[Mpi3-abi] telecon Thursday June 26

Supalov, Alexander alexander.supalov at [hidden]
Thu Jun 26 11:41:07 CDT 2008



Hi everybody,

Unfortunately, I was unable to connect under +1-505-606-1201. Here's what I'd like to say about the classes introduced into the spreadsheet:

This comes from Bill Gropp's paper on GMPI, a layer that would insulate MPI details from the client. Since ABI is going to act almost like GMPI, this classification seems to be applicable to our case as well.

The classes are as follows:

o	Compile-time values
§		Used in declarations (MPI_MAX_ERROR_STRING, etc.)

Since preprocessor and compiler are involved, there's no chance for ABI other than to agree on common values for all complying implementations.

§		Other (MPI_ANY_SOURCE, MPI_ERR_TRUNCATE, etc.)

These we can translate within the ABI wrapper, if any.

o	Init-time constants (MPI_INT, etc.)

Same here. These conceptually come to be after MPI_Init. Mind the array conversion in the collective *v calls, though.

o	Opaque objects (MPI_Comm, MPI_Datatype, etc.)

Same here.

o	Defined objects (MPI_Status)

This is the biggest problem of all, most likely requiring an extension to the MPI API.

o	Defined pointers (MPI_BOTTOM, MPI_STATUS_NULL, etc.)

I bet we can try to translate these in the wrapper. See Status above.

The next step would be to check the classification for correctness, and then set about defining the standard entity sizes and their encoding.

Best regards.

Alexander 

-----Original Message-----
From: mpi3-abi-bounces_at_[hidden] [mailto:mpi3-abi-bounces_at_[hidden]] On Behalf Of Jeff Brown
Sent: Wednesday, June 25, 2008 6:04 PM
To: MPI 3.0 ABI working group; MPI 3.0 ABI working group
Subject: Re: [Mpi3-abi] telecon Thursday June 26

eems I attached an old version - here's the latest

At 09:59 AM 6/25/2008, Jeff Brown wrote:
>reminder - telecon tomorrow (Thursday) at 10:00 MDT
>
>latest spreadsheet attached ...
>
>We will go over the "Class" column and start discussing populating 
>the ABI column.
>
>Jeff
>
>At 01:33 PM 6/19/2008, Jeff Brown wrote:
>>The conference call is scheduled.
>>
>>Dial in number: 665-2778 local  866-736-2112 toll free
>>
>>I hope the lines stay up this time!
>>
>>Jeff
>>
>>At 11:15 AM 6/19/2008, Jeff Brown wrote:
>>>folks,
>>>
>>>I'll schedule a telecon for Thursday 26 at 10:00 A.M. MDT.
>>>
>>>In the meantime, please take a look at the latest spreadsheet 
>>>posted on June 4.  Alexander added a "Class" column that 
>>>categorizes each item according to the taxonomy in the Gropp 
>>>paper.  The next step would be to agree on these categorizations 
>>>then start populating the ABI column.
>>>
>>>Jeff
>>>
>>>
>>>_______________________________________________
>>>mpi3-abi mailing list
>>>mpi3-abi_at_[hidden]
>>>http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-abi
>>
>>
>>_______________________________________________
>>mpi3-abi mailing list
>>mpi3-abi_at_[hidden]
>>http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-abi
>
>
>_______________________________________________
>mpi3-abi mailing list
>mpi3-abi_at_[hidden]
>http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-abi
---------------------------------------------------------------------
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.



More information about the Mpi3-abi mailing list