On Tue, Aug 12, 2008 at 23:50, Jeff Squyres <jsquyres_at_[hidden]> wrote:
> On Aug 8, 2008, at 5:35 PM, Jeff Squyres (jsquyres) wrote:
>
>> --> We know that linux processor ID's can have holes. I am pretty
>> sure that socket IDs can have holes, too. I'm not sure if core IDs
>> can have holes, but I guess it's conceivable that in the manycore
>> world, you could have a socket where 1 core is dead and other cores
>> are fine...?
>>
>
>
> To answer my own question, yes, core IDs *can* have holes -- one case is if
> you manually take a processor offline. E.g., if you have a 4 core socket
> and take core 1 offline (e.g., "echo 0 >
> /sys/devices/system/cpu/cpu1/online"), cores 0, 2, and 3 are still reported
> by the kernel.
>
> The bummer is that when you manually take a processor offline, the kernel no
> longer reports that processor's topology information (@#$@#$), so PLPA has
> no idea what (socket_id,core_id) it belongs to.
It can't. In x86, it needs to execute the CPUID instruction on this
cpu, and if you offline this, its stuck. Other archs, like power have
not this problem.
Bert
>
> --
> Jeff Squyres
> Cisco Systems
>
> _______________________________________________
> plpa-users mailing list
> plpa-users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/plpa-users
>
|
![[Flag]](../../../../../community/mirrors/flag-us.gif){kind=small}