I'd say (possibly being biased) that the avoidance of for-loops using https://en.wikipedia.org/wiki/Array_programming constructs both saves programmer time, reader/maintenance time, and (thanks to not having to enter/leave Perl land lots in a repetitive way) execution time. PDL provides such array-programming constructs, e.g. $out_pdl = $in1_pdl - $in2_pdl which works however large the two input ndarrays are.