Sorry about that; this was a copy-and-paste error, and it's basically Moodle's fault (OK it's my fault since I should have checked). Of course, int(1.5 * r) does not make sense as a loop length because all the problem sizes would be in the L1 cache. The whole point of the exercise is to make you produce a performance graph similar to what was shown in the lecture (last slide of lecture 05/06). For that, we need array lengths that span multiple orders of magnitude. The "weird" formula int(1.5^r) makes sure that the data points are equidistant if you plot them on a log-x scale.