I didn't realize it was possible to enable enough of ACPI to power off,
while still avoiding latency issues. I assumed that ACPI = power management
= latency problems -- I ignored the "configuration" part of Advanced
Configuration and Power Interface. :-)
In the .config at git.linuxcnc.org / [infrastructure.git] / livecd / kernels
/ hardy / config-2.6.24-16-rtai, I see the line
# CONFIG_PM is not set
... so the Ubuntu kernel doesn't use ACPI and (I assume) can't power itself
off. I have no experience with creating packages, or I would offer to
create a .deb for Hardy Heron.
Alexey, thanks for that info. I'm gonna have to experiment some now. :-)
Mark
On Wed, Oct 21, 2009 at 6:43 PM, EBo <[email protected]> wrote:
>
> Alex,
>
> When you get a chance I would ***love*** to take a look at your .config's
> if
> you would not mind either posting them or emailing them directly. I have a
> machine that I have not been able to quiet the latencies down sufficiently
> to
> trust. From this info I see where I might have gone wrong...
>
> Thanks for the information!!!
>
> EBo --
>
> Alexey Starikovskiy <[email protected]> said:
>
> > Stephen Wille Padnos пишет:
> > Alexey Starikovskiy wrote:
> >
> >
> >> Mark,
> >> You are solving problem "backwords". You should just enable ACPI on your
> >> realtime box to make it shutdown properly.
> >> ACPI by itself does not worsen real time capabilities of your machine,
> >> it is power management that do.
> >>
> >>
> >>
> > It seems to on some machines. Though it may work well for you or me,
> > enabling ACPI in the kernels we distribute does reduce the number of PCs
> > that have acceptable latency. If there have
> >
> >
> Actually, by not enabling ACPI, you leave your machine in APM (legacy
> power management) mode,
> which require BIOS/SMI for any task.
> >> You should definitely have CPU idle framework and ACPI processor driver
> >> disabled. ACPI, ACPI button are perfectly fine.
> >>
> >>
> >>
> > I don't recall if it's only on the SMP kernels, but I know that some
> > parts of ACPI have been enabled with good results on some machines.
> > Part of the power control system is required to bring up the other cores
> > in dual- and quad-core CPUs.
> >
> You are speaking about ACPI MADT table, which lists all processors and
> interrupt controllers (APICs) for multiprocessor configurations. Without
> it Linux can't configure anything but boot CPU.
> Still, ACPI processor driver is not required for SMP.
> >
> >> If you care about latencies, you should also disable C1E support in BIOS
> >> (if you have such option).
> >> Saying this, I am having ~2-3msec average latencies on Celeron box with
> >> the above settings and worst case is 8msec.
> >>
> >>
> >>
> > Did you intend to say milliseconds (0.001 seconds) or microseconds
> > (0.000001 seconds) here?
> >
> >
> Right, microseconds (us) :)
> > If your'e talking about milliseconds, that is just about the worst
> > latency I've ever heard of :)
> >
> > Assuming it's microseconds, that's one of the best. What hardware
> > (motherboard, chipset, disk controller, ethernet controller, video card)
> > are you using on that system? Also, what did you do to enable ACPI on
> > your machine, and which parts did you enable? Did you build your own
> > kernel? (if so, can you make the .config available online?) Did you
> > use boot-time kernel options?
> >
> I have only 3 machines running, all Intel Celeron 430 based.
> Shuttle K45
> ASUS P5M-VM SE
> Gigabyte GA31M -- this one I would not recommend, it requires SMI
> disabling trick.
> Onboard graphics and parallel and ethernet ports. SATA 3.5" disks.
>
> 15us delay on modern CPUs comes from C1E enabled by default in BIOS.
>
> In kernel config I enable ACPI, Local APIC on single CPU (may not work
> for everybody), and ACPI buttons.
> No special kernel parameters.
>
> I can't easily post my .config, as machines are in production
> environment and not nearby.
>
> Regards,
> Alex.
>
>
> ------------------------------------------------------------------------------
> Come build with us! The BlackBerry(R) Developer Conference in SF, CA
> is the only developer event you need to attend this year. Jumpstart your
> developing skills, take BlackBerry mobile applications to market and stay
> ahead of the curve. Join us from November 9 - 12, 2009. Register now!
> http://p.sf.net/sfu/devconference
> _______________________________________________
> Emc-developers mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/emc-developers
>
>
>
>
> --
>
>
>
>
>
>
> ------------------------------------------------------------------------------
> Come build with us! The BlackBerry(R) Developer Conference in SF, CA
> is the only developer event you need to attend this year. Jumpstart your
> developing skills, take BlackBerry mobile applications to market and stay
> ahead of the curve. Join us from November 9 - 12, 2009. Register now!
> http://p.sf.net/sfu/devconference
> _______________________________________________
> Emc-developers mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/emc-developers
>
>
------------------------------------------------------------------------------
Come build with us! The BlackBerry(R) Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay
ahead of the curve. Join us from November 9 - 12, 2009. Register now!
http://p.sf.net/sfu/devconference
_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers
