Open MPI logo

Hardware Locality Users' Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Hardware Locality Users mailing list

Subject: Re: [hwloc-users] No caches or hiearchy on RHEL 4.7 or 4.8
From: Dan Eaton (daniel.eaton_at_[hidden])
Date: 2010-01-28 13:03:07


If I go through the API (which I would do anyway, I just wanted a "taste" with lstopo), can I choose to override information reported by the OS (i.e. go straight to CPUID), or do I still require the patch in this case?

Unfortunately that version of x86info requires glibc 2.4 -- RHEL 4.x only has 2.3.4.

Kind regards,
Dan

-----Original Message-----
From: hwloc-users-bounces_at_[hidden] [mailto:hwloc-users-bounces_at_[hidden]] On Behalf Of Samuel Thibault
Sent: Thursday, January 28, 2010 10:55 AM
To: Hardware locality user list
Subject: Re: [hwloc-users] No caches or hiearchy on RHEL 4.7 or 4.8

Dan Eaton, le Thu 28 Jan 2010 10:32:46 -0700, a écrit :
> I just meant without resorting to CPUID, because it is so unportable
> (to consider implementing directly into our library).

Ok. Unfortunately, when the information reported by the OS is bogus (as it seems to be the case here), there is no other way than just asking the hardware directly, or fixing the lstopo xml output and using it.

> I've attached the output of x86info.

Ah, that's an old x86info, could you try building a newer one? (Like Debian's 1.25 available on http://ftp.debian.org/debian/pool/main/x/x86info/x86info_1.25.orig.tar.gz
)

> Is it possible to get at your CPUID backend now?

The quick & dirty way is the attached patch, to be applied on top of the svn trunk, which forces x86 cpuid detection on Linux. Please pass --enable-debug to ./configure so we get precise debugging output from lstopo.

Samuel