count execution time using clock() in milliseconds

i have implemented an algorithm and the execution time is really small and when I use clock() I get a 0 as a result

is it possible to make clock() display results in milliseconds or even smaller like microseconds?

if yes, how is this possible? what parameters should I change?

all I do is

Code:

---


time_t before, after;

before = clock();

algorithm

after = clock();

time <- (double)(after-before)/CLOCKS_PER_SEC

---

i need to gather statistical data in seconds in order to see how my implementation reacts on different inputs

If you know the time in seconds, then surely the time in milliseconds is 1000 times that. Right?

Quote:

---

Originally Posted by brewbuck

If you know the time in seconds, then surely the time in milliseconds is 1000 times that. Right?

---

the problem is that i dont know the time in seconds, it always returns 0

Quote:

---

Originally Posted by nik

the problem is that i dont know the time in seconds, it always returns 0

---

Then the code you are trying to time is taking less than a single clock tick. That means clock() isn't going to be a good way of timing it.

On UNIX, you can use gettimeofday() to get better resolution. On Windows, you can use QueryPerformanceCounter().

There is no standard way to do this. You need to use the functions mentioned or an external library.

Let me assure you it is possible to get really accurate time measurements (even more than microsecond resolution). I used them many times at a job I was working over the summer. You need to research tsc clocks and high frequency timers.

Quote:

---

Originally Posted by Mozza314

Let me assure you it is possible to get really accurate time measurements (even more than microsecond resolution). I used them many times at a job I was working over the summer. You need to research tsc clocks and high frequency timers.

---

Not on most MS OS systems, except CE.

Most of the time you will be +/- 50 msec, getting anywhere near a msec requires a filter driver (or realtime OS).

Both the methods you mention are unreliable, especially on multi core / variable speed CPUs.