Way to track the current time with good accuracy

Question

I was wondering how to track the current time in C to as much accuracy as possible. In particular, I was wondering if there was a function in C that mimics Python's time.time()?

I found the code below, but it rounds the seconds.

#include <time.h>
int main() {
  time_t seconds;
  seconds = time(NULL);
}

Oleg Utkin · Answer 1 · 2020-06-18T09:16:40.117

2

If you use POSIX-compatible system you may try clock_gettime function. It returns the current time in nanoseconds.

#include <time.h>
#include <stdint.h>
int main() {
    struct timespec ts;
    clock_gettime(CLOCK_REALTIME, &ts);
    int64_t time_ns = ts.tv_sec * 1000000000LL + ts.tv_nsec;
    return 0;
}

edited Jun 18 '20 at 09:16

answered Jun 17 '20 at 17:51

Oleg Utkin

171
1
3

2

`ts.tv_sec * 1000000000` can reallily overflow a [32-bit `time_t`](https://en.wikipedia.org/wiki/Year_2038_problem). Suggest `ts.tv_sec * 1000000000LL` or the like to insure at least a 64-bit multiplication. – chux - Reinstate Monica Jun 17 '20 at 19:02

score 0 · Answer 2 · answered Jun 17 '20 at 17:52

0

This should help:

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

int main()
{
    clock_t t1, t2;
    t1 = clock();
    float diff = (t1 / 1000000.0F );
    printf("%f",diff);

    return 0;
}

answered Jun 17 '20 at 17:52

Deepthought

118
9

1

Don't assume that `clock` has that resolution. Use `t1 / (float)CLOCKS_PER_SEC`. – Weather Vane Jun 17 '20 at 18:13
Deep Thought, `double` would make more sense here. After all, OP wants _good accuracy_, so no need to narrow what `clock_t` may provide. Also as `diff` might be small, consider `"%g"`. – chux - Reinstate Monica Jun 17 '20 at 19:07

score 0 · Answer 3 · answered Jun 17 '20 at 17:52

0

You can use the clock function from time.h to measure time with extreme precision.

It'll return the number of clocks elapsed since a reference frame (which is implementation defined but it's related to the program's execution).

#include <time.h>
clock_t now = clock();

Now if you want to turn this into an actually understandable time unit, you should divide the result of clock() by CLOCKS_PER_SEC, maybe even multiply it by 1000 to get the result in milliseconds (the precision should be preserved after dividing by CLOCKS_PER_SEC)

double secs = now / CLOCKS_PER_SEC; // Result in seconds
double msecs = secs * 1000; // Result in milliseconds

This will work on all systems.

More information on clock

answered Jun 17 '20 at 17:52

Chase

5,315
2
15
41

You need a cast `double secs = (double)now / CLOCKS_PER_SEC;` and also, on some systems you don't get millisecond resolution from `clock()`. – Weather Vane Jun 17 '20 at 18:11
@WeatherVane In practice, `(double)now` is OK, yet that could narrow the precision of `clock_t`. A more gentle alternative is `double secs = 1.0 * now / CLOCKS_PER_SEC;` to allow the division to occurs at best precision. – chux - Reinstate Monica Jun 17 '20 at 18:46
@chux-Reinstate Monica or `double secs = now / (double)CLOCKS_PER_SEC;` similar to another answer. Though `clock_t`'s precision isn't that great. – Weather Vane Jun 17 '20 at 18:48
@WeatherVane True in general. Yet `clock_t` and `time_t` are only specified as some scalar. It is just that casts can not only widen a type, sometimes they will narrow, so I try to avoid them unless no other good solution exists. IAC, your point about the weakness of `now / CLOCKS_PER_SEC;` remains valid. – chux - Reinstate Monica Jun 17 '20 at 18:59

Way to track the current time with good accuracy

3 Answers3