Comment by MarekDlugos

re: optimization for 1024b vectors — do you pad shorter ones, or fallback to a more general kernel?