[mpich-discuss] issue setting process-core mapping with hydra

[sreeram potluri]

Pavan,

The error output I had observed was because debug mode was enabled inside
hwloc. It appears that the hwloc debug mode is enabled by default when we do
an svn checkout. This does not happen if we do an export instead. I believe
you did an export and hence did not see the debug output.

Sreeram Potluri

On Wed, Nov 10, 2010 at 11:02 AM, sreeram potluri <
potluri at cse.ohio-state.edu> wrote:

> Pavan,
>
> This seems to be something to do with my environment. I tried a fresh
> check-out of MPICH2 but still see the same issue. I see the same error with
> any of the hwloc commands available from my MPICH2 build.
>
> However, none of the other folks here see it.
>
> I will try to figure out whats wrong.
>
> Thanks for your time and sorry for the trouble.
>
> Sreeram Potluri
>
>
> On Wed, Nov 10, 2010 at 9:11 AM, Pavan Balaji <balaji at mcs.anl.gov> wrote:
>
>>
>> Works fine for me.
>>
>>
>> /home/balaji/software/mpich2/mpich2-trunk-r7431
>> #wci-login# Wed Nov 10 09:10> ./bin/mpiexec -n 2 -binding user:0,1
>> ./examples/cpi
>> Process 0 of 2 is on wci37
>> Process 1 of 2 is on wci37
>>
>> pi is approximately 3.1415926544231318, Error is 0.0000000008333387
>> wall clock time = 0.000264
>>
>>
>> /home/balaji/software/mpich2/mpich2-trunk-r7431
>> #wci-login# Wed Nov 10 09:10> echo $SLURM_NODELIST
>> wci[37-38]
>>
>>
>> /home/balaji/software/mpich2/mpich2-trunk-r7431
>> #wci-login# Wed Nov 10 09:11> ./bin/mpiexec -n 16 -binding user:0,1
>> ./examples/cpi
>> Process 5 of 16 is on wci37
>> Process 6 of 16 is on wci37
>> Process 0 of 16 is on wci37
>> Process 1 of 16 is on wci37
>> Process 3 of 16 is on wci37
>> Process 4 of 16 is on wci37
>> Process 2 of 16 is on wci37
>> Process 7 of 16 is on wci37
>> Process 8 of 16 is on wci38
>> Process 9 of 16 is on wci38
>> Process 10 of 16 is on wci38
>> Process 11 of 16 is on wci38
>> Process 13 of 16 is on wci38
>> Process 14 of 16 is on wci38
>> Process 15 of 16 is on wci38
>> Process 12 of 16 is on wci38
>> pi is approximately 3.1415926544231274, Error is 0.0000000008333343
>> wall clock time = 0.002389
>>
>>  -- Pavan
>>
>>
>> On 11/10/2010 08:09 AM, Sayantan Sur wrote:
>>
>>> Hi Pavan,
>>>
>>> You should be able to 'salloc' them and use them.
>>>
>>> On Wed, Nov 10, 2010 at 9:05 AM, Pavan Balaji<balaji at mcs.anl.gov>
>>>  wrote:
>>>
>>>>
>>>> On 11/10/2010 08:02 AM, sreeram potluri wrote:
>>>>
>>>>>
>>>>> I am cc'ing Jonathan in this email. I think he should be able to get
>>>>> you
>>>>> access.
>>>>>
>>>>
>>>> I already have an account. I just need to know if I need to reserve
>>>> machines
>>>> in some way, or just salloc them and use them.
>>>>
>>>> Btw, it seems to work fine on the login node:
>>>>
>>>> /home/balaji/software/mpich2/mpich2-trunk-r7431
>>>> #wci-login# Wed Nov 10 09:01>  ./bin/mpiexec -f ~/hosts -n 2 -binding
>>>> user:0,1 ./examples/cpi
>>>> Process 0 of 2 is on wci-login.cse.ohio-state.edu
>>>> Process 1 of 2 is on wci-login.cse.ohio-state.edu
>>>> pi is approximately 3.1415926544231318, Error is 0.0000000008333387
>>>> wall clock time = 0.001009
>>>>
>>>> /home/balaji/software/mpich2/mpich2-trunk-r7431
>>>> #wci-login# Wed Nov 10 09:04>  cat ~/hosts
>>>> wci-login
>>>> 127.0.0.1
>>>>
>>>>  -- Pavan
>>>>
>>>> --
>>>> Pavan Balaji
>>>> http://www.mcs.anl.gov/~balaji
>>>>
>>>>
>>>>
>>>
>>>
>>>
>> --
>> Pavan Balaji
>> http://www.mcs.anl.gov/~balaji
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20101110/da5b16d6/attachment.htm>