Crate tegra_swizzle

tegra_swizzle

tegra_swizzle is an unofficial CPU implementation of the Tegra X1 block linear memory tiling for texture surfaces.

Getting Started

Tiled texture data is often stored in a single buffer containing all arrays and mipmaps. This memory layout can be untiled all at once using surface::deswizzle_surface.

use tegra_swizzle::surface::{BlockDim, deswizzle_surface};
use std::num::NonZeroUsize;

// 16x16 BC7 cube map with 5 mipmaps.
let surface = deswizzle_surface(
    16,
    16,
    1,
    &swizzled_surface,
    BlockDim::block_4x4(),
    None,
    16,
    5,
    6,
);

// 128x128 R8G8B8A8 2D texture with no mipmaps.
let surface = deswizzle_surface(
    128,
    128,
    1,
    &swizzled_surface,
    BlockDim::uncompressed(),
    None,
    4,
    1,
    1,
);

// 16x16x16 R8G8B8A8 3D texture with no mipmaps.
let surface = deswizzle_surface(
    16,
    16,
    16,
    &swizzled_surface,
    BlockDim::uncompressed(),
    None,
    4,
    1,
    1,
);

Block Linear Memory Tiling

The surface::swizzle_surface and surface::deswizzle_surface functions implement safe and efficient tiling and untiling for the Tegra X1’s block linear format.

Block linear arranges bytes of a texture surface into a 2D grid of blocks where blocks are arranged linearly in row-major order. The tiled surface size is padded to integral dimensions in blocks, so tiled surfaces may be larger than the corresponding data in row-major order.

Groups of 512 bytes form GOBs (“group of bytes”) where each GOB is 64x8 bytes. The block_height parameter determines how many GOBs stack vertically to form a block.

Modules

• ffi
  Documentation for the C API.
• surface
  Functions for working with surfaces stored in a combined buffer for all array layers and mipmaps.
• swizzle
  Functions for tiling and untiling a single mipmap of a surface.

Enums

• BlockHeight
  The height of each block in GOBs where each GOB is 8 bytes tall.
• SwizzleError
  Errors than can occur while tiling or untiling.

Functions

• block_height_mip0
  Calculates the block height parameter to use for the first mip level if no block height is specified.
• div_round_up
  Calculates the division of x by d but rounds up rather than truncating.
• mip_block_height
  Calculates the block height parameter for the given mip level.