Installation

Magetypes depends on archmage for capability tokens. Add both to your Cargo.toml:

[dependencies]
archmage = "0.8"
magetypes = "0.8"

Magetypes re-exports what it needs from archmage, but you'll want archmage directly for SimdToken, token types, and macros like #[arcane].

Feature Flags

magetypes

archmage

Enable AVX-512 types in both crates:

[dependencies]
archmage = { version = "0.8", features = ["avx512"] }
magetypes = { version = "0.8", features = ["avx512"] }

no_std

Both crates support no_std with alloc:

[dependencies]
archmage = { version = "0.8", default-features = false, features = ["macros"] }
magetypes = { version = "0.8", default-features = false }

Platform Requirements

x86-64

Works out of the box. Tokens detect CPU features at runtime via CPUID.

To compile for a known target CPU (detection compiles away):

RUSTFLAGS="-Ctarget-cpu=native" cargo build --release

AArch64

NEON is baseline on 64-bit ARM. NeonToken::summon() succeeds on all AArch64 hardware.

WASM

Enable SIMD128 in your build:

RUSTFLAGS="-Ctarget-feature=+simd128" cargo build --target wasm32-unknown-unknown

Verify It Works

use archmage::{X64V3Token, SimdToken};
use magetypes::simd::{
    generic::f32x8,
    backends::F32x8Backend,
};

#[inline(always)]
fn print_splat<T: F32x8Backend>(token: T, val: f32) {
    let v = f32x8::<T>::splat(token, val);
    println!("f32x8: {:?}", v.to_array());
}

fn main() {
    match X64V3Token::summon() {
        Some(token) => print_splat(token, 42.0),
        None => println!("AVX2+FMA not available on this CPU"),
    }
}

cargo run