Re: CVS commit: src/sys/kern

[Frank Kardel <kardel%netbsd.org@localhost>]

Hi *,

Izumi Tsutsui wrote:

> perry%piermont.com@localhost wrote:
>
>  
>
> > Yes. It tells you how many Hz the clock source is supposed to be. The
> > clock itself will certainly have some inaccuracy, but it tells you
> > exactly what the spec is.
> >    
>
> How is such Hz value taken?
>  
Either by definition (ACPU PM timer - some NTSC base frequency, or 
measurement).

> i.e. doesn't the measuring method (i8254_delay()?) have inaccuracy?
>  
Yes. Do you know an exact method when measuring ;-)?

>  
>
> > > timecounter: Timecounter "TSC" frequency 1300162690 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300166200 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300133700 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300126680 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300159180 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300128370 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300135390 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300171530 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300152420 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300130320 Hz quality 800
> > > timecounter: Timecounter "TSC" frequency 1300171140 Hz quality 800
> > >      
Just the cycles counted within a delay(9) period -  thus there is quite
some room  for variation .

> IMHO, it's better to show/use only significant digits defined
> by the quality (though I don't know how the quality is defined).
>  
Quality is just a number. Positive qualities allow for automatic
selection. Higher qualities get prefereed, on tie the higher frequency
wins. On same frequency the currently registered (last) is choosen.

> ---
> Izumi Tsutsui
>