From 57b44200a2bf37cc4775ea88b5ff8ed9f67bc145 Mon Sep 17 00:00:00 2001 From: FernandoS27 Date: Fri, 21 Sep 2018 12:09:59 -0400 Subject: Reverse stride align restriction on FastSwizzle due to lost performance --- src/video_core/textures/decoders.cpp | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src/video_core/textures/decoders.cpp') diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 5a2a0b84d..3d5476e5d 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -85,7 +85,6 @@ static void FastSwizzleData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_ const std::size_t pixel_base{y * width * out_bytes_per_pixel}; const auto& table = fast_swizzle_table[y % gobs_in_y]; for (std::size_t xb = 0; xb < stride; xb += copy_size) { - const std::size_t truncated_copy = std::min(copy_size, stride - xb); const std::size_t gob_address{initial_gob + (xb / gobs_in_x) * gobs_size * block_height}; const std::size_t swizzle_offset{gob_address + table[(xb / 16) % 4]}; @@ -93,14 +92,14 @@ static void FastSwizzleData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_ const std::size_t pixel_index{out_x + pixel_base}; data_ptrs[unswizzle] = swizzled_data + swizzle_offset; data_ptrs[!unswizzle] = unswizzled_data + pixel_index; - std::memcpy(data_ptrs[0], data_ptrs[1], truncated_copy); + std::memcpy(data_ptrs[0], data_ptrs[1], copy_size); } } } 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) { - if (bytes_per_pixel % 3 != 0) { + if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % 16 == 0) { FastSwizzleData(width, height, bytes_per_pixel, out_bytes_per_pixel, swizzled_data, unswizzled_data, unswizzle, block_height); } else { -- cgit v1.2.3