Timer in program is too slow - c

Question

Hello i am creating a speedcube timer i just got the time centered but then i noticed that the time on it was too slow, I tried changing the usleep function from 1000 but it was either to fast or to slow, any ideas?
#include <ncurses.h>
#include <stdlib.h>
#include <unistd.h>
int main()
{
int minutes = 0, milliseconds = 0, seconds = 0, x = 0, y = 0, text = 6, textminutes = 0, textseconds = 0, textmilliseconds = 0;
initscr();
while(1)
{
/*This block of code centers the text on the screen by incrementing each variable by one
for each number starting at ten, Then prints the time.*/
getmaxyx(stdscr,y,x);
if (seconds == 60 && minutes == 10){
textminutes += 1;
}
if (milliseconds == 1000 && seconds == 10){
textseconds += 0;
}
if (milliseconds == 10){
textmilliseconds += 1;
}
else if (milliseconds == 100)
{
textmilliseconds += 1;
}
else if(milliseconds == 1000)
{
textmilliseconds += 1;
}
int left_row = (x / 2) - (3 + textminutes + textseconds + textmilliseconds / 2);
mvprintw(y/2, left_row,"%d : %d : %d", minutes, seconds, milliseconds);
/*Sleep for 1 millisecond the increment the milliseconds
var i don't think that the timing is right though.
Then it refreshes and clears the screen to fetch the new contents.*/
usleep(1000);
milliseconds++;
if(milliseconds == 1000)
{
milliseconds = 0;
textmilliseconds -= 2;
seconds++;
if(seconds == 60)
{
seconds = 0;
textseconds -= 1;
minutes++;
}
}
refresh();
clear();
}
endwin();
return(0);
}

Answer 1

Your code seems to be written with the assumption that your process can run reliably once every millisecond, but you're probably running it on an operating system that has to perform other tasks, so you won't get to run every millisecond. Also, usleep might not be as accurate as you are hoping for it to be, and you're not accounting for the time it takes to do your calculations and output data to the terminal.
It's OK to use usleep to conserve CPU time. But when you want to figure out what time it is for displaying the time to the user, you should use a function that actually gets the time, like the C clock_gettime function perhaps.

Answer 2

You should never rely on an internal accumulator to accumulate elapsed wall time. Doing so not only captures the time consumed by usleep(), but also the execution time of formatting that string and any other calculations (the string format is likely to the big one).
Instead, get the system time at the start. Then, whenever you need to sample a new amount of time, get the system time again and subtract the start time from the now time. That'll give you the elapsed time and then you can adjust from there.