Unruh wrote:

> Juergen Kosel <[email protected]> writes:
> 
>>Hello,
> 
>>Harlan Stenn schrieb:
>>>>>> In article <[email protected]>, Juergen Kosel
>>>>>> <[email protected]> writes:
>>> 
>>> And I also don't understand what you mean by "computing time is a
>>> concern". Is the overhead of a subroutine call that significant in your
>>> application?
> 
>>it would be a waste of computing time to convert the reference clock
>>time into the format guessed year, month day, hour, ... to call the
>>subroutine refclock_process(), which would convert it back.
> 
> Oh dear. The time taken is about 1 ns on amodern CPU. Instead you waste
> seconds and possibly hours drinking your coffee and chatting to friends
> while you r computer waits for you, and you are worried about ns in order
> to make an orderly API for the refclocks.
> 
> 
> 
> 
>>Greetings
>>Juergen

Uh-Oh...

Moore's law: the transistor density on chips roughly doubles every two
years.

Wirth's law: The software decelerates faster than the hardware accelerates.

So basically I'm trying to be conservative with my CPU ressources (and 15
years of embedded system programming *might* have to with it), but I think
the Juergen Kosel had something different in mind: inaccuracies because of
processing delays. But the point is that once you have the sample pair
(clock time and associated local time) you have nearly all the time in the
world to process it. It's the acquisition delay variance that kills, not
the processing that follows.
-- 
juergen 'pearly' perlinger
"It's hard to make new errors!"

_______________________________________________
questions mailing list
[email protected]
https://lists.ntp.org/mailman/listinfo/questions

Reply via email to