How to check the rapidity of execution of a method during runtime ?

[yellowmat]

Hi everybody.

I have an application with some function, some of them are identical but coded differently. I want to test each of these identical functions in order to know which one is the fastest during runtime.

If I add some specific code in my applycation for the test, it will probably interfere with the results no ? Does someone know a way to process such tests, what should I take care during the developpment of the test architecture ?

Thanks in advance.

[Methedrine]

You are looking for a way to profile your code. Take a look at, for example gprof (which comes with gcc).

[yellowmat]

I am using VC++ 6.0 so it won't be possible to use gprof. Can you tell me more about code profiling ?

What I want to know is the time spended in some function in ordre to try, test and validate some kind of implementation for many similar functions.

[wysota]

Then use a profiler suited for MSVC There probably is one. You also probably have to pay for it, but I guess you're ready for that if you decided to use MSVC.

Of course you can make pseudo-tests using your code directly. Just use some way to measure time differences (for example using time() or clock()).

[Methedrine]

At work we used to use intel's v-tune for profiling purposes with visual c++ applications.

However, if you are only interested in how long a certain function runs you might as well just take the start and end time and substract start from end to get the runtime, for instance like this:

#include <iostream>
#include <ctime>
 
using namespace std;
 
void myFunction()
{
  int j = 0;
 
  for (int i = 0; i < 1000000; ++i)
  {
    ++j;
  }
}
 
void main(int argc, char** argv)
{
  // start benchmarking
  double truntime=0.0, tstart;
  tstart = clock();
 
  myFunction();
 
  // end benchmarking
  truntime += clock() - tstart;
 
  // rescale to seconds
  truntime /= CLOCKS_PER_SEC;
 
  cout << "Runtime was: " << truntime << " seconds." << endl;
 
  return 0;
}

#include <iostream>
#include <ctime>

using namespace std;

void myFunction()
{
  int j = 0;

  for (int i = 0; i < 1000000; ++i)
  {
    ++j;
  }
}

void main(int argc, char** argv)
{
  // start benchmarking
  double truntime=0.0, tstart;
  tstart = clock();

  myFunction();

  // end benchmarking
  truntime += clock() - tstart;

  // rescale to seconds
  truntime /= CLOCKS_PER_SEC;

  cout << "Runtime was: " << truntime << " seconds." << endl;

  return 0;
}

To copy to clipboard, switch view to plain text mode 

edit: wysota beat me

[yellowmat]

Ok, I know now what profiling is and found were to enable this feature in MSVC. I think it will be enough for my need.

Thanks to all.