Will,
On Fri, Nov 10, 2006 at 01:58:37PM -0500, William Cohen wrote:
> Hua-Ying Ling wrote:
> >Hi Stephane,
> >
> >On 11/10/06, Stephane Eranian <[EMAIL PROTECTED]> wrote:
> >
> >>> [EMAIL PROTECTED]:~/testc/debug/src> dmesg|grep perfmon
> >>> Linux version 2.6.18-perfmon2-smp ([EMAIL PROTECTED]) (gcc version 4.1.0
> >>> (SUSE Linux)) #2 SMP Wed Oct 11 22:02:37 EDT 2006
> >>> perfmon: CPU0 APIC mask=0x100ee
> >>> perfmon: version 2.2
> >>> perfmon: added sampling format default
> >>> perfmon: cores/package=2 threads/core=0
> >>^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> >>> perfmon: HyperThreading supported, status off
> >>> perfmon: Data Save Area (DS) supported
> >>> perfmon: PEBS supported, status off (because of HT)
> >> ^^^^^^^^^^^^^^^^^^^^^^^^^^
> >>those 2 messages look strange to me. Make sure you have Hyperthreading
> >>OFF.
> >
> >
> >I thought that was strange too but the Pentium D 820 doesn't support
> >Hyperthreading... is it possible perfmon isn't detecting HT correctly?
> >
> >Thanks,
> >Hua-Ying
>
> What does the ouptut of /proc/cpu infor look like? is there an "ht" on the
> flags line? Does the number of processors agree with how the machine is
> configured.
>
The /proc/pcpuinfo ht flag does not tell you if you have hyperthreading on.
On a Pentium D, just like on a Woodcrest (both simply dual-core), the ht flag
is present. It denotes that fact that the processor is capable of exposing
more than one logical CPU. This was designed that way for licensing reasons,
if I recall.
> Could the test for HT be confusing multicore processors with hyperthreaded
> processors, e.g. see more than one processor in the package assume HT.
The latest code base does the following test:
ht_capable = (cpus_weight(cpu_core_map[smp_processor_id()]) /
cpu_data->x86_max_cores) > 1;
I don't recall what is in 060926.
--
-Stephane
_______________________________________________
perfmon mailing list
[email protected]
http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/
