Arithmetic & Comparisons

Magetypes supports natural arithmetic inside the feature-enabled kernels shown in gain and SrcOver blending.

Operation	Calls / syntax	Notes
Arithmetic	`a + b`, `a - b`, `a * b`, `a / b`, `-a`	Availability depends on element type
Assignment	`+=`, `-=`, `*=`	Same vector semantics
Fused form	`a.mul_add(b, c)`, `a.mul_sub(b, c)`	One rounding, except relaxed WASM follows the engine; software fallback on strict WASM and scalar
Comparisons	`simd_eq`, `simd_ne`, `simd_lt`, `simd_le`, `simd_gt`, `simd_ge`	Produce vector lane masks, not Rust scalar booleans
Selection	`f32x8::blend(mask, yes, no)`	Use comparison-generated canonical masks
Min/max	`min`, `max`	NaN and signed-zero behavior needs the ISA contract
Magnitude	`abs`, `sqrt`	`sqrt` is rounded floating-point square root, not exact real arithmetic

This reference exercise shows compare/select in the same generated call-chain shape used by zen color kernels. It is not a complete tone-mapping algorithm.

use archmage::prelude::*;
#[magetypes(define(f32x8), v3, neon, wasm128, scalar)]
fn positive_impl(token: Token, values: [f32; 8]) -> [f32; 8] {
    let v = f32x8::from_array_t(token, values);
    let zero = f32x8::zero_t(token);
    f32x8::blend(v.simd_gt(zero), v, zero).to_array()
}
pub fn positive(values: [f32; 8]) -> [f32; 8] {
    incant!(positive_impl(values), [v3, neon, wasm128, scalar])
}
assert_eq!(positive([-1.0, 2.0, 0.0, 3.0, -4.0, 5.0, 6.0, 7.0]),
           [0.0, 2.0, 0.0, 3.0, 0.0, 5.0, 6.0, 7.0]);

Do not assume all boolean identities hold for NaNs: current simd_ne semantics vary. Do not infer bit-identical results from replacing multiply-plus-add with mul_add. The ISA tables specify these differences and the fixups the library does apply.

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.