[MPI3-IO] Non blocking collective I/O ticket

Fab Tillier ftillier at microsoft.com
Wed Dec 14 11:38:15 CST 2011


Hi Mohamad,

I read it over and it looks good to me.

Cheers,
-Fab

Mohamad Chaarawi wrote on Wed, 14 Dec 2011 at 08:17:27
> 
> I updated the draft and attached it to the ticket with Fab's comments,
> which included moving the entire split collective section to the
> deprecated chapter, and emphasizing more the progress rules.
> 
> I would like to get some feedback especially on the nonblocking
> Collective I/O section on page 484, which I updated to emphasize the
> progress rules which wasn't very clear in the first draft.
> 
> Right now it reads as:
> 
> \subsection{Nonblocking Collective File Operations}
> %-------------------------------------- ;
> \label{sec:io-semantics-nb-collective} Nonblocking collective file
> operations are subject to the same restrictions as blocking collective
> I/O operations. All processes, belonging to the communicator that was
> used to open the file, must call collective I/O operations (blocking and
> nonblocking) in the same order. In particular, once a process calls a
> collective I/O operation, all other processes, belonging to the
> communicator used to open the file, must eventually call the same
> collective I/O operation, and no other collective I/O operation on the
> same file handle in between.
> 
> Nonblocking collective I/O operations do not match with blocking
> collective I/O operations. Multiple nonblocking collective I/O
> operations can be outstanding on a single file handle.  Quality MPI
> implementations should be able to support a large number of pending
> nonblocking I/O operations.
> 
> All calls are local and return immediately, irrespective of the status
> of other processes. The call initiates the operation which may progress
> independently of any communication, computation, or I/O. The call
> returns a request handle, which must be passed to a completion call. All
> input arguments should not be modified, and all output buffers should
> not be accessed before the completion call. The same progress rules
> described for nonblocking collective operations apply for nonblocking
> collective I/O operations. For a complete discussion, please refer to
> the semantics set forth in Section~\ref{sec:nbcoll} on
> page~\pageref{sec:nbcoll}.
> 
> 
> Thanks,
> Mohamad
> 
> On 12/13/2011 04:49 PM, Mohamad Chaarawi wrote:
> Hi All,
> 
> I think enough people have joined to post to this list now..
> Quincey will present ticket 273 (immediate non-blocking collective I/O
> operations) for a first reading at the January meeting. I attached an
> updated draft for IO chapter authors (or others) to review before
> sending it out on the general mailing list.
> Last time there was an issue with deprecating the split collective
> routines. Let me know if you have any comments.
> 
> ticket url: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/273
> <https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/273>
> 
> Thanks,
> Mohamad
> 
> 
> 
> 
> 
> _______________________________________________ MPI3-IO mailing list
> MPI3-IO at lists.mpi-forum.org <mailto:MPI3-IO at lists.mpi-forum.org>
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-io
> <http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-io>
> 
>





More information about the mpiwg-io mailing list