Expand description
Fixed-width vector storage for the encoder.
The types here are plain arrays: Default + Copy, so they can live in the
encoder’s allocator-backed slices and be handed to Allocator<T>. They carry no
arithmetic of their own. Math is done on fearless_simd vectors instead: inside a
dispatch! region, Mem256f::to_simd (and friends) loads a register and
Mem256f::from_simd stores it back.
Structs§
Functions§
- detect_
level - The instruction set the vectorized encoder paths run on.
- min_
lane_ f32x8 - The smallest lane of
v, folded inlog2(8)steps. - min_
lane_ u32x8 - The smallest lane of
v, folded inlog2(8)steps.