clock() problem

[Nick_3596]

Hi all

I've been using Red Hat Linux gcc to compile my C program.
I've been trying to measure my function time by using <time.h>


:insert

#include <time.h>
int main()
{
clock_t t0,t1;
while(1)
{ t0=clock();
...
...
add(a,b)
...
t1=clock()
printf("%ld,%ld\n",t0,t1);
}
}

Output:

0,0
0,0
...
...
10000,10000
10000,10000
...
...
20000,20000
20000,20000...

It almost looks as if it's been rounded up to 5th digit.
This result stays doesn't matter how far apart i put the two clock() points.

Could you give me a suggestion on what I might be doing wrong?

Thank you very much.

[Dino]

See this tutorial: http://faq.cprogramming.com/cgi-bin/...&id=1043284392

Todd

[Nick_3596]

Thanks for your reply

However, when i tried the following code from the tutorial, the output was kept as 0s no matter how long i wait.

Anyidea?

I'm using "Linux p21.ecf 2.6.9-55.0.2.ELsmp #1 SMP Tue Jun 12 17:59:08 EDT 2007 i686 i686 i386 GNU/Linux"

Thanks

:insert

#include <stdio.h> 
#include <time.h> 

int main ( void )
{
  clock_t start, end;

  /* Start the timer */
  start = clock();

  printf ( "Please wait a few moments and hit return\n" );
  getchar();

  /* End the timer */
  end = clock();

  /* Print out the difference */
  printf ( "The interval was: %f seconds\n",
    (double)( end - start ) / (double)CLOCKS_PER_SEC );

  return 0;
}

[swoopy]

It works for me under windows. I get anything from fractions of a second to 3 seconds, depending on how long I wait. I wonder if clock() is measuring the process time used by the program, not run time (it seems like I've seen this referred to as wall time). To check if this is the case, you could print the value of start and end, as well as CLOCKS_PER_SEC:

printf("%ld,%ld %ld\n", start, end, (long) CLOCKS_PER_SEC);

And see if the difference between start and end is less than CLOCKS_PER_SEC.

[Salem]

Clock() measures CPU time used by the process, not user time spent waiting for the user to type something in.

[matsp]

In Linux, you will definitely get the time spent in the process running, rather than "wall clock" time. In Windows, I beleive it measures "wall time". I believe there is a way to find the processing time in Windows too.

And yes, the time given by clock is not necessarily so precise that all digits are "used", it may well be a precision of microseconds, but measured in increments of milliseconds or 10 milliseconds - this gives the choice of giving more fine-grained precision later on [e.g. if you configure a different clock-device] without changing the compile-time constant "CLOCKS_PER_SEC". This applies in both Windows and Linux.

--
Mats