diff options
Diffstat (limited to 'src/video_core')
-rw-r--r-- | src/video_core/engines/shader_bytecode.h | 10 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/gl_rasterizer_cache.cpp | 17 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/gl_rasterizer_cache.h | 24 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/gl_shader_decompiler.cpp | 21 | ||||
-rw-r--r-- | src/video_core/textures/decoders.cpp | 9 |
5 files changed, 66 insertions, 15 deletions
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h index d75de85e2..f32a17057 100644 --- a/src/video_core/engines/shader_bytecode.h +++ b/src/video_core/engines/shader_bytecode.h @@ -75,6 +75,10 @@ union Attribute { enum class Index : u64 { Position = 7, Attribute_0 = 8, + // This attribute contains a tuple of (~, ~, InstanceId, VertexId) when inside a vertex + // shader, and a tuple of (TessCoord.x, TessCoord.y, TessCoord.z, ~) when inside a Tess Eval + // shader. + TessCoordInstanceIDVertexID = 47, }; union { @@ -456,9 +460,9 @@ private: INST("00011110--------", Id::FMUL32_IMM, Type::Arithmetic, "FMUL32_IMM"), INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"), INST("0101110010010---", Id::RRO, Type::Arithmetic, "RRO"), - INST("0100110010101---", Id::F2F_C, Type::Arithmetic, "F2F_C"), - INST("0101110010101---", Id::F2F_R, Type::Arithmetic, "F2F_R"), - INST("0011100-10101---", Id::F2F_IMM, Type::Arithmetic, "F2F_IMM"), + INST("0100110010101---", Id::F2F_C, Type::Conversion, "F2F_C"), + INST("0101110010101---", Id::F2F_R, Type::Conversion, "F2F_R"), + INST("0011100-10101---", Id::F2F_IMM, Type::Conversion, "F2F_IMM"), INST("0100110010110---", Id::F2I_C, Type::Arithmetic, "F2I_C"), INST("0101110010110---", Id::F2I_R, Type::Arithmetic, "F2I_R"), INST("0011100-10110---", Id::F2I_IMM, Type::Arithmetic, "F2I_IMM"), diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp index 5bb34037b..65d643447 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp @@ -45,21 +45,23 @@ struct FormatTuple { static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_format_tuples = {{ {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false}, // ABGR8 - {GL_RGB, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, false}, // B5G6R5 + {GL_RGB, GL_RGB, GL_UNSIGNED_SHORT_5_6_5, false}, // B5G6R5 {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, false}, // A2B10G10R10 {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, false}, // A1B5G5R5 {GL_R8, GL_RED, GL_UNSIGNED_BYTE, false}, // R8 + {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, false}, // RGBA16F {GL_COMPRESSED_RGB_S3TC_DXT1_EXT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT1 {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT23 {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT45 + {GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, true}, // DXN1 }}; static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) { const SurfaceType type = SurfaceParams::GetFormatType(pixel_format); if (type == SurfaceType::ColorTexture) { ASSERT(static_cast<size_t>(pixel_format) < tex_format_tuples.size()); - // For now only UNORM components are supported - ASSERT(component_type == ComponentType::UNorm); + // For now only UNORM components are supported, or RGBA16F which is type FLOAT + ASSERT(component_type == ComponentType::UNorm || pixel_format == PixelFormat::RGBA16F); return tex_format_tuples[static_cast<unsigned int>(pixel_format)]; } else if (type == SurfaceType::Depth || type == SurfaceType::DepthStencil) { // TODO(Subv): Implement depth formats @@ -110,8 +112,9 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr, Tegra: morton_to_gl_fns = { MortonCopy<true, PixelFormat::ABGR8>, MortonCopy<true, PixelFormat::B5G6R5>, MortonCopy<true, PixelFormat::A2B10G10R10>, MortonCopy<true, PixelFormat::A1B5G5R5>, - MortonCopy<true, PixelFormat::R8>, MortonCopy<true, PixelFormat::DXT1>, - MortonCopy<true, PixelFormat::DXT23>, MortonCopy<true, PixelFormat::DXT45>, + MortonCopy<true, PixelFormat::R8>, MortonCopy<true, PixelFormat::RGBA16F>, + MortonCopy<true, PixelFormat::DXT1>, MortonCopy<true, PixelFormat::DXT23>, + MortonCopy<true, PixelFormat::DXT45>, MortonCopy<true, PixelFormat::DXN1>, }; static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr, Tegra::GPUVAddr, @@ -123,7 +126,9 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr, Tegra: MortonCopy<false, PixelFormat::A2B10G10R10>, MortonCopy<false, PixelFormat::A1B5G5R5>, MortonCopy<false, PixelFormat::R8>, - // TODO(Subv): Swizzling the DXT1/DXT23/DXT45 formats is not yet supported + MortonCopy<false, PixelFormat::RGBA16F>, + // TODO(Subv): Swizzling the DXT1/DXT23/DXT45/DXN1 formats is not yet supported + nullptr, nullptr, nullptr, nullptr, diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h index fc09f108c..6f08678ab 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h @@ -59,9 +59,11 @@ struct SurfaceParams { A2B10G10R10 = 2, A1B5G5R5 = 3, R8 = 4, - DXT1 = 5, - DXT23 = 6, - DXT45 = 7, + RGBA16F = 5, + DXT1 = 6, + DXT23 = 7, + DXT45 = 8, + DXN1 = 9, // This is also known as BC4 Max, Invalid = 255, @@ -102,9 +104,11 @@ struct SurfaceParams { 1, // A2B10G10R10 1, // A1B5G5R5 1, // R8 + 2, // RGBA16F 4, // DXT1 4, // DXT23 4, // DXT45 + 4, // DXN1 }}; ASSERT(static_cast<size_t>(format) < compression_factor_table.size()); @@ -124,9 +128,11 @@ struct SurfaceParams { 32, // A2B10G10R10 16, // A1B5G5R5 8, // R8 + 64, // RGBA16F 64, // DXT1 128, // DXT23 128, // DXT45 + 64, // DXN1 }}; ASSERT(static_cast<size_t>(format) < bpp_table.size()); @@ -143,6 +149,8 @@ struct SurfaceParams { return PixelFormat::ABGR8; case Tegra::RenderTargetFormat::RGB10_A2_UNORM: return PixelFormat::A2B10G10R10; + case Tegra::RenderTargetFormat::RGBA16_FLOAT: + return PixelFormat::RGBA16F; default: NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); UNREACHABLE(); @@ -172,12 +180,16 @@ struct SurfaceParams { return PixelFormat::A1B5G5R5; case Tegra::Texture::TextureFormat::R8: return PixelFormat::R8; + case Tegra::Texture::TextureFormat::R16_G16_B16_A16: + return PixelFormat::RGBA16F; case Tegra::Texture::TextureFormat::DXT1: return PixelFormat::DXT1; case Tegra::Texture::TextureFormat::DXT23: return PixelFormat::DXT23; case Tegra::Texture::TextureFormat::DXT45: return PixelFormat::DXT45; + case Tegra::Texture::TextureFormat::DXN1: + return PixelFormat::DXN1; default: NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); UNREACHABLE(); @@ -197,12 +209,16 @@ struct SurfaceParams { return Tegra::Texture::TextureFormat::A1B5G5R5; case PixelFormat::R8: return Tegra::Texture::TextureFormat::R8; + case PixelFormat::RGBA16F: + return Tegra::Texture::TextureFormat::R16_G16_B16_A16; case PixelFormat::DXT1: return Tegra::Texture::TextureFormat::DXT1; case PixelFormat::DXT23: return Tegra::Texture::TextureFormat::DXT23; case PixelFormat::DXT45: return Tegra::Texture::TextureFormat::DXT45; + case PixelFormat::DXN1: + return Tegra::Texture::TextureFormat::DXN1; default: UNREACHABLE(); } @@ -226,6 +242,8 @@ struct SurfaceParams { case Tegra::RenderTargetFormat::RGBA8_SRGB: case Tegra::RenderTargetFormat::RGB10_A2_UNORM: return ComponentType::UNorm; + case Tegra::RenderTargetFormat::RGBA16_FLOAT: + return ComponentType::Float; default: NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); UNREACHABLE(); diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index 75822e750..68efe74b8 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -299,7 +299,7 @@ public: * are stored as floats, so this may require conversion. * @param reg The destination register to use. * @param elem The element to use for the operation. - * @param attribute The input attibute to use as the source value. + * @param attribute The input attribute to use as the source value. */ void SetRegisterToInputAttibute(const Register& reg, u64 elem, Attribute::Index attribute) { std::string dest = GetRegisterAsFloat(reg); @@ -451,6 +451,12 @@ private: switch (attribute) { case Attribute::Index::Position: return "position"; + case Attribute::Index::TessCoordInstanceIDVertexID: + // TODO(Subv): Find out what the values are for the first two elements when inside a + // vertex shader, and what's the value of the fourth element when inside a Tess Eval + // shader. + ASSERT(stage == Maxwell3D::Regs::ShaderStage::Vertex); + return "vec4(0, 0, gl_InstanceID, gl_VertexID)"; default: const u32 index{static_cast<u32>(attribute) - static_cast<u32>(Attribute::Index::Attribute_0)}; @@ -834,13 +840,14 @@ private: } case OpCode::Type::Conversion: { ASSERT_MSG(instr.conversion.size == Register::Size::Word, "Unimplemented"); - ASSERT_MSG(!instr.conversion.selector, "Unimplemented"); ASSERT_MSG(!instr.conversion.negate_a, "Unimplemented"); ASSERT_MSG(!instr.conversion.saturate_a, "Unimplemented"); switch (opcode->GetId()) { case OpCode::Id::I2I_R: case OpCode::Id::I2F_R: { + ASSERT_MSG(!instr.conversion.selector, "Unimplemented"); + std::string op_a = regs.GetRegisterAsInteger(instr.gpr20, 0, instr.conversion.is_signed); @@ -851,6 +858,16 @@ private: regs.SetRegisterToInteger(instr.gpr0, instr.conversion.is_signed, 0, op_a, 1, 1); break; } + case OpCode::Id::F2F_R: { + std::string op_a = regs.GetRegisterAsFloat(instr.gpr20); + + if (instr.conversion.abs_a) { + op_a = "abs(" + op_a + ')'; + } + + regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1); + break; + } default: { NGLOG_CRITICAL(HW_GPU, "Unhandled conversion instruction: {}", opcode->GetName()); UNREACHABLE(); diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 9d7b73b73..2d2af5554 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -46,6 +46,7 @@ void CopySwizzledData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_bytes_ u32 BytesPerPixel(TextureFormat format) { switch (format) { case TextureFormat::DXT1: + case TextureFormat::DXN1: // In this case a 'pixel' actually refers to a 4x4 tile. return 8; case TextureFormat::DXT23: @@ -60,6 +61,8 @@ u32 BytesPerPixel(TextureFormat format) { return 2; case TextureFormat::R8: return 1; + case TextureFormat::R16_G16_B16_A16: + return 8; default: UNIMPLEMENTED_MSG("Format not implemented"); break; @@ -77,7 +80,9 @@ std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width, case TextureFormat::DXT1: case TextureFormat::DXT23: case TextureFormat::DXT45: - // In the DXT formats, each 4x4 tile is swizzled instead of just individual pixel values. + case TextureFormat::DXN1: + // In the DXT and DXN formats, each 4x4 tile is swizzled instead of just individual pixel + // values. CopySwizzledData(width / 4, height / 4, bytes_per_pixel, bytes_per_pixel, data, unswizzled_data.data(), true, block_height); break; @@ -86,6 +91,7 @@ std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width, case TextureFormat::A1B5G5R5: case TextureFormat::B5G6R5: case TextureFormat::R8: + case TextureFormat::R16_G16_B16_A16: CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data, unswizzled_data.data(), true, block_height); break; @@ -106,6 +112,7 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat case TextureFormat::DXT1: case TextureFormat::DXT23: case TextureFormat::DXT45: + case TextureFormat::DXN1: case TextureFormat::A8R8G8B8: case TextureFormat::A2B10G10R10: case TextureFormat::A1B5G5R5: |