is the resolution of do_gettimeofday in usec?

is the resolution of do_gettimeofday in usec?

[wael showair]

Hi All,
i have board that contains MPC8555 processor with linux 2.6.27 ported to it.
i want to use an accurate function to measure the time. i searched the
kernel code & i found several functions but i read that the do_gettimeofday
is the most accurate one since it has a timer resolution of usec.

my question is how this function give this accuracy while the kernel timer
is 4msec?
is this a real accuracy or does it convert from msec to usec?
or does it depend on another timer counter that has this resolution?
or may b it measure the clock cycles of the processor within certain
interval?

Note that the freq of the processor is 850MHz.

Unfortunately, the function is implemented in assembly & i cant understand
it?
so can u tell me how this accurate numbers can be got or they r just fake
numbers?

thanks
-- 
View this message in context: http://www.nabble.com/is-the-resolution-of-do_gettimeofday-in-usec--tp23807779p23807779.html
Sent from the linuxppc-dev mailing list archive at Nabble.com.

Re: is the resolution of do_gettimeofday in usec?

[Chris Friesen]

wael showair wrote:
> Hi All,
> i have board that contains MPC8555 processor with linux 2.6.27 ported to it.
> i want to use an accurate function to measure the time. i searched the
> kernel code & i found several functions but i read that the do_gettimeofday
> is the most accurate one since it has a timer resolution of usec.

You might also look at clock_gettime(), which has a theoretical accuracy
of nanoseconds.  Both this and gettimeofday are NTP-corrected if you are
synced to an NTP clock.

> my question is how this function give this accuracy while the kernel timer
> is 4msec?

The accuracy of the timestamp is not related to the tick interval.

Chris