pub fn element_wise_splat<S: Simd>(simd: S, input: f32x4<S>) -> f32x16<S>
Splat each single element in the vector to 4 lanes.