Thanks Jeff. Mostly just checking i was not missing something.
I am doing some checking with local state for the moment. Will put #1 in
the list of "like to haves" for the time being.
Jeff Squyres wrote On 03/21/07 13:10,:
> On Mar 21, 2007, at 10:24 AM, Donald Kerr wrote:
>> Case: A btl parameter is given a default setting and at some later
>> point is potentially adjusted depending on the number of peers. Now
>> if the MPI user sets this same parameter I don't won't to overright
>> what they have set, or at least I would like to warn them before
>> doing so.
> Gotcha. Right now, there's no way to check this (i.e., what "level"
> an MCA param was set from). I see two options for going forward:
> 1. Add an API call to check what level a given MCA param is currently
> resolved from (probably return some kind of enum?).
> 2. Don't use the MCA param API but rather keep your own state (e.g.,
> a global with the default, another with the adjusted value derived
> from the number of peers, etc.) and do your own comparisons to see if
> the value was changed.
> #2 is workable today, but obviously more work for you. #1 is a
> better long-term solution. It wouldn't be too much work to do, but I
> unfortunately do not have the cycles to do it at the moment. I'd be
> happy to point you in the right direction, if you want...