Construction & Extraction

Construct vectors inside a generated feature context. define(f32x8) makes the body-local spelling concise; explicit f32x8::<Token> works equally well.

This API exercise uses the load/multiply/store operations in the zenfilters gain chain, reduced to one array so the construction and extraction calls are visible.

use archmage::prelude::*;
#[magetypes(define(f32x8), v3, neon, wasm128, scalar)]
fn double_impl(token: Token, input: &[f32; 8]) -> [f32; 8] {
    let v = f32x8::load_t(token, input);
    (v * f32x8::splat_t(token, 2.0)).to_array()
}
pub fn double(input: &[f32; 8]) -> [f32; 8] {
    incant!(double_impl(input), [v3, neon, wasm128, scalar])
}
assert_eq!(double(&[3.0; 8]), [6.0; 8]);

Call	Contract
`f32x8::zero_t(token)`	All lanes zero
`f32x8::splat_t(token, value)`	Repeat a scalar
`f32x8::from_array_t(token, values)`	Consume a fixed-size array
`f32x8::load_t(token, &values)`	Load a fixed-size array reference
`f32x8::from_slice_t(token, values)`	Load the first vector; requires enough elements
`v.to_array()`	Return all scalar lanes by value
`v.store(&mut values)`	Store to a fixed-size array reference
`v[index]`	Scalar lane access with Rust bounds checking

Type inference often supplies T; turbofish is needed only when inference is insufficient. Fixed arrays carry the vector length in the type. Slice lengths and dynamic lane indices still need checks unless optimization proves them. See memory and bounds.

Found an error or it needs a clarification? Open an issue on GitHub.

Substantiated corrections will be incorporated with attribution.

Found a typo? Fork, modify and open a PR.