diff options
Diffstat (limited to 'src/video_core/renderer_opengl')
-rw-r--r-- | src/video_core/renderer_opengl/gl_rasterizer.cpp | 28 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/gl_rasterizer.h | 15 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/gl_rasterizer_cache.cpp | 27 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/gl_rasterizer_cache.h | 101 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/gl_shader_decompiler.cpp | 4 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/maxwell_to_gl.h | 5 |
6 files changed, 144 insertions, 36 deletions
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 56d9c575b..a1c47bae9 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -5,6 +5,7 @@ #include <algorithm> #include <memory> #include <string> +#include <string_view> #include <tuple> #include <utility> #include <glad/glad.h> @@ -37,11 +38,6 @@ MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255)); MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100)); 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; - // Create sampler objects for (size_t i = 0; i < texture_samplers.size(); ++i) { texture_samplers[i].Create(); @@ -59,7 +55,8 @@ RasterizerOpenGL::RasterizerOpenGL() { 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))}; + const std::string_view extension{ + reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, i))}; if (extension == "GL_ARB_buffer_storage") { has_ARB_buffer_storage = true; @@ -110,8 +107,6 @@ RasterizerOpenGL::RasterizerOpenGL() { glBindBufferBase(GL_UNIFORM_BUFFER, index, buffer.handle); } - accelerate_draw = AccelDraw::Disabled; - glEnable(GL_BLEND); LOG_CRITICAL(Render_OpenGL, "Sync fixed function OpenGL state here!"); @@ -601,7 +596,6 @@ void RasterizerOpenGL::SamplerInfo::Create() { sampler.Create(); mag_filter = min_filter = Tegra::Texture::TextureFilter::Linear; wrap_u = wrap_v = Tegra::Texture::WrapMode::Wrap; - border_color_r = border_color_g = border_color_b = border_color_a = 0; // default is GL_LINEAR_MIPMAP_LINEAR glSamplerParameteri(sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR); @@ -630,8 +624,12 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr } if (wrap_u == Tegra::Texture::WrapMode::Border || wrap_v == Tegra::Texture::WrapMode::Border) { - // TODO(Subv): Implement border color - ASSERT(false); + const GLvec4 new_border_color = {{config.border_color_r, config.border_color_g, + config.border_color_b, config.border_color_a}}; + if (border_color != new_border_color) { + border_color = new_border_color; + glSamplerParameterfv(s, GL_TEXTURE_BORDER_COLOR, border_color.data()); + } } } @@ -691,10 +689,12 @@ u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, GLuint progr glBindBuffer(GL_UNIFORM_BUFFER, 0); // Now configure the bindpoint of the buffer inside the shader - std::string buffer_name = used_buffer.GetName(); - GLuint index = glGetProgramResourceIndex(program, GL_UNIFORM_BLOCK, buffer_name.c_str()); - if (index != -1) + const std::string buffer_name = used_buffer.GetName(); + const GLuint index = + glGetProgramResourceIndex(program, GL_UNIFORM_BLOCK, buffer_name.c_str()); + if (index != GL_INVALID_INDEX) { glUniformBlockBinding(program, index, buffer_draw_state.bindpoint); + } } state.Apply(); diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index c406142e4..e150be58f 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -77,10 +77,7 @@ private: Tegra::Texture::TextureFilter min_filter; Tegra::Texture::WrapMode wrap_u; Tegra::Texture::WrapMode wrap_v; - u32 border_color_r; - u32 border_color_g; - u32 border_color_b; - u32 border_color_a; + GLvec4 border_color; }; /// Configures the color and depth framebuffer states and returns the dirty <Color, Depth> @@ -138,10 +135,10 @@ private: /// Syncs the blend state to match the guest state void SyncBlendState(); - bool has_ARB_buffer_storage; - bool has_ARB_direct_state_access; - bool has_ARB_separate_shader_objects; - bool has_ARB_vertex_attrib_binding; + bool has_ARB_buffer_storage = false; + bool has_ARB_direct_state_access = false; + bool has_ARB_separate_shader_objects = false; + bool has_ARB_vertex_attrib_binding = false; OpenGLState state; @@ -170,5 +167,5 @@ private: void SetupShaders(u8* buffer_ptr, GLintptr buffer_offset); enum class AccelDraw { Disabled, Arrays, Indexed }; - AccelDraw accelerate_draw; + AccelDraw accelerate_draw = AccelDraw::Disabled; }; diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp index 28f0bc379..91ce0357b 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp @@ -38,7 +38,8 @@ struct FormatTuple { params.addr = config.tic.Address(); params.is_tiled = config.tic.IsTiled(); params.block_height = params.is_tiled ? config.tic.BlockHeight() : 0, - params.pixel_format = PixelFormatFromTextureFormat(config.tic.format); + params.pixel_format = + PixelFormatFromTextureFormat(config.tic.format, config.tic.r_type.Value()); params.component_type = ComponentTypeFromTexture(config.tic.r_type.Value()); params.type = GetFormatType(params.pixel_format); params.width = Common::AlignUp(config.tic.Width(), GetCompressionFactor(params.pixel_format)); @@ -106,6 +107,12 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form true}, // BC7U {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4 {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // G8R8 + {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // BGRA8 + {GL_RGBA32F, GL_RGBA, GL_FLOAT, ComponentType::Float, false}, // RGBA32F + {GL_RG32F, GL_RG, GL_FLOAT, ComponentType::Float, false}, // RG32F + {GL_R32F, GL_RED, GL_FLOAT, ComponentType::Float, false}, // R32F + {GL_R16F, GL_RED, GL_HALF_FLOAT, ComponentType::Float, false}, // R16F + {GL_R16, GL_RED, GL_UNSIGNED_SHORT, ComponentType::UNorm, false}, // R16UNORM // DepthStencil formats {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, ComponentType::UNorm, @@ -115,6 +122,8 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT, ComponentType::Float, false}, // Z32F {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT, ComponentType::UNorm, false}, // Z16 + {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV, + ComponentType::Float, false}, // Z32FS8 }}; static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) { @@ -197,9 +206,12 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr), MortonCopy<true, PixelFormat::DXT1>, MortonCopy<true, PixelFormat::DXT23>, MortonCopy<true, PixelFormat::DXT45>, MortonCopy<true, PixelFormat::DXN1>, MortonCopy<true, PixelFormat::BC7U>, MortonCopy<true, PixelFormat::ASTC_2D_4X4>, - MortonCopy<true, PixelFormat::G8R8>, MortonCopy<true, PixelFormat::Z24S8>, + MortonCopy<true, PixelFormat::G8R8>, MortonCopy<true, PixelFormat::BGRA8>, + MortonCopy<true, PixelFormat::RGBA32F>, MortonCopy<true, PixelFormat::RG32F>, + MortonCopy<true, PixelFormat::R32F>, MortonCopy<true, PixelFormat::R16F>, + MortonCopy<true, PixelFormat::R16UNORM>, MortonCopy<true, PixelFormat::Z24S8>, MortonCopy<true, PixelFormat::S8Z24>, MortonCopy<true, PixelFormat::Z32F>, - MortonCopy<true, PixelFormat::Z16>, + MortonCopy<true, PixelFormat::Z16>, MortonCopy<true, PixelFormat::Z32FS8>, }; static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr), @@ -213,7 +225,7 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr), MortonCopy<false, PixelFormat::RGBA16F>, MortonCopy<false, PixelFormat::R11FG11FB10F>, MortonCopy<false, PixelFormat::RGBA32UI>, - // TODO(Subv): Swizzling the DXT1/DXT23/DXT45/DXN1/BC7U formats is not yet supported + // TODO(Subv): Swizzling DXT1/DXT23/DXT45/DXN1/BC7U/ASTC_2D_4X4 formats is not supported nullptr, nullptr, nullptr, @@ -221,10 +233,17 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr), nullptr, nullptr, MortonCopy<false, PixelFormat::G8R8>, + MortonCopy<false, PixelFormat::BGRA8>, + MortonCopy<false, PixelFormat::RGBA32F>, + MortonCopy<false, PixelFormat::RG32F>, + MortonCopy<false, PixelFormat::R32F>, + MortonCopy<false, PixelFormat::R16F>, + MortonCopy<false, PixelFormat::R16UNORM>, MortonCopy<false, PixelFormat::Z24S8>, MortonCopy<false, PixelFormat::S8Z24>, MortonCopy<false, PixelFormat::Z32F>, MortonCopy<false, PixelFormat::Z16>, + MortonCopy<false, PixelFormat::Z32FS8>, }; // Allocate an uninitialized texture of appropriate size and format for the surface diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h index b084c4db4..fc864c56f 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h @@ -37,14 +37,21 @@ struct SurfaceParams { BC7U = 12, ASTC_2D_4X4 = 13, G8R8 = 14, + BGRA8 = 15, + RGBA32F = 16, + RG32F = 17, + R32F = 18, + R16F = 19, + R16UNORM = 20, MaxColorFormat, // DepthStencil formats - Z24S8 = 15, - S8Z24 = 16, - Z32F = 17, - Z16 = 18, + Z24S8 = 21, + S8Z24 = 22, + Z32F = 23, + Z16 = 24, + Z32FS8 = 25, MaxDepthStencilFormat, @@ -97,10 +104,17 @@ struct SurfaceParams { 4, // BC7U 4, // ASTC_2D_4X4 1, // G8R8 + 1, // BGRA8 + 1, // RGBA32F + 1, // RG32F + 1, // R32F + 1, // R16F + 1, // R16UNORM 1, // Z24S8 1, // S8Z24 1, // Z32F 1, // Z16 + 1, // Z32FS8 }}; ASSERT(static_cast<size_t>(format) < compression_factor_table.size()); @@ -127,10 +141,17 @@ struct SurfaceParams { 128, // BC7U 32, // ASTC_2D_4X4 16, // G8R8 + 32, // BGRA8 + 128, // RGBA32F + 64, // RG32F + 32, // R32F + 16, // R16F + 16, // R16UNORM 32, // Z24S8 32, // S8Z24 32, // Z32F 16, // Z16 + 64, // Z32FS8 }}; ASSERT(static_cast<size_t>(format) < bpp_table.size()); @@ -151,6 +172,8 @@ struct SurfaceParams { return PixelFormat::Z32F; case Tegra::DepthFormat::Z16_UNORM: return PixelFormat::Z16; + case Tegra::DepthFormat::Z32_S8_X24_FLOAT: + return PixelFormat::Z32FS8; default: LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); UNREACHABLE(); @@ -162,21 +185,30 @@ struct SurfaceParams { case Tegra::RenderTargetFormat::RGBA8_UNORM: case Tegra::RenderTargetFormat::RGBA8_SRGB: return PixelFormat::ABGR8; + case Tegra::RenderTargetFormat::BGRA8_UNORM: + return PixelFormat::BGRA8; case Tegra::RenderTargetFormat::RGB10_A2_UNORM: return PixelFormat::A2B10G10R10; case Tegra::RenderTargetFormat::RGBA16_FLOAT: return PixelFormat::RGBA16F; + case Tegra::RenderTargetFormat::RGBA32_FLOAT: + return PixelFormat::RGBA32F; + case Tegra::RenderTargetFormat::RG32_FLOAT: + return PixelFormat::RG32F; case Tegra::RenderTargetFormat::R11G11B10_FLOAT: return PixelFormat::R11FG11FB10F; case Tegra::RenderTargetFormat::RGBA32_UINT: return PixelFormat::RGBA32UI; + case Tegra::RenderTargetFormat::R8_UNORM: + return PixelFormat::R8; default: LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); UNREACHABLE(); } } - static PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format) { + static PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format, + Tegra::Texture::ComponentType component_type) { // TODO(Subv): Properly implement this switch (format) { case Tegra::Texture::TextureFormat::A8R8G8B8: @@ -196,7 +228,33 @@ struct SurfaceParams { case Tegra::Texture::TextureFormat::BF10GF11RF11: return PixelFormat::R11FG11FB10F; case Tegra::Texture::TextureFormat::R32_G32_B32_A32: - return PixelFormat::RGBA32UI; + switch (component_type) { + case Tegra::Texture::ComponentType::FLOAT: + return PixelFormat::RGBA32F; + case Tegra::Texture::ComponentType::UINT: + return PixelFormat::RGBA32UI; + } + LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", + static_cast<u32>(component_type)); + UNREACHABLE(); + case Tegra::Texture::TextureFormat::R32_G32: + return PixelFormat::RG32F; + case Tegra::Texture::TextureFormat::R16: + switch (component_type) { + case Tegra::Texture::ComponentType::FLOAT: + return PixelFormat::R16F; + case Tegra::Texture::ComponentType::UNORM: + return PixelFormat::R16UNORM; + } + LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", + static_cast<u32>(component_type)); + UNREACHABLE(); + case Tegra::Texture::TextureFormat::R32: + return PixelFormat::R32F; + case Tegra::Texture::TextureFormat::ZF32: + return PixelFormat::Z32F; + case Tegra::Texture::TextureFormat::Z24S8: + return PixelFormat::Z24S8; case Tegra::Texture::TextureFormat::DXT1: return PixelFormat::DXT1; case Tegra::Texture::TextureFormat::DXT23: @@ -210,7 +268,8 @@ struct SurfaceParams { case Tegra::Texture::TextureFormat::ASTC_2D_4X4: return PixelFormat::ASTC_2D_4X4; default: - LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); + LOG_CRITICAL(HW_GPU, "Unimplemented format={}, component_type={}", + static_cast<u32>(format), static_cast<u32>(component_type)); UNREACHABLE(); } } @@ -248,7 +307,25 @@ struct SurfaceParams { return Tegra::Texture::TextureFormat::BC7U; case PixelFormat::ASTC_2D_4X4: return Tegra::Texture::TextureFormat::ASTC_2D_4X4; + case PixelFormat::BGRA8: + // TODO(bunnei): This is fine for unswizzling (since we just need the right component + // sizes), but could be a bug if we used this function in different ways. + return Tegra::Texture::TextureFormat::A8R8G8B8; + case PixelFormat::RGBA32F: + return Tegra::Texture::TextureFormat::R32_G32_B32_A32; + case PixelFormat::RG32F: + return Tegra::Texture::TextureFormat::R32_G32; + case PixelFormat::R32F: + return Tegra::Texture::TextureFormat::R32; + case PixelFormat::R16F: + case PixelFormat::R16UNORM: + return Tegra::Texture::TextureFormat::R16; + case PixelFormat::Z32F: + return Tegra::Texture::TextureFormat::ZF32; + case PixelFormat::Z24S8: + return Tegra::Texture::TextureFormat::Z24S8; default: + LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); UNREACHABLE(); } } @@ -263,7 +340,10 @@ struct SurfaceParams { return Tegra::DepthFormat::Z32_FLOAT; case PixelFormat::Z16: return Tegra::DepthFormat::Z16_UNORM; + case PixelFormat::Z32FS8: + return Tegra::DepthFormat::Z32_S8_X24_FLOAT; default: + LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); UNREACHABLE(); } } @@ -273,6 +353,8 @@ struct SurfaceParams { switch (type) { case Tegra::Texture::ComponentType::UNORM: return ComponentType::UNorm; + case Tegra::Texture::ComponentType::FLOAT: + return ComponentType::Float; default: LOG_CRITICAL(HW_GPU, "Unimplemented component type={}", static_cast<u32>(type)); UNREACHABLE(); @@ -284,10 +366,14 @@ struct SurfaceParams { switch (format) { case Tegra::RenderTargetFormat::RGBA8_UNORM: case Tegra::RenderTargetFormat::RGBA8_SRGB: + case Tegra::RenderTargetFormat::BGRA8_UNORM: case Tegra::RenderTargetFormat::RGB10_A2_UNORM: + case Tegra::RenderTargetFormat::R8_UNORM: return ComponentType::UNorm; case Tegra::RenderTargetFormat::RGBA16_FLOAT: case Tegra::RenderTargetFormat::R11G11B10_FLOAT: + case Tegra::RenderTargetFormat::RGBA32_FLOAT: + case Tegra::RenderTargetFormat::RG32_FLOAT: return ComponentType::Float; case Tegra::RenderTargetFormat::RGBA32_UINT: return ComponentType::UInt; @@ -314,6 +400,7 @@ struct SurfaceParams { case Tegra::DepthFormat::Z24_S8_UNORM: return ComponentType::UNorm; case Tegra::DepthFormat::Z32_FLOAT: + case Tegra::DepthFormat::Z32_S8_X24_FLOAT: return ComponentType::Float; default: LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index d4b1a6f33..acf067050 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -756,8 +756,8 @@ private: } } - std::string WriteTexsInstruction(const Instruction& instr, const std::string& coord, - const std::string& texture) { + void WriteTexsInstruction(const Instruction& instr, const std::string& coord, + const std::string& texture) { // Add an extra scope and declare the texture coords inside to prevent // overwriting them in case they are used as outputs of the texs instruction. shader.AddLine('{'); diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h index e19c3b280..16b1bd606 100644 --- a/src/video_core/renderer_opengl/maxwell_to_gl.h +++ b/src/video_core/renderer_opengl/maxwell_to_gl.h @@ -56,6 +56,9 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) { return {}; } + case Maxwell::VertexAttribute::Type::UnsignedInt: + return GL_UNSIGNED_INT; + case Maxwell::VertexAttribute::Type::Float: return GL_FLOAT; } @@ -112,6 +115,8 @@ inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) { return GL_MIRRORED_REPEAT; case Tegra::Texture::WrapMode::ClampToEdge: return GL_CLAMP_TO_EDGE; + case Tegra::Texture::WrapMode::Border: + return GL_CLAMP_TO_BORDER; case Tegra::Texture::WrapMode::ClampOGL: // TODO(Subv): GL_CLAMP was removed as of OpenGL 3.1, to implement GL_CLAMP, we can use // GL_CLAMP_TO_BORDER to get the border color of the texture, and then sample the edge to |