Hi! Currently, "timeit" often does not more than finding the "best of three". I suspect that computing a standard deviation from just three samples is not quite reliable. Hence, if one wants to get a statistic then one needs a lot more runs, thus, the test will take much more time.
Do we want that "timeit" will take much more time *by default*? Perhaps one could keep "timeit" as it is and introduce a new object "time_statistics" (say), that collects timings and may add additional data points, if the user so wish? For example: sage: s="some code to be tested" sage: TS = time_statistics(s, number=10000) sage: TS 10000 runs per sample 3 samples in total Mean: 451 ns Standard deviation: 103 ns The parameter "number" is supposed to have the same meaning (and the same default) as for timeit: It is the number of repetitions of the code string forming one data point. Since three samples (or whatever the default should be) is not enough, we wish to add samples: sage: TS.additional_samples(15) sage: TS 10000 runs per sample 18 samples in total Mean: 446 ns Standard deviation: 17 ns And of course TS has methods returning these (and other) statistical data. I guess that is similar to what William is suggesting for the "timeit" command; but I think "timeit" (for fast tests) and "time_statistics" (for detailed but more time consuming tests) should be separated. Cheers, Simon -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
