diff options
author | Fernando Sahmkow <fsahmkow27@gmail.com> | 2019-06-02 05:03:22 +0200 |
---|---|---|
committer | ReinUsesLisp <reinuseslisp@airmail.cc> | 2019-06-21 02:38:34 +0200 |
commit | 9f755218a1359cbd004e6c287f5fead0897c1d11 (patch) | |
tree | 9a0e4187602f3c650b7367d268b46e22892b5a5c /src/video_core/texture_cache | |
parent | texture_cache: Optimize GetSurface and use references on functions that don't change a surface. (diff) | |
download | yuzu-9f755218a1359cbd004e6c287f5fead0897c1d11.tar yuzu-9f755218a1359cbd004e6c287f5fead0897c1d11.tar.gz yuzu-9f755218a1359cbd004e6c287f5fead0897c1d11.tar.bz2 yuzu-9f755218a1359cbd004e6c287f5fead0897c1d11.tar.lz yuzu-9f755218a1359cbd004e6c287f5fead0897c1d11.tar.xz yuzu-9f755218a1359cbd004e6c287f5fead0897c1d11.tar.zst yuzu-9f755218a1359cbd004e6c287f5fead0897c1d11.zip |
Diffstat (limited to '')
-rw-r--r-- | src/video_core/texture_cache/surface_base.cpp | 103 | ||||
-rw-r--r-- | src/video_core/texture_cache/surface_base.h | 106 | ||||
-rw-r--r-- | src/video_core/texture_cache/surface_params.cpp | 33 | ||||
-rw-r--r-- | src/video_core/texture_cache/surface_params.h | 32 |
4 files changed, 135 insertions, 139 deletions
diff --git a/src/video_core/texture_cache/surface_base.cpp b/src/video_core/texture_cache/surface_base.cpp index 7e90960f7..8c6edb04f 100644 --- a/src/video_core/texture_cache/surface_base.cpp +++ b/src/video_core/texture_cache/surface_base.cpp @@ -42,6 +42,109 @@ SurfaceBaseImpl::SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params) } } +MatchTopologyResult SurfaceBaseImpl::MatchesTopology(const SurfaceParams& rhs) const { + const u32 src_bpp{params.GetBytesPerPixel()}; + const u32 dst_bpp{rhs.GetBytesPerPixel()}; + const bool ib1 = params.IsBuffer(); + const bool ib2 = rhs.IsBuffer(); + if (std::tie(src_bpp, params.is_tiled, ib1) == std::tie(dst_bpp, rhs.is_tiled, ib2)) { + const bool cb1 = params.IsCompressed(); + const bool cb2 = rhs.IsCompressed(); + if (cb1 == cb2) { + return MatchTopologyResult::FullMatch; + } + return MatchTopologyResult::CompressUnmatch; + } + return MatchTopologyResult::None; +} + +MatchStructureResult SurfaceBaseImpl::MatchesStructure(const SurfaceParams& rhs) const { + // Buffer surface Check + if (params.IsBuffer()) { + const std::size_t wd1 = params.width * params.GetBytesPerPixel(); + const std::size_t wd2 = rhs.width * rhs.GetBytesPerPixel(); + if (wd1 == wd2) { + return MatchStructureResult::FullMatch; + } + return MatchStructureResult::None; + } + + // Linear Surface check + if (!params.is_tiled) { + if (std::tie(params.width, params.height, params.pitch) == + std::tie(rhs.width, rhs.height, rhs.pitch)) { + return MatchStructureResult::FullMatch; + } + return MatchStructureResult::None; + } + + // Tiled Surface check + 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 (std::tie(params.width, params.height) == std::tie(rhs.width, rhs.height)) { + return MatchStructureResult::FullMatch; + } + 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; + } + } + return MatchStructureResult::None; +} + +std::optional<std::pair<u32, u32>> SurfaceBaseImpl::GetLayerMipmap( + const GPUVAddr candidate_gpu_addr) const { + if (candidate_gpu_addr < gpu_addr) { + return {}; + } + const auto relative_address{static_cast<GPUVAddr>(candidate_gpu_addr - gpu_addr)}; + const auto layer{static_cast<u32>(relative_address / layer_size)}; + const GPUVAddr mipmap_address = relative_address - layer_size * layer; + const auto mipmap_it = + Common::BinaryFind(mipmap_offsets.begin(), mipmap_offsets.end(), mipmap_address); + if (mipmap_it == mipmap_offsets.end()) { + return {}; + } + const auto level{static_cast<u32>(std::distance(mipmap_offsets.begin(), mipmap_it))}; + return std::make_pair(layer, level); +} + +std::vector<CopyParams> SurfaceBaseImpl::BreakDownLayered(const SurfaceParams& in_params) const { + const u32 layers{params.depth}; + const u32 mipmaps{params.num_levels}; + std::vector<CopyParams> result; + result.reserve(static_cast<std::size_t>(layers) * static_cast<std::size_t>(mipmaps)); + + for (u32 layer = 0; layer < layers; layer++) { + for (u32 level = 0; level < mipmaps; level++) { + const u32 width = SurfaceParams::IntersectWidth(params, in_params, level, level); + const u32 height = SurfaceParams::IntersectHeight(params, in_params, level, level); + result.emplace_back(width, height, layer, level); + } + } + return result; +} + +std::vector<CopyParams> SurfaceBaseImpl::BreakDownNonLayered(const SurfaceParams& in_params) const { + const u32 mipmaps{params.num_levels}; + std::vector<CopyParams> result; + result.reserve(mipmaps); + + for (u32 level = 0; level < mipmaps; level++) { + const u32 width = SurfaceParams::IntersectWidth(params, in_params, level, level); + const u32 height = SurfaceParams::IntersectHeight(params, in_params, level, level); + const u32 depth{std::min(params.GetMipDepth(level), in_params.GetMipDepth(level))}; + result.emplace_back(width, height, depth, level); + } + return result; +} + void SurfaceBaseImpl::SwizzleFunc(MortonSwizzleMode mode, u8* memory, const SurfaceParams& params, u8* buffer, u32 level) { const u32 width{params.GetMipWidth(level)}; diff --git a/src/video_core/texture_cache/surface_base.h b/src/video_core/texture_cache/surface_base.h index 70b5258c9..9d19ecd5f 100644 --- a/src/video_core/texture_cache/surface_base.h +++ b/src/video_core/texture_cache/surface_base.h @@ -136,83 +136,17 @@ public: return params.target == target; } + MatchTopologyResult MatchesTopology(const SurfaceParams& rhs) const; + + MatchStructureResult MatchesStructure(const SurfaceParams& rhs) const; + bool MatchesSubTexture(const SurfaceParams& rhs, const GPUVAddr other_gpu_addr) const { return std::tie(gpu_addr, params.target, params.num_levels) == std::tie(other_gpu_addr, rhs.target, rhs.num_levels) && params.target == SurfaceTarget::Texture2D && params.num_levels == 1; } - MatchTopologyResult MatchesTopology(const SurfaceParams& rhs) const { - const u32 src_bpp{params.GetBytesPerPixel()}; - const u32 dst_bpp{rhs.GetBytesPerPixel()}; - const bool ib1 = params.IsBuffer(); - const bool ib2 = rhs.IsBuffer(); - if (std::tie(src_bpp, params.is_tiled, ib1) == std::tie(dst_bpp, rhs.is_tiled, ib2)) { - const bool cb1 = params.IsCompressed(); - const bool cb2 = rhs.IsCompressed(); - if (cb1 == cb2) { - return MatchTopologyResult::FullMatch; - } - return MatchTopologyResult::CompressUnmatch; - } - return MatchTopologyResult::None; - } - - MatchStructureResult MatchesStructure(const SurfaceParams& rhs) const { - // Buffer surface Check - if (params.IsBuffer()) { - const std::size_t wd1 = params.width * params.GetBytesPerPixel(); - const std::size_t wd2 = rhs.width * rhs.GetBytesPerPixel(); - if (wd1 == wd2) { - return MatchStructureResult::FullMatch; - } - return MatchStructureResult::None; - } - - // Linear Surface check - if (!params.is_tiled) { - if (std::tie(params.width, params.height, params.pitch) == - std::tie(rhs.width, rhs.height, rhs.pitch)) { - return MatchStructureResult::FullMatch; - } - return MatchStructureResult::None; - } - - // Tiled Surface check - 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 (std::tie(params.width, params.height) == std::tie(rhs.width, rhs.height)) { - return MatchStructureResult::FullMatch; - } - 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; - } - } - return MatchStructureResult::None; - } - - std::optional<std::pair<u32, u32>> GetLayerMipmap(const GPUVAddr candidate_gpu_addr) const { - if (candidate_gpu_addr < gpu_addr) { - return {}; - } - const auto relative_address{static_cast<GPUVAddr>(candidate_gpu_addr - gpu_addr)}; - const auto layer{static_cast<u32>(relative_address / layer_size)}; - const GPUVAddr mipmap_address = relative_address - layer_size * layer; - const auto mipmap_it = - Common::BinaryFind(mipmap_offsets.begin(), mipmap_offsets.end(), mipmap_address); - if (mipmap_it == mipmap_offsets.end()) { - return {}; - } - const auto level{static_cast<u32>(std::distance(mipmap_offsets.begin(), mipmap_it))}; - return std::make_pair(layer, level); - } + std::optional<std::pair<u32, u32>> GetLayerMipmap(const GPUVAddr candidate_gpu_addr) const; std::vector<CopyParams> BreakDown(const SurfaceParams& in_params) const { return params.is_layered ? BreakDownLayered(in_params) : BreakDownNonLayered(in_params); @@ -241,35 +175,9 @@ private: void SwizzleFunc(MortonSwizzleMode mode, u8* memory, const SurfaceParams& params, u8* buffer, u32 level); - std::vector<CopyParams> BreakDownLayered(const SurfaceParams& in_params) const { - const u32 layers{params.depth}; - const u32 mipmaps{params.num_levels}; - std::vector<CopyParams> result; - result.reserve(static_cast<std::size_t>(layers) * static_cast<std::size_t>(mipmaps)); - - for (u32 layer = 0; layer < layers; layer++) { - for (u32 level = 0; level < mipmaps; level++) { - const u32 width = SurfaceParams::IntersectWidth(params, in_params, level, level); - const u32 height = SurfaceParams::IntersectHeight(params, in_params, level, level); - result.emplace_back(width, height, layer, level); - } - } - return result; - } - - std::vector<CopyParams> BreakDownNonLayered(const SurfaceParams& in_params) const { - const u32 mipmaps{params.num_levels}; - std::vector<CopyParams> result; - result.reserve(mipmaps); + std::vector<CopyParams> BreakDownLayered(const SurfaceParams& in_params) const; - for (u32 level = 0; level < mipmaps; level++) { - const u32 width = SurfaceParams::IntersectWidth(params, in_params, level, level); - const u32 height = SurfaceParams::IntersectHeight(params, in_params, level, level); - const u32 depth{std::min(params.GetMipDepth(level), in_params.GetMipDepth(level))}; - result.emplace_back(width, height, depth, level); - } - return result; - } + std::vector<CopyParams> BreakDownNonLayered(const SurfaceParams& in_params) const; }; template <typename TView> diff --git a/src/video_core/texture_cache/surface_params.cpp b/src/video_core/texture_cache/surface_params.cpp index 77c09264a..60a7356bb 100644 --- a/src/video_core/texture_cache/surface_params.cpp +++ b/src/video_core/texture_cache/surface_params.cpp @@ -6,7 +6,6 @@ #include "common/alignment.h" #include "common/bit_util.h" -#include "common/cityhash.h" #include "core/core.h" #include "video_core/engines/shader_bytecode.h" #include "video_core/surface.h" @@ -237,14 +236,6 @@ std::size_t SurfaceParams::GetConvertedMipmapOffset(u32 level) const { return offset; } -std::size_t SurfaceParams::GetGuestMipmapSize(u32 level) const { - return GetInnerMipmapMemorySize(level, false, false); -} - -std::size_t SurfaceParams::GetHostMipmapSize(u32 level) const { - return GetInnerMipmapMemorySize(level, true, false) * GetNumLayers(); -} - std::size_t SurfaceParams::GetConvertedMipmapSize(u32 level) const { constexpr std::size_t rgb8_bpp = 4ULL; const std::size_t width_t = GetMipWidth(level); @@ -253,10 +244,6 @@ std::size_t SurfaceParams::GetConvertedMipmapSize(u32 level) const { return width_t * height_t * depth_t * rgb8_bpp; } -std::size_t SurfaceParams::GetGuestLayerSize() const { - return GetLayerSize(false, false); -} - std::size_t SurfaceParams::GetLayerSize(bool as_host_size, bool uncompressed) const { std::size_t size = 0; for (u32 level = 0; level < num_levels; ++level) { @@ -269,16 +256,6 @@ std::size_t SurfaceParams::GetLayerSize(bool as_host_size, bool uncompressed) co return size; } -std::size_t SurfaceParams::GetHostLayerSize(u32 level) const { - ASSERT(target != SurfaceTarget::Texture3D); - return GetInnerMipmapMemorySize(level, true, false); -} - -bool SurfaceParams::IsPixelFormatZeta() const { - return pixel_format >= VideoCore::Surface::PixelFormat::MaxColorFormat && - pixel_format < VideoCore::Surface::PixelFormat::MaxDepthStencilFormat; -} - std::size_t SurfaceParams::GetInnerMipmapMemorySize(u32 level, bool as_host_size, bool uncompressed) const { const bool tiled{as_host_size ? false : is_tiled}; @@ -289,16 +266,6 @@ std::size_t SurfaceParams::GetInnerMipmapMemorySize(u32 level, bool as_host_size GetMipBlockHeight(level), GetMipBlockDepth(level)); } -std::size_t SurfaceParams::GetInnerMemorySize(bool as_host_size, bool layer_only, - bool uncompressed) const { - return GetLayerSize(as_host_size, uncompressed) * (layer_only ? 1U : depth); -} - -std::size_t SurfaceParams::Hash() const { - return static_cast<std::size_t>( - Common::CityHash64(reinterpret_cast<const char*>(this), sizeof(*this))); -} - bool SurfaceParams::operator==(const SurfaceParams& rhs) const { return std::tie(is_tiled, block_width, block_height, block_depth, tile_width_spacing, width, height, depth, pitch, num_levels, pixel_format, component_type, type, target) == diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h index 5fde695b6..c51e174cd 100644 --- a/src/video_core/texture_cache/surface_params.h +++ b/src/video_core/texture_cache/surface_params.h @@ -8,6 +8,7 @@ #include "common/alignment.h" #include "common/bit_util.h" +#include "common/cityhash.h" #include "common/common_types.h" #include "video_core/engines/fermi_2d.h" #include "video_core/engines/maxwell_3d.h" @@ -39,7 +40,10 @@ public: static SurfaceParams CreateForFermiCopySurface( const Tegra::Engines::Fermi2D::Regs::Surface& config); - std::size_t Hash() const; + std::size_t Hash() const { + return static_cast<std::size_t>( + Common::CityHash64(reinterpret_cast<const char*>(this), sizeof(*this))); + } bool operator==(const SurfaceParams& rhs) const; @@ -113,18 +117,27 @@ public: std::size_t GetConvertedMipmapOffset(u32 level) const; /// Returns the size in bytes in guest memory of a given mipmap level. - std::size_t GetGuestMipmapSize(u32 level) const; + std::size_t GetGuestMipmapSize(u32 level) const { + return GetInnerMipmapMemorySize(level, false, false); + } /// Returns the size in bytes in host memory (linear) of a given mipmap level. - std::size_t GetHostMipmapSize(u32 level) const; + std::size_t GetHostMipmapSize(u32 level) const { + return GetInnerMipmapMemorySize(level, true, false) * GetNumLayers(); + } std::size_t GetConvertedMipmapSize(u32 level) const; /// Returns the size of a layer in bytes in guest memory. - std::size_t GetGuestLayerSize() const; + std::size_t GetGuestLayerSize() const { + return GetLayerSize(false, false); + } /// Returns the size of a layer in bytes in host memory for a given mipmap level. - std::size_t GetHostLayerSize(u32 level) const; + std::size_t GetHostLayerSize(u32 level) const { + ASSERT(target != VideoCore::Surface::SurfaceTarget::Texture3D); + return GetInnerMipmapMemorySize(level, true, false); + } static u32 ConvertWidth(u32 width, VideoCore::Surface::PixelFormat pixel_format_from, VideoCore::Surface::PixelFormat pixel_format_to) { @@ -194,7 +207,10 @@ public: } /// Returns true if the pixel format is a depth and/or stencil format. - bool IsPixelFormatZeta() const; + bool IsPixelFormatZeta() const { + return pixel_format >= VideoCore::Surface::PixelFormat::MaxColorFormat && + pixel_format < VideoCore::Surface::PixelFormat::MaxDepthStencilFormat; + } SurfaceCompression GetCompressionType() const { return VideoCore::Surface::GetFormatCompressionType(pixel_format); @@ -229,7 +245,9 @@ private: std::size_t GetInnerMipmapMemorySize(u32 level, bool as_host_size, bool uncompressed) const; /// Returns the size of all mipmap levels and aligns as needed. - std::size_t GetInnerMemorySize(bool as_host_size, bool layer_only, bool uncompressed) const; + std::size_t GetInnerMemorySize(bool as_host_size, bool layer_only, bool uncompressed) const { + return GetLayerSize(as_host_size, uncompressed) * (layer_only ? 1U : depth); + } /// Returns the size of a layer std::size_t GetLayerSize(bool as_host_size, bool uncompressed) const; |