Open MPI logo

Hardware Locality Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |  

This web mail archive is frozen.

This page is part of a frozen web archive of this mailing list.

You can still navigate around this archive, but know that no new mails have been added to it since July of 2016.

Click here to be taken to the new web archives of this list; it includes all the mails that are in this frozen archive plus all new mails that have been sent to the list since it was migrated to the new archives.

Subject: Re: [hwloc-devel] hwloc 1.1 rc2 make check fails on SLES10SP1 on PPC64
From: Christopher Samuel (samuel_at_[hidden])
Date: 2010-11-20 12:43:35

Hash: SHA1


On 21/11/10 03:42, Brice Goglin wrote:

> Thanks. The tarball looks good. It seems to report everything as
> expected. By the way, it's funny to see a L4 cache without any L3,
> there's likely something to fix in the device-tree code, but it should
> not be very important.


> So, it looks like running lstopo on the tarball doesn't work fine. It
> doesn't see any caches (which may be related to /proc/device-tree/ too).

Yeah, it's also not showing the SMT's either.

> Could you extract the tarball under /tmp and run
> "HWLOC_FSROOT=/tmp/sles10sp1-ppc64 lstopo" ? I assume you will not see
> any caches.

Correct - this is what it looks like:

Machine (15GB)
  NUMANode #0 (phys=0 7680MB)
    PU #0 (phys=0)
    PU #1 (phys=1)
    PU #2 (phys=2)
    PU #3 (phys=3)
  NUMANode #1 (phys=1 8000MB)
    PU #4 (phys=4)
    PU #5 (phys=5)
    PU #6 (phys=6)
    PU #7 (phys=7)

This is what it's meant to look like:

Machine (15GB)
  NUMANode #0 (phys=0 7680MB) + L4 #0 (32MB)
    L2 #0 (4096KB) + L1 #0 (64KB) + Core #0
      PU #0 (phys=0)
      PU #1 (phys=1)
    L2 #1 (4096KB) + L1 #1 (64KB) + Core #1
      PU #2 (phys=2)
      PU #3 (phys=3)
  NUMANode #1 (phys=1 8000MB) + L4 #1 (32MB)
    L2 #2 (4096KB) + L1 #2 (64KB) + Core #2
      PU #4 (phys=4)
      PU #5 (phys=5)
    L2 #3 (4096KB) + L1 #3 (64KB) + Core #3
      PU #6 (phys=6)
      PU #7 (phys=7)

> If so, first check that /tmp/sles10sp1-ppc64/proc/device-tree/
> seems to contain what's also in /proc/device-tree/ . Then
> reconfigure with --enable-debug and send the whole stdout/stderr
> of "HWLOC_FSROOT=/tmp/sles10sp1-ppc64 lstopo"

OK - have to go and check out now so no more Internet access until
I get to LA tonight, so that will need to wait I'm afraid! :-(

> If you can get me access to this machine, it would help a lot too :)

I'll see what I can do when I'm back in Melbourne. :-)

- --
 Christopher Samuel - Senior Systems Administrator
 VLSCI - Victorian Life Sciences Computational Initiative
 Email: samuel_at_[hidden] Phone: +61 (0)3 903 55545

Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla -