[Mpi3-abi] ABI Working Group

Thu Feb 7 14:07:59 CST 2008

Hi everybody,

I've just uploaded an updated proposal version 0.3 to the ABI Wiki page
(see http://svn.mpi-forum.org/trac/mpi-forum-web/wiki/AbiWikiPage).

Best regards.

Alexander 

-----Original Message-----
From: Jeff Brown [mailto:jeffb_at_[hidden]] 
Sent: Thursday, February 07, 2008 7:17 PM
To: Supalov, Alexander; Edric Ellis; Narasimhan, Kannan; Erez Haba;
Christian.Bell_at_[hidden]; Solt, David George; sameer Shende
Cc: Nathan DeBardeleben; ddd_at_[hidden]; jeffb_at_[hidden]
Subject: RE: ABI Working Group

all,

Let's start using the ABI WG email list so that the discussions will 
be archived, etc.

To subscribe, link to:
http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-abi

Rich is encouraging all the working groups to have a mid term telecon 
some time within the next couple of weeks.  I think we should do that 
as well.  In preparation for the telecon, let's do some work on the 
draft proposal Alexander kicked off then discuss.

What's a good date/time for the telecon?  What are the time zone 
restrictions (i.e. where are all of you located)?

Jeff

At 03:43 AM 1/31/2008, Supalov, Alexander wrote:
>Dear Edric,
>
>Thank you. I accepted all changes and updated the document slightly.
>Note that I added version number to the file name to keep track of
draft
>editions. Please use this document for further contributions if you
can.
>
>Now, to your points. The idea of a common compatibility "outlet" may I
>think work well in practice, but this would be 2nd best solution if you
>ask me. We should keep this option in mind as plan B (in fact, this
>would be a standardized morph layer interface - a chance to save some
>work downstream by reusing one of the existing morph packages). I'd
>rather go for plan A first - agreeing on a "real" common interface if
>possible.
>
>That ABI cannot cover the practical side of the matter is no problem if
>one puts this into perspective: ABI is just another step on the way to
>MPI maturity. Agreeing on functionality may come next if the ABI bet
>pans out.
>
>Best regards.
>
>Alexander
>
>--
>Dr Alexander Supalov
>Intel GmbH
>Hermuelheimer Strasse 8a
>50321 Bruehl, Germany
>Phone:          +49 2232 209034
>Mobile:          +49 173 511 8735
>Fax:              +49 2232 209029
>
>
>-----Original Message-----
>From: Edric Ellis [mailto:Edric.Ellis_at_[hidden]]
>Sent: Thursday, January 31, 2008 11:28 AM
>To: Supalov, Alexander; Narasimhan, Kannan; Jeff Brown; Erez Haba;
>Christian.Bell_at_[hidden]; Solt, David George
>Cc: craig Rasmussen; Nathan DeBardeleben; ddd_at_[hidden];
>rlgraham_at_[hidden]
>Subject: RE: ABI Working Group
>
>Hi Alexander,
>
>Thanks for that - there's lots of good stuff in there. I've added a few
>things, and there are a couple more that I'm not entirely sure how best
>to include, or perhaps they're already covered elsewhere:
>
>* It's possible to write an application that can compile successfully
>with multiple MPI implementations, but works correctly with only one.
>This issue is not intended to be addressed by the ABI as it is present
>even with the current API.
>
>* In my mind, we could describe the ABI as effectively another language
>binding. This would mean that MPI implementations were free to keep
>their "native" mpi.h representation for situations where maximum
>performance is required and recompilation is possible, and the "abi"
>mpi.h library can be implemented using a morph layer or similar, and
>this is used by applications that cannot be recompiled.
>
>Cheers,
>
>Edric.
>
> > -----Original Message-----
> > From: Supalov, Alexander [mailto:alexander.supalov_at_[hidden]]
> > Sent: Wednesday, January 30, 2008 10:41 PM
> > To: Narasimhan, Kannan; Jeff Brown; Edric Ellis; Erez Haba;
> > Christian.Bell_at_[hidden]; Solt, David George
> > Cc: craig Rasmussen; Nathan DeBardeleben; ddd_at_[hidden];
> > rlgraham_at_[hidden]
> > Subject: RE: ABI Working Group
> >
> > Hi everybody,
> >
> > Here comes the promised draft. Comments and updates are
> > welcome. If you can, please use "Track Changes" in MS Word
> > when contributing.
> >
> > Best regards.
> >
> > Alexander
> >
> > --
> > Dr Alexander Supalov
> > Intel GmbH
> > Hermuelheimer Strasse 8a
> > 50321 Bruehl, Germany
> > Phone:          +49 2232 209034
> > Mobile:          +49 173 511 8735
> > Fax:              +49 2232 209029
> >
> >
> > -----Original Message-----
> > From: Narasimhan, Kannan [mailto:kannan.narasimhan_at_[hidden]]
> > Sent: Tuesday, January 29, 2008 10:50 PM
> > To: Jeff Brown; Supalov, Alexander; Edric Ellis; Erez Haba;
> > Christian.Bell_at_[hidden]; Solt, David George; Narasimhan, Kannan
> > Cc: craig Rasmussen; Nathan DeBardeleben; ddd_at_[hidden];
> > rlgraham_at_[hidden]
> > Subject: RE: ABI Working Group
> >
> > Adding David Solt to this Working Group (and email list).
> >
> > Thanx!
> > Kannan
> >
> > -----Original Message-----
> > From: Jeff Brown [mailto:jeffb_at_[hidden]]
> > Sent: Tuesday, January 29, 2008 3:41 PM
> > To: Supalov, Alexander; Edric Ellis; Erez Haba; Narasimhan,
> > Kannan; Christian.Bell_at_[hidden]
> > Cc: craig Rasmussen; Nathan DeBardeleben; ddd_at_[hidden];
> > rlgraham_at_[hidden]
> > Subject: RE: ABI Working Group
> >
> > Let's see what we can get done by email, then set up a
> > telecon for mid February.  I have a couple of questions for the
group:
> > 1. what is the high level goal of this WG?
> >
> > should be covered in the intro to the proposal - Alexander to draft
> >
> > 2. what do we want to accomplish by the next Chicago meeting?
> >
> > have a fairly well thought out proposal out there to present
> > to the larger group
> >
> > At 12:34 PM 1/29/2008, Supalov, Alexander wrote:
> > >Hi,
> > >
> > >Upon creation of the email list and some pause, I would suggest to
> > >organize a kickoff telecon with all who subscribe to the
> > list by then.
> > >Waiting till the next Forum meeting is also an option unless we're
> > >under pressure to report out earlier to stay afloat.
> > >
> > >As for the contents, I'd first clearly identify - for us and for
the
> > >future readers/voters - what this proposal is about, and
> > what is not in
> >
> > >its scope.  Here, the earlier contributions may help to
> > introduce some
> > >co-ordinate system and select a patch we can reasonably hope to
> > address.
> > >I can sketch this intro up in the coming days.
> >
> > that would be great!
> >
> >
> > >Then we have to select some procedure for editing the proposal, sit
> > >down and write it.
> >
> > Let's see if we can draft a proposal prior to the telecon
> > then send out to the larger group for review prior to the
> > Chicago meeting.
> >
> > Jeff
> >
> >
> >
> >
>---------------------------------------------------------------------
>Intel GmbH
>Dornacher Strasse 1
>85622 Feldkirchen/Muenchen Germany
>Sitz der Gesellschaft: Feldkirchen bei Muenchen
>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
>Registergericht: Muenchen HRB 47456 Ust.-IdNr.
>VAT Registration No.: DE129385895
>Citibank Frankfurt (BLZ 502 109 00) 600119052
>
>This e-mail and any attachments may contain confidential material for
>the sole use of the intended recipient(s). Any review or distribution
>by others is strictly prohibited. If you are not the intended
>recipient, please contact the sender and delete all copies.

---------------------------------------------------------------------
Intel GmbH
Dornacher Strasse 1
85622 Feldkirchen/Muenchen Germany
Sitz der Gesellschaft: Feldkirchen bei Muenchen
Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
Registergericht: Muenchen HRB 47456 Ust.-IdNr.
VAT Registration No.: DE129385895
Citibank Frankfurt (BLZ 502 109 00) 600119052

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.