Re: clock and timing.

[Der Herr Hofrat <der.herr@hofr.at>]

> 
> This has nothing to do with gcc. On many systems clock()'s resolution
> is only 1/100 of a second. This means that clock() is incremented
> every  10,000/CLOCKS_PER_SEC  seconds.
>
> 
> 
> "Tom Davids" <opentest@angelfire.com>@gcc.gnu.org on 22-10-2001 06:14:31
> 
> > I've just coded up this small program
> 
> > [ deleted a program that uses clock() ]
> >
> > The CLOCKS_PER_SEC is 1000000,
> > When I run the program I get the either
> > total = 10000
> > total = 20000
> > or
> > total = 0.
>
you are seeing the "ticks" of the OS not of the hardware - if you need a high
resolution clock then you need to directly read the hardware clock - for X86
on a Linux system this could be done by 


----snip---
#include <stdio.h>
#include <asm/io.h>

__inline__ unsigned long long int hwtime(void)
{
	unsigned long long int x;
	__asm__("rdtsc\n\t"
		"mov %%edx, %%ecx\n\t"
		:"=A" (x));
	return x;
}

main(){
	long long int hwtime1,hwtime2,jitt,jitt_max;

	jitt_max=0;
	while(1){
		hwtime1 = hwtime();
		hwtime2 = hwtime();
		jitt=hwtime2-hwtime1;
		if(jitt > jitt_max){
			jitt_max=jitt;
			printf("got %lx\n",jitt);
			}
		sleep(1);
		}
	return;
}

this would show you the "hardware-resolution" of the system if you let it run
for a while. On a normal linux system you will see worst-case jitter of up to
1 second (PIII/800 Running a 2.4.4 kernel).

hofrat