On Mon, 2006-05-08 at 12:09 -0600, Meswani, Mitesh wrote:
>
> Hello
Hi
>
> I am looking to use the SMT related parameters like Snooze delay, HMT
> thread priorities, SMT ON/Off and wanted to know how to invoke and set
> them. I am running Open Suse 10 with 2.6.16.rc4-3-ppc64 kernel on
> eServer p590 2-way POWER5 partition.
>
> I noticed the parameters in /sys/devices/system/cpu/cpu# with the
> following :
>
> mmcr0 online pmc2 pmc5 smt_snooze_delay
> mmcr1 physical_id pmc3 pmc6 topology
> crash_notes mmcra pmc1 pmc4 purr
>
Some of this is unique to Logical Partitions and cpu's on POWER5 pSeries
hardware.. I think you've already found the useful entry "online",
and the rest would be just trivial information.
On the kernel boot commandline, you can add parms such as
"smt-snooze-delay=xxxx" to set the snooze_delay value, and
"smt-enabled=off" to turn off the secondary threads.
As you've already found, you can echo values into the sys entries to
cause the cpus to go online or offline. 0=offline, 1=online. You can
also adjust the value for snooze-delay. This one defaults to '0'. This
controls the amount of time the processor thread spins before declaring
that it's got no useful work to do and cedes itself.
"mmcr*" and "pmc*" are performance counter registers and values. These
are used by oprofile.
"purr" is a Processor Utilization Resource Register, indicating the
number of ticks that the processor has been in use.
I believe "crash_notes" has to do with lkcd or crashdump.
No idea on the "topology" entry. possibly related to NUMA.
> setting the value in online to 0 seems to turn off the logical
> processor, but I am not sure what the others are for and the meaning
> of their hex values?
> It seems that there is include/asm-ppc64/processor.h with macros like
> HMT_very_low() , wonder if these can be set on command line
> since I am running unmodified app binaries.
the HMT_* macros are telling firmware that "this processor thread should
run at this priority". Typically used when we're waiting on a spinlock.
I.e. When we are waiting on a spinlock, we hit the HMT_low macro to drop
our threads priority, allowing the other thread to use those extra
cycles finish it's stuff quicker, and maybe even release the lock we're
waiting for. HMT_* is all within the kernel though, no exposure
to userspace apps.
>
> Thanks,
> Mitesh
>
>
Hope that is helpful..
-Will
> Mitesh R. Meswani
> Ph.D. Candidate
> Research Associate, PLS2 Group
> Room 106 F, Department of Computer Science
> The University of Texas at El Paso,
> El Paso, Texas 79968
> Tel: 915 747 8012 (O)
> Email: [email protected]
>
> ______________________________________________________________________
> From: [email protected] on behalf of
> Geoff Levand
> Sent: Fri 5/5/2006 6:00 PM
> To: Paul Mackerras
> Cc: Arnd Bergmann; Levand,Geoffrey; [email protected];
> [email protected]; Arnd Bergmann; [email protected]
> Subject: Re: [PATCH 04/13] cell: remove broken __setup_cpu_be function
>
<Snippage...>
> _______________________________________________
> Linuxppc-dev mailing list
> [email protected]
> https://ozlabs.org/mailman/listinfo/linuxppc-dev
> the HMT_* macros are telling firmware that "this processor thread
> should
> run at this priority". Typically used when we're waiting on a
> spinlock.
> I.e. When we are waiting on a spinlock, we hit the HMT_low macro to
> drop
> our threads priority, allowing the other thread to use those extra
> cycles finish it's stuff quicker, and maybe even release the lock
> we're
> waiting for. HMT_* is all within the kernel though, no
> exposure
> to userspace apps.
Actually, those macros translate straight into a single machine insn.
No firmware is involved. See include/asm-powerpc/processor.h. For
example:
#define HMT_very_low() asm volatile("or 31,31,31 # very low
priority")
You can use those same macros from user space, although it is CPU
implementation dependent which priorities you can actually set (you
probably can do low and medium priority).
Segher