video_core: Use un-shifted block sizes to avoid integer divisions

Instead of storing all block width, height and depths in their shifted
form:

block_width = 1U << block_shift;

Store them like they are provided by the emulated hardware (their
block_shift form). This way we can avoid doing the costly
Common::AlignUp operation to align texture sizes and drop CPU integer
divisions with bitwise logic (defined in Common::AlignBits).

4 files changed, 28 insertions, 24 deletions

diff --git a/src/video_core/texture_cache/surface_base.cpp b/src/video_core/texture_cache/surface_base.cpp
index 36ca72b4a..510d1aef5 100644
--- a/src/video_core/texture_cache/surface_base.cpp
+++ b/src/video_core/texture_cache/surface_base.cpp
@@ -22,7 +22,6 @@ SurfaceBaseImpl::SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params)
     : params{params}, mipmap_sizes(params.num_levels),
       mipmap_offsets(params.num_levels), gpu_addr{gpu_addr}, host_memory_size{
                                                                  params.GetHostSizeInBytes()} {
-
     std::size_t offset = 0;
     for (u32 level = 0; level < params.num_levels; ++level) {
         const std::size_t mipmap_size{params.GetGuestMipmapSize(level)};
@@ -75,7 +74,7 @@ void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager,
         return;
     }
     if (params.is_tiled) {
-        ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture target {}",
+        ASSERT_MSG(params.block_width == 0, "Block width is defined as {} on texture target {}",
                    params.block_width, static_cast<u32>(params.target));
         for (u32 level = 0; level < params.num_levels; ++level) {
             const std::size_t host_offset{params.GetHostMipmapLevelOffset(level)};
diff --git a/src/video_core/texture_cache/surface_params.cpp b/src/video_core/texture_cache/surface_params.cpp
index b537b26e2..3a47f404d 100644
--- a/src/video_core/texture_cache/surface_params.cpp
+++ b/src/video_core/texture_cache/surface_params.cpp
@@ -96,9 +96,9 @@ SurfaceParams SurfaceParams::CreateForDepthBuffer(
     SurfaceParams params;
     params.is_tiled = type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
     params.srgb_conversion = false;
-    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.block_width = std::min(block_width, 5U);
+    params.block_height = std::min(block_height, 5U);
+    params.block_depth = std::min(block_depth, 5U);
     params.tile_width_spacing = 1;
     params.pixel_format = PixelFormatFromDepthFormat(format);
     params.component_type = ComponentTypeFromDepthFormat(format);
@@ -120,9 +120,9 @@ SurfaceParams SurfaceParams::CreateForFramebuffer(Core::System& system, std::siz
         config.memory_layout.type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
     params.srgb_conversion = config.format == Tegra::RenderTargetFormat::BGRA8_SRGB ||
                              config.format == Tegra::RenderTargetFormat::RGBA8_SRGB;
-    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.block_width = config.memory_layout.block_width;
+    params.block_height = config.memory_layout.block_height;
+    params.block_depth = config.memory_layout.block_depth;
     params.tile_width_spacing = 1;
     params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
     params.component_type = ComponentTypeFromRenderTarget(config.format);
@@ -149,9 +149,9 @@ SurfaceParams SurfaceParams::CreateForFermiCopySurface(
     params.is_tiled = !config.linear;
     params.srgb_conversion = config.format == Tegra::RenderTargetFormat::BGRA8_SRGB ||
                              config.format == Tegra::RenderTargetFormat::RGBA8_SRGB;
-    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.block_width = params.is_tiled ? std::min(config.BlockWidth(), 5U) : 0,
+    params.block_height = params.is_tiled ? std::min(config.BlockHeight(), 5U) : 0,
+    params.block_depth = params.is_tiled ? std::min(config.BlockDepth(), 5U) : 0,
     params.tile_width_spacing = 1;
     params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
     params.component_type = ComponentTypeFromRenderTarget(config.format);
@@ -190,9 +190,9 @@ u32 SurfaceParams::GetMipBlockHeight(u32 level) const {
     const u32 height{GetMipHeight(level)};
     const u32 default_block_height{GetDefaultBlockHeight()};
     const u32 blocks_in_y{(height + default_block_height - 1) / default_block_height};
-    u32 block_height = 16;
-    while (block_height > 1 && blocks_in_y <= block_height * 4) {
-        block_height >>= 1;
+    u32 block_height = 4;
+    while (block_height > 0 && blocks_in_y <= (1U << block_height) * 4) {
+        --block_height;
     }
     return block_height;
 }
@@ -202,17 +202,17 @@ u32 SurfaceParams::GetMipBlockDepth(u32 level) const {
         return this->block_depth;
     }
     if (is_layered) {
-        return 1;
+        return 0;
     }
 
     const u32 depth{GetMipDepth(level)};
-    u32 block_depth = 32;
-    while (block_depth > 1 && depth * 2 <= block_depth) {
-        block_depth >>= 1;
+    u32 block_depth = 5;
+    while (block_depth > 0 && depth * 2 <= (1U << block_depth)) {
+        --block_depth;
     }
 
-    if (block_depth == 32 && GetMipBlockHeight(level) >= 4) {
-        return 16;
+    if (block_depth == 5 && GetMipBlockHeight(level) >= 2) {
+        return 4;
     }
 
     return block_depth;
@@ -252,7 +252,8 @@ std::size_t SurfaceParams::GetLayerSize(bool as_host_size, bool uncompressed) co
         size += GetInnerMipmapMemorySize(level, as_host_size, uncompressed);
     }
     if (is_tiled && is_layered) {
-        return Common::AlignUp(size, Tegra::Texture::GetGOBSize() * block_height * block_depth);
+        return Common::AlignBits(size,
+                                 Tegra::Texture::GetGOBSizeShift() + block_height + block_depth);
     }
     return size;
 }
diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h
index e0ec1be0e..7c48782c7 100644
--- a/src/video_core/texture_cache/surface_params.h
+++ b/src/video_core/texture_cache/surface_params.h
@@ -54,12 +54,12 @@ public:
             constexpr std::size_t rgb8_bpp = 4ULL;
             // ASTC is uncompressed in software, in emulated as RGBA8
             host_size_in_bytes = 0;
-            for (std::size_t level = 0; level < num_levels; level++) {
+            for (u32 level = 0; level < num_levels; ++level) {
                 const std::size_t width =
                     Common::AlignUp(GetMipWidth(level), GetDefaultBlockWidth());
                 const std::size_t height =
                     Common::AlignUp(GetMipHeight(level), GetDefaultBlockHeight());
-                const std::size_t depth = is_layered ? depth : GetMipDepth(level);
+                const std::size_t depth = is_layered ? this->depth : GetMipDepth(level);
                 host_size_in_bytes += width * height * depth * rgb8_bpp;
             }
         } else {
@@ -96,7 +96,8 @@ public:
     // Helper used for out of class size calculations
     static std::size_t AlignLayered(const std::size_t out_size, const u32 block_height,
                                     const u32 block_depth) {
-        return Common::AlignUp(out_size, Tegra::Texture::GetGOBSize() * block_height * block_depth);
+        return Common::AlignBits(out_size,
+                                 Tegra::Texture::GetGOBSizeShift() + block_height + block_depth);
     }
 
     /// Returns the offset in bytes in guest memory of a given mipmap level.
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 1c2b63dae..f35d0c88f 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -81,6 +81,9 @@ public:
         if (!gpu_addr) {
             return {};
         }
+        if (gpu_addr == 0x1b7ec0000) {
+            // __debugbreak();
+        }
         const auto params{SurfaceParams::CreateForTexture(system, config, entry)};
         return GetSurface(gpu_addr, params, true).second;
     }