diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/video_core/CMakeLists.txt | 2 | ||||
| -rw-r--r-- | src/video_core/compatible_formats.cpp | 162 | ||||
| -rw-r--r-- | src/video_core/compatible_formats.h | 32 | ||||
| -rw-r--r-- | src/video_core/texture_cache/texture_cache.h | 25 |
4 files changed, 216 insertions, 5 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 2dc752aa9..21c46a567 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt | |||
| @@ -3,6 +3,8 @@ add_library(video_core STATIC | |||
| 3 | buffer_cache/buffer_cache.h | 3 | buffer_cache/buffer_cache.h |
| 4 | buffer_cache/map_interval.cpp | 4 | buffer_cache/map_interval.cpp |
| 5 | buffer_cache/map_interval.h | 5 | buffer_cache/map_interval.h |
| 6 | compatible_formats.cpp | ||
| 7 | compatible_formats.h | ||
| 6 | dirty_flags.cpp | 8 | dirty_flags.cpp |
| 7 | dirty_flags.h | 9 | dirty_flags.h |
| 8 | dma_pusher.cpp | 10 | dma_pusher.cpp |
diff --git a/src/video_core/compatible_formats.cpp b/src/video_core/compatible_formats.cpp new file mode 100644 index 000000000..6c426b035 --- /dev/null +++ b/src/video_core/compatible_formats.cpp | |||
| @@ -0,0 +1,162 @@ | |||
| 1 | // Copyright 2020 yuzu Emulator Project | ||
| 2 | // Licensed under GPLv2 or any later version | ||
| 3 | // Refer to the license.txt file included. | ||
| 4 | |||
| 5 | #include <array> | ||
| 6 | #include <bitset> | ||
| 7 | #include <cstddef> | ||
| 8 | |||
| 9 | #include "video_core/compatible_formats.h" | ||
| 10 | #include "video_core/surface.h" | ||
| 11 | |||
| 12 | namespace VideoCore::Surface { | ||
| 13 | |||
| 14 | namespace { | ||
| 15 | |||
| 16 | // Compatibility table taken from Table 3.X.2 in: | ||
| 17 | // https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_texture_view.txt | ||
| 18 | |||
| 19 | constexpr std::array VIEW_CLASS_128_BITS = { | ||
| 20 | PixelFormat::RGBA32F, | ||
| 21 | PixelFormat::RGBA32UI, | ||
| 22 | }; | ||
| 23 | // Missing formats: | ||
| 24 | // PixelFormat::RGBA32I | ||
| 25 | |||
| 26 | constexpr std::array VIEW_CLASS_96_BITS = { | ||
| 27 | PixelFormat::RGB32F, | ||
| 28 | }; | ||
| 29 | // Missing formats: | ||
| 30 | // PixelFormat::RGB32UI, | ||
| 31 | // PixelFormat::RGB32I, | ||
| 32 | |||
| 33 | constexpr std::array VIEW_CLASS_64_BITS = { | ||
| 34 | PixelFormat::RGBA16F, PixelFormat::RG32F, PixelFormat::RGBA16UI, PixelFormat::RG32UI, | ||
| 35 | PixelFormat::RGBA16U, PixelFormat::RGBA16F, PixelFormat::RGBA16S, | ||
| 36 | }; | ||
| 37 | // Missing formats: | ||
| 38 | // PixelFormat::RGBA16I | ||
| 39 | // PixelFormat::RG32I | ||
| 40 | |||
| 41 | // TODO: How should we handle 48 bits? | ||
| 42 | |||
| 43 | constexpr std::array VIEW_CLASS_32_BITS = { | ||
| 44 | PixelFormat::RG16F, PixelFormat::R11FG11FB10F, PixelFormat::R32F, | ||
| 45 | PixelFormat::A2B10G10R10U, PixelFormat::RG16UI, PixelFormat::R32UI, | ||
| 46 | PixelFormat::RG16I, PixelFormat::R32I, PixelFormat::ABGR8U, | ||
| 47 | PixelFormat::RG16, PixelFormat::ABGR8S, PixelFormat::RG16S, | ||
| 48 | PixelFormat::RGBA8_SRGB, PixelFormat::E5B9G9R9F, PixelFormat::BGRA8, | ||
| 49 | PixelFormat::BGRA8_SRGB, | ||
| 50 | }; | ||
| 51 | // Missing formats: | ||
| 52 | // PixelFormat::RGBA8UI | ||
| 53 | // PixelFormat::RGBA8I | ||
| 54 | // PixelFormat::RGB10_A2_UI | ||
| 55 | |||
| 56 | // TODO: How should we handle 24 bits? | ||
| 57 | |||
| 58 | constexpr std::array VIEW_CLASS_16_BITS = { | ||
| 59 | PixelFormat::R16F, PixelFormat::RG8UI, PixelFormat::R16UI, PixelFormat::R16I, | ||
| 60 | PixelFormat::RG8U, PixelFormat::R16U, PixelFormat::RG8S, PixelFormat::R16S, | ||
| 61 | }; | ||
| 62 | // Missing formats: | ||
| 63 | // PixelFormat::RG8I | ||
| 64 | |||
| 65 | constexpr std::array VIEW_CLASS_8_BITS = { | ||
| 66 | PixelFormat::R8UI, | ||
| 67 | PixelFormat::R8U, | ||
| 68 | }; | ||
| 69 | // Missing formats: | ||
| 70 | // PixelFormat::R8I | ||
| 71 | // PixelFormat::R8S | ||
| 72 | |||
| 73 | constexpr std::array VIEW_CLASS_RGTC1_RED = { | ||
| 74 | PixelFormat::DXN1, | ||
| 75 | }; | ||
| 76 | // Missing formats: | ||
| 77 | // COMPRESSED_SIGNED_RED_RGTC1 | ||
| 78 | |||
| 79 | constexpr std::array VIEW_CLASS_RGTC2_RG = { | ||
| 80 | PixelFormat::DXN2UNORM, | ||
| 81 | PixelFormat::DXN2SNORM, | ||
| 82 | }; | ||
| 83 | |||
| 84 | constexpr std::array VIEW_CLASS_BPTC_UNORM = { | ||
| 85 | PixelFormat::BC7U, | ||
| 86 | PixelFormat::BC7U_SRGB, | ||
| 87 | }; | ||
| 88 | |||
| 89 | constexpr std::array VIEW_CLASS_BPTC_FLOAT = { | ||
| 90 | PixelFormat::BC6H_SF16, | ||
| 91 | PixelFormat::BC6H_UF16, | ||
| 92 | }; | ||
| 93 | |||
| 94 | // Compatibility table taken from Table 4.X.1 in: | ||
| 95 | // https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_copy_image.txt | ||
| 96 | |||
| 97 | constexpr std::array COPY_CLASS_128_BITS = { | ||
| 98 | PixelFormat::RGBA32UI, PixelFormat::RGBA32F, PixelFormat::DXT23, | ||
| 99 | PixelFormat::DXT23_SRGB, PixelFormat::DXT45, PixelFormat::DXT45_SRGB, | ||
| 100 | PixelFormat::DXN2SNORM, PixelFormat::BC7U, PixelFormat::BC7U_SRGB, | ||
| 101 | PixelFormat::BC6H_SF16, PixelFormat::BC6H_UF16, | ||
| 102 | }; | ||
| 103 | // Missing formats: | ||
| 104 | // PixelFormat::RGBA32I | ||
| 105 | // COMPRESSED_RG_RGTC2 | ||
| 106 | |||
| 107 | constexpr std::array COPY_CLASS_64_BITS = { | ||
| 108 | PixelFormat::RGBA16F, PixelFormat::RG32F, PixelFormat::RGBA16UI, PixelFormat::RG32UI, | ||
| 109 | PixelFormat::RGBA16U, PixelFormat::RGBA16S, PixelFormat::DXT1_SRGB, PixelFormat::DXT1, | ||
| 110 | |||
| 111 | }; | ||
| 112 | // Missing formats: | ||
| 113 | // PixelFormat::RGBA16I | ||
| 114 | // PixelFormat::RG32I, | ||
| 115 | // COMPRESSED_RGB_S3TC_DXT1_EXT | ||
| 116 | // COMPRESSED_SRGB_S3TC_DXT1_EXT | ||
| 117 | // COMPRESSED_RGBA_S3TC_DXT1_EXT | ||
| 118 | // COMPRESSED_SIGNED_RED_RGTC1 | ||
| 119 | |||
| 120 | void Enable(FormatCompatibility::Table& compatiblity, size_t format_a, size_t format_b) { | ||
| 121 | compatiblity[format_a][format_b] = true; | ||
| 122 | compatiblity[format_b][format_a] = true; | ||
| 123 | } | ||
| 124 | |||
| 125 | void Enable(FormatCompatibility::Table& compatibility, PixelFormat format_a, PixelFormat format_b) { | ||
| 126 | Enable(compatibility, static_cast<size_t>(format_a), static_cast<size_t>(format_b)); | ||
| 127 | } | ||
| 128 | |||
| 129 | template <typename Range> | ||
| 130 | void EnableRange(FormatCompatibility::Table& compatibility, const Range& range) { | ||
| 131 | for (auto it_a = range.begin(); it_a != range.end(); ++it_a) { | ||
| 132 | for (auto it_b = it_a; it_b != range.end(); ++it_b) { | ||
| 133 | Enable(compatibility, *it_a, *it_b); | ||
| 134 | } | ||
| 135 | } | ||
| 136 | } | ||
| 137 | |||
| 138 | } // Anonymous namespace | ||
| 139 | |||
| 140 | FormatCompatibility::FormatCompatibility() { | ||
| 141 | for (size_t i = 0; i < MaxPixelFormat; ++i) { | ||
| 142 | // Identity is allowed | ||
| 143 | Enable(view, i, i); | ||
| 144 | } | ||
| 145 | |||
| 146 | EnableRange(view, VIEW_CLASS_128_BITS); | ||
| 147 | EnableRange(view, VIEW_CLASS_96_BITS); | ||
| 148 | EnableRange(view, VIEW_CLASS_64_BITS); | ||
| 149 | EnableRange(view, VIEW_CLASS_32_BITS); | ||
| 150 | EnableRange(view, VIEW_CLASS_16_BITS); | ||
| 151 | EnableRange(view, VIEW_CLASS_8_BITS); | ||
| 152 | EnableRange(view, VIEW_CLASS_RGTC1_RED); | ||
| 153 | EnableRange(view, VIEW_CLASS_RGTC2_RG); | ||
| 154 | EnableRange(view, VIEW_CLASS_BPTC_UNORM); | ||
| 155 | EnableRange(view, VIEW_CLASS_BPTC_FLOAT); | ||
| 156 | |||
| 157 | copy = view; | ||
| 158 | EnableRange(copy, COPY_CLASS_128_BITS); | ||
| 159 | EnableRange(copy, COPY_CLASS_64_BITS); | ||
| 160 | } | ||
| 161 | |||
| 162 | } // namespace VideoCore::Surface | ||
diff --git a/src/video_core/compatible_formats.h b/src/video_core/compatible_formats.h new file mode 100644 index 000000000..d1082566d --- /dev/null +++ b/src/video_core/compatible_formats.h | |||
| @@ -0,0 +1,32 @@ | |||
| 1 | // Copyright 2020 yuzu Emulator Project | ||
| 2 | // Licensed under GPLv2 or any later version | ||
| 3 | // Refer to the license.txt file included. | ||
| 4 | |||
| 5 | #include <array> | ||
| 6 | #include <bitset> | ||
| 7 | #include <cstddef> | ||
| 8 | |||
| 9 | #include "video_core/surface.h" | ||
| 10 | |||
| 11 | namespace VideoCore::Surface { | ||
| 12 | |||
| 13 | class FormatCompatibility { | ||
| 14 | public: | ||
| 15 | using Table = std::array<std::bitset<MaxPixelFormat>, MaxPixelFormat>; | ||
| 16 | |||
| 17 | explicit FormatCompatibility(); | ||
| 18 | |||
| 19 | bool TestView(PixelFormat format_a, PixelFormat format_b) const noexcept { | ||
| 20 | return view[static_cast<size_t>(format_a)][static_cast<size_t>(format_b)]; | ||
| 21 | } | ||
| 22 | |||
| 23 | bool TestCopy(PixelFormat format_a, PixelFormat format_b) const noexcept { | ||
| 24 | return copy[static_cast<size_t>(format_a)][static_cast<size_t>(format_b)]; | ||
| 25 | } | ||
| 26 | |||
| 27 | private: | ||
| 28 | Table view; | ||
| 29 | Table copy; | ||
| 30 | }; | ||
| 31 | |||
| 32 | } // namespace VideoCore::Surface | ||
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 85075e868..6207d8dfe 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h | |||
| @@ -24,6 +24,7 @@ | |||
| 24 | #include "core/core.h" | 24 | #include "core/core.h" |
| 25 | #include "core/memory.h" | 25 | #include "core/memory.h" |
| 26 | #include "core/settings.h" | 26 | #include "core/settings.h" |
| 27 | #include "video_core/compatible_formats.h" | ||
| 27 | #include "video_core/dirty_flags.h" | 28 | #include "video_core/dirty_flags.h" |
| 28 | #include "video_core/engines/fermi_2d.h" | 29 | #include "video_core/engines/fermi_2d.h" |
| 29 | #include "video_core/engines/maxwell_3d.h" | 30 | #include "video_core/engines/maxwell_3d.h" |
| @@ -47,8 +48,8 @@ class RasterizerInterface; | |||
| 47 | 48 | ||
| 48 | namespace VideoCommon { | 49 | namespace VideoCommon { |
| 49 | 50 | ||
| 51 | using VideoCore::Surface::FormatCompatibility; | ||
| 50 | using VideoCore::Surface::PixelFormat; | 52 | using VideoCore::Surface::PixelFormat; |
| 51 | |||
| 52 | using VideoCore::Surface::SurfaceTarget; | 53 | using VideoCore::Surface::SurfaceTarget; |
| 53 | using RenderTargetConfig = Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig; | 54 | using RenderTargetConfig = Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig; |
| 54 | 55 | ||
| @@ -595,7 +596,7 @@ private: | |||
| 595 | } else { | 596 | } else { |
| 596 | new_surface = GetUncachedSurface(gpu_addr, params); | 597 | new_surface = GetUncachedSurface(gpu_addr, params); |
| 597 | } | 598 | } |
| 598 | const auto& final_params = new_surface->GetSurfaceParams(); | 599 | const SurfaceParams& final_params = new_surface->GetSurfaceParams(); |
| 599 | if (cr_params.type != final_params.type) { | 600 | if (cr_params.type != final_params.type) { |
| 600 | if (Settings::IsGPULevelExtreme()) { | 601 | if (Settings::IsGPULevelExtreme()) { |
| 601 | BufferCopy(current_surface, new_surface); | 602 | BufferCopy(current_surface, new_surface); |
| @@ -603,7 +604,7 @@ private: | |||
| 603 | } else { | 604 | } else { |
| 604 | std::vector<CopyParams> bricks = current_surface->BreakDown(final_params); | 605 | std::vector<CopyParams> bricks = current_surface->BreakDown(final_params); |
| 605 | for (auto& brick : bricks) { | 606 | for (auto& brick : bricks) { |
| 606 | ImageCopy(current_surface, new_surface, brick); | 607 | TryCopyImage(current_surface, new_surface, brick); |
| 607 | } | 608 | } |
| 608 | } | 609 | } |
| 609 | Unregister(current_surface); | 610 | Unregister(current_surface); |
| @@ -694,7 +695,7 @@ private: | |||
| 694 | } | 695 | } |
| 695 | const CopyParams copy_params(0, 0, 0, 0, 0, base_layer, 0, mipmap, width, height, | 696 | const CopyParams copy_params(0, 0, 0, 0, 0, base_layer, 0, mipmap, width, height, |
| 696 | src_params.depth); | 697 | src_params.depth); |
| 697 | ImageCopy(surface, new_surface, copy_params); | 698 | TryCopyImage(surface, new_surface, copy_params); |
| 698 | } | 699 | } |
| 699 | } | 700 | } |
| 700 | if (passed_tests == 0) { | 701 | if (passed_tests == 0) { |
| @@ -791,7 +792,7 @@ private: | |||
| 791 | const u32 width = params.width; | 792 | const u32 width = params.width; |
| 792 | const u32 height = params.height; | 793 | const u32 height = params.height; |
| 793 | const CopyParams copy_params(0, 0, 0, 0, 0, slice, 0, 0, width, height, 1); | 794 | const CopyParams copy_params(0, 0, 0, 0, 0, slice, 0, 0, width, height, 1); |
| 794 | ImageCopy(surface, new_surface, copy_params); | 795 | TryCopyImage(surface, new_surface, copy_params); |
| 795 | } | 796 | } |
| 796 | for (const auto& surface : overlaps) { | 797 | for (const auto& surface : overlaps) { |
| 797 | Unregister(surface); | 798 | Unregister(surface); |
| @@ -1192,6 +1193,19 @@ private: | |||
| 1192 | return {}; | 1193 | return {}; |
| 1193 | } | 1194 | } |
| 1194 | 1195 | ||
| 1196 | /// Try to do an image copy logging when formats are incompatible. | ||
| 1197 | void TryCopyImage(TSurface& src, TSurface& dst, const CopyParams& copy) { | ||
| 1198 | const SurfaceParams& src_params = src->GetSurfaceParams(); | ||
| 1199 | const SurfaceParams& dst_params = dst->GetSurfaceParams(); | ||
| 1200 | if (!format_compatibility.TestCopy(src_params.pixel_format, dst_params.pixel_format)) { | ||
| 1201 | LOG_ERROR(HW_GPU, "Illegal copy between formats={{{}, {}}}", | ||
| 1202 | static_cast<int>(dst_params.pixel_format), | ||
| 1203 | static_cast<int>(src_params.pixel_format)); | ||
| 1204 | return; | ||
| 1205 | } | ||
| 1206 | ImageCopy(src, dst, copy); | ||
| 1207 | } | ||
| 1208 | |||
| 1195 | constexpr PixelFormat GetSiblingFormat(PixelFormat format) const { | 1209 | constexpr PixelFormat GetSiblingFormat(PixelFormat format) const { |
| 1196 | return siblings_table[static_cast<std::size_t>(format)]; | 1210 | return siblings_table[static_cast<std::size_t>(format)]; |
| 1197 | } | 1211 | } |
| @@ -1241,6 +1255,7 @@ private: | |||
| 1241 | VideoCore::RasterizerInterface& rasterizer; | 1255 | VideoCore::RasterizerInterface& rasterizer; |
| 1242 | 1256 | ||
| 1243 | FormatLookupTable format_lookup_table; | 1257 | FormatLookupTable format_lookup_table; |
| 1258 | FormatCompatibility format_compatibility; | ||
| 1244 | 1259 | ||
| 1245 | u64 ticks{}; | 1260 | u64 ticks{}; |
| 1246 | 1261 | ||