diff options
Diffstat (limited to 'src/video_core/textures/decoders.cpp')
| -rw-r--r-- | src/video_core/textures/decoders.cpp | 33 |
1 files changed, 19 insertions, 14 deletions
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 25b0860b0f..ef50be156b 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -18,6 +18,11 @@ namespace Tegra::Texture { namespace { +constexpr u32 NativeGobOffset(u32 offset, bool native_tiling) { + if (!native_tiling) return offset; + const u32 difference = ((offset >> 5) ^ (offset >> 6)) & 1; + return offset ^ (difference << 5) ^ (difference << 6); +} template <u32 mask> constexpr u32 pdep(u32 value) { u32 result = 0; @@ -38,7 +43,7 @@ void incrpdep(u32& value) { template <bool TO_LINEAR, u32 BYTES_PER_PIXEL> void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 height, u32 depth, - u32 block_height, u32 block_depth, u32 stride) { + u32 block_height, u32 block_depth, u32 stride, bool native_tiling) { // The origin of the transformation can be configured here, leave it as zero as the current API // doesn't expose it. static constexpr u32 origin_x = 0; @@ -77,7 +82,7 @@ void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift; const u32 base_swizzled_offset = offset_z + offset_y + offset_x; - const u32 swizzled_offset = base_swizzled_offset + (swizzled_x | swizzled_y); + const u32 swizzled_offset = NativeGobOffset(base_swizzled_offset + (swizzled_x | swizzled_y), native_tiling); const u32 unswizzled_offset = slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL; @@ -94,7 +99,7 @@ void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 template <bool TO_LINEAR, u32 BYTES_PER_PIXEL> void SwizzleSubrectImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, u32 num_lines, - u32 block_height, u32 block_depth, u32 pitch_linear) { + u32 block_height, u32 block_depth, u32 pitch_linear, bool native_tiling) { // The origin of the transformation can be configured here, leave it as zero as the current API // doesn't expose it. static constexpr u32 origin_z = 0; @@ -136,7 +141,7 @@ void SwizzleSubrectImpl(std::span<u8> output, std::span<const u8> input, u32 wid const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift; const u32 base_swizzled_offset = offset_z + offset_y + offset_x; - const u32 swizzled_offset = base_swizzled_offset + (swizzled_x | swizzled_y); + const u32 swizzled_offset = NativeGobOffset(base_swizzled_offset + (swizzled_x | swizzled_y), native_tiling); const u32 unswizzled_offset = slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL; @@ -156,12 +161,12 @@ void SwizzleSubrectImpl(std::span<u8> output, std::span<const u8> input, u32 wid template <bool TO_LINEAR> void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width, - u32 height, u32 depth, u32 block_height, u32 block_depth, u32 stride_alignment) { + u32 height, u32 depth, u32 block_height, u32 block_depth, u32 stride_alignment, bool native_tiling) { switch (bytes_per_pixel) { #define BPP_CASE(x) \ case x: \ return SwizzleImpl<TO_LINEAR, x>(output, input, width, height, depth, block_height, \ - block_depth, stride_alignment); + block_depth, stride_alignment, native_tiling); BPP_CASE(1) BPP_CASE(2) BPP_CASE(3) @@ -181,35 +186,35 @@ void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixe void UnswizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth, - u32 stride_alignment) { + u32 stride_alignment, bool native_tiling) { const u32 stride = Common::AlignUpLog2(width, stride_alignment) * bytes_per_pixel; const u32 new_bpp = (std::min)(4U, static_cast<u32>(std::countr_zero(width * bytes_per_pixel))); width = (width * bytes_per_pixel) >> new_bpp; bytes_per_pixel = 1U << new_bpp; Swizzle<false>(output, input, bytes_per_pixel, width, height, depth, block_height, block_depth, - stride); + stride, native_tiling); } void SwizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth, - u32 stride_alignment) { + u32 stride_alignment, bool native_tiling) { const u32 stride = Common::AlignUpLog2(width, stride_alignment) * bytes_per_pixel; const u32 new_bpp = (std::min)(4U, static_cast<u32>(std::countr_zero(width * bytes_per_pixel))); width = (width * bytes_per_pixel) >> new_bpp; bytes_per_pixel = 1U << new_bpp; Swizzle<true>(output, input, bytes_per_pixel, width, height, depth, block_height, block_depth, - stride); + stride, native_tiling); } void SwizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, u32 extent_y, - u32 block_height, u32 block_depth, u32 pitch_linear) { + u32 block_height, u32 block_depth, u32 pitch_linear, bool native_tiling) { switch (bytes_per_pixel) { #define BPP_CASE(x) \ case x: \ return SwizzleSubrectImpl<true, x>(output, input, width, height, depth, origin_x, \ origin_y, extent_x, extent_y, block_height, \ - block_depth, pitch_linear); + block_depth, pitch_linear, native_tiling); BPP_CASE(1) BPP_CASE(2) BPP_CASE(3) @@ -227,13 +232,13 @@ void SwizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes_p void UnswizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, - u32 extent_y, u32 block_height, u32 block_depth, u32 pitch_linear) { + u32 extent_y, u32 block_height, u32 block_depth, u32 pitch_linear, bool native_tiling) { switch (bytes_per_pixel) { #define BPP_CASE(x) \ case x: \ return SwizzleSubrectImpl<false, x>(output, input, width, height, depth, origin_x, \ origin_y, extent_x, extent_y, block_height, \ - block_depth, pitch_linear); + block_depth, pitch_linear, native_tiling); BPP_CASE(1) BPP_CASE(2) BPP_CASE(3) |