Open MPI logo

Hardware Locality Development Mailing List Archives

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

Subject: Re: [hwloc-devel] mpich2 maintained patches for hwloc
From: Brice Goglin (Brice.Goglin_at_[hidden])
Date: 2012-09-06 01:22:03

No, we didn't apply anything to v1.5. In trunk, we're using sysconf
instead of getpagesize when it supports SC_PAGESIZE or SC_PAGE_SIZE
(very often).


Le 06/09/2012 05:33, Pavan Balaji a écrit :
> Brice,
> Sorry, I didn't get to try this patch out. Did it end up in any of
> the hwloc releases?
> -- Pavan
> On 07/24/2012 02:45 PM, Brice Goglin wrote:
>> Le 24/07/2012 13:39, Brice Goglin a écrit :
>>> Le 23/07/2012 18:40, Pavan Balaji a écrit :
>>>>>> 3. The changes to config/hwloc.m4 and include/private/private.h are
>>>>>> essentially a warning squash when getpagesize() requires an explicit
>>>>>> prototype declaration. But it's not clear how easy it is for you to
>>>>>> absorb it as it uses an MPICH2 internal m4 macro. Maybe there's a
>>>>>> cleaner way to integrate this patch.
>>>>> I will look at this. On which platform/system did you need this
>>>>> patch?
>>>> On regular x86_64 machines. We use strict builds in mpich2. I can
>>>> figure out the exact flags that trigger it, but my guess is that -Wall
>>>> would.
>>> Looking deeper into this, it looks like the right solution is to use
>>> sysconf(_SC_PAGESIZE) instead of getpagesize(). getpagesize() looks
>>> deprecated, that's why it's getting disabled when modern features are
>>> enabled. I'll send a patch to use sysconf when supported and
>>> fallback to
>>> the old getpagesize() otherwise.
>> Can you try this patch of some of your platforms?
>> I think I couldn't reproduce your problem because hwloc's
>> which defines __USE_BSD, which makes getpagesize() available without
>> looking at other XOPEN/POSIX defines.
>> Brice
>> Brice