[mpiwg-sessions] [EXTERNAL] Sessions/Spawn

[Pritchard Jr., Howard]

HI Wesley,

We don't think this is a show stopper for the MPI 4.0.  For 4.1 we need to add some clarifying text to indicate that there are certain limitations to using the dynamic process model and sessions.
I'll open an issue to track this for 4.1.

Howard


On 2/3/21, 8:32 AM, "Martin Schulz" <schulzm at in.tum.de> wrote:

    Sure, that is an option, of course, but I don't think it is that critical and I think Howard sees it the same way (please speak up if you disagree). However, if MPI should slip for other reasons, I think we should add a fix.
    
    Just my 2c,
    
    Martin
    
    
    -- 
    Prof. Dr. Martin Schulz, Chair of Computer Architecture and Parallel Systems
    Department of Informatics, TU-Munich, Boltzmannstraße 3, D-85748 Garching
    Member of the Board of Directors at the Leibniz Supercomputing Centre (LRZ)
    Email: schulzm at in.tum.de
     
     
    
    On 03.02.21, 16:21, "Wesley Bland" <work at wesbland.com> wrote:
    
        If that’s the solution, that’s fine. The other option is to hold MPI 4.0 if this is a showstopper. I’m not that familiar with the problem so I was trying to figure out which side this is on.
    
        Thanks,
        Wes
    
        > On Feb 3, 2021, at 9:18 AM, Martin Schulz <schulzm at in.tum.de> wrote:
        > 
        > Hi Wesley,
        > 
        > Sorry, we overlapped - based on the discussion we had on Monday, there is no way we can get this into MPI 4.0 the "right" way. So, the idea to put a stop-gap measure in that for now disallows certain parts, and that could be done as an errata. Hence, it would be nice to get this cleaned and ready, but I don't think we have a huge rush here.
        > 
        > Martin
        > 
        > 
        > -- 
        > Prof. Dr. Martin Schulz, Chair of Computer Architecture and Parallel Systems
        > Department of Informatics, TU-Munich, Boltzmannstraße 3, D-85748 Garching
        > Member of the Board of Directors at the Leibniz Supercomputing Centre (LRZ)
        > Email: schulzm at in.tum.de
        > 
        > 
        > 
        > On 03.02.21, 16:16, "Bland, Wesley" <wesley.bland at intel.com> wrote:
        > 
        >    Is this something that needs to be in MPI 4.0? If so, we need to at least decide what we want to do along that path. If this can be fixed by an errata later, that’s fine.
        > 
        >    Thanks,
        >    Wes
        > 
        >> On Feb 3, 2021, at 9:11 AM, Pritchard Jr., Howard <howardp at lanl.gov> wrote:
        >> 
        >> Hi Martin,
        >> 
        >> I mulled this over some more and think we need to discuss the implications here further within the WG before presenting something to the forum.
        >> So let's hold off bringing this topic up at today's MPI forum.
        >> 
        >> Howard
        >> 
        >> 
        >> On 2/2/21, 8:28 AM, "Schulz, Martin" <martin.w.j.schulz at tum.de> wrote:
        >> 
        >>   Hi Howard,
        >> 
        >>   Based on the discussion yesterday in the WG, should we mention this at the whole forum tomorrow and possible suggest the one sentence errata that Spawn is only of for the WPM, or is this too early?
        >> 
        >>   Thanks,
        >> 
        >>   Martin
        >> 
        >> 
        >>   -- 
        >>   Prof. Dr. Martin Schulz, Chair of Computer Architecture and Parallel Systems
        >>   Department of Informatics, TU-Munich, Boltzmannstraße 3, D-85748 Garching
        >>   Member of the Board of Directors at the Leibniz Supercomputing Centre (LRZ)
        >>   Email: schulzm at in.tum.de
        >> 
        >> 
        >> 
        >> 
        >> 
        > 
        > 
        >