Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

From: Rolf vandeVaart (Rolf.Vandevaart_at_[hidden])
Date: 2007-10-05 16:08:35


Jeff Squyres wrote:
> On Oct 2, 2007, at 3:43 PM, Rolf.Vandevaart_at_[hidden] wrote:
>
>
>> I propose that we drop the mca parameter "btl_base_debug". It is
>> an added
>> level of complexity that does not provide any benefit. In fact, the
>> debug/verbose code in the btl_base_open.c is such that we can never
>> turn on verbose messages.
>>
>
> Heh -- I assume that this is an RFC, right? :-)
>
It seemed kind of small for an RFC, but maybe it is.
>
>> A quick research of the other base components in the OMPI area show
>> that
>> none of them have a *_base_debug mca parameter. They only have a
>> *_base_verbose parameter.
>>
>> I will plan on making this change but just thought I would check
>> and see if
>> there is some historical or other reason why btl_base_debug should
>> *not*
>> be removed.
>> (This came up as I was trying to figure out how to turn on
>> verbosity for
>> the MX problem reported on the users list. )
>>
>
> I don't really have an opinion either way, but I think you can turn
> on *some* debug messages by setting btl_base_verbose to (for example)
> 100. It looks like these messages are mainly in the BTL base
> selection logic and the openib BTL.
>
Jeff, can you verify that you can get something out with
btl_base_verbose? My experience is that
it has no effect at all. After writing the earlier message I realized
that btl_base_debug is the one
that affects what comes out. But as I said, this is different then all
the other frameworks and was
a source of confusion for me.

Rolf