Can you measure the performance of a function that calls multiple functions within using 'timeit'?

I want to use timeit (or tic-toc) to measure the performance of a function. I am calling this function within a loop to measure the timing information multiple times and then averaging. When I run this function once, suppose the timing is measured to be m1±s1 (mean±standard deviation) across N runs (in the loop that I mentioned). The standard deviation is not very high, quite acceptable. However, if I execute this loop again, say in an hour, it becomes m2±s2, the mean changes to m2 which is quite different from m1. s2 however remains small. Can I know how to use timeit properly? What factors does it depend on?

P.S. I am NOT doing the following within the functions (as mentioned in the timeit documentation)

• Using timeit between tic and toc
• Using timeit to time a function that includes calls to tic and toc
• Using timeit recursively

[Rik]

Without a description of the function you're measuring I don't have an idea. The timeit function should take care of things like clear all being called in between runs of your performance measurement.

[shreya]

Hi, thanks for your response. The function (f1) iteratively refines an initial estimate. In doing so it calls another function f2 (and this function calls two other functions f3 and f4) that iterates in a loop. There are no 'clear all' commands anywhere. I am wondering whether I can measure the performance of f1 using 'timeit' in this case.

[Rik]

I don't think that should matter, as long as you run the function with the same input.

If your function contacts some server, there will be a dependency on your internet connection, but I don't see any such external factors in your description.

Also, your question is not urgent, the reason for that is explained here. I therefore removed that part from your question title.

[shreya]

Thanks a lot, Rik.

Sure, sorry for using that adjective.