pub fn matmul_blocked(
a: &[f32],
a_rows: usize,
a_cols: usize,
b: &[f32],
b_rows: usize,
b_cols: usize,
block: usize,
) -> Result<Vec<f32>, EngineError>Expand description
Blocked matmul used as Neon / SIMD-friendly path (portable; aarch64 may specialize later).