pub fn simd_normalize_f32(input: &ArrayView1<'_, f32>) -> Array1<f32>Expand description
SIMD-accelerated L2 normalization for f32 arrays
Normalizes the input array to unit L2 norm (Euclidean length = 1). If the norm is zero or NaN, returns a zero array.
§Arguments
input- Input array to normalize
§Returns
Normalized array where ||output||_2 = 1 (or zero array if input norm is zero)
§Performance
- f32 (100K elements): 0.44x-0.60x vs NumPy (limited by two-pass algorithm)
- f64 (100K elements): 0.39x vs NumPy
Note: Performance is limited by the two-pass nature of the algorithm (calculate norm, then divide). NumPy uses highly optimized BLAS routines.