diff options
Diffstat (limited to 'src/video_core/renderer_opengl')
6 files changed, 151 insertions, 204 deletions
diff --git a/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp b/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp index 7c926bd48..a5d69d78d 100644 --- a/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp @@ -35,21 +35,16 @@ OGLFramebuffer FramebufferCacheOpenGL::CreateFramebuffer(const FramebufferCacheK local_state.draw.draw_framebuffer = framebuffer.handle; local_state.ApplyFramebufferState(); - if (key.is_single_buffer) { - if (key.color_attachments[0] != GL_NONE && key.colors[0]) { - key.colors[0]->Attach(key.color_attachments[0], GL_DRAW_FRAMEBUFFER); - glDrawBuffer(key.color_attachments[0]); - } else { - glDrawBuffer(GL_NONE); - } - } else { - for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) { - if (key.colors[index]) { - key.colors[index]->Attach(GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(index), - GL_DRAW_FRAMEBUFFER); - } + for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) { + if (key.colors[index]) { + key.colors[index]->Attach(GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(index), + GL_DRAW_FRAMEBUFFER); } + } + if (key.colors_count) { glDrawBuffers(key.colors_count, key.color_attachments.data()); + } else { + glDrawBuffer(GL_NONE); } if (key.zeta) { @@ -67,9 +62,9 @@ std::size_t FramebufferCacheKey::Hash() const { } bool FramebufferCacheKey::operator==(const FramebufferCacheKey& rhs) const { - return std::tie(is_single_buffer, stencil_enable, colors_count, color_attachments, colors, - zeta) == std::tie(rhs.is_single_buffer, rhs.stencil_enable, rhs.colors_count, - rhs.color_attachments, rhs.colors, rhs.zeta); + return std::tie(stencil_enable, colors_count, color_attachments, colors, zeta) == + std::tie(rhs.stencil_enable, rhs.colors_count, rhs.color_attachments, rhs.colors, + rhs.zeta); } } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_framebuffer_cache.h b/src/video_core/renderer_opengl/gl_framebuffer_cache.h index a3a996353..424344c48 100644 --- a/src/video_core/renderer_opengl/gl_framebuffer_cache.h +++ b/src/video_core/renderer_opengl/gl_framebuffer_cache.h @@ -19,7 +19,6 @@ namespace OpenGL { struct alignas(sizeof(u64)) FramebufferCacheKey { - bool is_single_buffer = false; bool stencil_enable = false; u16 colors_count = 0; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 4dd08bccb..6a17bed72 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -49,40 +49,6 @@ MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(128, 128, 192)); MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100)); MICROPROFILE_DEFINE(OpenGL_PrimitiveAssembly, "OpenGL", "Prim Asmbl", MP_RGB(255, 100, 100)); -struct DrawParameters { - GLenum primitive_mode; - GLsizei count; - GLint current_instance; - bool use_indexed; - - GLint vertex_first; - - GLenum index_format; - GLint base_vertex; - GLintptr index_buffer_offset; - - void DispatchDraw() const { - if (use_indexed) { - const auto index_buffer_ptr = reinterpret_cast<const void*>(index_buffer_offset); - if (current_instance > 0) { - glDrawElementsInstancedBaseVertexBaseInstance(primitive_mode, count, index_format, - index_buffer_ptr, 1, base_vertex, - current_instance); - } else { - glDrawElementsBaseVertex(primitive_mode, count, index_format, index_buffer_ptr, - base_vertex); - } - } else { - if (current_instance > 0) { - glDrawArraysInstancedBaseInstance(primitive_mode, vertex_first, count, 1, - current_instance); - } else { - glDrawArrays(primitive_mode, vertex_first, count); - } - } - } -}; - static std::size_t GetConstBufferSize(const Tegra::Engines::ConstBufferInfo& buffer, const GLShader::ConstBufferEntry& entry) { if (!entry.IsIndirect()) { @@ -270,29 +236,6 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() { return offset; } -DrawParameters RasterizerOpenGL::SetupDraw(GLintptr index_buffer_offset) { - const auto& gpu = system.GPU().Maxwell3D(); - const auto& regs = gpu.regs; - const bool is_indexed = accelerate_draw == AccelDraw::Indexed; - - DrawParameters params{}; - params.current_instance = gpu.state.current_instance; - - params.use_indexed = is_indexed; - params.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology); - - if (is_indexed) { - params.index_format = MaxwellToGL::IndexFormat(regs.index_array.format); - params.count = regs.index_array.count; - params.index_buffer_offset = index_buffer_offset; - params.base_vertex = static_cast<GLint>(regs.vb_element_base); - } else { - params.count = regs.vertex_buffer.count; - params.vertex_first = regs.vertex_buffer.first; - } - return params; -} - void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) { MICROPROFILE_SCOPE(OpenGL_Shader); auto& gpu = system.GPU().Maxwell3D(); @@ -399,12 +342,6 @@ std::size_t RasterizerOpenGL::CalculateIndexBufferSize() const { static_cast<std::size_t>(regs.index_array.FormatSizeInBytes()); } -bool RasterizerOpenGL::AccelerateDrawBatch(bool is_indexed) { - accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays; - DrawArrays(); - return true; -} - template <typename Map, typename Interval> static constexpr auto RangeFromInterval(Map& map, const Interval& interval) { return boost::make_iterator_range(map.equal_range(interval)); @@ -445,99 +382,51 @@ void RasterizerOpenGL::LoadDiskResources(const std::atomic_bool& stop_loading, shader_cache.LoadDiskCache(stop_loading, callback); } -std::pair<bool, bool> RasterizerOpenGL::ConfigureFramebuffers( - OpenGLState& current_state, bool using_color_fb, bool using_depth_fb, bool preserve_contents, - std::optional<std::size_t> single_color_target) { +void RasterizerOpenGL::ConfigureFramebuffers() { MICROPROFILE_SCOPE(OpenGL_Framebuffer); auto& gpu = system.GPU().Maxwell3D(); - const auto& regs = gpu.regs; - - const FramebufferConfigState fb_config_state{using_color_fb, using_depth_fb, preserve_contents, - single_color_target}; - if (fb_config_state == current_framebuffer_config_state && !gpu.dirty.render_settings) { - // Only skip if the previous ConfigureFramebuffers call was from the same kind (multiple or - // single color targets). This is done because the guest registers may not change but the - // host framebuffer may contain different attachments - return current_depth_stencil_usage; + if (!gpu.dirty.render_settings) { + return; } gpu.dirty.render_settings = false; - current_framebuffer_config_state = fb_config_state; texture_cache.GuardRenderTargets(true); - View depth_surface{}; - if (using_depth_fb) { - depth_surface = texture_cache.GetDepthBufferSurface(preserve_contents); - } else { - texture_cache.SetEmptyDepthBuffer(); - } + View depth_surface = texture_cache.GetDepthBufferSurface(true); + const auto& regs = gpu.regs; + state.framebuffer_srgb.enabled = regs.framebuffer_srgb != 0; UNIMPLEMENTED_IF(regs.rt_separate_frag_data == 0); // Bind the framebuffer surfaces - current_state.framebuffer_srgb.enabled = regs.framebuffer_srgb != 0; - FramebufferCacheKey fbkey; + for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) { + View color_surface{texture_cache.GetColorBufferSurface(index, true)}; - if (using_color_fb) { - if (single_color_target) { - // Used when just a single color attachment is enabled, e.g. for clearing a color buffer - View color_surface{ - texture_cache.GetColorBufferSurface(*single_color_target, preserve_contents)}; - - if (color_surface) { - // Assume that a surface will be written to if it is used as a framebuffer, even if - // the shader doesn't actually write to it. - texture_cache.MarkColorBufferInUse(*single_color_target); - } - - fbkey.is_single_buffer = true; - fbkey.color_attachments[0] = - GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(*single_color_target); - fbkey.colors[0] = color_surface; - for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) { - if (index != *single_color_target) { - texture_cache.SetEmptyColorBuffer(index); - } - } - } else { - // Multiple color attachments are enabled - for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) { - View color_surface{texture_cache.GetColorBufferSurface(index, preserve_contents)}; - - if (color_surface) { - // Assume that a surface will be written to if it is used as a framebuffer, even - // if the shader doesn't actually write to it. - texture_cache.MarkColorBufferInUse(index); - } - - fbkey.color_attachments[index] = - GL_COLOR_ATTACHMENT0 + regs.rt_control.GetMap(index); - fbkey.colors[index] = color_surface; - } - fbkey.is_single_buffer = false; - fbkey.colors_count = regs.rt_control.count; + if (color_surface) { + // Assume that a surface will be written to if it is used as a framebuffer, even + // if the shader doesn't actually write to it. + texture_cache.MarkColorBufferInUse(index); } - } else { - // No color attachments are enabled - leave them as zero - fbkey.is_single_buffer = true; + + fbkey.color_attachments[index] = GL_COLOR_ATTACHMENT0 + regs.rt_control.GetMap(index); + fbkey.colors[index] = std::move(color_surface); } + fbkey.colors_count = regs.rt_control.count; if (depth_surface) { // Assume that a surface will be written to if it is used as a framebuffer, even if // the shader doesn't actually write to it. texture_cache.MarkDepthBufferInUse(); - fbkey.zeta = depth_surface; fbkey.stencil_enable = depth_surface->GetSurfaceParams().type == SurfaceType::DepthStencil; + fbkey.zeta = std::move(depth_surface); } texture_cache.GuardRenderTargets(false); - current_state.draw.draw_framebuffer = framebuffer_cache.GetFramebuffer(fbkey); - SyncViewport(current_state); - - return current_depth_stencil_usage = {static_cast<bool>(depth_surface), fbkey.stencil_enable}; + state.draw.draw_framebuffer = framebuffer_cache.GetFramebuffer(fbkey); + SyncViewport(state); } void RasterizerOpenGL::ConfigureClearFramebuffer(OpenGLState& current_state, bool using_color_fb, @@ -688,17 +577,9 @@ void RasterizerOpenGL::Clear() { } } -void RasterizerOpenGL::DrawArrays() { - if (accelerate_draw == AccelDraw::Disabled) - return; - - MICROPROFILE_SCOPE(OpenGL_Drawing); +void RasterizerOpenGL::DrawPrelude() { auto& gpu = system.GPU().Maxwell3D(); - if (!gpu.ShouldExecute()) { - return; - } - SyncColorMask(); SyncFragmentColorClampState(); SyncMultiSampleState(); @@ -743,10 +624,7 @@ void RasterizerOpenGL::DrawArrays() { // Upload vertex and index data. SetupVertexBuffer(vao); SetupVertexInstances(vao); - const GLintptr index_buffer_offset = SetupIndexBuffer(); - - // Setup draw parameters. It will automatically choose what glDraw* method to use. - const DrawParameters params = SetupDraw(index_buffer_offset); + index_buffer_offset = SetupIndexBuffer(); // Prepare packed bindings. bind_ubo_pushbuffer.Setup(0); @@ -754,10 +632,11 @@ void RasterizerOpenGL::DrawArrays() { // Setup shaders and their used resources. texture_cache.GuardSamplers(true); - SetupShaders(params.primitive_mode); + const auto primitive_mode = MaxwellToGL::PrimitiveTopology(gpu.regs.draw.topology); + SetupShaders(primitive_mode); texture_cache.GuardSamplers(false); - ConfigureFramebuffers(state); + ConfigureFramebuffers(); // Signal the buffer cache that we are not going to upload more things. const bool invalidate = buffer_cache.Unmap(); @@ -778,11 +657,107 @@ void RasterizerOpenGL::DrawArrays() { if (texture_cache.TextureBarrier()) { glTextureBarrier(); } +} + +struct DrawParams { + bool is_indexed{}; + bool is_instanced{}; + GLenum primitive_mode{}; + GLint count{}; + GLint base_vertex{}; + + // Indexed settings + GLenum index_format{}; + GLintptr index_buffer_offset{}; + + // Instanced setting + GLint num_instances{}; + GLint base_instance{}; + + void DispatchDraw() { + if (is_indexed) { + const auto index_buffer_ptr = reinterpret_cast<const void*>(index_buffer_offset); + if (is_instanced) { + glDrawElementsInstancedBaseVertexBaseInstance(primitive_mode, count, index_format, + index_buffer_ptr, num_instances, + base_vertex, base_instance); + } else { + glDrawElementsBaseVertex(primitive_mode, count, index_format, index_buffer_ptr, + base_vertex); + } + } else { + if (is_instanced) { + glDrawArraysInstancedBaseInstance(primitive_mode, base_vertex, count, num_instances, + base_instance); + } else { + glDrawArrays(primitive_mode, base_vertex, count); + } + } + } +}; - params.DispatchDraw(); +bool RasterizerOpenGL::DrawBatch(bool is_indexed) { + accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays; + MICROPROFILE_SCOPE(OpenGL_Drawing); + + DrawPrelude(); + + auto& maxwell3d = system.GPU().Maxwell3D(); + const auto& regs = maxwell3d.regs; + const auto current_instance = maxwell3d.state.current_instance; + DrawParams draw_call{}; + draw_call.is_indexed = is_indexed; + draw_call.num_instances = static_cast<GLint>(1); + draw_call.base_instance = static_cast<GLint>(current_instance); + draw_call.is_instanced = current_instance > 0; + draw_call.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology); + if (draw_call.is_indexed) { + draw_call.count = static_cast<GLint>(regs.index_array.count); + draw_call.base_vertex = static_cast<GLint>(regs.vb_element_base); + draw_call.index_format = MaxwellToGL::IndexFormat(regs.index_array.format); + draw_call.index_buffer_offset = index_buffer_offset; + } else { + draw_call.count = static_cast<GLint>(regs.vertex_buffer.count); + draw_call.base_vertex = static_cast<GLint>(regs.vertex_buffer.first); + } + draw_call.DispatchDraw(); + + maxwell3d.dirty.memory_general = false; accelerate_draw = AccelDraw::Disabled; - gpu.dirty.memory_general = false; + return true; +} + +bool RasterizerOpenGL::DrawMultiBatch(bool is_indexed) { + accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays; + + MICROPROFILE_SCOPE(OpenGL_Drawing); + + DrawPrelude(); + + auto& maxwell3d = system.GPU().Maxwell3D(); + const auto& regs = maxwell3d.regs; + const auto& draw_setup = maxwell3d.mme_draw; + DrawParams draw_call{}; + draw_call.is_indexed = is_indexed; + draw_call.num_instances = static_cast<GLint>(draw_setup.instance_count); + draw_call.base_instance = static_cast<GLint>(regs.vb_base_instance); + draw_call.is_instanced = draw_setup.instance_count > 1; + draw_call.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology); + if (draw_call.is_indexed) { + draw_call.count = static_cast<GLint>(regs.index_array.count); + draw_call.base_vertex = static_cast<GLint>(regs.vb_element_base); + draw_call.index_format = MaxwellToGL::IndexFormat(regs.index_array.format); + draw_call.index_buffer_offset = index_buffer_offset; + } else { + draw_call.count = static_cast<GLint>(regs.vertex_buffer.count); + draw_call.base_vertex = static_cast<GLint>(regs.vertex_buffer.first); + } + draw_call.DispatchDraw(); + + maxwell3d.dirty.memory_general = false; + accelerate_draw = AccelDraw::Disabled; + return true; } void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) { diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index eada752e0..9c10ebda3 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -57,7 +57,8 @@ public: ScreenInfo& info); ~RasterizerOpenGL() override; - void DrawArrays() override; + bool DrawBatch(bool is_indexed) override; + bool DrawMultiBatch(bool is_indexed) override; void Clear() override; void DispatchCompute(GPUVAddr code_addr) override; void FlushAll() override; @@ -71,45 +72,13 @@ public: const Tegra::Engines::Fermi2D::Config& copy_config) override; bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr, u32 pixel_stride) override; - bool AccelerateDrawBatch(bool is_indexed) override; void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) override; void LoadDiskResources(const std::atomic_bool& stop_loading, const VideoCore::DiskResourceLoadCallback& callback) override; private: - struct FramebufferConfigState { - bool using_color_fb{}; - bool using_depth_fb{}; - bool preserve_contents{}; - std::optional<std::size_t> single_color_target; - - bool operator==(const FramebufferConfigState& rhs) const { - return std::tie(using_color_fb, using_depth_fb, preserve_contents, - single_color_target) == std::tie(rhs.using_color_fb, rhs.using_depth_fb, - rhs.preserve_contents, - rhs.single_color_target); - } - bool operator!=(const FramebufferConfigState& rhs) const { - return !operator==(rhs); - } - }; - - /** - * Configures the color and depth framebuffer states. - * - * @param current_state The current OpenGL state. - * @param using_color_fb If true, configure color framebuffers. - * @param using_depth_fb If true, configure the depth/stencil framebuffer. - * @param preserve_contents If true, tries to preserve data from a previously used - * framebuffer. - * @param single_color_target Specifies if a single color buffer target should be used. - * - * @returns If depth (first) or stencil (second) are being stored in the bound zeta texture - * (requires using_depth_fb to be true) - */ - std::pair<bool, bool> ConfigureFramebuffers( - OpenGLState& current_state, bool using_color_fb = true, bool using_depth_fb = true, - bool preserve_contents = true, std::optional<std::size_t> single_color_target = {}); + /// Configures the color and depth framebuffer states. + void ConfigureFramebuffers(); void ConfigureClearFramebuffer(OpenGLState& current_state, bool using_color_fb, bool using_depth_fb, bool using_stencil_fb); @@ -136,6 +105,9 @@ private: void SetupGlobalMemory(const GLShader::GlobalMemoryEntry& entry, GPUVAddr gpu_addr, std::size_t size); + /// Syncs all the state, shaders, render targets and textures setting before a draw call. + void DrawPrelude(); + /// Configures the current textures to use for the draw command. Returns shaders texture buffer /// usage. TextureBufferUsage SetupDrawTextures(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, @@ -228,9 +200,6 @@ private: OGLVertexArray> vertex_array_cache; - FramebufferConfigState current_framebuffer_config_state; - std::pair<bool, bool> current_depth_stencil_usage{}; - static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024; OGLBufferCache buffer_cache; @@ -250,7 +219,7 @@ private: GLintptr SetupIndexBuffer(); - DrawParameters SetupDraw(GLintptr index_buffer_offset); + GLintptr index_buffer_offset; void SetupShaders(GLenum primitive_mode); diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index 76439e7ab..74cb59bc1 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -462,6 +462,14 @@ private: code.AddLine("float gl_PointSize;"); } + if (ir.UsesInstanceId()) { + code.AddLine("int gl_InstanceID;"); + } + + if (ir.UsesVertexId()) { + code.AddLine("int gl_VertexID;"); + } + --code.scope; code.AddLine("}};"); code.AddNewLine(); @@ -964,7 +972,7 @@ private: switch (element) { case 2: // Config pack's first value is instance_id. - return {"config_pack[0]", Type::Uint}; + return {"gl_InstanceID", Type::Int}; case 3: return {"gl_VertexID", Type::Int}; } diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index 4f135fe03..173b76c4e 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -97,6 +97,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // RG8U {GL_RG8, GL_RG, GL_BYTE, ComponentType::SNorm, false}, // RG8S {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // RG32UI + {GL_RGB16F, GL_RGBA16, GL_HALF_FLOAT, ComponentType::Float, false}, // RGBX16F {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // R32UI {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X8 {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X5 |