Scott Cheloha <[email protected]> wrote:

> On Sat, Dec 10, 2022 at 09:33:55AM -0700, Theo de Raadt wrote:
> > Mark Kettenis <[email protected]> wrote:
> > 
> > > So really the question is whether we should register the "tick"
> > > timecounter if we also provide the "stick" or "sys_tick" timecounter.
> > > 
> > > Personaly I think the code we currently have is fine.  If you make the
> > > sconscious decision to use "tick" on a system that provides "stick" or
> > > "sys_tick" you'd better make the conscious decision not to use the CPU
> > > frequency scaling stuff.
> > 
> > Right.  Is it really any different than choosing a clock with known or 
> > unknown
> > quirks on an x86 machine:
> > 
> > kern.timecounter.choice=i8254(0) tsc(2000) acpihpet0(1000) acpitimer0(1000)
> 
> I think this situation is different from the one on i386/amd64.
> 
> When we offer a timecounter to the end user, we're saying (1) this
> clock is monotonic, and (2) this clock counts at a fixed frequency.
> The quality number then gives you a hint about which clock is best.
> 
> So, yes, some of those x86 clocks are slow compared with the TSC, but
> they still count time correctly.

Really.  Are you willing to put money on that?  That on x86, none of the
clocks are variant anymore, especially after a resume?  How much money.
Anyone else want to put money on it?  We'd be talking a 1 year long bet to
discover a mistake...

Reply via email to