8 files changed, 55 insertions, 36 deletions

diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp
index 74e44c7fe..ee8ebf678 100644
--- a/src/video_core/engines/fermi_2d.cpp
+++ b/src/video_core/engines/fermi_2d.cpp
@@ -68,13 +68,13 @@ void Fermi2D::HandleSurfaceCopy() {
             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());
+                                      regs.src.BlockDepth(), 0);
         } 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, regs.src.depth,
                                       src_bytes_per_pixel, dst_bytes_per_pixel, dst_buffer,
                                       src_buffer, false, regs.dst.BlockHeight(),
-                                      regs.dst.BlockDepth());
+                                      regs.dst.BlockDepth(), 0);
         }
     }
 }
diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp
index f14abba7d..1cf27ee25 100644
--- a/src/video_core/morton.cpp
+++ b/src/video_core/morton.cpp
@@ -16,12 +16,12 @@ namespace VideoCore {
 using Surface::GetBytesPerPixel;
 using Surface::PixelFormat;
 
-using MortonCopyFn = void (*)(u32, u32, u32, u32, u32, u8*, std::size_t, VAddr);
+using MortonCopyFn = void (*)(u32, u32, u32, u32, u32, u32, u8*, std::size_t, VAddr);
 using ConversionArray = std::array<MortonCopyFn, Surface::MaxPixelFormat>;
 
 template <bool morton_to_linear, PixelFormat format>
 static void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth, u32 depth,
-                       u8* buffer, std::size_t buffer_size, VAddr addr) {
+                       u32 tile_width_spacing, u8* buffer, std::size_t buffer_size, VAddr addr) {
     constexpr u32 bytes_per_pixel = GetBytesPerPixel(format);
 
     // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
@@ -31,12 +31,13 @@ static void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth
 
     if constexpr (morton_to_linear) {
         Tegra::Texture::UnswizzleTexture(buffer, addr, tile_size_x, tile_size_y, bytes_per_pixel,
-                                         stride, height, depth, block_height, block_depth);
+                                         stride, height, depth, block_height, block_depth,
+                                         tile_width_spacing);
     } else {
-        Tegra::Texture::CopySwizzledData((stride + tile_size_x - 1) / tile_size_x,
-                                         (height + tile_size_y - 1) / tile_size_y, depth,
-                                         bytes_per_pixel, bytes_per_pixel, Memory::GetPointer(addr),
-                                         buffer, false, block_height, block_depth);
+        Tegra::Texture::CopySwizzledData(
+            (stride + tile_size_x - 1) / tile_size_x, (height + tile_size_y - 1) / tile_size_y,
+            depth, bytes_per_pixel, bytes_per_pixel, Memory::GetPointer(addr), buffer, false,
+            block_height, block_depth, tile_width_spacing);
     }
 }
 
@@ -325,11 +326,11 @@ static u32 GetMortonOffset128(u32 x, u32 y, u32 bytes_per_pixel) {
 }
 
 void MortonSwizzle(MortonSwizzleMode mode, Surface::PixelFormat format, u32 stride,
-                   u32 block_height, u32 height, u32 block_depth, u32 depth, u8* buffer,
-                   std::size_t buffer_size, VAddr addr) {
+                   u32 block_height, u32 height, u32 block_depth, u32 depth, u32 tile_width_spacing,
+                   u8* buffer, std::size_t buffer_size, VAddr addr) {
 
-    GetSwizzleFunction(mode, format)(stride, block_height, height, block_depth, depth, buffer,
-                                     buffer_size, addr);
+    GetSwizzleFunction(mode, format)(stride, block_height, height, block_depth, depth,
+                                     tile_width_spacing, buffer, buffer_size, addr);
 }
 
 void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixel, u32 linear_bytes_per_pixel,
@@ -350,4 +351,4 @@ void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixel, u32 linear_
     }
 }
 
-} // namespace VideoCore
\ No newline at end of file
+} // namespace VideoCore
diff --git a/src/video_core/morton.h b/src/video_core/morton.h
index b9b9eca86..065f59ce3 100644
--- a/src/video_core/morton.h
+++ b/src/video_core/morton.h
@@ -12,10 +12,10 @@ namespace VideoCore {
 enum class MortonSwizzleMode { MortonToLinear, LinearToMorton };
 
 void MortonSwizzle(MortonSwizzleMode mode, VideoCore::Surface::PixelFormat format, u32 stride,
-                   u32 block_height, u32 height, u32 block_depth, u32 depth, u8* buffer,
-                   std::size_t buffer_size, VAddr addr);
+                   u32 block_height, u32 height, u32 block_depth, u32 depth, u32 tile_width_spacing,
+                   u8* buffer, std::size_t buffer_size, VAddr addr);
 
 void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixel, u32 linear_bytes_per_pixel,
                          u8* morton_data, u8* linear_data, bool morton_to_linear);
 
-} // namespace VideoCore
\ No newline at end of file
+} // namespace VideoCore
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index d458f77e4..dde2f468d 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -97,6 +97,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
     params.block_width = params.is_tiled ? config.tic.BlockWidth() : 0,
     params.block_height = params.is_tiled ? config.tic.BlockHeight() : 0,
     params.block_depth = params.is_tiled ? config.tic.BlockDepth() : 0,
+    params.tile_width_spacing = params.is_tiled ? (1 << config.tic.tile_width_spacing.Value()) : 1;
     params.srgb_conversion = config.tic.IsSrgbConversionEnabled();
     params.pixel_format = PixelFormatFromTextureFormat(config.tic.format, config.tic.r_type.Value(),
                                                        params.srgb_conversion);
@@ -162,6 +163,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
     params.block_width = 1 << config.memory_layout.block_width;
     params.block_height = 1 << config.memory_layout.block_height;
     params.block_depth = 1 << config.memory_layout.block_depth;
+    params.tile_width_spacing = 1;
     params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
     params.srgb_conversion = config.format == Tegra::RenderTargetFormat::BGRA8_SRGB ||
                              config.format == Tegra::RenderTargetFormat::RGBA8_SRGB;
@@ -197,6 +199,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
     params.block_width = 1 << std::min(block_width, 5U);
     params.block_height = 1 << std::min(block_height, 5U);
     params.block_depth = 1 << std::min(block_depth, 5U);
+    params.tile_width_spacing = 1;
     params.pixel_format = PixelFormatFromDepthFormat(format);
     params.component_type = ComponentTypeFromDepthFormat(format);
     params.type = GetFormatType(params.pixel_format);
@@ -223,6 +226,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
     params.block_width = params.is_tiled ? std::min(config.BlockWidth(), 32U) : 0,
     params.block_height = params.is_tiled ? std::min(config.BlockHeight(), 32U) : 0,
     params.block_depth = params.is_tiled ? std::min(config.BlockDepth(), 32U) : 0,
+    params.tile_width_spacing = 1;
     params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
     params.srgb_conversion = config.format == Tegra::RenderTargetFormat::BGRA8_SRGB ||
                              config.format == Tegra::RenderTargetFormat::RGBA8_SRGB;
@@ -387,8 +391,8 @@ void SwizzleFunc(const MortonSwizzleMode& mode, const SurfaceParams& params,
         for (u32 i = 0; i < params.depth; i++) {
             MortonSwizzle(mode, params.pixel_format, params.MipWidth(mip_level),
                           params.MipBlockHeight(mip_level), params.MipHeight(mip_level),
-                          params.MipBlockDepth(mip_level), 1, gl_buffer.data() + offset_gl, gl_size,
-                          params.addr + offset);
+                          params.MipBlockDepth(mip_level), params.tile_width_spacing, 1,
+                          gl_buffer.data() + offset_gl, gl_size, params.addr + offset);
             offset += layer_size;
             offset_gl += gl_size;
         }
@@ -396,8 +400,8 @@ void SwizzleFunc(const MortonSwizzleMode& mode, const SurfaceParams& params,
         const u64 offset = params.GetMipmapLevelOffset(mip_level);
         MortonSwizzle(mode, params.pixel_format, params.MipWidth(mip_level),
                       params.MipBlockHeight(mip_level), params.MipHeight(mip_level),
-                      params.MipBlockDepth(mip_level), depth, gl_buffer.data(), gl_buffer.size(),
-                      params.addr + offset);
+                      params.MipBlockDepth(mip_level), depth, params.tile_width_spacing,
+                      gl_buffer.data(), gl_buffer.size(), params.addr + offset);
     }
 }
 
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index 9ac79c5a4..c710aa245 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -196,9 +196,15 @@ struct SurfaceParams {
 
     /// Checks if surfaces are compatible for caching
     bool IsCompatibleSurface(const SurfaceParams& other) const {
-        return std::tie(pixel_format, type, width, height, target, depth) ==
-               std::tie(other.pixel_format, other.type, other.width, other.height, other.target,
-                        other.depth);
+        if (std::tie(pixel_format, type, width, height, target, depth, is_tiled) ==
+            std::tie(other.pixel_format, other.type, other.width, other.height, other.target,
+                     other.depth, other.is_tiled)) {
+            if (!is_tiled)
+                return true;
+            return std::tie(block_height, block_depth, tile_width_spacing) ==
+                   std::tie(other.block_height, other.block_depth, other.tile_width_spacing);
+        }
+        return false;
     }
 
     /// Initializes parameters for caching, should be called after everything has been initialized
@@ -208,6 +214,7 @@ struct SurfaceParams {
     u32 block_width;
     u32 block_height;
     u32 block_depth;
+    u32 tile_width_spacing;
     PixelFormat pixel_format;
     ComponentType component_type;
     SurfaceType type;
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp
index 7eabd34f1..bbae9285f 100644
--- a/src/video_core/textures/decoders.cpp
+++ b/src/video_core/textures/decoders.cpp
@@ -127,7 +127,8 @@ void FastProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const
 template <bool fast>
 void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool unswizzle,
                   const u32 width, const u32 height, const u32 depth, const u32 bytes_per_pixel,
-                  const u32 out_bytes_per_pixel, const u32 block_height, const u32 block_depth) {
+                  const u32 out_bytes_per_pixel, const u32 block_height, const u32 block_depth,
+                  const u32 width_spacing) {
     auto div_ceil = [](const u32 x, const u32 y) { return ((x + y - 1) / y); };
     const u32 stride_x = width * out_bytes_per_pixel;
     const u32 layer_z = height * stride_x;
@@ -137,7 +138,8 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool
     const u32 block_x_elements = gob_elements_x;
     const u32 block_y_elements = gob_elements_y * block_height;
     const u32 block_z_elements = gob_elements_z * block_depth;
-    const u32 blocks_on_x = div_ceil(width, block_x_elements);
+    const u32 aligned_width = Common::AlignUp(width, gob_elements_x * width_spacing);
+    const u32 blocks_on_x = div_ceil(aligned_width, block_x_elements);
     const u32 blocks_on_y = div_ceil(height, block_y_elements);
     const u32 blocks_on_z = div_ceil(depth, block_z_elements);
     const u32 xy_block_size = gob_size * block_height;
@@ -169,13 +171,15 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool
 
 void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel,
                       u32 out_bytes_per_pixel, u8* const swizzled_data, u8* const unswizzled_data,
-                      bool unswizzle, u32 block_height, u32 block_depth) {
+                      bool unswizzle, u32 block_height, u32 block_depth, u32 width_spacing) {
     if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % fast_swizzle_align == 0) {
         SwizzledData<true>(swizzled_data, unswizzled_data, unswizzle, width, height, depth,
-                           bytes_per_pixel, out_bytes_per_pixel, block_height, block_depth);
+                           bytes_per_pixel, out_bytes_per_pixel, block_height, block_depth,
+                           width_spacing);
     } else {
         SwizzledData<false>(swizzled_data, unswizzled_data, unswizzle, width, height, depth,
-                            bytes_per_pixel, out_bytes_per_pixel, block_height, block_depth);
+                            bytes_per_pixel, out_bytes_per_pixel, block_height, block_depth,
+                            width_spacing);
     }
 }
 
@@ -228,19 +232,19 @@ u32 BytesPerPixel(TextureFormat format) {
 
 void UnswizzleTexture(u8* const unswizzled_data, VAddr address, u32 tile_size_x, u32 tile_size_y,
                       u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height,
-                      u32 block_depth) {
+                      u32 block_depth, u32 width_spacing) {
     CopySwizzledData((width + tile_size_x - 1) / tile_size_x,
                      (height + tile_size_y - 1) / tile_size_y, depth, bytes_per_pixel,
                      bytes_per_pixel, Memory::GetPointer(address), unswizzled_data, true,
-                     block_height, block_depth);
+                     block_height, block_depth, width_spacing);
 }
 
 std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y,
                                  u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
-                                 u32 block_height, u32 block_depth) {
+                                 u32 block_height, u32 block_depth, u32 width_spacing) {
     std::vector<u8> unswizzled_data(width * height * depth * bytes_per_pixel);
     UnswizzleTexture(unswizzled_data.data(), address, tile_size_x, tile_size_y, bytes_per_pixel,
-                     width, height, depth, block_height, block_depth);
+                     width, height, depth, block_height, block_depth, width_spacing);
     return unswizzled_data;
 }
 
diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h
index f4ef7c73e..85b7e9f7b 100644
--- a/src/video_core/textures/decoders.h
+++ b/src/video_core/textures/decoders.h
@@ -22,19 +22,20 @@ inline std::size_t GetGOBSize() {
 void UnswizzleTexture(u8* unswizzled_data, VAddr address, u32 tile_size_x, u32 tile_size_y,
                       u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                       u32 block_height = TICEntry::DefaultBlockHeight,
-                      u32 block_depth = TICEntry::DefaultBlockHeight);
+                      u32 block_depth = TICEntry::DefaultBlockHeight, u32 width_spacing = 0);
 /**
  * Unswizzles a swizzled texture without changing its format.
  */
 std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y,
                                  u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                                  u32 block_height = TICEntry::DefaultBlockHeight,
-                                 u32 block_depth = TICEntry::DefaultBlockHeight);
+                                 u32 block_depth = TICEntry::DefaultBlockHeight,
+                                 u32 width_spacing = 0);
 
 /// Copies texture data from a buffer and performs swizzling/unswizzling as necessary.
 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);
+                      bool unswizzle, u32 block_height, u32 block_depth, u32 width_spacing);
 
 /**
  * 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 ffa08f5c1..e7c78bee2 100644
--- a/src/video_core/textures/texture.h
+++ b/src/video_core/textures/texture.h
@@ -166,6 +166,8 @@ struct TICEntry {
         BitField<3, 3, u32> block_height;
         BitField<6, 3, u32> block_depth;
 
+        BitField<10, 3, u32> tile_width_spacing;
+
         // High 16 bits of the pitch value
         BitField<0, 16, u32> pitch_high;
         BitField<26, 1, u32> use_header_opt_control;