140, binding = EMULATION_UBO_BINDING) uniform fs_config {
     vec4 viewport_flip;
     uvec4 config_pack; // instance_id, flip_stage, y_direction, padding