irithyll_core/simd/mod.rs
1//! SIMD-accelerated math primitives for neural forward passes.
2//!
3//! Provides AVX2-accelerated dot product and matrix-vector multiply
4//! with automatic runtime feature detection and scalar fallback.
5//!
6//! On `x86_64` targets with the `std` feature enabled, these functions
7//! use runtime `is_x86_feature_detected!("avx2")` to select AVX2 when
8//! available. On all other targets (or without `std`), they fall back
9//! to scalar implementations.
10
11pub mod ops;
12
13pub use ops::{simd_dot, simd_mat_vec};