Messages in this thread | | | Date | Tue, 5 Mar 2002 20:31:16 -0800 (PST) | From | Davide Libenzi <> | Subject | Re: a faster way to gettimeofday? |
| |
On Tue, 5 Mar 2002, Ben Greear wrote:
> > > Davide Libenzi wrote: > > > On Tue, 5 Mar 2002, Ben Greear wrote: > > > > > >>I have a program that I very often need to calculate the current > >>time, with milisecond accuracy. I've been using gettimeofday(), > >>but gprof shows it's taking a significant (10% or so) amount of > >>time. Is there a faster (and perhaps less portable?) way to get > >>the time information on x86? My program runs as root, so should > >>have any permissions it needs to use some backdoor hack if that > >>helps! > >> > > > > If you're on x86 you can use collect rdtsc samples and convert them to ms. > > You'll get even more then ms accuracy. > > > Can I do this from user space? If so, any examples or docs > you can point me to? > > Also, I'm looking primarily for a speed increase, not an accuracy > increase.
#include <linux/timex.h>
unsigned long long mscurr; cycles_t cys, cye, mscycles; struct timespec ts1, ts2;
clock_gettime(CLOCK_REALTIME, &ts1); cys = get_cycles(); sleep(1); clock_gettime(CLOCK_REALTIME, &ts2); cye = get_cycles(); mscycles = (cye - cys) / ((ts2.tv_sec - ts1.tv_sec) * 1000 + (ts2.tv_nsec - ts1.tv_nsec) / 1000000);
mscurr = ts2.tv_sec * 1000 + ts2.tv_nsec * 1000000 + (get_cycles() - cye) / mscycles;
- Davide
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |