Re: [Deinterlace-discuss] Problems fixed

[Atsushi N. <at...@ch...>]

Rob Muller <rob...@gm...> wrote:
> On a P2 350 MHz QueryPerformanceCounter() can be 
> executed about 200,000 times per second, this is not fast enough to 
> accurately measure a 50 kHz event. Some more testing reveals that the
> rdtsc instruction can be executed about 10 million times per second on
> the same machine.

Isn't it only necessary that the code in question measure out a delay of
*at least* 20 us, and not necessarily an exact 50 kHz event?  If there
isn't accuracy, isn't the worst that'll happen that operations following
will be performed slightly later (in microsecond terms) than necessary?

If this is the case, I'm not sure if there's much benefit (over possible
risks) in attaining better accuracy by using inline assembly.  In other
places where I've needed microsecond sleeps, I've just used Sleep() even
though it's not that accurate.

(Aside: What criteria determines whether 200,000 times per second is
enough?  Since it's four times 50 kHz, isn't it sufficient?)

Regards,


-- 
Atsushi Nakagawa
<at...@ch...>
Changes are made when there is inconvenience.