diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp index 912e785b9..597b279b9 100644 --- a/src/video_core/engines/fermi_2d.cpp +++ b/src/video_core/engines/fermi_2d.cpp @@ -62,14 +62,16 @@ void Fermi2D::HandleSurfaceCopy() { u8* dst_buffer = Memory::GetPointer(dest_cpu); if (!regs.src.linear && regs.dst.linear) { // If the input is tiled and the output is linear, deswizzle the input and copy it over. - Texture::CopySwizzledData(regs.src.width, regs.src.height, src_bytes_per_pixel, - dst_bytes_per_pixel, src_buffer, dst_buffer, true, - regs.src.BlockHeight()); + Texture::CopySwizzledData(regs.src.width, regs.src.height, regs.src.depth, + src_bytes_per_pixel, dst_bytes_per_pixel, src_buffer, + dst_buffer, true, regs.src.BlockHeight(), + regs.src.BlockDepth()); } else { // If the input is linear and the output is tiled, swizzle the input and copy it over. - Texture::CopySwizzledData(regs.src.width, regs.src.height, src_bytes_per_pixel, - dst_bytes_per_pixel, dst_buffer, src_buffer, false, - regs.dst.BlockHeight()); + Texture::CopySwizzledData(regs.src.width, regs.src.height, regs.src.depth, + src_bytes_per_pixel, dst_bytes_per_pixel, dst_buffer, + src_buffer, false, regs.dst.BlockHeight(), + regs.dst.BlockDepth()); } } } diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp index aa7481b8c..bf2a21bb6 100644 --- a/src/video_core/engines/maxwell_dma.cpp +++ b/src/video_core/engines/maxwell_dma.cpp @@ -68,12 +68,14 @@ void MaxwellDMA::HandleCopy() { if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) { // If the input is tiled and the output is linear, deswizzle the input and copy it over. - Texture::CopySwizzledData(regs.src_params.size_x, regs.src_params.size_y, 1, 1, src_buffer, - dst_buffer, true, regs.src_params.BlockHeight()); + Texture::CopySwizzledData(regs.src_params.size_x, regs.src_params.size_y, + regs.src_params.size_z, 1, 1, src_buffer, dst_buffer, true, + regs.src_params.BlockHeight(), regs.src_params.BlockDepth()); } else { // If the input is linear and the output is tiled, swizzle the input and copy it over. - Texture::CopySwizzledData(regs.dst_params.size_x, regs.dst_params.size_y, 1, 1, dst_buffer, - src_buffer, false, regs.dst_params.BlockHeight()); + Texture::CopySwizzledData(regs.dst_params.size_x, regs.dst_params.size_y, + regs.dst_params.size_z, 1, 1, dst_buffer, src_buffer, false, + regs.dst_params.BlockHeight(), regs.dst_params.BlockDepth()); } } diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h index 311ccb616..df19e02e2 100644 --- a/src/video_core/engines/maxwell_dma.h +++ b/src/video_core/engines/maxwell_dma.h @@ -43,6 +43,10 @@ public: u32 BlockHeight() const { return 1 << block_height; } + + u32 BlockDepth() const { + return 1 << block_depth; + } }; static_assert(sizeof(Parameters) == 24, "Parameters has wrong size"); diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp index 65a220c41..86cb62dae 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp @@ -323,8 +323,8 @@ static bool IsFormatBCn(PixelFormat format) { } template -void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, std::size_t gl_buffer_size, - VAddr addr) { +void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth, u32 depth, u8* gl_buffer, + std::size_t gl_buffer_size, VAddr addr) { constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT; constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format); @@ -333,7 +333,7 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, std::si // pixel values. const u32 tile_size{IsFormatBCn(format) ? 4U : 1U}; const std::vector data = Tegra::Texture::UnswizzleTexture( - addr, tile_size, bytes_per_pixel, stride, height, block_height); + addr, tile_size, bytes_per_pixel, stride, height, depth, block_height, block_depth); const std::size_t size_to_copy{std::min(gl_buffer_size, data.size())}; memcpy(gl_buffer, data.data(), size_to_copy); } else { @@ -345,7 +345,7 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, std::si } } -static constexpr std::array morton_to_gl_fns = { // clang-format off @@ -403,7 +403,7 @@ static constexpr std::array gl_to_morton_fns = { // clang-format off @@ -827,25 +827,27 @@ void CachedSurface::LoadGLBuffer() { if (params.is_tiled) { gl_buffer.resize(total_size); + u32 depth = params.depth; + u32 block_depth = params.block_depth; ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}", params.block_width, static_cast(params.target)); - ASSERT_MSG(params.block_depth == 1, "Block depth is defined as {} on texture type {}", - params.block_depth, static_cast(params.target)); - // TODO(bunnei): This only unswizzles and copies a 2D texture - we do not yet know how to do - // this for 3D textures, etc. switch (params.target) { case SurfaceParams::SurfaceTarget::Texture2D: - // Pass impl. to the fallback code below + // TODO(Blinkhawk): Eliminate this condition once all texture types are implemented. + depth = 1U; + block_depth = 1U; break; case SurfaceParams::SurfaceTarget::Texture2DArray: case SurfaceParams::SurfaceTarget::TextureCubemap: + depth = 1U; + block_depth = 1U; for (std::size_t index = 0; index < params.depth; ++index) { const std::size_t offset{index * copy_size}; morton_to_gl_fns[static_cast(params.pixel_format)]( - params.width, params.block_height, params.height, gl_buffer.data() + offset, - copy_size, params.addr + offset); + params.width, params.block_height, params.height, 1U, 1U, + gl_buffer.data() + offset, copy_size, params.addr + offset); } break; default: @@ -854,9 +856,11 @@ void CachedSurface::LoadGLBuffer() { UNREACHABLE(); } + const std::size_t size = copy_size * depth; + morton_to_gl_fns[static_cast(params.pixel_format)]( - params.width, params.block_height, params.height, gl_buffer.data(), copy_size, - params.addr); + params.width, params.block_height, params.height, block_depth, depth, gl_buffer.data(), + size, params.addr); } else { const u8* const texture_src_data_end{texture_src_data + total_size}; gl_buffer.assign(texture_src_data, texture_src_data_end); diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 5e2d3ac32..6693067f8 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -51,18 +51,18 @@ void Precise3DProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool un const u32 xy_block_size, const u32 layer_z, const u32 stride_x, const u32 bytes_per_pixel, const u32 out_bytes_per_pixel) { std::array data_ptrs; - u32 z_adress = tile_offset; + u32 z_address = tile_offset; const u32 gob_size_x = 64; const u32 gob_size_y = 8; const u32 gob_size_z = 1; const u32 gob_size = gob_size_x * gob_size_y * gob_size_z; for (u32 z = z_start; z < z_end; z++) { - u32 y_adress = z_adress; + u32 y_address = z_address; u32 pixel_base = layer_z * z + y_start * stride_x; for (u32 y = y_start; y < y_end; y++) { const auto& table = legacy_swizzle_table[y % gob_size_y]; for (u32 x = x_start; x < x_end; x++) { - const u32 swizzle_offset{y_adress + table[x * bytes_per_pixel % gob_size_x]}; + const u32 swizzle_offset{y_address + table[x * bytes_per_pixel % gob_size_x]}; const u32 pixel_index{x * out_bytes_per_pixel + pixel_base}; data_ptrs[unswizzle] = swizzled_data + swizzle_offset; data_ptrs[!unswizzle] = unswizzled_data + pixel_index; @@ -70,9 +70,9 @@ void Precise3DProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool un } pixel_base += stride_x; if ((y + 1) % gob_size_y == 0) - y_adress += gob_size; + y_address += gob_size; } - z_adress += xy_block_size; + z_address += xy_block_size; } } @@ -136,7 +136,7 @@ void Fast3DProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool unswi const u32 xy_block_size, const u32 layer_z, const u32 stride_x, const u32 bytes_per_pixel, const u32 out_bytes_per_pixel) { std::array data_ptrs; - u32 z_adress = tile_offset; + u32 z_address = tile_offset; const u32 x_startb = x_start * bytes_per_pixel; const u32 x_endb = x_end * bytes_per_pixel; const u32 copy_size = 16; @@ -145,12 +145,12 @@ void Fast3DProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool unswi const u32 gob_size_z = 1; const u32 gob_size = gob_size_x * gob_size_y * gob_size_z; for (u32 z = z_start; z < z_end; z++) { - u32 y_adress = z_adress; + u32 y_address = z_address; u32 pixel_base = layer_z * z + y_start * stride_x; for (u32 y = y_start; y < y_end; y++) { const auto& table = fast_swizzle_table[y % gob_size_y]; for (u32 xb = x_startb; xb < x_endb; xb += copy_size) { - const u32 swizzle_offset{y_adress + table[(xb / copy_size) % 4]}; + const u32 swizzle_offset{y_address + table[(xb / copy_size) % 4]}; const u32 out_x = xb * out_bytes_per_pixel / bytes_per_pixel; const u32 pixel_index{out_x + pixel_base}; data_ptrs[unswizzle] = swizzled_data + swizzle_offset; @@ -159,9 +159,9 @@ void Fast3DProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool unswi } pixel_base += stride_x; if ((y + 1) % gob_size_y == 0) - y_adress += gob_size; + y_address += gob_size; } - z_adress += xy_block_size; + z_address += xy_block_size; } } @@ -214,14 +214,15 @@ void Fast3DSwizzledData(u8* swizzled_data, u8* unswizzled_data, const bool unswi } } -void CopySwizzledData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_bytes_per_pixel, - u8* swizzled_data, u8* unswizzled_data, bool unswizzle, u32 block_height) { +void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel, + u32 out_bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, + bool unswizzle, u32 block_height, u32 block_depth) { if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % 16 == 0) { - Fast3DSwizzledData(swizzled_data, unswizzled_data, unswizzle, width, height, 1U, - bytes_per_pixel, out_bytes_per_pixel, block_height, 1U); + Fast3DSwizzledData(swizzled_data, unswizzled_data, unswizzle, width, height, depth, + bytes_per_pixel, out_bytes_per_pixel, block_height, block_depth); } else { - Precise3DSwizzledData(swizzled_data, unswizzled_data, unswizzle, width, height, 1U, - bytes_per_pixel, out_bytes_per_pixel, block_height, 1U); + Precise3DSwizzledData(swizzled_data, unswizzled_data, unswizzle, width, height, depth, + bytes_per_pixel, out_bytes_per_pixel, block_height, block_depth); } } @@ -269,10 +270,11 @@ u32 BytesPerPixel(TextureFormat format) { } std::vector UnswizzleTexture(VAddr address, u32 tile_size, u32 bytes_per_pixel, u32 width, - u32 height, u32 block_height) { + u32 height, u32 depth, u32 block_height, u32 block_depth) { std::vector unswizzled_data(width * height * bytes_per_pixel); - CopySwizzledData(width / tile_size, height / tile_size, bytes_per_pixel, bytes_per_pixel, - Memory::GetPointer(address), unswizzled_data.data(), true, block_height); + CopySwizzledData(width / tile_size, height / tile_size, depth, bytes_per_pixel, bytes_per_pixel, + Memory::GetPointer(address), unswizzled_data.data(), true, block_height, + block_depth); return unswizzled_data; } diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h index 234d250af..aaf316947 100644 --- a/src/video_core/textures/decoders.h +++ b/src/video_core/textures/decoders.h @@ -14,17 +14,14 @@ namespace Tegra::Texture { * Unswizzles a swizzled texture without changing its format. */ std::vector UnswizzleTexture(VAddr address, u32 tile_size, u32 bytes_per_pixel, u32 width, - u32 height, u32 block_height = TICEntry::DefaultBlockHeight); - -/** - * Unswizzles a swizzled depth texture without changing its format. - */ -std::vector UnswizzleDepthTexture(VAddr address, DepthFormat format, u32 width, u32 height, - u32 block_height = TICEntry::DefaultBlockHeight); + u32 height, u32 depth, + u32 block_height = TICEntry::DefaultBlockHeight, + u32 block_depth = TICEntry::DefaultBlockHeight); /// Copies texture data from a buffer and performs swizzling/unswizzling as necessary. -void CopySwizzledData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_bytes_per_pixel, - u8* swizzled_data, u8* unswizzled_data, bool unswizzle, u32 block_height); +void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel, + u32 out_bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, + bool unswizzle, u32 block_height, u32 block_depth); /** * Decodes an unswizzled texture into a A8R8G8B8 texture. diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h index 58d17abcb..5947bd2b9 100644 --- a/src/video_core/textures/texture.h +++ b/src/video_core/textures/texture.h @@ -141,6 +141,7 @@ static_assert(sizeof(TextureHandle) == 4, "TextureHandle has wrong size"); struct TICEntry { static constexpr u32 DefaultBlockHeight = 16; + static constexpr u32 DefaultBlockDepth = 1; union { u32 raw; diff --git a/src/yuzu/debugger/graphics/graphics_surface.cpp b/src/yuzu/debugger/graphics/graphics_surface.cpp index cbcd5dd5f..44d423da2 100644 --- a/src/yuzu/debugger/graphics/graphics_surface.cpp +++ b/src/yuzu/debugger/graphics/graphics_surface.cpp @@ -386,8 +386,9 @@ void GraphicsSurfaceWidget::OnUpdate() { // TODO(bunnei): Will not work with BCn formats that swizzle 4x4 tiles. // Needs to be fixed if we plan to use this feature more, otherwise we may remove it. - auto unswizzled_data = Tegra::Texture::UnswizzleTexture( - *address, 1, Tegra::Texture::BytesPerPixel(surface_format), surface_width, surface_height); + auto unswizzled_data = + Tegra::Texture::UnswizzleTexture(*address, 1, Tegra::Texture::BytesPerPixel(surface_format), + surface_width, surface_height, 1U); auto texture_data = Tegra::Texture::DecodeTexture(unswizzled_data, surface_format, surface_width, surface_height);