diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/common/alignment.h | 29 | ||||
| -rw-r--r-- | src/video_core/texture_cache/accelerated_swizzle.cpp | 4 | ||||
| -rw-r--r-- | src/video_core/texture_cache/util.cpp | 10 | ||||
| -rw-r--r-- | src/video_core/textures/decoders.cpp | 8 |
4 files changed, 23 insertions, 28 deletions
diff --git a/src/common/alignment.h b/src/common/alignment.h index 5040043de..fb81f10d8 100644 --- a/src/common/alignment.h +++ b/src/common/alignment.h | |||
| @@ -9,50 +9,45 @@ | |||
| 9 | namespace Common { | 9 | namespace Common { |
| 10 | 10 | ||
| 11 | template <typename T> | 11 | template <typename T> |
| 12 | [[nodiscard]] constexpr T AlignUp(T value, std::size_t size) { | 12 | requires std::is_unsigned_v<T>[[nodiscard]] constexpr T AlignUp(T value, size_t size) { |
| 13 | static_assert(std::is_unsigned_v<T>, "T must be an unsigned value."); | ||
| 14 | auto mod{static_cast<T>(value % size)}; | 13 | auto mod{static_cast<T>(value % size)}; |
| 15 | value -= mod; | 14 | value -= mod; |
| 16 | return static_cast<T>(mod == T{0} ? value : value + size); | 15 | return static_cast<T>(mod == T{0} ? value : value + size); |
| 17 | } | 16 | } |
| 18 | 17 | ||
| 19 | template <typename T> | 18 | template <typename T> |
| 20 | [[nodiscard]] constexpr T AlignDown(T value, std::size_t size) { | 19 | requires std::is_unsigned_v<T>[[nodiscard]] constexpr T AlignUpLog2(T value, size_t align_log2) { |
| 21 | static_assert(std::is_unsigned_v<T>, "T must be an unsigned value."); | 20 | return static_cast<T>((value + ((1ULL << align_log2) - 1)) >> align_log2 << align_log2); |
| 22 | return static_cast<T>(value - value % size); | ||
| 23 | } | 21 | } |
| 24 | 22 | ||
| 25 | template <typename T> | 23 | template <typename T> |
| 26 | [[nodiscard]] constexpr T AlignBits(T value, std::size_t align) { | 24 | requires std::is_unsigned_v<T>[[nodiscard]] constexpr T AlignDown(T value, size_t size) { |
| 27 | static_assert(std::is_unsigned_v<T>, "T must be an unsigned value."); | 25 | return static_cast<T>(value - value % size); |
| 28 | return static_cast<T>((value + ((1ULL << align) - 1)) >> align << align); | ||
| 29 | } | 26 | } |
| 30 | 27 | ||
| 31 | template <typename T> | 28 | template <typename T> |
| 32 | [[nodiscard]] constexpr bool Is4KBAligned(T value) { | 29 | requires std::is_unsigned_v<T>[[nodiscard]] constexpr bool Is4KBAligned(T value) { |
| 33 | static_assert(std::is_unsigned_v<T>, "T must be an unsigned value."); | ||
| 34 | return (value & 0xFFF) == 0; | 30 | return (value & 0xFFF) == 0; |
| 35 | } | 31 | } |
| 36 | 32 | ||
| 37 | template <typename T> | 33 | template <typename T> |
| 38 | [[nodiscard]] constexpr bool IsWordAligned(T value) { | 34 | requires std::is_unsigned_v<T>[[nodiscard]] constexpr bool IsWordAligned(T value) { |
| 39 | static_assert(std::is_unsigned_v<T>, "T must be an unsigned value."); | ||
| 40 | return (value & 0b11) == 0; | 35 | return (value & 0b11) == 0; |
| 41 | } | 36 | } |
| 42 | 37 | ||
| 43 | template <typename T> | 38 | template <typename T> |
| 44 | [[nodiscard]] constexpr bool IsAligned(T value, std::size_t alignment) { | 39 | requires std::is_integral_v<T>[[nodiscard]] constexpr bool IsAligned(T value, size_t alignment) { |
| 45 | using U = typename std::make_unsigned<T>::type; | 40 | using U = typename std::make_unsigned_t<T>; |
| 46 | const U mask = static_cast<U>(alignment - 1); | 41 | const U mask = static_cast<U>(alignment - 1); |
| 47 | return (value & mask) == 0; | 42 | return (value & mask) == 0; |
| 48 | } | 43 | } |
| 49 | 44 | ||
| 50 | template <typename T, std::size_t Align = 16> | 45 | template <typename T, size_t Align = 16> |
| 51 | class AlignmentAllocator { | 46 | class AlignmentAllocator { |
| 52 | public: | 47 | public: |
| 53 | using value_type = T; | 48 | using value_type = T; |
| 54 | using size_type = std::size_t; | 49 | using size_type = size_t; |
| 55 | using difference_type = std::ptrdiff_t; | 50 | using difference_type = ptrdiff_t; |
| 56 | 51 | ||
| 57 | using propagate_on_container_copy_assignment = std::true_type; | 52 | using propagate_on_container_copy_assignment = std::true_type; |
| 58 | using propagate_on_container_move_assignment = std::true_type; | 53 | using propagate_on_container_move_assignment = std::true_type; |
diff --git a/src/video_core/texture_cache/accelerated_swizzle.cpp b/src/video_core/texture_cache/accelerated_swizzle.cpp index a4fc1184b..15585caeb 100644 --- a/src/video_core/texture_cache/accelerated_swizzle.cpp +++ b/src/video_core/texture_cache/accelerated_swizzle.cpp | |||
| @@ -27,7 +27,7 @@ BlockLinearSwizzle2DParams MakeBlockLinearSwizzle2DParams(const SwizzleParameter | |||
| 27 | const Extent3D num_tiles = swizzle.num_tiles; | 27 | const Extent3D num_tiles = swizzle.num_tiles; |
| 28 | const u32 bytes_per_block = BytesPerBlock(info.format); | 28 | const u32 bytes_per_block = BytesPerBlock(info.format); |
| 29 | const u32 stride_alignment = CalculateLevelStrideAlignment(info, swizzle.level); | 29 | const u32 stride_alignment = CalculateLevelStrideAlignment(info, swizzle.level); |
| 30 | const u32 stride = Common::AlignBits(num_tiles.width, stride_alignment) * bytes_per_block; | 30 | const u32 stride = Common::AlignUpLog2(num_tiles.width, stride_alignment) * bytes_per_block; |
| 31 | const u32 gobs_in_x = Common::DivCeilLog2(stride, GOB_SIZE_X_SHIFT); | 31 | const u32 gobs_in_x = Common::DivCeilLog2(stride, GOB_SIZE_X_SHIFT); |
| 32 | return BlockLinearSwizzle2DParams{ | 32 | return BlockLinearSwizzle2DParams{ |
| 33 | .origin{0, 0, 0}, | 33 | .origin{0, 0, 0}, |
| @@ -47,7 +47,7 @@ BlockLinearSwizzle3DParams MakeBlockLinearSwizzle3DParams(const SwizzleParameter | |||
| 47 | const Extent3D num_tiles = swizzle.num_tiles; | 47 | const Extent3D num_tiles = swizzle.num_tiles; |
| 48 | const u32 bytes_per_block = BytesPerBlock(info.format); | 48 | const u32 bytes_per_block = BytesPerBlock(info.format); |
| 49 | const u32 stride_alignment = CalculateLevelStrideAlignment(info, swizzle.level); | 49 | const u32 stride_alignment = CalculateLevelStrideAlignment(info, swizzle.level); |
| 50 | const u32 stride = Common::AlignBits(num_tiles.width, stride_alignment) * bytes_per_block; | 50 | const u32 stride = Common::AlignUpLog2(num_tiles.width, stride_alignment) * bytes_per_block; |
| 51 | 51 | ||
| 52 | const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) >> GOB_SIZE_X_SHIFT; | 52 | const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) >> GOB_SIZE_X_SHIFT; |
| 53 | const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block.height + block.depth); | 53 | const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block.height + block.depth); |
diff --git a/src/video_core/texture_cache/util.cpp b/src/video_core/texture_cache/util.cpp index 279932778..ce8fcfe0a 100644 --- a/src/video_core/texture_cache/util.cpp +++ b/src/video_core/texture_cache/util.cpp | |||
| @@ -279,7 +279,7 @@ template <u32 GOB_EXTENT> | |||
| 279 | const bool is_small = IsSmallerThanGobSize(blocks, gob, info.block.depth); | 279 | const bool is_small = IsSmallerThanGobSize(blocks, gob, info.block.depth); |
| 280 | const u32 alignment = is_small ? 0 : info.tile_width_spacing; | 280 | const u32 alignment = is_small ? 0 : info.tile_width_spacing; |
| 281 | return Extent2D{ | 281 | return Extent2D{ |
| 282 | .width = Common::AlignBits(gobs.width, alignment), | 282 | .width = Common::AlignUpLog2(gobs.width, alignment), |
| 283 | .height = gobs.height, | 283 | .height = gobs.height, |
| 284 | }; | 284 | }; |
| 285 | } | 285 | } |
| @@ -352,7 +352,7 @@ template <u32 GOB_EXTENT> | |||
| 352 | // https://github.com/Ryujinx/Ryujinx/blob/1c9aba6de1520aea5480c032e0ff5664ac1bb36f/Ryujinx.Graphics.Texture/SizeCalculator.cs#L134 | 352 | // https://github.com/Ryujinx/Ryujinx/blob/1c9aba6de1520aea5480c032e0ff5664ac1bb36f/Ryujinx.Graphics.Texture/SizeCalculator.cs#L134 |
| 353 | if (tile_width_spacing > 0) { | 353 | if (tile_width_spacing > 0) { |
| 354 | const u32 alignment_log2 = GOB_SIZE_SHIFT + tile_width_spacing + block.height + block.depth; | 354 | const u32 alignment_log2 = GOB_SIZE_SHIFT + tile_width_spacing + block.height + block.depth; |
| 355 | return Common::AlignBits(size_bytes, alignment_log2); | 355 | return Common::AlignUpLog2(size_bytes, alignment_log2); |
| 356 | } | 356 | } |
| 357 | const u32 aligned_height = Common::AlignUp(size.height, tile_size_y); | 357 | const u32 aligned_height = Common::AlignUp(size.height, tile_size_y); |
| 358 | while (block.height != 0 && aligned_height <= (1U << (block.height - 1)) * GOB_SIZE_Y) { | 358 | while (block.height != 0 && aligned_height <= (1U << (block.height - 1)) * GOB_SIZE_Y) { |
| @@ -528,9 +528,9 @@ template <u32 GOB_EXTENT> | |||
| 528 | const u32 alignment = StrideAlignment(num_tiles, info.block, bpp_log2, info.tile_width_spacing); | 528 | const u32 alignment = StrideAlignment(num_tiles, info.block, bpp_log2, info.tile_width_spacing); |
| 529 | const Extent3D mip_block = AdjustMipBlockSize(num_tiles, info.block, 0); | 529 | const Extent3D mip_block = AdjustMipBlockSize(num_tiles, info.block, 0); |
| 530 | return Extent3D{ | 530 | return Extent3D{ |
| 531 | .width = Common::AlignBits(num_tiles.width, alignment), | 531 | .width = Common::AlignUpLog2(num_tiles.width, alignment), |
| 532 | .height = Common::AlignBits(num_tiles.height, GOB_SIZE_Y_SHIFT + mip_block.height), | 532 | .height = Common::AlignUpLog2(num_tiles.height, GOB_SIZE_Y_SHIFT + mip_block.height), |
| 533 | .depth = Common::AlignBits(num_tiles.depth, GOB_SIZE_Z_SHIFT + mip_block.depth), | 533 | .depth = Common::AlignUpLog2(num_tiles.depth, GOB_SIZE_Z_SHIFT + mip_block.depth), |
| 534 | }; | 534 | }; |
| 535 | } | 535 | } |
| 536 | 536 | ||
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 9f5181318..62685a183 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp | |||
| @@ -49,7 +49,7 @@ void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixe | |||
| 49 | // We can configure here a custom pitch | 49 | // We can configure here a custom pitch |
| 50 | // As it's not exposed 'width * bpp' will be the expected pitch. | 50 | // As it's not exposed 'width * bpp' will be the expected pitch. |
| 51 | const u32 pitch = width * bytes_per_pixel; | 51 | const u32 pitch = width * bytes_per_pixel; |
| 52 | const u32 stride = Common::AlignBits(width, stride_alignment) * bytes_per_pixel; | 52 | const u32 stride = Common::AlignUpLog2(width, stride_alignment) * bytes_per_pixel; |
| 53 | 53 | ||
| 54 | const u32 gobs_in_x = Common::DivCeilLog2(stride, GOB_SIZE_X_SHIFT); | 54 | const u32 gobs_in_x = Common::DivCeilLog2(stride, GOB_SIZE_X_SHIFT); |
| 55 | const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth); | 55 | const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth); |
| @@ -217,9 +217,9 @@ void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32 | |||
| 217 | std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, | 217 | std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, |
| 218 | u32 block_height, u32 block_depth) { | 218 | u32 block_height, u32 block_depth) { |
| 219 | if (tiled) { | 219 | if (tiled) { |
| 220 | const u32 aligned_width = Common::AlignBits(width * bytes_per_pixel, GOB_SIZE_X_SHIFT); | 220 | const u32 aligned_width = Common::AlignUpLog2(width * bytes_per_pixel, GOB_SIZE_X_SHIFT); |
| 221 | const u32 aligned_height = Common::AlignBits(height, GOB_SIZE_Y_SHIFT + block_height); | 221 | const u32 aligned_height = Common::AlignUpLog2(height, GOB_SIZE_Y_SHIFT + block_height); |
| 222 | const u32 aligned_depth = Common::AlignBits(depth, GOB_SIZE_Z_SHIFT + block_depth); | 222 | const u32 aligned_depth = Common::AlignUpLog2(depth, GOB_SIZE_Z_SHIFT + block_depth); |
| 223 | return aligned_width * aligned_height * aligned_depth; | 223 | return aligned_width * aligned_height * aligned_depth; |
| 224 | } else { | 224 | } else { |
| 225 | return width * height * depth * bytes_per_pixel; | 225 | return width * height * depth * bytes_per_pixel; |