Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] [EXTERNAL] Consequence of bind-to-core by default
From: Barrett, Brian W (bwbarre_at_[hidden])
Date: 2013-12-19 10:54:24


On 12/19/13 8:43 AM, "Ralph Castain" <rhc_at_[hidden]> wrote:

>
>On Dec 19, 2013, at 6:27 AM, Barrett, Brian W <bwbarre_at_[hidden]> wrote:
>
>> On 12/19/13 6:59 AM, "Jeff Squyres (jsquyres)" <jsquyres_at_[hidden]>
>>wrote:
>>
>>> 3. Finally, we're giving a warning saying:
>>>
>>> -----
>>> WARNING: a request was made to bind a process. While the system
>>> supports binding the process itself, at least one node does NOT
>>> support binding memory to the process location.
>>> -----
>>>
>>> For both #1 and #3, I wonder if we shouldn't be warning if no binding
>>>was
>>> explicitly stated (i.e., we're just using the defaults). Specifically,
>>> if no binding is specified:
>>>
>>> - if we oversubscribe, (possibly) warn about the performance loss of
>>> oversubscription, and don't bind
>>> - don't warn about lack of memory binding
>>
>> We have a couple machines where memory binding is failing for one reason
>> or another. If we're binding by default, we really shouldn't throw
>>error
>> messages about not being able to bind memory. It's REALLY annoying.
>
>Just to help me understand a bit better - you are saying that the node
>supports process binding, but not memory binding? I don't see how the
>error appears otherwise, but want to ensure I understand the code path.

That appears to be the case, yes.

Brian

--
  Brian W. Barrett
  Scalable System Software Group
  Sandia National Laboratories