> If the cost of doing the "perfect power" test is reasonably > the same as doing > > # q: > > then the first step of finding "perfect power" of the input > variable too slow. > > Supposedly the "perfect power" test can be done in linear > time. See http://cr.yp.to/papers/powers-ams.pdf
I would not design programs this way. Dan Bernstein's observations about computational speed are heavily focussed on assembly language assumptions, with C sometimes being an acceptable substitute. If I wanted to be replicating his work, I would be working in assembly language, or in C, or in something along that line. I would not be working in a high level language. Alternatively, since I prefer working with high level languages, I regard a lot of Dan's work as academically interesting but of little immediate personal relevance. Working in J, my first goal is to get something working properly. If I can achieve that, I am usually done -- I only delve into resource concerns if they prevented me from accomplishing my goals. (And, often enough, I am happy to refine my goals based on my available resources.) FYI, -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
