File Exchange Pick of the Week

February 12th, 2008

It was about a year and a half ago when I posted about using the profiler to find bottlenecks in your MATLAB code [click here]. This week’s video covers two quick and easy ways of timing your code in MATLAB. This is how people would profile their code before the profiler, but it is still a useful set of tools to have.

In response to the survey, I have obtained new audio recording equipment. Let’s hope you like the results.

Video Content

10:03 UTC | Posted in MATLAB Basics, PodCast, Video | Permalink |

You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.

2 Responses to “Timing code in MATLAB”

1. Tim Davis replied on February 15th, 2008 at 07:16 UTC :
   

   Sometimes you want to time code that runs for a teeny fraction of a second, and

   tic ;
   dosomething ;
   t = toc

   is not accurate. The profiler will also be inaccurate. (You might want to time this teeny fraction of a second becaus you want to use the dosomething function zillions of times inside a larger application). Here is a trick I like to use

   tic ;
   t = 0 ;
   ntrials = 0 ;
   while (t < 1)
   dosomething ;
   ntrials = ntrials + 1 ;
   t = toc ;
   end
   t = t / ntrials ;

   That way, you run “dosomething” enough times to accumulate a whole second, and then you compute the average time per call to dosomething.

2. Doug replied on February 15th, 2008 at 08:42 UTC :
   

   Tim, an excellent suggestion. I would want to make sure that the time for dosomething >> the rest of the code in loop. For most applications it would be.

   Thanks,
   Doug

Leave a Reply

<pre class="code">
a = magic(3);
sum(a)
</pre>

These postings are the author's and don't necessarily represent the opinions of The MathWorks.