Module texture_compress 

Multi-format texture compression (BC1 and BC4) for OxiGDAL GPU.

This module provides both pure-Rust CPU implementations and GPU-accelerated (wgpu compute shader) paths for compressing raster data into BC1 (DXT1) and BC4 (ATI1) block-compressed texture formats.

Format overview

Format	Input	Output	Ratio
TextureFormat::Bc1RgbUnorm	RGBA8 (4 B/px)	8 B / 4×4 block	6:1
TextureFormat::Bc4RUnorm	R8 (1 B/px)	8 B / 4×4 block	2:1

BC1 block layout (8 bytes)

[u16 c0 LE] [u16 c1 LE] [u32 lookup LE]

When c0 > c1 four colours are interpolated (opaque mode, always used here). When c0 <= c1 three colours + transparent; this module never emits that mode.

BC4 block layout (8 bytes)

[u8 ep0] [u8 ep1] [48-bit LE packed 3-bit indices for 16 pixels]

When ep0 > ep1 eight interpolated values are used (always in this module).

GPU path

TextureCompressor drives a wgpu compute shader that processes one 4×4 block per workgroup thread. When wgpu is unavailable the compress method falls back silently to the CPU path.

Structs

TextureCompressor

GPU-accelerated texture compressor for BC1 and BC4 block formats.

Enums

TextureFormat

Target block-compressed texture format produced by TextureCompressor.

Functions

compress_bc1_block_cpu

Compress one 4×4 RGBA block (64 bytes, row-major) to an 8-byte BC1 block.

compress_bc4_block_cpu

Compress one 4×4 R-channel block (16 bytes) to an 8-byte BC4 block.

dequantize_rgb565

Reverse a packed RGB565 value back to approximate 8-bit R, G, B.

nearest_index_4

Return the index (0..4) of the endpoint in endpoints nearest to value_rgb under the squared-Euclidean distance metric in RGB space.

quantize_rgb565

Quantise 8-bit RGB components to packed 16-bit RGB565.

validate_texture_dimensions

Validate that width and height are both multiples of 4 (required for block-compressed textures).