Question

Is it possible to measure the resolution of the std::clock() call? Or is this a problem where observing without influencing isn’t possible?

I wrote the following naive benchmark:

#include <ctime>
#include <iostream>

int main() {
    std::clock_t initial = std::clock();
    std::clock_t current;
    while (initial == (current = std::clock()));
    std::cout << "Initial: " << initial << std::endl;
    std::cout << "Current: " << current << std::endl;
    std::cout << "Precision: " << (static_cast<double>(current - initial) / CLOCKS_PER_SEC) << "s" << std::endl;
}

I’ve run it a few hundred times and it always outputs 0.01s.

My questions are:

• Is above code a good way to measure the resolution of clock()?
• If no then what is wrong with it? And how could it be improved?

Answer 1

You can, sort of. Something like what you’re doing is a good first
approximation. But I’m not sure how useful it is: it determines the
resolution, but it still doesn’t tell you anything about the accuracy;
under Windows, for example, clock is so inaccurate as to render it
useless.