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] libpci: GPL
From: Brice Goglin (Brice.Goglin_at_[hidden])
Date: 2013-02-06 10:02:07

Le 06/02/2013 15:11, Brice Goglin a écrit :
> Le 06/02/2013 15:06, Jeff Squyres (jsquyres) a écrit :
>> On Feb 5, 2013, at 10:11 PM, Brice Goglin <Brice.Goglin_at_[hidden]> wrote:
>>> It's a Xorg dependency, so it's available on many platforms and
>>> prepackaged in many distributions, including darwin if I remember correctly.
>> I'm thinking that we should actually release a new version for all hwloc versions that defaulted to use libpci -- that's 1.3.x and beyond. :-\
>> I'll need to update the version that's bundled all versions of OMPI (1.6.x., 1.7.x, trunk), for example.
>> I'll volunteer to do the configury stuff. I see Brice just made a svn branch; do you want me to commit the updates there? Or did you want to just do it all?
> I am already doing it. Will push things within an hour.
>> Are we removing the possibility of using libpci?
> Not yet. Samuel's nvidia card isn't properly recognized, and I haven't
> tested on many machines yet.

Just pushed Samuel's patch + the configury + one fix for object names.
The code currently uses libpciaccess by default. pciutils is only used
when libpciaccess isn't available. I am not sure yet if we should add a
--disable-gpl or --enable-gpl, or just remove libpci entirely.

lstopo --whole-io exports the exact same XML on the all our local
machines (from Clovertown to E5, one AMD Bulldozer, with IB, 10G and
AMD/NVIDIA GPUs, from Dell, IBM and Supermicro, and one Qemu VM). In
short, it looks pretty good, but I'd like to understand Samuel's problem.