Re: [RFC] timer-changes_A0

[Alan Cox]

On Wed, 2002-08-14 at 00:58, john stultz wrote:
> where you have clock_tsc.c, clock_pit.c, clock_cyclone.c etc. Each one
> implementing init_clock_xxxx, mark_timeoffset_xxxx, and
> do_gettimeoffset_xxxx (possibly probing functions too?).  But for now I
> just kept it all in time.c 

Longer term a struct timer_ops would indeed be really clean. Splitting
it makes sense

>  }
>  
> +enum clock_type {CLOCK_PIT, CLOCK_TSC};
> +enum clock_type select_clock(void)
> +{
> +	if(cpu_has_tsc)
> +		return CLOCK_TSC;
> +	return CLOCK_PIT;
> +}

I put this in setup.c in my case so I could also cover the VISWS more
nicely, and also because of the 2.5 plans to do a nice architecture
split

Pulling the other code out of the switch into functions is definitely
good. I'll go apply that bit now