Avoiding conversion and rearrangement overhead in SIMD architectures

More Info
expand_more