Thread: Measuring Run Time

Currently I'm studying from 'The C Programming Language' and it asks you to modify some code and then compare the run times. What is the best way to do this?

Thanks for the reply. Any idea how I could do it on os x? :P

Originally Posted by BIOS

Thanks for the reply. Any idea how I could do it on os x? :P

I'm sure they have equivalent functions... Google is your friend.

Originally Posted by Salem

Can be pretty accurate.
gettimeofday

Portable.
clock

Anything sub-second needs a bit of care to make sure it is done well.
The accuracy and precision of any fast clock you can get at can vary from one machine to another.

Heck they can vary within the same machine... depending what else is running. But as a rule they're good enough for non-critical functions.

Originally Posted by CommonTater

Google is your friend.

Had a feeling you'd say this! :P I did actually google first but figured asking would be quicker than trawling.

@salem Thanks. Clock seems perfect

Just in case anyone else digs up this thread in future. Here is some example code showing how to use clock():

Code:

#include <stdio.h>
#include <time.h>
#include <math.h>
 
int main()
{
  clock_t start = clock();
  for (long i = 0; i < 100000000; ++i)
    exp(log((double)i));
  clock_t finish = clock();
 
  printf("It took %d seconds to execute the for loop.\n",
  (finish - start) / CLOCKS_PER_SEC);
 
  return 0;
}

outputs:

Code:

It took 23 seconds to execute the for loop.

+1 for using the standard clock() function. Good job, BIOS.

If you're trying to measure something short, like a single statement, you should subtract off the
loop overhead, as it can be a significant part of the execution time.

Just run the same time measurement on the the same loop empty.

Then subtract that off the loop with the code under test in it.

And might as well convert the result to time per single pass also.

Originally Posted by quzah

If the loop is empty, your compiler may optimize it out, throwing the loop out completely.
Quzah.

Also... compared to the time for 100 million floating point calculations, the loop overhead is probably trivial...

Originally Posted by BIOS

Just in case anyone else digs up this thread in future. Here is some example code showing how to use clock():

Code:

#include <stdio.h>
#include <time.h>
#include <math.h>
 
int main()
{
  clock_t start = clock();
  for (long i = 0; i < 100000000; ++i)
    exp(log((double)i));
  clock_t finish = clock();
 
  printf("It took %d seconds to execute the for loop.\n",
  (finish - start) / CLOCKS_PER_SEC);
 
  return 0;
}

outputs:

Code:

It took 23 seconds to execute the for loop.

Next actually assign the result to something... a = exp(log((double) i)); ... Optimization may have run an empty loop...

Well it does depend on the code under test.

But I was also including the loop counter increment and test. That's also done 100 million times.