From 345e73f2feb0701e3c3099d002a1c21fb524eae4 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Fri, 10 May 2019 04:17:48 -0300 Subject: video_core: Use un-shifted block sizes to avoid integer divisions Instead of storing all block width, height and depths in their shifted form: block_width = 1U << block_shift; Store them like they are provided by the emulated hardware (their block_shift form). This way we can avoid doing the costly Common::AlignUp operation to align texture sizes and drop CPU integer divisions with bitwise logic (defined in Common::AlignBits). --- src/video_core/engines/fermi_2d.h | 9 +++------ src/video_core/engines/maxwell_dma.h | 4 ++-- 2 files changed, 5 insertions(+), 8 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h index 45f59a4d9..3d28afa91 100644 --- a/src/video_core/engines/fermi_2d.h +++ b/src/video_core/engines/fermi_2d.h @@ -63,18 +63,15 @@ public: } u32 BlockWidth() const { - // The block width is stored in log2 format. - return 1 << block_width; + return block_width; } u32 BlockHeight() const { - // The block height is stored in log2 format. - return 1 << block_height; + return block_height; } u32 BlockDepth() const { - // The block depth is stored in log2 format. - return 1 << block_depth; + return block_depth; } }; static_assert(sizeof(Surface) == 0x28, "Surface has incorrect size"); diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h index e5942f671..522fa97dc 100644 --- a/src/video_core/engines/maxwell_dma.h +++ b/src/video_core/engines/maxwell_dma.h @@ -59,11 +59,11 @@ public: }; u32 BlockHeight() const { - return 1 << block_height; + return block_height; } u32 BlockDepth() const { - return 1 << block_depth; + return block_depth; } }; -- cgit v1.2.3 From 6bd034eae9ab48c00e1635e2d7059b3d99b764b0 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Sat, 11 May 2019 00:38:35 -0400 Subject: engine_upload: Addapt to new Texture Cache --- src/video_core/engines/engine_upload.cpp | 4 ++-- src/video_core/engines/engine_upload.h | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/engine_upload.cpp b/src/video_core/engines/engine_upload.cpp index 082a40cd9..c776b9a56 100644 --- a/src/video_core/engines/engine_upload.cpp +++ b/src/video_core/engines/engine_upload.cpp @@ -36,8 +36,8 @@ void State::ProcessData(const u32 data, const bool is_last_call) { } else { UNIMPLEMENTED_IF(regs.dest.z != 0); UNIMPLEMENTED_IF(regs.dest.depth != 1); - UNIMPLEMENTED_IF(regs.dest.BlockWidth() != 1); - UNIMPLEMENTED_IF(regs.dest.BlockDepth() != 1); + UNIMPLEMENTED_IF(regs.dest.BlockWidth() != 0); + UNIMPLEMENTED_IF(regs.dest.BlockDepth() != 0); const std::size_t dst_size = Tegra::Texture::CalculateSize( true, 1, regs.dest.width, regs.dest.height, 1, regs.dest.BlockHeight(), 1); tmp_buffer.resize(dst_size); diff --git a/src/video_core/engines/engine_upload.h b/src/video_core/engines/engine_upload.h index ef4f5839a..cb294aec3 100644 --- a/src/video_core/engines/engine_upload.h +++ b/src/video_core/engines/engine_upload.h @@ -39,15 +39,15 @@ struct Registers { } u32 BlockWidth() const { - return 1U << block_width.Value(); + return block_width; } u32 BlockHeight() const { - return 1U << block_height.Value(); + return block_height; } u32 BlockDepth() const { - return 1U << block_depth.Value(); + return block_depth; } } dest; }; -- cgit v1.2.3 From d267948a73d2364949660a24d07833ea05c9fcc8 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Tue, 14 May 2019 00:55:32 -0400 Subject: texture_cache: loose TryReconstructSurface when accurate GPU is not on. Also corrects some asserts. --- src/video_core/engines/maxwell_dma.cpp | 2 +- src/video_core/texture_cache/surface_base.cpp | 2 +- src/video_core/texture_cache/texture_cache.h | 20 ++++++++++++++++++-- 3 files changed, 20 insertions(+), 4 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp index 3a5dfef0c..afb9578d0 100644 --- a/src/video_core/engines/maxwell_dma.cpp +++ b/src/video_core/engines/maxwell_dma.cpp @@ -111,7 +111,7 @@ void MaxwellDMA::HandleCopy() { memory_manager.WriteBlock(dest, write_buffer.data(), dst_size); } else { - ASSERT(regs.dst_params.BlockDepth() == 1); + ASSERT(regs.dst_params.BlockDepth() == 0); const u32 src_bytes_per_pixel = regs.src_pitch / regs.x_count; diff --git a/src/video_core/texture_cache/surface_base.cpp b/src/video_core/texture_cache/surface_base.cpp index ceff51043..d4aa2c54b 100644 --- a/src/video_core/texture_cache/surface_base.cpp +++ b/src/video_core/texture_cache/surface_base.cpp @@ -130,7 +130,7 @@ void SurfaceBaseImpl::FlushBuffer(Tegra::MemoryManager& memory_manager, return; } if (params.is_tiled) { - ASSERT_MSG(params.block_width == 1, "Block width is defined as {}", params.block_width); + ASSERT_MSG(params.block_width == 0, "Block width is defined as {}", params.block_width); for (u32 level = 0; level < params.num_levels; ++level) { const std::size_t host_offset{params.GetHostMipmapLevelOffset(level)}; SwizzleFunc(MortonSwizzleMode::LinearToMorton, host_ptr, params, diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 04e9528b8..85c9160e0 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -425,6 +425,7 @@ private: } bool modified = false; TSurface new_surface = GetUncachedSurface(gpu_addr, params); + u32 passed_tests = 0; for (auto surface : overlaps) { const SurfaceParams& src_params = surface->GetSurfaceParams(); if (src_params.is_layered || src_params.num_levels > 1) { @@ -434,12 +435,12 @@ private: const std::size_t candidate_size = surface->GetSizeInBytes(); auto mipmap_layer{new_surface->GetLayerMipmap(surface->GetGpuAddr())}; if (!mipmap_layer) { - return {}; + continue; } const u32 layer{mipmap_layer->first}; const u32 mipmap{mipmap_layer->second}; if (new_surface->GetMipmapSize(mipmap) != candidate_size) { - return {}; + continue; } modified |= surface->IsModified(); // Now we got all the data set up @@ -448,8 +449,15 @@ private: const CopyParams copy_params(0, 0, 0, 0, 0, layer, 0, mipmap, std::min(src_params.width, dst_width), std::min(src_params.height, dst_height), 1); + passed_tests++; ImageCopy(surface, new_surface, copy_params); } + if (passed_tests == 0) { + return {}; + // In Accurate GPU all test should pass, else we recycle + } else if (Settings::values.use_accurate_gpu_emulation && passed_tests != overlaps.size()) { + return {}; + } for (auto surface : overlaps) { Unregister(surface); } @@ -548,6 +556,14 @@ private: } return {current_surface, *view}; } + // The next case is unsafe, so if we r in accurate GPU, just skip it + if (Settings::values.use_accurate_gpu_emulation) { + return RecycleSurface(overlaps, params, gpu_addr, preserve_contents, false); + } + // This is the case the texture is a part of the parent. + if (current_surface->MatchesSubTexture(params, gpu_addr)) { + return RebuildSurface(current_surface, params); + } } else { // If there are many overlaps, odds are they are subtextures of the candidate // surface. We try to construct a new surface based on the candidate parameters, -- cgit v1.2.3 From 4e81fc8296c6204645151bbaa23a7d80827a4293 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Sat, 27 Apr 2019 20:50:35 -0300 Subject: shader: Implement texture buffers --- src/video_core/engines/shader_bytecode.h | 16 ++++++++++++ src/video_core/shader/decode/texture.cpp | 44 ++++++++++++++++++++++++++++++++ src/video_core/shader/shader_ir.h | 2 ++ 3 files changed, 62 insertions(+) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h index ffb3ec3e0..5b32e1249 100644 --- a/src/video_core/engines/shader_bytecode.h +++ b/src/video_core/engines/shader_bytecode.h @@ -1231,6 +1231,20 @@ union Instruction { } } texs; + union { + BitField<28, 1, u64> is_array; + BitField<29, 2, TextureType> texture_type; + BitField<35, 1, u64> aoffi; + BitField<49, 1, u64> nodep_flag; + BitField<50, 1, u64> ms; // Multisample? + BitField<54, 1, u64> cl; + BitField<55, 1, u64> process_mode; + + TextureProcessMode GetTextureProcessMode() const { + return process_mode == 0 ? TextureProcessMode::LZ : TextureProcessMode::LL; + } + } tld; + union { BitField<49, 1, u64> nodep_flag; BitField<53, 4, u64> texture_info; @@ -1408,6 +1422,7 @@ public: TXQ, // Texture Query TXQ_B, // Texture Query Bindless TEXS, // Texture Fetch with scalar/non-vec4 source/destinations + TLD, // Texture Load TLDS, // Texture Load with scalar/non-vec4 source/destinations TLD4, // Texture Load 4 TLD4S, // Texture Load 4 with scalar / non - vec4 source / destinations @@ -1682,6 +1697,7 @@ private: INST("1101111101001---", Id::TXQ, Type::Texture, "TXQ"), INST("1101111101010---", Id::TXQ_B, Type::Texture, "TXQ_B"), INST("1101-00---------", Id::TEXS, Type::Texture, "TEXS"), + INST("11011100--11----", Id::TLD, Type::Texture, "TLD"), INST("1101101---------", Id::TLDS, Type::Texture, "TLDS"), INST("110010----111---", Id::TLD4, Type::Texture, "TLD4"), INST("1101111100------", Id::TLD4S, Type::Texture, "TLD4S"), diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index 4a356dbd4..b22831c64 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -245,6 +245,18 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { } break; } + case OpCode::Id::TLD: { + UNIMPLEMENTED_IF_MSG(instr.tld.aoffi, "AOFFI is not implemented"); + UNIMPLEMENTED_IF_MSG(instr.tld.ms, "MS is not implemented"); + UNIMPLEMENTED_IF_MSG(instr.tld.cl, "CL is not implemented"); + + if (instr.tld.nodep_flag) { + LOG_WARNING(HW_GPU, "TLD.NODEP implementation is incomplete"); + } + + WriteTexInstructionFloat(bb, instr, GetTldCode(instr)); + break; + } case OpCode::Id::TLDS: { const Tegra::Shader::TextureType texture_type{instr.tlds.GetTextureType()}; const bool is_array{instr.tlds.IsArrayTexture()}; @@ -575,6 +587,38 @@ Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool de return values; } +Node4 ShaderIR::GetTldCode(Tegra::Shader::Instruction instr) { + const auto texture_type{instr.tld.texture_type}; + const bool is_array{instr.tld.is_array}; + const bool lod_enabled{instr.tld.GetTextureProcessMode() == TextureProcessMode::LL}; + const std::size_t coord_count{GetCoordCount(texture_type)}; + + u64 gpr8_cursor{instr.gpr8.Value()}; + const Node array_register{is_array ? GetRegister(gpr8_cursor++) : nullptr}; + + std::vector coords; + for (std::size_t i = 0; i < coord_count; ++i) { + coords.push_back(GetRegister(gpr8_cursor++)); + } + + u64 gpr20_cursor{instr.gpr20.Value()}; + // const Node bindless_register{is_bindless ? GetRegister(gpr20_cursor++) : nullptr}; + const Node lod{lod_enabled ? GetRegister(gpr20_cursor++) : Immediate(0u)}; + // const Node aoffi_register{is_aoffi ? GetRegister(gpr20_cursor++) : nullptr}; + // const Node multisample{is_multisample ? GetRegister(gpr20_cursor++) : nullptr}; + + const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false); + + Node4 values; + for (u32 element = 0; element < values.size(); ++element) { + auto coords_copy = coords; + MetaTexture meta{sampler, array_register, {}, {}, {}, lod, {}, element}; + values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy)); + } + + return values; +} + Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is_array) { const std::size_t type_coord_count = GetCoordCount(texture_type); const bool lod_enabled = instr.tlds.GetTextureProcessMode() == TextureProcessMode::LL; diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h index edcf2288e..1b84c0672 100644 --- a/src/video_core/shader/shader_ir.h +++ b/src/video_core/shader/shader_ir.h @@ -277,6 +277,8 @@ private: Node4 GetTld4Code(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type, bool depth_compare, bool is_array, bool is_aoffi); + Node4 GetTldCode(Tegra::Shader::Instruction instr); + Node4 GetTldsCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type, bool is_array); -- cgit v1.2.3 From b8c75a845b1784045a10fa8b5f1f57f2ec53eeca Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Sun, 28 Apr 2019 01:01:22 -0300 Subject: maxwell_3d: Partially implement texture buffers as 1D textures --- src/video_core/engines/maxwell_3d.cpp | 12 ++++-------- src/video_core/renderer_opengl/gl_rasterizer_cache.cpp | 2 +- src/video_core/surface.cpp | 2 ++ src/video_core/textures/texture.h | 18 +++++++++++++++++- 4 files changed, 24 insertions(+), 10 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 08d553696..8755b8af4 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -430,14 +430,10 @@ Texture::TICEntry Maxwell3D::GetTICEntry(u32 tic_index) const { Texture::TICEntry tic_entry; memory_manager.ReadBlockUnsafe(tic_address_gpu, &tic_entry, sizeof(Texture::TICEntry)); - ASSERT_MSG(tic_entry.header_version == Texture::TICHeaderVersion::BlockLinear || - tic_entry.header_version == Texture::TICHeaderVersion::Pitch, - "TIC versions other than BlockLinear or Pitch are unimplemented"); - - const auto r_type = tic_entry.r_type.Value(); - const auto g_type = tic_entry.g_type.Value(); - const auto b_type = tic_entry.b_type.Value(); - const auto a_type = tic_entry.a_type.Value(); + const auto r_type{tic_entry.r_type.Value()}; + const auto g_type{tic_entry.g_type.Value()}; + const auto b_type{tic_entry.b_type.Value()}; + const auto a_type{tic_entry.a_type.Value()}; // TODO(Subv): Different data types for separate components are not supported DEBUG_ASSERT(r_type == g_type && r_type == b_type && r_type == a_type); diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp index a7681902e..543b36271 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp @@ -140,7 +140,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only, params.width = Common::AlignUp(config.tic.Width(), GetCompressionFactor(params.pixel_format)); params.height = Common::AlignUp(config.tic.Height(), GetCompressionFactor(params.pixel_format)); - if (!params.is_tiled) { + if (config.tic.IsLineal()) { params.pitch = config.tic.Pitch(); } params.unaligned_height = config.tic.Height(); diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp index 6384fa8d2..56c43af17 100644 --- a/src/video_core/surface.cpp +++ b/src/video_core/surface.cpp @@ -12,6 +12,8 @@ SurfaceTarget SurfaceTargetFromTextureType(Tegra::Texture::TextureType texture_t switch (texture_type) { case Tegra::Texture::TextureType::Texture1D: return SurfaceTarget::Texture1D; + case Tegra::Texture::TextureType::Texture1DBuffer: + return SurfaceTarget::Texture1D; // Fixme case Tegra::Texture::TextureType::Texture2D: case Tegra::Texture::TextureType::Texture2DNoMipmap: return SurfaceTarget::Texture2D; diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h index f22b4e7c7..ddeed73d0 100644 --- a/src/video_core/textures/texture.h +++ b/src/video_core/textures/texture.h @@ -172,12 +172,16 @@ struct TICEntry { BitField<26, 1, u32> use_header_opt_control; BitField<27, 1, u32> depth_texture; BitField<28, 4, u32> max_mip_level; + + BitField<0, 16, u32> buffer_high_width_minus_one; }; union { BitField<0, 16, u32> width_minus_1; BitField<22, 1, u32> srgb_conversion; BitField<23, 4, TextureType> texture_type; BitField<29, 3, u32> border_size; + + BitField<0, 16, u32> buffer_low_width_minus_one; }; union { BitField<0, 16, u32> height_minus_1; @@ -206,7 +210,10 @@ struct TICEntry { } u32 Width() const { - return width_minus_1 + 1; + if (header_version != TICHeaderVersion::OneDBuffer) { + return width_minus_1 + 1; + } + return (buffer_high_width_minus_one << 16) | buffer_low_width_minus_one; } u32 Height() const { @@ -237,6 +244,15 @@ struct TICEntry { header_version == TICHeaderVersion::BlockLinearColorKey; } + bool IsLineal() const { + return header_version == TICHeaderVersion::Pitch || + header_version == TICHeaderVersion::PitchColorKey; + } + + bool IsBuffer() const { + return header_version == TICHeaderVersion::OneDBuffer; + } + bool IsSrgbConversionEnabled() const { return srgb_conversion != 0; } -- cgit v1.2.3 From 06c4ce86458310870abec90ada68ac393256b9b6 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Sat, 27 Apr 2019 02:07:18 -0300 Subject: shader: Decode SUST and implement backing image functionality --- CMakeModules/GenerateSCMRev.cmake | 1 + src/common/CMakeLists.txt | 1 + src/video_core/CMakeLists.txt | 1 + src/video_core/engines/shader_bytecode.h | 66 +++++++++++++++- .../renderer_opengl/gl_shader_decompiler.cpp | 70 +++++++++++++++++ .../renderer_vulkan/vk_shader_decompiler.cpp | 7 ++ src/video_core/shader/decode.cpp | 1 + src/video_core/shader/decode/image.cpp | 89 ++++++++++++++++++++++ src/video_core/shader/node.h | 42 +++++++++- src/video_core/shader/shader_ir.h | 9 +++ 10 files changed, 284 insertions(+), 3 deletions(-) create mode 100644 src/video_core/shader/decode/image.cpp (limited to 'src/video_core/engines') diff --git a/CMakeModules/GenerateSCMRev.cmake b/CMakeModules/GenerateSCMRev.cmake index 31edeb63d..dd65cfe42 100644 --- a/CMakeModules/GenerateSCMRev.cmake +++ b/CMakeModules/GenerateSCMRev.cmake @@ -70,6 +70,7 @@ set(HASH_FILES "${VIDEO_CORE}/shader/decode/half_set.cpp" "${VIDEO_CORE}/shader/decode/half_set_predicate.cpp" "${VIDEO_CORE}/shader/decode/hfma2.cpp" + "${VIDEO_CORE}/shader/decode/image.cpp" "${VIDEO_CORE}/shader/decode/integer_set.cpp" "${VIDEO_CORE}/shader/decode/integer_set_predicate.cpp" "${VIDEO_CORE}/shader/decode/memory.cpp" diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt index 198b3fe07..8ae05137b 100644 --- a/src/common/CMakeLists.txt +++ b/src/common/CMakeLists.txt @@ -44,6 +44,7 @@ add_custom_command(OUTPUT scm_rev.cpp "${VIDEO_CORE}/shader/decode/half_set.cpp" "${VIDEO_CORE}/shader/decode/half_set_predicate.cpp" "${VIDEO_CORE}/shader/decode/hfma2.cpp" + "${VIDEO_CORE}/shader/decode/image.cpp" "${VIDEO_CORE}/shader/decode/integer_set.cpp" "${VIDEO_CORE}/shader/decode/integer_set_predicate.cpp" "${VIDEO_CORE}/shader/decode/memory.cpp" diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 9d43f03d2..6839abe71 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -90,6 +90,7 @@ add_library(video_core STATIC shader/decode/conversion.cpp shader/decode/memory.cpp shader/decode/texture.cpp + shader/decode/image.cpp shader/decode/float_set_predicate.cpp shader/decode/integer_set_predicate.cpp shader/decode/half_set_predicate.cpp diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h index 5b32e1249..54a1a04f9 100644 --- a/src/video_core/engines/shader_bytecode.h +++ b/src/video_core/engines/shader_bytecode.h @@ -126,6 +126,15 @@ union Sampler { u64 value{}; }; +union Image { + Image() = default; + + constexpr explicit Image(u64 value) : value{value} {} + + BitField<36, 13, u64> index; + u64 value; +}; + } // namespace Tegra::Shader namespace std { @@ -344,6 +353,26 @@ enum class TextureMiscMode : u64 { PTP, }; +enum class SurfaceDataMode : u64 { + P = 0, + D_BA = 1, +}; + +enum class OutOfBoundsStore : u64 { + Ignore = 0, + Clamp = 1, + Trap = 2, +}; + +enum class ImageType : u64 { + Texture1D = 0, + TextureBuffer = 1, + Texture1DArray = 2, + Texture2D = 3, + Texture2DArray = 4, + Texture3D = 5, +}; + enum class IsberdMode : u64 { None = 0, Patch = 1, @@ -398,7 +427,7 @@ enum class LmemLoadCacheManagement : u64 { CV = 3, }; -enum class LmemStoreCacheManagement : u64 { +enum class StoreCacheManagement : u64 { Default = 0, CG = 1, CS = 2, @@ -811,7 +840,7 @@ union Instruction { } ld_l; union { - BitField<44, 2, LmemStoreCacheManagement> cache_management; + BitField<44, 2, StoreCacheManagement> cache_management; } st_l; union { @@ -1294,6 +1323,35 @@ union Instruction { } } tlds; + union { + BitField<24, 2, StoreCacheManagement> cache_management; + BitField<33, 3, ImageType> image_type; + BitField<49, 2, OutOfBoundsStore> out_of_bounds_store; + BitField<51, 1, u64> is_immediate; + BitField<52, 1, SurfaceDataMode> mode; + + BitField<20, 3, StoreType> store_data_layout; + BitField<20, 4, u64> component_mask_selector; + + bool IsComponentEnabled(std::size_t component) const { + ASSERT(mode == SurfaceDataMode::P); + constexpr u8 R = 0b0001; + constexpr u8 G = 0b0010; + constexpr u8 B = 0b0100; + constexpr u8 A = 0b1000; + constexpr std::array mask = { + 0, (R), (G), (R | G), (B), (R | B), + (G | B), (R | G | B), (A), (R | A), (G | A), (R | G | A), + (B | A), (R | B | A), (G | B | A), (R | G | B | A)}; + return std::bitset<4>{mask.at(component_mask_selector)}.test(component); + } + + StoreType GetStoreDataLayout() const { + ASSERT(mode == SurfaceDataMode::D_BA); + return store_data_layout; + } + } sust; + union { BitField<20, 24, u64> target; BitField<5, 1, u64> constant_buffer; @@ -1385,6 +1443,7 @@ union Instruction { Attribute attribute; Sampler sampler; + Image image; u64 value; }; @@ -1428,6 +1487,7 @@ public: TLD4S, // Texture Load 4 with scalar / non - vec4 source / destinations TMML_B, // Texture Mip Map Level TMML, // Texture Mip Map Level + SUST, // Surface Store EXIT, IPA, OUT_R, // Emit vertex/primitive @@ -1558,6 +1618,7 @@ public: Synch, Memory, Texture, + Image, FloatSet, FloatSetPredicate, IntegerSet, @@ -1703,6 +1764,7 @@ private: INST("1101111100------", Id::TLD4S, Type::Texture, "TLD4S"), INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"), INST("1101111101011---", Id::TMML, Type::Texture, "TMML"), + INST("11101011001-----", Id::SUST, Type::Image, "SUST"), INST("11100000--------", Id::IPA, Type::Trivial, "IPA"), INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"), INST("1110111111010---", Id::ISBERD, Type::Trivial, "ISBERD"), diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index ece386cdc..2ae2f1db2 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -180,6 +180,7 @@ public: DeclareGlobalMemory(); DeclareSamplers(); DeclarePhysicalAttributeReader(); + DeclareImages(); code.AddLine("void execute_{}() {{", suffix); ++code.scope; @@ -531,6 +532,36 @@ private: code.AddNewLine(); } + void DeclareImages() { + const auto& images{ir.GetImages()}; + for (const auto& image : images) { + const std::string image_type = [&]() { + switch (image.GetType()) { + case Tegra::Shader::ImageType::Texture1D: + return "image1D"; + case Tegra::Shader::ImageType::TextureBuffer: + return "bufferImage"; + case Tegra::Shader::ImageType::Texture1DArray: + return "image1DArray"; + case Tegra::Shader::ImageType::Texture2D: + return "image2D"; + case Tegra::Shader::ImageType::Texture2DArray: + return "image2DArray"; + case Tegra::Shader::ImageType::Texture3D: + return "image3D"; + default: + UNREACHABLE(); + return "image1D"; + } + }(); + code.AddLine("layout (binding = IMAGE_BINDING_" + std::to_string(image.GetIndex()) + + ") coherent volatile writeonly uniform " + image_type + ' ' + + GetImage(image) + ';'); + } + if (!images.empty()) + code.AddNewLine(); + } + void VisitBlock(const NodeBlock& bb) { for (const auto& node : bb) { if (const std::string expr = Visit(node); !expr.empty()) { @@ -1478,6 +1509,39 @@ private: return tmp; } + std::string ImageStore(Operation operation) { + constexpr std::array constructors{"int(", "ivec2(", "ivec3(", "ivec4("}; + const auto meta{std::get(operation.GetMeta())}; + + std::string expr = "imageStore("; + expr += GetImage(meta.image); + expr += ", "; + + const std::size_t coords_count{operation.GetOperandsCount()}; + expr += constructors.at(coords_count - 1); + for (std::size_t i = 0; i < coords_count; ++i) { + expr += VisitOperand(operation, i, Type::Int); + if (i + 1 < coords_count) { + expr += ", "; + } + } + expr += "), "; + + const std::size_t values_count{meta.values.size()}; + UNIMPLEMENTED_IF(values_count != 4); + expr += "vec4("; + for (std::size_t i = 0; i < values_count; ++i) { + expr += Visit(meta.values.at(i)); + if (i + 1 < values_count) { + expr += ", "; + } + } + expr += "));"; + + code.AddLine(expr); + return {}; + } + std::string Branch(Operation operation) { const auto target = std::get_if(&*operation[0]); UNIMPLEMENTED_IF(!target); @@ -1718,6 +1782,8 @@ private: &GLSLDecompiler::TextureQueryLod, &GLSLDecompiler::TexelFetch, + &GLSLDecompiler::ImageStore, + &GLSLDecompiler::Branch, &GLSLDecompiler::PushFlowStack, &GLSLDecompiler::PopFlowStack, @@ -1786,6 +1852,10 @@ private: return GetDeclarationWithSuffix(static_cast(sampler.GetIndex()), "sampler"); } + std::string GetImage(const Image& image) const { + return GetDeclarationWithSuffix(static_cast(image.GetIndex()), "image"); + } + void EmitIfdefIsBuffer(const Sampler& sampler) { code.AddLine(fmt::format("#ifdef SAMPLER_{}_IS_BUFFER", sampler.GetIndex())); } diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp index 33ad9764a..97ce214b1 100644 --- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp @@ -935,6 +935,11 @@ private: return {}; } + Id ImageStore(Operation operation) { + UNIMPLEMENTED(); + return {}; + } + Id Branch(Operation operation) { const auto target = std::get_if(&*operation[0]); UNIMPLEMENTED_IF(!target); @@ -1326,6 +1331,8 @@ private: &SPIRVDecompiler::TextureQueryLod, &SPIRVDecompiler::TexelFetch, + &SPIRVDecompiler::ImageStore, + &SPIRVDecompiler::Branch, &SPIRVDecompiler::PushFlowStack, &SPIRVDecompiler::PopFlowStack, diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp index a0554c97e..2c9ff28f2 100644 --- a/src/video_core/shader/decode.cpp +++ b/src/video_core/shader/decode.cpp @@ -169,6 +169,7 @@ u32 ShaderIR::DecodeInstr(NodeBlock& bb, u32 pc) { {OpCode::Type::Conversion, &ShaderIR::DecodeConversion}, {OpCode::Type::Memory, &ShaderIR::DecodeMemory}, {OpCode::Type::Texture, &ShaderIR::DecodeTexture}, + {OpCode::Type::Image, &ShaderIR::DecodeImage}, {OpCode::Type::FloatSetPredicate, &ShaderIR::DecodeFloatSetPredicate}, {OpCode::Type::IntegerSetPredicate, &ShaderIR::DecodeIntegerSetPredicate}, {OpCode::Type::HalfSetPredicate, &ShaderIR::DecodeHalfSetPredicate}, diff --git a/src/video_core/shader/decode/image.cpp b/src/video_core/shader/decode/image.cpp new file mode 100644 index 000000000..66fdf5714 --- /dev/null +++ b/src/video_core/shader/decode/image.cpp @@ -0,0 +1,89 @@ +// Copyright 2019 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include + +#include "common/assert.h" +#include "common/common_types.h" +#include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/shader_ir.h" + +namespace VideoCommon::Shader { + +using Tegra::Shader::Instruction; +using Tegra::Shader::OpCode; + +namespace { +std::size_t GetImageTypeNumCoordinates(Tegra::Shader::ImageType image_type) { + switch (image_type) { + case Tegra::Shader::ImageType::Texture1D: + case Tegra::Shader::ImageType::TextureBuffer: + return 1; + case Tegra::Shader::ImageType::Texture1DArray: + case Tegra::Shader::ImageType::Texture2D: + return 2; + case Tegra::Shader::ImageType::Texture2DArray: + case Tegra::Shader::ImageType::Texture3D: + return 3; + } + UNREACHABLE(); + return 1; +} +} // Anonymous namespace + +u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) { + const Instruction instr = {program_code[pc]}; + const auto opcode = OpCode::Decode(instr); + + switch (opcode->get().GetId()) { + case OpCode::Id::SUST: { + UNIMPLEMENTED_IF(instr.sust.mode != Tegra::Shader::SurfaceDataMode::P); + UNIMPLEMENTED_IF(instr.sust.image_type == Tegra::Shader::ImageType::TextureBuffer); + UNIMPLEMENTED_IF(instr.sust.out_of_bounds_store != Tegra::Shader::OutOfBoundsStore::Ignore); + UNIMPLEMENTED_IF(instr.sust.component_mask_selector != 0xf); // Ensure we have an RGBA store + + std::vector values; + constexpr std::size_t hardcoded_size{4}; + for (std::size_t i = 0; i < hardcoded_size; ++i) { + values.push_back(GetRegister(instr.gpr0.Value() + i)); + } + + std::vector coords; + const std::size_t num_coords{GetImageTypeNumCoordinates(instr.sust.image_type)}; + for (std::size_t i = 0; i < num_coords; ++i) { + coords.push_back(GetRegister(instr.gpr8.Value() + i)); + } + + ASSERT(instr.sust.is_immediate); + const auto& image{GetImage(instr.image, instr.sust.image_type)}; + MetaImage meta{image, values}; + const Node store{Operation(OperationCode::ImageStore, meta, std::move(coords))}; + bb.push_back(store); + break; + } + default: + UNIMPLEMENTED_MSG("Unhandled conversion instruction: {}", opcode->get().GetName()); + } + + return pc; +} + +const Image& ShaderIR::GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type) { + const auto offset{static_cast(image.index.Value())}; + + // If this image has already been used, return the existing mapping. + const auto itr{std::find_if(used_images.begin(), used_images.end(), + [=](const Image& entry) { return entry.GetOffset() == offset; })}; + if (itr != used_images.end()) { + ASSERT(itr->GetType() == type); + return *itr; + } + + // Otherwise create a new mapping for this image. + const std::size_t next_index{used_images.size()}; + const Image entry{offset, next_index, type}; + return *used_images.emplace(entry).first; +} + +} // namespace VideoCommon::Shader diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h index 3cfb911bb..8b8d83ae7 100644 --- a/src/video_core/shader/node.h +++ b/src/video_core/shader/node.h @@ -146,6 +146,8 @@ enum class OperationCode { TextureQueryLod, /// (MetaTexture, float[N] coords) -> float4 TexelFetch, /// (MetaTexture, int[N], int) -> float4 + ImageStore, /// (MetaImage, float[N] coords) -> void + Branch, /// (uint branch_target) -> void PushFlowStack, /// (uint branch_target) -> void PopFlowStack, /// () -> void @@ -263,6 +265,39 @@ private: bool is_bindless{}; ///< Whether this sampler belongs to a bindless texture or not. }; +class Image { +public: + explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type) + : offset{offset}, index{index}, type{type}, is_bindless{false} {} + + std::size_t GetOffset() const { + return offset; + } + + std::size_t GetIndex() const { + return index; + } + + Tegra::Shader::ImageType GetType() const { + return type; + } + + bool IsBindless() const { + return is_bindless; + } + + bool operator<(const Image& rhs) const { + return std::tie(offset, index, type, is_bindless) < + std::tie(rhs.offset, rhs.index, rhs.type, rhs.is_bindless); + } + +private: + std::size_t offset{}; + std::size_t index{}; + Tegra::Shader::ImageType type{}; + bool is_bindless{}; +}; + struct GlobalMemoryBase { u32 cbuf_index{}; u32 cbuf_offset{}; @@ -289,8 +324,13 @@ struct MetaTexture { u32 element{}; }; +struct MetaImage { + const Image& image; + std::vector values; +}; + /// Parameters that modify an operation but are not part of any particular operand -using Meta = std::variant; +using Meta = std::variant; /// Holds any kind of operation that can be done in the IR class OperationNode final { diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h index 1b84c0672..c7f264371 100644 --- a/src/video_core/shader/shader_ir.h +++ b/src/video_core/shader/shader_ir.h @@ -104,6 +104,10 @@ public: return used_samplers; } + const std::set& GetImages() const { + return used_images; + } + const std::array& GetClipDistances() const { return used_clip_distances; @@ -154,6 +158,7 @@ private: u32 DecodeConversion(NodeBlock& bb, u32 pc); u32 DecodeMemory(NodeBlock& bb, u32 pc); u32 DecodeTexture(NodeBlock& bb, u32 pc); + u32 DecodeImage(NodeBlock& bb, u32 pc); u32 DecodeFloatSetPredicate(NodeBlock& bb, u32 pc); u32 DecodeIntegerSetPredicate(NodeBlock& bb, u32 pc); u32 DecodeHalfSetPredicate(NodeBlock& bb, u32 pc); @@ -254,6 +259,9 @@ private: Tegra::Shader::TextureType type, bool is_array, bool is_shadow); + /// Accesses an image. + const Image& GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type); + /// Extracts a sequence of bits from a node Node BitfieldExtract(Node value, u32 offset, u32 bits); @@ -329,6 +337,7 @@ private: std::set used_output_attributes; std::map used_cbufs; std::set used_samplers; + std::set used_images; std::array used_clip_distances{}; std::map used_global_memory; bool uses_physical_attributes{}; // Shader uses AL2P or physical attribute read/writes -- cgit v1.2.3 From 175aa343ff1c9f931b266caf2d19b8df943dab0d Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Sat, 18 May 2019 04:57:49 -0400 Subject: texture_cache: Fermi2D reform and implement View Mirage This also does some fixes on compressed textures reinterpret and on the Fermi2D engine in general. --- src/video_core/engines/fermi_2d.cpp | 10 ++-- src/video_core/engines/fermi_2d.h | 44 +++++++++++++----- src/video_core/rasterizer_interface.h | 3 +- .../renderer_opengl/gl_framebuffer_cache.cpp | 8 ++-- src/video_core/renderer_opengl/gl_rasterizer.cpp | 5 +- src/video_core/renderer_opengl/gl_rasterizer.h | 3 +- .../renderer_opengl/gl_texture_cache.cpp | 54 ++++++++++------------ src/video_core/renderer_opengl/gl_texture_cache.h | 6 +-- src/video_core/texture_cache/surface_base.h | 15 ++++-- src/video_core/texture_cache/surface_params.h | 14 ++++++ src/video_core/texture_cache/texture_cache.h | 40 ++++++++++------ 11 files changed, 125 insertions(+), 77 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp index 55966eef1..d63b82838 100644 --- a/src/video_core/engines/fermi_2d.cpp +++ b/src/video_core/engines/fermi_2d.cpp @@ -4,7 +4,6 @@ #include "common/assert.h" #include "common/logging/log.h" -#include "common/math_util.h" #include "video_core/engines/fermi_2d.h" #include "video_core/memory_manager.h" #include "video_core/rasterizer_interface.h" @@ -35,7 +34,7 @@ void Fermi2D::HandleSurfaceCopy() { static_cast(regs.operation)); // TODO(Subv): Only raw copies are implemented. - ASSERT(regs.operation == Regs::Operation::SrcCopy); + ASSERT(regs.operation == Operation::SrcCopy); const u32 src_blit_x1{static_cast(regs.blit_src_x >> 32)}; const u32 src_blit_y1{static_cast(regs.blit_src_y >> 32)}; @@ -48,8 +47,13 @@ void Fermi2D::HandleSurfaceCopy() { const Common::Rectangle dst_rect{regs.blit_dst_x, regs.blit_dst_y, regs.blit_dst_x + regs.blit_dst_width, regs.blit_dst_y + regs.blit_dst_height}; + Config copy_config; + copy_config.operation = regs.operation; + copy_config.filter = regs.blit_control.filter; + copy_config.src_rect = src_rect; + copy_config.dst_rect = dst_rect; - if (!rasterizer.AccelerateSurfaceCopy(regs.src, regs.dst, src_rect, dst_rect)) { + if (!rasterizer.AccelerateSurfaceCopy(regs.src, regs.dst, copy_config)) { UNIMPLEMENTED(); } } diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h index 3d28afa91..0a4c7c5ad 100644 --- a/src/video_core/engines/fermi_2d.h +++ b/src/video_core/engines/fermi_2d.h @@ -9,6 +9,7 @@ #include "common/bit_field.h" #include "common/common_funcs.h" #include "common/common_types.h" +#include "common/math_util.h" #include "video_core/gpu.h" namespace Tegra { @@ -38,6 +39,26 @@ public: /// Write the value to the register identified by method. void CallMethod(const GPU::MethodCall& method_call); + enum class Origin : u32 { + Center = 0, + Corner = 1, + }; + + enum class Filter : u32 { + PointSample = 0, // Nearest + Linear = 1, + }; + + enum class Operation : u32 { + SrcCopyAnd = 0, + ROPAnd = 1, + Blend = 2, + SrcCopy = 3, + ROP = 4, + SrcCopyPremult = 5, + BlendPremult = 6, + }; + struct Regs { static constexpr std::size_t NUM_REGS = 0x258; @@ -76,16 +97,6 @@ public: }; static_assert(sizeof(Surface) == 0x28, "Surface has incorrect size"); - enum class Operation : u32 { - SrcCopyAnd = 0, - ROPAnd = 1, - Blend = 2, - SrcCopy = 3, - ROP = 4, - SrcCopyPremult = 5, - BlendPremult = 6, - }; - union { struct { INSERT_PADDING_WORDS(0x80); @@ -102,7 +113,11 @@ public: INSERT_PADDING_WORDS(0x177); - u32 blit_control; + union { + u32 raw; + BitField<0, 1, Origin> origin; + BitField<4, 1, Filter> filter; + } blit_control; INSERT_PADDING_WORDS(0x8); @@ -121,6 +136,13 @@ public: }; } regs{}; + struct Config { + Operation operation; + Filter filter; + Common::Rectangle src_rect; + Common::Rectangle dst_rect; + }; + private: VideoCore::RasterizerInterface& rasterizer; MemoryManager& memory_manager; diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h index 3c18d3b1f..6007e8c2e 100644 --- a/src/video_core/rasterizer_interface.h +++ b/src/video_core/rasterizer_interface.h @@ -52,8 +52,7 @@ public: /// Attempt to use a faster method to perform a surface copy virtual bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src, const Tegra::Engines::Fermi2D::Regs::Surface& dst, - const Common::Rectangle& src_rect, - const Common::Rectangle& dst_rect) { + const Tegra::Engines::Fermi2D::Config& copy_config) { return false; } diff --git a/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp b/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp index bb9f9b81f..7c926bd48 100644 --- a/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp @@ -37,7 +37,7 @@ OGLFramebuffer FramebufferCacheOpenGL::CreateFramebuffer(const FramebufferCacheK if (key.is_single_buffer) { if (key.color_attachments[0] != GL_NONE && key.colors[0]) { - key.colors[0]->Attach(key.color_attachments[0]); + key.colors[0]->Attach(key.color_attachments[0], GL_DRAW_FRAMEBUFFER); glDrawBuffer(key.color_attachments[0]); } else { glDrawBuffer(GL_NONE); @@ -45,14 +45,16 @@ OGLFramebuffer FramebufferCacheOpenGL::CreateFramebuffer(const FramebufferCacheK } else { for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) { if (key.colors[index]) { - key.colors[index]->Attach(GL_COLOR_ATTACHMENT0 + static_cast(index)); + key.colors[index]->Attach(GL_COLOR_ATTACHMENT0 + static_cast(index), + GL_DRAW_FRAMEBUFFER); } } glDrawBuffers(key.colors_count, key.color_attachments.data()); } if (key.zeta) { - key.zeta->Attach(key.stencil_enable ? GL_DEPTH_STENCIL_ATTACHMENT : GL_DEPTH_ATTACHMENT); + key.zeta->Attach(key.stencil_enable ? GL_DEPTH_STENCIL_ATTACHMENT : GL_DEPTH_ATTACHMENT, + GL_DRAW_FRAMEBUFFER); } return framebuffer; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 4f7eeb22c..d613cb1dc 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -696,10 +696,9 @@ void RasterizerOpenGL::FlushAndInvalidateRegion(CacheAddr addr, u64 size) { bool RasterizerOpenGL::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src, const Tegra::Engines::Fermi2D::Regs::Surface& dst, - const Common::Rectangle& src_rect, - const Common::Rectangle& dst_rect) { + const Tegra::Engines::Fermi2D::Config& copy_config) { MICROPROFILE_SCOPE(OpenGL_Blits); - texture_cache.DoFermiCopy(src, dst, src_rect, dst_rect); + texture_cache.DoFermiCopy(src, dst, copy_config); return true; } diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 64c27660f..33582ac42 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -67,8 +67,7 @@ public: void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override; bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src, const Tegra::Engines::Fermi2D::Regs::Surface& dst, - const Common::Rectangle& src_rect, - const Common::Rectangle& dst_rect) override; + 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; diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index 197c9f02c..9e9734f9e 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -378,26 +378,26 @@ CachedSurfaceView::CachedSurfaceView(CachedSurface& surface, const ViewParams& p CachedSurfaceView::~CachedSurfaceView() = default; -void CachedSurfaceView::Attach(GLenum attachment) const { +void CachedSurfaceView::Attach(GLenum attachment, GLenum target) const { ASSERT(params.num_layers == 1 && params.num_levels == 1); const auto& owner_params = surface.GetSurfaceParams(); switch (owner_params.target) { case SurfaceTarget::Texture1D: - glFramebufferTexture1D(GL_DRAW_FRAMEBUFFER, attachment, surface.GetTarget(), - surface.GetTexture(), params.base_level); + glFramebufferTexture1D(target, attachment, surface.GetTarget(), surface.GetTexture(), + params.base_level); break; case SurfaceTarget::Texture2D: - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, attachment, surface.GetTarget(), - surface.GetTexture(), params.base_level); + glFramebufferTexture2D(target, attachment, surface.GetTarget(), surface.GetTexture(), + params.base_level); break; case SurfaceTarget::Texture1DArray: case SurfaceTarget::Texture2DArray: case SurfaceTarget::TextureCubemap: case SurfaceTarget::TextureCubeArray: - glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, attachment, surface.GetTexture(), - params.base_level, params.base_layer); + glFramebufferTextureLayer(target, attachment, surface.GetTexture(), params.base_level, + params.base_layer); break; default: UNIMPLEMENTED(); @@ -460,11 +460,10 @@ void TextureCacheOpenGL::ImageCopy(Surface src_surface, Surface dst_surface, copy_params.depth); } -void TextureCacheOpenGL::ImageBlit(Surface src_surface, Surface dst_surface, - const Common::Rectangle& src_rect, - const Common::Rectangle& dst_rect) { - const auto& src_params{src_surface->GetSurfaceParams()}; - const auto& dst_params{dst_surface->GetSurfaceParams()}; +void TextureCacheOpenGL::ImageBlit(View src_view, View dst_view, + const Tegra::Engines::Fermi2D::Config& copy_config) { + const auto& src_params{src_view->GetSurfaceParams()}; + const auto& dst_params{dst_view->GetSurfaceParams()}; OpenGLState prev_state{OpenGLState::GetCurState()}; SCOPE_EXIT({ prev_state.Apply(); }); @@ -476,51 +475,46 @@ void TextureCacheOpenGL::ImageBlit(Surface src_surface, Surface dst_surface, u32 buffers{}; - UNIMPLEMENTED_IF(src_params.target != SurfaceTarget::Texture2D); - UNIMPLEMENTED_IF(dst_params.target != SurfaceTarget::Texture2D); - - const GLuint src_texture{src_surface->GetTexture()}; - const GLuint dst_texture{dst_surface->GetTexture()}; + UNIMPLEMENTED_IF(src_params.target == SurfaceTarget::Texture3D); + UNIMPLEMENTED_IF(dst_params.target == SurfaceTarget::Texture3D); if (src_params.type == SurfaceType::ColorTexture) { - glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, - src_texture, 0); + src_view->Attach(GL_COLOR_ATTACHMENT0, GL_READ_FRAMEBUFFER); glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, - dst_texture, 0); + dst_view->Attach(GL_COLOR_ATTACHMENT0, GL_DRAW_FRAMEBUFFER); glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); buffers = GL_COLOR_BUFFER_BIT; } else if (src_params.type == SurfaceType::Depth) { glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0); - glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, src_texture, - 0); + src_view->Attach(GL_DEPTH_ATTACHMENT, GL_READ_FRAMEBUFFER); glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0); - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, dst_texture, - 0); + dst_view->Attach(GL_DEPTH_ATTACHMENT, GL_DRAW_FRAMEBUFFER); glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); buffers = GL_DEPTH_BUFFER_BIT; } else if (src_params.type == SurfaceType::DepthStencil) { glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0); - glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, - src_texture, 0); + src_view->Attach(GL_DEPTH_STENCIL_ATTACHMENT, GL_READ_FRAMEBUFFER); glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0); - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, - dst_texture, 0); + dst_view->Attach(GL_DEPTH_STENCIL_ATTACHMENT, GL_DRAW_FRAMEBUFFER); buffers = GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT; } + const Common::Rectangle& src_rect = copy_config.src_rect; + const Common::Rectangle& dst_rect = copy_config.dst_rect; + const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear; + glBlitFramebuffer(src_rect.left, src_rect.top, src_rect.right, src_rect.bottom, dst_rect.left, dst_rect.top, dst_rect.right, dst_rect.bottom, buffers, - buffers == GL_COLOR_BUFFER_BIT ? GL_LINEAR : GL_NEAREST); + is_linear ? GL_LINEAR : GL_NEAREST); } } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h index 0ba42dbab..0b333e9e3 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.h +++ b/src/video_core/renderer_opengl/gl_texture_cache.h @@ -73,7 +73,7 @@ public: ~CachedSurfaceView(); /// Attaches this texture view to the current bound GL_DRAW_FRAMEBUFFER - void Attach(GLenum attachment) const; + void Attach(GLenum attachment, GLenum target) const; GLuint GetTexture() { if (is_proxy) { @@ -138,8 +138,8 @@ protected: void ImageCopy(Surface src_surface, Surface dst_surface, const VideoCommon::CopyParams& copy_params) override; - void ImageBlit(Surface src_surface, Surface dst_surface, const Common::Rectangle& src_rect, - const Common::Rectangle& dst_rect) override; + void ImageBlit(View src_view, View dst_view, + const Tegra::Engines::Fermi2D::Config& copy_config) override; private: OGLFramebuffer src_framebuffer; diff --git a/src/video_core/texture_cache/surface_base.h b/src/video_core/texture_cache/surface_base.h index cb7f22706..a3dd1c607 100644 --- a/src/video_core/texture_cache/surface_base.h +++ b/src/video_core/texture_cache/surface_base.h @@ -126,14 +126,19 @@ public: return MatchStructureResult::None; } // Tiled surface - if (std::tie(params.height, params.depth, params.block_width, params.block_height, - params.block_depth, params.tile_width_spacing, params.num_levels) == - std::tie(rhs.height, rhs.depth, rhs.block_width, rhs.block_height, rhs.block_depth, + if (std::tie(params.depth, params.block_width, params.block_height, params.block_depth, + params.tile_width_spacing, params.num_levels) == + std::tie(rhs.depth, rhs.block_width, rhs.block_height, rhs.block_depth, rhs.tile_width_spacing, rhs.num_levels)) { - if (params.width == rhs.width) { + if (std::tie(params.width, params.height) == std::tie(rhs.width, rhs.height)) { return MatchStructureResult::FullMatch; } - if (params.GetBlockAlignedWidth() == rhs.GetBlockAlignedWidth()) { + const u32 ws = SurfaceParams::ConvertWidth(rhs.GetBlockAlignedWidth(), + params.pixel_format, rhs.pixel_format); + const u32 hs = + SurfaceParams::ConvertHeight(rhs.height, params.pixel_format, rhs.pixel_format); + const u32 w1 = params.GetBlockAlignedWidth(); + if (std::tie(w1, params.height) == std::tie(ws, hs)) { return MatchStructureResult::SemiMatch; } } diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h index b3082173f..13a08a60f 100644 --- a/src/video_core/texture_cache/surface_params.h +++ b/src/video_core/texture_cache/surface_params.h @@ -126,6 +126,20 @@ public: /// Returns the size of a layer in bytes in host memory for a given mipmap level. std::size_t GetHostLayerSize(u32 level) const; + static u32 ConvertWidth(u32 width, VideoCore::Surface::PixelFormat pixel_format_from, + VideoCore::Surface::PixelFormat pixel_format_to) { + const u32 bw1 = VideoCore::Surface::GetDefaultBlockWidth(pixel_format_from); + const u32 bw2 = VideoCore::Surface::GetDefaultBlockWidth(pixel_format_to); + return (width * bw2 + bw1 - 1) / bw1; + } + + static u32 ConvertHeight(u32 height, VideoCore::Surface::PixelFormat pixel_format_from, + VideoCore::Surface::PixelFormat pixel_format_to) { + const u32 bh1 = VideoCore::Surface::GetDefaultBlockHeight(pixel_format_from); + const u32 bh2 = VideoCore::Surface::GetDefaultBlockHeight(pixel_format_to); + return (height * bh2 + bh1 - 1) / bh1; + } + /// Returns the default block width. u32 GetDefaultBlockWidth() const { return VideoCore::Surface::GetDefaultBlockWidth(pixel_format); diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 85c9160e0..593ceeaf6 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -141,11 +141,6 @@ public: return {}; } - if (regs.color_mask[index].raw == 0) { - SetEmptyColorBuffer(index); - return {}; - } - const auto& config{regs.rt[index]}; const auto gpu_addr{config.Address()}; if (!gpu_addr) { @@ -192,11 +187,11 @@ public: void DoFermiCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src_config, const Tegra::Engines::Fermi2D::Regs::Surface& dst_config, - const Common::Rectangle& src_rect, - const Common::Rectangle& dst_rect) { - TSurface dst_surface = GetFermiSurface(dst_config); - ImageBlit(GetFermiSurface(src_config), dst_surface, src_rect, dst_rect); - dst_surface->MarkAsModified(true, Tick()); + const Tegra::Engines::Fermi2D::Config& copy_config) { + std::pair dst_surface = GetFermiSurface(dst_config); + std::pair src_surface = GetFermiSurface(src_config); + ImageBlit(src_surface.second, dst_surface.second, copy_config); + dst_surface.first->MarkAsModified(true, Tick()); } TSurface TryFindFramebufferSurface(const u8* host_ptr) { @@ -234,8 +229,8 @@ protected: virtual void ImageCopy(TSurface src_surface, TSurface dst_surface, const CopyParams& copy_params) = 0; - virtual void ImageBlit(TSurface src, TSurface dst, const Common::Rectangle& src_rect, - const Common::Rectangle& dst_rect) = 0; + virtual void ImageBlit(TView src_view, TView dst_view, + const Tegra::Engines::Fermi2D::Config& copy_config) = 0; void Register(TSurface surface) { std::lock_guard lock{mutex}; @@ -282,10 +277,11 @@ protected: return new_surface; } - TSurface GetFermiSurface(const Tegra::Engines::Fermi2D::Regs::Surface& config) { + std::pair GetFermiSurface( + const Tegra::Engines::Fermi2D::Regs::Surface& config) { SurfaceParams params = SurfaceParams::CreateForFermiCopySurface(config); const GPUVAddr gpu_addr = config.Address(); - return GetSurface(gpu_addr, params, true).first; + return GetSurface(gpu_addr, params, true); } Core::System& system; @@ -551,7 +547,21 @@ private: if (view.has_value()) { const bool is_mirage = !current_surface->MatchFormat(params.pixel_format); if (is_mirage) { - LOG_CRITICAL(HW_GPU, "Mirage View Unsupported"); + // On a mirage view, we need to recreate the surface under this new view + // and then obtain a view again. + SurfaceParams new_params = current_surface->GetSurfaceParams(); + const u32 wh = SurfaceParams::ConvertWidth( + new_params.width, new_params.pixel_format, params.pixel_format); + const u32 hh = SurfaceParams::ConvertHeight( + new_params.height, new_params.pixel_format, params.pixel_format); + new_params.width = wh; + new_params.height = hh; + new_params.pixel_format = params.pixel_format; + std::pair pair = RebuildSurface(current_surface, new_params); + std::optional mirage_view = + pair.first->EmplaceView(params, gpu_addr, candidate_size); + if (mirage_view) + return {pair.first, *mirage_view}; return RecycleSurface(overlaps, params, gpu_addr, preserve_contents, false); } return {current_surface, *view}; -- cgit v1.2.3 From cb728797b0cec79d756ba9395d24924882222103 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Wed, 12 Jun 2019 16:20:20 -0400 Subject: fermi2d: Correct Origin Mode --- src/video_core/engines/fermi_2d.cpp | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp index d63b82838..0ee228e28 100644 --- a/src/video_core/engines/fermi_2d.cpp +++ b/src/video_core/engines/fermi_2d.cpp @@ -38,11 +38,16 @@ void Fermi2D::HandleSurfaceCopy() { const u32 src_blit_x1{static_cast(regs.blit_src_x >> 32)}; const u32 src_blit_y1{static_cast(regs.blit_src_y >> 32)}; - const u32 src_blit_x2{ - static_cast((regs.blit_src_x + (regs.blit_dst_width * regs.blit_du_dx)) >> 32)}; - const u32 src_blit_y2{ - static_cast((regs.blit_src_y + (regs.blit_dst_height * regs.blit_dv_dy)) >> 32)}; - + u32 src_blit_x2, src_blit_y2; + if (regs.blit_control.origin == Origin::Corner) { + src_blit_x2 = + static_cast((regs.blit_src_x + (regs.blit_du_dx * regs.blit_dst_width)) >> 32); + src_blit_y2 = + static_cast((regs.blit_src_y + (regs.blit_dv_dy * regs.blit_dst_height)) >> 32); + } else { + src_blit_x2 = static_cast((regs.blit_src_x >> 32) + regs.blit_dst_width); + src_blit_y2 = static_cast((regs.blit_src_y >> 32) + regs.blit_dst_height); + } const Common::Rectangle src_rect{src_blit_x1, src_blit_y1, src_blit_x2, src_blit_y2}; const Common::Rectangle dst_rect{regs.blit_dst_x, regs.blit_dst_y, regs.blit_dst_x + regs.blit_dst_width, -- cgit v1.2.3 From 7232a1ed16e46715c29d781fb143bdf799090bec Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Thu, 13 Jun 2019 16:41:16 -0400 Subject: decoders: correct block calculation --- src/video_core/engines/engine_upload.cpp | 2 +- src/video_core/engines/engine_upload.h | 6 ++--- src/video_core/engines/fermi_2d.h | 6 ++--- src/video_core/engines/maxwell_dma.cpp | 2 +- src/video_core/engines/maxwell_dma.h | 4 ++-- src/video_core/texture_cache/texture_cache.h | 16 +++++++++++++ src/video_core/textures/decoders.cpp | 34 ++++++++++++---------------- 7 files changed, 41 insertions(+), 29 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/engine_upload.cpp b/src/video_core/engines/engine_upload.cpp index c776b9a56..d44ad0cd8 100644 --- a/src/video_core/engines/engine_upload.cpp +++ b/src/video_core/engines/engine_upload.cpp @@ -39,7 +39,7 @@ void State::ProcessData(const u32 data, const bool is_last_call) { UNIMPLEMENTED_IF(regs.dest.BlockWidth() != 0); UNIMPLEMENTED_IF(regs.dest.BlockDepth() != 0); const std::size_t dst_size = Tegra::Texture::CalculateSize( - true, 1, regs.dest.width, regs.dest.height, 1, regs.dest.BlockHeight(), 1); + true, 1, regs.dest.width, regs.dest.height, 1, regs.dest.BlockHeight(), 0); tmp_buffer.resize(dst_size); memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size); Tegra::Texture::SwizzleKepler(regs.dest.width, regs.dest.height, regs.dest.x, regs.dest.y, diff --git a/src/video_core/engines/engine_upload.h b/src/video_core/engines/engine_upload.h index cb294aec3..462da419e 100644 --- a/src/video_core/engines/engine_upload.h +++ b/src/video_core/engines/engine_upload.h @@ -39,15 +39,15 @@ struct Registers { } u32 BlockWidth() const { - return block_width; + return block_width.Value(); } u32 BlockHeight() const { - return block_height; + return block_height.Value(); } u32 BlockDepth() const { - return block_depth; + return block_depth.Value(); } } dest; }; diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h index 0a4c7c5ad..05421d185 100644 --- a/src/video_core/engines/fermi_2d.h +++ b/src/video_core/engines/fermi_2d.h @@ -84,15 +84,15 @@ public: } u32 BlockWidth() const { - return block_width; + return block_width.Value(); } u32 BlockHeight() const { - return block_height; + return block_height.Value(); } u32 BlockDepth() const { - return block_depth; + return block_depth.Value(); } }; static_assert(sizeof(Surface) == 0x28, "Surface has incorrect size"); diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp index afb9578d0..3a5dfef0c 100644 --- a/src/video_core/engines/maxwell_dma.cpp +++ b/src/video_core/engines/maxwell_dma.cpp @@ -111,7 +111,7 @@ void MaxwellDMA::HandleCopy() { memory_manager.WriteBlock(dest, write_buffer.data(), dst_size); } else { - ASSERT(regs.dst_params.BlockDepth() == 0); + ASSERT(regs.dst_params.BlockDepth() == 1); const u32 src_bytes_per_pixel = regs.src_pitch / regs.x_count; diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h index 522fa97dc..17b015ca7 100644 --- a/src/video_core/engines/maxwell_dma.h +++ b/src/video_core/engines/maxwell_dma.h @@ -59,11 +59,11 @@ public: }; u32 BlockHeight() const { - return block_height; + return block_height.Value(); } u32 BlockDepth() const { - return block_depth; + return block_depth.Value(); } }; diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 201c4d42e..7a9b4c27d 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -335,6 +335,9 @@ private: if (untopological == MatchTopologyResult::CompressUnmatch) { return RecycleStrategy::Flush; } + if (untopological == MatchTopologyResult::FullMatch && !params.is_tiled) { + return RecycleStrategy::Flush; + } return RecycleStrategy::Ignore; } @@ -372,6 +375,11 @@ private: } return InitializeSurface(gpu_addr, params, preserve_contents); } + case RecycleStrategy::BufferCopy: { + auto new_surface = GetUncachedSurface(gpu_addr, params); + BufferCopy(overlaps[0], new_surface); + return {new_surface, new_surface->GetMainView()}; + } default: { UNIMPLEMENTED_MSG("Unimplemented Texture Cache Recycling Strategy!"); return InitializeSurface(gpu_addr, params, do_load); @@ -520,6 +528,10 @@ private: const auto host_ptr{memory_manager->GetPointer(gpu_addr)}; const auto cache_addr{ToCacheAddr(host_ptr)}; + if (gpu_addr == 0x00000001682F0000ULL) { + LOG_CRITICAL(HW_GPU, "Here's the texture!"); + } + // Step 0: guarantee a valid surface if (!cache_addr) { // Return a null surface if it's invalid @@ -566,6 +578,10 @@ private: return InitializeSurface(gpu_addr, params, preserve_contents); } + if (!params.is_tiled) { + return RecycleSurface(overlaps, params, gpu_addr, preserve_contents, + MatchTopologyResult::FullMatch); + } // Step 3 // Now we need to figure the relationship between the texture and its overlaps // we do a topological test to ensure we can find some relationship. If it fails diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index f45fd175a..9a2f4198a 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -256,19 +256,18 @@ std::vector UnswizzleTexture(u8* address, u32 tile_size_x, u32 tile_size_y, } void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width, - u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height) { - const u32 block_height_size{1U << block_height}; + u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height_bit) { + const u32 block_height = 1U << block_height_bit; const u32 image_width_in_gobs{(swizzled_width * bytes_per_pixel + (gob_size_x - 1)) / gob_size_x}; for (u32 line = 0; line < subrect_height; ++line) { const u32 gob_address_y = - (line / (gob_size_y * block_height_size)) * gob_size * block_height_size * - image_width_in_gobs + - ((line % (gob_size_y * block_height_size)) / gob_size_y) * gob_size; + (line / (gob_size_y * block_height)) * gob_size * block_height * image_width_in_gobs + + ((line % (gob_size_y * block_height)) / gob_size_y) * gob_size; const auto& table = legacy_swizzle_table[line % gob_size_y]; for (u32 x = 0; x < subrect_width; ++x) { const u32 gob_address = - gob_address_y + (x * bytes_per_pixel / gob_size_x) * gob_size * block_height_size; + gob_address_y + (x * bytes_per_pixel / gob_size_x) * gob_size * block_height; const u32 swizzled_offset = gob_address + table[(x * bytes_per_pixel) % gob_size_x]; u8* source_line = unswizzled_data + line * source_pitch + x * bytes_per_pixel; u8* dest_addr = swizzled_data + swizzled_offset; @@ -279,19 +278,17 @@ void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 } void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 swizzled_width, - u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height, + u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height_bit, u32 offset_x, u32 offset_y) { - const u32 block_height_size{1U << block_height}; + const u32 block_height = 1U << block_height_bit; for (u32 line = 0; line < subrect_height; ++line) { const u32 y2 = line + offset_y; - const u32 gob_address_y = - (y2 / (gob_size_y * block_height_size)) * gob_size * block_height_size + - ((y2 % (gob_size_y * block_height_size)) / gob_size_y) * gob_size; + const u32 gob_address_y = (y2 / (gob_size_y * block_height)) * gob_size * block_height + + ((y2 % (gob_size_y * block_height)) / gob_size_y) * gob_size; const auto& table = legacy_swizzle_table[y2 % gob_size_y]; for (u32 x = 0; x < subrect_width; ++x) { const u32 x2 = (x + offset_x) * bytes_per_pixel; - const u32 gob_address = - gob_address_y + (x2 / gob_size_x) * gob_size * block_height_size; + const u32 gob_address = gob_address_y + (x2 / gob_size_x) * gob_size * block_height; const u32 swizzled_offset = gob_address + table[x2 % gob_size_x]; u8* dest_line = unswizzled_data + line * dest_pitch + x * bytes_per_pixel; u8* source_addr = swizzled_data + swizzled_offset; @@ -302,20 +299,19 @@ void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 } void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32 dst_y, - const u32 block_height, const std::size_t copy_size, const u8* source_data, + const u32 block_height_bit, const std::size_t copy_size, const u8* source_data, u8* swizzle_data) { - const u32 block_height_size{1U << block_height}; + const u32 block_height = 1U << block_height_bit; const u32 image_width_in_gobs{(width + gob_size_x - 1) / gob_size_x}; std::size_t count = 0; for (std::size_t y = dst_y; y < height && count < copy_size; ++y) { const std::size_t gob_address_y = - (y / (gob_size_y * block_height_size)) * gob_size * block_height_size * - image_width_in_gobs + - ((y % (gob_size_y * block_height_size)) / gob_size_y) * gob_size; + (y / (gob_size_y * block_height)) * gob_size * block_height * image_width_in_gobs + + ((y % (gob_size_y * block_height)) / gob_size_y) * gob_size; const auto& table = legacy_swizzle_table[y % gob_size_y]; for (std::size_t x = dst_x; x < width && count < copy_size; ++x) { const std::size_t gob_address = - gob_address_y + (x / gob_size_x) * gob_size * block_height_size; + gob_address_y + (x / gob_size_x) * gob_size * block_height; const std::size_t swizzled_offset = gob_address + table[x % gob_size_x]; const u8* source_line = source_data + count; u8* dest_addr = swizzle_data + swizzled_offset; -- cgit v1.2.3 From 082740d34db0996a0af73d7680c57e1abb31c712 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Fri, 14 Jun 2019 16:40:04 -0400 Subject: surface: Correct format S8Z24 --- src/video_core/engines/maxwell_dma.cpp | 2 +- src/video_core/surface.cpp | 4 ++-- src/video_core/texture_cache/texture_cache.h | 4 ---- src/video_core/textures/texture.h | 4 ++-- 4 files changed, 5 insertions(+), 9 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp index 3a5dfef0c..afb9578d0 100644 --- a/src/video_core/engines/maxwell_dma.cpp +++ b/src/video_core/engines/maxwell_dma.cpp @@ -111,7 +111,7 @@ void MaxwellDMA::HandleCopy() { memory_manager.WriteBlock(dest, write_buffer.data(), dst_size); } else { - ASSERT(regs.dst_params.BlockDepth() == 1); + ASSERT(regs.dst_params.BlockDepth() == 0); const u32 src_bytes_per_pixel = regs.src_pitch / regs.x_count; diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp index 52a79e4a7..c50f6354d 100644 --- a/src/video_core/surface.cpp +++ b/src/video_core/surface.cpp @@ -308,8 +308,8 @@ PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format, return PixelFormat::Z32F; case Tegra::Texture::TextureFormat::Z16: return PixelFormat::Z16; - case Tegra::Texture::TextureFormat::Z24S8: - return PixelFormat::Z24S8; + case Tegra::Texture::TextureFormat::S8Z24: + return PixelFormat::S8Z24; case Tegra::Texture::TextureFormat::ZF32_X24S8: return PixelFormat::Z32FS8; case Tegra::Texture::TextureFormat::DXT1: diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 8213f434d..a9e61cba1 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -534,10 +534,6 @@ private: const auto host_ptr{memory_manager->GetPointer(gpu_addr)}; const auto cache_addr{ToCacheAddr(host_ptr)}; - if (gpu_addr == 0x00000001682F0000ULL) { - LOG_CRITICAL(HW_GPU, "Here's the texture!"); - } - // Step 0: guarantee a valid surface if (!cache_addr) { // Return a null surface if it's invalid diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h index ddeed73d0..e3be018b9 100644 --- a/src/video_core/textures/texture.h +++ b/src/video_core/textures/texture.h @@ -52,9 +52,9 @@ enum class TextureFormat : u32 { DXT45 = 0x26, DXN1 = 0x27, DXN2 = 0x28, - Z24S8 = 0x29, + S8Z24 = 0x29, X8Z24 = 0x2a, - S8Z24 = 0x2b, + Z24S8 = 0x2b, X4V4Z24__COV4R4V = 0x2c, X4V4Z24__COV8R8V = 0x2d, V8Z24__COV4R12V = 0x2e, -- cgit v1.2.3 From 4d63f9794575f6180fbecb1d7dce7fcd3875439f Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Mon, 24 Jun 2019 01:51:02 -0300 Subject: shader_bytecode: Include missing --- src/video_core/engines/shader_bytecode.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h index 54a1a04f9..404d4f5aa 100644 --- a/src/video_core/engines/shader_bytecode.h +++ b/src/video_core/engines/shader_bytecode.h @@ -4,6 +4,7 @@ #pragma once +#include #include #include #include -- cgit v1.2.3