> Usually the time spent in the random number generator is a very small > portion of a real task. It is only when the real task is empty, like when > testing randon numbers, that the speed of the randomization is of much > concern. The difference between small and a bit bigger matters when > compared to nothing but is of no concern when compared to something real > that is much larger. Usually much faster is reserved for things like > 1000*1000 compared to 1000*log(1000)=1000*10. I believe that Python is > an interpreted langauge so is noticably slower that a compiled language > which one would use for a real task where timeing was important.
Yes, Python is normally interpreted, but IMHO an interpreted run could in some measure reflect the results in optimized runs in other languages (though this of course must remain speculative without hard facts). You are certainly right that my use of "much faster" is highly problematic. (In fact for common programmers a 50% speed increase of his codes may be something, but for persons in theoretical computer science generally only exponential time improvements of algorithms would be of significance.)