Loren on the Art of MATLAB

Turn ideas into MATLAB


Loren on the Art of MATLAB has been archived and will not be updated.

Nested Timing

Did you know that you can use tic and toc to measure times of operations that may have other timed operations nested inside?



Suppose I want to see how long some operations take, and the overall calculation as well. Let's see how long it takes to create some magic squares and how long it takes to find the rank for each one.

Nsizes = [10 20 50 100 200 500 1000];
N = numel(Nsizes);
magTimes = zeros(1,N);
rankTimes = zeros(1,N);
ranks = zeros(1,N);
timeTemp = tic;
for k = 1:N
    mTemp = tic;
    m = magic(Nsizes(k));
    magTimes(k) = toc(mTemp);
    rTemp = tic;
    ranks(k) = rank(m);
    rankTimes(k) = toc(rTemp);
allTimes = toc(timeTemp);

Compare Times

If I sum the times for creating the magic squares and computing their ranks, I should get the same time as the overall time calculation, right?

totTimes = sum(magTimes)+sum(rankTimes)
totTimes =
allTimes =

The totals don't match! That's because there's other stuff going on in the loop, including overhead for managing the loop.

Timing Tips

To get relative time estimates, you can use the profiler, tic/toc, or Steve's timeit function on the file exchange. Though I am not doing so in this post, you generally should time functionality written in functions, especially if getting the best performance is your goal.

What are Your Timing Patterns?

Do you find yourself timing nested calculations? Do you have other strategies? Please post them here.

Published with MATLAB® 7.7

  • print


To leave a comment, please click here to sign in to your MathWorks Account or create a new one.