And for static-ish-typed numerical processing like that shown here (and mentioned already), PDL is the way to achieve great performance (and since 2.062, using multi-cores by default for "broadcasting").