Re: Canonical (External?) Data Representation

Albert Cheng (acheng@ncsa.uiuc.edu)
Wed, 19 Feb 1997 16:32:11 -0600

Jean-Pierre,
I probably used some wrong wording. I didnot mean the
canonical formatted could carry the run-time view from
application to application. I meant for each run-time
view, all MPI-environments will use the same defined
bits pattern for the same data values. (hm... now I
see how I confused you. I used the word "unique" bits
pattern in my last mail. A unique pattern would mean one
can deduce the view from it. Wrong wording, Albert.)

At 01:50 PM 2/19/97 -0400, Jean-Pierre Prost wrote:
>Just one precision to what you are saying Albert.
>MPI views are dynamic in their nature, and there is not
>one-to-one mapping between MPI files and MPI views.
>There can even be several views for the same open file
>on a given process.
>Views are associated with open files when the MPI
>application runs, and they are usually different on different
>MPI tasks in order to achieve file data partitioning across
>tasks. Therefore, in most cases, there is more than one
>view associated with an MPI file.
>The purpose of the canonical representation of file data is
>to maintain persistence of data representation across runs
>in different MPI environments.
>Therefore, I do not think we can relate canonical representation
>to MPI views.
>Jean-Pierre
>
>##########################################
>Jean-Pierre Prost
>IBM T.J. Watson Research Center
>PO Box 218
>Yorktown Heights, NY 10598
>USA
>Phone: (914) 945 3225 - Fax: (914) 945 2141
>Lotus Notes: Jean-Pierre Prost @ IBM Research
>Internet: jpprost@watson.ibm.com
>URL: http://www.research.ibm.com/people/p/prost/
>##########################################
>
>
>
>