SMP status?

SMP status?

[John Clemens]

Out of curiousity, has anyone thought about what it would take get SMP
support into cpufreq?  I'm thinking about servers...

Is someone working on this, or has any work been don on the past about
this? I can imagine some of the timing issues, etc.. and some of them
should be addressed by the work that went on some time ago to allow
processor of different speeds to work in SMP (i guess they don't use the
TSC for clocks?).

how big of a rats nest is this, as bad as i think it is?

just curious :)
john.c

-- 
John Clemens          http://www.deater.net/john
john@deater.net     ICQ: 7175925, IM: PianoManO8
      "I Hate Quotes" -- Samuel L. Clemens

Re: SMP status?

[Dominik Brodowski]

On Mon, Jul 21, 2003 at 01:06:05AM -0400, John Clemens wrote:
> 
> Out of curiousity, has anyone thought about what it would take get SMP
> support into cpufreq?  I'm thinking about servers...

It's already done. The generic cpufreq core is perfectly SMP-capable, the 
x86 timing updates [in arch/i386/kernel/timers/timer_tsc.c] is 
SMP-capable, the /proc/sys/, /proc and /sys/devices interfaces are
SMP-capable. However, it's more a question of which CPUs support frequency
and voltage scaling. There's no "speedstep" for server-grade processors yet,
though AMD is reportedly thinking of including PowerNow in future releases
of their AMD64 processors. So, on x86 only the throttling-driver 
"p4-clockmod" does provide SMP [and Hyperthreading!] support.

On other architectures cpufreq + SMP is more widely available, e.g. on
sparc64.

> how big of a rats nest is this, as bad as i think it is?

It seems to work :-)

	Dominik